Power Systems

Amjad Anvari-Moghaddam
Hamdi Abdi
Behnam Mohammadi-Ivatloo
Nikos Hatziargyriou Editors

Microgrids
Advances in Operation, Control, and
Protection
Power Systems
Electrical power has been the technological foundation of industrial societies
for many years. Although the systems designed to provide and apply electrical
energy have reached a high degree of maturity, unforeseen problems are constantly
encountered, necessitating the design of more efficient and reliable systems based
on novel technologies. The book series Power Systems is aimed at providing
detailed, accurate and sound technical information about these new developments
in electrical power engineering. It includes topics on power generation, storage
and transmission as well as electrical machines. The monographs and advanced
textbooks in this series address researchers, lecturers, industrial engineers and
senior students in electrical engineering.

**Power Systems is indexed in Scopus**

More information about this series at http://www.springer.com/series/4622

Amjad Anvari-Moghaddam • Hamdi Abdi
Behnam Mohammadi-Ivatloo • Nikos Hatziargyriou
Editors

Microgrids
Advances in Operation, Control,
and Protection
Editors
Amjad Anvari-Moghaddam Hamdi Abdi
Department of Energy Technology Electrical Engineering Department
Aalborg University Razi University
Aalborg, Denmark Kermanshah, Iran

Behnam Mohammadi-Ivatloo Nikos Hatziargyriou

Faculty of Electrical Electrical and Computer Engineering
and Computer Engineering Department
University of Tabriz National Technical University of Athens
Tabriz, Iran Zografou, Attika, Greece

ISSN 1612-1287 ISSN 1860-4676 (electronic)

Power Systems
ISBN 978-3-030-59749-8 ISBN 978-3-030-59750-4 (eBook)
https://doi.org/10.1007/978-3-030-59750-4

© Springer Nature Switzerland AG 2021

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

Increased penetration of distributed generations (DGs) in the power system has

crucially changed the control and operation relevant concepts. In this way, the
microgrids are being expanded and developed as a fundamental and essential
building block for future smart power systems. The microgrid concept-introduced
by the Consortium for Electric Reliability Technology Solutions (CERTS) to
improve the reliability, sustainability, and efficiency of the modern power system-
is an aggregation of DG units, distributed energy storage (DES), sensitive and
nonsensitive loads, and centralized/decentralized control system, operating as a
controllable subsystem which can operate in grid-connected as well as in an islanded
mode. These new systems requested a revisited definition of the most well-known
issues in the power system control aspects. AC, DC, and AC/DC networks, or hybrid
microgrids are the most dominant nature in these new systems. Therefore, key
issues for the operation, control, and protection of these systems include integration
technologies, hierarchical control techniques, and optimization methods that should
be carefully updated, focusing on primary, secondary, and tertiary control layers
in both islanded and grid-connected modes. The importance of revising the relevant
operation studies in the presence of microgrids is very crucial and vital. The impacts
of different uncertainties arising from the increased penetration of renewables
sources, power market pricing policies, electric vehicles, storage system devices,
and demand-side management are the most important features.
This book covers a comprehensive study on the control, operation, and protection
of microgrids with related strategies to analyze and understand the salient features
of modern control and optimization techniques applied to these systems. It also
discusses emerging concepts, key drivers and new players in microgrids, and local
energy markets while addressing various aspects from day-ahead scheduling to real-
time testing of microgrids.

v
vi Preface

This book is divided into three parts, including operation (Chaps. 1–9), control
(Chaps. 10–15), and protection (Chaps. 16–21) of microgrids.
Part I covers the studies related to microgrid operation. The first chapter
provides an overview of the definitions and clustering of microgrids from different
perspectives. Chapter 2 studies the advantages, challenges, objectives, architecture,
control strategies, and operation of networked microgrid and those in clusters. This
chapter also proposes a model for daily energy management and scheduling of
networked microgrids considering different generation resources and loads. Chap-
ter 3 discusses energy management systems (EMSs) for microgrids in normal and
contingency conditions and appropriate objectives are defined for each condition.
This chapter also provides an overview of the different aspects that should be
considered in EMS design and implementation for microgrids like technical and
security issues, economic objectives, power flow management, reconfiguration,
etc. Optimal dispatch and unit commitment in microgrids are studied in Chap. 4,
considering economic and environmental aspects at the same time, which results
in a multi-objective optimization problem. The fuzzy decision-making method is
utilized to find the best compromise operational schedule of energy units. Chapter 5
focuses on energy storage systems in different microgrids, such as land-based
microgrids and mobile microgrids. The application of energy storage for load
leveling and power quality improvement in microgrids are also studied in this
chapter. The authors of Chap. 6 study the formation of local markets in microgrids
and provide an overview of definitions in this area, potential benefits, and objectives.
A summary of key enabler elements for local market implementation is given, and
different trading approaches, as well as, market settlement approaches for local
markets are presented with detailed case examples to help readers with outlining
attributes of different market models. Chapter 7 covers a summary of demand
response (DR) programs applicable to microgrids in different sectors such as
residential and commercial and explores the impact of customers’ participation
level in both price-based and incentive-based DR programs from the economic
point of view. Chapter 8 elaborates on a framework for operation management of
networked microgrids that have different owners. A combination of the alternating
direction method of multipliers and robust optimization methods is introduced and
implemented to solve effectively such a problem at the operating layer. Chapter 9
focuses on recent progress in the application of computational intelligence and
heuristic techniques in microgrids and provides an overview of the application of
evolutionary algorithms to the energy management problem of microgrids.
Part II covers the topics related to the control of microgrids ranging from
conventional droop methods at local levels to wide-area measurement system
(WAMS)-based hierarchical control techniques. Chapter 10 focuses on droop
control concept in microgrids and discusses the application of droop-based control
methods in both AC and DC microgrids, considering different characteristics
and features of inverter-based renewable energy sources, dispersed generation
units, and energy storage systems. Chapter 11 addresses the hierarchical control
structure of microgrids, where the primary, secondary, and tertiary control levels
are discussed in detail. The chapter also gives a focus on distributed control
Preface vii

of both AC and DC microgrids and covers the distributed control techniques

utilized for voltage/frequency control as well as active/reactive power-sharing.
Chapter 12 introduces the control and application of energy storage systems (ESSs)
in microgrid systems. The characteristics of energy storage techniques, power
electronic interfaces, and battery management systems are also discussed. Finally,
a comprehensive review of ESSs in both islanded and grid-connected microgrids
is conducted and future research roadmaps of multifunctional ESSs together with
cyber-security issues of ESSs are outlined. Chapter 13 elaborates on microgrid’s
stability definitions and analysis, followed by examples on voltage and frequency
stability improvement in islanded microgrids. The chapter also introduces a novel
multi-machine structure-based simulation model for the study of the dynamic
behavior of microgrids. Chapter 14 studies the control methods used to compensate
harmonics and voltage unbalance disturbances in microgrids as the main problems
of power quality in the steady state. Finally, Chap. 15 in Part II of the book presents a
WAMS-based hierarchical control for islanded microgrids where a stable operation
is an important concern due to the low-inertial nature of power electronics interfaced
units.
Part III covers recent advances in the protection of microgrids. This part of the
book elaborates on conventional and emerging schemes for fault identification,
protection, and restoration in microgrids, followed by various real-time testing
methods and the categories of simulation suitable for control and protection of
microgrids. In this regard, Chap. 16 is devoted to Fault Ride Through (FRT),
as the ability of distributed energy resources to stay connected during the faulty
conditions, and Fault Current Management (FCM) in microgrids. This chapter also
provides the microgrid FRT in different grid codes and proposes new methods for
FRT control and FCM. The proposed model is validated using real-time simulation
based on hardware in the loop tests. Chapter 17 provides an overview of the
challenges in microgrid protection and employed relays. The application of multi-
agent systems (MAS) to microgrid protection is investigated in this chapter, where
different studies related to relay coordination and islanding detection are carried
out. Chapter 18 provides an overview of requirements for the protection system
from the point of view of differential protection used in different elements of
microgrids and distribution systems such as bus bars, transformers, generators,
and reactors. As a complementary example to this chapter, the second central
moment (SCM) scheme is presented as a fast and low computation method for fault
detection in different zones. Chapter 19 discusses the conventional and adaptive
protection schemes for microgrids during grid-connected and islanded operation
conditions. The chapter also addresses the issues related to protection schemes in
systems with a high number of distributed energy sources, gives an overview of the
existing and new requirements of protection schemes, and analyzes their potentials
in microgrids. Chapter 20 mostly focuses on fault identification and restoration
of microgrids after fault clearance. The challenges related to direction detection,
blinding issues, and possible mal-operations are studied as well in this chapter.
Finally, Chap. 21 studies the real-time testing of microgrids. This chapter explores
the basics and makes a comparison of the various testing methods ranging from
viii Preface

off-line to real-time simulations, rapid controller prototyping, hardware in the loop

(HIL): both the signal level (CHIL) and power level (PHIL), real-time power level
emulation, test-bed platforms, hybrid approaches/combinations of these techniques
and novel solutions such as digital twin, blockchain and internet of things (IoT)
based approaches.

Aalborg, Denmark Amjad Anvari-Moghaddam

Kermanshah, Iran Hamdi Abdi
Tabriz, Iran Behnam Mohammadi-Ivatloo
Zografou, Attika, Greece Nikos Hatziargyriou
Contents

Part I Operation of Microgrids

1 An Introduction to Microgrids, Concepts, Definition,
and Classifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Maryam Shahbazitabar, Hamdi Abdi, Hossein Nourianfar,
Amjad Anvari-Moghaddam, Behnam Mohammadi-Ivatloo,
and Nikos Hatziargyriou
2 Operation Management of Microgrid Clusters . . . . . . . . . . . . . . . . . . . . . . . . . 17
Meisam Moradi and Asghar Akbari Foroud
3 Energy Management Systems for Microgrids . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Seyed Mohsen Hashemi and Vahid Vahidinasab
4 Optimal Dispatch and Unit Commitment in Microgrids . . . . . . . . . . . . . . . 97
Hossein Shayeghi and Masoud Alilou
5 The Role of Energy Storage Systems in Microgrids Operation . . . . . . . 127
Sidun Fang and Yu Wang
6 Microgrids and Local Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Mohsen Khorasany and Reza Razzaghi
7 An Economic Demand Management Strategy for Passive
Consumers Considering Demand-Side Management Schemes
and Microgrid Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
Mohammad Esmaeil Honarmand, Vahid Hosseinnezhad,
Barry Hayes, Behnam Mohammadi-Ivatloo, and Pierluigi Siano
8 Real-Time Perspective in Distributed Robust Operation
of Networked Microgrids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
Mehdi Jalali, Manijeh Alipour, and Kazem Zare

ix
x Contents

9 Application of Heuristic Techniques and Evolutionary

Algorithms in Microgrids Optimization Problems . . . . . . . . . . . . . . . . . . . . . 219
Amir Aminzadeh Ghavifekr

Part II Control of Microgrids

10 Conventional Droop Methods for Microgrids . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
Kwang Woo Joung and Jung-Wook Park
11 Distributed Control Approaches for Microgrids . . . . . . . . . . . . . . . . . . . . . . . . 275
Tohid Khalili and Ali Bidram
12 On Control of Energy Storage Systems in Microgrids . . . . . . . . . . . . . . . . . 289
Yu Wang, Sidun Fang, and Yan Xu
13 Microgrid Stability Definition, Analysis, and Examples . . . . . . . . . . . . . . . 305
Hossein Shayeghi, Hamzeh Aryanpour, Masoud Alilou,
and Aref Jalili
14 Voltage Unbalance Compensation in AC Microgrids . . . . . . . . . . . . . . . . . . . 337
Shahram Karimi, Mehdi Norianfar, and Josep M. Guerrero
15 WAM-Based Hierarchical Control of Islanded AC Microgrids . . . . . . . 375
E. S. N. Raju P and Trapti Jain

Part III Protection of Microgrids

16 Fault Ride Through and Fault Current Management for
Microgrids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
Wei Kou and Sung-Yeul Park
17 Microgrid Protection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 437
Arturo Conde Enríquez, Yendry González Cardoso,
and José Treviño Martínez
18 A New Second Central Moment-Based Algorithm for
Differential Protection in Micro-Grids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 489
Ernesto Vázquez, Héctor Esponda, and Manuel A. Andrade
19 Microgrid Protection with Conventional and Adaptive
Protection Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523
Aushiq Ali Memon, Hannu Laaksonen, and Kimmo Kauhaniemi
20 Fault Identification, Protection Schemes, and Restoration
Requirements of Microgrids. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 581
Amin Yazdaninejadi, Amir Hamidi, Sajjad Golshannavaz,
and Daryoush Nazarpour
Contents xi

21 Real-Time Testing of Microgrids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 615

A. S. Vijay and Suryanarayana Doolla

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
 Part I
Operation of Microgrids
Chapter 1
An Introduction to Microgrids, Concepts,
Definition, and Classifications

Maryam Shahbazitabar, Hamdi Abdi, Hossein Nourianfar,

Amjad Anvari-Moghaddam , Behnam Mohammadi-Ivatloo,
and Nikos Hatziargyriou

1.1 Introduction

Traditional electric power systems are rapidly transforming by increased renewable

energy sources (RESs) penetration resulting in more efficient and clean energy pro-
duction while requiring advanced control and management functions. Microgrids
(MGs) are significant parts of this transformation at the distribution level.
As a fact, since the year 2004, in which the MG was defined as “a better way to
realize the emerging potential of distributed generation in a systematic approach
which views generation and associated loads as a subsystem” [1], significant
improvements and innovations have been made.
The MG concept was firstly introduced by the USA’s Consortium for Elec-
tric Reliability Technology Solutions (CERTS) to reduce the cost, and increase
the power quality, effectively all around the world [2]. Among various definitions,
the U.S. Department of Energy (DOE) Microgrid Exchange Group (MEG) has used
the following [3]:
“A microgrid is a group of interconnected loads and distributed energy resources
within clearly defined electrical boundaries that acts as a single controllable entity

M. Shahbazitabar () · H. Abdi · H. Nourianfar

Department of Electrical Engineering, Faculty of Engineering, Razi University, Kermanshah, Iran
e-mail: m.sht@razi.ac.ir
A. Anvari-Moghaddam
Department of Energy Technology, Aalborg University, Aalborg, Denmark
B. Mohammadi-Ivatloo
Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, Iran
N. Hatziargyriou
Electrical and Computer Engineering Department, National Technical University of Athens
Zografou, Attika, Greece

© Springer Nature Switzerland AG 2021 3

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_1
4 M. Shahbazitabar et al.

with respect to the grid. A microgrid can connect and disconnect from the grid to
enable both grid-connected and island-modes of operation.” In a widely accepted
definition “Microgrids are electricity distribution systems containing loads and
distributed energy resources, (such as distributed generators, storage devices, or
controllable loads) that can be operated in a controlled, coordinated way, either
while connected to the main power network and/or while islanded” [4]. The MG
is a flexible and dispatchable system that is capable of operating in both modes
of grid-connected or stand-alone. It can potentially reduce the dependency of its
consumers on traditional generation systems by providing different types of energy,
such as electrical and thermal energy, and provide ancillary services trading activity
between the MG and the main grid. The MG configuration can be AC, DC, or
hybrid.
This chapter, as an introduction to the MG concept, tries to present some practical
and useful information for MG integration. Definitions, classifications, components,
control methods, and protection schemes of MGs are also addressed briefly along
with their merits or demerits.

1.2 Microgrid Components

Global warming and growing energy demand are the most significant drivers
spurring renewable energy sources (RESs) to reduce greenhouse gas (GHG)
emissions by fossil fuel-based electricity generation. Distributed energy resources
(DERs) such as solar photovoltaic (PV) modules, wind turbines (WTs), combined
heat and power (CHP) units, and controllable loads such as electric vehicles (EVs)
are expected to play a considerable role in future electricity supply because of
their significant benefits such as carbon emissions reduction, energy efficiency
enhancement, power quality and reliability (PQR) improvement, and line losses
reduction and deferral of grid expansion plans [5]. The intermittent nature of
renewable-based DERs is the main challenge for their integration into traditional
power systems.
The presence of MGs helps the increase of DGs penetration, more specifically
at low voltages (LV) distribution networks. DERs integration, along with energy
storage systems (ESSs) and controllable loads near power consumers within MGs,
provides economic and environmental benefits [6]. MGs as parts of the distribution
systems are connected to the upstream network at a single point of common
coupling (PCC) often by power electronic-based switchgear. The operation of MGs
in islanded mode is particularly challenging. In such an operation, the ESSs activity
is particularly essential to improve power quality, stability, and reliability of supply,
at least for critical loads.
The capability of MGs to switch into the islanded mode in case of faults in the
upstream network increases the reliability of customer supply and the resilience of
1 An Introduction to Microgrids, Concepts, Definition, and Classifications 5

Wind turbine Loads

Solar PV
MG
control center

Battery ESS

Micro-Grid

Main Grid

Conventional DG
Controllable Loads

Fig. 1.1 A simple MG structure

the local distribution networks. This is particularly important, in case of natural or

human-made disasters, when MGs can isolate and continue to provide electricity
to critical loads. In fact, resilience is nowadays one of the key drivers for the
development of MGs in developed countries, such as the United States, Japan, etc.
Perhaps, the most common application of MGs is found in rural electrification.
In developing countries, MGs can be used for the electricity supply of remote
communities or to support some facilities, such as healthcare, water use, food
preservation, waste treatment, telecommunication support, etc.
Figure 1.1 shows a simple MG comprising DGs, ESS, and flexible loads
connected via power converters. Extensive research is currently underway in MG
development and demonstration to solve several technical and economic challenges
such as accurate and integrated management of all energy sources, increasing the
penetration of hybrid AC/DC power networks in the various areas of planning,
operation, protection, and control [7].

1.3 Classification

According to Navigant Research, MG can be classified into different groups based

on various aspects [8, 9]:

1.3.1 Type

MGs could be categorized in different types scuh as campus, military, residential,

commercial, and industrial. The campus MG includes onsite generation, while it
is managed by a single owner. Military MGs are used for improved efficiency and
resilience.
6 M. Shahbazitabar et al.

1.3.2 Size

MGs can exist in different sizes: small and simple (about hundreds of kW) at low
or medium voltages [4] supplying just a few customers to large and complex ones
(few MWs) [10].

1.3.3 Application

MGs can be used to provide premium power or resilience-oriented services. The

premium power is the power that provides a stable level of voltage noise free to
its end consumers [11]. Also, the resilience-oriented MG is referred to as an MG
with the ability to withstand and recover from “high impact–low-frequency” events
[12]. The accidents, such as deliberate attacks, or naturally occurring incidents, are
considered in this regard, while the negative impacts during both long-term and
short-term horizons should be minimized [13].

1.3.4 Operation Mode

MGs can operate in two modes: grid-connected and islanded. In grid-connected

mode, the MG can exchange power with the upstream grid, depending on the elec-
tricity generated and its load demand [14]. The MG can be disconnected from the
utility grid due to faults or in planned maintenance and operate autonomously [15].
Unlike grid-connected mode, an islanded MG may face challenges in regulating
voltage and frequency or maintain the required quality of the power.

1.3.5 Configuration

The topologies in which components of an MG, namely loads, micro-sources, and

storage devices, are integrated lead to different configurations: AC network MGs,
DC network MGs, and hybrid AC–DC MGs. Emerging DC sources and loads have
given rise to the application of DC–MGs in recent years. Distribution in AC–MG
can be one of the following three types: single phase, three phase with neutral, and
three phase without neutral, while in DC–MG, it can be monopolar, bipolar, and
homopolar [16]. In AC-one, DC-based DERs, ESSs, and loads are connected to
a universal AC bus via DC-to-AC inverters. On the other hand, DC–MGs could
offer various merits compared to AC–MGs: more efficient supply of DC loads, loss
reduction via decreasing the multiple converters used for DC loads, facilitate various
DC-DERs integration such as fuel cells (FC) and photovoltaic systems (PV) to the
1 An Introduction to Microgrids, Concepts, Definition, and Classifications 7

Urban Rural Off-grid

Feeders are placed in a
Feeders are loaded in a rather populated area, the
By definition, an off-grid
congested industrial area distance between the main
microgrid always
with low imbalance body and the laterals is
operates in islanded
degree becouse of short fairly long becouse of
mode, remote area
distance between the main scattered loads; the voltage
assighned with no
body and the laterals. The profile is not flat. DERs
possibility for macrogrid,
voltage and frequency are have effects on voltage
where large-size DER
controled by the fluctuation and must be
integration is occurring
macrogrid so the voltage controlled to facilitate in
faster.
drop is low. the feeder voltage
regulation.

Fig. 1.2 Feeder-based classification

common node with simplified interfaces, and decreasing the need for synchronizing
generators and the buses versus several challenges in control, and operation
[8, 17, 18].

1.3.6 Characteristics/Properties of the Feeder

MGs can be categorized into three groups via feeder properties, as shown in Fig. 1.2.

1.4 Control

Control of MGs is the one significant feature that distinguishes them from simple
distribution lines with DER. This is further discussed in this section.

1.4.1 Hierarchical Control

Hierarchical, multilevel control is adopted for the effective control of MGs,

including the following three levels [19, 20].

1.4.1.1 Primary Control

Its task is the control of voltage, current, and local power. The set points of inverters
in this level are changed based on their droops.
8 M. Shahbazitabar et al.

1.4.1.2 Secondary Control

It deals with optimal load sharing, frequency restoration, voltage regulation at pilot
points, etc., and this is the level where the MGCC determines the set points needed
to be followed by local controllers at the primary level. Some important subjects,
such as forecasting functions and economic dispatch, could be also implemented at
this level.

1.4.1.3 Tertiary Control

This level deals with upstream networks, like MG synchronization, and electricity
market trading.
The time scales of MG control functions can be divided into different levels. For
example, primary level control actions such as voltage and current control should be
executed in a couple of seconds to meet the system’s security constraints. Secondary
and tertiary control functions require a couple of minutes, and hours, respectively.
Three-level control functions for the MG are illustrated in Fig. 1.3, with time scaling
definition.

Electricity Market
Trading
Terti ary

MG
Synchronization

Demand
Response

Economic Forecasting
Secondary

Load Sharing Functions

Frequency Economic
Restoration Dispatch

Power quality
Prim ary

Changing the Set -

points of Inverters

Frequency and Voltage

Control

s ms s min h Day

Fig. 1.3 Different control level for the MG

1 An Introduction to Microgrids, Concepts, Definition, and Classifications 9

Central Central Central

Controller Controller Controller
Communication Communication
Communication
Device Device
Device

Comm. Comm. Comm. Comm. Comm. Comm. Comm. Comm. Comm. Comm. Comm. Comm.
Device Device Device Device Device Device Device Device Device Device Device Device

LC LC LC LC LC LC LC LC LC LC LC LC

DG DG DG DG DG DG DG DG DG DG DG DG

Point of common coupling Point of common coupling Point of common coupling

a) Centralized MG b) Distributed MG c) Hybrid MG

Fig. 1.4 Different types of control coordination structures from the communication perspective
[23]

1.4.2 Control Coordination

Control coordination of MG can be divided into a centralized, distributed, or a

combination of both, named as a hybrid [21]. The type of coordinated control
structure designed for the MG depends on DER ownership, MG size, available
technologies, and communication infrastructures. The following is a classification
of coordinated control, as shown in Fig. 1.4.

1.4.2.1 Centralized Control

Building a central controller that can communicate with all controlled units requires
extensive communication infrastructures and significant computer resources. The
main advantage of the centralized control structure is that it can apply optimal
solutions. When the MG switches from grid-connected to islanded mode, one micro-
source can act as a master controller, providing voltage and frequency reference to
others [21]. It allows simple algorithms to be used in the MG energy management
unit. One of the major drawbacks of centralized control is that it suffers from a
single point of failure. A centralized control structure is typically recommended for
small environments such as educational centers and hospitals.

1.4.2.2 Distributed Control

In distributed control, each local controller operates on its own without instructions
from a central controller. Any appropriate control actions are specified locally based
on local evaluation and the information shared among neighboring local controllers
of the MG through peer-to-peer communication. Since limited information about
the entire MG status is communicated among neighboring nodes, optimal global
10 M. Shahbazitabar et al.

performance is not generally achieved in this topology as compared to the central-

ized control scheme. However, a “plug and play” feature is satisfied, meaning that
when a DER is connected or removed from the system, the MG will continuously
operate without reconfiguration. Multi-agent system (MAS) control is an example
of such topology that allows every component to exchange information with its
neighbors as an autonomous entity that can be decided based on its own status with
no external command [22].

1.4.2.3 Hybrid Control

A hybrid control scheme, as the name suggests, is a combination of central

and distributed controllers. Hybrid hierarchical control consists of several central
controllers with a distributed topology that is coordinated. Each central controller
contains several local controllers that can operate independently.

1.5 Stability

RES integration with stochastic, uncontrollable, and intermittent nature is one of

the attractive points of MGs, which in turn necessitates proper mechanisms in the
system to assure reliable operation in transient events such as a sudden drop in
wind speed or a cloud passing over a solar array [24]. DGs may be inverter-based
or directly connected to MGs. Since MGs typically use renewable energy sources,
most DGs are inverter-based. The existence of a variety of DGs can create diverse
features in MG stability problems. Another issue affecting MG stability is its small
inertia. An MG is stable if all the state variables are recovered to steady-state values
after being subjected to a disturbance so that all constraints are satisfied [10].
It should be mentioned that, in MGs which generally are equipped with the
inverter-based DER units, the inertia is zero or very low and the reference signal
is used to set their output frequency, internally [25, 26]. As a matter of fact, the
frequency will not inherently vary as the active power changes [27]. In traditional
power systems, there is the chance to store energy by using the rotating masses
of synchronous generators to deal with load changes. However, in MGs, some
form of energy storage is needed to deal with transients happening during the
islanded operation. As an effective manner, some load shedding strategies and
storage techniques can be applied in MGs [27].
MG stability is divided into grid-connected MG stability issues and islanded MG
stability subjects. DER integration with intermittent nature, which leads to stability
issues in MGs, has been addressed extensively in research works considering both
large and small-signal stabilities. Linear analysis tools such as Nyquist or Routh–
Hurwitz for synchronous generators, inverters, rectifiers, and motors are in common.
On the other hand, large-signal nonlinear stability studies such as the Lyapunov-
based technique are employed for intrinsically nonlinear converters for integrating
1 An Introduction to Microgrids, Concepts, Definition, and Classifications 11

Fig. 1.5 Stability definition

RES and energy storage devices. Instability in conventional power systems can be
classified into three general categories [28]: rotor angle stability, voltage stability,
and frequency stability. All these three mentioned categories are highly dependent
on the dynamic behavior of the synchronous generators [29].

1.5.1 Grid-Connected MG Stability

When the MG is connected, its voltage and frequency are maintained by the utility
grid. Therefore, stability studies of DGs with small capacities are not required. Since
the capacity of the MG is much smaller than the utility grid, the disturbances in the
MG will have little effect on the network frequency regulation. Therefore, rotor
angle and frequency stability are not relevant in the grid-connected mode.

1.5.2 Islanded MG Stability

When the MG is disconnected (islanded mode), it must support its voltage and
frequency. In other words, it must preserve the balance between power generation
and load. Therefore, it is crucial to investigate the voltage and frequency stability
in an islanded MG. Long transmission lines are one of the main causes of voltage
instability in conventional power systems, which limits the transmission of power
between loads and generation. However, in MGs, the feeders are relatively short,
resulting in relatively small voltage drops. Indeed, in an islanded MG, frequency
stability is more significant than voltage one, due to its small inertia. Figure 1.4
shows the classification of MG stability types. The study of the stability types is
beyond the scope of this chapter. A comprehensive review of MG stability can be
found in [30]. A brief stability definition is clarified in Fig. 1.5.
12 M. Shahbazitabar et al.

1.6 Protection

Electronically coupled unit control has significant effects on the MG transient

behavior, particularly subsequent to faults. According to its fault specification, the
fault types can be pole to pole or pole to the ground, bus or feeder. According to
the fault location, the fault would be more severe near the energy sources; so, the
bus fault is crucial in overall the system. Pole-to-pole short circuit may be occurring
inside the capacitors and batteries that generally cannot be quickly diagnosed. In
these cases, using fuses and circuit breakers along with device replacement could be
profitable. Another challenge that has to be gradually overcome is an arc, which is
created by the current interruption and extinguished hardly in a DC system without
the current crossing through zero [17]. Also, low fault currents due to the power
electronics interfaces, and adaptive protection because of the variety of generation
sources, are two main subjects which should be clearly discussed and addressed in
MGs.
We should note that unlike conventional distribution systems in which the power
flow is unidirectional and protection schemes are simple, in the MGs, the power
flow is bidirectional, so their protection schemes are more complex. The main
challenges of protection in DCMGs are related to following issues [31]: lack of
phasor, and frequency data making it difficult to detect and accurate location of
faults; absence of natural zero crossings to extinguish the arc occurring in circuit
breaker opening; rising the fast current imposing strict time limits needed for fault
interruption; protection coordination issues because of intermittent nature of DERs;
and need for suitable protection standards.

1.7 Microgrid’s Advantages and Challenges

In this section, the main advantages, and challenges of MGs are briefly addressed.

1.7.1 Advantages

The main advantages of MGs can be categorized as follows:

• Decreasing CO2 emission and fuel cost by using renewable energy supply instead
of conventional fossil-fueled energy sources.
• Increasing consumer reliability and power system flexibility.
• Increasing the power transmission capacity in DC–MGs.
• Loss reduction with local power delivery.
• Having much smaller financial commitments.
1 An Introduction to Microgrids, Concepts, Definition, and Classifications 13

Operation

Components and compability

Technical
MG challenges

Integration of DERs
Regulation
Protection
Economical

marketing

Fig. 1.6 MG challenges

• Requiring fewer technical skills for operation and relying more on a remote
control, and automation.
• Isolating from any grid disturbance or outage.

1.7.2 Challenges

Despite some clear advantages of MG, there are several challenges must be over-
come. The main types of MG challenges are illustrated in Fig. 1.6.and summarized
as follow:

1.7.2.1 Technical

• The electricity generation of some RESs, such as wind and solar, is highly depen-
dent on weather conditions; hence their generation is unpredictable. Because
their capacity is small, they are sensitive to unpredictable changes. This causes
problems with operational capability.
• Due to the smaller number of loads and interrelated changes in available energy
sources, the uncertainty of islanded MGs is much greater than that of the utility
grid. Even though the reliability of MG could be higher than the utility grid.
• One of the challenges of islanded MGs is their low inertia characteristic in
comparison with the bulk power systems due to numerous power electronic-
based units and lack of conventional synchronous generators. This low inertia
in the system can lead to intense frequency deviations.
• One of the major challenges faced by MGs in islanded mode is maintaining the
balance between generation and load continuously. Large disturbances can easily
lead to MG instability.
• In order to ensure the safety and reliability of the system, islanding conditions
must be quickly detected.
14 M. Shahbazitabar et al.

1.7.2.2 Regulation

The current regulatory framework was not designed to incorporate DERs or MGs.
Therefore, in some countries, changes have been made to the regulatory framework
to influence the benefits of MGs to the entire community. Some laws about
renewable DGs and energy storage systems have been incorporated into the new
regulatory framework [32]. As an example, some of the common grid services
provided by ESSs are categorized as power quality, transient stability, and regulation
services; spinning reserve, voltage control, arbitrage (energy), load balancing,
congestion relief, firm capacity, and upgrade deferral [33].

1.7.2.3 Economical

Despite advances in technology, the investment cost remains high in MGs. The cost
of energy storage systems, some of DGs such as photovoltaic (PV) and fuel cells,
is still high and not affordable. However, today in most countries, there are various
types of financial support to facilitate conditions for investment in this field. Another
economic challenge of MGs is its efficient energy management.

1.7.2.4 Marketing

MGs, in addition to supplying local loads, can sell their additional generation power
to the utility grid or purchase some power from the utility grid. Thus, MGs can
participate in the market by selling their products and services. MGs also play an
important role in developing free/local energy markets by encouraging energy users
to install DERs, offering new services, and supporting self-consumption [32]. While
regulatory gaps can be primarily followed to the origin of DSO (distribution system
operator) concerns, market gaps lying before the road of commercializing MG are
mainly related to the direct economic interests of end users and DG owners. Based
on these conditions, denial of local energy trading among the DG and demand, and
MG market positioning difficulty can be determined as two major market challenges
in this context [34].

References

1. Lasseter, R.H. and P. Paigi. Microgrid: A conceptual solution. In 2004 IEEE 35th Annual Power
Electronics Specialists Conference (IEEE Cat. No. 04CH37551). 2004. IEEE.
2. Planas, E., et al. (2015). AC and DC technology in microgrids: A review. Renewable and
Sustainable Energy Reviews, 43, 726–749.
3. Energy, U., DOE microgrid workshop report. 2018.
4. Hatziargyriou, N. (2014). Microgrids: Architectures and control. John Wiley & Sons.
1 An Introduction to Microgrids, Concepts, Definition, and Classifications 15

5. Asano, H., et al. (2007). Microgrids: An overview of ongoing research, development, and
demonstration projects. IEEE Power Energy Magazine, 78–94.
6. Shayeghi, H., et al. (2019). A survey on microgrid energy management considering flexible
energy sources. Energies, 12(11), 2156.
7. Abdi, H., & Shahbazitabar, M. (2020). Smart city: A review on concepts, definitions, standards,
experiments, and challenges. Journal of Energy Management and Technology, 4(3), 1–6.
8. Lotfi, H., & Khodaei, A. (2015). AC versus DC microgrid planning. IEEE Transactions on
Smart Grid, 8(1), 296–304.
9. Hooshyar, A., & Iravani, R. (2017). Microgrid protection. Proceedings of the IEEE, 105(7),
1332–1353.
10. Farrokhabadi, M., et al. (2019). Microgrid stability definitions, analysis, and examples. IEEE
Transactions on Power Systems, 35(1), 13–29.
11. Azar, K. Power consumption and generation in the electronics industry. A perspective. In
Sixteenth Annual IEEE Semiconductor Thermal Measurement and Management Symposium
(Cat. No. 00CH37068). 2000. IEEE.
12. Chi, Y. and Y. Xu. Resilience-oriented microgrids: A comprehensive literature review. In 2017
IEEE Innovative Smart Grid Technologies-Asia (ISGT-Asia). 2017. IEEE.
13. Watson, J.-P., et al., Conceptual framework for developing resilience metrics for the electricity
oil and gas sectors in the United States. Sandia national laboratories, albuquerque, nm (united
states), tech. rep, 2014.
14. Pedrasa, M.A. and T. Spooner. A survey of techniques used to control microgrid generation and
storage during island operation. In Proceedings of the 2006 Australasian Universities Power
Engineering Conference (AUPEC’06). 2006.
15. Lopes, J.P., et al. Control strategies for microgrids emergency operation. In 2005 International
Conference on Future Power Systems. 2005. IEEE.
16. Justo, J. J., et al. (2013). AC-microgrids versus DC-microgrids with distributed energy
resources: A review. Renewable and Sustainable Energy Reviews, 24, 387–405.
17. Zhang, L., et al. (2018). A review on protection of DC microgrids. Journal of Modern Power
Systems and Clean Energy, 6(6), 1113–1127.
18. Rodriguez-Diaz, E., et al. Multi-level energy management and optimal control of a residential
DC microgrid. In 2017 IEEE International Conference on Consumer Electronics (ICCE). 2017.
IEEE.
19. Meng, L., et al. (2017). Review on control of DC microgrids and multiple microgrid clusters.
IEEE Journal of Emerging and Selected Topics in Power Electronics, 5(3), 928–948.
20. Shotorbani, A. M., et al. (2018). Distributed secondary control of battery energy storage
systems in a stand-alone microgrid. IET Generation, Transmission & Distribution, 12(17),
3944–3953.
21. Yamashita, D. Y., Vechiu, I., & Gaubert, J.-P. (2020). A review of hierarchical control for
building microgrids. Renewable and Sustainable Energy Reviews, 118, 109523.
22. Zhou, Y. and C.N.-M. Ho. A review on microgrid architectures and control methods. In 2016
IEEE 8th International Power Electronics and Motion Control Conference (IPEMC-ECCE
Asia). 2016. IEEE.
23. Meng, L., Hierarchical control for optimal and distributed operation of microgrid systems.
2015, Ph. D. dissertation, 10 2015.
24. Kabalan, M., Singh, P., & Niebur, D. (2016). Large signal Lyapunov-based stability studies in
microgrids: A review. IEEE Transactions on Smart Grid, 8(5), 2287–2295.
25. De Brabandere, K., et al. (2007). A voltage and frequency droop control method for parallel
inverters. IEEE Transactions on Power Electronics, 22(4), 1107–1115.
26. Blaabjerg, F., et al. (2006). Overview of control and grid synchronization for distributed power
generation systems. IEEE Transactions on Industrial Electronics, 53(5), 1398–1409.
27. Dag, O. and B. Mirafzal. On stability of islanded low-inertia microgrids. In 2016 Clemson
University Power Systems Conference (PSC). 2016. IEEE.
16 M. Shahbazitabar et al.

28. Kundur, P., et al. (2004). Definition and classification of power system stability IEEE/CIGRE
joint task force on stability terms and definitions. IEEE Transactions on Power Systems, 19(3),
1387–1401.
29. Maniatopoulos, M., et al. (2017). Combined control and power hardware in-the-loop simula-
tion for testing smart grid control algorithms. IET Generation, Transmission & Distribution,
11(12), 3009–3018.
30. Shuai, Z., et al. (2016). Microgrid stability: Classification and a review. Renewable and
Sustainable Energy Reviews, 58, 167–179.
31. Jayamaha, D., Lidula, N., & Rajapakse, A. (2020). Protection and grounding methods in DC
microgrids: Comprehensive review and analysis. Renewable and Sustainable Energy Reviews,
120, 109631.
32. Bellido, M. H., et al. (2018). Barriers, challenges and opportunities for microgrid implemen-
tation: The case of Federal University of Rio de Janeiro. Journal of Cleaner Production, 188,
203–216.
33. Castillo, A., & Gayme, D. F. (2014). Grid-scale energy storage applications in renewable
energy integration: A survey. Energy Conversion and Management, 87, 885–894.
34. Tao, L., et al. From laboratory Microgrid to real markets—Challenges and opportunities. In 8th
International Conference on Power Electronics-ECCE Asia. 2011. IEEE.
Chapter 2
Operation Management of Microgrid
Clusters

Meisam Moradi and Asghar Akbari Foroud

2.1 Introduction

This chapter deals with the operation management of networked microgrid clusters
(NMCs) or networked microgrids (NMGs). The system that contains a connection
of two or more microgrids with the ability to exchange energy with each other and
with a distribution system (DS) is called NMCs. The NMCs differ from the DS
includes multi-microgrids which exchange energy only with DS. In NMCs, power
flow from one MG to another MG or DS is possible bidirectional and the topology
of the network in the NMCs can change continuously. NMCs in the normal mode
is similar to the DS with several MGs and have the benefits of these networks.
The main difference of this structure is adding the networked mode between MGs.
Designing and operation of a set of multiple MGs with DS as NMCs, will lead to
increasing the resiliency of the network significantly.
Networked mode operation in NMCs reduces system loss. The loss parameter
is defined as the loss of power network persistence and failure to supply demand
after an extreme event. Since the resiliency of the system is inversely related to this
parameter, so, the resiliency of NMG is higher than the resiliency of the distribution
network containing several MGs and traditional networks. Also, the improvement
of resiliency increases system reliability. Studies show that the average system
interruption duration index (SAIDI) and the average system interruption frequency
index (SAIFI) are decreased by adding MGs to the network [1]. As the resiliency
and reliability of NMCs are more than the DS with several MGs, the value of these
mentioned two indices also improves more in the NMCs.

M. Moradi · A. Akbari Foroud ()

Semnan University, Semnan, Iran
e-mail: mrd.meisam@semnan.ac.ir; aakbari@semnan.ac.ir

© Springer Nature Switzerland AG 2021 17

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_2
18 M. Moradi and A. Akbari Foroud

To indicate the difference between NMCs and DS containing several MGs, Fig.
2.1 and Fig. 2.2 are presented. Figure 2.1 demonstrates the DS with NMCs. There
are five MGs with one radial 33-bus IEEE DS. Each MG is capable of handling its
loads. Also, in NMCs, each MG has one or more PCCs which makes the possibility
of exchanging power with other MGs or DS. As can be seen in this figure, MG1 and
MG2 can operate as the grid-connected mode by closing switch SW1 via PCC at
bus 25 and switches SW2 and SW3 via PCCs at bus 33 or 14, respectively. These
two MGs can operate as islanded or grid-connected. In addition to the island and
grid-connected mode, the other three MGs can exchange power with each other in
networked mode. By closing switches SW7, SW8, or both, the networked mode of
MGs operation can be enabled. It is noteworthy that MG3 and MG5 with closing
SW8 and SW7 through PCC at bus 18 and bus 22 can exchange power with MG4,
respectively. Figure 2.2 shows the DS with two MGs. In this structure, the MGs
are connected to the network via PCC on buses 12 and 33. Also, in this network,
MGs are capable of supplying their loads and in order to have higher reliability,
they are connected to the DS. Clearly, the first network is more resilient to severe
contingencies than the second. In sever contingencies it is possible not to provide
critical loads of the entire network or MGs and it is impossible to return to the
initial conditions quickly. So, the resiliency of the system decreases against an
extreme event. As shown in Fig. 2.1, NMCs had a higher level of resiliency due
to its ability to change configuration against extreme events. Due to the flexible
NMCs configuration, any interruption factor outside of the MGs is simply cut off.
In this condition, the NMCs can exchange power through their physical network.
By MGs isolation, the repair process in power distribution systems becomes faster.
After removing the interruption factor, it is possible to start supplying the power
of critical loads from the MGs. Accordingly, system performance can be recovered
much faster with a simultaneous bidirectional supply method. Therefore, it can be

MG1
MG2
29 30 31 32 33 SW2
SW1

25 28

24 27
SW3
23 26
1 2 12 18
Main Grid
3 4 5 6 7 8 9 10 11 13 14 15 16 17
19

20 SW5 SW4

21
SW6 SW7 SW8
22

MG5 MG4 MG3

Fig. 2.1 Structure of NMCs

2 Operation Management of Microgrid Clusters 19

15

9 14 16 17
8

2 3 4 5 6 10 12 13
1
7
11
35

18 19

MG1
22 20 21

31
27 32 33 34
23
26 29

24 28 30

25
MG2

Fig. 2.2 Structure of DS with several MGs

concluded that the NMCs have more resiliency and less power return time than
several MGs or traditional DSs.
On the other hand, some researches are being done to develop the NMCs, such
as Bronzeville Community Microgrid (BCM)1 and Illinois Institute of Technology
(IIT)2 [1]. These researches demonstrate that NMCs can reduce contamination sig-
nificantly and improve ancillary services, such as sustainability, security, efficiency,
reliability, and cost reduction of providing customers demand. Also, according to
research done by the Los Alamos National Laboratory, the NMCs can reduce the
operating costs of MGs by at least 10% [1]. This is a significant reduction in
operating costs. Given these cases, further studies are necessary on the NMCs. The
major aspects of NMCs can be divided into four categories including architecture,
control strategy, communication, and operation. In this chapter, in addition to the
advantages and challenges of NMCs, the main objectives, planning, and operation
of NMCs and the motivations that lead to the evolution of this power system, will
be described. Also, the different types of energy management systems (EMS) and
problem-solving methods in the NMCs will be analyzed.
Microgrids can be “networked” in physical layers by distribution systems with
closing or opening one or more keys or controlling layers by independent local
controllers or both of them. It is important to note that the priority of any microgrid
is the balancing of power and economic operation in the management of its internal

1 https://bronzevillecommunityofthefuture.com/
2 http://www.iitmicrogrid.net/microgrid.aspx
20 M. Moradi and A. Akbari Foroud

resources, and then, if the microgrid is not capable of feeding its loads, it can buy
the amount of its energy deficit from the other microgrids. In emergency conditions,
the microgrid is separated from the upstream network and it uses the output power
within the microgrid or if needed, it uses the output power from the other microgrids.
In such conditions, it is observed that the EMS protects critical loads from being
outage. These structures are highly flexible in critical situations and also resistant to
unnecessary outages. However, distribution systems with multi-microgrids are not
always in networked mode and switch to networked mode according to predefined
conditions and planning. Figure 2.3 shows the operation of physical and control
layers.
It is to mention that the range of opportunities and potential architectures of
NMCs like the individual MGs are very diverse, however, there are very few
examples are deployed in NMCs. Also, there are very few tools to simulate or
analyze the behavior of NMCs. Nevertheless, several national laboratories are devel-
oping design and analysis software tools that partially address some components of
the quantitative evaluation of the economic, reliability, and resilience benefits of
networked microgrids.

Other loads and

Generations

Controller N
Controller 1

Controller 2
Gen 2
Gen 1

Loads

Gen 2
Gen 1

Loads
Gen 2
Gen 1

Loads

Network
Controller

Fig. 2.3 Performance of physical and control layers in NMCs

2 Operation Management of Microgrid Clusters 21

2.2 Benefits of NMCs

Considering NMCs advantages, the number of MGs will increase and NMCs in the
DS will be created. In developed countries, NMCS was considered as the transition
state of smart grids [2].

2.2.1 Best Utilization of DERs

In general, using DERs locally is more accessible in the microgrids network. These
resources can be connected to the nearest distribution network in larger networks. In
fact, the EMS in NMCs can network multiple microgrids and distributed generation
resources can also feed the intended consumer in this condition. Therefore, EMS
can reduce the total cost of the operating system in NMCs by integrating local
DERs. In other words, in distribution systems with microgrids, if the use of DERs is
done with a proper control or coordinator system, the uncertainty of wind and solar
renewable energy sources in different microgrids can also be reduced by information
aggregation and correlation.
Besides, microgrids that have energy storage systems can play an important role
in smoothing the power generation of DGs. For example, a storage system of one
MG can be entered at peak load times of other MGs and reduce the startup cost
and total operating cost of the system by preventing the installation of thermal
units. This work reduces the peak load and facilitates demand response. Also, at
the transmission level, the integration of DERs using the microgrids network can
increase the efficiency of the transmission system and postpone the investment of
transmission substruction and new substations.

2.2.2 Reduction of Overall Cost

The integration of DERs using the microgrids network can decrease the total cost
of the required generation power. For example, a study done by the Los Alamos
National Laboratory shows that operation in NMG mode reduces at least 10% of
costs compared to normal distribution network mode. This amount of reduction in
operating costs is significant. Also, it has been shown in [3] that the interaction
between interconnected and independent microgrids in the form of NMCs, with
planning strategy and energy trading development, can reduce the total operating
cost by 13.2% and reduce the cost by 29.4% for one MG.
22 M. Moradi and A. Akbari Foroud

2.2.3 Ancillary Services Improvement

Optimal use of DERs can improve ancillary services in NMCs. In fact, ancillary
services keep flowing and direction of power by maintaining a balance between
supply and demand and it can help the system returning to normal mode when faults
occur. Also, since EVs and ESSs have been used in NMCs and EVs are capable of
transferring stored power, proper planning of this equipment can further enhance the
ancillary services market. Therefore, at the planning stage of NMCs, the addition of
power storage equipment can add benefits to the ancillary market. Studies show that
using NMCs improves 10% of service costs and the ancillary services market [3].

2.2.4 Resiliency Improvement

Improving system resilience is one of the most important goals of power systems.
The use of NMCs allows the generation resources to be distributed between
microgrids and loads near to DSs, which can lead to additional resiliency and lower
investment costs. On the other hand, networked mode operation in NMCs reduces
system loss. The loss parameter is defined as the loss of power network persistence
and failure to supply demand after an extreme event. Since the resiliency of the
system is inversely related to this parameter, so, the resiliency of NMG is higher than
that of the distribution network containing several MGs and traditional networks.

2.2.5 Reliability Improvement

One of the benefits of NMCs is the increment of system reliability. This advantage
can be achieved by reducing the amount of load which is not provided or reducing
the utility outages. This reduction of blackouts in NMCs may increase costs
by balancing between generation and consumption, but there must be a balance
between the costs of removing remaining blackouts against the entire cost of the
network. This means that removing the load may be more economical for the
network. However, networking the MGs and sharing their resources reduce the total
investment costs of reservation generation and increment of reliability. It is usually
suggested that critical peak load be considered to account for approximately 80% of
the backup generation capacity [3].
In a simple example, suppose if two MGs have a peak load-to-average load ratio
of 1 MW to 0.5 MW and the critical peak load time of them are not the same, instead
of buying two generators with a nominal capacity (1 MW/0.80 = 1.25 MW), in
NMG mode, the total required capacity of 1.5 MW/0.80 = 1.88 MW can be used.
This capacity reduction will reduce the cost of $186,000 by considering the cost of
0.30 $/W for a diesel generator. However, if the capacity of the bought generators
2 Operation Management of Microgrid Clusters 23

is not reduced, a higher capacity of the generators can increase the reliability of
the system. Also, the improvement of resiliency increases system reliability. Studies
show that the average of system interruption duration index (SAIDI) and the average
of system interruption frequency index (SAIFI) are decreased by adding MGs to the
network. As the resiliency and reliability of NMCs are more than the DS with several
MGs, the value of these mentioned two indices also improves more in the NMCs.

2.2.6 Bilateral and out of Market Transactions

Since in NMCs, power exchanging between MGs exists, there is also the possibility
of power exchanging outside the wholesale market without controlling. Suppose
in the series structure of NMCs, upper MGs buy the required energy from the
distribution network with wholesale price. Since the only way to meet the shortage
of power required by the second MG is to purchase it from the first MG distribution
feeder. Therefore, the second MG has to buy a lack of power at the retail price
from the first MG, this bilateral transaction is not acceptable in NMCs. In other
words, if the pure measured energy from the upstream system is purchased with
wholesale rate while being sold with retail rate to the other consumers or MGs,
bilateral transaction rules in NMCs are allowed to prevent exporting energy. Because
in the retail transaction, it is possible to increase the price anomalistic price. In many
areas, this price increment can reach 50$/MWh. For example, if multiple microgrids
on a 1 MW grid can prevent 1 MW from exporting over 4 hours per day, the total
profit would be 73,000$/year [4].

2.3 Challenges of NMCs

2.3.1 Stability of the System

Disconnecting from the upstream network and switching to the network mode
can cause system stability problems. By isolating MG, the penetration level of
resources and uncertainties of them are increased compared to the previous one
and can cause challenges in the stability of the system. In Ref [5], the stability
problems associated with NMCs have been discussed exactly. By increasing the
penetration level of resources with uncertainty, there is the possibility of transition
from stability margin. Therefore, for any change in the structure of NMCs, stability
limits must be examined. As the topology of NMCs is changing constantly, stability
assessment is essential for this structure. Therefore, due to widespread and rapid
changes in NMCs, control systems should be able to analyze network stability with
high speed. In other words, any change in NMCs, especially when operating in the
island state, should be done to ensure sufficient stability margins of them. Once
the margin of proper stability is assured, economic dispatch can be done, and only
24 M. Moradi and A. Akbari Foroud

when networked MGs have sufficient stability margins, they can serve as resiliency
sources to actively and coordinately provide ancillary services that stabilize, restore,
or black start the main grid.

2.3.2 Protection Coordination

Proper protection coordination is one of the most essential requirements for using
high-reliability NMCs. As mentioned in the previous section, to achieve the best
performance, the network structure in NMCs has been modified frequently and this
can cause a challenge in coordinating protection equipment. It should be noted that
by changing the network topology, the amplitude and direction of fault currents are
constantly changing and can be quite different from the previous one in each case.
Also, the coordination of protective equipment in the grid-connected mode, isolate
mode, and networked mode is different from each other. Therefore, a large number
of possible topologies of NMCs changes must be preplanned. On the other hand, the
equipment fault current rate of each MG is usually considered above the nominal
limit, but it is not valid for upstream equipment. The expansion of grid-connected
MGs increases the level of system fault current and increases the nominal value of
equipment such as transformers, breakers, and protection equipment in the upstream
network (DS). Replacing equipment with higher fault current endurance is not an
economical solution. One of the proper methods is to use an adaptive protection
scheme, but it will increase the complexity of the protection system.

2.3.3 Privacy of MGs

In terms of ownership, MG is divided into three categories, utility MG, community

MG, and private MG. Privacy is a social concern and since all MG owners may not
want to share all the information except the details of power exchange, this can be
a challenge for NMCs. It should be noted that excess or deficit of power in NMCs
is usually compensated within the internal network of MGs and this will reduce the
privacy challenge partly.

2.3.4 Threat of Cyberattack

Information and Communication Technology (ICT) is one of the main elements of

NMCs. Therefore, there is always the possibility of a cyberattack that can prevent
the proper entire function of the system. Therefore, cybersecurity systems must be
strong enough to prevent any probable cyberattacks. Any disturbance of the process-
ing of information or sending it to unidirectional or bidirectional communication
2 Operation Management of Microgrid Clusters 25

protocols can endanger the entire security of the system. Both in the centralized
and distributed control scheme of NMCs, there is the possibility of cyberattacks
and serious damage to the system. Of course, centralized control systems are more
resistant to cyberattacks [6]. The vulnerability is relatively higher in the distributed
control scheme, because of requiring partial data of local controllers. For example,
a destructive factor can attack to one node or linkage between nodes and exchange
inaccurate data to local or central controllers and causes instability of the entire
system. So, NMCs that require extensive communications for optimal control are
very susceptible to future cyberattack.

2.3.5 Disallowed Transactions

As mentioned in Sect. 2.6, in systems with NMCs, it is possible to exchange power

out of the framework. This will significantly increase the price of energy for the
consumer. Therefore, the exchange of power between MGs in the grid mode and
disconnected from the main grid requires an appropriate regulatory framework to
minimize illegal transactions or nodal prices increment.

2.4 Main Objectives and Constraints of NMCs

The main objectives of NMCs are divided into two levels of the distribution network
and the MG level. The objective at the distribution network level is to achieve
economic dispatch and maintain power quality throughout the studied system. To
achieve this objective, using distributed generation resources should be optimized
and the EMS should be implemented effectively in both the distribution system
and all MGs. The main operation of EMS in distribution systems includes optimal
economic performance and energy quality throughout the system.
The main objective at the level of MGs is to control voltage frequency, power
supply, and generation balance, and to manage energy storage and reservation at
each MG. In NMCs, EMS for each MG supplies the power of each MG. The
extra power generation of MG is stored in the storage system or delivered directly
to the DS or adjacent MG through EMS coordination at the distribution level.
Similarly, the lack of power of MG is supplied either through the distribution
network or through a grid connection between the MGs, directly from the MGs
of itself. Therefore, with EMS, the power supply throughout the system continues
uninterrupted and economically. This economic dispatch program in NMCs is done
by EMS at predetermined time intervals.
It should be noted that the studied system can be grid-connected, disconnected
from the grid, and a combination of both of them (Only NMCs should be
disconnected from the grid). NMCs are capable of sharing power between different
MGs of one network or between MG and DS. In an emergency condition and
26 M. Moradi and A. Akbari Foroud

isolation state from the distribution network, all MGs or some of them with more
significant loads operate as a grid by EMS. In this condition, only critical loads are
initially provided in MG. Extra power can be sent to the main grid after providing
significant loads in the DS depending on the communication between the DSO
and the transmission system operator. However, if power generation is insufficient
to supply all the loads on the DS, only significant system loads can be supplied
by activating energy storage equipment in each MG. Therefore, NMCs can feed
critical loads for longer periods by DERs and ESSs in critical conditions. In Table
2.1, general considerations, the main purposes, and constraints used in NMCs are
discussed.

2.5 Typical Architecture of NMCs

In general, the connection of MGs to each other or connection of them to a

distribution network is different. Different connections can have advantages and
disadvantages for NMCs. Since the structure of NMCs has many variations for
optimal control, it is necessary to know the types of NMCs architecture. In different
architectures, the role of each MG can vary depending on the laws of interaction
between them. There are generally two methods of interaction. For example, in one
structure, one MG may be considered as a generating and controlling member but
in another structure, it can be a controllable member [1, 3, 31, and].
In general, the architecture of NMCs can be divided into three overall categories:
serial MGs on a single feeder, parallel MGs on a single feeder, and interconnected
MGs with multiple feeders. Depending on the architecture considered for NMCs,
a set of unique communication, control, protection, and economic requirements are
considered. Figure 2.4 shows the different types of architectures used for NMCs.

2.5.1 Serial MGs on a Single Distribution Feeder

If there is no external grid, the interconnected system must be able to control the
voltage and frequency. To achieve this goal, coordination between MGs is essential
to balance power. Unlike parallel architecture, in the event of an occurrence or
disconnection, the system can be divided into subclusters with similar architecture.
According to this topology, MGs within the subcluster do not lose completely
external support. So, this topology performs better than the off-grid operation.
The serial architecture needs further evaluation and study due to its potential
performance and benefits.
According to Fig. 2.4a, if two or more MGs connect together by a single
interconnection and install on one feeder, they will form the structure of the
serial MGs on a single feeder. In this configuration, there is only one way of
communication through MG A to MG B. As MGs are interconnected and have
2 Operation Management of Microgrid Clusters 27

Table 2.1 General considerations and the main objective in NMCs

General considerations Main objective Methodology Highlights Ref
Internal communications of Develop Software PSCAD Offline study [19]
MGs hierarchical
Minimize load shading control
Stability issues and high Stability analysis FARs and Gersgorin Offline studies [5]
penetration of DERs theorem without ESS
with LP
MG syncs issues and pay Pay attention to Observer based Limited to [6]
attention to cyberattack cyberattack inverter-based
MGs
Incentive-based model with Ancillary Convex programming The role of the [7]
the presence of PV, WTs, service analysis with ADMM network operator
and ESSs has been ignored
Priority indexing of sources Economic OBS and ODD Uncertainty not [8]
and loads scheduling seen
Optimum energy scheduling
Coordinated energy Energy Stochastic ESS has not been [9]
management management programming seen.
In the presence of KKT and big-M
uncertainty and scenario method
reduction methods
Two-layered stochastic Minimize Stochastic Only active [10]
energy modeling with loads operating costs optimization power flow
and DG uncertainties
Distributed energy Energy Robust optimization High-volume [11]
management management ADMM computing
Uncertainty in DERs and
loads
Energy management Energy ADMM online High-volume [12]
Online monitoring management computing
Controller for a wide range Develop Software simulation PV and wind [13]
of functions and hierarchical uncertainties are
communication failure is control not seen.
modeled.
Resiliency improvement NMCs planning Robust optimization ESS has not been [14]
N–k contingency modeling CPLEX seen.
Self-healing and Energy Consensus algorithm Dynamic [16]
autonomous operation management stability is not
Energy storage and MTs guaranteed.
Dynamic electricity pricing Economic Fuzzy adaptive Uncertainty not [17]
based on incentive model, scheduling MOPSO seen
PV, WTs, MTs, FCs, and
ESSs
Service and maintenance Planning service MILP stochastic High- [18]
With DG and ESSs and maintenance performance PC
is needed.
Voltage control and power Voltage stability Simulink modeling Delay in [20]
management communication
causes instability.
(continuted)
28 M. Moradi and A. Akbari Foroud

Table 2.1 (continued)

General considerations Main objective Methodology Highlights Ref
Outage detection Economic MIQP An advanced [21]
Reconfiguration and load scheduling metering
curtailment infrastructure
system is
required.
V/F stability Develop Software PSCAD The study is [22]
hierarchical offline and
control requires large
ESS.
P/F/V control Economic MILP It has an [23]
With linear approximation scheduling uncertain
solution.
Combined AC/DC MGs Develop Simulink Privacy issues are [24]
with various DGs hierarchical not considered
control
Economic dispatch problem Economic MOIA Suitable only for [25]
Non-dispatchable DG and scheduling MGs with
ESS nondispatchable
DGs.
Hybrid energy management Energy CPLEX It is a day-ahead [26]
MGs privacy management scheduling
Auction-based production Automatic OPC toolbox in Reliable [27]
control short-term energy generation MATLAB communication
management control is required.
Energy modeling with MPC Energy MILP Uncertainty is [28]
framework management not considered.
Different types of DGs and
ESSs
Minimize operating costs NMCs planning ICA ESS not seen. [29]
Uncertainties in DERs and
loads
Maximize the profits of Demand PSO It is a day-ahead [30]
customers and energy response and scheduling
producers energy
Reduce production management
costs/pollution/uncertainty programs
and maximize load supply
Minimize the operating Energy Self-healing control Load and [15]
costs of each MG and management generation
automatic power supply in Exchange of uncertainties are
NMCs and charge control of active and not considered.
energy storage before and reactive power
after fault
NMCs resiliency NMCs planning Robust optimization High-volume [15]
DC/AC MGs and consensus algorithm computing
uncertainties
2 Operation Management of Microgrid Clusters 29

1 2 3
Main Grid

SW1

Communication line SW2

Power line
MG A MG B

DMS
(a) Serial MGs on a single feeder of DS

1 2 3
Main Grid

SW1 SW3

Communication line SW2

Power line
MG A MG B

DMS
(b) Parallel MGs on a single feeder of DS

1 2
SW1

MG A

Main Grid SW2

DMS

MG B

SW3

3 4
(c) Interconnected MGs on multiple feeders of DS

Fig. 2.4 Types of architectures used in NMCs

only one way of communication, a communication protocol with a master-slave

architecture or hierarchical communication system can be used to control this
structure. In this structure, the distribution management system (DMS) provides
the communication requirements for the main controller of MG A, and the MG
A controller (MCA) also provides the communication requirements for the MCB
similarly. Since there is only one connector between the equipment, each DMS
needs to exchange information with the MCB.
30 M. Moradi and A. Akbari Foroud

The following situations may occur for this architecture:

If both SW1 and SW2 switches are closed. The MGs are connected to the
distribution feeder. In this case, each MG controller is optimized and controlled
according to its information and orders and the other MGs. In other words, for
optimization of the MGA, lack or extra of MGB power should also be considered.
If both SW1 and SW2 switches are opened. MGs are isolated from the distri-
bution feeder and operates as an island mode. In this case, each MG controller is
optimized and controlled according to its information and orders.
If the SW1 switch is closed and SW2 is opened, MG A is connected to the grid
and MG B operates as an island mode. MCA controls MG A based on information
obtained from DMS, while MCB controls only MG B through the information of
internal resources.
If the SW2 switch is closed and SW1 is opened, MG A is disconnected from
the grid and MG B is connected to MG A through internal communication. In the
hierarchical control architecture, MG A can consider MG B as part of its system
and optimize it individually or optimize MG B with terms of total cost objective
function and voltage and power limit of point common coupling (PCC). For these
conditions, MCA should have plug-and-play capability. But, if a distributed control
system is used, MCA and MCB will interact to reach an agreement about the voltage
and power limitation of the AB at the PCC.

2.5.2 Parallel MGs on a Single Distribution Feeder

According to Fig. 2.4b, if two or more MGs are installed individually like tie
interconnection on one feeder and connected together, they will form a parallel MG
structure on a single feeder. In this structure, all microgrids are connected to the
same external network and each MG needs a PCC for connecting to the network.
Therefore, any electrical path between MGs is performed through an external grid.
Regarding architecture and the grid-connected mode, the MGs can provide ancillary
services to the external grid. Also, the operator of an external grid can send orders
to the MG with the DMS controller. When one MG operates in island mode, it must
be self-sufficient because it has no other external electrical connection to support it.
In this structure, the master-slave or hierarchical control architecture can also
be used and different modes of this connection to the distribution system can be
considered:
If SW1 and SW3 are closed and SW2 is opened, both MGs A and B are connected
separately to the distribution network. In this case, the controller of each MG is
optimized and controlled by DMS based on their information or orders.
If all three switches are opened, both MGs A and B are disconnected from the grid
and each MG is controlled and optimized based on its information and internal
resources.
2 Operation Management of Microgrid Clusters 31

If SW1 is closed and SW2 and SW3 are opened, MG A is connected to the grid,
but MG B is island mode. In this case, the controller of MG A is optimized and
controlled based on the information or order obtained from the DMS, while MG
B is controlled based on its information and internal resources.
If SW1or SW3 is opened and SW2 is closed, MG A or MG B is connected to the
grid and the second MG is connected to the other MG by internal connection
and they form an internal grid. In this case, the grid structure becomes a serial
architecture connected to the single feeder.

2.5.3 Interconnected MGs on Multiple Distribution Feeders

Interconnected MGs in multi-feeders’ structure, that two or more MGs are con-
nected separately to different feeders, are interconnected MGs on multiple distribu-
tion feeders. In this architecture, MGs can be connected directly to the external grid
or can form clusters of series interconnected MGs. Each of these clusters has, at
least, one interconnection with the external grid. It allows MGs to get support from
other grids when they are disconnected or when the external network is overloaded.
Besides, in grid-connected mode, they can provide the ancillary services required
by the main grid operator. Assuming a master-slave architecture for this structure,
the following modes associated with the distribution network may be formed:
1. Both MGs A and B are connected separately to the grid with different feeders.
Each MG controller is optimized and controlled by the DMS based on its
information and orders.
2. Both MGs A and B are disconnected from the grid and are optimized and
controlled based on their information and internal resources.
3. If MG A is connected to the grid but MG B is island mode or vice versa. In this
case, control and optimization are similar to the previous sections.
4. If MG A or MG B are connected to the grid, the other is connected via PCC
AB and communication switch. By coordinating the separator switches in the
feeders, different paths can be provided to transfer power between the feeders or
other MGs. This is a unique feature.
A comparison between different architectures of NMCs is not easy as two
compared systems must be similar in terms of amount and rating. For this purpose,
a comparison is made between two parallel and serial structures in Ref [31]. To
compare both architectures, similar features must be considered. So, according to
Fig. 2.5, both examples contain 4 MGs in the form of a ring.
To reduce this effect, the size of the transmission cables should be large enough.
However, in both plans, cable size increment should be considered. But, in serial
structure, interface cables size between MGs should be increased. It can be noted
that power can flow both through the cables and the interface device. But in the
parallel structure, the interface elements are designed solely on their input and
output power and they have fewer elements than the series structure.
32 M. Moradi and A. Akbari Foroud

PCC 1 PCC 2

Link 1-2
MG1 MG2 MG1 MG2

Link 1-3 Link 2-4

MG3 Link 3-4 MG4 MG3 MG4

PCC 3 PCC 4

Fig. 2.5 Series architectures (left) and parallel (right) with 4 MGs

For example, in Fig. 2.5, the serial structure requires eight interface devices,
but the parallel structure requires four interface devices. Therefore, according to
the required higher rating equipment and the cost of interface devices, the serial
structure has higher costs than the parallel structure. The cost of the hybrid structure
is also expected to be between parallel and serial structures.

2.6 Control Strategy of NMCs

Coordinated control strategies for MGs are divided into four general categories:
(a) peer-to-peer control strategy; (b) master-slave control strategy; (c) hierarchical
control strategy; and (d) distributed control strategy.

2.6.1 Peer-to-Peer Control Strategy in NMCs

Peer-to-peer control systems are recommended for the plugging and playing of the
MG controller. The peer-to-peer control strategy in NMCs has been developed to
remove the problem of centralized control. This method avoids the increment of
communication links and it can develop well and decrease the cost. This type of
control strategy has not a central controller and it is inspired by P2P computer
networking. As shown in Fig. 2.6, all agents or local DERs can communicate with
other agents and have the same importance.
In peer-to-peer, MGs are autonomous due to the absence of a central controller,
and its communication is used for dissemination of the grid states to all required
agents in the MG. The grid-supporting agents can then act according to the received
information and in cooperation with each other. In this way, they should be able to
reach an optimal operation of the considered microgrid. When a single agent fails,
the others can still stably manage the grid. Also, when a single communication
2 Operation Management of Microgrid Clusters 33

Fig. 2.6 P2P control strategy

channel fails, the required information can still reach all necessary participants,
through other agents. Because of these properties, this strategy is a robust way to
control an MG. Also, the elimination of possible privacy concerns is reached due to
keeping all information locally. So, all agents need a considerable amount of local
intelligence, as they need to be able to execute the necessary optimizations.

2.6.2 Master-Slave Control Strategy in NMCs

In master-slaves strategy, one section is known as Master and the other is known as
Slave. The information is transmitted between the master controller and the slave
controllers. The technical difficulties and risks of this control system are low for
NMCs, but if they fail, the main controller of the MGCs and the entire control
system of the NMCs will not be able to perform well. Therefore, if the system
relies too much on the main controller, MGC reliability will get into trouble. It is
noteworthy that the master-slave control is mostly used in the island mode of NMCs
and the peer-to-peer control is mostly used in grid-connected mode.

2.6.3 Hierarchical Control Strategy in NMCs

A hierarchical control strategy (HCS) is used to solve the stability problem when
changing the control mode. Hierarchical control is the most common method in
MGs and MGCs and is also suitable for more complex systems including NMCs.
A multilayer HCS is commonly used to control NMCs. Figure 2.7 represents the
34 M. Moradi and A. Akbari Foroud

Primary Control Secondary Control Tertiary Control

(Distributed) (Quasi-Centralized) (Centralized)

Optimized Energy Management Regular Error Correction Fast Load Sharing

• Economic dispatch • Real-time load management • Droop control
• Unit commitment • Load-frequency control • Local protection
• Optimal power flow • Voltage control
• Reactive power control • Synchronization (MG level only)
(Distributed/networked level) • Automatic generation control
(Networked/ MG level)

Measurement Signal
Control Signal

Milliseconds Minutes

Fig. 2.7 Hierarchical control strategy for networked MGs

HCS for controlling the performance of NMCs. This control structure has primary,
secondary, and tertiary layers.
The main objectives for each control layer are shown in Fig. 2.7. The main
objective function of the primary (first) control layer is the exclusive and local
control of the performance of each MG equipment such as V/f, P/Q, and P/V.
Power quality, power flow, and frequency synchronization control are performed
in the second layer. Optimal energy management such as economic load dispatch,
load forecasting, and resource optimization at the distribution network level is
performed in the third layer of the hierarchical control structure. Also, DMS or
EMS, in general, has direct control over all separator switches and controls the PCC
through MG controllers. It is noteworthy that a reliable communication channel is
an important section of this type of control strategy [32].
In the grid-connected mode of NMCs, DSO communicates with all grid MGs to
control and optimize the economic dispatch and internal resources. In fact, in this
control system, DSO monitors the power flow of all PCCs in the grid. This moni-
toring is done based on price signals/incentives and general objectives/optimization
requirements and is based on solving the economic dispatch problem. Also, when
island mode occurs, DSO sends an islanding mode signal to MG or the MG may
also send the DSO disconnection signal after the disturbance occurs.
In such cases, the DSO may also request from neighboring MGs to switch the
island mode. The synchronization of MGs to the distribution network or switching to
the networked mode is performed by the DSO and the request of the incoming MG
controller. However, the reconnection of the two islanded MGs is done only when
the agreement is reached between the two MGs. First, the V/f control responsibility
is assigned to one MG and the PCC connection point voltage is kept constant, then
the PCC is closed and the networking signal is sent to the two MG sources. HCS
2 Operation Management of Microgrid Clusters 35

requires data collection from all the major components of MGs, and links need to
be established between the DSO and each local controller.

2.6.4 Distributed Control Strategy in NMCs

In distributed control strategy (DCS), each microgrid controls all performances

locally and independently and shares all important information with the other
MGs and DSO controllers. Figure 2.8 represents a typical DCS for controlling
NMCs. Performances such as V/f control, P/Q and P/V control, power distribution,
frequency synchronization, and optimal EM at the MG level are fully distributed. To
achieve optimal performance across the control system, information is shared with
all local controllers in each MG as well as with the DSO. Local controllers continue
their performance by sharing information to optimize performance at the MG level.
This control method can be used in NMCs because of its high resiliency and “plug-
and-play” capability. Because NMCs have high changes in performance level, this
control strategy can maintain optimal system performance in consecutive changes.
Figure 2.9, shows the adopted DCS to control networked MGs. In this strategy,
the primary (distributed) control is performed in the shortest possible time to
maintain always a balance between load and generation, and the secondary control
that is quasi-centralized is implemented in the next step. Primary control is similar in
both distributed and hierarchical control strategies. By changing the system loads,
the PCC voltage and MG frequency will be changed. The voltage and frequency

Primary Control Secondary Control

(Distributed) (Quasi-Centralized)

Information Monitoring Load Sharing and Error

• Information collection from Correction and Energy Management
agents of the other MGs • Droop control
• Local-frequency control
• Information sharing with the • Synchronization
agents of the other MGs • Real-time load management
(Distribution/ networked level) • Optimal power flow
(MG level only)

Local MG information
Other MG information

Milliseconds Seconds

Fig. 2.8 Distributed control strategy for NMCs

36 M. Moradi and A. Akbari Foroud

Tertiary Control DG 1

Global
Consensus
Networked Microgrids Communication Network

Secondary Control

Networked Microgrids Power Network

Local Primary Control

Local Microgrid Power Network

Local Communication Network

Consensus

Droop Reference V&I PWM + UPS

Control Generator Control Inverter
f&V
Restoration
Virtual
Impedance

Secondary Control
DG 2
Primary Control
Local
Consensus
Droop Reference V&I PWM + UPS
Control Generator Control Inverter
F&V
Restoration
Virtual
Impedance

PCC 1

Fig. 2.9 Proper operation of NMCs based on an overall adopted control strategy

Table 2.2 Comparison of control strategies

Features Hierarchical control Distributed control
Reliability Moderate High
Plug-and-play Low High
Flexibility/expandability Low High
Communication bandwidth Low High
Time/space complexity High Low
Design complexity Complex Simple
Economic operation Optimal Suboptimal
Hardware platform Powerful computer Embedded controller

correction in secondary control is performed in both control strategies to return them

to their previous states. A comparison of some features of both control strategies is
presented in Table 2.2 [33].
In the DCS, the master control gives the command to the power generation
units with the aim of setting and supplying the required demand. However, in this
strategy, the economic objectives are arriving through the consensus of all local
controllers per MG. In primary control, the amount of power of the generation units
and in secondary control, the amount of voltage and frequency in each MG are kept
within the allowable range. For this purpose, power electronic converters (PECs)
are responsible for restoring v/f. The adopted distributed control strategy of NMCs
with utilizing PECs is shown in Fig. 2.9.
2 Operation Management of Microgrid Clusters 37

2.7 Energy Management and Operation of NMCs

This section deals with the topic of Energy Management System in NMCs. In
systems with one MG, the MG Central Controller (MGCC) is responsible for energy
management and it has the highest priority in hierarchical control. But if two or
more MGs form the structure of NMCs, a NMG central controller (NMGCC) can
be considered to take the task of energy management and the highest priority of
hierarchical control and coordinate with all existing MGCCs. As stated (mentioned)
in Sect. 6.3, in control systems, control is performed in millisecond up to a few
minutes. The EMS manages three levels of control during this period. At the first and
second levels, voltage and frequency control, power balance, and load management
are performed. But at the third level, EMS has the task of feeding critical loads
under any circumstances by optimally managing power transmission with other
DGs, distribution networks, feeders, or neighboring MGs. The time framework of
the EMS is shown in Fig. 2.10a. If the EMS is also used in a distributed control
system, according to Sect. 6.4, functions such as V/f control, P/Q, and P/V control,
power distribution, frequency synchronization, and optimal energy management at
MG level are done fully distributed. Then, in the networked mode of the MGs,
to achieve optimal performance across the system, the information is shared with
all local controllers in each MG as well as with the DSO. This is the task of the
NMGCC.
In fact, in systems with NMCs, another control layer called NMGCC is added to
communicate with MGCC, DMS, and tie-line control and optimize the network in
this case. This framework is shown in Fig. 2.10b.

2.7.1 Energy Management Strategies of NMCs

Achieving optimal economic dispatch and making the best use of DERs are the main
objectives of NMCs. To achieve these objectives, the EMS must operate properly at
distribution and MG levels. The major objective of EMS at the DS level is improving
overall economic performance and power quality of the entire system. Also, the
main purpose of EMS at the MGs level is monitoring and improving power balance
performance, optimizing the use of ESSs, and controlling effectively V and F across
all MGs. Figure 2.11 shows the overall performance of EMS in NMCs [1].
In NMCs, the EMS maintains the power supply for each MG. An excess power
generated in the MGs is stored in a storage or delivered directly to the DS or adjacent
MGs via EMS coordination at the distribution level. Similarly, if one MG has a
lack of power, the EMS system supplies the required power to the MG through the
distribution network or adjacent MGs. It is noteworthy that the MG controller, due
to economic considerations, can buy its requested power deficit from the distribution
network or other MGs, or it can regard buying this lack of power by using the DR
program and eliminating the load. Therefore, according to economic purposes, EMS
38 M. Moradi and A. Akbari Foroud

a
Several minutes - Day Several Seconds- Minute Millisecond - Second

CO2 Emission Electricity Load Fast-Dynamic

limit Market Management Storage Availability

Tie- Line Converter(s),

Dispatchable DG(s),

Switch(es)
Unit Economic
Power Balancing
Commitment Dispatch

Power
Lond RES Production prediction Energy Storage
Capability
Forecasting Availability of RE-DG

Long Term Short Term

b
DMS
Tie – Line
Control

Network MG
Central Controller
MGCC MGCC

Local Local Local Local

Control Control Control Control
MGCC

MG1 MG3

DG Hybrid PV –
Control Bat Control
Power Transfer Information
MG2

Fig. 2.10 (a) Time framework of the energy management system, (b) NMGCC location in the
control system
2 Operation Management of Microgrid Clusters 39

Main Grid

EMS
Loads (a) Economic power dispatch DGs
(b) Power quality monitoring

MG1 P1 Q1 PN QN MGN
EMS1 MG1 to MGN
EMSN
(a) Voltage-frequency control (a) Voltage-frequency control
(b) Load generation monitoring (b) Load generation monitoring
(c) Energy-storage management (c) Energy-storage management

Loads Generators Storage Loads Generators Storage

Fig. 2.11 Overall performance of EMS in NMCs

can maintain the power supply of the entire system. This optimal economic dispatch
of the entire system is done by the EMS at regular intervals and the EMS can change
MGs to the island mode, grid-connected, or NMCs depending on the calculated
optimal solution.
In an emergency, if the connection is separated from the main grid, all MGs will
be connected to the network throughout PCC by predetermined operation done by
the EMS. In this condition, only the critical loads in each MG are supplied and
the extra power is sent to the DS. If power generation is not enough to supply all
the existing loads in the DS, only the critical loads of the distribution system are
supplied by the ESSs of MGs. This will allow the EMS to handle critical loads of
MG for more hours. Of course, different scenarios can be programmed to provide
power depending on predetermined conditions and rules. Different types of EMS
strategies in NMCs are discussed in the following.

2.7.2 Compare EMS Structures

In recent years, various types of EMS have been studied and used in various
MGs researches. EMS can be divided into three general categories. Centralized,
decentralized (distributed), and hybrid EMS. In a centralized structure, all MGs
are controlled through a single management system. This system reduces the total
operating cost by preventing the removal of critical loads under any possible
conditions. The system has a simple structure and in island mode, it has acceptable
reliability. Because of requiring high communication infrastructure, this structure
40 M. Moradi and A. Akbari Foroud

can impose heavy costs on the system. It also has less resiliency in faults and is more
useful in slow-changing situations. In a decentralized EMS, each MG has its local
control center that can operate independently [2]. Each MG balances its generation
and loads through power sharing with the distribution network or other adjacent
MGs. Compared to centralized EMS, this type of EMS can work properly in NMCs
mode. But this structure, in grid-connected structure mode, is highly dependent
on the main network. This increases uneconomical operating costs. Therefore, the
decentralized or distributed EMS structure in the island mode of MGs or grid-
connected mode is not very useful and resilient. According to the advantages and
disadvantages of both energy management systems, a hybrid management system
can solve the problems of centralized and decentralized structures. Various hybrid
systems have been proposed in different references.
The hybrid EMS structure is usually such that local resources in each MG are
optimized and the central EMS modifies the extra or lack of power in each MG. This
hybrid system significantly reduces operating costs compared to the decentralized
EMS strategy. However, these hybrid systems also have problems in the island
mode. In these EMSs, because of the parallel performance of the hybrid system in
the MGs (optimize the MGs in parallel) and since the MGs are unaware of the local
data of the other MGs, it may not achieve the economic benefit of the entire network.
So, especially when MGs are used as NMCs, the hybrid EMS structure must receive
all local data of MGS and manage the system according to their analysis. In this
regard, the EMS structure based on NMCs provides an opportunity for all MGs to
become aware of data from other MGs such as power generation level, required
load for customers, and the required amount of buying or selling of extra or lack
of power. Thus, in the hybrid EMSs used in NMCs, each MG is responsible not
only for balancing its power but also for providing optimal economic performance
in the interaction of energy with other MGs. In this case, slight changes in the
performance of each MG can make a big difference in the power generation of
other MGs. Accordingly, different hybrid EMS strategies are observed in different
references. Various EMS strategies are shown in Fig. 2.12. A summary of studies
done on EMS strategies is shown in Table 2.3. Also, the merits and demerits of EMS
types are shown in Table 2.4 [26].

2.7.3 Overview of Energy Management Modeling and Solution

Methods in NMCs

As shown in Fig. 2.13a, several cases can be included in the EMS problem
formulation in NMCs. Usually, in articles, the type of EMS strategy and selected
control and items shown in Fig. 2.13a, are added to the problem formulation based
on the studied case. Then, based on the type of created problem, and according to the
objective functions and constraints are shown in Fig. 2.13b and c, the appropriate
solution is chosen. The methods and solutions used for problems with MGs are
shown in Fig. 2.13d.
2 Operation Management of Microgrid Clusters 41

Centralized Decentralized
Central External
EMS EMS

Main Grid

Central External
EMS EMS

Hybrid EMS

Fig. 2.12 Various EMS strategies

2.8 Objective Functions Formulation

In the following, one example of a simple formulation of objectivefunctions and

constraints in the EMS problem in NMCs is shown. In this formulation, a model is
presented for the operation of NMCs that are connected to each other and fed loads.
In this model, it is possible to communicate several MGs continuously, and one MG
can buy a lack of its power from other MGs or sell its extra power to other MGs. This
will improve the reliability of the MGs and their better stability. On the other hand,
the addition of renewable units, which have uncertain and imbalance nature, cause
problems such as load mismatch and voltage instability in using of MGs. Therefore,
the simultaneous use of heat and renewable units together contributes greatly to
reducing these problems. Therefore, units such as energy storage systems and
batteries can overcome this uncertainty in scheduling. But due to the high level of
uncertainty caused by wind and solar units, the reservation issue is also considered
in situations where the forecasting power generation of these units does not match
Table 2.3 Summary of studies on EMS strategies
42

EMS type Optimization objective Optimization algorithm Major considerations Limitations

Centralized Cost minimization Particle swarm optimization 1. Load and renewable power 1. Increase in the computational
uncertainties burden of central EMS
2. Only grid-connected mode 2. Failure to preserve customer
privacy of microgrids
3. Requirement of extensive
communication setup
Cost minimization and Imperialistic competitive 1. Load-generation uncertainties
reliability maximization algorithm 2. Islanded and grid-connected
mode
System reliability maximization Mixed integer linear 1. Unbalanced systems with
programming (MILP) detailed modeling
2. Islanded mode only
Centralized Cost minimization 1. Distribution of computational 1. Failure to preserve customer
with sequential load privacy of microgrids
operation 2. Grid-connected mode only 2. Requirement of extensive
communication setup
Cost minimization Sequential quadratic 1. Three-level control and
programming demand bidding
2. Grid-connected mode only
Decentralized Cost minimization Multi-agent system with 1. Autonomous operation and 1. Unawareness of system-level
contract net protocol privacy of customer resources
2. Grid-connected mode only 2. Due to individual objectives,
equilibrium may exist and
further optimization may not be
possible
Minimization of load shedding 1. Competitive environment
among suppliers
2. Islanded mode only
(continued)
M. Moradi and A. Akbari Foroud
Table 2.3 (continued)
EMS type Optimization objective Optimization algorithm Major considerations Limitations
Cost minimization and Stochastic optimization 1. Two-layered stochastic model and
reliability maximization uncertainties
2. Islanded and grid-connected modes
Supply adequacy maximization Average consensus algorithm 1. Autonomous operation and
self-healing
. Islanded and grid-connected modes
Hybrid Cost minimization and MILP 1. Three-level control and dynamic 1. Parallel operation of
reliability maximization conditions microgrids
2. Islanded and grid-connected modes 2. Single-level privacy of
customers, easy to reveal
. Reduction in the resiliency
of disconnected microgrids
4. Central EMS failure results
2 Operation Management of Microgrid Clusters

in autonomous operation
Cost minimization 1. Control of voltage, frequency, and
power
2. Islanded and grid-connected modes
Cost minimization and 1. Hybrid microgrids and diversity
reliability maximization gain
2. Islanded and grid-connected modes
Cost minimization 1. Load smoothening and cooperative
operation
2. Islanded mode only
Nested Cost minimization 1. Hybrid MGs 1. Upper operation cost vs.
hybrid Privacy maximization 2. Island, grid-connected and NMCs centralized EMSs
3. Trade-off between various 2. Lesser flexibility
contradictory objectives such as
operation cost, privacy of the
customer, and network resiliency
43
44 M. Moradi and A. Akbari Foroud

Table 2.4 Merits and demerits of EMS types

EMS type Merits Demerits
Centralized 1. Minimized operation cost with 1. Extensive communication
global optimization. infrastructure with powerful central
2. Efficient components of each MG is EMS is needed.
used. 2. Weak plug-and-play capability.
3. Standardize and implement are easy. 3. Without ensuring preserve the
4. External trading is reduced. privacy of customer
5. Higher reliability in islanded mode. 4. After a small change in the system
onerous testing is needed.
5. Lower flexibility, adaptability, and
propagation of forecasting errors
Decentralized 1. Preserve the privacy of the customer. 1. Increased operation cost
2. Ensures plug-and-play flexibility. 2. The resiliency in islanded mode is
3. Distribution of computational load. decreased;
4. Influence of forecasting errors on 3. Excessive power trading with utility
the MG level only. grid in grid-connected mode
Hybrid 1. Preserve single level the privacy of 1. Parallel operation of MGs, unaware
the customer. of other MGs.
2. Higher flexibility vs. centralized 2. Easy to disclose the consumers’
EMSs. privacy due to single-level privacy.
3. Ensure plug-and-play flexibility. 3. Resilient performance of
4. Distribution of computational load. disconnected MGs is decreased.
5. Lesser operation cost vs. 4. If central EMS is compromised, all
decentralized EMSs. MGs will operate in an autonomously
decentralized mode.

with the actual value. The scheduling problem can be considered with the set of
objectives shown in Fig. 2.13b and optimized by one of the methods shown in Fig.
2.13d. For example, objective functions can be considered to minimize operating
costs connected to grid and island mode, minimize the costs due to environmental
pollutant emissions, and optimally exchange of power between MGs and with the
distribution network. It is noteworthy that the formulation and solution method
vary depending on the selected objective function and different constraints. In this
formulation, we can consider the uncertainties of wind and solar power generation
units and loads [30].
 
MinOF = Cost t,m + Emissiont,m ∀t ∈ T , ∀m ∈ {MG1 , . . . , MGm }
t
(2.1)
 Formulation for EMS in NMGs

Demand Side Modeling of power Modeling of power

Energy Market Generation Units
Load Modeling ESS Modeling Management exchange between exchange betwwen
Modeling Modeling
Modeling MGs MGs & DS

(a)
2 Operation Management of Microgrid Clusters

Main objectives in EMS

Maximized Minimized

Resiliency Privacy of MGs Reliability Emission Cost ESS Cost Load Shedding Operation Cost Reserve Cost

(b)
Fig. 2.13 Overview of EMS modeling and solution methods in NMCs. (a) EMS problem formulation in NMCs, (b) EMS objective functions in NMCs, (c)
45

EMS constraints in NMCs, and (d) EMS solution method in NMCs

 46

Constraints in NMGs EMS

Max & Min

Reserve Power Balance Demand Operation Storage Carbon Emission Price Self efficiency
generation unit

(c)

NMGs EMS Solution Method

Multi-Agent Linear and Non-

Decomposition Heuristic Dynamic
Stochastic & Robust Artificial Meta-Heuristic Model Predictive System Linear
Methods Methods Programming
Programming Intelligence Methods Control Programming Programming

Two-level
ADMM Game Theory Priority-based MILP
Optimization

Chance Markov Scenario-

Benders MINLP
Constraint Decision Based

MIQP

(d)
Fig. 2.13 (continued)
M. Moradi and A. Akbari Foroud
2 Operation Management of Microgrid Clusters 47

2.8.1 Cost Operation Modeling

The operation cost of the MGs includes the fuel cost of the generating units, the cost
of energy storage units, and the cost of power exchanged between the MGs and the
main grid. This function is formulated in Eq. (2.2):

 T  T ⎧Ng,m
     ⎨
F Pi,m = Cost t,m = PGi,m (t)BGi,m (t)
m
⎩
t=1 t=1 i=1
⎫

Ns,m ⎬
+ PSj,m (t)BSj,m (t) +PGrid,m (t)BGrid,m (t)
⎭
j =1

∀m ∈ {MG1 , . . . , MGm }
(2.2)

where F(Pi ) is the total operating cost (in term of $), Trepresents the total studied
hours,Ng ,NS represent the number of generating and storing units, respectively.
PGi (t) andPSj (t)are output power of ith generation unit and ith storage unit in the
time, respectively.BGi (t)and BSj (t)are the proposed energy price of ith generation
unit and ith storage unit in the time, respectively. PGrid (t) and BGrid (t) are the amount
of exchanging power with the main grid and its proposed price at the time t and the
index m represents the desired MG.

2.8.2 Pollution Emission Modeling

Pollution emission level is proportional to the output power of units. The formula-
tion of this function is shown in Eq. (2.3):
⎧
T ⎨N
  T  g,m

C Pi,m = Emissiont,m = PGi,m (t)EGi,m (t)

⎩
t=1 t=1 i=1
⎫

Ns,m ⎬
+ PSj,m (t)ESj,m (t) + PGrid,m (t)EGrid,m (t)
⎭
j =1

∀m ∈ {MG1 , . . . , MGm }
(2.3)
48 M. Moradi and A. Akbari Foroud

In Eq. (2.3), EGi (t) ESj (t) and EGrid (t) indicate the amount of pollution caused
by the ith generation unit, jth storage,
 and the global network in time t, respectively,
Kg
and they are in terms of MW h .

2.8.3 Problem Constraints

2.8.3.1 Power Balance Constraint

In the studied MGs, the sum of the power of generation units, storages, and the main
distribution network should be equal to the required load of the network.
⎛ ⎞
Ng,m
  
Ns,m
⎝ PGi,m (t) + PSj,m (t)⎠ + PGrid (t) = PL (t) (2.4)
m i=1 i=1

where p is the load demanded by the system at time t. (In the case of load response
programs or shedding of the load, this limitation is violated. However, the penalty
coefficient can be considered a penalty of this limitation).

2.8.3.2 Generation Capacity Limit

All generation units in the MG, storages, and distribution networks have their own
generation power limit. These limitations are described in Eq. (2.5).

PGi,min (t) ≤ PGi (t) ≤ PGi,max (t)

PSj,min (t) ≤ PSj (t) ≤ PSj,max (t) (2.5)
PGrid,min (t) ≤ PGrid (t) ≤ PGrid,max (t)

2.8.3.3 Charge and Discharge Rate Limit

Batteries as a storage system have limitations such as charging and discharging rate
and stored energy that is given in Eq. (2.6).

SoCsj (t) = SoC sj (t − 1) + Pchg/Dchg (t)

  (2.6)
0 ≤ Pchg/Dchgj (t) ≤ PCDSj ,max

where SoCsj (t) and SoCsj (t − 1) denote the amount of charge storage at the present
time and the hour ago. Pchg/Dchg (t)displays the amount of charge during tth hour and
PCDSj ,max denotes the maximum charge rate.
2 Operation Management of Microgrid Clusters 49

2.8.3.4 Self-Sufficiency Constraint

NMCs, under certain conditions, may be disconnected from the distribution network
and operated as an island. Under these conditions, operator decisions and planning
must be such that the distribution generation resources can feed the loads on the MG
completely without interruption. In the proposed model, self-sufficiency is added to
the optimization operation as an index and it has been discussed. In the following
equation, the influence of MG self-sufficiency index on problem formulation is
expressed mathematically in Eq. (2.7).

max
PM,t = (1 − SSI) Dd,t (2.7)
d

In Eq. (2.7), SSI is self-sufficiency index, PM,t max is the capacity of MG lines for

power transmission and Dd, t is the amount of power demand associated with the
load d at time t. It is found in this equation that if the SI index is greater, the capacity
of the lines will be lower and MG can receive less power from the distribution
network. As a result, it is more self-sufficient.

2.8.3.5 Reserve Constraint

In MGs, both loads and distributed generation resources have uncertainty. In a

comprehensive (general) model for 24-hour scheduling of MG, some power must be
considered as a reservation to overcome these faults. A probable index called PSS
is considered for reservation modeling. This index indicates the probable capability
of MG to feed existing loads. To introduce the PSS index, it is assumed that the
prediction fault of the load consumption power and the generation power of the
scattered generation resources
 are modeled asnormal distributed
 variables. On the
other hand, Δd ∼ N μed , σed 2 andΔ ∼ N μ , σ 2 . Where, μ
w ew ew ew and μed are
the average generation power of wind units and the consumption power of loads,
2 and σ 2 are the standard deviation of the generation power of wind
respectively. σew ed
units and the consumption power of loads, respectively. N represents the normal
distribution of a probable variable. However, the required reservation of MG to
overcome the prediction error can be added to problem formulation as one of the
optimization constraints. The mathematical form of these explanations is shown in
the Eq. (2.8).

     
f  −1
Pi,t ≥ Di,t −pwt − (μew − μed ) + 2 σew +σed × erf (1−2P SS)
2 2

i d
(2.8)
50 M. Moradi and A. Akbari Foroud

In Eq. (2.8), Pi, t is the output power of ith unit at time t, Di, t is the predicted
f
consumption power in MG, pwt is the predicted generation power at time t, erf is
the error function and PSS is a probable index used in reservation modeling.

2.8.4 Demand Response Pogromming (DRP)

Economic models that are presented for the DR are trying to measure demand from
parameters that are important from a marketing point of view [34]. This is based
on microeconomic theories. Economic models are derived from data gained from
actual experience and are used to evaluate DRPs. The elasticity of the demand is
defined as the demand sensitivity to the price:

ρ0 ∂d new
E= × (2.9)
do ∂ρ new

where, ρ 0 is an initial energy price regardless of DRP and do is the initial demand for
energy before considering and applying DRP. Also, dnew is the amount of demand
after applying DRP and ρ new is the energy price after applying the charge.
According to Eq. (2.9), this demand elasticity proportion to price can be divided
into two parts:
1. The sensitivity of a consumer’s demand at a given time proportion to consumer
energy price changes at the same time elasticity.
2. The sensitivity of a consumer’s demand at a given time proportion to energy
price change at the other times (mutual elasticity).
The self-elasticity and mutual elasticity can be written mathematically in the
form of Eq. (2.10) and Eq. (2.11), respectively. Self and mutual elasticity are also
considered to be −0.2 and 0.01, respectively.

ρ0 (i) ∂d new (i)

E (i, i) = × (2.10)
do (i) ∂ρ new (i)

ρ0 (j ) ∂d new (i)
E (i, j ) = × (2.11)
do (i) ∂ρ new (j )

It can be noted that self-elasticity and mutual elasticity have negative and positive
values, respectively.

E (i, j ) ≤ 0 i = j, ∀i, j ∈ T
(2.12)
E (i, j ) ≥ 0 i = j, ∀i, j ∈ T
2 Operation Management of Microgrid Clusters 51

According to the demand response relative to the price change, two models can
be presented:
• One-time model.
• Multi-time model.
In a one-time model, the loads are only sensitive to the price changes at the
same time and so, do not have the ability to transfer to another time. In this model,
whenever the price of energy changes, the loads react to it. Therefore, only self-
elasticity is involved in this model. The economic one-time model will be as follows:
 
ρnew (i) − ρ0 (i)
dnew (i) = d0 (i) × 1 + E (i, i) ∀i ∈ T (2.13)
ρ0 (i)

In the multi-time model, the mutual elasticity which is mentioned previously is

used. According to this model, the demand changes at a special hour depends not
only on the price change at that time but also on price change at the other hours. If
we want to see the effect of encouragement and penalty on determining the level of
demand we have:


24
d0 (i)
d(i) = d0 (i) + E (i, j ) [ρ(j ) − ρnew (i)] (2.14)
ρ0 (i)
i=1
i = j

By combining the model of one time and multi times, a comprehensive economic
model of the load is obtained [34].
⎧ ⎫
⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
⎨ ρnew (i)−ρ0 (i)  24
1 ⎬
d (i)= d0 (i) 1+E (i, i) + E (i, j ) [ρ(j )−ρnew (i)]
⎪
⎪ ρ0 (i) ρ0 (i) ⎪
⎪
⎪
⎪ ⎪
⎪
⎪
⎪ i=1 ⎪
⎪
⎩ ⎭
i = j
(2.15)

The Eq. (2.15) shows that a consumer needs to adjust his consumption at any
time of the day to get the most profit.

2.8.5 Modeling of Generation Units in the MGs

In this section, all units in the MGs are modeled mathematically. These units include
fuel cell units, micro-turbines, and CHP units.
52 M. Moradi and A. Akbari Foroud

2.8.5.1 Fuel Cell Model

The cost of fuel for each fuel cell is calculated as a function of its efficiency. The fuel
cell efficiency is related to its operating point and is referenced in terms of output
power ratio to the input energy capacity in natural gas, which must be in the same
unit (W). Fuel cost for each fuel cells is defined as follows:
 PJ
CF C = Cnl (2.16)
ηJ
J

where Cnl is the price of natural gas for feeding fuel cell, PJ is the active power
generated at the time interval J, ηJ is the efficiency of the cell at the time interval J
and CFC is the cost of fuel consumption of fuel cells.

2.8.5.2 Micro-Turbines Model

Unlike fuel cells, the efficiency of the micro-turbine increases with increasing
operating power. The overall and electrical efficiency of the micro-turbine is
expressed as:

PJ + Qdh
ηtot = (2.17)
Qf uel

PJ
ηJ = (2.18)
Qf uel

where, PJ is pure generated electric power, Qdh is the generated heat of each units
and Qfuel is input fuel energy. According to Eq. (2.19), the cost of MTs fuel is
calculated as follows:
 PJ
CMT = Cnl (2.19)
ηlJ
J

where PJ is pure generated electric power at the time interval J, Cnl is the cost of
natural gas for feeding the MT and CMT is the cost of fuel consumption of MT.

2.8.5.3 Combined Heat and Power Model

The large-scale old generators have an efficiency of about 35%. While combined
heat and power (CHP) units are able to increase the efficiency of micro-turbines
to 80–85%. Without considering CHP units, MTs will have less efficiency than
2 Operation Management of Microgrid Clusters 53

traditional power generation, and the operation of MGs is not economically feasible.
The fuel cost of CHPs is as follows:
  t 
P − ηe
rec ηCH t
CCH P = CMT × 1 − (2.20)
ηb

where,  rec is the heat recovery factor in CHP and is assumed 0.95. Also, ηb and ηet
are assumed 0.8 and 0.35, respectively.

2.9 Numerical Result

In this section, three scenarios are assumed. To calculate the costs and emission,
all of the required parameters are shown in Table 2.5 and Table 2.6. Also, the
amount of loads and the TOU method pertinent prices are given in Table 2.7. As
mentioned earlier, NMCs should be able to perform well in different operational
conditions, such as island mode. For this purpose, three scenarios with different SSI
indicators have been considered. The transfer power limit between MGs and the
upstream network is assumed 300 kW. In general, as the SSI increases, the amount
of transmission power purchased between the MGs and the upstream network
decreases. The reason for the increase in the cost of operation is that considering the
SSI, the capacity of the lines between the MGs is limited and as a result, the volume
of energy exchanges between the MGs is more limited. Thus, an MG has to use more
expensive units to meet its power deficit, and therefore increase operating costs.
Also, the DGs of each MG must generate extra power to prevent shedding of the
load. Figures 2.14, 2.15 and 2.16 (includes a, b, and c) show the effect of increasing
the SSI on unit commitment and battery charge/discharge status per MG (includes 1,
2, and 3). The comparison of energy scheduling costs in different scenarios is shown
in Table 2.8. As shown in Table 2.8, with the increase of SSI and the decrease in
the transfer power purchased from the upstream network, the cost of reservation has
also increased. As the output power of DGs increases, the uncertainty of the system,
and the number of required reservations to overcome the prediction error will be
increased. To demonstrate the effect of the SSI on unit scheduling, the generation of
WT and PV units are obtained in the first scenario and are not changed for the rest
of the scenarios. As the price of power changes throughout the day, the amount of
demand reacts to price changes per hour. Also, by discharging the batteries during
peak hours, the purchase of additional energy is prevented. In fact, batteries are very
effective to reduce the dependence of MGs on the upstream network and improve
the reliability and resilience of the network in the critical event.
54 M. Moradi and A. Akbari Foroud

Table 2.5 Required parameters to calculate the operation costs of MGs

MG No. MG 1 MG 2 MG 3
Unit type CHP PV WT FC MT WT CHP PV FC
Operation and 0.088 0.1095 0.1095 0.016 0.088 0.1095 0.088 0.1095 0.016
maintenance cost
(λOM )
Efficiency (η) 0.85 – – 0.6 0.8 – 0.85 – 0.6
Min power 50 0 0 30 50 0 50 0 30
generation (kw)
Max power 500 150 250 250 500 250 500 150 250
generation (kw)
Sold power costs d1 = 0.14 d2 = 0.13 d3 = 0.12
($/kwh)
Purchased power C1 = 0.18 C2 = 0.17 C3 = 0.17
costs ($/kwh)

Table 2.6 Required parameters to calculate emissions costs

           
kg kg kg kg kg
Gas Type γ kg$
ρMT kwh ρCH P kwh ρF C kwh ρP V kwh ρW T kwh
NOx 10.0714 0.0001 0.00003 0.00044 0 0
SO2 2.3747 0.000007 0.000006 0.000008 0 0
CO2 0.0336 0.00137  0.001078
 0.001596 0   0
Natural gas low-hot value: L = 9.7 KW h
m3
, natural gas price: Cnl = 0.76 $
m3

Table 2.7 Mean value of MGs load and consumption prices based on TOU
Load (Kw) Load (Kw)
Hours MG1 MG2 MG3 Price ($/KWh) Hours MG1 MG2 MG3 Price ($/KWh)
1 100.36 77.87 111.08 0.023 13 143.89 144.77 246 0.034
2 87.13 64.17 108.31 0.023 14 174.59 117.35 204.46 0.034
3 83.55 53.42 90.09 0.023 15 146.8 113.31 190.61 0.034
4 105.25 74.16 134.2 0.023 16 172.82 159.38 207.45 0.034
5 73.5 55.15 100.48 0.023 17 258.05 252.04 402.02 0.04
6 72.16 57.67 93.78 0.023 18 318.7 264.33 375.03 0.04
7 137.35 94.58 148.98 0.023 19 367.4 268.67 509.36 0.04
8 171.8 133.41 189.42 0.034 20 372.66 222.92 373.83 0.04
9 231.08 172.55 319.1 0.04 21 294.51 251.99 381.99 0.04
10 266.28 218.85 342.73 0.04 22 238.72 223.91 340.36 0.04
11 148.19 128.19 215.65 0.034 23 214.65 159.97 276.6 0.04
12 164.7 125.97 214.86 0.034 24 135.92 137.86 223.7 0.023

2.10 Conclusion

The NMCs are cost-effective and beneficial for both owners of resource and energy
consumers. In this chapter, the 24-hour scheduling of the NMCs, consisting of
several DGs and loads has been discussed. This programming is an optimization
2 Operation Management of Microgrid Clusters 55

250 MG1
400
Demand
Generated Power
200 350
Power(kw) in MG1

150 300

Power (KW)
250
100
200
50
150
0
100
CHP Wind PV Battery
-50 50
0 5 10 15 20 25 0 5 10 15 20 25
Hour Hour
(a)
70 MG2
300
Demand
60 Generated

50 250
Power(kw) in MG2

40

Power (KW)
30 200
20
10 150
0
-10 100
-20
FC Wind MT Battery 50
-30 0 5 10 15 20 25
0 5 10 15 20 25
Hour Hour
(b)
300 MG3
550
Demand
250 500 Generated
Power(kw) in MG3

450
200 400
Power (KW)

150 350
300
100 250
50 200
150
0 100
FC CHP PV Battery 50
-50 0 5 10 15 20 25
0 5 10 15 20 25
Hour (c) Hour

Fig. 2.14 Units generation scheduling with considering of DRPs and SSI = 0 in (a) MG1 (b)
MG2 (c) MG3

250 MG1
400
Demand
Generated
200 350
Power(kw) in MG1

150 300
Power (KW)

250
100
200
50
150

0 100
CHP Wind PV Battery
-50 50
0 5 10 15 20 25 0 5 10 15 20 25
Hour Hour
70
(a)
MG2
60 300
Demand
Generated
50
Power(kw) in MG2

250
40
30
Power (KW)

200
20
10
150
0
-10 100
-20
FC Wind MT Battery
-30 50
0 5 10 15 20 25 0 5 10 15 20 25
Hour (b) Hour

300 MG3
550
Demand
250 500 Generated
Power(kw) in MG3

450
200
400
Power (KW)

150 350
300
100
250
50 200
150
0
100
FC CHP PV Battery
-50 50
0 5 10 15 20 25 0 5 10 15 20 25
Hour (c) Hour

Fig. 2.15 Units generation scheduling in with considering of DRPs and SSI = 0.5 in (a) MG1 (b)
MG2 (c) MG3
56 M. Moradi and A. Akbari Foroud

250 MG1
400
Demand
Generated
200 350
Power(kw) in MG1

300
150

Power (KW)
250
100
200
50
150
0 100
CHP Wind PV battery
-50 50
0 5 10 15 20 25 0 5 10 15 20 25
Hour (a) Hour
MG2
80 300
Demand
Generated
60
Power(kw) in MG2

250
40

Power (KW)
200
20
150
0

-20 100
FC Wind MT battery
-40 50
0 5 10 15 20 25 0 5 10 15 20 25
Hour (b) Hour

300 MG3
550
Demand
500 Generated
250
450
Power(kw) in MG3

200 400
Power (KW)

350
150
300
100 250
200
50
150
0 100
FC CHP PV battery
50
-50 0 5 10 15 20 25
0 5 10 15 20 25
Hour (c) Hour

Fig. 2.16 Units generation scheduling with considering of DRPs and SSI = 0.7 in (a) MG1 (b)
MG2 (c) MG3

problem that contains two objective functions, operating costs, and the emission
cost. Also, SSI, reservation constraint, and active and reactivepower exchange have
been considered in the optimization problem. In this programming, a structure
for the operation of NMCs is described, that structure provides the possibility of
exchanging power between each MG, and MGs with the upstream distribution
network. The uncertainty of the wind generation and solar units is investigated using
the probabilistic distribution function of the Monte Carlo method. To overcome
the difference caused by uncertainty, the required reserve is generated and the
cost of it is added to the objective function. Also, the DRP based on time of use
(TOU) is employed due to that customers tend to reduce some of their power
consumption during peak hours and transfer some of their power consumption
to off-peak hours. On the other hand, the SSI has been added to the scheduling
constraint to increase the reliability of the MGs, especially in islanding mode. Also,
the limitations of the operating systems are considered in all operating conditions
based on the uncertainty of units and DGs and load variation (in terms of quantity,
power factor, etc.). In addition, the ability to exchange power between any MG with
the others is possible. In the networked mode, a local EMS located on each of the
MGs optimizes the DERs in the MGs according to the information of other MGs
and it can generate excessive power requirements. So, it can compensate for the
power shortages of the other MGs. This is done in coordination with the distribution
network operator and the external EMS. It is necessary to note that the priority of
any MGs is to balance the power and economic operation to manage its resources. In
Table 2.8 Comparison of energy scheduling costsin different scenarios
Purchased Power cost ($) Sold power cost ($)
Scenario No. SSI Reserve cost ($) Operation cost ($) Emission cost ($) MG1 MG2 MG3 MG1 MG2 MG3 Total cost ($)
2 Operation Management of Microgrid Clusters

1 0 379.681 1155.794 33.772 −27.36 −45.99 −108.43 318.970 70.094 229.791 1132.181
2 0.5 390.037 1174.379 32.338 −32.64 −45.60 −97.60 267.245 69.648 198.245 1237.468
3 0.7 394.073 1225.997 30.184 −35.91 −44.22 −128.08 217.616 70.255 163.304 1407.305
57
58 M. Moradi and A. Akbari Foroud

the next step, if the MG cannot be able to feed its demands, it should buy its energy
shortage from the other MGs. In order to examine the uncertainty of loads, normal
distribution function with an average value obtained from DRP, and the standard
deviation value equals to 0.3 are considered. The simulation results are validated by
MATLAB software and with the PSO algorithm. The simulation results show that
NMCs can prevent load shedding by reducing the power dependence by exchanging
power between MGs. This will increase resilience and reliability in NMCs.

References

1. Alam, M. N., Chakrabarti, S., & Ghosh, A. (2019). Networked microgrids: State-of-the-art and
future perspectives. IEEE Transactions on Industrial Informatics, 15(3), 1238–1250.
2. Tang, C., Liu, M., Dai, Y., Wang, Z., & Xie, M. (2019). Decentralized saddle-point dynamics
solution for optimal power flow of distribution systems with multi-microgrids. Applied Energy,
252.
3. Liu, G., Starke, M. R., Ollis, B., & Xue, Y. (2016). Networked Microgrids Scoping Study. Oak
Ridge National Laboratory (ORNL/TM-2016/294).
4. Backhaus, S., Dobriansky, L., et al. (2016). Networked microgrids scoping study. Los Alamos
National Laboratory.
5. Hussain, A., Bui, V., & Kim, H. (2019). Resilience-oriented optimal operation of networked
hybrid microgrids. IEEE Transactions on Smart Grid, 10(1), 204–215.
6. Zamora, R., & Srivastava, A. K. (2018). Multi-layer architecture for voltage and frequency
control in networked microgrids. IEEE Transactions on Smart Grid, 9(3), 2076–2085.
7. Wang, H., & Huang, J. (July). Incentivizing energy trading for interconnected microgrids. IEEE
Transactions on Smart Grid, 9(4), 2647–2657.
8. Jadhav, A. M., & Patne, N. R. (2017). Priority-based energy scheduling in a smart distributed
network with multiple microgrids. IEEE Transactions on Industrial Informatics, 13(6), 3134–
3143.
9. Wang, Z., Chen, B., Wang, J., Begovic, M. M., & Chen, C. (2015). Coordinated energy
Management of Networked Microgrids in distribution systems. IEEE Transactions on Smart
Grid, 6(1), 45–53.
10. Wang, Z., Chen, B., Wang, J., & kim, J. (2016). Decentralized energy management system
for networked microgrids in grid-connected and islanded modes. IEEE Transactions on Smart
Grid, 7(2), 1097–1105.
11. Gao, H., Liu, J., Wang, L., & Wei, Z. (2018). Decentralized energy Management for Networked
Microgrids in future distribution systems. IEEE Transactions on Power Systems, 33(4), 3599–
3610.
12. Ma, W., Wang, J., Gupta, V., & Chen, C. (2018). Distributed energy Management for
Networked Microgrids Using Online ADMM with regret. IEEE Transactions on Smart Grid,
9(2), 847–856.
13. Yuan, W., Wang, J., Qiu, F., Chen, C., Kang, C., & Zeng, B. (2016). Robust optimization-based
resilient distribution network planning against natural disasters. IEEE Transactions on Smart
Grid, 7(6), 2817–2826.
14. Abhinav, S., Modares, H., Lewis, F. L., Ferrese, F., & Davoudi, A. (2018). Synchrony in
networked microgrids under attacks. IEEE Transactions on Smart Grid, 9(6), 6731–6741.
15. Wang, Z., Chen, B., Wang, J., & Chen, C. (2016). Networked microgrids for self-healing power
systems. IEEE Transactions on Smart Grid, 7(1), 310–319.
16. Li, Y., Zhang, P., & Luh, P. B. (2018). Formal analysis of networked microgrids dynamics.
IEEE Transactions on Power Systems, 33(3), 3418–3427.
2 Operation Management of Microgrid Clusters 59

17. Hu, X., & Liu, T. (2017). Co-optimisation for distribution networks with multi-microgrids
based on a two-stage optimisation model with dynamic electricity pricing. IET generation,
Transmission & Distribution, 11(9), 2251–2259.
18. Arif, A., & Wang, Z. (2017). Networked microgrids for service restoration in resilient
distribution systems. IET generation, Transmission & Distribution, 11(14), 3612–3619.
19. Shahnia, F., Chandrasena, R. P. S., Rajakaruna, S., & Ghosh, A. (2014). Primary control
level of parallel distributed energy resources converters in system of multiple interconnected
autonomous microgrids within self-healing networks. IET Generation, Transmission & Distri-
bution, 8(2), 203–222.
20. Golsorkhi, M. S., Hill, D. J., & Karshenas, H. R. (2018). Distributed voltage control and power
Management of Networked Microgrids. IEEE Journal of Emerging and Selected Topics in
Power Electronics, 6(4), 1892–1902.
21. Wang, Z., & Wang, J. (2017). Service restoration based on AMI and networked MGs under
extreme weather events. IET generation, Transmission & Distribution, 11(2), 401–408.
22. Zhang, F., Zhao, H., & Hong, M. (2015). Operation of networked microgrids in a distribution
system. CSEE Journal of Power and Energy Systems, 1(4), 12–21.
23. Tian, P., Xiao, X., Wang, K., & Ding, R. (2016). A hierarchical energy management system
based on hierarchical optimization for microgrid community economic operation. IEEE
Transactions on Smart Grid, 7(5), 2230–2241.
24. Che, L., Shahidehpour, M., Alabdulwahab, A., & Al-Turki, Y. (2015). Hierarchical coordina-
tion of a community microgrid with AC and DC microgrids. IEEE Transactions on Smart Grid,
6(6), 3042–3051.
25. Chiu, W., Sun, H., & Poor, H. V. (2015). A multiobjective approach to Multimicrogrid system
design. IEEE Transactions on Smart Grid, 6(5), 2263–2272.
26. Hussain, A., Bui, V., & Kim, H. (2018). A resilient and privacy-preserving energy management
strategy for networked microgrids. IEEE Transactions on Smart Grid, 9(3), 2127–2139.
27. Cintuglu, M. H., & Mohammed, O. A. (2017). Behavior modeling and auction architecture
of networked microgrids for frequency support. IEEE Transactions on Industrial Informatics,
13(4), 1772–1782.
28. Parisio, A., Wiezorek, C., Kyntaja, T., Elo, J., Strunz, K., & Johansson, K. H. (2017).
Cooperative MPC-based energy management for networked microgrids. IEEE Transactions
on Smart Grid, 8(6), 3066–3074.
29. Mojtahedzadeh, S., Ravadanegh, S. N., & Haghifam, M. (2017). Optimal multiple microgrids
based forming of greenfield distribution network under uncertainty. IET Renewable Power
Generation, 11(7), 1059–1068.
30. Nikmehr, N., Najafi-Ravadanegh, S., & Khodaei, A. (2017). Probabilistic optimal scheduling
of networked microgrids considering time-based demand response programs under uncertainty.
Applied Energy, 198, 267–269.15.
31. Bullich-Massague, E., Díaz-González, F., et al. (2018). Microgrid clustering architectures.
Applied Energy, 212, 340–361.
32. Hu, M., Wang, Y.-W., Xiao, J.-W., & Lin, X. (2019). Multi-energy management with
hierarchical distributed multi-scale strategy for pelagic islanded microgrid clusters. Energy,
185, 910–92115.
33. Antoniadou-Plytaria, K. E., Kouveliotis-Lysikatos, I. N., Georgilakis, P. S., & Hatziargyriou,
N. D. (2017). Distributed and decentralized voltage control of smart distribution networks:
Models, methods, and future research. IEEE Transactions on Smart Grid, 8(6), 2999–3008.
34. Ajoulabadi, A., Ravadanegh, S. N., & Ivatloo, B. M. (2020). Flexible scheduling of reconfig-
urable microgrid-based distribution networks considering demand response program. Energy,
196, 1.
Chapter 3
Energy Management Systems
for Microgrids

Seyed Mohsen Hashemi and Vahid Vahidinasab

3.1 Introduction

Increasing use of the small-scale generation resources and energy storage in the
distribution networks and close to the end users has led to several technical
and economic advantages. Microgrids (MGs) provide a systematic approach for
operating an energy system with these features. A strong Energy management
system (EMS) enables the MG to monitor and control the resources in the time
steps near the real operation time. The EMS of an MG can be introduced as an
integrated software and hardware system that brings a decision support system for
the MGs’ operator in which all of the tasks of monitoring, decision-making, and
controls could be done using it.
Furthermore, it determines how to exchange energy with the main grid consider-
ing the operational constraints and the system conditions such as load consumption,
electricity market price, generation capability of the units, and the stored energy of
the energy storage systems. In addition to the normal operation state, EMS should
be attentive to the critical conditions of the contingencies. For this purpose, security
constraints are considered in the mathematical model of the MG’s operation using
the frequency-related constraints.

S. M. Hashemi · V. Vahidinasab ()

Department of Electrical Engineering, Abbaspour School of Engineering, Shahid Beheshti
University, Tehran, Iran
e-mail: Moh_hashemi@sbu.ac.ir; v_vahidinasab@sbu.ac.ir

© Springer Nature Switzerland AG 2021 61

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_3
62 S. M. Hashemi and V. Vahidinasab

3.2 An Overview of the EMS of the MGs

The EMS of an MG enables different functionalities such as:

• Monitoring of the MG in different conditions.
• Analyzing the system’s condition in different operational states.
• Ability to deal with various threats.
• Making quick decisions in critical situations.
• Performing control actions.
Different input signals are provided for an EMS and are analyzed through its
computational core. The output results or decisions can be used either for the
assessment of system conditions through the graphical reports or be automatically
sent to the different agents as the dispatching orders. EMS has a critical function in
real-time control, where the input data has high variability and, in such conditions,
it would be impossible for manual control to handle the system. As indicated in Fig.
3.1, an MG has multiple interactions with different agents that should be handled by
the EMS.
As previously mentioned, EMS has three tasks of monitoring, decision making,
and control that are indicated in Fig. 3.2. They will be analyzed in the following
subsections.

Fig. 3.1 Structure of a typical MG [1]

3 Energy Management Systems for Microgrids 63

EMS
Decision making system
Assessment tools

Network/component Remedial action

Security assessment State estimation
Monitoring system

modeler system

Control system
Load forecast/ Short circuit Uncertainty
Load flow
estimation calculation assessment

Optimization tools
Network
Unit commitment Economic dispatch Optimal power flow
reconfiguration

Decision making
under uncertainty

Restoration tools

FLISR

MG
Fig. 3.2 General structure of EMS in MG

3.3 Monitoring System

The monitoring system of EMS provides a detailed and comprehensive picture of

the microgrid at any time. This information includes the status and performance
of switches, lines, and transformers. As well as the output power of DGs, the
output power of storages and their state of charge (SOC), available fuel of the
Microturbines (MT), and the information of the reactors and capacitors of the
network. This information is gathered by the meters and sensors and sent to EMS
by the remote terminal units (RTU) through the communication infrastructure.

3.4 Control System

EMS controls the MG in different manners. They are determined considering the
complexity and importance of the MG’s procedures. In some cases, the automatic
control is applied in which EMS directly controls the components. For example, in
the MGs equipped with the self-healing strategies, locating and isolating the fault
and restoring the system are automatically performed. Another control scheme is
64 S. M. Hashemi and V. Vahidinasab

the supervisory control in which the system operator can directly open or close the
switches by the EMS if needed. In cases where there is no remote access to the
components, the manual control is used in which the system operator calls the plant
operator to close or open the switches.

3.5 Decision-Making System

Decision-making can be considered as the most valuable feature of the EMS.

It enables the EMS with the capabilities of the assessment, optimization, and
restoration of the MG. They will be analyzed in the following subsections.

3.5.1 Assessment Tools

Assessment tools enable the operator to find out about the MGs condition in
the current operation regime or any possible operation scenario. These tools are
presented in the following.

3.5.1.1 Network and Component Modeler

An important tool including the mathematical model of the system components’

behaviors. With an extension of the MG by the addition of the new components, it
provides the capability of mathematical modeling.

3.5.1.2 Security Assessment Tool

This tool empowers the system operator to assess the system condition in different
contingencies, in order to identify the most critical contingencies. According to the
low inertia of MGs, this tool considers the dynamic aspects of the contingencies
such as the frequency deviation. Sudden islanding is usually considered as the
contingency of the MGs. Also, in the isolated MGs, EMS considers events such
as the resource outage that affect the dynamic variables of the system.

3.5.1.3 Remedial Action Scheduling Tool

Using this tool, the system operator can deal with the system’s contingencies. It
provides solutions such as system reconfiguration, generation unit startup, and so
on. Some of the demand response programs such as the emergency demand response
3 Energy Management Systems for Microgrids 65

program (EDRP) are valuable resources to provide remedial actions. The response
rate of the mentioned resources is an important factor in MGs.

3.5.1.4 State Estimation

A useful tool to estimate the state variables of the system by the limited number of
metered data.

3.5.1.5 Load Forecast/Estimation

This tool uses the current load consumption, the historical data, and the forecasted
weather condition to determine the load consumption in the coming hours or
minutes. Also, it approximately calculates the load of the downstream buses of the
metered transformers. This capability is based on the load pattern of the different
load groups.

3.5.1.6 Load Flow

Using this tool, the system operator will be able to calculate the line flows and bus
voltage of the network in different configurations or load levels.

3.5.1.7 Short Circuit Calculation

This tool calculates the short circuit current of the circuit breakers and switches and
compares them with their allowed values.

3.5.1.8 Uncertainty Assessment

This tool studies uncertain parameters such as renewable power generation and
proposes an uncertainty model for them. For example, in the form of some of
the scenarios and their probabilities. Moreover, it provides scenario generation and
scenario reduction tools.

3.5.2 Optimization Tools

Decision-making in MGs is a complex process that requires a systematic approach

as there are different resources and strategies to supply the load. Depending on the
operation strategy and considering different economic and technical aspects, the
66 S. M. Hashemi and V. Vahidinasab

MG’s operator requires an optimization tool that provides different applications as

will be expressed in the following subsections.

3.5.2.1 Unit Commitment

An optimization model to determine the online generation units in the operation

horizon. Usually, it considers the economic objectives and includes the technical
properties of different generation units and storages.

3.5.2.2 Economic Dispatch

This optimization module determines the optimal output power of the online
resources using the linear models that can be solved very fast.

3.5.2.3 Optimal Power Flow

An optimization tool with different technical or economic objectives including the

network constraints. Loss minimization, cost minimization, and profit maximization
can be different objectives of the OPF module.

3.5.2.4 Network Reconfiguration Tool

This tool determines the best supplying path of the loads in different viewpoints.
Although the network is designed in a mesh structure this module opens or closes
the switches to achieve a radial structure with low power loss or high resiliency.
The other objectives can be considered for reconfiguration such as sharing the load
between different lines.

3.5.2.5 Decision Making under Uncertainty

This tool is able to provide different decision-making methods considering uncer-

tainty. The system operator applies each of these methods based on the uncertainty
model of the parameters. This module includes different methods of stochastic
programming, robust optimization, information gap decision theory, and so on.
3 Energy Management Systems for Microgrids 67

3.5.3 Restoration Tools

This tool empowers the system operator to deal with system faults. Locating and
isolation of the network faults are two important tasks that can be performed in the
automatic or manual manners. As well as, the interrupted loads should be restored
fast. In the MGs equipped with the smart meters and switches, the process of
finding the fault location, isolation, and system restoration (FLISR) is automatically
performed.

3.6 Interaction with Other Systems

EMS interacts with the other systems in order to perform its mentioned functionali-
ties. These systems are introduced in the following (see Fig. 3.3).

3.6.1 Distribution Management System (DMS)

This system similarly manages the main distribution grid to the EMS. Information
such as the scheduled outage of the microgrid’s connecting feeder can be received
from DMS. In this condition, EMS should prepare the MG for the islanded mode
operation.

Distribution Outage
management management
system system

EMS
Advanced Maintenance
Decision
metering Monitoring Control scheduling
making
infrastructure system

Weather
Bid/offer Electricity
forecasting
interface market
system

Fig. 3.3 Interaction of EMS with other systems

68 S. M. Hashemi and V. Vahidinasab

3.6.2 Advanced Metering Infrastructure (AMI)

AMI interacts with EMS by providing different facilities such as smart meters,
two-way communication infrastructure, and other data gathering and transmission
facilities to interact with EMS in real-time operation.

3.6.3 Outage Management System (OMS)

In OMS, data of the components’ outage and load interruption are collected
and analyzed based on different data resources such as the customer information
system (CIS). The outage’s information such as fault location, duration of the load
interruption, and the number of the affected customers are determined in OMS.

3.6.4 Maintenance Scheduling System

This system schedules the preventive maintenance of the lines, transformers,

generation resources, and the storages, leading to their planned outages. EMS
considers these schedules in its scheduling procedure.

3.6.5 Weather Forecasting System

This system forecasts the data such as environmental temperature and wind speed
that are used in EMS to determine the load consumption and the power generation
of renewable resources in the coming hours or minutes.

3.6.6 Electricity Market

EMS can participate in the electricity market as a buyer or seller to determine its
energy exchange with the main grid. Additionally, EMS can be a player of the
ancillary services market by providing the reserve capacity.
3 Energy Management Systems for Microgrids 69

3.6.7 Bid/Offer Interface

Using this system, generation resources and consumers can declare the supply and
demand curves indicating their cost function or price sensitivity.
Although different aspects of the EMS were previously analyzed, the next
sections will focus on EMS functionality in short-term scheduling of MG, in which
EMS dispatches/commits the existing resources for the next minutes/hours to supply
the load in the most economic manner. Hereinafter, we mean EMS a decision-
making module that considers different technical and economic aspects to schedule
the resources.

3.7 Centralized and Decentralized Energy Management

Generally, an EMS can be implemented in centralized and decentralized ways.

Different local controllers are used in both of these methods. In a centralized
manner, a computational core sends the dispatching commands to the local con-
trollers to operate different components. In return, in a decentralized structure,
the local controllers are intelligent agents analyzing the conditions to dispatch
their associated components. The centralized manner applies optimization proce-
dures using different gathered information to economically operate the MG. Unit
commitment (UC) and economic dispatch (ED) are two popular functions of the
centralized energy management system which are applied in different time scales.
In order to reduce the effect of uncertainties, scheduling should be continuously
renewed by updating the input data. This procedure can be applied to EMS as a
rolling horizon (RH) strategy [2]. There are different decision-making methods for
the implementation of UC and ED. For instance, they can be considered as the
optimization problems containing continuous or discrete variables [3–10]. Also,
the machine learning systems [11, 12], and rule-based systems [13] can be used
as decision-making strategies. Centralized EMS is appropriate for the real-time
application [14] because different parts of MG are monitored and analyzed in a
central agent. In this structure, EMS should be equipped with a powerful computing
unit to process a high amount of input data. Decentralized energy management has a
lower computational burden in which different agents schedules a part of MG. This
method is appropriate for large MGs, where the resources and loads are dispersed
and centralized data gathering is impossible or costly. A multi-agent system (MAS)
is one of the important techniques in decentralized decision making for power
management in different time scales [15, 16]. In MAS, each component is managed
by an agent that makes decisions to achieve its desired goals. The agent uses its
local information and the received data from the other agents. In this method, the
final operation scheme is achieved by passing an iterative process. For applying
MAS to real-time scheduling, the convergence speed is a vital factor [17].
70 S. M. Hashemi and V. Vahidinasab

3.8 The Necessity of EMS in the Scheduling of MG

With the advancement of technology and the rise of the smart grid concept,
MGs have become more and more popular. MG is a small power system that
supplies loads using different distributed energy resources (DER). Besides, there
are advanced monitoring and control facilities. Distinguishing features of the MGs
are as follows:
• These power systems contain different distributed generation (DG) technologies.
• The number of end users and energy resources of the MGs are very lower than
that of the bulk power systems. Therefore, each consumer has a higher impact on
the system’s load, and the system condition is so variable.
• MG’s inertia constant is very low. The bulk power systems containing many
resources and loads have high inertia constant. So, sudden outage of the lines
or generators has no significant impact on the system frequency. However, the
MGs having low inertia are very vulnerable to disturbances.
• Having the radial structure, MGs are so vulnerable against the line outages. Bulk
power systems have a circular structure and supply the load points from different
paths. On the other hand, MGs’ access to the resources is reduced in the line
outage conditions.
• Considering the appropriate access to the end users, MGs are capable to exercise
different control actions based on the load’s properties.
According to the mentioned properties, only a powerful EMS can operate the
MG efficiently and reliably.

3.9 EMS Functions in the MG Scheduling

EMS is the main decision module of the MG scheduling the resources to serve
the loads and exchange the energy with the main grid based on several input
parameters indicating the MG’s condition. Figure 3.4 presents the inputs and outputs
of the EMS. Some of these input data such as network configuration and technical
properties of the resources are about constant during the scheduling horizon. On the
other hand, load, electricity market price, and the renewables’ power generation are
rapidly changing during the operation horizon.
Example 3.1 A simple MG containing two MTs (with the capacity of 10 MW) and
a photovoltaic (PV) unit (with an installed capacity of 5 MW) is depicted in Fig.
3.5. As indicated in Fig. 3.6a, EMS schedules the resources to serve the load during
a 24-hours horizon aiming the minimum operation cost. The operation cost of MTs
is 5 (cent/kWh) and PV’s generation cost is about 0. Also, the hourly price forecast
of the electricity market and the forecasted power generation of the PV unit are
respectively indicated in Fig. 3.6b–c. As can be observed, the PV unit is operated
in its maximum available capacity during the hours 8–21. During the hours 1:00 to
3 Energy Management Systems for Microgrids 71

Input data
Technical properties of
resources

Electricity market price

Optimization module of EMS Scheduling decisions
Load forecast Output power of generation
Operation strategies: resources
Power generation forecast
Cost minimization, Profit maximization
of the renewable resources Charge/discharge power of energy
Constraints: storages
Power generation cost of Technical constraints of components,
the resources Reliability constraints Consumption scheduling of the
controllable loads
Internal network topology

System operation strategy

System operation horizon

Fig. 3.4 The structure of MG scheduling by EMS

Fig. 3.5 A simple MG Main grid

Microgrid

PV MT2 MT1

11:00, all of the consumption demand is supplied by the main grid and both of the
MTs are off. By increasing the electricity market price beyond the 5 cents/kWh after
the 12:00, MTs are started to supply the load.
It was a very simple example showing the function of EMS, and it may not be
so complicated to schedule the MG without the EMS, in this example. However, by
increasing the number of resources (generation and storage units) and consumers,
considering the network constraints and need to fast decisions near the real operation
time, the presence of a strong EMS will be essential. In the following, EMS
functions will be analyzed in the aspects of the planning hierarchy and operation
strategies.
72 S. M. Hashemi and V. Vahidinasab

Fig. 3.6 The inputs and outputs of EMS in Example 3.1: (a) Power generation scheduling. (b)
Generation cost of MTs and the electricity market price. (c) Forecasted power generation of the
PV units

3.9.1 Microgrid’s Hierarchical Scheduling

EMS schedules the MG using forecasted data of the loads and power generation of
the renewable generation units. According to the variable condition of the system
during the operation horizon, results of the hour-ahead scheduling (HAS) such
as 24-hour scheduling of Example 3.1, are not valid for the real-time system
operation. Therefore, EMS should rerun an optimum scheduling problem near the
real operation time, based on the results of the HAS and using the updated input
data. Indeed, a hierarchical scheduling framework is needed that firstly applies
HAS over the hours of the operation horizon and then uses real-time scheduling
(RTS) module for energy management within the small time steps near the real
operation time. Although RTS as the latest operational planning of the system has
a significant role in the system operation procedure it still needs the HAS for the
following reasons:
• Some of the decision variables should be inherently considered in the HAS
problem and they cannot be determined in real-time scheduling. For example,
scheduling of shiftable loads should be performed considering the whole opera-
3 Energy Management Systems for Microgrids 73

HAS1
HAS2
. . . . . .
M G s c h ed u li n g

HASk
by EM S

RTS2,m
RTS1,m

RTSk,m
RTS2,2
RTS1,2

RTSk,2
... ... ...
RTS1,1

RTS2,1

RTSk,1
U p d a tin g th e
fo r ec a s te d

hk+1 h24
input data

h1 h2 h3 hk
... ... ... . . .

Fig. 3.7 Hierarchical scheduling of MG in EMS

tion horizon [18]. Also, the state of charge (SOC) of the energy storage should
be determined considering the whole operation horizon [19].
• Repeated start and shut down of microturbines in the small time steps is
not desired and amortizes them. Therefore, the status of the microturbine is
determined in the HAS and is not changed in RTS.
• RTS should be fast because it is run repeatedly near the real operation time.
HAS can provide an initial solution for the optimization model of the RTS
module that reduces the time of solving the problem. Also, to further reduce
the problem’s solving time, some of the decision variables can be fixed into the
values calculated by HAS.
Generally, the module of HAS can be hourly run for the next hours up to the
end of the operation horizon, applying the updated data, as indicated in Fig. 3.7.
Then, the results are applied to the RTS which is repeatedly run in 5-minute time
steps. Although the RTS schedules the MG only for the next time steps to the end of
the current hour it indirectly considers the system’s forecasted condition of the next
hours through applying the results of the latest HAS problem.
Example 3.2 Near the real operation time, the electricity market price is determined
by the real-time market. It deviates from the forecasted values. This deviation can be
modeled by the Normal probability distribution. Figure 3.8 indicates the real-time
market prices in the 15-minute time steps. Without the HAS, the on/off status of the
MTs should be determined in RTS. In this condition, severe price fluctuations can
result in the start or shut down of the MTs, as indicated in Fig. 3.8. As a result, EMS
should benefit from both HAS and RTS modules to appropriately schedule the MG.
74 S. M. Hashemi and V. Vahidinasab

Fig. 3.8 The inefficiency of RTS to determine the on/off status of MTs

3.9.2 System Operation Strategies

Different operation strategies can be considered in the EMS, such as [1] operational
and maintenance cost of resources, energy transaction cost, battery degradation
cost [20], outages and interruption cost, demand response incentives, losses cost,
load shedding penalty cost, emission cost and levelized cost of renewable energy
resources. Additionally, the EMS may include technical constraints such as elec-
trical network loading, energy balancing, output limits of the renewable energy
resources, demand response, reactive power support, reliability, and physical limits
of resources. In summary, the operation strategies can be categorized into the
economic strategies and technical strategies that will be presented in the following
subsections.

3.9.2.1 Economic Aspects

Minimizing the total operation cost or maximizing the profit are the most popular
economic objectives of EMS. When the MG is connected to the main grid, it can
export its surplus power to the main grid to make a profit. In return, it can import
power from the main grid to supply internal consumers. There is an economic
competition between the internal resources of the MG and the upstream network
to meet the consumers. The electricity market price will be a vital input parameter
of the decision-making, and a decision with a higher profit is the best choice in this
condition. In a market-oriented decision-making, EMS generates the price bids to
3 Energy Management Systems for Microgrids 75

Exported power
to the main grid

Fig. 3.9 Power generation scheduling of a profit maximizer EMS

participate in the power market [13]. MG’s profitability is not constrained to the
energy exchange with the main grid. According to the fast response resources and
the controllable loads, EMS can make a profit by providing reserve capacity for
the main grid [21]. For the isolated MGs, cost minimization is the main economic
strategy.
Example 3.3 In Example 3.2, When EMS operates the MG for the profit maximiza-
tion, the scheduling outputs will be different from the cost minimization strategy, as
indicated in Fig. 3.9. As can be observed, MG exports energy to the main grid in
some hours. When the. market price is higher than the generation cost of MTs and
there is surplus generation capacity.

3.9.2.2 Technical Aspects

Aside from the economic aspects, there are some technical criteria for the system
operation problem. Maximizing the reliability indices of the system and minimizing
the network loss or fuel consumption [22] are important technical decision strate-
gies. The two later strategies may be in line with the cost minimization or profit
maximization. While the reliability maximization may be against the cost/profit-
oriented decisions. Suppose a sudden contingency leads to the load shedding.
Considering the main task of the power system is supplying the load, the system
operation strategy is to minimize the load shedding or to maximize the amount
of the restored loads. In fact, in the contingency condition, minimum cost or
maximum profit is not considered by the MG’s operator. Although the MGs are
operated in the radial structure, they are designed in the mesh structure, using the
76 S. M. Hashemi and V. Vahidinasab

System operation strategies

Reliability
Cost Profit Technical
and
minimization maximization aspects
Resiliency

 Exporting energy
to main grid
 Operation cost  Providing  Minimizing the  Minimum load
 Maintenance cost ancillary services energy loss shedding.
 Startup/shut for main grid  Minimizing the  Maximum load
down cost  Strategic bidding fuel consumption restoration
to the electricity
markets

Fig. 3.10 System operation strategies of EMS

normally open and normally closed switches. In some scheduling problems, system
reconfiguration variables are determined to minimize the system loss [23]. As well
as, reconfiguration can be applied for the reliability or resiliency goals. For example,
in some of the isolated MGs such as the electric ships, EMS should select the best
supplying path of the loads during the critical condition aiming the minimum load
curtailment [24]. As well, EMS may be designed to deal with the physical attacks
or cyberattacks using the system reconfiguration [25]. Figure 3.10 summarizes the
operation strategies of the EMS.

3.10 Mathematical Modeling of Different MG’s Components

To design and implementation of the EMS, the mathematical models of different

elements of the MG are presented in the following.

3.10.1 Loads

There are different electrical and thermal loads in the MG. Generally, the electrical
load model is expressed by the ZIP model indicated in (3.1) and (3.2) [26]. It means
that each load can be considered as the combination of three load models of constant
impedance, constant current, and constant power.
3 Energy Management Systems for Microgrids 77

  2   
V V
P = P0 Zp + Ip + Pp , Zp + Ip + Pp = 1 (3.1)
V0 V0

  2   
V V
Q = Q0 Zq + Iq + Pq , Zq + Iq + Pq = 1 (3.2)
V0 V0

The hourly thermal load is generally modeled as a constant energy consumption

within an hour. In terms of controllability, loads are categorized into controllable
and uncontrollable loads. In the controllable loads, the power consumption (P) can
be controlled in the range of (3.3), in which PL is the forecasted power consumption.
Eq. (3.4) guarantees that the total power consumption of the load is constant during
the operation horizon.

0 ≤ P ≤ PL (3.3)


Tn 
Tn
Ph = P Lh (3.4)
h=1 h=1

3.10.2 Dispatchable Generation Resources

Dispatchable generation (DG) units have the main function of energy generation
in MG. Diesel generators and microturbines (MT) that consume natural gas and
produce electrical energy are categorized as the DG units. The generation cost of
DGs is a quadratic function as indicated in (3.5). To linearize the optimization
model, this cost function can be replaced by some of the estimating linear functions
[27] as indicated in Fig. 3.11.

Cg = AP 2g + BP g + C (3.5)

The startup cost of the generation units is another cost term that is indicated in
(3.6).
 
SC g,t ≥ 0, SC g,t ≥ ug,t -ug,t -1 scg (3.6)

The active and reactive power of the online DGs should be within the allowable
range of (3.7) and (3.8). In these inequalities, u is a binary variable indicating the
status of the units.
78 S. M. Hashemi and V. Vahidinasab

Cg

ln

l3

l2

l1

Pg

Fig. 3.11 Cost function of the dispatchable gas-burning generation units

Pgmin ug,t ≤ Pg,t ≤ Pgmax ug,t (3.7)

g ug,t ≤ Qg,t ≤ Qg ug,t

Qmin max
(3.8)

There are ramping limits between the consecutive time intervals as indicated in
(3.9)–(3.14). URg and DRg are respectively the up and down ramping capabilities.
z and y are binary variables respectively showing the start and stop of the units.
Also, P0 indicates the initial power generation of the units at the beginning of
the scheduling. The inequalities of (3.9) and (3.10) are the up-ramping constraints.
After starting a unit, its generation should be fixed to Pmin . As well as, (3.11) and
(3.12) are the down ramping constraints. Before stopping a generation unit its power
should be fixed to Pmin .
 
Pg,t − Pg,t−1 ≤ 1 − zg,t U R g + zg,t Pgmin ∀ t >1 (3.9)

 
Pg,t − Pg0 ≤ 1 − zg,t U R g + zg,t Pgmin ∀t =1 (3.10)

 
Pg,t−1 − Pg,t ≤ 1 − yg,t DR g + yg,t Pgmin ∀ t >1 (3.11)

 
Pg0 − Pg,t ≤ 1 − yg,t DR g + yg,t Pgmin ∀t =1 (3.12)
3 Energy Management Systems for Microgrids 79

zg,t + yg,t ≤ 1 ∀ t (3.13)

zg,t − yg,t = ug,t − ug,t−1 ∀ g, t (3.14)

Repeated starting and stopping of DGs in short time intervals amortizes their
mechanical elements. Hence, the minimum up/down time is defined. According to
(3.15) and (3.16), when a unit is started/stopped, its status should not change up to
the MUT/MDT time interval.


T !
ug,t+k−1 ≥ zg,t T , T = min MU T , T h − t + 1 (3.15)
k=1


T
  !
1 − ug,t+k−1 ≥ yg,t T , T = min MDT , T h − t + 1 (3.16)
k=1

3.10.3 Renewable-Based Units with MPPT

Renewable resources are popular generation units of MG. Their generation cost
is very low and they are categorized as clean energy technologies. Considering
their primary energy resources, they have lower controllability than conventional
DGs. For example, the power generation ability of the PV units is directly related
to the solar irradiance and ambient temperature. In fact, they cannot generate
power as much as their installed capacity at any time. Generally, the wind turbines
and PV panels are equipped with the maximum power point tracking (MPPT)
algorithm [28]. MPPT trackers find the best working point of the mentioned
generation units in which the maximum power is produced based on different
environmental conditions. In fact, MPPT trackers improve the controllability of
the renewable resources. Considering P* as the maximum forecasted power of
renewable resources, their actual power generation should be within the specified
range of (3.21).
∗
0 ≤ Pg,h ≤ Pg,h , g ∈ ΩW (3.17)

3.10.4 Energy Storages

Energy storages have a vital function in the MGs. They can be operated as generators
or consumers. They are charged in some hours and generate energy in the next hours.
By the storage units, it will be possible to benefit from the energy price difference
80 S. M. Hashemi and V. Vahidinasab

in different hours. As well as, in the light load condition, they can be charged by
the renewable resources and supply the load at peak hours. SOC of the storages
is calculated by (3.18) which considers the initial stored energy (E0) and charge
or discharge power (Psc/d ) at different time intervals. Since a part of the energy is
wasted during the power charge or discharge, energy efficiency (ηc/d ) is included in
(3.18). Energy storage capacity and charge and discharge rate should be considered
as indicated in (3.19)–(3.22).


t 
t d
P sst,k
Est,t = E0st + c
ηst · P s cst,k − d
(3.18)
k=1 k=1
ηst

min
Est ≤ Est,h ≤ Est
max
(3.19)

0 ≤ P s cst,h ≤ P s max
st (3.20)

0 ≤ P s dst,h ≤ P s max
st (3.21)

P s st,h = P s dst,h − P s cst,h (3.22)

The operation of energy storage has no significant operational cost.

3.10.5 Reactive Power Resources

Generally, capacitors and reactors of the electrical network participate in the voltage
regulation as the reactive power resources. Their reactive power generation should
be constrained by their installed capacities as indicated in (3.23).

Qcmin
c ≤ Qcc,t ≤ Qcmax
c (3.23)

3.10.6 Combined Heat and Power (CHP) and Boiler

CHPs having high energy efficiency are usually installed in the MGs containing the
thermal loads. CHP simultaneously generates electrical and thermal power using
natural gas. The allowed operational zone of the CHPs has been indicated in Fig.
3 Energy Management Systems for Microgrids 81

PCHP

Maximum heat
generation

HCHP

Fig. 3.12 Allowable thermal-electrical operating zone of CHP

3.12 and (3.28). Operational constraints presented for DGs can be applied for CHPs.
The constraints such as ramping limitations.

chp
P Pch,t + βCH P Hch,t ≤ γCH P ug,t th ∈ {1, 2, 3}
th CH P th CH P th
αCH (3.24)

In general, the operation cost of the CHP unit is a nonlinear function as indicated
in (3.25) [29].
p p
Cch,t = αch Pch,t
2
+ βch Pch,t + αch
h 2
Hch,t + βch
h
Hch,t + λch Pch,t Hch,t + σch
(3.25)

In [30] a linearized version has been proposed for operation cost of CHP units in
which piecewise linearization has been applied for two variables: output electrical
and thermal power. Besides, some researches [31] use a linear cost function for CHP
using an average energy efficiency factor (ηchCH P ) and gas price (π gas ) as indicated

(3.30)–(3.31). In these constraints, σ MWh/m3 is a conversion factor that includes the

thermal value of the natural gas.

Cch,t = π gas gas CH

ch,t
P
(3.26)

 
CH P + H CH P
Pch,t ch,t
ch,t = σ
gas CH P MW h/m3
CH P
(3.27)
ηch
To increase the flexibility and to serve the thermal load, auxiliary boilers are
installed beside the CHP units. Their thermal power generation should be within the
allowed range as indicated in (3.32).
82 S. M. Hashemi and V. Vahidinasab

boi,min boi,max
uboi
b,t Hb,t ≤ Hb,t
boi
≤ uboi
b,t Hb,t (3.28)

Similar to the operation cost of CHP units in (3.30)–(3.31), the operation cost
of the boilers is modeled by (3.29) and (3.30) in which uboi
b,t is a binary variable
indicating the on/off status of the boiler.

Cb,t = π gas gas boi

b,t (3.29)

boi
Hb,t
b,t = σ
gas boi MW h/m3
(3.30)
ηbboi

3.10.7 Electrical Network

Operational variables of the electrical network such as line flows and bus voltages
should be modeled in EMS. They can be computed using the injected power of the
buses and the properties of the lines through the power flow equations. Although
methods such as forward–backward sweep [32] calculate the line flows and bus
voltages of the radial networks in an iterative procedure they are not appropriate
for the centralized optimization models of EMS. Direct current (DC) power flow
is a popular method for large power systems. Although it is a linear and simple
method, it is not appropriate for the MGs and distribution networks in which the R/X
ratio is high. On the other hand, using the conventional AC power flow complicates
the optimization problem because it contains the non-convex constraints. In the
following subsections, two mathematical models are presented for the power flow.

3.10.7.1 Second-Order Cone Programming (SOCP)

According to [33], an exact second-order cone relaxation method can be defined

for the convexification of the AC load flow of the radial systems. The convexified
AC load flow equations are indicated in (3.27)–(3.37). The Eqs. (3.31) and (3.32)
are respectively the active and reactive power balance in each bus and (3.33)-(3.34)
are the injected active and reactive power of the buses. According to (3.35), any
bus voltage is determined based on the input power from the upstream bus and its
voltage. Eq. (3.36) relates the active and reactive flows to the bus voltage and line
currents. In the conventional formulation, (3.36) is a non-convex equality constraint.
In return, the indicated inequality constraint is convex. This relaxation will be exact,
if the mentioned constraint is active, after solving the optimization problem. In
p q
these equations, Fn,t and Fn,t are active or reactive input powers of bus n from
its upstream line. r and x are respectively the resistance and reactance of the lines.
Pg, t , Psst, t and PLn, t are respectively the output active power of generators and
3 Energy Management Systems for Microgrids 83

storage units and the energy consumption of the load. As well as Qg, t , Qcc, t and
QLn, t are respectively the reactive power generation of the generators and reactive
power resources, and the reactive loads. Vn, t and In, t are the bus voltages and line
currents that are indicated in Fig. 3.13.

p inj
  p  p
Fn,t + Pn,t = Fm,t + rm m,t , Fn=0,t = 0 (3.31)
m∈Ωn

q inj
  q  q
Fn,t + Qn,t = Fm,t + xm m,t , Fn=0,t = 0 (3.32)
m∈Ωn

inj
 
Pn,t = Pg,t + P s st,t − P Ln,t (3.33)
g∈Ωgn st∈Ωst n

inj
 
Qn,t = Qg,t + Qcc,t − QLn,t (3.34)
g∈Ωgn c∈Ωcn

  
p q 
υn,t = υm,t + 2 rm Fm,t + xm Fm,t + rm
2
+ xm
2
m,t , m ∈ Ωnn (3.35)

 p 2  q 2
Fn,t + Fn,t ≤ υn,t n,t , n = 0 υn,t = Vn,t
2
, n,t = In,t
2
(3.36)

V 02 (1 − 0 )2 ≤ υn,t ≤ V 02 (1 + 0 )2 (3.37)

3.10.7.2 Linear DistFlow

The presented SOCP method is an exact convex model for power flow. However,
it is a MIQCP model. There is another power flow method that is a linear model.
The mathematical formulation of the Linear DistFlow [34] method is indicated in

Fig. 3.13 Line flow and bus

voltage
84 S. M. Hashemi and V. Vahidinasab

(3.34)–(3.44). As can be observed, its accuracy is higher than the DC load flow
method. It determines the bus voltages, while it does not model the power loss of
the lines. It should be noted that Eq. (3.44) is not included in the optimization model.
It is used to calculate the bus voltages after solving the optimization problem.

p inj
 p p
Fn,t + Pn,t = Fm,t , Fn=0,t = 0 (3.38)
m∈Ωn

q inj
 q q
Fn,t + Qn,t = Fm,t , Fn=0,t = 0 (3.39)
m∈Ωn

inj
 
Pn,t = Pg,t + P sst,t − P Ln,t (3.40)
g∈Ωgn st∈Ωst n

inj
 
Qn,t = Qg,t + Qcc,t − QLn,t (3.41)
g∈Ωgn c∈Ωcn

 p q 
υn,t = υm,t + 2 rm Fm,t + xm Fm,t , m ∈ Ωnn (3.42)

   
V 02 0 2 − 20 ≤ υn,t ≤ V 02 0 2 + 20 (3.43)

υn,t = Vn,t
2
− V 02 (3.44)

3.10.8 Energy Exchange with the Main Grid

MG can export/import electrical energy to/from the main grid. The cost of the
maingrid
energy exchange with the main grid is calculated by (3.45) in which Pt is the
electrical imported power from the main grid and prct is the electricity market price.
maingrid
The negative value of Pt means the power export to the main grid resulting
in the negative cost that means the profit.

grid maingrid
Ct = Pt prct (3.45)
3 Energy Management Systems for Microgrids 85

3.11 Mathematical Modeling of System Security

Power system reliability is a concept introducing different indices showing the

system capability to deal with the planned and unplanned outages of the equipment.
It is divided into two popular terms of adequacy and security. The reliability of
the system in a short-term scheduling problem is analyzed through the concept of
system security. n − 1 is a well-known criterion in security modeling. It means that
the normal state working point of the system should be robust against the outage of
any single element of line or power plant so that the operating limits are satisfied and
all the loads are supplied. Usually, MG’s security is analyzed when it is suddenly
isolated from the main grid [35]. In comparison to the large power systems, MGs
have lower inertia. So, the power disturbance severely affects the MGs frequency.
In fact, in addition to the reserve constraints, some frequency-related constraints
should be considered in the scheduling procedure as dynamic security. Here, two
formulations are presented for the dynamic security which is compatible with HAS
and RTS.

3.11.1 Security Modeling in HAS

It is not possible to include the exact frequency control models in the HAS because
the on/off status of the resources is unknown and the frequency response model is
undefined. HAS should provide enough online generators to increase the system’s
inertia and its capability to respond to the power disturbance. Here, an energy-
related frequency response model is presented [36, 37]. The kinetic energy of the
MG is indicated in (3.46) in which JT and f are respectively the total inertia moment
of MG and the frequency of MG.

1 T
E= J (2πf )2 (3.46)
2

JT is the sum of inertia moments of the online generators as indicated in (3.47).

In this equation, Hk is the inertia constant of the kth generator.
   2P max Hk
JT = Jk = uk Jk = uk (3.47)
4π 2 f 2
k∈online generators k k

A power mismatch of D > 0 (kW) for t seconds reduces the system’s energy as
"t
much as ΔE1 = Ddτ = Dt. Generation units change their production to deal with
0
"t rt 2
the disturbance. This changes the systems’ energy as much as ΔE2 = rτ dτ = 2 ,
0
in the opposite direction of E1 , in which r is the equal ramp rate of the generators.
86 S. M. Hashemi and V. Vahidinasab

Considering E = E1 − E2 as the net energy deviation of the system, the time
of minimum kinetic energy (tnadir ) can be calculated by (3.48).

d D
(ΔE) = 0 → t nadir = (3.48)
dt r

According to the minimum allowed value of the system frequency (fmin ), a lower
bound is determined for the systems’ kinetic energy in t = tnadir as indicated in
(3.49).

1 T D2 1  2
E − ΔE ≥ E min → J (2πf )2 − ≥ J T 2πf min (3.49)
2 2r 2
Replacing (3.47) and (3.48) into (3.49) results in (3.50).
 2
2 f − f min 
t nadir
≤ uk Pkmax Hk (3.50)
f 2D
k

On the other hand, it takes tr minutes to deliver Ri MW of reserve capacity by

the generation unit of i. It is determined considering its ramp rate (ai ) as indicated
in (3.51).

Ri
tr = (3.51)
ai

When tr of the reserve units is lower than tnadir , the system frequency won’t
be lower than fmin . Considering (3.50) and (3.51), reserve allocation should be
constrained as indicated in (3.52).
 2
Ri 2 f − f min 
≤ uk Pkmax Hk (3.52)
ai f 2D
k

Also, the sum of the reserve capacity of the generators should be equal to the
power mismatch (3.53).

Ri = D (3.53)
i

In the connected mode operation of MG, the power mismatch of D is equal to

the amount of power exchange with the main grid. In the isolated MGs, any load
connection or the generation outage can be considered as the power disturbance.
Example 3.4 An isolated MG is analyzed in this example. If in Example 3.1
generation capacity of MTs is increased to 40 kW, MG will be able to serve the loads
3 Energy Management Systems for Microgrids 87

Table 3.1 Technical and economic properties of the generation units of MG

H (sec) Pmax (kW) Pmin (kW) Cost (cent/kW) Ramp rate (kW/sec)
MT1 1.5 40 4 4 10
MT2 1.5 40 4 4.4 12
PV 0 Variable 0 0 50

in the islanded mode. The technical and economic specifications of the generation
units are indicated in Table 3.1.
As indicated in Fig. 3.14a, if the frequency deviation is not considered in the
system scheduling, PV and MT1 will supply the load during the whole operation
horizon. Given the generation cost of resources, that is the right decision. By adding
the frequency-related constraints, different factors affect the power generation
schedule. Factors such as inertia constants of the online generators, minimum
allowed frequency, ramp rate of the resources, and the power mismatch. Assuming
fmin = 49 Hz and D = 5% load, generation scheduling will be as Fig. 3.14b. It
is worth noting that the PV power plant can participate in the primary frequency
regulation procedure [38–41]. As can be observed, MT2 that is an expensive
generation unit is online in many hours, and it generates as much as its minimum
power. In fact, on the one hand, it increases the system’s inertia and on the other
hand, it provides a fast reserve resource. Figure 3.14c indicates the reserve capacity
of different resources. PV unit having low operation cost is often operated on its
maximum generation ability and does not provide reserve capacity. Figure 3.14d
indicates tr of the units and tnadir of the system. Reserve capacities are deployed
before tnadir .
Example 3.5 Increasing the generation cost of MT2 to 6 (cent/kW) changes the
scheduling results. In this condition, as indicated in Fig. 3.15a MT2 is turned off
in many hours that results in the low inertia constant of the MG. To deal with this
problem, the PV unit having a very high ramp rate reduces the power generation
(Fig. 3.14d) and provides a large share of the reserve capacity (Fig. 3.14b). Despite
the reduction of the system’s inertia during the hours 9–19, the reserve capacity is
deployed very fast as indicated in Fig. 3.14c.

3.11.2 Security Modeling in RTS

In RTS, online generators are determined by the latest HAS. Hence, the frequency
response model can be constructed. It can be used to limit the disturbance power of
the MG within a specified margin avoiding frequency deviation beyond the allowed
values. For instance, in the connected operation mode of MG, the sudden islanding
severely affects the MG’s frequency. Especially when the power exchange with the
main grid is high, the RTS should limit the power exchange. Frequency response
 88

Fig. 3.14 Results of HAS for islanded MG of Example 3.4: (a) Power generation of different resources in the insecure operation. (b) Power generation of
different resources in the secure operation. (c) Reserve capacity of different resources. (d) The reaction time of the reserve resources versus the frequency drop
time
S. M. Hashemi and V. Vahidinasab
 3 Energy Management Systems for Microgrids

Fig. 3.15 Results of HAS for islanded MG of Example 3.5: (a) Power generation of different resources. (b) Reserve capacity of the MG’s resources. (c) The
reaction time of the reserve resources versus the frequency drop time. (d) PV unit’s generated power versus its available power
89
90 S. M. Hashemi and V. Vahidinasab

−k Secondary
control
1 Primary control
loop (MT)
s loop
R Micro turine

_
1 1 + 1 f
+ 1+ T S v
1+ T S t
_ + 2Hs + D
Governer Turbine

kss
ΔP=P main grid 1 + TsS
Virtual inertia
control system

Fig. 3.16 Frequency response diagram of MG

0
Frequency response for power
disturbance of pα
Frequency

-f α

-Hzmax

0
Time (sec)

Fig. 3.17 The impact of power disturbance on the MG’s frequency deviation

in the downside of the network can be calculated by the linear frequency response
model of [42] that is indicated in Fig. 3.16. As indicated, the input power of MG is
considered as the disturbance power of the frequency response model.
According to Fig. 3.17, Assuming fα as the maximum frequency deviation for
power disturbance of pα , power exchange should be within the specified margin of
(3.54) avoiding frequency to go beyond HZmax [43].

maingrid
H zmax P H zmax
− ≤ t α ≤ (3.54)
ftα p ftα

In order to calculate the dynamic parameters of the frequency response model of

the MG’s equal machine, equations of (3.55)–(3.61) should be considered [44].
3 Energy Management Systems for Microgrids 91

#
Pkmax Hk
k∈online generator
H = # (3.55)
Pkmax
k∈online generator

#
Pkmax R1k
1 k∈online generator
= # (3.56)
R Pkmax
k∈online generator

#
Pkmax Tkv
k∈online generator
Tv = # (3.57)
Pkmax
k∈online generator

#
Pkmax Tkt
k∈online generator
Tt = # (3.58)
Pkmax
k∈online generator

#
Pimax ki
i∈online generator
k= # (3.59)
Pimax
i∈online generator

#
Pimax kis
i∈online storages
ks = # (3.60)
Pimax
i∈online storages

#
Pkmax Tks
k∈online storages
Ts = # (3.61)
Pkmax
k∈online storages

In the islanded operation mode in which the outage of generation units leads to
power disturbance, the frequency limit of (3.54) can be considered in several single
contingencies of the online generators.
Example 3.6 Suppose that in Example 3.4, MG is connected to the main grid.
Dynamic parameters of MTs are indicated in Table 3.2. Also, ks and Ts are
respectively 0.5 and 10 seconds.
92 S. M. Hashemi and V. Vahidinasab

Table 3.2 Dynamic parameters of the MTs

H (p.u. Tv (sec) Tt (sec) R (Hz D K (p.u. ks (sec) T3 (sec)
MW/sec) /p.u. MW) (MW/Hz) MW/Hz)
MT1, 1.5 0.1 0.4 2.4 0.015 0.05 0.5 10
MT2

a b

d c

Fig. 3.18 The impact of frequency constraints on the RTS: (a) Given the status of MTs from
HAS. (b) generation cost of MTS and the electricity market price in RTS. (c) Power generation
scheduling in an insecure operation. (d) Power generation scheduling in the secure operation

We want to schedule the MG in real time during 04:00–08:45. The results of

the latest HAS are provided and determine the online generators in different hours
as indicated in Fig. 3.18a. The electricity market price and generation cost of MTs
are indicated in Fig. 3.18b and RTS is run in 15 minutes intervals. When there
is no attention to the frequency response of the MG in the sudden islanding, MG
imports power from the main grid and reduces the power generation of MTs in most
of the operation horizon as indicated in Fig. 3.18c. According to the market price
variations, it is the best decision from the economic viewpoint. Adding security
constraints changes the MG’s scheduling. When the maximum frequency deviation
is limited to 0.1 Hz, input power from the main grid will be reduced as indicated
in Fig. 3.18d. This scheduling guarantees the secure operation of MG in the case of
sudden islanding. As can be observed, during the hours 7 and 8 in which both of
the MTs are online, power scheduling has no significant sensitivity to the frequency
constraints because in this condition high inertia constant of MG avoids the severe
frequency deviation. In return, the low inertia constant of MG during the hours
4 to 6, activates the frequency constraint and changes the generation scheduling
decisions.
3 Energy Management Systems for Microgrids 93

3.12 Conclusion

The EMS is a vital element in an MG that monitors and controls all the interactions
and operates different resources based on the determined strategies. To analyze the
MG operation, EMS provides different analytical tools, the most widely used of
which is the optimal operation tool. This chapter focused on this aspect to present
the scheduling models of MG. Cost minimization and profit maximization are two
popular operation strategies of the MG. According to the intrinsic properties of some
of the decision variables and the uncertainties of the input parameters, a two-stage
scheduling structure was used for EMS which includes the HAS and RTS. HAS
includes a UC model and is hourly run to schedule the MG from the running time to
the end of the operation horizon. Hence, all of the forecasted parameters of the next
hours are considered in HAS. According to the variability of the input parameters
such as load, renewable power generation, and electricity market price, the results
of HAS are not appropriate for the system operation and MG should be rescheduled
near the real operation time by RTS. By receiving the on/off status of the resources
from the results of the latest HAS, the RTS only focuses on the current hour using
an ED model. To deal with contingencies, security was considered in EMS using
the frequency response model. MG’s inertia is very low and sudden contingencies
severely deviate the frequency. An energy-based frequency response model was
applied to HAS to the commitment of fast response generators in accordance with
the MG’s inertia. In RTS, the online generators are given by the latest HAS. So, the
exact frequency deviation was calculated by the frequency response diagram, and an
allowed range was determined for it. The mathematical formulations and technical
details of this chapter were included both the islanded and connected operation
mode of MG.

References

1. Zia, M. F., Elbouchikhi, E., & Benbouzid, M. (2018). Microgrids energy management systems:
A critical review on methods, solutions, and prospects. Applied Energy, 222, 1033–1055.
2. Palma-Behnke, R., et al. (2013). A microgrid energy management system based on the rolling
horizon strategy. IEEE Transactions on Smart Grid, 4(2), 996–1006.
3. G. Graditi, M. G. Ippolito, E. R. Sanseverino, and G. Zizzo, Optimal set points regulation
of distributed generation units in micro-grids under islanded operation, In 2010 IEEE
International Symposium on Industrial Electronics, 2010, pp. 2253–2260.
4. Chen, C., Duan, S., Cai, T., Liu, B., & Hu, G. (2011). Smart energy management system for
optimal microgrid economic operation. IET Renewable Power Generation, 5(3), 258–267.
5. M. L. Di Silvestre, G. Graditi, M. G. Ippolito, E. R. Sanseverino, and G. Zizzo, “Robust
multi-objective optimal dispatch of distributed energy resources in micro-grids,” In 2011 IEEE
Trondheim Power Tech, 2011, pp. 1–5.
6. Sanseverino, E. R., Di Silvestre, M. L., Ippolito, M. G., De Paola, A., & Re, G. L. (2011). An
execution, monitoring and replanning approach for optimal energy management in microgrids.
Energy, 36(5), 3429–3436.
94 S. M. Hashemi and V. Vahidinasab

7. R. Palma-Behnke, C. Benavides, E. Aranda, J. Llanos, and D. Saez, “Energy management

system for a renewable based microgrid with a demand side management mechanism,” In 2011
IEEE symposium on computational intelligence applications in smart grid (CIASG), 2011, pp.
1–8.
8. Bustos, C., Watts, D., & Ren, H. (2012). Microgrid operation and design optimization with
synthetic wins and solar resources. IEEE Latin America Transactions, 10(2), 1550–1562.
9. X. Zhang, R. Sharma, and Y. He, “Optimal energy management of a rural microgrid system
using multi-objective optimization,” In 2012 IEEE PES innovative smart grid technologies
(ISGT), 2012, pp. 1–8.
10. Mohamed, F. A., & Koivo, H. N. (2012). Multiobjective optimization using mesh adaptive
direct search for power dispatch problem of microgrid. International Journal of Electrical
Power & Energy Systems, 42(1), 728–735.
11. G. Celli, F. Pilo, G. Pisano, and G. G. Soma, “Optimal participation of a microgrid to the energy
market with an intelligent EMS,” In 2005 International Power Engineering Conference, 2005,
pp. 663–668.
12. Siano, P., Cecati, C., Yu, H., & Kolbusz, J. (2012). Real time operation of smart grids via FCN
networks and optimal power flow. IEEE Trans Ind Inform, 8(4), 944–952.
13. A. G. Tsikalakis and N. D. Hatziargyriou, “Centralized control for optimizing microgrids
operation,” In 2011 IEEE power and energy society general meeting, 2011, pp. 1–8.
14. Meng, L., Sanseverino, E. R., Luna, A., Dragicevic, T., Vasquez, J. C., & Guerrero, J.
M. (2016). Microgrid supervisory controllers and energy management systems: A literature
review. Renewable and Sustainable Energy Reviews, 60, 1263–1273.
15. Logenthiran, T., Srinivasan, D., Khambadkone, A. M., & Aung, H. N. (2012). Multiagent
system for real-time operation of a microgrid in real-time digital simulator. IEEE Transactions
on Smart Grid, 3(2), 925–933.
16. Colson, C. M., & Nehrir, M. H. (2013). Comprehensive real-time microgrid power manage-
ment and control with distributed agents. IEEE Transactions on Smart Grid, 4(1), 617–627.
17. Liang, H., Choi, B. J., Abdrabou, A., Zhuang, W., & Shen, X. S. (2012). Decentralized
economic dispatch in microgrids via heterogeneous wireless networks. IEEE Journal on
Selected Areas in Communications, 30(6), 1061–1074.
18. Kanchev, H., Lu, D., Colas, F., Lazarov, V., & Francois, B. (2011). Energy management
and operational planning of a microgrid with a PV-based active generator for smart grid
applications. IEEE Transactions on Industrial Electronics, 58(10), 4583–4592.
19. Olivares, D. E., Lara, J. D., Cañizares, C. A., & Kazerani, M. (2015). Stochastic-predictive
energy management system for isolated microgrids. IEEE Transactions on Smart Grid, 6(6),
2681–2693.
20. L. N. An and T. Quoc-Tuan, “Optimal energy management for grid connected microgrid
by using dynamic programming method,” In 2015 IEEE Power & Energy Society General
Meeting, 2015, pp. 1–5.
21. Shen, J., Jiang, C., Liu, Y., & Qian, J. (2016). A microgrid energy management system with
demand response for providing grid peak shaving. Electric Power Components & Systems,
44(8), 843–852.
22. Anglani, N., Oriti, G., & Colombini, M. (2017). Optimized energy management system
to reduce fuel consumption in remote military microgrids. IEEE Transactions on Industry
Applications, 53(6), 5777–5785.
23. Zidan, A., & El-Saadany, E. F. (2013). Distribution system reconfiguration for energy loss
reduction considering the variability of load and local renewable generation. Energy, 59, 698–
707.
24. S. B. Ganesh, K. Schoder, H.-J. Lai, A. Al-Hinai, and A. Feliachi, “Energy management
system with automatic reconfiguration for electric shipboard power systems,” In Proc. of ASNE
Reconfiguration and Survivability Symposium, RSS 2005, 2005.
25. Paridari, K., O’Mahony, N., Mady, A. E.-D., Chabukswar, R., Boubekeur, M., & Sandberg,
H. (2017). A framework for attack-resilient industrial control systems: Attack detection and
controller reconfiguration. Proceedings of the IEEE, 106(1), 113–128.
3 Energy Management Systems for Microgrids 95

26. M. Sadeghi, “Determination of ZIP parameters with least squares optimization method,” In
2009 IEEE Electrical Power & Energy Conference (EPEC), 2009, pp. 1–6.
27. H. Ahmadi, J. R. Martí, and A. Moshref, “Piecewise linear approximation of generators
cost functions using max-affine functions,” In 2013 IEEE Power & Energy Society General
Meeting, 2013, pp. 1–5.
28. M. A. G. De Brito, L. Galotto, L. P. Sampaio, G. de A. e Melo, and C. A. Canesin, “Evaluation
of the main MPPT techniques for photovoltaic applications,” IEEE Transactions on Industrial
Electronics, vol. 60, no. 3, pp. 1156–1167, 2012.
29. Sashirekha, A., Pasupuleti, J., Moin, N. H., & Tan, C. S. (2013). Combined heat and power
(CHP) economic dispatch solved using Lagrangian relaxation with surrogate subgradient
multiplier updates. International Journal of Electrical Power & Energy Systems, 44(1), 421–
430.
30. Moradi, S., Ghaffarpour, R., Ranjbar, A. M., & Mozaffari, B. (2017). Optimal integrated sizing
and planning of hubs with midsize/large CHP units considering reliability of supply. Energy
Conversion and Management, 148, 974–992.
31. Zhang, Y., Wang, R., Zhang, T., Liu, Y., & Guo, B. (2016). Model predictive control-based
operation management for a residential microgrid with considering forecast uncertainties and
demand response strategies. IET Generation Transmission and Distribution, 10(10), 2367–
2378.
32. Panwar, L. K., Konda, S. R., Verma, A., Panigrahi, B. K., & Kumar, R. (2017). Operation
window constrained strategic energy management of microgrid with electric vehicle and
distributed resources. IET Generation Transmission and Distribution, 11(3), 615–626.
33. Gan, L., Li, N., Topcu, U., & Low, S. H. (2015). Exact convex relaxation of optimal power
flow in radial networks. IEEE Transactions on Automatic Control, 60(1), 72–87.
34. Šulc, P., Backhaus, S., & Chertkov, M. (2014). Optimal distributed control of reactive power
via the alternating direction method of multipliers. IEEE Transactions on Energy Conversion,
29(4), 968–977.
35. Vergara, P. P., López, J. C., da Silva, L. C., & Rider, M. J. (2017). Security-constrained optimal
energy management system for three-phase residential microgrids. Electric Power Systems
Research, 146, 371–382.
36. S. M. Hashemi, V. Vahidinasab, M. S. Ghazizadeh, and J. Aghaei, “Reliability-oriented DG
allocation in radial Microgrids equipped with smart consumer switching capability,” In 2019
International Conference on Smart Energy Systems and Technologies (SEST), 2019, pp. 1–6.
37. Lee, Y.-Y., & Baldick, R. (2013). A frequency-constrained stochastic economic dispatch
model. IEEE Transactions on Power Apparatus and Systems, 28(3), 2301–2312.
38. A. Hoke and D. Maksimović, “Active power control of photovoltaic power systems,” In 2013
1st IEEE Conference on Technologies for Sustainability (SusTech), 2013, pp. 70–77.
39. A. S. Dobakhshari, S. Azizi, and A. M. Ranjbar, “Control of microgrids: Aspects and
prospects,” In 2011 International Conference on Networking, Sensing and Control, 2011, pp.
38–43.
40. Rahmann, C., & Castillo, A. (2014). Fast frequency response capability of photovoltaic power
plants: The necessity of new grid requirements and definitions. Energies, 7(10), 6306–6322.
41. P. Pourbeik, S. Soni, A. Gaikwad, and V. Chadliev, “Providing primary frequency response
from photovoltaic power plants,” in Proc. CIGRE Symp., 2017, pp. 1–10.
42. Kerdphol, T., Rahman, F. S., Mitani, Y., Watanabe, M., & Küfeoǧlu, S. (2018). Robust virtual
inertia control of an islanded microgrid considering high penetration of renewable energy.
IEEE Access, 6, 625–636.
43. Hashemi, S. M., Vahidinasab, V., Ghazizadeh, M. S., & Aghaei, J. (2018). Valuing consumer
participation in security enhancement of microgrids. IET Gener Transm Distrib.
44. Shekari, T., Aminifar, F., & Sanaye-Pasand, M. (2016). An analytical adaptive load shedding
scheme against severe combinational disturbances. IEEE Transactions on Power Apparatus
and Systems, 31(5), 4135–4143.
Chapter 4
Optimal Dispatch and Unit Commitment
in Microgrids

Hossein Shayeghi and Masoud Alilou

4.1 Introduction

In the last years, the concept of microgrid (MG) has been introduced for better
managing an extensive and complex power network. In other words, the total power
system operates efficiently if each of the microgrids is managed properly. The
microgrid is defined as a group of loads, distributed generation (DG) units, and
electrical energy storage systems (ESS) [1]. The microgrid is connected to the
upstream network for buying energy from the grid when the produced power of
DG units and ESS is lower than the demand of the MG. Moreover, MG can sell
energy to the power network when there is extra power of energy sources. Therefore,
the microgrid has the ability for increasing the local reliability and flexibility of
the electric power network [2]. Optimal management of the energy sources of
the microgrid has a high effect on the performance of both microgrid and power
network. The operator of the microgrid has to schedule the operational time of
renewable and nonrenewable DG units and energy storage system of the MG in
order to improve the technical, economic, and environmental indices of the MG. The
variations of market price, load, and weather affect the operational time of energy
sources [3].
In microgrids, demand-side management (DSM) is a practical method to increase
the efficiency of the microgrid. DSM is the planning, implementation, and monitor-
ing of utility activities that are designed to control the consumption of customers.
DSM causes to change the daily pattern and magnitude of various loads [4]. One
of the useful technologies of DSM is the demand response (DR). Demand response

H. Shayeghi () · M. Alilou

Energy Management Research Center, University of Mohaghegh Ardabili, Ardabil, Iran

© Springer Nature Switzerland AG 2021 97

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_4
98 H. Shayeghi and M. Alilou

programs can be utilized due to the two-way communication between the suppliers
and consumers. These programs are practical to better match the consumption
pattern with the production curve of the microgrid [5]. So it can be said that
the performance of the microgrid improves significantly when a demand response
program and an optimal unit commitment are simultaneously implemented on the
microgrid.
In the last years, some researchers have studied the unit commitment and
demand response programs in microgrids. For instance, optimizing the operational
schedule of multi-DG and grid has been studied in [6] in the presence of a time-
of-use demand response program. In this study, the considered objective functions
were optimized by the combination of a non-dominated sorting firefly algorithm
and a fuzzy method. In another study, the authors have proposed an intelligent
energy management framework for optimizing the operational schedule of units
[7]. The minimization of electrical peak load and reduction of electricity costs
are considered as objective functions in this research. Wu et al. have studied the
optimal unit commitment in a grid-connected photovoltaic-battery hybrid system in
the [8]. The main purpose of this research is to sufficiently explore solar energy
and benefit customers at microgrid. Kotur and Durisic in [9] have studied on
managing the microgrid through the spatial and temporal demand-side management.
The considered objective function of this research consists of daily energy losses
and daily operating costs of units. In another research, unit commitment has been
investigated in the residential microgrid in the presence of rooftop photovoltaic units
[10]. An autonomous energy consumption scheduling algorithm has been proposed
to decrease the peak load and reverse power flow. In some studies, the optimization
of the capacity of local energy sources has been investigated. For instance, in
[11], the simultaneous capacity optimization of distributed generation and energy
storage system has been solved by a sequential quadratic programming method.
Both the grid-connected and stand-alone microgrid has been considered as the
main network of study. The minimization of the total annual energy losses and the
reduction of the cost of the energy have been investigated during the optimization.
In another study, a bi-level programming model has been proposed to optimize the
location and size of the battery energy storage system by a numerical optimization
algorithm [12]. The microgrid has been considered in the presence of a wind farm.
The result of numerical simulations indicates that the located battery at the load
center has higher performance than the located battery at the wind farm. In [13],
an offshore wind farm has been considered. In this work, the unit commitment
has been studied in the presence of high wind penetration scenarios in order to
increase hourly spinning reserve capacity for covering the uncertainty of wind
power and load. Garlik and Krivan formulated a unit commitment optimization
problem for renewable energy units in a microgrid in [14]. The proposed unit
commitment has been done using a simulated annealing heuristic optimization
technique. In another study, the modified particle swarm optimization algorithm has
been used for optimal unit commitment of renewable energy sources including wind
turbines, photovoltaic panels, and combined heat and power plants in a microgrid
[15]. Minimizing the cost of the microgrid is the main objective function of the
4 Optimal Dispatch and Unit Commitment in Microgrids 99

proposed unit commitment in this study. Experimental studies of microgrid and its
devices have been presented in some papers. For instance, in [16], multi-objective
optimization has been presented for simultaneously minimizing the total cost of
the distribution system and maximizing the stability of the microgrid. The particle
swarm optimization has been utilized to find the best combination of battery and
hydrogen storage system based on the total cost, the occupied area, legislation, and
local pollution.
According to previous studies, it can be said that optimizing as a multi-objective
function and considering both supplier and consumer sides of the microgrid are the
topics that have been less studied. The innovative contributions of this study are
highlighted as follows:
• Simultaneously optimizing both local energy sources and loads of the microgrid.
• Multi-objective optimization of the proposed method by considering an
economic–environmental objective function.
• Considering renewable and nonrenewable energy sources.
• Utilizing a new intelligent algorithm based on the grey wolf optimization algo-
rithm and the fuzzy method for multi-objective optimizing the unit commitment.
So in this chapter, a multi-objective optimization algorithm is proposed to
optimal manage the operational schedule of the energy sources of the microgrid
in order to improve the performance of the MG. Renewable distributed generation
such as wind turbine and photovoltaic and nonrenewable units such as micro
turbine and fuel cell and also energy storage system are energy sources of the
microgrid. For better managing the MG, a price-based demand response program
is also considered in the proposed method. The considered objective function of
the optimization consists of economic and environmental indices. A new method
based on the grey wolf optimization algorithm and fuzzy method is introduced
for multi-objective optimizing the hourly performance of energy sources of the
microgrid in the presence of a demand response program in order to improve the
efficiency of the MG. Finally, the proposed unit commitment is applied to a sample
microgrid. Ultimately, the efficiency of the proposed method is pondered based on
the simulation results.

4.2 Microgrid

In the last decade, microgrid has been introduced for better controlling the total
power network. Figure 4.1 shows a sample power system. This grid is divided into
the number of microgrids in order to increase the performance of the system. Each
microgrid has some energy sources and loads that should be managed optimally
based on technical, economic, and environmental parameters.
Thus, microgrids are utilized in the power network for improving the local
reliability and flexibility of electric power systems so that the total grid is operated
efficiently if each of the microgrids is managed and operated optimally. Although
100 H. Shayeghi and M. Alilou

Microgrid Microgrid

Upper network

Microgrid Microgrid

Fig. 4.1 A sample power system with the number of microgrids

microgrids can operate in islanded mode, microgrids in grid-connected mode have

a useful ability for buying power from the upstream network when the demand of
the microgrid is more than the produced power of local energy sources. Moreover,
the microgrid can sell its extra energy to the upstream network. In this chapter, the
grid-connected microgrid is studied.

4.3 Demand Response Program

Demand response is a method for increasing the participation of consumers in

managing the microgrid. The proper applying the DR programs to the microgrid
causes to shave the peak demand, manage risk and reliability, and reduce carbon
emission and energy cost.
Totally, DR programs can be divided into Time/Price-based programs and
Incentive-based programs. The various methods for applying the DR program are
demonstrated in Fig. 4.2 [17].
The time or price-based DR programs use the variation of electricity cost during
the 24-hour as a control signal to affect the electricity consumption of consumers
while Incentive-based DR programs utilize the incentive and penalty to encourage
customers for managing the load profile. In this chapter, the real-time pricing (RTP)
4 Optimal Dispatch and Unit Commitment in Microgrids 101

Time/Price
DR programs
Emergency DR Time of use

Capacity market Real time pricing

Ancillary service Critical peak pricing

Demand response
programs
Demand bidding

Interruptible service

Direct load control

Incentive
DR programs

Fig. 4.2 The different methods of demand response programs

method, which is a price based DR program, is utilized to participate the consumers

in the demand side management.
Electricity tariff varies continuously in the RTP program based on the climate
zone, seasonal weather, and time. The dynamic price is available to the consumers
one day ahead or an hour. This DR program causes that the distribution company
of MG better distribute the electricity tariff reflecting the demand-supply elastics.
Thus, the RTP program encourages the consumers to optimally manage the load
profile and decrease their energy consumption during peak times or move the period
of energy use to off-peak hours.
Therefore, the RTP program is used as a demand response program in this chapter
because it is one of the common and useful DR programs in microgrids. Equation
(4.1) presents the considered model of demand response program [17].
⎡ ⎤
⎢ P (i) − P0 (i) 24 P (j ) − P0 (j ) ⎥
d(i) = d0 (i) × ⎣1 + E (i, i) + j = 1 E (i, j ) ⎦
P0 (i) P0 (j )
j = i
(4.1)

In Eq. (4.1), E(i, i) and E(i, j) are the self-elasticity and cross-elasticity of demand
response, respectively. d0 (i) and d(i) are the demand of system (Consumers) before
and after applying for the RTP program, respectively. Moreover, the initial price of
each hour is represented by using P0 (i) while P(i) demonstrates the price of an hour
i after applying the RTP program.
102 H. Shayeghi and M. Alilou

4.4 Distributed Generation Units

Distributed generation units are one of the important technologies of microgrids

because the local production of electricity is the main proviso for calling a system
as a microgrid. The produced power of DG sources increases the reliability and
independence of the microgrid. The distribution company of MG utilizes the energy
of these technologies for supplying the demand of the MG and selling energy to
the upstream network. Of course, the distribution company has to buy energy from
the upstream grid when the local sources cannot provide the MG’s demand. Totally,
distributed generation units can be divided into nonrenewable and renewable units.
Both types of DG units are described in the following [18, 19].

4.4.1 Nonrenewable Units

The produced energy of nonrenewable units is stable and dispatchable because they
can generate the electricity until their initial energy is supplied. So, this type of DG
unit has a high effect on the reliability and stability of microgrids. In this chapter,
two technologies of nonrenewable units including micro turbine (MT) and fuel cell
(FC) are considered in the microgrids. The description and modeling of MT and FC
are presented as following [18, 19]:

4.4.1.1 Micro Turbine

Micro turbine is one of the useful technologies of nonrenewable units. An overview

of micro turbine is shown in Fig. 4.3. Totally, the micro turbine is a small turbine
like a jet engine that has the capability to operate on a variety of gaseous and liquid
fuels. This small turbine is connected to an electric generator. The combination of a
small turbine, electric generator, power electronic devices, and control equipment is
called micro turbine.
The not nature dependent behavior of micro turbine helps the operator of the
microgrid for providing the demand of the grid in urgently times using local energy
sources instead of buying energy from the upstream network. Moreover, the MT has
the unique ability to simultaneously produce electricity and heat. Micro turbine can
run on a variety of fuels, including natural gas, propane, and fuel oil. Also, MTs
have the ability to inject both active and reactive powers into the network.

4.4.1.2 Fuel Cell

A fuel cell is a nonrenewable DG technology that converts the chemical energy

from a fuel into electricity. Figure 4.4 demonstrates an overview of fuel cells.
4 Optimal Dispatch and Unit Commitment in Microgrids 103

Exhaust Heat

Gas Heat Exchanger

compressor

Fuel input

Produced
electricity Combustor
Air

Generator Compressor Turbine

Fig. 4.3 An overview of micro turbine

Produced Electricity

e– e–

Hydrogen In Oxygen In
e–

e–
e– e–
e–
e–

e–
e–
e–
e–
Water and Heat
Out
Anode Cathode
Electrolyte

Fig. 4.4 An overview of fuel cells

The fuel cell can provide highly reliable electricity so that it produces almost no
pollutants. The efficiency of fuel cells is more than 85%. Fuel cells produce only
water and heat in their energy-generating process. Thus, the fuel cell can be said as
an environmentally friendly technology.
104 H. Shayeghi and M. Alilou

It is worth mentioning that the technology of a fuel cell is different from the
technology of a battery because the FC can produce electricity continuously for
as long as fuel is supplied while in the battery, the electrical energy comes from
chemicals already present. The fuel cell produces just active power.

4.4.2 Renewable Units

Renewable distributed generation units are called to technologies that their initial
energy such as solar irradiance will never be exhausted. These units are also called
Green because they have an eco-friendly technology. Although renewable units
improve environmental indices of the microgrid, one of the main disadvantages
of these technologies is their unstable behavior due to their initial energy. In the
following, description and modeling of two common types of renewable units,
photovoltaic panel (PV) and wind turbine (WT), are presented [18, 19].

4.4.2.1 Photovoltaic Panel

Photovoltaic is one of the popular technologies of renewable DGs, especially in rural

and urban areas. PV is a solar power unit that uses solar cells or solar photovoltaic
arrays to convert the light of the sun (solar irradiance) directly into electrical power.
An overview of PV panels is shown in Fig. 4.5.
PV panels are the practical choice for providing the electricity demand of remote
areas due to the availability of solar energy approximately all points of the world.
The output power of each PV panel relates to the amount of solar irradiance, the
area, and the efficiency of the solar panel. Mathematically, the active power of the
photovoltaic panel can be calculated by Eq. (4.2).

PP V = Apv βμ (4.2)

In this equation, β is solar irradiance. The parameters Apv and μ show the area
and efficiency of the solar panel, respectively. PV panels can produce only active
power.

4.4.2.2 Wind Turbine

Renewable wind energy is approximately available in most parts of the world during
the year. For this reason, it is one of the proper sources for producing electricity. The
technology which is used to convert wind energy into electrical energy is called the
wind turbine. Figure 4.6 demonstrates an overview of wind turbines.
The output power of WT has a direct relation to wind speed and swept area
of the turbine. Moreover, air density and power coefficient affect the power of the
4 Optimal Dispatch and Unit Commitment in Microgrids 105

Produced
electricity

Solar
irradiance

Photons

Electron
flow

N-type silicon

Junction
P-type silicon

Fig. 4.5 An overview of photovoltaic panels

Propeller leg
Brake Electricity
Mechanical regulation system
transfer

Housing
Generator
Rotor management
system Turning system

Tower

Connection to the
Foundation grid

Fig. 4.6 An overview of wind turbine

106 H. Shayeghi and M. Alilou

wind turbine. Therefore, Eq. (4.3) can be utilized to calculate the active power of
WT units.
1
PW T = ρAwt Vw3 Cp (4.3)
2
Here, ρ and VW are air density and wind speed, respectively. The parameters
Awt and Cp demonstrate swept area and power coefficient of the wind turbine,
respectively. It is worth mentioning that the wind turbine consumes reactive power
to produce active power due to its induction generator.

4.5 Energy Storage System

Energy storage system (ESS) is a device that has the ability to store electrical energy
in the form of chemical energy and convert chemical energy into electricity when
required. The optimal charging and discharging schedule of energy storage systems
based on the condition of market price, the demand of the grid, and state of charge
can significantly improve the technical, economic, and environmental indices of the
microgrid. An overview of the energy storage system is shown in Fig. 4.7.
As mentioned earlier, one of the main advantages of microgrids is the ability
of local electricity production. The combination of distributed generation units
and energy storage system increases the level of self-consumed electricity in the
microgrid. With an ESS, the extra electrical energy of the MG is stored and used
later. Totally, the microgrid equipped with both DG and ESS units can reduce the
energy drawn from the upstream network. Therefore, the self-sufficiency of the
microgrid is increased. Moreover, using the ESS beside the DG units decrease the
dependence of the microgrids on the upper grid as well as reducing carbon emissions
[20].

DC current

Temperature
controller

DC switch AC AC transformer
breaker

AC current
Monitor and
Storage control Power conversion
system

Fig. 4.7 An overview of the energy storage system

4 Optimal Dispatch and Unit Commitment in Microgrids 107

4.6 Objective Functions and Constraints

In this chapter, the operational schedule of energy sources is optimized as a multi-

objective problem with considering some constraints and predetermined parameters.
In the following, the objective functions and constraints are explained completely.

4.6.1 Objective Functions

An economic–environmental objective function is considered for finding the best

unit commitment in the microgrid. Mathematically, the main objective function is
presented in Eq. (4.4).

Objectivefunction = max {PMGDC } , min {EMG } (4.4)

Here, PMGDC is the profit of the microgrid distribution company and EMG is the
pollution emission of the microgrid. They are formulated as follows.

4.6.1.1 The Profit of MGDC

The profit of the microgrid distribution company (MGDC) is considered as the

technical index of the unit commitment. The MGDC has to provide the demand
of the microgrid including a load of customers and the energy storage system.
It is presumed that the MGDC owns the renewable and nonrenewable distributed
generation units and the electrical storage system. So, the MGDC can use the
power of DGs and ESS to provide the part of the total demand of the grid;
another part of the demand is purchased from the upstream network. According
to the abovementioned data, the profit of the MGDC is composed of the following
terms:
1. The income from selling the energy to the customers.
2. The income from selling the energy to the upstream network.
3. The cost of purchased energy from the upstream network.
4. The cost of the produced energy of DG units.
5. The cost of saving energy to the ESS in charging mode.
Therefore, the profit of the MGDC is formulated based on the incomes and costs
in Eq. (4.5).

PMGDC = ID + ISB − CW M − CDG − CESS (4.5)

ID is the income from selling the energy to the customers. This income is presented
in Eq. (4.6). In this equation, nc and PijD are the number of customers and their
108 H. Shayeghi and M. Alilou

consumption at each interval, respectively. PrM j is the electricity price (market price)
at interval j. H is the number of time intervals in each day. It can be noted that the
time interval in this chapter is 1 h; so H is equal to 24.


nc 
H
ID = PijD × PrM
j (4.6)
i=1 j =1

ISB is the income from selling the extra produced energy to the upstream network.
So, when the sum of the produced power of DG units and the discharged energy of
ESS is more than a load of the microgrid, the MGDC could sell the extra power to
the upstream network. This income can be calculated by Eq. (4.7). In this equation,
PjSB and PrSBj are the amount of sold power to the upstream network and the price
of power at interval j, respectively.


H
ISB = PjSB × PrSB
j (4.7)
j =1

CWM is the cost of purchased energy from the upstream network. It is worth noting
that the MGDC, firstly, provides the possible amount of energy of the microgrid
using DGs and ESS; then it purchases extra energy from the upstream grid. This
cost is formulated in Eq. (4.8). In this equation, PjW M and PrW
j
M
are the purchased
power from the upstream network and the price of purchased electricity at time
interval j, respectively.


H
CW M = PjW M × PrW
j
M
(4.8)
j =1

CDG is the cost of the produced power of distributed generation units. This cost can
be calculated by Eq. (4.9). In this equation, nt is the number of different types of DG
units (Renewable and nonrenewable). nDG and PsDG i,j
are the number of DGs from
the considered type and their capacity at time interval j, respectively. PrDG
s is the
price of the production active power of the different types of DGs.

 
nt n DG 
H
CDG = PsDG
i,j
× PrDG
s (4.9)
s=1 i=1 j =1

CESS is the cost of the saved energy to the electrical storage system. It is worth
noting here that the charging of the ESS unit has only cost for the MGDC. In other
words, the discharging of ESS doesn’t have an extra cost for the MGDC because the
MGDC is the owner of the ESS unit. Mathematically, the cost of charging the ESS
Ch are the number of electrical
is given by Eq. (4.10). In this equation, nESS and PESS ij
storage systems and their charging energy at time interval j, respectively.
4 Optimal Dispatch and Unit Commitment in Microgrids 109

n
ESS 
H
CESS = Ch
PESS ij
× PrW
j
M
(4.10)
i=1 j =1

4.6.1.2 Pollution Emission of MG

Nowadays, most countries have special attention to the environmental aspects of

their decisions because the environmental condition gets worse every year. In this
chapter, pollution emission of the microgrid is considered as the environmental
issue of the optimization. Moreover, the pollution emission of central power plants
is combined with the environmental index of the MG when the MGDC has to
buy energy from the upstream network. The following five pollutant gasses are
considered for calculating the pollutant emission of the MG:
1. Carbon Monoxide (CO).
2. Carbon Dioxide (CO2 ).
3. Sulfur Dioxide (SO2 ).
4. Nitrogen Oxides (NOx ).
5. Particulate Matter (PM10 ).
Mathematically, the pollution emission of the microgrid considering the above
pollutant gasses is calculated by Eq. (4.11).
H nunit nP G
EMG = Puniti (h) × P Gij (4.11)
h=1 i=1 j =1

Here, nunit and nPG are the number of energy sources (DG, ESS, and upstream
network) and the number of pollutant gasses, respectively. Punit _ i (h) shows the
power of unit i at hour h while PGij is the rate of pollutant gas j of unit i.

4.6.2 Constraints

There are the following constraints during the implementation of the proposed unit
commitment in the microgrid.

4.6.2.1 Power Balance Constraint

This constraint presents the sum of the total power of various types of distributed
generation units (PDG ), the power energy storage system in discharge mode
(PESS, Dis ) and the purchased power from the upstream grid (PG2MG ) should be
equal to the sum of the total demand of the microgrid (Pdemand ), the power of the
energy storage system in charge mode (PESS, Ch ) and the sold power to the upstream
110 H. Shayeghi and M. Alilou

network (PMG2G ). Mathematically, the power balance constraint is calculated by Eq.

(4.12).
nDG n
PDGi + PG2MG + PESS,Dis = Pdemandj + PMG2G + PESS,Ch
i=1 j =1
(4.12)

4.6.2.2 Distributed Generation Constraint

The produced power of each type of DG unit should be in allowable size as the
following range:

PDG min ≤ PDGi ≤ PDG max (4.13)

Here, PDG min and PDG max are the minimum and maximum power of each DG
unit for producing the electricity, respectively.

4.6.2.3 Energy Storage System Constraint

There are limitations of charging and discharging in the energy storage system
during each hour. The Eqs. (4.14–4.18) are considered as constraints of the ESS.

XtESS,Ch + XtESS,Dis ≤ 1 (4.14)

ESS ≤ SOC t
SOC min ≤ SOC max
ESS
ESS (4.15)

0 ≤ PtESS,Ch ≤ RESS
Ch
(4.16)

0 ≤ PtESS,Dis ≤ RESS
Dis
(4.17)

   
ESS,Ch
SOC ESS
t = SOC ESS
t−1 + Pt × ηESS
Ch
− PtESS,Dis × ηESS
Dis
(4.18)

where, Eq. (4.14) shows that the charge and discharge of ESS are not simultaneous.
In this equation, XtESS,Ch and XtESS,Dis indicate a binary variable that shows the
charge and discharge state of ESS, respectively (0 or 1). The limitation of the total
state of charge (SOC) of ESS is shown in Eq. (4.15). In this equation, SOC ESS t
demonstrates SOC of ESS at time interval t while SOC min max
ESS and SOC ESS indicate
4 Optimal Dispatch and Unit Commitment in Microgrids 111

the minimum and maximum rate of SOC of ESS, respectively. The maximum
amount of charged power of the ESS from the grid and the maximum amount of
discharged power of ESS to the grid are satisfied by Eq. (4.16) and Eq. (4.17),
Ch and R Dis show the maximum charge and discharge rate of
respectively. RESS ESS
the ESS, respectively. Eq. (4.18) indicates that SOC of ESS at each time interval
(SOC ESS
t ) consists of the remaining SOC of ESS from the previous interval
ESS,Ch
(SOC ESS
t−1 ), the amount of power exchanged with the grid and ESS (Pt and
ESS,Dis Ch Dis
Pt ) and the charge and discharge efficiency (ηESS and ηESS ).

4.6.3 Optimization Algorithm

In this chapter, the combination of multi-objective grey wolf optimization

(MOGWO) and fuzzy method is utilized to multi-objective optimize the operational
schedule of energy sources of the microgrid. For getting the best result, firstly, the
MOGWO is used to optimize the multi-objective functions including economic
and environmental indices, and create the optimal Pareto-front. After applying the
MOGWO algorithm, the fuzzy method is used to find the optimal particle from the
non-dominated particles.

4.6.3.1 Multi-Objective Grey Wolf Optimization Algorithm

Intelligent algorithms are usually inspired by existing natural behaviors of nature.

The optimization method of MOGWO is inspired by the grey wolves. Grey wolves
that belong to the Canidae family are considered apex predators. This means that
they are at the top of the food chain. Grey wolves mostly prefer to live in a pack
with group size.
The MOGWO is based on the social behavior of Wolves. In this meta-heuristic
algorithm, the best solution is considered as α wolf. β and δ wolves are the second
and third solutions, respectively. The rest of the particles are assumed as ω wolves.
The optimization method of the MOGWO is managed by alpha, beta, and delta so
that the omega wolves follow α, β, and δ wolves for reaching the best result. The
position updating of search agents (ω wolves) based on the position of alpha, beta,
and delta wolves is demonstrated in Fig. 4.8.
So in this algorithm, the position of each particle is updated using Eq. (4.19) [21].

X (t + 1) = XP (t) − A.D (4.19)

Here,

D = |C.XP (t) − X(t)| (4.20)

112 H. Shayeghi and M. Alilou

a1
C2
a

a2
C2
b
R

prey
Dalpha

Dbeta

Move

d Ddelta w

a3
C3

Fig. 4.8 Position updating of search agents (ω wolves) based on α, β, and δ in MOGWO algorithm

A = 2a.r1 − a (4.21)

C = 2r2 (4.22)

Here, XP and X show the position vector of the prey and the position vector of a
grey wolf, respectively. Vectors of A and C are the coefficients. Moreover, a linearly
decrease from 2 to 0 over the iterations while vectors of r1 and r2 are random in [0
and 1].
The extra detail of the multi-objective grey wolf optimization algorithm is
available in reference [21].

4.6.3.2 Fuzzy Method

After optimizing the economic–environmental objective functions of unit com-

mitment in the microgrid, the fuzzy decision-making method run to select the
optimal compromise solution that represents the optimal amount of economic and
4 Optimal Dispatch and Unit Commitment in Microgrids 113

environmental indices equal to the best operational schedule of various type of

distributed generation units and energy storage system in the microgrid.
In the fuzzy decision-making method, the best result is selected in three steps
[22]:
• Firstly, the membership values of each particle based on various objective
functions are calculated by Eq. (4.23).
⎧
⎪
⎨ 1 Fi k ≤ Fi min
Fi max −Fi k
μi k = Fi min < Fi k < Fi max (4.23)
⎪
⎩ Fi max −Fi min
0 Fi max ≤ Fi k

Here, Fi min and Fi max demonstrate the lower and upper bound of index i,
respectively. Fi k shows the amount of particle k based on the objective function i.
• Secondly, Eq. (4.24) is utilized to calculate the total membership value of each
non-dominated particle which is in the Pareto-front.

#N O k
μi
μk = #N K i=1
#N O (4.24)
k=1 i=1 μi k

• Thirdly, each particle that has the highest amount of total membership value is
selected as the best compromise solution.
Consequently, the complete method for optimizing the hourly energy schedule
of various energy sources of the microgrid is demonstrated in Fig. 4.9. This figure
presents that the best unit commitment in the microgrid is selected in four sections
including (1) input details of the problem; (2) initial steps; (3) MOGWO steps; (4)
fuzzy decision steps.

4.7 Numerical Results

In this section, the proposed unit commitment method is tested on a sample

microgrid. The hourly electricity consumption of the microgrid is demonstrated in
Fig. 4.10. The distribution company of microgrid has a micro turbine, fuel cell,
wind turbine, photovoltaic panel, and energy storage system. Thus, the distribution
company of microgrid can utilize from the produced power of energy sources of the
microgrid and the bought energy from the upstream grid for providing the electricity
demand of the microgrid.
The micro turbine is a CAPSTONE C1000S with a capacity of 1000 KW. The
HD85 type of BALLARD Company is considered as fuel cell units of the microgrid.
114 H. Shayeghi and M. Alilou

Start

Input details of the problem including:

• Hourly loads of the microgrid
• Details of DG units and ESS
• Hourly variation of solar irradiance and wind speed
• Economic and environmental parameters
• Parameters of considered constraints
• Details of demand response program

Initial steps including:

• Create random particles
• Evaluate the constraints
• Calculate the amounts of objective functions

MOGWO steps including:

• Select the alpha, beta and delta particles
• Update position of particles based on MOGWO method
• Evaluate the constraints
• Calculate the amount of objective functions
• Update the iteration counter
• Repeat above steps until reaching the maximum iteration

Fuzzy decision steps including:

• Calculate the membership values of each non-dominated particle
• Calculate the total membership value of each particle
• Select the best particle with the highest amount of total membership value
• Determine the best unit commitment in the microgrid
• Calculate the amount of economic and environmental objective functions

End

Fig. 4.9 Flowchart of the unit commitment in the microgrid

This type of fuel cell unit can inject 85 KW to the grid. The used wind turbine in this
chapter is AW1500 type of NORDEX Company with 1500 KW capacity. Moreover,
the cut-in, normal, and cutout speeds are 4, 13, and 25 m/s, respectively [23, 24].
The 310 OS5 types of IBC-SOLAR Company are utilized for producing electri-
cal energy from solar power. The capacity of the PV panel is 310 KW. Moreover,
the energy storage system of ECCINC Company with a capacity of 400 KWh is
operated as an ESS unit of the microgrid. The hourly charging and discharging
limit of ESS battery is 100 KW. According to variations of wind speed and solar
irradiance [6], the hourly abilities of wind turbines and photovoltaic panels for
producing electrical energy are shown in Figs. 4.11 and 4.12, respectively.
Figure 4.13 shows the hourly market price of the microgrid. This price is based
on the RTP method. Moreover, The costs of producing 1 Kwh using a micro turbine,
fuel cell, wind turbine, and photovoltaic are 0.043, 0.039, 0.037, and 0.038 $,
4 Optimal Dispatch and Unit Commitment in Microgrids 115

4000
3500
3000
2500
Kw

2000
1500
1000
500
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour

Fig. 4.10 Hourly demand for considered microgrid

1600

1400

1200

1000
Kw

800

600

400

200

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour

Fig. 4.11 Hourly ability of WT for producing electricity

respectively. Microgrid buys energy from the upstream network with a price of 0.12
$/Kwh. Moreover, microgrid sells energy to the upstream grid with the mean price
of the daily market price.
The amounts of emitted pollutant gases (carbon monoxide, carbon dioxide, sulfur
dioxide, nitrogen oxides, and particulate matter) of renewable and nonrenewable
DGs, ESS and central power plant are presented in Table 4.1.
For better evaluating the results, the proposed unit commitment method is
evaluated in the sample microgrid without and with considering the demand
response program in the sections 4.8.1 and 4.8.2, respectively.
116 H. Shayeghi and M. Alilou

350
325
300
275
250
225
200
Kw

175
150
125
100
75
50
25
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour

Fig. 4.12 Hourly ability of PV for producing electricity

0.6
0.55
0.5
0.45
0.4
$/Kwh

0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour

Fig. 4.13 Hourly market price of electricity

4.7.1 Without DR Program

In this section, it is considered that the demand response program is not applied
to the microgrid. So, the hourly load variation of the microgrid is equal to Fig.
4.10. As mentioned earlier, the combination of the MOGWO and fuzzy method is
utilized to optimize the economic–environmental objective function and find the
best unit commitment in the considered microgrid. The optimal hourly power of
various energy sources of the microgrid without real-time pricing DR program is
presented in Table 4.2.
4 Optimal Dispatch and Unit Commitment in Microgrids 117

Table 4.1 The environmental details of various energy sources

Pollution gases rate (Kg/Kwh)
Type of energy source CO2 SO2 NOx CO PM10
Micro turbine 0.72 0.002 0.091 0.247 0.018
Fuel cell 0.46 0.012 0.006 0.002 0
Photovoltaic panel 0 0 0 0 0
Wind turbine 0 0 0 0 0
Energy storage system 0.02 0 0.00001 0.0003 0.001
Central power plant (received 0.85 2.14 9.723 6.043 0.87
energy from upstream
network)

Table 4.2 The optimal unit Power of energy resources (KW)

commitment in the microgrid
Hour MT FC PV WT ESS Grid
without RTP program
1 0.0000 0.00 0.000 1411.80 100.0 0.00000
2 21.000 10.0 0.000 1429.35 0.000 0.00000
3 221.00 71.0 0.000 1376.40 −100.0 0.00000
4 156.00 21.0 0.000 1411.80 −100.0 0.00000
5 69.000 74.0 0.000 1235.25 0.000 0.00000
6 115.00 85.0 15.50 904.270 −100.0 0.00000
7 0.0000 0.00 46.50 1052.12 66.78 0.00000
8 615.80 2.00 104.0 1235.25 47.00 0.00000
9 1000.0 41.0 201.5 1147.05 −37.00 0.00000
10 1000.0 12.0 279.0 882.300 29.42 419.700
11 1000.0 85.0 303.8 705.900 0.000 1256.23
12 1000.0 85.0 310.0 794.100 0.000 1307.46
13 1000.0 85.0 279.0 882.300 0.000 1177.44
14 1000.0 85.0 217.0 1023.60 100.0 1216.59
15 1000.0 85.0 170.5 1058.85 100.0 1300.65
16 1000.0 85.0 93.00 1235.25 100.0 1056.12
17 1000.0 85.0 62.00 1500.00 −100.0 803.930
18 1000.0 85.0 15.50 1358.85 0.000 745.580
19 1000.0 85.0 0.000 1235.25 0.000 1030.68
20 1000.0 85.0 0.000 1411.80 100.0 681.316
21 1000.0 85.0 0.000 1376.40 100.0 716.716
22 1000.0 85.0 0.000 1235.25 100.0 493.424
23 1000.0 85.0 0.000 1129.35 −100.0 435.255
24 1000.0 59.0 0.000 1200.00 −14.00 0.00000

According to Table 4.2, the renewable DG units have an important role in power
generation; the produced power of wind turbines and photovoltaic panels is about
48.97% of total provided energy. Of course, the wind turbine unit produces more
electricity than a photovoltaic panel. According to the result, the produced power of
wind turbines is 93% of total daily renewable energy while the produced power of
118 H. Shayeghi and M. Alilou

48.97
50
45
40
35 30.15
30
% 25 20.41
20
15
10
5 0.47
0
Nonrenewable Renewable DGs ESS grid
DGs

Fig. 4.14 The percentage participation of various types of sources in unit commitment

the photovoltaic panel is only 7% of total renewable electrical power. The 24.36%
of the load is produced by nonrenewable DG units. The energy storage system
provides only 0.47% of total electricity. Although the upstream grid provided the
total electricity demand of the microgrid before utilizing DG units and the optimal
unit commitment, the provided power by the grid is about 20% of total demand after
applying the proposed method to the microgrid. In other words, energy sources
of the microgrid can provide the load of system at most times; therefore, it can
be said that the upstream grid is the backup source after doing the proposed unit
commitment. Moreover, it is worth mentioning that the microgrid sells electrical
power to the upstream network at some hours. According to statistics, the microgrid
has the ability to inject about 2718 KWh electrical energy into the upstream
grid during the day. The percentage participation of various energy sources in
the proposed unit commitment in the microgrid without considering the demand
response program is demonstrated in Fig. 4.14.
Table 4.3 demonstrates hourly amounts of economic and environmental indices
of the microgrid with and without considering the proposed unit commitment
method. Results demonstrate that the economic and environmental indices of the
microgrid are improved after applying the proposed unit commitment method.
The hourly profits of the distribution company of the microgrid increase between
14% and 677% after utilizing the energy sources using the optimal unit commitment.
The daily statistic data of the economic index is presented in Table 4.4. As can be
shown in this table, the daily profit of the microgrid is about 4500$ more than the
initial case. Moreover, the minimum and maximum amounts of hourly profit of the
microgrid are increased so that the daily average of the considered economic index
is improved by about 24.95% after utilizing the proposed unit commitment. Thus,
the distribution company of the microgrid earns more profit after optimal utilizing
the nonrenewable and renewable DG sources and energy storage system.
4 Optimal Dispatch and Unit Commitment in Microgrids 119

Table 4.3 Hourly amounts of economic and environmental indices of the microgrid without RTP
program
Without DGs and ESS With unit commitment
Hour Profit ($) Pollution emission (Kg) Profit ($) Pollution emission (Kg)
1 346.69866 15633.63315 550.374473 2.1310000
2 209.77119 14714.28358 411.147892 27.438000
3 87.406520 13794.93402 302.747966 270.18700
4 30.593025 12875.58446 237.969899 176.11700
5 52.082443 11956.23489 283.353635 109.90200
6 158.06396 12875.58446 198.899998 162.63900
7 238.90607 14714.28358 338.058257 1.4229752
8 515.74230 22073.77065 749.641502 665.74007
9 655.54890 27589.86804 874.221344 1096.8915
10 812.94974 33110.65598 990.560855 6383.5191
11 1072.2976 42308.84218 1239.68390 16979.959
12 1153.8641 44147.54131 1329.07944 17626.765
13 1095.5981 43228.19174 1275.59198 15985.208
14 1147.2886 45986.24044 1345.92639 16481.546
15 1133.0750 46905.59000 1330.82555 17542.937
16 1088.6585 45066.89087 1294.69521 14455.527
17 1139.3162 42308.84218 1301.55603 11267.089
18 1217.8736 40465.45249 1415.81414 10532.499
19 1440.8999 42308.84218 1627.31065 14132.165
20 1163.7312 41389.49262 1376.79558 9723.2268
21 1081.7783 41389.49262 1291.90448 10170.187
22 961.51258 36788.05424 1159.92334 7350.9087
23 777.62937 32191.30642 904.021256 6612.1923
24 530.29396 23912.46978 800.100575 1106.0217

Table 4.4 Daily statistics parameters of the economic index of the microgrid
Profit ($)
Without DGs and ESS With unit commitment
Sum of hourly profits 18111.58 22630.21
Minimum of hourly profits 30.59000 198.9100
Maximum of hourly profits 1440.900 1627.310
Average of hourly profits 754.6500 942.9300

The environmental index of the microgrid is improved considerably after apply-

ing the proposed unit commitment method. The impact of renewable DG sources
in this improvement is significant. According to the results, the emitted pollutant
gasses of the microgrid are reduced by about 76% after utilizing the DGs and
ESS. Minimum hourly amounts of the environmental index without and with unit
commitment are 11.96 Mg and 1.42 Kg, respectively. On the other hand, the
120 H. Shayeghi and M. Alilou

maximum hourly amounts of this index without and with unit commitment are
46.91 Mg and 17.62 Mg, respectively.
Totally, it can be said that the proposed unit commitment method considerably
improves the economic and environmental indices of the microgrid without con-
sidering the demand response program using the nonrenewable and renewable DG
units and energy storage system.

4.7.2 With DR Program

In this section, the unit commitment is applied to the microgrid in the presence of
a demand response program. So, in this situation, the consumers of the microgrid
are encouraged to improve their consumption curve based on the real-time pricing
method.
After applying for the demand response program, the hourly load variation of
the microgrid becomes more linear than without considering the demand response
program. The consumption pattern of the microgrid in the presence of the RTP
program is demonstrated in Fig. 4.15. As can be shown in this figure, the difference
between the minimum and maximum electricity demand of microgrid is consider-
ably decreased by participating the customers in a demand response program. This
significant reduction causes to improve the efficiency of the electricity and increase
the performance of the microgrid.
As mentioned earlier, the unit commitment of the microgrid in the presence of
the RTP program is optimized using the combination method of multi-objective
grey wolf optimization and fuzzy method. The hourly produced power of various
energy sources is presented in Table 4.5. According to this table, the impact of

3500

3000

2500

2000
Kw

1500

1000

500

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour

Fig. 4.15 Improved demand of the microgrid after applying for the DR program
4 Optimal Dispatch and Unit Commitment in Microgrids 121

Table 4.5 The optimal unit Hour Power of energy resources (KW)
commitment in the microgrid
MT FC PV WT ESS Grid
with RTP program
1 651.76 85 0.000 1399.19 73.19 0.00000
2 656.00 18 0.000 1429.35 0.000 0.00000
3 1000.0 85 0.000 1376.40 44.00 0.00000
4 569.92 85 0.000 1411.80 −100.0 0.00000
5 1000.0 85 0.000 1235.25 −100.0 0.00000
6 1000.0 85 15.50 1252.95 −100.0 0.00000
7 1000.0 85 46.50 1252.95 47.00 0.00000
8 254.75 82 124.0 1235.25 89.00 0.00000
9 914.45 81 201.5 1147.05 34.00 0.00000
10 1000.0 85 279.0 882.300 100.0 348.932
11 1000.0 85 303.8 705.900 70.00 312.091
12 1000.0 85 310.0 794.100 0.000 578.946
13 1000.0 85 279.0 882.300 100.0 858.631
14 1000.0 85 217.0 1023.60 100.0 415.260
15 1000.0 85 170.5 1058.85 100.0 644.951
16 1000.0 85 93.00 1235.25 0.000 573.238
17 1000.0 85 62.00 1500.00 −100.0 148.233
18 1000.0 85 15.50 1358.85 100.0 427.138
19 1000.0 85 0.000 1235.25 18.12 793.750
20 1000.0 85 0.000 1411.80 −100.0 0.00000
21 984.39 16 0.000 1376.40 100.0 0.00000
22 1000.0 85 0.000 1235.25 56.00 173.355
23 1000.0 85 0.000 1129.35 0.000 335.254
24 1000.0 85 0.000 1200.00 100.0 310.232

renewable energy sources is increased than the microgrid without the demand
response program. Totally, the produced power of wind turbines and photovoltaic
panels is about 50.34% of all provided energy. The participation of renewable units
is increased by about 3% than the situation that the RTP program is not considered.
Moreover, the produced power of wind turbines is 93.14% of total daily renewable
energy while the produced power of photovoltaic panels is only 6.86% of total
renewable electrical power. The difference between the produced power of the wind
turbine and the photovoltaic panel is related to their capacity and availability of their
initial energy including wind speed and solar irradiance.
The participation of nonrenewable DG units is also increased when the demand
response program is utilized in the microgrid. The 38.99% of total electrical energy
is produced by nonrenewable DG units. Therefore, the produced power of this
type of DG technologies is increased by about 30% than the microgrid without the
demand response program. The injected power of the energy storage system is also
raised more than 100% so that it provides 1.03% of total electrical energy.
Although the upstream grid provided about 20% of the total demand of the
microgrid without applying for the demand response program, the participation of it
122 H. Shayeghi and M. Alilou

46.88
50
45
40 35.91
35
30
% 25
20
15 9.64
10
3.09 3.45
5 1.03
0
MT FC PV WT ESS Grid

Fig. 4.16 The percentage participation of energy sources in unit commitment in the presence of
DR program

in providing the daily demand is only 9% after utilizing both unit commitment and
demand response program. Thus the impact of the upstream grid in providing the
demand of the microgrid is reduced 53% using the demand response program.
The percentage participation of nonrenewable and renewable distributed genera-
tion units, energy storage systems and grid in the proposed unit commitment in the
microgrid with considering the real-time pricing program is demonstrated in Fig.
4.16. The wind turbine has the highest impact on providing the daily demand of the
microgrid.
The hourly amounts of economic and environmental indices of the microgrid
in the presence of a demand response program with and without considering the
proposed unit commitment method are presented in Table 4.6. According to the
results, the profit of the distribution company and the pollution gasses of the
microgrid are improved after utilizing the proposed unit commitment method and
demand response program.
According to the results of Table 4.6, the hourly profits of the distribution
company of the microgrid increase between 15% and 250% after utilizing the energy
sources using the optimal unit commitment. The daily statistic data of the economic
index is presented in Table 4.7.
As can be shown in Table 4.7, the daily profit of the microgrid is about 4914.8$
more than the initial case. Moreover, the minimum and maximum amounts of hourly
profit of the microgrid are increased so that the daily average of the considered
economic index is improved by about 29.11% after utilizing the proposed unit
commitment. Thus, the distribution company of the microgrid earns more profit
after optimal utilizing the nonrenewable and renewable DG sources and energy
storage system using the proposed unit commitment method. Moreover, the demand
response program increases the effect of the energy sources of the microgrid on the
energy management of the microgrid.
The environmental index of the microgrid is also improved considerably after
applying for both the unit commitment method and demand response program.
4 Optimal Dispatch and Unit Commitment in Microgrids 123

Table 4.6 Hourly amount of economic and environmental indices of the microgrid with RTP
program
Without DGs and ESS With unit commitment
Hour Profit ($) Pollution emission (Kg) Profit ($) Pollution emission (Kg)
1 611.84564 27589.8681 800.36088 744.95089
2 377.61489 26487.5866 549.71568 715.80800
3 151.51256 23912.4698 521.42321 1119.7376
4 59.001630 24831.8193 175.72078 653.04383
5 100.16011 22993.1202 344.02882 1116.6690
6 277.77798 22627.2566 541.31707 1116.6690
7 367.38378 22627.2566 738.80195 1119.8015
8 515.74231 22073.7706 675.37263 315.87709
9 655.54890 27589.8681 900.83861 1025.3816
10 835.52208 34030.0055 1027.5160 5526.5527
11 792.57296 31271.9569 968.35926 5060.7463
12 913.45534 34949.3551 1088.6706 8428.5785
13 1025.5777 40465.4525 1217.5716 11961.999
14 894.87106 35868.7047 1093.5089 6364.0100
15 933.08726 38626.7534 1130.8378 9264.1012
16 910.87899 37707.4038 1104.9157 8356.5093
17 916.37905 34030.0056 1078.6189 2988.2525
18 1134.8656 37707.4038 1344.8062 6513.9817
19 1346.8101 39546.1029 1535.3948 11141.073
20 879.26063 31271.9569 1007.3151 1116.6690
21 817.34086 31271.9569 1020.6761 1070.9839
22 841.36949 32191.3064 1034.5002 3308.7672
23 777.62937 32191.3064 955.25042 5351.7233
24 754.66510 34030.0056 950.15010 5037.9265

Table 4.7 Daily statistics parameters of the economic index of the microgrid in the presence of
the RTP program
Profit ($)
Without DGs and ESS With unit commitment
Sum of hourly profits 16890.87 21805.67
Minimum of hourly profits 59.00000 175.7200
Maximum of hourly profits 1346.810 1535.390
Average of hourly profits 703.7800 908.5600

The impact of renewable DG sources in this improvement is significant because

their pollution is zero. According to the results, the emitted pollutant gasses of the
microgrid are reduced by about 86.67% after utilizing the DGs and ESS. Moreover,
the environmental index is also decreased by about 13.91% than the situation
that unit commitment is utilized in the microgrid without considering the demand
response program. Minimum hourly amounts of the environmental index without
124 H. Shayeghi and M. Alilou

and with unit commitment are 22.07 Mg and 0.31 Mg, respectively. On the other
hand, the maximum hourly amounts of this index without and with unit commitment
are 40.46 Mg and 11.96 Mg, respectively.
Therefore, the combination of the proposed unit commitment method and
demand response program significantly increases the economic index and decreases
the environmental index of the microgrid.

4.8 Conclusion

Optimal dispatch and unit commitment of various energy sources of the microgrid
has a high effect on the efficiency of both microgrid and upstream grid. Moreover,
increasing the interest of consumers for optimal managing the consumption using
demand response programs increases the performance of the microgrid. For these
reasons, the hourly operational schedule of energy sources of the microgrid was
optimized with and without the demand response program in this chapter. Micro
turbine, fuel cell, wind turbine, photovoltaic panel, and energy storage system were
the energy sources of the microgrid which provide the demand of the microgrid.
Also, the microgrid could buy/sell energy from/to the upstream grid when the
produced power of energy sources of the microgrid is lower/more than the demand
of the microgrid. The combination of multi-objective grey wolf optimization algo-
rithm and the fuzzy set method was used to optimize the economic–environmental
objective function and select the best operational schedule of energy units.
The numerical results demonstrate that the proposed method can properly
optimize a schedule for unit commitment in the microgrid so that considered indices
of the microgrid are improved considerably after applying the proposed method. The
economic index of the microgrid is increased by about 25–29% in various cases. On
the other hand, the considered environmental index is decreased by about 76–87%
in different tests. Based on the participation of energy sources in unit commitment,
it can be said that renewable units have a high effect on providing the demand of the
microgrid. The wind turbine produces the highest amount of electricity during the
day. The demand response program has also a positive effect on the performance of
the microgrid. Totally, the results prove that the proposed unit commitment method
has a suitable performance in increasing the efficiency of the microgrid.

References

1. Shayeghi, H., & Alilou, M. (2015). Application of multi objective HFAPSO algorithm for
simultaneous placement of DG, capacitor and protective device in radial distribution network.
Journal of Operation and Automation in Power Engineering, 3, 131–146.
4 Optimal Dispatch and Unit Commitment in Microgrids 125

2. Shayeghi, H., & Shahryari, E. (2017). Integration and management technique of renewable
energy resources in microgrid. Energy Harvesting and Energy Efficiency, 37, 393–421.
3. Gelazanskas, L., & Gamage, A. (2014). Demand side Management in Smart Grid: A review
and proposals for future direction. Sustainable Cities and Society, 11, 22–30.
4. Behrangrad, M. (2015). A review of demand side management business models in the
electricity market. Renewable and Sustainable Energy Reviews, 47, 270–283.
5. Heydarian, E., & Aalami, H. A. (2016). Multi objective scheduling of utility-scale energy
storages and demand response programs portfolio for grid integration of wind power. Journal
of Operation and Automation in Power Engineering, 4, 104–116.
6. Alilou, M., Nazarpour, D., & Shayeghi, H. (2018). Multi-objective optimization of demand
side management and multi DG in the distribution system with demand response. Journal of
Operation and Automation in Power Engineering, 6, 230–242.
7. Shakouri, H., & Kazemi, A. (2017). Multi-objective cost-load optimization for demand side
Management of a Residential Area in smart grids. Sustainable Cities and Society, 32, 171–
180.
8. Wu, Z., Tazvinga, H., & Xia, X. (2015). Demand side Management of Photovoltaic-Battery
Hybrid System. Applied Energy, 148, 294–304.
9. Yao, E., Samadi, P., Wong, V., & Schober, R. (2016). Residential demand side management
under high penetration of rooftop photovoltaic units. IEEE Transactions on Smart Grid, 7,
1597–1608.
10. Kotur, D., & Durisic, Z. (2017). Optimal spatial and temporal demand side Management in a
Power System Comprising Renewable Energy Sources. Renewable Energy, 108, 533–547.
11. Sfikas, E., Katsigiannis, Y., & Georgilakis, P. (2015). Simultaneous capacity optimization of
distributed generation and storage in medium voltage microgrids. Electrical Power and Energy
Systems, 67, 101–113.
12. Xu, G., Cheng, H., Fang, S., Ma, Z., Zeng, P., & Yao, L. (2018). Optimal size and location
of battery energy storage Systems for Reducing the wind power curtailments. Electric Power
Components and Systems, 46, 342–352.
13. Y. Wu, Sh. Chang, L. Chang, D. Viet, “Unit commitment in a high wind-power penetration
system”, Energy Procedia, Vo. 156, pp.18–22, 2019.
14. Garlik, B., & Krivan, M. (2013). Renewable energy unit commitment, with different acceptance
of balanced power, solved by simulated annealing. Energy and Buildings, 67, 392–402.
15. Gholami, K., & Dehnavi, E. (2019). A modified particle swarm optimization algorithm
for scheduling renewable generation in a micro-grid under load uncertainty. Applied Soft
Computing, 78, 496–514.
16. Avril, S., Arnaud, G., Florentin, A., & Vinard, M. (2010). Multi-objective optimization of
batteries and hydrogen storage Technologies for Remote Photovoltaic Systems. Energy, 35,
5300–5308.
17. Aalami, H., Moghaddam, M., & Yousefi, G. (2010). Modeling and prioritizing demand
response programs in power markets. Electric Power Systems Research, 80, 426–435.
18. Viral, R., & Khatod, D. (2012). Optimal planning of distributed generation systems in
distribution system: A review. Renewable and Sustainable Energy Reviews, 16, 5146–5165.
19. Maczulak, A. (2010). Renewable energy sources and methods. Green Technology - Facts on
File, Inc..
20. Akbari, H., Browne, M. C., Ortega, A., Huang, M. J., Hewitt, N. J., Norton, B., & McCormack,
S. J. (2019). Efficient energy storage Technologies for Photovoltaic Systems. Solar Energy,
192, 144–168.
21. Mirjalili, S., Saremi, S., Mirjalili, M., & Coelho, L. (2016). Multi-objective grey wolf opti-
mizer: A novel algorithm for multi-criterion optimization. Expert Systems with Applications,
47, 106–119.
22. Alilou, M., Talavat, V., & Shayeghi, H. (2018). Simultaneous placement of renewable DGs
and protective devices for improving the loss, reliability and economic indices of distribution
system with nonlinear load model. International Journal of Ambient Energy, In Press.
126 H. Shayeghi and M. Alilou

23. Modi, A., Bühler, F., Andreasen, J., & Haglind, F. (2017). A review of solar energy based heat
and power generation systems. Renewable and Sustainable Energy Reviews, 67, 1047–1064.
24. Stathopoulos, T., Alrawashdeh, H., Al-Quraan, A., Blocken, B., Dilimulati, A., Paraschivoiu,
M., & Pilay, P. (2018). Urban wind energy: Some views on potential and challenges. Journal
of Wind Engineering and Industrial Aerodynamics, 179, 146–157.
Chapter 5
The Role of Energy Storage Systems
in Microgrids Operation

Sidun Fang and Yu Wang

5.1 Introduction

5.1.1 Background

Generally, a microgrid can be defined as a local energy district that incorporates

electricity, heat/cooling power, and other energy forms, and can work in connection
with the traditional wide area synchronous grid (macrogrid) or “isolated mode” [1].
The flexible operation pattern makes the microgrid become an effective and efficient
interface to integrate multiple energy sources, such as distributed generators,
energy storage, and so on [2]. Additionally, with the development of transportation
electrification, electrified vehicles, ships, or even aircraft become available, which
introduces another type of special microgrids, that is, mobile microgrids [3]. This
type of microgrids mostly work in autonomous conditions and sometimes they will
connect to the main grid, that is, when electrified vehicles are charging, or electrified
ships berth into a seaport.
No matter which type of microgrid is, the grid-connected and islanded modes
are two typical operation patterns, and to accomplish different tasks and needs,
microgrids will supply power to different equipment, such as the cooling/heat
equipment in a residential area, or the air-conditioning power for a data center, or
the power consumed by port cranes in a seaport, or charging the parked electrical
vehicles. In this way, the energy storage system (ESS) is an important component
in a microgrid to act as an energy/power buffer between the generation side and
demand side. Lots of literature focus on this topic and fundamentally prove the
great effects of ESS in microgrid operation, that is, to facilitate the mitigation of

S. Fang () · Y. Wang

School of Electrical and Electronics Engineering, Nanyang Technological University, Singapore,
Singapore

© Springer Nature Switzerland AG 2021 127

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_5
128 S. Fang and Y. Wang

Fig. 5.1 Classification of microgrids

renewable energy source, or to benefit the autonomous operation, or to provide

uninterruptible power to critical components, and so on [4].
In this chapter, the role of ESS in different types of microgrids will be illustrated
in detail, that is, in both conventional land-based microgrids and mobile microgrids,
and the microgrids discussed in this chapter are classified as the following Fig. 5.1.

5.1.2 Land-based Microgrids

5.1.2.1 Residential Microgrid

Residential microgrid is the most conventional type of microgrid since the concept
of microgrid is raised. It is designed to supply power to a residential area, which may
consist of several buildings and blocks. The following figure gives the topology of
a typical AC/DC multi-energy microgrid (MEMG) [5].
From Fig. 5.2, the energy of MEMG is supplied from three sources, that is,
photovoltaics(PVs), electrical substation, and gas pressure house. The energy from
PVs is collected by the DC bus and the substation injects electricity to the AC bus.
Additionally, to improve the system flexibility, a battery ESS, two thermal storages,
and gas storage are incorporated. In the MEMG, multiple energy forms, that is,
electricity, heat/cooling power, and gas are coupled together. Excess electricity can
convert to gas by the power to gas (P2G) equipment and heating power by the power
to thermal (P2T) equipment. The gas can convert to electricity by a combined heat
plant (CHP) and the by-product is thermal power, or directly convert to thermal
power by the gas boiler.
In the MEMG, the generation side and demand side may not be always matched
to each other, which motivates the investment of ESS. Additionally, with the
development of transportation electrification, electric vehicles will become more
popular in the future, and the charging power will become an important service
5 The Role of Energy Storage Systems in Microgrids Operation 129

Fig. 5.2 A typical AC/DC multi-energy residential microgrid [5] (Permission for usage from the
author)

load in the MEMG. However, the arrivals/departures of electric vehicles are usually
uncertain and cannot be accurately forecasted, which further inquires the optimal
operation of ESS.

5.1.2.2 Industrial Microgrids

Industrial microgrids provide power to different commercial consumers, such as

the data center, industrial park, and seaport. The following figure gives a typical
structure of seaport microgrid [3].
Compared with MEMG, seaport microgrid is a newly proposed concept for
seaport management after a high level of electrification [6]. It generally defines as a
harbor territory that has the own energy plan and involves multiple renewable energy
penetrations [7].
From Fig. 5.3, the seaport microgrid is connected with the main grid by a
transformer. The harbor wind farm and PV are integrated to enhance the energy
efficiency of the seaport. Electricity is the only secondary energy form to drive
various logistic equipment, such as cold-ironing, and port cranes. Two practical
seaport microgrid projects in Hamburg (German) and Genoa (Italy) [8] have proved
its validity. Similar to the seaport microgrid, other industrial microgrids also have
similar topologies. The main difference lies in the service load demand type. For
example, seaport microgrid serves the logistic load demand, and the data center
microgrid supplies power to the data center and the cooling power. Generally, for
industrial applications, reliability is highly correlated with the economy, which
motivates the ESS installment in industrial microgrids.
130 S. Fang and Y. Wang

Fig. 5.3 A typical industrial microgrid (seaport) [3] (Permission for usage from the author)

5.1.3 Mobile Microgrids

The mobile microgrid is a new type of microgrids in the trend of transportation

electrification, including various electric vehicles, ships, and aircrafts [3, 9]. Mobile
microgrids mostly work in isolated mode and also can connect to the main grid
in some operating conditions, such as charging of electrical vehicles, and berthed
in of ships. Since the mobile microgrids mostly work in an isolated mode, the
integration of ESS is essential for the operation and control of mobile microgrids.
In the following, an all-electric ship (AES) is illustrated as a representative case of
a mobile microgrid.
AES is a new type of ship recently which replaces the mechanically driven
propulsion system by the electrically driven propulsion system. AES has proved to
have high energy efficiency than conventional ships and becomes the future trend of
shipping design. As illustrated in Fig. 5.4, AES dispatches energy via an integrated
power system onboard (shipboard microgrid), which consists of an energy network
(blue lines and arrows) and a communication network (green lines and arrows).
The power sources of an energy network include generators and battery, and some
ships may have renewable energy integrations (PV modules in Fig. 5.4). The load
demands of energy networks include the propulsion load and service load, that is, the
onboard radar, navigation system, air conditioning, and so on. The communication
network is used to send dispatch signals to each component to control the energy
network.
5 The Role of Energy Storage Systems in Microgrids Operation 131

Fig. 5.4 Typical structure of all-electric ship [3] (Permission for usage from the author)

5.1.4 Comparisons between Different Types of Microgrids

From above, microgrids are defined as a local energy district to integrate various
energy sources to supply multiple energy demands, but there are still some
differences in the following two aspects.

5.1.4.1 Operation Modes

Generally, microgrids can work in both grid-connected mode and isolated mode.
However, different types of microgrids have different durations of operation modes,
which will influence the planning, operation, and control of ESS. For example,
the residential microgrids mostly work in grid-connected mode and the ESS is
therefore acting as auxiliary equipment to adjust the power demand. For the
mobile microgrids, such as AESs, they mostly work in isolated mode and the
ESS is an important energy source, which needs to sustain the system operations
independently for a certain period. The difference in operation modes among various
types of microgrids also has great impacts on the ESS operation, which is further
declared in the following context.
132 S. Fang and Y. Wang

5.1.4.2 Load Demand Types

The load demand types are diversified in different microgrids. For example, electric
vehicle charging power is an important load demand type in residential microgrids,
and the propulsion load is an important load demand type in AES. Those different
load demand types certainly have different characteristics, such as the time-
dependent charging power profile and propulsion fluctuation, which gives different
requirements on the ESS.
From above, ESS gradually becomes an indispensable component in different
types of microgrids. To comprehensively illustrate the role of ESS, this chapter
is organized into three parts: at first, this chapter reviews current energy storage
technologies and compares their differences. Secondly, two typical application
scenarios are selected to show the roles of energy storage in microgrids, that is,
load leveling and the power quality issues. At last, the conclusions are summarized.

5.2 Energy Storage Technologies

In this chapter, the nomenclature of various energy storage technologies is shown in

Table 5.1.

Table 5.1 Nomenclature of BES: Battery energy storage

different Energy Storage
CAES: Compressed air energy storage
technologies
FBES: Flow battery energy storage
FESS: Flywheel energy storage
Li ion: Lithium ion
SMES: Superconducting magnetic energy storage
SCES: Supercapacitor energy storage
VRB: Vanadium redox battery
ZBB: Zinc–bromine flow battery
NaS: Sodium–sulfur
Ni–Cd: Nickel–cadmium
Ni–Mn: Nickel–manganese
PHS: Pumped hydro storage
PEM: Proton exchange membrane
MCFC: Molten carbonate fuel cell
SOFC: Solid oxide fuel cell
5 The Role of Energy Storage Systems in Microgrids Operation 133

5.2.1 Classification of Energy Storage Technologies

Nowadays, there already exist many energy storage technologies, which are suitable
for microgrid usage or not. In this section, several energy storage technologies
available now are reviewed for clarifying their applications. Generally, electricity
can be converted to many different forms for storage, which are shown as following
Fig. 5.5, and Table 5.2 shows the parameters of different ESSs collected from several
publications and manufacturers.

5.2.2 Single Energy Storage Technologies

5.2.2.1 Pumped Hydro Storage (PHS)

Currently, PHS is the most mature one among various energy storage technologies.
In 2018, the global PHS capacity is already more than 1675GW. The typical
structure of PHS is shown as the following Fig. 5.6.
From Fig. 5.6, a PHS usually consists of two reservoirs, that is, upper reservoir
and lower reservoir. The energy stored is proportional to the water volume in the
upper reservoir and the height of the waterfall. The water is pumped upwards to the
upper reservoir or flowed to the lower reservoir when low/high electricity price.
An illustrative PHS example is in operation by the State Grid of China, named as
the Tianhuangping PHS [29], which is the biggest one in Asia and the third biggest
around the world. The installation was commissioned in 1992 and completed in
2000. The rated power of this PHS is 1800 MW. It locates in Zhejiang Province
and is near the load central of eastern China, that is, Shanghai. Every summer,

Fig. 5.5 Classification of Energy Storage

 134

Table 5.2 Characteristics of different ESS

Technologies Investment (US$/kWh) Energy rating (MWh) Power rating (MW) Specific energy (kWh/kg) Specific power (kW/kg)
PHS 10–15 [10] 500–8000 [11] 10–1000 [11] – –
CAES 2–4 [10] 580, 2860 [12] 50–300 [12] 3.2–5.5 [13] –
VRB 600 [14] 1.2–60 [15] 0.2–10 [15] 25–35 [16] 166 [17]
ZBB 500 [18] 0.1–4 [19] 0.1–1 [19] 70–90 [13] 45 [20]
PSB 450 [14] 0.005–120 [16] 0.1–15 [16] – –
NaS 170–200 [10] 0.4–244.8 [21] 0.05–34 [21] 100 [14] 90–230 [22]
Lead-acid 50–100 [13] 0.001–40 [23] 0.05–10 [23] 30–50 [24] 180–200 [13]
Ni–cd 400–2400 [13] 6.75 [25] 45 [25] 30–80 [26] 100–150 [26]
Li ion 900–1300 [13] 0.001–50 [27] 0.01–50 [27] 80–200 [13] 200–2000 [13]
SMES 200–300 [16] 0.015 [13] 1–100 10–70 [28] 400–2000
FES 400–800 [13] 0.025–5 [16] 0.1–20[12] 5–100 [14] 10,000+ [25]
SCES 100–300 [28] 0.01 [25] 0.05–0.2 [16] 5–15 [12] 10,000+ [28]
FC 10,000+ [28] – 0.5–5 [12] 800–10,000 [16] 500+ [14]
S. Fang and Y. Wang
5 The Role of Energy Storage Systems in Microgrids Operation 135

Fig. 5.6 Illustration of pumped hydro storage (PHS)

Fig. 5.7 Illustration of compressed air energy storage (CAES)

Tianhuangping PHS can shift more than 500 MW peak load per day. However, the
construction of PHS highly depends on the geographic conditions, and may not
suitable for microgrid usage.

5.2.2.2 Compressed Air Energy Storage (CAES)

CAES stores energy in the form of compressed air and the air is usually stored in
an underground cavern or man-made storage tank. A typical structure of CAES is
shown as Fig. 5.7.
136 S. Fang and Y. Wang

When CAES outputs electricity, the compressed air is released from the air tank,
then the released air is heated and expanded to drive the high/low pressure turbines.
After that, the air is combusted with the natural gas to drive the generator. The
by-produced waste heat is used to heat the compressed air in the recuperator. The
overall lifetime of CAES is approximately 40 years, and the energy efficiency is
about 71% [13].
Currently, CAES systems are not widely used. As far as the authors’ knowledge,
until now there are only two practical projects of CAES globally so far, the one
is invested in Germany (290 MW) and the other one is constructed in the USA
(110 MW) [30]. However, the size of CAES is generally too large for microgrid
usage and the installment of CAES also needs a very large space. Since the
capacities of conventional microgrids are no more than100MW, CAES can be used
only in some special microgrids which have large energy storage requirement, such
as wind farm microgrids and tidal energy microgrids [28].

5.2.2.3 Battery Energy Storage (BES)

Due to high energy density and flexibility for installment, the battery is the most
commonly used ESS on the market. The energy is stored as the electrochemical
energy in each battery cell, and multiple battery cells are accumulated in series/in
parallel to make up the desired voltage and capacity. A typical battery structure is
shown in Fig. 5.8.
From Fig. 5.8, a battery pack consists of several battery cells and each battery
cell consists of two electrodes immersed in the electrolyte. All the battery packs are
sealed in a container and then integrated into the external grid.

Fig. 5.8 Illustration of battery energy storage packs

5 The Role of Energy Storage Systems in Microgrids Operation 137

In the last decade, the technologies of the battery have become much more
mature. The main representatives are the lead–acid battery, nickel–cadmium battery,
and lithium–ion battery. The lead–acid batteries, which have been researched
for more than 140 years, is the most mature battery technology now. Currently,
tremendous efforts have been devoted to the technologies like nickel–cadmium
and lithium–ion batteries to achieve more cost-effective options in power grid
applications.

5.2.2.4 Flow Battery Energy Storage (FBES)

FBES is an emerging technology only recently. The principle of FBES is based

on the reversible electrochemical reactions similar to conventional BES. The main
difference between the flow FBES and conventional BES is two different aqueous
electrolytic of the flow battery are in the liquid solutions and stored in separate
tanks. These aqueous solutions are pumped to the cell when operation [31]. There
are three types of commercially available flow batteries currently, that is, Vanadium
Redox Battery (VRB), Zinc Bromine Battery (ZBB), and Polysulphide Bromide
Battery (PSB). Their illustrations are presented in Fig. 5.9.
The main advantage of the FBES is the scalable energy capacity. The volume of
the stored electrolyte can be designed to meet the desired energy requirement [31],

Fig. 5.9 Illustration of flow battery energy storage

138 S. Fang and Y. Wang

Fig. 5.10 Illustration of

flywheel energy storage

and the power capacity of the FBES depends on the cell number and the size of the
electrodes. FBES also have significant advantages over conventional batteries, (1)
FBES can be fully discharged without any damage and (2) FBES have very low self-
discharge since the electrolytes are stored in separate sealed tanks. In this manner,
FBES has a quite long life and low maintenance cost than conventional batteries and
is also able to store energy over a long period with slight energy loss.

5.2.2.5 Flywheel Energy Storage (FES)

FES stores electricity in kinetic energy of rotating mass or rotor. The stored energy
is proportional to the rotor mass, location of the mass, and the rotor’s rotational
speed. When FES charges, it absorbs the energy from outside and accelerates the
rotational speed of the mass. On the other side, when FES discharges, the rotating
mass drives a generator to produce electricity, and the rotational speed slows down.
An illustration of FES is shown in Fig. 5.10.
Generally, FES can quickly respond to the power demand, and therefore widely
used in the fast-acting scenario, such as for the power quality improvement, fast load
demand shaving, or uninterrupted power supply (UPS) [32], frequency response
[33], power smoothing [34], and port crane locomotivations [35].
The advantage of FES is the intermediate characteristics between the batteries
and supercapacitors, that is, the FES has much higher power density than batteries
and much higher energy density than supercapacitors. In addition, FES also has
many advantages compared with prior other energy storage technologies, such as
less sensitivity to temperature, chemical hazardless, higher life cycle, reduced space
and weight, which is suitable for many applications. But the FES also has its
shortcomings, that is, the complex maintenance process for rotating mass and the
high self-discharging rate.
5 The Role of Energy Storage Systems in Microgrids Operation 139

Fig. 5.11 Illustration of superconducting magnetics energy storage

5.2.2.6 Superconducting Magnetics Energy Storage (SMES)

SMES uses a magnetic field created by the superconducting coil to store energy
[36]. The structure of SMES is shown in Fig. 5.11. However, in order to keep the
coil in superconductive states, the working temperature of SMES should below the
critical temperature by an external cooling system.
Since zero resistance, SMES storage devices have very high energy efficiency,
that is, usually more than 95%. The main energy loss is due to the power electronic
interfaces, which accounts for about 2–3% loss in both charging/discharging. The
advantage of SMES is the high lifecycles and the disadvantages are (1) high rate
of self-discharge, (2) very large installment space because of the external cooling
system, and (3) mechanical stability issues led by the cryogenic liquid.

5.2.2.7 Supercapacitor Energy Storage (SES)

Capacitors store energy in the electric field, and therefore, have very short respond-
ing time. Currently, the supercapacitor is the most commonly used capacitor, and
the corresponding SES is shown in the following Fig. 5.12. The main advantages
of SES include high power density, shorter charging/discharging time, quite long
lifecycles compared with other ESSs. The disadvantages are the low voltage of each
cell, and much higher investment cost per watt-hour, that is, more than 10 times
compared with a lithium battery. Other drawbacks of SES include relatively low
energy density compared with battery and very high self-discharge.
140 S. Fang and Y. Wang

Fig. 5.12 Illustration of super-capacitor energy storage

Fig. 5.13 Illustration of fuel

cell

Fuel Oxygen

Electrolyte

Waste

Anode Cathode

5.2.2.8 Fuel Cell

Different from the above energy storage technologies, a fuel cell is more like
a generator that directly transforms chemical energy into electrical energy. But
since there is no combustion process and no rotating equipment, a fuel cell has
similar operating characteristics with conventional ESS and superior efficiency than
generators. A typical structure of the fuel cell is shown in Fig. 5.13.
Since the reduced space, scalable capacity, fuel cells are viewed as promising
energy source for mobile microgrids. In this field, polymer exchange membrane
(PEM) fuel cell is the most mature technology and has been already used in
submarine applications and other propulsion usages, ranging from 30–40 kW [37].
The main advantage of the fuel cell in a maritime application is the great reductions
5 The Role of Energy Storage Systems in Microgrids Operation 141

on gas emissions [38], but the storage technologies for hydrogen fuel is the main
challenging features since it is hard to store the hydrogen fuel as conventional
hydrocarbon fuels, that is, liquid natural gas (LNG) or Marine Diesel Oil (MDO)
[39]. Currently, fuel cell technologies using conventional hydrocarbon fuels have
gained high concerns, such as Molten Carbonate Fuel Cell (MCFC) and Solid Oxide
Fuel Cell (SOFC).

5.2.3 Hybrid Energy Storage Technologies

In current microgrid usage, the battery is the most commonly used energy storage
technology to act as an energy buffer. However, the battery usually has high energy
density but the power density is low. Therefore, hybrid ESSs are used to combine
the advantages of different energy storage technology. Three combinations are
frequently used in microgrid operation.

5.2.3.1 Battery Supercapacitor

Among all the hybridization technologies, the battery-supercapacitor combination

has been studied quite extensively. In battery-supercapacitor hybrid ESS, the
batteries serve as an energy buffer, which generally has high energy density but
relatively low power density, and the life cycle of the battery is also limited, only
1500 to 4500 full cycles, meanwhile the supercapacitor has 100,000+ lifecycles. On
the other side, the supercapacitor energy storage device has low energy density but
very high power density. Therefore, the supercapacitor serves as a power buffer to
undertake the high power demand, and the hybridization with battery can combine
the advantages of high energy density and high power density. Currently, there are
two types of hybridization for the battery-supercapacitor hybrid energy storage, that
is, internal hybridization and external hybridization, which are shown as Fig. 5.14
a, b, respectively.

5.2.3.2 Battery–Fuel Cell

Since the reduced scale and convenient maintenance, battery–fuel cell hybrid ESS
has shown great advantages in the maritime applications. The main advantages
include the reduction of fuel consumption, reduced emissions, lower noise, lower
maintenance requirements, and minimal vibration. For example, fuel cells are first
integrated into a tourist vessel in 2008, named as Alster-Touristik, which has two
50 kW fuel cells in combination with lead–acid batteries [40]. The vessel can hold
100 passengers at maximum and the cruising speed is 8 knots. Compared with the
traditional diesel-powered vessel, the practical operation of this fuel–cell driven
vessel demonstrates that the integration of battery–fuel cell hybrid ESS benefits the
142 S. Fang and Y. Wang

Fig. 5.14 Illustration of battery-supercapacitor hybridization

Fig. 5.15 Typical topologies of fuel cell integrated shipboard microgrid

gas emission control. A typical topology of fuel cell integrated shipboard microgrid
is shown in Fig. 5.15.
Other applications include “Nemo H2”, a zero-emission ferry launched in 2009,
of which the propulsion system comprises of 60–70 kW PEM-based fuel cells
and 30–50 kW batteries [41]. In the future, since those great advantages, the
hybridization of battery and fuel cell will certainly have wider applications in
microgrid operation.
5 The Role of Energy Storage Systems in Microgrids Operation 143

5.2.3.3 Battery–Flywheel

From above, the battery has a relatively low power density and is suitable to
undertake the long timescale power demand, and the flywheel can quickly respond
to the power demand fluctuations. In this sense, the combination of battery and
flywheel can gain both high power density and high energy density. In [33],
the author uses hybrid battery–flywheel energy storage to mitigate the long-term
and short-term load fluctuations. The proposed method can be applied in wind
power integration, renewable uncertainty mitigation, as well as the shocking load
mitigation [34, 35]. Other applications also proposed different control methods to
balance the power demand on battery and flywheel for optimal microgrid control,
battery lifetime extension, or facilitating PV integration.

5.3 Energy Storage Applications in Microgrids

According to timescales, the applications of ESS in microgrids can be classified as

long-term timescale and short-term timescale. The long-term timescale applications
are shifting or shaving the load hourly or even longer, and the short-term timescale
applications are the quick load response to mitigate the fluctuations, which are
illustrated as follow:

5.3.1 Load Leveling

In microgrid operation, ESS acts as an energy/power buffer to keep the power

balance, and proper management of ESS can shave the peak load or level the load
demand, which is illustrated in Fig. 5.16.
From the above Fig. 5.16, when the load demand is high, the ESS discharges
power to share the demand of load, and when the load demand is low, the ESS can
absorb power. With the iteration of charging and discharging, the load demand is
leveled or shaved. The load leveling and peak shaving can bring lots of benefits to
microgrid operations. For example, the ESS can charge when the electricity price is
low and discharge when the electricity price is high, to gain extra benefits.
In land-based microgrids, there are many researches addressing the load leveling
and peak shaving. In Ref. [42], a practical energy storage project is analyzed to show
the advantages of an integrated energy storage system, which shows the integrated
battery ESS can gain extra economic benefits. In Ref. [43], the authors proposed
an effective energy storage sizing method and an optimal peak shaving strategy
to reduce the peak load of a residential area. In Ref. [44], a novel algorithm to
control the charging/discharging of a battery is presented, for the peak load shaving,
power curve smoothing, and voltage regulation of a distribution transformer. In
Ref. [45], the battery energy storage system is used to shave the peak load of a
144 S. Fang and Y. Wang

Fig. 5.16 Illustration of peak load shaving

residential building. In Ref. [46], battery planning is used to enhance the generation
hosting capacity of a distribution network. In Ref. [47], the vanadium redox flow
battery is used to coordinately control the frequency fluctuations and shave the
peak load. In [48], the centralized and distributed energy storage systems are both
used to facilitate the expansion of an active distribution network for maximizing the
overall economic benefits. In Ref. [49], a cost–benefit analysis of Na–S battery is
conducted for the peak load shaving. In [50], the battery energy storage anticipates
the electricity market to conduct a price-aware optimal dispatch. In [51], the authors
propose a model predictive control method to make battery tracking and shaving the
power demand.
In mobile microgrids, Ref. [52, 53] proposes energy storage management
methods to reduce the fuel consumption and gas emission of ships, respectively.
Ref. [54] uses battery energy management in a multi-objective framework, that is,
both the fuel consumption and gas emission. Ref. [55] uses hybrid energy storage to
mitigate the load fluctuations. Ref. [56] uses the hybrid energy storage for reducing
the fuel consumption and gas emission and extends the lifetime of the battery. Ref.
[57] proposes a battery planning method in a ferry and demonstrates the integration
of battery is necessary for the optimal operation of mobile microgrids. Ref. [58,
59] use a battery energy storage system to facilitate the integration of shipboard
photovoltaic modules.
In summary, the integration of energy storage into microgrids greatly facilitates
the optimal operation. The peak shaving and load leveling can make the generation
system of microgrids works in a more economic and environmental way. The energy
storage with high energy density usually serves in these scenarios, that is, the battery
or flow battery.
5 The Role of Energy Storage Systems in Microgrids Operation 145

5.3.2 Power Quality

There exist many load fluctuations in microgrid operation, such as renewable

energy uncertainties and pulsed loads, which might result in voltage and frequency
fluctuations, that is, voltage and frequency fluctuations and harmonic contamination,
and so on. Since the ability to keep power balance and the convenient deployment
requirement, energy storage systems are viewed as promising routes to mitigate
power quality issues. Various high power density energy storage system, such as
supercapacitor, flywheel, are used to handle the power quality issues.
In land-based microgrids, Ref. [60] comprehensively analyzes the control archi-
tecture for microgrids which addresses the power quality by energy storage.
Ref. [61] uses expanded energy storage planning to address the power quality
issues introduced by renewable power generation. Ref. [62] coordinately uses
DSTATCOM and energy storage to mitigate the power quality issues. Ref. [63] uses
the battery to reduce the fluctuations of photovoltaic integration. Ref. [64] proposes
a control strategy to integrate battery to support the frequency in microgrids.
In mobile microgrids, Ref. [65] proposes a hybrid battery–flywheel to address
the power quality issue in a shipboard microgrid, which can store energy up to
80 MJ. Ref. [66] uses FESS to mitigate the voltage sags for improving the ship’s
survivability. Ref. [67] proposed a multi-modular DC–DC configuration for the
HESS integration into the shipboard microgrid.

5.4 Conclusions

This chapter introduces the role of energy storage systems in microgrids operation.
The main types of microgrids, and the requirements on the ESS, and the operation
characteristics of ESS are comprehensively illustrated in this chapter. The main
conclusions of this chapter can be summarized as follow:
1. There exist many different types of microgrids and their operating characteristics
will greatly influence the ESS, that is, requirements on the power/energy
densities, operation conditions, and so on.
2. Different ESS technologies have their applications and the hybrid ESS technol-
ogy is promising for the future microgrid applications since it can combine the
high power density and high energy density.
3. Since its high efficiency, small scales, and less noise, the combination of fuel cell
and other ESS will provide a promising energy source configuration for future
microgrids.
146 S. Fang and Y. Wang

References

1. Vandoorn, T., Vasquez, J., et al. (2013). Microgrids: Hierarchical control and an overview of
the control and reserve management strategies. IEEE Industrial Electronics Magazine, 7(4),
42–55.
2. Hafez, O., & Bhattacharya, K. (2012). Optimal planning and design of a renewable energy
based supply system for microgrids. Renewable Energy, 45(9), 7–15.
3. Fang, S., Gou, B., et al. (2020). Towards future green maritime transportation: Overview of
seaport microgrids and all-electric ships. IEEE Transactions on Vehicular Technology, 69(1),
207–219.
4. Tan, X., Li, Q., et al. (2013). Advances and trends of energy storage technology in microgrid.
International Journal of Electrical Power & Energy Systems, 44(1), 179–191.
5. Fang, S., Zhao, T., et al. (2020). Coordinated Chance Constrained Programming of a Multi-
energy Microgrid System for Balancing Operational Efficiency and Quality-of-service. Modern
Power System and Clean Energy, In press.
6. Parise, G., Parise, L., et al. (2016). Wise-port business energy management: Port facilities,
electrical power distribution. IEEE Transactions on Industry Application, 52(1), 18–24.
7. Lamberti, T., Sorce, A., et al. (2015). Smart port: Exploiting renewable energy and storage
potential of moored boats. Genova, Italy: Oceans-2015.
8. Acciaro, M., Ghiara, H., et al. (2014). Energy management in seaports: A new role for port
authorities. Energy Policy, 71, 4–12.
9. Fang, S., & Xu, Y. (2019). Multi-objective coordinated scheduling of energy and flight for
hybrid electric unmanned aircrafts. IEEE Transactions on Industrial Electronics, 66(7), 5686–
5695.
10. Kaldellis, J. K., & Zafirakis, D. (2007). Optimum energy storage techniques for the improve-
ment of renewable energy sources-based electricity generation economic efficiency. Energy,
32, 2295–2305.
11. Kazempour, J., Moghaddam, P., et al. (2009). Electric energy storage systems in a market-based
economy: Comparison of emerging and traditional technologies. Renewable Energy, 34(12),
2630–2639.
12. Ridge Energy Storage & Grid Services L.P. The economic impact of CAES on wind in TX,
OK, and NM, http://www.ridgeenergystorage.com/re wind projects-compressed2005.pdf.
13. Hadjipaschalis, I., Poullikkas, A., et al. (2009). Overview of current and future energy storage
technologies for electric power applications. Renewable and Sustainable Energy Reviews,
13(6), 1513–1522.
14. Beck, F., & Ruëtschi, P. (2000). Rechargeable batteries with aqueous electrolytes.
Electrochimica-Acta, 45, 2467–2482.
15. Prudent Energy Corp. website, http://www.pdenergy.com/.
16. Ponce-de-León, C., Frías-Ferrer, A., et al. (2006). Redox flow cells for energy conversion.
Journal of Power Sources, 160, 716–732.
17. L. Barote, R. Weissbach, et al. Stand-alone wind system with vanadium redox battery
energy storage. In 11th international conference on optimization of electrical and electronic
equipment, 407–412, 2008.
18. D. Ton, C. Hanley, et al. Solar energy grid integration systems-energy storage (SEGIS-ES)
http://prod.sandia.gov/techlib/accesscontrol. cgi/2008/084247.pdf.
19. G. Corey. An assessment of the state of the zinc–bromine battery development effort,
www.redflow.com.
20. Emerging electric technologies whitepaper. Parcon website, http://www.parcon.uci.edu/paper/
EmergingElectricalTechnologies/eeenergy.htm.
21. N. Kawakami, Y. Iijima, et al. Development and field experiences of stabilization system using
34 MW NaS batteries for a 51 MW wind farm IEEE International Symposium on Industrial
Electronics (ISIE), 2010.
5 The Role of Energy Storage Systems in Microgrids Operation 147

22. EA Technology Review of electrical energy storage technologies and systems and of
their potential for the UK [R], http://www.wearemichigan.com/JobsAndEnergy/documents/
file15185.pdf, 2004.
23. McDowall, J. (2006). Integrating energy storage with wind power in weak electricity grids.
Journal of Power Sources, 162, 959–964.
24. Baker, J. (2008). New technology and possible advances in energy storage. Energy Policy, 36,
4368–4373.
25. Alaska Energy Authority Energy Storage Review, http://www.aidea.org/aea/.
26. Harding Energy Inc. website, http://www.hardingenergy.com/.
27. Saft Batteries website, http://www.saftbatteries.com/.
28. Francisco, D., Andreas, S., et al. (2012). A review of energy storage technologies for wind
power applications. Renewable and Sustainable Energy Reviews, 16(4), 2154–2171.
29. Tianhuangping pumped storage power station., https://en.wikipedia.org/wiki/
Tianhuangping_Pumped_Storage_Power_Station.
30. Ibrahim, H., Ilinca, A., et al. (2008). Energy storage systems—Characteristics and compar-
isons. Renewable and Sustainable Energy Reviews, 12(5), 1221–1250.
31. Skyllas-Kazacos, M., Chakrabarti, M., et al. (2011). Progress in flow battery research and
development. Journal of the Electrochemical Society, 158(8), 55–79.
32. Park, J., Kalev, C., et al. (2008). Control of high-speed solid-rotor synchronous reluctance
motor/generator for flywheel-based uninterruptible power supplies. IEEE Transactions on
Industrial Electronics, 55(8), 3038–3046.
33. Cheng, M., Sami, S., et al. (2017). Benefits of using virtual energy storage system for power
system frequency response. Applied Energy, 194, 376–385.
34. Díaz-González, F., Sumper, A., et al. (2013). Energy management of flywheel-based energy
storage device for wind power smoothing. Applied Energy, 110, 207–219.
35. Spiryagin, M., Wolfs, P., et al. (2015). Application of flywheel energy storage for heavy haul
locomotives. Applied Energy, 157, 607–618.
36. Hsony, W., & Dodds, S. (1993). Applied superconductivity developments in Japan. Power
Engineering Journal, 7(4), 170–176.
37. German HDW Submarine. Accessed https://www.naval-technology.com/projects/type_212/ on
27 August 2018.
38. Fuel Cell Ship Alsterwasser. Accessed https://www.drewsmarine.com/en/references/
passenger-ferries/fcs-alsterwasser on 27 August 2018.
39. Symington, W., Belle, A., et al. (2016). Emerging technologies in marine electric propulsion.
Journal of Engineering for the Maritime Environment, 230(1), 187–198.
40. A. Anvari-Moghaddam, T. Dragicevic, et al. Optimal planning and operation management of
a ship electrical power system with energy storage system. In Proceedings of the IECON 42nd
annual conference of the IEEE industrial electronics society, Florence, Italy, 24–27 October
2016, pp. 2095–2099.
41. Fuel Cell Boat (Nemo H2). Accessed http://www.opr-advies.nl/fuelcellboat/efcbboot.html on
27 August 2018.
42. Koller, M., Borsche, T., et al. (2015). Review of grid applications with the Zurich 1 MW battery
energy storage system. Electric Power Systems Research, 120, 128–135.
43. Chua, K., Lim, Y., et al. (2016). Energy storage system for peak shaving. International Journal
of Energy Sector Management, 10(1), 3–18.
44. Reihani, E., Sepasi, S., et al. (2016). Energy management at the distribution grid using a battery
energy storage system (BESS). International Journal of Electrical Power & Energy Systems,
77, 337–344.
45. Leadbetter, J., & Swan, L. (2012). Battery storage system for residential electricity peak
demand shaving. Energy and Buildings, 55, 685–692.
46. Jayasekara, N., Masoum, M., et al. (2015). Optimal operation of distributed energy storage
systems to improve distribution network load and generation hosting capability. IEEE Trans-
actions on Sustainable Energy, 7(1), 250–261.
148 S. Fang and Y. Wang

47. Lucas, A., & Chondrogiannis, S. (2016). Smart grid energy storage controller for frequency
regulation and peak shaving, using a vanadium redox flow battery. International Journal of
Electrical Power & Energy Systems, 80, 26–36.
48. Shen, X., Shahidehpour, M., et al. (2016). Expansion planning of active distribution networks
with centralized and distributed energy storage systems. IEEE Transactions on Sustainable
Energy, 8(1), 126–134.
49. Liao, Q., Sun, B., et al. (2016). A techno-economic analysis on NaS battery energy storage
system supporting peak shaving. International Journal of Energy Research, 40(2), 241–247.
50. Awad, A., Fuller, J., et al. (2014). Impact of energy storage systems on electricity market
equilibrium. IEEE Transactions on Sustainable Energy, 5(3), 875–885.
51. Di-Giorgio, A., Liberati, F., et al. (2016). Model predictive control of energy storage systems
for power tracking and shaving in distribution grids. IEEE Transactions on Sustainable Energy,
8(2), 496–504.
52. Kanellos, F. D. (2014). Optimal power management with GHG emissions limitation in all-
electric ship power systems comprising energy storage systems. IEEE Transactions on Power
System, 29(1), 330–339.
53. Kanellos, F. D., Tsekouras, G. J., & Hatziargyriou, N. D. (2014). Optimal demand-side
management and power generation scheduling in an all-electric ship. IEEE Transactions on
Sustainable Energy, 5(4), 1166–1175.
54. Shang, C., Srinivasan, D., & Reindl, T. (2016). Economic and environmental generation and
voyage scheduling of all-electric ships. IEEE Transactions on Power Systems, 31(5), 4087–
4096.
55. Hou, J., Sun, J., et al. (2018). Mitigating power fluctuations in electric ship propulsion with
hybrid energy storage system: Design and analysis. IEEE Journal of Oceanic Engineering,
43(1), 93–107.
56. Fang, S., Xu, Y., Li, Z., et al. (2019). Two-step multi-objective management of hybrid energy
storage system in all-electric ship microgrids. IEEE Transactions on Vehicular Technology,
68(4), 3361–3372.
57. Boveri, A., Silvestro, F., Molinas, M., et al. (2018). Optimal sizing of energy storage systems
for shipboard applications. IEEE Transactions on Energy Conversion, 34(2), 801–811.
58. Lan, H., Wen, S., et al. (2015). Optimal sizing of hybrid PV/diesel/battery in ship power system.
Applied Energy, 158, 26–34.
59. Wen, S., Lan, H., et al. (2016). Allocation of ESS by interval optimization method considering
impact of ship swinging on hybrid PV/diesel ship power system. Applied Energy, 175, 158–
167.
60. Guerrero, J., Loh, P., et al. (2012). Advanced control architectures for intelligent microgrids-
part II: Power quality, energy storage, and AC/DC microgrids. IEEE Transactions on Industrial
Electronics, 60(4), 1263–1270.
61. Fu, Q., Montoya, L., et al. (2012). Microgrid generation capacity design with renewables and
energy storage addressing power quality and surety. IEEE Transactions on Smart Grid, 3(4),
2019–2027.
62. Mahela, O., & Shaik, A. (2016). Power quality improvement in distribution network using
DSTATCOM with battery energy storage system. International Journal of Electrical Power &
Energy Systems, 83, 229–240.
63. Li, X., Hui, D., et al. (2013). Battery energy storage station (BESS)-based smoothing control of
photovoltaic (PV) and wind power generation fluctuations. IEEE Transactions on Sustainable
Energy, 4(2), 464–473.
64. Serban, I., & Marinescu, C. (2013). Control strategy of three-phase battery energy storage
systems for frequency support in microgrids and with uninterrupted supply of local loads.
IEEE Transactions on Power Electronics, 29(9), 5010–5020.
65. C. Xie, C. Zhang. Research on the ship electric propulsion system network power quality with
flywheel energy storage. In Proceedings of the Power and Energy Engineering Conference
(APPEEC) Asia-Pacific, Chengdu, China, 28–31 March 2010; pp. 1–3.
5 The Role of Energy Storage Systems in Microgrids Operation 149

66. Samineni, S., Johnson, B., et al. (2006). Modeling and analysis of a flywheel energy storage
system for voltage sag correction. IEEE Transations Industry Application, 42(1), 42–52.
67. Mo, R., & Li, H. (2017). Hybrid energy storage system with active filter function for shipboard
MVDC system applications based on isolated modular multilevel DC/DC converter. IEEE
Journal of Emerging and Selected Topics in Power Electronics, 5(1), 79–87.
Chapter 6
Microgrids and Local Markets

Mohsen Khorasany and Reza Razzaghi

6.1 Introduction

Progressive integration of Distributed Energy Resources (DERs) aligned with recent

advances in smart technologies are accelerating the pace of the global energy
transition to a decentralized, decarbonized, and digitalized network. These trends
in the energy sector are challenging well-established practices for supplying and
selling electricity by introducing new models for DER integration such as virtual
power plants, smart grids, and microgrids. The efficient integration of DERs requires
appropriate operational and control layers, which coordinate these resources. This
coordination provides substantial benefits for the stability of power networks and
increases the value of the local management of these resources [1]. Through
participating in new markets, DER owners can benefit from this coordination
by maximizing the return on their investment. From the networks perspective,
coordinated and controlled use of DERs, in combination with load management
technologies and services such as demand-side response can provide significant
benefits for power systems stability [2]. The flexibility of DERs in generating
and absorbing energy is a key value, which prevents or relieves localized network
performance issues such as voltage fluctuation, where the network is at the risk of
breaching its performance limits.
Nonetheless, the complexity of the energy system avoids DER owners to access
values from their assets and investments. Hence, there is a need to develop new
systems such as active distribution networks that are capable of providing value
streams to customers to incentivize them to share their flexibilities [3]. In a network
with high penetration of DERs, new market models can emerge that favor a local

M. Khorasany () · R. Razzaghi

Department of Electrical and Computer Systems Engineering, Monash University, Clayton, VIC,
Australia
e-mail: Mohsen.khorasany@monash.edu

© Springer Nature Switzerland AG 2021 151

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_6
152 M. Khorasany and R. Razzaghi

Distribution Transmission

On-site Storage
DNSP DSO TNSP

Prosumers Consumers Microgrid

Operator Market
MICROGRID Operator

Retailer Generation

On-site Generation

Local Market Retail Market Wholesale Market

Fig. 6.1 Interaction of local market with existing systems

usage of the generated electricity [4]. Different services can be provided through
local markets, such as aggregation of local resources, Peer-to-Peer (P2P) trading,
and demand-side flexibility, which can help to overcome system balancing issues at
the local level. The provided flexibility by local market participants can be utilized
by network operators to manage capacity and voltage constraints in networks.
Through implementing local markets, the expensive grid reinforcement costs can
be avoided by employing the flexibility of local resources. Also, it can add value to
the whole energy system by enabling more intelligent management of local energy
resources.
In this context, microgrids can be used as an architecture capable of making
a local market to utilize resources at the local level and to maximize the local
consumption of electricity generated in a distributed manner [5]. Microgrids provide
the opportunity for sharing and monetizing values from the increasing volume of
DERs and the emergence of prosumers (proactive consumers) in the electricity
system.
Through a monitoring and control platform, the microgrid operator can coor-
dinate the supply and demand of customers connected to the microgrid. Through
coordination, a microgrid can maximize the value of the connected DERs for
microgrid participants, the network, and the broader market. Microgrids, as a
decentralized business model, help customers to participate in the market and access
values from services they provide to the grid. A microgrid as a benchmark for
local markets and its interaction with different stakeholders is shown in Fig. 6.1.
Local market in microgrids fulfills customers’ demand from local energy resources
and reduces the need for expensive and inefficient energy transportation with
substantial losses [6]. It increases social welfare in the community and encourages
reinvestments in an additional renewable generation [7].
This chapter is focused on local markets for microgrid provision. An overview
of local markets’ definition, their potential benefits, and objectives are provided.
6 Microgrids and Local Markets 153

A summary of key enabler elements for local market implementation is given.

Different trading approaches for local markets are presented with a detailed
literature review on each model. Also, an overview of different market settlement
approaches for local markets is given. Case examples of local market business
models and market settlements in local markets are provided to help the reader with
outlining the attributes of different market models.

6.2 Local Markets

6.2.1 Definition

Different references have used diverse definitions for local markets. The authors
of [8] define the local market as a marketplace, which allows prosumers and
consumers to trade electricity directly within their community at variable prices.
In [9], the local market is defined as a group of interconnected loads and DERs
that provide a market platform for locally generated energy to their participants.
According to the definition in [10], a local community including different types
of prosumers, consumers, and producers, as well as storage facilities constitute a
local market, where community members are engaged in an array of commercial
activities that serve to create a better and more sustainable energy experience for
all involved parties. In [11], the local market is defined as a highly flexible market
platform to coordinate self-interested energy agents. A local market is rooted in
a residential area or similar and is based on a micro-market [12] concept that
includes prosumers and consumers, as well as storage facilities within such a
community [13]. Munné-Collado et al. [14] define a local market as a trading arena
located within a local energy community, operated in a public grid to provide two
different services: energy and flexibility. A local market is a platform on which
individual consumers and prosumers trade energy supporting regional scopes such
as a neighborhood environment [15]. The authors of [16] state that local markets
can promote renewable energy development more efficiently and define the local
market as a marketplace, where local cogeneration plants, wind power plants, and
consumers are incentivized to trade internally and to cope locally with the wind
power fluctuations.
All of these definitions have two common aspects: customer engagement and
the locality of the market. Customer engagement means that customers in the local
market are active players who participate in market activities in different ways. The
locality of activities underlines the location of customers as a key point in forming
local markets. Local markets can counteract current issues in energy systems and
solve grid imbalances on a local level with a market-based approach [17]. The local
market can be interpreted as an approach for energy management based on market
rules, where price signals can be developed to manage players’ behavior out with
a market or contractual obligation. Transactive energy management is an example
154 M. Khorasany and R. Razzaghi

of market-based energy management, which uses the value as a key operational

parameter to balance demand and supply in the network [18]. Transactive energy
provides economic signals aligned with operational goals to encourage demand-side
activities based on economic incentives [19].

6.2.2 Benefits

Developing local markets can provide a range of different benefits for different
stakeholders. Customers in local markets are able to consume self-generated
electricity and trade their energy surplus within their neighborhood area. Consuming
locally generated electricity reduces transport losses and the risk of backfeeding
at medium voltage or low voltage transformers [20]. Local markets strengthen
customers’ position by enabling them to play a more active role in the market instead
of being submissive ratepayers. Through aggregating and coordinating small-scale
resources, local markets can provide different services, such as the sale of electricity
into wholesale and retail markets, and the provision of voltage and frequency control
services to the local network. Improvement in system efficiency due to a reduction
in losses and improved network agility due to reduced dependency on the main
grid are other benefits of local markets. Furthermore, local markets support the
development of a smart grid [21]. Energy consumers may have different preferences
on the type of energy source based on factors like environmental preferences, such
as being carbon-free, pollution-free, exclusively renewable, and locally generated
[22]. Consumers’ preferences can be incorporated in local markets by providing
those customers with an opportunity to choose the source of their energy. Also,
local markets support the local economy, which provides new opportunities for local
industries and regional businesses. Sharing energy in a community of prosumers and
consumers increases social cohesion and improves the sense of community [23].

6.2.3 Objectives

Local markets can be established to achieve different objectives. Designing an

appropriate model for local markets is a challenging task, as it needs to incorporate
several inconsistent objectives [24]. Several principles need to be considered in the
market design, including (1) maximizing market surplus for involved parties, (2)
incentivizing market players to follow their commitments, (3) clarifying the market
price details for all players, and (4) providing an opportunity for all suppliers to
recover their costs. According to these principles and the specified list of objectives
in [14], the following objectives can be considered for a local market:
• Maximizing social welfare of market players.
6 Microgrids and Local Markets 155

• Minimizing energy costs by providing access to cheaper locally generated

energy.
• Increasing utilization of renewable energy resources by creating an attractive and
competitive marketplace that incentives to buy energy from local and renewable
resource.
• Supporting power systems by providing services in energy and flexibility
markets.

6.2.4 Services

Different services can be provided in local markets, which can be categorized as

energy and flexibility services exchange [14].
Energy services aim to manage load and generation resources to minimize total
energy costs. Local markets can provide energy related services by facilitating local
energy trading and managing local resources for participation in external energy
markets. On the other side, flexibility services aim to adjust power in a given
moment for a given duration in a specific location within a network. This service
can be provided in local markets by modifying the generation and consumption
patterns of local resources in reaction to market signals.
In the literature, there are several examples of local market design for both
energy and flexibility services. In [25], the local market is defined as a flexibility
market, which employs regional flexibility to resolve grid violations issues. Authors
of [26] presents an aggregator-based flexibility market for the participation of small
scale DERs in flexibility services trading. A framework for flexibility trading by
prosumers is presented in [27], where a local market operator plans the flexibility
to avoid congestion problems in the distribution grids. Authors of [28] propose a
market design, which allows aggregators and prosumers to provide flexibility in
response to distribution system operator requests. Different market mechanisms
for coordinating the retailer, transmission system operator, and distribution system
operator as buyers of flexibility are studied in [29]. The proposed local flexibility
market in [30] is designed as a market-based management mechanism for aggrega-
tors, in which multiple participants compete for selling or buying flexibility.
Market design for local energy trading is addressed in several studies. A
blockchain-based platform for the microgrid energy market is presented in [31],
where a classification of required components for an efficient market design is
presented. In [32], the energy trading of microgrids in a transactive system is
modeled as an auction-based electricity market, which manages the local demand
and supply. Authors of [33] proposes a double-sided auction for energy trading
in an islanded microgrid, where a distributed algorithm is implemented by an
aggregator to clear the market. The presented framework in [34] is an energy
exchange model for several microgrids, which aims to lower the exchanged energy
between the microgrids and the utility. A contribution-based mechanism is used
in [35] for energy trading among microgrids, in which a noncooperative energy
156 M. Khorasany and R. Razzaghi

competition game among the consumers is modeled that allows them to maximize
their satisfaction. In [36], the proposed trading framework enables microgrids to
trade energy in an independent market to maximize their average revenue, in which
a trading scheme is developed employing learning automation and noncooperative
repeated game. In [37], the energy trading is formulated using the Stackelberg game
model and a reward concept is proposed to promote energy trading for consumers.
Energy trading and scheduling of several interconnected microgrids modeled by
Nash bargaining theory are presented in [38]. In [39], a game-theoretic approach
is employed to design pricing mechanisms for energy trading in a prosumer-based
community microgrid. Dynamic price-based demand response for energy sharing
model among prosumers in a microgrid is presented in [40], where through a
distributed approach, prosumers can co-decide on internal prices.

6.2.5 Value Streams for Microgrids

Developing local markets in microgrids unlocks several value streams for micro-
grids such as cost saving on electricity bills, increasing energy efficiency, and
reducing grid dependency. Through the establishment of a local market, microgrids
can coordinate local resources for providing different services. For instance, micro-
grids can participate in Frequency Control Ancillary Services (FCAS) to provide
services requested by network operators. Peak demand is a significant challenge in
power networks, increasing infrastructure costs, and making balancing of supply
and demand difficult. Peak demand charges can be reduced in microgrids through
local markets. Sale of electricity into wholesale and retail markets, sale of voltage
control services to the local network, and avoiding network augmentation are other
value steams available to microgrids. Table 6.1 summarizes the types of value
streams and the ways in which economic and commercial value are delivered.

6.3 Key Elements in Local Markets Framework

In setting the framework for a local market, several key elements should be
considered. Key elements in the local markets framework can be grouped into five
layers as discussed in [41].
– Component layer: Active participation of end users is essential in local market
development. Key enablers for active participation are smart meters, energy
management systems, and communication devices. Smart meters provide data
about demand and generation at different times. Energy management system
enables market participants to determine the amount of energy for trade at
different times of the day based on the energy requirements and preferences,
weather conditions, and market prices. A communication device assists market
players with connecting to the market platform through the communication layer.
Table 6.1 Local Markets value streams for Microgrids
Type of value stream Market impact Source of economic value Source of commercial value
Demand side response Control demand side to maintain NSP: Cost saving from avoided Reduces energy costs for
system reliability network investment and associated customers. Potential increase in
maintenance costs NSP’s profit by deferring capital
investment
6 Microgrids and Local Markets

Peak demand management Reduce peak demand, peak power NSP: Cost saving from avoided Reduced network charges for
flow in the network network investment and associated customers by lowering peak
maintenance costs demand, and importing energy
during peak periods
Frequency control services to the Lowering ancillary service costs by Reduced costs and increased Increased profit for prosumers who
power system providing distributed provision of revenue for generators by provide frequency control which is
frequency control services providing frequency control more than offsets the lost value
from reduced power production
from solar power or battery
Trading energy in wholesale It can change the wholesale price, Increased market efficiency Customers benefit from pooling
market and with retailer depending on scale, time of day their energy for these markets
Providing voltage control services Substituting centralized voltage Deferred investment in centralized Customers benefit from providing
to the local network control with distributed control voltage control equipment network ancillary services for
defers the investment for reactors Improved quality of service in voltage control
and capacitor banks weak networks
NSP: network service provider
157
158 M. Khorasany and R. Razzaghi

– Information layer: This layer includes a secure information system that enables
all market participants to exchange required information during the negotiation
and energy trading. In implementing local markets, a secure information system
is an essential element, which enables market players to be integrated with the
market mechanism.
– Communication layer: This layer embraces Information and Communication
Technology (ICT) devices that are essential for the communication of all market
participants.
– Function layer: The key elements of this layer are market settlement and pricing
mechanism, trading approach, and market participants trading strategies. The
market settlement indicates the allocation of energy, pricing mechanism, and
payment rule. Trading approach guides the market operation by specifying the
way market participants interact in the local market. The strategy of participants
in reacting to market signals is another element of the function layer.
– Business layer: The role of stakeholders and their interaction are defined in the
business layer. The main stakeholders involved in the local market are consumers,
producers, prosumers, local market operators, distribution system operators, and
retailers. In the case of microgrid local market, the role of the market operator can
be fulfilled by a microgrid operator. In order to leverage economic benefits from
local markets, appropriate business models need to be designed [42]. A business
model describes benefits that an enterprise will deliver to customers, how it will
do so, and how it will capture a portion of the value it delivers [43]. Local markets
change the traditional unidirectional energy value chain to a multidirectional by
allowing the participation of new stakeholders, and hence, the business models
for local markets are known to be disruptive [44]. Different business models for
local markets can be considered, and for each model, a unique market model
needs to be designed.

6.4 Local Market Models

The market model for microgrids depends on the trading approaches, and the way
market participants interact during market settlement. Different trading floors can
be formed in microgrids, in which market participants can trade at the microgrid
level, trade with other microgrids at a higher level, or trade with upstream markets.
The formation of different trading floors enables market participants to participate
in virtual bidding for trading at different levels. During the bidding process, market
players interact with each other and can negotiate to agree on the price and amount
of energy to be traded. The virtual bidding is enabled by online services based on the
communication and information layers, which provide the required infrastructure
for communication and information exchange between market players.
The trading approaches in local markets can be classified based on the interaction
of market participants as pool-based, P2P, and hybrid trading, which are explained
in the rest of this section. The approaches implemented for market settlement can
6 Microgrids and Local Markets 159

Trading approach Pool-based P2P Hybrid

Local Market Model for Microgrids

Market settlement
approach Centralized Distributed Decentralized

Fig. 6.2 Formation of different local market models based on the trading and market settlement
approaches

also change the local market model, as the interaction of market players in the
market depends on the market settlement mechanism. The combination of different
trading approaches and market settlement mechanisms yields different models for
local markets, as illustrated in Fig. 6.2.

6.5 Trading Approaches in Local Markets

6.5.1 Pool-Based Trading

In pool-based trading (Fig. 6.3a), which is also known as indirect customer-to-

customer trading [45], there is a community manager who coordinates trading
among sellers and buyers and clears the market. Pool-based trading allows market
players to pool their resources to achieve their objectives such as reducing their
costs, increasing their revenues, and utilizing their assets more efficiently [46].
The community manager can be a nonprofit operator, responsible for providing
the platform for market participants and facilitating trading activities [47]. Also,
it can be a third party that aggregates and orchestrates players in the local market
and interacts with external markets to maximize its own profit as well as market
players’ profits [28]. Microgrids can be used as a benchmark for pool-based trading.
Each microgrid can form a local community, where community members can
collaborate with each other to share their investments and reduce their costs. For
instance, [46] proposes a framework for a community microgrid to share revenues
and costs among members, where a marginal pricing scheme is implemented as a
pricing mechanism in the local market. A community market to support residents
in a community microgrid is presented in [48], which allows agents to effectively
dispatch their resources by estimating the operation of the market.
160 M. Khorasany and R. Razzaghi

Fig. 6.3 Different trading approaches in local markets

6.5.2 P2P Trading

The other trading approach is P2P trading illustrated in Fig. 6.3b, which is also
known as “Energy AirBnB” [49]. P2P trading allows market players to directly
interact with each other without the intervention of an intermediary entity. It
provides a democratized environment, in which players can freely communicate
with each other. Microgrids provide infrastructures and technologies required for
monitoring, communication, and control that are important enablers for P2P markets
[50]. P2P trading among prosumers and consumers in a microgrid increases cost
saving in the microgrid. DER owners can achieve effective cost saving compared
with directly trading with the utility grid under the feed-in-tariff scheme.
Microgrids as the benchmark for P2P trading have recently received increasing
momentum as an innovative solution for allowing prosumers to engage directly in
energy trading. For example, a P2P market for a community microgrid is proposed
in [39], where the game-theoretic approach was employed to model the interaction
of sellers and buyers. Key drivers for prosumers’ participation in P2P microgrids
are identified in [51], where motivating initial user participation and ensuring the
long-term sustainability of microgrid development are introduced as fundamental
goals of P2P microgrid development. Authors of [52] propose a motivational game-
theoretic approach for P2P trading in which different motivational models that a
prosumer needs to satisfy before being convinced to participate in energy trading are
considered. Different market paradigms for P2P trading in a community microgrid
is investigated in [46]. The main difference between community-based and full P2P
markets is in the pricing, where in a P2P structure each transaction can have a
singular price, whereas in a community-based market all transactions usually have
a uniform price. Nonetheless, there are several P2P market models, which consider
a uniform price for all transactions [53].
6 Microgrids and Local Markets 161

6.5.3 Hybrid Trading

Besides pool-based and P2P market models, there are other models, which allow
both pool-based trading and bilateral trading with neighborhood markets, named
as hybrid models (Fig. 6.3c). In the hybrid market structure, players can trade
within a community (e.g., a microgrid) in a pool-based market at the lower level,
and communities (e.g., microgrids) can trade with each other in a P2P based
market at the upper level. For instance, [54] presents a three-level energy trading
platform for P2P trading within a microgrid, trading between multi microgrids, and
trading between cells (multi microgrids). A feeder-based local market is proposed
in [55], where prosumers can trade energy locally and with neighborhood areas.
The hierarchical trading scheme in [56] enables energy trading among nanogrids,
community microgrids, and multi-microgrid systems. A hybrid trading scheme is
presented in [57], which enables market players to trade energy in transactive energy
markets at different levels, including community, trading with neighborhood areas,
and traditional trading with the upstream grid.

6.6 Market Settlement Approaches in Local Markets

In local markets, the market settlement and supply and demand balancing can
be carried out in three ways: centralized, distributed, and decentralized. In the
centralized method, a central operator gathers information and data from all market
players to settle the market. The central operator is the decision maker for all players
and sends control signals to them to manage their actions in the market. Hence,
prosumers and consumers cannot actively engage in the market. This approach is
easy to implement and does not need to install new infrastructure in market players’
premises. However, centralized methods can endanger the privacy of players, as
they need to reveal their private information with the central operator. Also, there
are concerns about the scalability of these approaches, when employing for a large-
scale market.
These issues can be resolved by using distributed and decentralized methods.
In these methods, all market players are individual decision makers who share
information and reach an agreement once they agree on the value of the shared
information. These approaches can be implemented as an iterative price negotiation
mechanism, where a signal such as voltage, estimated power mismatches, and
the market price will be shared among market players to coordinate them. By
applying distributed and decentralized methods, the computation can be distributed
among players, which reduces computation complexity. These approaches protect
market players’ privacy and decrease the number of communication links. The
main difference between distributed and decentralized methods is that decentralized
methods remove the need for a coordinator and there is no need for communicating
with a third party. Hence, decentralized methods are appropriate for market structure
162 M. Khorasany and R. Razzaghi

with P2P trading, in which all peers directly negotiate to reach agreement on
their activities in the market. In the following sections, three different algorithms
for market settlement in local markets are introduced, namely, auction-based,
distributed, and decentralized approach.

6.6.1 Auction-Based Approach

Auctions are widely used for market settlements in different markets. An auction
is a negotiation mechanism, where market players can express their interest in
energy trading by submitting their offers or bids. Depending on the market model,
different types of auctions can be implemented for local markets. One-sided auction
is appropriate for flexibility markets with one seller and several buyers. Two-
sided auctions can be applied for energy markets with several sellers and buyers
to enhance the social welfare of the market players through participation in a
competitive market [58]. Double auction has received significant attention in the
literature as a market settlement method in local markets and microgrids [59–
62]. In the auction-based approach, a market settlement has three steps, namely
Determination where the number of sellers and buyers who win the auction is
identified, Allocation where energy transfer from sellers to buyers is resolved. The
third step is Payment, in which the auction prices for buyers and sellers for traded
energy are settled [63]. Both uniform price and differential pricing can be applied
for the payment step. Figure 6.4 illustrates an auction-based settlement for a double
auction.
Determination: In the double auction, sellers and buyers submit their offers/bids
to the market operator. Let denote seller’s offer with Oi = (pi , λi ), where pi is the
amount of offered energy by seller i, and λi is the reservation price for this offer. The
buyer’s bid is denoted by Oj = (pj , λj ), where pj is demanded energy by buyer j, and
λj is the corresponding price for this bid. After collecting all offers and bids, sellers’

Sellers Buyers

Offers Bids
Auctioneer

Payment/Allocation Payment/Allocation

Fig. 6.4 Auction-based settlement

6 Microgrids and Local Markets 163

offers are arranged in ascending order and buyers’ bids are arranged in descending
order to generate aggregated demand and supply curves. The intersection of these
curves indicates players who can participate in the market, such that the first K
buyers with λj higher than the price in intersection point and the first L sellers with
λi lower than the price at the intersection point can participate in the market.
 Allocation:
 This step determines the allocated energy from sellers to buyers
pi∗ , pj∗ . Energy allocation from sellers to buyers is in a greedy manner, which
means that energy from the seller with the lowest reservation price should be
allocated to the buyer with the highest bid. Depending on market players energy
requirements, they can participate in the market in two ways: (i) Fractional
participation: the allocated energy to sellers/buyers can be a fraction of their
offers/bids, (ii) Non-fractional participation: Players can only win for their total
offers/bids energy or lose. Energy allocation can be considered as a knapsack
problem and can be done from sellers’ or buyers’ perspectives.
Payment: After allocating energy from sellers to buyers, the next step is to
indicate the price for transactions in the market (λ∗ ). Different auction mechanisms
use diverse rules for payment and these rules can be divided into two groups; (1)
Fixed price: in which the price for all players in the market is the same and (2)
Variable price: in which based on the allocation rule, different buyers should pay
different prices and sellers will be paid at different prices. Different scenarios for
payment rules are tabulated in Table 6.2.

Table 6.2 Different payment rules

Payment mechanism Description
Fixed price Uniform price Auction price for all buyers and sellers in the
same and is equal to lowest winning bid or
highest losing bid
VCG mechanism Each buyer pays the lowest equilibrium price
max (λL , λK + 1 ), and each seller receives the
highest equilibrium price min (λK , λL + 1 )
Average mechanism Auction price for all buyers and sellers in the
same and is equal to average of all bids and
offers of sellers and buyers
Vickery All buyers pay the price equal to the second
highest winning offer
Trade reduction mechanism The first L − 1 sellers sell energy with price λL
and the first K − 1 buyers receive the item and
pay λK to the auctioneer
Variable price Pay as bid Buyers pay their bid and sellers receive what
buyers pay
Generalized second price The highest bidder pays the price bid by the
second highest bidder, the second-highest pays
the price bid by the third-highest, and so on.
Sellers get what the buyers pay
164 M. Khorasany and R. Razzaghi

6.6.2 Optimization-Based Approach

6.6.2.1 Distributed Clearing

In the distributed market clearing, market players optimize their own objectives
and participate in the market to maximize their profits. Market players iteratively
communicate through two-way communication links with a coordinator [64] to
reach an agreement on their actions in the market. The coordinator is responsible
for clearing the market and determines the clearing price to balance demand and
supply. The coordinator needs only the demanded/supplied power by each player to
clear the market such that utility and cost function parameters would remain private.
To explain the distributed method for market settlement, we consider a local
market consists of NS sellers indexed by i ∈ NS  {1, . . . , NS } , and NB buyers
indexed by j ∈ NB  {1, . . . , NB }. We consider a general model for market players
equipped with different assets and capability to manage these assets in response to
market signals. Market players are defined by the range of flexibility that they can
provide in the local market, which allows us to model different DERs as market
players. The distributed market settlement is visualized in Fig. 6.5.
The market objective is to maximize the social welfare of all players in the
local market. Social welfare is the sum of all sellers and buyers surplus, which
contributes to more user comfort with lower utility company costs [65]. The total
welfare of players in the local market can be maximized through social welfare
maximization, considering that each individual player’s profit is maximized too
[66]. Social welfare can be formulated as the sum of the welfare of all players as
in (6.1), and welfare of sellers and buyers can be modeled with (6.2) and (6.3),
respectively:

NS 
NB
SW tot = SW i + SW j (6.1)
i=1 j =1

Fig. 6.5 Optimization-based settlement: distributed

6 Microgrids and Local Markets 165

SW i = λpi − Ci (pi ) (6.2a)

1
Ci (pi ) = αi pi2 + βi pi + γi (6.2b)
2

p i ≤ pi ≤ p i (6.2c)

 
SW j = Uj pj − λpj (6.3a)

*
  αj pj2
Uj yj = βj pj − 2 , 0 ≤ pj ≤ βj /αj (6.3b)
βj2 /2αj , pj ≥ βj /αj

pj ≤ pj ≤ p j (6.3c)

where, pi and pj are traded energy by seller i and buyer j, respectively and λ is the
market clearing price. We employ a quadratic cost function to approximate the cost
of providing energy by the seller to the market as in (6.2b), with α i , β i , and γ i as
predetermined positive constants, denoting the seller’s willingness to sell energy at
different prices. These parameters depend on the type of generation and are private
information of the seller, which should not be revealed during the market clearing
process. The satisfaction level of the buyers from consuming the power is modeled
by a quadratic utility function as in (6.3b), with α j and β j as positive constants
determining the willingness of the buyer to pay for different levels of energy [68].
Substituting (6.2a) and (6.3a) in (6.1), yields (6.4) which is a convex optimization
problem. Market clearing should manage demand and supply such that the total
demanded energy by buyers is equal to the total supplied energy by sellers at the end
of the market clearing as in (6.5), which is the constraint of the objective function
in (6.4).


NB
   NS
max Uj p j − Ci (pi ) (6.4)
pi ,pj
j =1 i=1

Subject to. (6.2c), (6.3c), and


NS 
NB
pi = pj . (6.5)
i=1 j =1
166 M. Khorasany and R. Razzaghi

The optimization problem is a convex problem and can be relaxed using

Lagrangian multipliers. Then, dual decomposition can be employed to develop a
distributed iterative approach to maximize (6.4) subject to constraints without any
need to have individual parameters of all market players. The updating rules for
primal variables (supply and demand), and dual variable (price), are developed using
the primal-dual gradient descent method [67] as presented in (6.6), (6.7), and (6.8),
respectively
   
pik+1 = argmin Ci pik − λk pik (6.6)
pi ≤pi ≤pi

  
pjk+1 = argmin λk pjk − Uj pjk (6.7)
pj ≤pj ≤pj

⎡ ⎛ ⎞ ⎤+
NS 
NB
λk+1 = ⎣λk − ρ ⎝ pik = pjk ⎠⎦ (6.8)
i=1 j =1

 
 k+1 
λ − λk  <  (6.9)

where λ represents the Lagrangian multiplier and is the same as the market clearing
price, k is iteration index, ρ denotes step size and the notation [.]+ denotes max (0,
.). In each iteration, sellers and buyers update their supply and demand using (6.6),
(6.7) respectively and send their updated supply/demand to the market operator.
Then, the market operator updates the market price using (6.8) and sends the updated
price to all market players. This algorithm repeats to meet the stopping criterion in
(6.9).

6.6.2.2 Decentralized Clearing

Another approach for market settlement in local markets is decentralized clearing,

where market players can directly negotiate without the interaction with any central
entity, as represented in Fig. 6.6. This method is different from the distributed
method as there is no need for a supervisory node. In this section, a decentralized
market clearing method using a primal-dual gradient algorithm is explained. This
algorithm is a resource allocation problem to dispatch generation and demand of
market players in the local market. In the decentralized clearing, each transaction
can have a different price depending on players’ agreements in each bilateral trade.
Hence, (6.2a) and (6.3a) can be rewritten as (6.10) and (6.11) respectively:
6 Microgrids and Local Markets 167

Fig. 6.6 Optimization-based settlement: decentralized


NB
 
SW i = λij pij − Ci (pi ) (6.10a)
j =1


NB
pi = pij (6.10b)
j =1

   NS
 
SW j = Uj pj − λij pij (6.11a)
i=1


NS
pj = pj i (6.11b)
i=1

where, pij is the traded energy between seller i and buyer j, which should be the
same as pji , and λij is the price for this transaction. The market objective is the same
as (6.1) to maximize the social welfare of all players subject to constraints. In each
bilateral trade, the offered energy by the seller should be equal to the demanded
energy by the buyer, at the end of the market settlement as in (6.12):

pij = pj i , ∀i ∈ NS , ∀j ∈ NB (6.12)

The optimization problem can be solved centrally or in a distributed manner.

Central optimization needs a central controller who is aware of all market players’
characteristics and endangers the privacy of the players. Distributed optimization
resolves this issue by decomposing the problem into several locally solvable
subproblems. However, it needs a coordinator to coordinate local optimizations.
168 M. Khorasany and R. Razzaghi

The decentralized optimization eliminates the need for a coordinator, as all data
and related computations are handled locally by agents. To design a decentralized
solution for the optimization problem, constraints in (6.2c), (6.3c), and (6.12c)
should be relaxed using Lagrangian multipliers as in:

  NB
   NS
L pij , pj i , λij , μi , μi , μj , μj = Uj p j − Ci (pi )
j =1 i=1


NS 
NB 
NS 
NS  
 
+ λij pij − pj i − μi (pi − pi ) − μi pi − pi
i=1 j =1 i=1 i=1

  
NB
  NB
− μj pj − pj − μj pi − pj (6.13)
j =1 j =1

where, λij , μi , μi , μj , μj are Lagrangian multipliers associated with constraints.

Updating rules for both primal and dual variables can be developed using the
proposed method in [69]. In this method, sellers update their prices for different
buyers and send these prices to them, and buyers need to respond to these prices by
updating their demanded energy from different sellers. Sellers and buyers use (6.14)
and (6.15) respectively to update their parameters in the negotiation process.
+  ,+
λk+1
ij = λkij − ρλk pij
k
− pjki (6.14a)

+  ,+
μk+1
i = μk
i + ρμ
k
p k
i − p i (6.14b)

  +
μk+1
i = μk
i + ρμ
k
p i − p k
i (6.14c)

+  ,+
k+1 k+1
pij = pij
k
+ ζijk p̃ij − pik (6.14d)

k+1
λk+1 k+1
ij − μi + μk+1
i − βi
p̃ij = (6.14e)
αi

k + vk
pij
ζijk = #   (6.14f)
NB k + vk
j =1 p ij

+  ,+
μk+1
j = μ k
j + ρμ
k
p k
j − p j (6.15a)
6 Microgrids and Local Markets 169

  +
μk+1
j = μkj + ρμk pj − pjk (6.15b)

+  ,+
pjk+1 k k k+1
i = pj i + ζj i p̃j i − pj
k
(6.15c)

βj − λk+1 k+1
ij − μj + μk+1
j
p̃jk+1
i = (6.15d)
αj

pjki + v k
ζijk = #   (6.15e)
NS k + vk
i=1 p ji

k+1 k+1
where, p̃ij , p̃ij are power set points of sellers and buyers in each transaction,
which are obtained by taking the first-order derivative of the relaxed problem in
(6.13) with respect to pij , and pji , and ζijk , ζijk are asymptotically proportional factors.
In each iteration, sellers update their prices for different transactions using (6.14a)
and their offered energy using (6.14b–f). Then, all buyers in parallel update their
demanded energy from different sellers using (6.15a–e). The stopping criteria for
the negotiation process are as
 
 k+1 
λij − λkij  <  (6.16a)

 
 k+1 
μi − μki  <  (6.16b)

 
 k+1 
μi − μki  <  (6.16c)

 
 k+1 
μj − μkj  <  (6.16d)

 
 k+1 
μj − μkj  < . (6.16e)

6.7 Case Example of Local Markets for Microgrids: The

Monash Microgrid

The Monash Microgrid is a case example of real-world implementation of local

markets for microgrids [70]. The microgrid system is a fully functioning local
electricity network and trading market with dynamic optimization of resources
170 M. Khorasany and R. Razzaghi

interacting with external energy markets. The Monash Microgrid is grounded on the
idea of orchestrating and coordinating DERs through a transactive energy market
(TEM), where each DER will act as an independent customer that will, where it
chooses to, offer and commit to providing their flexibility as a commercial service
to the TEM. The TEM will then be able to use this internal market functionality to
aggregate the microgrid’s available flexibility. The microgrid provides the complete
hardware and software requirements to implement a marketplace at the local level
to trade energy and flexibility in both internal and external markets. A distributed
and ICT-oriented real-time integration platform is implemented to perform the
monitoring and control of the microgrid and all connected components to support a
real market implementation. The Monash Microgrid includes a range of customers
and DER assets such as flexible buildings, a battery storage system, solar generation
units, and EV chargers that participate in the market independently to provide
the requested services. The local market enables microgrid’s participants to trade
electricity and access revenues from the market and ancillary services they provide
to the broader electricity network (such as demand response and frequency and
voltage control). The establishing of a competitive internal market provides the
Monash Microgrid with the opportunity to reduce pressure on the network during
high and peak demand, sell renewable generation in the wholesale market, provide
frequency and voltage control services to the grid through the ancillary services
market, and help the grid respond to emergencies.

6.8 Case Examples of Market Settlement in Local Markets

This section provides illustrative examples of different market settlement

approaches for a local market. A microgrid with four sellers and five buyers is
considered, where market players participate in a forward market with a duration of
one hour to trade energy locally. Market players’ parameters are given in Table 6.3.
First, a double auction with a uniform price is implemented to clear the market.
Sellers and buyers submit their offers/bids to participate in the auction. Each
seller/buyer participates in the market by offering/bidding its maximum genera-
tion/demand and uses (6.2a)/(6.3a) to calculate its marginal price to maximize its
welfare. After receiving all offers and bids, an auctioneer generates the aggregated
demand-supply curves and allocates energy from sellers to buyers at the clearing

Table 6.3 Market players i pi pi αi βi j pj pj αj βj

parameters
1 0 6 0.168 4.21 1 0 6 0.144 8.21
2 0 5 0.15 6.35 2 0 4 0.168 11.21
3 0 5 0.102 8.98 3 0 4 0.102 12.98
4 0 3 0.158 10.25 4 0 3 0.158 13.24
5 0 2 0.122 15.21
6 Microgrids and Local Markets 171

Fig. 6.7 Simulation results: auction-based settlement

price. The market clearing price for all players is λ∗ = 9.49 ¢/kWh. Market
settlement results are illustrated in Fig. 6.7. In this case study, the first three sellers
and first four buyers win the auction, as their offers/bids are lower/higher than
market clearing price. The total welfare of market players is 83.72 ¢, where among
sellers, the first seller has the highest welfare in the market due to having the lowest
marginal cost as λ1 =5.21 ¢/kWh.
The second implemented method for market clearing is a distributed method,
where market players use (6.6) and (6.7) to update their supply and demand and
the coordinator updates the price at each iteration to clear the market. This method
can be interpreted as an iterative auction process, where market players actively
negotiate on their actions in the market. The results for this approach are shown
in Fig. 6.8. The total welfare of all players is 83.77 ¢. This approach allows
market players to optimize their actions at any iteration, while in the auction-based
approach, players cannot update their actions once the market is settled. Although
auction-based and distributed methods are easy to implement, the need for a central
coordinator makes it challenging to develop these methods for the market with a
large number of players. Hence, decentralized market settlements can be employed
to clear the market without any interaction of a third party.
Similar to the distributed approach, the decentralized method allows market
players to actively negotiate their actions in the market. This method is implemented
for the presented market and results are illustrated in Fig. 6.9. Figure 6.9a shows
results from the seller 1 perspective, where allocated energy from seller 1 to all
buyers is indicated. As in this approach, sellers are price makers, they use the
same price for all buyers, and buyers demand different amounts of energy from
sellers based on their preferences. Market settlement from buyer 3 perspective is
presented in Fig. 6.9b. Due to the fact that buyers can negotiate with different
sellers simultaneously, the offered prices by sellers 1 to 3 are converged to the same
value, while seller 4 is not selling any energy to the market due to its high offered
172 M. Khorasany and R. Razzaghi

Fig. 6.8 Simulation results: distributed settlement

price. Results verify that the decentralized method can reach approximately the
same results as the distributed method. However, it needs a two-way communication
network between all market players. Also, a higher number of iterations is needed
for market settlement in a decentralized method. As a comparison, the required
number of iterations for players to reach an agreement in a distributed method is
9, while this value for a decentralized method is 293.

6.9 Conclusions

Deployment of local markets in microgrids facilitates the integration of DERs in the

power systems. Local markets enable DERs to provide services to networks and to
be rewarded for these services. Microgrids can be integrated with communication
networks, smart devices, and monitoring systems to provide the required platform
for local markets. Through establishing a local market, microgrids can play the role
of a flexible power supply or a flexible load by aggregating flexibilities of different
resources.
This chapter has reviewed local market attributes including, definition, benefits,
objective, and services which can be provided through local markets. Different
trading approaches for local markets are presented. Three different methods for mar-
ket settlement are discussed, namely auction-based, distributed, and decentralized
clearing. For each method, a detailed formulation is presented. Also, case examples
are provided to demonstrate market settlement methods for local markets.
Research in the field of local markets for microgrids is a new trend, emerged
from interest to employ more DERs and decentralization of energy markets. Future
6 Microgrids and Local Markets 173

Fig. 6.9 Simulation results: decentralized settlement, 1) Seller 1 perspective, 2) Buyer 3 perspec-
tive

studies should analyze regulatory regimes to determine how microgrid local markets
fit into current energy policies. Compared to traditional markets, local markets are
smaller in size. Thus, they would have a different dynamic, which might require
different regulations. An important aspect of market regulation is to allocate costs
such as network utilization costs to market players at a reasonable rate to motivate
them for active participation in the market.
Communication networks are the key enablers for local market implementation.
Without having a two-way communication network, the active participation of mar-
174 M. Khorasany and R. Razzaghi

ket players in the local market is not feasible. An appropriate network architecture
needs to be designed for microgrids to serve as a benchmark for local markets.
Information flow among microgrid components raises privacy protection concerns
due to cyberattack issues, which raises the need for countermeasure techniques to
keep private data secure with minimal loss and latency.

Acknowledgments Icons in Figs. 6.1 and 6.3 to 6.6 are made by Freepik, Eucalyp, Pixel Perfect,
and Smashicons from www.flaticon.com.

References

1. Pudjianto, D., Ramsay, C., & Strbac, G. (2007). Virtual power plant and system integration of
distributed energy resources. IET Renewable Power Generation, 1(1), 10–16.
2. Eftekharnejad, S., et al. (2013). Impact of increased penetration of photovoltaic generation on
power systems. IEEE Transactions on Power Systems, 28(2), 893–901.
3. Energy Networks Australia, AEMO, Interim Report: Required Capabilities and Recommended
Actions. Technical Report, 2019.
4. Morales, J. M., et al. (2013). Integrating renewables in electricity markets: Operational
problems (Vol. 205, pp. 1–13). Springer Science & Business Media.
5. Hirsch, A., Parag, Y., & Guerrero, J. (2018). Microgrids: A review of technologies, key drivers,
and outstanding issues. Renewable and Sustainable Energy Reviews, 90, 402–411.
6. Jimeno, J., et al. (2011). Architecture of a microgrid energy management system. European
Transactions on Electrical Power, 21(2), 1142–1158.
7. Koirala, B. P., et al. (2016). Local alternative for energy supply: Performance assessment of
integrated community energy systems. Energies, 9(12), 981.
8. Mengelkamp, E., et al. Trading on local energy markets: A comparison of market designs and
bidding strategies. in 14th International Conference on the European Energy Market (EEM).
IEEE, June 2017. Dresden, Germany.
9. Ton, D. T., & Smith, M. A. (2012). The US department of energy’s microgrid initiative. The
Electricity Journal, 25(8), 84–94.
10. Ilieva, I., et al. Design characteristics of a smart grid dominated local market. in CIRED
Workshop. IEEE, June 2016, Helsinki, Finland.
11. Lamparter, S., S. Becher, and J.-G. Fischer. An agent-based market platform for smart grids. in
Proceedings of the 9th international conference on autonomous agents and multiagent systems:
industry track. May 2010, Toronto, Canada.
12. Cox, W. and T. Considine, Energy, micromarkets, and microgrids. Grid-Interop 2011, Decem-
ber 2011, Phoenix, AZ, USA: p. 1-8.
13. Bremdal, B.A., P. Olivella, and J. Rajasekharan. EMPOWER: A network market approach for
local energy trade. in 2017 IEEE Manchester PowerTech. IEEE, June 2017, Manchester, United
Kingdom.
14. Munné-Collado, ´i., et al. (2019). Local and micro power markets. In Micro and local power
markets (pp. 37–96). Wiley.
15. Da Silva, P. G., Ilić, D., & Karnouskos, S. (2013). The impact of smart grid prosumer grouping
on forecasting accuracy and its benefits for local electricity market trading. IEEE Transactions
on Smart Grid, 5(1), 402–410.
16. Hvelplund, F. (2006). Renewable energy and the need for local energy markets. Energy, 31(13),
2293–2302.
17. Olivella-Rosell, P., et al. (2018). Optimization problem for meeting distribution system
operator requests in local flexibility markets with distributed energy resources. Applied Energy,
210, 881–895.
6 Microgrids and Local Markets 175

18. Melton, R. (2015). Pacific northwest smart grid demonstration project technology performance
report volume 1: Technology performance. Richland, WA: Pacific Northwest National Lab.
(PNNL).
19. Holmberg, D. G., et al. (2016). Transactive energy application landscape scenarios. Technical
Report.
20. Kok, J.K., C.J. Warmer, and I. Kamphuis. PowerMatcher: multiagent control in the electricity
infrastructure. in Proceedings of the fourth international joint conference on Autonomous
agents and multiagent systems. July 2005. Ultrecht, The Netherlands.
21. Rahimiyan, M., & Baringo, L. (2015). Strategic bidding for a virtual power plant in the day-
ahead and real-time markets: A price-taker robust optimization approach. IEEE Transactions
on Power Systems, 31(4), 2676–2687.
22. Lane, C. W., et al. (2013). Costing consumer preferences for a micro energy market. Center for
nation reconstruction and capacity development, Technical Report.
23. Jogunola, O., et al. (2017). State-of-the-art and prospects for peer-to-peer transaction-based
energy system. Energies, 10(12), 2106.
24. Ela, E., et al. (2017). Electricity markets and renewables: A survey of potential design changes
and their consequences. IEEE Power and Energy Magazine, 15(6), 70–82.
25. Pavlovic, M., et al., SGAM business layer for a local flexibility market. in CIRED Workshop.
June 2016. Helsinki, Finland.
26. Zhang, C., et al. A flex-market design for flexibility services through DERs. in IEEE PES ISGT
Europe. IEEE, October, 2013. Copenhagen, Denmark.
27. Torbaghan, S.S., et al. Local market framework for exploiting flexibility from the end users. in
13th International Conference on the European Energy Market (EEM). IEEE, June 2016, Porto,
Portugal.
28. Morstyn, T., Teytelboym, A., & McCulloch, M. D. (2018). Designing decentralized markets
for distribution system flexibility. IEEE Transactions on Power Systems, 34(3), 2128–2139.
29. Vicente-Pastor, A., et al. (2018). Evaluation of flexibility Markets for Retailer–DSO–TSO
coordination. IEEE Transactions on Power Systems, 34(3), 2003–2012.
30. Olivella-Rosell, P., et al. (2018). Local flexibility market design for aggregators providing
multiple flexibility services at distribution network level. Energies, 11(4), 822.
31. Mengelkamp, E., et al. (2018). Designing microgrid energy markets: A case study: The
Brooklyn microgrid. Applied Energy, 210, 870–880.
32. Nunna, H. K., & Srinivasan, D. (2017). Multiagent-based transactive energy framework for
distribution systems with smart microgrids. IEEE Transactions on Industrial Informatics,
13(5), 2241–2250.
33. Faqiry, M. N., & Das, S. (2016). Double-sided energy auction in microgrid: Equilibrium under
price anticipation. IEEE Access, 4, 3794–3805.
34. Li, P., Zan, P., & Xu, H. (2016). Integrated energy exchange scheduling for multi-microgrid
system with electric vehicles. IEEE Transactions on Smart Grid, 7(4), 1762–1774.
35. Park, S., et al. (2016). Contribution-based energy-trading mechanism in microgrids for future
smart grid: A game theoretic approach. IEEE Transactions on Industrial Electronics, 63(7),
4255–4265.
36. Wang, H., et al. (2016). Reinforcement learning in energy trading game among smart
microgrids. IEEE Transactions on Industrial Electronics, 63(8), 5109–5119.
37. Park, S., et al. (2017). Event-driven energy trading system in microgrids: Aperiodic market
model analysis with a game theoretic approach. IEEE Access, 5, 26291–26302.
38. Wang, H., & Huang, J. (2016). Incentivizing energy trading for interconnected microgrids.
IEEE Transactions on Smart Grid, 9(4), 2647–2657.
39. Paudel, A., et al. (2018). Peer-to-peer energy trading in a prosumer-based community
microgrid: A game-theoretic model. IEEE Transactions on Industrial Electronics, 66(8), 6087–
6097.
40. Liu, N., et al. (2017). Energy-sharing model with price-based demand response for microgrids
of peer-to-peer prosumers. IEEE Transactions on Power Systems, 32(5), 3569–3583.
176 M. Khorasany and R. Razzaghi

41. Kirpes, B., et al. (2019). Design of a microgrid local energy market on a blockchain-based
information system. Information Technology, 61(2-3), 87–99.
42. Mengelkamp, E., Schlund, D., & Weinhardt, C. (2019). Development and real-world applica-
tion of a taxonomy for business models in local energy markets. Applied Energy, 256, 113913.
43. Chesbrough, H., & Rosenbloom, R. S. (2002). The role of the business model in capturing
value from innovation: Evidence from Xerox Corporation’s technology spin-off companies.
Industrial and Corporate Change, 11(3), 529–555.
44. Richter, M. (2012). Utilities’ business models for renewable energy: A review. Renewable and
Sustainable Energy Reviews, 16(5), 2483–2493.
45. Chen, T., Pourbabak, H., & Su, W. (2019). Electricity market reform. In The energy internet
(pp. 97–121). Elsevier.
46. Cornélusse, B., et al. (2019). A community microgrid architecture with an internal local
market. Applied Energy, 242, 547–560.
47. Khorasany, M., Y. Mishra, and G. Ledwich. Distributed market clearing approach for local
energy trading in transactive market. in IEEE Power & Energy Society General Meeting
(PESGM). IEEE, August 2018. Portland, OR, USA.
48. Shamsi, P., et al. (2015). Economic dispatch for an agent-based community microgrid. IEEE
Transactions on Smart Grid, 7(5), 2317–2324.
49. Luo, J., W. Su, and A.Q. Huang. Bit-energy: An innovative bitcoin-style distributed transac-
tional model for a competitive electricity market. in IEEE Power & Energy Society General
Meeting (PESGM). IEEE, July 2017. Chicago, IL, USA.
50. Sousa, T., et al. (2019). Peer-to-peer and community-based markets: A comprehensive review.
Renewable and Sustainable Energy Reviews, 104, 367–378.
51. Kirchhoff, H., & Strunz, K. (2019). Key drivers for successful development of peer-to-peer
microgrids for swarm electrification. Applied Energy, 244, 46–62.
52. Tushar, W., et al. (2019). A motivational game-theoretic approach for peer-to-peer energy
trading in the smart grid. Applied Energy, 243, 10–20.
53. Tushar, W., et al. (2018). Peer-to-peer energy trading with sustainable user participation: A
game theoretic approach. IEEE Access, 6, 62932–62943.
54. Long, C., et al. (2017). Feasibility of peer-to-peer energy trading in low voltage electrical
distribution networks. Energy Procedia, 105, 2227–2232.
55. Khorasany, M., Mishra, Y., & Ledwich, G. (2019). Two-step market clearing for local energy
trading in feeder-based markets. The Journal of Engineering, 2019(18), 4775–4779.
56. Paudel, A. and G.H. Beng. A hierarchical peer-to-peer energy trading in community microgrid
distribution systems. in IEEE Power & Energy Society General Meeting (PESGM). IEEE,
August 2018. Portland, OR, USA.
57. Khorasany, M., Mishra, Y., & Ledwich, G. (2019). Hybrid trading scheme for peer-to-peer
energy trading in transactive energy markets. IET Generation, Transmission & Distribution,
14(2), 245–253.
58. Munné-Collado, ´i., P. Olivella-Rosell, and A. Sumper, Power Market Fundamentals. Micro
and Local Power Markets. Wiley, 2019: p. 1.35.
59. Bhattacharjee, S. and S.B. Raha. Double auction bidding based economics for a microgrid
model. in Innovations in Power and Advanced Computing Technologies (i-PACT). IEEE, April
2017. Vellore, India.
60. Tushar, W., et al. (2016). Energy storage sharing in smart grid: A modified auction-based
approach. IEEE Transactions on Smart Grid, 7(3), 1462–1475.
61. Khorasany, M., Y. Mishra, and G. Ledwich. Auction based energy trading in transactive energy
market with active participation of prosumers and consumers. in Australasian Universities
Power Engineering Conference (AUPEC). IEEE, November 2017. Melbourne, Australia.
62. El-Baz, W., Tzscheutschler, P., & Wagner, U. (2019). Evaluation of energy market platforms
potential in microgrids: Scenario analysis based on a double-sided auction. Frontiers in Energy
Research, 7, 41.
6 Microgrids and Local Markets 177

63. Khorasany, M., Mishra, Y., & Ledwich, G. (2019). Design of auction-based approach for
market clearing in peer-to-peer market platform. The Journal of Engineering, 2019(18), 4813–
4818.
64. Pourbabak, H., et al. (2017). A novel consensus-based distributed algorithm for economic
dispatch based on local estimation of power mismatch. IEEE Transactions on Smart Grid,
9(6), 5930–5942.
65. Zou, X. (2009). Double-sided auction mechanism design in electricity based on maximizing
social welfare. Energy Policy, 37(11), 4231–4239.
66. Stiglitz, J.E., Whither socialism?. MIT press. 1996. (1).
67. Boyd, S., et al. (2011). Distributed optimization and statistical learning via the alternating
direction method of multipliers. Foundations and Trends® in Machine learning, 3(1), 1–122.
68. Samadi, P., et al. Optimal Real-Time Pricing Algorithm Based on Utility Maximization for
Smart Grid. in 1st IEEE International Conference on Smart Grid Communications. IEEE,
October 2010. Gaithersburg, MD, USA.
69. Khorasany, M., Mishra, Y., & Ledwich, G. (2020). A decentralised bilateral energy trading
system for peer-to-peer electricity markets. IEEE Transactions on Industrial Electronics, 67(6),
5930–5942.
70. Khorasany, M., et al. (2020). Transactive energy market for energy Management in Microgrids:
The Monash microgrid case study. Energies, 13(8), 1–23.
Chapter 7
An Economic Demand Management
Strategy for Passive Consumers
Considering Demand-Side Management
Schemes and Microgrid Operation

Mohammad Esmaeil Honarmand, Vahid Hosseinnezhad, Barry Hayes,

Behnam Mohammadi-Ivatloo, and Pierluigi Siano

7.1 Introduction

In recent years, there are strong incentives to utilize electricity end users for reducing
greenhouse gases, competitive energy policies, and participate in consumption
management. The increased penetration of microgrids and the extended strategies
of demand-side management have created an excellent opportunity for consumers
to acquire these benefits in smart systems. Consequently, in this circumstance, a
passive consumer, in addition to purchasing energy from the electricity market
directly, can manage to supply its demand economically by choosing the right
strategy. To this end, it can utilize the new concept of microgrids and local
production to meet the need. Consumers can also participate in load management
programs independently or integrated with the microgrid concept.
Nowadays, with the development of smart infrastructures, microgrids can bridge
the gaps between electricity market prices and consumer behavior. A microgrid can
be considered as a single electrical load from a utility’s viewpoint, and from behind
the consumer meter, a microgrid can function as a distributed energy resource [1].
Due to seasonal peak loads, increased network reliability, and energy shortages, this

M. E. Honarmand
Gilan Electric Power Distribution Co., Rasht, Iran
V. Hosseinnezhad () · B. Hayes
School of Engineering, University College Cork (UCC), Cork, Ireland
e-mail: vahid.hosseinnezhad@ucc.ie
B. Mohammadi-Ivatloo
Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, Iran
P. Siano
Department of Management and Innovation Systems, University of Salerno, Fisciano, Italy

© Springer Nature Switzerland AG 2021 179

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_7
180 M. E. Honarmand et al.

controllable load can be an excellent platform for the implementation of demand-

side management schemes.
Demand response (DR) programs as a part of demand-side management are
powerful tools that facilitate the process of transforming conventional microgrids
into green systems by consumption management and efficient utilization of renew-
able sources. These programs involve the modification of customer’s demand for
energy by various means, such as financial incentives and behavioral change through
educational approaches. Accordingly, microgrid and customer reliability will be
improved by implementing the DR programs in microgrids. Besides, this benefit
can be achieved through the reduction of demand during critical times.
Recently, due to the propulsion of the smart grid paradigm, numerous efforts
have been made to integrate the schemes of DR and microgrid operation. In [2],
the impact of the customer participation level in emergency DR for microgrid
operation is analyzed in the presence of different uncertainties. In addition, this
work examines the effect of different incentives on total costs of operation by
a model that is presented based on price elasticity and customer benefit. The
scheduling problem based on various incentive rewards and constraints related to
microgrid operation is presented in [3]. The authors of [4] propose an optimization
function of microgrid operation with the purpose of minimizing operational costs
and emissions by considering the DR programs. Due to the economic dispatch of a
renewable microgrid, the operational costs are investigated in [5] by considering the
participation of the consumers in DR schemes. To solve the economic dispatch to
minimize the operation cost of microgrid, in [6], the authors proposed the model of
different DR programs to prioritize running the plans in the presence of microgrid.
In [7], the optimal operation of a microgrid is assessed by the combination of various
DR programs. A mathematical model has been developed for the microgrid system
considering the impact of different DR schemes to minimize the objective function.
In order to maximize net income in [8], a cost–benefit analysis is presented to plan
the operation of a grid-connected microgrid in combination with DR programs. The
authors in [9] introduced an incentive payment oriented DR scheme for microgrid
operational planning. A stochastic optimization approach is proposed to consider the
presence of different types of customers. In [10], a comprehensive DR framework
in a microgrid environment is proposed to mitigate peak demand and energy saving.
Considering DR schemes for a microgrid retailer, an optimal strategy for energy
dispatch and pricing is presented in [11]. An economic optimization model is
introduced in [12] to manage microgrid and allocate the shiftable loads in the
residential sector. The authors in [13] proposed a dynamic optimization model based
on DR schemes to minimize the operation cost and maintain the supply–demand
balance in microgrids.
Despite the reviewed literature, the question that authors are interested in here is
quite different. This work explores which strategy would be best suited for a typical
passive load to manage its demand economically; purchasing electricity from the
grid and participating in load management plans, or transferring to a microgrid and
integrating with DR programs. To this end, developing an integrated procedure that
covers both microgrid establishment and DR strategy effects from the economic
7 An Economic Demand Management Strategy for Passive Consumers. . . 181

viewpoint, can be vital. The proposed methodology here is concentrated to find a

tradeoff point between the costs of microgrid and the DR.
The scope of this work is to propose a procedure that can address this economic
problem as much as possible. Accordingly, the intended method is aimed to consider
microgrid costs and DR scheme costs regardless of the market mechanisms to
manage the network load economically. On this base, first, the corresponding
cost model for the installation and operation of a microgrid is formulated. For
simplicity, in the formation of the microgrid, renewable energy-based units are not
considered. Then, the output is evaluated alongside DR cost with the viewpoint of
reaching the compromise point. Finally, the proper DR program related to the best
performance is selected as the final strategy. Two DR programs including price-
based DR (PDR) and incentive-based DR (IDR) are considered in the studies. The
introduced procedure is implemented on several real loads and is investigated under
different case studies. The detailed results are presented in the analysis of significant
contributions and benefits of the proposed method.
This chapter covers a summary of DR programs in Sect. 2. Classification of
microgrid applications is provided in Sect. 3. The cost model for installation and
operation of the microgrid, the cost function related to the run of the various
DR programs, and the proposed decision-making method are presented in Sect. 4.
Finally, to investigate and analyze the proposed algorithm, three case studies are
presented in Sect. 5.

7.2 Types of DR Programs

DR can be defined as changes in electricity usage by end use customers from

their usual consumption patterns in response to changes in prices. DR programs
may be classified either according to how the enrolled consumers respond or by
their type (motivation procedure and trigger criteria) regarding the characterization
of their load. The U.S. Department of Energy (DOE) categorizes these programs
into PDR and IDR [14]. PDR offers collaboration in time-varying rates that reflect
the value and cost of electricity for different periods. However, in IDR consumers
voluntarily provide load reductions by responding to economic signals. Indeed, the
PDR includes the actual cost for the electricity, while the IDR provides customers
with peak shaving incentives [15]. This section includes detailed discussions on
some of the DR strategies for both categories. In the end, the overall impact of these
strategies on microgrid operation is investigated.

7.2.1 PDR Strategies

In different articles, various strategies have been studied for PDR. However, here,
only three schemes are considered. These include time of use (TOU), critical peak
182 M. E. Honarmand et al.

pricing (CPP), and real-time pricing (RTP) [16]. These methods are used for cost
modeling and economic evaluation in the proposed model.
• Time of Use (TOU) pricing. Electricity consumers that are charged with flat
prices are not aware of the varying cost of electricity. One way in which
consumers will be incentivized to change their consumption patterns is through
price signals delivered via TOU tariffs, in which the price of electricity varies
depending on factors such as electricity network constraints and the wholesale
price of electricity [17]. These tariffs are designed to more closely reflect the
investment and the production cost structure, so key issues such as the duration
of individual periods and related price levels are involved in the design of
TOU rates. Many countries divide a year into peak periods and valley periods
according to summer or non-summer months to charge differently. This pricing
is easy to be implemented and has a great effect on load shifting, although it
cannot be strictly called a dynamic pricing strategy for its high consistency [18].
Static TOU tariffs are the stepped rate structure, which varies into several periods
(usually less than five periods) during the day fixedly and regularly. The pricing
scheme of TOU tariff can be introduced as follows:
⎧
⎪
⎪ F1 , i ∈ (h1 , h2 )
⎪
⎨ F2 , i ∈ (h2 , h3 )
PT OU −i = .. (7.1)
⎪
⎪ .
⎪
⎩
Fk , i ∈ (hk−1 , hk )

In (7.1), a day is separated into k periods (h stands for hour) and a certain price
level is provided for each period known as Fi .
• Critical Peak Pricing (CPP). The long-term electricity supply costs associated
with using electricity during a specific period of the day are reflected by TOU
tariffs. In order to capture the short-term costs of considered critical periods
for the power system, CPP tariffs may be employed [19]. This strategy, also
called peak-load pricing, has both characteristics of TOU tariffs and emergency
load control, therefore, it can be a supplement to TOU pricing which has some
mandatory restrictions for electricity demand in critical peak periods. A CPP
period is announced ahead of time, typically day-ahead, and customers on the
CPP rate can reduce their bill by shifting or reducing their loads during these
peak times. In this way, the higher rates on emergency or critical peak periods
(e.g., unavailability of reserves, extreme outages, etc.) are charged by CPP tariffs,
while the prices in other times remain the same [20]. The power utility may sign
contracts with consumers to specify the maximum number of days per year that
may be considered critical and the number of periods for which the CPP rate is
applied otherwise; they will get some punishments. Equation (7.2) provides the
pricing scheme of CPP tariff:
7 An Economic Demand Management Strategy for Passive Consumers. . . 183

⎧
⎪
⎪ RN 1 , i ∈ (h1 , h2 )
⎪
⎪
⎪
⎪ ..
⎪
⎪ . If not Critical Periods
⎪
⎪
⎪
⎨ RN k , i ∈ (hk−1 , hk )
PCP P −i = (7.2)
⎪
⎪
⎪
⎪ RC1 , i ∈ (hC−1 , hC−2 )
⎪
⎪
⎪
⎪ .. Otherwise
⎪
⎪ .
⎪
⎩  
RCm , i ∈ hC−(m−1) , hC−m

where RNk and RCm are the separated price levels for normal and critical tariffs,
respectively. A day is split into k periods and also m part is introduced for critical
times.
• Real-time pricing (RTP). This strategy separates a day into several short time
slots similar to the TOU tariff. Generally, RTP is an electricity pricing strategy
that directly according to the real-time supply and demand situation, reflects the
marginal value of electricity [21]. Therefore, prices vary in real time (e.g., an hour
or a half-hour) depending on the current wholesale cost of electricity so it can be
said that the scheme is theoretically most reasonable. Furthermore, RTP and real-
time power load have a positive correlation for the general operating state. A typical
relationship of the pricing scheme is obtained as follows [22]:

PRT P −i = α × ALoad i (7.3)

In (7.3), α is the rate between load and price, and A_Loadi is the overall con-
sumed load of a certain end user. This correlation depends on different factors such
as real-time status of the operation, line losses, wholesale price, and so on. Besides,
electricity regulations and the policy of governments and other organizations, for
example, ISOs, limit the setting of electricity prices in practice.

7.2.2 IDR Strategies

In dynamic pricing, the load control scheme is defined without a third-party operator
that manages load-shedding. However, IDR programs are employed by the power
utility to control the load of the consumer directly or based on the response of the
incentive measures by consumers. Here, three schemes of IDR include direct load
control (DLC), emergency DR (EDR), and interruptible load program (ILP), are
discussed.
• Direct Load Control (DLC). Ordinarily, DLC programs involve a utility or system
operator that allows them to switch on and off specific customers’ appliances for
184 M. E. Honarmand et al.

a short time during peak periods and critical events. In return for participating,
consumers are usually rewarded by way of a financial incentive such as a
one–off sign-up payment, recurring annual payment, and ongoing electricity
bill discounts [23]. This type of program is mostly applied to residential or
small commercial customers. PDR schemes could be supplemented by DLC
to reinforce gains of these schemes. This service of contract-based reliability
enhancing can be planned to control loads in 10–14 hours.
• Interruptible load program (ILP). This program considers curtailment options
to a predefined level. In order to turn off specific loads by participants in these
services, customers receive a discount or bill credit in exchange to reduce load
during system contingencies. Besides, participants may face penalties in case
they fail to respond to a DR event [24]. These are offered for typical customer
size from 200 kW up to 3 MW so that customers on these tariffs must curtail
within 30–60 minutes when being notified by the utility. Also, the total amount
of period that a utility can call interruption often is not more than 200 hours per
year [25].
• Emergency DR (EDR) program. The incentive payments are considered to
consumers due to the reducing power consumption during reliability triggered
events. These programs can be also known as a combination of DLC and ILP
programs. In contrast to ILP, since there is no contractual obligation, this scheme
does not impose any penalties if consumers cannot participate [26]. However,
these programs have a narrow application, and they are called a very limited
number of times per year (less than 5).

7.2.3 DR Programs and Microgrid Operation

As mentioned earlier, DR programs are divided into two categories. PDR programs
depend on the behavior and response of electricity consumers to the suggested
prices. Therefore, from the viewpoint of participants, the utilization of microgrids
by the consumers is related to the associated cost–benefit. Due to the type of PDR
programs, the operation of microgrid will be relatively long term (at least monthly).
On this way, if the microgrid costs involved are higher than the cost of PDR
implementation, the participant’s preference will be to use the DR scheme into the
desired period. Assuming a variety of PDR schemes, the participation priority can
be a criterion to choose the economic microgrid operation.
Since the period and times of participation in IDR programs are predefined,
it is possible to combine them with PDR programs. Generally, if the microgrid
costs involved are lower than the cost of PDR implementation, then these costs
can equilibrate with IDR program cost. With each IDR scheme implemented, the
microgrid can be run for the customer’s electricity availability. Accordingly, when
the IDR contract is signed, the microgrid is engaged if needed. In this case, the
microgrid is run at the specific periods then turns off.
7 An Economic Demand Management Strategy for Passive Consumers. . . 185

7.3 Classification of Microgrid Applications

In the literature, conventional micro-generators, renewable energy resources, and

energy storage systems are often described as distributed energy resources (DERs)
that are on-site generation sources in the distribution system [27]. Generally, a
microgrid is well-known for the integration of DERs into the power grid, as well
as its ability to operate in an islanded mode during certain cases. Therefore, the
microgrid is described as a low-voltage distribution network of interconnected
DERs, controllable loads, and critical loads that can operate in either grid-connected
or islanded mode [28]. Besides, the microgrid may be used as a remote/off-grid case
that is not considered in this article.
In general, the microgrid may comprise several distributed generation systems,
renewable (such as wind power, photovoltaic, hydro, and fuel-cell devices) or
conventional generation (such as micro-turbines, diesel generators, and internal
combustion engines) and a cluster of loads [29]. In this way, the customers can
utilize from on-site generators, with the intent of adding additional resources over
time, such as energy storage or other renewable sources [30]. However, due to
the specific conditions of use of renewable technologies (availability of wind or
solar radiation in the time of load demand), in this study, it is assumed that the
conventional generation is used only as a microgrid.
Therefore, customers can deploy this solution to manage their electric load
consumption. Whereas DR is primarily focused on loads of consumer side, features
analysis of various loads in the presence of a microgrid can facilitate to operate
microgrid integrated with DR.
The consumption pattern of loads may be indicated as urban, semi-urban, and
rural or island. However, the practical application of microgrid for various loads
can be mostly classified into eight sectors: industrial, military, campus/institutional,
commercial, healthcare, residential, remote or rural, and others (such as data centers
and cell phone towers). These applications are explained from the viewpoint of DR
as follows.

7.3.1 Industrial Sector

Industrial facilities, which are increasingly being established in remote locations,

may not have continuous access to the main electricity grid. Therefore, this sector
may be dependent on fuel and tend to use microgrid. Suppose an industrial consumer
can manage electrical demand by utilizing distributed generation, energy storage,
and load shifting [31]. Furthermore, the motivation of microgrid operation is the
increased security and reliability needs in a grid-connected industrial site.
On the other hand, the industrial sector is suitable for developing DR programs;
however, adopting DR programs may be challenging for industrial firms. For
instance, temporarily interrupting one or more processes may result in significant
186 M. E. Honarmand et al.

load reductions [32]. On this base, certain industrial sectors are only suitable for load
management due to their technical restricts. By deploying microgrids in place, the
industrial customer may solve this problem to integrate microgrid and DR strategies.
Therefore, to provide flexible load solutions, a smart microgrid scheme can establish
continuity of the industrial performance by implementing DR strategies.

7.3.2 Military Sector

Military sector bases require reliable and resilient power to accommodate a variety
of missions. Microgrids are easy to communicate on a community level but have
more specific benefits when installed in military applications. Indeed, the military
sector can enhance the security of critical electrical loads against the threat of grid
outages by microgrids and this can be useful for DR schemes implementation.
In order to develop a better DR management system, smart technologies can be
used to communicate critical loads performance in real time. This can help make
decisions and participate in DR programs, using microgrid without military sector
interruption. Besides, cost-effective energy security is a driver to use the other
military microgrids such as renewable resources [33]. It should be noted if these
technologies are implemented in a secure procedure and well protected from cyber
threats, it can be an opportunity for the military sector [34].

7.3.3 Campus/Institutional Sector

So far, deploying on-site generation on a campus with multiple loads has been a
successful procedure. Typically, the operation of microgrids with capacities ranging
from 4 to over 40 MW has been common in this sector [35]. Furthermore, various
abilities and numerous advantages to the management of net load shape during grid
needs driven development of smart microgrids in the institutional sector. Generally,
campus systems may include university and government campuses, and corporate
parks. These are geographically large systems covering many buildings (residential,
commercial, and/or industrial) but within a single ownership boundary that does
not cross public rights of way [36]. Therefore, the potential for automated demand
response is one of the key benefits of this sector along with the use of microgrids.

7.3.4 Commercial Sector

Nowadays, microgrids have gained popularity to provide for economic requirements

and demand management of commercial building installations, thereby supplement-
ing the conventional grid. Practical applications of microgrids in smart commercial
7 An Economic Demand Management Strategy for Passive Consumers. . . 187

buildings are to increase renewable generation contribution and provide a high level
of reliability and resiliency in response to grid outages [37]. One of the most direct
ways a commercial microgrid can be used to cut costs is as a means to hedge power
prices so that the system controls can be programmed to optimize for price. The
usage peaks in commercial buildings typically align with an electric utility’s overall
demand peaks. This means that utilities are particularly incentivized to participate
with commercial customers to reduce or shift load through DR. Accordingly, a
microgrid could use utility power until prices rise and then switch to its own, lower
cost power by participating in DR Schemes [38].

7.3.5 Healthcare Sector

A healthcare facility or medical center requires reliable electricity, heating, and cool-
ing for running high-tech equipment and keeping patients healthy and comfortable.
The microgrids system provides significant economic and environmental benefits to
the most advanced healthcare sectors, ensuring the medical center’s sustainability
and reducing its carbon footprint [39]. These sectors must care for patients 24/7,
which creates greater demand for lights, heat, and cooling so that their consumption
is much more than a commercial building of the same size. Due to the special
energy requirements of this sector, huge opportunities can be offered in healthcare
facilities by new technologies and procedures. These may be constituted to adopt
new enablers and install advanced systems such as smart microgrids, which can
empower them to participate in DR programs.

7.3.6 Residential Sector

The main challenges in this sector are in making decisions for integrating individual
home residential customers into large microgrids, and the deployment of microgrid
technology at the level of individual homes. In the first case, it is possible to serve
anywhere from a few up to thousands of customers and to support the penetration
of local energy sources (electricity, heating, and cooling). In this situation, some
houses may have some renewable sources that can supply their demand as well
as that of their neighbors within the same community. In addition, this microgrid
may have centralized or several distributed energy storage [40]. However, a
decentralized building-integrated microgrid approach has the advantages of control
over energy resources by customers. Besides, by adding microgrid capabilities, any
changes performed behind the utility meter will likely not introduce significant
legal or regulatory complications beyond because individual homes are already
connected to the electrical distribution network [41]. Accordingly, using a variety
of DR strategies, an appropriate framework for the management of energy can be
developed for a residential microgrid [42].
188 M. E. Honarmand et al.

7.3.7 Remote or Rural Microgrids

These microgrids never connect to the utility grid and instead operate in an island
mode at all times because of economic issues or geographical position. To incor-
porate renewable energy sources as an add-on to diesel generator-based systems
that so-called hybrid microgrids, provide great potential to diversify generation and
lower microgrid operating costs in rural areas [43]. The careful resource assessment
and understanding of demand profiles based on local conditions should be employed
for the selection of remote microgrids. On this basis, the effective and economic
operation of a microgrid is vital for sustained development; therefore, some DR
strategies may be used as an appropriate method to operate rural microgrids [44].

7.3.8 Other Microgrids

In addition to the above sectors, microgrids can also be used in other cases. For
example, today’s data centers are trying to make their operations more resilient
and efficient, and this is creating the perfect environment for new technologies
like advanced microgrids to flourish. Because of data center investments in backup
power equipment such as battery storage, a participant between a data center and
its local microgrid may improve the situation for all parties. In this way, the DR
schemes can be employed by interacting with data center facilities and microgrids
[45].
Furthermore, the electric vehicle (EV) is to be viewed as a distributed resource
and is becoming an enabling technology for microgrids. The integration EV-
based microgrid, and operation planning strategies can be created under different
vehicle behaviors with the minimum total cost goal [46]. Accordingly, the EV-based
microgrid, as well as renewable resources, can present new opportunities for DR
strategies so that can be employed to store energy when electricity consumption is
low and discharge it in times of peak demands [47].
Microgrids connected to cell phone towers could help nearby communities gain
access to electricity. In this way, energy service utilities can provide cell phone
tower owners and operators with electricity at a competitive price while also
providing electricity to nearby communities, with everyone being connected via
the microgrid. These towers typically rely on expensive diesel generators, but now,
renewable microgrids can offer the less expensive electricity prices, reliable and
clean alternative [48].
7 An Economic Demand Management Strategy for Passive Consumers. . . 189

7.4 The Decision Procedure for Operating

of a Microgrid Integrated with Demand Response

In this section, details of the proposed decision procedure are provided to select DR
schemes. For this purpose, the main concepts involved in this procedure are initially
discussed. Therefore, this section begins with the cost model of investment and
operation corresponding to the typical microgrid. Then, the impact of DR programs
on microgrid costs is investigated by modeling the cost of these programs. Finally,
the detailed descriptions about the decision procedure are presented.

7.4.1 Microgrid Cost Modeling

This section presents the costs related to installation, maintenance, operation, and
start-up of a microgrid. In order to compare these to the annual DR costs, microgrid
costs are expressed annually.

7.4.1.1 Microgrid Installation Cost

The cost of microgrid installation is included in the purchased cost of distributed

generation (DG) with the specified capacity and distribution infrastructure costs.
The first cost element can be formulated as the following equation [49]:

CDG−I = CIDG × P DG (7.4)

where CDG-I is the total installation cost of the microgrid. Also CIDG and PDG
are installation cost of DG ($/kW) and capacity of DG unit (kW), respectively.
Moreover, the distribution infrastructure costs are comprised of the network costs
and likely transformer cost. These can be calculated as follows:

CN T −I = CIN T × LN T (7.5)

CT R−I = CIT R × P T R (7.6)

where CNT-I , CIN T , and LNT indicate the overall installation cost of a private network,
the installation cost of a private network ($/m), and length of the network unit (m),
respectively. Furthermore, CTR-I ,CIT R , and PTR show the total installation cost of
the private transformer, the installation cost of a private transformer ($/kVA), and
capacity of the transformer unit (kVA), respectively.
190 M. E. Honarmand et al.

These costs should be converted to the annualized cost (CIAnn ) for a payback
period of n years and interest rate r, using the following equation:

r(r + 1)n
CIAnn = × (CDG−I + CN T −I + CT R−I ) (7.7)
(r + 1)n − 1

7.4.1.2 Microgrid Maintenance Cost

This cost includes the annual mechanical and electrical reformation costs. Generally,
this term is presented as a percentage of installation cost that can be calculated as
[50]:

CM = CI × ρ (7.8)

where ρ is a constant value in terms of percentage and CM is the maintenance cost

per year.

7.4.1.3 Microgrid Operation Cost

The cost of microgrid operation can consist of fuel cost, workforce, and so on. This
cost depends on the duration of microgrid operation; therefore, this equation can be
written as [49]:

CO = PAv
DG
× CO
DG
× T DG /24 (7.9)

DG is average generated power by DG. C DG and TDG are the operation cost
where PAv O
of DG source and duration of operating hours in a year, respectively.

7.4.1.4 Microgrid Start-Up Cost

Generally, this cost is considered only for fuel-consuming DG units. By definition,

the start-up cost can be shown as a function of two parts, that is, the hot and cold
start-up cost. This function is expressed as follows [51]:
+  ,
CS = CSH + CSC 1 − e(Toff /TC ) × N (7.10)

where CSH and CSC are hot and cold start-up costs ($), respectively. In addition, Toff is
the shutdown time of the DG unit, TC is the time constant of DG cooling and N
indicates the number of start-ups.
7 An Economic Demand Management Strategy for Passive Consumers. . . 191

Fig. 7.1 The typical curves of load and TOU price

Finally, the overall cost of microgrid deployment can be calculated as:

i
CMG = CIAnn + CM + CO + CS (7.11)

7.4.2 DR Cost Modeling

The costs of participating in DR schemes are considered as DR costs. In practice, the

customers will participate in these programs according to the comparison of the cost
of microgrid utilization and DR strategies. As mentioned previously, these programs
are categorized into PDR and IDR. In order to make proper decisions about the use
of microgrid or the participant in DR schemes, the cost model should be developed
for each scheme.

7.4.2.1 PDR-Based Cost

In PDR programs, electricity tariffs are defined based on different hours of

consumption in various sectors. The typical curves of power load and TOU prices
in each period are shown in Fig. 7.1. The load curve may be divided into three
parts: peak, mid-peak, and off-peak. On this base, TOU prices are selected as a
fixed tariff for each part. By predicting the consumption of each part, the participant
can calculate the related cost using different tariffs as follows:


3
CTi OU = pr T OU −m × WT OU −m (7.12)
m=1

where WTOU − 1 , WTOU − 2 and WTOU − 3 are the forecasted consumptions in

total periods of the peak, mid-peak, and off-peak, respectively. Also, prTOU − 1 ,
prTOU − 2 , and prTOU − 3 are the tariffs corresponding in the TOU scheme.
192 M. E. Honarmand et al.

Fig. 7.2 The typical curves of load and CPP price

CPP pricing is similar to the TOU strategy. In this scheme, the price level of
different periods is slightly lower than the TOU scheme. However, if an event
happens, the price is suddenly raised. Therefore, in this case, two values of
electricity consumption, that is, the normal and the event period, should be estimated
to calculate the total cost. A typical CPP scheme is illustrated in Fig. 7.2.
As mentioned earlier, if participants in this scheme fail to abide by their mandate,
they have to pay the penalty. Assuming all obligations are met, the estimated cost of
participating in the scheme is calculated as follows:


3 
3
’ ’
P = pr CP P −m × WCP P −m + P −m × WCP P −m
i
CCP pr CP (7.13)
m=1 m=1

where prCPP − m and pr ’CP P −m are related to the normal and event tariffs in each
period, respectively. Also WCPP − m and WCP ’
P −m indicate the estimated consump-
tions in the normal and the event periods, respectively.
Since RTP pricing expresses better flexibility than TOU and CPP schemes, this
tariff is indicated usually based on the consumption of the electric load. In this study,
the common RTP is described in relation to the load of microgrid, given as [13]:

pr RT P −k = ak WRT
2
P −k + bk WRT P −k + ck (7.14)

where prRTP − k and WRTP − k indicate the RTP price and the general consumption of
microgrid at time step k, respectively. Besides, the different values for ak , bk , and
ck can be selected based on the actual demand at various time steps. Therefore, the
cost of this scheme can be calculated to estimate the consumption at K periods as
follows:


K

P = pr RT P −k × WRT P −k
i
CRT (7.15)
k=1
7 An Economic Demand Management Strategy for Passive Consumers. . . 193

7.4.2.2 IDR-Based Cost

Despite the popularity of these programs, the determination of the incentive amount
is often arbitrary. Generally, these schemes are bilateral contracts in which the
participants can attend at certain periods and specific tariffs. Depending on the type
of scheme, there may be occasional penalties if the program does not run. In fact,
if the IDR contract seems profitable, the customers will be eager to participate in
the scheme. As a result, the costs of these programs are seen from the viewpoint of
electricity utility.
In order to increase the motivation to participate in the scheme, the related
incentives are often constituted from two terms: readiness charge and participation
incentive. The readiness charge is related to the fee of capability for the predefined
demand interrupt or curtailment that is specified in the contract. But after participat-
ing in the scheme, the incentive quantity is calculated based on the content of the
reduction in demand. Considering these points, by assuming the participation in the
total contract, the cost function is expressed as follows:
N 

CI DRk = PI DRk −i × TI DRk −i pr I DRk + PI’DRk × pr I’ DRk (7.16)
i=1

where CI DRk can be related to kth scheme, that is, DLC, ILP, and EDR. Also
PI DRk −i and TI DRk −i are ith demand amount and ith period participating in kth
scheme. Besides, pr I DRk indicates the price of participating in kth scheme per
$/kWh. The second element is introduced for readiness charge. Therefore, P ’ is I DRk
associated with the agreed interrupted/curtailment power and pr I’ DRk is related to
power price per $/kW, in kth scheme.

7.4.3 The Decision Algorithm

In order to provide flexible load side services, there are various options, such as the
use of microgrids and DR schemes, namely PDR and IDR. Because of the economic
savings resulting from the use of these options, customers have to consider one of
these choices. This is particularly visible and palpable to large consumers.
Generally, demand side plans are implemented according to the overall policies
of each distribution company. The PDR program, usually defined as a one- or several
month (almost long-term) period, seeks to modify or reduce end user consumption
at different times. By saving and managing consumption at different time slots,
consumers can control their costs at PDR prices. However, in practice, an electricity
consumer may not be willing to cooperate in reducing or disrupting electricity due to
the sensitivity of production lines or activities, specific requirements for a high level
of service, or the effort required on the consumer’s part to participate. Therefore, the
194 M. E. Honarmand et al.

consumer should either pay the costs in full by participating in the PDR program or
cover its entire load through the microgrid. The IDR program, which is often offered
to customers in critical condition of the network, results in short interruptions (1 to
4 hours). Contracts for these programs usually offer attractive tariffs to entice more
consumer participation. Despite these attractions, in practice, these plans may also
cause problems for some customers to reduce or disconnect, which may prevent
them from fully cooperating. In these cases, too, the microgrid can be used as a
backup to prevent customers from being interrupted at a given time. Therefore,
comparing the rewards of this program with the cost of microgrids will be crucial for
selecting the appropriate IDR program and the related interval. Thus, a combination
of IDR and microgrid can be an economic option for consumers.
Therefore, the integration of microgrid technology and demand-side man-
agement plans can increase customer satisfaction. Here, to plan the economic
demanding strategy of passive load considering DR programs and operating in
the form of microgrid, a procedure based on different cost evaluation is proposed.
In fact, in this approach, the cost of DR plans is examined along with the costs
of microgrid technology for a typical load. The flow chart of this procedure is
illustrated in Fig. 7.3. As observed, the investigation of two DR schemes is carried
out separately in the presence of microgrids. To this end, the PDR scheme is first
evaluated alongside the microgrid. Given the cost of the types of PDR programs,
the consumer can decide on whether or not to cooperate, in other words, to use
microgrid or not. However, by participating in IDR programs, the consumer can
utilize the microgrid to eliminate the problems of cutting or reducing the load due
to participate in these programs.
Regarding the proposed tariffs, this algorithm assumes that the cost of a DR
scheme is comparable to the microgrid operation costs. As it is observed, this
flowchart is divided into two steps, and each step consists of three levels (L-I, L-
II, and L-III), which are described as follows:
Step I: Microgrid operation and PDR programs.
At this step, the cost of PDR schemes is compared to the costs of microgrids. For
this purpose, first, one of the plans is selected, and the cost of one-day collaboration
in that PDR scheme is calculated. Then, the cost of one-day performance in the form
of a microgrid is obtained (L-I). These two costs are compared and if the cost of
cooperating in the PDR were less than the cost of the microgrid, the day’s counter
would increase. This increase will continue to a threshold value, M1, defined by
the operating company unless the cost of cooperating in the PDR exceeds the cost
of the imaginary microgrid for a day. From that day on, consumer performance in
the form of a microgrid would be economical. This is the point of decision and
discrimination between the choice of transferring into the form of a microgrid or
participating merely in DR programs (L-2). These calculations are performed for
other PDR schemes in the same manner and compared with the cost of microgrids,
and a discrimination point is specified for all schemes. Becoming a microgrid for
the number of days beyond this point will be the economical choice of the passive
load (L-3). It should be mentioned that there may be no discrimination point for any
7 An Economic Demand Management Strategy for Passive Consumers. . . 195

Start

Select C1 , C2 = 0
Step I Step II

Select a PDR scheme Select a IDR scheme

ii =1 kk =1

Number of day Number of day

ii = ii +1 kk = kk +1
i =1 k =1

Calculate Calculate
i = i +1 k = k +1
i
C MG , i
–T CPDR
k
C MG , k
ii –T CIDR
kk
L-I
Yes Yes
No No
–T < C PDR ii –T < C PDR ii
i i i i
C MG i+ 1 ≤ M 1 C MG k+ 1 ≤ M 2

No No
Yes Yes
Yes Yes
ii + 1 ≤ 3 kk + 1 ≤ 3
No No No
–T < C 1 ≠ 0
i
–T < C 2 ≠ 0
k
C MG C MG No

Yes Yes

Select C 1 = C MG
i
–T ; day = i ; PDR = ii Select C 2 =C MG
k
–T ; day = k ;IDR = kk

Yes Yes
ii + 1 ≤ 3 kk + 1 ≤ 3
L-II No No

Planning of MG Planning of MG
operation for i days operation for k days
integrated with No use of MG
integrated with
PDRii scheme IDRkk scheme
L-III

END

Fig. 7.3 The flowchart of the proposed framework to decide about economic demanding strategy
for a passive consumer

of the PDR schemes, which means that becoming a microgrid versus PDR plans is
not economical.
Step II: Microgrid operation and IDR programs.
At this stage, similar to the previous step, computation is executed at the first
level, and at the second level, comparisons are completed, and finally, at the third
level, the selection is made. The only difference is the nature of the IDR scheme.
Thus, for each time of execution of a particular IDR program in a defined interval
(one hour or more per interval), the value of the reward to the plan is calculated.
Accordingly, the rewards for IDR should be greater than the cost of transferring
196 M. E. Honarmand et al.

the consumer into the microgrid at each stage to benefit customers. Therefore, the
counter increases to the threshold value, M2, specified by the company, and the
discrimination points are specified for all the schemes. It is ultimately up to the user
to decide on the best strategy. Ultimately, the consumer makes the best choice by
comparing costs.

7.5 Numerical Studies

To assess the efficiency of the proposed procedure, three case studies of industrial,
commercial, and hospital customers are evaluated while they can be operated
as a microgrid and numerical results for DR strategies are analyzed thoroughly.
Customer data from the Guilan electrical distribution company are used in this study.
This company is responsible for providing distribution services for an approximate
of 1.5 million customers to the south of the Caspian Sea. The load curves of selected
customers are illustrated in Fig. 7.4. Also, the cost parameters of a typical microgrid
are presented in Table 7.1.
Furthermore, in all cases, it is assumed that seven events in peak, two events in
mid-peak, and one event in off-peak have been considered in the CPP scheme. In
addition, the penalties in the schemes are ignored.
For each load, the microgrid is connected to the external grid with 5800, 450, and
1600 kVA capacity, respectively, which can be operated at 90% of its capacity due

Fig. 7.4 The typical load curves (a) industrial, (b) commercial, and (c) hospital
7 An Economic Demand Management Strategy for Passive Consumers. . . 197

Table 7.1 Typical Parameter Unit Value

information of microgrid
Installation cost $/MW 320
Operation cost $/MWh 29
Maintenance cost $/MWh 7
Startup cost $ 0.15
Interest rate % 12.5
Planning period Year 15

Table 7.2 Tariffs of TOU and CPP schemes in case I

CPP price ($/kWh)
Period Hours TOU price ($/kWh) Nonevent day Event day
Peak (8 h) 12–17 0.14 0.096 0.46
20–24
Mid-peak (6 h) 9–12 0.09 0.065 0.15
17–20
Off-peak (10 h) 00–9 0.055 0.039 0.039

Table 7.3 Tariffs of IDR schemes in case I

For load reduction For readiness For consumption reduction
Scheme Hour ($/kW) ($/kW) ($/kWh)
DLC 4h 1.52 – –
ILP 2h – 3.04 0.09
EDR 1h 4.56 – –

to operational constraints. In this study, considering centralized-style consumption

of case studies, the cost terms related to the private distribution network and
transformer are ignored. First, the effect of running PDR schemes on microgrid
cost is investigated. Then an economic assessment of IDR program implementation
is provided. All the schemes compared with the state that the load can act as a
microgrid. In the end, the best option of the DR program is selected to operate the
microgrid.

7.5.1 Case I: Industrial Load

The peak load curve of a typical industrial firm is illustrated in Fig. 7.4a. This large
electrical load has various industrial production lines and a connected microgrid
to medium voltage (MV) with 5.8 MVA capacity. The different tariffs of TOU and
CPP schemes are shown in Table 7.2, whereas the proposed prices for IDR programs
are introduced in Table 7.3. The RTP rates are calculated using (7.14) by constant
coefficients 5 × 10−9 , 2 × 10−5 , and 0.01 for ak , bk , and ck , respectively.
198 M. E. Honarmand et al.

Fig. 7.5 The various costs for (a) PDR and (b) IDR, compared with microgrid cost in case I

It is assumed that the participation occurs during the summer season, and the
overall electric load is charged by microgrid when the customer participates in the
DR scheme. Accordingly, the costs of different DR schemes should be calculated
and compared with the deploying microgrid cost. In order to estimate the industrial
load consumption, historical data are used.
In this way, firstly, the cost of various PDR programs for different days are
calculated and depicted as Fig. 7.5a in comparison with microgrid utilization
cost. As it is observed, if the TOU program is planned for less than 55 days,
the participation in this program is economical compared to microgrid running;
whereas, the use of microgrid can be more economical for more days. Besides, this
number of days for the CPP program is equal to 75 days, while in RTP program is
roughly the same as the TOU scheme.
In this way, the similar results of the IDR program running are shown in Fig. 7.5b.
Generally, if the DLC contract is conducted 36 times a year, this contract may be
economical for this industrial consumer. Obviously, the contract is more economical
with more times a year. Furthermore, this value is for the ILP scheme equal to 17
times yearly, whereas, EDR contract can be economical annually for more 11 times.

7.5.2 Case II: Commercial Load

In this section, the impact of DR programs in the presence of microgrid for a

hypermarket with the peak load curve shown in Fig. 7.4b is investigated. The
microgrid is also connected to the MV grid at 450 kVA capacity. The tariffs of
different PDR and IDR schemes are shown in Table 7.4 and Table 7.5, respectively.
Besides, the constant coefficients 8 × 10−7 , 4 × 10−4 , and 0.05 for ak , bk , and
ck , respectively, are used to calculate RTP rates. Cost calculations are considered
by assuming the participation of total electric load in the summer and the use of
microgrid.
Similar to case I, the costs of DR program implementation are compared to
microgrid cost. The results are shown in Fig. 7.6. As can be seen in Fig. 7.6a,
7 An Economic Demand Management Strategy for Passive Consumers. . . 199

Table 7.4 Tariffs of TOU and CPP schemes in case II

CPP price ($/kWh)
Period Hours TOU price ($/kWh) Nonevent day Event day
Peak (7 h) 17–24 0.3 0.2 0.94
Mid-peak (7 h) 9–17 0.22 0.15 0.7
Off-peak (10 h) 00–9 0.13 0.09 0.09

Table 7.5 Tariffs of IDR schemes in case II

For load reduction For readiness For consumption reduction
Scheme Hour ($/kW) ($/kW) ($/kWh)
DLC 4h 1.73 – –
ILP 2h – 3.00 0.057
EDR 1h 3.46 – –

Fig. 7.6 The various costs for (a) PDR and (b) IDR, compared with microgrid cost in case II

the TOU, CPP, and RTP schemes are economical for 14, 21, and 11 days in a
year, respectively, compared with microgrid cost. Indeed, the microgrid can be used
instead of participating in the above schemes with longer intervals.
Accordingly, the results of IDR programs running are shown in Fig. 7.6b. In
this way, the contract of DLC, ILP, and EDR schemes are economical for more
29, 16, and 14 times per year, respectively. Therefore, for a commercial consumer
is economical to use a microgrid if periods of IDR contract are greater than the
mentioned values.

7.5.3 Case III: Hospital Load

In this case, the deployment cost of the microgrid in a hospital integrated with
DR programs that has the peak load curve in Fig. 7.4c, is assessed. The microgrid
is connected to the MV grid at 1.6 MVA capacity. The TOU and CPP tariffs in
200 M. E. Honarmand et al.

Table 7.6 Tariffs of TOU and CPP schemes in case III

CPP price ($/kWh)
Period Hours TOU price ($/kWh) Nonevent day Event day
Peak (8 h) 12–17 0.14 0.096 0.46
20–23
Mid-peak (13 h) 7–12 0.09 0.065 0.15
17–20 23–4
Off-peak (3 h) 4–7 0.055 0.039 0.039

Table 7.7 Tariffs of IDR schemes in case III

For load reduction For readiness For consumption reduction
Scheme Hour ($/kW) ($/kW) ($/kWh)
DLC 4h 2.6 – –
ILP 2h – 3.04 0.0176
EDR 1h 5.2 – –

Fig. 7.7 The various costs for (a) PDR and (b) IDR, compared with microgrid cost in case III

Table 7.6 and the IDR tariffs in Table 7.7 are shown. The RTP rates are calculated
by the constant coefficients 8 × 10−7 , 5 × 10−5 , and 0.022 for ak , bk , and ck ,
respectively. In the presence of a microgrid, if the overall electric load in the summer
can contribute to DR schemes, different costs are calculated.
The calculation results for PDR and IDR schemes are depicted in Fig. 7.7. The
CPP program for less than 58 days is economical compared to microgrid utilization,
whereas the TOU scheme is more suitable for less than 36 days. Similarly, for less
than 32 days, the RTP program is an economic scheme. These results are shown in
Fig. 7.7a. Obviously, the use of microgrid may be cost-effective if there are more
days in the proposed schemes.
Figure 7.7b illustrates the cost of IDR schemes compared to microgrid cost.
Accordingly, the contract of DLC, ILP, and EDR schemes is economical for more
19, 16, and 9 times per year, respectively. As a result, in order to be economical for
7 An Economic Demand Management Strategy for Passive Consumers. . . 201

Table 7.8 A summary of PDR scheme (days) IDR scheme (hours)

results to use different
Load TOU CPP RTP DLC ILP EDR
schemes integrated with
microgrid Industrial 55 75 52 36 17 11
Commercial 14 21 11 29 16 14
Hospital 36 58 32 19 16 9

microgrid utilization, this consumer should conclude a contract with more than the
mentioned values for each scheme

7.5.4 Final Deduction

The results presented in the previous sections are analyzed here. Table 7.8 sum-
marizes these results. By comparing the cost of microgrid and PDR schemes, it
is economical to use these schemes for the maximum number of days specified in
Table 7.8. Since PDR schemes should be deployed for the summer season (90 days),
the operation of a microgrid for the commercial load is entirely economical instead
of PDR tariffs. For hospital load, microgrid operation is more suitable compared to
TOU and RTP schemes; however, with these tariffs, the least economic benefit of
microgrid utilization is for the industrial load.
An IDR scheme alongside with microgrid operation is economical when the
number of times running the program in contract exceeds the mentioned values
in Table 7.8. Although the microgrid is operated when a load is interrupted, an
important limitation of these schemes is the execution number; the load may be
disconnected or connected time after time which can cause problems for some
customers. On the whole, the participation of all consumers in ILP scheme can be
economical for values greater than those presented in the Table 7.8.

7.6 Conclusions

In this chapter, an economic demand management strategy for a passive consumer

considering demand-side management schemes and microgrid operation was pro-
posed. The main categories of DR programs, including PDR and IDR schemes
were examined. Microgrid utilization for different types of loads was investigated in
detail. To this end, the cost models for microgrid and DR schemes were extended.
Based on these models, a decision criterion for determining the best choice for
supplying the consumer demand was developed. In this regard, the cost of microgrid
utilization alongside DR schemes was analyzed and compared. Several practical
examples were provided and the results of the proposed method in the real case
studies were presented.
202 M. E. Honarmand et al.

References

1. Zia, M. F., Elbouchikhi, E., & Benbouzid, M. (2018). Microgrids energy management systems:
A critical review on methods, solutions, and prospects. Applied Energy, 222, 1033–1055.
2. Hosseini Imani, M., Niknejad, P., & Barzegaran, M. R. (2018). The impact of customers’
participation level and various incentive values on implementing emergency demand response
program in microgrid operation. Electrical Power and Energy Systems, 96, 114–125.
3. Shen, J., Jiang, C., Liu, Y., & Qian, J. (2016). A microgrid energy management system with
demand response for providing grid peak shaving. Electric Power Components and Systems,
44(8), 843–852.
4. Aghajani, G. R., Shayanfar, H. A., & Shayeghi, H. (2017). Demand-side management in a
smart micro-grid in the presence of renewable generation and demand response. Energy, 126,
622–637.
5. Kumar, D., Verma, Y. P., & Khanna, R. (2019). Demand response-based dynamic dispatch
of microgrid system in hybrid electricity market. International Journal of Energy Sector
Management, 13(2), 318–340.
6. Hosseini Imani, M., Jabbari Ghadi, M., Ghavidel, S., & Li, L. (2018). Demand response
modeling in microgrid operation: A review and application for incentive-based and time-based
programs. Renewable and Sustainable Energy Reviews, 94, 486–499.
7. Nguyen, A. D., Bui, V. H., Hussain, A., Nguyen, D. H., & Kim, H. M. (2018). Impact of
demand response programs on optimal operation of multi-microgrid system. Energies, 11(6)
1452, 1–18.
8. Han, X., Zhang, H., Yu, X., & Wang, L. (2016). Economic evaluation of grid-connected micro-
grid system with photovoltaic and energy storage under different investment and financing
models. Applied Energy, 184, 103–118.
9. Zakariazadeh, A., Jadid, S., & Siano, P. (2014). Smart microgrid energy and reserve scheduling
with demand response using stochastic optimization. Electrical Power and Energy Systems, 63,
523–533.
10. Zunnurain, I., Maruf, M. N. I., Rahman, M. M., & Shafiullah, G. M. (2018). Implementation of
advanced demand-side management for microgrid incorporating demand response and home
energy management system. Infrustructures, 3(4), 1–25.
11. Jin, M., Feng, W., Marrnay, C., & Spanos, C. (2018). Microgrid to enable optimal distributed
energy retail and end-user demand response. Applied Energy, 210, 1321–1335.
12. Ferruzzi, G., Graditi, G., Rossi, F., & Russo, A. (2015). Optimal operation of a residential
microgrid: The role of demand-side management. Intelligent Industrial Systems, 1, 61–82.
13. Yang, X., Zhang, Y., He, H., & Weng, G. (2019). Real-time demand-side management for a
microgrid considering uncertainties. IEEE Trans. on Smart Grid, 10(3), 3401–3414.
14. Department of Energy, Benefits of demand response in electricity markets and recommenda-
tions for achieving them, a report to the United States congress pursuant to section 1252 of the
energy policy act of 2005, 2006.
15. Hussain, M., & Gao, Y. (2018). A review of demand response in an efficient smart grid
environment. The Electricity Journal, 31(5), 55–63.
16. Siano, P. (2014). Demand response and smart grids- a survey. Renewable and Sustainable
Energy Reviews, 30, 461–478.
17. Nicolson, M. L., Fell, J., & Huebner, M. (2018). Consumer demand for time of use electricity
tariffs: A systematized review of the empirical evidence. Renewable and Sustainable Energy
Reviews, 97, 276–289.
18. Hung, M. F., & Huang, T. H. (2015). Dynamic demand for residential electricity in Taiwan
under seasonality and increasing-block pricing. Energy Economics, 48, 168–177.
19. Paterakis, N. G., Erdinc, O., & Catalao, J. P. S. (2017). An overview of demand response:
Key-elements and international experience. Renewable and Sustainable Energy Reviews, 69,
871–891.
7 An Economic Demand Management Strategy for Passive Consumers. . . 203

20. Yang, C., Meng, C., & Zhaou, K. (2018). Residential electricity pricing in China: The context
of price-based demand response. Renewable and Sustainable Energy Reviews, 81, part 2,
2870–2878.
21. Xu, F. Y., Zhang, T., Lai, L. L., & Zhou, H. (2015). Shifting boundary for price-based
residential demand response and applications. Applied Energy, 146, 353–370.
22. Davito, B., Tai, H., & Uhlaner, R. (2010). The smart grid and the promise of demand-side
management. McKinsey on Smart Grid Summer, 2010, 38–44.
23. Shen, B., Ghatikar, G., Lei, Z., Li, J., Wikler, G., & Martin, P. (2014). The role of regulatory
reforms, market changes, and technology development to make demand response a viable
resource in meeting energy challenges. Applied Energy, 130, 814–823.
24. Vardakas, J. S., Zorba, N., & Verikoukis, C. V. (2015). A survey on demand response programs
in smart grids: Pricing methods and optimization algorithms. IEEE Communications Surveys
and Tutorials, 17(1), 152–178.
25. Aghaei, J., Alizadeh, M.-I., Siano, P., & Heidari, A. (2016). Contribution of emergency demand
response programs in power system reliability. Energy, 103, 688–696.
26. Parhizi, S., Lotfi, H., Khodaei, A., & Bahramirad, S. (2015). State of the art in research on
microgrids: A review. IEEE Access, 3, 890–925.
27. Hatziargyrion, N., Asano, H., Iravani, R., & Marnay, C. (2007). Microgrids: An overview of
ongoing research, development, and demonstration projects. IEEE Power Energy Magazine,
5(4), 78–94.
28. Lasseter, R. H., & Paigi, P. (2004). Microgrid: A conceptual solution. IEEE 35th Annual Power
Electronics Specialists Conference, 4258–4290.
29. Microgrid Knowledge, Microgrids and distributed energy: How customers can get full
value, 2019. Available at: https://microgridknowledge.com/asset-backed-demand-response-
microgrids/.
30. Shoreh, M. H., Siano, P., Shafie-khah, M., Loia, V., & Catalao, J. P. S. (2016). A survey of
industrial applications of demand response. Electric Power Systems Research, 141, 31–49.
31. Mohagheghi, S., & Raji, S. (2014). Managing industrial energy intelligently: Demand response
scheme. IEEE Industrial Applications Magazine, 20(2), 53–62.
32. Smith, D. M., St. Leger, A., & Severson, B. (2015). Automated demand response of thermal
load with a photovoltaic source for military microgrids. IEEE Clemson University Power
Systems Conference, 1–8.
33. Van Broekhoven, S., Judson, N., Galvin, J., & Marqusee, J. (2013). Leading the charge:
Microgrids for domestic military installations. IEEE Power and Energy Magazine, 11(4), 40–
45.
34. Hirsch, A., Parag, Y., & Guerrero, J. (2018). Microgrids: A review of technologies, key drivers,
and outstanding issues. Renewable and Sustainable Energy Reviews, 90, 402–411.
35. Hanna, R., Ghonima, M., Kleissl, J., Tynan, G., & Victor, D. G. (2017). Evaluating business
models for microgrids: Interactions of technology and policy. Energy Policy, 103, 47–61.
36. Meegahapola, L. G., Robinson, D., Agalgaonkar, A. P., Perera, S., & Ciufo, P. (2014).
Microgrids of commercial buildings: Strategies to manage mode transfer from grid connected
to islanded mode. IEEE Trans. on Sustainable Energy, 5(4), 1337–1347.
37. Wang, Y., Wang, B., Chu, C. C., Pota, H., & Gadh, R. (2016). Energy management for
a commercial building microgrid with stationary and mobile battery storage. Energy and
Buildings, 116, 141–150.
38. Cotto, S. V., & Lee, W. J. (2016). Challenges and opportunities: Microgrid modular design for
tribal healthcare facilities. IEEE North American Power Symposium, 1–6.
39. Dincer, I., & Abu-Rayash, A. (2019). Energy sustainability. Elsevier Science & Technology.
40. Sechilariu, M., Wang, B., & Locment, F. (2013). Building-integrated microgrid: Advanced
local energy management for forthcoming smart power grid communication. Energy and
Buildings, 59, 236–243.
41. Jamshidi Monfared, H., Ghasemi, A., Loni, A., & Marzband, M. (2019). A hybrid price-based
demand response program for the residential micro-grid. Energy, 185, 274–285.
204 M. E. Honarmand et al.

42. Kuang, Y., Zhang, Y., Zhou, B., Li, C., Cao, Y., Li, Y., & Zeng, L. (2016). A review of
renewable energy utilization in islands. Renewable and Sustainable Energy Reviews, 59, 504–
513.
43. Anand, H., & Ramasubbu, R. (2018). A real time pricing strategy for remote micro-grid with
economic emission dispatch and stochastic renewable energy sources. Renewable Energy, 127,
779–789.
44. Thompson, C. C., Oikonomou, P. E. K., Etemadi, A. H., & Sorger, V. J. (2015). Optimization
of data center battery storage investments for microgrid cost saving, emissions reduction, and
reliability enhancement. IEEE Industry Applications Society Annual Meeting, 1–7.
45. Aluisio, B., Dicorato, M., Ferrini, I., Forte, G., Sbrizzai, R., & Trovato, M. (2019). Optimal
sizing procedure for electric vehicle supply infrastructure based on DC microgrid with station
commitment. Energies, 12(10), 1–19.
46. Tushar, M. H. K., Zeineddine, A. W., & Assi, C. (2018). Demand-side management by
regulating charging and discharging of the EV, ESS, and utilizing renewable energy. IEEE
Trans. on Industrial Informatics, 14(1), 117–126.
47. Aliasghari, P., Mohammadi-Ivatloo, B., Alipour, M., Abapour, M., & Zare, K. (2018). Optimal
scheduling of plug-in electric vehicles and renewable micro-grid in energy and reserve markets
considering demand response program. Journal of Cleaner Production, 186, 293–303.
48. Kansal, S., Tyagi, B., & Kumar, V. (2017). Cost-benefit analysis for optimal DG placement in
distribution system. International Journal of Ambient Energy, 38(1), 45–54.
49. Chen, C., & Duan, S. (2014). Optimal allocation of distributed generation and energy storage
system in microgrids. IET Renewable Power Generation, 8(6), 581–589.
50. An, L. N., Quoc-Tuan, T., Seddik, B., & Van-Linh, N. (2015). Optimal sizing of a grid-
connected microgrid, IEEE International Conference on Industrial Technology (ICIT), 1–6.
Chapter 8
Real-Time Perspective in Distributed
Robust Operation of Networked
Microgrids

Mehdi Jalali, Manijeh Alipour, and Kazem Zare

8.1 Introduction

By extending distributed energy resources, the structure of traditional distribution

networks is moving toward distribution systems. Furthermore, some consumers are
equipping to the local generation units and controlling elements to obtain more
reliable electricity and more economic benefits by constructing microgrids (MGs).
According to CIGRE WG C6.21, microgrids are defined as “small controllable
energy systems including loads and energy resources (such as micro-turbines,
storage devices, or controllable loads) that can be operated in a coordinated way
either while connected to the main network or while islanded” [1]. In the future
distribution systems, there will be multiple distinct microgrids that each of them
is responsible to provide its own consumption and can operate in grid connected
and islanded modes based on the owner’s strategies. At the same time, microgrids
will consider participating in day-ahead (DA) and real-time (RT) markets besides
managing their own distributed energy resources (DERs). Therefore, designing a
trading mechanism among networked microgrids is an urgent task in which energy
exchange between MGs will be done by the agreement of both the seller and buyer.
In this respect, real-time scheduling based on real-time forecasts tackle balance
between supply and demand at each entity.
Some operation management models for distribution systems with individual
microgrids are presented in literature which can be distinguished in the terms
of dependency in the objective function and ownerships and market clearing
entity. In [2], based on the transactive energy framework, an energy management
model for DERs in the form of a virtual power plant is presented. In [3], the

M. Jalali () · M. Alipour · K. Zare

Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, Iran
e-mail: mehdi.jalali@tabrizu.ac.ir; alipour@tabrizu.ac.ir; kazem.zare@tabrizu.ac.ir

© Springer Nature Switzerland AG 2021 205

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_8
206 M. Jalali et al.

distribution system operator (DSO) determines distributed locational marginal price

in the distribution network, then calculates transactive energy incentive signals and
shares with participants. A distributed approach for energy sharing between several
residential microgrids is presented in [4]. The optimal energy sharing problem
in [6] to tackle the less computational burden and sharing private information is
decomposed into one master problem and two subproblems. The decision variables
of subproblems are sent to the master problem which controls the energy interaction
between microgrids. In [5], a multi-follower bi-level programming approach is
presented which by using KKT conditions from the DSO’s point of view could
not be implemented for the case of independent microgrids [5].
In [6], the alternating direction method of multiplier (ADMM) is implemented to
relax the complicating power balance between the distribution system operator and
microgrids. In fact, the result of this approach is corresponding to the optimal sum
of DSO and MGs’ costs. Obviously, the results of [5] would not lead to the Nash
equilibrium point.
The main idea lied in this chapter is presenting a framework for networked MGs
that have different owners. Compared with the discussed methods, we designed
a new third-part economic entity (TPEE) that it is not on the domination of
any operator (DSO or MGs) and does not allow the utility operators to access
the other players’ information and their economic values. Microgrids’ bids and
offers, pairs of price, and corresponding quantity that are named as economic
values, are submitted to the TPEE. Then, TPEE clears the bids and offers by
determining feasible transactions and sellers and buyers. TPEE determines the
feasible transactions’ social welfare values, provide integrated buying and selling
bid and offers. At the next stage, the physical power between sellers and buyers
corresponding to the approved transactions and satisfaction would be exchanged
by satisfying the equality of technical parameters. To have optimal self-scheduling
within the MGs, ADMM as a distributed optimization technique is implemented.
In other words, distributed optimization provides a framework for the satisfaction
of voltage magnitude and voltage angle equality at the connection nodes. The
presented mechanism feeds by updated forecasts and prepares real-time energy
management schemes through rolling horizons.

8.2 Methodology Description

In this section, the functions of the proposed TPEE and energy management within
the MGs are described in detail. To satisfy the secure information sharing and
equality in terms of voltage magnitudes and angles, the trading mechanism and
physical interactions have been distinguished.
TPEE is considered as an intermediary in exchanging economic values (selling
offers and buying bids) between individual microgrid owners. This mechanism does
not allow any operator to have knowledge of the behavior of other ones and cannot
impose its volition on other participants. TPEE receives economic values from the
8 Real-Time Perspective in Distributed Robust Operation of Networked Microgrids 207

Fig. 8.1 Information sharing

in the presented framework TPEE

MG1 MGn

MG2 MGi

Power delivering signal Economic value signals

MGs and clears the offers and bids. The economic values can be determined by
using different methods. We supposed that MGs provide economic values (i.e., bid
and offer curves) respect to their neighborhood entities and submit to the TPEE.
TPEE determines sets of sellers and buyers based on submitted economic values.
It should be noted that TPEE would not consider the MGs’ technical constraints.
The summary of proposed energy trading is presented in Algorithm 8.1.

Algorithm 8.1 The proposed transactive energy trading

1. Each entity submits economic (selling offers and buying bids) to TPEE.
2. TPEE selects the transaction corresponds which are profitable for both of
the participants.
3. Each entity sends its own final satisfactory by a signal to TPEE.
4. TPEE finalizes all deliverable and approved transactions.
5. MGs schedule their own resources considering approved transactions via
ADMM.

The proposed energy trading microgrids and utility is illustrated in Fig. 8.1.
According to Fig. 8.1, participants share their economic values with TPEE and
after local market clearing by TPEE, participants share power delivering signal (dual
variable of complicated constraints) with each other.
The power exchange and self-scheduling of resources within the microgrids and
distribution network are tackled using distributed optimization and is described in
the next section.
208 M. Jalali et al.

8.3 Problem Formulation

In this section, the physical energy exchange between entities and self-scheduling
of resources with the individual entities have been addressed.

8.3.1 Microgrids’ Real-Time Energy Management

According to Fig. 8.1, the interconnected MGs can participate in the main electricity
DA and RT markets. Based on the DA market’s gate at a specific time prior to the
trading day will be closed and after a short delay, the DA market’s price will be
published. The RT market opens at each time slot of the trading day and closes
specific minutes before the start of the trading hour and results will be published
prior to the trading start time. Since MGs do not have any information about prices
in the DA and RT markets before the markets’ clearing times, it needs to have
appropriate forecasts as far as possible. After the DA market’s clearing time, MGs
provides an optimal operating scheme considering updated forecasts. On the other
hand, MGs by participating in the proposed energy framework can sell or buy power
from each other. To have proper operation scheduling, the entities need to have an
accurate prediction of their electricity consumption and price of DA and RT market.
Based on rolling horizon load and electricity markets’ prices, the objective function
from the MG i’s point of views can be formulated as:
⎡
 ⎢ DA,i DA,i  
Min ⎣ λ̂t|k Pt|k + λ̂RT
t|k
,i RT ,i
Pt|k + ai,g Pi,g,t|k + bi,g
MT MT MT
MG
Ψi k=t:T g∈ΩiMT
,
+SU C MT
i,g,t|k + SDC i,g,t|k
MT

(8.1)

The first term in (8.1) denotes the cost of purchased power from day-ahead
and real-time markets. Price uncertain parameters (λ̂DA,i
t|k andλ̂RT ,i
t|k ) are forecasted
day-ahead and real-time markets’ prices at time slot t for the k-hour ahead. The
second term of (8.1) is used to model operation, start-up, and shut-down costs
of micro-turbines within the MG i. Since TPEE clears trades before real-time,
the transactions’ cost of the next time horizon for the player are predetermined.
Therefore, these terms are not included in the objective function. The operating
constraint and load follow equations in the microgrid can be expressed as:
 
i,g,t|k ≥ SU i,g
SU C MT − δi,g,t|k−1
MT MT MT
δi,g,t|k (8.2)
8 Real-Time Perspective in Distributed Robust Operation of Networked Microgrids 209

 
SDC MT
i,g,t|k ≥ SD MT
i,g δ MT
i,g,t|k−1 − δ MT
i,g,t|k (8.3)

 
on
Ti,g,t|k ≥ U T MT
i,g δ MT
i,g,t|k − δ MT
i,g,t|k−1 (8.4)

 
off
Ti,g,t|k ≥ DT MT
i,g δ MT
i,g,t|k−1 − δ MT
i,g,t|k (8.5)

MT
δi,g,t|k × Pi,g
Min
≤ Pi,g,t|k
MT
≤ δi,g,t|k
MT
× Pi,g
Max
(8.6)

MT
δi,g,t|k × QMin
i,g ≤ Qi,g,t|k ≤ δi,g,t|k × Qi,g
MT MT Max
(8.7)

I nj
 DA,i RT ,i
 
Pi,n,t|k = Pt|k + Pt|k + MT
Pi,g,t|k − P̂i,n,t|k
d
− P Linm,t|k
s∈ΩnU,i g∈ΩnMT ,i m∈ΩnBus,i
(8.8)

I nj
  
Qi,n,t|k = t|k +
QG i,g,t|k − Q̂i,n,t|k −
QMT d
QLinm,t|k (8.9)
s∈ΩnU,i g∈ΩnMT ,i m∈ΩnBus,i

      
I nj
Pi,n,t|k = Ginn 2Vi,t|k
n −1 + i
Bnm n − θm
θi,t|k i n m
i,t|k + Gnm Vi,t|k + Vi,t|k − 1 (8.10)
m∈ΩnBus,i

      
I nj i n
Qi,n,t|k = Bnn 1 − 2Vi,t|k + Ginm θi,t|k
n − θm i n m
i,t|k − Bnm Vi,t|k + Vi,t|k − 1 (8.11)
m∈ΩnBus,i

   
P Linm,t|k = Ginm Vi,t|k
n
− Vi,t|k
m
+ Bnm
i n
θi,t|k − θi,t|k
m
(8.12)

   
QLinm,t|k = −Bnm
i n
Vi,t|k − Vi,t|k
m
+ Ginm θi,t|k
n
− θi,t|k
m
(8.13)
210 M. Jalali et al.

SLinm,t|k ∼
= P Linm,t|k + ξnm,t|k
i
.QLinm,t|k (8.14)

−SLi,Max
nm ≤ SLinm,t|k ≤ SLi,Max
nm (8.15)

ViMin ≤ Vi,t|k
n
≤ ViMax (8.16)

The set of MG i’s independent decision-making variables is defined as

DA,i RT ,i MT
ΨiMG ={Pt|k i,g,t|k ,δi,g,t|k ,Vi,t|k ,θi,t|k }. The start-up and shout-
,Pt|k ,Pi,g,t|k ,QMT MT n n

down costs are formulated as (8.2) and (8.3), respectively. The micro-turbines’
minimum uptime and downtime constraints are subjected to (8.4) and (8.5),
respectively. Constraints (8.6) and (8.7) satisfy the minimum and maximum active
and reactive power generation of micro-turbines. The details of the implemented
power flow method can be followed in [6]. According to the power flow method
presented in [6], the active and reactive power balances can be considered in (8.8)
and (8.9). The injected active and reactive powers into each node are equal to (8.10)
and (8.11), respectively. The linearized form of active and reactive powers which
flow in-line between node n and m are modeled according to (8.12) and (8.13),
respectively. By implementing Taylor series, the apparent power flow through
the line between node n and node m is approximated introducing an auxiliary
i
parameter (ξnm,t|k ), in (8.14), and is enforced to be limited according to the line
capacity in (8.15). The maximum and minimum limits on voltage magnitudes are
shown in (8.16). As addressed in microgrid planning related works, microgrids
within the distribution systems can be operated in islanded and grid-connected
modes. It should be noted that considered power flow eqs. (8.2)–(8.16) would be
duplicated for each microgrid. The connection point of two entities corresponding
to each transaction is imagined as a zero impedance line between MG i and MG
j. By exchanging approved power between seller and buyer at nodes i and j, the
constraints corresponding to the voltage magnitudes and voltage angles can be
written as follows:
CPj CPi
Vi,t = Vj,t (8.17)

CPj CPi
θi,t = θj,t (8.18)

The Eqs. (8.17) and (8.18) satisfy the voltage magnitude and voltage angle
equality at connection points (CP) of microgrid i and j. It should be noted that the
value of power exchange between entities is predetermined by TPEE.
8 Real-Time Perspective in Distributed Robust Operation of Networked Microgrids 211

8.3.2 Distributed Energy Management

As mentioned in stage 5 of Algorithm 8.1, ADMM is implemented to distribute

the operation scheduling and physical exchanging of the approved transactions. The
general form of implemented ADMM for the relaxing complicated constraints (see.
(8.17) and (8.18)) between individual MGs is presented [7]. Using ADMM, the
modified objective functions of each MG by relaxing complicated constraints (8.17)
and (8.18) at iteration κ + 1can be written as:

⎡ ⎤
#   #
Min ⎣ λ̂ DA,i DA,i RT ,i
+ λ̂t|k Pt|k RT ,i
+ ai,g Pi,g,t|k + bi,g + +SDC i,g,t|k ⎦
MT MT MT MT
t|k Pt|k
ΨiMG k=t:T g∈ΩiMT
 - 
      -2
# V ,κ CPj CPi κ ρV -
- CPj CPi κ --
+ λij,t Vi,t − V̂j,t + 2 -Vi,t − V̂j,t - (8.19)
j ∈ΩiMG 2
 - 
     κ -
# CPj CPi κ ρθ -
-θ CPj − θ̂ CPi -
-2
+ λθ,κ
ij,t θ i,t − θ̂ j,t + 2 - i,t j,t -
j ∈ΩiMG 2

In comparison with (8.1), the first and second terms added in (8.19) are
augmented lagrangian terms based on ADMM. The objective function (8.19) subject
to constraints (8.3)–(8.16) would be minimized by MG i at each time t, before
real-time market gate closing time and after publishing TPEE. Parameter ρ V and
ρ θ are positive values. The constraints (8.17) and (8.18) duality variables as power
delivering signals between MG i and j will be updated at iteration κ + 1 at time slot
t as:
 CP κ   
V ,κ+1 CPi κ
λij,t = λVij,t,κ + ρV V̂i,t j − V̂j,t (8.20)

 CP κ   
θ,κ+1 CPi κ
λij,t = λθ,κ
ij,t + ρ
θ
θ̂i,t j − θ̂j,t (8.21)

To have a robustness vision about the uncertain parameters, uncertain parameters

are fixed at their worst scenarios.

8.3.3 Numerical Studies and Result Analysis

A case study is conducted based on a test system with four microgrids which are
shown in Fig. 8.2. Each MG consists of several micro-turbines with the location and
operation characteristics including feasible maximum, minimum active, and reactive
powers as well as coefficient of cost functions are presented in Table 8.1.
212 M. Jalali et al.

Fig. 8.2 Characteristics test networked MGs

Table 8.1 Characteristics of Entity Bus Pmax [MW] Qmax [MW] MT [$/MW]
ai,g
the micro-turbines
MG1 2 3 1.2 0.025
3 3 1.2 0.026
5 3 1.2 0.027
6 2 0.8 0.028
MG2 6 2 0.8 0.029
9 2 0.8 0.027
10 1.5 0.6 0.027
MG3 3 3 1.2 0.024
5 2.5 1 0.024
MG 4 5 3 1.2 0.021
6 2.5 1 0.021

The magnitudes of base apparent power and base voltage are considered to be 10
MVAr and 20 kV, respectively. The lines resistance and reactance are 0.05 p.u and
0.1 p.u, respectively.
The implemented forecasting framework for the electricity demand, price is
inspired by the rolling horizon technique [8]. To provide a simple and applicable
forecasting method, auto-regressive with exogenous input variables (ARX) is used
to train forecasting models. The most informative input features are selected based
on the technique which is presented in [9]. The candidate inputs for the load
forecasting consist of two sets. The first set includes the lagged value corresponding
to 1, 24, 48, 72, 96, 120, 144, and 168 hours ago of the electricity load and the second
set is corresponding to the exogenous variables which are minimum and maximum
loads of the previous day and the same day in the previous week, and the hours
corresponding to occurrences of peak load and ramp-up hours of the previous day
8 Real-Time Perspective in Distributed Robust Operation of Networked Microgrids 213

Table 8.2 1-Hour ahead load Time MG1 MG2 MG3 MG4
forecasts in the distribution
network and individual 1 1158.50 737.20 462.53 2640.74
microgrids 2 1134.64 704.53 455.79 2582.51
3 1184.24 691.70 447.97 2564.61
4 1253.37 730.86 443.11 2755.95
5 1348.92 805.24 444.25 3034.38
6 1409.19 873.10 450.55 3070.22
7 1495.77 908.14 462.90 3153.46
8 1497.29 907.17 469.36 3131.52
9 1436.91 872.26 478.03 3073.04
10 1408.13 851.57 478.74 2941.85
11 1335.63 809.79 491.14 2814.18
12 1241.29 741.67 490.46 2729.38
13 1198.63 711.07 497.04 2664.55
14 1155.17 686.46 493.18 2622.46
15 1144.44 680.17 491.44 2574.77
16 1144.29 734.05 503.98 2578.03
17 1167.25 813.55 518.13 2624.61
18 1197.32 879.85 563.51 2503.48
19 1219.82 928.34 616.12 2619.79
20 1244.99 939.39 641.90 2647.13
21 1231.34 922.57 648.50 2625.74
22 1220.24 871.46 639.96 2627.73
23 1179.76 793.23 628.21 2525.90
24 1147.55 741.95 643.28 2448.08

and the same day in the previous week. The 1-hour ahead electricity load forecasts
of microgrids are reported in Table 8.2. According to the feature selection technique
which is presented in [10], the considered features for the electricity price of day-
ahead market consist eight lagged prices (1, 2, 12, 23, 24, 48, 72, and 168 hour
ago prices) plus four lagged loads (1, 12, 24, and 73 hour ago loads) as exogenous
variables. The historic data for price forecasting is obtained from [11].
As described in Algorithm 1, the entities submit their selling and buying bids
and offers to the IEE. IEE collects the offers and determines the transactions. IEE
determines the seller and buyers as well as the price of each transaction.
TPEE sorts buyers’ bids and sellers’ offers in the highest and lowest bids and
offers, respectively. This mechanism has been illustrated in Fig. 8.3. As shown in
Fig. 8.3a, at time slot t = 1, the bids of buyers are 0.0295, 0.029, 0.027, and 0.021
respectively belong to MG2, MG3, MG1, and MG4. The offered price for the selling
are 0.0232 and 0.0335 belonging to MG4 and MG3. Therefore, TPEE based on the
value of feasible transactions selects participants in an ordered pair form (SELLER,
BUYER). The selected transactions for time slot t = 1 are: (MG4, MG1), (MG4,
MG2), (MG4, MG3). For time slot t = 23, based on the suggested price for the
214 M. Jalali et al.

Fig. 8.3 Transaction clearing by TPEE at time slot (a) t = 1 and (b) t = 23

(a) - Price and participating in DA and RT markets (b) - Price and energy exchange MG4 with other MGs

Fig. 8.4 MG4’s participating in electricity markets and interactions with other MGs. (a) Price and
participating in DA and RT markets. (b) Price and energy exchange MG4 with other MGs

buying and selling as well as the quantities, TPEE accepts the transaction between
TPEE and MG1.
Since the proposed energy management is established for real-time applications,
TPEE should clear transactions between MGs before the final scheduling of entities.
Entities by receiving their own approved transactions for the next time slot, solve
the operation problem. Due to handling the approved transactions, the alternating
direction method of multipliers according to Algorithm 8.1 has been implemented
in an iterative manner until achieving convergence criteria.
MGs in the connection with upper grids can participate in day-ahead and
real-time markets. According to results, MG4 provides the economic values cor-
responding to the energy trading with other MGs based on its own forecasts of
day-ahead and real-time markets’ prices as well as the operation cost of its own
micro-turbines. It should be mentioned that the real-time and day-ahead electricity
markets’ prices are published after their closing times. Therefore, microgrids’
strategies are dependent on the price and load forecasts’ accuracy. In this case, study
decisions are made based on rolling forecasting of load and prices. In Fig. 8.4a, the
last updates forecasts of day-ahead (DA) and real-time (RT) markets’ prices and
purchased powers by MG4 are illustrated.
8 Real-Time Perspective in Distributed Robust Operation of Networked Microgrids 215

As shown in Fig. 8.4a, at time slots between t = 1 until t = 9, the prices of RT

or DA are less than the generation cost of micro-turbines (see. Table 8.1), MG4
purchase power from the upper grid by participating in DA and RT markets. At time
slots t = 3 until t = 5, the price of DA was less than the predicted RT market’s price,
therefore, MG4 purchases power from the DA. In real-time decision making fashion,
microgrids similar to MG4 submit selling offers and buying bids to the TPEE.
It should be noted that all the accepted transactions are double-checked by the
corresponding sellers and buyers. The double-checking stage validates deliverable
transactions and prevents technical challenges such as congestion, overvoltage, and
stability problems in microgrids. For our test system, all the feasible transactions
which are approved by TPEE and corresponding seller and buyer are shown in
Fig.8.4b. At each time slot, the number of transactions and buyers’ bids are depicted
in Fig. 8.4b. As can be concluded from Fig. 8.4b, buyers with maximum bids are
selected by TPEE as approved buyers. Since MG4’s offered selling price is lower
than other MGs, MG4 is selected as the seller over the operating horizon. At time
slots between t = 1 and t = 9, the price of electricity markets is low. Therefore,
MG4 benefits from arbitrage of buying power from the DA or RT and selling to the
microgrids. MG4 sells to all the microgrids at time slots t 2 [1; 9]. Because MG4
by buying from markets at a low price, it obtains more capacity respect to time slots
t = 2 [10; 24]. By comparing exchanged power at time slot t = 23 and t = 24, it can
figure out that buyers are selected according to their bids. At time slot t = 23, MG4
sells to MG1 but at time slot t = 24, MG4 sells to MG3.
The real-time scheduling of energy resources within the entities are presented
in Fig. 8.5. Operation schedules of micro-turbines within the MG1, MG2, MG3,
and MG4 are illustrated in Figs. 8.5a–d, respectively. Since the generation costs of
MG1–MG3 is expensive than the purchasing cost from MG4, MG1–MG3 prefers
to buy from MG4. On the other hand, MG4 by comparing the price of DA and RT
markets with selling prices just sells the production of its own micro-turbines at time
slots t ∈ [10, 24].

Fig. 8.5 Output power of micro-turbines (a) MG1, (b) MG 2, (c) MG 3, and (d) MG 4
216 M. Jalali et al.

8.4 Conclusion

In this chapter, an energy trading framework for a networked microgrid framework

is presented. In this framework, a third-part economic entity by collecting selling
and buying economic values determines deliverable and feasible transactions and
corresponding sellers and buyers. Furthermore, physical exchanging power between
entities is tackled by the entities’ operators. This aim is achieved by implementing
ADMM as a distributed optimization technique. Since the exchange power between
entities is determined by TPEE, entities should satisfy the equality constraints
of voltage magnitude and voltage angle at the connection points. Therefore,
these complicated constraints are relaxed by implementing ADMM. Moreover, the
performance of the presented method is studied on the real-time operation of a test
system by using updated forecasts of electricity markets’ prices as well as electricity
loads. The results show that microgrids select operating their resources when the
buying price is high or when their suggested quantity is not available in the market.
Moreover, MGs by using updated forecasts of day-ahead and real-time market, at
low prices buys from electricity markets and sells to the microgrids in the market.

Nomenclature

Sets

MT Set of micro-turbine

MT
i Set of micro-turbines within the MG i
U,i
n Indicator of connection node n to the upper grid in MG i
MT
n
,i
Set of connected micro-turbines into bus n within the MG i
Bus,i
n Sec of connected bussed to bus n within the MG i
MG
i Set of connected MGs to MG i

Indexes

t, k Time slots
i, j Microgrids
n, m Buses
g Micro-turbines
κ Iteration
8 Real-Time Perspective in Distributed Robust Operation of Networked Microgrids 217

Parameters

λ̂ Markets’ price
P̂ d Electricity load demand
aMT , bMT Micro-turbines cost coefficients
SUMT , SDMT Micro-turbines start-up and shut-down costs
UTMT , DTMT Micro-turbines minimum up and downtimes
PMin , PMax Micro-turbines minimum and maximum active power capacity
QMin , QMax Micro-turbines minimum and maximum reactive power capacity
G, B Conductance and susceptance components of the admittance matrix
ξ Auxiliary parameter in linearized aperient power flow
SLMax Maximum aperient power flow
VMax Maximum voltage magnitude

Variables

PDA ,PRT Purchased power from DA and RT markets

PMT ,QMT Active and reactive generated power of micro-turbine
SUCMT ,SDCMT Realized Start-up and shut-down cost
δ MT Commitment status of micro-turbine
PInj ,QInj Injected active and reactive power
PL,QL,SL Active, reactive, and aperient power flow in lines
V,θ Magnitude and angle of voltage

References

1. Wang, Z., Chen, B., Wang, J., Begovic, M. M., & Chen, C. (2014). Coordinated energy
management of networked microgrids in distribution systems. IEEE Transactions on Smart
Grid, 6(1), 45–53.
2. Qiu, J., Meng, K., Zheng, Y., & Dong, Z. Y. (2017). Optimal scheduling of distributed
energy resources as a virtual power plant in a transactive energy framework. IET Generation,
Transmission & Distribution, 11(13), 3417–3427.
3. Sajjadi, Sayyid Mohssen, Paras Mandal, Tzu-Liang Bill Tseng, and Miguel Velez-Reyes.
“Transactive energy market in distribution systems: A case study of energy trading between
transactive nodes.” In 2016 North American Power Symposium (NAPS), pp. 1–6. IEEE, 2016.
4. Akter, Most N., Md Apel Mahmud, and Amanullah MT Oo. “An optimal distributed transactive
energy sharing approach for residential microgrids.” In 2017 IEEE Power & Energy Society
General Meeting, pp. 1–5. IEEE, 2017.
5. Jalali, M., Zare, K., & Seyedi, H. (2017). Strategic decision-making of distribution net-
work operator with multi-microgrids considering demand response program. Energy, 141,
1059–1071.
6. Gao, H., Liu, J., Wang, L., & Wei, Z. (2017). Decentralized energy management for networked
microgrids in future distribution systems. IEEE Transactions on Power Systems, 33(4), 3599–
3610.
7. Boyd, S., Parikh, N., Chu, E., Peleato, B., & Eckstein, J. (2011). Distributed optimization and
statistical learning via the alternating direction method of multipliers. Foundations and Trends®
in Machine learning, 3(1), 1–122.
218 M. Jalali et al.

8. Chitsaz, H., Zamani-Dehkordi, P., Zareipour, H., & Parikh, P. P. (2017). Electricity price
forecasting for operational scheduling of behind-the-meter storage systems. IEEE Transactions
on Smart Grid, 9(6), 6612–6622.
9. Chitsaz, H., Shaker, H., Zareipour, H., Wood, D., & Amjady, N. (2015). Short-term electricity
load forecasting of buildings in microgrids. Energy and Buildings, 99, 50–60.
10. Amjady, N., & Keynia, F. (2008). Day-ahead price forecasting of electricity markets by
mutual information technique and cascaded neuro-evolutionary algorithm. IEEE Transactions
on Power Systems, 24(1), 306–318.
11. Historic price data,“ http://www.energyonline.com/Data/GenericData.aspx.
Chapter 9
Application of Heuristic Techniques
and Evolutionary Algorithms
in Microgrids Optimization Problems

Amir Aminzadeh Ghavifekr

9.1 Introduction

Microgrids can efficiently solve numerous problems of the traditional power

systems and provide better management of peak loads, acceptable reliability, and
feeder losses reduction [1]. Two primary goals of each commercial microgrid are
customer satisfaction and cost-efficiency. Besides all environmental and economic
benefits, there are numerous technical challenges in energy management, control,
and protection of microgrids that cause designing a cost-effective microgrid system
considered as a complicated problem. Every planning process suffers from social,
regulatory, geographical, and environmental uncertainties that can be external such
as lack of knowledge and the nature of the environment or be internal, which is
mostly accrued in the process of identification. Thus, the planning process for
microgrids is commonly based on the trade-off in solution searching and can be
considered as a multi-objective problem. Despite all achievements of the microgrids,
designing a cost-effective structure is a complicated problem due to the different
parameters that should be taken into account at any decision level.
Different optimization techniques have been presented in the literature to solve
microgrids optimization problems. These methods can be compared based on
criteria such as calculation complexity, ability to consider predictions, model
dependency, and flexibility concerning microgrid expansion. According to the
intrinsic features of the optimization problem, each of these algorithms can offer
a successful solution.
The classical optimization approaches, such as the numerical and analytical
methods have been applied to several optimization problems of microgrids. These

A. A. Ghavifekr ()
Department of Control Engineering, University of Tabriz, Tabriz, Iran
e-mail: aa.ghavifekr@tabrizu.ac.ir

© Springer Nature Switzerland AG 2021 219

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_9
220 A. A. Ghavifekr

methods provide desirable results when the accurate starting point is set in advance.
They provide fast convergence to the best local optimum. However, it is very
challenging to escape from the local optimum solution in these algorithms, and the
selection of the appropriate starting point is not straightforward. The accurate esti-
mation of this point needs intuition, experience, and lots of theoretical knowledge.
Heuristic and metaheuristic algorithms that have been established since the 1980s
are nature-inspired methods that include trial and error solutions to find strategies
for complex problems. These algorithms refer to calculating the minimum and
maximum of a function with systematic inputs chosen through an initial set. It
cannot be proved that these methods always find the global optimum, but they can
provide sufficiently good results near to the global minimum. These algorithms
are best suited for problems with unknown starting points and large optimization
space. Although these methods have slower convergence, they can search whole
optimization space, which mostly cannot be reached by the classical algorithms.
These data-driven methods can cover different challenges of this area, including
the sizing and management optimization, predictive maintenance, estimation of
exploitable energy, a real-time self-tuning system, siting, operation scheduling, and
a variety of other applications. The extension of these results in home microgrids (H-
MGs), besides improvement in energy performance, can lead to the minimization
of the product cost and market clearing price (MCP). Also, the optimization cost
function can be formulated for system losses considering the balance of the load
and generation.
Numerous software tools have been developed for a predefined optimization
problem both in real time and also offline. These software tools can model a
wide range of nonlinear complicated optimization problems. Also, they can provide
precise simulations due to the flexibility in the definition of time intervals. However,
these tools still need to be updated for promoted usage in stability analysis and
energy management problems. In [2], a detailed discussion and comparison of
available software tools for analyzing various energy systems have been presented.
Homer is one of these software tools, which was developed by NREL. It can model
the hybrid systems and is designed for a large range of generators, loads, and
converters. GAMS is another modeling software that can be utilized for a wide range
of linear and nonlinear optimization problems. In HYBRID2, which was developed
by the RERL, the simulation is very accurate due to the variable time intervals. One
of the most downloaded optimization tools is RETSCREEN, which is based on MS
Excel. It provides a platform to compare the base case models of classical methods
with the financial and technical viability of renewable projects.
This chapter will focus on recent progress in the application of computational
intelligence and heuristic techniques in microgrids. However, regarding the prob-
abilistic nature of these algorithms, they present variable performance in solving
different optimization problems. This is the main motivation of presenting this
chapter to give a general view of the recent achievements in solving microgrid
optimization problems with evolutionary algorithms (EA). First, a brief review
9 Application of Heuristic Techniques and Evolutionary Algorithms. . . 221

of some recent and promising EAs is presented. In the rest of this chapter,
the applications of these algorithms in energy management systems, operation
scheduling, voltage and frequency control, and sizing optimization problems are
discussed.

9.2 Brief Introduction to Evolutionary Algorithms

There are many types of EAs that have been progressed based on the behavior
of nature and can be either trajectory-based or population-based algorithms. This
section gives an insight into the structure and basis of some EA algorithms which
have been applied for microgrid problem optimizations. Some of the well-known
EAs and their related techniques are introduced in this section. These methods try
to explore the best available answer among a broad set of feasible solutions for
objective functions with practical constraints.
Genetic Algorithm (GA), Ant Colony Optimization (ACO), Particle Swarm Opti-
mization (PSO), Biogeography Based Optimization (BBO), Harmony Search Algo-
rithm (HSA), Cuckoo Search Algorithm (CSA), Artificial Bee Colonies (ABC),
Grey Wolf Optimization (GWO), Firefly Algorithm (FA), and Grasshopper opti-
mization algorithm (GOA) are among popular widely used heuristics methods that
will be discussed in this chapter. These methods are employed to solve different
optimization problems and energy planning issues in microgrids literature such
as the optimal partition of power grids, replacement, maintenance, and monthly
income by selling power to the grid. Using these methods leads to a reliable network
partitioning with less CPU effort and save the operating costs of the distributed
nodes and interruption costs. Also, they provide more mobility to add further
restrictions to the issues mentioned earlier, such as sizing and scheduling of power-
generation sources.

9.2.1 Genetic Algorithm (GA)

Genetic Algorithm (GA) is one of the well-known and effective techniques of the
EAs and is inspired by Darwin’s theorem of survival of the best. The principles of
the genetic and nature’s evolution structure have been utilized in this algorithm to
eject the weak population and let only the high-quality ones survive the process.
The initial population should be considered as chromosomes, and the operators of
the algorithm (Crossover and Mutation) should be applied several times to meet one
of the convergence criteria. GA can escape from the local minimum; however, its
complexity increases with the number of parameters. Simple pseudo code for this
algorithm is presented in Fig. 9.1
This algorithm can be used for nonlinear problems with numerous discrete
variables such as microgrids reconfiguration [3], which satisfies the operational
222 A. A. Ghavifekr

Fig. 9.1 Pseudo code of GA

constraints and considers load priorities. Also, this algorithm has been proposed
to obtain favorable solutions for operation scheduling and energy management
problems [4]. The energy management system should solve the optimization
problem with the classification of all the necessary information and resend it to each
distributed energy resources (DERs). This concept can be extended to the optimal
power flow (OPF) problems in a local microgrid setup [5].
Microgrids include several distributed generators (DGs), which are mostly
utilized to supply the local loads. GA covers the sizing optimization problem of
microgrids. This technique has been applied for hybrid AC–DC microgrids [6],
battery energy storage system (BESS) optimal sizing [7], optimization of storage
devices [8], combined heat and power (CHP) microgrid systems [9], determination
of optimal sizing of combined wind and gas generators [10], and capacitor place-
ment and sizing [11]. Although distribution system protection has technical aspects
in microgrids, their insertion leads to violating the relay coordination. GA can be
used for optimization problems of the distribution systems to optimize the operating
time, optimal DG placement, and maximize the penetration level of DGs [12]. Also,
the distribution network can be partitioned to minimize the energy exchange via the
microgrids using GA-based methods [13, 14].
Utilizing decentralized controllers provides more reliability and robustness of
the microgrid systems. In the case of any failure in one of the controllers, it can be
ensured that the frequency regulation can be achieved via others. GA can be used
for optimizing the parameters of these controllers. Such an application can also be
extended to voltage and load frequency control [15]. In some studies, more than one
optimization problem have been taken into account to define an objective function
and solve it with GA [16].
9 Application of Heuristic Techniques and Evolutionary Algorithms. . . 223

9.2.2 Particle Swarm Optimization Algorithm (PSO)

Particle Swarm Optimization (PSO) is one of the oldest intelligent optimization

algorithms. The algorithm has been introduced in 1995 [17] and is inspired by the
social behavior of animals such as a school of fish or flock of birds living in small
and large groups. In the PSO algorithm, the population members interact straightly
with each other by transforming data and recalling valuable experiences of the past.
The PSO algorithm is suitable for a variety of continuous and discrete problems and
provides efficient solutions to various optimization problems.
Since the PSO starts with a random initial population, it is similar to many other
EAs, such as the continuous GA. Unlike the GA, PSO has no evolutionary operator,
such as mutation and crossover. Each population member is called a particle (which
is similar to the chromosome in GA). These particles randomly take an initial
value, and position and velocity are defined for each of them. The dimension of
the problem is defined according to the number of parameters. Two memories are
dedicated to registering the best position of each particle in the past, and the best
position for all particles. Velocities and positions of particles are updated in each
iteration. It is deduced that PSO has faster and better results compared with the
other classical methods. The flowchart of this algorithm has been depicted in Fig.
9.2.
PSO has been utilized in optimal energy management and operation scheduling
of microgrids to overcome constraints such as environmental aspects [18, 19]. In
[20], a comprehensive study has been done on the usage of PSO in choosing the
storage characteristics and voltage regulation. PSO can be used for optimization
problems on the battery storage systems and distributed battery systems (DBS)
to compensate imbalanced active and reactive power flows [21, 22]. Also, this
algorithm can be applied for tuning of controller parameters, which are designed to
control the reactive power flow between the main grid and the microgrids [23]. PSO
can be utilized to find the optimal coefficients for controllers to enhance voltage
unbalance factors (VUFs) [24]. Also, it covers the sizing problems and optimal
placement of DG units in microgrids [25] as well as the siting and optimization
of power-sharing Schemes [26].

9.2.3 Ant Colony Optimization (ACO)

Ant Colony Optimization (ACO), which is introduced in [27], is inspired by the

social behavior of ants. The ants work together to find the shortest path between the
nest and the food source so that they can transport food to the nest in the shortest
time. The ants communicate with each other by laying down pheromone trails. If an
ant follows a path, it can identify the path of its return by sensing the pheromone it
has placed on the ground. Other ants can find such a path, and instead of traveling
at random, follow the trail. These pheromones do not always stay on the ground
224 A. A. Ghavifekr

Fig. 9.2 General flowchart of PSO

Fig. 9.3 Pseudo code of

ACO

and evaporate slowly after a while. Thus, two main operators in this algorithm are
pheromone placing and pheromone evaporation. Applying these operators helps to
avoid convergence to the locally optimal solutions. The simple pseudo code for this
algorithm is presented in Fig. 9.3.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . . 225

Fig. 9.4 Pseudo code of select H i with probability αλi if H i selected

habitat migration in BBO
for k =1: N
if rndreal(0,1)< λi
select H j with probability αμ i
if H j is selected
Randomly select an SIV σ from H j
Replace a random SIV in H i with σ
end if
end if
end for
end if

The voltage deviations could be minimized through appropriate control methods

whose parameters could be optimally tuned by ACO [28]. Also, it can provide a
rapid microgrid power management system, including numerous constraints and
objectives such as economic, environmental, and fuel availability considerations
[29, 30].

9.2.4 Biogeography Based Optimization (BBO)

Biogeography based optimization (BBO) was introduced in [31] and inspired by

how species are distributed in multiple habitats. This algorithm includes speciation,
the migration of species between habitats, and the extinction of species. The Habitat
suitability index (HIS) is defined for each island that is affected by parameters such
as vegetative diversity, rainfall, topographic diversity, temperature, and so on. The
determiner features are called suitability index variables (SIVs) that, unlike HIS
variables, are independent parameters. Islands that have high HIS can support many
species, unlike low HIS islands. In these islands, many species try to emigrate
to nearby habitats due to the large population that causes much competition for
recourses. These islands have high emigration and low immigration rates. If N
denotes the dimension of the search space, the semi-pseudo code for habitat
migration is depicted in Fig. 9.4.
This algorithm has been used for designing a Linear Quadratic Regulator (LQR)
and minimizing the frequency excursion following a disturbance in microgrids [32,
33].
226 A. A. Ghavifekr

9.2.5 Harmony Search Algorithm (HSA)

Harmony is one of the main components of music that is added to enhance the
content of music, and its proper design requires knowledge of the principles of (at
least empirically) harmony science. Even ordinary music audiences, who do not
listen to music professionally, will notice a deficiency in music if the harmony
is removed from many pieces of music. Many famous musicians are well known
for designing powerful chords and have spent a great deal of time creating and
discovering the right harmonies. From the modeling and simulation process that a
composer goes through to harmonize a piece of music, an algorithm is extracted,
known as Harmony Search (HS) [34]. This algorithm explores all feasible solutions
to reach a new one; unlike the genetic algorithm that only explores two-generation
vectors. This feature makes the algorithm more flexible to search for various
solution spaces. The pseudo code of this algorithm is illustrated in Fig. 9.5.
From the application point of view, this algorithm has also been utilized to
optimize the droop control coefficients such that the current sharing error and error
due to the voltage drop are minimized [35]. HSA can find the optimum locations
of energy storage systems (ESS) or distributed energy recourses (DERs) [36]. Also,
it can be utilized for optimal operation of the biomass, solar, and geothermal units
considering the minimum functional cost of the system [37], and for a day-ahead
scheduling model for the optimal operations of microgrids [38]. HSA can also
be used to cover the multi-objective sizing optimization of microgrids [39] or to
optimally adjust an islanded microgrid performance in terms of frequency and
voltage regulation [40].

Fig. 9.5 Pseudo code of HSA

9 Application of Heuristic Techniques and Evolutionary Algorithms. . . 227

9.2.6 Cuckoo Search Algorithm (CSA)

The cuckoo search algorithm (CSA) is inspired by the behavior of a particular

species of bird called cuckoo. Specifically, the unique features of this bird in
laying and reproduction have been used to implement the cuckoo optimization
algorithm [41]. Each cuckoo has a habitat for life and lays eggs in their habitat.
Cuckoo migrates to areas with the most food and best living conditions for
reproduction and lays eggs around the optimal habitats. This procedure allows
for more environmentally optimal habitat to be searched for an optimal global
response. After several migration cycles, most cuckoo populations converge to the
optimal solution of the optimization problem. The evaluation results show that this
algorithm has high speed and accuracy in convergence to the optimal solution of the
benchmark functions. Even where the target function has a large number of locally
optimal solutions, this algorithm is able to converge to a near approximation of
the optimal global solution in a few iterations. The cuckoo optimization algorithm
can be considered as one of the successful implementations of the nature-inspired
process. One of the advantages of this algorithm is finding an accurate value in fewer
iterations. The pseudo code of this algorithm is as follows (Fig. 9.6):
This algorithm can be used for solving the multimodal optimization problems in
microgrid systems and operation scheduling optimization [42, 43]. CSA can also
be developed to optimize the performance of the BESS to mitigate the voltage
fluctuation in microgrids [44]. It can be utilized to minimize the emission and
generation costs of the microgrids while meeting system constraints and its hourly
demands [45] as well as solving nonlinear sizing problems [46].

Fig. 9.6 Pseudo code of CSA

228 A. A. Ghavifekr

9.2.7 Artificial Bee Colonies (ABC)

Bees are insects that live in relatively large colonies. Besides the benefits of bees
in agriculture and horticulture, the regular social behavior of these organisms has
always been the source of inspiration for scientific studies. The mathematical model
for this algorithm is proposed in [47] based on the intelligent foraging behavior of
bee swarms.
In a bee colony, adopted by the ABC algorithm, there are three types of bees:
employed bees, onlooker bees, and scouts. Half of the colonies contain employed
bees, and the other half contains onlooker bees. Employed bees are responsible for
exploiting previously discovered food sources, as well as providing data to other
onlooker bees in the hive about the quality of the food they are extracting. Onlooker
bees stay in the hive and, according to the information shared by the employed bees,
decide on a food source to explore. Scouts randomly search the environment to find
a new food source based on intrinsic motivation or external evidence. The general
flowchart of this algorithm is presented in Fig. 9.7.
This algorithm can also be used for solving a variety of optimization problems
in microgrids ranging from economic dispatch (ED) and network reconfiguration
of microgrids related to planning and sizing problems [48, 49]. In some literature,

Fig. 9.7 General flowchart

of ABC
9 Application of Heuristic Techniques and Evolutionary Algorithms. . . 229

it has also been utilized to minimize the market cleaning price (MCP), production
cost, and provide better utilization of renewable energy resources [50].

9.2.8 Grey Wolf Optimization (GWO)

Gray wolves are animals that live in a semi-democracy environment, and each wolf
has its position in society. Gray wolves live and hunt collectively. In each group of
gray wolves, there are, on average, between five and 12 wolves. The gray wolves first
surround the prey and begin to tame it by narrowing the siege, then according to the
order of the leader wolf attack the victim and eventually kill it. The aforementioned
process is mathematically modeled in [51]. In implementing the GWO algorithm,
four types of gray wolves, such as Alpha, Beta, Delta, and Omega, have been used
to model a hierarchical gray wolf algorithm, which includes three steps of the search
for prey, prey siege, and bait attacks.
Alpha pairs, known as group leaders, make decisions about hunting, sleeping
time, waking up, and so on. Alpha decisions apply to the whole group. The second
class belongs to the Beta wolves, which help alpha wolves in decision making and
other group activities. These wolves are the best candidate for substituting when the
alphas are very old or dying. The lowest ranked wolves are Omega wolves, which
play the role of scapegoat. They must obey all other wolves and are the last groups
that are allowed to eat. Wolves not mentioned in the aforementioned groups are
called Delta wolves. Delta wolves are more lower order than alpha and beta but
superior to omega. The flowchart of this algorithm is depicted in Fig. 9.8.
GWO has been frequently used in the literature to handle different optimization
tasks in microgrids. Special attention has been made in its application to determine
the optimal size of battery energy storage (BES) considering various constraints
such as power and energy capacity [52]. Also, GWO has been used to reduce
the operating time of directional overcurrent relays (DOCRs) and to do energy
management scenarios [53]. This algorithm has also been utilized for the distributed
hierarchical control of microgrids to solve the dispatching problem [54].

9.2.9 Firefly Algorithm (FA)

The Firefly Algorithm (FA) has been introduced in 2008 [55]. The main idea is the
flashing communication between fireflies, where each of them can be characterized
by their flashing light produced by bioluminescence or biochemical structure.
This population-based algorithm can be considered as Swarm Intelligence, where
higher levels of intelligence are created by the cooperation and competition of
the simple and the less intelligent members, which is certainly not obtainable by
any of the components. The flashing light in this algorithm is the main concept
for mating, and the degree of the attractiveness of each firefly is proportional to
230 A. A. Ghavifekr

Fig. 9.8 General flowchart

of GWO

the related brightness and intensity. This persuades less bright fireflies to move
toward the brighter one. Thus, it can be deduced that more brightness causes less
distance between fireflies. This brightness consists of the objective function, and the
evolutionary process is performed to optimize it. The simple pseudo code for this
algorithm is presented in Fig. 9.9.
This algorithm can be used for energy management and operation scheduling of
microgrids, such as determining the optimal power output of each generator at min-
imum cost [56, 57]. Also, it can be utilized for tuning of controller parameters in a
decentralized control scheme to mitigate varying load perturbations and improve the
frequency and voltage stability of microgrids [15, 58, 59]. FA has been applied for
calculation of the payoff function of each coalition group between microgrids and
enables them to maximize their utilities [60]. It presents an acceptable performance
in energy storage system management to optimize operation systems [61].

9.2.10 Grasshopper Optimization Algorithm (GOA)

This optimization algorithm was presented in 2017 [62]. Grasshoppers have usually
a destructive role in nature and can cause damage to the agricultural produces and
crop productions. The lifecycle of the full-grown adult grasshopper is depicted
in Fig. 9.10. The algorithm is inspired by the food searching behaviors of the
grasshoppers and their reactions to the surrendering environment. In this algorithm,
updating the location of each grasshopper depends on the distance from the entire
9 Application of Heuristic Techniques and Evolutionary Algorithms. . . 231

Fig. 9.9 Pseudo code of FA

Fig. 9.10 The lifecycle of

the full-grown adult
grasshopper

population in the current generation, and the location of the best grasshopper.
Besides simplicity in implementation, the main feature of this algorithm consists
of only one parameter to adjust. In [62], the swarm behavior of these insects is
modeled mathematically for solving a variety of optimization problems.
The pseudo code of this algorithm is presented in Fig. 9.11.
This algorithm can be used for tuning of controller parameters for load frequency
control of interconnected microgrid power systems [63, 64]. Also, it is able to size
the autonomous microgrid systems, optimally [65].
232 A. A. Ghavifekr

Fig. 9.11 Pseudo code of GOA

9.2.11 Whale Optimization Algorithm (WOA)

This algorithm is inspired by the hunting mechanism of humpback whales in nature

[66]. These kinds of whales can recognize the position of prey and surrendered them.
Since the optimal design position in the search space is not known, the algorithm
considers the current best candidate solution as the target or a close solution to the
optimum. After the definition of the best search agent, the other agents will try
to update their positions regarding the best search agent. The pseudo code of this
algorithm is depicted in Fig. 9.12.
This algorithm has been utilized for the optimal tuning of PI controllers in an
autonomous microgrid system. This can enhance the flow of active and reactive
power during load variation [67]. WOA presents an acceptable performance in
solving the combined economic emission dispatch problem that is one of the well-
known problems of the energy management systems [68].
Table 9.1 presents a timeline of the development of the aforementioned nature-
based evolutionary algorithms.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . . 233

Fig. 9.12 Pseudo code of WOA

Table 9.1 EA optimization techniques development timeline

Method Developer Year
Genetic algorithm (GA) [69] John. R. Koza 1988
Particle swarm optimization (PSO) [17] James Kennedy and Russel. C. Earhart 1995
Harmony search algorithm (HAS) [34] Z. W. Geem, J. H. Kim, and G. V. 2001
Loganathan
Ant colony optimization (ACO) [27] M. Dorigo, M. Birattari, and T. Stutzle 2006
Biogeography-based optimization D. Simon 2008
(BBO) [31]
Firefly algorithm (FA) [55] X.-S. Yang 2008
Artificial bee colonies (ABC) [47] D. Karaboga 2010
Cuckoo search algorithm (CSA) [41] R. Rajabioun 2011
Grey wolf optimization (GWO) [51] S. Mirjalili, S. M. Mirjalili, and A. Lewis 2014
Whale optimization algorithm (WOA) S. Mirjalili and A. Lewis 2016
[66]
Grasshopper optimization algorithm S. Saremi, S. Mirjalili, and A. Lewis 2017
(GOA) [62]

9.3 Illustrative Examples on Application of EAs

in Microgrids

9.3.1 Energy Management and Operation Scheduling

EA methods play a significant role in the scheduling of microgrids to obtain

better quality power at minimum cost. Since the past decade, EA methods have
234 A. A. Ghavifekr

been employed extensively in unit commitment, economic dispatch, and generation

scheduling problems [70].
A genetic algorithm has been proposed to obtain favorable solutions for operation
scheduling and energy management problems [71]. The energy management system
should solve the optimization problem with the classification of all the necessary
information and resend it to each distributed energy resources (DERs). This concept
can be extended to the optimal power flow (OPF) problems in a local microgrid
setup [72]. PSO has been utilized in optimal energy management and operation
scheduling of microgrids to overcome the constraints such as environmental aspects
[73]. ACO can provide a rapid microgrid power management system, including
numerous constraints and objectives such as economic, environmental, and fuel
availability considerations [29, 30]. HSA can be utilized for optimal operation of
the biomass, solar, and geothermal units considering the minimum functional cost
of the system [37]. In [38], the cost function is defined as the sum of the total
generation and operation costs of PV arrays, WT, battery, and DGs. This is such a
nonlinear mixed-integer programming problem for smart microgrids, which is hard
to solve with conventional methods. HSA with modified mutation and selection
operators and adaptive parameters is utilized to solve this optimization problem.
It can be concluded that for optimal day-ahead scheduling of microgrids, HSA
is more reliable under both fault and normal operation situations. CSA can be
used for solving the multimodal optimization problems in microgrid systems and
operation scheduling optimization [74, 75]. In [76], the improved version of CSA
is applied on a microgrid, including numerous renewable and conventional energy
power plants such as two wind powers, two diesels, and three fuel-cell plants. The
system is presumed to be isolated from the electric network, and the local generation
cost optimization problem is solved with CSA. Also, in [74], CSA is utilized for
different scenarios, including MG with all sources, all sources without wind energy,
all sources without solar energy, and all sources without solar and wind energy.
ABC can be utilized to minimize the market cleaning price (MCP), production
cost, and providing better utilization of renewable energy resources [77]. The main
objective in [78] is providing optimal scheduling for a real system of 24-hours
load demand with the least operation cost of a microgrid, which includes PV
array, wind turbine, fuel cell, diesel engine, and microturbine. Applying ABC for
solving the proposed optimization problem demonstrates that the minimum cost
is obtained when the diesel engine shares less power in comparison with the fuel
cell, and microturbine. GWO can help to reduce the operating time of directional
overcurrent relays (DOCRs) and other energy management problems [79]. In [80],
a different cost function has been defined that leads to optimization of the total
grid loss, minimization of the system pollution and voltage deviation, and reduction
of the total energy costs in the operation stage. Thus, the multi-objective grey wolf
algorithm has been utilized for solving this complex problem. It can be extracted that
the integration of storage units and DGs in microgrids has been provided economic
and environmental advantages besides power loss reduction and voltage deviation
improvement. The improved GWO method helps to accurate operation modeling of
DGs and storage units.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . . 235

FA can be used for energy management and operation scheduling of microgrids,

such as determining the optimal power output of each generator at minimum cost
[81]. These studies demonstrate that FA can achieve the lowest operation cost in
comparison with all the aforementioned algorithms [56]. Assume that the MG
system included various DG units such as FC, WT, PV, and MT. All units are
operated in the unity power factor. The main goal is the reduction of operating costs
through the optimal operation of a variety of DG units. The objective function can
be defined as:


T
Min F = ft + OM DG , t = 1, 2, . . . , T (9.1)
t=0

where

ft = Cgrid,t + CDG,t + SU C F C,t + SU C MT ,t + SDC MT ,t + SDC F C,t

(9.2)

and Cgrid, t can be defined as:

⎧
⎨ Pgrid,t Bgrid,t if Pgrid,t > 0
Cgrid,t = (1 − tax) Pgrid,t Bgrid,t if Pgrid,t < 0 (9.3)
⎩
0 if Pgrid,t = 0

Also, we have
 
SU C F C,t = max 0, uF C,t − uF C,t−1 ∗ SU F C
  (9.4)
SU C MT ,t = max 0, uMT ,t − uMT ,t−1 ∗ SU MT

where OMDG is the total maintenance, and operation cost of DG units connected to
the microgrid, CDG, t is the cost of operating power, and fuel of Dg units Cgrid, t is
the grid cost at time t, SUCFC, t and SUCMT, t are the start-up costs of FC and MT at
time t, respectively. SDCFC, t and SDCMT, t are the shut-down costs of FC and MT at
time t, respectively. Pgrid, t is the power of the grid at time t. OMDG can be calculated
as

OM DG = T ∗ (OM F C + OM MT + OM W T + OM P V ) (9.5)

where OMFC , OMMT , OMWT and OMPV are operation costs of FC, MT, WT, and PV,
respectively. T denotes the total time in hours.
Now some constraints should be considered for this optimization problem.
1. The load demand balance constraints should be considered as:
236 A. A. Ghavifekr

Table 9.2 Power limitations and bids of grid and DG units

Type Min Power (kW) Max Power (kW) Bid ($/kW) OM ($/kW) SUC/SDC ($/kW)
FC 3 30 0.326 0.095 1.833
PV 0 25 2.87 0.231 0
WT 0 25 1.19 0.583 0
MT 6 30 0.507 0.049 1.06
GRID −30 30 – – –

PD,t = PF C,t uF C,t + PMT ,t uMT ,t + PP V ,t uP V ,t + Pgrid,t ugrid,t + PW T ,t uW T ,t

(9.6)

where Pφ and uφ demonstrate the power output of φ and the operating status
at time t, respectively.
2. Constraint of the output power of DG units (PDG, t ) is formulated as

PDG,min ≤ PDG,t ≤ PDG,max (9.7)

where PDG, max and PDG, min are maximum and minimum power limits of DG
units, respectively.
3. Constraint of grid output power (PDG , t) can be formulated as:

Pgrid,min ≤ Pgrid,t ≤ Pgrid,max (9.8)

where Pgrid, max and Pgrid, min are the maximum and minimum power limits of
the utility grid.
The optimization problem has been defined for a typical microgrid system [56]
which is connected with other DG units such as FC, PV, WT, and MT. The minimum
and maximum power limits, maintenance and operation cost, bidding cost, SDC, and
SUC for units are defined in the following Table 9.2.
Now the objective function is ready and the FA algorithm can be applied for
solving the optimization problem. The number of iterations is defined as 500. The
initial parameters of the algorithm are chosen as step size (α) = 0.5, attractiveness
(β) = 0.2, and absorption coefficient (γ ) = 1. In [56], the results will be compared
with other EA techniques to validate the priority of FA for the cost minimization
problem of microgrids. For PSO, the inertia weight is w = 1, and the acceleration
coefficients are chosen as C1 = 0.5 and C2 = 1.5. The intensity of attraction for
GOA is 0.5 and the minimum and maximum shrinking factors are 0.1 and 0.9,
respectively. For GA, the crossover rate, and mutation rate are selected as Pc = 0.9
and Pm = 0.01,respectively. Emigration and immigration rates for BBO are E = 1
and I = 1, respectively. For comparison purposes, the similar swarm size, number of
the habitats, and number of particles are chosen for the aforementioned algorithms.
The results are given in Table 9.3.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . . 237

Table 9.3 Comparative study of total operation cost in energy management problem
Methods Total operation cost ($) Iteration number of reaching to the min value
PSO 1075.576 420
GA 1157.623 480
BBO 935.412 325
FA 903.674 285
GOA 929.325 315

The total operation cost obtained by FA is compared with other EAs in Table
9.3 and demonstrates the better performance of the FA algorithm. If we increase the
number of population, the total generation cost will have more reduction. However,
the computation time is also growing.

9.3.2 Optimal Placement and Sizing of Energy-Related Devices

The sizing optimization approaches can provide the lowest investment with full use
of DG units. The type of suitable fuels for the power plant is an important problem
and has a critical impact on reliability and cost-efficiency. The microgrid consists
of a wind turbine (WT), photovoltaic (PV), diesel generator, and battery storage
system (BSS). The main objective is the optimal construction that can satisfy the
demand of the residential housing reliability based on the DPSP while minimizing
the cost of energy.
GA can be used for solving the optimal sizing problems of microgrids. This
technique has been applied for hybrid AC–DC microgrids [6], battery energy
storage system (BESS) optimal sizing [82, 83], optimization of storage devices
[8], combined heat and power (CHP) microgrid systems [9], determination of
optimal sizing of combined wind and gas generators [10], and capacitor placement
and sizing [11]. GA can be utilized for optimization problems of the distribution
systems to optimize the operating time, optimal DG placement, and maximize the
penetration level of DGs [12]. PSO can be used for optimization problems on
the battery storage systems and distributed battery systems (DBS) to compensate
imbalanced active and reactive power flows [84, 85]. Also, it can cover sizing
problems and optimal placement of DG units in a microgrid [86]. The approaches
based on HSA are presented for multi-objective sizing optimization of microgrids
in [87]. CSA is another evolutionary algorithm that can be developed to optimize
the performance of the BESS to mitigate the voltage fluctuation in microgrids [44].
It can be utilized for solving nonlinear sizing problems [46].
The problem of economic dispatch (ED) and network reconfiguration of micro-
grids can be solved by ABC [48, 49]. Recently, it is deduced that GWO produces
the most optimal solution for the BESS sizing problem and determines the optimal
size of battery energy storage (BES) considering various constraints, such as power
238 A. A. Ghavifekr

and energy capacity of BES [88, 89]. GOA can optimally size the autonomous
microgrid systems [65], and comparing the results with all of the aforementioned
papers demonstrates that GOA has better performance and very fast convergence as
well as a promising balance between exploration and exploitation. We will present
it with more details hereafter.
The objective functions for sizing problems can be defined based on the cost of
energy (COE) and the deficiency of power supply probability (DPSP). Optimization
of these parameters would guarantee minimum cost and reliable power supply. COE
(in $/kWh) is calculated as:


8760
COE = T NP C/ Pl (h) × CRF (9.9)
t=1

where TNPC is the total net present cost, includes replacement, capital, mainte-
nance, and operation cost. Pl (h)is the hourly consumption of energy and capital
recovery factor (CRF) is obtained by:

CRF = ri (1 + ro )n /(1 + ri )n − 1 (9.10)

where ri and ro present the real and optimal interest rates and n shows the system life
span. Another adjustable factor is DPSP that can be assumed as the reliability index
of microgrids that denotes the probability of the power supply leakage for energy
demand. It is formulated as:

DP SP = (Pl − PP V − PW T + PSOC + PDG ) /Pl (9.11)

where Pl , PPV , PWT , PSOC and PDG denote energy demand, PV power, the power
output of the wind turbine, power related to minimum battery state of charge,
and generated power by a diesel generator, respectively. The constraints of the
optimization problem can be defined based on the number of PV and WT as:

0 ≤ N P V ≤ Nmax
PV
(9.12)
0≤N W T ≤ Nmax
WT

P V and N W T show the maximum number of PV and WTs, and assumed to

where Nmax max
be 45 and 10, respectively. Also, we have three autonomy days.
Now, the optimization problem can be restated in its traditional form, which can
be generally formulated as:

min U (x) = [u1 (x), u2 (x), . . . . uk (x)]

(9.13)
subj ect to c(x) ≤ 0 and g(x) = 0

where, x is the vector of the search space, U(x)is a vector of objective functions, and
g(x) and c(x) present the equality and inequality constraints.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . . 239

Fig. 9.13 A comparison of algorithms used for the cost of energy reduction in microgrids [65]

GOA algorithm is employed to solve this optimization problem and is demon-

strated better performance when compared with other well-known methods such as
GA and CSA [65]. Like its counterpart methods, it starts with random particles in
the space of the landscape with user-defined limitations. The initial values of GOA
parameters are chosen as population size = 5, the number of iterations = 100, and
intensity of attraction = 0.5. The minimum and maximum shrinking factors are 0.1
and 0.9, respectively. For comparison purposes, the similar swarm size and number
of the nest are chosen for PSO and CS algorithms. Also, the maximum generation of
PSO equals the number of iterations of CS, and GOA. The abandoned eggs fraction
and smart eggs fraction for CS are Pa = 0.2 and Pc = 0.5, respectively. For PSO,
the inertia weight is w = 1,and the acceleration coefficients are chosen as C1 = 0.25
and C2 = 1.75.
The convergence characteristic of GOA, CSA, and PSO is depicted in Fig. 9.13,
and the optimal size of the system is presented in Table 9.4. It is noticeable that,
by applying GOA, the optimal configuration includes 26 PV panels and a 4 kW
WT. Furthermore, the COE is minimized to 0.365$/kWh which is less than its
counterparts, PSO, and CS algorithms. Zero DPSP means that energy demand is
fully satisfied.
The results demonstrate that GOA minimizes computer memory usage, and
reduces the computation time due to the fast convergence. Also, the CS algorithm
has better performance than PSO.
240 A. A. Ghavifekr

Table 9.4 Optimal sizing PSO CS GOA

results obtained by applying
EA [65] DPSP 0 0 0
Number of WT 7 6 4
Number of PV 30 29 26
BSS capacity (kW) 40 40 40
DG capacity (kW) 4 4 4
COE ($/kWh) 0.3674 0.3662 0.3656

9.3.3 Microgrid Optimal Voltage and Frequency Control

The interconnection of microgrids can be adjusted and controlled to provide them to

work in both islanded and grid-connected modes of operation. The control concern
in the grid-connected mode is to regulate the reactive and active power flow among
DGs connected within MG and between the main grid and MG. In the latter, the
frequency and voltage of the system are controlled by the power system, and there
is no control objective to optimize. But in the islanding mode, controlling the
frequency and voltage of the whole system is a crustal problem.
The control system must guarantee that there are not any considerable circulating
currents from the micro sources. These currents even can be produced by the small
mismatch in voltage and frequency set points. In order to provide smooth operation
of microgrids, an optimized control plan is mostly required for islanded operation.
Any uncertainties or inappropriate selection for the gain of controller leads to a large
variation in frequency and voltage levels in the operation mode. This architecture
includes a DC–DC boost converter, two PV panels, a three-phase voltage source
inverter (VSI), a coupling inductor, an RLC filter, and a three-phase load.
Utilizing decentralized controllers provides more reliability and robustness of
the microgrid systems. In the case of any failure in one of the controllers, it can be
ensured that the frequency regulation can be achieved via others. GA can be used for
optimizing the parameters of these controllers and can be extended to voltage and
load frequency control [90, 91]. PSO is another evolutionary algorithm that can be
used for tuning of controller parameters, which are designed to control the reactive
power flow between the main grid and the microgrids [23]. Using DGs leads to
active provide more flexible microgrids. PSO can be utilized to find the optimal
coefficients for controllers to enhance voltage unbalance factors (VUFs) [92, 93].
BBO [33], and HSA [40] have promising results in designing a Linear Quadratic
Regulator (LQR) and minimize the frequency excursion following a disturbance
in microgrids to improve the islanded microgrid frequency and voltage. FA [58,
59] and GOA [63, 64, 94] can be utilized for tuning of controller parameters in a
decentralized control scheme and load frequency control to mitigate varying load
perturbations and improve the frequency and voltage stability of interconnected
microgrid power systems.
The control circuit generates the controlled pulses for the VSI for the production
of pure sinusoidal voltage waves to provide high-quality power for loads. GOA
9 Application of Heuristic Techniques and Evolutionary Algorithms. . . 241

algorithm is utilized in this step to promote the dynamic output of the microgrid with
the efficient tuning of controller gains. As the proposed method in [95] minimize the
voltage controller corresponding errors, it is released from parameter optimization
for the current controller. However, using constant parameters for controller gains
cannot cause an optimal operation. Thus, proper tuning of the parameters can
enhance the power quality and performance of the system during the load changes
and disturbances. The results demonstrate that GOA provides a minimum output
current to optimize frequency and voltage overshoots. ITAE is selected as the
optimization index since it allows for the smoother implementation in comparison
with its counterparts such as an integral square error and absolute error. ITAE can
be defined as:
.∞
I T AE = t |e| dt (9.14)
0

where e(t) is the error signal that shows the difference between the controlled value
and the reference value. Using Eq. (9.14), the fitness function can be assumed as the
summation of the frequency and voltage error functions and formulated as:
⎛∞ ⎞
. .∞
 
Min ⎝ t × |ev | dt + t × ef  dt ⎠ (9.15)
0 0

Minimization of the aforementioned objective function leads to the optimal

selection of controller gains, which enhance the dynamic performance of the
islanded microgrid system. The GOA algorithm has been employed for parameter
tuning of PI controllers and its performance compared with PSO and WOA. The
numbers of iterations and particles are set to 50 for each of the algorithms to have a
reasonable comparison. The cognitive and social constants of the PSO are selected
as C1 = C2 = 2, and inertial weight is C = 0.5. The control parameters of WOA
are d = 3, and 0 < r < 1. Where d is the number of variables, and r is the random
number. For GOA, the intensity of attraction is 0.5, and the minimum and maximum
shrinking factors are 0.1 and 0.9, respectively.
The convergence behavior of these algorithms is depicted in Fig. 9.14.
Smaller final value and higher convergence rate represent the better performance
of the controller. The number of iterations and final optimized values for the
proposed objective function is given in Table 9.5.
As it is deduced from Table 9.5, the GOA algorithm provides a faster and more
accurate optimal solution in comparison with the PSO and WOA. The searching
process stops in the given criteria for the objective function or in the given iteration
numbers. The final values for the voltage controller gains (Kpv and Kiv ) and
frequency controller gains(Kpf and Kif ) after applying the GOA algorithm are given
in Table 9.6.
242 A. A. Ghavifekr

Fig. 9.14 The convergence plot of the EA methods [95]

Table 9.5 The minimum fitness function and convergence values of proposed EA algorithms [95]
EA method Min value of FF Iteration number of reaching to the min value
PSO 1.00145 21
WOA 0.87710 24
GOA 0.49635 16

Table 9.6 Final values of EA method Kpv Kiv Kpf Kif

controller gains [95]
PSO 0.25 25.64 0.09 23.64
WOA 0.94 1.60 0.04 26.74
GOA 17.09 27.87 0.08 12.26

The voltage and frequency regulation comparison between the EA algorithms are
depicted in Fig. 9.15. The aforementioned designed controller restores the nominal
voltage and frequency of the system as soon as the voltage falls.
GOA based controller provides the frequency (50 Hz) and nominal voltage
(375 V) as well as minimum overshoot and settling time and can keep the frequency
and voltage in the standard limits. Also, it is deduced that the proposed control
method can provide a better situation for the high power generation of DGs and is
faster and more efficient.
Moreover, it is concluded from Fig. 9.15 that, the EA-based control strategy
keeps the frequency value within its ±1%tolerance and is able to reach its rated
voltage within 0.07 s.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . . 243

(a) 450
PSO WOA GOA
400
350
300
Voltage (Volt)

250
200
150
100
50
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (sec)

(b)
PSO WOA GOA
50.3

50.3
Frequency (Hz)

50.1

50

49.9

49.8
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (sec)

Fig. 9.15 Voltage and frequency regulations comparison after applying EA methods [95]

9.4 Conclusion

This chapter provides an overview of the latest studies concerning the use of EAs
in microgrids planning. Approaches for optimal sizing, operational scheduling,
and voltage and frequency control of autonomous microgrids based on recently
introduced EAs have been presented. Regarding the probabilistic nature of these
algorithms, they present different performance in solving optimization problems.
GWO produces the most optimal solution for the BESS sizing problem and
determines the optimal size of battery energy storage (BES) considering various
constraints such as power and energy capacity of BES. On the other hand, GOA has
better performance in optimal sizing and optimal control problems of microgrids in
comparison with other algorithms. It can minimize computer memory usage, and
reduce the computation time due to the fast convergence as well as a promising
244 A. A. Ghavifekr

Table 9.7 Microgrids planning approaches and references using evolutionary algorithms
Evolutionary Algorithm Related Problem References
Genetic algorithm Operation scheduling [4, 71, 96–98]
Scheduling of power generation [6, 9–11]
Sources and sizing
BESS and storage devices sizing [7, 8, 82, 83]
Siting [13, 14, 99]
Voltage and frequency control [15, 90, 91, 100]
Particle swarm optimization Operation scheduling [18, 19, 73, 101, 102]
BESS and storage devices sizing [20–22, 84, 85]
Scheduling of power generation [25, 86]
Sources and sizing
Siting [26, 103]
Voltage and frequency control [24, 92, 93, 104, 105]
Ant colony optimization Operation scheduling [29, 30, 106]
Scheduling of power generation [28]
Sources and sizing
Biogeography-based Voltage and frequency control [32, 33]
optimization
Harmony search algorithm Siting [35, 36]
Operation scheduling [37, 38]
Scheduling of power generation [39, 87, 107]
Sources and sizing
Voltage and frequency control [40, 108]
Cuckoo search algorithm (CSA) BESS and storage devices sizing [44]
Operation scheduling [42, 43, 74–76]
Scheduling of power generation [45, 46]
Sources and sizing
Artificial bee colonies (ABC) Operation scheduling [50, 77, 78, 109]
Scheduling of power generation [48, 49]
Sources and sizing
Grey wolf optimization (GWO) Scheduling of power generation [52, 88, 89]
Sources and sizing
Operation scheduling [53, 79, 80]
Siting [54]
Firefly algorithm (FA) Voltage and frequency control [15, 58, 59]
Operation scheduling [56, 57, 81]
BESS and storage devices sizing [61, 110, 111]
Siting [60]
Grasshopper optimization Voltage and frequency control [63, 64, 94]
algorithm (GWO)
Scheduling of power generation [65]
Sources and sizing
Whale optimization algorithm Operation scheduling [68]
(WOA)
Voltage and frequency control [67]
9 Application of Heuristic Techniques and Evolutionary Algorithms. . . 245

balance between exploration and exploitation. Also, it is deduced that FA has better
performance in energy management and operation scheduling of microgrids, such
as determining the optimal power output of each generator at a minimum cost.
This review of the application of EA techniques for microgrid systems can be
instructive for microgrid planners and power system engineers. All reviewed papers
are classified in Table 9.7, regarding their optimization algorithm and the related
problem.

References

1. Kang, X., Nuworklo, C. E., Tekpeti, B. S., & Kheshti, M. (2017). Protection of micro-grid
systems: A comprehensive survey. The Journal of Engineering, 2017, 1515–1518.
2. Connolly, D., Lund, H., Mathiesen, B. V., & Leahy, M. (2010). A review of computer tools for
analysing the integration of renewable energy into various energy systems. Applied Energy,
87, 1059–1082.
3. F. Shariatzadeh, N. Kumar, and A. K. Srivastava, ”Intelligent control algorithms for optimal
reconfiguration of microgrid distribution system,“ in 2015 IEEE Industry Applications Society
Annual Meeting, 2015, pp. 1–8.
4. Leonori, S., Paschero, M., Mascioli, F. M. F., & Rizzi, A. (2020). Optimization strategies for
microgrid energy management systems by genetic algorithms. Applied Soft Computing, 86,
105903.
5. V. Suresh and P. Janik, ”Optimal Power Flow in Microgrids using Genetic Algorithm,“ in
2019 2nd International Conference on Power and Embedded Drive Control (ICPEDC), 2019,
pp. 483–487.
6. Rousis, A. O., Konstantelos, I., & Strbac, G. (2019). A planning model for a hybrid AC–DC
microgrid using a novel GA/AC OPF algorithm. IEEE Transactions on Power Systems, 35,
227–237.
7. Luo, L., Abdulkareem, S. S., Rezvani, A., Miveh, M. R., Samad, S., Aljojo, N., et al. (2020).
Optimal scheduling of a renewable based microgrid considering photovoltaic system and
battery energy storage under uncertainty. Journal of Energy Storage, 28, 101306.
8. T. Adefarati, S. Potgieter, R. Bansal, R. Naidoo, R. Rizzo, and P. Sanjeevikumar, ”Optimiza-
tion of PV-Wind-Battery Storage Microgrid System Utilizing a Genetic Algorithm,“ in 2019
International Conference on Clean Electrical Power (ICCEP), 2019, pp. 633–638.
9. J. Zhao, H. Nian, and L. Kong, ”multiobjective sizing optimization for combined heat
and power microgrids using a triangular evaluation model and the self-adaptive genetic
algorithm,“ in 2019 IEEE Innovative Smart Grid Technologies-Asia (ISGT Asia), 2019, pp.
1840-1845.
10. A. S. Eldessouky and H. A. Gabbar, ”Micro grid renewables dynamic and static performance
optimization using genetic algorithm,“ in 2015 IEEE International Conference on Smart
Energy Grid Engineering (SEGE), 2015, pp. 1–6.
11. M. Mahdavian, M. H. Kafi, A. Movahedi, and M. Janghorbani, ”improve performance in
electrical power distribution system by optimal capacitor placement using genetic algorithm,“
in 2017 14th International Conference on Electrical Engineering/Electronics, Computer,
Telecommunications and Information Technology (ECTI-CON), 2017, pp. 749–752.
12. P. P. Bagul and S. M. Akolkar, ”Relay coordination in microgrid,“ in 2017 International
Conference on Computing Methodologies and Communication (ICCMC), 2017, pp. 784–789.
246 A. A. Ghavifekr

13. M. Kamal, I. Ashraf, and E. Fernandez, ”Energy Resource Planning for a Rural Microgrid:
Comparison of Results using Different Optimization Algorithms,“ in 2019 Second Interna-
tional Conference on Advanced Computational and Communication Paradigms (ICACCP),
2019, pp. 1–6.
14. S. Korjani, A. Facchini, M. Mureddu, and A. Damiano, ”A genetic algorithm approach for
the identification of microgrids partitioning into distribution networks,“ in IECON 2017-43rd
Annual Conference of the IEEE Industrial Electronics Society, 2017, pp. 21–25.
15. A. Sawarni, K. Dhamane, D. Kumar, and H. Mathur, ”Decentralized Frequency Control
for an Isolated Microgrid using Nature Inspired Algorithms,“ in 2019 IEEE International
Conference on Electrical, Computer and Communication Technologies (ICECCT), 2019, pp.
1–6.
16. Parol, M., Wójtowicz, T., Ksi˛ez̈yk, K., Wenge, C., Balischewski, S., & Arendarski, B. (2020).
Optimum management of power and energy in low voltage microgrids using evolutionary
algorithms and energy storage. International Journal of Electrical Power & Energy Systems,
119, 105886.
17. J. Kennedy and R. Eberhart, ”Particle swarm optimization,“ in Proceedings of ICNN’95-
International Conference on Neural Networks, 1995, pp. 1942-1948.
18. L. Ravi, C. V. Kumar, and M. R. Babu, ”Stochastic optimal management of renewable
microgrid using simplified Particle Swarm Optimization algorithm,“ in 2018 4th International
Conference on Electrical Energy Systems (ICEES), 2018, pp. 261–266.
19. Hossain, M. A., Pota, H. R., Squartini, S., & Abdou, A. F. (2019). Modified PSO algorithm
for real-time energy management in grid-connected microgrids. Renewable Energy, 136, 746–
757.
20. D. Kaczorowska and J. Rezmer, ”Particle swarm algorithm for microgrid optimization,“ in
2018 Innovative Materials and Technologies in Electrical Engineering (i-MITEL), 2018, pp.
1–4.
21. M. W. Alzahlan, F. S. El-Faouri, M. G. Batarseh, A. Mohammad, and M. E. Za’ter, ”Particle
Swarm Optimization of a Microgrid’s Cost Function Involving Distributed Generation and
Highly Fluctuating Load,“ in 2019 IEEE Jordan International Joint Conference on Electrical
Engineering and Information Technology (JEEIT), 2019, pp. 319–324.
22. Y. Jiang, Y. Yang, S.-C. Tan, and S.-Y. R. Hui, ”Mitigating distribution power losses of
standalone AC microgrids using particle-swarm-optimization control for distributed battery
systems,“ in 2019 IEEE Applied Power Electronics Conference and Exposition (APEC), 2019,
pp. 3224–3229.
23. P. Malathi and M. Saravanan, ”Power management in microgrid using PI controller tuned
by particle swarm optimization,“ in 2017 Innovations in Power and Advanced Computing
Technologies (i-PACT), 2017, pp. 1–6.
24. S. H. Qazi, M. Mustafa, N. Hussain, and U. Sultana, ”Performance evaluation of PI and PI-
PSO in improving power quality of an autonomous microgrid,“ 2017.
25. R. R. Nivedha, J. G. Singh, and W. Ongsakul, ”PSO based economic dispatch of a
hybrid microgrid system,“ in 2018 International Conference on Power, Signals, Control and
Computation (EPSCICON), 2018, pp. 1–5.
26. L. K. Vedulla and M. K. Mishra, ”PSO based power sharing scheme for an islanded
DC microgrid system,“ in IECON 2017-43rd Annual Conference of the IEEE Industrial
Electronics Society, 2017, pp. 392–397.
27. Dorigo, M., Birattari, M., & Stutzle, T. (2006). Ant colony optimization. IEEE Computational
Intelligence Magazine, 1, 28–39.
28. N. Emerson and S. Srinivasan, ”Integrating hybrid power source into islanded microgrid using
ant colony optimization,“ in 2015 International Conference on Advanced Computing and
Communication Systems, 2015, pp. 1–4.
29. I. N. Trivedi, D. K. Thesiya, A. Esmat, and P. Jangir, ”A multiple environment dispatch
problem solution using ant colony optimization for micro-grids,“ in 2015 International
Conference on Power and Advanced Control Engineering (ICPACE), 2015, pp. 109–115.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . . 247

30. Marzband, M., Yousefnejad, E., Sumper, A., & Domínguez-García, J. L. (2016). Real
time experimental implementation of optimum energy management system in standalone
microgrid by using multi-layer ant colony optimization. International Journal of Electrical
Power & Energy Systems, 75, 265–274.
31. Simon, D. (2008). Biogeography-based optimization. IEEE Transactions on Evolutionary
Computation, 12, 702–713.
32. Mishra, S., Mallesham, G., & Sekhar, P. (2013). Biogeography based optimal state feedback
controller for frequency regulation of a smart microgrid. IEEE Transactions on Smart Grid,
4, 628–637.
33. R. H. Kumar and S. Ushakumari, ”Biogeography based tuning of PID controllers for Load
Frequency Control in microgrid,“ in 2014 International Conference on Circuits, Power and
Computing Technologies [ICCPCT-2014], 2014, pp. 797–802.
34. Geem, Z. W., Kim, J. H., & Loganathan, G. V. (2001). A new heuristic optimization algorithm:
harmony search. simulation, 76, 60–68.
35. P. Monica and M. Kowsalya, ”Improved Harmony Search Based Optimization of Droop
Control Parameters for Load Sharing in DC Microgrids,“ in TENCON 2019–2019 IEEE
Region 10 Conference (TENCON), 2019, pp. 2199–2203.
36. N. A. Ashtiani, M. Gholami, and G. Gharehpetian, ”Optimal allocation of energy storage
systems in connected microgrid to minimize the energy cost,“ in 2014 19th Conference on
Electrical Power Distribution Networks (EPDC), 2014, pp. 25–28.
37. E. Lazar, A. Ignat, D. Petreus, and R. Etz, ”Energy Management for an Islanded Microgrid
Based on Harmony Search Algorithm,“ in 2018 41st International Spring Seminar on
Electronics Technology (ISSE), 2018, pp. 1–6.
38. Zhang, J., Wu, Y., Guo, Y., Wang, B., Wang, H., & Liu, H. (2016). A hybrid harmony search
algorithm with differential evolution for day-ahead scheduling problem of a microgrid with
consideration of power flow constraints. Applied Energy, 183, 791–804.
39. Y. Li, W. Dong, X. Feng, and M. Zhang, ”The Expansion Planning of Wind-Thermal Co-
generation System Based on Harmony Search Algorithm under Smart Grid,“ in 2019 IEEE
International Conference on Communications, Control, and Computing Technologies for
Smart Grids (SmartGridComm), 2019, pp. 1–5.
40. Sedhom, B. E., El-Saadawi, M. M., Hatata, A. Y., & Alsayyari, A. S. (2020). Hierarchical
control technique-based harmony search optimization algorithm versus model predictive
control for autonomous smart microgrids. International Journal of Electrical Power & Energy
Systems, 115, 105511.
41. Rajabioun, R. (2011). Cuckoo optimization algorithm. Applied Soft Computing, 11, 5508–
5518.
42. J. Singh, S. Poddar, S. P. Ramalingam, P. K. Shanmugam, and A. Kalam, ”investigation
on dynamic economic dispatch problem of micro grid using cuckoo search algorithm-
grid connected and islanded mode,“ in TENCON 2019–2019 IEEE Region 10 Conference
(TENCON), 2019, pp. 1886-1891.
43. D. Arcos-Aviles, G. García-Gutièrrez, F. Guinjoan, E. V. Carrera, J. Pascual, P. Ayala, et al.,
”Adjustment of the Fuzzy Logic controller parameters of the energy management strategy of a
grid-tied domestic electro-thermal microgrid using the Cuckoo search algorithm,“ in IECON
2019-45th Annual Conference of the IEEE Industrial Electronics Society, 2019, pp. 279–285.
44. K. Tantrapon, P. Jirapong, and P. Thararak, ”Optimal Operation of Battery Energy Storage
System for Mitigating Voltage Fluctuation in Microgrid Using Cuckoo Search Optimiza-
tion,“ in 2019 16th international conference on electrical engineering/electronics, computer,
Telecommunications and Information Technology (ECTI-CON), 2019, pp. 877–880.
45. S. Vasanthakumar, N. Kumarappan, R. Arulraj, and T. Vigneysh, ”Cuckoo search algorithm
based environmental economic dispatch of microgrid system with distributed generation,“
in 2015 International Conference on Smart Technologies and Management for Computing,
Communication, Controls, Energy and Materials (ICSTM), 2015, pp. 575–580.
248 A. A. Ghavifekr

46. Kumar, J. S., Raja, S. C., Nesamalar, J. J. D., & Venkatesh, P. (2018). Optimizing renewable
based generations in AC/DC microgrid system using hybrid Nelder-Mead–cuckoo search
algorithm. Energy, 158, 204–215.
47. Karaboga, D. (2010). Artificial bee colony algorithm. scholarpedia, 5, 6915.
48. P. Chaweewat, J. G. Singh, W. Ongsakul, and A. K. Srivastrava, ”Economic and environ-
mental impact assessment with network reconfiguration in microgrid by using artificial bee
colony,“ in 2016 International Conference on Cogeneration, Small Power Plants and District
Energy (ICUE), 2016, pp. 1–7.
49. L. Ciabattoni, F. Ferracuti, G. Ippoliti, and S. Longhi, ”Artificial bee colonies based optimal
sizing of microgrid components: A profit maximization approach,“ in 2016 IEEE Congress
on Evolutionary Computation (CEC), 2016, pp. 2036-2042.
50. M. M. Moghaddam, M. Marzband, and F. Azarinejadian, ”Optimal energy management for a
home Microgrid based on multi-period artificial bee colony,“ in 2017 Iranian Conference on
Electrical Engineering (ICEE), 2017, pp. 1446–1451.
51. Mirjalili, S., Mirjalili, S. M., & Lewis, A. (2014). Grey wolf optimizer. Advances in
Engineering Software, 69, 46–61.
52. Miao, D., & Hossain, S. (2020). Improved gray wolf optimization algorithm for solving
placement and sizing of electrical energy storage system in micro-grids. ISA Transactions.
53. Y. Cao, T. Chen, X. Qing, J. Zhang, and L. Sun, ”Energy Management for a Microgrid
With Different Charging and Discharging Priorities of Batteries Using Modified Grey Wolf
Optimizer,“ in 2019 IEEE PES Asia-Pacific Power and Energy Engineering Conference
(APPEEC), 2019, pp. 1–5.
54. Zhang, J., Wang, X., & Ma, L. (2019). An optimal power allocation scheme of microgrid
using Grey wolf optimizer. IEEE Access, 7, 137608–137619.
55. Yang, X.-S. (2008). Firefly algorithm. Nature-inspired Metaheuristic Algorithms, 20, 79–90.
56. J. D. Vasanth, N. Kumarappan, R. Arulraj, and T. Vigneysh, ”Minimization of operation cost
of a microgrid using firefly algorithm,“ in 2017 IEEE International Conference on Intelligent
Techniques in Control, Optimization and Signal Processing (INCOS), 2017, pp. 1–6.
57. M. Abdullah, N. L. Abdullah, and J. Jamian, ”Optimal power generation in microgrid system
using firefly algorithm,“ in 2017 6th International Conference on Electrical Engineering and
Informatics (ICEEI), 2017, pp. 1–6.
58. R. Khan, N. Gogoi, J. Barman, A. Latif, and D. C. Das, ”Virtual Power Plant enabled Co-
ordinated Frequency Control of a grid connected Independent Hybrid Microgrid using Firefly
algorithm,“ in 2019 IEEE Region 10 Symposium (TENSYMP), 2019, pp. 795–800.
59. M. Akbari, M. Golkar, and S. M. Tafreshi, ”Firefly algorithm-based voltage and frequency
control of a hybrid ac-dc microgrid,“ in 2012 Proceedings of 17th Conference on Electrical
Power Distribution, 2012, pp. 1–7.
60. S. Sun, J. Ni, Y. Cheng, P. Yu, Z. Xin, and Q. Ai, ”Firefly algorithm based dynamic coalition
formation strategy in micro-grids,“ 2015.
61. Satapathy, P., Dhar, S., & Dash, P. K. (2017). Stability improvement of PV-BESS diesel
generator-based microgrid with a new modified harmony search-based hybrid firefly algo-
rithm. IET Renewable Power Generation, 11, 566–577.
62. Saremi, S., Mirjalili, S., & Lewis, A. (2017). Grasshopper optimisation algorithm: Theory
and application. Advances in Engineering Software, 105, 30–47.
63. A. Annamraju and S. Nandiraju, ”Frequency Control in an Autonomous Two-area Hybrid
Microgrid using Grasshopper Optimization based Robust PID Controller,“ in 2018 8th IEEE
India International Conference on Power Electronics (IICPE), 2018, pp. 1–6.
64. Barik, A. K., & Das, D. C. (2018). Expeditious frequency control of solar photo-
voltaic/biogas/biodiesel generator based isolated renewable microgrid using grasshopper
optimisation algorithm. IET Renewable Power Generation, 12, 1659–1667.
65. Bukar, A. L., Tan, C. W., & Lau, K. Y. (2019). Optimal sizing of an autonomous photo-
voltaic/wind/battery/diesel generator microgrid using grasshopper optimization algorithm.
Solar Energy, 188, 685–696.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . . 249

66. Mirjalili, S., & Lewis, A. (2016). The whale optimization algorithm. Advances in Engineering
Software, 95, 51–67.
67. S. H. Qazi, M. Uqaili, and U. Sultana, ”Whales Optimization Algorithm Based Enhanced
Power Controller for an Autonomous Microgrid System,“ in 2019 8th International Confer-
ence on Modern Power Systems (MPS), 2019, pp. 1–7.
68. I. N. Trivedi, M. Bhoye, R. Bhesdadiya, P. Jangir, N. Jangir, and A. Kumar, ”An emission
constraint environment dispatch problem solution with microgrid using Whale Optimization
Algorithm,“ in 2016 National Power Systems Conference (NPSC), 2016, pp. 1–6.
69. J. R. Koza, ”Survey of genetic algorithms and genetic programming,“ in Wescon conference
record, 1995, pp. 589–594.
70. Yamashita, D. Y., Vechiu, I., & Gaubert, J.-P. (2020). A review of hierarchical control for
building microgrids. Renewable and Sustainable Energy Reviews, 118, 109523.
71. Askarzadeh, A. (2017). A memory-based genetic algorithm for optimization of power
generation in a microgrid. IEEE Transactions on Sustainable Energy, 9, 1081–1089.
72. V. Guliashki and G. Marinova, ”An Accelerated Genetic Single Objective Algorithm for
Optimization of Energy Flows in Microgrids,“ in 2018 25th international conference on
systems, Signals and Image Processing (IWSSIP), 2018, pp. 1–5.
73. Moghaddam, A. A., Seifi, A., & Niknam, T. (2012). Multi-operation management of a
typical micro-grids using particle swarm optimization: A comparative study. Renewable and
Sustainable Energy Reviews, 16, 1268–1281.
74. M. Bhoye, S. N. Purohit, I. N. Trivedi, M. H. Pandya, P. Jangir, and N. Jangir, ”Energy
management of renewable energy sources in a microgrid using cuckoo search algorithm,“ in
2016 IEEE Students’ Conference on Electrical, Electronics and Computer Science (SCEECS),
2016, pp. 1–6.
75. M. De, G. Das, S. Mandal, and K. Mandal, ”A reliable energy management and generation
scheduling model in microgrids using modified cuckoo search algorithm,“ in 2018 Emerging
Trends in Electronic Devices and Computational Techniques (EDCT), 2018, pp. 1–6.
76. M. Modiri-Delsha, S. S. Taheri, S. J. Seyed-Shenava, and N. Rahim, ”Optimal generation
scheduling in microgrids by cuckoo search algorithm,“ 2014.
77. N. K. Paliwal, N. K. Singh, and A. K. Singh, ”Optimal power flow in grid connected microgrid
using artificial bee colony algorithm,“ in 2016 IEEE Region 10 Conference (TENCON), 2016,
pp. 671–675.
78. M. Govardhan and R. Roy, ”Artificial Bee Colony based optimal management of microgrid,“
in 2012 11th International Conference on Environment and Electrical Engineering, 2012, pp.
334–339.
79. J. Yi and D. Shi, ”Parameter Optimization of DOCR and FCL in Microgrids Based on Grey
Wolf Optimizer,“ in 2018 53rd International Universities Power Engineering Conference
(UPEC), 2018, pp. 1–5.
80. M. Moazzami, M. Ghanbari, H. Shahinzadeh, J. Moradi, and G. B. Gharehpetian, ”Appli-
cation of multi-objective grey wolf algorithm on energy management of microgrids with
techno-economic and environmental considerations,“ in 2018 3rd Conference on Swarm
Intelligence and Evolutionary Computation (CSIEC), 2018, pp. 1–9.
81. Mohammadi, S., Mozafari, B., Solimani, S., & Niknam, T. (2013). An adaptive modified
firefly optimisation algorithm based on Hong’s point estimate method to optimal operation
management in a microgrid with consideration of uncertainties. Energy, 51, 339–348.
82. D.-O. Sidea, L. Toma, M. Sanduleac, I.-I. Picioroaga, and V.-A. Boicea, ”Optimal BESS
Scheduling Strategy in Microgrids Based on Genetic Algorithms,“ in 2019 IEEE Milan
PowerTech, 2019, pp. 1–6.
83. Mahmoud, T. S., Ahmed, B. S., & Hassan, M. Y. (2019). The role of intelligent generation
control algorithms in optimizing battery energy storage systems size in microgrids: A case
study from Western Australia. Energy Conversion and Management, 196, 1335–1352.
84. Kerdphol, T., Qudaih, Y., & Mitani, Y. (2016). Optimum battery energy storage system
using PSO considering dynamic demand response for microgrids. International Journal of
Electrical Power & Energy Systems, 83, 58–66.
250 A. A. Ghavifekr

85. Kerdphol, T., Fuji, K., Mitani, Y., Watanabe, M., & Qudaih, Y. (2016). Optimization of a
battery energy storage system using particle swarm optimization for stand-alone microgrids.
International Journal of Electrical Power & Energy Systems, 81, 32–39.
86. F. Tooryan, E. R. Collins, A. Ahmadi, and S. S. Rangarajan, ”Distributed generators optimal
sizing and placement in a microgrid using PSO,“ in 2017 IEEE 6th International Conference
on Renewable Energy Research and Applications (ICRERA), 2017, pp. 614–619.
87. Elattar, E. E. (2018). Modified harmony search algorithm for combined economic emission
dispatch of microgrid incorporating renewable sources. Energy, 159, 496–507.
88. S. Sukumar, M. Marsadek, A. Ramasamy, and H. Mokhlis, ”Grey Wolf Optimizer Based
Battery Energy Storage System Sizing for Economic Operation of Microgrid,“ in 2018
IEEE International Conference on Environment and Electrical Engineering and 2018 IEEE
Industrial and Commercial Power Systems Europe (EEEIC/I&CPS Europe), 2018, pp. 1–5.
89. Sharma, S., Bhattacharjee, S., & Bhattacharya, A. (2016). ”Grey wolf optimisation for
optimal sizing of battery energy storage device to minimise operation cost of microgrid,“
IET Generation. Transmission and Distribution, 10, 625–637.
90. P. C. Nayak, S. Bisoi, R. C. Prusty, and S. Panda, ”Performance Analysis of PDF+ (1+
PI) Controller for Load Frequency Control of the Multi Microgrid System Using Genetic
Algorithm,“ in 2019 International Conference on Information Technology (ICIT), 2019, pp.
448–453.
91. R. Wang, S. Wu, C. Wang, S. An, Z. Sun, W. Li, et al., ”optimized Operation and Control of
Microgrid based on Multi-objective Genetic Algorithm,“ in 2018 International Conference
on Power System Technology (POWERCON), 2018, pp. 1539–1544.
92. S. F. Eilaghi, A. Ahmadian, and M. A. Golkar, ”Optimal voltage unbalance compensation in
a microgrid using PSO algorithm,“ in 2016 IEEE 7th Power India International Conference
(PIICON), 2016, pp. 1–6.
93. Al-Saedi, W., Lachowicz, S. W., Habibi, D., & Bass, O. (2013). Voltage and frequency
regulation based DG unit in an autonomous microgrid operation using particle swarm
optimization. International Journal of Electrical Power & Energy Systems, 53, 742–751.
94. D. K. Lal, A. K. Barisal, and M. Tripathy, ”Load frequency control of multi area intercon-
nected microgrid power system using grasshopper optimization algorithm optimized fuzzy
PID controller,“ in 2018 Recent Advances on Engineering, Technology and Computational
Sciences (RAETCS), 2018, pp. 1–6.
95. Jumani, T. A., Mustafa, M. W., Md Rasid, M., Mirjat, N. H., Leghari, Z. H., & Saeed, M. S.
(2018). Optimal voltage and frequency control of an islanded microgrid using grasshopper
optimization algorithm. Energies, 11, 3191.
96. H. Li, C. Zang, P. Zeng, H. Yu, and Z. Li, ”A genetic algorithm-based hybrid optimization
approach for microgrid energy management,“ in 2015 IEEE International Conference on
Cyber Technology in Automation, Control, and Intelligent Systems (CYBER), 2015, pp. 1474–
1478.
97. P. P. Vergara, R. Torquato, and L. C. Da Silva, ”Towards a real-time Energy Management
System for a Microgrid using a multi-objective genetic algorithm,“ in 2015 IEEE Power &
Energy Society General Meeting, 2015, pp. 1–5.
98. C. Changsong, D. Shanxu, C. Tao, L. Bangyin, and Y. Jinjun, ”Energy trading model for
optimal microgrid scheduling based on genetic algorithm,“ in 2009 IEEE 6th International
Power Electronics and Motion Control Conference, 2009, pp. 2136–2139.
99. Kirthiga, M. V., Daniel, S. A., & Gurunathan, S. (2012). A methodology for transforming an
existing distribution network into a sustainable autonomous micro-grid. IEEE Transactions
on Sustainable Energy, 4, 31–41.
100. A. Singh, ”GA optimized PID controller for frequency regulation in standalone AC micro-
grid,“ in 2016 7th India International Conference on Power Electronics (IICPE), 2016, pp.
1–5.
101. Li, P., Xu, D., Zhou, Z., Lee, W.-J., & Zhao, B. (2015). Stochastic optimal operation of
microgrid based on chaotic binary particle swarm optimization. IEEE Transactions on Smart
Grid, 7, 66–73.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . . 251

102. M. A. Djema and M. Boudour, ”Autonomous microgrid power management using Particle
Swarm Optimization,“ in 2016 International Conference on Control, Decision and Informa-
tion Technologies (CoDIT), 2016, pp. 222–226.
103. Oboudi, M., Hooshmand, R., & Karamad, A. (2017). A feasible method for controlled
intentional islanding in microgrids based on PSO algorithm. Swarm and Evolutionary
Computation, 35, 14–25.
104. Y. Du, D. Gong, K. Ding, R. Chen, H. Wang, W. Xi, et al., ”Research on Autonomous
Frequency Modulation and Voltage Regulation Method of Microgrid Based on Immunity
Selection-particle Swarm Optimization (ISPSO),“ in 2019 IEEE PES Asia-Pacific Power and
Energy Engineering Conference (APPEEC), 2019, pp. 1–5.
105. M. Akbari, M. Golkar, and S. Tafreshi, ”A PSO solution for improved voltage stability of a
hybrid ac-dc microgrid,“ in ISGT2011-India, 2011, pp. 352–357.
106. C. Colson, M. Nehrir, and C. Wang, ”Ant colony optimization for microgrid multi-objective
power management,“ in 2009 IEEE/PES Power Systems Conference and Exposition, 2009,
pp. 1–7.
107. Li, P., Li, R.-X., Cao, Y., Li, D.-Y., & Xie, G. (2017). Multiobjective sizing optimization
for island microgrids using a triangular aggregation model and the levy-harmony algorithm.
IEEE Transactions on Industrial Informatics, 14, 3495–3505.
108. Sedhom, B. E., El-Saadawi, M. M., Elhosseini, M. A., Saeed, M. A., & Abd-Raboh, E.
E. (2019). A harmony search-based H-infinity control method for islanded microgrid. ISA
Transactions.
109. Marzband, M., Azarinejadian, F., Savaghebi, M., & Guerrero, J. M. (2015). An optimal
energy management system for islanded microgrids based on multiperiod artificial bee colony
combined with Markov chain. IEEE Systems Journal, 11, 1712–1722.
110. O. Penangsang, A. Soeprijanto, A. A. Fitriana, and E. S. Ningrum, ”Operation optimization
stand-alone microgrid using firefly algorithm considering lifetime characteristics of battery,“
in 2016 International Seminar on Intelligent Technology and Its Applications (ISITIA), 2016,
pp. 565–570.
111. Satapathy, P., Dhar, S., & Dash, P. (2016). A mutated hybrid firefly approach to mitigate
dynamic oscillations of second order PLL based PV-battery system for microgrid applica-
tions. Sustainable Energy Technologies and Assessments, 16, 69–83.
 Part II
Control of Microgrids
Chapter 10
Conventional Droop Methods
for Microgrids

Kwang Woo Joung and Jung-Wook Park

10.1 Introduction

Recently, renewable energy sources (RESs) have been connected worldwide to

power grids in the form of distributed generation (DG) because of environmental
and economic reasons. In order to efficiently operate and control DGs in the
decentralized small-scale power grid, the concept of a microgrid is used. In a
microgrid, the hierarchical control system is required to control various types of
generators in parallel. Then, the droop control is widely used for the most local
controller, which is also called as primary control [1–5].
In AC microgrid, the conventional droop control is divided into the frequency and
voltage droop controls [6, 7]. The characteristic of P–f droop control is based on the
relationship between the active power from the conventional synchronous generator
and the frequency of the system. In other words, if the amount of P from the DGs
is suddenly increased, f is reduced and vice versa. Similarly, the characteristic of
Q–V droop control is applied for restoring V to its nominal value by providing the
negative Q when the value of V drops.
The synchronous generator has these droop features inherently by rotational
inertia. To improve its control performance, additional droop control methods are
often used. However, the RESs, which are usually based on the inverter, have no
inertia. Without grid supporting controls, they will generate constant power output
even when the frequency and voltage of the system are changed [8, 9]. Therefore, the
droop control is required for the RESs to maintain a stable operation of microgrid
while having the effect of inertia [10]. On the other hand, the grid frequency is not
considered in DC microgrid, and there is no need for DGs to be synchronized. Thus,

K. W. Joung · J.-W. Park ()

School of Electrical & Electronic Engineering, Yonsei University, Seoul, South Korea
e-mail: jungpark@yonsei.ac.kr

© Springer Nature Switzerland AG 2021 255

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_10
256 K. W. Joung and J.-W. Prank

in DC microgrid, the voltage becomes the only major factor in implementing the
droop control [11–13]. In general, the voltage–current (V–I) droop control method,
which is based on the relationship between line resistance and current, is used. The
generators with droop control in both AC and DC microgrids have the following
advantages.
• The parallel operation among the DGs is achieved.
• The power allocation is possible by considering the capacity or characteristics of
DGs.
• The DG can be controlled as a grid supporting unit enhancing the reliability of
the microgrid.
• No communication lines are needed.
In this chapter, the conventional droop control methods used in microgrids
are firstly described. They can be implemented for generators in AC microgrid
with the “self-synchronizing” characteristic of a synchronous generator. However,
particularly for inverter-based renewables (IBRs) among the RESs, their power
outputs can be varied according to environmental conditions. For example, the
power of the wind turbine generator and the photovoltaic generator will vary
according to wind speed and irradiance, respectively. Therefore, to support a grid
with many IBRs is difficult because their maximum power output depends on the
external environment. This makes the energy storage system (ESS) necessary in
microgrid since it can reduce the variability of RES as a control unit for enhancing
the flexibility and stability of microgrid. The main advantage of ESS is that it
can mitigate the power fluctuation faster than conventional synchronous generators.
Also, unlike other generators, the ESS enables absorbing the active power from the
grid by its charging operation.
By considering several different features of DGs, the power sharing among DGs
will be successfully achieved when the suitable droop methods are applied to each
type of generators. Furthermore, to use the IBRs in the DC system, the conventional
DC droop control methods can be applied. In this case, the droop coefficient will
act as a “virtual resistance” to imitate the voltage drop caused by actual resistance.
The mathematical analysis and simulation study are given below for both AC and
DC microgrids to verify the effectiveness of existing droop control methods.

10.2 Conventional Droop Method for AC Microgrid

10.2.1 Mathematical Analysis

The generation and consumption of active powers must be balanced to maintain the
grid frequency within its proper ranges. When this power balance in a microgrid is
10 Conventional Droop Methods for Microgrids 257

broken, the powers from some generators will be increased (or decreased) abruptly.
Then, the rotating speed of generators is determined as.

2H d 2 δ 2H dω
2
= = Pm − Pe (10.1)
ωs dt ωs dt

where
H is the inertia constant.
δ is the power angle of the generator.
ωs is the grid frequency (at synchronous speed).
ω is the grid frequency.
Pm and Pe are the mechanical and electrical powers of the generator, respectively.
In Eq. (10.1), whenever the values of Pm and Pe are not the same, the derivative
of ωm changes. This means that ω has a deviation from ωs . After the power balance
of the microgrid is broken, the outputs from each generator will change until the
new equilibrium state is achieved. For example, assume that one of the generators
in a microgrid is suddenly disconnected. Then, Pm of the entire system is decreased,
and the derivative of ω becomes negative while resulting in the reduction of the
grid frequency. Thereafter, the other generators will increase their power outputs to
mitigate the power deviations. The droop control is widely used to implement this
physical characteristic in order to control synchronous generators more precisely.
In other words, the power output deviation of the generator can be determined by
using the frequency deviation as.

 
fi − f ∗ = −Rp−f · Pi − P ∗ (10.2)

where
fi and f* are the actual and nominal frequencies of ith generator, respectively.
Pi and P* are the actual and nominal active power outputs from ith generator,
respectively.
RP–f is the droop constant of P–f droop control.
In general, the droop constant is set by considering the physical limitation of
the generator in its output power variations. From Eq. (10.2), it can be shown that
the deviations of both active power and grid frequency have a negative proportional
relationship. The characteristic of such a droop control method is shown in Fig.
10.1. Assume that the reference values of grid frequency and active power output
in a particular operating point are f1 and P1 , respectively. When the grid frequency
is reduced from f1 to f2 , the output from the generator will change from P1 to P2
according to its droop characteristic.
For the microgrid with the high renewable penetration, however, this droop
control method is difficult to be applied. This is because many RESs are controlled
258 K. W. Joung and J.-W. Prank

f
Storage Genearation
P<0 P>0

f1 RP – f
f2

P
– Pmax P1 P2 P
max

Fig. 10.1 Characteristic of P–f droop control method

to generate the active power close to their power output limit, Pmax . If the droop
control of Eq. (10.2) is used, the deviation of grid frequency can be calculated by
the droop constant of the entire system as.

Δfsys = −Rsys · ΔP (10.3)

#n
i=1 Ri × Pi,cap
Rsys = #n (10.4)
i=1 Pi,cap

where
fsys is the deviation of grid frequency for the entire microgrid system.
P is the deviation of active power generation caused by a disturbance.
Rsys is the droop constant of the entire microgrid system.
Ri is the droop constant of ith generator.
Pi,cap is the capacity of ith generator.
The value of Rsys in Eq. (10.4) is affected by the operating status of RESs, which
can generate more outputs only if they have enough reserve power [6]. In other
words, the droop control can be only applied to the RESs, which generate less power
when compared to their maximum power point (MPP). After a disturbance, the
power deviation is mitigated by generators with the droop control according to Eq.
(10.2). If the power output of one RES is increased to its maximum (or conversely
if its maximum power is reduced due to the change of environmental condition),
the RES is unable to participate in mitigating the power deviation anymore because
it only generates the constant maximum power. After at least one RES turns into
uncontrollable, Eq. (10.3) can be changed as.
10 Conventional Droop Methods for Microgrids 259

Fig. 10.2 The reduction of

droop constant due to the
change in the operating status
of RES

#
Ri × Pi,cap ⎛ ⎞
i=G,drp 
Δfsys = # ⎝Pi − ΔPi,m ⎠ + Δfm (10.5)
Pi,cap
i=G,drp
i=G,drp

where
Pi,m is the deviation of active power before at least one RES becomes uncontrol-
lable.
fm is the deviation of grid frequency before at least one RES becomes uncontrol-
lable.
The subscript, G, drp indicates the generator, which can be controlled by the droop
control method.
From Eq. (10.5) the droop constant for the entire microgrid will be reduced after
the RES is unable to control, as illustrated in Fig. 10.2. This usually happens to the
RESs because their reserve powers are not always determined precisely.
Similar to grid frequency, to maintain the voltage stable is also important. Note
that the reactive power is generated depending on the voltage difference between
buses as.

V12 V1 · V2
Q= · cos (δ) (10.6)
X12 X12

where
Q is the reactive power from the generator.
V1 and V2 are the magnitudes of voltage at buses, 1 and 2, respectively.
X12 is the line impedance between buses, 1 and 2.
260 K. W. Joung and J.-W. Prank

It is known from Eq. (10.6) that the voltage can be regulated by controlling the
reactive power. That is, as the reactive power output becomes greater, the bus voltage
is increased more. When the generations from RESs are increased without any
voltage control, it might cause the voltage stability problem. In particular, it becomes
serious in microgrid because the distance among generators is relatively close when
compared to a large-scale power system. Then, this voltage problem can be solved
by controlling the reactive power deviation, which is inversely proportional to the
voltage deviation as.

 
Vi − V ∗ = −RQ−V · Qi − Q∗ (10.7)

where
Vi and V* are the actual and nominal voltages of ith generator, respectively.
Qi and Q* are the actual and nominal reactive power outputs from ith generator,
respectively.
RQ–V is the droop constant of Q–V droop control.
The performance of power sharing depends on the droop constant. The use of
a high droop constant might still cause to fluctuate the voltage. In general, it is
set by considering the maximum reactive power output from the generator and the
maximum voltage variations at the bus as.

Qmax − Qmin
RQ−V = (10.8)
Vhigh − Vlow

The characteristic of Q–V droop control method by Eq. (10.7) is shown in Fig.
10.3. Assume that the reference values of bus voltage and reactive power output in
a particular operating point are V1 and Q1 , respectively. When the bus voltage is
reduced from V1 to V2 , the generator increases the reactive power output from Q1 to
Q2 according to its droop characteristic. The combined P–f and Q–V droop control
method is shown in Fig. 10.4. The gray area (circle) represents the operating range
of the generator with the apparent power, S. The shapes of this area are different
depending on the type of generator.

10.2.2 Synchronous Generator

In AC microgrid, the diesel synchronous generator is usually used because of its fast
ramp-up capability, reliability, durability, and so on. It is the dispatchable source,
which can fully control the active and reactive powers with its governor and exciter,
respectively. For its governor control, the model of “DEGOV1” as shown in Fig.
10.5 can be used to carry out its P–f droop control. It can be implemented through
10 Conventional Droop Methods for Microgrids 261

V
Inductive Capacitive
Q<0 Q>0
V1

V2
RQ – V

Q
Q1 Q2
–Qmax Qmax
Fig. 10.3 Characteristic of Q–V droop control method

Q Q

Qmax

Q*
Capacitive
P
*
V V
Inductive

–Qmax

f*
Storage Genearation
P
– Pmax P * Pmax

Fig. 10.4 Operating range of generator by the combined P–f and Q–V droop control method in
AC microgrid

the droop constant, RP–f in practice. Typically, the value of RP–f is set to 5% by
considering the ramp-up speed of power from the diesel generator. On the other
hand, the Q–V droop control is implemented by adding the virtual line impedance
(the generator will control the virtual bus voltage instead of real terminal voltage). If
the value of virtual line impedance is negative, the terminal voltage is compensated
262 K. W. Joung and J.-W. Prank

f Actuator
Tmax
*
f Electric control box
Engine
*
– (1 + sT3 ) K (1 + sT4 )
P
1 + sT1 + s 2T2T1 e – sTD P
s (1 + sT5 ) + (1 + sT6 )

Tmin
Droop Control
RP – f

Fig. 10.5 Governor model of diesel generator (DEGOV1)

Comp IEEEVC
V V
Vc = V – j . X e . I Vc Vc = V + ( Rc + j . X c ) . I Vc
I I
(a) (b)

Fig. 10.6 Voltage droop control model (a) Comp (b) IEEEVC

by controlling the reactive power output. In other words, the capacitive reactive
power is released when the terminal voltage is decreased. In contrast, when it is
increased, the inductive reactive power from the generator is provided. The model
“Comp” or “IEEEVC” in Fig. 10.6 are used to carry out the Q–V droop control. The
maximum value of reactive power is determined by the power factor of output from
a diesel generator. Typically, the droop constant, RQ–V is set to 4 ~ 6%.
The gray area in Fig. 10.7 represents the operating range of diesel generators.
Because the diesel generator cannot generate negative active power, its operating
range is limited to the right hemisphere area in Fig. 10.4. Also, the diesel generator
has some operating power reserve such that the value of output reference is reduced
by 5% from the maximum power output. Moreover, the idle operation (or generating
too small output) must be avoided because of its poor operating efficiency. Finally,
the other physical limits such as field current, armature heating, winding end region
heating, and under excitation may exist [7].

10.2.3 Renewable Energy Sources

The power output from RES is very sensitive to environmental conditions. It

can be divided into two types, which are the IBR and non-IBR depending on
whether it is connected to the microgrid through an inverter or not. The IBR is
difficult to control because the RES must be synchronized with the main grid.
In particular, this becomes the major issue when the penetration ratio of RESs to
10 Conventional Droop Methods for Microgrids 263

Q
Field current limit

Qmax,cap Armature heating

constraints

P
V

Qmax,ind Winding end region

heating limit

Under excitation limit

Pmin Pmax

Fig. 10.7 Operating range of diesel generator

microgrid is very high. On the contrary, the non-IBR such as hydroelectric, which
is physically synchronized with the grid, is easy to control because it is identical to
the synchronous generator.

10.2.3.1 IBR

The replacement of diesel generators with the RESs in the form of IBR reduces
the system inertia because they do not have any rotational inertia. For example,
the photovoltaic (PV) generator uses a DC–AC inverter, and the wind turbine (WT)
generator requires a back-to-back converter for connecting it to the microgrid. Then,
the power generations from such IBRs are subject to makes the system unstable,
when the unbalance between power supply and demand occurs. Therefore, the
proper control method for grid support must be applied to keep increasing the
penetration level of IBRs.
The structure of the power controller applied to the IBR is shown in Fig.
10.8. The output frequency is synchronized by the phase-locked loop (PLL) with
the frequency of the main microgrid, and it is first measured. Then, the external
power control loop is used to control P and Q to their reference values with its
droop control. In other words, it changes P and Q properly based on its droop
characteristics. In normal operation, the active power output of IBR is determined
by the maximum power point tracking (MPPT) control method. However, when this
droop control method is applied to the IBR, the power deviation will be added to
264 K. W. Joung and J.-W. Prank

Inverter
Vabc
I abc
Inner loops
f
Voltage Current PWM
Control Loop Control Loop
V

P
Droop Power
Control Calculation
Q

Fig. 10.8 Structure of control block diagram for IBR

RES
Characteristics
Environmental
factors
P
P* f* f
Droop
f

P
P

Q*
V* V
Droop
V
Q
Q

Fig. 10.9 Structure of droop control for IBR

the MPPT signal, as shown in Fig. 10.9. Next, the inner loops with the voltage and
current controls are used to regulate the output from the inverter to the d and q-axes
reference signals.
However, there are many constraints to operate IBRs with droop control in
practice. For example, for the P–f droop control, it is difficult to allocate the reserve
power to each IBR. As mentioned earlier, MPPT control is mainly used for the
IBR due to economic reasons. In this case, the active power from IBR cannot be
increased even when the grid frequency is reduced, and therefore the additional
active power is required. In other words, it can be only decreased at MPP. Therefore,
in order to fully apply the P–f droop control, the IBR must be capable of adjusting
10 Conventional Droop Methods for Microgrids 265

Power Factor

Qmax,cap

P
V

Qmax,ind
Pmin PMPPT
determined by
environmental factors

Fig. 10.10 Operating range of IBR

the reference signal such that it can reduce its maximum output while obtaining the
reserve power. Also, the IBR is not dispatchable because of the continuous change
of MPP. This problem can be solved by forecasting the environmental condition,
calculating the probability of power output from RES, and compensating the power
fluctuations by the ESS.
The reactive power control of IBR is more easily achieved than that of a diesel
generator. It is possible to generate its reactive power at its related value simply by
the inverter control because it has less physical constraint than a diesel generator.
Therefore, it enables to supply the reactive power even when the active power cannot
be generated. The Q–V droop control method is easy to be implemented for the IBR.
However, the reactive power from IBR must be limited depending on the amount of
active power and power factor according to the system condition [8]. The operating
range of IBR is shown in Fig. 10.10.

10.2.3.2 Non-IBR

The control of non-IBR is similar to that of diesel generator. For example, the
hydro turbine can be connected to a microgrid, and it is synchronized with the grid
frequency. Also, it has rotational inertia like a diesel generator. Therefore, the non-
IBR can be connected to the grid without considering the careful grid code. Its
operating range is the same as that of a diesel generator, as shown in Fig. 10.7.
266 K. W. Joung and J.-W. Prank

10.2.4 Energy Storage System

The ESS is important to improve the controllability of microgrid. As mentioned

previously, the ESS can be linked with IBR to reduce the variability and uncertainty
of output depending on the weather condition. In an emergency, it can be also
served as an effective reserve power source by injecting the additional active power
to the grid quickly. This type of ESS is called as the renewable integration (RI)-
ESS. On the other hand, the ESS can be used for frequency regulation (FR). In
this case, it operates directly to satisfy the requirement of a microgrid. This type
of FR-ESS plays an important role in mitigating power imbalance, and therefore
improving the flexibility and stability of microgrid with its charging and discharging
operations. The stored energy in ESS is released when the power generated is less
than consumption. Otherwise, the surplus energy is saved to the ESS from the grid.
Moreover, the ESS has a great frequency control capability, similarly to the IBR
for voltage regulation. In other words, it can enhance the stability of the microgrid
because it has a fast ramp rate than the diesel generator [9]. This property can
be applied to the droop control method in Eqs. (10.2), and (10.7) by setting the
droop constant more sensitive than that of other generators. However, the trade-
off between the stability of the system and dependency on the ESS must be also
considered. That is, if the droop constant of ESS is very different from that of
other generators, most of the power imbalance in the microgrid must be mainly
solved by the ESS. Therefore, it is important to set the appropriate droop constant
by considering the generation profile of the microgrid.
Furthermore, the ESS might be unable to participate in supporting the microgrid
with its droop control method if the amount of stored energy is not enough. Thus, the
additional control is required to regulate the state-of-charge (SOC) properly while
guaranteeing the reliable and safe operation of ESS. Assuming that the SOC is well-
regulated, the ESS can be operated in the widest range when compared to all types
of generators used in a microgrid, as shown in Fig. 10.11.

10.2.5 Frequency and Voltage Responses

To evaluate the performance of droop control, the case studies are carried out by
®
using the DIgSILENT PowerFactory software. The conventional droop control
method is implemented in the AC microgrid shown in Fig. 10.12. It has one diesel
generator, two IBRs, one ESS, and six loads.
The diesel generator of 300 kW operates as a slack bus with a droop constant of
5%. Although the rated power of IBR1 (PV generator) is 300 kW, it is assumed that
it generates the real power of 250 kW by its MPPT control. Also, the rated power
of IBR2 (wind turbine generator) is still 300 kW. However, its actual power output
is assumed to be about 200 kW without its MPPT control to have the reserve power
10 Conventional Droop Methods for Microgrids 267

Power Factor

Qmax,cap

P
V

Qmax,ind

Pmax,discharge Pmax,charge

Fig. 10.11 Operating range of ESS

Fig. 10.12 AC microgrid

with one diesel generator, two
Main Grid
IBRs, and one ESS

L1 L3 L5
ESS1

Diesel L2 IBR1 L4 IBR2 L6

for an emergency. Finally, the ESS of 300 kWh operates in an idle mode unless a
disturbance occurs. The total amount of load demand is 600 kW.
For Case 1, the amount of P and Q in the load, L5 is increased by 100 kW
and 45 kVAR, respectively, at 1 s. The P–f and Q–V droop constants of diesel
generator, IBR1, and IBR2 are set to 5%. Because the ESS is used only to regulate
the frequency, it has the steeper droop constant, which is 2%. The results are shown
268 K. W. Joung and J.-W. Prank

Fig. 10.13 Results of Case 1 with the ESS when the load, L5 is increased at 1 s: (a) frequency,
(b) bus voltage, (c) active power, (d) reactive power

in Fig. 10.13. It is observed that the frequency drops to 59.72 Hz at maximum, and is
restored to 59.78 Hz. After the load is increased at 1 s, the active power outputs from
a diesel generator, IBR2, and ESS are increased. Among them, the increased amount
of active power from ESS is highest because it has the steepest droop constant. The
IBR1 still generates the same amount of P, which is 250 kW, due to its MPPT
control, as shown in Fig. 10.10. Also, the bus voltage of IBR2 is decreased, and its
reactive power output is increased. In contrast, the bus voltage of IBR1 is increased,
and its reactive power output is reduced after the load is changed. This is because
their reactive powers are overcompensated by too high Q–V droop constant. Then,
the remaining reactive power is compensated by the diesel generator since it is a
slack generator.
For Case 2, the system configuration and load event are the same as those of
Case 1. However, it is assumed that the ESS does not operate because its SOC is in
the emergency state. The results are shown in Fig. 10.14. It is clearly observed that
the frequency stability of AC microgrid becomes worse without the operation of
ESS. In other words, the frequency drops to 59.36 Hz at maximum, and is restored
to 59.5 Hz. This is because the droop control effect of the overall AC microgrid
becomes lower. However, the reactive power outputs are not overcompensated as
10 Conventional Droop Methods for Microgrids 269

Fig. 10.14 Results of Case 2 without the ESS when the load, L5 is increased at 1 s: (a) frequency,
(b) bus voltage, (c) active power, (d) reactive power

the Q–V droop constant is reduced. Therefore, all voltages are decreased after the
load is increased.

10.3 Conventional Droop Control Method in DC Microgrid

10.3.1 Mathematical Analysis

The role of generators in DC microgrid is to supply the active power to loads and
to regulate the DC bus voltage at the same time. To operate multiple generators
together without any communication links, the droop control method is widely
used. In the DC unit circuit shown in Fig. 10.15, the injected active power from
the generator is defined by multiplying the injected current and bus voltage. Also,
the amount of power injection to the bus, i is determined by the voltage difference
between two buses as.
270 K. W. Joung and J.-W. Prank

Vi Vj
Pi , I i

Fig. 10.15 Simplified unit circuit between two DC buses

Pi = Ii · Vi (10.9)

 
Vi − Vj · Vi
Pi = (10.10)
R
where
Pi is the active power injected at ith bus.
Ii is the DC current injected at ith bus.
Vi is the DC voltage of ith bus.
R is the resistance between the ith and jth buses.
Based on the fact that the injected current (or the active power) is proportional to
the voltage difference, the V–I droop control for the DC microgrid is carried out as.

Vi = V ∗ − RV −I · Ii (10.11)

where
V* is the nominal voltage.
RV–I is the droop constant of V–I droop control.
The concept of this conventional V–I droop control method is to use the voltage
drop property by adding a virtual resistance [12]. Then, the generator will inject
more power to the grid when the magnitude of voltage is less than its nominal value,
and vice versa. The characteristic of V–I droop control by Eq. (10.11) is shown in
Fig. 10.16. Assume that the nominal values of bus voltage and current are V1 and
I1 , respectively. According to the V–I droop control, the current is increased from
I1 to I2 when the bus voltage is reduced from V1 to V2 .

10.3.2 Converter-Based Generator

Unlike the AC microgrid, most converter-based generators (CBGs) are connected

to the DC microgrid with DC–DC converter. Then, it may have a connection to
10 Conventional Droop Methods for Microgrids 271

V1
RV – I
V2

I
I1 I2 I max

Fig. 10.16 Characteristic of V–I droop method

AC Grid

DC
Load

PV ESS WT

Fig. 10.17 Example of DC microgrid

other AC systems, which require AC–DC inverters. However, most components

such as PV, WT, ESS, and DC load are connected to the DC bus by using DC–DC
converters, as illustrated in Fig. 10.17.
The controller of CBG is shown in Fig. 10.18. The inner current control loop
regulates the output current of the converter, which is determined by the voltage
control loop. Then, the droop control method can be applied before the voltage
control loop by adding the V–I droop characteristic to determine the reference signal
272 K. W. Joung and J.-W. Prank

Inner loops
I dc
V* Voltage Current DC/DC
Vdc
Control Loop Control Loop Converter

I
Droop

Fig. 10.18 Structure of the control block diagram of CBG

Fig. 10.19 Results in DC microgrid without the V–I droop control: (a) active power and (b) bus
voltage

of voltage. If there is some difference in output current, the voltage reference is

adjusted by the V–I droop control, and it is provided to inner control loops.
To verify the effectiveness of V–I droop control in the DC microgrid of Fig.
10.17, two case studies with and without droop control are carried out. It has three
CBGs. Their capacities are all set to 300 kW. The amount of DC load demand is
initially 200 kW, and it is increased from 200 kW to 300 kW at 1 s. When the
DC microgrid operates by the CBGs without the V–I droop control, the results
are shown in Fig. 10.19. The power required for the load is equally shared after
the change of load. In other words, it is allocated among three CBGs according to
the resistances between each generator and load, which are identical. Also, the bus
voltage at the load bus is slightly decreased from 1 pu after the load is increased.
Also, when it operates with the V–I droop control applied to the CBGs, the results
are shown in Fig. 10.20. The ratio of power sharing among CBGs is determined
mainly by their virtual resistances (or droop constants). The ratios of droop constant
in CBG1, CBG2, and CBG3 are set to 1, 2, and 4, respectively. Thus, the CBG1
generates the greatest active power, and the CBG3 supplies the smallest active
10 Conventional Droop Methods for Microgrids 273

Fig. 10.20 Results in DC microgrid with the V–I droop control: (a) active power and (b) bus
voltage

V
Small droop
Large droop
V*
RV – I
V1

I
I* I1 I2

Fig. 10.21 Characteristic of V–I droop control depending on the droop constant

power. According to the droop characteristic shown in Fig. 10.21, the injected
current (or active power) is increased when the reference voltage is reduced. Also,
if the droop constant is larger, the injected current is increased smaller (from I* to
I1 ). In contrast, the amount of increase in the injected current becomes larger (from
I* to I2 ) when the droop constant is smaller.

10.4 Conclusion

This chapter describes the conventional droop control methods used in both AC and
DC microgrids. For the stable operation of AC microgrid, the P–f and Q–V droop
control methods regulate the grid frequency and bus voltage by changing active
and reactive power output, respectively. Also, with the droop control, the power
274 K. W. Joung and J.-W. Prank

sharing among the generators can be achieved. Also, the operating range of each
generator type is discussed. The diesel generator has the most constraints because
of its complex physical structure. The inverter-based renewables (IBR) and energy
storage system (ESS) have the wider operating ranges than the diesel generator or
non-IBR because they have no rotational inertia and faster ramp-rate capability.
These features are verified by simulation results on the case studies.
For the DC microgrid, the V–I droop control method is discussed. Here, the
droop constant is used as a virtual resistance, which is making the additional voltage
drop. The effectiveness of droop control is also verified by a simulation test with and
without droop control. The results show that the converter-based generator (CBG)
with the droop control enables to adjust the ratio of power sharing among multiple
CBGs by changing their droop constants.

References

1. Wu, D., Tang, F., Dragicevic, T., Vasquez, J. C., & Guerrero, J. M. (2015). A control
architecture to coordinate renewable energy sources and energy storage systems in islanded
microgrids. IEEE Transactions on Smart Grid, 6(3), 1156–1166.
2. Marwali, M. N., Jin-Woo, J., & Keyhani, A. (2004). Control of distributed generation systems
- part II: Load sharing control. IEEE Transactions on Power Electronics, 19(6), 1551–1561.
3. Rocabert, J., Luna, A., Blaabjerg, F., & Rodríguez, P. (2012). Control of power converters in
AC microgrids. IEEE Transactions on Power Electronics, 27(11), 4734–4749.
4. Joung, K. W., Kim, T., & Park, J.-W. (2019). Decoupled frequency and voltage control
for stand-alone microgrid with high renewable penetration. IEEE Transactions on Industry
Applications, 55(1), 122–133.
5. Guerrero, J. M., Vasquez, J. C., Matas, J., Vicuna, L. G. d., & Castilla, M. (2011). Hierarchical
control of droop-controlled AC and DC microgrids – A general approach toward standardiza-
tion. IEEE Transactions on Industrial Electronics, 58(1), 158–172.
6. Joung, K. W., Lee, H.-J., & Park, J.-W. (2019). Assessment of maximum penetration capacity
of photovoltaic generator considering frequency stability in practical stand-alone microgrid.
Energies, 12(8), 1445.
7. Samimi, A., Kazemi, A., & Siano, P. (2015). Economic-environmental active and reactive
power scheduling of modern distribution systems in presence of wind generations: A distri-
bution market-based approach. Energy Conversion and Management, 106, 495–509.
8. Yang, Y., Enjeti, P., Blaabjerg, F., & Wang, H. (2015). Wide-scale adoption of photovoltaic
energy: Grid code modifications are explored in the distribution grid. IEEE Industry Applica-
tions Magazine, 21(5), 21–31.
9. Venkataraman, S., Ziesler, C., Johnson, P., & Kempen, S. V. (2018). Integrated wind, solar, and
energy storage: Designing plants with a better generation profile and lower overall cost. IEEE
Power and Energy Magazine, 16(3), 74–83.
10. Van de Vyver, J., De Kooning, J. D., Meersman, B., Vandevelde, L., & Vandoorn, T. L. (2016).
Droop control as an alternative inertial response strategy for the synthetic inertia on wind
turbines. IEEE Transactions on Power Systems, 31(2), 1129–1138.
11. Che, L., Shahidehpour, M., Alabdulwahab, A., & Al-Turki, Y. (2015). Hierarchical coordina-
tion of a community microgrid with AC and DC microgrids. IEEE Transactions on Smart Grid,
6(6), 3042–3051.
12. Gao, F., Kang, R., Cao, J., & Yang, T. (2019). Primary and secondary control in DC microgrids:
A review. Journal of Modern Power Systems and Clean Energy, 7(2), 227–242.
13. Lee, S. H. (2019). Adaptive droop based virtual slack control of multiple DGs in practical DC
distribution system to improve voltage profile. Energies, 12(8), 1541.
Chapter 11
Distributed Control Approaches
for Microgrids

Tohid Khalili and Ali Bidram

11.1 Introduction

In deregulated power systems, several new challenges have emerged to effectively

integrate renewable energy sources (RESs) [1] and provide a balance between
the generation and the demanded load of the consumers. The RESs, for example,
photovoltaic systems and wind turbine generators are intermittent resources where
their output is drastically probabilistic and causes some problems for the load
dispatching of the power systems [2, 3]. Moreover, inertias of RESs are low and
they don’t have enough capability to improve the power quality of the network.
Utilization of the microgrids (MGs) as small-scale power systems that can operate
in the grid-connected and islanded modes helps with the mitigation of challenges
associated with RESs [4]. The unique feature of MGs is their ability to operate
in both grid-connected and islanded modes. The MG control system plays a
critical role in accommodating its reliable operation during grid-connected and
islanded modes. The MG control system deploys a hierarchical control structure
including primary, secondary, and tertiary control levels. These control hierarchies
are responsible for the voltage and frequency control and regulation and the optimal
operation of the MG. In this chapter, these control hierarchies are elaborated. The
MG control system can either adopt a centralized or distributed control structure.
The distributed control structure has rendered more advantages compared to the
centralized one in terms of reliability and resilience [5]. In a distributed control
platform, a peer to peer communication network facilitates the communication
of information among Distributed Generators (DGs). DGs make consensus-based
control decisions to satisfy a specific MG control target. This chapter addresses

T. Khalili () · A. Bidram

Department of Electrical and Computer Engineering, University of New Mexico, Albuquerque,
NM, USA
e-mail: khalili@unm.edu; bidram@unm.edu

© Springer Nature Switzerland AG 2021 275

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_11
276 T. Khalili and A. Bidram

the distributed control of both AC and DC MGs and covers the distributed control
techniques utilized for voltage/frequency control as well as active/reactive power
sharing [6–11]. AC MGs are common types of MGs. However, DC MGs are gaining
more attention due to their inherent advantages. DC MGs facilitate the integration
of RESs with DC nature. The DC MG’s control system is less complex compared
to an AC MG’s as there is no requirement for controlling the frequency and reactive
power. Moreover, DC MGs have fewer power losses compared to AC ones [12–18].

11.2 Hierarchical Control Structure of AC and DC MGs

Considering the fact that MGs could be operated in both grid-connected and
autonomous modes, the stable and economical exploitation of the MGs requires
appropriate control approaches [19]. In this section, the hierarchical control struc-
ture of DC and AC MGs is elaborated.

11.2.1 DC MGs

By having two orders-of-magnitude, DC MGs possess high availability against

AC MGs; so, they are a good choice for mission-critical applications [13, 20]. In
addition, several challenges of the AC MGs could be solved by using the DC MGs
such as frequency synchronization, reactive power control, handling unbalanced
loads, and so on [21]. A combination of DC loads which are connected to DC
power generators by transmission or distribution systems could create a DC MG.
considering the fact that most of the demanded electric loads of the consumers are
alternative, power generators have to implement the dynamic control to supply more
load instantly respect to the voltage limitations’ criteria. Regarding the required load
of the consumers, power generators share the demanded load among each other
respect to their rated power which is called proportional load sharing. This concept
facilitates the prevention of the power generators’ overstressing and increases their
lifetime as well in the MGs. Considering the power generators’ voltage as the single
variable to control the power flow in the power systems, they have to be regulated
to gain the appropriate voltage profile in the system. To manage and control the DC
power generators, a three-level control framework is utilized. The highest and the
lowest bandwidth are for the primary and tertiary levels, respectively [22].
The primary level of the control for managing the load sharing utilizes the droop
mechanism which relates the DC generator output voltage to its current considering
the virtual resistance, RD . This virtual resistance does not have any effect on power
loss, and it just has several advantages for load sharing. Accordingly, by using the
droop mechanism, and the voltage reference, the voltage controller of the sources is
calculated as follows:
11 Distributed Control Approaches for Microgrids 277

vo∗ = vref − RD .io (11.1)

In (11.1), the voltage controller for the inner loop is shown by vo∗ . In addition,
the droop coefficient is indicated by RD . Also, the MG’s rated voltage is shown by
vref . Furthermore, the symbol of the source’s output current is io .
For voltage restoration of the MGs, the secondary control level is presented.
The secondary controller checks the voltage and compares it with the considered
voltage by the controller. A Proportional-Integral (PI) module is usually used as the
controller. Depending on all of the sources the controller corrects the voltage. By
utilizing the corrected voltage as the reference voltage, for performing the droop
mechanism, and they do not use the previous voltage reference.
To control the power flow between the main grid and the MG, the tertiary control
is proposed. Regarding the power flow, the tertiary control compares the reference
value with the power flow between the two grids. Then it updates the MG’s reference
voltage. Commonly, when the MG’s voltage increases, the DC MG transfers the
extra power and vice versa.

11.2.2 AC MGs

The MG control structure’s main duties are as follows [23–26]:

• Managing the restoration and the transients of the considered situations in the
switching mode,
• Optimizing of the MG’s operating cost,
• Controlling the power flow between the MG and the upstream grid,
• Resynchronizing the MG with the upstream grid,
• Appropriate load sharing and coordination of the distributed generators (DGs),
• Regulating the frequency and the voltage for different conditions.
These MGs’ significant roles are so important from the time scale viewpoint.
Consequently, these duties are addressed by utilizing a hierarchical control method
for answering the mentioned requirements in each level of the control. The
hierarchical control method of the MGs has three levels shown in Fig. 11.1. The
first level of the control is the fastest from the time duration point of view and helps
the MG be stable according to its frequency and the voltage profile in the islanding
process and switching between the main grid and the MG. Regarding the nonlinear
and the linear loads, it is vital for the DGs to have reactive and active power-
sharing control which is independent. Additionally, unwanted circulating currents
are avoided by using the power-sharing control. The first layer of the control has the
basic control infrastructures, which is known as zero level. This includes the internal
current and voltage control loops in DGs. The second level of the control operates
for compensating the frequency and voltage deviations made by the exploitation of
the first level of the control and helps the voltage and frequency synchronization.
278 T. Khalili and A. Bidram

Fig. 11.1 AC MG’s hierarchical control levels [8]

The tertiary control level operates at the highest level and least time scale by
managing the power flow between the main grid and the MG and helps to have
an optimal economic operation [8, 22].
11 Distributed Control Approaches for Microgrids 279

11.3 Distributed Control of DC MGs

11.3.1 DG Model in a DC MG

Figure 11.2 shows the model of a DG integrated into a DC MG. Each DG is

equipped with the internal droop and voltage control loops to accommodate a proper
voltage regulation and power sharing in the DC MG. The droop control presents the
primary control of a DC microgrid and relates DG’s voltage, vo , to its output current,
io . The droop controller stabilizes the MG voltage by proportionally sharing power
among DGs according to their ratings. The DC droop characteristic can be described
as [27, 28].

vo∗ = Vn − rd io , (11.2)

where v* o is the voltage reference for the converter and is translated to a duty cycle
which creates a similar voltage value across DG’s terminal; Vn is the voltage droop
reference; rd is the droop coefficient; rd descries DG as a virtual resistance. It should
be noted that the droop coefficients, rd , are chosen according to the DGs current
ratings, i.e.,

rd1 i1,max = · · · = rdN iN,max , (11.3)

where ik,max is the kth DG’s current rating. Equation (11.3) facilitates the propor-
tional sharing of power among DGs based on their current ratings, that is,

Fig. 11.2 DC MG’s DG model

280 T. Khalili and A. Bidram

io1 ioN
= ··· = . (11.4)
i1,max iN,max

The DG’s terminal voltage can be limited to the prespecified threshold vk, max ,
if the droop coefficients satisfy

Δvk,max
rdk ≤ . (11.5)
ik,max

11.3.2 Distributed Secondary Control of DC MGs

The DC MG primary control can maintain the voltage of MG in a stable range

through the local droop controllers. After the primary controller facilitates the MG
transition to the islanded mode, its voltage is less than the nominal MG’s voltage.
For critical loads, it is of paramount value to operate them at the nominal voltage.
To this end, secondary control is required to regulate the voltage of the critical bus
back to MG’s nominal voltage. This section of the chapter presents a distributed
DC MG secondary control as seen in Fig. 11.3. The secondary control deploys
distributed control protocols on each DG. It is assumed that DG’s control units
can communicate with each other in a distributed fashion. The goal of distributed
control protocols is to regulate the voltage of a critical bus while proportionally
sharing power among DGs according to their current ratings. This relationship is
shown in (11.4) [29].
In the distributed secondary control, a PI controller is utilized to define a leader
voltage, vleader , for the distributed control agents as

Fig. 11.3 DC MG distributed secondary control [29]

11 Distributed Control Approaches for Microgrids 281

.
   
vleader = kp vref − vcrit + ki vref − vcrit dt, (11.6)

where vref is the DC MG’s nominal voltage. vref can be also defined as a reference
voltage dictated by MG’s control center. vcrit is DC MG’s critical bus voltage. kp
and ki denote the proportional and integral control gains in the PI controller. The
vleader only needs to be shared with one DG.
The distributed secondary control protocol at each DG calculates the droop
reference, Vn in (11.2). These control protocols are extracted by defining first-order
droop control dynamics for each DG as

d d  ∗ d  
(Vnk ) = v + rdk ik,max ik,ratio , (11.7)
dt dt ok dt

where ik,ratio describes the kth DG’s current ratio as

iok
ik,ratio = . (11.8)
ik,max

The droop reference for kth DG is calculated as [29].

.
Vnk = uvk dt, (11.9)

where

uvk = Ck (δvk + δik ) , (11.10)

  
δvk = akj voj − vok + gk (vleader − vok ) , (11.11)
j ∈Nk

  
δik = akj rdk ik,max ij,ratio − ik,ratio , (11.12)
j ∈Nk

where δ vk and δ ik are the local voltage and current neighborhood tracking errors,
respectively. akj is the communication link gain between kth and ith DGs. Ck is
a control parameter. The pinning gain gk ≥ 0 is nonzero for only one DG that
possesses the leader voltage information in (11.6).
282 T. Khalili and A. Bidram

11.4 Distributed Control of AC MGs

The conventional secondary level of the MGs’ control operates a centralized control
model. Central controller rules based on accumulated data from the system and
needs a complicated and centralized communication system [30, 31]. All DGs are
controlled through a centralized controller. This negatively influences the system’s
configurability and flexibility and enhances the concerns regarding the reliability by
posing a single point of failure. The single point of failure indicates that the control
system can collapse if the mentioned central controller does not work correctly.
MGs could be supposed as the multi-agent systems where their DGs are the agents.
A distributed model of the considered communications enhances the reliability of
the system. In this controlling model, the control procedures are separated among
all of the existing DGs. Thus, the need for a centralized controller is removed, and
the control system is not impacted after the failure of a single agent.
The control model for a voltage source inverter (VSI) may change with respect
to the control targets such as frequency and voltage, or reactive and active power.
Distributed control models for autonomous MGs with voltage controlled voltage
source inverters (VCVSIs) contains internal current and voltage controller loops
which help DGs control their frequency and voltage properly.
MGs are capable of exploiting both autonomous and grid-connected states.
After the islanding operation, MGs miss the frequency and voltage support by
the upstream grid and the reactive and active power equilibrium among the whole
consumption and power generation of the system. Thus, several DGs are needed
to change their mode to the VCVSI model to make up the differences among the
consumption and power generation and to have a fast frequency and voltage support
[8].
The VCVSIs are reinforced with the primary droop controllers to keep the MG’s
frequency and voltage stability. Primary control stops the frequency and voltage
instability by preserving these amounts in safe and stable zones. The primary local
controllers’ coordinated control could be gained by the frequency and voltage droop
methods as
*
ωi = ωi∗ − DP i Pi ,
∗ , (11.13)
vo,magi = Ei∗ − DQi Qi

where Ei∗ and ωi∗ indicate the references for the primary control. The droop
coefficients are indicated by the DQi and DPi . These droop coefficients are chosen
regarding each of VCVSIs reactive and active power ratings. Also, Qi and Pi are the
obtained reactive and active power considering the DG’s terminal. The VCVSI’s
angular frequency is presented by ωi which the primary control determines. v* o,magi
is the reference signal for the magnitude of the terminal voltage of inverter which is
considered for controlling the VCVSI’s internal voltage controller.
Additionally, the primary control accommodates seamless frequency and voltage
support for the MGs. Droop controllers can only stabilize the MG voltage and
11 Distributed Control Approaches for Microgrids 283

frequency. They fail to maintain these quantities at their nominal values. To this
end, secondary control is needed to force the frequency and the voltage back to their
nominal values. The secondary control of AC MGs includes voltage and frequency
controls. ωi∗ in (11.13) is selected by the secondary frequency control to synchronize
VCVSIs’ angular frequency to MG’s nominal angular frequency, i.e. ωi → ωref .
Furthermore, it finds the VCVSI’s output active powers considering

P1 PNV
= ··· = , (11.14)
Pmax 1 Pmax NV

where NV is the VCVSIs’ number in the autonomous MG. The active power rating
of ith VCVSI is indicated by the Pmaxi . So, DPi is selected regarding VCVSIs’ active
power ratings as

DP 1 P1 = · · · = DP NV PNV . (11.15)

In (11.13), Ei∗ is selected by the secondary voltage control to regulate VCVSI’s

terminal voltages to a reference voltage value, that is, vo,magi →vref . If vref is adjusted
to the reference voltage of the MG vnominal , the VCVSIs’ output voltage magnitude
synchronizes to the MG’s reference voltage. Meanwhile, vref could be selected to
synchronize a vital bus of the MG’s voltage magnitude to the vnominal .
The other target of the secondary voltage control could be finding the VCVSI’s
output reactive powers considering

Q1 Q NV
= ··· = , (11.16)
Qmax 1 Qmax NV

where NV is the VCVSIs’ number in the autonomous MG. The reactive power rating
of ith VCVSI is indicated by the Qmaxi . So, DQi is selected regarding VCVSIs’
reactive power ratings as

DQ1 Q1 = · · · = DQNV QNV . (11.17)

The characteristics of the voltage droop in (11.13) are built by assuming an MG

that is totally inductive. Thus, (11.16) and (11.17) could be exactly fulfilled in a
totally inductive MG. However, reactive power ratio mismatches could be seen in
the noninductive MGs.

11.4.1 Frequency Control

This section elaborates on how to attain the two aims of the secondary frequency
control by utilizing just a single distributed control procedure for each of DGs.
The distributed control procedure is designed to get the DGs’ frequency, ωi ,
284 T. Khalili and A. Bidram

synchronized with the nominal frequency, ωref . Also, it is designed to share the
active power fairly among DGs with respect to their related power ratings.
Differencing with the frequency-droop characteristic results in

ω̇i∗ = ω̇i + DP i Ṗi = ui , (11.18)

where ui is an additional variable to be modeled. In (11.18), a dynamic model

for calculating the ωi∗ from ui is presented. The additional control variable has
to be modeled subject to synchronizing the frequency of the DGs to the nominal
frequency ωref . Regarding (11.18), an MG’s secondary frequency control having N
DGs is changed to a synchronization problem which its objective is to synchronize
a first-order linear multi-agent system as.
⎧
⎪
⎪ ω̇1 + DP 1 Ṗ1 = u1 ,
⎪
⎨ ω̇2 + DP 2 Ṗ2 = u2 ,
. (11.19)
⎪
⎪ ..
⎪
⎩
ω̇N + DP N ṖN = uN .

To get synchronized, it is considered that DGs could send and receive information
among themselves via the determined connection graph Gr (representing communi-
cation network). The additional controls ui are selected respect to the DG’s self-data,
and the data of existing nearby DGs in the connection graph as [5, 19].
⎛ ⎞
       
ui = −cf ⎝ aij ωi − ωj + bi ωi − ωref + aij DP i Pi − DPj Pj ⎠ ,
j ∈Ni j ∈Ni
(11.20)

where the coupling gain is indicated by cf ∈ R. It is considered that the pinning gain
bi ≥ 0 is nonzero for just a single DG which has the nominal frequency ωref .
Figure 11.4 illustrates the secondary frequency control’s schematic with respect
to the distributed consensus-based control. As shown in Fig. 11.4, the ωi∗ , control
input, is indicated as
.
ωi∗ = ui dt. (11.21)

cf as the coupling gain has an effect on the convergence speed of frequency

restoration.
11 Distributed Control Approaches for Microgrids 285

Fig. 11.4 The schematic of the secondary frequency control in a distributed fashion [5, 19]

11.4.2 Voltage Control

The voltage control’s aim is to select suitable Ei∗ , control inputs, in (11.13) to get
vo,magi , VCVSIs’ voltage magnitudes, synchronized to the reference voltage vref ,
and find the VCVSIs’ generated reactive powers with respect to their reactive power
ratings, that is, they meet (11.16) and (11.17). It is notable that (11.13)‘s voltage-
droop characteristic is modeled considering an MG contains transmission lines that
are inductive. Thus, little differences between voltage magnitudes and the reactive
power ratios could be seen in MGs that have not fully inductive transmission lines.
The VCVSIs’ voltage magnitudes, vo,magi , synchronization is the same as the
synchronization of the vodi . For the concurrent reactive power and VCVSIs’
voltage control, the rapid dynamics of internal current and voltage controllers are
disregarded. As a result, v* o,magi and vodi can be considered equal. Due to the stated
fact, the (11.13)‘s differentiated voltage-droop specification is written as.

Ėi∗ = v̇odi + DQi Q̇i = vvi , (11.22)

where vvi is an additional variable to be modeled. In (11.22), a dynamic system for

calculating the Ei∗ from vvi is presented. The additional control variable has to be
modeled subject to synchronizing the VCVSI’s voltage magnitudes to the reference
voltage vref and satisfying the (11.16). Regarding (11.22), an MG’s voltage control
having NV VCVSIs is changed to a problem wherein its objective is to synchronize
a first-order linear multi-agent system as.
⎧
⎪
⎪ v̇od1 + DQ1 Q̇1 = vv1 ,
⎪
⎨ v̇od2 + DQ2 Q̇2 = vv2 ,
. (11.23)
⎪
⎪ ..
⎪
⎩
v̇odN V + DQNV Q̇NV = vvNV .
286 T. Khalili and A. Bidram

Fig. 11.5 The schematic of distributed secondary voltage control [5, 19]

To get synchronized, it is considered that VCVSIs could send and receive

information among themselves via the determined connection graph Gr. The
additional controls vvi are selected with respect to the VCVSI’s self-data, and the
data of existing nearby VCVSIs on the connection graph as [5, 19].
⎛ ⎞
       
vvi =−cv ⎝ aij vodi −vodj + bi vodi −vref + aij DQi Qi −DQj Qj ⎠ ,
j ∈Ni j ∈Ni
(11.24)

where the coupling gain is indicated by cv ∈ R. It is considered that the pinning gain
bi ≥ 0 is nonzero for just a single VCVSI which has the data of the nominal voltage
vref . It could be observed that all VCVSI voltage amplitudes get synchronized to vref
by the (11.24). Moreover, (11.24) allocates VCVSIs’ reactive power with respect to
their reactive power ratings. The nominal voltage vref could be adjusted to the MG’s
reference voltage for synchronizing the VCVSI voltage magnitudes to the reference
voltage or could be selected to control the voltage magnitude of a vital bus of the
MG.
Figure 11.5 shows the schematic of the presented voltage control. As illustrated
in Fig. 11.5, the Ei∗ , control input, is indicated as
.
Ei∗ = vvi dt. (11.25)

11.5 Conclusion and Future Trend

This chapter addressed the hierarchical control structure of MGs. The primary,
secondary, and tertiary control levels were discussed in detail. The distributed
11 Distributed Control Approaches for Microgrids 287

control techniques for both AC and DC MGs were covered. Although the reliability
of the MG control system is enhanced through a distributed control structure
increases, it is exposed to cyberattacks due to the deployment of communication
and control devices. The future trends in the area of MG control will focus on
improving the reliability, security, data integrity, and efficiency of the control
system. As one of the major challenges, the cybersecurity of the MG control
system should be addressed to tackle both denial-of-service and false data injection
attacks. Moreover, with the deployment of intelligent electronic devices, the MG
communication network transmits a vast amount of data among different entities.
To reduce the burden on the communication network, event-triggered distributed
control techniques should be deployed to ensure that the data is only transmitted
when the control system is triggered by an event.

References

1. Jafari, A., Khalili, T., Ganjehlou, H. G., & Bidram, A. (Feb. 2020). Optimal integration of
renewable energy sources, diesel generators, and demand response program from pollution,
financial, and reliability viewpoints: A multi-objective approach. Journal of Cleaner Produc-
tion, 247, 119100.
2. Hagh, M. T., & Khalili, T. (2019). A review of fault ride through of PV and wind renewable
energies in grid codes. International Journal of Energy Research, 43(4).
3. Khalili, T., Nojavan, S., & Zare, K. (Mar. 2019). Optimal performance of microgrid in the
presence of demand response exchange: A stochastic multi-objective model. Computers and
Electrical Engineering, 74, 429–450.
4. Khalili, T., Jafari, A., Abapour, M., & Mohammadi-Ivatloo, B. (Feb. 2019). Optimal battery
technology selection and incentive-based demand response program utilization for reliability
improvement of an insular microgrid. Energy, 169, 92–104.
5. Bidram, A., Davoudi, A., & Lewis, F. L. (Aug. 2014). A multi-objective distributed control
framework for islanded microgrids. IEEE Trans. Industrial informatics, 10, 1785–1798.
6. T. Khalili, A. Jafari, and E. Babaei, “Scheduling and siting of storages considering power peak
shaving and loss reduction by exchange market algorithm,” in IEEE Proceedings 2017 Smart
Grid Conference, SGC 2017, 2018, vol. 2018.
7. Khalili, T., Hagh, M. T., Zadeh, S. G., & Maleki, S. (Jul. 2019). Optimal reliable and
resilient construction of dynamic self-adequate multi-microgrids under large-scale events. IET
Renewable Power Generation, 13(10), 1750–1760.
8. Bidram, A., & Davoudi, A. (Dec. 2012). Hierarchical structure of microgrids control system.
IEEE Transactions on Smart Grid, 3, 1963–1976.
9. Hatziargyriou, N., Asano, H., Iravani, R., & Marnay, C. (2007). Microgrids. IEEE Power &
Energy Magazine, 5, 78–94.
10. R. H. Lasseter, “Microgrid,” in Proc. IEEE Power Eng. Soc. winter meeting, vol. 1, New York,
2002, pp. 305–308.
11. Lopes, J. A. P., Moreira, C. L., & Madureira, A. G. (May 2006). Defining control strategies
for microgrids islanded operation. IEEE Transactions on Power Apparatus and Systems, 21,
916–924.
12. Kwasinski, A., & Onwuchekwa, C. N. (March 2011). Dynamic behavior and stabilization of
dc microgrids with instantaneous constant-power loads. IEEE Trans. Power Electronics, 26,
822–834.
288 T. Khalili and A. Bidram

13. Kwasinski, A. (2011). Quantitative evaluation of dc microgrids availability: Effects of system

architecture and converter topology design choices. IEEE Transactions on Power Electronics,
26(3), 835–851.
14. Sanchez, S., & Molinas, M. (2014). Degree of influence of system state transition on the
stability of a dc microgrid. IEEE Transactions on Smart Grid, 30, 2535–2542.
15. Farhadi, M., & Mohammed, O. (2015). Adaptive energy management in redundant hybrid dc
microgrid for pulse load mitigation. IEEE Transactions on Smart Grid, 6, 54–62.
16. Inthamoussou, F. A., Queralt, J. P., & Bianchi, F. D. (2013). Control of a supercapacitor energy
storage system for microgrid applications. IEEE Transactions on Energy Conversion, 28, 690–
697.
17. Xu, G., Xu, L., Morrow, D. J., & Chen, D. (2012). Coordinated dc voltage control of wind
turbine with embedded energy storage system. IEEE Transactions on Energy Conversion, 27,
1036–1045.
18. Tummuru, N. R., Mishra, M. K., & Srinivas, S. (2015). Dynamic energy management of hybrid
energy storage system with high-gain PV converter. IEEE Transactions on Energy Conversion,
30, 150–160.
19. Bidram, A., Nasirian, V., Davoudi, A., & Lewis, F. L. (2017). Cooperative synchronization
in distributed microgrid control. Springer International Publishing. ISBN:978-3-319-50807-8
(Print) and 978-3-319-50808-5 (Online).
20. H. Ikebe, “Power systems for telecommunications in the IT age,” in Proc. of IEEE INTELEC,
2003, pp. 1–8.
21. Balog, R. S., Weaver, W., & Krein, P. T. (2012). The load as an energy asset in a distributed dc
Smartgrid architecture. IEEE Transactions on Smart Grid, 3, 253–260.
22. Guerrero, J. M., Vasquez, J. C., Matas, J., de Vincuña, L. G., & Castilla, M. (2011).
Hierarchical control of droop-controlled AC and DC microgrids – A general approach toward
standardization. IEEE Transactions on Industrial Electronics, 58, 158–172.
23. Guerrero, J. M., Matas, J., Vicuna, L. G. D., Castilla, M., & Miret, J. (2007). Decentralized con-
trol for parallel operation of distributed generation inverters using resistive output impedance.
IEEE Transactions on Industrial Electronics, 54, 994–1004.
24. Guerrero, J. M., Vicuna, L. G. D., Matas, J., Castilla, M., & Miret, J. (2005). Output
impedance design of parallel-connected UPS inverters with wireless load-sharing control.
IEEE Transactions on Industrial Electronics, 52, 1126–1135.
25. Katiraei, F., Iravani, M. R., & Lehn, P. W. (2005). Microgrid autonomous operation during and
subsequent to islanding process. IEEE Trans. Power Del, 20, 248–257.
26. Katiraei, F., & Iravani, M. R. (2005). Power management strategies for a microgrid with
multiple distributed generation units. IEEE Transactions on Power Apparatus and Systems,
21, 1821–1831.
27. Chen, Y. K., Wu, Y. C., Song, C. C., & Chen, Y. S. (2013). Design and implementation of
energy management system with fuzzy control for dc microgrid systems. IEEE Transactions
on Power Electronics, 28(4), 1563–1570.
28. Li, Y., Vilathgamuwa, D. M., & Loh, P. C. (2004). Design, analysis, and real-time testing of a
controller for multibus microgrid system. IEEE Transactions on Power Electronics, 19, 1195–
1204.
29. Poudel, B., Mustafa, A., Modares, H., & Bidram, A. (2020). Detection and mitigation of cyber-
threats in the DC microgrid distributed control system. International Journal of Electrical
Power & energy Systems, 120. https://doi.org/10.1016/j.ijepes.2020.105968.
30. Barklund, E., Pogaku, N., Prodanovic´, M., Hernandez-Aramburo, C., & Green, T. C. (2008).
Energy management in autonomous microgrid using stability-constrained droop control of
inverters. IEEE Transactions on Power Electronics, 23, 2346–2352.
31. Borup, U., Blaabjerg, F., & Enjeti, P. N. (2001). Sharing of nonlinear load in parallel-connected
three-phase converters. IEEE Transactions on Industry Applications, 37, 1817–1823.
Chapter 12
On Control of Energy Storage Systems
in Microgrids

Yu Wang, Sidun Fang, and Yan Xu

12.1 Introduction

The traditional energy structure highly depending on fossil fuels such as coal and oil
has become a major concern of climate change and air pollution in modern society.
These environmental concerns and energy crises of fossil fuels lead to the rapid
development of renewable energy technologies [1, 2]. A large number of renewable
energy sources (RESs) together with energy storage systems (ESSs) have been
penetrated into existing power systems especially distribution sides through power
electronics interfaces. The power generation becomes more and more decentralized
other than the conventional centralized generation, which results in the concept
of distributed generation. In power distribution systems, a cluster of demand-side
loads and distributed energy resources can be connected and disconnected from
the main grid to operate in grid-connected or islanded mode. These small-scale
power systems are named as microgrids. The original idea of microgrids emerges at
the beginning of this century, which aims to benefit the integration of distributed
generators (DGs) and enhance grid resilience [3]. In grid-connected mode, the
microgrid can be viewed as one entity, which exchanges power with the main grid to
realize energy trading and provide ancillary services [4, 5]. In the islanded mode, the
major target of the microgrid is to maintain local generation/demand balance, while
providing stable and high-quality power supply [6]. According to the current flow,
microgrids are usually categorized into alternative current (AC), direct current (DC),
and hybrid AC/DC microgrids [7, 8]. The residential, commercial, and industry
microgrids have been built all over the world to benefit the renewable penetrations,
grid resilience enhancement as well as traffic electrification [9–11].

Y. Wang () · S. Fang · Y. Xu

School of Electrical and Electronics Engineering, Nanyang Technological University, Singapore,
Singapore

© Springer Nature Switzerland AG 2021 289

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_12
290 Y. Wang et al.

Energy storage systems are relatively new units in microgrids or power distribu-
tion systems following in the wake of increased installation of renewable energy
generation in the twenty-first century. One typical feature of renewable energy
generation is the inherent nature of uncertainties. For example, the power generation
of photovoltaic (PV) fluctuates along with the daily solar irradiance variations, and
the power generation of wind energy is influenced largely by the seasonal variation
of wind at the installed location [11, 12]. PV and wind generation cannot ensure
constant power supply as their stochastic and intermittent characteristics, which
will influence the stability, reliability, and power quality of power systems [13].
As various problems caused by the penetration of solar panels and wind turbines
have become very common, ESSs are considered as one promising solution for
such problems in microgrid systems [14]. Typically, ESSs can be categorized by
the form of stored energy into five groups: mechanical, chemical, electrochemical,
electrical, and thermal energy storage [15]. The energy storage devices belonging
to each classification are shown in Fig. 12.1. Among all energy storage categories,
electrochemical energy storage with different kinds of batteries is the most widely
used in low-voltage electrical systems like microgrids.
In microgrids, the ESSs can be installed in a centralized way by the utility
company at the point of common coupling (PCC) in the substation [16]. Besides,
the ESSs can also be integrated in a distributed way such as plug-in electric
vehicles (PEV) and building/home ESSs [17, 18]. Depending on the operation
modes of microgrids, the ESSs can be operated for various functionalities. In
the grid-connected mode of microgrids, the ESSs with various capacities can be
applied for different kinds of ancillary services. According to the time dimension,
the primary frequency/voltage control and stability enhancement will be concerned

Fig. 12.1 Classification of energy storage technologies according to energy form [15]
12 On Control of Energy Storage Systems in Microgrids 291

Fig. 12.2 Timescales of energy storage systems for grid ancillary services

in the millisecond to second timeframe, the secondary frequency/voltage control,

and reserve service will be concerned in the minute to hour timeframe, peak
shaving, and energy trading will be concerned in the hour to day timeframe
[16–24]. Figure 12.2 shows the timescales of multiple usages of energy storage
systems according to timescale. In the islanded mode of microgrids, the primary
objective of ESSs is to balance the generation and demand mismatch, which is
particularly important for microgrids with high penetration of RESs. Therefore, the
SoC balancing control among ESSs becomes a commonly adopted strategy, in order
to effectively utilized ESSs capacity and maintain an uninterrupted power supply of
the microgrid. Besides, the ESSs can also be utilized for other functionalities in
islanded microgrids such as frequency/voltage control, power quality improvement,
renewable generation compensation, and economic dispatch.

12.2 Overview of Energy Storage Systems

12.2.1 Characteristics of ESSs

In microgrid applications, the main technical characteristics of ESSs include power

density, energy density, life cycle (lifetime), energy efficiency, and self-discharge.
According to [25, 26], these metrics are further explained as follows:
(a) Energy density (Wh/kg) refers to the energy to weight ratio of one energy
storage device. Energy density indicates the capability of continuous energy
supply over a period of time. The ESS with higher energy density can discharge
energy for a longer period.
(b) Power density (W/kg) refers to the power to weight ratio of one energy storage
device. Power density indicates the capability of ESS to provide instantaneous
power. The ESS with higher power density can discharge a larger amount of
power when needed.
(c) Energy efficiency (%) refers to the ratio of released energy to stored energy. It
also indicates the output and input electricity regardless of self-discharge.
292 Y. Wang et al.

(d) Life cycle refers to the number of times where ESS can provide the designed
energy level after each recharge. One cycle means one full charge and discharge
operation for the battery.
(e) Self-discharge refers to the ratio of initial stored energy to dissipated energy
during a nonuse time.
Batteries and super-capacitors are two typical ESSs in microgrids, which are
further discussed in this paper. Compared to super-capacitors, batteries usually
have high energy density but low power density. Among all kinds of batteries
(i.e., Pb–Acid, Ni–Cd, Ni–MH, Lithium-ion, Li–polymer, NaS, and VRB), Ni–MH
has the lowest energy efficiency. Besides, battery ESSs usually have a short life
cycle, while Pb–Acid has the shortest. On the contrary, super-capacitors usually
have low energy density but high power density. Although super-capacitors have a
long lifetime, their self-discharge is quite high. It means super-capacitors are very
suitable for short-term storage applications, considering their high self-discharge
and high energy efficiency. In some cases, hybrid ESSs are applied which can
combine the positive features of batteries and super-capacitors. However, more
complicated power electronics interfaces and control systems are required for
the power management of hybrid ESSs. Subsequently, the basic power electronic
interface for typical battery ESSs and the battery management system is further
discussed.

12.2.2 Power Electronic Interface

The power converters serve as the interface between the battery ESSs and the
microgrid. The interface should ensure grid codes and electrical standards are
satisfied while providing ancillary services to the electricity market [27–29]. The
charging and discharging of the battery ESS and power flow of the power conversion
system are dependent on the microgrid requirement. In the meantime, the operating
points (e.g., terminal voltage, charging/discharging current, state-of-charge (SOC))
of the batteries are regulated, which provides efficient charging/discharging and
protects the health of the battery ESS.
Depending on the application scenarios, various topologies are proposed to
connect battery ESS into the microgrids [27–29]. A simple way is to connect the
battery bank directly to the DC link of the DC/AC converter among battery ESS and
microgrid. However, as the battery voltages vary with the SOC, the DC-link voltage
will be influenced. In this condition, the grid interfacing converters should have
the capability to provide a wide range of operating DC voltage. In the meantime,
the wide range of modulation indices is also required to address the DC voltage
variation, which will cause lower efficiency and higher harmonics. One solution
to control the DC-link voltage is to add a bidirectional DC/DC converter between
the battery bank and the DC link of the grid interfacing converters, as shown in
Fig. 12.3. This conversion system design enables that the battery terminal voltage
12 On Control of Energy Storage Systems in Microgrids 293

Fig. 12.3 A general power electronic interface of grid-connected battery ESS

connecting to the dc link can be adjusted flexibly. In addition, the bidirectional

DC/DC converters also contribute to eliminating the low-order harmonics current
in the battery bank, which will benefit the total lifetime of the battery bank.
Particularly, when the battery ESS is connected to a medium voltage (MV)
microgrid system, a transformer should be included in between. Battery cells can
be connected in parallel and series at the low-voltage side to build up a battery ESS
from hundreds of kWs to tens of MWs. The transformer is installed to boost the
voltage from hundreds of Volts to tens of kVs.

12.2.3 Battery Management System

The battery ESS usually consists of hundreds to thousands of battery cells. The
battery management system (BMS) plays a vital role to manage battery ESS
[30, 31]. From an electrical perspective, three major objectives of BMS are: (i)
monitoring and estimating the battery states; (ii) protecting the batteries; and (iii)
managing the charging/discharging of the batteries.
Battery monitoring and state estimation. The battery terminal voltage, current,
and surface temperature are key parameters that can be directly measured from
sensors. The state estimation is needed to estimate other important indices, including
SOC or depth of discharge (DOD), state of health (SOH). SOC or DOD of batteries
can be estimated according to the operating condition of battery current, voltage,
and temperature [32]. SOH can be calculated by the extent of abuse and performance
degradation of batteries [33].
Battery protection and safety control. The protection of the battery should
consider both thermal and electrical parameters. When the faults occurs in the
battery, the system should diagnose the fault and prevent or mitigate the damage
or injury to people caused by overcurrent, overvoltage, overcharge, overdischarge,
over-temperature, under-temperature, and so on [34].
Battery cell balancing control. BMS should have the cell equalization function
including equalizing charging/discharging to maintain the SOC balance among cells
294 Y. Wang et al.

Table 12.1 ESS Control Strategies in Islanded Microgrids

Multiple ESSs RESs and ESSs
Feature Droop Consensus Hybrid ESSs Decentralized Hierarchical
Ref. [37–43] [44–50] [51–56] [57–60] [61–66]

[35]. This strategy can prolong the entire battery ESS health by avoiding the over-
usage of certain cells.

12.3 ESS Control Strategies in Islanded Microgrids

In islanded microgrids, the major objective is to maintain power supply while

enhancing system stability and resilience. The ESSs play key roles in compensating
the short-term power mismatch as well as long-term energy management. This
section will review the typical control strategies of ESSs in islanded microgrids
by functionalities. The terminologies from hierarchical control architecture in
microgrids is adopted [36]. This chapter mainly focuses on the system-level, real-
time, coordinated control of ESSs. The control strategies of battery cells and power
electronics in ESSs, as well as the system-level optimization of ESSs, are not
covered. The coordinated control of islanded microgrids with ESSs summarized
in this section are listed in Table 12.1 below.

12.3.1 Coordinated Control of Multiple ESSs

The energy limitation can be ignored for DG units supplied by fossil fuels, as it
is assumed that the fuel supply is of a sufficient amount. However, the energy
stored in ESSs (e.g., the SOC of lithium-ion battery) is limited, which should be
considered in the microgrid level control system. The SOC balancing becomes a
commonly adopted strategy for multiple ESSs in islanded microgrids, due to the
following reasons: (1) the power mismatch between RESs and loads can be buffered
by an islanded microgrid with balanced SOC among ESSs; (2) the prevention of
unintentionally switch-off batteries caused by their energy depletion or saturation;
and (3) the overcharged or overdischarged of a certain battery that damage the
battery health can also be avoided.
Droop based SOC Balancing. The droop based SOC balancing control has been
widely discussed in state-of-the-art. The main idea is by involving SOC into the
droop coefficients. The droop coefficients will be changed adaptively according to
the SOC of each ESS so that SOC levels in the microgrid will be balanced.
Sufficient work has been conducted to investigate the droop based SOC balancing
methods. In [37], the authors propose an adaptive droop control for SOC balancing
12 On Control of Energy Storage Systems in Microgrids 295

of distributed ESSs in DC microgrids. Later on, this method is improved with a

double-quadrant SOC-based droop control method in [38]. In [39], a multifunctional
droop control scheme is proposed, which can provide localized SOC balancing. In
[40], a decentralized power management strategy is reported for a hybrid ESS with
SOC recovery and autonomous bus voltage restoration. Considering the different
capacities of multiple ESSs, the authors in [41] proposes an improved droop control.
A voltage scheduling droop control for SOC balancing of distributed ESSs in
DC microgrids is reported in [42]. Fuzzy droop control is designed for balancing
stored energy in distributed ESSs in [43]. The main concern of decentralized SOC
balancing control is the lack of global coordination. The system power quality
is sacrificed (e.g., frequency and voltage deviations due to droop control). It is
hard to design localized control for frequency and voltage restoration without any
communication.
Consensus-based SOC Balancing. The distributed SOC balancing control has
been included in the secondary level of microgrid control due to the limitations
above. Based on neighboring communication, the secondary control objectives can
be achieved together with SOC balancing. The power quality in the islanded micro-
grid system is improved. Therefore, researchers are working on SOC balancing
among a group of ESSs in the secondary level by distributed control methods [44–
50]. The voltage or frequency deviation problem caused by primary droop control
can be effectively avoided.
In the early stage, the authors in [44] propose a distributed multi-agent coopera-
tive control for frequency regulation and energy level balancing in AC microgrids.
Later, this method is extended for heterogeneous ESS in DC microgrids in [45] and
by using sliding mode control for SOC balancing in [46]. In later research, [47]
proposes a SOC balancing for ESSs in grid-connected AC microgrids, while [48]
proposes a distributed secondary control including SOC balancing in islanded AC
microgrids. Recently, researchers are trying to use novel control strategies to further
improve the control performance. In [49], a distributed finite-time consensus control
for heterogeneous ESSs in AC microgrids is proposed considering cyber-physical
implementation. In [50], a distributed cyber-resilient secondary control is presented
for SOC balancing among multiple ESSs.
Coordination of Hybrid ESSs. Energy storages with different characteristics
can collaboratively operate for power management in microgrids. Typically, super-
capacitors are units with high power density but low energy density, while lithium-
ion batteries have low power density but high energy density [50]. The philosophy
is to control HESSs to deal with power variation events with different frequency
bandwidth. Research efforts have been made on coordinated control strategies of
HESSs. In [52], a droop-based decentralized power-sharing strategy is proposed for
hybrid ESSs in DC microgrids. In [53], a hierarchical control scheme for hybrid
ESSs in DC microgrids is introduced. In [54], a robust frequency regulation with
hybrid ESSs in multi-area microgrids is proposed. In recent research, researchers
are also making efforts to including SOC balancing using consensus in hybrid ESS
in DC microgrids with different topologies [55, 56].
296 Y. Wang et al.

12.3.2 Coordinated Control of RESs and ESSs

RESs like PV and wind turbines are widely integrated into microgrids. The
coordination problem of RESs and ESSs have raised much attention in the research
community. Many research works have been conducted to provide solutions for
RESs and ESSs coordination in islanded microgrids.
Fully Decentralized Control. Various kinds of droop control curves are designed
for localized coordination of RESs and ESSs in microgrids. In [57], a real power-
sharing control strategy is proposed for islanded AC microgrids by using frequency
bus signaling to achieve power allocation in a decentralized manner. In [58], a
smooth switching droop control is introduced to coordinate RESs and ESSs in
islanded AC microgrids. In [59], a real power-sharing method by the dynamic droop
factors to control charge/discharge power allocation between the superconducting
magnetic energy storage and the battery is proposed. Decentralized power control in
AC islanded microgrids with RESs and ESSs is also reported in [60]. The frequency
and voltage restoration are still the major concerns of decentralized coordination.
Hierarchical Control. Influenced by the idea of hierarchical control of traditional
power grids. Hieratical control has been proposed for the coordination of RESs
and ESSs in microgrids. The centralized secondary control can be adopted for
frequency/voltage restoration. The master-slave hierarchical control scheme for DG
units has been proposed by some research works. In [61], a master-slave coordinated
control strategy for DGs and the ESSs for islanded operation is proposed. In [62],
a localized model predictive control (MPC) for DG units and rule-based centralized
coordination is introduced. On the other hand, the DG units can be governed by
a peer-to-peer hierarchical control scheme. In [63], hierarchical power control of
various units in DC microgrids is proposed. Hierarchical control of the hybrid
ESSs in DC microgrids is proposed in [64]. In [65], a centralized architecture
is proposed for real power curtailment of generation to avoid overcharge of the
ESSs, load shedding for avoiding deep discharge of the ESSs, and balancing of
the SOCs among all ESSs to prevent battery degradation. The major concerns
of the centralized architecture are that the central controller will have large
computation and communication burdens for large-scale systems, and it is prone to
communication failures and central device failures. In [66], a distributed coordinated
control strategy is proposed for managing the multiple ESSs to balance the power
generation and load demand while minimizing the system power loss during the
charging/discharging.

12.4 ESS Control Strategies in Grid-Connected Microgrids

In grid-connected microgrids or power distribution networks, the major objective

is to maintain acceptable power quality while providing grid ancillary services.
The ESSs play key roles in improving the short-term power quality requirement
12 On Control of Energy Storage Systems in Microgrids 297

Table 12.2 ESS Control Strategies in Grid-Connected Microgrids

Voltage regulation Frequency regulation
Feature Decentralized Hierarchical Primary Secondary
Ref [75, 76] [77–81] [85–92] [93–95]

as well as satisfying long-term grid dispatch order. This section presents the control
strategies of ESSs for voltage and frequency regulation in grid-connected microgrids
and power distribution networks. The literature reviewed in this section is listed in
Table 12.2 below.

12.4.1 Voltage Regulation

A power distribution network can be viewed as a grid-connected microgrid with

RESs, ESSs, and local load demands. Since the last few years, the fast developments
of renewable technology have increased the penetration level of DGs in distribution
networks. However, it results in two kinds of voltage problems: voltage rise and
drop issues. In a community with a high penetration level of rooftop PVs, PV
generation in the day time can significantly exceed load demand. The reverse power
flow from the customer end to the utility grid can potentially result in voltage rise
issues in the distribution network [67–70]. In the meantime, the large share of PEV
will cause additional power demand, which will cause voltage drop issues in the
distribution network [71–74]. So voltage limits violation could occur during both
peak PV generation and load demand periods, which leads to poor power quality
and even equipment failure.
Decentralized Control. The decentralized/local voltage control by power invert-
ers of ESSs can be usually divided into P-dependent and V-dependent methods [75,
76]. In the P-dependent method, the real/reactive power output of ESSs is dependent
on the real power generation of PV units. This control is also known as ramp-
rate control to mitigate voltage fluctuations [75]. In the V-dependent method, the
power output from the ESS is dependent on the local bus voltage, that is, voltage
droop control [76]. However, the system statuses are changing rapidly considering
the high uncertainty of renewables and plug-and-play ability of PEVs, It makes the
decentralized control itself hard to deal with the voltage regulation in such a system.
Even with the optimized droop settings and set points from upper level optimization,
the system voltages can still exceed the operational limits. In this condition, the real-
time coordination between local controllers is required.
Hierarchical Control. As fully decentralized control cannot provide well coor-
dination of ESSs in microgrids, the hierarchical control has been proposed which
incorporates upper level coordination. The centralized coordinated control scheme
of distributed ESSs with tap changer transformers to mitigate voltage rise in a
system with high PV penetration is addressed in [77]. Many works have been
298 Y. Wang et al.

conducted for the distributed coordination of ESSs in power distribution networks.

Multi-agent-based voltage control for clustered active distribution networks by
means of ESSs is proposed in [78]. In [79], a local and distributed coordinated
control of distributed ESSs is proposed for voltage and SOC management. In [80],
a distributed MPC method for battery ESSs is introduced for voltage regulation in
a high renewable penetrated distribution network. In [81], a hierarchical control
framework is introduced to coordinate several groups of virtual energy storage
(thermostatically controlled loads), to control network loading and voltage in power
distribution networks.

12.4.2 Frequency Regulation

Traditionally, the synchronous generators from bulk power plants function to

balance power mismatch between generation and load consumption. However,
the system inertia and the frequency reserves are decreased due to the large
integration of RESs and subsequent replacement of rotational generators. Generally,
when a disturbance occurred, the power system with less inertia and frequency
regulation reserves will suffer from more severe frequency deviations. Under
this circumstance, ESSs have been viewed as a good candidate for solving such
problems. As the capability to contribute to power system operation and control,
ESSs have been gradually deployed in modern power systems [82–84].
Primary frequency control. As the fast response speed of ESSs, they have been
viewed as suitable candidates for system primary frequency control, from both
industry and academic perspective [85–87]. In [88], a fuzzy logic frequency control
strategy is presented by utilizing the large capacity distributed PV systems and
PEVs. The sizing problem of ESSs for grid inertial response and primary frequency
reserve is studied in [89]. From an economic point of view, the planning of battery
ESSs for primary frequency control is investigated in [90]. In [91], the battery life
issues during the operation of a grid-tied lithium-ion battery ESS for the provision
of primary frequency regulation is studied. In [92], a data-driven method is reported
to predict the real-time power fluctuations in power systems, which is used by ESSs
for power system frequency support.
Secondary Frequency Control. In high renewable penetrated power systems,
an emerging but important task to be addressed in the increasing requirement of
automatic generation control (AGC) capacity [93]. The centralized ESSs installed
by the grid operator is able to provide AGC services. However, the limited capacity
of utility ESSs makes them more suitable for primary control. In the meantime,
the demand-side ESSs such as PEVs and home ESSs have a high potential for
the secondary frequency regulation services. Although the demand-side ESSs are
usually of small capacity, the number of them will be very large. Herein, it is needed
to design a coordinated control scheme to aggregate ESSs. In [94], a coordinated
control framework is designed for large-scale PEV charging aggregators and
ESSs to provide the frequency regulation. In [95], an MPC is proposed for load
12 On Control of Energy Storage Systems in Microgrids 299

aggregators and utility-scale ESSs for power system frequency regulation. In [18], a
distributed finite-time consensus control scheme is proposed to aggregate demand-
side ESSs for power system secondary control.

12.5 Research Trends and Opportunity

Multifunctional ESSs. In recent studies, it is found that the net economic benefit of
ESSs is not exploited with only one single grid ancillary service. For example, only
1–50% of a battery’s lifetime capacity will be taken, if it only participates in primary
services. Therefore, there are commercial opportunities from the ESS providing
additional functions and services. Recent studies have suggested more applications,
that is, stack service, would reduce the idling duration of ESS, increasing the
economic value and payback period [96]. In future work, multiple time-scale
coordinated control scheme of ESSs is to be developed. The multifunctional
operation of ESSs without conflicting each functionality is the bottleneck of this
design.
Large-scale Control and Optimization of ESSs. In recent years, a distributed
control framework has proved its merits for real-time coordination without limita-
tions of centralized control. The computation and communication burden are fairly
shared by the distributed controllers. It also offers more robustness to single point
and communication failures. For large-scale distributed ESSs integration in power
distribution level networks, the distributed coordinated control and optimization
is recommended. The distributed algorithms considering the dynamics, functions,
economics, health, as well as other working conditions, are to be investigated.
Both networked dynamic control and networked optimization techniques are highly
demanded in order to address large-scale penetration of ESSs in the power networks.
In the cyber networks, the distributed algorithms are realized on multi-agent systems
(MASs) with embedded systems [97]. The realization of distributed algorithms is
also an important issue to promote future development.
Cybersecurity of ESSs. Large-scale distributed control and optimization of ESSs
highly relies on the performance of communication networks as well as its security.
Yet so far, the cybersecurity issues of ESSs have not raised much attention.
However, in future power systems with large-scale networked controlled ESSs, the
security problem should be taken into account. To protect the security of networked
controlled ESSs under lethal cyberattacks, proper defense mechanisms are to be
investigated. Three aspects of defense mechanisms can be investigated in future
works [98]. (1) The prevention methods to protect the ESS from cyberattacks.
(2) The detection and isolation methods to detect the cyberattacks and isolate
the corrupted subsystems. (3) The resilience algorithms to overcome the negative
impact of an attack and operate the system closer to its normal state.
300 Y. Wang et al.

12.6 Conclusion

This chapter introduces the control and application of ESSs in microgrid systems.
The characteristics of energy storage techniques, power electronic interfaces, and
battery management systems are introduced. A comprehensive review of ESSs
in both islanded microgrids and grid-connected microgrids has been conducted.
The future research roadmaps of multifunctional ESSs, distributed control of
optimization of ESSs, and cybersecurity issues of ESSs are discussed.

References

1. Bose, B. K. (2010). Global warming: Energy, environmental pollution, and the impact of power
electronics. IEEE Industrial Electronics Magazine, 4(1), 6–17.
2. Panwar, N. L., Kaushik, S. C., & Kothari, S. (2011). Role of renewable energy sources in
environmental protection: A review. Renewable and Sustainable Energy Reviews, 15(3), 1513–
1524.
3. Lasseter, R. H. (2002). Microgrids. Proc. IEEE Power Engineering Society Winter Meeting,
305–308.
4. Mengelkamp, E., et al. (2018). Designing microgrid energy markets: A case study: The
Brooklyn microgrid. Applied Energy, 210, 870–880.
5. Wang, Y., Xu, Y., Tang, Y., & Xu, Y. L. (2020). A distributed control scheme of thermostatically
controlled loads for the building-microgrid community. IEEE Transactions on Sustainable
Energy, 11(1), 350–360.
6. Rocabert, J., et al. (2012). Control of power converters in AC microgrids. IEEE Transactions
on Power Electronics, 27(11), 4734–4749.
7. Liu, X., Wang, P., & Loh, P. C. (2011). A hybrid AC/DC microgrid and its coordination control.
IEEE Transactions on Smart Grid, 2(2), 278–286.
8. Wang, P., et al. (2013). Harmonizing AC and DC: A hybrid AC/DC future grid solution. IEEE
Power and Energy Magazine, 11(3), 76–83.
9. Ustun, T. S., Ozansoy, C., & Zayegh, A. (2011). Recent developments in microgrids and
example cases around the world—A review. Renewable and Sustainable Energy Reviews,
15(8), 4030–4041.
10. Wang, Y., Xu, Y., & Tang, Y. (2019). Distributed Aggregation Control of Grid-Interactive Smart
Buildings for Power System Frequency Support. Applied Energy, 251(1).
11. Fang, S., Wang, Y., Gou, B., & Xu, Y. (2020). Towards future green maritime transportation:
Seaport microgrids and all-electric ships. IEEE Transactions on Vehicular Technology, 69(1),
207–219.
12. Dincer, I. (2000). Renewable energy and sustainable development: A crucial review. Renewable
and Sustainable Energy Reviews, 4(2), 157–175.
13. Arul, P. G., Ramachandaramurthy, V. K., & Rajkumar, R. K. (2015). Control strategies for a
hybrid renewable energy system: A review. Renewable and Sustainable Energy Reviews, 42,
597–608.
14. Vazquez, S., Lukic, S. M., Galvan, E., Franquelo, L. G., & Carrasco, J. M. (Dec. 2010). Energy
storage systems for transport and grid applications. IEEE Trans. Industrial Electronics, 57(12),
3881–3895.
15. Electrical energy storage project team, “Electrical energy storage.” International Electrotech-
nical Commission, 2011.
16. Gyuk, I., Johnson, M., Vetrano, J., Lynn, K., et al. (2013). Grid energy storage. U.S Department
of Energy.
12 On Control of Energy Storage Systems in Microgrids 301

17. Carpinelli, G., Celli, G., Mocci, S., Mottola, F., Pilo, F., & Proto, D. (June 2013). Optimal
integration of distributed energy storage devices in smart grids. IEEE Transactions on Smart
Grid, 4(2), 985–995.
18. Wang, Y., et al. (July 2019). Aggregated energy storage for power system frequency control: A
finite-time consensus approach. IEEE Transactions on Smart Grid, 10(4), 3675–3686.
19. Energy Market Authority, “Energy Storage Programme,” 2015. Available https://
www.ema.gov.sg/Energy_Storage_Programme.aspx.
20. Hill, C. A., Such, M. C., Chen, D., Gonzalez, J., & Grady, W. M. (2012). Battery energy
storage for enabling integration of distributed solar power generation. IEEE Transactions on
Smart Grid, 3(2), 850–857.
21. Subburaj, A. S., Pushpakaran, B. N., & Bayne, S. B. (2015). Overview of grid connected
renewable energy based battery projects in USA. Renewable and Sustainable Energy Reviews,
45, 219–234.
22. Zhang, Y., Ren, S., Dong, Z. Y., Xu, Y., Meng, K., & Zheng, Y. (2017). Optimal placement
of battery energy storage in distribution networks considering conservation voltage reduction
and stochastic load composition. IET Generation, Transmission & Distribution, 11(15), 3862–
3870.
23. Fang, S., et al. (2019). Optimal hierarchical Management of Shipboard Multibattery Energy
Storage System Using a data-driven degradation model. IEEE Transactions on Transportation
Electrification, 5(4), 1306–1318.
24. Dong, C., Jia, H., Xu, Q., Xiao, J., Xu, Y., et al. (2018). Time-delay Stability Analysis for
Hybrid Energy Storage System with Hierarchical Control in DC Microgrids. IEEE Trans.
Smart Grid, 99(99).
25. Grid integration of large-capacity renewable energy sources and use of large-capacity electrical
energy storage. International Electrotechnical Commission, 2012.
26. Ibrahim, H., Ilinca, A., & Perron, J. (2008). Energy storage systems-characteristics and
comparisons. Renewable Sustainable Energy Reviews, 12(5), 1221–1250.
27. Wang, G., et al. (2016). A review of power electronics for grid connection of utility-scale
battery energy storage systems. IEEE Transactions on Sustainable Energy, 7(4), 1778–1790.
28. Yilmaz, M., & Krein, P. T. (May 2013). Review of battery charger topologies, charging power
levels, and infrastructure for plug-in electric and hybrid vehicles. IEEE Transactions on Power
Electronics, 28(5), 2151–2169.
29. Grainger, B. M., Reed, G. F., Sparacino, A. R., & Lewis, P. T. (June 2014). Power electronics
for grid-scale energy storage. Proceedings of the IEEE, 102(6), 1000–1013.
30. Lu, L., et al. (2013). A review on the key issues for lithium-ion battery management in electric
vehicles. Journal of Power Sources, 226, 272–288.
31. Xiong, R., Li, L., & Tian, J. (2018). Towards a smarter battery management system: A critical
review on battery state of health monitoring methods. Journal of Power Sources, 405, 18–29.
32. Xiong, B., et al. (2014). Extended Kalman filter method for state of charge estimation of
vanadium redox flow battery using thermal-dependent electrical model. Journal of Power
Sources, 262, 50–61.
33. Wei, Z., et al. (2018). Online state of charge and state of health estimation for a lithium-ion
battery based on a data–model fusion method. Energies, 11(7), 1810.
34. Sidhu, A., Izadian, A., & Anwar, S. (2014). Adaptive nonlinear model-based fault diagnosis of
Li-ion batteries. IEEE Transactions on Industrial Electronics, 62(2), 1002–1011.
35. Lee, Y., & Cheng, M. W. (2005). Intelligent control battery equalization for series connected
lithium-ion battery strings. IEEE Transactions on Industrial Electronics, 52(5), 1297–1307.
36. Guerrero, J. M., et al. (2010). Hierarchical control of droop-controlled AC and DC
microgrids—A general approach toward standardization. IEEE Transactions on Industrial
Electronics, 58(1), 158–172.
37. Lu, X., et al. (2013). State-of-charge balance using adaptive droop control for distributed
energy storage systems in DC microgrid applications. IEEE Transactions on Industrial
Electronics, 61(6), 2804–2815.
302 Y. Wang et al.

38. Lu, X., et al. (2014). Double-quadrant state-of-charge-based droop control method for dis-
tributed energy storage systems in autonomous DC microgrids. IEEE Transactions on Smart
Grid, 6(1), 147–157.
39. Sun, X., et al. (2016). A multifunctional and wireless droop control for distributed energy
storage units in islanded AC microgrid applications. IEEE Transactions on Power Electronics,
32(1), 736–751.
40. Xu, Q., et al. (2017). A decentralized power management strategy for hybrid energy storage
system with autonomous bus voltage restoration and state-of-charge recovery. IEEE Transac-
tions on Industrial Electronics, 64(9), 7098–7108.
41. Wu, Q., et al. (2018). SoC balancing strategy for multiple energy storage units with different
capacities in islanded microgrids based on droop control. IEEE Journal of Emerging and
Selected Topics in Power Electronics, 6(4), 1932–1941.
42. C. Li, et al. ”Voltage scheduling droop control for State-of-Charge balance of distributed energy
storage in DC microgrids.“ 2014 IEEE International Energy Conference (ENERGYCON).
IEEE, 2014.
43. N. Díaz, et al. ”Fuzzy droop control loops adjustment for stored energy balance in distributed
energy storage system.“ 2015 9th International Conference on Power Electronics and ECCE
Asia (ICPE-ECCE Asia). IEEE, 2015.
44. Morstyn, T., Hredzak, B., & Agelidis, V. G. (2015). Distributed cooperative control of
microgrid storage. IEEE Transactions on Power Systems, 30(5), 2780–2789.
45. Morstyn, T., Hredzak, B., & Agelidis, V. G. (2016). Cooperative multi-agent control of
heterogeneous storage devices distributed in a DC microgrid. IEEE Transactions on Power
Systems, 31(4), 2974–2986.
46. Morstyn, T., Savkin, A. V., Hredzak, B., & Agelidis, V. G. (2018). Multi-agent sliding mode
control for state of charge balancing between battery energy storage systems distributed in a
DC microgrid. IEEE Transactions on Smart Grid, 9(5), 4735–4743.
47. Cai, H., & Hu, G. (2016). Distributed control scheme for package-level state-of-charge
balancing of grid-connected battery energy storage system. IEEE Transactions on Industrial
Informatics, 12(5), 1919–1929.
48. Y. Wang, et al. ”Distributed Secondary Control of Energy Storage Systems in Islanded
AC Microgrids.“ 2018 Asian conference on energy, power and transportation electrification
(ACEPT). IEEE.
49. Wang, Y., et al. (2020). Distributed control of heterogeneous energy storage systems in islanded
microgrids: Finite-time approach and cyber-physical implementation. International Journal of
Electrical Power & Energy Systems, 119, 105898.
50. Deng, C., et al. (2020, in press). Distributed resilient control for energy storage Systems in
Cyber-Physical Microgrids. IEEE Transactions on Industrial Informatics, 1.
51. Jing, W., et al. (2016). Battery-supercapacitor hybrid energy storage system in standalone DC
microgrids: A review. IET Renewable Power Generation, 11(4), 461–469.
52. Xu, Q., et al. (2016). A decentralized dynamic power sharing strategy for hybrid energy storage
system in autonomous DC microgrid. IEEE Transactions on Industrial Electronics, 64(7),
5930–5941.
53. Xiao, J., Wang, P., & Setyawan, L. (2015). Hierarchical control of hybrid energy storage system
in DC microgrids. IEEE Transactions on Industrial Electronics, 62(8), 4915–4924.
54. C. Ju, Y. Tang, Y. Wang, “Robust Frequency Regulation with Hybrid Energy Storage
Systems in Islanded Microgrids,” 2018 Asian conference on energy, power and transportation
electrification (ACEPT), Oct. 2018.
55. Lin, P., et al. (2019). A semi-consensus strategy toward multi-functional hybrid energy storage
system in DC microgrids. IEEE Transactions on Energy Conversion.
56. B. Wang et al., ”Consensus-based Control of Hybrid Energy Storage System with a Cascaded
Multiport Converter in DC Microgrids,“ IEEE Transactions on Sustainable Energy. doi: https:/
/doi.org/10.1109/TSTE.2019.2956054.
57. Wu, D., et al. (2014). Autonomous active power control for islanded ac microgrids with
photovoltaic generation and energy storage system. IEEE Transactions on Energy Conversion,
29(4), 882–892.
12 On Control of Energy Storage Systems in Microgrids 303

58. Wu, D., et al. (2014). A control architecture to coordinate renewable energy sources and energy
storage systems in islanded microgrids. IEEE Transactions on Smart Grid, 6(3), 1156–1166.
59. Li, J., et al. (2017). Design/test of a hybrid energy storage system for primary frequency control
using a dynamic droop method in an isolated microgrid power system. Applied Energy, 201,
257–269.
60. de Matos, J. G., Silva, F. S. F. E., & Ribeiro, L. A. D. S. (2015). Power control in AC isolated
microgrids with renewable energy sources and energy storage systems. IEEE Transactions on
Industrial Electronics, 62(6), 3490–3498.
61. Kim, J. Y., et al. (2010). Cooperative control strategy of energy storage system and
microsources for stabilizing the microgrid during islanded operation. IEEE Transactions on
Power Electronics, 25(12), 3037–3048.
62. Y. Wang, K. T. Tan, and P. L. So. ”Coordinated control of battery energy storage system
in a microgrid.“ 2013 IEEE PES Asia-Pacific power and energy engineering conference
(APPEEC). IEEE, 2013.
63. Jin, C., et al. (2013). Implementation of hierarchical control in DC microgrids. IEEE
Transactions on Industrial Electronics, 61(8), 4032–4042.
64. Xiao, J., Wang, P., & Setyawan, L. (2015). Hierarchical control of hybrid energy storage system
in DC microgrids. IEEE Transactions on Industrial Electronics, 62(8), 4915–4924.
65. Diaz, N. L., et al. (2016). Centralized control architecture for coordination of distributed
renewable generation and energy storage in islanded AC microgrids. IEEE Transactions on
Power Electronics, 32(7), 5202–5213.
66. Xu, Y., et al. (2014). Cooperative control of distributed energy storage systems in a microgrid.
IEEE Transactions on Smart Grid, 6(1), 238–248.
67. Tonkoski, R., Turcotte, D., & El-Fouly, T. H. M. (Jul. 2012). Impact of high PV penetration on
voltage profiles in residential neighborhoods. IEEE Transactions on Sustainable Energy, 3(3),
518–527.
68. Y. Wang, T. Zhao, C. Ju, Y. Xu, P. Wang. “Two-level distributed voltage/Var control of
aggregated PV inverters in distribution networks,” IEEE Transactions on Power Delivery, Nov.
2019.
69. Wang, Y., Xu, Y., Tang, Y., Syed, M. H., Guillo-Sansano, E., & Burt, G. (Feb. 2019).
Decentralised-distributed hybrid voltage regulation of power distribution networks based on
power inverters. IET Generation, Transmission & Distribution, 13(3), 444–451.
70. Wang, Y., Syed, M. H., Guillo-Sansano, E., Xu, Y., & Burt, G. (Nov. 2019). Inverter-
based voltage control of distribution networks: A three-level coordinated method and power
hardware-in-the-loop validation. IEEE Transactions on Sustainable Energy.
71. C. Zhang, C. Chen, J. W. Sun, P. W. Zheng, et al., “Impacts of electric vehicles on the transient
voltage stability of distribution network and the study of improvement measures,” in Proc.
2014 IEEE PES eAsia-Pacific power and energy engineering conference (APPEEC), pp. 1–6,
Dec. 2014.
72. Luo, X., & Chan, K. W. (2014). Real-time scheduling of electric vehicles charging in low-
voltage residential distribution systems to minimise power losses and improve voltage profile.
IET Generation, Transmission and Distribution, 8(3), 516–529.
73. Richardson, P., Flynn, D., & Keane, A. (2011). Optimal charging of electric vehicles in low-
voltage distribution systems. IEEE Trans. Power Systems, 27(1), 268–279.
74. Wang, Y., et al. (2019). A coordinated voltage control scheme using electric vehicle chargers
in low-voltage distribution networks. Electric Power Systems Research, 168, 218–227.
75. Alam, M. J. E., Muttaqi, K. M., & Sutanto, D. (2014). A novel approach for ramp-rate control
of solar PV using energy storage to mitigate output fluctuations caused by cloud passing. IEEE
Transactions on Energy Conversion, 29(2), 507–518.
76. Kabir, M. N., Mishra, Y., Ledwich, G., Dong, Z. Y., & Wong, K. P. (2014). Coordinated
control of grid-connected photovoltaic reactive power and battery energy storage systems to
improve the voltage profile of a residential distribution feeder. IEEE Transactions on Industrial
Informatics, 10(2), 967–977.
304 Y. Wang et al.

77. Liu, X., et al. (2012). Coordinated control of distributed energy storage system with tap
changer transformers for voltage rise mitigation under high photovoltaic penetration. IEEE
Transactions on Smart Grid, 3(2), 897–906.
78. Bahramipanah, M., Cherkaoui, R., & Paolone, M. (2016). Decentralized voltage control of
clustered active distribution network by means of energy storage systems. Electric Power
Systems Research, 136, 370–382.
79. Wang, Y., Tan, K. T., Peng, X., & So, P. L. (2016). Coordinated control of distributed energy
storage systems for voltage regulation in distribution networks. IEEE Transactions on Power
Delivery, 31(3), 1132–1141.
80. Meng, K., et al. (2015). Cooperation-driven distributed model predictive control for energy
storage systems. IEEE Transactions on Smart Grid, 6(6), 2583–2585.
81. Meng, K., et al. (2017). Distributed control of thermostatically controlled loads in distribution
network with high penetration of solar PV. CSEE Journal of Power and Energy Systems, 3(1),
53–62.
82. Kottick, D., Blau, M., & Edelstein, D. (1993). Battery energy storage for frequency regulation
in an island power system. IEEE Transactions on Energy Conversion, 8(3), 455–459.
83. Roberts, B. P., & Sandberg, C. (2011). The role of energy storage in development of smart
grids. Proceedings of the IEEE, 99(6), 1139–1144.
84. Wen, S., et al. (2019). Proactive frequency control based on ultra-short-term power fluctuation
forecasting for high renewables penetrated power systems. IET Renewable Power Generation.
85. M. Swierczynski, et al. ”Field tests experience from 1.6 MW/400kWh Li-ion battery energy
storage system providing primary frequency regulation service.“ IEEE PES ISGT Europe 2013.
IEEE, 2013.
86. Y. Wang, H. Li, Y. Xu, and Y. Tang, “System frequency regulation in Singapore using dis-
tributed energy storage systems,” 2017 Asian Conference on Energy, Power and Transportation
Electrification (ACEPT), Oct. 2017.
87. Oudalov, A., Chartouni, D., & Ohler, C. (2007). Optimizing a battery energy storage system
for primary frequency control. IEEE Transactions on Power Systems, 22(3), 1259–1266.
88. Datta, M., & Senjyu, T. (2013). Fuzzy control of distributed PV inverters/energy storage
systems/electric vehicles for frequency regulation in a large power system. IEEE Transactions
on Smart Grid, 4(1), 479–488.
89. Knap, V., et al. (2015). Sizing of an energy storage system for grid inertial response and primary
frequency reserve. IEEE Transactions on Power Systems, 31(5), 3447–3456.
90. Zhang, Y. J., et al. (2016). Profit-maximizing planning and control of battery energy storage
systems for primary frequency control. IEEE Transactions on Smart Grid, 9(2), 712–723.
91. Stroe, D. I., et al. (2016). Operation of a grid-connected lithium-ion battery energy storage
system for primary frequency regulation: A battery lifetime perspective. IEEE Transactions on
Industry Applications, 53(1), 430–438.
92. S. Wen, et al. “Identification of real-time power fluctuations based on LSTM recurrent neural
network and a case study on Singapore power system,” IEEE Trans. Industrial Informatics,
2019.
93. Donadee, J., & Wang, J. (2014). AGC signal modeling for energy storage operations. IEEE
Transactions on Power Systems, 29(5), 2567–2568.
94. Zhong, J., et al. (2014). Coordinated control for large-scale EV charging facilities and energy
storage devices participating in frequency regulation. Applied Energy, 123, 253–262.
95. Wen, G., et al. (2015). Frequency regulation of source-grid-load systems: A compound control
strategy. IEEE Transactions on Industrial Informatics, 12(1), 69–78.
96. F. Teng, M. Aunedi, R. Moreira, et al, “Business case for distributed energy storage,” in Proc.
24th International Conference on Electricity Distribution (CIRED), Glasgow, 2017.
97. Wang, Y., et al. (2020). A distributed control scheme of microgrids in energy internet and its
multi-site implementation. IEEE Transactions on Industrial Informatics. in press.
98. Dibaji, S. M., Pirani, M., Flamholz, D. B., Annaswamya, A. M., Johanssonb, K. H., &
Chakraborttyc, A. (2019). A systems and control perspective of CPS security. Annual Reviews
in Control, 47, 394–411.
Chapter 13
Microgrid Stability Definition, Analysis,
and Examples

Hossein Shayeghi, Hamzeh Aryanpour, Masoud Alilou, and Aref Jalili

13.1 Introduction

Microgrids, as a new type of network in power distribution systems, have been

developed with the advent of distributed generation to increase system reliability
and address economic and environmental issues [1]. To build a microgrid, renewable
energy is usually applied as much as possible so inverter interfaced distributed
generations are used widely in the microgrid, which makes the operating character-
istics of microgrid quite different from the traditional grid. Similar to synchronous
generators that play the most key roles in the traditional grid, inverter interfaced
distributed generations are also as the same important roles in a microgrid. However,
the dynamic process of microgrid stability is more complicated due to the decrease
in the inertia of the microgrid in islanded mode.
Due to the microgrid operation mode, its stability problems are categorized
into grid-connected and islanded stability issues. In the grid-connected mode, the
stability issues of the microgrid in transient and small signal studies are focused
more on voltage stability. The researches on small signal stability of islanded
microgrid have drawn much attention. Because maintaining power supply and load
balance are very vital by microgrid itself. In the islanded mode, microgrid stability is
categorized into the voltage stability and frequency stability in both the transient and
small signal studies. A linearized model of the network is used for the analysis of
small signal stability in the microgrid. Also, the time domain and eigenvalue-based
analysis and droop gain optimization are the common methods to study small-
disturbance stability.

H. Shayeghi () · H. Aryanpour · M. Alilou

Energy Management Research Center, University of Mohaghegh Ardabili, Ardabil, Iran
A. Jalili
Department of Electrical Engineering, Islamic Azad University, Ardabil, Iran

© Springer Nature Switzerland AG 2021 305

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_13
306 H. Shayeghi et al.

Microgrid distribution systems comprise different distributed energy resources

(DERs) and operate in isolation from or while connected to the main grid. In
islanded mode, the voltage and frequency of microgrids should be controlled
by different DERs [2] given the numerous disturbances and load uncertainties
experienced by these systems in the real world. Such problems reduce the stability
of voltage and frequency owing to the consequent imbalance between power
generation and demand. Voltage and frequency control cannot be guaranteed
by traditional controllers in the presence of different DERs, disturbances, and
uncertainties because these controllers are not always suitable for all operating
conditions. A controller that robustly performs across a wide range of system
operating conditions is necessary for an islanded microgrid (a requirement that has
been addressed with the development of hierarchical control). As indicated in the
IC/ISO 62264 standard [3], the goal of fast-response primary control is to adjust the
frequency and amplitude of voltage references against any variations in sources and
loads. Secondary control is intended to regulate frequency and voltage to ensure
that they are of acceptable levels and to compensate for the imbalance in the
two. In tertiary control, realizing acceptable and cost-effective operating conditions
necessitates economic dispatch and power optimization.
A review of the literature indicated that many droop control structures have
been presented for voltage and frequency compensations, as has been done in
[4–6]. The imbalance between total demand and generation causes voltage and
frequency to deviate from admissible limits. To improve voltage [7] and frequency
[8] responses, researchers implement load shedding as a separation scheme. In [9],
a method was developed on the basis of a real-time load shedding computation
for the simultaneous recovery of voltage and frequency. In [10], a battery inverter
with rapid response was examined to control voltage and frequency in an islanded
microgrid. Li et al. presented active and reactive power management for an inverter-
based microgrid to enhance the system’s transient stability [11]. In [12], the
coordinated control of inverter-based distributed generators and distributed energy
storage was provided to evaluate dynamic stability in the microgrid. In islanded
microgrids, matching demand and generation, regulating voltage and frequency,
and sharing power between sources are critical issues. Correspondingly, researchers
recommended an approach to operating islanded microgrids based on voltage source
inverters that operate in current-regulated mode [13]. To consider nonlinearity and
uncertainties in nonconventional energy sources, the researchers provided a robust
energy management system that uses predictive control theory as mathematical
bases. This method was verified as robust against variations in the wind power
generated by islanded microgrids [14]. In [15], a battery energy storage system was
suggested to ensure the independence of system frequency from the inertia of syn-
chronous generators, and an active/reactive power droop controller was employed as
the voltage controller. In [16], the authors recommended the coordination of DERs
and demand responses. In this coordination, loads are classified on the basis of their
13 Microgrid Stability Definition, Analysis, and Examples 307

associated sensitivity values, and total operating cost in grid-connected mode is

minimized using a particle swarm optimization algorithm. Tang et al. proposed a
hybrid method based on the combination of a voltage/frequency droop controller
and an active/reactive droop controller to improve small-signal stability [17].
One of the most important procedures in the simultaneous control of voltage and
frequency is the complete modeling of microgrids which facilitates the design of
acceptable controllers. The study, in which this modeling was conducted, increases
running time because of rising complexity, experts cannot design a controller with
good performance. As a solution, Kunjumhummed modeled a power system based
on a multi-machine structure by relating system equations to one another using a
network impedance matrix [18]. A survey of microgrid modeling approaches was
conducted in [19] to detail dynamic models of main microgrid components. The
discussion of the models reflected that some adaptive and robust controllers are
used to develop secondary voltage and frequency control schemes; these controllers
include PI [20], sliding mode [21], robust mixed H2/H∞ [22], robust H∞ and μ-
synthesis [23] and fuzzy logic [24] controllers.
For the islanded operation of microgrids, it is necessary to develop a new
simulation model for the simultaneous control of voltage and frequency as a
secondary control scheme. Correspondingly, this chapter develops a secondary
control simulation model that is based on a multi-machine structure and expresses
the relationship between different units through an admittance matrix. The model
exhibits a highly accurate and fast simulation and considers all kinds of renewable
energy sources and energy storage devices. On the basis of the model, a new
adaptive fuzzy proportional integral-derivative (AFPID) controller is developed to
ensure effective control in the presence of renewable energy sources and robust
performance over various operating conditions. The proposed controller is equipped
with two independent parts that work in parallel with others. The first part is a
PID controller that guarantees stability and eliminates steady-state errors in nominal
operating conditions. The second part is a fuzzy logic-based mechanism that tunes
the gains of the PID controller. This fuzzy component increases flexibility against
uncertainties and nonlinearity. The design of the AFPID controller is inspired
by the tuning of controller output based on nonlinear and voltage responses.
However, precisely designing a fuzzy controller is more complex and based on
human knowledge. Thus, the nondominated sorting improved differential evolution
(NSIDE) algorithm is used to optimally design the controller. The advantages of the
NSIDE algorithm include its easy understandability, robustness, and requirement
for only a few parameters in promptly adjusting and handling non-differentiable,
nonlinear, and multi-model functions. It also presents diverse solutions and effective
uniform distributions of Pareto solutions. In the optimization process, the Analytical
Hierarchy Process (AHP) mechanism is used to ensure the selection of the best
solutions. The AHP is a multi-criteria decision-making method that helps decision-
makers for addressing a complex problem with multiple conflicting and subjective
criteria. The method adopted in this work is the combination of the NSIDE
algorithm, fuzzy concepts, and the AHP increases convergence speed and optimal
search capabilities. The algorithm is employed to avoid the need for trial-and-
308 H. Shayeghi et al.

error processes, augment the performance of the AFPID controller, and accordingly
ascertain that it comprises parameters such as PID gains, fuzzy rule bases, and fuzzy
membership functions. The AFPID controller has a simple structure and robustly
performs against uncertainties and nonlinearity. For the validation of effectiveness,
the controller is applied in an actual microgrid equipped with three combined heat
and power (CHP) units and three wind units. The controller is installed in the CHP
units and the dynamic behavior of the system is tested under different scenarios.
Its performance is then compared with that of the conventional fuzzy PID and
classical PID controllers based on performance indices. The results show that the
proposed controller exhibits robust performance under different load changes and
disturbances and is therefore recommended for use in the actual microgrids as a
secondary control scheme.
• Motivations and contributions
The literature review showed that an adaptive method has been implemented
by researchers to enable simultaneous voltage/frequency control in an islanded
microgrid. Major factors that prompted the present work are the development of a
novel model and the design of a simultaneous fuzzy PID controller. Most traditional
controllers do not appropriately function against nonlinearity and uncertainty issues.
The current scheme involves employing the NSIDE algorithm to design the AFPID
controller which demonstrates to be acceptably productive. The main motivation of
the study is the design of the AFPID controller as a secondary control scheme for
microgrids. The proposed control strategy is applied in an islanded microgrid for
simultaneous controlling of voltage and frequency.
Therefore, the contributions of this work are summarized below.
• The development of a novel multi-machine structure-based simulation model for
the study of the dynamic behavior of microgrids.
• The design of an AFPID voltage and frequency controller for islanded micro-
grids.
• The use of the NSIDE algorithm as an effective tool for the optimization and
automatic design of an AFPID controller for islanded microgrids.
• The effective minimization of the integral of squared time multiplied by square
error (ISTSE) against uncertainty and nonlinearity.

13.2 Microgrid Modeling

In the studies of microgrid, researchers develop many methods and software

programs but they also encounter certain problems such as long simulation times
and changes to microgrid structures because of expansion initiatives. In this chapter,
a multi-machine model for the study of microgrid dynamics is developed. The
model is suitable for use in MATLAB/Simulink software and can be considered
multiple types of distributed generators, various loading conditions, and different
13 Microgrid Stability Definition, Analysis, and Examples 309

Synchronous generator
differential equation

Asynchronous generator
differential equation
Voltage Transmission network model

Reactive compensator IDQ=YBUS× VDQ

differential equation

Current
Other system components
differential equation

Fig. 13.1 The main structure of the proposed model

compensators. The model also links various load models for microgrids through
a transmission admittance matrix. Figure 13.1 illustrates the main structure of the
proposed model.

13.2.1 The Fixed-Speed Wind Turbine Model

A fixed-speed wind turbine consists of three parts, namely, a turbine blade, a

squirrel-cage asynchronous generator, and a capacitor bank. Turbine blades convert
kinetic energy into mechanical energy which is then converted into electrical
energy by squirrel-cage asynchronous generators. Capacitor banks are installed
near the terminal of asynchronous generators to compensate for the use of reactive
power. The establishment of an asynchronous generator model depends on the
level of details in the model; that is, the model that needs to be constructed
hinges on whether it is of fifth-order, third-order, or one-order type. These models
are described in [25, 26]. The present research uses a third-order asynchronous
generator model.
In the aerodynamic component shown in Fig. 13.2, Cp, i , λi , β i , and Trot, i are the
aerodynamic power coefficient, speed ratio, pith angle, and aerodynamic torque,
respectively, in a fixed-speed wind turbine i. This component can be modeled
according to Eqs. (13.1)–(13.3) [26].

Ri ωrot,i
λi = (13.1)
υ
310 H. Shayeghi et al.

Speed ratio Aerodynamic power

calculation coefficient
Eq. (7.1) calculation Eq. (7.2)
,

, Aerodynamic , Mechanical Asynchronous

torque calculation model generator
Eq. (7.3) Eq. (7.4) model

Fig. 13.2 Simple model of constant speed wind turbo generator [26]

 
π (λi − 3)
Cp,i = (0.44 − 0.0167βi ) sin − 0.0184 (λi − 3) βi (13.2)
15 − 0.3βi

 
1 ρi π Ri2 υ 3
Trot,i = Cp,i (13.3)
ωrot,i 2

where ρ i , Ri , υ, and ωrot, i denote the air density, rotor radius, wind speed, and rotor
speed, respectively. The mechanical part of the asynchronous generator is described
by Eq. (13.4) as follows:
 
dωrot,i ω0 ωrot,i
= Trot,i − Te,i − Di (13.4)
dt 2Hi ω0

Here, ω0 , Te, i , Hi , and Di represent the synchronous speed, electrical torque,

inertia, and damping coefficient of the asynchronous generator, respectively.

13.2.2 CHP Model

CHP is the most important energy technology with respect to microgrids; it is

efficient and presents environmental, economic, and reliability benefits. These
advantages have prompted the increased use of this energy technology. In CHP
units, gas turbines are mechanically connected to a synchronous generator. In other
words, energy conversion in the units is carried out by a synchronous generator. In
this chapter, a gas turbine is used in the CHP unit that has been modeled in [27],
with some modifications. In a gas turbine, a droop controller is typically used to
regulate frequency but such a controller is inappropriate for islanded microgrids.
This deficiency requires the development of a new extra control loop through which
13 Microgrid Stability Definition, Analysis, and Examples 311

DTurb
Pref
AFPID Controller
(Frequency)
min
- Vmax
+
W - 1/R Droop 1 1 -
+ + Pm
Control Select T1.s+1 T2.s+1
Wref
1 + + Vmin
- K
+
T3.s+1

AT

Fig. 13.3 Gas turbine model

Qg
Vref VRmax
PF ref PF Error PI PF Control
Measurement
+ AFPID Ka 1 Efd
Pg + +
- (Voltage) Ta.s+1 -
1 Te.s
Vt -1 Voltage Control
VRmin
T.s+1
Control Select
Vf Kfs s+1/Tf3
+ Ke
s+1/Tf1 s+1/Tf3 +

Se

Fig. 13.4 Modified AC5A IEEE excitation model

secondary control is implemented. In consideration of this issue, therefore, we

incorporated an extra frequency control loop into our gas turbine. Figure 13.3 shows
the modified gas turbine structure of the CHP unit.
The IEEE’s type AC5A excitation system is used as the excitation system for
the synchronous generator [27]. Figure 13.4 shows the excitation system which
has an extra voltage controller (AFPID) as a second voltage control tool. The
model developed for the CHP unit can be controlled and regulate voltage and
frequency deviations during the islanded operation of microgrids. These controllers
can function simultaneously as a secondary control measure.

13.2.3 Synchronous Reference Frame

Examining multi-machine stability necessitates identifying a relationship between

a machine’s terminal voltage and current in terms of transmission admittance.
Individual machine coordinates di–qi may be related to common system coordinates
D–Q [28] because the transmission admittance matrix in static coordinates can be
expressed by rotating coordinates. Details of this reference frame conversion are
presented in [29].
312 H. Shayeghi et al.

13.2.4 Reduced Network Model

In this chapter, loads, the capacitor, and reactor compensation are modeled as
constant admittance. The admittance of each component can be determined using
Eq. (13.5).

Pi − j Qi
yCi = (13.5)
Vbase

In this equation, Pi , Qi , and Vbase are the active power, reactive power at bus i,
and voltage base, respectively. The variable yCi is incorporated into the transmission
admittance matrix Ybus . If n is the number of buses, then the dimensions of Ybus
are n × n. The size of Ybus can be reduced by adopting only generator buses and
disregarding non-generator buses. That is, the effects of non-generator buses are
transferred into generator buses. If Ybus is divided into four sub-matrices (YGG , YGN ,
red. is written as:
YNG , and YNN ), then Ybus
 
YGG YGN
Ybus = ⇒ Ybus
red.
= YGG − YGN .YN−1N .YN G (13.6)
YN G YN N

13.3 AFPID Controller

Simultaneous voltage/frequency control for islanded microgrids is complex because

of the presence of many parametric uncertainties, load changes, and nonlinearity.
Therefore, a robust controller should be developed. Given that a fuzzy controller
considers uncertainties and nonlinearities in implementing voltage/frequency con-
trol, this research is designed as an AFPID controller (Fig. 13.5). AFPID-based
control is an approach that combines the traditional PID control and fuzzy logic
which does not require an exact mathematical model of a controller object and
adopts a fast, small overshoot, and short settling time. In this approach, PID
parameters can be adjusted in real-time under adaptive fuzzy PID control, thereby
enabling updates to control knowledge and enhancements to the electrical behavior
of a system. The proposed AFPID controller has two independent parts that act in
parallel with others. The first part is a simple PID controller, whose gains are tuned
in nominal load conditions as offline. The second is a fuzzy logic-based mechanism
intended to update the gains of PID controllers online due to the dynamical behavior
of the microgrid. The mechanism has two inputs with coefficients α (j) and β (j)
and automatically tunes the gains of the first part (i.e., the PID controller). As
indicated in Fig. 13.5, the proposed controller has two inputs and three outputs,
whose membership functions are shown in Fig. 13.6.
Seven triangular membership functions are used for each input/output. These are
the NL (negative large), NM (negative medium), NS (negative small), ZR (zero), PS
13 Microgrid Stability Definition, Analysis, and Examples 313

Fig. 13.5 Structure of the AFPID controller in CHP

Fig. 13.6 Membership functions of the fuzzy PID controller

Table 13.1 The fuzzy rules e

of fuzzy PID controllers for
de NL NM NS ZR PS PM PL
all outputs
NL ZR PS PM PL PL PL PL
NM NS ZR PS PM PL PL PL
NS NM NS ZR PS PM PL PL
ZR NL NM NS ZR PS PM PL
PS NL NL NM NS ZR PS PM
PM NL NL NL NL NS ZR PS
PL NL NL NL NL NM NS ZR

(positive small), PM (positive medium), and PL (positive large) functions. The basic
role base for the proposed controller is given in Table 13.1, which is depicted as a
7 × 7 matrix given that the inputs have seven membership functions. The parameter
x(k)(h)(j) in Fig. 13.6 denotes interval changes in a membership function and its value
should be obtained from optimization. In the notation, k denotes the input/output.
If K is the input, then h is an error or Δ error; otherwise, h denotes ΔKp , ΔK i , or
ΔK d . In this notation, j denotes the voltage or frequency controller.
314 H. Shayeghi et al.

13.4 Global Design of the Optimization

To derive an optimal design for the AFPID controller, the NSIDE algorithm is
used as this avoids a trial-and-error design process and enables the simultaneous
control of voltage and frequency in islanded microgrids. Optimally designing a
fuzzy system requires the appropriate selection of gains, membership functions,
fuzzy weighting coefficients, and rule bases. This design also depends on accurate
knowledge of a system, previous experiences, and observations. Such complexity
renders the simultaneous optimal adjustment of all parameters a highly complex and
time-consuming process. Additionally, it is an unreasonable approach because of the
availability of numerous parameters. To overcome these problems, the optimization
of this study is divided into three cascaded stages:
Stage I: Tuning gains and membership functions.
Stage II: Tuning fuzzy rule weighting coefficients.
Stage III: Tuning fuzzy rule bases.
In every optimization stage, the NSIDE algorithm is adopted to achieve optimal
performance. The cost function, optimization variables, and NSIDE algorithm must
also be refined to realize the aims of optimal design.

13.4.1 Cost Function

One of the aims of this study is to minimize voltage and frequency deviations during
the islanded operation of microgrids. Therefore, the ISTSE index based on voltage
and frequency deviations is regarded as the cost function, which is described as:
*. /
(tsim )
f1 = max t 2 2
VCH P (i) i = 1, 2 and 3 (13.7)
0

*. /
(tsim )
f2 = max t 2 2
FCH P (i) i = 1, 2 and 3 (13.8)
0

In these equations, ΔVCHP(i) and ΔFCHP(i) are the voltage and frequency changes
in CHP (i), respectively.
13 Microgrid Stability Definition, Analysis, and Examples 315

13.4.2 System Optimization Variables

13.4.2.1 Stage I

In this stage, the gains of the controller and the membership functions must be
optimized. Figure 13.5 shows that five gains exist for each controller (Kp(j ) , Ki(j ) ,
Kd(j ) , α (j) , and β (j) ). In the notation, j denotes the voltage or frequency controller.
As illustrated in Fig. 13.6, to reduce the number of optimization variables for fuzzy
membership functions, each member function is described by x(k)(h)(j) . The linguistic
variables are defined in Eq. (13.9).
x(k)(h)(j ) x(k)(h)(j )
N L = −x(k)(h)(j ) , NM = − , NS = −
3 3

x(k)(h)(j ) x(k)(h)(j )
ZR = 0, P S = ,PM = , P L = x(k)(h)(j ) (13.9)
3 3

13.4.2.2 Stage II

In this stage, for each rule (n is the number of rules, equal to 49), a weight factor
(j )
w(h) (i) (i = 1 : 49) is examined to determine the importance and influence of rules
on the controller’s final voltage and frequency responses. To employ this type of
coding, we can provide a condition for the removal of ineffective rules. In this case,
a number of fuzzy rules become optimal because the zero weighting coefficients of
the rules do not affect the final response and can be removed from the set of rules.
The importance of this method is that in addition to improving the performance of
the fuzzy controller, it eliminates some ineffectual laws, decreases the number of
variables in the next step (fuzzy rules set), and advances improved optimization.

13.4.2.3 Stage III

In this stage, the set of fuzzy logic rules need to be encoded, regulated, and then
decoded. To encode fuzzy rules, an index (a positive integer number) is assigned
to each law. First, any logical expression determined according to Eq. (13.10) is
converted into T, thus:

n + 1
T = max (mi )  (13.10)
i=1

where T and mi are the maximum number of logical expressions and the logical
expressions related to the ith variable, respectively. In this conversion, zero is
assigned to the lowest logic expression, and the number of membership functions
316 H. Shayeghi et al.

minus one is assigned to the highest logic expression. Then, the obtained numbers
are converted on the basis of a positive integer T according to Eq. (13.11):


n+1
I= Li T (n+1−i) Li ∈ (0, 1, . . . , mi−1 ) (13.11)
i=1

Here, I is the index that corresponds to a given fuzzy rule and serves as an
optimization variable. The decoding of fuzzy rules contrary to the above-mentioned
steps is demonstrated in Eq. (13.12):
+ ,  −(i−1)  
T n + 1
L(n+2−i) = I.T −(i−1) − T  (13.12)
T i = 1

In this equation, [.] denotes that only the integer part of the operational result
is taken [30]. In the proposed controller, each variable has seven triangular output
membership functions. According to the fuzzy rules table, therefore, the number of
fuzzy rules for each output variable is 49. If an optimization variable is allocated
to each rule, the number of optimization variables for each variable is 147. In the
proposed method, however, this number is reduced to 49, resulting in more effective
optimization.

13.4.3 NSIDE Algorithm

Nowadays, the differential evolutionary (DE) algorithm is regarded as a very

effective method for solving complex multi-objective problems. Its advantages
include fast convergence rate, robustness, and global optimization. The basic DE
algorithm involves mutation, crossover, and selection operations, through which
vectors of trial parameters are generated. The implementation of the DE algorithm
has been presented in greater detail in [31].

13.4.3.1 Improved Differential Evolution Algorithms

The DE algorithm is improved by implementing changes to the four stages of

initialization, mutation, selection, and updating of the mutation coefficient. This
improvement is described as follows:
I. Initialization: Randomly construct population OP of NP individuals, with the
dimension of each vector being n, using the following rule [32]:

Xi+N P = V ar min + rand (0, 1) × (V ar max − Xi ) (13.13)

13 Microgrid Stability Definition, Analysis, and Examples 317

where Varmin and Varmax denote the lower and upper bound of the ith component,
respectively, and rand(0, 1) is a uniform random number falling between 0 and 1.
II. Mutation: Randomly select three distinct individuals Xr1 , Xr2 , and Xr2 from
population S and perform mutation using the following formula:
 
Vi,G+1 = Xtb + F × Xr1,G − Xr2,G (13.14)

Here, Xtb is the best of the three individuals (Xr1 , Xr2 , and Xr2 from population
S).
III. Selection: Calculate the objective function value at newly generated indi-
viduals and select NP individuals from the population of the latest generated
individuals, mutation individuals, and crossover individuals to diversify the newly
generated ones.
IV. Updating the mutation coefficient: After each generation, the mutation rate
is updated using the expressions below so that mutation parameter F increases
from an initial value F2i to a final value F2f with the iterative progression of the
optimization algorithm [33].

F1 (t) = μ × F1 (t − 1) × [1 − F1 (t − 1)] (13.15)

   
  iter
F (t) = F2f − F2i + F2i F1 (t) 0 ≤ μ ≤ 4 (13.16)
iter max

In these equations, F1(0) lies between [0,1]. The index “t” is the current iteration,
and F1 (t) is the new mutation factor.

13.4.3.2 Nondominated Sorting-Based Multi-Objective Algorithm

A nondominated sorting-based multi-objective algorithm was proposed and used in

the nondominated sorting genetic algorithm II by Deb et al. [34]. In this algorithm,
every step of the optimization process involves the selection of favorable responses
to produce the next generation of individuals. The domination concept can be
defined as Eq. (13.17).

∀i : xi ≤ yi
x dominate ⇐⇒ (13.17)
∃i0 : xi0< yi0

In this relationship, x dominates y if the two conditions indicated in the equation

above are satisfied. Responses are sorted in a Pareto sequence (Pareto I, Pareto
II, and . . . ) according to the number of dominations by other responses. For the
next generation of a Pareto’s parent, the first priority is the Pareto number (quality
318 H. Shayeghi et al.

criterion). If the parent needs to select the number of Pareto members, the criterion
is used to call the discipline.
The discipline criterion is determined on the basis of the responses in a Pareto
distribution. The discipline criteria are the dispersion of the responses to be satisfied
in the search space and increase the chances of reaching an improved response.
Crowding distance is used according to Eq. (13.18) to rank the discipline with the
benchmark index.
 k 
n f 
 (i+1) − f(i−1) 
k
CD i = cd i , i = 1, 2, . . . , np, cd i = 
k k

(k=1)  fm ax k − fm ink 
(13.18)

where i is the number of individuals in each A Pareto, k is the number of objective

functions, n denotes the number of objective functions, and f represents the objective
function value.
Finally, the AHP mechanism (see [35] for more details) is used to select the best
response from the first Pareto member of the multi-objective optimization as output
and complete the setting of parameters for the AFPID controller. Figures 13.7a–c
show the flowchart of the three phases of parameter setting for the AFPID controller.

13.5 Simulation Results

Figure 13.8 displays the actual microgrid examined to validate the proposed method
[36]. The microgrid is part of the Himmerlands Elforsyning power system in
Aalborg, Denmark. It encompasses a total of 10 loads, three fixed-speed wind
turbine units, and CHP units with three gas turbo generators. The wind turbine
generator close to the unity power factor and the capacitor bank used for reactive
power compensation is operated in islanded mode. The data on loads and lines are
extracted from [36]. Because the wind turbines in the microgrid run at a fixed speed
and the CHPs are the main power producers, the AFPID controller is installed in the
CHPs. The reduced admittance matrix of the microgrid is equal to:

⎡ ⎤
+1.0157 − 1.1498i −0.3961 + 0.4972i −0.3961 + 0.4972i −0.0928 + 0.0721i −0.0573 + 0.0426i −0.0535 + 0.0355i
⎢ ⎥
⎢ − 0.3961 + 0.4972i +1.0157 − 1.1498i −0.3961 + 0.4972i − 0.0928 + 0.0721i −0.0573 + 0.0426i −0.0535 + 0.0355i ⎥
⎢ ⎥
⎢ − 0.3961 + 0.4972i −0.3961 + 0.4972i +1.0157 − 1.1498i − 0.0928 + 0.0721i −0.0573 + 0.0426i −0.0535 + 0.0355i ⎥
2⎢ ⎥
red.
Ybus = 10 ⎢ ⎥
⎢ −0.0928 + 0.0721i −0.0928 + 0.0721i −0.0928 + 0.0721i +0.4332 − 0.3624i −0.0742 + 0.0785i −0.0742 + 0.0785i ⎥
⎢ ⎥
⎢ ⎥
⎣ − 0.0573 + 0.0426i −0.0573 + 0.0426i −0.0573 + 0.0426i − 0.0742 + 0.0785i +0.3983 − 0.3381i −0.1458 + 0.1328i ⎦
− 0.0535 + 0.0355i −0.0535 + 0.0355i −0.0535 + 0.0355i − 0.0703 + 0.0670i −0.1458 + 0.1328i +0.3824 − 0.3049i

Power flow results for islanded microgrid, the calculated initial conditions for
islanded microgrid, and the parameters of the NSIDE algorithm are presented in
Tables 13.2, 13.3, and 13.4, respectively.
13 Microgrid Stability Definition, Analysis, and Examples 319

Fig. 13.7 (a) The tuning process of AFPID controller by NSMDE algorithm. Stage 1: Fuzzy
coefficients and membership function tuning. (b) The tuning process of AFPID controller by
NSMDE algorithm. Stage 2: Fuzzy rule weights tuning. (c) The tuning process of AFPID controller
by NSMDE algorithm. Stage 3: Fuzzy rules tuning

Figure 13.9 shows the nondominated responses (Pareto front) of the three-
stage optimization of the AFPID controller. The objective functions are the ISTSE,
applied according to Eq. (13.7) and Eq. (13.8). As indicated in Fig. 13.9, the
best solution is selected by the AHP mechanism. The results of the three-stage
optimization including optimized coefficients of controllers by NSIDE algorithm,
optimized membership function parameters, optimized rules for voltage controller,
and optimized rules for frequency controller are presented in Tables 13.5, 13.6, 13.7,
and 13.8, respectively.
To demonstrate the effectiveness of the proposed controller, four scenarios with
different disturbances are considered. The disturbances are a symmetrical three-
320 H. Shayeghi et al.

Fig. 13.7 (continued)

phase fault, load shedding, and decreasing and increasing loads. In all the scenarios,
the worst time-domain values for each of the CHPs are selected to examine
the worst-case condition. The performance of the proposed controller in all the
scenarios is assessed against that of the fuzzy PID controller [37] and the classical
PID controller. The fuzzy PID controller does not work online, but its output signal,
which was derived after applying coefficients Kp , Ki , and Kd , is used directly in
the system. This controller does not adjust the fuzzy PID coefficients but adjusts its
output.
13 Microgrid Stability Definition, Analysis, and Examples 321

Fig. 13.7 (continued)

13.5.1 Scenario 1: Symmetrical Three-Phase Fault

A three-phase fault over the lines and near the buses affects the admittance
matrix. Therefore, the row and column of the admittance matrix of the microgrid
experiencing the fault bus must be omitted, after which the lower columns and
rows must be pulled one step back. This process generates a small-order admittance
matrix. After the fault is cleared, replacing the current admittance matrix with the
initial admittance matrix returns the microgrid to its normal state. In this scenario,
a symmetrical three-phase fault occurs in t = 1 s at bus 11, and after two cycles of
t = 1.04 s, the fault is cleared. The microgrid voltage and frequency responses to
this fault are depicted in Figs. 13.10 and 13.11, respectively.
The AFPID controller damps the transient responses in a better manner in terms
of settling time and steady-state error than do the comparison controllers. The results
also indicate that the AFPID controller is considerably more suitable than the two
322 H. Shayeghi et al.

Fig. 13.8 The real microgrid

Table 13.2 Power flow Bus |Vi |  θi Pgi Qgi

results for islanded microgrid
1 1 0 3.07 0.69
2 1 −0.5 2.9 0.67
3 1 −0.5 2.9 0.67
4 0.98 −11.34 0.1 0
5 0.97 −11.35 0.1 0
6 0.97 −11.35 0.1 0

others. To confirm the results obtained through the simulation, the values of some
response characteristics (e.g., ISTSE, overshoot, undershoot, and settling time) are
determined on the basis of Figs. 13.10 and 13.11.
The result of voltage and frequency is listed in Table 13.9, which shows that
AFPID improvements in the time domain of voltage and frequency in terms of
the ISTSE are 76.75% and 72.23% rather than PID respectively, and 62.83%
and 37.14% rather than FPID. The table also indicates that the percentage of
improvement achieved with the AFPID controller is higher than that realized using
the fuzzy PID controller. We can, therefore, conclude that the AFPID controller is
the best control option among all the evaluated controllers.
13 Microgrid Stability Definition, Analysis, and Examples 323

Table 13.3 Calculated initial conditions for islanded microgrid

Parameter CHP 1 CHP 2 CHP3 WTG 1 WTG 2 WTG 3
Ia0 3.147 2.976 2.976 0.090 0.901 0.090
0 −0.221 −0.236 −0.236 −0.198 −0.198 −0.198
Eq0 1.386 1.369 1.369 1.122 1.122 1.122
δ0 0.519 0.499 0.499 0 0 0
Id0 −2.121 −1.996 −1.996 −0.018 −0.018 −0.018
Iq0 2.324 2.208 2.208 0.088 0.088 0.088
Vd0 −0.496 −0.487 −0.487 −0.218 −0.219 −0.219
Vq0 0.868 0.874 0.874 1.088 1.088 1.088
E’d0 0.278 0.264 0.264 0.011 0.011 0.011
E’q0 1.069 1.071 1.071 0.002 0.002 0.002
Te0 3.044 2.971 2.971 0.122 0.122 0.122
Tm0 3.044 2.971 2.971 0.122 0.122 0.122

Table 13.4 The parameters Parameter Value Parameter Value

of the NSIDE algorithm
NC 5 F1t 0.8
NM 5 F2i 0.5
NW 50 F2f 1.5
NR 50 μ 4
Npop 50 CR 0.6
OP 50 CRmin 0.5
Iteration 50 CRmax 0.9

13.5.2 Scenario 2: Load Shedding in Bus 8

In this scenario, to ensure the effective performance of the designed controller under
load shedding, the load produced by bus 8 (load 08) at t = 1 s is separated from the
system. Load shedding causes to happen an imbalance between power generation
and consumption and fluctuations in the microgrid voltage and frequency. The
microgrid voltage and frequency responses to this condition are demonstrated in
Figs. 13.12 and 13.13, respectively. The values of the time domain characteristics of
voltage and frequency are presented in Table 13.10.
Table 13.10 reflects that the AFPID controller’s AFPID improvements in the time
domain characteristics of voltage and frequency in terms of the ISTSE are 76.64%
and 72.53% rather than PID, respectively, and 64.54% and 55.99% rather than FPID.
The simulation results on voltage and frequency in the time domain suggested that
the AFPID controller improves these characteristics to a greater extent than do
the comparison controllers. The ISTSE and overshoot improvements in frequency
responses are negative because, in this scenario, the fuzzy PID controller cannot
exhibit a suitable performance.
324 H. Shayeghi et al.

Best paretto of FPID Optimization with NSMDE in 3Stage

15
stage1: Ceoffs & MFs
stage2: Weights
stage3: Roles

10
ISTSE (DV)

AHP Selected
ISTSE(Dv)=5.0606 , ISTSE(Df)=20.3813

0
0 10 20 30 40 50 60 70 80 90 100
ISTSE (Df)

Fig. 13.9 First Pareto (nondominated response) for three-stage controller optimization

Table 13.5 Optimized Voltage Controller Frequency Controller

coefficients of controllers by
Parameter PID AFPID PID AFPID
NSIDE algorithm
Kp 0.4600 0.0717 0.0114 0.0308
KI 1 0.0786 0.0501 0.7473
KD 0.4905 0 0.0001 0.6587
α – 0.5091 – 0.9992
β – 0.8763 – 0.0703

13.5.3 Scenario 3: 30% Load Decrease

In this scenario, 30% of load STNO (see Fig. 13.8) at t = 1 s is reduced.

Figures 13.14 and 13.15 show the voltage and frequency responses of the grid
in this operating condition of the microgrid, respectively. The other simulation
results for this scenario, including those on the voltage and frequency responses
to load reduction, are presented in Table 13.11. The reduction in active and reactive
loads distorts frequency and voltage; after the occurrence of damping oscillations,
frequency and voltage slightly increase.
The AFPID controller improves the time domain characteristics of voltage
and frequency in terms of the ISTSE to 89.53% and 48.34% rather than PID,
respectively, and 80.06% and 34.09% rather than FPID. This result indicates that
the proposed controller is superior to the fuzzy PID and PID controllers in terms of
achieving system stability.
13 Microgrid Stability Definition, Analysis, and Examples 325

Table 13.6 Optimized membership function parameters

x(input) (error) (voltage) 0.0001 x(input) (error) (voltage) 0.2216
x(input) (Δerror) (voltage) 0.05679 x(input) (Δerror) (voltage) 0.3520
   
x(output) KP (voltage) 0.0995 x(output) KP (voltage) 0.5130
   
x(output) KI (voltage) 0.1024 x(output) KI (voltage) 0.9253
   
x(output) KD (voltage) 0.0219 x(output) KD (voltage) 0.7264

( AFPID, FPID and PID)

Fig. 13.10 Voltage responses of islanded microgrid in scenario 1. (AFPID, FPID, and PID)

13.5.4 Scenario 4: 30% Load Increase

In this scenario, the effects of a 30% increase in busload STNO (see Fig. 13.8)
produced by bus 15 at t = 1 s is examined. The increased load reduces speed
because of an increase in the electric generator’s torque. Figures 13.16 and 13.17
demonstrate the voltage and frequency responses of the microgrid.
The time-domain characteristics of this scenario are presented in Table 13.12.
After microgrid oscillation damping, frequency considerably decreases. With
increasing network reactive power, microgrid voltage is disturbed; efforts to
reach sustainability slightly decrease the microgrid voltage. The AFPID controller
improves the time domain characteristics of voltage and frequency in terms of
the ISTSE to 85.55% and 55.30% rather than PID, respectively, and 70.09% and
35.26% rather than FPID. The results for this scenario are similar to those for
scenario 3. The superiority of the AFPID controller in terms of fast and convenient
damping of voltage and frequency fluctuations is clearly noticeable. In this scenario,
the time domain results of all the three controllers showed that the conventional
 326

Table 13.7 Optimized rules for voltage controller

a) ΔKP(voltage) b) ΔKI(voltage) c) ΔKD(voltage)
de de de
e NL NM NS ZR PS PM PL NL NL NM NS ZR PS PM PL NL NL NM NS ZR PS PM PL
NL – PS PS NS PL NL PL – PS PS ZR PL PS PL – – NM PM ZR PL ZR PL –
NM NS ZR PL ZR – NS PL NS PS PL PL – NL PL NS NS ZR NS NL – NS PL NS
NS NM NS ZR PL ZR PL PM NM NS ZR PM PM PL PS NM NM NS ZR PM ZR PL ZR NM
ZR NL – – ZR – PS PL NL – – ZR – PL PL NL NL – – ZR – PL PL NL
PS NL NL PS ZR NS PM PM NL NL PL PL PS ZR PM NL NL NL NM NS PM PM PM NL
PM NL ZR PS PL ZR PM PS NL PS PL ZR NS ZR PS NL NL NM NM PM PM PL PS NL
PL ZR PM NL P PS ZR – NL NS NL NL PS NM – NL NS PL NL NS NS PM – NS
H. Shayeghi et al.
Table 13.8 Optimized rules for frequency controller
a) ΔKP(frequency) b) ΔKI(frequency) c) ΔKD(frequency)
de de de
e NL NM NS ZR PS PM PL NL NL NM NS ZR PS PM PL NL NL NM NS ZR PS PM PL
NL NL NM NM NL PS PL PL NL PL ZR NL NM PS NM NL NL ZR NS PS NM – NS NL
NM NS ZR PS PS NL – NM ZR ZR PS NS NM – NM ZR PS ZR PS NS NM PL PS PS
NS ZR NS NS PS – PM PS NL NS NM PS – NM ZR NL NL NS PS PS – PL PL NL
ZR NL NL NS ZR PS PS NM NL NS NS ZR PS PL NM NL NL NS NS ZR PS PM PS NL
13 Microgrid Stability Definition, Analysis, and Examples

PS NS NL – NL PS PM NS ZR NL – ZR NL NL PM ZR NM NL NM NM NM ZR PS NM
PM PM NL ZR – PM ZR PS NM NL NM – NL ZR PS NM PS NL – – NM ZR PS PS
PL NL NL ZR NL NM PL – NL NL ZR NL NM NS – NL NL NL PM NL NM PL – NL
327
328 H. Shayeghi et al.

( AFPID, FPID and PID)

Fig. 13.11 Frequency responses of islanded microgrid in scenario 1. (AFPID, FPID, and PID)

Table 13.9 Time domain characteristics of voltage and frequency in scenario 1

Value
Parameter PID (NSIDE) FPID [37] AFPID (NSIDE)
Voltage ISTSE (pu) 26.0163 69.9972 111.9122
Overshoot (pu) 0.2448 0.2871 0.3106
Undershoot (pu) −0.0532 −0.0793 −0.0905
Settling time (s) 7.8250 12.1054 12.8424
Frequency ISTSE (pu) 50.6243 80.5385 182.3305
Overshoot (pu) 0.459 0.0420 0.0686
Undershoot (pu) −0.0387 −0.0498 −0.0781
Settling time (s) 14.9450 18.7554 19.4224

Table 13.10 Time domain characteristics of voltage and frequency in scenario 2

Value
Parameter PID (NSIDE) FPID [37] AFPID (NSIDE)
Voltage ISTSE (pu) 0.7372 2.0792 3.1344
Overshoot (pu) 0.0083 0.0155 0.0189
Undershoot (pu) −0.0413 −0.0478 −0.0576
Settling time (s) 5.1550 7.6905 7.8413
Frequency ISTSE (pu) 3.7590 8.5422 13.6843
Overshoot (pu) 0.0155 0.0212 0.0267
Undershoot (pu) −0.0161 −0.0267 −0.0336
Settling time (s) 9.2150 10.8905 12.2713
13 Microgrid Stability Definition, Analysis, and Examples 329

Fig. 13.12 Voltage responses of islanded microgrid in scenario 2. (AFPID, FPID, and PID)

Fig. 13.13 Frequency responses of islanded microgrid in scenario 2. (AFPID, FPID, and PID)

fuzzy PID exhibits the worst operation; the two other controllers perform well
because of the appropriate functioning of the NSIDE algorithm in the optimization
of these controller’s parameters. Among the three, however, the AFPID controller
remains the most robust and dynamic in performance.
330 H. Shayeghi et al.

Fig. 13.14 Voltage responses of islanded microgrid in scenario 3. (AFPID, FPID, and PID)

( AFPID, FPID and PID)

Fig. 13.15 Frequency responses of islanded microgrid in scenario 3. (AFPID, FPID, and PID)

13.6 Conclusion

In this chapter, the operation of an AFPID controller was investigated for the
simultaneous control of voltage and frequency in islanded microgrids through a
newly established system simulation model. The model was based on a multi-
machine structure, wherein the relationship between different units was expressed
13 Microgrid Stability Definition, Analysis, and Examples 331

Table 13.11 Time domain characteristics of voltage and frequency in scenario 3

Value
Parameter PID (NSIDE) FPID [37] AFPID (NSIDE)
Voltage ISTSE (pu) 0.0967 0.4997 0.9236
Overshoot (pu) 0.0034 0.0083 0.0111
Undershoot (pu) −0.0166 −0.0246 −0.0339
Settling time (s) 5.0950 5.0991 5.8740
Frequency ISTSE (pu) 3.4194 5.1883 6.6195
Overshoot (pu) 0.0112 0.0126 0.0144
Undershoot (pu) −0.0111 −0.0133 −0.0150
Settling time (s) 10.1050 9.5876 9.6740

( AFPID, FPID and PID)

Fig. 13.16 Voltage responses of islanded microgrid in scenario 4. (AFPID, FPID, and PID)

Table 13.12 Time domain characteristics of voltage and frequency in scenario 4

Value
Parameter PID (NSIDE) FPID [37] AFPID (NSIDE)
Voltage ISTSE (pu) 0.0485 0.1622 0.3349
Overshoot (pu) 0.0115 0.0138 0.0200
Undershoot (pu) −0.0024 −0.0046 −0.0065
Settling time (s) 5.0950 5.0997 5.1047
Frequency ISTSE (pu) 2.9553 4.5649 6.6123
Overshoot (pu) 0.0103 0.0126 0.0155
Undershoot (pu) −0.0105 −0.0109 −0.0129
Settling time (s) 10.035 9.5197 9.6447
332 H. Shayeghi et al.

( AFPID, FPID and PID)

Fig. 13.17 Frequency responses of islanded microgrid in scenario 4. (AFPID, FPID, and PID)

through an admittance matrix. The AFPID controller functions as a secondary

control scheme that covers all conditions and damps oscillations in the voltage and
frequency of actual microgrids. The proposed control strategy was equipped with a
constant-gain PID controller and a fuzzy inference system, which adjusts PID gains.
The AFPID parameters, fuzzy rule weights, and rule bases were automatically tuned
to minimize design efforts and identify an improved fuzzy system control through
the developed NSIDE optimization algorithm. The simulation results indicated that
tuning the fuzzy rule weights eliminates some fuzzy rules and reduces the need to
optimize fuzzy rule variables. These effects, in turn, advance improved optimization
results. The simulation results for different disturbance scenarios reflected the
superior performance of the proposed controller over the conventional fuzzy and
classical PID controllers. In all the scenarios, the time domain characteristics
achieved by the AFPID controller were superior to those generated by the two
others. The microgrid modeling based on the multi-machine system modeling
showed that applying a variety of changes in microgrid loads and the optimal
design of the AFPID controller is easily possible. It also confirmed the accuracy
of the proposed method. The advantages of the proposed method are summarized as
follows:
• The newly developed model of dynamic microgrid behavior considers all kinds
of distributed generators and energy storage devices.
• Different faults and load changes are easily considerable in system performance
analysis using the developed model.
13 Microgrid Stability Definition, Analysis, and Examples 333

• Because of the gain adjustment for the PID controller, the AFPID controller
generates suitable simultaneous voltage and frequency responses against dif-
ferent faults and load uncertainties. The results presented in Sect. 5 indicated
that the conventional fuzzy and classical PID controllers exhibit unsuitable and
unacceptable performance in all the islanded microgrid operation scenarios.
• The newly developed controller based on a fuzzy system can satisfactorily func-
tion in all operating conditions. The ISTSE performance index was calculated
to ensure the accuracy and speed of voltage and frequency responses, thereby
reinforcing the validity of the controller.

References

1. Khalghani, M., Khooban, M., Moghaddam, E., Vafamand, N., & Goodarzi, M. (2016). A
self - tuning load frequency control strategy for microgrids: Human brain emotional learning.
International Journal of Electrical Power Energy Systems, 75(2), 311–319.
2. Sanjarani, M., & Gharehpetian, G. (2014). Game-theoretic approach to cooperative control of
distributed energy resources in islanded microgrid considering voltage and frequency stability.
Neural Computing and Applications, 25(2), 343–351.
3. Palizban, O., Kauhaniemi, K., & Guerrero, J. (2014). Microgrids in active network
management-part I: Hierarchical control, energy storage, virtual power plants, and market
participation. Renewable and Sustainable Energy Reviews, 36(7), 428–439.
4. Schiffer, J., Ortega, R., Astolfi, A., Raisch, J., & Sezi, T. (2014). Conditions for stability of
droop controlled inverter based microgrids. Automatica, 50(10), 2457–2469.
5. Bidram, A., & Davoudi, A. (2012). Hierarchical structure of microgrids control system. IEEE
transactions on smart grids, 3(4), 1963–1976.
6. Lu, L., & Chu, C. (2015). Consnsus based secondary frequency and voltage droop control
of virtual synchronous generators for isolated AC micro grids. IEEE Journal on Emerging
Selected Topics in Circuits and Systems, 5(3), 443–455.
7. Wang, Y., Pordanjani, I., & Li, W. (2011). Strategy to minimize the load shedding amount for
voltage collapse prevention. IET Generation, Transmission and Distribution, 5(3), 307–313.
8. Gu, W., & Zhu, J. (2014). Adaptive decentralized under frequency load shedding for islanded
smart distribution networks. IEEE Transactions on Transmission, Sustain Energy, 5(3), 886–
895.
9. Nourollah, S., Piraysh, A., & Aminifar, F. (2016). Combinational scheme for voltage and
frequency recivery in an island distribution system. IET Generation, Transmission and
Distribution, 10(12), 2899–2906.
10. H. Laaksonen, P. Saari, R. Komulainen, “Voltage and frequency control of inverter based LV
network microgrid”, in Future power systems, 2005 international conference on, 2005.
11. Li, Y., & Li, Y. (2011). Power management of inverter interfaced autonomous microgrids based
on virtual frequency voltage frame. IEEE transactions on smart grids, 2(1), 30–40.
12. Mari, M. I., & Soliman, M. H. (2013). A coordinated voltage and frequency control of inverter
based distributed generation and distributed energy storage system for autonomous microgrids.
Electric Power Components and Systems, 41(4), 383–400.
13. B. Shoeiby, R. Davoodnezhad, D. Holms, B. McGrath, “Voltage - frequency control of an
islanded microgrids using the intrinsic droop characteristics of resonant current regulators”, in
Energy conversion congress and exposition, 2014.
14. Valencia, F., Collado, J., Saez, D., & Martin, L. (2016). Robust energy management system
for a microgrid based on a fuzzy prediction interval model. IEEE Transactions on Smart Grid,
7(3), 1486–1494.
334 H. Shayeghi et al.

15. Kim, Y., Kim, E., & Moon, S. (2015). Frequency and voltage control strategy of standalone
microgrids with high penetration of intermittent renewable generation systems. IEEE Transac-
tions on Power Systems, 31(1), 718–728.
16. Bayat, M., Sheshyekani, K., Hamzeh, M., & Rzazadeh, A. (2015). Coordination of distributed
energy resources and demand response for voltage and frequency support of MV microgrids.
IEEE Transactions on Power Systems, 31(2), 1506–1516.
17. Tang, X., Hu, X., Li, N., Deng, W., & Zhang, G. (2016). A novel frequency and voltage control
method for islanded microgrid based on multi energy storage. IEEE Transactions on Smart
Grids, 7(1), 410–419.
18. Kunjumhummed, L. (2013). A power system dynamics simulation program using matlab/
Simulink. International Journal of Advanced Research in Electrical, Electronics and Instru-
mentation Engineering, 2(1), 111–121.
19. Schiffer, J., Zonetti, D., Ortega, R., Stancovic, A., Sezi, T., & Raisch, J. (2016). A survey on
modeling of microgrids- from fundamental physics to phasors and voltage sources. Automatica,
74(12), 135–150.
20. Hassan, M., & Abido, M. (2014). Real time implementation and optimal design of autonomous
microgrids. Electric Power Systems Research, 109(4), 118–127.
21. S. Pati, K. Mohanty, S. Kar, S. Mishra, “A sliding mode controller based STATCOM with
battery storage for voltage and frequency stabilization in a micro grids”, in 2016 International
conference on circuit, power and computing technologies, 2016.
22. Vachirasricirikul, S., & Nagmaroo, I. (2012). Robust controller design of micro turbine and
electrolyzer for frequency stabilization in a microgrid system with plug-in hybrid electric
vehicles. International Journal of Electrical Power & Energy Systems, 4(1), 804–811.
23. Bevrani, H., Feizi, M., & Ataeee, S. (2016). Robust frequency control in an islanded micro
grid: H and u synthesis approach. IEEE Transactions on Smart Grid, 7(2), 706–717.
24. Ahmadi, S., Shokoohi, S., & Bevrani, H. (2015). Fuzzy logic based droop control for
simultaneous voltage and frequency regulation in an AC microgrids. International Journal
of Electrical Power & Energy Systems, 64(1), 148–155.
25. A. Perdana, “Dynamic models of wind turbines”, Thesis of PhD., Goteborg, Sweden, 2008.
26. Abbas, F., & Abdulsada, A. (2010). Simulation of wind turbine speed control by MATLAB.
International Journal of Computer and Electrical Engineering, 2(2), 912–915.
27. Mahat, P., Chen, Z., & Jensen, B. (2011). Control and operation of distributed generation in
distribution systems. Electric Power Systems Research, 81, 495–502.
28. Y. Yu, “Electric power system dynamics”, 1983.
29. Robert, J., Ian, A., & Hiskens, A. (1997). Lyapunov functions for multi machine power systems
with dynamic loads. IEEE Transactions on Circuits and Systems I: Fundamental Theory and
Applications, 44, 796–812.
30. Zamani, A., Bijami, E., sheikholeslami, F., & jafrasteh, B. (2014). Optimal fuzzy load
frequency controller with simultaneous auto-tuned membership functions and fuzzy control
rules. Turkish Journal of Electrical Engineering & Computer Sciences, 22, 66–86.
31. Storn, R., & Price, K. (1997). Differential evolution-a simple and efficient heuristic for global
optimization over continuous spaces. Journal of Global Optimization, 11, 341–359.
32. A. Musrrat, P. Millie, A. Ajith, “A modified differential evolution algorithm and its application
to engineering problems”, in International conference of soft computing and pattern recogni-
tion, 2009.
33. Singh, H., & Srivastava, L. (2014). Modified differential evolution algorithm for multi-
objective VAR management. Electrical Power and Energy Systems, 55, 731–740.
34. Deb, K., Pratap, A., Agarwal, S., & Meyarivan, T. (2002). A fast and elitist multi objective
genetic algorithm: NSGA-II. IEEE Transactions on Evolutionary Computation, 6, 182–197.
35. Handfield, R., Walton, S., Sroufe, R., & Melnyk, S. (2002). Applying environmental criteria to
supplier assessment: A study in the application of the analytical hierarchy process. European
Journal of Operational Research, 141(1), 70–87.
36. Mahat, P., Chen, Z., & Jensen, B. (2011). Control and operation of distributed generation in
distribution systems. Electric Power Systems Research, 81, 495–502.
13 Microgrid Stability Definition, Analysis, and Examples 335

37. Pradhan, P., Sahu, R., & Panda, S. (2016). Firefly algorithm optimized fuzzy PID controller for
AGC of multi-area multi-source power systems with UPFC and SMES. Engineering Science
and Technology, 19(1), 338–354.
38. Shayeghi, H., & Aryanpour, H. (2016). Robust online fuzzy PID design based on improved
differential evolution for islanded microgrid frequency control considering nonlinear factors
and uncertainties. Tabriz Journal of Electrical Engineering, 46, 241–256.
39. Ghasemi, A., Shayeghi, H., & Alkhati, H. (2013). Robust Design of Multi machine power sys-
tem stabilizers using fuzzy gravitational search algorithm. International Journal of Electrical
Power & Energy Systems, 51, 190–200.
40. Shayeghi, H., & Alilou, M. (2015). Application of multi objective HFAPSO algorithm for
simultaneous placement of DG, capacitor and protective device in radial distribution network.
Journal of Operation and Automation in Power Engineering, 3, 131–146.
41. Shayeghi, H., Shayanfar, H. A., & Jalili, A. (2007). Multi stage fuzzy PID load frequency
controller in a restructured power system. Journal of Electrical Engineering, 58, 61–70.
42. Shayeghi, H., Ghasemi, A., & Shayanfar, H. A. (2011). PID type stabilizer Design for Multi
Machine Power System Using IPSO procedure. Computer Science and Engineering, 1, 36–42.
43. Hashemi, Y., Shayeghi, H., & Moradzadeh, M. (2017). Design of Dual-Dimensional Controller
Based on multi-objective gravitational search optimization algorithm for amelioration of
impact of oscillation in power generated by large-scale wind farms. Applied Soft Computing,
53, 236–261.
Chapter 14
Voltage Unbalance Compensation in AC
Microgrids

Shahram Karimi, Mehdi Norianfar, and Josep M. Guerrero

14.1 Introduction

Nowadays, the continuity of the electricity supply and the voltage quality with
considering the environmental and economic aspects are the important issues related
to the power distribution operation. Adopting the distributed generations (DGs) in
the electrical distribution networks are the main strategy of the electrical engineers
to guarantee the continuous supply of the local and sensitive loads, improving
the power quality, decreasing the emission of greenhouse gases, through applying
renewable energy resources, and decreasing the power losses. The presence of DGs
in electrical distribution systems has introduced a new structure called microgrid
(MG) [1]. MGs can operate in grid-connected or islanded mode. In the islanded
condition, there is a greater probability of the power quality problems due to the
lack of the main grid support. Power quality can be handled for harmonics, voltage
unbalance, interruptions, sags, swells, and transients [2].
This chapter focuses on the harmonics and voltage unbalance disturbances as
the main problems of power quality in the steady state. Harmonics and voltage
unbalance can have negative impacts on the customer equipment and the network
performance if they exceed the permissible limit. So, satisfying the constraints of the
total harmonic distortion (THD) and voltage unbalance factor (VUF) is noteworthy
due to the negative effects of these disturbances on the sensitive loads. The IEEE
and IEC standards recommend the limitations for the power quality parameters [3].
An example of these standards is given in Table 14.1.

S. Karimi () · M. Norianfar

Department of Electrical Engineering, Faculty of Engineering, Razi University, Kermanshah, Iran
e-mail: shahramkarimi@razi.ac.ir
J. M. Guerrero
Center for Research on Microgrids (CROM), Department of Energy Technology, Aalborg
University, Aalborg, Denmark

© Springer Nature Switzerland AG 2021 337

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_14
338 S. Karimi et al.

Table 14.1 IEE voltage Bus voltage VPCC Individual harmonics (%) THD (%)
distortion limits [6]
V < 69 kV 3.0 5.0
69 kV ≤ V < 161 kV 1.5 2.5
161 kV ≤ V 1.0 1.5

The unbalanced voltage and harmonics can be compensated by using equipment

such as active power filters, dynamic voltage restorers (DVR), and static syn-
chronous compensator (STATCOM). However, all these solutions impose additional
costs. Therefore, researchers used the surplus capacity of the inverter-interfaced
DGs, recently, to improve the power quality in order to reduce the additional costs.
In fact, the inverter-interfaced DGs are capable of performing multiple tasks at
a time, the so-called multifunctional [4, 5]. Indeed, one can use an appropriate
control method in the control system of inverter-interfaced DGs to improve the
power quality at the consumer’s side. This chapter involves the control methods
proposed to compensate unbalanced voltage and harmonics The performance
of these methods is shown analytically and its effectiveness and feasibility are
illustrated at the end of the chapter by simulating a simple islanded MG.

14.2 Inverter-Interfaced DG

In this chapter, we study the general structure of a three-phase inverter-interfaced

DG based on voltage source converter (VSC) allowing to compensate the current
harmonics and imbalance of a three-phase three-wire electrical network. The block
diagram of a grid-connected VSC is shown in Fig. 14.1.

14.2.1 VSC Control

As shown in Fig. 14.1, the control part of a VSC generally performs three main
functions:
• Identification of the reference currents,
• Current controller,
• DC voltage regulation.
One of the control blocks in the VSC control system is the DC voltage regulation
which is outside the scope of this chapter. Reference currents need to be well
identified to reduce the current harmonics and compensate the voltage imbalance.
In the following sections, current identification methods for reference currents
generation will be analyzed and described in detail.
14 Voltage Unbalance Compensation in AC Microgrids 339

Electrical
network
Linear and
non-linear
load

Identification of Output
the reference filter
current

Current Energy
controller resource

Regulation of Inverter
the DC voltage.

Control part Power part

Fig. 14.1 General structure of a VSC

14.3 Harmonic Compensation

The harmonic loads and the nonlinear nature of the converters are the main factors
in producing the harmonic currents. Therefore, it is not possible to avoid harmonic
voltage and current in the microgrids. However, as mentioned earlier, THD should
be limited in an acceptable range. There are various methods to reduce harmonic
distortion. The harmonic distortion caused by the output voltage of the converters
can be reduced by choosing a higher switching frequency. But increasing the
switching frequency will increase the power losses. As a result, the frequency of
the converter, especially for high powers, cannot be selected too large. Another
common solution is to use an LC or LCL filter. By using these common methods
and since the inverters are used in the microgrid at their linear operating point, the
problem of harmonic distortion resulting from them is largely solved. For harmonic
compensation due to nonlinear loads, in the simplest case, a low-pass filter can
be used. Since harmonics are close to the fundamental component, for appropriate
filtering, the filter bandwidth should be small enough, which affects the fundamental
component. The following is a present active solution to this issue.
340 S. Karimi et al.

14.3.1 Identification of Reference Currents

The quality of the current harmonics compensation strongly depends on the

performance of the chosen identification technique. In fact, a control system, even a
very efficient one, cannot perform satisfactory filtering if the harmonic currents are
badly identified. Hence, various identification techniques have been proposed and
extended. They can be grouped according to two approaches [7, 8].
• Identification in the frequency domain.
This type of approach uses the fast Fourier transform, for extracting the
harmonics of the load current. This approach is particularly suitable for loads that
harmonic content varies slowly. It also has the advantage of selecting each harmonic
individually and thus only compensates for the preponderant harmonic currents.
However, this method requires heavy calculations in order to identify the harmonic
currents. For this reason, frequency methods are not used in practice since the
harmonic currents can generally vary rapidly over time.
• Identification in the time domain.
To identify the harmonic currents, various methods in the time domain have
been proposed [7, 9–11]. The basic methods will be the subject of the following
paragraphs.

14.3.1.1 Instantaneous Active and Reactive Power Method

The instantaneous active and reactive power method (generally so-called pq method)
was initially extended in [9]. In this method, the instantaneous active and reactive
powers are calculated using the Concordia transformation. After transformation,
the instantaneous powers include a DC component and AC components. The DC
component is related to the fundamental frequency of the load current and the AC
components are associated with the harmonic components of the load current. By
eliminating the DC component of the instantaneous active power through a low-pass
filter (LPF), the harmonic components can be identified. The basis of pq method is
presented in the following.
Let vs1 , vs2 , vs3 and ic1 , ic2 , ic3 be the phase voltages of a three-phase network
without homopolar, and the load currents, respectively. The transformation of
Concordia makes it possible to bring this balanced three-phase system to a two-
phase system that axes are quadrature:
⎡ ⎤
ic1
⎡ ⎤ 0  ⎢ ⎥
iα
1− 1
−√12 ⎢ ⎥
⎣ ⎦= 2 √2 ⎢ ⎥
⎢ ic2 ⎥ (14.1)
3 0 23 − 23 ⎢ ⎥
iβ ⎣ ⎦
ic3
14 Voltage Unbalance Compensation in AC Microgrids 341

⎡ ⎤
vs1
⎡ ⎤ 0  ⎢ ⎥
vα
1− 1
−√12 ⎢ ⎥
⎣ ⎦= 2 √2 ⎢ ⎥
⎢ vs2 ⎥ (14.2)
3 0 23 − 23 ⎢ ⎥
vβ ⎣ ⎦
vs3

The instantaneous active power p and the instantaneous reactive power q are
defined by:
⎡ ⎤ ⎡ ⎤⎡ ⎤
p vα vβ iα
⎣ ⎦=⎣ ⎦⎣ ⎦ (14.3)
−q − vβ vα iβ

The instantaneous active and reactive powers can be written as follows:

⎡ ⎤ ⎡ ∼⎤
p p+p
⎣ ⎦=⎢
⎣
⎥
⎦ (14.4)
∼
q q+q

∼ ∼
wherein p and q are the DC components of p and q, and p and q are the AC
components of p and q. From eq. (14.3), we can deduce the expressions of the
components of the load current along the axes αβ:
⎡ ⎤ ⎡ ⎤−1 ⎡ ⎤ ⎡ ⎤⎡ ⎤
iα vα vβ p vα −vβ p
⎣ ⎦=⎣ ⎦ ⎣ ⎦= 1 ⎣ ⎦⎣ ⎦ (14.5)
vα2 + vβ2
iβ − vβ vα −q vβ vα −q

By replacing (14.4) in (14.5), these currents are expressed along the axes αβ by:
⎡ ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤⎡ ∼ ⎤
iα vα −vβ p vα −vβ p
⎣ ⎦= 1 ⎣ ⎦⎣ ⎦+ 1 ⎣ ⎦⎢
⎣
⎥
⎦
vα2 + vβ2 vα2 + vβ2 ∼
iβ vβ vα −q vβ vα −q
(14.6)

Depending on the control objectives, one can simultaneously compensate for the
current harmonics and the reactive power, or just the current harmonics compensa-
tion is selected as the control objective. If for example the current harmonics and
the reactive power compensation are considered as the control objectives, the DC
component of p is eliminated by using a LPF. Then, the active power pc (necessary
for the regulation of the DC voltage vdc ) is added to the AC component of the
instantaneous active power. The reference currents along the αβ axes are obtained
as follows:
342 S. Karimi et al.

Regulator

abc LPF
Eq. 3 −
αβ αβ
Eq. 7
abc
abc
αβ

Fig. 14.2 Reference currents generation by using the instantaneous powers method

⎡ ⎤ ⎡ ⎤⎡∼ ⎤
iref α vα −vβ p + pc
⎣ ⎦= 1 ⎣ ⎦⎢
⎣
⎥
⎦ (14.7)
vα2 + vβ2
irefβ vβ vα −q

Finally, by the inverse transformation of Concordia, the reference currents along

the abc axes are achieved:
⎡ ⎤
iref 1 ⎡ ⎤⎡
⎢ ⎥ 0 ⎤
⎢ ⎥ 1 √
0 iref α
⎢ ⎥ 2 ⎢ 1 3 ⎥⎣ ⎦
⎢ iref 3 ⎥ = ⎣−2 2√ ⎦
(14.8)
⎢ ⎥ 3
⎣ ⎦ −2 − 2
1 3 irefβ
iref 3

Figure 14.2 illustrates the reference currents generation when the instantaneous
powers method is used to simultaneously compensate for the current harmonics and
the reactive power.

14.3.1.2 Synchronous Reference Frame (SRF) Method

This method, introduced in [12], also exploits the transformation of Concordia but
it is applied only to load currents ic1 , ic2 , and ic3 . A second transformation is then
carried out to obtain the load currents along the dq axes. This makes it possible
to transform the fundamental component of the load current into a DC component
and the harmonic components of the load current into AC components. The DC
component of the load current can then be eliminated by using a LPF. The advantage
of the SRF method compared to the pq method lies in the fact that the voltage
harmonics no longer have a significant influence on the identified currents. The
principle of the SRF method is described in the following.
Let ic1 , ic2 , and ic3 be the load currents of a three-phase system without a zero
sequence component. The transformation of Concordia makes it possible to obtain
14 Voltage Unbalance Compensation in AC Microgrids 343

the load currents in the stationary reference frame, as illustrated previously by the
relation (14.1).    
By generating the signals cos θ̂ and sin θ̂ from the network voltage, using a
phase-locked loop (PLL), the load currents in the dq reference frame are obtained:
  ⎡  ⎤
⎡ ⎤
sin θ̂ − cos θ̂ ⎡ ⎤
id
⎢ ⎥ iα
⎣ ⎦=⎢ ⎥⎣ ⎦ (14.9)
⎣     ⎦
iq cos θ̂ sin θ̂ iβ

with θ̂ the angular position of the fundamental network voltage, estimated by the
PLL. These components can then be expressed as the sum of a DC component and
an AC component:

⎡ ⎤ ⎡ ∼ ⎤
id id + id
⎣ ⎦=⎢
⎣
⎥
⎦ (14.10)
∼
iq iq + iq

∼ ∼
with id and iq the DC components of id and iq , and id and iq the AC components
of id and iq . From (14.9), the load currents in the stationary reference frame can be
expressed as follows:
  ⎡   ⎤−1 ⎡    ⎤
⎡ ⎤
sin θ̂ − cos θ̂ ⎡ ⎤ sin θ̂ cos θ̂ ⎡ ⎤
iα
⎢ ⎥ i d ⎢ ⎥ id
⎣ ⎦=⎢ ⎥ ⎣ ⎦=⎢ ⎥⎣ ⎦
⎣     ⎦ ⎣    ⎦
iβ cos θ̂ sin θ̂ iq − cos θ̂ sin θ̂ iq
(14.11)

Or again:
 ⎡  ⎤ ⎡    ⎤⎡ ⎤
⎡ ⎤
sin θ̂ cos θ̂ ⎡ ⎤ sin θ̂ cos θ̂
∼
iα
⎢ i
⎥ d ⎢ ⎥ ⎢ id ⎥
⎣ ⎦=⎢ ⎥⎣ ⎦ + ⎢ ⎥
⎣    ⎦ ⎣     ⎦⎣ ∼ ⎦
iβ − cos θ̂ sin θ̂ iq − cos θ̂ sin θ̂ iq
(14.12)

Considering the simultaneous compensation of the current harmonics and the

reactive power, after adding the current icd necessary for the regulation of the DC
∼
voltage vdc , to the AC component id , the Eq. (14.12) becomes:
344 S. Karimi et al.

Regulator

abc LPF
Eq. 9 −
αβ αβ
Eq. 13
abc

PLL

Fig. 14.3 Generation of reference currents by the SRF method

 
⎡  ⎤⎡ ⎤
⎡ ⎤
sin θ̂ cos θ̂ ∼
iref α⎢ ⎥⎢ d
i + i
⎣ ⎦=⎢ ⎥
cd
⎥
⎣     ⎦⎣ ⎦ (14.13)
irefβ − cos θ̂ sin θ̂ iq

Then, using (14.8) the reference currents in the abc reference frame are obtained.
Figure 14.3 illustrates the generation of reference currents for the simultaneous
compensation of the current harmonics and the reactive power by the SRF method.

14.3.2 Performance of Pq and SRF Methods in Ideal

and Nonideal Conditions

In this subsection, the comparative analytical study of the performance of pq and

SRF methods in ideal and nonideal conditions is presented. To conduct this study,
we will consider the classic example of a three-phase network connected to a three-
phase thyristor rectifier bridge. The rectifier current ic can be represented using
Fourier series according to the following equation:
 
√ sin (ωt−α) − 15 sin [5 (ωt−α)] − 17 sin [7 (ωt−α)] + 11
1 sin [11 (ωt−α)] +
ic = 2I1 1 sin [13 (ωt−α)] − 1 sin [17 (ωt−α)] − 1 sin [19 (ωt−α)] + . . .
13 17 19
(14.14)

with α the thyristor commutation angle and I1 the effective value of the fundamental
of current defined by:
√
6
I1 = Id (14.15)
π
where Id is the load average current.
14 Voltage Unbalance Compensation in AC Microgrids 345

As seen in (14.14), the harmonic currents are of rank (h = 6 k ± 1) with k integer,

and that the effective value of each harmonic current is inversely proportional to its
rank (Ih = I1 /h).

14.3.2.1 Case of the pq Method

Ideal Case: Balanced Harmonic Currents and Balanced Sinusoidal Voltages

In the ideal case, the nonlinear load currents are balanced and defined by the
Eq. (14.14). As for the network voltages, they are assumed to be sinusoidal and
balanced:
⎡ ⎤
⎡ ⎤ sin (ωt)
vs1 ⎢ ⎥
⎢ ⎥ ⎢  ⎥
⎢ ⎥ √ ⎢ 2π ⎥
⎢ ⎥ ⎢ sin ωt − 3 ⎥
⎢ vs2 ⎥ = 2Vs ⎢ ⎥ (14.16)
⎢ ⎥ ⎢ ⎥
⎣ ⎦ ⎢ ⎥
⎢ ⎥
vs3 ⎣   ⎦
sin ωt + 2π
3

By applying the Concordia transformation, the currents and voltages along the
axes αβ are obtained:
⎡ ⎤ ⎡ ⎤
vsα √ sin (ωt)
⎣ ⎦ = 3Vs ⎣ ⎦ (14.17)
vsβ − cos (ωt)

⎡ ⎤⎛⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎞
iα
√ sin (ωt) sin (ωt) sin (ωt)
⎣ ⎦ = 3I1 ⎝⎣ ⎦ −1 ⎣ ⎦− 1⎣ ⎦ + ...⎠
5 7
iβ − cos (ωt) − cos (ωt) − cos (ωt)
(14.18)

The instantaneous active power, p, can be calculated from (14.3), (14.17), and
(14.18) and is defined by Eq. (14.19):

3Vs I1 3Vs I1
p = 3Vs I1 cos α + cos (6ωt − 5α) − cos (6ωt − 7α) + . . .
5 7
(14.19)

The first term of this equation represents the average active power, noted p,
related to the fundamental of the current, while the sum of the other terms represents
∼
an AC power, noted p, generated by the harmonic currents. From (14.19), we can
346 S. Karimi et al.

h
0 1 5 6 7 11 12 13 17 18 19
Harmonic rank

Fig. 14.4 Harmonic components of the instantaneous active power in the ideal conditions

distinguish the harmonic components of the instantaneous active power p. Figure

14.4 shows these harmonic components and their origins. A similar figure can
be obtained for instantaneous reactive power. As seen in Fig. 14.4, the current
harmonics of ranks (h = 6 k ± 1) give rise to alternative powers with multiple
pulses of 6. In this case, to generate the reference currents, it suffices to filter the
DC components of p and q.

Balanced Harmonic Currents and Unbalanced Sinusoidal Voltages

The network voltages are now considered to be unbalanced and defined by the
following matrix relation:
⎡ ⎤ ⎡ ⎤
⎡ ⎤ sin (ωt) sin (ωt)
vs1 ⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢  ⎥ ⎢   ⎥
⎢ ⎥ √ ⎢ 2π ⎥ ⎢ ⎥
⎢ ⎥ ⎢ sin ωt − 3 ⎥ √ − ⎢ sin ωt + 2π ⎥
⎢ vs2 ⎥ = 2V + ⎢ ⎥ + 2V ⎢ 3 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ ⎦ ⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
vs3 ⎣   ⎦ ⎣   ⎦
sin ωt + 2π
3 sin ωt − 2π
3
(14.20)

where V+ and V− represent the positive and negative sequence components of

the network voltage, respectively. The load current has been defined according to
(14.14). By applying the Concordia transformation, we obtain along the αβ axes the
currents already defined in (14.18) and the following voltages:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
vsα √ + sin (ωt) √ − sin (ωt)
⎣ ⎦ = 3V ⎣ ⎦ + 3V ⎣ ⎦ (14.21)
vsβ − cos (ωt) cos (ωt)

In this case, we will also establish the analytical expression of the active power
p and generalize the results obtained to the reactive power. Active power can be
calculated from Eqs. (14.3), (14.18), and (14.21) and is expressed according to the
relation (14.22):
14 Voltage Unbalance Compensation in AC Microgrids 347

h
0 1 2 4 5 6 7 8 10 11 12 13 14 16 17 18 19
Harmonic rank

Fig. 14.5 Harmonic components of the instantaneous active power (Balanced harmonic currents
and unbalanced sinusoidal voltages)

 3V + I1 3V − I

+ 3V − I1 cos (2ωt − α) + cos (6ωt − 5α) − 5 1 cos (4ωt − 5α)
p = 3V I1 cos α + 3V + I1
5
3V − I1
− 7 cos (6ωt − 7α) + 7 cos (8ωt−7α) + . . .
(14.22)

The first term of this equation represents the average active power, while the
sum of the other terms represents an AC power originating from the composition
of the harmonic currents and the positive and negative sequence voltages. From
(14.22), we obtain the harmonic components of the instantaneous active power.
Figure 14.5 shows these harmonic components and their origins. In Fig. 14.5, we
find in red the contribution of the harmonic currents of the load, composed with
the positive sequence voltage (see Fig. 14.4). Furthermore, the composition of the
negative sequence voltage with the fundamental of the load current generates an
AC power 2ω. In addition, the negative sequence voltage composed with harmonic
currents induces alternative powers of (4ω, 8ω, 10ω, 14ω, 16ω, . . . ). An identical
result can be obtained for instant reactive power.
From these analyzes, it appears that the unbalanced voltages are at the origin
of the harmonic component of rank 2 of the instantaneous powers. This can
lead to an incorrect value when identifying the reference currents because a LPF
(conventionally used in the pq method) is not effective for eliminating this harmonic
component close to its cutoff frequency and has a significant residue.

14.3.2.2 Case of the SRF Method

The performance of the SRF method strongly depends on the performance of the
PLL implemented and intended to generate sine and cosine signals, synchronous
with the network. In this subsection, we will analyze the behavior of the con-
ventional PLL under different conditions: balanced or unbalanced sinusoidal or
non-sinusoidal source voltages. The block diagram of conventional PLL is presented
in Fig. 14.6.
348 S. Karimi et al.

∧ ∧

− Regulator ⁄
−

∧ ∧

abc
αβ

Fig. 14.6 Structure of PLL

Balanced Sinusoidal Source Voltages

In this case, the network voltages along the axes αβ are defined by (14.17). The
expression of these voltages along the axes dq is then as follows:
  ⎡  ⎤
⎡ ⎤
cos θ̂ sin θ̂ ⎡ ⎤
vsd ⎢ ⎥ vsα
⎣ ⎦=⎢ ⎥⎣ ⎦ (14.23)
⎣    ⎦
vsq − sin θ̂ cos θ̂ vsβ

From (14.17) and (14.23), we obtain:

√  
vsd = 3Vs sin θ − θ̂ (14.24)

with θ = ωt.
Knowing that the value of (θ − θ̂ ) is small, (14–24) can be approximated by:
√  
vsd ≈ 3Vs θ − θ̂ (14.25)

The eq. (14.25) shows that in the case where the network voltages are sinusoidal
and balanced, an efficient regulator allows the PLL to accurately estimate the
angular position.

Unbalanced Sinusoidal Source Voltages

In this case, the network voltages along the axes αβ are defined by (14.21). From
eqs. (14.21) and (14.23), we get:
14 Voltage Unbalance Compensation in AC Microgrids 349

√   √  
vsd = 3V + sin θ − θ̂ + 3V − sin θ + θ̂ (14.26)

The first term in (14.26) represents the influence of the positive sequence
component of the voltage, while the second term represents the influence of its
negative sequence component. Considering (14.26), we can conclude that vsd , and
therefore the angular position estimated by the PLL, will both be affected by the
negative sequence component of the voltage.

Balanced Source Voltages Containing Harmonics

In this case, we consider that the network voltages contain harmonics. They can
therefore be written:
 
vs1
vs2 vs3
⎛⎡ ⎤ ⎡ 1 sin (5ωt) ⎤ ⎡ 1 sin (7ωt) ⎤ ⎞
sin (ωt) 5 7
⎜⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎟
⎜⎢   ⎥ ⎢ +  , ⎥ ⎢ +  , ⎥ ⎟
⎜⎢ ⎥ ⎢ 1 ⎥ ⎢ 1 ⎥ ⎟
√ ⎜ ⎢ sin ωt − 2π ⎥ ⎢ 2π ⎥ ⎢ 2π ⎥ ⎟
⎜⎢ 3 ⎥ ⎢ 5 sin 5 ωt − 3 ⎥ ⎢ 7 sin 7 ωt − 3 ⎥ ⎟
= 2Vs ⎜ ⎢ +
⎥ ⎢ +
⎥ ⎢ ⎥ + . . . ⎟ (14.27)
⎜⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎟
⎜⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎟
⎜⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎟
⎝⎣   ⎦ ⎣ +  , ⎦ ⎣ +  , ⎦ ⎠
sin ωt + 2π 3
1 sin 7 ωt + 2π
5 3
1 sin 7 ωt + 2π
7 3

By applying the Concordia transformation, we obtain the following equation:

⎡ ⎤ ⎛⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎞
vsα √ sin (ωt) sin (5ωt) sin (7ωt)
⎣ ⎦ = 3Vs ⎝ ⎣ 1
⎦+ ⎣ 1
⎦+ ⎣ ⎦+...⎠
5 7
vsβ − cos (ωt) cos (5ωt) cos (7ωt)
(14.28)

From (14.28), one can deduce:

  √   √3V  
√ 3Vs s
vsd = 3Vs sin θ − θ̂ + sin 5θ + θ̂ + sin 5θ − θ̂ + . . .
5 7
(14.29)

As seen from (14.29), the voltage harmonics induce alternative components that
affect the performance of the conventional PLL. Some solutions are introduced to
improve the performances of PLL in the nonideal conditions which are beyond the
reach of this chapter. However, interested readers can refer to the references [13–
16].
350 S. Karimi et al.

14.4 Imbalance Compensation

The voltage unbalance has negative impacts on the equipment such as induction
motors, electronic converters, and adjustable speed drives. Additionally, the voltage
unbalance causes more power losses in the distribution networks. Therefore, to
compensate for the unbalanced voltage, the negative sequence component of the
unbalanced loads must be effectively compensated. In order to achieve this purpose,
it is necessary to properly extract the negative sequence component of the unbal-
anced loads. In the literature, various methods have been presented for controlling
the power converters under unbalanced conditions, such as resonant controllers [17,
18], hysteresis current controllers [19, 20], direct power control methods [21, 22],
and model-based predictive controllers [23]. In this section, however, some methods
that are introduced for extracting of the positive and negative sequence components
are briefly described. Generally, positive and negative sequence extraction could be
grouped into two main approaches: direct extraction methods and indirect extraction
methods. In direct extraction methods, the instantaneous positive and negative
sequences are directly extracted by using Lyon transform. In indirect extraction
methods, the instantaneous positive and negative sequences are estimated by using
PLL or Frequency Locked-Loop (FLL).

14.4.1 Direct Extraction Methods

14.4.1.1 In Abc Frame

Direct extraction methods are based on the symmetric components, which are per-
formed in the frequency domain by the Fortescue transform and in the time domain
by the Lyon transform [24, 25]. The symmetrical components can decompose the
steady-state phasors of a generic three-phase system into the positive, negative, and
zero sequence components.
Using this principle, the positive, negative, and zero sequence phasors of phase a
of unbalanced currents can be calculated by the Fortescue transform as follows:
⎡−
→ ⎤ ⎡ ⎤⎡−→ ⎤
I +(a) 1 a a2 I a
−
→ −
→ ⎢−
→ ⎥ 1 ⎢−
→ ⎥
I +−0(a) = [T+−0 ] I abc ; ⎣ I −(a) ⎦ = ⎣ 1 a 2 a ⎦ ⎣ I b ⎦ (14.30)
−
→ 3
1 1 1 −
→
I 0(a) I c

where a = ej2π /3 = 1  120◦ is Fortescue operator.

Lyon extended the work of Fortescue and applied the symmetrical components
in the time domain. Using this principle, the instantaneous positive and negative
sequence components of the following set of three-phase unbalanced sinusoidal
waveforms are given by
14 Voltage Unbalance Compensation in AC Microgrids 351

⎡ ⎤
ia
iabc = ⎣ ib ⎦ = iabc
+ −
+ iabc + iabc
0
(14.31)
ic

⎡ ⎤ ⎡ ⎤⎡ ⎤
ia+ 1 a a2 ia
+ 1
iabc = [T+ ] iabc ; ⎣ ib+ ⎦ = ⎣ a 2 1 a ⎦ ⎣ ib ⎦ (14.32)
3
ic+ a a2 1 ic

⎡ ⎤ ⎡ ⎤⎡ ⎤
ia− 1 a2 a ia
− 1
iabc = [T− ] iabc ; ⎣ ib− ⎦ = ⎣ a 1 a 2 ⎦ ⎣ ib ⎦ (14.33)
3
ic− a2 a 1 ic

14.4.1.2 In αβ Frame

Regarding positive and negative sequence components and by using the non-
normalized Clarke transformation, the unbalanced currents in the αβ reference
frame can be calculated as follows:
+ ,T
Iαβ = Iα Iβ = Tαβ Iabc (14.34)

with
0  
2 1−√
1
−√12
Tαβ = 2 (14.35)
3 0 23 − 23

Hence, the instantaneous positive and negative sequence currents in the αβ

reference frame can be obtained by:
 
+ + T 1 1 −q
iαβ = Tαβ iabc = Tαβ [T+ ] iabc = Tαβ [T+ ] Tαβ iαβ = iαβ
2 q 1
(14.36)

 
− − T 1 1 q
iαβ = Tαβ iabc = Tαβ [T− ] iabc = Tαβ [T− ] Tαβ iαβ = iαβ
2 −q 1
(14.37)
π
where q = e−j 2 is a phase-shift time-domain operator to obtain in the quadrature
version (90◦ -lagging) of an original waveform. It is necessary to mention the
352 S. Karimi et al.

+
- [ −
]
++

[ ]

−
+
+ −
−
− [ ]
+
-

Fig. 14.7 Instantaneous positive and negative sequence calculation in αβ reference frame

homopolar component does not have effect on the instantaneous positive and
negative sequence components in the stationary reference frame. Figure 14.7 shows
the block diagram of the instantaneous positive and negative sequence calculation
in the αβ reference frame.

14.4.1.3 In dq Frame

The positive and negative sequences of the unbalanced currents in the dq reference
frame are obtained by using the Park’s transformation:
+ ,T
idq = id iq = Tdq iαβ (14.38)

 
cos θ sin θ
Tdq = (14.39)
− sin θ cos θ

Thus, the instantaneous positive and negative sequence currents in the dq

reference frame can be deduced by:
 
+ + + cos θ + sin θ + +
idq = Tdq+ iαβ = Tdq θ + iαβ = i (14.40)
− sin θ + cos θ + αβ

 
− − − cos θ − sin θ − −
vdq = Tdq− iαβ = Tdq θ − iαβ = i (14.41)
− sin θ − cos θ − αβ
14 Voltage Unbalance Compensation in AC Microgrids 353

+
- [ ]
++

[ ]

−
−
+
+
− [ ]
+ −
-

Fig. 14.8 Instantaneous positive and negative sequence calculation in dq reference frame

Figure 14.8 shows the block diagram of the instantaneous positive and negative
sequence calculation in dq reference frame.

14.4.2 Indirect Extraction Methods

In direct extraction methods, it is supposed that the unbalanced currents are

sinusoidal. Therefore, the presence of harmonic components causes inappropriate
separation of positive and negative sequences. In the presence of harmonic com-
ponents, the use of indirect extraction methods is more efficient. Some of these
methods are briefly described in the following.

14.4.2.1 Double Synchronous Reference Frame (DSRF)

In the DSRF method, to separate the positive sequence component, the current
signal in the αβ frame must be rotated with the system angular frequency ω in
the counterclockwise direction by applying the Park’s transformation. Also, for
extracting the negative sequence component, the current signal in the αβ frame must
be rotated in the clockwise direction (see Fig. 14.9). After rotating, the positive and
negative sequence components in the dq frame comprise two components, a DC
component and an AC component as follows [26]:

+ +
, + −
+ + −
idq = e−j θ · iαβ = idq + e−j (θ −θ ) · idq (14.42)
3456 3 45 6
dc term AC term
354 S. Karimi et al.

abc

− − −

Fig. 14.9 DSRF structure

+ −
, −j (θ − −θ + )
− − +
idq = e−j θ · iαβ = idq + ee · idq (14.43)
3456 3 45 6
dc term AC term

where
 
id+ = I + cos δ + − ϕ +
 
iq+ = I + sin δ + − ϕ +
  (14.44)
id− = I − cos δ − − ϕ −
 
iq− = I − sin δ − − ϕ −

which that

θ + = ωt + ϕ +
(14.45)
θ − = −ωt + ϕ −

And the park’s transformation is given by

+ ,  cos θ sin θ

ej θ = (14.46)
− sin θ cos θ

In these equations, ϕ+ and ϕ− are the initial phases of the positive and negative
sequence of the measured voltage, respectively. Also, δ + and δ − are the initial phases
of the positive and negative sequence of the current.
According to (14.42), (14.43), and (14.45), from the positive sequence reference
frame point of view, the negative sequence is rotating at a double frequency and vice
versa. Generally, to eliminate this oscillatory component a low pass filter is used.
However, the low pass filter cannot completely eliminate the oscillatory component.
Some methods, such as decoupled double synchronous reference frame (DDSRF),
have been proposed for eliminating the oscillating term and decoupling the positive
and negative sequences [26, 27].
14 Voltage Unbalance Compensation in AC Microgrids 355

+
- +- ′

SOGI

Fig. 14.10 SOGI scheme [29]

14.4.2.2 DSOGI-FLL

There are various ways for quadrature-signals generation. One of the easy and
effective methods is using the second-order generalized integrator (SOGI). The
SOGI scheme is shown in Fig. 14.10. The transfer functions of a SOGI are as follows
[28]:

i (s) ks ω0 s
D(s) = = 2 (14.47)
i(s) s + ks ω0 s + ω0 2

qi (s) ks ω0 2
Δ(s) = = 2 (14.48)
i(s) s + ks ω0 s + ω0 2

where ω0 and ks set resonance frequency and damping factor, respectively.

Transfer functions (14.47) and (14.48) reveal that if i is a sinusoidal signal,
i and qi will be sinusoidal as well. Moreover, qi will be always 90-lagging
i , independently of both the frequency of i and the values of ω0 and ks . These
characteristics make the SOGI ideal for the quadrature signals generation.
A dual SOGI (DSOGI) is necessary to perform the positive and negative
sequences estimation. On the other hand, only when the ω0 corresponds to the
grid frequency, the estimated components will be correct. Therefore, it is necessary
for a frequency adaptation system to tune the SOGIs with the grid frequency. For
this purpose, the frequency locked-loop (FLL) can be used [29]. DSOGI-FLL has
good performance, even under imbalance and harmonic conditions. The basis of
the DSOGI is the concept of symmetrical components that was discussed earlier.
The DSOGI-FLL diagram is shown in Fig. 14.11. As can be seen in Fig. 14.11,
DSOGI-FLL consists of two SOGI and one FLL.
356 S. Karimi et al.

′
′

FLL +
- ′

′ ′ ++
′
′
′

′
SOGI-QSG (α)
′

′ −′
′ +
+
′
−′
+
-
SOGI-QSG (β)

DSOGI

DSOGI-FLL

Fig. 14.11 DSOGI-FLL block diagram

′
+ PI ++
- ′
[ ]
′
++
′
′
′

SRF-PLL
′
−
′
−′
− −
+
+
DSOGI-QSG − [ ] −
+
-

Fig. 14.12 Block diagram of the DSOGI-PLL [29]

14.4.2.3 DSOGI-PLL

Figure 14.12 shows the block diagram of DSOGI-PLL. It is clear from Fig. 14.12
that the DSOGI-PLL is similar to its DSOGI-FLL counterpart, except that it is in
the rotating reference frame. Hence PLL is used instead of FLL in the DSOGI-PLL
structure.
14 Voltage Unbalance Compensation in AC Microgrids 357

14.4.3 Instantaneous Power under Unbalanced Conditions

According to instantaneous power theory, complex power injected into the grid is
defined and given by:

s = p + j q = V I ∗ = Vsdq Idq
∗
(14.49)

where V = [va . vb . vc ], I = [ia . ib . ic ]T , Vsdq = [vsd . vsq . vs0 ], Idq = [id . iq . i0 ]T .

Considering the voltage and current sequence components, (14.49) can be
represented as:
  ∗
+ − + −
s = ej ωt Vsdq + e−j ωt Vsdq ej ωt Idq + e−j ωt Idq (14.50)

By separating the real and imaginary parts:

∼
p = real {s} = P + p; p(t) = P + Pc cos (2ωt) + Ps sin (2ωt) (14.51)

∼
q = img {s} = Q + q; q(t) = Q + Qc cos (2ωt) + Qs sin (2ωt) (14.52)

where P and Q are the average parts of the active and reactive powers, respectively,
∼ ∼
and p and q are the oscillatory parts of the active and reactive powers, respectively.
The average active and reactive power parts are as follows:
+ + + + − − − −
P = Vsd Id + Vsq Iq + Vsd Id + Vsq Iq (14.53)

+ + + + − − − −
Q = −Vsd Iq + Vsq Id − Vsd Iq + Vsq Id (14.54)

And the oscillatory parts are deduced using the following equations:
+ − + − − + − +
Pc = Vsd Id + Vsq Iq + Vsd Id + Vsq Iq (14.55)

+ − + − − + − +
Ps = Vsd Iq − Vsq Id − Vsd Iq + Vsq Id (14.56)

+ − + − − + − +
Qc = −Vsd Iq + Vsq Id − Vsd Iq + Vsq Id (14.57)

+ − + − − + − +
Qs = Vsd Id + Vsq Iq − Vsd Id − Vsq Iq (14.58)
358 S. Karimi et al.

As one can see from (14.51) and (14.52), the frequency of the oscillatory terms
is two times the grid frequency. In addition, as can be seen from (14.55) to (14.58),
there are interactions between positive and negative sequence current and voltages.
Since, for compensating the unbalanced load currents, the negative sequence current
must be injected by the inverter-interfaced DG, the power oscillation is inevitable,
even if the voltage becomes quite balanced.

14.5 Case Study

In this section, simulation demonstrations are presented to analyze the characteris-

tics of the approaches previously described. In Fig. 14.13, the structure of the studied
MG is shown. The studied MG is operated in islanded mode and comprises two
inverter-interfaced DGs. DG#1 acts as a grid-forming converter and DG#2 operates
as a grid-feeding converter that provides the ancillary services for power quality
improvement. DG#1 regulates the voltage magnitude and frequency, and DG#2 is
responsible for compensating the harmonic and imbalanced currents. In addition,
DG#2 delivers active power to the MG. The harmonic load LH considered in this
simulation is a nonlinear load consists of a three-phase diode bridge rectifier with
the constant DC current equal to 80 A. To simulate the three-phases unbalanced load
(LUB ), two sinusoidal current sources are considered for phase a and phase b, and
phase c is unconnected. The other parameters of the studied MG are listed in Table
14.2.
The block diagram of the control system for the grid-forming and grid-feeding
converters are illustrated in Figs. 14.14 and 14.15, respectively. In the grid-

PCC

Grid

DG #2
DG #1

Grid-feeding
Converter Grid-forming
Converter

Fig. 14.13 Studied MG

14 Voltage Unbalance Compensation in AC Microgrids 359

Table 14.2 System Parameters Value Parameters Value

parameters
Rf 1.19 m Vdc 1500 V
Lf 100 μH vdref 391 V
Cf 2500 μF vqref 0
RL 1 m LB1 40 kW
LL 4 mH LB2 200 kW
f 50 Hz Pref 100 kW

abc
dq

abc
dq

dq Pulses
Voltage Control Current Control
PWM
Loop Loop abc

Fig. 14.14 The block diagram of the control system for DG#1

dq
⁄
abc

+
abc - PI -1
dq
+
Negative Sequence
+
Extraction Pulses
+ Hysteresis

Fig. 3

Fig. 14.15 The block diagram of the control system for DG#2

forming converter control system (Fig. 14.14), the desirable voltage magnitude and
frequency are sent to the voltage and current control loops, and then translated using
dq/abc transformation to the abc reference frame and given to the PWM block to
generate suitable switching signals.
The grid-feeding converter control system consists of four main parts (Fig.
14.15). The first and the second parts are the conventional control blocks that are
used in the grid-feeding converters, and the third and the fourth parts perform the
desired ancillary controls. The first part is used to deliver the specified active power
to MG. In this part, the reference power is converted to the reference current in
d axis after division by 3/2vdref . The voltage magnitude control is also the task
360 S. Karimi et al.

Fig. 14.16 Unbalanced load current

of the second part. In this part, the reference current in q axis is generated by
using the voltage error and a PI controller. The third part deals with the harmonic
current compensation caused by the nonlinear load. This part generates the reference
currents required to compensate for the harmonic currents, using the scheme shown
in Fig. 14.3. The fourth part is responsible for compensating the unbalanced currents
caused by an unbalanced load. In this part, using one of the indirect extraction
methods, the negative sequence reference currents are generated. Finally, these
reference currents are added together and given to the hysteresis controller to
generate the suitable switching signals for the inverter-interfaced DG#2.
The unbalanced load is connected to the MG at t = 0.25 Sec. The unbalanced
load current is shown in Fig. 14.16. The peak of the unbalanced current is 80 A. The
nonlinear load also is connected at t = 0.4 Sec. Figure 14.17 depicts the current of
this load. Figure 14.18 to Fig. 14.20 illustrate the DG#2 output voltage when DG#2
operates with the conventional control system (without ancillary controls). Figure
14.18 shows the DG#2 output voltage before the unbalanced and nonlinear loads
are connected. As one can see, the voltage is properly controlled and the voltage
peak is about 391 V. Figure 14.19 and Fig. 14.20 present the DG#2 output voltage
after connecting the unbalanced and nonlinear loads, respectively. As can be seen,
in the absence of ancillary controls, the voltage becomes unbalanced and distorted.
The VUF and THD of the DG#2 output voltage under these conditions are presented
in Fig. 14.21. The THD and VUF are calculated using the following formulas:
#
M 2
h=2 Vh
T H D% = × 100% (14.59)
V1

where Vh is the RMS of h-order harmonic.

14 Voltage Unbalance Compensation in AC Microgrids 361

Fig. 14.17 Nonlinear load current

Fig. 14.18 Voltage before unbalanced and nonlinear load connection


vd− + vq−
2 2
V UF% =  × 100% (14.60)
vd+ + vq+
2 2

Figure 14.21 shows that both VUF and THD exceeded their permissible limit,
and the VUF and THD reach about 8% in the steady state. It is important to
remark that in the distribution networks, the THD and VUF must be limited under
362 S. Karimi et al.

Fig. 14.19 Voltage after unbalanced load connection

Fig. 14.20 Voltage after nonlinear load connection

5% and 2%, respectively. Figure 14.22 presents the results of the FFT analysis of
the DG#2 output voltage in the steady state. Active and reactive output power of
the DG#1 and DG#2 are depicted in Fig. 14.23 and Fig. 14.24, respectively. As
seen, the unbalanced and nonlinear load affects the active and reactive powers and
causes power oscillations. In addition, under unbalanced and distorted conditions,
the DG#2 does not properly follow its reference to active power.
Figure 14.25 to Fig. 14.29 illustrate the results after adding ancillary control.
Figure 14.25 shows the DG#2 output voltage after compensation. As seen, the
14 Voltage Unbalance Compensation in AC Microgrids 363

(a)

(b)

Fig. 14.21 VUF and THD of the DG#2 output voltage

voltage is appropriately compensated and controlled. The VUF and THD of the
DG#2 output voltage after compensation are depicted in Fig. 14.26. As one can see,
the mentioned ancillary control can effectively reduce the VUF and THD, and they
are less than the permitted amount. Figure 14.27 shows the results of FFT analysis in
the steady state under these conditions. This figure illustrates that using the ancillary
control the THD in the steady state is 1.45%. Active and reactive output power of
the DG#1 and DG#2 after compensation are presented in Fig. 14.28 and Fig. 14.29.
As indicated in Fig. 14.28, since the unbalanced voltage is compensated, the power
364 S. Karimi et al.

Fig. 14.22 Voltage harmonic components

oscillations no longer exist in the active and reactive output power of the DG#1.
It is important to note that the power oscillations in the active and reactive output
power of DG#2 are generated due to the injection of the negative sequence current
by DG#2.
14 Voltage Unbalance Compensation in AC Microgrids 365

(a)

(b)

Fig. 14.23 Active and reactive output power of the DG#1

366 S. Karimi et al.

(a)

(b)

Fig. 14.24 Active and reactive output power of the DG#2

14 Voltage Unbalance Compensation in AC Microgrids 367

Fig. 14.25 Voltage after compensation

368 S. Karimi et al.

(a)

(b)

Fig. 14.26 VUF and THD after compensation

14 Voltage Unbalance Compensation in AC Microgrids 369

Fig. 14.27 Voltage harmonic components after compensation

370 S. Karimi et al.

(a)

(b)

Fig. 14.28 Active and reactive output power of the DG#1 after compensation
14 Voltage Unbalance Compensation in AC Microgrids 371

(a)

(b)

Fig. 14.29 Active and reactive output power of the DG#2 after compensation
372 S. Karimi et al.

References

1. Sabzevari, K., Karimi, S., Khosravi, F., & Abdi, H. (2019). Modified droop control for
improving adaptive virtual impedance strategy for parallel distributed generation units in
islanded microgrids. International Transactions on Electrical Energy Systems, 113, 758–771.
2. Pal, R., & Gupta, S. (2019). Topologies and control strategies implicated in dynamic voltage
restorer (DVR) for power quality improvement. Iranian Journal of Science and Technology,
Transactions of Electrical Engineering. https://doi.org/10.1007/s40998-019-00287-3.
3. Guerrero, J. M., Chandorkar, M., Lee, T. L., & Loh, P. C. (2012). Advanced control
architectures for intelligent microgrids—Part I: Decentralized and hierarchical control. IEEE
Transactions on Industrial Electronics, 60(4), 1254–1262.
4. Naderi, Y., Hosseini, S. H., Ghassem Zadeh, S., Mohammadi-Ivatloo, B., Vasquez, J. C., &
Guerrero, J. M. (2018). An overview of power quality enhancement techniques applied to
distributed generation in electrical distribution networks. Renewable and Sustainable Energy
Reviews, 93, 201–214.
5. Y. Naderi, S.H. Hosseini, S.G. Zadeh, B. Mohammadi-Ivatlo, J.C. Vasquez, J.M. Guerrero,
“Distributed power quality improvement in residential microgrids”, 10th International Confer-
ence on Electrical and Electronics Engineering (ELECO), 2017, pp. 90–94.
6. IEEE recommended practice and requirements for harmonic control in electric power systems,
IEEE Std. 519-2014 (Revision of IEEE Std 519-1992), 2014, pp. 1-213.
7. Singh, B., Al-Haddad, K., & Chandra, A. (1999). A review of active filters for power quality
improvement. IEEE Transactions on Industrial Electronics, 46(5), 960–971.
8. Asiminoaei L, Blaabjerg F, Hansen S. “Evaluation of harmonic detection methods for active
power filter applications”, Twentieth Annual IEEE Applied Power Electronics Conference and
Exposition, (APEC 2005), 2005 (Vol. 1, pp. 635-641).
9. Akagi H. “Generalized theory of the instantaneous reactive power in three-phase circuits”,
IEEJ IPEC-Tokyo’83. 1983.
10. Chang, G. W., & Shee, T. C. (2002). A comparative study of active power filter reference
compensation approaches. IEEE Power Engineering Society Summer Meeting, 2, 1017–1021.
11. Ordonez R, Karimi C, Oueidat M, Sadarnac D. “Comparison of current reference generation
techniques for non-active power compensation under distorted and unbalanced conditions”,
32nd Annual Conference on IEEE Industrial Electronics, (IECON 2006), 2006.
12. Bhattacharya S, Divan DM, Banerjee B. “Synchronous frame harmonic isolator using active
series filter”, European conference on power electronics and applications 1992.
13. Freijedo, F. D., Yepes, A. G., Lopez, O., Vidal, A., & Doval-Gandoy, J. (2010). Three-
phase PLLs with fast post fault retracking and steady-state rejection of voltage unbalance and
harmonics by means of lead compensation. IEEE Transactions on Power Electronics, 26(1),
85–97.
14. Campanhol, L. B., da Silva, S. A., & Goedtel, A. (2014). Application of shunt active
power filter for harmonic reduction and reactive power compensation in three-phase four-wire
systems. IET Power Electronics., 7(11), 2825–2836.
15. Han, Y., Luo, M., Zhao, X., Guerrero, J. M., & Xu, L. (2015). Comparative performance
evaluation of orthogonal-signal-generators-based single-phase PLL algorithms—A survey.
IEEE Transactions on Power Electronics., 31(5), 3932–3944.
16. Golestan, S., Guerrero, J. M., & Vasquez, J. C. (2016). Three-phase PLLs: A review of recent
advances. IEEE Transactions on Power Electronics., 32(3), 1894–1907.
17. Teodorescu, R., Blaabjerg, F., Liserre, M., & Loh, P. C. (2006). Proportional resonant con-
trollers and filters for grid-connected voltage-source converters. Electric Power Applications,
IEE Proceedings, 153(5), 750–762.
18. Lascu, C., Asiminoaei, L., Boldea, I., & Blaabjerg, F. (2007). High performance current
controller for selective harmonic compensation in active power filters. IEEE Transactions on
Power Electronics, 22(5), 1826–1835.
14 Voltage Unbalance Compensation in AC Microgrids 373

19. Stankovic, A. V., & Lipo, T. A. (2001). A novel control method for input output harmonic
elimination of the PWM boost type rectifier under unbalanced operating conditions. IEEE
Transactions on Power Electronics, 16(5), 603–611.
20. Serpa, L. A., Round, S. D., & Kolar, J. W. (2007). A virtual-flux decoupling hysteresis current
controller for mains connected inverter systems. IEEE Transactions on Power Electronics,
22(5), 1766–1777.
21. Serpa, L. A., Ponnaluri, S., Barbosa, P. M., & Kolar, J. W. (2007). A modified direct power
control strategy allowing the connection of three- phase inverters to the grid through LCL
filters. IEEE Transactions on Industry Applications, 43(5), 1388–1400.
22. Vazquez, S., Sanchez, J. A., Carrasco, J. M., Leon, J. I., & Galvan, E. (2008). A model-
based direct power control for three-phase power converters. IEEE Transactions on Industrial
Electronics, 55(4), 1647–1657.
23. Cortés, P., Rodríguez, J., Antoniewicz, P., & Kazmierkowski, M. (2008). Direct power control
of an AFE using predictive control. IEEE Transactions on Power Electronics, 23(5), 2516–
2553.
24. Lyon, W. V. (1937). Applications of the method of symmetrical components. McGraw-Hill book
Company, Incorporated.
25. Fortescue, C. L. (1918). Method of symmetrical co-ordinates applied to the solution of poly
phase networks. Transactions of the American Institute of Electrical Engineers, 37(2), 1027–
1140.
26. Rodríguez, P., Pou, J., Bergas, J., Candela, J. I., Burgos, R. P., & Boroyevich, D. (2007).
Decoupled double synchronous reference frame PLL for power converters control. IEEE
Transactions on Power Electronics., 22(2), 584–592.
27. Reyes, M., Rodriguez, P., Vazquez, S., Luna, A., Teodorescu, R., & Carrasco, J. M. (2012).
Enhanced decoupled double synchronous reference frame current controller for unbalanced
grid-voltage conditions. IEEE Transactions on power electronics, 27(9), 3934–3943.
28. Rodriguez P, Teodorescu R, Candela I, Timbus AV, Liserre M, Blaabjerg F. “New positive-
sequence voltage detector for grid synchronization of power converters under faulty grid
conditions”, 37th IEEE Power Electronics Specialists Conference 2006.
29. Rodriguez, A. L., Candela, I., Mujal, R., Teodorescu, R., & Blaabjerg, F. (Jan). Multi-resonant
frequency-locked loop for grid synchronization of power converters under distorted grid
conditions. IEEE Transactions on Industrial Electronics, 58(1), 127–138.
Chapter 15
WAM-Based Hierarchical Control of
Islanded AC Microgrids

E. S. N. Raju P and Trapti Jain

15.1 Introduction

A microgrid comprises of low voltage distributed systems with DG units, storage

devices, loads, and interconnecting switches [1, 2]. Microgrids (MGs) can be
operated either in an island mode or in a grid connected mode of operation [1, 2].
Based on the type of grid voltage, MGs can be classified into AC MicroGrids
(ACMGs), DC MicroGrids (DCMGs), and Hybrid AC/DC MicroGrids (HADMGs).
ACMGs have been proposed to utilize the existing AC grid technologies, protection,
and standards. However, power generation from various DGs such as photo-voltaic
arrays and fuel cells is DC, which needs to be converted into AC power through
power electronic interface for connecting them with the AC utility grid. This
AC power is again converted back into DC power required by today’s electrical
loads such as Uninterrupted Power Supply (UPS), fluorescent lights, variable
motor drives, and hybrid electric vehicles. Thus, an individual ACMG may be less
efficient due to more power losses occurring in multiple conversions. Besides this,
synchronization, stability, and reactive power requirement are its inherent demerits.
DCMGs are emerging as a better alternative due to the abovementioned reasons
for renewable energy based DGs. However, power generation from the sources such
as diesel generator, small hydro turbine with synchronous generator and photo-
voltaic panels, etc., as well as the electrical loads are a mix of AC and DC power.
Thus, an individual DCMG may not completely eliminate the losses occurring in
multiple stage conversions, though the losses occurring in DC/DC conversions are
lesser than those occurring in AC/DC or DC/AC conversions.

E. S. N. Raju P () · T. Jain

Discipline of Electrical Engineering, Indian Institute of Technology Indore, Indore,
Madhya Pradesh, India
e-mail: pesnraju88@gmail.com; traptij@iiti.ac.in

© Springer Nature Switzerland AG 2021 375

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_15
376 E. S. N. Raju P and T. Jain

Fig. 15.1 Typical architecture of hybrid AC/DC microgrid

In an individual ACMG, AC loads require single stage conversion and DC loads

require multiple conversions. Similarly in an individual DCMG, DC loads require
single stage conversions and AC loads require multiple stage conversions. There-
fore, a HADMG, which is a cluster of ACMG, DCMG, bidirectional converters,
control equipment, and energy management system, as shown in Fig. 15.1, has
been proposed. ACMG consists of DGs such as diesel generators, small hydro
turbine with synchronous generator, biomass based power generation, etc., which
generate AC power output, AC loads, fly wheel energy storage system with AC/AC
interface, and utility grid connection through bidirectional power electronic based
switch at the PCC. DCMG includes DGs such as photo-voltaic panels, fuel cell
tracks, etc., which produce DC power output, loads requiring DC power input
such as UPS, fluorescents lighting, etc., DC energy storage system such as battery,
super-capacitor, etc., and hybrid electric vehicles. These components are connected
to DC bus through DC/DC buck or boost converter. A bidirectional AC-DC/DC-
AC converter is required to interface ACMG and DCMG. A back-up converter is
also included to avoid any islanding of ACMG and DCMG. The main objective
of the bidirectional AC/DC converter is to maintain the smooth power transfer
between ACMG and DCMG, to maintain stable voltages of AC bus and DC bus
under varying generation and load conditions. Power transfer would take place from
ACMG to DCMG if the power generation in ACMG is more than in DCMG and vice
versa in grid connected as well as islanding modes of operation.
The various microsources in the HADMG are interfaced through five different
types of power electronic converters viz. AC/AC, AC/DC/DC, DC/DC, DC/AC, and
AC-DC/DC-AC bidirectional converter. These power electronic converters are used
to control the microsources in the grid connected as well as in the island mode
of operation. Different types of control methods are used, as shown in Fig. 15.2 to
15 WAM-Based Hierarchical Control of Islanded AC Microgrids 377

HADMG Micro Sources Control Methods

ACMG Micro Sources Control DCMG Micro Sources Control

Methods Methods

Island Mode Grid Mode

-Droop Control and

PQ control Method -Voltage/frequency control
Method

Analog Control Techniques Digital Control Techniques

-Current control
-Predictive digital current programmed
-Voltage Mode Control and control
-Current Mode Control -Variable frequency predictive control
-Sensor less current mode control and
-Predictive digital dead-beat control

Fig. 15.2 Microsources control methods in hybrid AC/DC microgrid

control the microsources in the ACMG and those in the DCMG. The microsources in
the ACMG are controlled by active and reactive power (PQ) control, droop control,
and voltage/frequency control, while the microsources in the DCMG are controlled
using analog and digital control techniques. In PQ control method, which is applied
in grid connected mode of operation, the reference values for the active and reactive
power are given by the utility grid controller to the controllers of microsources in the
ACMG. Droop control method and v/f control method, which are employed in island
mode of operation, are like primary control and secondary control, respectively, in
the transmission network. There are two types of droop control method viz. P-f
droop control method and Q-V droop control method. P-f control method controls
the frequency by controlling the active power supplied by DGs whereas the QV
droop control method controls the voltage magnitude by controlling the reactive
power supplied by the DGs. These methods can minimize the fluctuations in the
voltage and the frequency for small disturbances only. In order to minimize large
fluctuations in the voltage and frequency, v/f control method is used. V/f control
method uses PI controllers and PID controllers to damp out the oscillations in the
voltage and the frequency. The controllers shall be designed such that in island mode
of operation, it turns from PQ control to droop control for small disturbances and
from droop control to v/f control for large disturbances.
Analog control techniques utilize the comparison of output voltage and output
current with the reference value and known as voltage mode control and current
mode control, respectively. Analog control techniques are simple and cheap but
378 E. S. N. Raju P and T. Jain

suffer from low flexibility, low reliability, high complexity, and are more sensitive.
Digital control techniques are current control method, predictive digital current pro-
grammed control method, variable frequency predictive control method, sensorless
current mode control method, and predictive digital dead-beat controller for dc-dc
converters.
Thus, the HADMG seems more beneficial to facilitate the connection of various
renewable AC and DC power sources and loads with the power system in order
to minimize the conversion losses. However, the practical implementation of the
HADMG needs to overcome several technical and economical challenges as listed
below.
• Building a new DC grid and upgrading the existing AC grid is a long-term
process.
• Development of appropriate control strategies and protection schemes for safe
and reliable operation of the HADMG in grid connected as well as islanded mode
of operation.
• Designing the control coordination between the various types of power electronic
converters for power sharing among various types of DGs under varying
operating conditions.
• Redesigning of home and office products to remove the embedded AC/DC
rectifiers.
• Optimal voltage levels need to be determined for easy connection of various types
of DC loads.
Due to the above technical and economical challenges, ACMGs are still domi-
nant because of their similar intrinsic characteristics to the traditional Low Voltage
(LV) as well as Medium Voltage (MV) distribution networks. The main challenges
in ACMGs include stability issues and development of control schemes to enhance
their stability and dynamic performance.
Power system stability is defined as “the ability of an electric power system, for
a given initial operating condition, to regain a state of operating equilibrium after
being subjected to a physical disturbance, with most system variables bounded so
that practically the entire system remains intact.” [3, 4]. It has been considered as
an important issue since 1920. Stability of a power system mainly depends on the
amount of physical disturbance and operating condition prior to the disturbance.
Depending upon the amount of disturbance, viz. small-signal and large-signal,
stability can be classified as small-signal stability and transient stability, respectively
[3, 4]. The small-signal disturbances are in the form of continuous load variations,
whereas, large-signal disturbances are severe disturbances, line outages, and faults.
Power system stability is mainly classified into three categories: rotor angle
stability, frequency stability, and voltage stability, as shown in Fig. 15.3 [3, 4].
The rotor angle stability and voltage stability are further categorized into small-
signal/disturbance stability and large-signal/disturbance stability, while frequency
stability is divided into the short-term and long-term stability. This classification
has been done based on the following considerations [3, 4].
15 WAM-Based Hierarchical Control of Islanded AC Microgrids 379

Fig. 15.3 Classification of power system stability [3, 4]

• The physical nature of the resulting mode of instability as indicated by the main
system variable in which instability can be observed.
• The size of the disturbance considered, which influences the method of calcula-
tion and prediction of stability.
• The devices, processes, and the time span that must be taken into consideration
in order to assess stability.
Stability issues in ACMGs become more vital concern due to the presence
of multi-energy source based IIDG units, interaction between different types of
power electronic converters and energy storage system. Stability analysis in ACMGs
follows the similar concepts as in the existing AC grid. The stability issues in
ACMGs can be divided into small-signal, transient, and voltage stability. The
reasons behind their occurrence, shown in Fig. 15.4, and improvement methods,
shown in Fig. 15.5, are briefly given below [5].
• Small-signal stability: The small-signal stability in ACMGs can be analyzed
with a linearized model of microsources, network lines, and loads. Small-signal
instability in ACMGs may occur due to the feedback controller, continuous
load switching, system damping, and power limit of the IIDG units. Small-
signal stability can be improved by adding various supplementary control loops
around the existing control loops of IIDG units, using stabilizers with IIDG units,
coordinated control of the microsources, and energy management system.
• Transient stability: The transient stability analysis of ACMGs can be performed
with their nonlinear models [6]. The construction of the Lyapunov function [7]
is one of the methods to analyze the transient stability. A fault with subsequent
island, loss of IIDG unit, faults in the main grid or in the ACMG, large changes in
the load posses most of the transient instability problems. The improvement in the
transient stability can be achieved by controlling of Energy Storage System (ESS)
to inject active/reactive power during shortage of generation, tripping of IIDG
units, load dynamics, and islanding. Load shedding, control of power electronic
converters, and adaptive protection devices also help in improving the stability.
380 E. S. N. Raju P and T. Jain

Fig. 15.4 Reasons behind the stability problems of ACMGs

Fig. 15.5 Stability improvement methods of ACMGs

• Voltage stability: The voltage stability problem in ACMGs can be revealed from
the P-V and Q-V curves. The P-V curve shows the maximum loadability, while
the Q-V curve indicates the necessary amount of reactive power required at
the load end for the desired voltage. The main issues in the voltage stability
analysis include characteristics of the load, control strategy of reactive power,
slow increase in the power demand, and outage of one of the parts of the ACMG
network [5]. Reactive power limits/current limiters, load dynamics associated
15 WAM-Based Hierarchical Control of Islanded AC Microgrids 381

with the induction motor loads, tap changers, and voltage regulators create most
of the voltage instability problems in ACMGs. Voltage regulation with IIDG
units, reactive compensation with distributed FACT devices like DSTATCOM
[8], advanced load controller, load shedding, and modified current limiters of the
microsources can improve the voltage stability in ACMGs.
Stability of ACMGs is not a critical issue in grid connected mode of operation as
the stiff grid would be responsible for its stable operation. However, in an island
mode of operation, it is an important concern due to the low-inertial nature of
power electronics interfaced DER units. At a given steady-state operating condition,
the islanded AC microgrids (IACMGs) may be unstable when it is subjected to
the large disturbance. However, the IACMG should operate satisfactorily if it is
subjected to small-signal disturbance at a given steady-state operating condition.
Therefore, small-signal stability enhancement is a fundamental requirement for
the satisfactory and reliable operation of IACMGs. Stability of IACMGs feeding
passive, active, and dynamic loads has been enhanced by developing secondary
controllers [9] such as decentralized [10–13], centralized [14], distributed [15, 16],
and hierarchical controllers [17–19]. However, these controllers have been validated
only on a particular type of static or dynamic load and provide relatively good
performance for those load dynamics. Future microgrids (MGs) are expected to
have multiple types of static and dynamic loads such as constant power load (CPL),
rectifier interfaced active load (RIAL), and dynamic induction motor (IM) load.
Further, the stability analysis of IACMGs feeding these multiple types of static and
dynamic loads, simultaneously, addressed in [20], revealed that the presence of these
multiple types of loads introduces impedance unbalance caused by the negative
incremental impedance characteristics offered by both the CPL as well as RIAL. In
addition, the dynamic IM load introduces new low-damping high frequency, inter-
area, and unstable modes in IACMGs. Moreover, the aforementioned controllers are
not found suitable for the stability enhancement of IACMGs when these multiple
types of static and dynamic loads were present. Therefore, it is essential to design
the controller considering the dynamics of these multiple types of static and dynamic
loads in IACMGs.
Recent advancements in synchronized phasor measurement technology (SMT)
led to developing a wide area measurement system (WAMS) based hierarchical
controllers, shown in Fig. 15.6, for the stability enhancement of conventional power
system [21–31]. These hierarchical controllers have been proved to be one of the
most potentially effective solutions to enhance the stability of a conventional power
system. Traditional supervisory control and data acquisition (SCADA)/energy
management system (EMS), based on voltage and current measurements from
conventional remote terminal units (RTUs), are becoming increasingly unreliable
for real-time operations of microgrids because they cannot fully anticipate all the
conditions faced by operators. New technologies, which rely on accurate, high
resolution, real-time monitoring of actual system conditions using SMT, are needed
to support the real-time operations. For these reasons, nowadays, industries are
beginning to explore the use of SMT at the distribution level for distribution system
382 E. S. N. Raju P and T. Jain

Fig. 15.6 General structure of wide area measurement system (WAMS) based control system

visibility enhancement, integration of distributed generation, microgrid operation

and control, demand response, and power quality applications [32]. Motivated by
these factors, PMUs supported two-level hierarchical controller has been explored
for the stability enhancement of IACMGs as well [33]. However, delay-free
communication channels have been considered during the design procedure, which
may not be pragmatic in real WAMS.
In WAMS-based stabilization application, transmitting signals from PMUs to
the upper-level centralized controller, through the phasor data concentrator (PDC),
and then back to generator’s lower-level decentralized controllers involve some
time delays [23–31]. These signal transmission time delays mainly depend on the
type of communication link used. In the Western Electricity Coordinating Council
(WECC) system, these time delays vary from 25 ms (one-way from PMUs to the
upper-level centralized controller) for fiber optics cables to 250 ms for satellite-
based communication links [23]. Moreover, these time delays can deteriorate the
effectiveness of the WAMS-based two-level hierarchical controllers [23–31]. Thus,
it is essential to verify the effectiveness of PMUs supported two-level hierarchical
controller for a wide range of signal transmission time delays encountered in real
WAMS. Therefore, in this chapter, the impact of signal transmission time delays
on the performance of PMUs supported two-level hierarchical controller, designed
for the stability enhancement of IACMGs with static and dynamic loads, has been
investigated.
15 WAM-Based Hierarchical Control of Islanded AC Microgrids 383

Fig. 15.7 Schematic diagram of a studied IACMG system with the proposed hierarchical con-
troller considering signal transmission time delays

The proposed hierarchical controller is composed of a lower-level decentralized

controller for each IIDG unit helped by a multi-input-multi-output (MIMO) cen-
tralized controller at the upper level, as shown in Fig. 15.7. Both lower-level and
upper-level controllers at the two levels work together for the stability enhancement
of IACMGs. Further, the proposed hierarchical controller relies on synchronized
measurements supplied by PMUs through the WAMS. The modal-based extended
linear-quadratic-Gaussian (LQG) approach has been used to design the proposed
hierarchical controller and its performance has been compared with the hierarchical
controller based on state-based extended LQG approach. The proposed design
approach also includes time delays experienced in the transmission of signals from
PMUs to the upper-level centralized controller and from the upper-level centralized
controller to both the lower-level decentralized controller of each inverter-interfaced
384 E. S. N. Raju P and T. Jain

distributed generation (IIDG) unit as well as local controller of RIAL. Finally, the
performance of the proposed hierarchical controller has been assessed for different
time delays using both eigenvalue analysis and time-domain simulations performed
on the studied IACMG system, as shown in Fig. 15.7.

15.2 Design of WAMS-Based Hierarchical Controller

Considering Signal Transmission Time Delays

Figure 15.8 shows the proposed hierarchical controller’s design procedure con-
sidering signal transmission time delays. It is worth mentioning that the detailed
design procedure of the proposed hierarchical controller considering the delay-free
communication channels has been produced in [33]. However, the same design
procedure of the proposed hierarchical controller considering signal transmission
time delays has been discussed in this paper for enhanced readability. In the design
procedure, first, a lower-level decentralized controller is incorporated in the IACMG
system model. The lower-level decentralized controller adds additional auxiliary

Fig. 15.8 Design procedure of the proposed hierarchical controller considering signal transmis-
sion time delays
15 WAM-Based Hierarchical Control of Islanded AC Microgrids 385

control terms to the conventional power sharing/droop controller. The auxiliary

control terms are based on the desired power sharing of each IIDG unit and the total
active and reactive power generations information acquired from PMUs through
the WAMS [33]. The control commands, ωi−dec∗ ∗
and v0di−dec , from the lower-level
decentralized controller modulates the output of conventional droop controller to
generate the modified reference frequency (ωi∗ ) and d-axis voltage (v0dqi
∗ ) of i th
IIDG unit, as given in (15.1)–(15.3).
∗
ωi−dec
5  63 4

nG
ωi∗ = ωi−dr
∗
+ KP deci αP i Pi − Pi ∀i (15.1)
i=1

.  
nG

∗ ∗
v0di = v0di−dr + KQdeci αQi Qi − Qi dt ∀i (15.2)
i=1
3 45 6
∗
v0di−dec

∗ ∗
v0qi = v0qi−dr = 0 ∀i (15.3)

where nG represents total number of IIDG units, αP i and αQi , are the desired
real and reactive power sharing by the i th IIDG unit, respectively, KP deci and
KQdeci , are the real and reactive power control gains of lower-level decentralized
controller, respectively. Therefore, these are the tuning parameters of the lower-
level decentralized controller of the i th IIDG unit. Furthermore, it is evident that
#nG #
nG
∗
αP i = 1 and αQi = 1. In (15.1) and (15.2), reference frequency, ωi−dr , and
i=1 i=1
∗
reference voltage, v0di−dr , are set by the conventional droop controller according
to the active and reactive power droop characteristics [20, 33, 34]. The main
advantage of the lower-level decentralized controller is that the stability and basic
operation of the IIMG system are maintained when the centralized controller fails
to communicate with the local controllers.
In the next step, a small-signal linearized model has been developed to produce
frequency and voltage responses in the presence of load disturbances [33, 34]. Then,
an optimal set of input/output signals have been obtained using geometric measures
approach to reduce the costs associated with the installation of PMUs [21, 22]. It is
worth mentioning that the I/O signals are selected from and to all IIDG units in the
studied IACMG system due to its small size and for the purpose of demonstrating
the design procedure of the proposed hierarchical controller, thoroughly. Thereafter,
model-order reduction technique based on the Schur balanced model reduction
technique [14, 21, 35] has been used to obtain a reduced-order model and selected
input/output signals have been incorporated in the reduced-order model.
The most important communication issue coupling with PMUs is a signal trans-
mission time delay. Transmitting signals from PMUs to the upper-level centralized
386 E. S. N. Raju P and T. Jain

controller, through the PDC, and then back to both the lower-level decentralized
controller of each IIDG unit as well as local controller of RIAL involves some time
delays, as shown in Fig. 15.7. These signal transmission time delays mainly depend
on the type of communication link used and could vary from 25 ms (one-way) for
fiber optic based links to 250 ms (one-way) for satellite-based links [23]. Basically,
time delays involved in the processing and routing of signals at both PMUs and
the controller are small (especially in slow communication links) compared to
communication link delays and can be neglected without loss of generality [24].
In general, time delays may reduce the effectiveness of the controller, especially,
when relatively slow communication links such as satellite-based links are used.
Therefore, signal transmission time delays, experienced in practical WAMs, have
been incorporated.
Padé approximation method is widely used in WAMS-based stabilization appli-
cation for the modeling of signal transmission time delays [23–28]. Figure 15.9
shows the phase response of 1st and 2nd order Padé approximations, given in (15.4)
and (15.5), respectively, and compares them with the exact response of the time
delay (e−τ s ) [24]. It is evident that the 1st order Padé approximation provides
good accuracy in modeling short time delays (25–100 ms). However, its accuracy
is reduced for longer time delays greater than 100 ms. Whereas, 2nd order Padé
approximation is more accurate than the 1st order Padé approximation for modeling
longer time delays (250 ms) encountered in slow satellite-based communication
links. Therefore, in this paper, the 2nd order Padé approximation has been used
for the modeling of longer time delays.

Phase response
0
=25 ms

=100 ms

-50

=250 ms

-100

e- s

P2(s)
P1(s)
-150 0 1
10 10
Frequency, rad/s

Fig. 15.9 Phase response of 1st and 2nd order Padé approximations
15 WAM-Based Hierarchical Control of Islanded AC Microgrids 387

−τ s + 2
P1 (s) = (15.4)
τs + 2

τ 2 s 2 − 6τ s + 12
P2 (s) = (15.5)
τ 2 s 2 + 6τ s + 12

where P1 (s) and P2 (s) are 1st and 2nd order Padé approximation functions in
Laplace domain, respectively, and τ is signal transmission time delay for one-way
communication. It is worth mentioning that time delays are assumed to be identical
and fixed for all channels i.e., τ = τin = τout .
Next, noise and disturbance models are augmented to the reduced-order model
with selected input/output signals to consider the impact of measurement noises
and model mismatch, respectively. Then, the augmented model has been used to
design the upper-level centralized controller, which generates control commands to
the lower-level controllers of IIDG units and RIAL, as shown in Fig. 15.7. It was
observed that the robust control approaches, available in the existing literature [21–
31], such as conventional LQG, H2 and H∞ syntheses, and mixed H2 /H∞ synthesis
with pole placement constraints [35] are not suitable to stabilize IACMGs when
multiple types of static and dynamic loads are present. The reason behind this is
that the low-inertial nature of IIDG units and the presence of multiple types of static
and dynamic loads make IACMGs more vulnerable to instability [20]. Therefore,
the upper-level centralized controller has been designed based on a modal-based
extended LQG approach, shown in Fig. 15.10, to enable direct damping of the
targeted modes while keeping other modes unaffected. This feature makes it highly
suitable to PMUs supported two-level hierarchical controller.
In the modal-based extended LQG approach, first, the augmented model is
transformed into the modal canonical form using the real Schur decomposition [24].
The modal-augmented model has been used to design the upper-level centralized
controller. Next, the separation principle [21] has been used to solve the modal-
based extended LQG approach. This principle divides the central control problem
into two subproblems: an optimal estimate of the state and optimal output-feedback
control problems. The former one is based on the Kalman estimator theory [14,
21, 35] and the latter one is based on the linear quadratic regulator with prescribed

Fig. 15.10 Upper-level centralized controller based on modal-based extended LQG control
approach
388 E. S. N. Raju P and T. Jain

degree of stability (LQRPDS) with output weighting [13, 14, 35]. The gain matrix
of both the optimal output-feedback control and the optimal state estimator is
dependent on their diagonal weighting matrices, represented by QfMG & RfMG
and QcMG & RcMG , respectively. Therefore, these are the tuning parameters of the
upper-level centralized controller. Further, the prescribed degree-of-stability value,
α, can be used to force the proposed hierarchical controller to act faster and then
prevent any slowing down of the response caused by the additional supplementary
controller.
The optimal parameters of both lower-level decentralized as well as upper-
level centralized controllers have been obtained by formulating a bi-objective
optimization problem, as given in (15.6). It is desirable that the controller should
be able to provide a good transient response while sustaining small parameter
perturbations so that the stability of IACMGs is maintained under small-signal
disturbances. This can be achieved by minimizing the condition number of the
eigenvector matrix of the closed-loop IACMG as well as the second norm of the
optimal output-feedback control gain matrix, as given in (15.6).

minimize f1 (x) = VMG (x)2 VMG −1 (x)2

x

f2 (x) = KcMG (x)2 (15.6)

subject to xmin ≤ x ≤ xmax

where KcMG is the optimal output-feedback control gain matrix, VMG is the
eigenvector of the closed-loop IACMG, and x is a vector of the real and reactive
power control gains of decentralized controllers of nG number of IIDG units and
diagonal elements of QfMG & RfMG and QcMG & RcMG , as given in (15.7).

x = x1 x2 x3 (15.7)

x1 = KPDEC
(15.8)
= KP dec1 KP dec2 · · · KP decnG

x2 = KQDEC
(15.9)
= KQdec1 KQdec2 · · · KQdecnG

x3 = diag(QfMG )diag(RfMG ) diag(QcMG )diag(RcMG ) (15.10)

The two objective functions, given in (15.6), are conflicting in nature [14]. There-
fore, a fast and elitist multi-objective non-dominated sorting genetic algorithm
(NSGA-II) [36], which produce Pareto-optimal solutions, has been used to solve
the above optimization problem. Post-Pareto analysis [14] has been performed to
identify the most effective Pareto-optimal solution.
15 WAM-Based Hierarchical Control of Islanded AC Microgrids 389

Fig. 15.11 Block diagram of an IIDG unit with control signals from the centralized controller
considering time delay

Fig. 15.12 Block diagram of a RIAL with the control signals from the centralized controller
considering time delay

Finally, the control commands generated by the upper-level centralized controller

are given to the lower-level modified power-sharing controller of each IIDG unit and
AC current as well as DC voltage controllers of the RIAL, as shown in Figs. 15.11
and 15.12. The i th IIDG unit with consideration of the generated control command
including signal transmission delays for round-trip communication, e−2τ s ucci , is
shown in Fig. 15.11. The control command, ucci , modulates the dq-axis voltage
reference output of the modified power-sharing controller, v∗0dqi , to generate a
modified dq-axis voltage reference, v∗0dqcci , as given in (15.11).

v∗0dqcci = v∗0dqi + e−2τ s ucci ∀i (15.11)

where ucci = v dqi is a voltage correction (angle and magnitude of voltage) signal,
which is responsible to compensate voltage deviations caused by the small-signal
disturbances. Further, the modulated voltage signal, v∗0dqcci , is used as setpoint
values for the inner voltage controller. This modulates the dq-axis reference current
signal, i∗ldqi , of the current controller to generate a modified dq-axis reference
390 E. S. N. Raju P and T. Jain

input voltage signal, v∗idqi , to pulse width modulation (PWM) based voltage source
inverter (VSI), as shown in Fig. 15.11.
Figure 15.12 shows the AC current and DC voltage controllers of the i th RIAL
with consideration of the generated control commands including signal transmission
delays for round-trip communication, e−2τ s ilqALi∗ ∗
and e−2τ s vdcALi , as given
in (15.12) and (15.13), respectively.
∗ ∗
ilqccALi = ilqALi + e−2τ s ilqALi
∗
(15.12)

∗ ∗
vdcccALi = vdcALi + e−2τ s vdcALi
∗
(15.13)

∗
where ilqALi ∗
and vdcALi are reference current and voltage correction signals
generated by the upper-level centralized controller, respectively. Further, these
∗
correction signals modulate the q-axis reference current, ilqALi , of the AC current
∗
controller and DC voltage reference, vdcALi , of DC voltage controller, respectively,
to generate a modified dq-axis reference voltage input signal, v∗idqALi , as shown in
Fig. 15.12.

15.3 Results and Discussions

The robustness of the proposed hierarchical controller has been assessed for
different signal transmission time delays (τ ) such as 25 ms (for fiber optics cables)
and 250 ms (for satellite-based communication links) for one-way communication,
i.e., 50 ms and 500 ms for round-trip communication [23]. Furthermore, due to
the wide range of time delays encountered in real WAMS projects [23–31], the
effectiveness of the proposed hierarchical controller has also been verified for a/an
longer/increased signal transmission time delays such as 500 ms, 525 ms and 530 ms
(maximum admissible value) for one-way communication (i.e., 1000 ms, 1050 ms
and 1060 ms for round-trip communication). The maximum admissible value
(530 ms) was chosen as the time delay for which the stability of the closed-loop
IACMG (with the proposed hierarchical controller) system has been deteriorated.
The assessment has been carried out on the studied IACMG system, shown in
Fig. 15.7, using both eigenvalue analysis and time-domain simulations implemented
in MATLAB/SIMULINK environment. The studied IACMG system’s parameters
are given in Appendix [20]. The tuned parameters of the proposed hierarchical
controller, given in Table 15.1, have been obtained from the design procedure,
given in Sect. 15.2, considering the signal transmission time delay of 25 ms for
one-way communication. The rationale behind this consideration is that microgrids
are assumed to be established with the fiber optics communication links since
they cover a smaller area as compared to the conventional power system, which
is usually spread over a wide area and may be established with the satellite-
based communication links. However, it is possible to retune the parameters of the
15 WAM-Based Hierarchical Control of Islanded AC Microgrids 391

Table 15.1 Tuned parameters of the proposed hierarchical controller for the signal transmission
time delay (τ ) of 25 ms
Decentralized controller
Gain Parameters
matrix
KPDEC [607.81e−9 675.99e−9 531.23e−9 548.17e−9]
KQDEC [30.16e−3 34.24e−3 14.83e−3 33.22e−3]
Centralized controller
Matrix Parameters
Optimal state estimator
QfMG [40.46e−3 46.55e−3 49.21e−3 88.98e−3 55.91e−3 39.02e−3
36.83e−3 59.84e−3 63.31e−3 22.38e−3 81.31e−3 61.47e−3]
RfMG [29.07e−3 36.08e−3 68.05e−3 20.77e−3 74.78e−3 36.19e−3
81.88e−3 35.47e−3 50.85e−3 34.36e−3 46.83e−3 66.34e−3
66.43e−3 64.51e−3 39.27e−3 54.04e−3 74.81e−3 64.84e−3
51.97e−3 27.98e−3 89.15e−3 28.89e−3 34.53e−3 81.87e−3
13.59e−3 62.02e−3 78.76e−3 25.45e−3 68.14e−3]
Optimal state-feedback control
QcMG [875.00 226.22 667.02 150.62 687.34 821.94 714.56 472.97
130.42 817.17 338.88 343.30 293.42 342.75 522.58 578.79
688.78 951.79 190.18 788.18 942.00 611.32 151.89 290.82 342.82
744.16 219.41 318.93 631.93]
RcMG [59.58e−3 336.72e−3 594.33e−3 563.39e−3 586.98e−3 156.81e−3
996.66e−3 467.93e−3 630.52e−3 616.27e−3 611.45e−3 663.39e−3]

proposed hierarchical controller by considering the different amount of time delays

at the design stage and can evaluate its validity for increased time delays. It can
be observed from Table 15.1 that the tuned diagonal elements of QcMG are very
large as compared to the diagonal elements of RcMG . This indicates that the cost
of the proposed hierarchical controller is quite small, i.e., a cheap control strategy
[35]. Further, the effectiveness of the proposed hierarchical controller has been
verified by comparing the results obtained with those of a hierarchical controller
based on state-based extended LQG [14]. In order to have a fair comparison, both
the hierarchical controllers are designed with the same tuned parameters given in
Table 15.1. It is worth mentioning that the value of the prescribed degree of stability,
α, has been chosen as 10 to minimize both the excessive control energy cost or
controller complexity cost as well as to force the controller to act faster [35].

15.3.1 Eigenvalue Analysis

To test the impact of signal transmission time delays on the performance of the
proposed hierarchical controller, an eigenvalue analysis of the open-loop IACMG
(without controller) and closed-loop IACMG has been performed, as given in
 392

Table 15.2 Eigenvalue analysis based performance validation of the proposed hierarchical controller for different signal transmission time delays
Without controller With the state-based hierarchical controller With the proposed hierarchical controller
S.No Unstable modes Eigenvalue Damping factor Eigenvalue Damping factor Eigenvalue Damping factor
Delay-free communication channels (τ = 0)
1 iCP LQ &φAL 62.98 ± j 39.21 −0.8489 −46.71 ± j 23.10 0.8964 −60.01 ± j21.55 0.9411
Signal transmission time delay (τ ) = 25 ms (one-way communication)
2 iCP LQ &φAL 62.98 ± j 39.21 −0.8489 −36.00 ± j 20.79 0.8660 −41.88 ± j22.99 0.8765
Signal transmission time delay (τ ) = 250 ms (one-way communication)
3 iCP LQ &φAL 62.98 ± j39.21 −0.8489 −23.70 ± j24.35 0.6975 −32.82 ± j21.06 0.8422
Signal transmission time delay (τ ) = 500 ms (one-way communication)
4 iCP LQ &φAL 62.98 ± j39.21 −0.8489 4.45 ± j5.41 −0.6355 −7.06 ± j7.94 0.6646
Signal transmission time delay (τ ) = 525 ms (one-way communication)
5 iCP LQ &φAL 62.98 ± j39.21 −0.8489 8.87 ± j6.61 −0.8022 −3.19 ± j6.41 0.4451
Signal transmission time delay (τ ) = 530 ms (one-way communication)
6 iCP LQ &φAL 62.98 ± j39.21 −0.8489 17.46 ± j7.92 −0.9101 4.74 ± j4.87 -0.6975
E. S. N. Raju P and T. Jain
15 WAM-Based Hierarchical Control of Islanded AC Microgrids 393

Table 15.2. It should be noted that the seven (3 complex conjugate and one real)
modes lie in the unstable region (right-hand plane (RHP)) for the open-loop IACMG
system [33]. However, the most critical complex conjugate mode, affected by
the increased signal transmission time delays, has been considered for a better
comparison of its performance. It can be observed that an increase in time delay
degrades both the degree of stability as well as the damping ratios of the critical
complex conjugate mode.
Further, it can be seen that the proposed hierarchical controller shifts the critical
complex conjugate mode to the stable region for signal transmission time delays
less than 525 ms, while the critical mode shifts to the unstable region for signal
transmission time delays equal to 530 ms and above (the exact value is 527 ms
and above). Whereas, the state-based hierarchical controller shifts the critical mode
for signal transmission time delays less than 250 ms (it was found that the exact
value is 400 ms and below), while the critical mode shifts to the unstable region
for signal transmission time delays greater than 400 ms. Moreover, it can be
noticed that the degree of stability and damping ratios of unstable modes stabilized
by the proposed hierarchical controller are more as compared to the state-based
hierarchical controller. In addition to this, although the damping ratios are degraded
by the increased time delays, the proposed hierarchical controller is still effectively
mitigating the instability for longer signal transmission time delays up to 526
ms. Thus it can be concluded that the maximum signal transmission time delay
that can be tolerated by the proposed hierarchical controller is 526 ms (one-way
communication), whereas, the state-based hierarchical controller can tolerate up to
400 ms (one-way communication).

15.3.2 Time-Domain Simulations

Time-domain simulation results have also been performed to assess the impact
of different signal transmission time delays (25 ms, 250 ms, 500 ms, 525 ms,
and 530 ms for one-way communication) on the robustness of the proposed
hierarchical controller in the presence of small-signal disturbances. These small-
signal disturbances are associated with the small load disturbances at the individual
bus as well as multiple buses at a time. Considering the worst case of adding
a step signal to the input load disturbance at all the buses, simultaneously, as
shown in Fig. 15.7, time-domain simulations of the open-loop IACMG and closed-
loop IACMG are shown in Fig. 15.13 and Figs. 15.14, 15.15, 15.16, 15.17, 15.18,
respectively. It should be noted that, in order to show the detailed oscillations, the
figures show the results only for periods of interest. However, the settling time is
given in Table 15.3 to identify the time required for the response to reach and stay
within a range of certain percentage (usually 5% or 2%) of the final value.
Figure 15.13 shows the dynamic step response of the deviation in output
variables, such as frequency and voltage at all buses, of the open-loop IACMG.
It can be observed that the step response of the deviation in output variables is
394 E. S. N. Raju P and T. Jain

Fig. 15.13 Dynamic step response of the deviation in the outputs of open-loop IACMG. (a)
Frequency. (b) Voltage at buses

Fig. 15.14 Dynamic step response of the deviation in frequency. (a) With the state-based
hierarchical controller. (b) With the proposed hierarchical controller

increasing continuously with respect to time, reflecting instability of the open-

loop IACMG. It is worth reiterating that the detailed information about causes
behind the instability of the open-loop IACMG can be found in [20]. Figures 15.14,
15.15, 15.16, 15.17, and 15.18 show the dynamic step response of the deviation
in output variables of the closed-loop IACMG, using both the state-based as well
as the proposed hierarchical controllers. It can be seen from Figs. 15.14a, 15.15a,
15.16a, 15.17a, and 15.18a that the state-based hierarchical controller has produced
the stable response for signal transmission time delays varying from 25–250 ms,
while the response is unstable for signal transmission time delays equal to 500
15 WAM-Based Hierarchical Control of Islanded AC Microgrids 395

Fig. 15.15 Dynamic step response of the deviation in voltage at bus 1. (a) With the state-based
hierarchical controller. (b) With the proposed hierarchical controller

Fig. 15.16 Dynamic step response of the deviation in voltage at bus 2. (a) With the state-based
hierarchical controller. (b) With the proposed hierarchical controller

ms and above (it was found that the exact value is 400 ms and above). Whereas,
the proposed hierarchical controller has produced the stable response for signal
transmission time delays varying from 25–525 ms, while the response is unstable for
signal transmission time delays equal to 530 ms and above, as shown in Figs. 15.14b,
15.15b, 15.16b, 15.17b, and 15.18b.
Furthermore, the time-domain simulation specification, i.e., the settling time of
the dynamic step response of the deviation in output variables, shown in Figs. 15.14,
396 E. S. N. Raju P and T. Jain

Fig. 15.17 Dynamic step response of the deviation in voltage at bus 3. (a) With the state-based
hierarchical controller. (b) With the proposed hierarchical controller

Fig. 15.18 Dynamic step response of the deviation in voltage at bus 4. (a) With the state-based
hierarchical controller. (b) With the proposed hierarchical controller

15.15, 15.16, 15.17, and 15.18, is given in Table 15.3 for better performance
comparison of the two hierarchical controllers.It can be seen that the settling
time increases for an increase in signal transmission time delay, especially, for
the longer time delays, i.e., 400 ms for the state-based hierarchical controller,
while in case of the proposed hierarchical controller, it increases gradually for
the longer time delays of 500 ms and 526 ms. It can also be observed that
the proposed hierarchical controller has produced a fast settling time than the
Table 15.3 Performance comparison of the proposed hierarchical controller with the state-based hierarchical controller for different signal transmission time
delays
With the state-based two-level hierarchical With the proposed two-level hierarchical controller:
Deviation in controller: settling time (ms) settling time (ms)
S.no. output Without delay 25 ms delay 250 ms delay 400 ms delay Without delay 25 ms delay 250 ms delay 500 ms delay 526 ms delay
1 f (H z) 259 302.1 375.6 2360 250 287.1 361.1 1253 1830
2 V1 (V ) 168.1 404.0 472.4 3600 129 399.3 466.6 1150 2100
3 V2 (V ) 295 400.0 459.0 3400 152 392.5 452.3 1088 1990
4 V3 (V ) 229 359.1 428.5 3250 189 352.3 426.1 1042 1970
5 V4 (V ) 319 344.7 460.4 3476 314 338.5 453.9 1056 2020
15 WAM-Based Hierarchical Control of Islanded AC Microgrids
397
398 E. S. N. Raju P and T. Jain

state-based hierarchical controller. The rationale behind this is that the proposed
hierarchical controller is based on modal-based extended LQG approach, which
enables the addition of damping only to modes of interest while keeping other
modes unaffected [24].
Thus, it can be concluded from Figs. 15.14, 15.15, 15.16, 15.17, 15.18 and
Tables 15.2, 15.3 that both the proposed as well as state-based hierarchical
controllers have an adequate capability to mitigate the undamped oscillations for
time delays varying from 25–250 ms due to their optimal Kalman estimator that
takes care of the delay in signal transmission. Furthermore, the increase in time
delay degrades the damping ratios of the critical modes. In addition, time delay
has a significant influence on the boundary of the small-signal stability region
of microgrids, especially, when the time lag is large. Moreover, the proposed
hierarchical controller is effective in enhancing the stability even with longer signal
transmission time delays (i.e., 526 ms (one-way communication)) encountered in
real WAMS.

15.4 Conclusion

In this chapter, a wide area measurement system (WAMS) based hierarchical

control of islanded AC microgrids (IACMGs) with static and dynamic loads has
been presented. The proposed controller has a hierarchical (two-level) structure
comprising of an upper-level centralized controller built on the top of lower-
level decentralized/local controllers and incorporates signal transmission time
delays encountered in the two-way communication. Finally, the simulation results
concluded that the proposed hierarchical controller is effective in enhancing the
stability of IACMGs for a wide range of time delays encountered in real WAMS.

Appendix

IIDG Units Ratings: IIDG1 -(10 + j6) kVA; IIDG2 -(15 + j9) kVA; IIDG3 -(20 + j12)
kVA; IIDG4 -(25 + j15) kVA. Static Active and Reactive Power Droop Gains: mP 1 =
6.28e−4 rad/s/W, mP 2 = 4.18e−4 rad/s/W, mP 3 = 3.14e−4 rad/s/W, mP 4 =
2.52e−4, nQ1 = 1.66e−3 V/VAR, nQ2 = 1.11e−3 V/VAR, nQ3 = 8.33e−4
V/VAR and nQ4 = 6.66e−4 V/VAR. IIDG unit Parameters: Lf = 1.35 mH, Cf = 50
μF, Rf = 0.1 , fsw = 8 kHz, wc = 31.41 rad/s, Kpv = 0.05, Kiv = 390, Kpi = 10.5,
Kii = 16e−3, F = 0.75, fnl = 50.5 Hz, Rc = 0.03 , Lc = 0.35 mH. RIAL Parameters:
Lf = 2.3mH, Cf = 8.8 μF, Rf = 0.1 , fsw = 10 kHz, wc = 31.41 rad/s, Kpv = 0.5,
Kiv = 150, Kpi = 7, Kii = 25e3, Rc = 0.03 , Lc = 0.93 mH. Line Parameters: Line 1:
(0.23 + j0.11) , Line 2: (0.35 + j0.58) , Line 3: (0.30 + j0.47) . Load Parameters:
Induction Motor Load: 10 hp, 400 V, 50 Hz, rs = 0.7834, Lss = 127.1 mH,
rr = 0.7402, Lrr = 127.1 mH, Lm = 124.1 mH, P=4, TL = 47.75 N m; CP L:
15 WAM-Based Hierarchical Control of Islanded AC Microgrids 399

12 kVA, rCP L =13.224  /phase and cosα=0.85; RI AL: 12 kW and RRI AL =40.833
; R Load: 25 kW, RRLoad =6.347 /phase and VDC = 700 V.

References

1. Parhizi, S., Lotfi; H., Khodaei, A., & Bahramirad. S., (2015). State of the art in research on
microgrids: A review. IEEE Access, 3, 890–925.
2. Chowdhury, S., Chowdhury, S. P., & Crossley, P. (2009). Microgrids and active distribution
networks. London: Institution of Engineering and Technology.
3. Kundur, P. (1994). Power system stability and control. New York: McGrawHill.
4. Kundur, P., et al. (2004). Definition and classification of power system stability IEEE/CIGRE
joint task force on stability terms and definitions. IEEE Transactions on Power Systems, 19(3),
1387–1401.
5. Majumder, R. (2010). Some aspects of stability in microgrids. IEEE Transactions on Power
Systems, 28(3), 3242–3253.
6. Xinhe, C., Wei, P., & Xisheng, T. (2010). Transient stability analyses of micro-grids with
multiple distributed generations. In Proceedings of the International Conference Power System
Technology (POWERCON).
7. Narendra, K. S., & Balakrishnan, J. (1994). A common Lyapunov function for stable LTI
systems with commuting A-matrices. IEEE Transactions on Industrial Electronics, 39(12),
2469–2471.
8. Majumder, R. (2013). Reactive power compensation in single-phase operation of microgrid.
IEEE Transactions on Industrial Electronics, 60(4), 1403–1416.
9. Olivares, D. E., et al. (2014). Trends in microgrid control. IEEE Transactions on Smart Grid,
5(4), 905–1919.
10. Liang, H., Choi, B. J., Zhuang, W., & Shen, X. (2013). Stability enhancement of decentralized
inverter control through wireless communications in microgrids. IEEE Transactions on Smart
Grid, 4(1), 21–331.
11. Kahrobaeian, A., & Mohamed, Y. A.-R. I. (2014). Analysis and mitigation of low-frequency
instabilities in autonomous medium-voltage converter-based microgrids with dynamic loads.
IEEE Transactions on Industrial Electronics, 61(4), 1643–1658.
12. Ovalle, A., Ramos, G., Bacha, S., Hably, A., & Rumeau, A. (2015). Decentralized control of
voltage source converters in microgrids based on the application of instantaneous power theory.
IEEE Transactions on Industrial Electronics, 62(2), 1152–1162.
13. Raju, P. E. S. N., & Jain, T. (2017). Optimal decentralized supplementary inverter control loop
to mitigate instability in an islanded microgrid with active and passive loads. International
Journal of Emerging Electric Power Systems, 18(01).
14. Raju, P. E. S. N., & Jain, T. (2017). Robust optimal centralized controller to mitigate the
small signal instability in an islanded inverter based microgrid with active and passive loads.
International Journal of Electrical Power & Energy Systems, 90, 225–236.
15. Guo, F., Wen, C., Mao, J., & Song, Y. D. (2015). Distributed secondary voltage and frequency
restoration control of droop-controlled inverter-based microgrids. IEEE Transactions on
Industrial Electronics, 62(7), 4355–4364.
16. Wang, Y., Wang, X., Chen, Z., & Blaabjerg, F. (2017). Distributed optimal control of reactive
power and voltage in islanded microgrids. IEEE Transactions on Industrial Electronics, 53(01),
340–349.
17. Zhao, Z., Yang, P., Guerrero, J. M., Xu, Z., & Green, T. C. (2016). Multiple-time-scales
hierarchical frequency stability control strategy of medium-voltage isolated microgrid. IEEE
Transactions on Power Electronics, 31(8), 5974–5991.
400 E. S. N. Raju P and T. Jain

18. Baghaee, H. R., Mirsalim, M., & Gharehpetian, G. B. (2016). Power calculation using RBF
neural networks to improve power sharing of hierarchical control scheme in multi-DER
microgrids. IEEE Journal of Emerging and Selected Topics in Power Electronics, 04(4), 1217–
1225.
19. Li, Z., Zang, C., Zeng, P., Yu, H., & Li, S. (2018). Fully distributed hierarchical control of
parallel grid-supporting inverters in islanded AC microgrids. IEEE Transactions on Industrial
Informatics, 14(2), 679–690.
20. Raju, P. E. S. N., & Jain, T. (2018). Impact of load dynamics and load sharing among distributed
generations on stability and dynamic performance of islanded AC microgrids. Electric Power
Systems Research, 157, 200–210.
21. Zolotas, A. C., Chaudhuri, B., Jaimoukha, I. M., & Korba, P. (2007). A study on LQG/LTR
control for damping inter-area oscillations in power systems. IEEE Transactions on Control
Systems Technology, 15(1), 151–160.
22. Surinkaew, T., & Ngamroo, I. (2016). Hierarchical co-ordinated wide area and local controls
of DFIG wind turbine and PSS for robust power oscillation damping. IEEE Transactions on
Sustainable Energy, 7(3), 943–955.
23. Stahlhut, J. W., Browne, T. J., Heydt, G. T., & Vittal, V. (2008). Latency viewed as a stochastic
process and its impact on wide area power system control signals. IEEE Transactions on Power
Systems, 23(1), 84–91
24. Almutairi, A. (2010). Enhancement of Power System Stability Using Wide Area Measurement
System Based Damping Controller, Ph.D. dissertation, The University of Manchester, Manch-
ester.
25. Cheng, L., Chen, G., Zhang, F., Li, G., & Gao, W. (2014). Adaptive time delay compensator
(ATDC) design for wide-area power system stabilizer. IEEE Transactions on Smart Grid, 5(6),
2957–2966.
26. Milano, F. (2016). Small-signal stability analysis of large power systems with inclusion of
multiple delays. IEEE Transactions on Power Systems, 31(4), 3257–3266.
27. Beiraghi, M., & Ranjbar, A. M. (2016). Adaptive delay compensator for the robust wide-area
damping controller design. IEEE Transactions on Power Systems, 31(6), 4966–4976.
28. Ghosh, S., Folly, K. A., & Patel, A. (2018). Synchronized versus non-synchronized feedback
for speed-based wide-area PSS: Effect of time-delay. IEEE Transactions on Smart Grid, 9(5),
3976–3985.
29. Aminifar, F., Fotuhi-Firuzabad, M., Safdarian, A., Davoudi, A., & Shahidehpour, M. (2015).
Synchrophasor measurement technology in power systems: Panorama and state-of-the-art.
IEEE Access, 2, 1607–1628.
30. Li, J., Chen, Z., Cai, D., Zhen, W., & Huang, Q. (2016). Delay-dependent stability control for
power system with multiple time-delays. IEEE Transactions on Power Systems, 31(3), 2316–
2326.
31. Li, M., & Chen, Y. (2018). A wide-area dynamic damping controller based on robust H∞
control for wide-area power systems with random delay and packet dropout. IEEE Transactions
on Power Systems, 33(4), 4026–4037.
32. National Academies of Sciences, Engineering, and Medicine and Others (2016). Analytic
Research Foundations for the Next-Generation Electric Grid. Washington, DC: National
Academies Press.
33. Raju, P. E. S. N., & Jain, T. (2018). A two-level hierarchical controller to enhance stability and
dynamic performance of islanded inverter-based microgrids with static and dynamic loads.
IEEE Transactions on Industrial Informatics, 15(5), 2786–2797.
34. Raju, P. E. S. N., & Jain, T. (2019). Development and validation of a generalized modeling
approach for islanded inverter-based microgrids with static and dynamic loads. International
Journal of Electrical Power & Energy Systems, 108, 177–190.
35. Xue, D., Chen, Y. Q., & Atherton, D. P., et al. (2007). Linear Feedback Control: Analysis and
Design with MATLAB. New York: SIAM.
36. Deb, K., Pratap, A., Agarwal, S., & Meyarivan, T. (2002). A fast and elitist multiobjective
genetic algorithm: NSGA-II. IEEE Transactions on Evolutionary Computation, 06(2), 182–
197.
 Part III
Protection of Microgrids
Chapter 16
Fault Ride Through and Fault Current
Management for Microgrids

Wei Kou and Sung-Yeul Park

16.1 Introduction

Considerable distributed energy resources (DERs) integrated into the transmission

system poses many questions about the fact that the system cannot be assisted in
disturbances. A DER’s response to a voltage disturbance can affect the stability of
the system as defined in [1–4]. The development of stringent grid code requirement
to maintain security of supply is dictated by the increased DER integration into
bulk power systems. Meanwhile, microgrids (MGs) as an essential element of the
modern power system have gradually used a mixture of different DERs. Potential
applications for MGs can provide service to a single customer, for example, a
residential (also called the nanogrid) MG or a campus MG; or to a group of
customer, for example, secondary MGs, partial and full feeder MGs, and substation
MGs. The industry has meanwhile become more interested in the application of MG
to take advantage of the proliferation of DER to address planning objectives, for
example, by improving resilience and efficiency. Fault ride through (FRT), which
mainly reflects the ability of DERs to be connected during any fault occurrence,
is one of the main requirements. Smart inverters play a critical role in enabling
the FRT implementation in MGs by providing advanced control capabilities (for
example, voltage and frequency control).
Unlike the latest FRT technologies used in DERs such as large-scale solar power
plants and wind power stations [5], the FRT strategy on MGs has a higher power
control criterion. One is to remove power ripples, especially the double-frequency
power ripples, created by the unbalanced faults. A high proportion of the double-
frequency power ripples in the power exchange between MGs and the host grid

W. Kou () · S.-Y. Park

Department of Electrical and Computer Engineering, University of Connecticut, Storrs, CT, USA
e-mail: wei.kou@uconn.edu

© Springer Nature Switzerland AG 2021 403

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_16
404 W. Kou and S.-Y. Park

will contaminate the inverter-based device safety, the host grid power quality as
well as the MGs local loading system reliability in the event of system failures.
Furthermore, MG is run as a group with a range of loads and DGs, and it manages
the output power fairly passively. It is unlike solar and wind plants, which could
decrease output power rapidly by using the maximum power point tracking (MPPT)
control and the wind-blade diverter, respectively. The IB-MG FRT difficulty in
managing the power flow between the MG and the host network is responsible for
implementing the dynamic energy management on the inverter-based interfaces.
The universal FRT power control scheme is based on the dq frame of the
symmetrical components of the three-phase injection currents to regulate both
active and reactive power. It comprises both positive and negative sequence loops,
and each sequence loop has the d- and q-axis frame current control loops [6–
9]. The symmetrical components’ decomposition and the synchronous reference
frame transition must be carried out on voltages and currents, and at least four
PI controllers have to be controlled in this system. Therefore, the amplitudes
and phase angles of the three-phase currents cannot be openly modified with the
reference current calculation algorithm based on the double synchronous reference
frame. Recently, Kou et al. [10–13] developed a novel FRT strategy for directly
regulating active output power in the abc reference frame without any electrical
variable transformations. The decomposition and transformations of the three-phase
electrical measurements, which are conducted in the double synchronous reference
frame, are not needed in this new method. Its effectiveness on power flow balance
and power ripple elimination is comparable with the double synchronous reference
frame vector control.
With the FRT technique being adapted to DER integration at a fast pace,
the associated fault current increase due to the DER connections during faults
becomes an issue. Especially in places where the old aging grid with fast-growing
concentrated loads, the substation circuit breakers in the distribution network will
be at the risk of exceeding its short-circuit duty limitation because of the fault
current injections from DERs. Even if the increase in fault current does not exceed
the protection limits of the installed devices, coordination of the primary and
secondary protective devices may be disturbed due to excessive DER fault current.
Concerning the effects of DERs on fault current increase, the capacity of DER
integration through FRT will be limited, and the reliability of the whole system
during fault will be decreased. Overcoming these problems clearly requires fault
current management (FCM) to eliminate the effects of the fault current injection
from MGs on the distributed networks.
FCM is controlling the fault current injection from DERs to eliminate the fault
current increase in the main grid, which is applied at the system level. FRT is
controlling the three-phase current injections from DERs to maintain the power
exchange between DERs and main grid during the fault, which is applied at the
device level. To overcome the fault current control conflict between FCM and FRT,
FCM should be designed with the compatibility with FRT on each MG. Fault current
limiter (FCL) is a mature FCM device, which is deployed at the point of common
coupling (PCC) and used to limit the fault current inputs of MGs through a rapid
16 FRT and FCM for MGs 405

increase in impedance [14, 15]. FCL installation provides additional investments in

the integration of MGs with energy losses, triggering and retrieval times, steady-
state impedance as well as the installation costs of the FCLs, and the system
operators have to consider them all. The inverter-based interfaces of MGs that
could achieve the same effect as FCLs are recently taken into account by flexibly
modifying the fault currents. The magnitudes of the three-phase output currents
could be limited to a reasonable value by using a hard limit on current referrals in
a dq frame. But in unbalanced conditions of fault, the three-phase fault currents are
connected and cannot be separately controlled. It is not a feasible solution. An FCM
technique to preserve the present fault current level by adjusting both amplitude
and phase angles from the inversion interface output current is proposed by Rajaei
et al. [16, 17]. When the output current changes in the same step, the use of this
FCM technique on the IB-MG directly results in a sudden power imbalance in the
PCC and also affects the IB-MG’s inner device operation. The FCM analysis in this
chapter was motivated by the need to encourage IB-MG FRT to provide FCM in
different fault situations, especially unbalanced faults.

16.2 MG FRT in Grid Codes

The large number and different types of power generation units and their connection
complexity makes the transmission system operators form an operative collection
of regulations to manage their integration and service in the grid, which is called
grid code [18]. The incorporation of DERs within the bulk power network has
gained substantial interest from grid operators when building productive distributed
networks by environmental, economic, and technological opportunities [19–21].
Some basic specifications, including IEEE 1547, IEEE 2030 [22], VDE-AR-N 4105
[23], C22.2 NO.257, C22.3 NO.9, and NB/T 32015-2013 [24], have been developed
to govern the operating process of the integrated DERs.
For international grid code, IEEE 1574 is the earliest standard for distributed
resource integration, which was published by IEEE in 2003. It can be applied to
all distributed resources and has been extended to a series of standards, which
cover testing, monitoring, information exchanges and control, etc. IEC/IEEE/PAS
63547-2011, the standard of distributed resources interconnecting with electric
power systems, was initially converted from IEEE 1547 but not published under
the standards IEC. It was used at last as PAS (Publicly Available Specification),
and its specification is close to that of IEEE 1547. In Germany, the guidelines
for power generation systems integrated to the medium-voltage and the low-
voltage distribution networks were released in January 2008 and August 2011,
respectively. All specifications are essential for the incorporation of DERs, which
refer to all power generation systems connected by synchronously motor and
asynchronously engaged or transforming devices, such as wind, hydropower, and
photovoltaic electricity, to the medium- or low-voltage distribution networks. Two
major incorporation standards exist in Canada, for example. The DER services
406 W. Kou and S.-Y. Park

incorporated into networks below 50 kV and within total integration capability

of 10 MW are specified by C22.3 NO.9; the requirements for connecting DER
based on inverter to low-voltage distribution networks under 0.6 kV are specified
by C22.3 NO.9. The Chinese technical standard NB/T32015-2013 was released
in 2013. The standard applies to new infrastructure, restoration, and extension of
distributed resources linked to power systems no more than 35 kV.
While DER penetration has increased in recent years, some specifications in
these standards have passive effects on the dynamic operation of the power system.
IEEE 1547, the first DER integration standard published in 2003 by IEEE, for
instance, has “must trip” requirements of the grid-connected DERs against system
voltage deviations, whereas the host grid disturbance is worsened and the safety of
power supply is at risk. In 2014, IEEE 1547 was revised to make the “must ride”
specifications more flexible. Amendment 1 to IEEE 1547 requires DER to “pass
around” grid tension anomalies and “grid operators and DER can agree on other
tensile voltage and time settings” [25].
Inverter-based DERs have been developed and used to help maintain the relia-
bility of the bulk power system during faults by FRT technology that allows DER
exchange power to the host system under the voltage deviation case [26, 27]. As FRT
technology matures, the ability to integrate FRT in DER standards such as VDE-
AR-N 4105 and NB/T 32015 begins to be required. It shows that DER is expected
to support the bulk network in the event of failures, and FRT technology should play
a critical role in this dynamic process. DERs are typically grouped into MGs with
their local loads and connected to the distribution networks in urban areas. Such
DERs also face technical challenges in their adaptability to their FRT capabilities.
These problems could be categorized as flexible power flow control, meeting the
varied failure scenarios, distributed network reliability criteria, and communication
support for timely measurements.
DER-related ride-through specifications are laid out in draft IEEE P1547,
and they explicitly focus on voltage and frequency ride-through, as well as
improvements in the voltage cycle angle and frequency. Abnormal performance
requirements for ride-through disturbance are classified into categories I, II, and III.
Category I is intended to meet minimum BPS reliability needs and to be achievable
by all DER technologies, including rotating machines. Category II is designed to
align with the requirement in NERC PRC-024-2 on the ride-through specifications
for the whole converter-coupled device and other performance criteria to enable the
delayed post-fault stress retrieval at the distribution level (adding some ride-through
curve margins). In high DER penetration or low-inertia, distribution-related grids
such as California and Hawaii, Category III is developed to align with California
Rule 21(19), Hawaii Rule 14(20), and other similar rules. It focuses on the additional
requirements for high DER penetration systems such as California and Hawaii,
where both distribution network and BPS reliability depend considerably on the
performance of DER.
FRT requirements of BPS generating resources are also defined in the NERC
quality standard PRC-024-2, which describes a “no trip” zone for generator
frequency and voltages relay used to trip the applicable generating units. Footnote
16 FRT and FCM for MGs 407

1 of the standard clearly describes that these protective functions are part of “multi-
function protective devices or protective functions” within the control systems
directly operating or transmitting trip signals to the generator based on the inputs
of frequencies and voltages. It is essential to realize that the control actions that
perform the protective functions are part of the requirements in PRC-024-2 for BES-
connected converter-coupled resources.

16.3 FRT Control in Microgrids

MG is an organization of DERs capable of islanding and grid-connected operation

[28–30]. AC and DC are two main categories of the microgrid networks. The DER
capabilities of AC MG are combined with the AC grid. It is cheaper to construct and
can be introduced without modifying the current network. If the sources and loads
are both AC natured, AC MGs are more effective. DC MGs shown as in Fig. 16.2
are mostly combined with distributed energy resource and storage systems. DC MGs
have the benefits of offering higher reliability, efficiency, and simplified monitoring
topology over AC MGs. An IB-MG system consists of the utility, the microgrid,
and the inverter-based interface for implementing FRT control. The inverter-based
interface is developed to balance the total power flow between the utility and MG in
the regular operation and FRT operation, and it is disabled when MG is running in
the islanded operation. In this session, FRT controls for IB-MG in both AC and DC
microgrids are introduced.

16.3.1 AC Microgrids

To achieve a versatile power flow control between the utility and the MG under the
usual and even abnormal load, the grid-connected interface of IB-MG adopts a back-
to-back (B2B) configuration comprised of two traditional pulse-width-modulated
(PWM) voltage source converters (VSCs) with their DC sides linked via a DC
link capacitor [5, 12]. Figure 16.1 displays the B2B framework and its control
implementation. The utility side converter, VSC-1, is the master converter of IB-
MG FRT, and the MG side converter, VSC-2, is the slave converter operated by the
MG frequency regulation. When a failure exists on the utility side, the fault current
injected by the utilities Ig , together with the DC-link voltage Vdc , the three-phase
voltages uM , and the three-phase currents at VSC-1 side iM , is detected and sent to
VSC-1 controller for calculating the FRT reference current iMr ef .
The three-phase voltages at MG local load side, uL , is fed in phase-locked loop
(PLL) for collecting the control variable inputs, the AC local bus frequency fL , and
the voltage angle θL of the dq reference frame current controller of VSC-2. The PI
control on frequency variance gives the d-axis part of the corresponding reference
currents iL_ref . For generating PWM signals, SM _abc and SL_abc , for both VSC-1
408 W. Kou and S.-Y. Park

Fig. 16.1 Schematic diagram of AC microgrid FRT control

and VSC-2, a deadbeat control algorithm is used [31]. See Appendix for details on
the adopted controls of FRT reactive power injection Q∗ in section “FRT Reactive
Power Injection,” PLL in section “Synchronous Reference Frame PLL,” and the
deadbeat control algorithm in section “Deadbeat Current Control.” Before faults
appear in utility, the B2B interface works under the regular operation mode and
keeps the power flow balance between the utility side and the MG side. Under the
normal operation, VSC-1 is controlled by the DC-link voltage control in the dq
reference frame, which adopts the structure of the VSC-2 frequency PI regulator
[10, 32].

16.3.2 DC Microgrids

Although DC MGs have no power system frequency, the control systems of DC

MGs can be studied in a framework similar to AC MGs . The control is divided
into two timescales: the primary controls on the order of milliseconds, and the
secondary control on the order of a few seconds, or even a minute. The “secondary”
controllers typically use relatively slow communication to control power generators,
storage, loads, and voltages to minimize losses. The secondary control designs in
the AC or DC architectures are not substantially different. The primary control,
however, reacts to quick MG transients or other upsets. In DC architecture, voltage
replaces frequency as a physical variable, which can be used to communicate the
imbalance of generation and burden on the “primary” control timescale [29]. The
required speed of the primary response and critical nature of the control makes
a communication-based control less reliable. In comparison to AC architecture,
voltage is a “local” variable in the DC architectures, i.e., the voltage droop depends
on the position of the charge changes relative to the generation location and the
measuring point. As for the inverter interfaces connected with DERs in DC MG,
in the regular operation conditions, they usually use MPPT to control the output
16 FRT and FCM for MGs 409

DC BUS
DC/AC
Utility
DC/DC

IB-MG FRT

MPPT
DC/DC
I-V Droop
ENABLE Energy
Storage
AC/DC

Adaptive Droop

DC/AC AC Bus
MPPT
Load
I-V Droop
ENABLE

frequecy PI Motor
control

Fig. 16.2 Schematic diagram of DC microgrid FRT control

power from solar or wind power. At the moment of the grid voltage fault, MPPT
function is disabled and is replaced with I–V droop control, and the active current
injection should be regulated according to the fault severity level, as shown in
Fig. 16.2. The generation or storage source response across DC MG is uniform when
the voltage is used as an indication of generation/load imbalance. Although it may
not be optimal to configure the response in the primary control timescale, it can
be modified with a secondary control at a slower time frame. In the faster primary
control, however, the reaction comes exclusively from power electronic devices with
more or less equivalent dynamics to prevent mixed MG control complications in
AC architectures. In Fig. 16.2, the energy storage adopts adaptive droop control by
adjusting droop constant rather than merely selecting a constant droop at the cost of
achieving voltage regulation and load sharing simultaneously.
410 W. Kou and S.-Y. Park

16.3.3 Traditional Reference Current Control Methods

The traditional methods of FRT current control involve both positive and negative
sequence loops and the grid voltage synchronous reference system, which is used
to regulate the active and reactive power through dq transformation [31]. The
control strategy is complicated because both positive and negative frames are
necessary to feed the two control loops, which are converted independently from
the three-phase voltages and currents [33]. Others [32, 34] obtained the active
and reactive power expressions under the framework of d-and q-axis currents and
voltages. The current references can then be obtained based on the required power
output. Nevertheless, there are common issues with these strategies. First of all,
they take dual current control schemes for positive and negative sequences, which
increases the difficulty of controls. Second, they require accurate online symmetric
component decomposition in time domain, which compromises the solidity and
reliability of the FRT. In what follows, two reference current control methods in
dq frame are introduced.

16.3.3.1 Balanced Positive Sequence Control Method

Adopting balanced positive sequence control (BPSC) method, the goal is to control
grid-connected current injection with the positive sequence components only. Thus
the corresponding active power and reactive power reference values Pref and Qref
can be simplified as
*
Pref = 3V+ ip∗
(16.1)
Qref = 3V+ iq∗

This method is useful for maintaining the quality of the currents injected [35].
Besides, it has less technique demanding on PLL, by which the balanced currents
can be generated using simple synchronous controllers, provided that the synchro-
nization system accurately estimates the amplitudes and the phase angles of the
positive sequence components of the grid voltages. The BPSC strategy calculates
the active and reactive reference currents as follows:
Pref
ip∗ = · υ+ , (16.2)
|ν+ |2

Qref
iq∗ = · υ+ , (16.3)
|ν+ |2
16 FRT and FCM for MGs 411

where
⎡ ⎤ ⎡ √ ⎤
va+ + φ+ )
√ 2V+ sin(ωt 2π
v+ = ⎣vb+ ⎦ = ⎣ 2V+ sin(ωt − 3 + φ+ )⎦ , (16.4)
√
vc+ 2V+ sin(ωt + 2π
3 + φ+ )

3
|v+ |2 = va+
2
+ vb+
2
+ vc+
2
= |V+ |2 . (16.5)
2

16.3.3.2 Decoupled Double Synchronous Reference Control Method

To minimize double-frequency ripples on DC-link voltage and the three-phase

output voltages caused by the unbalanced faults, the negative sequence components
of output currents are involved in the decoupled double synchronous reference
control method. Using instantaneous power theory and Park’s transformation [36],
one can obtain the instantaneous active and reactive power outputs P and Q as

P = P0 + Pcos cos(2ωt) + Psin (2ωt), (16.6)

Q = Q0 + Qcos cos(2ωt) + Qsin (2ωt). (16.7)

Here, P0 and Q0 are the constant components of P and Q, Pcos and Psin are the
amplitudes of real power ripple injection, Qcos and Qsin are the amplitudes of
reactive power ripple injection, which can be determined by
⎡ ⎤ ⎡ + ⎤
P0 ed eq+ ed− eq− ⎡ ⎤
⎢P ⎥ ⎢ e− eq− ed+ eq+ ⎥ +
⎢ cos ⎥ ⎢ d ⎥ id
⎢ ⎥ ⎢ − −eq− −eq+ + ⎥ ⎢ +⎥
⎢ Psin ⎥ 3 ⎢ ed ed ⎥ ⎢iq ⎥
⎢ ⎥= ⎢ + ⎥⎢ ⎥, (16.8)
⎢ Q0 ⎥ 2 ⎢ eq −ed+ eq− −ed− ⎥ ⎣id− ⎦
⎢ ⎥ ⎢ − ⎥
⎣Qcos ⎦ ⎣ eq −ed− eq+ −ed+ ⎦ iq−
Qsin −ed− −eq− ed+ eq+

where ed+ , eq+ and ed− , eq− are the constant signals being transformed from positive
and negative sequence components of three-phase grid-tied voltage by the Park’s
transformation. id+ , iq+ and id− , iq− are transformed from three-phase grid-tied cur-
rent. Equation (16.8) shows that unbalanced sags induce double-frequency ripples
of power output. For instance, ed− and eq− would not be zeros under unbalanced
sags, which gives a nonzero Pcos . If eliminating the active power fluctuations is the
control priority, P0 , Pcos , Psin , and Q0 should be selected as the control priority,
and the appropriate dq rotation coordinates should be set as eq+ = 0 and eq− = 0.
Therefore, the expression of the dq-axis reference currents could be deduced as
412 W. Kou and S.-Y. Park

⎧  
⎪
⎪ ed+ ed−
⎪
⎪ id+ = 2
P∗
(ed ) −(ed− )2 0
+ 2 − P∗
(ed ) −(ed− )2 cos
+ 2
⎪
⎪
3
 
⎪
⎪ +
⎪
⎪ ed+ ed−
⎨ iq = 2
3 (ed ) +(ed− )2
+ 2 Q∗0 + P∗
(ed ) +(ed− )2 sin
+ 2
  (16.9)
⎪
⎪ ed− ed+
⎪
⎪ id− = 23 (ed ) −(ed− )2
+ 2 P0∗ + P∗
(ed ) −(ed− )2 cos
+ 2
⎪
⎪  
⎪
⎪
⎪
⎪ ed− ed−
−
⎩i =
q
2
3 (ed ) +(ed− )2
+ 2 Q∗0 − P∗
(ed ) +(ed− )2 sin
+ 2 .

The iα and iβ could be derived through inverse Park transformation in positive and
negative reference frame as
⎡ +⎤
i
   ⎢ d ⎥
iα cos ωt − sin ωt cos ωt sin ωt ⎢iq+ ⎥
= ⎢ ⎥. (16.10)
iβ sin ωt cos ωt − sin ωt cos ωt ⎣id− ⎦
iq−

Finally the reference currents in αβ-axis are transformed into three-phase reference
currents through the inverse Clark transformation.
⎡ ⎤ 0 ⎡ ⎤
ia 1 √0  
⎣ ib ⎦ = 2 ⎢
⎣− 2 2√ ⎦
1 3 ⎥ iα (16.11)
3 iβ
ic − 12 − 23

Figure 16.3 shows the operation procedures of the reference currents in the
decoupled double synchronous reference control method.

Fig. 16.3 The decoupled double synchronous reference current calculation block
16 FRT and FCM for MGs 413

16.3.4 Natural Phase-Coordinates Approach

Standard regulation involves both positive and negative sequence loops and uses a
synchronous grid voltage reference framework to monitor both active and reactive
energy by switching to dq-axis [32, 37]. The d-axis average negative sequence
reference value is set to zero to eliminate double-frequency ripples on the DC-
link voltage and the three-phase output voltage induced by unbalanced faults. The
control strategy is complicated since two control loops, transformed separately
by the three-phase voltages and currents, are supplied both with the positive and
negative dq frames. This section presents a new method, natural phase-coordinates
(NPC) approach, for calculating reference current under abc phase coordinates
rather than with symmetric component decomposition to address these problems
that appeared in applying the unbalanced FRT technology. The relationship between
active power ripples and current injections has been theoretically established and
analyzed under phase coordinates in the unbalanced faults. The reference currents
can be derived under abc phase coordinates for regulating the active base power,
eliminating the active double-frequency power ripples and stabilizing DC bus
voltage.
The natural phase-coordinates approach is developed based on the instantaneous
power theory [35]. The instantaneous active power of a three-phase system can be
expressed as (16.12)

p3φ (t) = pa + pb + pc = ua ia + ub ib + uc ic (16.12)


= Ui cos(ωt − ψi )Ii cos (ωt − δi ) (16.13)
i=a,b,c
 Ui Ii
= [cos (2ωt − ψi − δi ) + cos(δi − ψi )] (16.14)
2
i=a,b,c
 
= P0,i + P2ω,i . (16.15)
i=a,b,c i=a,b,c

where uabc and iabc are the three-phase instantaneous voltage and current phasors at
the PCC of the integrated DER, respectively, under the abc phases. Equation (16.12)
could be expanded into the time-domain expression with the amplitude values and
the phase angle values, respectively, as shown in (16.13)–(16.15). Uabc and ψabc are
the amplitudes and phase angels of uabc . Iabc and δabc are the amplitudes and phase
angels of iabc . Under the balanced fault conditions, the amplitudes and phase angles
of uabc are usually kept as

U a = Ub = Uc , (16.16)
2π 2π
ψa = ψb + = ψc − . (16.17)
3 3
414 W. Kou and S.-Y. Park

If the output current is generated as the same balanced structure, the vector addition
of the three-phase double-frequency power components of p3φ (t) is zero while the
constant component of p3φ (t) remained as same as PM 0 , which is the active power
injected by MG based on its local load balance.

p2ω,a + p2ω,b + p2ω,c = 0, (16.18)

p3φ = P0,a + P0,b + P0,c = PM _0 . (16.19)

The balanced relationship of uabc in Eqs. (16.16) and (16.17) does not exist under
the unbalanced fault conditions, like the single-phase-to-ground or the double-
phase-to-ground faults. If iabc stays balanced with the unbalanced vabc , it breaks
the balance among three-phase output power in (16.18) and generates the double-
frequency active power components in p3φ . Thus, an unbalanced reference current
frame is designed in NPC approach. For a single-phase-to-ground fault, the faulted
phase is set as the benchmark phase. Without loss of the generality, here phase A is
set as the benchmark phase. The currents on the other two phases, iM _b and iM _c ,
both have δ phase angle difference to iM _a as shown in Fig. 16.5b. Considering
the application field of the three-wire connection systems, the zero-sequence
symmetrical components should be avoided in iM _abc . To generate symmetric three-
phase currents for the balanced loading system, the amplitude ratios among the
symmetric three-phase currents should satisfy

1
Icref = Ibref = Ia . (16.20)
2 cos(π − δ) ref

Thus, the generating structure of iref on the abc frame is designed as

ia _ref = Iaref cos (ωt) , (16.21)

1
ib_ref = Ia cos (ωt − δ) , (16.22)
2 cos(π − δ) ref
1
ic_ref = Ia cos (ωt + δ) . (16.23)
2 cos(π − δ) ref

As seen, the phasor function of iref is calculated by δ. In [5], δ is determined

with the assumption that uabc remains symmetrical but no longer balanced. Here,
two words must be clarified: “balanced” represents the amplitudes and phase angles
of the three-phase phasors, which exactly fulfill a relationship given in Eqs. (16.16)
and (16.17); and “symmetrical” describes a kind of disarrangement against the
“balanced” uabc , which is depicted in Fig. 16.4a, that is in mirror symmetry.
After taking into consideration the irregular loading, zero-sequence impedance on
transmission lines, as well as the grounding impedance, the three-phase system
voltages at the PCC after the unbalanced fault typically will not be symmetrical
but be asymmetrical as shown in Fig. 16.4b. In what follows, the symmetric NPC
16 FRT and FCM for MGs 415

Fig. 16.4 Phasor graphs of uabc : (a) symmetrical uabc , and (b) asymmetrical uabc

method for calculating δ under the unbalanced but symmetric faults will be
introduced at first. Then, it will be expanded to a generalized NPC method for
tackling the asymmetric cases.

16.3.4.1 Symmetric NPC Method

The following discussion is assumed that a single-phase-to-ground fault is happened

on phase A. The amplitude ratio of uM _abc after the fault is

UM _a = μUM _b = μUM _c , (16.24)

where μ ⊂ (0, 1) represents the depth of the voltage sag on phase A. The phase
angle differences of uM _abc after the fault is still kept as 23 π as shown in Fig. 16.5a.
From Eqs. (16.13)–(16.15), the double-frequency power components of pM _abc
could be extracted as follows:
416 W. Kou and S.-Y. Park

Fig. 16.5 Vector diagrams for (a) uM _abc , (b) iM _abc , and (c) p2ω,abc

UM _a IM _a
p2ω,a = cos(2ωt), (16.25)
2
 
UM _b IM _b 2
p2ω,b = cos 2ωt − π − δ , (16.26)
2 3
 
UM _c IM _c 2
p2ω,c = cos 2ωt + π + δ , (16.27)
2 3

whose vector diagram is shown in Fig. 16.5c. To eliminate the power ripple, the
amplitudes of p2ω_abc should have the following relationship:

1
P2ω,c = P2ω,b = P2ωa , (16.28)
2 cos(− 13 π + δ)

to make the phasor sum of p2ω equal to zero. Combining Eqs. (16.24), (16.20),
and (16.28), δ is derived as
16 FRT and FCM for MGs 417

Fig. 16.6 The MG FRT reference current calculation block

 
2μ + 1
δ = tan−1 − √ + π. (16.29)
3

Here, δ has the same expression as Eq. (16.29) in the double-phase-to-ground

fault with the voltage sag ratio μ in both B and C phases of uM _abc and their phase
angle offsets against A phase are still even kept as 23 π . The outcome can be applied
to any symmetric three-phase fault voltages cases with any predefined phase angle
offsets ϕ. Thus, (16.29) could be rewritten as
 
−1 cos ϕ − μ
δ = tan + π. (16.30)
sin ϕ

When ϕ = 23 π , Eq. (16.30) has the same expression of δ as in Eq. (16.29)

(Fig. 16.6 and 16.15).

16.3.4.2 Generalized NPC Method

For the asymmetric three-phase fault voltages, their amplitudes and phase angles
could be expressed as

Ua = μb Ub = μc Uc , (16.31)
ϕa = ϕb + θb = ϕc − θc , (16.32)

where μb and μc represent the voltage sag depth on phases B and C, respectively, as
shown in Fig. 16.4b. θb represents the phase angle difference between phase A and
B, and θc represents between phase A and C. By applying Eqs. (16.21)–(16.23) into
Eq. (16.14), the three-phase double-frequency ripples are expressed as
418 W. Kou and S.-Y. Park

Ua Ia
p2ω,a = cos(2ωt), (16.33)
2
Ua Ia
p2ω,b = cos(2ωt − θb − δ), (16.34)
4μb cos(π − δ)
Ua Ia
p2ω,c = cos(2ωt + θc + δ). (16.35)
4μc cos(π − δ)

In order to satisfy (16.19), the following relationship is built up:

P2ω,b cos(π − θb − δ) + P2ω,c cos(π − θc − δ) = P2ω,a , (16.36)

and the amplitude values in (16.33)–(16.35) are substituted to yield

cos(π − θb − δ) cos(π − θc − δ)

+ = 1. (16.37)
2μb cos(π − δ) 2μc cos(π − δ)

Through the trigonometric function transformation, δ is obtained as

cos θb cos θc
μb + μc − 2
δ = arctan sin θb sin θc
+ π. (16.38)
μb + μc

By comparing (16.38) with the expression of δ shown in (16.29), it is concluded

that the δ calculation for the asymmetrical uabc needs four measurement inputs,
θb , θc , μb , μc , which are more than the only one measurement input, μ in the δ
calculation for the symmetrical uabc . Utilizing the WAMS system and the PMU
devices in modern power systems, these measurements are accessible. When θb =
θc = 2π 3 , μb = μc , Eq. (16.38) will degrade to Eq. (16.29) and its compatibility is
proven.

16.4 Fault Current Management for Microgrids

Due to the fast-growing concentrated loads, IB-MGs usually connect to low-voltage

distribution networks to power the loads nearby. The substation circuit breakers of
those distribution networks are usually reaching their short-circuit duty limitations.
To ride through a grid fault, the current contribution of IB-MGs causes the substation
breakers not to cut off the short currents with the large fault currents from the utility
side, especially in the first few cycles after the fault happens [38]. MGs with FRT
ability can inject significant fault currents beyond current fault limit of the grid
equipment and then result in catastrophic damage and possible grid instability [39].
For further analysis, Fig. 16.7 shows an equivalent circuit of a system connected
with an IB-MG when a fault occurs in the upstream side of the MG at Bus_4.
Utility power system transfers power through network equivalent impedance Znet
16 FRT and FCM for MGs 419

HV/MV 1 2 3
Z net Z1 Z2 MV/LV
MG
2
T1 DS
I MG
Utility substation Ig MV/LV Zf 4

Circuit breaker If
Fig. 16.7 Short-circuit current injection of IB-MG system

and transformer T1 to the distributed power system at the left side of the substation,
Bus_1. The fault currents from the utility side Ig flow through the distribution
transmission line with equivalent impedance Z1 and fault impedance Zf . On the
right side of Bus_3, IB-MG behaves as a current source to inject IMG into fault at
Bus_4 through Z2 .
In this system, the three-phase short-circuit current If before implementing DG
can be estimated as the fault current from the utility side

E E
Ig = = . (16.39)
Znet + Z1 + Zf Z + Zf

If it is an unbalanced fault, like a single-phase-to-ground fault on phase A, the fault

current of A phase is

3E
Ig,a = , (16.40)
Z(1) + Z(2) + Z(0) + 3Zf

where Z(1) , Z(2) , and Z(0) are the positive, negative, and zero sequences short-
circuit impedance of transmission line. And the angle difference between steady-
state short current and utility source is

ωL
ϕg,a = tan−1 . (16.41)
R

After adding the injecting current from IB-MG, IM , the fault current can be obtained
as [40]

E Znet + Z1
If = + IM . (16.42)
Znet + Z1 + Zf Znet + Z1 + Zf

For the low-impedance fault (Zf  Znet + Z1 ), the fault current can be further
approximated as [17]
420 W. Kou and S.-Y. Park

Fig. 16.8 Fault current waveforms for IB-MG system. (a) IB-MG supplies power to the utility,
(b) IB-MG absorbs power from the utility

E
If ≈ + IM . (16.43)
Znet + Z1 + Zf

Here, IM is positive if IB-MG supplies power to the utility during the fault;
otherwise, it is negative. With the incorporation of IM , the magnitude of the short-
circuit current If could be higher than Ig as shown in Fig. 16.8a or lower as shown
in Fig. 16.8b.
Thus, If is at the risk of breaking the safety limits planned for the protection
system. The FCM design for IB-MG interface during FRT is manipulating the
current phase angle to satisfy
   
ig + im  = ig  , (16.44)

which is realized by shifting the phase angle of im to eliminate the magnitude

increase of if . The magnitude Ig _a and phase angle ϕg _a of ig in the fault phase
are known as shown in Fig. 16.9, with a single-phase-to-ground fault occurred in
phase A as an example. Along with changing the amplitude of im , the range of
ϕM _a variation could be any point on the circle with the radius Ig . Thus, the effect
of extra current injection from MG on increasing the short-circuit level of the utility
grid could be neutralized. Considering the direction of power flow, ϕM _a belongs to
(− π2 , π2 ). Based on the reference current structure of FRT in Eqs. (16.21)–(16.23),
a compatible FCM strategy is proposed for single MG FCM in (16.4.1). Then, the
FCM strategy illustrated in (16.44) is generalized in (16.4.2) for a multiple MG
16 FRT and FCM for MGs 421

Fig. 16.9 Phasor illustration

of FCM strategy on the
single-phase-to-ground fault

FCM case: a utility network with multiple MG connections, where im represents the
phasor sum of the fault currents from each MG.

16.4.1 Single MG Fault Current Management

To achieve Eqs. (16.44), (16.18), and (16.19) at the same time, the structural form
of the three-phase reference current, iM _ref , is designed as follows:

iref _a = M cos (ωt + R ) , (16.45)

1
iref _b = M cos (ωt − δ + R ) , (16.46)
2 cos (π − δ)
1
iref _c = M cos (ωt + δ + R ) , (16.47)
2 cos (π − δ)

of which δ is given by Eq. (16.29) to determine the relative relationship on

amplitudes and phase angles of iM _ref . Based on δ, iM _b and iM _c are changed
to iM _b and iM _c , respectively, as shown in Fig. 16.10.
With the three-phase reference currents for FRT designed in Eqs. (16.21)–
(16.23), the formulation of iM ref for FCM is modified to satisfy Eqs. (16.44)
and (16.19) by introducing two controllable variables (R and M ). iM _a and
422 W. Kou and S.-Y. Park

Fig. 16.10 Phasor

illustration of the MG FRT
reference current construction

iM _bc are expanded by the same amplitude ratio M and are rotated by the same
phase angles R . Thus, the active power flow balance described in (16.27) can be
represented as
  
1 UM _ b 2
M UM _a cos R + cos R − δ + π
2 2 cos (π − δ) 3
  (16.48)
UM _c 2
+ cos R + δ − π = PM _0 .
2 cos (π − δ) 3

M is usually configured to fulfill (16.19), as a time-varied value regulated by the

close-loop DC-link voltage control. Given M , R could be obtained from the FCM
strategy in Eq. (16.44). The calculation of R is explained in further detail below for
the different kinds of unbalanced faults. We still take phase A as the faulted phase
to discuss a single-phase-to-ground fault case. The existing short circuit occurs only
in phase A, and phase B and C have no short-circuit injection. The FCM strategy
shown in Fig. 16.10 gets M as follows:
 
. M
R = ϕM _a = π − ϕg _a − arccos . (16.49)
2Ig

FCM also remains aware of the short-circuit currents in both fault phases for the
double-phase-to-ground faults. Suppose phase B and phase C were the failure
16 FRT and FCM for MGs 423

Fig. 16.11 Phasor illustration of FCM for the double-phase-to-ground fault

phases, the total fault current injecting in the fault location is if _b , if _c . The fault
currents on the utility side are defined as ig _b , ig _c for the two fault phases with
.
phase angles ϕg _b , ϕg _c as shown in Fig. 16.11. R = αM bc can be calculated as
follows:

. αM _b + αM _c
R = ϕM _bc = , (16.50)
2
where
 
−1 M
αM _b = π − cos + δ − ϕg _b , (16.51)
4 (π − cos δ) Ig _b
 
−1 M
αM _c = cos + δ + ϕg,c , (16.52)
4 (π − cos δ) Ig _c

to eliminate the increase of the short-circuit current on phase B and C at the same
time.
424 W. Kou and S.-Y. Park

Fig. 16.12 Schematic diagram of multiple MGs interconnected to a 27-kV distributed grid

16.4.2 Multiple Microgrids Fault Current Management

The current FRT development is focused primarily on the single MG integration

scenario. The related fault current increase due to the numerous MG connections
during faults is becoming a concern by applying the FRT technique to a power
system with the fast-paced MG integration. The FCM organizational architecture
should be applied to the multi-MGs deployment scenarios. The pictorial depiction of
a distributed network connected to N MGs is provided in Fig. 16.12. The distribution
system is a conventional 10-kV radial network in one way and the MG connection
is power independent from the other networks; that is, the only power exchange link
for its MG is managed by the inverter-based device. If Bus_4 has a fault situated
on the top of the MG links, the remote utility source transfers electricity to the
distribution grid on the left side of Bus_1 through a transmission network with the
lumped impedance Zg to the transformer T1. The short current ig injected by the
remote utility grid flows through Z12 and contributes to the fault current if flowing
from Bus_2 to Bus_4. On the right side of the Bus_3, MGs serve as numerous
current sources, and a total fault current IMG is generated at Bus_3 and injected into
the Bus_4 through Z23 and Z24 . With multiple MGs integration, we approximate the
short current if as [17]

i f ≈ ig + iMG,k . (16.53)

The multi-MG FRT strategy is adjusting iMG,i of each MG within its operation
range, and the N MGs are cooperating together to eliminate the amplitude change
of if , that is,
 
Minimize If − Ig 
(16.54)
Subjectto iMG,k ⊆ Ik , k = 1, 2, · · · , N,
16 FRT and FCM for MGs 425

where Ik reflects an MG k operational scope under faults. Each MG has its inverter-
interfaced FRT control by the NPC method, which produces the fault current seen
in Eqs. (16.45)–(16.47). Section 16.4.1 has demonstrated how to achieve δ, which
provides the profile information of iref _abc on the amplitudes and phase angles.
Here, we rewrite Eqs. (16.45)–(16.47) as follows:

ia _ref = IMG cos (ωt + ϕ) , (16.55)

1
ib_ref = IMG cos (ωt − δ + ϕ) , (16.56)
2 cos (π − δ)
1
ic_ref = IMG cos (ωt + δ + ϕ) . (16.57)
2 cos (π − δ)

The other two controllable#parameters, IMG and ϕ, of iref _abc must, therefore,
be employed to manage iMG,k to realize the optimization goal function in
Eq. (16.54). Thus, by changing IMG and ϕ, the three-phase reference currents
iabc_ref are magnified by the same times as IMG and rotated by the same angles
as ϕ. It should be noted that each MG has the individual operation condition: the
inverter-based interface capacity, time-varied output power, as well as the distance
to a specific fault location. It defines the scope of the fault current, iref _abc , that
each MG could generate during the FRT operation mode. It is denoted as the “FRT
zone” of MG fault current iMG .
Let N = {1, 2, · · · , N} be the set of MG indices with FRT ability,
C = {C1 , C2 , · · · , CN } be the set of capacity indices of each MG, and
P = {P1 , P2 , · · · , PN } be the set of active output power for each MG before a
host grid fault. If Pk > 0 (k ⊆ N), MGi is injecting Pk to the connected host
network; otherwise, if Pk < 0 (k ⊆ N), MGk is absorbing Pk from the host
network. The power balance constraint for each MG is
  
1 Ub 2
IMG,k Ua cos ϕk + cos ϕk + δ − π
2 2 cos (π − δ) 3
 
Uc 2
+ cos ϕk − δ + π = Pk
2 cos (π − δ) 3
(16.58)

with the power electronic interface operation constraint:

  √
IMG,k 3
max IMG,k , × UN,k × √ ≤ Ck , (16.59)
2 cos (π − δ) 2

where UN,k represents the utility side nominal voltage of the bus connected with
MG k. The allowable amount of fault current injection from MG k, (IMG,k , ϕk ),
which is identified in Eqs. (16.58) and (16.59), is displayed as the related FRT
zones in Fig. 16.13. It shows all the phasor positions of iMG,k that MG k could
produce under the faults. Those current phasors start at O, end at any point on the
426 W. Kou and S.-Y. Park

Fig. 16.13 FRT zone for the fault current injection

line segments, Ak Bk , and form a triangular track, Ak OBk , which is the visualization
of MG k’s FRT zone. For those FRT zones located in the right plane (Ok > 0),
the associated MG injects power Pk to the utility during faults and vice versa. The
intersection point on IDER cos ϕ axis, Ok , is decided by the exchange power Pk . The
distance of OOk is related to the power capacity exchanging between the utility and
MG k during the faults. The inverter-based interface capacity, Ck , determines the
maximum value of IMGk in the FRT zone.

16.5 Hardware-in-the-Loop Platform for MG FRT

Given the fact that setting up the in-field electrical power system test system is a
time-consuming and weather-dependent process, which needs high investment and
lacks flexibility, it is challenging to meet the experimental requirements of the FRT
controller design, upgrades, and certifications test based on the aim of academics
or business. Hardware-in-the-loop (HIL) is an emergent research field that provides
both component-level and system-level testing a new approach. The physical or
rapid control prototyping for the HIL application is linked to, instead of a physical
plant, a virtual plant performed on a real-time simulator.
Real-time digital simulator (RTDS) is a simulator developed by RTDS Technolo-
gies to carry out the real-time power system simulations. It is designed to simulate
16 FRT and FCM for MGs 427

Fig. 16.14 Co-simulation HIL test setup

Fig. 16.15 The power system configuration in RSCAD

in real-time versions of the control network in 50 μs and to simulate electronic

devices in the time steps smaller than 2 μs. RTDS is a mix of software and hardware
experimental platforms. The architecture of the hardware consists of optical signal
processors, RISC processors, I/O cards, and power supply, arranged into single
racks. The simulator can communicate with external devices through analog and
digital I/O ports. RSCAD, a personalized app package, supports the user-to-user
interface. RTDS, OP4510 simulator, level shifting circuit, and DLP750P scope
recorder are used for building this co-simulation testbed. RTDS has been used in
real-time simulating grid and failure phases. The inverter was powered by OP4510.
In the GTAO card of RTDS, the voltage and current measurements are scaled down
and used on the OP4510 analog display. Six PWM signals are transmitted from
OP4510 to RTDS GTDI. The fault-triggering condition is detected by the GPC pin
in RTDS. For voltage and current waveform monitoring, DL750P scope recorder
was used. The testbed layout is shown in Fig. 16.14 [13] (Fig. 16.15).
428 W. Kou and S.-Y. Park

Fig. 16.16 The control system diagram in OPAL-RT

On the other side, by using OPAL-RT OP4510 [41], the FRT control device on
the HIL testbed has been incorporated. The analog inputs of the FRT control model
are the scaled-down voltage and current measurements, Vabc and Iabc , which are
within ±20-volt scale. At the same time, it feeds RTDS six PWM signals. Platforms
of 16-bit resolution [42] are available on ADC platforms. By using the additional
circuit, the PWM signals are created between 0 V and 5 V, and up to 180 kHz. The
OP4510 is equipped with the Intel quad-core 3.3 GHz Intel Xeon processor, which
can run up to 50 kHz without any overrun. The Opal-RT simulation control models
are depicted in Fig. 16.16, and the process flowchart of the RTDS and OPAL-RT
co-simulation platform is shown in Fig. 16.17.

16.6 Analysis and Results

On the distributed power grid network, as seen in Fig 16.12, the developed MG
FRT methods under the utility faults with the unbalanced voltage sag are tested.
To have the high switch frequency inverter simulation model, we use the RTLAB
real-time simulation OPAL-RT toolbox, which is completely combined with MAT-
LAB/Simulink. The device specifications and parameters of the simulation are listed
in Table 16.1.
16 FRT and FCM for MGs 429

Fig. 16.17 Co-simulation workflow

Table 16.1 Simulation test configuration

Network specifications
Utility system voltage and frequency 27 kV, 60 Hz
IB-MG interface capacity 1 MW
Line posi-/zero-sequence resistance 0.2153/0.813 /km
Line posi-/zero-sequence inductance 1.05e−3/3.02e−3 H/km
Length of line 1−2, 2−4, 2−3 20 km, 20 km, 5 km
Simulation settings
Simulation step size 50 μs
IB-MG interface converter PWM frequency 10 kHz

At t = 1 s, Bus_3 induces a single-phase-to-ground fault, and the inverter-

interfaced DER device switches its output current control from the standard
operating mode to FRT mode by utilizing three separate approaches: (1) the
traditional method under the dq frame, (2) the symmetric NPC method, and (3)
the generalized NPC method. The comparison results of these three methods on the
three-phase output voltages and currents vabc , iabc , the active power p0 , and the
double-frequency active power p2ω on Bus_4 are shown in Fig. 16.18.
When the fault arises, an asymmetrical vabc deviated from the balanced three-
phase voltage profile at the Bus_3 has occurred. Apart from the apparent decrease
in the amplitude of va , there are different sag ratios, μb = 0.225 and μc = 0.2, in
430 W. Kou and S.-Y. Park

Fig. 16.18 Left: FRT operation under the dq frame; middle: FRT operation with symmetrical
NPC method; right: FRT operation with generalized NPC method. (a) vabc . (b) iabc . (c) p0 . (d)
p2ω

nondefault phase voltages vb and vc . The discrepancies in phases for vabc are not the
same as 120◦ : va and vb have 131◦ phase variation, and vc and vb have 238◦ . Thus,
the voltage profile on Bus_3 connected with DER is asymmetrical. During the fault,
the output currents of the traditional dq frame method and the symmetrical NPC
method are unbalanced, as shown in Fig. 16.18 right(b) and Fig. 16.18 middle(b),
while the DER holds the p0 steady, as seen in Fig. 16.18 right(c) and Fig. 16.18
middle(c), following the fast fluctuation in the first two frequency cycles.
By comparing to the other two approaches, the efficiency of the generalized
NPC approach for the reduction of power rips through unsymmetrical failures is
validated. The double-frequency ripples, p2ω , have been eliminated in Fig. 16.18
right(d), while the peak-to-peak value of p2ω is 2.1 × 105 W in the symmetrical
NPC as shown in Fig. 16.18 middle(d) and is increased to 4 × 105 W in the
traditional dq frame method in Fig. 16.18 left(d). For asymmetrical voltage failures,
the generalized NPC method works better at power ripples elimination.
The proposed FRT control scheme is also realized on the co-simulated platform
consisting of RT-LAB and Opal-RT simulator. Figure 16.19 displays the three-phase
output voltages vabc and currents iabc of the inverter-based interface at the PCC to
the utility. After the unbalanced failures, the proposed NPC FRT regulation produces
the unbalanced three-phase fault currents. Phase A’s magnitude is higher than that
of phase B and phase C in the single-phase-to-ground fault case. The magnitude of
phase C is higher than that of phase A and phase B in the double-phase-to-ground
fault case.
16 FRT and FCM for MGs 431

Fig. 16.19 (a) Real-time HIL simulation results on single-phase-to-ground fault at phase A; (b)
Real-time HIL simulation results on double-phase-to-ground fault at phase A and phase B

16.7 Summary

This chapter discussed available IB-MG integration requirements and control

technologies from FRT capability viewpoint and system aspects. A new FRT control
strategy embedded IB-MG to realize flexible NPC current control during faults is
proposed. The related FCM strategies for single MG FRT, as well as multiple MGs
FRT, are formulated. A co-simulation HIL testbed comprised of OPAL-RT and
RTDS real-time simulators is developed. The motivation of this work was to explore
DER integration in a bulk power system where MG FRT development plays an
essential role in system resilience improvement and coordination management. The
real-time experimental results also identify the control performance of the proposed
MG FRT on output active power and three-phase currents.

Appendix

16.7.1 FRT Reactive Power Injection

In recent years, many utilities and network system operators are posing requirements
for large distributed generators to support grid voltage in a specific voltage drop
range (the grid voltage sag depth μ ≥10%) [18]. In FRT control, the MPPT feature
of DER is disabled and is replaced by the DC-link voltage control. The IB-MG
interface has the ability to provide not only the injection of active power but also
the contribution of reactive power. This requires the grid-connected MG injecting
a reactive current iq∗ in proportion to μ, as shown in (16.60). Under specific severe
scenarios (μ ≥50%), the current injection can even be purely reactive.
432 W. Kou and S.-Y. Park

⎧
⎨ 0 0.9  μ < 1.1
iq∗ = 2 · μ · IN 0.5  0.9  μ < 1.1 (16.60)
⎩
IN μ < 0.5

16.7.2 Synchronous Reference Frame PLL

The PLL based on the synchronous reference frame (SRF-PLL) is the most
popular technique used for frequency-insensitive grid synchronization in three-
phase structures. By using the Park’s transformation, as shown in Fig. 16.20, the
standard SRF-PLL converts the three-phase voltage vector from the abc natural
reference frame to the dq rotating reference frame. This dq reference frame’s
angular position is controlled by a feedback loop, which controls the q element
to zero.
⎡ ⎤   ⎡v ⎤
  va
vα 2 1− 1
−√2
1 a
= Tαβ · ⎣vb ⎦ = √2 · ⎣vb ⎦ (16.61)
vβ 3 0 2 − 2
3 3
vc vc
     
vd+ √ 1 √ cos(−2ωt)
vdq+ = = Tdq+ vαβ = 2V+ + 2V− (16.62)
vq+ 0 sin(−2ωt)
     
vd− √ 1 √ cos(−2ωt)
vdq− = = Tdq− vαβ = 2V− + 2V+ , (16.63)
vq− 0 sin(−2ωt)

where
 
cos θ sin θ
Tdq = . (16.64)
− sin θ cos θ

Fig. 16.20 Block diagram of the SRF-PLL

16 FRT and FCM for MGs 433

16.7.3 Deadbeat Current Control

The deadbeat controller is part of the predictive controller family. These are based
on a common principle: choose the state of the converter, ON–OFF predictive status,
or the average voltage generated by the converter (predictive with a PWM) based
on foreseeing the evolution of the controlled quantity (the current) [31]. According
to the Kirchhoff voltage law of a three-phase symmetrical system, the loop voltage
equation is
⎧ di di
⎨ Ua − Ub = −L dta + L dtb + Uuv
dib dic
U − Uc = −L dt + L dt + Uvw (16.65)
⎩ b
Uc − Ua = −L di dia
dt + L dt + Uwu
c

The related loop voltage equation is

⎧ di di
⎨ ua − ub = −L dta + L dtb + (du Udc − dv Udc )
dib dic
u − uc = −L dt + L dt + (dv Udc − dw Udc ) (16.66)
⎩ b
uc − ua = −L di dia
dt + L dt + (dw Udc − du Udc )
c

dk = 1 is defined as k-phase upper bridge switch is on and lower bridge switch is

off. dk = 0 is defined as k-phase upper bridge switch is off and lower bridge switch
is on. We name the control period as T and the grid base period as Tg . Let T  Tg
to discretize the loop voltage equation, and it can be considered that the three-phase
grid voltage and the DC bus voltage are constant within a control period. Then, the
loop voltage equation in a control cycle could be deduced as
⎧ ia∗ −ia ib∗ −ib
⎪
⎨ ua − ub = −L ∗ T + L T + (du Udc − dv Udc )
i −i i ∗ −i
⎪ ub − uc = −L b T b + L c T c + (dv Udc − dw Udc ) (16.67)
⎩ ∗
i −i i ∗ −i
uc − ua = −L c T c + L a T a + (dw Udc − du Udc )

Among them, ia∗ , ib∗ , and ic∗ are three-phase current reference values; du , dv ,
and dw are the three-phase switching duty cycles of the inverter. Since ua + ub +
uc = 0, the system of (16.67) has only two independent equations. At the same time,
in a switching cycle, the total on-time of the three switching tubes on the upper arm
of the three-phase inverter and the corresponding total on-time of the lower arm are
equal. Thus, we could have
⎧ i ∗ −ib
⎪ i ∗ −ia
⎨ua − ub = −L a T + L b T + (du Udc − dv Udc )
∗
i −i i ∗ −i
⎪ ub − uc = −L b T b + L c T c + (dv Udc − dw Udc ) (16.68)
⎩
du + dv + dw = 1.5.

The duty cycle of the PWM signal of the three-phase inverter could be solved as
follows:
434 W. Kou and S.-Y. Park

⎧    
i ∗ −i i ∗ −i i ∗ −i i ∗ −i
⎪
⎪ 1.5Udc +2 ua −ub +L a T a −L b T b + ub −uc +L b T b −L c T c
⎪
⎪
⎪
⎪ du =
⎪
⎨  3Udc 
i ∗ −i i ∗ −i

i ∗ −i i ∗ −i
1.5Udc − ua −ub +L a T a −L b T b + ub −uc +L b T b −L c T c
(16.69)
⎪
⎪ dv =
⎪
⎪  3Udc 
ib∗ −ib ib∗ −ib

⎪
⎪ − −u +L
ia∗ −ia
−L −2 −u +L −L
ic∗ −ic
⎪
⎩d =
1.5Udc ua b T T ub c T T
w 3Udc .

References

1. Ackermann, T. (2005). Wind power in power systems. Hoboken, NJ: Wiley.

2. Jauch, C., Sørensen, P., Norheim, I., & Rasmussen, C. (2007). Simulation of the impact of
wind power on the transient fault behavior of the Nordic power system. Electric Power Systems
Research, 77(2), 135–144.
3. Ellis, A., & Gonzalez, S. (2014). Implementation of Voltage and Frequency Ride-Through
Requirements in Distributed Energy Resources Interconnection Standards
4. Nair, N.-K. C., & Qureshi, W. A. (2014). Fault ride-through criteria development. In Renewable
energy integration (pp. 41–67). New York: Springer.
5. Kou, W., Wei, D., Zhang, P., & Xiao, W. (2015). A direct phase-coordinates approach to fault
ride through of unbalanced faults in large-scale photovoltaic power systems. Electric Power
Components and Systems, 43(8–10), 902–913.
6. Gkavanoudis, S. I., Oureilidis, K. O., & Demoulias, C. S. (2013). Fault Ride-Through
Capability of a Microgrid with WTGs and Supercapacitor Storage During Balanced and
Unbalanced Utility Voltage Sags (pp. 231–236).
7. Moursi, E., Shawky, M., Xiao, W., & Kirtley, J. L. (2013). Fault ride through capability for
grid interfacing large scale PV power plants. Generation, Transmission & Distribution, IET,
7(9), 1027–1036.
8. Rodriguez, P., Timbus, A. V., Teodorescu, R., Liserre, M., & Blaabjerg, F. (2007). Flexible
active power control of distributed power generation systems during grid faults. IEEE
Transactions on Industrial Electronics, 54(5), 2583–2592.
9. Wang, F., Duarte, J. L., & Hendrix, M. A. M. (2011). Pliant active and reactive power control
for grid-interactive converters under unbalanced voltage dips. IEEE Transactions on Power
Electronics, 26(5), 1511–1521.
10. Kou, W., Wei, D., Zhang, P., & Xiao, W. (2015). A direct phase-coordinates approach to fault
ride through of unbalanced faults in large-scale photovoltaic power systems. Electric Power
Components and Systems, 43(8–10), 902–913.
11. Kou, W., & Park, S.-Y. (2017). Generalized direct phase-coordinates control of inverter-
interfaced distributed energy resources on fault ride through strategies. In Power Symposium
(NAPS), 2017 North American (pp. 1–6). New York: IEEE.
12. Kou, W., & Wei, D. (2018). Fault ride through strategy of inverter-interfaced microgrids
embedded in distributed network considering fault current management. Sustainable Energy,
Grids and Networks 15, 43–52
13. Kou, W., Rakiul Islam, S. M., & Park, S.-Y. (2018). Co-simulation testbed for unbalanced fault
ride through hardware in the loop validation. In 2018 IEEE Electronic Power Grid (eGrid)
(pp. 1–5). New York: IEEE.
14. Zheng, F., Deng, C., Chen, L., Li, S., Liu, Y., & Liao, Y. (2015). Transient performance
improvement of microgrid by a resistive superconducting fault current limiter. IEEE Trans-
actions on Applied Superconductivity, 25(3), 1–5.
15. Ghanbari, T., & Farjah, E. (2013). Unidirectional fault current limiter: an efficient interface
between the microgrid and main network. IEEE Transactions on Power Systems, 28(2), 1591–
1598.
16 FRT and FCM for MGs 435

16. Rajaei, N., Ahmed, M. H., Salama, M., & Varma, R. (2014). Fault current management using
inverter-based distributed generators in smart grids. IEEE Transactions on Smart Grid, 5(5),
2183–2193.
17. Yazdanpanahi, H., Li, Y. W., & Xu, W. (2012). A new control strategy to mitigate the impact of
inverter-based DGs on protection system. IEEE Transactions on Smart Grid, 3(3), 1427–1436.
18. Tarafdar Hagh, M., & Khalili, T. (2019). A review of fault ride through of PV and wind
renewable energies in grid codes. International Journal of Energy Research, 43(4), 1342–
1356.
19. Colombo, E., Bologna, S., & Masera, D. (2013). Renewable energy for unleashing sustainable
development (Vol. 10, pp. 978–983). Cham: Springer International Publishing.
20. Guttromson, R., & Glover, S. (2014). The Advanced Micro-Grid Integration and Interoper-
ability.
21. Hossain, J., & Pota, H. R. (2014). Robust control for grid voltage stability: High penetration
of renewable energy. New York: Springer.
22. Basso, T. S. (2014). IEEE 1547 and 2030 standards for distributed energy resources
interconnection and interoperability with the electricity grid (Vol. 15013). Golden, CO:
National Renewable Energy Laboratory.
23. Crăciun, B.-I., Kerekes, T., Séra, D., & Teodorescu, R. (2012). Overview of Recent Grid Codes
for PV Power Integration (pp. 959–965).
24. Jiang, L., Wang, C., Huang, Y., Pei, Z., Xin, S., Wang, W., et al. (2015). Growth in wind
and sun: Integrating variable generation in China. IEEE Power and Energy Magazine, 13(6),
40–49.
25. IEEE standard for interconnecting distributed resources with electric power systems - amend-
ment 1. IEEE Std 1547a-2014 (Amendment to IEEE Std 1547–2003), May 2014 (pp. 1–16)
26. El-Naggar, A., & Erlich, I. (2017). Short-circuit current reduction techniques of the doubly-fed
induction generator based wind turbines for fault ride through enhancement. IET Renewable
Power Generation, 11, 1033–1040.
27. Baghaee, H. R., Mirsalim, M., Gharehpetian, G. B., & Talebi, H. A. (2017). A new current
limiting strategy and fault model to improve fault ride-through capability of inverter interfaced
DERs in autonomous microgrids. Sustainable Energy Technologies and Assessments, 24(C),
71–81.
28. Dahale, S., Das, A., Pindoriya, N. M., & Rajendran, S. (2017). An overview of DC-DC
converter topologies and controls in DC microgrid. In 2017 7th International Conference
on Power Systems (ICPS), pp. 410–415. New York: IEEE.
29. Backhaus, S. N., William Swift, G., Chatzivasileiadis, S., Tschudi, W., Glover, S., Starke, M.,
et al. (2015). DC Microgrids Scoping Study. Estimate of Technical and Economic Benefits.
Technical report, Los Alamos National Lab (LANL), Los Alamos, NM.
30. Fei, G., Ren, K., Jun, C., & Tao, Y. (2019). Primary and secondary control in dc microgrids: A
review. Journal of Modern Power Systems and Clean Energy, 7(2), 227–242.
31. Teodorescu, R., Liserre, M., et al. (2011). Grid converters for photovoltaic and wind power
systems (Vol. 29). Hoboken, NJ: Wiley.
32. Junyent-Ferré, A., Gomis-Bellmunt, O., Green, T. C., & Soto-Sanchez, D. E. (2011). Current
control reference calculation issues for the operation of renewable source grid interface VSCs
under unbalanced voltage sags. IEEE Transactions on Power Electronics, 26(12), 3744–3753.
33. Shawky, M., Moursi, E., Xiao, W., & Kirtley, J. L. Jr. Fault ride through capability for grid
interfacing large scale PV power plants. IET Generation, Transmission & Distribution, 7(9),
1027–1036.
34. Eloy-Garcia Carrasco, J., Tena, J. M., Ugena, D., Alonso-Martinez, J., Santos-Martin, D., &
Arnaltes, S. (2013). Testing low voltage ride through capabilities of solar inverters. Electric
Power Systems Research, 96, 111–118.
35. Akagi, H., Watanabe, E. H., & Aredes, M. (2017). Instantaneous power theory and
applications to power conditioning. Hoboken, NJ: Wiley.
36. Song, H.-S., & Nam, K. (1999). Dual current control scheme for PWM converter under
unbalanced input voltage conditions. IEEE Transactions on Industrial Electronics, 46(5), 953–
959.
436 W. Kou and S.-Y. Park

37. Camacho, A., Castilla, M., Miret, J., Vasquez, J. C., & Alarcón-Gallo, E. (2012). Flexible
voltage support control for three-phase distributed generation inverters under grid fault. IEEE
Transactions on Industrial Electronics, 60(4), 1429–1441.
38. Baran, M. E., & Mahajan, N. R. (2007). Overcurrent protection on voltage-source-converter-
based multiterminal dc distribution systems. IEEE Transactions on Power Delivery, 22(1),
406–412.
39. Salcedo, R., Bokhari, A., Diaz-Aguiló, M., Lin, N., Hong, T., de León, F., Czarkowski, D.,
Flank, S., McDonnell, A., & Uosef, R. E. (2016). Benefits of a nonsynchronous microgrid on
dense-load LV secondary networks. IEEE Transactions on Power Delivery, 31(3), 1076–1084.
40. Wang, Q., Zhou, N., & Ye, L. (2015). Fault analysis for distribution networks with current-
controlled three-phase inverter-interfaced distributed generators. IEEE Transactions on Power
Delivery, 30(3), 1532–1542.
41. OPAL-RT Technologies (2015). Op4510 RT-LAB-RCP/HIL Systems User Guide.
42. OPAL-RT Technologies (2015). Op5340-1/op5340-2 User Guide: Analog to Digital Converter
Module.
Chapter 17
Microgrid Protection

Arturo Conde Enríquez, Yendry González Cardoso,

and José Treviño Martínez

17.1 Introduction

Developments in the field of electricity generation in recent decades have been ori-
ented toward environmental conservation and greater efficiency, due to the increase
in fossil energy costs. This has modified the vertical structure of electrical networks
mainly affecting medium- and low-voltage networks, resulting in isolated electrical
systems or weak connection electrical networks. In recent years, the installation
of renewable generation sources (RASs) in medium- and low-voltage networks
has undergone a great deal of growth, and the horizontal structure of electrical
networks has led to the formation of microgrids (MGs). These electrical networks
have operational benefits since the power of the load is local and the problems of
network congestion, power losses, demand peaks are mitigated, hence reducing the
operation and generation costs. These networks also have the characteristic that they
integrate highly dynamic elements with significant limitations on controllability.
The connection of generation sources with an intermittent nature determines both
the topological and dynamic power conditions of the power grid.
The energy supplied by the MG must be reliable, and of sufficient quality that
the electrical network can continue to operate under adequate conditions. This has
prompted the development of distributed control systems with hybrid architectures
combining local and distributed control systems. These can improve the operation
of the power grid and the quality of protection.
MG protection systems are particularly affected by highly dynamic topologies
and operating conditions. The traditional philosophy for adjusting protections con-
siders the worst operating conditions of the electrical network. These adjustments,

A. C. Enríquez () · Y. G. Cardoso · J. T. Martínez

Faculty of Mechanical and Electrical Engineering, Autonomous University of Nuevo Leon,
Nuevo Leon, Mexico

© Springer Nature Switzerland AG 2021 437

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_17
438 A. C. Enríquez et al.

which are traditionally constant, often result in problems with sensitivity, selectivity,
or inadequate operating times. Adaptive relay systems offer an opportunity to
increase the protection of MG systems. The online settings of protection schemes
can be modified depending on the topological and operational conditions of the
electrical system. A multi-agent system (MAS) can be used to improve protection
performance for electrical networks under dynamic operating conditions.

17.2 Microgrids

MGs show highly dynamic behavior in terms of both their topology and their active
elements. The integration of an RAS into the power grid has multiple impacts, as
non-dispatchable generation sources with an intermittent nature require both voltage
and energy support elements. These sources are decoupled from the frequency of
the AC network, and since they do not involve energy storage do not provide inertia
to the electricity network; it is, therefore, necessary to supply the low inertia for
these systems by other means, for example through the use of storage systems or
converters. In addition, the proximity of the loads implies an additional dynamic
that must be considered. The reactive power demand of the electronic elements
imposes a condition of high reactive power demand, and since this is not provided
by the utility company, it must be provided by converters or storage systems.
In addition, the connections of the shunt elements must be analyzed to avoid
harmonic amplification when generating frequency resonance points. These aspects
of operation must be considered in the design of protection systems for the MG.
In the operation of the MG, it is necessary to use DMS to implement strategies
for the operation of the network from the economic and operational points of
view, such as the dispatch of units and energy management. The design of DMS
systems should consider the contribution of converters as one of the sources of
voltage support, an important aspect of which is the characterization of converters in
terms of their diversity of design or operation and their behavior under short circuit
conditions. It is also necessary to establish operation criteria in real time, due to the
high operational intermittency. This condition also affects protection systems, since
conventional criteria do not meet the requirements of the MG in many cases. The
use of adaptable online adjustment systems, a centralized control system, and the
activation/deactivation of protection schemes will ensure the creation of dynamic
protection zones that offer both sensitivity and acceptable operating times.

17.2.1 MG Protection

Low-voltage electrical systems have traditionally been designed to operate radially,

meaning that protection techniques are generally shaped by overcurrent protection.
These protection schemes are traditionally adjusted to maximum or minimum
17 Microgrid Protection 439

border values to ensure dependable and/or safe operation, requiring constant

regulation, and without monitoring the operational state of the power grid.
The topological changes arising from interconnected configurations, bidirec-
tional flows, and the highly intermittent operation of MGs mean that it is difficult
for conventional schemes to meet the protection requirements of an MG. We must
therefore consider the design of online protection systems with centralized control
systems and continuous monitoring of the operational state of the network, with
systems for readjustment of the relays and the use of dynamic protection zones.
It is important that these protection schemes are consistent with the mode of
operation of the MG. When operating in connected mode, the utility company
provides the reactive power required for voltage support and the main contribution
of the fault current. Although the sensitivity of the relays is not compromised, the
location of the fault can be problematic due to the flat values of the magnitude of the
fault currents. A detection system for island operation is required if the MG cannot
operate while temporarily isolated. In the case of disconnected operation, the need
for a protection system is even greater due to the reduced fault current, temporary
loss of voltage regulation, and the highly dynamic and intermittent operation of the
MG.

17.2.2 Problems and Functional Solutions for Relays

The use of communication and measurement channels through μPMU is generally

required as part of the design of the protection scheme. The algorithms like
Newton Raphson for power flow compute can be omitted and replaced by the
phasor measurement. The phasor update must be less than those used to calculate
the demand. The fault current must be updated before topological changes are
made. In particular, the calculation of n-1 contingencies in flows and faults for the
readjustment of relays must be considered. Data updates must consider the latency
of the communication system and the dynamics of the variables that will be updated,
either due to topological changes or variations in the load current.
The functional limitations are different for each aspect of protection, and
simpler protections such as overcurrent will be more strongly affected by the
dynamic conditions of the MG, while those that require more information about
the electrical system with multipoint measurements (such as differential protection)
or the measurement of a greater number of signals (such as distance) will be
less affected. It is important to evaluate the coordination between the different
protection principles. An adjustment and collaboration strategy between relays can
be performed in an MAS, where each relay can be an agent. The MAS system can be
integrated using a collaborative structure between protection agents and measuring
agents, for example. Agents can also be integrated via communication networks to
form the protection system of the electrical network. The limitations and areas of
opportunity for each aspect of protection are described below.
440 A. C. Enríquez et al.

Overcurrent relays. Directional overcurrent relays (DOCRs) must be used in

mesh systems with selective operation against bidirectional currents in the protected
line. For protection in networks with a limited current contribution to the fault,
voltage retention schemes may be required in order to increase the selectivity.
Traditionally, the use of standardized time curves has been a good solution for
the adjustment of relays, and the adjustment of the initial current should be carried
out via the maximum value of the load current, power transfer scenarios, and n-
1 contingencies. However, due to the coordination requirements of an MG, these
criteria are generally inadequate. The use of digital relays that can change the setting
parameters remotely through communication channels, in pre-established periods or
when detecting any topological change in the network, allows various adaptable
solutions to be incorporated, thus minimizing the limitations and increasing the
applications of MGs. The design of nonconventional flexible curves with more
degrees of freedom allows a time curve to be designed for each specific coordination
problem. In an MG, the time curve for an overcurrent relay may change according
to the coordination requirements under these conditions of operation. Likewise,
updating the pickup current with the measured load current flow will allow the
sensitivity of the relay to be increased and adaptable settings can be updated, to
ensure that safety is not compromised.
The coordination of DOCRs has been formulated as an optimization problem,
and metaheuristic methods have generally been used due to their robustness in
finding solutions in schemes lacking an initial condition. In this approach, an
objective function is established and used to perform a search, in order to minimize
the operation time of the relays and thus ensure that the coordination requirements
are met. In this way, equality and inequality restrictions are established to comply
with the relay time calculation and the sequence of operation between the primary
and backup relays. Heuristic optimization methods can search for solutions over
a larger area than deterministic ones, which gives them an advantage in terms of
finding better solutions, although they are typically expected to be slower. In the
search for new solution methods, approaches such as differential evolution (DE)
and gray wolf optimization (GWO) have been proposed, which offer adequate
calculation time for online coordination applications. In the weighting of the
optimization method used for online coordination, the dispersion of the results
obtained must be considered, since heuristic methods will give different results from
each execution. Some approaches such as genetic algorithms (GAs) have a high
level of dispersion in the results, that is, with substantial differences in each run,
while DE gives a small standard deviation, with very similar results in each run.
Relay tuning is commonly done off-line and it is a constant under the dynamic
operation of the electrical grid. However, in applications for MG protection, due
to dynamic operating conditions such as topological changes and the intermittent
contributions from renewable sources, applications are needed that give solutions
within a few seconds to enable online coordination. A flow chart for online
coordination applications is presented in Fig. 17.1. Updating of the relay settings
must be carried out in a stable state, so topological changes and the demand currents
must be considered when updating the pickup current and the coordination of relays.
17 Microgrid Protection 441

Fig. 17.1 On-line coordination diagram

The use of a micro phasor (μPMU) in the measurement system can simplify these
calculations and avoid the need for computation of power flow. It is important to
constantly calculate fault currents since the contributions from renewable sources
and topological changes have a great impact on the criteria for setting relays.
Differential relays. Differential protection is one of the most efficient types
of protection; it has existed for more than 100 years and has a wide range of
applications. The differential principle has absolute selectivity and has been used
442 A. C. Enríquez et al.

both for the protection of primary equipment and for the protection of transmission
lines and distribution feeders. The trip decision is based on a comparison of the input
currents and output currents of the equipment to be protected. The operation of the
differential protection system can be affected by conditions that increase the error
of differential current, and factors such as a change of tap in transformers, unequal
saturation of the current transformers, and internal high impedance can compromise
its performance.
The sensitivity of differential protection in active distribution network applica-
tions may be compromised under conditions of high fault impedance and weak
source (WS) inputs. In general, under external fault conditions where the flow of
current entering and leaving the scheme does not make a significant difference,
protection is not affected. However, connection via electronic converters generates
nonfundamental frequency components, and together with the connection of non-
linear loads and the presence of shunt capacitors, which are used to compensate
for the associated low voltage in this type of network, high-frequency pollution
scenarios commonly arise. In feeder schemes, the magnitude of the nonfundamental
frequency components changes depending on the impedance of the transformer,
generating a differential current of error that is associated with inter-harmonics and
subharmonics (Fig. 17.2). Designs for new differential protection schemes based
on positive sequence fault components have been reported to be able to mitigate
the effect of active distribution networks due to the high penetration of distributed
generators (DGs).
The design of dynamic protection zones, in which the nodes of the protection
zone are modified depending on the topology or operating conditions of the MG,
will allow for greater sensitivity. A differential scheme will be designed in a control
center, based on the measured synchronized signals. The communication channels
may be wired or wireless; this gives great operational flexibility, and the advantages
or disadvantages of communication must be evaluated. A further important aspect
to consider is that differential protection requires a backup system.
Distance relay. The principle underlying this form of protection involves the
apparent impedance of the line, which results from the measurement of two input
signals, voltage, and current. The apparent impedance calculated by the relay
coincides with the impedance of the line when power is supplied to a single source;
as the voltage is a function of the fault current, the impedance is independent of
the operational state or the Thevenin equivalent. However, when there is a bilateral
supply, an offset is generated between the current measured by the relay and the
current from the other end, causing an imaginary component in the fault impedance.
Relay detection will depend on the characteristics of the design and their
adjustment, so the reach for its first zone in short lines with small impedance is
difficult to obtain. In addition, the characteristics of the distance relay are affected by
the presence of intermediate sources, with underreach or overreach depending on the
strongest contribution and the pre-fault status (Fig. 17.3). Thus, operation times may
be longer or short than desired. The presence of inter-harmonics or subharmonics,
which are generated mainly by wind power plants (WPPs), can cause erroneous
impedance signals.
17 Microgrid Protection 443

Fig. 17.2 (a) Differential relay in an MG; (b) external fault current contaminated with subhar-
monic and inter-harmonic components

In networks where sensitivity to fault detection is lost, the use of distance relays
can be an attractive alternative. Their coordination with overcurrent relays has been
well studied and is feasible. In this approach, the intermittency of an RAS does not
affect the range of the distance relay as in the overcurrent relay, although the effects
of intermediate sources and contamination of the input signals to the relay must be
evaluated.

17.2.3 Discussion

The operation of the MG in protection systems needs to be evaluated. The operation

in mode connected to the utility offers operational support to the MG through the
444 A. C. Enríquez et al.

Second zone
X
4

Without
infeed

With
3 infeed

2 R

Fig. 17.3 Infeed effect in distance relays

contribution of reactive power and the contribution of fault current, the intermittent
dynamics of the MG are less representative as well as the performance of the
protection system. On the other hand, island operation, in which the system is
disconnected from the utility or its connection through high-impedance links,
imposes greater requirements on the protection system, island detection schemes
must be activated in order to switch the settings of the protection schemes. If the
fault is in the utility, the protection system may not detect this contribution, and
weak source detection is necessary for the separation of the MG.
An MAS can improve the function of protection schemes in an MG. Compu-
tational platforms must allow for the integration of various dynamics of elements
of the electrical network in a centralized way. Through the real-time monitoring of
electrical quantities and topological detectors, a multi-agent environment can carry
out the different protection actions that are required in the power grid.

17.3 Multi-Agent System Proposals

Distributed control is one method that has been proposed to make energy distribu-
tion more efficient. The development of distributed control is closely associated with
communication, and it is increasingly necessary to use intelligent devices to control
or remotely monitor distribution networks. The development of microprocessors,
microcontrollers, and programmable logic controllers (PLCs) has resulted in the
emergence of distributed control [1].
17 Microgrid Protection 445

Applications currently under investigation for use in distributed control are:

• Restoration of the electrical system;
• The operation of active distribution networks;
• Micro network control; and.
• Control of electrical systems.
In modern intelligent networks, there are ever more types of new elements,
making them increasingly complex [2]. Cooperative distribution has been receiving
a great deal of attention as a promising technology for the growth of wireless
networks [3, 4]. The authors of [2] propose a new protection scheme for distributed
control for a distribution system using an MAS with a rapid relay response and high
efficiency under different conditions.
One of the most common phenomena that occur during abnormalities in an
electrical power system, and especially in short circuits, is an increase in current
over normal operating values [5]. The operation of a protection system must also be
safe and must avoid tripping under normal operating conditions, such as overloads,
power transfer due to topological changes and intermittency, and a lack of control
over RASs. Due to the significant interest in the use of MAS, the design of
solutions the standards and methodologies are considered [6]. MASs are now used in
many applications such as diagnosis, monitoring, power system restoration, market
simulation, network control, and automation. This technology has matured to the
point where the first MAS is being migrated from the laboratory to a physical
system, allowing the industry to gain experience in the use of MASs and to evaluate
their effectiveness [6].
The IEEE Power Engineering Society (PES) formed a working group in conjunc-
tion with the PSACE committee to investigate these questions in relation to MASs.
In this section, a comparison is made with existing technologies for the design and
implementation of MASs, such as web services and grid computing.

17.3.1 Definition of a Smart Agent

There are many definitions of an agent, and although these differ, they share a basic
set of concepts: the notions of an agent, its environment, and autonomy. According
to Wooldridge [7], an agent is merely “a software entity (or hardware) that is placed
in an environment and is capable of reacting to changes in those environments.” The
environment is simply defined as everything external to the agent. At least in order
to be placed in an environment, one part of it could be observable or that could be
altered by the agent. An agent can alter the environment by taking an action: either
physical (such as closing a normally open point in a network), or computational
(saving diagnostic information in a database).
Obtaining an answer back of the agent should not affect the actions that are
taken or the goals for which it was designed. Autonomy means that agents “execute
control over their own actions,” meaning that they can initiate or schedule certain
446 A. C. Enríquez et al.

actions for execution based on environmental observations. From an engineering

perspective, this definition is problematic, as it is not easy to distinguish an agent
from another element of the network. For example, protection from an overcurrent
relay could be considered as an agent, as it is located within an environment (that
is, the power system), and reacts to changes in its environment (that is, changes in
voltage or current). It also has a degree of autonomy.
An intelligent agent has the following three characteristics:
• Reactivity: An intelligent agent is able to react to changes in its environment and
take action based on these changes and the function for which it was designed.
• Proactivity: Intelligent agents exhibit behavior with targeted goals. This means
that an agent can dynamically change its behavior in order to achieve its goals.
For example, if an agent loses communication with another agent that was
providing a service required to achieve its goal, it can look for another agent
that can provide the same service. Wooldrige describes this proactivity as the
ability to “take the initiative” [7].
• Social skill: Smart agents are able to interact with other smart agents. Social
ability involves more than simply passing data between different entities, and
requires the ability to negotiate and interact cooperatively. This is usually
performed using an agent communication language (ACL), which allows agents
to chat rather than simply passing data.

17.3.2 Multi-Agents in the Electricity Sector

Container architecture is proposed in which the participating agents with their

different skills are configured in them. Participating agents are grouped into crews
to integrate the protection system (Fig. 17.4). Within the JADE platform, there is a
flat hierarchy in which each agent communicates with the others to achieve good
coordination.
The assigned neighborhood for each agent is an electric system component, for
example, a source, load, or protection system. Any agent can take a request from
another agent within their neighborhood if it does not have the ability to make any
changes.
• An agent can detect and warn of any change in the power system.
• An external agent can block a request.
• More than one algorithm can be associated with an agent.
Relays can have different functions, such as measuring or relay protection, and
a relay agent may have different skills (see Fig. 17.5) depending on the protection
request.
The proposed scenario is composed of two agents: a relay agent, which runs the
relay algorithms, and a measurement agent, with the cluster of measurement signals.
Agents coordinate based on their abilities to obtain better results.
17 Microgrid Protection 447

Fig. 17.4 Proposed

architecture for agent gang

Fig. 17.5 Proposed architecture

17.3.3 Gang of Agents

A measurement agent has a detection system for registering a certain bounded

variation in measured signals. It communicates via TCP/IP with the distribution
network and another agent, and in the Jade environment, this is achieved through
agent communication language (ACL). If a violation is detected, the system can
request support or raise an alarm. For example, it ensures that the initial contribution
of the wind generator is not considered a fault and that incorrect protection action is
not generated.
A relay agent contains the relay algorithms. Its response will not only depend
on the local signal but will also be a reply to other relays or measurements received
from other points of the electrical network. An agent may send a remote trip signal
or request re-coordination if necessary.
448 A. C. Enríquez et al.

Specialist agents can be designed to handle special requests from the system.
For example, a relay agent can take remedial actions when a low voltage or
low frequency is detected, and load trip schemes can be defined for the current
conditions of operation of the MG. A reinitiating agent may reset the relay to its
initial conditions in the case of intermittency in communication.

17.3.3.1 Agents in JADE

JADE was used as a development and simulation platform for the construction
of the agents (Figs. 17.6 and 17.7). As mentioned earlier, TCP/IP is used for
communication with the power network, which is modeled in Simulink. The
measuring agent continuously monitors the network, and in the event of a change
in the micro network, it warns the relay agent to make the necessary adjustments.
The relay agent allows the relay to operate normally in the event of an abnormal
condition such as a fault and notifies the measuring agent to avoid the readout during
a fault condition.
Figure 17.7 shows an agent operation request and the different support requests,
using Foundation for Intelligent Physical Agents (FIPA) architecture. Communi-
cation between the electrical network, the measuring, and relay agents within the
JADE tool is shown.
Figure 17.8 shows how support requests are made between agents. First, a
support request is made by the measurement agent is done if a change is needed. The

Fig. 17.6 Activation of the

measurement agent in Jade

Fig. 17.7 Gang of agents in Jade

17 Microgrid Protection 449

Fig. 17.8 Support requests

between agents

Fig. 17.9 Algorithm for general operation

relay agent accepts the request for support. The exchange between the measurement
agent and the relay agent is based on the following general algorithm (Fig. 17.9).
The advantage of the use of agents in these algorithms is that they can be
replicated with the same characteristics without the need for additional code, simply
by invoking the agents necessary to enable efficient monitoring of the assigned
neighborhood. Figure 17.10 shows a group of two measuring agents and a relay,
which communicate through ASL in JADE to achieve cooperation.
450 A. C. Enríquez et al.

Fig. 17.10 Gang of two

measuring agents and a relay
agent

17.4 Implementation in JADE Architecture

Agent-oriented programming (AOP) is a relatively new software paradigm that

introduces concepts from theories related to artificial intelligence to the domain
of distributed systems. AOP essentially involves the application of models using
a collection of components called agents that are characterized by aspects such as
autonomy, proactivity, and capacity for communication [8].
This framework facilitates the development of complete agent-based applications
through an execution environment that implements the life cycle support features
required by the agents, the central logic of the agents themselves, and a rich set
17 Microgrid Protection 451

of graphic tools. Since JADE is written entirely in Java, it benefits from a wide
range of language features and third-party libraries and therefore offers numerous
programming approaches and abstractions that allow developers to build multi-agent
JADE systems with minimal experience in agent theory. Jade was initially developed
by the research and development department of Telecom Italia s.p.a., but is now
a community project and is distributed as open-source software under the LGPL
license (http://jade.tilab.com/). It is important to mention that in all aspects related
to the interoperability of the nucleus, Jade complies with the standards proposed by
FIPA [8].

17.4.1 Communication Language Used by Agents

The language used by agents involves a number of elements, including:

• The messenger.
• The receiver.
• The act of communicating.
• The content.
The agent communication language (ACL) allows the transmission of a set of
knowledge that will be expressed in a content language. The terms of the content
language that represent this knowledge will belong to a vocabulary common to the
different agents, called an ontology. More information on the FIPA-ACL ontology
is given in the Appendix. The content of a message within JADE has the following
structure:
(request
:sender (agent-identifier :m1@mydomain.com)
:receiver (agent-identifier :a1@a1domain.com)
:ontology Rele-Setting
:language FIPA-SL
:protocol fipa-request
:content
”“((action
(agent-measuring : a1@mydomain.com)
(book-Agent :service 15/06/2017
:departure 05/07/2017 ... )
))”“

17.4.2 Structure of the JADE Work Environment

The work environment in JADE is handled by containers. JADE has a main con-
tainer that has several sub-containers. The two most important of these containers
are defined in the FIPA standards for the handling of agents as follows:
452 A. C. Enríquez et al.

1. The agent management system (AMS) is the agent that supervises the JADE
platform. The AMS controls the cycle from when it is invoked until the agent
service ends. The AMS generates an ID for each agent to identify it from the
others.
2. The facilitator agent (directory facilitator, DF) is the “Yellow Pages” agent. Each
agent must register with the DF to list the service they can provide. An agent may
use the DF to seek support from agents with certain characteristics and to work
together.
In order for the agent to perform various actions and the activities registered in
the DF, it uses internal JADE functions, for example, Setup and AddBehavior. The
Setup function serves to initialize the conditions, while the AddBehavior function
serves to generate the behavior.
A simple example of an agent is given below [8]:
import jade.core.Agent;
public class HelloWorldAgent extends Agent {
protected void setup() {
// Printout a welcome message
System.out.println(”Hello World. I’m an agent!“);
}
}

To generate the executable from the agent console, the following instruction is
used:
javac -classpath <JADE-classes> HelloWorldAgent.java

and to run the agent:

java -classpath <JADE-classes>;. jade.Boot Peter:HelloWorldAgent

For example, the a1 monitoring agent used in the graphical environment is

illustrated in Fig. 17.11. The console indicates that the agent is waiting for the first
signal, as shown in Fig. 17.12. The agent code is used to communicate via both
Matlab and Simulink. Once the agent is finished, it is invoked in JADE from the
operating system console.

17.4.3 Testing and Measurement Tool

This tool allows the user to open a new host, create a host server, and receive or
send data via a client port, which in this case would be in Simulink. The testing
and measurement tool displays the resources (hardware, drivers, interfaces, etc.)
accessible from the toolboxes associated with the tool, and allows for configuration
and communication with those resources (Fig. 17.13). One of the tools in the
Instrument Control toolbox is the Query Instrument block.
The Query Instrument block (Fig. 17.14) configures and opens an interface
for an instrument, initializes it, and consults it to obtain data. Configuration and
17 Microgrid Protection 453

Fig. 17.11 Monitoring agent

Fig. 17.12 Agent a1 waiting for a message

initialization are carried out at the start of execution, and the block queries the device
during execution. The block has no input ports but has an output port corresponding
to the data received from the device [9].
454 A. C. Enríquez et al.

Fig. 17.13 Graphic interface tool

Fig. 17.14 Block for

receiving information

17.4.4 Communication Between Matlab/Simulink and JADE

JADE cannot connect directly with Matlab since it does not support parallel
operations, which are essential for MASs. MACSimJx is an extension of MACSim
(Multi-Agent Control Simulation), a simulator that allows the user to work with
Simulink to implement MASs; however, unlike JADE, it does not use the FIPA
standard [10]. FIPA was developed by the IEEE Computer Society to promote
standards for agent-based technology. To solve this problem, MACSimJx can be
used. Based on MACSim and an extension in Java, it allows the user to interconnect
with JADE to create agents that can interact with the models in Simulink [11]. One
17 Microgrid Protection 455

disadvantage of this approach is that the MACSim and the Jx extension are only for
32-bit operating systems, and it is difficult to configure them for 64 bits; in addition,
they are incompatible with the newer versions of Matlab.
A more direct approach is to take advantage of the TCP/IP protocol sockets via
which JADE and the Instrument Control Toolbox communicate in Simulink, and
this is the method proposed in this work. The authors of [12] analyzed different
ways of creating sockets with Matlab/Simulink, and in [13], JADE and PowerWorld
were also connected via sockets. To make a connection in JADE, we take advantage
of the fact that the code is Java, and the following libraries are added to agents:
import java.net.ServerSocket;
import java.net.Socket;

A power grid with distributed generation is then represented in Simulink. The

multi-agent environment was built in JADE, and in order to communicate with both
environments, it was necessary to establish a communication protocol. The TCP/IP
protocol was used to achieve this, and the model is illustrated in Fig. 17.15.
In this model, a software layer is removed by removing the interface between
JADE and Simulink, and connecting JADE directly via TCP/IP, as shown in
Fig. 17.15. Remote communication can be implemented with blocks. The TCP/IP
architecture is client-based, and client blocks allow data to be sent from Simulink
models to an application or other computers using TCP/IP. It is also possible to
send data to TCP/IP server blocks. The server block accepts data from the network
sockets using the TCP/IP protocol and the block mode. Data are received at fixed
intervals based on Simulink scheme cycles. The basic element of the block is the
S-function block, which uses a C MEX file [13].

Fig. 17.15 Proposed

communication model
456 A. C. Enríquez et al.

17.5 Test Scenarios

In this section, we discuss the application of protection schemes to MGs, in which

MAS architectures are used to make the proposals feasible. It is assumed that com-
munication between the relay and measurement agents takes place in the different
nodes that make up the MG. The latency times of these communication channels are
generally low (less than half a cycle), meaning that their application in the proposed
protection scheme is feasible. It is assumed that there are communication platforms
for remote programming that will allow us to define parameters to adjust the relays.
There are currently different packages on the market that offer this possibility.

17.5.1 Detection of Weak Infeed Conditions

Energy management systems (EMSs) must be used to control the network since the
operation of the electricity grid with unregulated sources is complicated and the
efficiency of the network is degraded by the lack of dispatch of these sources.
The operation of interconnection relays is based on the protection of unregulated
sources to avoid damage during sustained operation in unfavorable conditions. The
disconnection criterion is based on the measurement of electrical quantities that
are sensitive to an energy imbalance, such as the function of loss of synchronism,
frequency variation, and commonly for fault detection, voltage depression, and
overcurrent. Remote faults, where the GD makes only a small contribution (weak
source), can be very difficult to detect by conventional or interconnection protection
systems. The weak and sustained contribution from the network will continue to
feed the fault, maintaining the potentials, and avoiding reclosing operations to
maintain continuity of the electrical service.
Figure 17.16 shows a probable WS condition. Note that the interconnect relay
(bus C) may not detect the contribution of the WS to the remote fault F.
Due to the architecture of MAS and the relay agents, it is possible to obtain
sensitive and fast operation. Agents have particular functions that are integrated into
the MG protection scheme. Relay agents can also perform additional functions in
response to changes in demand or the topology of the electricity grid.
The proposed structure for the detection of weak sources is not affected by the
configuration of the power grid, and the affected area can be represented by means
of equivalents of the main contributions and through the WS link. Two operational
conditions are considered: a stable state and a fault condition.
Stable state. The actualization of relays is based on the topological conditions
detected by sensors. Several operating conditions can be established depending on
the load or generation trip, and the power flow through the link can change direction
based on specific coordination requirements. The relay settings will depend on the
measured current flow, which is used to identify steady state events (Fig. 17.17).
17 Microgrid Protection 457

Fig. 17.16 Weak source contribution

Fig. 17.17 Dynamic pickup settings of the relay agent

Relay coordination can be difficult due to the low contribution from the WS. The
operating criterion for the GD is therefore different from that in power grids. For
low-voltage networks, the GD must be disconnected using island detection schemes,
while for high-voltage networks, a disconnection criterion must be established based
on the utility’s network code. The relay agents must be able to be reconfigured
depending on the different topological states of the power grid. The use of an
MAS can complement these functions by communicating with neighboring agents
to ensure coordination between relays.
Failure Status. The low contribution from the WS represents a problem in terms
of loss of sensitivity of the protection system. This contribution will depend on the
458 A. C. Enríquez et al.

type of WS. In conventional generators, the contribution to the fault is sustained,

meaning that through continuous pre-fault readjustment, detection can be achieved
by varying the pickup current and re-coordination between relays. In unconventional
generators, the contribution is limited in magnitude and time, and a transferred trip
is necessary since the local relays cannot be reset.

17.5.1.1 Small Conventional Generators

The MAS performs actions based on the switch position, which is used to modify
the pickup current and verify online coordination with other relays, giving greater
sensitivity and a reduction in operating times.
In Fig. 17.18, the relay BC has no sensitivity, and source D will continue to
contribute to the fault even when the backup AB is operational. Although the
reduced source contribution from D does not mitigate damage to the equipment,
it prevents the reclosing of the BC line. The pickup current must be updated to

Fig. 17.18 Re-coordination of RDB for a fault on the line BC: (a) electrical network; (b)
coordination chart
17 Microgrid Protection 459

improve the sensitivity. When the starting current is modified, relay coordination on
the BD line is necessary when the WS is triggered, and on the DB line when the
D-load is triggered. Coordination between the relay on the BC line and the relay on
the DB line for a fault on the BC line is shown in Fig. 17.18. This coordination is
achieved only for the topological scenario described here, and in other scenarios,
the relay agent will achieve adjustment via the MAS.
In this operational scenario, the requirement is that the relay on the DB line must
operate slowly, while the relay on the BC line must be fast. To achieve this, the relay
coordination must be as shown in Fig. 17.18b. In the case where the relay on line
BC does not operate, the backup of relay AB will have, but a very long backup time
of Relay DB will have for the low fault contribution. The WS, therefore, continues
to feed the fault, preventing successful closure. The use of nonconventional curves
can reduce backup times. The information exchanged between agents is illustrated
in Fig. 17.19.

17.5.1.2 DG Generators

In electrical systems with large fault magnitudes, relays generally have adequate
levels of sensitivity and allow for proper operation by correctly discriminating
between normal operation and failure. The use of DG, wind (WP), or photovoltaic
(PV) generators will compromise protection, due to their low contributions [14, 15].
For instance, the fault current for a double-powered induction generator is only three
cycles at 3 p.u., while the failure contribution of a PV is only 1.1 p.u. A loss of
sensitivity of the protection system is very likely, and the operation of the relay on
the DB line cannot be determined for conventional DOCRs.

Fig. 17.19 Information exchanged during online coordination

460 A. C. Enríquez et al.

The proposed MAS enables the execution of a transferred trip that emulates
the relay operation on the BC line (Fig. 17.20). An integration process is used to
compute the operating time of the overcurrent relay. The coordination time interval
(CTI) is added to achieve coordination, and the transferred trip is then performed.
The relay on the DB line emulates the operation of the relay on the BC line
by measuring the MAS of the measured fault current and uses the CTI to ensure
coordination. Throughout this process, it is avoided to transfer scenarios with fixed
times, which are large since they are determined for the worst-case scenario, and
this affects the WS trigger times. When implementing the detection system, it is
necessary to consider the latency times and computational capacity.

Fig. 17.20 Overcurrent relay using MAS: (a) electrical system, (b) transfer trip signal
17 Microgrid Protection 461

17.5.2 Detection of Islanding Condition

Island operation results when the MG does not have the capacity to feed the loads in
the distribution network without the utility supply. This results in a loss of network
voltage support, and reconnection could result from synchronism, causing damage
to the RAS and loads; finally, the island results in an isolated system. The MG
should therefore be disconnected as soon as possible immediately after the loss of
the utility [16]. The IEEE 929-1988 standard specifies the disconnection of the DG
once it is on the island, while the IEEE 1547-2003 standard stipulates a maximum
delay of 2 s for the detection of non-intentional island operation [17, 18]. Due to
the aforementioned factors, there are various works in the literature dedicated to the
detection of electric island operation.
The development of anti-island protection systems has been widely reported,
but only frequency and voltage variation approaches have been implemented
using relays with explicit functions for the separation of the MG (Table 17.1).
Disconnection of the island represents losses both for the users and for the owners
of the MG.
The solution to the problems of island operation is based on finding a balance
between generation and load, with a range of applications that are limited by the
capabilities of the MG but that considerably reduces the number of cases in which
the system is completely lost. Island operation can be addressed by keeping the
system running until reconnection to the network is possible while giving priority to
the supply of energy in ranges of acceptable quality. The period of island operation
can be reduced to a few minutes due to the conditions of the primary source and the
variation in solar radiation and wind speed.

17.5.2.1 Detection Schemes

The main objectives of anti-island protection are to detect an island condition and
to actuate the switch at the point of common coupling (PCC) between the MG and
the distribution network. The aims are to avoid damage to the system isolated by the
operation due to frequency values that exceed the limits of the system, to allow for
restoration without complications from the supply of the main network, and to avoid
reconnection arising from synchronism of the two systems, which can cause severe
damage to the electrical network, the RAS and the connected loads. In general,
island detection methods are classified into two main groups (Fig. 17.21).
Firstly, remote methods are based on the use of a communication channel
between the main network switches and generators. Local methods are classified
into active and passive approaches. The former interact directly with the operation
of the power system by injecting disturbances that will result in a significant change
in the system parameters when the MG is isolated, but an insignificant change
when connected to the network. The method used in [33] involves a frequency shift
algorithm in sliding mode, which injects a small disturbance into the phase shift
462 A. C. Enríquez et al.

Table 17.1 Most effective anti-island protection methods

Classification Title Year Method
Passive Frequency change over power 2001 Based on the effect of the
[19] frequency change rate on load
power
[20] Voltage change rate and power 2001 Monitoring of instantaneous
factor voltage and current signals
measured
[17] Voltage imbalance and THD in 2004 Monitoring of the voltage
the current balance between the phases of
the load connected to the DG
and the harmonic current of the
network
[21] ROCOF relay performance for 2005 Proposes the concepts of
integrated generation detection time against power
applications imbalance curve
[22] Over/low-frequency relays 2006 Frequency estimation for
protection of generator detection
on island and disconnection
based on the frequency
[23] Transient signals: (i) design and 20102012 Extraction of the energy
implementation; coefficients of transient signals
(ii) performance evaluation of V and C using a wavelet
transform
[24] Model similar to Thevenin’s 2015 Equivalent linear network model
method of the system based on local
measurements, voltage, and
current phasors at the
fundamental frequency
[25] ROCOF y THD 2015 Uses the rate of change of
frequency and harmonic content
of the equivalent reactance seen
in the location of the DG
[26] Relay ROCOF active 2017 The performance of the existing
relay is evaluated and
modifications are suggested in
relation to the NDZ
[27] WTP in negative voltage 2018 Analysis of the negative
sequence sequence component in the PCC
using a wavelet transform
package (WTP)
Active Automatic phase shift method 2003 Based on the phase displacement
[28] of the inverter’s output
sinusoidal current
(continued)
17 Microgrid Protection 463

Table 17.1 (continued)

Classification Title Year Method
[29] SFS island detection method 2011 Analyzes the impact of the active
power frequency dependence of
the load on the performance of
the FSS method
[30] Scheme based on an adaptable 2016 Parameters are estimated
threshold adaptively according to the level
of operational penetration
Local hybrid Power line signaling-based 2007 Estimation of voltage imbalance
[31] technique for anti-islanding given by sequences 1 and 2 and
protection frequency
[32] Economic dispatch application of 2016 Uses a two-step genetic
power system algorithm; a load-based scheme
based on priorities is proposed

Fig. 17.21 Classification of Island

island detection methods Detection

Remote Local

Active

Passive

Hybrids

mode. The authors of [34] used a method of active frequency deviation, which works
by adding zero amplitude for a short period in the waveform of the inverter-based
output current.
Passive local methods operate based on the monitoring of signals in the terminals
of the MG tp detect island formation without neglecting over/low-frequency
conditions. The ROCOF method is presented [22], which uses a relay to monitor
the change in frequency with respect to time.

17.5.2.2 Implementation

The model implemented in this work is illustrated in Fig. 17.22. The system includes
a wind farm of 20 wind turbines, which add a nominal generation power of 4 MW.
The wind farm is connected to the electricity grid at the distribution level via a
small ring network that becomes part of the distribution system. The values of the
464 A. C. Enríquez et al.

Fig. 17.22 Model of the MG system

Fig. 17.23 Principle of

ROCOF anti-island protection

line impedances are typical distribution values with sections 1 km long, the losses in
the micro network are considered. The load blocks are distributed to different areas
using four buses.
Frequency change rate method (ROCOF). The basic principle is the monitor-
ing either of the frequency at the terminals of the MG or the local loads. Figure
17.23 shows the equivalent circuit for a synchronous generator equipped with a
ROCOF relay, operating in parallel with the distribution network. The generator
feeds a local load, and if the system is disconnected by a fault, the MG is isolated
using the CB switch and a power imbalance is generated due to a loss of power of
the main network. This imbalance causes the system frequency to change, and the
rate of frequency change df/dt can be used for island detection.
The ROCOF relay calculates the rate of frequency change using a measurement
window of a few cycles on the nodal voltage wave (typically between 2 and
40 cycles). The signal is processed using low pass filters, and the resulting signal
is used to detect the island. If the rate of change of the frequency is greater than
an adjustment threshold, a trip signal is immediately sent to the generator switch.
Typical ROCOF settings installed in 60 Hz systems are between 0.1 and 1.2 Hz/s.
The input parameter is the estimated system frequency, where the effective rate of
frequency change used by the relay is calculated based on the average value over
five cycles using the equation:
17 Microgrid Protection 465

Fig. 17.24 Operation of the ROCOF anti-island protection scheme

1  fi
5
df
= (17.1)
dt 5 ti
i=1

The results can be seen in Fig. 17.24. The system is simulated for 60 s, and the
switch that disconnects the MG from the main system is activated in the second 30 s.
The procedure used by the programmed ROCOF scheme is as follows:
1. The signal from the voltage phasor is continuously processed.
2. The time frequency of the signal is estimated using a frequency estimation
algorithm that performs a discretization of the waveform of the original signal
and uses convolution of two orthogonal components, the sine, and cosine.
3. Once the frequency is estimated, it is processed again using the ROCOF algo-
rithm, which calculates the frequency change for every five values of the signal
using Eq. 17.1. If the disturbance exceeds a limiting value for the adjustment,
the magnitude of the voltage is verified. There will be a considerable dejection to
shoot by the island. If the frequency disturbance is due to a fault or the starting
of a generator, the trip signal will be blocked.
Voltage vector phase shift method. The vector shift relay is a passive local
island detection technique that can detect an island operating condition by monitor-
ing the voltage signal at the MG terminals. The principle of operation is based on
the phase shift in the voltage signal monitored by the relay in relation to a reference
466 A. C. Enríquez et al.

Fig. 17.25 Voltage drop arising from the current demanded by the local load

Fig. 17.26 Internal and terminal voltage before and after opening the breaker

signal. This phase shift can constantly give slight increases or decreases due to the
variable behavior of the loads since it is directly proportional to the voltage drop
caused by the current demanded by the local load flowing from the MG. However,
the island condition causes a change in the phase. Figure 17.25 shows a synchronous
generator operating in parallel with a distribution network. It can be seen that a
voltage drop ΔV arises between the voltage at the terminal VT and the internal
voltage of the generator EI , due to the current demanded by the local ISG load
passing through the reactor reactance Xd .
This voltage drop causes a displacement in the phase of the voltage difference
between the terminal and reference voltages. This can be seen in the phasor diagram
shown in Fig. 17.26a. If the CB switch is opened due to a disturbance, the system
consisting of the generator and the load will be isolated; at that moment, the
generator will feed a greater or lesser load, since the current flow ISYS to or from
the network is abruptly interrupted. Consequently, the angular difference between
VT and EI is suddenly increased or decreased, and the phasor voltage at the terminals
changes direction, as shown in Fig. 17.26b.
When this phenomenon is analyzed in the time domain, the instantaneous value
of the voltage at the terminals jumps to another value, and the position of the phase
changes, as seen in Fig. 17.26. Point A indicates the instant of network loss, and it
17 Microgrid Protection 467

can be seen that the cycle time also changes. The voltage vector offset relay is based
on this phenomenon.
The relays, which are designed to provide anti-island protection based on this
phenomenon, monitor the voltage waveform at the generator terminals and make
incremental measurements of the duration of two complete cycles. The duration of
both cycles will have to be equal or less on the island condition. This variation in
the cycle time results in a proportional variation in the voltage angle at terminals
Δθ, and this final value forms the input parameter to the relay. A predetermined
adjustment threshold value in the range 2◦ to 20◦ is used as a reference signal, and
if the measurement at each cycle exceeds this value, a trip signal is sent directly to
the MG switch.
In the same way, the voltage vector shift trigger algorithm is programmed using
the simulated system signals. This algorithm works with the same voltage input
signal. The results of this scheme show the correct operation in Fig. 17.27. The
procedure used by the VSR scheme is as follows:
1. The signal from the voltage phasor is continuously processed.
2. The phase angle of the signal over time is estimated. To do this, the scheme
measures the duration of each cycle of the signal. The time obtained for each
cycle is compared with the duration of the previous cycle, and the difference
is directly proportional to an offset angle. This offset angle forms the input to
the relay that processes the operation algorithm based on the change in voltage
vector.
3. The offset angle obtained in this way is compared with an adjustment value,
and when a disturbance occurs, this value exceeds a certain limit. If the signal
indicates island operation, the magnitude of the voltage at the terminals of the
load will undergo a considerable change, which is the safety condition and the
trip of the switches is sent. If the disturbance in the voltage phase is due to a
fault or the starting of a generator, the trip signal will be blocked by the voltage
verification.

17.5.3 Coordination of DOCRs

One of the problems to consider in the implementation of an MG [35] is the design

of the protection Scheme [36]. The scheme must be able to meet the basic protection
requirements such as selectivity, sensitivity, and reliability, both while connected to
the main network and in island operation mode [37]. The protection scheme must
be adaptable [38], should take into account the operational modes of the MG, and
should provide an online coordination system that also incorporates the necessary
configurations of the protection relays depending on load variations, penetration
of renewable generation [39], topology changes and WSs, resulting in better fault
detection, reduction of false trips and shorter operating times.
468 A. C. Enríquez et al.

Fig. 17.27 Operation of the VSR anti-island protection scheme

A protection scheme based on DOCRs [40] is proposed. DOCRs are used to

protect the lines during the operation connected to the main network and in island
mode. The relay coordination problem is formulated as a non-linear programming
problem [41]. From the characteristic inverse time equation, standardized by
the IEEE, it is sought to provide protection with greater flexibility, the use of
nonconventional inverse time curves is presented [42], considering each adjustment
of the equation as a degree of freedom. The optimal configuration of the relays is
determined using the GWO [43] and DE [44] algorithms. DOCRs coordinate with
each other to ensure the selectivity and reliability of the protection scheme.

17.5.3.1 Optimization Problem

A. Objective function. The objectives of the coordination problem are to minimize

the operation time of each relay and ensure compliance with the coordination time
interval between the primary relay and its backup.
17 Microgrid Protection 469

⎧ ⎛⎞ ⎛ ⎞ ⎫
⎪
⎪ NCP
#
NCP
# ⎪
⎪
⎪
⎪ ⎜ ⎪
⎪
⎪ ⎜ tp ⎟
⎟
⎜
⎜ tb ⎟
⎟ ⎛ ⎞⎪⎪
⎪
⎪
⎨ ⎜p=1⎟ ⎜b=1⎟ NCP ⎪
⎬
NV ⎜ ⎟ ⎜ ⎟ #
OF = min + ω1 ⎜ ⎟ + ω2 ⎜ ⎟ + ω3 ⎝ CT I pb ⎠
⎪
⎪ N CP ⎜ NCP ⎟ ⎜ NCP ⎟ ⎪
⎪
⎪
⎪ ⎜ ⎟ ⎜ ⎟ pb = 1 ⎪
⎪
⎪
⎪ ⎝ ⎠ ⎝ ⎠ ⎪
⎪
⎪
⎩ ⎪
⎭
(17.2)

where NCP is the number of coordination pairs present in the system, NV is the
number of coordination pairs violated, tp is the operating time of the primary, tb is
the operating time of the backup relay, and CTIpb is the coordination time interval
(tb − tp ≥ CTI). The variables ω1, ω2, and ω3 are the Pareto front weights, which
are obtained from variations in Ipickup and Isc .
B. Time function. The operating time for each relay is defined with noncon-
ventional curves (NCCs), taking as a reference to the curves of the IEEE standard
[45]:

 
A
T p, b (s) =  p + B ∗ T DS (17.3)
I sc
P SM∗I load −1

where TDS is the time dial setting, Isc is the coordination current, PSM is the plug
setting multiplier, Iload is the load current, and A, B, p are characteristic curve
parameters.
The time curve model considers each coefficient of the vector as a degree of
freedom to obtain the generalized formulation of the overcurrent relay, and the
design obtained in this way is different for each relay. The resulting adjustment,
therefore, increases the possibility of complying with the restrictions of the relay.
This implementation allows us to obtain greater flexibility for the generation of the
time curves.
C. Coordination constraint. The use of conventional curves (CCs) is initially
used to find the solution, and standardized IEEE (VI) curve parameters are used.
The inequality restrictions for this case are given by:

P SM min ≤ P SM ≤ P SM max (17.4)

T DS min ≤ T DS ≤ T DS max (17.5)

470 A. C. Enríquez et al.

Table 17.2 Time curve Parameter Minimum range Maximum range

parameter intervals
TDS 0.5 3
PMS 1.2 1.6
A 0.05 28.2
B 0.11 0.49
p 0.02 2.00

NCCs are also used; in this case, the parameters A, B, and p are variable and
begin to be part of the inequality restrictions, defining the minimum and maximum
ranges that will be assigned to the relay.

Amin ≤ A ≤ Amax (17.6)

Bmin ≤ B ≤ Bmax (17.7)

pmin ≤ p ≤ pmax (17.8)

These variables determine the shape of the time curve. Based on its variability
(from moderately inverse to extremely inverse), it is possible to obtain a wide
diversity of time curves with the allowed parametric variation. Table 17.2 shows
the intervals considered for each of the parameters that make up the time function
of the relay.
To achieve good coordination, the primary relay must operate after a fault occurs
in its protected zone, but a sufficient time delay is necessary in order to ensure
backup for the adjacent line in the forward direction when the primary relay does
not operate. The coordination must be evaluated for a minimal fault to line-end
because crossing curves can be present. Hence, coordination must be achieved in
both fault locations.
D. Sensitivity analysis. A sensitivity filter detects coordination pairs that cannot
be coordinated. The backup relay must be able to detect minimal faults at the remote
end protected by the primary relay. The sensitivity is determined as follows:

I sc2Ø
Sb = ¿1.5 (17.9)
P SM ∗ Iload

where Isc2φ is the minimum fault detected by the backup relay, and PSM * Iload is
the pickup current of the backup relay. A lower limit of 1.5 is considered as among
the current multiples of 1 to 1.5, the times resulting from the operation of the relays
are large, and are thus useless for protection purposes.
17 Microgrid Protection 471

Fig. 17.28 Classification of wolves based on their distance from the prey

17.5.3.2 Optimization Algorithms

Gray wolf optimizer. The GWO is a bio-inspired algorithm that imitates the
hunting mechanism of gray wolves. It is used in various engineering problems [43].
The algorithm achieves optimization in three stages: (1) hunting; (2) surrounding;
and (3) attacking the prey.
(1) Hunting process. The hunting process is led by the alpha wolf, while
the other wolves (e.g., beta and delta wolves) help by recognizing the prey and
surrounding it (Fig. 17.28). In an optimization problem, a knowledge of the location
of the dam is established like the best solution of the wolves. Other wolves (search
agents) will carry out an exploration in the solution plane and are able to obtain
better solutions. The best positions are updated to give the best solution. Each wolf
contains the possible solution for the relay coordination.
(2) Surrounding the prey. The final positions of the wolves X α, β, δ define
−
→ − → −
→
the estimated area of the prey. Wolves X α, X β and X δ then surround the prey.
Surrounding is modeled mathematically as follows:

−
→ − →
→ −
→ −
D α,β,δ =  C 1,2,3 ∗ X α,β,δ − X  (17.10)

−
→ −
→ −
→ −
→
X 1,2,3 (t) = X α,β,δ − A 1,2,3 ∗ D α,β,δ (17.11)

−
→ −
→
where X α,β,δ are the positions of the gray wolves, X is the prey, t is the current
−
→ −
→
iteration, and the coefficient vectors A and C are used to avoid falling into local
minima and to define the surrounding of the prey.

−
→ −
→ → − → −
→
A = 2−
→
a ∗ r1 − −
a , C = 2 ∗ r2 (17.12)

The hunting strategy is implemented using variables r1 and r2 (Fig. 17.29).

(3) Attacking the prey. If | A > 1 | and C > 1, GWO executes a search, while if
| A < 1 | and C < 1, GWO attacks the prey. Gray wolves end a hunt by attacking the
prey when it stops moving (Fig. 17.30).
472 A. C. Enríquez et al.

Fig. 17.29 Updating of positions in GWO

Fig. 17.30 Attacking prey

A specific gene mutation (SGM) can be used to improve the exploration in the
search space. Better search solutions and a reduction in the algorithm execution time
are obtained (Fig. 17.31).
(4) Updating of the wolves’ positions. The best candidate solution is obtained
based on the distance from the best solutions:

−→ −→ −→
−
→ X1 (t) + X2 (t) + X3 (t)
X (t + 1) = (17.13)
3
17 Microgrid Protection 473

Fig. 17.31 Flow chart for gray wolf optimization

The stop criterion is the maximum defined iteration and indicates that the wolves
have converged and attacked the prey.
Differential evolution algorithm. The DE algorithm is a type of evolutionary
algorithm and is a search-based algorithm based on natural evolution and gene
selection. The DE uses genetic operators in a similar way to other evolutionary
algorithms (Fig. 17.32); however, unlike traditional evolutionary algorithms, it
randomly disturbs the members of the current generation population. There is
therefore an additional distribution of probability in terms of generating offspring.
This feature means that the algorithm involves fewer mathematical operations and
consequently has a shorter computational runtime than comparable algorithms.
(1) Initial population. An initial population is created in which the genes of all
individuals are initialized to some number in a range that is feasible with respect to
474 A. C. Enríquez et al.

Fig. 17.32 Main stages of the DE algorithm

Fig. 17.33 Mutation in DE

the relay settings. Each row represents an individual, and each column represents a
gene/variable/relay setting. The population size is (NP, D*NR) where NP represents
the number of individuals, D the number of control variables, and NR the number
of relays. For example, in a system with 12 relays and two degrees of freedom (dial
and k), the population size for 60 individuals will be (60,24). The initial population
is represented as follows:

⎡ ⎤
dial (1,1) · · · dial (1,N R) k(1,N R+1) · · · k(1,N R∗2)
⎢ .. .. .. .. .. .. ⎥
P =⎣ . . . . . . ⎦
dial (N P ,1) · · · dial (N P ,N R) k(N P ,N R+1) · · · k(N P ,N R∗2)
(17.14)

(2) Mutation. The mutation operator is based on the arithmetic difference

between pairs of randomly selected vectors (Fig. 17.33). A mutant vector is
calculated as the scaled difference between two vectors (r1 , r2 ) of three randomly
selected vectors, added to the value of the third vector (r0), which is called the base
vector. F is a user-defined parameter.
17 Microgrid Protection 475

Fig. 17.34 Crossing in DE

−
→ −
→ −
→ −
→ 
V i,g = X r 0,g +F Xr 1,g − Xr 2,g (17.15)

r0 = r1 = r2 = I (17.16)

0<F >2 (17.17)

(3) Crossing. To enrich the diversity of the population, the crossover operation
is applied after generating the mutant vector by the middle of the mutation. The
−
→
mutant vector exchanges its components with the parent vector X i,g to form a child
vector. Crossing is carried out for each variable D if a randomly generated number
in the range [0, 1] is less than or equal to the Cr value. In this case, the number
of parameters inherited from the mutant vector has an almost binomial distribution.
This scheme can be expressed as follows:

  
Vj,i,G if randi,j [0, 1] ≤ Cr or j = jrand
Uj,i,G = (17.18)
Xj,i,G otherwise

where rand i, j (0,1) is a uniformly distributed random number calculated for the
j-th component of the i-th individual. j rand ∈rand [1, D] is a random position of
−
→ −
→
the vector, which ensures that U i,g obtains at least one component of V i,g . This
operation is performed once for each individual in each generation (Fig. 17.34).
476 A. C. Enríquez et al.

(4) Selection. To keep the population size constant in subsequent generations, the
next step of the algorithm carries out a selection process to determine if the parent
vector or the child vector survives to the next generation, referred to as G = G + 1.
The selection operation can be represented as:
⎧− → 
− −
→ 
−
→ ⎨→U i,G if f U i,G ≤ f X i,G
X i,G+1 = −→ → 
− −
→  (17.19)
⎩ X i,G if f U i,G > f X i,G

→
−
where f ( X is the ability to minimize the function. Hence, if the new child vector
has an aptitude function value that is equal to or less than the parent vector,
it replaces the parent vector in the next generation; otherwise, the parent vector
persists in the population. The population may therefore improve (with respect to
the minimization of the fitness function) or retain the same fitness status, but does
not get worse. A flow chart for this process is shown in Fig. 17.35.

17.5.3.3 On-Line Implementation

The optimization algorithms selected for online execution are sufficiently robust
to achieve coordination. The adjustment update scheme must be implemented in
centralized mode, with continuous real-time measurement processing and moni-
toring of the system topology. Updating settings almost in real time before any
topological change must take place, and updating based on variation in demand
should be carried out at intervals of no more than 15 min, a similar length to the
intervals used to calculate the demand. The latency of the communication available
in the power grid must be considered. Ipickup is updated in a stable state within the
time interval considered, and the Isc is updated in the case of a topological change or
connection of a generation source. The use of phasor measurement is recommended,
for example, μPMU and the amount of calculation can be greatly reduced.

17.5.3.4 Test Systems

The proposed test system shown below allows us to study and simulate the behavior
of the protection system in an MG, and the performance of the algorithm is analyzed
in terms of the coordination between relays.
The system consists of 13 phase relays with a total of 12 coordination pairs (Fig.
17.36). The number of relays is obtained after excluding the relay protecting the
transformers. The relay names consist of two digits, the first of which represents the
nearby bus and the second represents the remote bus. The voltages are adjusted to
34.5 kV.
17 Microgrid Protection 477

Fig. 17.35 Optimized flow chart for differential evolution

The termination criterion is 1000 iterations and the system is simulated with 60
individuals. The maximum faults at the near end and the minimum faults at the far
end were obtained by opening the remote end.
478 A. C. Enríquez et al.

Fig. 17.36 Test system (microgrid)

17.5.3.5 Results and Discussion

The GWO with 60 individuals and DE with 100 individuals were used to solve the
coordination problem. Twenty algorithm repetitions of 1000 iterations each were
carried out to obtain the results. The results were evaluated using the standard
deviation (Std) to indicate the similarity of the results at each execution and to
analyze the execution time as an indication of the viability of the coordination for
online applications.
Test case I: An MG connected to the main network and operating in island
mode was simulated. The maximum generation conditions were considered and a
conventional IEEE very inverse (VI) curve is used. Based on the settings (dial, fk)
established in the algorithm, the primary time (tp) and the backup time (tb) could be
obtained. We also aimed to establish correct coordination between the times of each
relay, accomplish the time interval (CTI). The results are shown in Figs. 17.37 and
17.38.
The contribution of Isc of the utility to the MG (Fig. 17.37) means that it is not
possible that the system is properly coordinated with conventional curves. There are
17 Microgrid Protection 479

Fig. 17.37 Settings and parameters obtained (microgrid connected to the utility)

Fig. 17.38 Adjustments and parameters obtained (operating in island mode)

at least two instances of violation or loss of coordination. From Fig. 17.38, with
the MG operating in island mode, it can be seen that the value of Isc decreases
considerably, and therefore the loss of coordination will be more frequent. In this
case, four coordination pairs are affected. Correct operation and loss of coordination
are shown in Figs. 17.39 and 17.40, respectively.
Test case II: Nonconventional time curves are used with the aim of ensuring
coordination between all devices in the network. In these simulations, we consider
both an MG connected to the main network and operating in island mode. The
results show no loss of coordination (Figs. 17.41 and 17.42).
With the use of nonconventional curves (NCCs), correct coordination between
all the elements of the MG is guaranteed. Coordination between relays R12-R31
and R31-R53 in both directions is achieved. The coordination is different for each
480 A. C. Enríquez et al.

Fig. 17.39 Pair of correctly coordinated relays

mode of operation of MG, that is, connected (Fig. 17.43) and island modes (Fig.
17.44).
Figure 17.45 shows the performance of the algorithm using both conventional
curves (CCs) and NCCs. The advantages of NCCs can clearly be seen.
Test case III: Comparison between GWO and DE, using nonconventional time
curves. Simulations were carried out with the two algorithms in parallel, in order to
check their robustness in terms of solving the protection coordination problem for
both mesh and MG networks. Figure 17.46 shows the fitness of the GWO and DE
algorithms, and the average values for each algorithm are presented in Table 17.3.
17 Microgrid Protection 481

Fig. 17.40 Pair with loss of coordination

17.6 Conclusions

The dynamic conditions of the MG mean that new strategies are required to enable
the electricity network to remain reliable and to deliver energy to users. In this work,
the use of interconnection relays at the base of MAS is proposed to support the
reconfiguration of the relays, making them more sensitive to the changes in current
that is observed in the energy inputs from sources.
482 A. C. Enríquez et al.

Fig. 17.41 Microgrid connected to the utility

Fig. 17.42 Microgrid operating in island mode

17 Microgrid Protection 483

Fig. 17.43 Coordinated pair (connected to utility)

The proposed MAS is feasible for use in detecting the power supply conditions of
a WS through the online settings of the relays, considering topological changes and
operations due to faults. The relay settings were actualized through the continuous
detection of the switch positions and measurement of the demand current. WS
detection was achieved using a different strategy depending on the type of source,
for small conventional generators and DG. The performance of the protective relays
for the detection of a WS condition was improved, and an increase in sensitivity and
better coordination were obtained.
Non-standardized inverse time curves are used for MG protection, allowing for a
reduction in the operating times of each relay and providing the necessary flexibility
that the time is within preset ranges for the particular problem to be analyzed or
represented in the electricity grid. The results obtained from the test systems were
acceptable, as they were within predefined intervals with zero violations.
Fig. 17.44 Coordinated pair (island mode)

Fig. 17.45 Performance of the type of time curve

17 Microgrid Protection 485

Fig. 17.46 Performance of algorithms: (a) GWO and (b) DE

Table 17.3 Average values of the results for each algorithm

Average GWO Average DE
Tp Tb CTI Fit NV Tp Tb CTI Fit NV
0.552 0.866 0.313 17.302 0 0.493 0.841 0.347 13.489 0

References

1. Ferrari, J. P. (2005). Sistemas de Control Distribuido. Universidad Nacional de Rosario.

2. K. H. Kato T., S. Y. and T. Funabashi, “Multi-agent based control and protection of
power distributed system-protection scheme with simplified information utilization,” IEEE
Proceedings of the 13th International Conference on Intelligent Systems Application to Power
Systems, pp. 49-54. November 2005.
3. Lanemen, J., Tse, D., & Wornell, G. (2004). Cooperative diversity in wireless networks:
Efficient protocols and outage behavior. IEEE Transactions on Information Theory, 50(2),
3062–3080.
4. Sendonaris, A., Aazhang, B., & Erkip, E. (2003). User cooperation diversity, part I: System
description. IEEE Transactions on Communications, 51(11), 1927–1938.
5. Lotfi-fard, S., Faiz, J., & Iravani, R. (April 2007). Improved overcurrent protection using
symmetrical components. IEEE Transactions on Power Delivery, 22(2), 843–850.
6. McArthur, S. D. J., Davidson, E. M., Catterson, V. M., Dimeas, A. L., Hatziargyriou, N. D.,
Ponci, F., & Funabashi, T. (2007). Multi-agent systems for power engineering applications-part
ii: Technologies, standards, and tools for building multi-agent systems. IEEE Transactions on
Power Systems, 22(4), 1753–1759.
7. Wooldridge, M. (2009). An introduction to multiagent systems. John Wiley & Sons.
8. Bellifemine, F., Caire, G., & Greenwood, D. (2007). Developing Multi-Agent Systems with
JADE. Hoboken, NJ: Wiley series in agent Technology.
9. Patel, R., Bhatti, T. S., & Kothari, D. P. (2002). MATLAB/Simulink-based transient stability
analysis of a multimachine power system. International Journal of Electrical Engineering
Education, 39(4), 320–336.
10. Robinson, C. R., Mendham, P., & Clarke, T. (2010). MACSimJX: A tool for enabling agent
modelling with simulink using JADE. Journal of Physical Agents, 4(3).
486 A. C. Enríquez et al.

11. Silva, A. C. R. D., & Grout, I. A. (2011). MS2SV: Tradução de modelos em Simulink para
VHDL-AMS. IEEE Latin America Transactions.
12. MathWorks, Instrument Control Toolbox 2.7, Mathworks Inc., Natick, USA, 2008.
13. MathWorks, Simulink Simulation and Model-Based Design, MathWorks, Inc., Natick, MA,
2004.
14. J. Barsch, G. Bartok, G. BenmouyaI, O. Bolado, B. Boysen, S. Brahma and J. Burnworth,
“Fault current contributions from wind plants,” Report to the T&D Committee, IEEE PES
Electric Machinery Committee and Power System Relaying Committee, 2012.
15. J. Keller, B. Kroposki, R. Bravo and S. Robles, “Fault current contribution from single-phase
PV inverters,” IEEE 37th Photovoltaic Specialists Conference (PVSC), pp. 1822-1826, June
2011.
16. S. Chowdhury, C. F. Ten and P. Crossley, “Islanding protection of distribution systems with
distributed generators: A comprehensive survey report,” IEEE Power and Energy Society
General Meeting, pp. 1–8, IEEE, 2008.
17. Jang, S. I., & Kim, K. H. (2004). An islanding detection method for distributed generations
using voltage unbalance and total harmonic distortion of current. IEEE Transactions on Power
Delivery, 19, 745–752.
18. IEEE 929-2000 - IEEE Recommended Practice for Utility Interface of Photovoltaic (PV)
Systems.
19. Pai, F.-S., & Huang, S.-J. (2001). A detection algorithm for islanding-prevention of dispersed
consumer-owned storage and generating units. IEEE Transactions on Energy Conversion,
16(4), 346–351.
20. Salman, S. (2001). New loss of mains detection algorithm for embedded generation using
rate of change of voltage and changes in power factors. In 7th international conference on
developments in power systems protection (DPSP2001) (pp. 82–85).
21. Affonso, C., Freitas, W., Xu, W., & daSilva, L. (2005). Performance of ROCOF relays for
embedded generation applications. IEE Proceedings—Generation, Transmission and Distribu-
tion, 152(1), 109.
22. Vieira, J., Freitas, W., Xu, W., & Morelato, A. (2006). Efficient coordination of ROCOF and
frequency relays for distributed generation protection by using the application region. IEEE
Transactions on Power Delivery, 21, 1878–1884.
23. Lidula, N. W. A., & Rajapakse, A. D. (2012). A pattern recognition approach for detecting
power islands using transient signals part II: Performance evaluation. IEEE Transactions on
Power Delivery, 27, 1071–1080.
24. Fusco, G., DiFazio, A. R., & Russo, M. (2015). Islanding detection method based on a
Thevenin-like model. IET Generation, Transmission & Distribution, 9, 1747–1754.
25. Merino, J., Mendoza-Araya, P., Venkataramanan, G., & Baysal, M. (2015). Islanding detection
in microgrids using harmonic signatures. IEEE Transactions on Power Delivery, 30, 2102–
2109.
26. Gupta, P., Bhatia, R. S., & Jain, D. K. (2017). Active ROCOF relay for islanding detection.
IEEE Transactions on Power Delivery, 32, 420–429.
27. Gupta, N., & Garg, R. (2018). Algorithm for islanding detection in photovoltaic generator
network connected to low-voltage grid. IET Generation, Transmission & Distribution, 12,
2280–2287.
28. Holdsworth, L., Wu, X., Ekanayake, J., & Jenkins, N. (2003). Comparison of fixed speed
and doubly fed induction wind turbines during power system disturbances. IEE Proceedings:
Generation, Transmission & Distribution, 150(3), 343.
29. L. W. Arachchige and A. Rajapakse, “A pattern recognition approach for detecting power
islands using transient signals—Part I: Design and implementation,” 2011 IEEE power and
energy society general meeting, 2011.
30. R. C. Dugan, J. A. Taylor and D. Montenegro, “Energy storage modeling for distribution
planning,” 2016 IEEE Rural Electric Power Conference (REPC), 53, pp.12–20, IEEE, 2016.
31. W. Xu, G. Zhang, C. Li, W. Wang, G. Wang and J. Kliber, “A power line signaling based
technique for anti-islanding protection of distributed generators: Part I: Scheme and analysis,”
2007 IEEE power engineering society general meeting, Jun 2007.
17 Microgrid Protection 487

32. Yan, N., Xing, Z. X., Li, W., & Zhang, B. (2016). Economic dispatch application of power
system. IEEE Transactions on Applied Superconductivity, 26(7), 1–5.
33. Smith, G., Onions, P., & Infield, D. (2000). Predicting islanding operation of grid connected
PV inverters. IEE Proceedings—Electric Power Applications, 147(1), 1.
34. Ropp, M., Begovic, M., & Rohatgi, A. (1999). Analysis and performance assessment of
the active frequency drift method of islanding prevention. IEEE Transactions on Energy
Conversion, 14(3), 810–816.
35. J. A. Ocampo-Wilches, A. J. Ustariz-Farfan and E. A. Cano-Plata, “Modeling of a centralized
microgrid protection scheme,” 2017 IEEE Workshop on Power Electronics and Power Quality
Applications (PEPQA), pp. 1-6, May 2017.
36. Ali Memon, A., & Kauhaniemi, K. (2015). A critical review of AC microgrid protection issues
and available solutions. Electric Power Systems Research, 25, 23–31.
37. A. Sharma and B. Panigrahi, “Optimal relay coordination suitable for grid-connected and
islanded operational modes of microgrid,” 2015 Annual IEEE India Conference (INDICON),
2015.
38. Awaad, M., Mekhamer, S., & Abdelaziz, Y. (2018). Design of an adaptive overcurrent pro-
tection scheme for microgrids. International Journal of Engineering, Science and Technology,
10(1), 1–12.
39. Jain, D. K., Gupta, P., & Singho, M. (November 2015). Overcurrent protection of distribution
network with distributed generation. In 2015 IEEE Innovative Smart Grid Technologies-Asia
(ISGT ASIA) (pp. 1–6).
40. Sahoo, A. K. (September 2014). Protection of microgrid through coordinated directional
over-current relays. In 2014 IEEE Global Humanitarian Technology Conference—South Asia
Satellite (GHTC-SAS) (pp. 129–134).
41. H. Zeineldin, E. El-saadany and M. Salama, “Protective relay coordination for microgrid
operation using particle swarm optimization,” 2006 Large Engineering Systems Conference
on Power Engineering, pp. 152-157, July 2006.
42. Arreola, O., Conde, A., & Trujillo, L. A. (2014). Overcurrent relay with unconventional curves
and its application in industrial power systems. Electric Power Systems Research, 110, 113–
121.
43. Mirjalili, S., Mirjalili, S. M., & Lewis, A. (2014). Grey wolf optimizer. Advances in
Engineering Software, 69, 46–61.
44. Storn, R., & Price, K. (1997). Differential evolution—A simple and efficient heuristic for global
optimization over continuous spaces. Journal of Global Optimization, 11(4), 341–359.
45. IEEE C37.112: IEEE Standard Inverse-Time Characteristic Equations for Overcurrent Relays,
2018.
Chapter 18
A New Second Central Moment-Based
Algorithm for Differential Protection in
Micro-Grids

Ernesto Vázquez, Héctor Esponda, and Manuel A. Andrade

18.1 Introduction

Although the literature presents different definitions for Smart Grid, the charac-
teristics remain the same: it can skillfully integrate the actions of all stakeholders
connected to it—generators, consumers, and those that do both—in order to deliver
sustainable, economic, and reliable electricity supplies efficiently. A reliable and
secure Smart Grid requires fast and accurate fault localization and intelligent
automation functions to take care of restoring the power to healthy parts of the
network.
Smart grids typically consist of various amounts of local generation, as well as
energy storage systems and dynamic loads (see Fig. 18.1). The increased penetration
of DG challenges conventional protection schemes because fault currents now
come from both traditional sources and inverted-driven sources. However, many
DG technologies are based on DC/AC converters, making conventional protection
schemes infeasible. Moreover, storage systems operating in discharge mode are,
from the grid point of view, similar to DG [11].
The paradigm of protection for Smart Grid should be adapted to the flexibility
inherent to the Smart Grid concept. An essential requirement is “adaptability”: a
protection system needs to adapt to the changes in the primary system.
The protection system for Smart Grids should consider the amount and type of
DG (and for some units, its capacity of short-circuit current) and the amount of
power from the primary system. If part of the Smart Grid DG is capable of supply
the load, it can be disconnected in case of a fault occurs. However, intermittency
may force the DG to be disconnected to avoid damage. Thus, protection schemes

E. Vázquez · H. Esponda · M. A. Andrade ()

Universidad Autónoma de Nuevo León, School of Mechanical and Electrical Engineering,
Monterrey, Nuevo León, Mexico

© Springer Nature Switzerland AG 2021 489

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_18
490 E. Vázquez et al.

Industrial Residential
load load
Grid
controller

Hydro Diesel
generation Solar and wind
generation generation

Fig. 18.1 Smart Grid with distributed generation components feeding power to local loads

are needed to ensure the faults that are external to the island, and in some extents
to the faulted equipment, do not cause unnecessary disconnection of the DGs and
transformers, so the Smart Grid has possibilities of remain energized.
Fault protection methods must incorporate more than one approach to coping
with the issues arising in Smart Grid protection schemes.
Adaptive protection schemes change its response according to the system dynam-
ics. The most straightforward approach is to have two sets of relay settings (for
islanded and grid-connected operating modes). Although this is a simple approach,
the protection system must cope with changes in fault current levels by shifting
through characteristic curves [17].
Adaptability also requires having a piece of prior knowledge on the configuration
of the Smart Grid to update the settings for each possible scenario.
Another approach to Smart Grid protection is the differential scheme, mostly in
scenarios with low levels of fault current. Traditional differential protection cannot
be used as a complete solution because it is better suited to detect downstream
earth faults and requires additional techniques to detect symmetrical faults and faults
without ground component [4].
The classic overcurrent protection is not well suited to cope with DG varying
fault levels and intermittency. Traditional overcurrent schemes have problems
identifying not-fault conditions as inrush currents and may require directional units.
Adaptive protection systems have evolved to wide-area schemes where supervi-
sory systems (SCADA) and IED are integrated to measure over broad areas and
18 A New Second Central Moment-Based Algorithm. . . 491

perform protective actions based on a high volume of data [20] over intensive
communication systems. Sometimes it is necessary to use distributed intelligence
in the form of multi-agent systems based on fuzzy networks, decision trees, support
vector machines, machine learning, etc. [3, 7, 11, 12, 16]. Having such independent
agents, with awareness about their component or substation states through sensor
links, permits the agents to perform various functions that are not implemented by
either the protection systems or the central control systems.
To add smartness to a power system, independent protective functions in each
component are a must [13], and these protective elements need to adapt itself to the
dynamics of the power system. The Smart Grid adaptive protection also requires
to be smart. Further sections will discuss some approaches to implement flexible,
adaptive protection systems for Smart Grid components.
In this chapter, the Smart Grid requirements for the protection system are
reviewed from the point of view of differential protection used in different elements
of electricity distribution systems such as busbar, transformer, generator, reactor,
and the impact of the high harmonic content due to the non-linear loads.
The present method is based on the analytical second central moment (SCM)
to characterize differential current’s patterns to detect faults occurring within the
differential zone. The method calculates the SCM magnitude on the base of
waveforms of a half and full sinusoidal signal to define the similitude between fault
current and any other current due to transient phenomena. The method applies a
filtering and normalization stage to eliminate redundant information and to highlight
any change in the differential current signals. On the other hand, a universal
threshold was proposed and set up to identify the faults currents. This limit is a
universal unique threshold for any element in a smart grid (busbar, transformer,
generator, reactor, capacitor bank, motors, and transmission lines), and it is based
on the maximum magnitude of the SCM that it can achieve in a half-sinusoidal
waveform.
The method offers a straightforward implementation with a low computational
cost; detecting faults faster than traditional differential protections. In the case of
transformers, the method carries on turn-to-turn short-circuits detection even during
energization cases.

18.2 Differential Protection Based on SCM

Differential protection schemes have been used for 20 years to protect generators,
transformers, motors, reactors, buses, lines, and any other equipment where inputs
and outputs current are compared. This scheme is selective, and it only responds
to faults within protected elements. Nevertheless, the leading difficulty this scheme
suffers is to distinguish between inrush current and fault currents correctly. Inrush
currents are transient phenomena arising when the transformer is being energized.
Although inrush currents have large magnitudes, they do not reflect in the sec-
ondary of the transformer. This phenomenon may contribute to the misoperation
492 E. Vázquez et al.

of differential elements because of the false differential current generated. Thus,

the reliability of the scheme could be compromised. Diverse methods—most of
them based on the harmonic content and the waveform of the signal—have been
presented in the literature to overcome this problem. Nonetheless, some conditions
as current transformers (CT) saturation, changes in the transformer’s parameters,
overexcitation, and frequency variations may impair the performance of these
methods.

18.2.1 Differential Protection Principle

Differential protection is based on Kirchhoff’s current law [21]. The principle of

operation calculates the sum of all currents flowing from the input to the output
terminals of the protected zone. The CTs delimit the protected area. Under ideal
conditions, steady-state or faults outside of the protected zone, the differential
current will be zero as is shown in Fig. 18.2a. On the other hand, if a fault occurs
inside of the protection zone as is shown in Fig. 18.2b, the differential current will
be different from zero, and the relay will send a trip signal to disconnect the power
transformer.

18.2.1.1 Percentage Differential Protection

Although differential current principle, in theory, works well, the error introduced by
the mismatch of CT ratios, tap changers, and other factors such as the power trans-
former phase-shift, may cause a false differential current that must be considered
to define the settings of the differential protection scheme. Therefore, percentage
differential protection was introduced to decrease the magnitude of differential
current in steady-state and gives a high sensibility when CTs are saturated. This
scheme adds a restraint current Irest defined as

Irest = k (|IW1 | + |IW2 |) , (18.1)

or

Irest = max (|IW1 |, |IW2 |) , (18.2)

where IW1 and IW2 are the currents flowing through the CT secondary at the high
and low voltage windings, respectively.
The restraint current is compared to the differential current Idiff defined as

Idiff = |IW1 + IW2 | . (18.3)

18 A New Second Central Moment-Based Algorithm. . . 493

Fig. 18.2 Differential Fault

protection principle. (a)
1 2
During an external fault. (b)
During an internal fault Protected
element

1 2

Relay Δ = 1 + 2 =0

a
Fault

1 2
Protected
element

1 2

Relay Δ = 1 + 2

The relay will operate if the differential current is greater than a percentage of the
restraint current defined by a slope setting (SLP) according to

Idiff > SLP × Irest . (18.4)

Graphically, the operation characteristic of the percentage differential protection

is shown in Fig. 18.3. This characteristic consists of a straight line with a slope
equal to SLP1 . The intersection point of this characteristic with the vertical axis
Iop defines the minimum pickup current of the relay Ipu min . Also, a dual-slope
characteristic (SLP2 ) can be added to increase the relay safety in conditions where
the CTs experience a heavy saturation as is shown in Fig. 18.3. The SLP values are
common expressed in percentage and is defined as
494 E. Vázquez et al.

Fig. 18.3 Relay

characteristic with double
slope

op 0

ret 0 ret 1 ret 2

Idiff
SLP = . (18.5)
Irest

Some typical values for SLP are 30% and 80% [21]. Figure 18.4 shows an
example of a differential protection performance in a power transformer during
steady-state and a fault inside of the differential protection zone, respectively. It
was used a dual-slope characteristic with the SLP1 and SLP2 set in 0.3 and 0.6,
respectively. Figure 18.4a shows the trajectory of the operation current which
is on the non-operation zone and it means the transformer is working correctly.
On the other hand, Fig. 18.4b illustrates the current operation crossed the SLP
characteristic from the non-operation to the operation zone. This behavior implies
that the transformer must be disconnected to avoid damage to the equipment.

18.2.1.2 Negative-Sequence Differential Principle

Recently, the negative-sequence differential protection (87Q) [10, 19] has been
introduced to avoid unnecessary relay tripping during external faults. This method
has an excellent performance against external faults and turn-to-turn faults where
high sensitivity is required. Using the power system as is shown in Fig. 18.5, it is
possible to explain the principle of the negative-sequence protection. This system
comprises a transformer without phase shifting and equal ratio on both sides. If a
fault occurs outside the differential protection zone, a fictitious negative-sequence
source can be placed at the fault point as shown in Fig. 18.5a. Therefore, the
negative-sequence currents in each side of the transformer will flow in the same
direction having a phase-shift of 0 degrees. On the other hand, if a fault occurs inside
the differential protection zone as is shown in Fig. 18.5b, the negative-sequence
currents will flow in opposite directions with a phase-shift of 180 degrees. These
18 A New Second Central Moment-Based Algorithm. . . 495

Fig. 18.4 Differential 1.5

protection performance. (a)
Steady-state condition. (b)
Fault inside the differential
protection zone
1

(pu))
op
0.5

0
0 1 2 3 4
ret (pu)
a

1.5

1
(pu))
op

0.5

0
0 1 2 3 4
ret (pu)
b

characteristics have been useful to implement a method to discriminate between

an internal and an external fault using the magnitude and the phase shifting of the
negative-sequence currents.
The negative-sequence method calculates the negative-sequence currents from
the primary and secondary side of the element protected. After that, both negative-
sequence currents are compared with a threshold defined as minimum pickup
Ipu min Q . If either or both are greater than this threshold, the method will compare
the phases of both negative-sequence currents. The resulting phase will place in an
α plane as shown in Fig. 18.6. If the phase falls between ±60 degrees (also called
496 E. Vázquez et al.

Fault
2S 2R
2S 2R

+
2S 2R − 2

Relay

Fault

2S 2R
2S 2R

+
2S − 2
2R

Relay

Fig. 18.5 Direction of negative-sequence currents. (a) External fault. (b) Internal fault

the relay operating angle (ROA) zone), the relay will send a trip signal to disconnect
the element. Otherwise, the event will be determined as a steady-state condition or
external fault.
To show the performance of the differential negative-sequence scheme, an
external three-phase fault, and a turn-to-turn fault at 10% of the winding short-
circuited were evaluated as are shown in Fig. 18.7. During the external fault, the
magnitude of negative-sequence current did not enter to the ROA zone. On the
contrary, at the time the internal fault occurred from a steady-state condition, the
magnitude of the negative-sequence current crossed to the ROA zone.
Traditional percentage differential protection and negative-sequence method
usually have a good performance discriminating transient events and internal faults.
However, these conventional methods may malfunction during the inrush current,
remanent flux in the core of the power transformer, CT saturation, incipient faults,
non-linear loads, etc. The next section will address the main issues the percentage
differential protection faces.
18 A New Second Central Moment-Based Algorithm. . . 497

Fig. 18.6 Negative-sequence 90

alpha plane 120 60

150 30
No trip
zone ROA

180 0
mín

210 330

240 300
270
90 90
120 60 120 60

150 30 150 30

Fault
10 20 30 40 2 3 4 5
180 0 180 0

210 330 210 330

240 300 240 300

270 270
a b

Fig. 18.7 Performance of the differential negative-sequence method. (a) External fault. (b)
Internal fault

18.2.1.3 Current Transformer Saturation

Current transformers usually are exposed to be saturated when a fault occurs. The
high levels of DC component when a fault happens, they do saturate the core of
the CTs distorting the waveform of the CT secondary current. The saturation of the
CTs introduces an error in the differential currents that may cause the misoperation
of the differential elements. The CT saturation will produce an increase in the
differential current once the fault already had occurred, even though, the fault was
498 E. Vázquez et al.

Current

Current

Current
Time Time Time

a b c

Fig. 18.8 Impact of CT saturation in different levels: primary current (solid lines) and secondary
current (dashed lines). (a) Light saturation. (b) Normal saturation. (c) Heavy saturation

outside the differential protection zone. Also, the CT saturation introduces harmonic
content in the differential signals that may cause a time delay in the operation of
the differential elements. Figure 18.8 illustrates how the CT secondary currents are
distorted in comparison with the CT primary currents when the core of the CTs is
under saturation [6]. Also, it is seen the saturation level directly will impact on the
CT secondary current distortion. Figure 18.9 shows an example of the performance
of the differential elements during a fault inside the differential protection zone. It
is seen the CT saturation caused time delay of the differential elements because of
the harmonic content.

18.2.1.4 Methods Summary

All different methods before mentioned have shown different advantages and
disadvantages depending on the power system conditions such as inrush current, CT
saturation, incipient faults, degradation of the transformer parameters, etc. Also, the
different kind of input signals used for the methods such as current, voltage, phasors,
symmetrical components, etc., may impact on the security, speed, and reliability of
the differential protection. Therefore, to bring a better understanding of what are the
events and the operation-basis of the methods, Tables 18.1 and 18.2 are introduced.

18.3 Second Central Moment Applied to Differentiate Inrush

Currents from Internal Faults

The proposed method is based on the magnitude of the second central moment
from the differential currents to differentiate internal faults from transient conditions
as well as inrush currents. Figure 18.10 shows the flow chart of the proposed
method which is divided into different stages. The method calculates the SCM
magnitude of each differential current previously normalized and filtered to identify
the kind of event. The SCM magnitude is compared to an identification threshold
based on the maximum variance that a half-sinusoidal signal could achieve. If the
18 A New Second Central Moment-Based Algorithm. . . 499

CT saturation
1
Current (A)
0.5

−0.5
0.4 0.45 0.5 0.55 0.6 0.65
Time (s)
a

12 Misoperation
10
(pu)

8
op

6
4
Impact of CT saturation
2
0
0 5 10 15 20 25 30
ret (pu)
b

Fig. 18.9 Misoperation of conventional differential protection in presence of heavy CT saturation.

(a) CT secondary current. (b) Differential relay performance

SCM magnitude is higher than the limit, the method identifies a fault condition.
Otherwise, if the SCM magnitude is lower or equal to the threshold, the method
determines a transient event or steady-state condition.
Table 18.1 presents which are the power conditions where each method correctly
operates. All different scenarios were selected from the most representative cases
that may lead to the misoperation of the differential elements. On the other hand,
Table 18.2 exposes which kind of signal is used for the method such as voltage,
current, or both of them. Also, Table 18.2 shows if the methods use empirical
thresholds to detect internal faults and if they are either online or offline versions.
As the presented method is intended to be implemented in a digital relay, it
forms the differential currents from the CT secondary currents. Accordingly, the
CT secondary currents are angle shift-compensated using a set of compensation
matrices adapted depending on the transformer vector group [1, 5]. In this case,
for the primary side, the matrix 0 was selected. For the secondary side, the matrix
 500

Table 18.1 Events where each method shows a correct operation

System condition
Non-linear Recovery Sympathetic Cross-country
Method CT saturation loads inrush inrush faults Incipient faults Degradation Over excitation
Buchholz relaya,b  
Harmonic restraint   
Harmonic cross-blocking   
Negative sequence     
Transformer parameters   
Two-terminal network 
Differential power  
PCA    
Differential current gradient   
Empirical Fourier transform   
Non-saturation zone  
Linkage-flux     
High-order statistics   
Wavelet transform     
Current/voltage ratio   
Superimposed differential     
Kalman filter  
Park’s vector    
Frequency responsea,b  
a The method does not detect inrush currents
b The method does not detect residual flux
E. Vázquez et al.
Table 18.2 Operation base and the kind of measurements each method uses to operate
Operation base
Symmetrical Superimposed
Method Phasors Instantaneous values components quantities Statistical distribution Sensors Empirical threshold
Buchholz relaya 
Harmonic restraint 
Harmonic cross-blocking 
Negative sequence  
Transformer parameters   
Two-terminal network  
Differential power  
PCA  
Differential current gradient   
Empirical Fourier transform  
Non-saturation zone   
18 A New Second Central Moment-Based Algorithm. . .

Linkage-fluxa   
High-order statistics   
Wavelet transform    
Current/voltage ratio   
Superimposed differential     
Kalman filter  
Park’s vector  
Frequency responseb   
a Method not based on voltage/current signals
b Offline method
501
502 E. Vázquez et al.

Fig. 18.10 Flow chart of the

proposed method’s algorithm Start
to identify transient events
from internal faults Data acquisition
from CTs

Data window

Differential
current

Normalization
[−1, 1]

Delta filter

Calculation
of SCM

SCM > 0.25

F
T
Trip

End

1 was chosen. A modification on the transformer vector group requires updated

compensation matrices. The angle-compensated differential currents are
⎡ ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤⎡ ⎤
Idiff A 1 0 0 Ip A 1 −1 0 Is A
⎣ Idiff B ⎦ = ⎣ 0 1 0 ⎦ ⎣ Ip B ⎦ + aT √1 ⎣ 0 1 −1 ⎦ ⎣ Is B ⎦ , (18.6)
3
Idiff C 0 0 1 Ip C −1 0 1 Is C
Idiff ABC = I∗p ABC + I∗s ABC , (18.7)

where aT is a phase compensation parameter for tap changers [9] and I∗ are the
angle-compensated CT secondary currents.
The introduced method utilizes a sliding window (see Fig. 18.11) to form a matrix
Idiff ABC ∈ M m,n (R) with the phase differential currents:
18 A New Second Central Moment-Based Algorithm. . . 503

Fig. 18.11 Example of sliding window

⎡ ⎤
Idiff A1 Idiff B1 Idiff C1
⎢ .. .. .. ⎥
⎢ . . . ⎥
⎢ ⎥
Idiff ABC = ⎢Idiff A32 Idiff B32 Idiff C32 ⎥
⎢
⎥, (18.8)
⎢ . . . ⎥
⎣ .. .. .. ⎦
Idiff A64 Idiff B64 Idiff C64

where m = 64 corresponds to 64 samples/cycle for n = 3 differential currents:

Idiff A , Idiff B , and Idiff C . The method does not need a minimum pickup value to
start.
Once the differential currents matrix is formed and organized in a sliding
window, these currents are filtered using a filter named Delta filter. This filter aims
to eliminate all irrelevant data and periodicity from differential currents to identify
variations. After applying Delta filter on differential currents, the incremental
differential currents can be represented as

Idiff ABC (i) = Idiff ABC (i) − Idiff ABC (i − nT ) , (18.9)

where i denotes the actual sample, and nT is n periods of the primary frequency.
The number of periods selected was nT = 1. An example of the application of the
504 E. Vázquez et al.

2
Current (A)
0

−2

−4
0 0.05 0.1 0.15 0.2 0.25 0.3
Time (s)
a

2
Current (A)

−2

0 0.05 0.1 0.15 0.2 0.25 0.3

Time (s)
b

Fig. 18.12 Example of Delta filter application. (a) Original signal. (b) Delta filtered signal

Delta filter to a sinusoidal signal is illustrated in Fig. 18.12. The sinusoidal signal
was sampled at 32 samples of one cycle of 60 Hz, and the number of periods chosen
was one. Figure 18.12a shows Delta filter eliminated the redundant information after
the first cycle. Besides, if any alteration occurs during the filtering as it happened in
the signal at ∼0.116 s (see Fig. 18.12b) the filtering will also highlight this change.
Details of Delta filter can be found in [2, 8].
Once incremental differential currents are calculated, these currents must be
normalized to apply the method to any power transformer independent on the
transformer parameters. The normalization will scale the signals to have amplitudes
between [−1, 1] dividing each sample by the maximum absolute value in each
window as
Idiff ABC (i)
IN (i) = , (18.10)
|max (Idiff ABC )|

where IN are the normalized differential currents, i is the actual sample, Idiff ABC
are the differential currents, and |max (Idiff ABC )| is the maximum absolute value in
each window.
18 A New Second Central Moment-Based Algorithm. . . 505

Table 18.3 Thresholds Magnitude of SMC Events

criteria
0 ≤ SCM ≤ 0.25 No fault condition
inrush current
load
SCM > 0.25 Turn-to-turn or
inter-turn-to-ground faults

Once the differential currents are formed and normalized, they serve as input to
determine the second central moment for each window data according to (18.10) as

1
n
SCM = (IN (i) − x̄(i))2 , (18.11)
n
i=1

where x̄ is the average of x obtained as

#
n
IN (i)
i=1
x̄ = , (18.12)
n
where n represents the samples per window, and x is the actual sample. For this
application n = 64; i. e., the selected sample rate was 64 samples/cycle.
A threshold value, ε = 0.25 is compared with the SCM magnitude to identify the
kind of event. The identification criteria are summarized in Table 18.3. If either one
sample of the SCM is in the interval [0, 0.25], the event is identified as a steady-state
condition or an inrush current, and the method obtains a new sample re-starting the
algorithm. Otherwise, the event is detected as an internal fault, and a trip signal is
sent. This kind of data processing does not require a pickup value to start.

18.3.1 Computational Complexity

The complexity of the second central moment algorithm was estimated employing
the online One-Pass method [18]. The total computational complexity of the
SCM method is O (N ) where N is the window’s length. Phasor-based differential
protection
  uses discrete Fourier transform (DFT) with an associated complexity of
O N 2 [15]. Further,
 the fast Fourier transform (FFT) has a computational cost of
O N log2 (N ) . For N = 64, the SCM method requires O (64) = 64 computations.
   
In contrast, the DFT and FFT need O 642 = 4096 and O 64 log2 (64) =
384 calculations, respectively. Hence, the SCM method is more computationally
efficient than the DFT and FFT. These results make the real-time implementation of
the method feasible using a sampling period based on a multiple of 60 Hz.
506 E. Vázquez et al.

18.4 Results

This section introduces the results of the differential protection based on the SCM.
The method was tested in different elements, e.g., power transformers, generators,
power reactors, and busbar. Furthermore, there were considered two different cases
focused on how the harmonic content impacts the differential relay due to the power
electronics used in renewable energies. In this section, the results are organized as
follows: each figure shows the CT secondary currents from the high and low side,
normalized differential currents, normalized incremental differential currents, and
the magnitude of the SCM for each phase.

18.4.1 Power Transformer Protection

A demonstrative example of the discrimination between transient events and

internal faults using the second central moment as a basis in a power transformer
is addressed. The example is divided into two different cases: a transformer
energization and a fault within the differential zone. The study case setup consists
of a -Y connected, 100 MVA, 230/115 kV, 60 Hz power transformer. The selected
CTs were 500 : 5 and 1000 : 5 for the high and low voltage side, respectively.
Figure 18.13 shows the power transformer energization at t = 1 s. Figure 18.13e
shows that the magnitude of the SCM remained below the 0.25 threshold level
(dashed line). Hence, the method recognized the event as a no fault event and
blocked the operation of differential elements.
For the second case, Fig. 18.14 shows the inception of a B-g single-phase fault
inside the differential protection zone on the high side at t = 0.5 s. Although the
CTs experienced saturation when the fault occurred, the SCM magnitude did cross
the identification threshold at 14.6 ms as is shown in Fig. 18.14e. Therefore, the
method correctly identified the event as an internal fault, and a trip signal must
be sent. Similar tests such as the previously presented example were conducted to
demonstrate the capability of the SCM method to differentiate transient conditions
and internal faults. Results showed that the method could be used to protect the
power transformer accurately.

18.4.2 Generator Protection

Synchronous generators (SG) are probably the most critical elements in a power
system. High repair costs and the undesirability of loss of generation are among
the main reasons to protect this element adequately. Generally, the SG is protected
with percentage differential protection relays to face internal phase-faults inside the
differential protection zone. Other schemes usually implemented for the protection
18 A New Second Central Moment-Based Algorithm. . . 507

PA PB PC
Current (A)

5
0
−5
1 1.1 1.2 1.3 1.4 1.5
a
SA SB SC
Current (A)

5
0
−5
1 1.1 1.2 1.3 1.4 1.5
b
diff A diff B diff C
1
Current (A)

0.5
0
−0.5
1 1.1 1.2 1.3 1.4 1.5
c
Δ diff A Δ diff B Δ diff C
1
Current (A)

0.5
0
−0.5
1 1.1 1.2 1.3 1.4 1.5
d
SCMA SCMB SCMC
Magnitude

0.25

0
1 1.1 1.2 1.3 1.4
Time (s)
e

Fig. 18.13 Method performance during a power transformer energization at time t = 1 s. (a–e)
Identification threshold (dashed line)
508 E. Vázquez et al.

PA PB PC
Current (A)

50
0
−50
−100
0.5 0.6 0.7 0.8 0.9 1
a
SA SB SC
5
Current (A)

−5
0.5 0.6 0.7 0.8 0.9 1
b
diff A diff B diff C
Current (A)

0.5
0
−0.5
−1
0.5 0.6 0.7 0.8 0.9 1
c
Δ diff A Δ diff B Δ diff C
1
Current (A)

0.5
0
−0.5
−1
0.5 0.6 0.7 0.8 0.9 1
d
SCMA SCMB SCMC
0.5
Magnitude

0.25

0
0.5 0.6 0.7 0.8 0.9 1
Time (s)
e

Fig. 18.14 Method performance during a fault inside the transformer differential protection zone
at time t = 0.5 s. (a–e) Identification threshold (dashed line)
18 A New Second Central Moment-Based Algorithm. . . 509

of the SGs are time-overcurrent relays, impedance relays, negative-sequence relays,

over-flux relays, loss of field relays, and stator earth-fault relays. However, these
schemes have issues detecting TTFs in the stator as well as phase-to-ground faults
adjacent to the neutral.
Accordingly, a variation of the SCM method based on differential residual
currents

3I0H = IHA + IHB + IHC (18.13)

3I0h = Iha0 + Ihb0 + Ihc0 (18.14)

is proposed for identifying turn-to-ground and turn-to-turn faults within the SG.
Two cases are studied to demonstrate the performance of the SCM-based
method. For the first case, Fig. 18.15 shows a turn-to-ground fault from steady-state
occurring at time t = 0.1 s. When the fault occurred, phase A current magnitude
increased as illustrated in Fig. 18.15a. Figure 18.15d shows how the Delta filter
highlighted the internal fault. Eventually, the magnitude of the SCM crossed the
threshold recognizing the internal fault as illustrated in Fig. 18.15e. The proposed
method detected the internal fault 9 ms after the fault inception.
The second case does correspond to a turn-to-turn fault from steady-state.
Fig. 18.16 shows the fault inception at time t = 0.105 ms. Figure 18.16a and b
show the internal fault is reflected on the currents of both sides of the transformer.
However, as Fig. 18.16c shows, the residual currents evidenced an internal fault
occurred. Moreover, the SCM magnitude exceeded the threshold when the internal
fault occurs (see Fig. 18.16e). The SCM method identified the internal fault
correctly, 10 ms after the fault had happened.

18.4.3 Reactor Differential Protection

Voltage control on transmission lines is sometimes achieved through high voltage

(HV) shunt reactors connected on the receiving end of transmission lines to
absorb reactive power. The continuous change in line loading implies that HV
reactors are switched several times per day. These switching events generate inrush
currents with decaying DC component appearing in one or more phases. Reactor
and transformer inrush current share characteristics such as harmonic content and
highly non-linear magnetization current. Conventional protection schemes (namely,
overcurrent, differential, REF (restricted-earth-fault), and distance relays) typically
used to protect reactors are useful to detect high-magnitude faults. However, these
protections are blind to turn-to-turn faults.
The SCM method performance was assessed on a reactor energization event.
Figure 18.17 shows a reactor being energized at time t = 0.2 s . For this case, the
algorithm outlined in Sect. 18.3 was evaluated on the residual currents shown in
Fig. 18.17c. As mentioned above, the residual currents of reactors share distinctive
510 E. Vázquez et al.

PA PB PC
10
Current (A)

5
0
−5
−10
0 0.1 0.2 0.3 0.4 0.5
a
SA SB SC
Current (A)

5
0
−5
0 0.1 0.2 0.3 0.4 0.5
b
res prim res sec
Current (A)

1
0
−1
0 0.1 0.2 0.3 0.4 0.5
c
Δ N res prim Δ N res sec
Current (pu)

1
0
−1
0 0.1 0.2 0.3 0.4 0.5
d
SCMprim SCMsec
0.5
Magnitude

0.25

0
0.1 0.2 0.3 0.4 0.5
Time (s)
e

Fig. 18.15 Method performance during a turn-to-ground fault into the generator differential
protection zone at time t = 0.1 s. (a–e) Identification threshold (dashed line)
18 A New Second Central Moment-Based Algorithm. . . 511

PA PB PC
Current (A)

20
0
−20
−40
0 0.1 0.2 0.3 0.4 0.5
a
SA SB SC
Current (A)

40
20
0
−20
0 0.1 0.2 0.3 0.4 0.5
b
res prim res sec
Current (A)

1
0
−1
0 0.1 0.2 0.3 0.4 0.5
c
Δ N res prim Δ N res sec
Current (pu)

1
0
−1
0 0.1 0.2 0.3 0.4 0.5
d
SCMprim SCMsec
0.5
Magnitude

0.25

0
0.1 0.2 0.3 0.4 0.5
Time (s)
e

Fig. 18.16 Method performance during a turn-to-turn fault into the generator differential protec-
tion zone at time t = 0.105 s. (a–e) Identification threshold (dashed line)
512 E. Vázquez et al.

PA PB PC
10
Current (A)

5
0
−5
0.15 0.2 0.25 0.3 0.35 0.4 0.45
a
SA SB SC
Current (A)

5
0
−5
−10
0.15 0.2 0.25 0.3 0.35 0.4 0.45
b
res prim res sec
Current (A)

0
−0.5
0.15 0.2 0.25 0.3 0.35 0.4 0.45
c
Δ N res prim Δ N res sec
0.5
Current (pu)

0
−0.5
−1
0.15 0.2 0.25 0.3 0.35 0.4 0.45
d
SCMprim SCMsec
Magnitude

0.25

0
0.2 0.25 0.3 0.35 0.4
Time (s)
e

Fig. 18.17 SCM method performance during a power reactor energization. (a) CT primary
currents. (b) CT secondary currents. (c) Residual currents. (d) Filtered and normalized residual
currents. (e) SCM magnitude (solid lines) and identification threshold (dashed line)
18 A New Second Central Moment-Based Algorithm. . . 513

features with transformer inrush currents. After a filtering stage (see Fig. 18.17d)
the method identified the event as a non-fault condition. Figure 18.17e shows that
the magnitude of the SCM did not exceed the threshold of 0.25.
Figure 18.18 presents a second case where a turn-to-turn fault occurred at time
t = 0.1 s. Because the reactor is a serial element, differential relays have problems
detecting turn-to-turn faults. Nevertheless, the SCM method detected the residual
current change, as shown in Fig. 18.18c. Moreover, the Delta filter highlighted the
internal fault waveform (see Fig. 18.18d) allowing the method to identify the event
as an internal fault correctly: the magnitude of the SCM exceeded the threshold of
0.25 as shown in Fig. 18.18e.

18.4.4 Busbar Differential Protection

Busbar protection could be considered one of the most critical elements of pro-
tection. The high fault current levels and the costly damages to the busbar, to all
associated elements, and the rest of the system in case of fault are the main reasons.
Moreover, the clearance of a fault in a busbar reduces the power transfer capability.
This is undesirable effect where the system capability is at limit.
Busbar protection depends on several factors (e.g., the rated voltage or the
complexity of network configuration such as the number of bays, sections, tie-
breakers, and disconnects). Further, the availability of CTs must be considered. The
most typical schemes used in busbar protection are overcurrent, percent differential,
partial differential, high-impedance, directional comparison differential, and time-
coordinated. However, the critical issue busbar protection schemes face is the
heavy saturation occurring in current transformers. Usually, current transformers
have similar CT ratios but are insufficiently rated. CT saturation may lead to the
misoperation of protection relays because of residual currents seen as differential
currents by the relays. Hence, the security of protection schemes is compromised.
On this basis, the SCM method performance was evaluated in two cases: a fault
inside and a fault outside the protection zone. Figure 18.19 shows the results for a
fault inside the differential protection zone occurring at time t = 1 s from steady-
state. The CTs experienced heavy saturation (see Fig. 18.19b) because of the high
DC decay component when the fault occurred. However, Fig. 18.19c shows an
increase in the magnitude of the differential currents. As seen in Fig. 18.19e, the
magnitude of the SCM exceeded the 0.25 limit of 0.25 classifying the event as an
internal fault 26 ms after the fault occurred.
Protection of busbar elements against external faults is one of the critical
scenarios because of the high CT saturation level that CTs are exposed. The
CT saturation will distort the waveform of the CT secondary currents leading
the misoperation of the protection elements. Consequently, the SCM method was
assessed for a fault outside the differential zone. Figure 18.20 shows an ABC-g
three-phase external fault occurring at time t = 0.5 s. Even as CTs are heavily
saturated (see Fig. 18.20a and b), the magnitude of the SCM stayed below the 0.25
514 E. Vázquez et al.

PA PB PC
Current (A)
5
0
−5
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
a
SA SB SC
Current (A)

5
0
−5
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
b
res prim res sec
Current (A)

1
0
−1
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
c
Δ N res prim Δ N res sec
Current (pu)

1
0
−1
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
d
SCMprim SCMsec
0.5
Magnitude

0.25

0
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
Time (s)
e

Fig. 18.18 SCM method performance identifying a turn-to-turn fault in an HV reactor. (a) CT
primary currents. (b) CT secondary currents. (c) Residual currents. (d) Filtered and normalized
residual currents. (e) SCM magnitude (solid lines) and identification threshold (dashed line)
18 A New Second Central Moment-Based Algorithm. . . 515

PA PB PC
Current (A)

20
0
−20
0.9 1 1.1 1.2 1.3 1.4 1.5
a
SA SB SC
Current (A)

20
0
−20
0.9 1 1.1 1.2 1.3 1.4 1.5
b
diff A diff B diff C
Current (A)

1
0.5
0
0.9 1 1.1 1.2 1.3 1.4 1.5
c
Δ diff A Δ diff B Δ diff C
Current (A)

50
0

0.9 1 1.1 1.2 1.3 1.4 1.5

d
SCMA SCMB SCMC
0.5
Magnitude

0.25

0
1 1.1 1.2 1.3 1.4
Time (s)
e

Fig. 18.19 SCM method performance during a fault inside the busbar differential protection
zone. (a) CT primary currents. (b) CT secondary currents. (c) Residual currents. (d) Filtered and
normalized residual currents. (e) SCM magnitude (solid lines) and identification threshold (dashed
line)
516 E. Vázquez et al.

level as seen in Fig. 18.20e. Therefore, the SCM method identified the fault as an
external fault condition.

18.4.5 Power Transformer Energization with Harmonic

Content

As high harmonic distortion produced by non-linear loads may cause the malfunc-
tion of the differential relays, the SCM method was tested for a power transformer
supplying an IGBT-based non-linear load [14]. Figure 18.21a and b show CT
secondary currents for the HV and LV sides, respectively, during a transformer
energization occurring at time t = 0.3 s. Figure 18.21a shows that the magnitude
of the differential currents is affected negatively by the IGBT-based load. Despite
the effects of the non-linear load, the filter stage removed most of the errors (see
Fig. 18.21d). Figure 18.21e shows that the magnitude of the SCM remained below
the 0.25 threshold for the entire duration of the event.

18.4.6 Power Reactor Energization with High Harmonic

Distortion

For the sake of assessing the robustness of the SCM in the presence of highly
harmonic contaminated currents, it is considered the case of a reactor energization
with harmonic content. Figure 18.22a shows the reactor energization occurring at
time t = 0.1 s. The currents measured by the CT display a high harmonic content
(1, 4, 61, 80, 120, 160.3, 180, 200, 210, 240, 300, 340, and 360 Hz) as seen in
Fig. 18.23. Notwithstanding, the filtering stage removes the periodicity present in
the current signals (see Fig. 18.22d), allowing the SCM method to identify the event
correctly. Figure 18.22e shows that the SCM magnitude did not cross the threshold.
18 A New Second Central Moment-Based Algorithm. . . 517

PA PB PC
100
Current (A)

50
0
−50
0.45 0.5 0.55 0.6 0.65 0.7
a
SA SB SC
Current (A)

20
0
−20
0.45 0.5 0.55 0.6 0.65 0.7
b
diff A diff B diff C
Current (A)

1
0.5
0
0.45 0.5 0.55 0.6 0.65 0.7
c
Δ diff A Δ diff B Δ diff C
Current (A)

1
0
−1
0.45 0.5 0.55 0.6 0.65 0.7
d
SCMA SCMB SCMC
Magnitude

0.25

0
0.5 0.55 0.6 0.65
Time (s)
e

Fig. 18.20 Method performance during an external three-phase fault of the busbar differential
protection zone. (a) CT primary currents. (b) CT secondary currents. (c) Residual currents. (d)
Filtered and normalized residual currents. (e) SCM magnitude (solid lines) and identification
threshold (dashed line)
518 E. Vázquez et al.

PA PB PC
Current (A)

10
0

0.3 0.32 0.34 0.36 0.38 0.4

a
SA SB SC
Current (A)

4
2
0
−2
−4
0.3 0.32 0.34 0.36 0.38 0.4
b
res prim res sec
Current (A)

1
0.5
0
−0.5
0.3 0.32 0.34 0.36 0.38 0.4
c
Δ N res prim Δ N res sec
Current (A)

1
0.5
0
−0.5
0.3 0.32 0.34 0.36 0.38 0.4
d
SCMprim SCMsec
Magnitude

0.25

0
0.3 0.32 0.34 0.36 0.38 0.4
Time (s)
e

Fig. 18.21 Method performance during a transformer energization event supplying a non-linear
load. (a) Primary side currents. (b) Secondary side currents. (c) Differential currents. (d) Delta
filtered differential currents. (e) SCM magnitude (solid lines) and identification threshold (dashed
line)
18 A New Second Central Moment-Based Algorithm. . . 519

PA PB PC
Current (A)

10
0
−10
0.1 0.15 0.2 0.25 0.3
a
SA SB SC
10
Current (A)

0
−10
0.1 0.15 0.2 0.25 0.3
b
res prim res sec
5
Current (A)

−5
0.1 0.15 0.2 0.25 0.3
c
Δ N res prim Δ N res sec
Current (pu)

1
0
−1
0.1 0.15 0.2 0.25 0.3
d
SCMprim SCMsec
Magnitude

0.25

0
0.1 0.15 0.2 0.25
Time (s)
e

Fig. 18.22 Method performance during a reactor energization with high harmonic currents. (a)
CT primary currents. (b) CT secondary currents. (c) Residual currents. (d) Filtered and normalized
residual currents. (e) SCM magnitude (solid lines) and identification threshold (dashed line)
520 E. Vázquez et al.

120 Hz

8,000
FFT spectrum amplitude

6,000
1 Hz

180 Hz
4,000 80 Hz
160 Hz

4 Hz 210 Hz
2,000 61 Hz 200 Hz
240 Hz 300 Hz
360 Hz

0
0 50 100 150 200 250 300 350
Frequency (Hz)
Fig. 18.23 Fourier spectrum from a reactor energization with high harmonic content

References

1. Behrendt, K., Fischer, N., & Labuschagne, C. (2011). Considerations for using harmonic
blocking and harmonic restraint techniques on transformer differential relays. Journal of
Reliable Power, 2(3), 36–52 (2011).
2. Benmouyal, G., & Roberts, J. (1999). Superimposed quantities: Their true nature and applica-
tion in relays. In 1999 26th Annual Western Protective Relay Conference (pp. 1–18).
3. Chaitanya, B. K., Soni, A. K., & Yadav, A. (2018). Communication assisted fuzzy based
adaptive protective relaying scheme for microgrid. Journal of Power Technologies, 98(1),
57–69.
4. Che, L., Khodayar, M. E., & Shahidehpour, M. (2014). Adaptive protection system for
microgrids: Protection practices of a functional microgrid system. IEEE Electrification
Magazine, 2(1), 66–80.
5. Edwards, B., Williams, D. G., Hargrave, A., Watkins, M., & Yedidi, V. K. (2017). Beyond the
nameplate – Selecting transformer compensation settings for secure differential protection. In
2017 70th Annual Conference for Protective Relay Engineers (CPRE) (pp. 1–23).
6. Hargrave, A., Thompson, M. J., & Heilman, B. (2018). Beyond the knee point: A practical
guide to CT saturation. In 2018 71st Annual Conference for Protective Relay Engineers
(CPRE) (pp. 1–23).
7. He, M., Zhang, J., & Vittal, V. (2013). Robust online dynamic security assessment using
adaptive ensemble decision-tree learning. IEEE Transactions on Power Systems, 28(4), 4089–
4098.
8. Hensler, T., Pritchard, C., Fischer, N., & Kasztenny, B. (2018). Testing superimposed-
component and traveling-wave line protection. In 2018 72nd Annual Georgia Tech Protective
Relaying Conference (pp. 1–12).
18 A New Second Central Moment-Based Algorithm. . . 521

9. IEEE guide for the application of current transformers used for protective relaying purposes
(2008). IEEE Std C37.110-2007 (Revision of Std C37.110-1996) (pp. 1–90).
10. Kasztenny, B., Fischer, N., & Altuve, H. J. (2015). Negative-sequence differential protection
– Principles, sensitivity, and security. In 2015 68th Annual Conference for Protective Relay
Engineers (pp. 364–378).
11. Kauhaniemi, K., & Voima, S. (2012). Adaptive relay protection concept for smart grids. In:
Proceedings of the Renewable Efficient Energy II Conference (pp. 1–10).
12. Lin, H., Sun, K., Tan, Z., Liu, C., Guerrero, J.M., & Vasquez, J.C. (2019). Adaptive protection
combined with machine learning for microgrids. IET Generation, Transmission Distribution,
13(6), 770–779.
13. Massoud Amin, S., & Wollenberg, B. F. (2005). Toward a smart grid: Power delivery for the
21st century. IEEE Power and Energy Magazine, 3(5), 34–41.
14. Mishra, M. K., & Karthikeyan, K. (2009). An investigation on design and switching dynamics
of a voltage source inverter to compensate unbalanced and nonlinear loads. IEEE Transactions
on Industrial Electronics, 56(8), 2802–2810.
15. Proakis, J. G., & Manolakis, D. G. (1996). Digital signal processing: Principles, algorithms,
and applications (3rd ed.) Upper Saddle River, NJ: Prentice Hall.
16. Samantaray, S., Mishra, D., & Joos, G. (2018). A combined wavelet and data-mining based
intelligent protection scheme for microgrid. In: 2018 IEEE Power Energy Society General
Meeting (PESGM) (pp. 1–1).
17. Senarathna, T. S., & Udayanga-Hemapala, K. T. M. (2019). Review of adaptive protection
methods for microgrids. AIMS Energy, 7(5), 557–578.
18. Welford, B. P. (1962). Note on a method for calculating corrected sums of squares and products.
Technometrics, 4(3), 419–420.
19. Zacharias, D., & Gokaraju, R. (2016). Prototype of a negative-sequence turn-to-turn fault
detection scheme for transformers. IEEE Transactions on Power Delivery, 31(1), 122–129.
20. Zhang, Y., Huang, T., & Bompard, E. F. (2018). Big data analytics in smart grids: A review.
Energy Informatics, 1(8), 1–24.
21. Ziegler, G. (2012). Numerical differential protection: Principles and applications (2nd ed.)
Erlangen: Publicis Publishing.
Chapter 19
Microgrid Protection with Conventional
and Adaptive Protection Schemes

Aushiq Ali Memon, Hannu Laaksonen, and Kimmo Kauhaniemi

19.1 Introduction

Microgrid is considered as the building block for the future smart grids, therefore
a properly designed microgrid is essential for the proper functioning of the entire
smart grid. Microgrid has many definitions but in general, it is a well-designed
local distribution system of electricity that facilitates the local integration of many
small-scale renewable energy sources and energy storage systems to meet the local
energy demand of consumers in a smart, secure, efficient, controlled, protected,
and managed environment. Unlike the traditional distribution systems of electricity,
microgrids can operate in islanded mode when the main grid is disconnected due
to faults. The islanded mode operation of the microgrid will enhance the reliability
level of the existing distribution system if some sections of distribution networks
are planned as microgrids. However, the islanded mode operation of microgrid
has many challenges and one important challenge is related to the design of the
protection scheme. This chapter addresses the issues related to protection schemes
in a microgrid, gives an overview of the existing and new requirements of protection
schemes, and analyses the potential of the existing and adaptive protection schemes
of a microgrid.

A. A. Memon () · H. Laaksonen · K. Kauhaniemi

School of Technology and Innovations, University of Vaasa, Vaasa, Finland
e-mail: aushiq.memon@uva.fi

© Springer Nature Switzerland AG 2021 523

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_19
524 A. A. Memon et al.

19.2 Microgrid Protection Issues

Microgrid protection issues can be classified into two broad categories depending
on its operational modes [1]:
(1) Microgrid protection issues in grid-connected mode (2) Microgrid protection
issues in islanded mode.
For the grid-connected mode of microgrid, the faults inside the microgrid as
well as faults outside the microgrid are considered. In the grid-connected mode
protection schemes of microgrids should not operate unnecessarily for faults outside
the microgrid for example faults upstream of the circuit breaker (CB) at the point
of common coupling (PCC). All faults insides the microgrid should be detected and
selectively isolated for the minimum interruption to other parts. During external
faults at the main grid, microgrid should be able to disconnect quickly to protect
its loads and start operating in islanded mode and as soon as the external fault is
removed it should be reconnected to the main grid. The external faults at the main
grid could be close-in faults or other far-end faults resulting in the loss of mains,
hence both types of faults should be differentiated. In the grid-connected mode fault
contribution from both the main grid and microgrid sources is being considered.
If microgrid has the majority of synchronous generator-based distributed energy
resources (DERs), then fault contribution from the main grid could be reduced
and overcurrent relays may experience selectivity issues like more time to issue
a trip command. On the other hand, if the majority of converter-based DERs are
connected inside the microgrid, then it will not have significant issues during the
grid-connected mode since fault contribution from the converter-based DERs could
be limited to twice the rated current capacity or even less depending on the converter
settings or rating.
For the islanded mode of operation, the faults inside the microgrid are only
considered and fault contribution from local DERs and energy storage systems
is taken into account. Microgrid protection should detect and isolate the faults
selectively even in an islanded mode of operation. During island operation, for
example, in low-voltage (LV) microgrids, large fault currents from the upstream
power system grid are not available. In addition, a large share from the DER
units in LV microgrids will be inverter-based with low fault currents. Therefore,
traditional one-directional protection schemes, assuming a large difference between
fault and load currents, are not applicable during island operation. These traditional
protection methods could also have slower fault clearing time and reduced sensitiv-
ity and selectivity. This means that the system reliability is expected to decrease
if the protection schemes are not adapted [2]. For these reasons, conventional
overcurrent (OC) protection based on fuses and one setting group will not be
able to guarantee selectivity during different types of possible faults. Therefore,
LV network conventional protection will not be compatible with island operated
LV microgrids and new protection schemes with adaptivity must be created. On
the other hand, the new LV microgrid protection scheme needs to be economical
and simple [3, 4]. For medium-voltage (MV) microgrids after isolation from the
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 525

utility grid, local DERs are the only fault current sources in the electric island,
and the fault current level depends on the types, sizes, and locations of the DER.
However, it is generally lower than the fault current from the utility grid [5]. The
reduced fault current contributions from microgrid DERs require revised protection
settings with the reduced pickup or threshold values of currents for islanded mode or
protection schemes based on other protection principles like differential current [6],
symmetrical components and residual current based [7], voltage based [8], harmonic
content based [9], or a suitable combination of them should be employed to detect
and clear the fault in islanded mode. An adaptive protection scheme [10] using high-
speed communication links and numerical directional overcurrent protection relays
could also be a suitable protection scheme to change protection settings adaptively
according to the grid-connected mode or islanded mode operations.
In the creation of a new protection system for microgrids, multiple issues need
to be taken into account, like
• The number of zones for protection,
• Operation speed specifications for the different operation modes and configura-
tions of microgrid and.
• Protection methods for microgrid normal grid-connected and islanded operation
[4].
The created microgrid protection system also needs to be compatible with the
microgrid operation and control solutions. Some key issues related to the LV
microgrid protection are briefly reviewed based on [11] from which more detailed
information can be found. The extent and number of microgrid protection zones
will determine the required number of protective devices (PDs) for microgrid
protection. However, the protection system simultaneously needs to fulfill customer
requirements and be economically feasible [4].
The essential structural choices will define the operation speed needs and
principles for the protection of LV microgrid and correspondingly the operation
speed requirements will determine some of the structural choices required to fulfill
the operation speed needs. The main reasons for the operation speed requirements of
LV microgrid protection are stability and customer sensitivity. The stability needs to
be maintained after fast disturbances like after islanding due to fault in the upstream
network during normal grid-connected operation or after a fault in the LV microgrid
during islanded operation. One important issue related to the operation principles
of LV microgrid protection is the fault behavior of the converter-based DER units.
The fault behavior needs to be compatible with the developed microgrid protection
Scheme [4].
As stated in [12], the microgrid protection issues cannot be solved without a
complete understanding about the microgrid dynamics before, during, and after
islanding or fault. Related to this, for example, directly connected rotating genera-
tors or motors are very sensitive from the stability viewpoint in voltage dips caused
by faults during microgrid island operation, and so they may endanger the stability
of the microgrid. Therefore, if directly connected rotating machines are connected
to a microgrid, protection should operate rapidly during all kinds of faults. For
526 A. A. Memon et al.

example, if microgrid customers have fuses with high nominal currents, there can
be a risk that customer protection may operate too slowly during island operation
due to small fault currents and that may lead to instability after clearance of the fault
[4].
In cases where overcurrent based protection is utilized during island operation,
the protection and control functions of IEDs (intelligent electronic devices) in
microgrids may need real-time information about network topology, the status of
DER units (on or off), the state of charge of storage systems, and also number and
size of loads connected to the microgrid. These conditions have to be updated and
checked continuously in order to guarantee that protection settings are suitable for
the actual configuration [3, 4].
Based on the above and as mentioned in [12], the high-speed operation of the
protection devices is very crucial for reliable operation of the microgrid protection
system. Utilization of high-speed telecommunication is expected to be an essential
part of future smart grid protection systems to achieve fast and selective protection
both in grid-connected and islanded modes of operation. The same communication
protocols and standards used in HV/MV network can be applied directly to the
LV microgrids. However, due to the smaller scale of LV microgrids, the costs of
protection devices must also be lower than the cost of devices used in the HV/MV
network [4].
One important issue, which is required to enable a stable transition from the
normal grid-connected operation to island operation, is the coordination of IED
protection settings with DER unit fault-ride-through (FRT) requirements (especially
low-voltage-ride-through, LVRT). In order to avoid unwanted tripping, faulty lines
must be disconnected first by the protection system and after that, the DER units
should be disconnected according to their FRT or LVRT functionality. Rapid pro-
tection operation is needed especially if there are many protection zones. Therefore,
communication-based protection methods and schemes are often required to achieve
selectivity [13]. Some further discussion about issues related to the protection of
microgrids can be also found in [2, 4].

19.2.1 DER Unit Fault Behavior and Effect on Microgrid

Protection

In the future, proper coordination between distributed generation (DG) unit grid
codes and distribution network protection schemes will be increasingly important
during both grid-connected and islanded operation of microgrids. The operation
time settings of short-circuit and earth-fault protection must be selective with DER
unit fault-ride-through settings during normal operation. In MV network short-
circuit protection operation time delays have traditionally been dependent on the
fault-current magnitude or measured impedance with fixed time delays or inverse
time curves. In the future, MV networks will be increasingly divided into multiple
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 527

protection zones to improve supply reliability. Therefore, short-circuit protection

operation times may become too long if high-speed communication based, and
for example, IEC 61850 GOOSE signals-based interlocking/blocking schemes are
not utilized. On the other hand, the communication may fail or is not available,
and therefore also grid code compatible protection schemes that are not based
on high-speed communication are needed in the future at least as a backup for
communication-based schemes.
Regarding the fault behavior required from DER units during faults, it is
important to ensure that the fault behavior is compatible with the developed LV
microgrid protection scheme and considers also the including FRT needs. This
means that when the protection for an island operated microgrid is determined, one
of the most important considerations is related to the fault current contribution of
the converter-based DER units [4, 14].
Fault behavior and fault current feeding capability of a DER unit is also highly
dependent on the type of the DER unit. For example, a synchronous generator
is usually able to feed prolonged fault current (about 200 to 400% of nominal
current). An induction generator, in the initial stage, feeds almost as big a fault
current as a synchronous generator, but the feeding is reduced quickly. Fault current
fed by the inverter-connected generating unit is typically limited to 1.2–1.3 p.u.
and highly dependent on the control system and control principles as well as
grid code requirements regarding fault-ride-through and reactive current feeding
requirements.
During normal grid-connected operation of microgrid with different types of
DER units, the grid codes can require FRT capability from the DER units in terms of
frequency (f ), rate-of-change-of-frequency (df/dt), voltage (U), and voltage support.
Traditionally, in grid-connected operation, larger DER units have more FRT and
frequency as well as voltage support related requirements in the grid codes. For
example, regarding voltage support, the FRT capability of DER units is defined with
a voltage-against-time-profile (LVRT curve). In addition, also additional voltage
support by capacitive/reactive, positive sequence, current injection during faults is
required from MV and HV network connected converter−/doubly-fed-induction-
generator (DFIG)-based DER units during grid-connected operation. Synchronous
generator DER units naturally provide voltage support during faults by feeding
reactive current. The required dynamic response of the reactive current feeding is
usually also defined in the grid codes for the converter-based DER units. However,
LVRT capability and additional reactive current feeding of the converter-based
DER units, like wind turbines, is not just a control issue. It also requires suitable
technology to be applied in the DC-link of the converter, like for example, DC-link
chopper or supercapacitor and crowbar possibly as a backup.
Most grid codes for the grid-connected operation of DER units do not include any
special requirements for the supply of negative sequence current. In synchronous
generators, the negative sequence current is fed naturally and there are no effective
measures to influence it. In contrast, with voltage source converters (VSCs) it is
possible to individually control positive and negative sequence quantities. The main
reason for the minimization of negative sequence current has been the impact of
528 A. A. Memon et al.

asymmetrical voltages on the DER unit (e.g., wind turbine) without considering
the impact of it on the network voltages or network protection. The negative
sequence may be controlled to mitigate twice the fundamental frequency oscillations
appearing in the converter/inverter DC-link during asymmetrical faults. Therefore,
in many applications the negative sequence component injection has been reduced
partially or fully. However, full negative sequence current reduction control also
reduces the line-to-line (2-phase) short circuit current to the level of the load
current or even to zero, which means that it may prevent the correct operation
of network protection in 2-phase short-circuit faults. To overcome this problem,
extensions to grid code have been proposed, which would require DG to inject
a clearly defined level of negative sequence current. This is also a requirement
already in Germany. From the network voltages point of view, the effect of negative
sequence fault current feeding during asymmetrical faults is beneficial because
it reduces the negative sequence voltage, improves the voltage phase symmetry,
and reduces the overvoltages in the healthy phases during 2-phase short-circuits.
However, a negative sequence current injection will limit the control capability of
the DG in the positive sequence. It can be expected that the future grid codes will
increasingly specify requirements for asymmetrical/imbalanced negative sequence
current injection during asymmetrical faults.
Usually, the control mode change of one or more DER units connected to the
distribution network is required after changing from the normal grid-connected
operation to island mode. Traditionally, this means that under normal operation the
DER unit uses active(P)/reactive(Q) power control and after islanding, the control
mode is changed to voltage(U)/frequency(f ) control (or voltage/speed control).
However, control schemes that do not require changing after transition to/from
island operation have also been proposed. For example, in [15], an enhanced
control strategy was proposed that improves the performance of a DER unit under
network faults and transient disturbances, in a multiunit microgrid setting. The
proposed control strategy does not require the detection of the mode of operation and
switching between different controllers (for grid-connected and islanded) modes,
and it enables the adopted DER units to ride through network faults, irrespective of
whether they take place within the host microgrid or impact the upstream grid [15].
LVRT, high voltage ride-through (HVRT) as well as f and df/dt related FRT
capabilities are required during LV microgrid island operation also from the small-
scale DER units, which typically is not the case during grid-connected operation.
Typically, in grid-connected operation the small-scale DER units, like PV units,
are only required to support frequency stability during over-frequency situations
by their active power-frequency (Pf )-droop control. From the island operated LV
microgrid protection viewpoint, it is important to know exactly how the converter-
based DER units behave during the faults and what kind of fault current (active,
reactive, positive sequence, negative sequence, etc.) they will feed. Therefore, the
so-called microgrid grid codes for island operated networks are necessary for the
development of future smart grid protection solutions to reduce complexity and to
avoid the need for too many case specific alternatives [4, 14].
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 529

Based on the simulations done in [14], the increased reactive power feeding
with the converter-based DER units was found to be beneficial for the possible
overcurrent-based protection in LV microgrid. However, it did not significantly
reduce the usability of undervoltage-based protection either due to the resistive
characteristics of LV lines. On the other hand, the reactive power feeding during the
fault did not significantly reduce the magnitude of the voltage dip, that is, support
microgrid voltage during the fault. In the end, the excessive reactive power feeding
of the converter-based DER units during the fault in island operated LV microgrid
was not justified based on the simulations in [14]. Therefore, it was suggested that
during faults in LV island operated microgrid the fault current fed by the converter-
based DER units is recommended to be active instead of reactive if possible. In
addition, the control of the converters during possible faults was not recommended
to change due to the increased possibility of instabilities after fault clearance.

19.2.2 Example—Microgrid Transition to Islanded Operation

In the following, the example from [16] is used to define the protection needs (func-
tions, time selectivity) when intentional island operation and especially successful
transition to island operation are considered.
From the island operation perspective (in addition to grid code FRT requirements
from DG units during the normal grid-connected operation) it is required that the
wind farm (green) in Fig. 19.1 as well as DER (1) and (2) have sufficient FRT
capabilities. Figure 19.1 shows possible intended islands and MV feeder short-
circuit protection at CB1-CB4 which is assumed to be directional like the earth-fault
protection. In the following, the time selectivity issues (Fig. 19.2) are discussed,
with different fault scenarios (faults A-E in Fig. 19.1), regarding the successful
transition to island operation.
In Fig. 19.2 protection time selectivity issues, general time delay setting princi-
ples, and the role of high-speed communication is shown when (a) islanding is not
allowed and (b) islanding is possible. Figure 19.2 also illustrates the role of high-
speed communication-based interlockings/blockings (as well as transfer trip-based
islanding detection) in the reliable and selective operation of future distribution
networks with many sequential protection zones and the possibility for the intended
island operation. In Fig. 19.1 and 19.2, the idea is that the possible operation
principles of directional short-circuit protection in the forward direction can be
1. Directional overcurrent protection with a fixed time delay (and high-stage/low-
stage settings).
2. Distance protection with a fixed time delay (in a forward direction).
Similarly, in Fig. 19.1 and 19.2 the possible operation principles of directional
protection in the reverse direction (for intentional islanding) can be
530 A. A. Memon et al.

Fig. 19.1 Possible intended islands 1.-4. b (see also Fig. 19.2) [16]

1. Undervoltage with a fixed time delay (and high-stage/low-stage settings) AND

current direction detection in the reverse direction. Function pick-up/start is only
based on undervoltage (i.e., not in overcurrent, because fault current levels of
inverter-based DER units can be fairly low as discussed in previous chapters).
2. Distance protection with a fixed time delay (in the reverse direction).
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 531

Fig. 19.2 Protection time selectivity issues, setting principles, and the role of high-speed com-
munication when (a) islanding is NOT allowed and (b) islanding IS possible (see also Fig. 19.1)
[16]
532 A. A. Memon et al.

From Fig. 19.1 and 19.2 it can be seen that selectivity problems are possible
if communication-based interlockings/blockings, for example, are NOT used (Fig.
19.2a), because coordination between LVRT curve of DG units (defined by grid
codes) and required time differences between CB2 and CB3 in the forward direction
may be hard to achieve. This naturally depends on the number of consecutive
protection zones and the allowed time difference between the operation time delays
of CB2 and CB3.
Transition to intentional island operation IS only possible (Fig. 19.2b) if active
and reactive power unbalance at CB1, CB2, CB3, or CB4 is small enough (or
enough, rapidly controllable active, and reactive power units exist in the possible
island (1.-4.b in Fig. 19.1) before the protection start/operation of CB1-CB4 in the
reverse direction). If this is not the case transition to island operation should not be
allowed. Here it is worth mentioning that the recent grid codes enable/support transi-
tion to intentional island operation because of the P/f -droop control requirements of
DER units during over-frequency situations (under-frequency based load shedding
schemes could have a similar kind of effect) and possibly also due to voltage control
(Q/U-control) requirements.
In the above discussion and Fig. 19.2, only short-circuit protection has been
considered, but naturally also earth-fault protection principles and settings must be
proper during both normal and island operation. Therefore, it should be noted here
that after opening CB2, CB3, or CB4, the MV network neutral earthing method may
change, for example, from compensated to isolated and MV feeder IEDs earth-fault
protection settings and protection principles also need to adapt to these changes.
In the following, the protection operation principles during different faults A-E in
the example network shown in Fig. 19.1 are described. It is assumed that high-speed
communication is available/possible, and islanding IS possible (Fig. 19.2b) using
time selectivity (with the communication) if power generation and consumption are
close to each other behind the possible island connection point CB.
In case of fault A in Fig. 19.1 (see also Fig. 19.2):
• CB1 will operate in a reverse direction and disconnect the LV microgrid
intentionally from the utility network.
– If active and reactive power unbalance at CB1 is small enough (i.e., stable
transition to island operation is possible) as stated earlier.
– CB1 could also send signals to the LV microgrid DER units to change their
control mode, and so on after the operation.
• CB2 will operate in a forward direction.
– CB2 sends a simultaneously interlocking signal to CB3 (and CB4) to prevent
their false operation and.
– CB2 can also send a communication-based transfer trip (faulty island) discon-
nection signal to the DER unit (2).
Wind farm and DER unit (1) will remain connected according to the LVRT curve
of DG units (Fig. 19.2).
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 533

In case of fault B in Fig. 19.1 (see also Fig. 19.2):

• CB1 will operate in a reverse direction and disconnect the LV microgrid
intentionally from the utility network.
– If active and reactive power unbalance at CB1 is small enough.
– CB1 could also send signals to the LV microgrid DER units to change their
control mode, and so on after the operation.
• CB2 may also operate in a reverse direction and disconnect part of the MV feeder
intentionally from the utility network to island operation.
– If active and reactive power unbalance at CB2 is small enough.
– CB2 could also send a signal to the DER unit (2) to change the control mode,
and so on after the operation.
• CB3 will operate in a forward direction.
– CB3 sends a simultaneously interlocking signal to CB4 to prevent false
operation and.
– CB3 can also send a communication-based transfer trip (faulty island) discon-
nection signal to the DER unit (1).
Wind farm will remain connected according to the LVRT curve (Fig. 19.2).
In case of fault C in Fig. 19.1 (see also Fig. 19.2):
• MV busbar fault.
• CB1 will operate in a reverse direction and disconnect the LV microgrid
intentionally from the utility network.
– If active and reactive power unbalance at CB1 is small enough.
– CB1 could also send signals to the LV microgrid DER units to change their
control mode, and so on after the operation.
• CB2 may also operate in a reverse direction and disconnect part of the MV feeder
intentionally from the utility network to island operation.
– If active and reactive power unbalance at CB2 is small enough.
– CB2 could also send a signal to the DER unit (2) to change the control mode,
and so on after the operation.
• CB3 may also operate in a reverse direction and disconnect either part of the MV
feeder (beginning of the feeder, 3. in Fig. 19.1) OR the whole MV feeder (2. and
3. in Fig. 19.1) to island operation.
– Depending on the active and reactive power unbalance at CB3 and CB2.
– Coordination with islanding of part of MV feeder by opening CB2 may be
beneficial/required.
– CB3 could also send a signal to the DER unit (1) (AND/OR DER unit (2)
depending on the power balance situation) to change control mode, and so on
or to disconnect.
• CB4 will operate in a forward direction.
534 A. A. Memon et al.

The wind farm will be disconnected according to LVRT curve (Fig. 19.2). Also,
directional short-circuit protection in the reverse direction could be included in
CB3c in order to disconnect the wind farm more rapidly in case of busbar faults
like fault C (Fig. 19.1). However, in case of upstream faults (like fault E and
F in Fig. 19.1), this reverse direction protection at CB3c should be blocked by
communication to enable fault-ride-through support from the wind farm according
to grid codes (e.g., LVRT curve).
In case of fault D in Fig. 19.1 (see also Fig. 19.2):
– Parallel MV feeder fault.
– CB3b will operate in a forward direction (Fig. 19.1) and simultaneously send
the interlocking signal to other CBs (e.g.) to avoid unnecessary islanding and to
ensure selectivity.
DER units (1) and (2) as well as the wind farm will remain connected according
to the LVRT curve (Fig. 19.2).
In case of fault E in Fig. 19.1 (see also Fig. 19.2):
• HV/MV transformer fault => intentional islanding can take place.
– Possible indication about intentional islanding possibility from HV/MV
transformer protection IED to MV feeder IEDs.
• CB1 will operate in a reverse direction and disconnect the LV microgrid
intentionally from the utility network.
– If active and reactive power unbalance at CB1 is small enough.
– CB1 could also send signals to the LV micro-grid DER units to change their
control mode, and so on after the operation.
• CB2 may also operate in a reverse direction and disconnect part of the MV feeder
intentionally from the utility network to island operation.
– If active and reactive power unbalance at CB2 is small enough.
– CB2 could also send a signal to the DER unit (2) to change the control mode,
and so on.
• CB3 may also operate in a reverse direction and disconnect either part of the MV
feeder (beginning of the feeder, 3. in Fig. 19.1) OR the whole MV feeder (2. And
3. in Fig. 19.1) to island operation.
– Depending on the active and reactive power unbalance at CB3 and CB2.
– Coordination with islanding of part of MV feeder by opening CB2 may be
beneficial/required.
– CB3 could also send a signal to the DG unit (1) (AND/OR DER unit (2)
depending on the power balance situation) to change the control mode, and so
on. or to disconnect.
• Alternatively, CB4 may also operate in reverse direction and disconnect.
– All MV feeders (4.b in Fig. 19.1) to island operation.
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 535

– Some of the MV feeders (e.g., (4.a in Fig. 19.1)) to island operation.

Simultaneously, the disconnection signal from CB4 is sent to MV feeder
CBs (e.g., (CB3c in Fig. 19.1)) which cannot be included into an intentional
island.
Wind farm remains connected according to the LVRT curve (Fig. 19.2)
unless the disconnection signal to CB3c (Fig. 19.1) is sent by CB4.
– This intentional island scheme is somewhat more complex due to the increased
number of possible island sizes.
– Size of the island depends on the active and reactive power unbalance at CB4,
CB3, and CB2 and.
Needs (central) coordination with islanding of part of or whole MV feeder
by opening CB3 or CB2.
Signals to DER unit (1) and (2) to change the control mode, and so on after
the operation or to disconnect will be sent based on the planned island size
(which depends from the power balance situation before fault C).
In case of fault F in Fig. 19.1 (see also Fig. 19.2):
• HV network fault = > intentional islanding should not take place (if there is an
alternative HV network supply route available), MV and LV network connected
DER units should ride-through HV network faults and possibly also support
the HV network according to grid code requirements (e.g., by reactive power
injection).
– Possible indication about HV network fault could be sent from the HV/MV
transformer protection IED to MV feeder IEDs to indicate that in this case
intentional islanding is not possible/allowed.
Instead, FRT and HV network support of DER units is preferred.
– However, if very sensitive customers (sensitive to voltage dips) are connected
to MV or LV network then similar actions could take place as described earlier
for the other faults.

19.3 Protection Requirements

The traditional protection scheme requirements include sensitivity, selectivity, and

reliability. However, the capability of a microgrid to work in an islanded mode
demands the additional requirement of adaptivity for the protection scheme. More-
over, microgrid transition from the islanded mode to the grid-connected mode or
even from the isolated mode to the islanded mode requires re-synchronization func-
tion to avoid wrong protection tripping during transition periods. Since the transi-
tions from the grid-connected to the islanded mode and vice versa are mainly depen-
536 A. A. Memon et al.

dent on the connection and disconnection status of microgrid switches/breakers,

therefore the selection of proper switching technology is also important. In this
section, the existing and new protection requirements of a microgrid are discussed.

19.3.1 Sensitivity

Sensitivity is one of the essential requirements of a protection scheme. Any protec-

tion scheme should be sensitive enough to sense and detect the abnormality or fault
in the power system components and respond to clear the fault as soon as possible.
Sensitivity can be defined as “the ability of the protection system to detect even the
smallest faults within the protected zone” [17]. Sensitivity is related to the minimum
pickup or threshold value of measured or sensed quantity which is by some margin
above or below the boundary of normal operating value. Different protection
methods have different sensitivity levels depending on operational characteristics or
settings and the magnitude of electrical or physical quantity on which the protection
scheme is working. Fuses, for example, are the traditional and simple overcurrent
protection devices that depend on the magnitude of the current flowing through
the fusible element which blows or melts due to thermal effects produced by the
current. Fuses are both sensing and interrupting devices with inverse operational
characteristics, which means less time of fusing operation at a high magnitude
of current and more time of fusing operation at a low or minimum magnitude of
a current flowing through the fusible element. The operational characteristics of
fuses greatly depend on the material of the fusible element through which electric
current flows during the fault. Therefore, operational characteristics of fuses can be
changed only by changing the material of the fusible element, but once the fuse type
is selected and installed the operational characteristics can no longer be changed,
this makes fuse a non-resettable device. Different types of fuses have different
time-current characteristics and therefore different levels of operating sensitivities.
The other protection device working on the overcurrent principle is the overcurrent
relay. The sensitivity of definite time overcurrent relay depends on its capability
or settings to detect the lowest possible magnitude of the current flowing through
the circuit during a fault. Usually, the lowest possible magnitude of the current is
observed in case of single-line to ground (SLG) short circuit fault with high fault
impedance and the maximum current is observed during three-phase (LLL) short
circuit fault. In conventional radial distribution systems with power flow in one
direction, overcurrent protection schemes for the detection of SLG faults need to be
more sensitive in comparison with overcurrent protection schemes for the detection
of three-phase faults. However, with the increasing connection of small-scale DER
units particularly the converter-based DERs result in very reduced three-phase fault
current levels which greatly affect the sensitivities of existing overcurrent relays and
both definite-time and inverse-time overcurrent relays may experience the blinding
of overcurrent protection. The blinding of protection scheme occurs when a fault
exists but the protection scheme is unable to detect the fault due to wrong or
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 537

less sensitive settings. Moreover, transient events like the starting of motor loads,
the energization of transformers, and switching of capacitor banks may affect the
sensitivity of overcurrent protection and result in wrong or unnecessary trips due to
more sensitive settings. Additional filtering, signal processing algorithms, or even
changed protection philosophies or functions can be employed to avoid these situ-
ations. Microgrids require different levels of sensitivities in overcurrent protection
schemes to differentiate between faults in the grid-connected and islanded mode
of operation. In the grid-connected mode, less sensitive settings are required to
avoid the unnecessary trips of protection schemes, and in the islanded mode more
sensitive settings are necessary to avoid the blinding of protection schemes. Other
protection functions based on the measurement of voltage, frequency, or impedance
may have different levels of sensitivities in different operational modes and different
fault categories, and these functions should be carefully selected and configured. For
example, the measurement of fault resistance (RF ) coverage is used as a means of
evaluating sensitivity of directional overcurrent, distance, and differential protection
for SLG faults [17].

19.3.2 Selectivity

Selectivity or coordination is an important requirement of the traditional protection

system which ensures that only the section of the power system close to the fault is
isolated and the minimum portion of the power system is interrupted. For a complete
selective or a coordinated protection scheme, primary protection operates first after
fault detection inside the protection zone and backup protection operates only after
primary protection fails to detect and isolate the fault after a predetermined time
delay. For definite time OC relays definite time-based coordination is done which
starts, for example, from the extreme load end of the feeder toward the source side
of the feeder or substation. It means for faults at the extreme end of the feeder on
the load side, OC relay near the load operates first and then a coordination interval
of 0.2 s or so is used between each upstream OC relay toward the substation end for
the backup protection of downstream relays. Usually, fast acting relays and breakers
or instantaneous adjustable devices like miniature circuit breaker (MCB) or molded
case circuit breakers (MCCB) are used on the load side for short circuit protection.
If due to any problem, load side primary protection fails, the first upstream relay
near the load will operate as a backup after a coordination interval of 0.2 s, the same
is valid for all subsequent upstream faults. This definite time-based coordination
is commonly used for three-phase faults in radial distribution feeders. Definite time
relay coordination is simple, independent of fault current magnitude, and it provides
complete coordination as long as each relay is able to detect the fault. The only
drawback for definite time coordination is longer tripping times for faults near the
source or substation side of the feeder. The inverse-time OC relays, commonly
known as inverse definite minimum time (IDMT) relays have operating times that
are inversely proportional to the magnitude of the current, the higher the fault
538 A. A. Memon et al.

current magnitude, the faster the IDMT relay operates. The IDMT relays have
several families of inverse characteristic curves with different degrees of inversity
like standard inverse, very inverse, and extremely inverse characteristics. For a given
pickup current above the minimum pickup value and an identical time-dial setting,
an extremely inverse IDMT relay operates faster than a very inverse relay and a
very inverse relay operates faster than the standard inverse. The IDMT relays with an
inverse or very inverse characteristics are the most commonly used types and ideally,
the relays with the same inverse characteristics are used throughout the system [18].
The IDMT relays operate faster for faults near the source or substation and slower
for faults away from the substation. Compared with the coordination of definite time
relays, the coordination of IDMT relays is a much complicated and time intensive
job, however, it all depends on the method used. The general methods used for the
coordination of IDMT relays include trial and error, curve fitting, and optimization
techniques using different algorithms [19]. In traditional radial distribution systems
with power flow in one direction, the coordination of OC relays may not be affected,
however, when a considerable amount of DERs are connected the coordination is
either altered or completely lost depending on the capacity, type, and location of
DERs [20, 21]. For microgrids with nearly 100% DER supply with the majority
of the converter-based DERs, the loss of protection coordination will result in the
reduced security of supply to consumers due to the disconnection of large portions,
which is not in line with the very fundamental purpose of microgrids. Therefore,
microgrid protection must be coordinated in both the grid-connected and islanded
mode of operation. This could be done by the separate coordination study and
settings of grid-connected and islanded mode protections or by providing sources of
high fault current also in islanded mode. A case study of protection coordination in
the grid-connected and islanded mode of microgrid using definite-time and inverse-
time OC relays is presented in Sect. 4 for further understanding about selectivity
issues.

19.3.3 Reliability

Reliability is the ability of a protection scheme to operate correctly and is usually

defined in terms of dependability and security of relay operations. Dependability is
defined as the measure of certainty that the protection relay will operate and trip
for all faults for which it is designed and security is the measure of certainty that
the protection relay will not operate and trip incorrectly. Traditionally, protection
schemes have been designed to provide high dependability at some degree of
compromise of security that may increase the false operations of protection schemes
resulting in the unwanted trippings of power system elements. Traditional large
interconnected power systems provide some degree of redundancy due to many
alternative paths of power flow and therefore the loss of a generator or a line (n-
1 criterion) due to a false trip is less objectionable in comparison with the sustained
fault which may damage the faulty component [17, 22, 23]. But, the false trips of
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 539

a distribution line due to the unsecured protection scheme in radial, grid-connected

microgrids is less acceptable because it may jeopardize the stability of microgrid.
The false trips of the breaker at the microgrid connection point to the main grid
may result in unwanted islanding, increased re-synchronization operations for the
restoration of grid-connected operations, unwarranted outage to non-priority loads,
and microgrid exposure to power quality problems [12, 24]. In an islanded mode
of operation, a false trip of a grid-forming generator may result in a complete
blackout due to the consequent tripping of grid-following generators. The reliability
of modern multifunction numerical relays using communication links is not only
dependent on hardware and software-based failures but also the dependability and
security of the communication system. To analyze the dependability and security
of protection schemes different fault trees can be used. Fault tree analysis is a
useful tool for the comparison of the relative reliability of protection schemes. The
construction of a fault tree starts with the identification of component failures which
may cause a failure to trip (a dependability problem) or an unwanted trip (a security
problem). The AND, OR, or other gates are used to represent the combinations
of failure rates. The idea behind using the OR gate is that any of several failures
can cause the protection scheme to fail, whereas, the AND gate expresses the idea
that all component failures happen simultaneously to cause a protection scheme to
fail. Various combinations of protection schemes using fiber-optic channels have
been analyzed for dependability and security using fault trees [17]. Such type of
reliability analysis will be useful for protection schemes in microgrids.

19.3.4 Adaptivity

The adaptivity of the microgrid protection scheme is the new requirement that is
the ability of the protection scheme to adapt its settings according to changing
operational modes from the grid-connected to an islanded mode and vice versa.
An adaptive protection is defined as an online activity that changes the preferable
response of protection devices according to changing states of the system or its
requirements. Adaptive protection is usually automated, but some necessary human
interventions can also be included. An adaptive relay is a protection device or relay
that includes different setting groups, characteristics, or logic functions that can be
altered or changed online very quickly by using external signals or control com-
mands [25]. The modern numerical relays, also called intelligent electronic devices
(IEDs), not only provide various protection functions (overcurrent, over/under
voltage, etc.) integrated into a single physical device, but also offer various settings
groups for each of the available protection functions. These setting groups can be
changed in an adaptive manner using the communication link between IEDs and
IEDs and circuit breakers (CBs). Adaptivity of protection scheme in a microgrid is
mainly required due to different magnitudes of fault current sensed by OC relays in
grid-connected and islanded modes and due to the connection and disconnection of
DERs. Further detailed discussion on adaptive protection is given in Sect. 5.
540 A. A. Memon et al.

19.3.5 Re-Synchronization

Re-synchronization is the ability of a well-planned microgrid to reconnect back

to the main grid soon after the clearance of faults on the main grid side. The
availability of re-synchronization means at the microgrid point of connection to the
main grid is necessary for the smooth transition of microgrid from the islanded
mode to the grid-connected mode. Re-synchronization is the process of connecting
islanded microgrid back to the main grid after checking or measuring the voltage,
frequency, and phase angle of both systems and closing the breaker contacts for
parallel operation only if these parameters are within acceptable limits as per Table
5 of IEEE Std 1547–2003 [26]. Three types of synchronization schemes have
been mentioned in [27]: active, passive, and open transition synchronization. In
active synchronization, there is a control mechanism that can be used to match
voltage, frequency, and phase angle of an islanded microgrid to the main grid before
closing the breaker contacts. Active synchronization requires collection or sensing
of conditions for both the main grid and islanded microgrid and then communicating
this information to the control mechanism. Passive synchronization uses traditional
synchrocheck relay for closing breaker contacts if the voltage, frequency, and
phase angle of both the main grid and islanded microgrid reach within specified
limits. Passive technique also requires sensing of conditions for both the main grid
and islanded microgrid, however, it is slower than active synchronization. Open
transition synchronization requires the disconnection of loads and DGs inside the
microgrid before reconnection and does not require any sensing or measurement
of conditions. Both active and passive synchronization methods maintain high
reliability of microgrid as no load or DG disconnection is required. The same
procedure of synchronization is applicable for the reconnection of any synchronous
DG, the converter-based DG, or isolated zones with one or more DGs back to the
large portion of the islanded microgrid in islanded mode.

19.3.5.1 LV Microgrid Synchronized Reconnection

One important issue in enabling future Smart LV Grids with island operation
capability is that challenges related to the synchronized reconnection (i.e., re-
synchronization) of island operated microgrid are solved. Islanded microgrid may
be synchronized with the utility system directly after islanding, but later the
synchronism is lost due to generation and load variations inside the microgrid. This
means that the voltage phase angle difference across the microgrid interconnection
switch or circuit-breaker will change. Microgrid re-synchronization or synchronized
reconnection of microgrid means that the voltage angle difference between the
utility grid and microgrid needs to be minimized before reconnection [4].
In the HV network where line reactance X is much larger than line resistance R,
the active power P depends mainly on load angle δ and reactive power Q depends
mainly on voltage difference. This means that the active power P control directly
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 541

controls the load angle δ and frequency f. Generators in the HV network are typically
directly connected with synchronous generators (SGs) which can be controlled, for
example, during synchronization of separate power system areas. In HV network
synchronism check relays have typically such settings that frequency difference over
open CB needs to be less than 55 mHz and phase difference 20◦ –45◦ before CB
connection [28].
On the other hand, in LV microgrids a large share of the DER units are connected
through the inverter- or converter-based interfaces and microgrid synchronized
reconnection can be done by the control of these DER units. In LV networks the
line resistance R is much larger than the line reactance X and therefore, the active
power P depends mainly on voltage difference, while the load angle δ and frequency
is mainly dependent on reactive power Q. This means that one possibility to manage
the phase difference across microgrid interconnection CB could be coordinated
reactive power control of the DER units.
The chosen strategy is dependent on the chosen microgrid concept. In [29],
it was proposed that the control of the grid-forming energy storage unit (master
unit) could slowly shift the microgrid frequency reference closer to the utility grid
frequency before reconnection. However, for example with P/f -droop controlled
DER units re-synchronization requires the coordinated control of all DER units.
This coordination should be done by an external central controller, for example,
microgrid management system or controller, which manages all the DG units during
the synchronization process [30, 31].
Voltage imbalance due to asymmetrical loads and single-phase DG units affects
the voltage phase difference across open microgrid interconnection CB so that the
phase difference deviation can be different in phases A, B, and C. This asymmetry
between phases may also require to be reduced before microgrid re-synchronization.
Active components in the connection point of microgrids, such as microgrid
interconnection switch, central energy storage unit, and microgrid management
system, are responsible for microgrid synchronized reconnection. In [32] it was
proposed that synchronous island operation could be done using a reference
signal with phase and frequency information to the microgrid master unit. The
phase difference before re-synchronization should be within acceptable levels, for
example, less than 60◦ [32]. Based on [33], the microgrid re-synchronizing function
has to meet a more stringent requirement than the one defined by IEEE 1547 which
requires that the phase difference between a microgrid and the utility grid needs to
be smaller than 20◦ before closing the interconnection CB. In [34], LV microgrid
re-synchronization was studied and different synchronized reconnection enabling
functionalities were developed and simulated.
In the simulations [34] either
1. Master unit voltage phase angle or.
2. Reactive power output of DG units was modified to enable synchronized
reconnection and in addition also.
3. Controllable single-phase loads were used in for phase asymmetry compensation
at MV/LV distribution substation.
542 A. A. Memon et al.

The simulation results in [34] showed that both re-synchronization functions

1. Voltage phase angle adjustment by master unit control and.
2. DER unit reactive power feeding can be utilized to enable successful LV
microgrid re-synchronization.
However, the voltage phase difference deviation or asymmetry between phases
A, B, and C across the microgrid interconnection switch still existed with these
re-synchronization functions. If this phase difference deviation is too large, it must
be compensated before LV microgrid re-synchronization. In simulations of [34],
this phase difference deviation was well corrected by the connection of resistive
or capacitive single-phase loads at MV/LV distribution substation. However, quite
large frequency and voltage oscillations after the connection of single-phase
capacitive loads were detected when compared to the connection of purely resistive
loads.
In general, the simulation results of [34] clearly showed that re-synchronization
is not necessarily a significant issue with small, for example, <10◦ , phase difference
across microgrid interconnection CB when only the converter-based DER units
are connected to the microgrid. The reason for this was that the phase-locked-
loop (PLL) component will draw grid-following/−supporting converters into phase
with the utility grid frequency after reconnection. On the other hand, with directly
connected synchronous generators even small phase difference across intercon-
nection CB during re-synchronization was found to be challenging. Therefore,
re-synchronization functions for minimizing phase angle difference and possibly
also voltage unbalance before LV microgrid reconnection will be needed and in
practice, these functions should be coordinated by microgrid central controller or
management system [4].

19.3.6 Circuit Breaker Technology

The circuit breaker (CB) technology for AC power systems has been well-developed
over the years and a variety of CB and switching technologies are available
nowadays. AC circuit breakers range from miniature circuit breakers (MCB),
molded case circuit breakers (MCCB), and air circuit breaker (ACB) for low voltage
levels up to 600 V, these devices include relays and CBs combined in one physical
device. For medium voltage (MV) applications mechanical CBs are available using
mostly vacuum and SF6 as arc extinguishing medium and spring, pneumatic,
hydraulic, and electromagnetic mechanisms for electrode separation to interrupt the
circuit. Most of the MV mechanical circuit breakers operate within 3–5 cycles of
supply voltage after getting trip command from protection relays. However, some
efforts have been made to still decrease the time of MV mechanical circuit breaker
operation by developing new breaker prototypes using a combination of Thomson
coil actuator and permanent magnet actuators which give one cycle interruption for
the largest current duty of T100a, but it is an expensive solution. The other prototype
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 543

using the Thomson coil actuator and spring mechanism reduced the normal trip
time delay of vacuum circuit breaker (VCB) from 15 ms to 13.5–14 ms [35]. AC
solid state circuit breakers (AC-SSCB) provide faster tripping response compared
with mechanical circuit breakers. AC-SSCB can be classified into two broad
categories: Noncurrent-limiting and current limiting types. Noncurrent limiting AC-
SSCBs are constructed using inverse parallel SCRs (Silicon controlled rectifiers)
and current limiting AC-SSCBs are constructed using faster power semiconductors
switches like IGBTs (insulated gate bipolar transistors) and IGCTs (integrated
gate-commutated thyristors). Noncurrent limiting SCR based AC-SSCBs can be
compatible with traditional protection coordination with downstream breakers or
fuses and offer advantages like silent operations, long life, and half-cycle tripping
response. However, for the worst-case asymmetrical fault currents of 200kA peak
for 85kA rms current, SCRs should be protected by fuses. On the other hand,
current limiting AC-SSCBs result in higher power losses than noncurrent limiting
AC-SSCBs and can be problematic for the protection coordination of traditional
systems even if a few of these breakers are connected in the system. Therefore, entire
protection schemes need to be designed based on only current limiting AC-SSCBs,
this may be well suited to microgrids with only the converter-based DERs. Current
limiting AC-SSCBs have many design challenges like employing suitable snubbers,
metal-oxide varistors (MOVs), and dynamic braking resistors for energy dissipation
and voltage limiting during interruption as well as the capability to differentiate
between fault current and load transients to avoid false trips [36].

19.4 Conventional Protections for AC Microgrids

Conventional protection of microgrids is usually based on the overcurrent principle

using either definite time or inverse definite OC relays. In addition, voltage-based
(over/under voltage) and frequency-based (over/under frequency) protections are
also used for the protection of DERs, for detection of islanding situation, or load-
frequency control in microgrids. However, only overcurrent protection schemes are
discussed in this section.

19.4.1 Overcurrent Protection (Definite-Time Versus Inverse

Time)

The most common protection scheme used in traditional distribution systems is

based on the overcurrent principle. Both definite-time and inverse-time (IDMT) OC
relays are used in a coordinated manner so that interruption should be limited to
only the faulty section. For both static and microprocessor-based relays the working
time of OC relays is decided based on two details: CB opening time (0.04–0.1 s) and
544 A. A. Memon et al.

security factor (0.12–0.22 s). In this way, total time of OC relay operations is in the
range of 0.2–0.4 s [18]. For definite-time, OC relays only pickup current (IP ), CB
opening time, and security factors are required to decide the total operating time of
primary and backup relays. However, for IDMT OC relays the relay operating time
is decided based on different relay characteristics defined by Eq. 19.1 according to
IEC 255–3 [37]:

ka
t= b (19.1)
I
IP −1

where t is the relay operating time, k is the time dial setting (TDS), I is the current
detected by relay, IP is the pickup current and a, b are the constants defining relay
characteristics.
Mainly two sets of standard inverse-time relay characteristics are applied, IEC
and ANSI (in the USA). Additionally, relays may include options to have some
manufacturer specific characteristics or even freely programmable characteristics. It
means based on the types of characteristics curve (inverse, very inverse, extremely
inverse, etc.) different relay manufacturers may have different/additional constants
for inverse-time OC relay characteristics [38] in addition to the standard IEC
or ANSI constants. Generally, the settings of inverse-time OC relays should be
adjusted so that there is always some coordination time interval (CTI) between
primary and backup relays to ensure correct selective operation. Figure 19.3 shows
the time current characteristic curves for primary and backup inverse-time OC
relays. For fault current of magnitude A1 (Fig. 19.3) primary relay will operate
first at time t1 and if the primary relay fails to operate than the backup relay will
operate at time t2 which is equal to t1 + CTI. If adequate CTI (0.2–0.3 s) is not
ensured, then nuisance trips may happen. For a given current, the higher the TDS is
the greater the time to tripping contact closure is. Since TDS provides up and down
adjustment of inverse-time OC relay characteristics and pickup current provides left
and right adjustment of characteristics on the coordination plots, therefore proper
CTI can be established through proper selection of TDS and IP values [18].
The coordination practice of using definite-time characteristics or inverse-time
characteristics vary between Europe/Japan and North America. In Europe and
Japan, it is common that primary distribution systems are operated as impedance
grounded or ungrounded three-wire systems and hence no presence of single-phase
laterals protected by fuses. Therefore, relay coordination can be achieved using
definite-time characteristics. In North America, the usual practice is to operate
grounded four-wire distribution systems with loads served by single-phase laterals
protected by fuses. Therefore, coordination is done using inverse-time current
characteristics which is suitable for fuse coordination [39]. However, in this section,
both definite-time and inverse-time coordination characteristics are used and the
performance of both is compared for traditional three-phase distribution system
operation without DERs, for the grid-connected operation of AC microgrid with
DERs, and islanded mode operation of AC microgrid with DERs and central battery
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 545

Fig. 19.3 Time current

characteristics curve of
primary and backup
inverse-time OC relays

energy storage system (BESS). The single-line diagrams of the grid-connected AC

microgrid (Fig. 19.4) and islanded AC microgrid (Fig. 19.5) are used to verify the
coordination of OC relays at CB1, CB2, CB6, CB7, and CB8 locations for three-
phase short-circuit faults F1, F6, and F8 as explained in the following subsections.
The circuit breakers CB1-CB9 in Fig. 19.4 are operated by corresponding
OC relays (OCR1–9) for faults in corresponding zones (zone1–9), and therefore
OC relay-1 (OCR-1) means the relay at CB1 location and so on. Definite-time
coordination starts from the relay OCR-8 at the extreme load end which is set to
trip instantaneously after 0.04 s of the fault. OCR-7 is set to trip after 0.2 s of the
fault, OCR-6 is set to trip after 0.4 s of the fault, OCR-2 is set to trip after 0.6 s of the
fault, and OCR-1 is set to trip after 0.8 s of the fault. The pickup current setting of
each relay (OCR1–2 and OCR6–8) is set at 2.25 p.u. of maximum current a relay can
experience during normal operation without DERs (wind turbine and photovoltaic).
For inverse-time current coordination, the same type of relay characteristic is chosen
for all inverse-time OCRs (inverse-time OCR1–2, OCR6–8), which is a very inverse
characteristic with constants of relays a = 13.5 and b = 1 (Eq. (19.1)). The selected
values of TDS for inverse-time OCRs are 0.05, 0.15, 0.25, 0.28, and 0.29 for OCR8,
OCR7, OCR6, OCR2, and OCR1, respectively, to ensure proper coordination of
relays. The pickup current for each inverse-time OCR is chosen to be the same as
chosen for definite-time OCRs which is 2.25 p.u. of maximum current a relay can
experience during normal operation without DERs. The performance of definite-
time and inverse-time OCRs with regard to the speed of operation and selectivity
for three different operational modes is analyzed in the following subsections. It
is worth here mentioning that the settings of both definite-time and inverse-time
OCRs are kept the same in all three modes to observe the issues in different modes
when single-settings OCRs are used. The real-time modeling of relays and other
components is done by using Matlab/Simulink and RT-Lab software of OPAL-RT.
546 A. A. Memon et al.

Fig. 19.4 Grid-connected mode of AC Microgrid with single setting OC relay protection

19.4.1.1 Protection Coordination in a Grid-Connected Mode without

DERs

In this subsection, the protection coordination of definite-time OCRs and inverse-

time OCRs with very inverse characteristic is analyzed for the grid-connected mode
of AC microgrid without DERs. It means, in this subsection the coordination of
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 547

Fig. 19.5 Islanded mode of AC Microgrid with single setting OC relay protection

OCRs in path CB1–2–6-7-8 is observed without considering green zones in Fig.

19.4. This makes the system of Fig. 19.4 a traditional MV/LV distribution system
protected by OCRs. A three-phase permanent short-circuit fault is applied separately
at locations F1, F6, and F8 (RF1 = RF6 = 5.001 Ohm, RF8 = 0.001 Ohm) for
a duration of 18 s (starting from 2 s to 20 s in simulations) and tripping time of
primary and backup OCRs was noted for both very inverse-time and definite-time
characteristics. During fault F1 only the tripping time of primary protection can
be observed. Table 19.1 presents the tripping time durations (top ) for primary and
backup relays for the faults F1, F6, and F8. The CTI values in Table 19.1 (given
in brackets) in columns 4 and 5 indicate the coordination time intervals between
primary and backup-1 OCRs, between backup-1 and backup-2 OCRs, and between
backup-2 and backup-3 OCRs. Also, fault currents in per unit (p.u.) of rated current
at corresponding OCRs in every fault case is given. During three-phase short-circuit
fault F1, the tripping of primary inverse-time OCR1 is much faster than definite-time
OCR1 because of inverse-time OCR1 trips after 0.6843 s of the fault F1 whereas
definite-time OCR1 trips after 0.8 s of the fault F1. During fault F6 also, primary
inverse-time OCR6 trips faster in 0.0732 s compared with primary definite-time
OCR6 which trips in 0.4 s. To observe the tripping times of backup-1 OCR2 and
backup-2 OCR1 protections of primary OCR6 during fault F6, the tripping of CB6
548 A. A. Memon et al.

and CB2 was disabled to simulate breaker failures. It is observed that backup-1
inverse-time OCR2 and backup-2 inverse-time OCR1 of inverse-time OCR6 also
trip faster than backup-1 definite-time OCR2 and backup-2 definite-time OCR1 of
definite-time OCR6 while maintaining a good CTI of 0.2638 s between primary
and backup-1 and CTI of 0.428 s between backup-1 and backup-2 during fault
F6. However, during fault F8 not only the primary inverse-time OCR8 is slower
than the primary definite-time OCR8, but also all backup inverse-time OCRs are
slower than backup definite-time OCRs. The CTI between primary inverse-time
OCR8 and backup-1 inverse-time OCR7 is 0.2757 s and CTI between backup-1
inverse-time OCR7 and backup-2 inverse-time OCR6 is 0.3062 s, which are longer
than the corresponding CTIs of 0.16 s and 0.2 s of definite-time OCRs during
fault F8. It can be observed from Table 19.1, that backup-3 inverse-time OCR2
operates considerably slower (top = 6.9592 s) than backup-3 definite-time OCR2
(top = 0.6 s) due to reduced fault current of 3.45 p.u. at OCR2 during the fault
F8. It is also observed that due to very low fault current (1.65 p.u.) at OCR1
during 18 s of fault F8, neither inverse-time OCR1 nor definite-time OCR1 provides
backup-4 for OCR8 due to higher pickup current settings. Therefore, during fault
F8 limited selective coordination only up to OCR2 is possible. In conclusion for
this case, inverse-time OCRs provide faster tripping times while maintaining good
coordination between primary and backup relays during faults F1 and F6. However,
during fault F8, inverse-time OCRs are much slower and provide lesser coordination
compared with definite-time OCRs during fault F8. So during fault F8, only definite-
time OCRs provide faster tripping times and good selective coordination.

19.4.1.2 Protection Coordination in Grid-Connected Mode with DERs

In this subsection, the protection coordination of definite-time OCRs and inverse-

time OCRs with very inverse characteristic is analyzed for the grid-connected
mode of AC microgrid with the converter-based DERs. It means in this section
the coordination of OCRs in path CB1–2–6-7-8 is observed with green zones of
DERs (Zones 3,4, and 9) also included (Fig. 19.4). This makes the section below
PCC breaker CB2 in Fig. 19.4 as a grid-connected AC microgrid with DERs
(when CB1–2 closed) where a wind turbine DER is connected at 20 kV MV bus-
2 and photovoltaic DER is connected at 0.4 kV LV bus. The main purpose of this
case study is to observe how the speed and selectivity of inverse-time OCRs are
affected after the connection of DERs. It should be noted here that both DERs
(wind turbine and photovoltaic systems) provide FRT capability, which means no
trip until the fault is cleared and provide 1.2 p.u. of rated current during faults
when voltage is below 0.5 p.u. Fault types and durations, in this case, are the
same as mentioned in previous Sect. 4.1.1, and the performance of inverse-time
and definite-time OCRs is observed in terms of speed and selectivity. In addition to
tripping times, CTI and fault currents of OCRs, DERs fault contribution, and DERs
terminal voltage at connection point are also noted in this case (Table 19.2). During
fault F1 primary inverse-time OCR1 operates faster at top = 0.7068 s than primary
Table 19.1 Protection coordination in a grid-connected mode without DERs (No green zones, 3, 4, and 9)
Fault (duration) Relay No. Primary/Backup Very inverse-time top - (CTI) Definite-time top (CTI) Fault current at relay (p.u.)
F1 (18 s) OCR1 Primary 0.6843 0.8 19.9
F6 (18 s) OCR6 Primary 0.0732 0.4 159
OCR2 Backup-1 0.337 (CTI6–2 = 0.2638 s) 0.6 (CTI6–2 = 0.2 s) 27.48
OCR1 Backup-2 0.765 (CTI2–1 = 0.428 s) 0.8 (CTI2–1 = 0.2 s) 18.11
F8 (18 s) OCR8 Primary 0.1368 0.04 19.72
OCR7 Backup-1 0.4125 (CTI8–7 = 0.2757 s) 0.2 (CTI8–7 = 0.16 s) 19.72
OCR6 Backup-2 0.7187 (CTI7–6 = 0.3062 s) 0.4 (CTI7–6 = 0.2 s) 19.05
OCR2 Backup-3 6.9592 (CTI6–2 = 6.2405 s) 0.6 (CTI6–2 = 0.2 s) 3.45
OCR1 Backup-4 No trip in 18 s No trip in 18 s 1.65
19 Microgrid Protection with Conventional and Adaptive Protection Schemes
549
550 A. A. Memon et al.

definite-time OCR1 which operates at top = 0.8 s. However, due to reduced fault
current experienced by inverse-time OCR1 in this case, its tripping time is 22.5 ms
slower than it was in the previous case without DERs (Sect. 4.1.1). During fault
F6, primary inverse-time OCR6 operates at the same time (top = 0.0732 s) as it
operated during the same fault in the previous case without DERs (Sect. 4.1.1) even
if the fault current at OCR6 has been reduced by 10 p.u.(from 159 p.u. to 149 p.u.)
after the connection of DERs. However, backup-1 inverse-time OCR2 and backup-2
inverse-time OCR-1 are observed with 11 ms and 26.7 ms slower tripping responses
respectively during fault F6 compared with the same fault in the previous case
without DERs (Sect. 4.1.1) due to comparatively reduced fault currents. However,
still faster tripping times and good CTI is observed for primary and backup inverse-
time OCRs compared with definite-time OCRs during faults F1 and F6. During
fault F8, primary inverse-time OCR8 operates slower than primary definite-time
OCR8, but its tripping time is faster compared with inverse-time OCR8 in the
previous case without DERs (Sect. 4.1.1) due to increased fault current observed
at OCR8 because some fault current contribution comes from photovoltaic DER
in this case. Whereas backup-1 inverse-time OCR7, backup-2 inverse-time OCR6,
and backup-3 inverse-time OCR2 are slower compared with backup-1 definite-time
OCR7, backup-2 definite-time OCR6, and backup-3 definite-time OCR3 during
fault F8 in this case. However, backup-1 inverse-time OCR7 and backup-2 inverse-
time OCR6 have nearly the same tripping times during fault F8 in this case (Table
19.2) compared with tripping times during the same fault in the previous case
without DERs (Table 19.1) even if the fault current seen at both OCRs is somehow
reduced in this case after DER connections (Table 19.2). The tripping time of
backup-3 inverse-time OCR2 during fault F8 has also increased from top = 6.9592 s
in the previous case without DERs (Table 19.1) to top = 8.5675 s in this case
due to comparatively reduced fault current seen by OCR2 (Table 19.2). It is also
observed that due to a very low fault current of 1.0851 p.u. at OCR1 during
fault F8 (Table 19.2), neither inverse-time OCR1 nor definite-time OCR1 provides
backup-4 for OCR8 during fault F8 in this case with DER connections. Therefore,
like the previous case (Sect. 4.1.1) in this case also limited selective coordination
only up to OCR2 is possible during fault F8. In conclusion for this case, inverse-
time OCRs provide faster tripping times compared with definite-time OCRs while
maintaining good coordination between primary and backup relays during faults F1
and F6. However, during fault F8, inverse-time OCRs are much slower and provide
lesser coordination compared with definite-time OCRs during fault F8. So during
fault F8, only definite-time OCRs provide faster tripping times and good selective
coordination. Comparing this case with DERs to the previous case without DERs, it
can be said that the inclusion of DERs somehow makes inverse-time OCRs slower
due to reduced fault current contribution from the main grid. Some inverse-time
relays, for example, OCR8 sense increased fault current due to DER connection and
operate faster as compared to their operation without DERs. Definite-time OCRs
show the same performance for the grid-connected mode with DERs and without
DERs.
Table 19.2 Protection coordination in grid-connected mode with DERs (Green zones included)
Primary/Backup Fault current at
Fault (duration) Relay No. zone Very inverse-time top (CTI) Definite-time top (CTI) relay (p.u.) DERs fault contribution
F1 (18 s) OCR1 Primary 0.7068 0.8 14.6 IF, pv, wtg ≤ 1.0 p.u.
(Upv,wtg = 0.9 p.u.)
F6 (18 s) OCR6 Primary 0.0732 0.4 149 IF,wtg ≤ 1.04 p.u.
IF,pv ≤ 1.01 p.u.
(Uwtg = 0.872 p.u.)
(Upv = 0.837 p.u.)
OCR2 Backup-1 0.348 (CTI6–2 = 0.2748 s) 0.6 (CTI6–2 = 0.2) 26.74
OCR1 Backup-2 0.7917 (CTI2–1 = 0.4437 s) 0.8 (CTI2–1 = 0.2) 13.31
F8 (18 s) OCR8 Primary 0.1277 0.04 20.89 IF,pv = 1.17 p.u.,
(Upv < 0.5 p.u.)
IF,wtg ≤ 1.0 p.u.
(Uwtg ≤ 1.0 p.u.)
OCR7 Backup-1 0.412 (CTI8–7 = 0.2843 s) 0.2 (CTI8–7 = 0.16 s) 18.095
OCR6 Backup-2 0.7175 (CTI7–6 = 0.3055 s) 0.4 (CTI7–6 = 0.2 s) 17.806
OCR2 Backup-3 8.5675 (CTI6–2 = 7.85 s) 0.6 (CTI6–2 = 0.2 s) 3.215
OCR1 Backup-4 No trip in 18 s No trip in 18 s 1.0851
19 Microgrid Protection with Conventional and Adaptive Protection Schemes
551
552 A. A. Memon et al.

19.4.1.3 Protection Coordination in Islanded Mode with DERs and BESS

In this subsection, the protection coordination of definite-time OCRs and inverse-

time OCRs with very inverse characteristic is analyzed for the islanded mode of
AC microgrid with DERs and BESS (battery energy storage system). In this case,
islanding is created by opening CB2 assuming that fault F1 had happened earlier
and OCR1 had tripped CB1 and transfer tripped CB2 within the next 0.5–0.8 s.
Then 2.4 MW BESS has been activated quickly as a grid-forming DER after closing
CB10. An islanded AC microgrid with BESS, DERs, and single-settings OCRs is
shown in Fig. 19.5. It is assumed that BESS is provided with a converter of higher
current rating (≥ 3 p.u.) and BESS can act as fault current source during any fault
in islanded mode. Such battery converters are installed in Bronsbergen microgrid,
Netherlands that have a continuous current of 290 A and provide a fault current of
1100 A [40]. Such fault current sources can provide fault current up to 5 s [41].
A fault current source using energy storage devices like a flywheel, a battery, or
ultracapacitor is proposed in [40, 41] which is a modular system and multiple units
can be connected in parallel to suit the local fault level. In this case, it is studied
that if such type of fault current source is provided at PCC then how it serves the
purpose of achieving faster tripping times and maintaining protection coordination
in the islanded mode.
In this subsection, the coordination of OCRs in path CB6–7-8 is observed with
the green zones of DERs (Zone 3,4 and 9) and BESS zone (Zone10) included (Fig.
19.5). In this case, the performance of OCRs has been observed for three-phase
permanent short-circuit faults F6 and F8 (RF6 = 5.001 Ohm, RF8 = 0.001 Ohm)
applied separately for a duration of 14 s (starting from 6 s to 20 s) and relay tripping
times are noted for different primary and backup OCRs in path CB6–7-8. All DERs
and BESS have FRT capability during faults, and provide fault current until the
fault is cleared. During fault F6 in islanded mode, primary inverse-time OCR6
trips after 0.6163 s of fault which is 0.2163 s slower than primary definite-time
OCR-6 which trips after 0.4 s of the fault F6. The OCR6 experiences 20.3 p.u. of
fault current during fault F6 and all DERs including BESS provide maximum fault
current contribution during this fault (Table 19.3). As compared to trip times in the
previous two grid-connected mode cases (Sect. 4.1.1 and 4.1.2), the tripping time of
primary inverse-time OCR6 during fault F6, in this case, is 0.5431 s slower than the
tripping time of the same inverse-time OCR6 in the previous two grid-connected
cases (Table 19.1 and 19.2). However, the tripping time of primary definite-time
OCR6 during fault F6, in this case, is the same as in the previous two grid-connected
mode cases. During fault F8 in islanded mode, primary inverse-time OCR8 trips at
0.1977 s, which is much slower than the tripping time of 0.04 s of primary definite-
time OCR8. It means primary inverse-time OCR8 is 0.1577 s slower than primary
definite-time OCR8 during fault F8 in islanded mode. Compared with trip times of
primary inverse-time OCR8 in previous grid-connected mode cases (Sect. 4.1.1 and
4.1.2) tripping time of primary inverse-time OCR8 is 0.061–0.07 s slower during
fault F8 in islanded mode. The tripping times of backup-1 inverse-time OCR7
and backup-2 inverse-time OCR6 are also slower than tripping times of backup-
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 553

1 definite-time OCR7 and backup-2 definite-time OCR6, respectively during fault

F8 in this case. Compared with the previous grid-connected case without DERs
(Sect. 4.1.1), the tripping times of backup-1 inverse-time OCR7 and backup-2
inverse-time OCR6 are 0.3485–0.6423 s slower during fault F8 in islanded mode.
When compared with the previous grid-connected case with DERs (Sect. 4.1.2), the
tripping times of backup-1 inverse-time OCR7 and backup-2 inverse-time OCR6 in
islanded mode are also 0.349–0.6435 s slower during fault F8 in islanded mode.
In conclusion to this islanded mode case, primary and backup definite-time OCRs
provide faster tripping times and maintain good protection coordination during
faults F6 and F8 compared with primary and backup inverse-time OCRs during
the same faults. The primary inverse-time OCR8 tripping time during fault F8 in
islanded mode is acceptable. However, tripping times of primary inverse-time OCR6
during fault F6, and backup-1 inverse-time OCR7 and backup-2 inverse-time OCR6
during fault F8 in islanded mode are quite slow. Therefore, slower fault clearance at
LV load with inverse-time OCRs can potentially damage the load if a fault current
source is connected in islanded mode. Hence, for the islanded mode of operation
definite-time OCRs perform better with faster tripping times and good CTI between
primary and backup protection as compared to inverse-time OCRs which become
slower to operate with increased CTI between primary and backup protection.
It should be noted here that in both the grid-connected mode with DERs and the
islanded mode of AC microgrid (cases 4.1.2 and 4.1.3) during fault F6 if primary
OCR6 trips CB6 then it should transfer trip CB7 for complete F6 fault clearance,
otherwise photovoltaic DER may still feed the fault in the reverse direction. The
second method is to trip photovoltaic DER after fault F6 is cleared by OCR6 by
anti-islanding protection. The conventional method is to trip all DERs immediately
after the fault is detected inside the microgrid. This conventional method reduces the
reliability of supply to the load. The other case of islanding can be the creation of
islands within islanded mode after fault F6 is cleared by tripping CB6 and CB7. In
this case, both LV bus and MV bus-2 sections will be isolated, and in this case, it will
be very difficult to clear the fault F8 in the islanded LV bus section with only fault
contribution from photovoltaic DER. In this situation, lower adaptive trip settings
will be required even for definite-time OCR8. Adaptive protection of microgrid is
discussed in Sect. 5.

19.5 Adaptive Protection for AC Microgrids

As discussed in previous sections microgrids can operate in both the grid-connected

and islanded mode. For the grid-connected mode, sufficient fault current is available
from the main grid in order to trip OCRs in case of any fault inside the microgrid.
However, in the islanded mode of operation, only limited fault current contribution
from local DERs is available which is not sufficient for the proper operation of
OCRs inside a microgrid. The reduction of fault current contributions from DERs
in islanded mode requires lower pickup currents for OCRs. If the OCRs designed
 554

Table 19.3 Protection coordination in islanded mode with DERs (Green zones included) and BESS as grid-forming source
Primary/Backup Fault current DERs fault
Fault (duration) Relay No. zone Very inverse-time top (CTI) Definite-time top (CTI) at relay (p.u.) contribution
F6 (14 s) OCR6 Primary 0.6163 0.4 20.3 IF,wtg = 1.2 p.u.,
IF,pv = 1.17 p.u.
Upv, wtg < 0.12 p.u.
IF,BESS = 3.17 p.u.
UBESS < 0.12 p.u.
F8 (14 s) OCR8 Primary 0.1977 0.04 13.45 IF,wtg = 1.2 p.u.
IF,pv = 1.17 p.u.,
Upv < 0.1 p.u.,
Uwtg < 0.45 p.u.
IF,BESS = 1.15p.u.,
UBESS ≤ 0.5 p.u.
OCR7 Backup-1 0.761(CTI8–7 = 0.563 s) 0.2 (CTI8–7 = 0.16 s) 11.25
OCR6 Backup-2 1.361 (CTI7–6 = 0.6 s) 0.4 (CTI7–6 = 0.2 s) 11.12
A. A. Memon et al.
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 555

with lower pickup values are used in both the grid-connected and islanded modes,
then false operations will increase, and OCRs may even trip during switching
transients and momentary overloads in the grid-connected mode. To overcome
this OCR sensitivity problem two general solutions are available: (1) Connect
some high fault current providing DERs inside the microgrid in islanded mode
and use OCRs with only grid-connected mode pickup settings in both modes (2)
Use two different OCRs with different pickup settings, one for the grid-connected
and other for the islanded mode. First solutions can be useful to some degree if a
large number of synchronous generators are available inside a microgrid. If a large
number of converter-based DERs inside microgrid are available, then the second
solution is the most favorite choice. However, to implement different OCRs with
different settings for the grid-connected and islanded modes, OCRs will need to
be alternatively activated and deactivated or their logic or settings changed after
knowing the present mode of operation for the correct detection of the faults and
avoiding false operations. The information about microgrid mode is usually in
terms of open or close status signals from circuit breakers, particularly breakers
at the connection point of the main grid and microgrid. The adaptive protection
can be implemented with a central controller or in an autonomous/decentralized
manner after getting microgrid mode information. It is obvious that communication-
based adaptive protections will be quick and dynamic, however, delays in data
transmission, reception, and processing and failures of communication links also
require noncommunication based adaptive protections.

19.5.1 Communication-Based Adaptive Protection

Communication-based adaptive protections can be broadly classified into two cat-

egories: (1) Centralized adaptive protections (2) Decentralized (autonomous) adap-
tive protections. Centralized and decentralized adaptive protections are explained
separately in the following subsections.

19.5.1.1 Centralized Adaptive Protection

The centralized communication-based adaptive protection is the conventional

approach in which centralized control architecture is used. In this method, a central
controller adapts the protection settings in a coordinated manner after collecting
information about the system changes at one central location. This method requires
a very powerful central controller that can save, process, and communicate data with
other components or systems. The centralized controller reduces the computation
burden on individual devices, and therefore, simple devices can be installed for
protection and control purposes. However, this method has one drawback that if
the central controller fails, it causes a complete loss of adaptive protection, and
hence a redundant central controller is a must for increased reliability. Various
556 A. A. Memon et al.

communication protocols like Modbus, DNP3, IEC 60870–5-101/104, IEC 61850

support the centralized communication architecture, and its implementation can be
done using serial/bus communication, over PLC (power-line carrier) or via Ethernet
network [42].

19.5.1.2 Decentralized Adaptive Protection

The decentralized communication-based adaptive protection uses distributed intel-

ligent electronic devices (IEDs) or multi-agent systems. Each IED acts in an
autonomous way after getting information from other IED to modify its active set-
ting group. The decentralized adaptive protection reduces the burden on the central
controller, however, IEDs with increased data storage capacity, more computational
and processing power will be needed in addition to fast and reliable communication
links between IEDs. The feasibility of decentralized architecture is only possible
if direct communication between IEDs is allowed by the communication protocol.
Presently, the focus of the industry is mainly on IEC 61850 as the standard protocol
for horizontal communication between IEDs. Although a bus or Ethernet network is
required to implement the decentralized architecture, but 4G/5G wireless network
or PLC can be the other useful options for its implementation [42]. The IEC
61850 communication standard is getting popular for application in electric power
substation automation. The IEC 61850 standard offers the possibility that IEDs of
different manufacturers can interoperate or interact with each other. This standard
is initially focused on communication between IEDs within a single substation but
can be extended for communication between various substations in the future. The
IEC 61850 standard provides a set of standard model structures for data and rules
defining how to exchange these data. The IEDs from different manufacturers that
comply with these model definitions can then communicate, understand, and interact
with each other [43]. The IEC 61850 communication standard, by one common
protocol, enables the integration of several functions including protection, control,
measurement, and monitoring, and so on [44].

19.5.2 Review of Communication-Based Protection Schemes

for AC Microgrids

In the literature, several solutions have been suggested for the protection of islanded
microgrids. Some main solutions are reviewed in the following.
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 557

19.5.2.1 Directional Overcurrent Protection

Directional overcurrent protection can be used to protect the distribution system

with bidirectional current flow. One problem with the overcurrent protection is
that faults closer to the source might take a longer time to clear. However, this
problem can be overcome with modern microprocessor-based devices, by having
shorter coordination margin, and instantaneous overcurrent protection. Still, there
could be protection co-ordination issues because of changing fault current due
to DER or change in the network configuration. An advanced solution could be
adaptive directional overcurrent protection based on the status of the generators and
networks’ topology.

Adaptive Directional Overcurrent Protection

In adaptive protection, relays still carry out the protection function using local mea-
surements. However, their settings are updated locally or remotely via communica-
tion links. Apart from the additional voltage transformer (VT) and microprocessor-
based relays, this scheme may also require a central controller and a communication
link to the relays; but this communication does not need to be very fast nor very
reliable.
In the solution of [45], relays detect the islanding condition themselves and
reconfigure their neighbors if they detect a topology change [46]. Paper [45]
suggests local information utilizing adaptive protection to handle the overcurrent
protection challenges in the distribution network with DG. The settings of the relays
are updated based on monitoring the operating states (grid-connected or island) and
detected faulted section. In [45] it is proposed that the faulted section detection
could be done by using time overcurrent characteristics of the protective relays.
The adaptive feature was also proposed in [3] for the protection of microgrids;
where directional overcurrent relays are reconfigured by a remote central unit in
case of grid separation (islanding) or grid reconnection to consider for the change of
short-circuit level. The central unit is constantly aware of the network topology and
the connected generators [46]. Paper [47] has also presented an adaptive protection
scheme for distribution networks which consists of multiple DG units. According to
real-time monitoring of the state and topology of the distribution network, the active
setting groups (Fig. 19.6) of the protection relays are adjusted. A programmable
logic application is called at the central controller to perform adjustments after
changes in circuit breaker (CB) status (Fig. 19.6). A controller which is centralized
is integrated into the substation with IEC 61850 based communications to execute
the control functions. The programmable logic in this central controller (substation
computer or RTU) is used to modify the settings with IEC 61131–3 compatible
programming languages. The adaptive protection scheme presented in paper [47]
together with the microgrid control system has also been developed and adapted for
a real demonstration at Hailuoto island in Finland [48].
558 A. A. Memon et al.

Fig. 19.6 Adaptive protection by the change of IED setting group

Based on the literature, it has been stated in [49] that the major challenges related
to the implementation of an adaptive protection system can be:
• The requirement for communication infrastructure.
• The requirement for previous knowledge about all possible microgrid configura-
tions.
• The requirement to update or upgrade a large amount of protection devices (e.g.,
fuses) which are currently used in the distribution networks (especially on LV
network).

19.5.2.2 Current Symmetrical Components Based Protection

Protection scheme proposals based on current symmetrical components try to

enhance the performance of traditional overcurrent protection [49]. For example,
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 559

in [50], it was suggested that fault detection in islanded microgrids could be based
on current symmetrical components. In [50] it was proposed to utilize in the event of
an upstream phase-to-earth fault zero-sequence current detection (coordinated with
unbalanced loads) and negative sequence current for phase-to-phase faults. In [51], a
pilot instantaneous overcurrent protection scheme was briefly presented to perform
instantaneous protection of a local line and remote bus-bar independent of the
DG unit location. In [52], a three-level, communication-based protection selectivity
scheme was proposed to be applied with voltage-restrained directional overcurrent
protection. In [53], a fault current magnitude independent strategy for protection
in LV microgrids with microprocessor-based overcurrent relays and directional
elements was suggested considering both operation modes (grid-connected and
islanded).

19.5.2.3 Distance Protection

Distance protection uses measured impedance or admittance values to detect the

fault. In reference [54], it has been stated that distance protection seems to be a
potential protection scheme for island operated MV networks. On the other hand,
papers [55–57] have proposed directional inverse time-based admittance protection
(Inverse Time Admittance, ITA), for grid-connected and islanded operation. This
means that it can operate both in forward and reverse faults, but the reach settings
should be different in forward and reverse directions. In the paper [13], comparison
of distance and directional overcurrent protection was done considering only 3-
phase short circuit faults. It was concluded that distance protection could selectively
separate faults in MV and LV networks. Further, based on the simulations and the
analysis, it was stated that overcurrent protection cannot be used if fault current
levels within the island are close to the maximum load current. Based on this
distance protection could be more suitable for being used in island operation.
However, when two-phase and earth faults are considered the viability of distance
protection must be further studied [13].

19.5.2.4 Voltage Based Protection Schemes

Due to a lack of high fault currents, it has been proposed by [58, 59] that voltages
could be used for the protection of an islanded microgrid. For example, paper
[59] proposed a method that was based on a voltage measurement comparison:
the location where the lowest voltage level is measured is tripped [46]. However,
it is difficult to realize selective microgrid protection during island operation with
voltage or current relays alone [60]. In [9] the same authors as in [59] also proposed
to use the total harmonic distortion (THD) of voltage to enhance the protection
of microgrids with the inverter-based DG units during earth faults. After detection
fault type based on the variation of the fundamental frequency (50 Hz), the voltage
THD seen by different feeder protection relays was analyzed to determine the
560 A. A. Memon et al.

faulted part of the network. To prevent the challenges of previous methods related
to detecting the oscillation waveform of the voltage variation, instead of voltage
magnitude, [61] suggested using only the positive sequence voltage. In [62], it was
stated that a distinction between three fault types is possible by only considering the
positive/direct and negative/inverse sequence voltage components, without using the
zero sequence/homopolar information. Reference [63] proposed the same type of
method to detect the fault and faulted network part, based on a voltage measurement
at the busbar and its transformation from abc coordinates to dq coordinates. In [64],
the reduction in system voltage has been also used to implement an under-voltage
backup protection scheme for current differential protection.
Main problems with voltage-based methodologies during island operation, based
on [49], are:
• Small differences in voltage drop seen by the relays at both ends of short lines
can lead to protection maloperation, due to reduced voltage gradient.
• Challenges with the practical application of some of these methods and also with
the potentially required communication system, when a large amount of DG units
are present.
• Solution can be highly dependent on the network architecture as well as on the
defined relay protection zone related to every DG unit.
• Challenges in high-impedance-fault (HIF) detection.

19.5.2.5 Current Differential

In [64], line differential protection based on current measurements was chosen for
the microgrid because it is nonsensitive to bidirectional power flows, changing fault
current levels, and the amount of DG units. It is also stated that current differential
protection offers the needed protection during both grid-connected and islanded
operation and it is not affected by a weak infeed. This means that current line
differential protection can detect internal faults even without having any DG units
connected. The use of differential protection for microgrids with low fault current
levels has been suggested also in [65–69] to protect inverter-dominated microgrids.
However, differential protection might be expensive since protective devices must
be placed on every line segment of the network. Therefore, [69] proposes to form
protection zones consisting of several line segments.
In addition, it is worth mentioning that the topology of the sample network, that
is, a schematic diagram in [64] is chosen to be very well suited for differential
protection also during island operation. In general, current differential protection is
not very suitable for the protection of the islanded part of the distribution network
having a radial topology and many protection zones (i.e., for protection of “last”
protection zones with an open-end/CB).
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 561

19.5.2.6 Protection Based on Voltage and Directional Overcurrent

In [62], a microgrid protection strategy based on voltage and current measurements

was proposed in addition to a voltage-based protection scheme. Also, in [70, 71],
the use of voltage measurement-based fault detection has been considered and a
potential solution for small microgrids is presented in the form of voltage-controlled
overcurrent devices to enable the use of lower current threshold settings.
In addition, in references [54, 4, 11, 72, 73] microgrid protection schemes based
on the use of both voltage and directional overcurrent are analyzed. However, these
schemes are based on the utilization of high-speed communication for interlock-
ing/blocking purposes to ensure selective operation of protection during island
operation. In the following, these proposed schemes for LV and MV microgrids
are presented.

19.5.3 Protection of LV AC Microgrids

Different kinds of protection methods and principles for microgrids have been
proposed. One problem in some proposed solutions for LV microgrid protection
is that their applicability is limited to microgrids with only converter-based DG
units. Therefore, these solutions may overlook others, for example, requirements on
the operational speed of protection to maintain the stability of LV microgrid with
directly connected rotating machines after fault clearance. According to [4] key
fundamental properties required from the future LV microgrid protection systems
include,
1. Adaptivity,
2. Utilization of fast standard-based communication (IEC 61850),
3. Fast operation in deep voltage dips due to faults to maintain stability in the
healthy part of LV microgrid,
4. Fast operation to fulfill the needs of very sensitive customers,
5. Selective operation in faults of every kind and.
6. Unnecessary operation of protection devices (PDs) and disconnection of DER
units must be avoided.
In the following, two different proposed protection schemes for LV microgrids
based on the utilization of both voltage and current (with direction detection) are
shortly presented and more details can be found from the references. The main
difference of the proposed protection schemes is that the first one (Proposed LV
Scheme 1) relies on the extensive use of high-speed communication and the other
(Proposed LV Scheme 2) is not based on the use of communication.
562 A. A. Memon et al.

19.5.3.1 Proposed LV Microgrid Protection Scheme 1

In [4, 11, 72, 73], the following scheme for the protection of LV microgrids has
been proposed. The main structural choices of the proposed LV microgrid protection
system is summarized in Fig. 19.7 showing the type of protection devices (PD 1–
4). When the measurements of active and reactive power flow between the utility
grid and the LV microgrid are needed during normal operation, then also current
measurements need to be included in PD 1. However, from the proposed protection
system point of view, the current measurement at PD 1 is not necessarily needed.
Properties of the examined LV microgrid, for example, type, number, and location
of fault current feeding DER units made it difficult to realize selective protection
for PD 2 s during island operation, which is only based on current or voltage relays.
Therefore, the protection algorithm of PD 2 s during the island operation of LV
microgrid was chosen to be multi-criteria based where both voltage and current
measurements have been utilized (Fig. 19.7). The protection of PD 2 s should also
be able to adapt to the present network configuration as well as to the current states
of the DG units during island operation. This adaptation of settings and pick-up
limits of PD 2 s could be realized, for example, by microgrid management system
(MMS) when the microgrid configuration changes. During island operation, PD 1
is changed to be ready for re-synchronization of the microgrid back to the utility
grid which requires that phase voltages be measured from both sides of PD 1. The
protection settings of PD 2 s are changed to the ones needed in island operation.
To avoid the malfunction of PD 2 s, the protection settings of PD 2 s are not
changed from normal to island operation settings before all possible transients and
oscillations in voltages, currents, and frequency are stabilized after the transition
to island operation. MMS will also send the state changed from island to normal
operation signals to PD 2 s and PD 4 s after successful reconnection back to the
utility grid. From island operated microgrid power balance viewpoint, the role of
MMS is also important. For example, after fault F2 at LV feeder, the MMS must
immediately, after operation of LV feeder protection (PD 2), send new set point
values for those DGs or DER units that are still connected to the healthy part of the
microgrid or disconnect the less critical customer loads.
During the islanded operation of a microgrid, possible oscillations due to sudden
changes in the microgrid configuration need to be taken into account for the protec-
tion concept to achieve selective protection and to avoid the unnecessary tripping of
protection. This could be done by using communication-based interlocking signals.
During islanded operation rapid communication is required for microgrid protection
needs between protection devices (PD 1 and 2) as well as between master unit and
DER units. MMS must also be able to communicate with all the other components
and customer loads in the microgrid. Communication should preferably be based
on commonly used standards like IEC 61850. Active microgrid components at the
connection point of the microgrid (PD 1, master unit, and MMS) are also responsible
for microgrid re-synchronization. More information for example about the functions
needed for LV microgrid protection in the proposed scheme as well as details about
the operation curves of PDs in the proposed LV microgrid protection system during
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 563

Fig. 19.7 Number of protection zones and types of protection devices (PD 1–4) needed in normal
and island operation of LV microgrid. [4, 11, 72, 73]

grid-connected and islanded operation can be found in [4, 11]. It is enough from
the proposed LV microgrid protection system point of view that the converter-based
DER units will feed 2 × In active current (In is nominal current) during faults in
LV microgrid for the required FRT time defined by the operation curves of different
PDs.

19.5.3.2 Proposed LV Microgrid Protection Scheme 2

In [15, 74], strategies for the coordination of protective devices, in typical radial
distribution networks with DER, were proposed. Expanding on the idea presented
564 A. A. Memon et al.

in [15], in [53, 74] protection strategies based on microprocessor-based relays for

LV microgrids has been proposed. One of the salient features of this protection
scheme is that it does not require communications or adaptive protective devices. In
addition, it is stated in [53] that the proposed scheme is to a large extent independent
of the fault current magnitude, the microgrid operational mode, and the type and
size of the DER, subject to the modified relay setting for the grid-connected mode
of operation [15].

19.5.4 Protection of MV AC Microgrids

In [15, 75], communication-based protection methods have been suggested, which

should be implementable by commercially available protection relays, for the
protection of inverter-based medium-voltage (MV) microgrids. Proposed scheme
also includes a backup protection method to handle the failure of the communication
network. In [15] it was also stated that the suggested protection method is
independent of the operation mode of the microgrid as well as of the fault current
level, type, size, and location of the DER units.

19.5.4.1 Proposed MV Microgrid Protection Scheme 1

In [76] one potential protection scheme without communication was proposed for
short-circuit and earth-fault protection of an island operated medium-voltage (MV)
microgrid with many protection zones. The suggested protection method was not
dependent on the fault current level fed by the DER units and was compatible
and selective with DG unit low-voltage-ride-through (LVRT) curves. However, the
LVRT curve of the inverter- and synchronous generator-based DER units may
be different. Because rotating generators typically cannot have as long fault-ride-
through (FRT) / LVRT capability as the inverter-based DER units may have,
and the islanded MV microgrid with synchronous generators can be divided into
fewer protection zones than with inverter-based DERs. The suggested protection
strategy is only for islanded microgrid operation, but there are some similarities
with previously, in [77–80], presented short-circuit protection strategies for the grid-
connected operation.
The short-circuit and earth-fault protection operation speed in the proposed
scheme are not dependent on the fault-current level or measured impedance. In
contrast, they are based on voltage measurements from that network point (also
considering the direction of the current). This means that it is easier to achieve
selectivity with DER units LVRT requirements especially without fast communica-
tion and when there are many protections zones in the MV feeder. The suggested
protection method could be also used as a backup method for communication
assisted strategies. It can also be mentioned that the earth-fault protection part of the
proposed protection scheme is only applicable to neutral isolated MV microgrids.
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 565

Fig. 19.8 Suggested directional short-circuit protection strategy for island operated MV networks
[76]

The main principles of this proposed protection scheme are presented broadly in Fig.
19.8, 19.9, and 19.10. From [76], more detailed descriptions with case examples can
be found.
Table 19.4 summarizes the differences between the proposed protection strategy
[76] and a few other previously proposed protection methods for island operated
microgrids. From Table 19.4 one can see that the key advantage of the proposed
strategy (6. in Table 19.4), when compared with other previously proposed methods
(1.-5. in Table 19.4), is that it is selective with DG unit LVRT curves/FRT
requirements. In addition, it can be applied to microgrids with many protection
zones without the requirement for fast communication.
566 A. A. Memon et al.

Fig. 19.9 Suggested directional earth-fault protection strategy for island operated (earthing
isolated) MV networks [76]

19.5.4.2 Enhanced MV Microgrid Protection Scheme 1—HIF Detection

Included

In [76] (Sect. 3.8.1) a new protection strategy was suggested including short-circuit
and earth-fault protection for island operated, neutral isolated MV microgrid. In
[82], the MV microgrid protection strategy, with mainly OH lines and the inverter-
based DER units, was enhanced so that it also included high-impedance-fault (HIF)
detection (Fig. 19.11). Additionally, other improvement possibilities were suggested
in [82], like DER unit high-voltage-ride-through (HVRT) curve compatibility and
use of adaptive start/pick-up setting. More details can be found in [82].
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 567

Fig. 19.10 Protection voltage dependent operation time delay curve during island operation with
inverter-based DG units [76]

Table 19.4 [76] Differences of protection schemes for island operated microgrids
Requirement for fast
communication to achieve
Sensitivity / Applicability selectivity in microgrids
also to microgrids with Selectivity of with DER unit FRT
Protection only the converter-based MV and LV curves and with many
Scheme DER units faults protection zones
1. NO NO/YESa YES
Directional
Over-current
(DOC)
2. Distance YES YES YES
3. Current YES YES YES
differentialb
4. Volt- YES YES YES
age + DOC
5. ITA [57, YES YES YES
81]
6. YES YES NO
Proposed
Scheme [76]
a Casedependent, but not if only the converter-based DER units and low pick-up settings
b Network topology dependent, not suited for microgrids with radial feeders without DER units at
the end of the feeders
568 A. A. Memon et al.

Fig. 19.11 Improved MV microgrid protection strategy [82]

19.5.5 Communication-Less Adaptive Protection

The communication-based adaptive protections may become ineffective due to com-

munication system failure. In this situation protection settings can not be changed
according to changing network configurations. Therefore, some communication-
less adaptive protections should be developed to take over in case of communication
failures. A method for knowing the present condition or state of grid operation (grid-
connected or islanded) has been proposed in [83] which uses some local schemes
installed near the protection devices. The proposed scheme uses the thyristor, which
is fired during a short waveform interval of voltage, then thyristor voltage and cur-
rents are monitored. The monitored voltage and currents are then used to calculate
the grid equivalent impedance. From the value of grid equivalent impedance, it is
determined whether the main grid is connected or not. After determining the grid
operating condition, a command is sent locally to the protection device to change
its active group settings. The scheme can be used as a communication-less adaptive
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 569

protection scheme or can be used as a backup of communication-based schemes to

provide resiliency during communication failures. However, the proposed scheme
may not be effective for weak grids. The use of energy storage systems as fault
current sources like batteries, flywheels, supercapacitors, and so on, for providing
increased fault current by deactivating the fault current limiters of DERs locally or
activating extra energy storage devices during the events of communication failures
or cyberattacks is proposed in [84]. However, this scheme will require overrated
converters of energy storage systems to sustain increased fault currents.

19.5.6 Islanding Detection during Island Operation of Nested

Microgrid

Functionalities like islanding detection should also operate during the islanded
operation of (MV + MV or MV + LV) nested microgrids. In [85], islanding
detection of LV network connected generation unit during nested (MV + LV)
microgrid islanded operation was studied by simulations with PSCAD model from
a real-life smart grid pilot (Fig. 19.12). The main focus of the simulations was
on the study and comparison of the usage possibilities of combined (high-speed
communication-based transfer trip and fault detection/direction + voltage vector
shift) (Fig. 19.13) and multi-criteria (voltage total harmonic distortion UTHD and
voltage unbalance VU) based islanding detection schemes in nested microgrid
consisting only of converter-based DER units [85] (Fig. 19.14).
Based on the simulations it was concluded that combined islanding detection
scheme (Fig. 19.13) seems to be very feasible for the islanding detection with
LV network connected inverter-based DG units in nested (MV + LV) microgrids.
However, the detection logic for faults must be adapted. In the future, nested
microgrid control and protection functionality could be centralized/de-centralized
in control and protection units at HV/MV and MV/LV substations. For example, the
fault detection and location determination could be coordinated between these units
by using high-speed communication, real-time synchronized measurements from
multiple locations simultaneously as well as knowledge about the type, status, and
location of different DER units to always ensure selective islanding detection.

19.5.7 Need for Microgrid Grid Codes

As stated in Sect. 2.1 from the islanded microgrid (LV or MV) protection viewpoint,
the FRT capabilities and requirements as well as fault current contribution of
the converter-based DER units during the faults is of importance. Microgrid grid
codes for the island operated networks could “standardize” the requirements more
570 A. A. Memon et al.

Fig. 19.12 Sundom Smart Grid (SSG) to study islanding detection of LV network connected DER
unit during nested (MV + LV) microgrid islanded operation [85]

Fig. 19.13 Combined islanding detection Scheme [85]

specifically for LV and MV microgrid by considering their typical features, and so

on, in a more detailed manner. This would also reduce complexity and the need
for many case-specific alternatives could be avoided. Compatibility of the DER
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 571

Fig. 19.14 LV network DG unit (PV) primary islanding detection scheme example case (healthy
LV islanding after MV fault) during islanded operation of SSG (Fig. 19.12) [85]

unit FRT requirements, fault behavior, and island operated microgrid protection
principles, as well as operation speed requirements are naturally very critical from
microgrid frequency and voltage stability viewpoint and these could be also defined
as part of the microgrid grid codes.
For example, in LV microgrids the fault current fed by converter-based DER
units could be required to be active and less than 2xIn active current (In is nominal
current) during faults in LV microgrid for the required FRT time defined by the
operation curves of different PDs (Fig. 19.7, 19.15, and 19.16).
On the other hand, in MV microgrids the HVRT curve during island operation for
converter-based DER units could be defined in microgrid grid codes as shown in Fig.
19.17. It must be always simultaneously confirmed that the HVRT curve require-
ment of the DER unit, as part of the protection scheme during islanded microgrid
operation, is compatible with safety regulations (e.g., touch voltages during earth-
faults). Otherwise, faster MV microgrid protection schemes, for example, based on
the use of high-speed communication may need to be utilized.

19.6 Protection of DC Microgrids

In DC microgrids, all DERs and loads are connected to a common DC bus. The
DERs generating AC power (wind turbine, micro-turbine) are interfaced into the
DC bus by using AC-DC converters and AC loads are interfaced by using DC-AC
converters. On the other hand, all DGs generating DC power inherently (PVs, fuel
cells, batteries) and DC loads are connected to the DC bus using DC-DC converters.
Typically, DC microgrids are categorized into different topological configurations,
like multi-terminal, zonal, and DC looped. The chosen topology of DC microgrid
is dependent on the application, reliability level, and voltage level. Independent
of the topology, there are two types of DC bus architectures, that is, (1) unipolar
572 A. A. Memon et al.

Fig. 19.15 Operation curves for voltage relays (PD 1 in normal operation and PD 4 in normal and
island operation) (see Fig. 19.7) [4, 11, 72, 73]

DC bus topology using two-level voltage source converters (VSCs) and (2) bipolar
bus topology using three-level neutral-point-clamped VSCs. The DC microgrid
protection schemes are typically divided into (a) unit based and (b) non-unit based,
that is, protective device/breaker based. For example, in ships and DC homes,
the most typical protection scheme has been unit-based protection. A unit-based
protection scheme means that protection functionality exists within the DER units
and tries to either limit the DC-side fault current or drive it to zero.
The common DC bus can be interfaced into the main AC distribution grid
via a bidirectional AC-DC converter. Depending on the voltage level of DC bus,
DC microgrids can be classified as low voltage DC (LVDC) or medium-voltage
DC (MVDC) microgrids. Protective devices of DC microgrid with non-unit-based
protection schemes are represented by a generic block “PD” representing protective
devices as these are different from conventional CBs operated by protective relays
[86]. The main requirements of selectivity, sensitivity, and speed of an effective AC
protection system are also expected from a DC protection system. However, DC
protection cannot meet all requirements of AC protection in a very straightforward
way. Because in an AC system sufficient protection speed is required to maintain
the stability of synchronous generators while allowing a combined relay and breaker
operating times for several cycles of the fundamental frequency. For DC systems, in
contrast, the protection system should be substantially faster because a contribution
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 573

Fig. 19.16 Operation curves frequency relays of PD 1 and PD 4 in normal and island operation of
microgrid and operation curves for OC relays of PD 2 (directional low-set stage and nondirectional
high-set stage) in normal operation and PD 3 in normal and island operation (see Fig. 19.7) [4, 11,
72, 73]

Fig. 19.17 Proposed HVRT curve during MV microgrid island operation for converter-based DER
units [82]

to the fault current by a voltage source converter (VSC) even for a duration of more
than a few milliseconds can damage the converter unless some DC fault current
limiting/blocking mechanism is available to protect the converter. Additionally, DC
protection should isolate the fault before the blockage of IGBT switches of the
converter happens because due to the blockage of IGBTs, the converter control
will be lost, and more time will be required for the post fault restoration process.
574 A. A. Memon et al.

The requirement to prevent IGBTs blocking imposes strict requirements for the
protection of DC microgrids. The technology of DC breakers and off-the-shelf
digital DC relays is at a very low advancement level. Therefore, it is common
practice to trip the breakers on the AC side of all converters during a DC fault in
a DC system and then the fault is isolated using DC disconnectors This approach
is fast enough to protect the converter, but it is not selective because the loss of
the complete DC system happens during any fault. Overcurrent protection is not
the optimum selection for DC microgrid protection because the relay should be
capable of differentiating between in-zone and out-of-zone well before the fault
current reaches its final value. Moreover, the fault resistance also severely affects
the selectivity and sensitivity of DC overcurrent relays [87]. The common method
used for the location of DC fault is based on the traveling-waves principle [88],
which uses a communication link to find the arrival times of fault-induced traveling
waves at two ends of a line. Using the difference between two arrival times and wave
propagation velocity, the location of the fault can be identified. Due to the presence
of noise and short lines in DC microgrids, this method faces problems. Moreover,
traveling-waves relays are very costly for implementation in DC microgrids [87].
The other protection methods proposed for DC microgrids as reported in [19]
include overcurrent, current derivative, directional overcurrent, distance, and differ-
ential protection. The common protective/interrupting devices for DC microgrids
are fuses, molded case circuit breakers, no load switches, and SSCBs [19, 86].
DC microgrids are evolving and so are DC protection systems. Until adequate DC
protection devices are available the applications of DC microgrids will be limited to
only special applications like avionics, automotive, marine, the international space
station (ISS), spacecraft, aircraft, electronic computers, and servers in data centers.

Acknowledgments This work has been carried out as part of SolarX project at the University of
Vaasa, Finland. SolarX project is mainly funded by Business Finland (grant No. 6844/31/2018).
Some part of this work has been done in VINPOWER project and the main funding for
VINPOWER project came from European Regional Development Fund, ERDF (project No.
A73094). The financial support provided through these projects is greatly acknowledged.

References

1. Memon, A. A., & Kauhaniemi, K. (2015). A critical review of AC Microgrid protection issues
and available Solutions. Electric Power Systems Research, 129, 23–31.
2. Pierre Janssen, Monitoring, protection and fault location in power distribution
networks using system-wide measurements, PhD thesis, 2013, available http://
theses.ulb.ac.be/ETD-db/collection/available/ULBetd-10152013-111608/unrestricted/
PhD_thesis_Pierre_Janssen_final.pdf
3. Oudalov, A., Fidigatti, A., Degner, T., Valov, B., Hardt, C., Yarza, J. M., Li, R., Jenkins, N.,
Awad, B., Van Overbeeke, F., Hatziargyriou, N. & Lorentzou, M., Novel protection systems
for microgrids, final version of partial report for WORK PACKAGE C (TC2: Technical
requirements for network protection) on more MicroGrids EU-project, 2009.
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 575

4. Laaksonen, H., Technical Solutions for Low-Voltage Microgrid Concept, Ph.D. thesis, Dept.
of Electrical and Energy Engineering, University of Vaasa, Acta Wasaensia, 241, Vaasa 2011.
271 pages. Available at: http://www.uva.fi/materiaali/pdf/isbn_978-952-476-345-5.pdf
5. A. Oudalov, New Technologies for Microgrid Protection, Santiago 2013 symposium on
microgrids, 2013., Available at: http://der.lbl.gov/sites/der.lbl.gov/files/santiago_oudalov.pdf.
6. S. Conti, L. Raffa, U. Vagliasindi, Innovative solutions for protection schemes in autonomous
MV micro-grids, in International Conference on Clean Electrical Power, 2009, pp. 647–654.
7. H. Nikkhajoei, R.H. Lasseter, Microgrid protection, in: IEEE Power Engineering Society
General Meeting, 2007, pp. 1–6.
8. H. Al-Nasseri, M.A. Redfern, F. Li, A voltage based protection for micro-grids containing
power electronic converters, in: IEEE Power Engineering Society General Meeting, 2006, p. 7.
9. H. Al-Nasseri, M.A. Redfern, Harmonics content based protection scheme for Micro-grids
dominated by solid state converters, in: Power System Conference, 2008. MEPCON 2008. 12th
International Middle-East, 2008, pp. 50–56.
10. Oudalov, A., & Fidigatti, A. (2009). Adaptive network protection in microgrids. Int J Distrib
Energy Resour, 5, 201–226.
11. Laaksonen, H. & Kauhaniemi, K., Smart Protection Concept for LV Microgrid, International
Review of Electrical Engineering (IREE), Vol. 5 No. 2, 2010.
12. Chowdhury, S., Chowdhury, S. P., & Crossley, P. (2009). Microgrids and active distribution
networks, published by the Institution of Engineering and Technology. London: Athenaeum
Press Ltd, Gateshead, Tyne & Wear.
13. Voima, S., Laaksonen, H. & Kauhaniemi, K., Adaptive Protection Scheme for Smart Grids.
The IET 12th International Conference on Developments in Power System Protection (DPSP)
2014, Copenhagen, Denmark, March 31 - April 3, 2014.
14. Laaksonen, H. & Kauhaniemi, K. & Voima, S., DG Unit Fault Behavior and Protection of LV
Microgrid, International Review on Modelling and Simulations (IREMOS), Vol. 3 No. 3, June
2010.
15. Zamani, M. A., Protection and control of active distribution networks and microgrids, PhD
thesis, The University of Western Ontario, London, Ontario, Canada, 2012.
16. CIGRE (Joint Working Group B5/C6.26/CIRED). Protection of Distribution Systems with
Distributed Energy Resources. Final Report. March 2015.
17. H. J. Altuve, K. Zimmerman and D. Tziouvaras, Maximizing line protection reliability, speed,
and sensitivity,” In 2016 69th annual conference for protective relay engineers (CPRE),
College Station, TX, 2016, pp. 1–28.
18. IEEE Std-242-2001 Recommended Practice for Protection and Coordination of Industrial and
Commercial Power Systems (IEEE Buff Book).
19. Beheshtaein, S., Cuzner, R., Savaghebi, M., & Guerrero, J. M. (2019). Review on microgrid
protection. IET Gener. Transm. Distrib., 13(6), 743–759.
20. Brahma, S. M., & Girgis, A. A. (2004). Development of adaptive protection scheme for
distribution systems with high penetration of distributed generation. IEEE Transactions on
Power Delivery, 19, 56–63.
21. M.T. Doyle, Reviewing the impacts of distributed generation on distribution system protection,
in: IEEE Power Engineering Society Summer Meeting, 2002,pp. 103–105, vol. 1.
22. Horowitz, S. H., & Phadke, A. G. (2008). Power System Relaying (3rd ed.). West Sussex: John
Wiley & Sons Ltd.
23. Gers, J. M., & Holmes, E. J. (2004). Protection of electricity distribution networks - 2nd ed.
London: The Institution of Electrical Engineers.
24. Lasseter, R., et al. (2002). Integration of distributed energy resources, in: The CERTS microgrid
concept. Lawrence Berkeley National Lab.
25. Rockefeller, G. D., Wagner, C. L., Linders, J. R., Hicks, K. L., & Rizy, D. T. (1988).
Adaptive transmission relaying concepts for improved performance. IEEE Transactions on
Power Delivery, 3(4), 1446–1458.
26. IEEE Std 1547–2003 IEEE Standard for Interconnecting Distributed Resources with Electric
Power Systems.
576 A. A. Memon et al.

27. IEEE, IEEE Guide for Design, Operation, and Integration of Distributed Resource Island
Systems with Electric Power Systems, IEEE Std 1547. 4–2011, 2011, pp.1–54.
28. AREVA T&D. (2002). Network Protection & Automation Guide. First edition.
29. Arulampalam, A., Barnes, M., Engler, A., Goodwin, A., & Jenkins, N. (2004). Control of power
electronic interfaces in distributed generation microgrids. International Journal of Electronics,
91(9), 503–523.
30. Nuñez, J., Gil de Muro, A. & Oyarzabal, J. (2010). Development and evaluation of innovative
local controls to improve stability and islanding detection. WPA: Design of Microsource and
Load Controllers for efficient integration, TA1: Requirements for various DGs in supporting
MicroGrid operation, advanced architectures and control concepts for more MicroGrids,
STREP project funded by the EC under 6FP, SES6–019864, Labein Tecnalia, Derio, Spain.
31. Van Overbeeke, F. & Cobben, S. (2010). Operational aspects of a microgrid with battery
storage. In proceedings of IEEE PES conference on innovative smart grid technologies Europe.
Gothenburg, Sweden.
32. Oyarzabal, J., Jimeno, J., Agnostos, D., Arnold, G., Berg, A., Mustermann, E., Agnostos, T.,
Mustermann, H. & Yarza, J. M. (2009). Report on applied data structures and mapping to
communication means. Report as part of WORK PACKAGE E (standardization of technical
and commercial protocols and hardware) on more MicroGrids EU-project.
33. Eto, J., Lasseter, R., Schenkman, B., Stevens, J., Klapp, D., Volkommer, H., Linton, E.,
Hurtado, H. & Roy, J. (2009). Overview of the CERTS Microgrid Laboratory Test Bed.
In Proceedings of Integration of Wide-Scale Renewable Resources into the Power Delivery
System, 2009 CIGRE/IEEE PES joint symposium. Calgary, Canada.
34. Laaksonen, H. & Kauhaniemi, K. Synchronized re-connection of island operated LV microgrid
Back to utility grid. Proc. IEEE PES ISGT Europe, Gothenburg, Sweden. October 2010.
35. Young-woo Jeong, Hyun-wook Lee, Seok-won Lee, Young-geun Kim, High-speed AC circuit
breaker and high-speed OCR, cired 22nd International Conference on Electricity Distribution
Stockholm, 10–13 June 2013, Paper 0608.
36. S. Krstic, E. L. Wellner, A. R. Bendre and B. Semenov, “Circuit Breaker Technologies for
Advanced Ship Power Systems,” 2007 IEEE electric ship technologies symposium, Arlington,
VA, 2007, pp. 201–20e8.
37. Kannuppaiyan, S., & Chenniappan, V. (2015). Numerical inverse definite minimum time
overcurrent relay for microgrid power system Protection. IEEJ Trans, 10, 50–54.
38. M. Pujiantara, D. C. Riawan, A. Indrasaputra, T. P. Sari and V. Raki Mahindara, “The
Automation of Time Dial Setting Calculation and Inverse Type Curve Selection for Over
Current Relay Based on Numerical Computation in Real Industrial Electrical System,” 2018
Conference on power engineering and renewable energy (ICPERE), Solo, Indonesia, 2018, pp.
1–6.
39. IEEE Std C37.112-1996, IEEE Standard Inverse-Time Characteristic Equations for Overcur-
rent Relays.
40. F. van Overbeeke, “Fault current source to ensure the fault level in inverter dominated
networks,” CIRED 2009 - The 20th International Conference and Exhibition on Electricity
Distribution - Part 2, Prague, 2009, pp. 1–13.
41. F. van Overbeeke, “Fault current source to ensure the fault level in inverter-dominated net-
works,” CIRED 2009 - 20th international conference and exhibition on electricity distribution
- part 1, Prague, Czech Republic, 2009, pp. 1–4.
42. Hatziargyriou, N. (2013). Microgrids: Architectures and control. West Sussex: John Wiley &
Sons.
43. Hou, D., & Dolezilek, D. (2010). IEC 61850-what it can and cannot offer to traditional
Protection schemes, Schweitzer Engineering Laboratories, Inc. SEL Journal of Reliable Power,
1(2).
44. Relion Protection and Control, 650 Series IEC 61850 Communication Protocol Manual,
Document ID: 1MRK 511 242-UEN, Issued: February 2011, Revision: - Product version: 1.1,
ABB 2011.
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 577

45. Mahat, P., Chen, Z., Bak-Jensen, B., & Bak, C. L. (2011). A Simple Adaptive Overcurrent
Protection of Distribution Systems with Distributed Generation. IEEE Transactions on Smart
Grid, 2(3).
46. Pierre Janssen, Monitoring, protection and fault location in power distribution networks
using system-wide measurements, PhD thesis 2013, available at: http://theses.ulb.ac.be/ETD-
db/collection/available/ULBetd-10152013-111608/unrestricted/PhD_thesis_Pierre_Janssen_
final.pdf
47. D. Ishchenko, A. Oudalov, and J. Stoupis, Protection coordination in active distribution grids
with IEC 61850, IEEE 2012.
48. Laaksonen, H., Ishchenko, D., & Oudalov, A. (2014). Adaptive Protection and Microgrid
Control Design for Hailuoto Island. IEEE Transactions on Smart Grid, 6(3).
49. G. Buigues, A. Dyśko, V. Valverde, I. Zamora, and E. Fernández Microgrid Protection: Tech-
nical challenges and existing techniques, International Conference on Renewable Energies
and Power Quality (ICREPQ’13), Bilbao (Spain), 20–22 March, 2013: http://www.icrepq.com/
icrepq’13/262-buigues.pdf
50. H. Nikkhajoei, et al., Microgrid fault protection based on symmetrical and differential current
components, 2006.
51. L. Bin, et al., Design of protection and control scheme for microgrid systems, in Universities
Power Engineering Conference (UPEC), 2009 Proceedings of the 44th International, 2009, pp.
1–5.
52. R. J. Best, et al., Communication assisted protection selectivity for reconfigurable and islanded
power networks, in Universities Power Engineering Conference (UPEC), 2009 Proceedings of
the 44th International, 2009, pp. 1–5.
53. Zamani, M. A., Sidhu, T. S., & Yazdani, A. (2011). A protection strategy and microprocessor-
based relay for low-voltage microgrids. IEEE Trans. on Power Delivery, 26(3), 1873–1883.
54. Voima, S., Kauhaniemi, K., & Laaksonen, H. Novel Protection Approach for MV Microgrid
(p. 2011). Frankfurt, Germany: CIRED 2011.
55. M. Dewadasa, Protection for distributed generation interfaced networks, Electrical Engineer-
ing, Faculty of Built Environment and Engineering, Queensland University of Technology,
Queensland (Australia), 2010.
56. M. Dewadasa, et al., Control and protection of a microgrid with converter interfaced micro
sources, in Power Systems, 2009. ICPS ’09. International conference on, 2009, pp. 1–6.
57. Dewadasa, M. Protection of distributed generation interfaced networks, Thesis, Queensland
University of Technology, Queensland, Australia, 2010.
58. Laaksonen, H. & Kauhaniemi, K., Fault type and location detection in islanded microgrid with
different control methods based converters, in Proceedings of 19th international conference
and exhibition on electricity distribution. Vienna, Austria, 2007.
59. Al-Nasseri, H. & Redfern, M. A., A new voltage based relay scheme to protect micro-
grids dominated by embedded generation using solid state converters, in Proceedings of 19th
international conference on electricity distribution. Vienna, Austria, 2007.
60. Oudalov, A. & Fidigatti, A., Microgrid protection and modern protection devices, final report
as part of WORK PACKAGE C (TC2: Technical requirements for network protection) on more
MicroGrids EU-project, 2008.
61. C. Hou, et al., A study of voltage detection based fault judgement method in micro-grid with
inverter interfaced power source, in The International Conference on Electrical Engineering,
Shenyang (China), 2009, pp. 1–5.
62. T. Loix, et al., Protection of microgrids with a high penetration of inverter-coupled energy
sources, in Integration of Wide-Scale Renewable Resources into the Power Delivery System,
2009 CIGRE/IEEE PES Joint Symposium, 2009, pp. 1–6.
63. W. Xiao-ping, et al., Research on the relay protection system for a small laboratory-
scale microgrid system, in Industrial Electronics and Applications (ICIEA), 2011 6th IEEE
Conference on, 2011, pp. 2712–2716.
578 A. A. Memon et al.

64. M. Dewadasa, Member, A. Ghosh, and G. Ledwich, Protection of Microgrids Using Differ-
ential Relays, (2011) Protection of microgrids using differential relays. In Proceedings from
AUPEC 2011. https://eprints.qut.edu.au/47434/1/2011011990.ePrints.Dewadasa.pdf
65. Conti Stefania, Raffa Lorenzo and Vagliasindi Umberto, Analysis of protection issues in
autonomous MV micro-grids, 20th International Conference on Electricity Distribution.
CIRED 2009. - pp. 1–5.
66. Casagrande, E., Woon, W. L., Zeineldin, H. H., & Svetinovic, D. (2014). A differential
sequence component Protection scheme for microgrids with inverter-based distributed gen-
erators. IEEE transactions on smart grid, 5(1). https://ieeexplore.ieee.org/abstract/document/
6678658.
67. Sortomme, E., Venkata, S. S., & Mitra, J. (2010). Microgrid protection using communication
assisted digital relays. IEEE Transactions on Power Delivery, 2789–2796.
68. Brahma, S. M. (2011). Fault location in power distribution system with penetration of
distributed generation. IEEE Transactions on Power Delivery, 1, 99.
69. Venkata S.S. et al., Advanced and adaptive protection for active distribution grid, 22nd
International Conference on Electricity Distribution (CIRED 2013). – Stockholm, Sweden,
2013.
70. R. M. Tumilty, et al., Approaches to Network Protection for Inverter Dominated Electrical Dis-
tribution Systems, in Power Electronics, Machines and Drives, 2006. The 3rd IET International
Conference on, 2006, pp. 622–626.
71. Tumilty, R. M., Elders, I. M., Burt, G. M., & McDonald, J. R. (2007). Coordinated Protection,
Control & Automation Schemes for microgrids. International Journal of Distributed Energy
Resources, 3(3), 225–241.
72. Laaksonen, H., Protection Principles for Future Microgrids, IEEE Transactions on Power
Electronics, Vol. 25, No. 12, December 2010.
73. Laaksonen, H., Kauhaniemi, K. & Voima, S., Protection System for Future LV Microgrids,
CIRED 2011, Frankfurt, Germany, June 6–9, 2011.
74. M. A. Zamani, T. S. Sidhu and A. Yazdani, A strategy for protection coordination in radial
distribution networks with distributed generators, in Proc. IEEE Power and Energy Society
General Meeting, 8 pages, Jul. 2010.
75. M. A. Zamani, A. Yazdani and T. S. Sidhu, A communication-assisted protection strategy for
inverter-based medium-voltage microgrids, in IEEE Trans. on Smart Grid, 12 pages, Vol 3, No.
4. Dec. 2012.
76. Laaksonen, H. Protection Scheme for Island Operated Medium-Voltage Microgrid, Interna-
tional Review of Electrical Engineering (IREE), Vol. 10 No. 4, August 2015.
77. Laaksonen, H. (2015). Future-proof MV distribution network short-circuit Protection scheme.
International Review of Electrical Engineering (IREE), 10, 98–108.
78. H. Laaksonen, Protection of Future Active Distribution Networks with Distributed Generation,
CIGRE Conference ~SEAPAC 2015, 2015, Sydney, Australia.
79. H. Laaksonen, Grid code compatible Protection scheme for smart grids, The 23rd international
conference on electricity distribution ~CIRED 2015~, 2015, Lyon, France.
80. Laaksonen, H. (2015). Comparison of new LVRT compatible MV network short-circuit
Protection schemes. International Review on Modelling and Simulations (IREMOS), 8, 122–
131.
81. M. Dewadasa, A. Ghosh, Microgrid Operation and Control: DELIVERABLE 6: Protection of
Meshed Microgrids, CSIRO Intelligent Grid Research Cluster- Project 7, Queensland Univ.
of Technology, Queensland, Australia. Available: http://igrid.net.au/resources/downloads/
project7/P7%20M6%20Meshed%20Microgrid%20Protection.pdf
82. Laaksonen, H., Hovila, P. Enhanced MV Microgrid Protection Scheme for Detecting High-
Impedance Faults. 12th IEEE PES PowerTech Conference, PowerTech Manchester 2017,
Manchester, UK, June 18-22, 2017.
19 Microgrid Protection with Conventional and Adaptive Protection Schemes 579

83. Ferreira, R. R., Colorado, P. J., Grilo, A. P., Teixeira, J. C., & Santos, R. C. (2019). Method for
identification of grid operating conditions for adaptive overcurrent protection during intentional
islanding operation. International Journal of Electrical Power & Energy Systems, Volume, 105,
632–641.
84. H. F. Habib, C. R. Lashway and O. A. Mohammed, “On the adaptive protection of microgrids:
A review on how to mitigate cyber-attacks and communication failures,” 2017 IEEE industry
applications society annual meeting, Cincinnati, OH, 2017, pp. 1–8.
85. H. Laaksonen and P. Hovila, “Islanding Detection During Intended Island Operation of Nested
Microgrid,” 2018 IEEE PES innovative smart grid technologies conference Europe (ISGT-
Europe), Sarajevo, 2018, pp. 1–6.
86. R. M. Cuzner and G. Venkataramanan, “The Status of DC Micro-Grid Protection,” 2008 IEEE
industry applications society annual meeting, Edmonton, AB, 2008, pp. 1–8.
87. Hooshyar, A., & Iravani, R. (2017). Microgrid Protection. Proceedings of the IEEE, 105(7),
1332–1353.
88. Saha, M. M., Izykowski, J. J., & Rosolowski, E. (2010). Fault Location on Power Networks.
London: Springer-Verlag.
Chapter 20
Fault Identification, Protection Schemes,
and Restoration Requirements
of Microgrids

Amin Yazdaninejadi, Amir Hamidi, Sajjad Golshannavaz,

and Daryoush Nazarpour

20.1 Introduction

Within the last few decades and with the appearance of distributed energy resources
(DERs), there has been a significant evolution in traditional distribution networks,
ending to the emergence of active distribution systems and microgrids (MGs) [1].
Limited to the specified geographical boundaries, MG is typically characterized
as a set of DERs which are coordinated for supplying local loads both in grid-
connected and islanded operation modes. As an eminent feature of MG, it can
maintain the continuity of power supply in the event of faults; accordingly, it reduces
the frequency and duration of outages in demand-side territory and improves
the resiliency of the electrification process against disconnections. However, the
offered resiliency by MGs, as a tremendous feature, can be easily imperiled if
they are not properly restored during power grid faults. The restoration process
by MGs is targeted to restore the healthy but disconnected parts of a system
due to possible isolation following a fault. Accordingly, optimal placement of
fault indicators (FIs) could facilitate easier and faster identification of faults and
isolation of faulty feeders. It should be noted that the existing protection devices
could be also deployed as FIs in this process. After isolation of faulty feeders,
it is well-known that power supply restoration is mainly provided by means of
DERs; however, transient stability and fault ride-through (FRT) of these resources
should be carefully observed. To do so, protection systems should be designed
to be selective, fast, and sensitive. That is, although MGs are established at the

A. Yazdaninejadi
Department of Electrical Engineering, Shahid Rajaee Teacher Training University, Tehran, Iran
e-mail: a.yazdaninejadi@sru.ac.ir
A. Hamidi · S. Golshannavaz () · D. Nazarpour
Faculty of Electrical and Computer Engineering, Urmia University, Urmia, Iran
e-mail: st_a.hamidi@urmia.ac.ir; s.golshannavaz@urmia.ac.ir; d.nazarpour@urmia.ac.ir

© Springer Nature Switzerland AG 2021 581

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_20
582 A. Yazdaninejadi et al.

distribution level of conventional electrical grids, they require different protection

schemes for averting the unintentional disconnection of DERs which could be
triggered coincidently at the fault occurrence time. This chapter puts a special focus
on both of these issues and develops effective models. If based on developed models,
faulty feeders are identified and isolated promptly and DERs transient stability and
FRT are satisfactorily preserved and realized, then it is a crucial need to restore the
MG promptly to an optimal target configuration. In this way, the existing protection
devices besides FIs can be jointly used as controllable switches for fast restoration
of MG. Besides, optimal switch allocation would definitely help to form new
optimal maneuver points and reconfigure the network to an optimal configuration;
accordingly, preserving the equipment health and enhancing the reliability of MG
which would shrink the restoration costs. In view of that, the optimal switch
allocation problem as the last step of successful and effective restoration is also
discussed in this chapter.

20.2 Fault Locating and Restoration Requirements in MG

As discussed, fast and precise identification of faulty feeders is the first step of
the restoration process. This task is highly hinged on FI placement and deploying
the existing protective devices as possible FIs. Placement of new FIs is itself an
important step that would be addressed in the following subsections, separately.
MGs are the nearest end users’ infrastructures; accordingly, the operational conti-
nuity of these infrastructures is of importance. Since facing with faults in MG is
inevitable, the restoration techniques should be extended properly in the operations
of these electrification systems. In this manner, MG operators should locate and
isolate the faults along with restoration swiftly to feed the maximum possible end
users. Prior to the definition of the smart grid concept, fault location, isolation, and
service restoration (FLISR) is carried out by manual tracking the feeders which
ends to locating the fault [2]. Then, the operational crew isolates the fault by
opening the corresponding switches. Afterward, in case of possibility, the maximum
number of consumers is restored to electrical service through closing tie switches
of maneuver lines. This procedure should be performed in minimum time which
is done by human-heuristic decisions and based on experience. The representative
schematic of the conventional procedure is shown in Fig. 20.1. Generally, some
crucial factors are analyzed to enhance the FLISR functionality in MGs. The fault
detecting and locating with the conventional procedure stands as a time-consuming
task of restoration. Although, finding the approximate location of faults can be
easily performed by the calls received from unserved consumers, accurate fault
locating brings a long process. What’s more, patrolling the operation crew with the
aim of fault locating along with making optimal decisions to isolate the fault and
restore healthy consumers increases this time. Consequently, it drastically increases
20 Fault Identification, Protection Schemes, and Restoration Requirements. . . 583

Fault detection with outage reports by

consumers (Calls)

Patrolling the lines of distribution grids by

operational crew to detect the fault nearby Manual switching by crew to isolate the
the unserved consumers fault

Manual switching by crew to restore the

healthy consumers

Fig. 20.1 The conventional service restoration procedure

the system average interruption duration index (SAIDI). This index is measured to
determine how long restoration takes for a consumer that is unserved. Likewise, the
time is taken for locating faults, and restoring the healthy consumers also affects the
energy not supplied (ENS) index. Therefore, improving allocated time for locating
the fault and service restoration reduces the SAIDI and ENS indices which enhances
the reliability indices and also the social welfare of consumers [3–6]. In this regard,
other indices are also utilized to analyze the functionality of the conducted FLISR
programs which are discussed thoroughly in the following sections.
The advent of new technologies to realize the smart grid concept facilitates the
implementation of automation in MGs. In such a manner, the key equipment are
indicators, switches, and controllers which are operated by supervisory control and
data acquisition (SCADA) and distribution management system (DMS). One of
the main tasks of the SCADA/DMS is the Outage Management System (OMS).
The FLISR function of an OMS can communicate with all sensors, switches, and
controllers of the network to locate any faults and restore the healthy consumers,
rapidly and accurately. Since this system is based on a supervisory control platform,
the resultant improvement in reliability indices of the MG could be further
enhanced. Different types of FIs, automated switches and reclosers, smart relays,
and automated Volt/VAr controllers are deployed to realize the supervisory FLISR
function. Despite the manual restoration, the FLISR function of a supervisory OMS
produces different feasible strategies for restoring healthy consumers. Each of these
feasible strategies causes different subsidiary objectives such as voltage security,
power quality, network security, and so on. The operator can then select different
solutions based on the operation priorities. This advantage in an automated and
supervisory FLISR function originates from analytical and mathematical solving
methods for restoration problems. These methods are established based on the
human experience for the manual FLISR and the technical knowledge of the
operational crew.
584 A. Yazdaninejadi et al.

20.2.1 Fault Indication and Restoration Devices

The performance of the FLISR in an OMS is highly influenced by the fault

indication and restoration devices. Different types of devices are deployed to tackle
this goal. Some devices are only utilized for the FLISR task and the others have
ancillary functionality to assist the OMS which aims at the improvement of this
task. The main devices are as follow [2]:
• Remote fault indicators.
• Circuit breakers, switches, and reclosers.
• Smart relays.
• Outage detectors.
Remote fault indicators: Remote fault indicators detect faults rapidly and
accurately by sensing the voltage and current passed through the power line
[7]. Then, FIs inform faults to OMS through the communication infrastructure.
Moreover, these devices have a visual representation for their status like the lights
to help the operational crew in detecting faults. Recognizing the distinction between
fault current and overload current of electrical motors is an important functionality
of FIs. Furthermore, FIs should also detect the faults with high impedance.
Circuit breakers, switches, and reclosers: The circuit breakers are assumed to
be operated by protection relays to isolate the occurred faults. Besides, the remotely
controllable switches (RCSs) can be directly opened and closed by the OMS to
isolate the fault and restore the service, respectively. However, these switches
are generally cooperated by smart relays. Likewise, these switches are used to
reconfigure the grid topology to facilitate the flexibility of the MG operator for
performing different FLISR strategies. The reclosers can frequently re-energize the
fault point to restore the service in case of transient faults like bird collision with the
MG feeders. Since the recloser distinguishes permanent and temporary faults, it can
efficiently improve the reliability of the MG and FLISR task [8]. However, it can be
used to isolate temporary faults in the FLISR task.
Smart relays: Smart relays are based on accurate algorithms to detect, isolate,
and assess the occurred faults. Therefore, they are used in MGs for protection and
automation tasks, concurrently. Smart relays can analyze the faults current and
voltage data and transfer these data to OMS to adopt automation tasks for FLISR.
Outage detectors: These devices recognize the power loose of consumers
besides facilitating and accelerating the outage information which is important to
enhance the restoration process. As revealed earlier, before devising automated
FLISR techniques, the calls from interrupted consumers inform the operator about
service outage. This time-consuming procedure is not accurate which deteriorates
the restoration. This is while; outage detectors provide the OMS with a rapid and
accurate response to outages.
To realize an efficient FLISR task, different frameworks, devices arrangement,
and algorithms are used by utilities. Table 20.1 summaries the applied arrangements
of the FLISR task by different utilities which were reported by the department of
20 Fault Identification, Protection Schemes, and Restoration Requirements. . . 585

Table 20.1 The specific arrangement of automation devices and the level of automation for
different utilities [9]
Feature Center point Duke NSTAR PHI Southern
Field devices Intelligent Electronic Telemetry Substation Automated
involved grid reclosers, communica- breakers, switches/
switching circuit tion, line field reclosers,
devices breakers, and sensors, and switches, and fault
(IGSDs) act line sensors smart reclosers, indicators
as switching switches and smart
devices and relays
monitoring
equipment
Mode of Manual Fully Transitioned Fully Fully
FLISR validation automated to full automated automated
operation required automation
during the
project

energy (DOE) of the US [9]. As can be seen, each utility has chosen a specific
arrangement to implement the FLISR task. These projects can realize a fully
automated FLISR framework by their own arrangement. To this end, the appropriate
arrangement of the devices to perform automated actions for outage management
should be assessed for MG structure. It is important for the operator to maximize
the automation level of the grid to achieve improvement in reliability indices,
economic profits of service, and investment costs. The techno-economic analysis
of MG planning to enhance the FLISR functionality is assessed in the next part.

20.2.2 FI Placement

The first step of the FLISR is allocated to the role of FIs in the detection of the
faulty paths of the network. As is obvious, the higher number of FIs is installed
in feeders, the higher observability of MG is possible. However, due to economic
considerations, this approach is not affordable. Locating faults with a limited
and sufficient number of FIs is possible. Therefore, the allocation of the FIs in
the electrical grids stands as a planning problem that should be considered by
the MG planner and operator from economic and reliability aspects. Different
allocation methods have been proposed in the literature for tackling the FI placement
problem [10]. In an accurate model of FI allocation, the investment, installation,
and maintenance costs, as well as interruption cost of the consumers, should be
considered as follow:

Minimize C inst + C main + C int (20.1)

586 A. Yazdaninejadi et al.

   
C inst = Xf,s CI f,s + Xf,p CIf,p (20.2)
f ∈Nf s∈Ns f ∈Nf p∈Np

   1    1
C main = t Xf,s MC f,s + X MCf,p
(1+DR) (1+DR)t f,p
t∈Nt f ∈Nf s∈Ns t∈Nt f ∈Nf p∈Np
(20.3)

     1
C int = λt,f,i Lot,f,j,k Ct,f,i,j,k (20.4)
(1 + DR)t
t∈Nt f ∈Nf i∈Ni j ∈Nj k∈Nk

In Eqs. (20.1)–(20.4) the Xf, s and Xf,p are the binary variables indicating
whether FI is installed on location i in the main feeder of feeder f and whether
FI is installed on location p in laterals of feeder f, respectively. CIf, s and CIf,p
are the capital investment cost of FI in feeder f and location s in the main feeder
and for FI in feeder f and location p in laterals, respectively. MCf, s and MCf,p are
maintenance cost of FI in feeder f and location s in the main feeder and FI in feeder
f and location p in laterals. Also, DR, λt, f, i , Lot, f, j, k , and Ct, f, i, j, k are the annual
discount rate, the failure rate at sect. i and feeder f at year t, the average load of
customers with type k in load point j and feeder f at year t, and interruption cost of
customers with type k in load point j for a fault at location i and feeder f at year t,
respectively. All of the values in (20.4) are certain parameters except the consumer
interruption cost, which is impacted by the location of FIs.
FIs allocation in MGs by this model imposes a numbering rule for feeders. Main
feeders and laterals are labeled by odd and even numbers, respectively. Figure 20.2
shows the modeling process of FI allocation. As can be seen, both sides of the
main feeder and the starting point of the laterals are defined as a candidate for FI
allocation. Here, (20.5) is utilized to recognize whether a section is suspicious of the
faulty zone or not. In this equation, bf, i, l shows the number of fault detective devices

1:1 s:1 s:2 1:3 s:3 s:4 1:5

p(p ′):1 p(p ′):2

i:3
1:2 1:4
i:4
j:1 j:2
Candidate place for FI Normally close CB Normally open switch

Fig. 20.2 The modeling process of FI allocation in [10]

20 Fault Identification, Protection Schemes, and Restoration Requirements. . . 587

Table 20.2 Assessing the functionality of the Eq. (20.5) [10]

Fault detective device exists between
location i and l bf, i, l Lf, i, l Section l is involved in the faulty zone
Yes 1 0 No
No 0 1 Yes

between faulty section i and l. The functionality of this equation is assessed in Table
20.2.

1 1
ε− bf,i,l ≤ Lf,i,l ≤ 1 − bf,i,l + ε (20.5)
ζ ζ

In (20.5), the value of bf, i, l is calculated by using (20.6)–(20.8). In these

equations, the Yf, s and Yf,p are the binary parameters that indicate any other fault
detective device except for FI installed on the location.
 l−1 
     
l−1
i l
bf,i,l = Xf,p +Yf,p + Xf,p +Yf,p + Xf,s + Yf,s ∀i<l , p= , p =
2 2
s=i s=i
(20.6)
 l−1 
     
l−1
i l
bf,i,l = Xf,p +Yf,p + Xf,p +Yf,p + Xf,s + Yf,s ∀i>l , p= , p =
2 2
s=i s=i
(20.7)

bf,i,l = 1 ∀i = l (20.8)

The required time for fault locating based on all fault detective devices can be
calculated by (20.9) and (20.10). In these equations, the Lef, l and Vf, l are the length
of section l and patrolling speed of operational crew for that section, respectively.
Moreover, Lf, i, l indicates that whether the feeder is in the faulty zone or not. If it is
not in the faulty zone, this feeder is excluded from patrolling.

patrol
 Lef,l
Tf,i = Lf,i,l ∀f ∈ Nf , ∀i ∈ Ni | odd (20.9)
Vf,l
l∈Nl |odd

patrol
 Lef,l
Tf,i = Lf,i,l ∀f ∈ Nf , ∀i ∈ Ni | even (20.10)
Vf,l
l∈Nl
588 A. Yazdaninejadi et al.

As discussed earlier, maneuvering of the feeders by the crew in conventional

FLISR task should be manually performed based on trial-and-error actions. This
process can be estimated by (20.11)–(20.13). In these equations, numf, i is the
number of main feeder sections in the faulty zone. In this process, only the main
feeder sections are involved in the switching actions to find out the faulty section.
This procedure is modeled by odd numbers in (20.11). Increasing the suspected
sections of the main feeder for the faulty zone gives a binary value of 1 to df, i, n
t−e
which adds additional time to Tf,i as the maneuver time. Finally, the total time
to inspect and trial-and-error actions of the faulty zone is calculated by (20.14)
m−sw
and (20.15). In these equations, the Tf,i is the time required for isolating the
recognized faulty section.

numf,i = Lf,i,l ∀f ∈ Nf , ∀i ∈ Ni (20.11)
l∈Nl |odd

numf,i − n numf,i
+ ε ≤ df,i,n ≤ + ε ∀f ∈ Nf , ∀i ∈ Ni , ∀n ∈ Nn
ζ n
(20.12)


t−e maneuver ∀f ∈ N ,
Tf,i = df,i,l Tf,i,n f ∀i ∈ Ni (20.13)
n∈Nn

patrol m−sw
total
Tf,i = Tf,i + Tf,i
travel
+ Tf,i (20.14)

t−e m−sw
total
Tf,i = Tf,i + Tf,i
travel
+ Tf,i (20.15)

Finally, the interruption cost of the consumers is calculated for three types of
interruptions that are experienced by them. In this way, (20.16) models the inter-
ruption cost for consumers whose load can be restored immediately by switching
actions after the fault and before the fault locating process. Equations (20.17) and
(20.18) models this value for consumers whose load can be restored after fault
locating by switching actions. These equations are different for the consumers which
are downstream and upstream of the faulted section. Equation (20.19) models the
interruption cost of consumers who need to be interrupted until repairs are finished.
 
Ct,f,i,j,k ≥ cdf t,f,i,j,k Tf,i,j
res
∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , ∀j ∈ Nj , ∀k ∈ Nk
(20.16)
20 Fault Identification, Protection Schemes, and Restoration Requirements. . . 589

 
  −1
2j# −1
2j#
Ct,f,i,j,k ≥ cdf t,f,i,j,k Tf,i
total 1 − Xf,p − Yf,p − Xf,s − Yf,s
s=i s=i
∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , ∀j > 2i , ∀k ∈ Nk , p = i
2
(20.17)

 
  #
i−1 #
i−1
Ct,f,i,j,k ≥ cdf t,f,i,j,k total
Tf,i 1 − Xf,p − Yf,p − Xf,s − Yf,s
s=2j s=2j
∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , ∀j < 2i , ∀k ∈ Nk , p = i
2
(20.18)

 
total + T repair
Ct,f,i,j,k ≥ cdf t,f,i,j,k Tf,i f,l (20.19)
∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , j = 2i , l = i, ∀k ∈ Nk

20.2.3 Simultaneous Placement of FIs and RCSs

To enhance the automation ability of the MG, RCSs are installed in some sections
of the feeders. They are classified as sectionalizing switches that are able to be
monitored and controlled by the OMS. This capability paves the way for OMS to
increase the speed of the FLISR operation either in the case of fully automated
or near-automated actions. This rapidness can be achieved in fault isolation and
service restoration phases [11]. After inspecting the faulty section by OMS based
on the received signals from FIs, the isolation process can be remotely performed
by the RCSs. This process is accelerated by remotely opening the nearest RCS to
the fault in upstream of the faulty section. Then, healthy consumers in the upstream
network can be restored by closing the circuit breakers which were tripped. Due to
the higher cost of these switches in comparison with manual switches (MSs), the
suitable analysis should be conducted to evaluate the value of the investment cost of
them for improving the reliability indices. The main feature of the RCSs is that they
have the functionality of both FI and switch as the fault indicator and fault isolator,
respectively. This capability makes RCSs comparable to FIs and MSs. Following the
model for FI placement in the previous section, authors in [12] proposed a model
for simultaneous placement of the FIs, RCSs, and MSs in an MG. This model aims
to reduce the fault locating and isolation time by appropriate placement of FIs and
RCSs besides reducing the restoration time by appropriate placement of MS and
RCS. Therefore, the RCSs are helpful in both phases. Here, the previous model
is developed to include the sectionalizing switches. The model is demonstrated by
(20.20)–(20.35).
590 A. Yazdaninejadi et al.

Minimize C inst + C main + C int (20.20)

  
C inst = FI
Xf,s CI Ff,sI + Xf,s
RCS
f,s + Xf,s CI f,s
CI RCS MS MS
(20.21)
f ∈Nf s∈Ns

   1  
C main = X FI
f,s MC FI
f,s + X RCS
f,s MC RCS
f,s + X MS
f,s MC MS
f,s
(1 + DR)t
t∈Nt f ∈Nf s∈Ns
(20.22)

     1
C int = λt,f,i Lot,f,j,k Ct,f,i,j,k (20.23)
(1 + DR)t
t∈Nt f ∈Nf i∈Ni j ∈Nj k∈Nk

Ct,f,i,j,k = cdf t,f,i,j,k (T ) (20.24)

1 1
ε− bf,i,l ≤ Lf,i,l ≤ 1 − bf,i,l + ε (20.25)
ζ ζ


l−1 
l−1
bf,i,l = FI
Xf,s + RCS
Xf,s ∀i < l (20.26)
s=i s=i


l−1 
l−1
bf,i,l = FI
Xf,s + RCS
Xf,s ∀i > l (20.27)
s=i s=i

bf,i,l = 1 ∀i = 1 (20.28)

 Lef,l pre
loc
Tf,i = Lf,i,l + Tf ∀f ∈ Nf , ∀i ∈ Ni (20.29)
Vf,l
l∈Nl
20 Fault Identification, Protection Schemes, and Restoration Requirements. . . 591

pre
where Tf , cdft, f, i, j, k are the preparation time for the crew to reach the faulty
section of the feeder and Customer damage function, respectively [13].
 
Ct,f,i,j,k ≥ cdf t,f,i,j,k Tf,i,j
aut
∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , ∀j ∈ Nj , ∀k ∈ Nk
(20.30)

 
  −1
2j#
Ct,f,i,j,k ≥ cdf t,f,i,j,k Tf,i
loc 1− RCS
Xf,s
s=i (20.31)
∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , ∀j > 2 , ∀k
i
∈ Nk

 
  # RCS
i−1
Ct,f,i,j,k ≥ cdf t,f,i,j,k Tf,i
loc 1− Xf,s
s=2j (20.32)
∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , ∀j < 2i , ∀k ∈ Nk

 
  −1 
2j# 
rep
Ct,f,i,j,k ≥ cdf t,f,i,j,k Tf,i + Tf,l
loc 1− Xf,s + Xf,s
RCS MS
s=i (20.33)
∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , j > i
2, l = i, ∀k ∈ Nk

 
  #
i−1  
rep
Ct,f,i,j,k ≥ cdf t,f,i,j,k Tf,i
loc + T
f,l 1− RCS + X MS
Xf,s f,s
s=2j (20.34)
∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , j < i
2, l = i, ∀k ∈ Nk

 
total + T rep
Ct,f,i,j,k ≥ cdf t,f,i,j,k Tf,i f,l (20.35)
∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , j = 2i , l = i, ∀k ∈ Nk

The different arrangements of the FI, RCS, and MS are included in the planning
problem which was simulated on the RBTS-bus4 test system. The cost-benefit
analysis showed that placement of the FIs and MSs results in lower cost which is
a better option in case of the limited budget. While in case of an unlimited budget,
investment in FIs, MS, and RCSs causes to better reliability indices where average
energy not supplied (AENS) and SAIDI are reduced from 2.074 (kWh/yr.cust)
and 0.822 (hr/yr.cust) to 0.911 (kWh/yr.cust) and 0.392 (hr/yr.cust) in case of FI-
MS and FI-MS-RCS placement, respectively. These results prove that increasing
the automation level of the FLISR task can improve the reliability and economic
indices. This improvement is in line with enhancing the economic benefits taken
592 A. Yazdaninejadi et al.

Fig. 20.3 Variation of

equipment number by
variation of customer damage
function multiplier [12]

by the MG operator and also social welfare experienced by the consumers. From
the social welfare point of view, Fig. 20.3 shows the impact of customer damage
function (CDF) on the number of required FIs, MSs, and RCSs. This figure declares
that increasing the CDF, implies higher number of equipment. This trend is not clear
for FI, since the indication functionality of this device can be also performed by the
RCS. However, in the case of higher investment on the RCSs, the required number
of FIs is obviously decreased.
It is worth mentioning that all of the models which are described are linear
models. This issue is important since the optimality of the results derived by the
linear models can be guaranteed by the solvers. Also, the switch placement problem
is one of the hardest problems tackling this problem by the linear models, alleviates
the required computational burden.

20.3 The Need for Selective and Fast Protection Schemes

in MGs

Besides the fault identification process, at the fault occurrence time, protection
schemes of MG and its ingredients such as DERs could face different challenges.
As discussed, protection schemes should be selective and fast to preserve the DERs
stability and realize a successful FRT. Meanwhile, the essence of selectivity is first
discussed and then the concept of fast-response protection is tailored. Effective
solutions are then devised to annihilate these issues.
20 Fault Identification, Protection Schemes, and Restoration Requirements. . . 593

20.3.1 Importance of Selective Protection

MGs can maintain the power supply in the event of faults which reduce the
frequency and duration of load outages. In this case, the protection system should
be selective to avert the unintentional disconnection of loads and DERs. Satisfying
selectivity conditions in MGs entails an adaptive protection scheme for meeting
different issues.

20.3.1.1 Selective Dead Time in Reclosing Scheme of MGs

The fixed dead time associated with the conventional auto-reclosing methods may
imperil the transient stability of synchronous-based DERs [14]. During faults,
energy is rapidly absorbed by the deployed synchronous-based DERs and if the
energy cannot be given back to the network, the fault may bring instability. In this
regard, reclosers may accelerate the instability of MGs. Simulation results show that
auto-reclosing requires accurate time to prevent this issue [14]. In MGs and for the
case of three-phase faults, as critical clearing times (CCT) varies with dead time and
loading of synchronous-based DERs, providing adaptive reclosing time seems to be
essential.
A bit faster or slower dead time may cause stability issues in MGs. To clarify
this issue, the rotor angle of a synchronous-based DER in a simple MG is illustrated
in Fig. 20.4 considering different dead times. A three-phase fault is performed at
0 ms and is cleared at 220 ms. As can be seen, the synchronous-based DER is
stable for the dead times 200 ms and 400 ms. This is while; results show that the
DER would be unstable for the dead time of 300 ms. Additionally, the loading of
the synchronous-based DER is another factor in meeting transient stability. For
example, by increasing the load of MG, the synchronous-based DER would not
be stable for a dead time of 200 ms which is shown in Fig. 20.5. Therefore,
an adaptive auto-reclosing method is required to determine the most appropriate
reclosing instant. In [14], based on numerical protective devices, a proper adaptive
auto-reclosing method is presented to avert transient stability of the synchronous-
based DER in MGs. However, minimizing load interruptions for temporary faults is
still a challenging task.

Fig. 20.4 The rotor angle of

synchronous-based DER with
fault clearing time of 200 ms
in a simple MG [14]
594 A. Yazdaninejadi et al.

Fig. 20.5 The rotor angle of synchronous-based DER with fault clearing time of 200 ms in a
simple MG with different loading [14]

20.3.1.2 Nuisance Tripping of Protective Devices in MG

The presence of DERs in MGs may bring nuisance tripping of protective devices.
Current research trends pointed out that traditional protection schemes cannot guar-
antee selectivity tasks among protective devices in MGs which requires devising
new protection solutions. Here, this issue is illustrated among relay-relay and fuse-
recloser.

Relay-Relay

The power flow in MGs could be bidirectional which needs to deploy directional
protective devices. Time-current characteristics of the coordinated directional over-
current relays (DOCRs) are shown in Fig. 20.6. These relays are coordinated in the
presence of a DER connected to Bus A. As can be seen, for an occurred fault at
point P, the coordinated relays satisfy the selectivity task by meeting coordination
time interval (CTI). However, after the disconnection of DER, the discrimination
time among primary and backup relay is reduced. As can be seen, the current seen
by the primary relay gets increased. On the other hand, the current seen by the
backup relay is decreased. Eventually, the currents which are seen by those relays
reach together at a new point. Therefore, by disconnecting DER, the relays would
have different operation times for the same fault which may imperil the selectivity
task. Thereby, due care should be contemplated in the protection of MGs. A simple
glance in Fig. 20.6 shows that by increasing the difference between fault currents
before and after disconnecting the DER, this issue seems to be more sophisticated.
Subsequently, DERs can affect the coordination of DOCRs via their contributions
in fault currents. The contribution of each DER depends on its specific technology
[15]. Inverter-based DERs contribute to the fault currents in the range of 1–2 per unit
rated currents. Contrarily, synchronous-based DERs can influence the short-circuit
levels, significantly. Therefore, synchronous-based DERs affect the coordination of
DOCRs, remarkably.
20 Fault Identification, Protection Schemes, and Restoration Requirements. . . 595

Fig. 20.6 Relay-relay coordination in presence of DER

In MGs and during the faults, inverter-based DERs can also bring misco-
ordinations in the dominated fashion with the variation of short-circuit current.
Since inverter-based DERs are sensitive to bus voltage and frequency, during
abnormal conditions of frequency or voltage, the associated standards recommend
that inverter-based DERs should be disconnected from the MG. As an example, the
standard IEEE 929 recommends isolating photovoltaics from the MG based on the
abnormal situation presented in Table 20.3 [16]. From the viewpoint of the reported
scheme in this table, the closest inverter-based DER to the occurred fault should
be disconnected faster than the others. Therefore, the current seen by relays varies
during the same fault. In this manner, experiencing currents in staircase fashion by
protection devices brings a selectivity issue. To clarify this issue, consider a fault on
the photovoltaic-dominated network which is shown in Fig. 20.7. As can be seen,
the system consists of n independent photovoltaics. In this testbed, a fault incident
occurs at the end of the feeder. The primary protective device PD2 experience faults
current IF . During the fault, the backup protective device PD1 sees the current IS . In
this case, IS , IPV and IF are as follow:

U − V1
IS = (20.36)
Zs
596 A. Yazdaninejadi et al.

Table 20.3 Disconnection Terminal voltage Delay

scheme for photovoltaic
systems V < 0.5 6 cycles
0.5 ≤ V < 0.88 120 cycles
0.88 ≤ V < 1.1 Normal operation
1.1 ≤ V < 1.37 120 cycles
1.37 ≤ V 2 cycles
The grid frequency Delay
f < 59.3 6 cycles
59.3 ≤ f ≤ 60.5 Normal operation
60.5 < f 6 cycles

UP Vi − Vi
IP Vi = (20.37)
ZP V

US − V1 UP V1 − V1 UP Vi − Vi UP Vn − Vn
IF = + + ··· + + ··· +
Zs ZP V1 ZP Vi ZP Vn
(20.38)

Therefore:


n
IS = IF − IP Vi (20.39)
i=1

By disconnecting PVs, IS approaches to IF . In [17], it is proven that deployment

of any source unit between two protective devices increases fault currents seen by
primary relay. However, the current behind the deployed source unit is decreased.
According to (20.39), it can be concluded that disconnecting PVs decreases IF ,
though increases IS which puts the selectivity tasks at risk.

Fuse—Re-Closer

Unsuccessful reclosing endangers the reliability of MGs. For an occurred fault,

for example, F1 in Fig. 20.8a, DER does not allow the corresponding relay to
eliminate the fault by reclosing action, and hence, the fault arc remains to burn.
Typically, the DER is disconnected from the networks to avert such schemes. What’s
more, a probable delay in the disconnection of DER may lead to second reclosing
performance which increases the stress on the upstream equipment. Therefore,
devising a reliable reclosing scheme is of crucial importance.
Furthermore, fuse and recloser are coordinated when the characteristic of the
fuse is placed between the fast and slow characteristics of the recloser. Therefore,
the fuse should clear the permanent fault as the primary protection. In this regard,
20 Fault Identification, Protection Schemes, and Restoration Requirements. . . 597

PD: Protective device

PV: Photovoltaic
PD1 PD2

DC/AC DC/AC DC/AC DC/AC

PV1 PV2 PVn-1 PVn

ZS V1 ZL V2 Vn-1 ZL Vn IF
IS IPV1 IPV2 IPVn-1 IPVn
ZPV ZPV ZPV ZPV

PV1 PV2 PVn-1 PVn

I PD2 I PD1

t
t
V1 V2 V3 Vn
Is

PD2 PD1
Vs

PV1 PV2 PVn-1 PVn

Fig. 20.7 Photovoltaic-dominated MG

the fast characteristic of the recloser should clear the transient faults. By doing so,
the fuse would be saved. This is while; DER disconnection may cause a change
in fault current which poses miscoordination. For clarification purposes, Fig. 20.8b
exemplifies the characteristics of fuse and recloser in two different scenarios. At
the outset, DER is connected to the MG and the coordination task is met. In this
scenario, the fuse characteristic is placed between the fast and slow characteristics
of the recloser. On the other hand, in the second scenario, by disconnecting the DER,
the fault currents seen be protective devices are settled down in new points. In this
598 A. Yazdaninejadi et al.

= Fault
Primary Recloser
substation
Fuse
Common
coupling
HV/MV F1 point F2 F3
Transformer

DER

(a)
Fuse Before disconnection of
Recloser slow the DER
Recloser fast After disconnection of
F2 F3 the ERS
Ti m e

Distance

(b)
Fig. 20.8 Effect of DER on fuse-recloser coordination. (a): a simple testbed including fuse and
recloser, and (b): Time-current characteristics of fuse and recloser

case, the fuse clears the fault after the slow characteristic of the recloser and hence,
the fuse would not be saved for transient faults.

20.3.1.3 Blinding of Protective Devices along with Inverter-Based DERs

In MG protection, inverter-based DERs may bring blinding issues for the employed
protective devices. Usually, directions of fault currents are determined by employing
torque angles. Here, the aim is to show that sometimes the conventional direction
detection module cannot determine fault directions in MGs with inverter-based
20 Fault Identification, Protection Schemes, and Restoration Requirements. . . 599

Fig. 20.9 A simple MG in

islanding mode with an
inverter-based DER [18]

DERs. Simulation studies are used to show the effect of inverter-based DERs on
directional elements that work with a positive sequence angle. When the positive
sequence torque angle is in the first and fourth quadrants, the fault is in the forward
direction. On the contrary, when the positive sequence torque angle is in second
and third quadrants, the fault is in the reverse direction. Here, this issue is explained
with a counterexample. A simple MG is shown in Fig. 20.9. This testbed includes an
inverter-based DER which is connected to Bus-6. When the inverter-based DER is
disconnected from the MG, for a fault at bus 10, PD1 and PD2 can detect the fault in
the reverse direction. This fault is injected at 2 s and the presented torque angles for
PD1 and PD2 in Fig. 20.10 are placed in the second quadrant. By a simple glance
in Fig. 20.10, it can be easily concluded that PD1 and PD2 can easily determine
the direction of the fault current. On the contrary, real-time simulation results show
that when the inverter-based DER is connected and the MG is operated in islanding
mode, for the same fault, the torque angle of PD1 is settled in the first quadrant.
Therefore, in this case, PD1 is blinded and detects the fault in an incorrect direction.
Base on the abovementioned explanations, devising an efficient direction detection
method is of priority in MG protection.
600 A. Yazdaninejadi et al.

PD1 in case of inverter-based DER

PD2 in case of inverter-based DER
PD1 in case of synchronous-based DER
PD2 in case of synchronous-based DER

Torque angle
(degree)
90
0

1990 2050 2090

Time (ms)

Fig. 20.10 Torque angle of PD1 and PD2 when inverter-based DER is connected and disconnected
[18]

20.3.2 Essence of Fast-Response Protection in Restoration

DERs play a vital role in restoring power supply in MGs. Despite advantages,
the low inertia time constant of synchronous-based DERs and FRT requirements
of inverter-based DERs makes them more prone to instability due to power grid
faults. Therefore, the faults should be cleared swiftly for meeting transient stability
constraints in MGs. That is, the protection system should clear the fault faster than
CCTs of synchronous-based DERs and the required standard for inverter-based
DERs. For the sake of clarity, Fig. 20.11 shows a simple MG with two synchronous
DERs. In this figure, each DER has a CCT curve. In each point of the line, the
minimum CCT of DERs is considered as the total CCT. This CCT curve is shown
with a red curve. When the operation curve of the protective device is under this
curve, the MG is stable. For example, for a fault at point F, the curve associated with
the protective device is under the CCT curve. Thereby, this fault would be cleared
faster than the associated CCT. Accordingly, after clearing the fault, DERs remain
connected to the MG and restoration would be possible. Likewise, a protection
scheme in the presence of inverter-based DERs necessitates satisfying different
requirements such as FRT [19]. Based on the recommended standard for inverter-
based DERs, they are allowed to be connected during faults in a limited time. In this
case, the protection system should clear the fault before the allowed time. Therefore,
as well as meeting transient stability, meeting this requirement entails establishing
a fast-response protection scheme. However, measuring the bus voltage as another
index is necessary for meeting the FRT requirement. For example, the voltage profile
for FRT capability in different countries for wind power integration is shown in Fig.
20.12.
As pointed earlier, meeting the transient stability of DERs in MGs calls the need
for faster protection solutions. This is while; considering fault impedance in the
coordination process of protective devices is also recommended to keep selectivity
20 Fault Identification, Protection Schemes, and Restoration Requirements. . . 601

Synchronous -based DER SB-DER Stable Stable Unstable Unstable

CCT Total
CCT SB-DER 2
CCT SB-DER 1
PD 1

F
PD
HV/MV
Transformer
Protected line
: Fault

PD : Protective device
SB-DER 1 SB-DER 2

Fig. 20.11 Simple MG with two synchronous-based DERs

Voltage (%)
Germany
90
Spain
Denmark
75

US

Keep connected
25 above the curves

Time (ms)
150 500 750 1000 1500

Fig. 20.12 Voltage profile for FRT capability in grid codes for wind power integration

and hence, avoid possible miscoordination among these devices. Considering fault
resistance increases the number of selectivity constraints in the coordination process
of protective devices which makes it complicated. This complexity may result
in higher operation time for protective devices, especially in looped structure
MGs. On the other hand, simulation results on IEEE 14-bus test system with
high-level penetration of synchronous-based DERs [20] show that by increasing
fault impedance, stability issue is attenuated. This fact is exemplified by different
faults with different impedances which are explored in Table 20.4. Eight different
scenarios are considered with faults at the midpoint of lines in the IEEE 14-bus
602 A. Yazdaninejadi et al.

Table 20.4 Impact of fault Faults at F1-F8

impedance on CCT [20]
Rf = 0.2 Rf = 1 Rf = 2
Relay No. CCT CCT CCT
1 0.712 Stable Stable
2 0.712 Stable Stable
3 0.633 1.537 Stable
4 0.633 1.537 Stable
5 0.706 Stable Stable
6 0.706 Stable Stable
7 Stable Stable Stable
8 Stable Stable Stable
9 Stable Stable Stable
10 Stable Stable Stable
11 0.439 0.726 Stable
12 0.439 0.726 Stable
13 Stable Stable Stable
14 0.712 Stable Stable
15 Stable Stable Stable
16 Stable Stable Stable

test system. As can be seen, the associated CCTs are decreased by increasing
impedance. For example, R3 should clear the fault with Rf = 0.2  faster than
712 ms. Meanwhile; it should eliminate the fault with Rf = 1  faster than 1537 ms.
Moreover, for the impedance of more than 2 , there is not stability concern.
Typically, the coordination process of protective devices is followed through an
optimization model. In the light of abovementioned points, this model should be
devised in a way that the obtained settings for protective devices can meet fault
impedance and transient stability constraints, simultaneously.

20.3.3 Selective and Fast Solutions

Aiming at satisfying transient stability of synchronous-based DERs and meeting

FRT of inverter-based DERs, a fast-response protection scheme is required. In
this manner, numerical relays with various functionalities help to accomplish this
complicated task. Here, different schemes, strategies, and solutions are explored to
show the advantage of numerical relaying in MGs protection.

20.3.3.1 Multi Agent-Based Protection

Multi-agent approaches can be employed for MGs protection based on overcurrent,

differential, and phase comparison schemes. In a multi agent system, the protection
20 Fault Identification, Protection Schemes, and Restoration Requirements. . . 603

Agents

A For zone A-C For zone C-D For zone D-E

S1 S1 S2 S3 S5 S2 S3 S4 S5 S6 S7 S4 S6

S2 C D E
Backup layer
Primary
S3 S4 S7 S8
layer
B
layer
Bus

S5 S6

Synchronous-based DER1 Synchronous-based DER2

Fig. 20.13 Multi agent-based protection

area is divided into different zones. Each agent is responsible for the primary
protection of each zone and the backup protection of nearby zones. In each zone,
three layers are ready to clear faults with a different priority. In the first layer, the
corresponded agent measures the current in the near-end and far-end side of the
protected zoon, as depicted in Fig. 20.13. This layer is called the primary layer. The
secondary layer is responsible for protecting buses and is called the bus layer. Bus
layer is also seen in Fig. 20.13. In this way, the required measurements should be
deployed in the near-end sides of the connected lines to the associated bus and the
point of common coupling of each DER. The third layer corresponds to the backup
scheme for the primary and bus layers. In the case of failure in the primary layer or
bus layer operation, the third layer would clear the faults. As shown in this figure,
the backup layer includes the measurements of the primary layer and bus layer,
concurrently.
Furthermore, each measurement is also sent to adjacent agents for processing.
Agents by measuring different currents detect the faults and clear them, swiftly. As
pointed out before, multi-agent approaches can deploy for electrical grids based on
a differential scheme. To do so, different logics can be developed for clearing faults.
For example, when a fault occurs at zone C-D, the current seen by S3 and S4 are
not in the same direction. Therefore, the agent associated with this zone detects the
fault between S3 and S4. Moreover, this fault is also detected between S3 and S7.
This is while; the priority is with S3 and S4 as the primary protection layer. If the
fault is not cleared by the associated breaker in the side of S4, the breaker associated
with S7 clears the fault as the backup scheme. What is more, during faults at any
bus, the bus layer measures the currents for the considered differential scheme. To
illustrate this scheme, consider a simple fault at Bus D where the measurements
are S4, S6, and S7 to detect the occurred faults. In this case, the differential logic
604 A. Yazdaninejadi et al.

that is considered among these measurements is in priority, and the logic associated
with S3, S6, and S7 from the left side and the logic associated with S4, S6, and S8
from the right side would act as the backup schemes for the bus layer. As can be
seen, clearing faults with a multi-agent approach can be done swiftly and without
any delays. Thereby, it seems a proper solution for meeting requirements associated
with transient stability and FRT.
Multi-agent approach besides laterals protection cannot act without any delays.
In this case, multi-agent approach should be coordinated with downstream pro-
tections. Likewise, laterals protective devices entail satisfying coordination tasks
with upstream protective devices. In this case, inverse characteristics are employed
for the operation function of a differential scheme. Although employing inverse
characteristics increases the clearing time of faults, the coordination time interval
is not considered among the deployed characteristics in the multi-agent layer,
which offers faster protection compared with non-communication-based overcur-
rent schemes. In [21], based on the multi-agent approach, a new variable tripping
time differential protection scheme is proposed to meet the transient stability
requirements. Moreover, proper logic is considered among different layers to satisfy
the selectivity tasks.

20.3.3.2 Overcurrent Protection Multi-Inverse Characteristic

As pointed out earlier, meeting transient stability requirements entails faster pro-
tection scheme. On the other hand, numerical relays can pave the way for defining
arbitrary characteristics by the users which helps to alleviate the operation time of
relays in clearing faults [22]. As an example, the double-inverse characteristic which
includes two inverse curves in series is depicted in Fig. 20.14. Employing double-
inverse characteristics yields higher flexibility and extensibility in the relaying
process. For the sake of clarity, consider Fig. 20.15 which shows the superiority of
double-inverse characteristics in a simple coordination example. As can be seen,
the presented double-inverse characteristic not only satisfies the CTI with other
characteristics but also yields lower operation time compared with typical inverse
characteristics. However, due to employing two or more inverse characteristics,
coordination tasks must be also performed at intersection points.
Each overcurrent relay with double-inverse characteristic is governed by:
⎛ ⎞
  
⎜ ⎟ min 0, I F do,f − I F do,F 3
do,f ⎜ A ⎟
t1 = T DS do ⎜  + C⎟ × ;
1
⎝ I F do,f B ⎠ I F do,f − I F do,F 3
−1
Ip1do
(20.40)
20 Fault Identification, Protection Schemes, and Restoration Requirements. . . 605

Time (s)
Double inverse characteristics for Rb
Double inverse characteristics for Rp >CTI
CTI <
CTI <
t min
Distance
Rb Rp F1 F3 F2

Fig. 20.14 Double-inverse characteristic

Time (s)

>CTI
>CTI
t min
Distance
R B Fault Fault RC Fault Fault

Fig. 20.15 Simple example of coordination tasks with double-inverse characteristic

⎛ ⎞
  
⎜ ⎟ min 0, I F do,F 3 − I F do,f
do,f ⎜ A ⎟
t2 = T DS do ⎜  + C⎟ × ;
2
⎝ I F do,f B ⎠ I F do,F 3 − I F do,f
−1
Ip2do
(20.41)

do,f do,f
t do,f = t1 + t2 ; (20.42)
606 A. Yazdaninejadi et al.

where tdo, f reveals the operation time of overcurrent relays with double-inverse
characteristics. Moreover, IFdo, F3 is the current seen by relay for fault at F3. When
do,f
the fault current IFdo, f is more than IFdo, F3 , t1 is zero and when the current is
do,f
less than IFdo, F3 , t2 is zero. In this manner, the two inverse curves intersect at
fault point F3 which is presented in Fig. 20.14. Normally, the operation time of an
overcurrent relay is formulated by the first term of Eq. (20.40). The second term
in this equation acts as a binary variable. When the fault current is bigger than the
current at F3, it is zero. Likewise, for the other faults, it is one. The second term
in Eq. (20.41) works in the opposite. Due to providing faster protection schemes,
double-inverse characteristics are deployed in [22] for meeting the transient stability
of distributed generation.
Each section of a multi-inverse characteristic is controlled by separate and inde-
pendent variables. Hence, the number of coordination variables and coordination
points are increased which may lead to a complicated optimization problem. Hence,
a proper process is required to find optimal set points to determine the best number
of inverse curves in a multi-inverse characteristic. Another key point to notice is
the multi-inverse characteristics that should demonstrate the decreasing fashion.
Therefore, the proper constraint should be considered in intersection points like F3.

20.3.3.3 Overcurrent Protection with Multi-Function Characteristic

A DOCR detects the direction of faults and operates just for faults in the forward
directions. These relays are blocked for the faults in reverse directions. This is while;
these relays can be developed with another characteristic in reverse directions of
DOCRs as shown in Fig. 20.16. By doing so, each DOCR can act as two independent
DOCRs in different directions. This feature is realized based on programming
code on numerical DOCRs. These multi-function relay with two independent
characteristics are called dual-setting DOCR [23, 24].

Normal inverse
Very inverse
Time (s)
Extremely inverse

Reverse direction Forward direction

Multiple of pickup current

Fig. 20.16 Dual-setting directional overcurrent relays characteristics

20 Fault Identification, Protection Schemes, and Restoration Requirements. . . 607

Primary
Substation
DOCR

Protection area

HV/MV
Transformer Line A
R3 R2 R1

Fault

(a) Synchronous-based DER

Primary Dual-setting DOCR

Substation
Low bandwidth
communication link

Protection area
HV/MV
Transformer Line A
R3 R2 R1

Fault

Synchronous-based DER
(b)

Fig. 20.17 Protection scheme based on (a): conventional, and (b): dual-setting characteristics

Application of these relays increases the number of protective relays. Therefore,

satisfying CTI in the coordination process leads to a complicated problem. This
is while; the reverse side characteristic of these relays can be employed instead
of backup protection for the primary relay in the same direction. In this case, the
conventional backups can be blocked through the communication links which helps
to relax the conventional selectivity constraints and hence, alleviates the complexity
of the coordination process. Figure 20.17 illustrates the protection scheme with
conventional and dual-setting DOCRs. In the case of a conventional scheme, for
a fault at Line A, R1 is the primary relay which is backed up with R3. The concept
of this scheme is depicted in Fig. 20.17a. However, in the case of employing a
multi-function relay and for the same fault, the forward direction of R1 is the
primary protection which is backed up with the reverse direction of R2. In this
case, the forward direction of R3 as the conventional backup is blocked through
the communication link which is shown in Fig. 20.17b. Deploying multi-function
relays helps to diminish the tripping time of relays. Therefore, dual-setting schemes
608 A. Yazdaninejadi et al.

Time (s)

Reverse direction Forward direction

Fig. 20.18 Dual-setting DOCRs with double-inverse characteristics

are employed for meeting transient stability requirements of MGs in [20]. As

shown in Fig. 20.18, the aforementioned scheme can be developed by multi-inverse
characteristics to provide further alleviation in the optimization complexity and
hence reduces operation time of relays. It is worth noting that forward direction
and reverse direction characteristics of dual-setting relays acts irrespective of each
other. Consequently, the number of inverse curves in the forward direction can be
different from the reverse direction.

20.3.3.4 Time-Current-Voltage Characteristics

As discussed earlier, the conventional overcurrent relaying operates based on

conventional inverse-current characteristics. According to [25], microprocessor-
based DOCRs facilities devising new tripping characteristics based on voltage
index. Typically, a DOCR determines directions of the occurred faults by using the
phase of the voltage and current. In this regard, in addition to the three existing
current transformers; these relays also deploy three voltage transformers to measure
voltage and current of three-phases [25]. Therefore, based on these relays, it is
possible to deploy a voltage index in tripping characteristics. In the developed
characteristics, by increasing voltage drop, the operation time of relays should
be decreased. Therefore, the conventional characteristic should be multiplied to
decreasing function as follow:
 K
1 T DS × A
t= × B (20.43)
e1−V If ault
Ip −1

where the first term reduces the operation time of relays with an increasing voltage
drop. Here, V is the bus voltage and K is a constant coefficient. This function is
called time-current-voltage characteristic and is plotted in Fig. 20.19. As can be
seen, the voltage index decreases the characteristic by keeping inverse behavior
20 Fault Identification, Protection Schemes, and Restoration Requirements. . . 609

1.8

1.6

1.4
Operation time (s)

1.2

1.0

0.8

0.6

0.4

0.2

0
0
10 0.2 0
0.8 0.6 0.4
20 1
Vf (p.u)

Fig. 20.19 Time-current-voltage characteristics

which facilities the way for using this characteristic in the coordination process.
The conducted study in [26] shows that this characteristic offers a suitable scheme
for meeting the FRT requirement.

20.3.3.5 Dual Time–Current–Voltage Characteristics

As explained in previous section, time-inverse characteristics can be developed

based on numerical relays as a multi-function scheme like the dual-setting approach.
In this manner, time-current-voltage characteristics of numerical DOCRs can be also
developed as a dual-setting approach with the following formulations:
 K
1 T DSf w × A
tf w = ×  (20.44)
e1−V If ault B
Ipf w −1

 K
1 T DSrv × A
trv = ×  (20.45)
e1−V If ault B
Iprv −1

where fw and rv indicate forward and reverse directions, respectively. Deploying

dual characteristics and communication links helps to relax the conventional
constraints in the coordination process which yields fast response protection.
610 A. Yazdaninejadi et al.

Fig. 20.20 Simple VI EI

piece-wise linear

Time
characteristic [27]

NI

Piece-wise linear

Multiple of pickup current

The developed scheme in [26] based on dual time-current-voltage characteristics

considers FRT requirements in transmission level interconnected wind parks. The
proposed scheme not only reduces the operation time of relays but also eliminates
miscoordinations. These schemes can be also enhanced by a user-defined strategy.
To do so, in addition to TDS and Ip; the other coefficients A, B, and K should be also
considered in the optimization process as new variables.

20.3.3.6 Piece-Wise Linear Characteristic

The typical characteristics for overcurrent relaying are standard (normal) inverse
characteristic, very inverse characteristic, and extremely inverse characteristic.
However, numerical relays allow the user to follow arbitrary characteristics. This
arbitrary characteristic can include some linear lines that are so-called piece-wise
linear characteristic. This feature helps to obtain a characteristic with a positive
property of standard inverse, very inverse, and extremely inverse characteristics like
the characteristic that is portrayed in Fig. 20.20. This figure shows that this kind
of characteristic helps to reduce the operation time of relays. These characteristics
are suitable for radial networks [27]. As can be seen, the presented characteristic
in Fig. 20.20 is piece-wised on the fault currents. Moreover, fault currents belong
to parameters set. Therefore, this characteristic cannot be used in the coordination
process of looped MG as a linear approach and suitable optimization model would
be of interest.
20 Fault Identification, Protection Schemes, and Restoration Requirements. . . 611

20.4 Reconfigurable Topology of MG and Enhancing

the Restoration Capability

The FLISR task of the OMS ensures a fast fault locating, isolating, and restoration
phases. However, it may be possible to restore more interrupted consumers by some
switching actions, until the repair operations are accomplished. For this task, it is
necessary to consider the technical and operational constraints of the MG. Efficient
feeder reconfiguration of the MG which ensures the operation constraints like feeder
congestion, voltage security, and radiality of the grid, can enhance the restoration
capability of the OMS. In this field, the linear model proposed by the authors in [28,
29], is developed to make it compatible with the restoration requirement of the grid.
The final model is represented as below:

Minimize C EN S = K EN S NS
Pi,t (20.46)
i∈B t∈T


in
Pi,t + Pi,t
NS
− Pi,t
D
= L
Pi,j,t (20.47)
j ∈B,j =i


i,t + Qi,t − Qi,t =
Qin NS D
QL
i,j,t (20.48)
j ∈B,j =i

    
L
Pi,j,t = −Gij 2Vi,t − 1 + Bij θij,t + Gij Vi,t + Vj,t + ωi,j,t − 2 ξi,j,t + γi,j,t
(20.49)

    
i,j,t = Bij 2Vi,t − 1 + Gij θij,t − Bij Vi,t + Vj,t + ωi,j,t − 2 ξi,j,t + γi,j,t
QL
(20.50)

θij,t = −θj i,t (20.51)

   
−M 1 − ξi,j,t ≤ γi,j,t ≤ M 1 − ξi,j,t (20.52)
612 A. Yazdaninejadi et al.

1 
−Mξi,j,t ≤ γi,j,t − Vj,t − Vi,t + 2 ≤ Mξi,j,t (20.53)
2

−Mξi,j,t ≤ θij,t ≤ Mξi,j,t (20.54)

  
ξi,j,t ≤ 2 N Bus − 1 ∀t (20.55)
i,j ∈B,i=j

ξi,j,t = ξj,i,t (20.57)


ξi,j,t = 0 ∀ (i, t) | i ∈ F B, t < t rep (20.58)
j ∈B,j =i

The connectivity of a section which can be controlled by the RCSs is modeled

by the binary variable ξ i, j, t . In the last equation, the set FB, are the adjacent buses
of the faulty feeder section which should be remained de-energized until the repair
process of this section is completely accomplished in trep . This model can be further
developed to include other reliability indices, too. The operation constraints that
should be considered in the model are as follow:
 
V ≤ Vi,t  ≤ V (20.59)

  G2 + B 2
sqr i,j i,j
Ii,j,t = Pi,j,t
L
+ Pj,i,t
L
(20.60)
Gi,j

sqr  2
Ii,j,t ≤ I i,j (20.61)

Note that the linearity of the model is preserved.

20.5 Conclusion

This chapter tailored the fault identification, protection requirements, and restora-
tion of MGs. In this way, the essence of meeting transient stability constraints
and FRT requirements by protection systems besides the necessity of enhancing
20 Fault Identification, Protection Schemes, and Restoration Requirements. . . 613

fault detection and isolations solutions were explored. It was shown that meeting
transient stability and FRT of DERs paves the way for preventing the unintentional
disconnection of DERs which calls the need for providing selective and fast-
response protection schemes. In this regard, the main issues were introduced,
some suitable solutions were reported, and in particular, a general research line on
eminent protection issues was established. Moreover, it was shown that conventional
protection devices besides equipment for FLSRT can be jointly used for fast
restoration of MGs, fast restoration of DERs, improving equipment health, and
enhancing the reliability of MG. Thereby, the developed model of fault detection
and isolation techniques along with conventional protections were also discussed.

References

1. Martins, V. F., & Borges, C. L. (2011). Active distribution network integrated planning
incorporating distributed generation and load response uncertainties. IEEE Transactions on
Power Systems, 26(4), 2164–2172.
2. Le, D., Bui, D., Ngo, C., & Le, A. (2018). FLISR approach for smart distribution networks
using E-Terra software—A case study. Energies, 11(12), 3333.
3. Andishgar, M. H., Fereidunian, A., & Lesani, H. (2016). Healer reinforcement for smart grid
using discrete event models of FLISR in distribution automation. Journal of Intelligent & Fuzzy
Systems, 30(5), 2939–2951.
4. Dimitrijevic, S., & Rajakovic, N. (2015). Service restoration of distribution networks consider-
ing switching operation costs and actual status of the switching equipment. IEEE Transactions
on Smart Grid, 6(3), 1227–1232.
5. Abniki, H., Taghvaei, S. M., & Mohammadi-Hosseininejad, S. M. (2019). Reliability improve-
ment in smart grid through incorporating energy storage systems in service restoration.
International Transactions on Electrical Energy Systems, 29(1), e2661.
6. Reliability improvement form the application of distribution automation technologies- initial
results, Available www.smartgrid.gov.
7. Teng, J. H., Huang, W. H., & Luan, S. W. (2014). Automatic and fast faulted line-section
location method for distribution systems based on fault indicators. IEEE Transactions on Power
Systems, 29(4), 1653–1662.
8. Lwin, M., Guo, J., Dimitrov, N., & Santoso, S. (2018). Protective device and switch allocation
for reliability optimization with distributed generators. IEEE Transactions on Sustainable
Energy, 10(1), 449–458.
9. Fault location, isolation, and service restoration technologies reduce outage impact and
duration, Available www.smartgrid.gov.
10. Farajollahi, M., Fotuhi-Firuzabad, M., & Safdarian, A. (2016). Deployment of fault indicator in
distribution networks: A MIP-based approach. IEEE Transactions on Smart Grid, 9(3), 2259–
2267.
11. Izadi, M., & Safdarian, A. (2018). A MIP model for risk constrained switch placement in
distribution networks. IEEE Transactions on Smart Grid.
12. Farajollahi, M., Fotuhi-Firuzabad, M., & Safdarian, A. (2018). Simultaneous placement of fault
indicator and sectionalizing switch in distribution networks. IEEE Transactions on Smart Grid,
10(2), 2278–2287.
13. Billinton, R., & Wangdee, W. (2005). Approximate methods for event-based customer
interruption cost evaluation. IEEE Transactions on Power Systems, 20(2), 1103–1110.
14. Teimourzadeh, S., Davarpanah, M., Aminifar, F., & Shahidehpour, M. (2016). An adaptive
auto-reclosing scheme to preserve transient stability of microgrids. IEEE Transactions on
Smart Grid, 9(4), 2638–2646.
614 A. Yazdaninejadi et al.

15. Yazdaninejadi, A., Nazarpour, D., & Golshannavaz, S. (2017). Dual-setting directional over-
current relays: An optimal coordination in multiple source meshed distribution networks.
International Journal of Electrical Power & Energy Systems, 86, 163–176.
16. IEEE. (2000). IEEE recommended practice for utility interface of photovoltaic (PV) systems.
IEEE.
17. Papaspiliotopoulos, V. A., Korres, G. N., Kleftakis, V. A., & Hatziargyriou, N. D. (2015).
Hardware-in-the-loop design and optimal setting of adaptive protection schemes for distri-
bution systems with distributed generation. IEEE Transactions on Power Delivery, 32(1),
393–400.
18. Hooshyar, A., & Iravani, R. (2017). Microgrid protection. Proceedings of the IEEE, 105(7),
1332–1353.
19. Jia, K., Chen, J., Xuan, Z., Wang, C., & Bi, T. (2019). Active protection for photovoltaic DC-
boosting integration system during FRT. IET Generation, Transmission & Distribution, 13(18),
4081–4088.
20. Yazdaninejadi, A., Nazarpour, D., & Talavat, V. (2018). Optimal coordination of dual-setting
directional over-current relays in multi-source meshed active distribution networks considering
transient stability. IET Generation, Transmission & Distribution, 13(2), 157–170.
21. Aghdam, T. S., Karegar, H. K., & Zeineldin, H. H. (2018). Variable tripping time differential
protection for microgrids considering DG stability. IEEE Transactions on Smart Grid, 10(3),
2407–2415.
22. Yazdaninejadi, A., Nazarpour, D., & Talavat, V. (2019). Coordination of mixed distance and
directional overcurrent relays: Miscoordination elimination by utilizing dual characteristics for
DOCR s. International Transactions on Electrical Energy Systems, 29(3).
23. Sharaf, H. M., Zeineldin, H. H., & El-Saadany, E. (2016). Protection coordination for
microgrids with grid-connected and islanded capabilities using communication assisted dual
setting directional overcurrent relays. IEEE Transactions on Smart Grid, 9(1), 143–151.
24. Yazdaninejadi, A., Golshannavaz, S., Nazarpour, D., Teimourzadeh, S., & Aminifar, F. (2018).
Dual-setting directional overcurrent relays for protecting automated distribution networks.
IEEE Transactions on Industrial Informatics, 15(2), 730–740.
25. Saleh, K. A., Zeineldin, H. H., Al-Hinai, A., & El-Saadany, E. F. (2014). Optimal coordination
of directional overcurrent relays using a new time–current–voltage characteristic. IEEE
Transactions on Power Delivery, 30(2), 537–544.
26. Saleh, K. A., El Moursi, M. S., & Zeineldin, H. H. (2015). A new protection scheme
considering fault ride through requirements for transmission level interconnected wind parks.
IEEE Transactions on Industrial Informatics, 11(6), 1324–1333.
27. Ojaghi, M., & Ghahremani, R. (2016). Piece–wise linear characteristic for coordinating
numerical overcurrent relays. IEEE Transactions on Power Delivery, 32(1), 145–151.
28. Hamidi, A., Golshannavaz, S., & Nazarpour, D. (2017). D-FACTS cooperation in renewable
integrated microgrids: A linear multiobjective approach. IEEE Transactions on Sustainable
Energy, 10(1), 355–363.
29. Jianfeng, D. A. I., Yi, T. A. N. G., Yuqian, L. I. U., Jia, N. I. N. G., Qi, W. A. N. G., Ninghui, Z.
H. U., & Jingbo, Z. H. A. O. (2019). Optimal configuration of distributed power flow controller
to enhance system loadability via mixed integer linear programming. Journal of Modern Power
Systems and Clean Energy, 7(6), 1484–1494.
Chapter 21
Real-Time Testing of Microgrids

A. S. Vijay and Suryanarayana Doolla

21.1 Introduction

Power electronic (PE) converters and equipment used for the conversion, control,
and conditioning of electric power are ubiquitous in our everyday lives. PE
converters are also increasingly being used in the integration of renewable sources
[1] and in electric vehicles [2]. The advancements in semiconductor technology
and microelectronics/ computational capability have been the primary cause for
this. In today’s competitive market scenario, designers of such equipment need to
ensure that the equipment are rigorously tested before field deployment, to meet
various performance standards, at the same time ensuring minimum time to market
the products. Therefore, the increasing need arises for evolving novel techniques
for testing, which has been actively taken up by various research labs around the
world. The testing methodology developed has to provide accurate and reproducible
results for the complex equipment being tested, and yet at the same time should
be economical, scalable, and easy to implement. Safety, testing time, and space
requirements typically pose constraints and demand careful consideration as they
may be in conflict with each other [3]. Important aspect to the testing process is
the degree of detail of the model of the system under test: the model should be
detailed enough so that the dynamic metrics of interest are captured accurately
through the testing process, and at the same time be simple enough to enable
quick testing and meet real-time (RT) constraints. The very nature of PE converters
makes them nonlinear due to the switching of semiconductor devices and this
poses enormous challenges in their testing. The modern distribution system faces
the increasing penetration of these low inertia and nonlinear PE equipment, which

A. S. Vijay · S. Doolla ()

Department of Energy Science and Engineering, IIT Bombay, Mumbai, India
e-mail: suryad@iitb.ac.in

© Springer Nature Switzerland AG 2021 615

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4_21
616 A. S. Vijay and S. Doolla

makes the modeling, design, and testing of these systems a nontrivial task. The smart
active distribution networks and microgrids are undergoing major technological
transformations, and in this context, the system level interactions between the
various PE converters and the controllers and the communication networks used
for their management pose uphill challenges for their modeling and testing.
The typical stages in the design and validation of a system involve: (1) stating
down the objectives/requirements, (2) analysis and system design, (3) modeling,
(4) prototyping, (5) testing, (6) validation, and then (7) iterations depending on the
results. The testing of the product development can be done at various stages of the
process depending on the complexity and the degree of detail required. This chapter
will cover the topics related to the testing of microgrids. Section. 2 will outline
the various RT testing methods and the categories of simulations, while Sect. 3
highlights the concepts in digital RT testing. Sections. 4 and 5 elaborate on the
popular real-time emulation and HIL techniques. Section 6 presents the novel state
of the art and hybrid testing approaches which are becoming increasingly popular
and Sect. 7 presents the challenges and conclusions in the aspects of microgrid
testing.

21.2 Real-Time Testing Methods

Approaches for testing power electronic equipment include simulations (off-line,

online, and real-time), model in the loop (MIL), processor in the loop (PIL), rapid
controller prototyping, software in the loop (SIL), hardware in the loop (HIL)—both
controller HIL (CHIL) and power HIL (PHIL), RT power level emulation, hardware
test beds (HTBs) or platforms and newer hybrid or combined approaches. Of these,
off-line simulations, MIL, PIL, and SIL are purely software-based approaches and
typically based on a fixed time step and need not be carried out in RT. Online
simulators are transient stability simulators that typically interact and exchange
data with an online system (such as supervisory control and data acquisition
(SCADA), energy management systems (EMS), state estimators, etc.) and these
may also participate in decision-making process and control the real/physical
electrical network. The integration of hardware in the testing process demands
a RT simulation. For large systems such as smart distribution networks, off-line
digital simulations are commonly performed to verify the performance. Previously
large analog simulators were built for carrying out tasks and verifying the system
performance. The rapid growth of computational power and increase in processing
speed [4] had given rise to digital simulation which made data analysis, storage, and
processing extremely convenient. Off-line digital simulation is carried out through
an iterative process of solving the differential equations of the system through
suitable numerical integration techniques. Typically, this is implemented by the use
of a high-level language-based code on a computer. Normally, single processors
were sufficient to carry out the task, but with the increasingly complex systems
21 Real-Time Testing of Microgrids 617

of today, newer software having multiple processors (cores) make possible faster
testing and provide graphical user interface (GUI) blocks to assemble the system
components and carry out the simulation task.
Any approach to testing involves the modeling of the system as the first step.
Model-based design (MBD) is a popular approach in the industry which prioritizes
the model of the system in the design process. Model-based Systems Engineering
(MBSE) and simulations are becoming increasingly popular to cope with more
and more complex systems. Off-line simulation permits the modeling to be carried
out for any depth of detail, but at the expense of the computational burden and
simulation time. This is proven to be the cheapest. Issues related to safety are
least concern in this type. This time step (fixed or variable) is very critical and
depends on objective of the simulation [5]. Also critical is the integration method.
The time required to simulate larger systems is very high. The time taken by the
processor to completely solve one integration instance is known as the simulation
step time or execution time (or the time step). In off-line simulation, the execution
time is more than the numerical integration time step [6]. In the context of PE
converters, capturing the fast switching transients demands appropriate modeling
of the semiconductor devices and needs smaller time steps for the simulation (of the
order of nanoseconds) which can be achieved with modern field programmable gate
arrays (FPGAs). Typically, off-line simulations form the first step in the design and
development stage.
The next step in the testing process would be involving hardware in the testing
process. But this requires the simulation to be carried out in real-time (RT), as the
hardware has to interact with the simulation as if it were in the actual world. Thus,
RT simulation tries to capture real-world behavior in the process of testing. Real-
world time constraints impose stringent conditions on the time sequence of events
and demand the use of a simpler model (lesser degree of detail) giving rise to a
tradeoff between accuracy and hardware integration. In RT simulation, the time
taken for execution is less than the integration step size. RT simulations can be
classified into three types depending on the applications: (a) Hard RT applications,
which use a fixed time step since the real-time constraint must strictly be met
in every simulation time step: cases in which the hardware integration interface
sampling period is comparable (small) to the minimum simulation time step, (b)
Firm RT applications, wherein a variable simulation time step can be used with
restriction: cases in which the interface sampling period is larger than the minimum
simulation time step, yet lower than a certain value (such as 1 ms), and (c) Soft
RT applications which employ variable time-stepping fully (minimum restrictions):
cases with no hardware interfacing (or a very large interface sampling period). In
digital real-time simulators (DRTS), a large system is broken down into smaller
subsystems and simulation is performed (simultaneous) on parallel processors. This
enables the accurate replication of the time scale of events [7]. The digital real-time
simulation can be classified into two main categories: (1) a fully software based
DRTS and (2) HIL RT simulation or HILS. Figure 21.1 portrays the various types
of simulation approaches. In some studies, the EUT is required to be connected to
618 A. S. Vijay and S. Doolla

Fig. 21.1 Classification (types) of simulations

electric drive system or a distribution network making. These are very large systems
(especially the distribution network). Since this may not be practically realizable
due to economic and space constraints, the sections of hardware are connected to
much more complex systems by the RT simulation of these systems.
HILS can be categorized into CHIL (or software HIL (SHIL)), PHIL, and
mechanical HIL (used in the context of electric drives). The difference between
CHIL and PHIL approaches is primarily on the voltage and current values at the
interface. This interface is for the RT simulation and EUT. In CHIL, the power level
at the interface is very small (in the order of mW to W), where as in PHIL the
power levels can easily be in the range of tens to even hundreds of kW. PHIL makes
possible the easy and flexible realization experiments which may otherwise be risky,
difficult, and expensive. Figure 21.2 shows interconnection of various systems in
HILS. The next stage in the development process involves the experimentation
on a hardware prototype of the system. The prototyping stage gives conclusive
test results. This prototyping stage serves as final validation for the system under
consideration. For prototype testing, test conditions are close to the real-world
conditions. Table 21.1 summarizes the comparative analysis of the various testing
techniques against different metrics such as flexibility, cost, ease of implementation,
degree of detail, safety, stability, and space requirements.
RT emulation holds the promise for providing the realistic environments for
testing systems. An emulator consists of a PE converter, which is controlled such
that it mimics a desired source or load. Since the hardware is fixed and the software
(model) is completely reconfigurable, it is possible to study the effect of aging
components like sensors, noise issues, tolerances, etc. This is not possible with
21 Real-Time Testing of Microgrids 619

Fig. 21.2 Concept of HILS:

hardware (device or
equipment) interfaced to an
RT simulator

Table 21.1 A comparison of various testing approaches [1]

Parameter Off-line simulation RT simulation HIL Emulation
Degree of detail Highest (I) Medium (IV) High (III) Higher (II)
Ease of implementation Easiest (I) Easy (II) Medium (III) Tough
Flexibility/scalability Highest (I) Medium (III) Medium (III) High (II)
Safety Safest (I) Safe (II) Medium (III) Medium (III)
Stability High (I) Medium (II) Least (IV) Medium (II)
Space requirements Least (I) Less (III) Less (III) Medium (II)
Cost Least (I) Medium (II) Medium (II) High (III)

simulation unless we model them in detail. For testing large systems, parallel use
of small emulators is a good option. Hardware test benches and a combination of
the abovementioned testing techniques (hybrid approaches) have also been reported
in the literature for modeling testing and analyzing the control algorithms and
complexities associated with microgrid systems. Yet another way of classifying
simulations would be as multi-domain and cooperative/co-simulation [8]. The final
stage of testing could involve a fully functioning interactive virtual prototype, which
allows for an effective testing, and verification on both the component and system
levels and may even involve virtual test beds (VTB) based platforms.
620 A. S. Vijay and S. Doolla

21.3 Digital Real-Time Testing: Concept

In off-line simulations, detailed modeling of each and every component of the

system is possible and hence the output is highly accurate. It is known to consume
more time if the system is large and complex. Real-time simulation with hardware
uses a reduced model. Thus, there exists a tradeoff between the simulation accuracy
and run time. With today’s state-of-the-art processors it is possible to create a
balance between these two aspects [9]. The time scale of the simulation is crucial to
preserve the dynamic characteristics of the system. The difference between the two
types of simulations: off-line and RT, is highlighted in Fig. 21.3. For the off-line

Fig. 21.3 Types of simulation. (a) Off-line accelerated simulation (faster than real-time), (b) off-
line simulation (slower than real-time), and (c) RT simulation
21 Real-Time Testing of Microgrids 621

simulation case, the time for execution for each integration step can either be lesser
or (but is normally) greater than the integration time (RT clock or wall clock time)
[6]. Respectively these are either faster than RT simulation—known as accelerated
simulation, or slower than RT simulation as illustrated in Fig. 21.3a, b. As can
be seen from the figures, idle time and over runs occur in the faster and slower
cases, respectively [10]. The RT simulator should reproduce the internal variables
and outputs of the simulation accurately, within the same duration of time that the
actual physical system would do. This is a necessary condition if we want an RT
simulation to be valid. Therefore, for RT simulations the execution time should
always be lesser than the integration time step and hence the processor will always
have some idle time as is shown in Fig. 21.3c. This time permits the processor
to perform all necessary operations such as driving I/O to and from externally
interfaced equipment and devices.
The RT simulator waits until the clock ticks to the next time step to start the
computation, as against accelerated simulation in which the idle time is used to
compute the states at the next time step. Hence, RT simulation aims to replicate
the dynamics of the simulated system at the natural time scale of occurrence of
the events. However, due to the presence of nonlinear events (such as switching) in
the RT simulation, numerical instability occurs. This can be attributed to the nature
of the discrete time step solvers. As we all know successfully bringing about RT
synchronously is an uphill task.
The real-time simulators available in the market come with advanced I/O
cards. The sampling rates of these system are much higher than the fixed-step
simulations. This helps in overcoming the challenges. For a detailed comparison on
various commercial RT platforms one can refer to [7]. RT simulation platforms are
commonly used today to obtain closer to real-world results for the designed systems.
In the context of PE applications and power systems, smart grid technologies and its
development has been summarized in [11], which highlights that the way to address
the challenges is through RT testing, probably involving the combination of HILS
and RT simulations. Due to the immense computing power of today’s DSCs (digital
signal controllers) and FPGAs (field programmable gate arrays), computationally
intensive studies involving nonlinearities and magnetic saturation, etc. in the area of
PE and electric machines have also been carried out [12].

21.4 Hardware in the Loop Testing

A hardware interfaced with real-time simulator is termed as hardware in the loop

simulation (HILS). CHIL, PHIL, and MHIL are three classifications in HIL. The
controller hardware in the loop simulation involves the testing of the controller
board while the plant is modeled and simulated in the real-time simulator. Thus,
only signal level power exists at the interface (feedback signals from the plant and
command signals from the processor—which complete the loop). In the past, HILS
was mainly used as a testing method for electrical drives in the aircraft industry.
622 A. S. Vijay and S. Doolla

Fig. 21.4 CHIL and PHIL: (a) signal level exchange in CHIL, (b) power level exchange through
the interface in PHIL

Nevertheless, now it has become an industry practice for power electronic and power
system apparatus testing. HILS has become a popular approach in microgrids too
[13]. A detailed comparison of simulation and HIL alternatives is presented in [14].
The work emphasizes that in digital simulations, the simulation run time is the main
constraint. This is critical for systems involving high-frequency switching. These
systems demand use of small time steps, thus increasing the computational burden
on the processor.
PHIL involves actual power transfer at the interface through a power electronic
converter, unlike its counterpart (CHIL). The differences between the two types of
HILS are as indicated in Figure 21.4. The RT simulator consists of the simulation of
the rest of system (ROS) which may have a human machine interface (HMI) as well.
Employing PHIL would reduce the cost and late changes to design specifications
[3]. It is because realistic and detailed testing environments can be provided
through PHIL, which can permit real-world scenario studies before the equipment
is installed in the field. The details of various applications, including high-speed
generator testing and fault current limiter testing, are provide in the work. Also
available are experimental applications pertaining to ship propulsion and rotor
heating in motors (superconducting). All these applications use the PHIL approach.
Research presented in [15] summarizes the modeling and simulation aspects for
PHIL-based testing.
The issue of stability in PHIL testing and the relevant analysis of the same
has been discussed widely in the literature. The ideal interface will have unity
gain, infinite bandwidth, zero delay, and will be linear over the complete range of
21 Real-Time Testing of Microgrids 623

inputs. The methodology or approach setup for the exchange of signals between
the EUT and the RT simulator (or through the interface) is known as the interface
algorithm (IA). Two popular algorithms typically used in PHIL are compared in
terms of the accuracy and stability in [16]. Attempts have been made to address
the stability issues in the PHIL systems using virtual impedance implementation
in software [17], impedance measurement using online wideband system [18], etc.
Stability issues arising out of the type of interface (voltage or current) are discussed
extensively in [19]. Issues arising out of sensor noises and bandwidth time delay
are also highlighted. Other known issues of numerical instability caused by splitting
of system into multiple subsystems for simulation are discussed. The impedances
of the equipment under test and the virtual aspects of it play a critical role in
determining the stability.

21.5 Real-Time Emulation

The block diagram of a general-purpose real-time emulator is shown in Fig. 21.5.

The power electronic converter or emulator converter presents itself as source
or load to rest of the system. The load emulation is done by current generation
through the real-time simulation in closed loop control while source emulation

Fig. 21.5 Generalized emulator (source/load)

624 A. S. Vijay and S. Doolla

is through voltage generation. The characteristics of the required load or source

are implemented through differential /difference equations, which describe the
mathematical model of the particular entity. Load emulation involves the tracking
of currents while its dual-source emulation involves the tracking of voltages.
Emulators have predominantly been used for the testing of individual com-
ponents like sources, loads, impedances, etc.; however recent applications in PE
testing include converter impedance control emulation, energy storage emulation,
and even the emulation of nonlinearities in the system [20–22]. Separate load and
source emulation is very well known and accepted by the industry. A new area of
system emulation through power electronic interface has gained recent attention.
For microgrid and distribution system studies, having a hierarchical system in
laboratory scale is very challenging. Same is the case with distributed controls. It
is almost impossible to put together all of them on a single system. If we like to
reproduce the hierarchical layers of the control scheme, proper emulation methods
need to be employed which may not be easily achieved on off-line simulators. A
detailed review of renewable energy system emulators is reported in [23]. The work
presented in [24] reports a microgrid system emulator based on a single bidirectional
converter. Another recent area where emulation is being widely used is in the testing
of the drive trains of electric vehicles [25, 26].

21.6 Novel/Hybrid Testing Approaches

A combination of real-time, PHIL, and emulation approaches to enhance the testing

performance is gaining attention of both industry and academia. Software in the loop
(SIL) is another approach which refers to the software-only simulation of two or
more subsystems on the same RT simulator. Closed loop simulation of two or more
subsystems on different real-time simulation platforms is known as Co-operative
simulation or Co- Simulation [27]. It refers to the case in which the subsystems
(of a larger complex system) with different modeling and run-time environments
are solved using a coordinated approach. This technique is employed for the
simulation of complex systems involving multi-domain and multi-rate components
(the power and communication layers) such as in smart grids [28]. The driving
force behind co-simulation is due to the fact that the subsystems individually
employ numerous established technologies and simulation tools. Depending on
the framework used the co-simulation employed can be further categorized as (1)
generic co-simulation and (2) specific co-simulation. If the model of the system and
the run-time environment/solver are integrated on the same platform, the simulation
is said to be the classical simulation. However, if the model of the complete
system is developed on one platform but executed on different solvers, it falls
under the category of parallel or distributed simulation. An alternative to this
approach is the hybrid or mixed simulation approach in which separately developed
representations (models) of various subsystems are solved on the same solver. Apart
from the abovementioned methods, hardware test benches/beds and high-power
21 Real-Time Testing of Microgrids 625

Fig. 21.6 Illustration of a generalized testing platform for a microgrid, using a combination of
testing approaches

level prototypes are also used extensively to test designed microgrid controllers and
equipment [14, 29]. Figure 21.6 depicts a scenario in which a large system is tested
using a combination of methods such as co-simulation, RT emulation, and HIL [30,
31].
With the increased smartness in the system due to intelligent electronic devices
(IED) and information and communication technologies (ICT), the testing frame-
work needs to account for the interaction between various communication layers
and associated latencies, which render complexities. In the Internet of Things (IoT)
scenario, smart microgrids are becoming ever more intelligent and self-resilient
compared to the conventional system [32]. A new concept called the internet of
energy (IoE) has emerged [33], which is a subsector of the internet of things and
involves the use of advanced digital controllers, sensors, meters, and actuators with
the ability of information exchange through IT networks. However, these result
in increased structural and communication network complexities. To address the
microgrid management and protection objectives an efficient and highly reliable
communication network is necessary. The latency requirements and security of
the communication network are vital aspects that need to be addressed. Artificial
626 A. S. Vijay and S. Doolla

Intelligence (AI) and deep learning algorithms and fog computing/ clouds are
being explored to constantly enhance existing services, and technologies such
as edge computing and smart systems have already been proposed in literature
to improve the system functionality and power quality. Transaction Management
Platforms (TMP) based on the Internet of Things (IoT) is a recent approach wherein
individual prosumers can engage in interactions, negotiate with each other, enter
agreements, and make proactive run-time decisions—individually and collectively
depending upon the energy demands and environmental conditions. The challenges
of privacy, trust and providing resilience is challenging—and rigorous algorithms
which seek to provide solutions need to be extensively tested, which demand that
the testing platforms provide both the cyber physical and power level capability and
flexibility. Another vital issue that needs to be addressed is that of the vulnerabilities
caused by cyber-attacks and data corruption, which again demand sophisticated
communication layers that need to be carefully modeled and emulated and validated
for their robust performance. To achieve a safe, resilient, efficient, and dispatchable
energy network, the cyber-security solution may need both hardware and software-
based mechanisms providing several layers of defense against cyber intrusions.
The use of blockchain-based mechanisms as a distributed solution for the
management of energy transactions in modern distributed systems is another recent
trend [34, 35]. Blockchains are distributed databases that maintain time stamped
lists of records and permit transactions among peers (P2P) without intermediary or
central institutions. Using crypto-currencies for monetary transactions is one of the
several possibilities offered by blockchain in the energy field. The key properties of
the blockchain include the maintenance of the consensus mechanism throughout the
network, storage of the data as a ledger into blocks, synchronization of the whole
ledger throughout the network, and provision of decentralized data. The blockchain
technology provides good cyber resilience to the microgrid, with its auditable and
robust data management characteristics and elimination of monopoly and providing
transparency and security. The testing of these management algorithms also needs
highly flexible reconfigurable platforms, which can capture the interaction between
the communication and hardware worlds.
Another trend which is quite popular in industry is a platform involving a real-
time digital replica of the product with HIL capability. This type of system, which
is known as a “Digital Twin,” is an approach compatible with model-based design.
Digital twin is a significant step toward the achievement of smart manufacturing and
provides a new paradigm for test validations and fault diagnosis. It simulates and
maps the complete product life cycle through realistic and dynamically optimized
simulation models [36, 37], thus reducing the cost of the design and maintenance
cycles. However, manufacturing companies face an uphill challenge to incorporate
digital twins due to the absence of verified and validated simulation models.
The main advantage of carrying out “virtual validation and commissioning” is
reduced physical design and commissioning time. The information generated by the
digital twin can be used to improve future predictive simulations, and the obtained
analytics can be used to carry out predictive and condition-based maintenance. The
digital twin is constantly optimized during the product design, development, and
21 Real-Time Testing of Microgrids 627

production phases. These experiment-able Digital Twins (EDTs) have a promising

potential to ease the model generation process for a complex system of systems
such as a smart microgrid and may offer the required degree of interoperability
and flexibility necessary for models spanning different disciplines and models.
Agent-based modeling (ABM) and AI-based deep learning algorithms may be used
to enhance the optimization of the digital twins. Thus, the analysis and design
phases are mixed with the anticipatory predictive diagnostics and this could well
hold the key for the efficient and effective testing of the complex smart microgrid
subsystems.

21.7 Conclusions

Microgrid systems and their applications have undeniable importance in the modern
smart distribution network which is becoming ever more decentralized, digitized,
decarbonized, and distributed. The testing of such systems is important in their
design and development cycle. A very flexible and accurate testing environment is
provided by both HIL and RT emulators. Figure 21.7 depicts the various approaches
adopted for the testing of microgrids around the world. Some of these systems
are widely deployed in academic institutions and research labs. However, issues
related to bandwidth, time step, stability, sensors and their interface have yet to be

Fig. 21.7 Testing platforms for microgrids around the world

628 A. S. Vijay and S. Doolla

overcome. As microgrids continue to evolve into much faster systems with growth
in size and complexity, and with the ever-evolving stringent standards governing
their control, various events and modifications may occur of a sudden, and if a
conventional centralized control is used, it may hit the limits. A possibility in the
area lies in the development of approaches to remotely perform the hardware in
the loop testing. This could be addressed from more research and exploration on
distributed HIL testing based on model reference adaptive control (MRAC).
The degree of detail required in the final results, constrain the testing method.
This would indirectly effect the cost of the system. Choice of a method is
case specific and hence a general comment of the superiority of one approach
over the other cannot be made. The RT simulation methods discussed in this
chapter are typically used independently for development and testing. These could
then be combined together to form a powerful tool for the testing of integrated
systems. Interoperability and intellectual property sharing of the models developed
is necessary to overcome the limitations. A thorough analysis of the models,
the assumptions involved in their simplification (model order reduction) and the
limitations or uncertainty quantification is the need of the hour. Therefore, a
collective effort between different research facilities and manufacturers would
prove beneficial. This will possibly accelerate the development of these testing
technologies, reducing the costs and time scales involved. More focus on the
interoperability and standardization of various tools and methods with attention to
their possible combinations for optimizing the performance of the overall testing of
complex systems is required. As microgrids and active distribution systems continue
to evolve into more efficient, smart, and complex systems, the multidimensional
challenges which arise must be handled using new methods and comprehensive
tools.

Acknowledgments The authors would like to thank the support from India-UK Centre for
Education and Research in Clean Energy (IUCERCE), a project funded by Department of Science
and Technology, India.

References

1. Vijay, A. S., Doolla, S., & Chandorkar, M. (2017). Real-time testing approaches for micro-
grids. IEEE J. Emerg. Sel. Topics Power Electron., 5(3), 1356–1376. https://doi.org/10.1109/
JESTPE.2017.2695486.
2. Adzic, E. M., Adzic, M. S., Katic, V. A., Marcetic, D. P., & Celanovic, N. L. (2013). Develop-
ment of high-reliability EV and HEV IM propulsion drive with ultra-low latency HIL environ-
ment. IEEE Trans. Ind. Informat., 9(2), 630–639. https://doi.org/10.1109/TII.2012.2222649.
3. Steurer, M., Meeker, R., Schoder, K., & McLaren, P. (2011). Power hardware in- the-loop: A
value proposition for early stage prototype testing. In Proceedings of the IEEE international
37th annual conference of industrial electronic society (IECON ‘11), November 2011, IEEE
(pp. 3731–3735).
4. Issacs, A. (2017). Simulation technology: The evolution of power system modelling. IEEE
Power and Energy Magz., 15(4), 88–102. https://doi.org/10.1109/MPE.2017.2690527.
21 Real-Time Testing of Microgrids 629

5. Siqueira, J. C. G., Bonatto, B. D., Marti, J. R., Hollman, J. A., & Dommel, H. W.
(2015). A discussion about optimum time step size and maximum simulation time in
EMTP-based programs. Int. J. of Electric Power and Energy Syst., Elsevier., 72, 24–32.
https://doi.org/10.1016/j.ijepes. 2015. 02.007.
6. Popovici, K., & Mosterman, P. J. (2013). Real-time simulation Technologies: Principles,
methodologies, and applications (p. 645). Boca Raton: CRC Press. https://doi.org/10.1201/
b12667-1.
7. Faruque, M. D. O., et al. (2015). Real-time simulation technologies for power systems design,
testing, and analysis. IEEE Power Energy Technol. Syst. J., 2(2), 63–73. https://doi.org/
10.1109/JPETS.2015.2427370.
8. Lauss, G. F., Faruque, M. O., Schoder, K., Dufour, C., Viehweider, A., & Langston, J. (2016).
Characteristics and design of power hardware-in-the-loop simulations for electrical power
systems. IEEE Transactions on Industrial Electronics, 63(1), 406–417. https://doi.org/10.1109/
TIE.2015.2464308.
9. Chen, Y., & Dinavahi, V. (2013). Multi-FPGA digital hardware design for detailed large-scale
real-time electromagnetic transient simulation of power systems. IET Genr. Transm. Distrib.,
7(5), 451–463. https://doi.org/10.1049/iet-gtd.2012.0374.
10. Marti, J. R., & Lin, J. (1989). Suppression of numerical oscillations in the EMTP power
systems. IEEE Transactions on Power Apparatus and Systems, 4(2), 739–747. https://doi.org/
10.1109/59.193849.
11. Dufour, C., & Belanger, J. (2014). On the use of real-time simulation technology in smart grid
research and development. IEEE Transactions on Industry Applications, 50(6), 3963–3970.
https://doi.org/10.1109/TIA.2014.2315507.
12. Chen, Y., & Dinavahi, V. (2012). Digital hardware emulation of universal machine and
universal line models for real-time electromagnetic transient simulation. IEEE Transactions
on Industrial Electronics, 59(2), 1300–1309. https://doi.org/10.1109/TIE.2011.2157296.
13. Panwar, M., Lundstrom, B., Langston, J., Suryanarayanan, S., & Chakraborty, S. (2013).
An overview of real-time hardware-in-the-loop capabilities in digital simulation for electric
microgrids. In Proceedings of the IEEE North Amer. Power Symp. (NAPS ‘13), September
2013, IEEE (pp. 1–6).
14. Sanchez, A., Castro, A., & Garrido, J. (2012). A comparison of simulation and hardware-in-
the- loop alternatives for digital control of power converters. IEEE Trans. Ind. Informat., 8(3),
491–500. https://doi.org/10.1109/TII.2012.2192281.
15. Edrington, C. S., Steurer, M., Langston, J., Mezyani, T. E., & Schoder, K. (2016).
Characteristics and design of power hardware-in-the-loop simulations for electrical power
systems. IEEE Transactions on Industrial Electronics, 63(1), 406–417. https://doi.org/10.1109/
TIE.2015.2464308.
16. Hatakeyama T, Riccobono A, Monti A. Stability and accuracy analysis of power hardware in
the loop system with different interface algorithms. In: Proceedings of the IEEE 17th workshop
control modeling power electron. (COMPEL ‘16), June 2016, IEEE; 2016. p. 1–8.
17. Yin, C., et al. (2016). Virtual impedance method of the power hardware-in-the loop simulation
to improve its stability and accuracy. In Proceedings of the IEEE 8th international power
electron. Motion control conference (IPEMC-ECCE Asia ‘16), may 2016, IEEE (pp. 2752–
2758).
18. Liegmann, E., Riccobono, A., & Monti, A. (2016). Wideband identification of impedance to
improve accuracy and stability of power-hardware-in the loop simulations. In Proceedings of
the IEEE international workshop Appl. Meas. Power Syst. (AMPS ‘16), September 2016, IEEE
(pp. 1–6).
19. Dargahi, M., Ghosh, A., & Ledwich, G. (2014). Stability synthesis of power hardware-in-
the-loop (PHIL) simulation. In Proceedings of the IEEE PES general meeting conference
exposition, July 2014, IEEE (pp. 1–5).
20. Daniil, N., & Drury, D. (2016). Improving the stability of the battery emulator - pulsed current
load Interface in a power hardware in-the-loop simulation. In Proceedings of 42nd annual
conference of the IEEE industrial electronics society (IECON ‘16), October 2016, IEEE (pp.
2064–2069).
630 A. S. Vijay and S. Doolla

21. Li, S., Qi, W., Tan, S. C., Hui, S. Y., & Tse, C. K. (2018). A general approach to programmable
and reconfigurable emulation of power impedances. IEEE Transactions on Power Electronics,
33(1), 259–271. https://doi.org/10.1109/TPEL.2017.2663424.
22. Shen, Z., & Dinavahi, V. (2017). Dynamic variable time-stepping schemes for real-time
FPGA-based nonlinear electromagnetic transient emulation. IEEE Transactions on Industrial
Electronics, 64(5), 4006–4016. https://doi.org/10.1109/TIE.2017.2652403.
23. Araujo, E. P., Rosell, P. O., Mane, M. C., Robles, R. V., & Bellmunt, O. G. (2015). Renewable
energy emulation concepts for microgrids. Renew. Sustain. Energy Rev. Elsevier., 50, 325–345.
https://doi.org/10.1016/j.rser.2015.04.101.
24. Vijay, A. S., Chandorkar, M. C., & Doolla, S. A system emulator for AC microgrid
testing. IEEE Trans. Ind. Electron. 2019; IEEE Early Access. https://doi.org/10.1109/
TIA.2019.2942275.
25. Fajri, P., Lee, S., Prabhala, V. A. K., & Ferdowsi, M. (2016). Modeling and integration
of electric vehicle regenerative and friction braking for motor/dynamometer test bench
emulation. IEEE Transactions on Vehicular Technology, 65(6), 4264–4273. https://doi.org/
10.1109/TVT.2015.2504363.
26. Rabia, S., & Tang, S. (2013). Control Laws for the emulation of an electric vehicle drivetrain
by two electric machines. In Proceedings of the IEEE vehicle power and propulsion conference
(VPPC ‘13), October 2013, IEEE (pp. 1–7).
27. Alvarez, A. F., Garcia, M. P., Valderas, M. G., Lopez, J., & Sanz, M. (2017). HW/SW
co-simulation system for enhancing hardware in the loop of power converter digital con-
trollers. IEEE J. Emerg. Sel. Topics Power Electron., 5(4), 1779–1786. https://doi.org/10.1109/
JESTPE.2017.2739710.
28. Lundstrom B, Chakraborty S, Lauss G, Brundlinger, Conklin R. evaluation of system-
integrated smart grid devices using software- and hardware-in-the-loop. In: Proceedings of
the IEEE power energy society innovative smart grid technology conference (ISGT ‘16),
September 2016, IEEE; 2016. p. 1–5.
29. Caldognetto, T., Santa, L. D., Magnone, P., & Mattavelli, P. (2017). Power electronics
based active load for unintentional islanding Testbenches. IEEE Transactions on Industry
Applications, 53(4), 3831–3839. https://doi.org/10.1109/TIA.2017.2694384.
30. Technologies, O., & Electrification Magazine, I. E. E. E. (2017). Vol., 2, 5.
31. Donahue, E. J. (2019). Microgrids: Applications, Solutions, Case Studies, and Demonstrations,
InTech Open Access Chapter. https://doi.org/10.5772/intechopen.83560.
32. Shahinzadeh H, Moradi J, Gharehpetian G B, Nafisi H, Abedi M, Internet of Energy (IoE) in
Smart Power Systems In: Proceedings of 5th Conference on Knowledge Based Engineering
and Innovation (KBEI ‘19) 2019. p. 627–636.
33. Moness, M., & Moustafa, A. M. (2016). A survey of cyber-physical advances and challenges
of wind energy conversion systems: Prospects for internet of energy. IEEE J Int. of Things.,
3(2), 134–145. https://doi.org/10.1109/JIOT.2015.2478381.
34. Di Silvestre, M. L., Gallo, P., Ippolito, M. G., Sanseverino, E. R., & Zizzo, G. (2018). A
technical approach to the energy Blockchain in microgrids. IEEE Trans. Ind. Informat., 14(11),
4792–4803. https://doi.org/10.1109/TII.2018.2806357.
35. Musleh, A. S., Yao, G., & Muyeen, S. M. (2019). Blockchain applications in smart
grid–review and frameworks. IEEE Access, 7, 86746–86757. https://doi.org/10.1109/
ACCESS.2019.2920682.
36. Xu, Y., Sun, Y., Liu, X., & Zheng, Y. (2019). A digital-twin-assisted fault diagno-
sis using deep transfer learning. IEEE Access, 7, 19990–19999. https://doi.org/10.1109/
ACCESS.2018.2890566.
37. Khan A, Dahl M, Falkman P, Fabian M, Digital Twin for Legacy Systems: Simulation Model
Testing and Validation In: Proceedings of IEEE 14th International Conference on Automation
Science and Engineering (CASE ‘18) 2018. p. 421–426.
Index

A Artificial intelligence (AI), 626

ABC, see Artificial bee colonies (ABC) Automation devices
ACMGs, see AC microgrids (ACMGs) and level of automation, 584, 585
AC microgrids (ACMGs), 407–408 Average energy not supplied (AENS), 591
DGs, 376
power losses, 375
stability analysis, 379, 380 B
ACO, see Ant colony optimization (ACO) Balanced positive sequence control (BPSC)
AC solid state circuit breakers (AC-SSCB), method, 410–411
543 Battery energy storage (BES), 98, 136–137,
Adaptive directional overcurrent protection, 143, 144, 222, 229, 237–238, 407,
557–558 552
Adaptive fuzzy proportional integral-derivative Battery–flywheel, 143, 145
(AFPID) controller, 312–313, 319, Battery–fuel cell, 141–142
322 Battery supercapacitor, 141, 142
design, 307 BBO, see Biogeography based optimization
NSIDE algorithm, 308 (BBO)
renewable energy sources, 307 BES, see Battery energy storage (BES)
voltage and frequency, 322, 325 Biogeography based optimization (BBO), 221,
Adaptive protection for AC microgrids, 553, 225, 233, 236, 237, 240
555–571 Blockchain-based mechanisms, 626
Adaptivity, protection scheme, 539, 592 Blockchain technology, 626
Advanced metering infrastructure (AMI), 68 BPSC method, see Balanced positive sequence
AFPID controller, see Adaptive fuzzy control (BPSC) method
proportional integral-derivative
(AFPID) controller
Agent-based modeling (ABM), 627 C
AI, see Artificial intelligence (AI) CAES, see Compressed air energy storage
AI-based deep learning algorithms, 627 (CAES)
AMI, see Advanced metering infrastructure Centralized communication-based adaptive
(AMI) protection, 555–556
Ant colony optimization (ACO), 221, 223–225, Central moment-based algorithm
244 adaptive protection schemes, 490
Artificial bee colonies (ABC), 221, 228–229, DC/AC converters, 489
234, 244, 513 differential protection, 492–498

© Springer Nature Switzerland AG 2021 631

A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems,
https://doi.org/10.1007/978-3-030-59750-4
632 Index

Central moment-based algorithm (cont.) frequency and voltage responses,

power system, 491 266–269
Smart Grids, 489, 490 mathematical analysis, 256–260
CERTS, see Consortium for Electric Reliability RESs, 262–265
Technology Solutions (CERTS) synchronous generator, 260–262
Circuit breaker (CB) technology, 542–543, 584 DC microgrid
Closed loop simulation, 624 converter-based generator, 270–273
Communication-based adaptive protections mathematical analysis, 269–270
for AC microgrids, 556–561 Conventional protection of microgrids,
centralized, 555–556 543–553
decentralized, 556 Conventional service restoration procedure,
Communication-less adaptive protection, 582, 583
568–569 Co-operative simulation, 624
Communication networks, 616 Co-simulation, 624
Compressed air energy storage (CAES), 132, Cost-benefit analysis, 591
134–136 Critical clearing times (CCT), 593–594
Consortium for Electric Reliability Technology Crypto-currencies for monetary transactions,
Solutions (CERTS), 3 626
Constraints Cuckoo search algorithm (CSA), 221, 227,
ADMM, 211 233, 234, 239, 244
charge and discharge rate limit, 48 Current symmetrical components based
in coordination process, 601 protection, 558–559
distributed generation, 110 Customer damage function (CDF), 592
economic and space, 618 Cyber-attacks, 626
energy storage system, 110–111 Cyber resilience, 626
frequency, 92
generation capacity limit, 48
NMCs, 25–26 D
objective functions (see Objective Data corruption, 626
functions) DCMGs, see DC microgrids (DCMGs)
power balance, 48, 109–110 DC microgrids (DCMGs), 375–377, 408–409,
reserve, 49–50 571–574
Control in MGs DCS, see Distributed control strategy (DCS)
AC MGs, 407–408 Deadbeat current control, 408, 433–434
DC MGs, 408–409 Decentralized communication-based adaptive
natural phase-coordinates approach protection, 556
amplitudes and phase angles, 413–414 Decision-making system
generalized NPC method, 417–418 assessment tools
phase-coordinates approach, 413 load flow, 65
phasor graphs, 415 load forecast/estimation, 65
symmetric NPC method, 415–417 network and component modeler, 64
three-phase currents, 414 remedial action scheduling tool, 64–65
traditional reference current control security assessment tool, 64
methods short circuit calculation, 65
BPSC, 410–411 state estimation, 65
decoupled double synchronous uncertainty assessment, 65
reference control method, 411–412 optimization tools
Controller hardware in the loop (CHIL), 616, decision making under uncertainty, 66
618 economic dispatch, 66
signal level exchange, 622 network reconfiguration tool, 66
Control mode change, DER units, 528 optimal power flow, 66
Conventional droop methods unit commitment, 66
AC microgrid restoration tools, 67
ESS, 266 UC and ED, 69
Index 633

Decision procedure Directional overcurrent relays (DOCRs)

decision algorithm, 193–196 coordination
DR cost modeling, 191–193 on-line implementation, 476
microgrid cost modeling, 189–191 optimization problem, 468–477
Decoupled double synchronous reference results and discussion, 478–485
control method, 411–412 test systems, 476–478
Deep learning algorithms, 626 dual-setting, 608
Demand response (DR) programs, 50–51, energy management scenarios, 229
100–101 Distance protection, 559
cost modeling with a fixed time delay, 529, 530
IDR-based, 193 Distributed control, 9–10
PDR-based, 191–192 AC MGs
decision algorithm, 193–196 frequency control, 283–285
with DR program, 120–124 voltage control, 285–286
economic problem, 181 applications, 445
green systems, 180 DC MGs
mathematical model, 180 DG model, 279–280
numerical studies distributed secondary control, 280–281
commercial load, 198–199 MCA and MCB, 30
final deduction, 201 MG control system, 275
hospital load, 199–201 in NMCs, 35–36
industrial load, 197–198 single system, 624
price-based, 99 SOC balancing, 295
types Distributed control strategy (DCS), 32, 35–36
IDR strategies, 183–184 Distributed energy resources (DERs), 581
PDR strategies, 181–183 and BESS, 552–553
programs and microgrid operation, 184 best utilization of, 21
without DR program, 116–120 converter-based, 536
See also Decision procedure energy users, 14
Department of energy (DOE), 584–585 fault current injections, 404
DERs, see Distributed energy resources flexibility, 151
(DERs) FRT technique, 404, 581
DER unit fault-ride-through (FRT) grid-connected mode, 548, 550
requirements, 526 integration, 151
DG units, see Distributed generation (DG) inverter-based, 406, 598–600
units KKT conditions, 206
Differential protection methodology description, 206–207
busbar differential protection, 513, 515–516 MGs, 70
factors, 442 on-site generation sources, 185
multipoint measurements, 439 optimal switch allocation, 582
principle optimal target configuration, 582
current transformer saturation, 497–498 optimization problem, 234
negative-sequence, 494–497 power
percentage, 492–494 grid, 185
reactor, 509, 513, 514 systems, 4
variable tripping time, 604 problem formulation
Digital real-time simulators (DRTS), 617 distributed energy management, 211
HIL RT simulation, 617, 618 microgrids’ real-time energy
software based, 617, 618 management, 208–210
Digital real-time testing, 620–621 numerical studies and result analysis,
Digital signal controllers (DSCs), 621 211–215
Digital twin, 626 protection coordination
Directional overcurrent protection, 529, grid-connected mode with DERs,
557–558 548–551
634 Index

Distributed energy resources (DERs) (cont.) Dual SOGI (DSOGI)

grid-connected mode without DERs, DSOGI-FLL, 355–356
546–548 DSOGI-PLL, 356
islanded mode with BESS, 552–553 Dual time–current–voltage characteristics,
relay-relay coordination, 595 609–610
and storage systems, 407
synchronous-based, 593, 601 E
transient stability, 581 EA, see Evolutionary algorithms (EA)
unit fault behavior, 526–529 Economic dispatch (ED), 23, 25, 28, 34, 39,
utilization of, 21 66, 69, 93, 180, 228, 237, 306
Distributed generation (DG) units Electrification process, 581
active and reactive output power, 364–366, EMS, see Energy management systems (EMS)
370, 371 Energy management systems (EMS), 616
capacity optimization, 98 centralized, 69
constraint, 110 control system, 63–64
DC MG, 279–280 decentralized, 69
economic and environmental issues, decision-making system, 64–67
305 functions, MG scheduling
and ESS, 106 microgrid’s hierarchical scheduling,
fault behavior and effect, 526–529 72–74
generators, 459–460 system operation strategies, 74–76
grid codes and distribution network interaction with other systems
protection schemes, 526 AMI, 68
inverter-interfaced, 338, 339 bid/offer interface, 68
network multiple microgrids, 21 DMS, 67
nonrenewable units electricity market, 68
fuel cell, 102–104 maintenance scheduling system, 68
micro turbine, 102 OMS, 68
power systems, 70 weather forecasting system, 68
renewable units mathematical modeling
photovoltaic panel, 104 MG’s components, 76–84
wind turbine, 104–106 system security, 85–92
Simulink, 455 and MGs, 62, 63
Distributed HIL, 628 monitoring system, 63
Distributed simulation, 624 NMCs (see Networked microgrid clusters
Distribution management system (DMS), (NMCs))
29–31, 34, 37, 67, 438, 583 scheduling of MG, 70
Distribution network protection schemes, structures, 39–40
526 studies, strategies, 42–43
Distribution networks, 581 time framework, 37, 38
DMS, see Distribution management system Energy not supplied (ENS) index, 583
(DMS) Energy storage systems (ESS), 106, 266
DOCRs, see Directional overcurrent relays AC and DC, 289
(DOCRs) applications in microgrids
Double synchronous reference frame (DSRF), load leveling, 143–144
353–354 power quality, 145
Drive trains of electric vehicles, 624 BMS, 293
DR programs, see Demand response (DR) characteristics, 291–292
programs classification, 133, 290
DSOGI, see Dual SOGI (DSOGI) constraint, 110–111
DSRF, see Double synchronous reference energy/power buffer, 127
frame (DSRF) in grid-connected microgrids, 296–299
Index 635

hybrid energy storage technologies F

battery–flywheel, 143 FA, see Firefly algorithm (FA)
battery–fuel cell, 141–142 Fast-response protection in restoration,
battery supercapacitor, 141 600–602
in islanded microgrids, 294–296 Fault behavior and fault current feeding
land-based microgrids, 128–130 capability of a DER unit, 526–529
load demand types, 132 Fault current management (FCM)
microgrids classification, 128 angle difference, 419
mobile microgrids, 130–131 IB-MG interface, 420
operation modes, 131 multiple microgrids, 424–426
opportunity, 299 phasor illustration, 421
PCC, 290 short-circuit currents, 418, 419, 422
power single MG, 421–423
electronic interface, 292–293 waveforms, 420
generation of DGs, 21, 289 See also Fault ride through (FRT)
research trends, 299 Fault indicators (FIs), 581
single energy storage technologies allocation, modeling process, 586
BES, 136–137 fault detective devices, 587
CAES, 135–136 fault locating process, 588
FBES, 137–138 feeders and laterals, 586
FES, 138 placement, 585–589
fuel cell, 140–141 and restoration devices, 584–585
PHS, 133–135 simultaneous placement, 589–592
SES, 139, 140 Fault location
SMES, 139 isolation and service restoration (FLISR),
timescales, 291 582
See also Microgrids (MGs) and restoration requirements, 582–592
Enhanced MV microgrid protection scheme Fault ride through (FRT)
1—HIF detection, 566–568 analysis and results, 428–431
ESS, see Energy storage systems (ESS) capabilities and requirements, 552, 569
Evolutionary algorithms (EA) control in MGs, 407–418
ABC, 228–229 deadbeat current control, 433–434
ACO, 223–225 DER integration, 404
application FCM, 418–426
energy management, 233–237 hardware-in-the-loop platform, 426–428
energy-related devices sizing, 237–240 implementation, 403
frequency control, 240–243 large-scale solar power plants, 403
microgrid optimal voltage, 240–243 MG FRT in grid codes, 405–407
operation scheduling, 233–237 power control scheme, 404
optimal placement, 237–240 reactive power injection, 431–432
BBO, 225 requirements, 526
cost-effective microgrid system, 219 SRF-PLL, 432
CSA, 227 voltage profile, 601
FA, 229–230 FBES, see Flow battery energy storage (FBES)
GA, 221–222 FC, see Fuel cells (FC)
GOA, 230–232 FCM, see Fault current management (FCM)
GWO, 229 FES, see Flywheel energy storage (FES)
heuristic and metaheuristic algorithms, 220 Field programmable gate arrays (FPGAs), 617,
HSA, 226 621
microgrids planning approaches and Firefly algorithm (FA), 98, 221, 229–230
references, 244 Firm RT
PSI, 223 applications, 617
WOA, 232–233 Flow battery energy storage (FBES), 132,
Experimentable Digital Twins (EDTs), 627 137–138
636 Index

Flywheel energy storage (FES), 132, 138, 143 structure mode, 40

Fog computing/clouds, 626 voltage regulation, 297–298
FPGAs, see Field programmable gate arrays Grid-connected mode of AC microgrid with
(FPGAs) single setting OC relay protection,
Frequency locked-loop (FLL), 350, 355–356 545, 546
FRT, see Fault ride through (FRT) GWO, see Grey wolf optimization (GWO)
Fuel cells (FC), 6, 14, 52, 102–104, 124,
140–141, 234, 375, 376
H
Hard RT
G applications, 617
GA, see Genetic algorithm (GA) Hardware in the loop (HIL), 616
Generalized emulator (source/load), Hardware in the loop RT simulation (HILS),
623–624 621–623
Generator-based distributed energy resources CHIL, 618
(DERs), 524 comparative analysis, testing approaches,
Generic co-simulation, 624 618, 619
Genetic algorithm (GA), 221–222, 226, 234, electrical drives in aircraft industry, 621
317, 440 hybrid approaches, 619
Global design of the optimization interactive virtual prototype, 619
cost function, 314 interconnection of various systems, 618,
NSIDE algorithm 619
improved differential evolution mechanical HIL, 618
algorithms, 316–317 multi-domain and cooperative/co-
nondominated sorting-based multi- simulation, 619
objective algorithm, 317–318 PHIL, 618
system optimization variables prototyping stage, 618
stage I, 315 Hardware test beds (HTBs), 616
stage II, 315 Harmonic compensation
stage III, 315–316 pq method, 345–347
GOA, see Grasshopper optimization algorithm reference currents identification
(GOA) instantaneous active, 340–342
Graphical user interface (GUI) blocks, 617 reactive power method, 340–342
Grasshopper optimization algorithm (GOA), SRF method, 342–344
230–232 SRF method, 347–349
Grey wolf optimization (GWO), 99, 111–112, Harmony search algorithm (HSA), 226, 233,
120, 124, 221, 229, 244 244
Grid codes, 569–571 HCS, see Hierarchical control structures (HCS)
Grid-connected microgrids Hierarchical control structures (HCS)
DERs, 548–551 AC MGs, 277–278
ESS control strategies DC MGs, 276–277
frequency regulation, 298–299 MG control system, 275
voltage regulation, 297–298 primary control, 35
frequency regulation, 298–299 Hierarchical system in laboratory scale, 624
islanded mode, 4, 289 High-speed telecommunication, 526
MG stability, 10, 11 High voltage ride-through (HVRT), 528
microgrid protection, 524 HSA, see Harmony search algorithm (HSA)
OC relay protection, 546 HV network fault, 535
photovoltaic-battery hybrid system, 98 Hybrid AC/DC microgrids (HADMGs), 6,
protection coordination, 546–548 222, 237, 289, 375–377
reactive and active power flow, 240 Hybrid simulation approach, 624
re-synchronization, 540 Hybrid testing approaches, 624–627
Index 637

I RESs and ESSs, 296

IACMGs, see Islanded AC microgrids double-sided auction, 155
(IACMGs) ESS control strategies, 294
IBRs, see Inverter-based renewables (IBRs) power flow, 322
IDR, see Incentive-based DR (IDR) proposed control strategy, 308
IED, see Intelligent electronic devices (IED) voltage/frequency control, 312
Imbalance compensation voltage responses, 325
direct extraction methods Islanding detection during island operation of
in abc frame, 350–351 nested microgrid, 569
αβ reference, 351–352
dq frame, 352–353
indirect extraction methods J
DSOGI-FLL, 355–356 JADE
DSOGI-PLL, 356 agents, 448–450
DSRF, 353–354 architecture
instantaneous power under unbalanced communication language used by
conditions, 357–358 agents, 451
Incentive-based DR (IDR), 183–184, 193–195, flat hierarchy, 446
197–201 Matlab/Simulink, 454–455
Information and communication technologies testing and measurement tool, 452–454
(ICT), 625 work environment structure, 451–452
Intelligent electronic devices (IED), 490, 526,
534, 556, 558, 625
Internet of energy (IoE), 625 L
Internet of things (IoT), 625 Land-based microgrids, 143, 145
Inverter-based renewables (IBRs), 256, industrial, 129, 130
262–266, 274 residential, 128–129
Inverter-interfaced DG Line differential protection based on current
switching signals, 360 measurements, 560
VSC control, 338, 339 Load emulation, 623–624
Islanded AC microgrids (IACMGs) Local markets
ACMGs, 379–381 benefits, 154
DCMGs, 375 definition, 153–154
dynamic step response, 394 key elements, 156, 158
HADMG, 376, 378 market settlement approaches, 161–169
hybrid AC/DC microgrid, 376 models, 158–159
microsources control methods, 377 objectives, 154–155
MIMO, 383 services, 155–156
open-loop, 393 trading approaches
PDC, 382 hybrid, 161
power system stability, 378, 379 pool-based, 159, 160
Q-V droop control method, 377 P2P trading, 160
results and discussions value streams for microgrids, 156, 157
eigenvalue analysis, 391–393 Low voltage ride-through (LVRT), 528
time-domain simulations, 393–398 LV microgrid protection, 525
signal transmission time delays, 383 LV microgrid synchronized reconnection,
small-signal disturbance, 381 540–542
SMT and WAMS, 381
stability enhancement, 382
WAMS, 382 M
Islanded microgrids Manual switches (MSs), 589
calculated initial conditions, 323 Market settlement approaches
coordinated control of auction-based approach, 162–163
of multiple ESSs, 294–295 case examples, 170–172
638 Index

Market settlement approaches (cont.) application, 6

in local markets, 170–172 campus/institutional sector, 186
the Monash microgrid, 169–170 characteristics/properties, 7
optimization-based approach commercial sector, 186–187
decentralized clearing, 166–169 configuration, 6–7
distributed clearing, 164–166 healthcare sector, 187
and pricing, 158 industrial sector, 185–186
MAS proposals, see Multi-agent system military sector, 186
(MAS) proposals operation mode, 6
Mathematical modeling, MG’s components other microgrids, 188
CHP and boiler, 80–82 remote/rural microgrids, 188
DG, 77–79 residential sector, 187
electrical network size, 6
linear distflow, 83–84 type, 5
SOCP, 82–83 combined heat and power model, 52–53
energy components, 4–5
exchange, 84 control coordination
storages, 79–80 centralized, 9
loads, 76–77 distributed, 9–10
reactive power resources, 80 hybrid, 10
renewable-based units with MPPT, 79 demand response programs, 98
Maximum power point tracking (MPPT), 79, discussion, 443–444
263–266, 268, 404, 408, 409, 431 DOCRs coordination
MCA, see MG A controller (MCA) on-line implementation, 476
MCB, see Miniature circuit breaker (MCB) optimization algorithms, 471–477
Medium-voltage (MV) optimization problem, 468–471
microgrid protection, 564–569 results and discussion, 478–485
microgrids after isolation, 524–525 test systems, 476–478
MEG, see Microgrid Exchange Group (MEG) ESS (see Energy storage systems (ESS))
MG A controller (MCA), 29, 30 fuel cell model, 52
MGs, see Microgrids (MGs) hierarchical control
Microelectronics/computational capability, 615 primary, 7
Microgrid cost modeling secondary, 8
installation, 189–190 tertiary, 8
maintenance, 190 islanding condition
operation, 190 detection schemes, 461–463
start-up, 190–191 implementation, 463–468
Microgrid Exchange Group (MEG), 3 micro-turbines model, 52
Microgrid protection issues, 524–535 multi-agent system proposals, 444–450
in grid-connected mode, 524 multi-objective function, 99
in islanded mode, 524 objective functions (see Objective functions
number of zones for protection, 525 formulation)
operation speed specifications, 525 operation management (see Operation
protection methods, 525 management)
reduced fault current contributions, 525 optimal management, 97
Microgrids (MGs), 99–100 problems and functional solutions, 439–443
advantages, 12–13 protection, 12, 438–439
CERTS, 3 unit commitment, 98
challenges weak infeed conditions detection
economical, 14 DG generators, 459–460
marketing, 14 small conventional generators, 458–459
regulation, 14 Microgrid transition to islanded operation,
technical, 13 529–535
classification MIMO, see Multi-input-multi-output (MIMO)
Index 639

Miniature circuit breaker (MCB), 29, 30, 537, reliability, 22–23

542 resiliency, 22
Mixed simulation approach, 624 objectives and constraints, 25–30
Model-based design (MBD), 617 physical and control layers, 20
Model-based systems engineering (MBSE) and reduction of overall cost, 21
simulations, 617 structure, 18
Model in the loop (MIL), 616 typical architecture of
Model reference adaptive control (MRAC), interconnected MGs, 31–32
628 multiple distribution feeders, 31–32
The Monash microgrid, 169–170 parallel MGs, 30–31
MPPT, see Maximum power point tracking serial MGs, 26, 29–30
(MPPT) single distribution feeder, 26, 29–31
MRAC, see Model reference adaptive control Networked microgrids (NMGs)
(MRAC) energy trading framework, 216
Multi agent-based protection, MGs, 602–604 KKT conditions, 206
Multi-agent system (MAS) proposals methodology description, 206–207
decentralized decision making, 69 multiple MGs, 424
distribution system, 445 nominal capacity, 22
electricity sector, 446, 447 problem formulation
first-order linear, 285 distributed energy management, 211
gang of agents, 448–450 microgrids’ real-time energy
IED, 556 management, 208–210
overcurrent relay, 460 numerical studies and result analysis,
power supply conditions, 483 211–215
smart agent definition, 445–446 resilience benefits, 20
switch position, 458 RT markets (see Real-time (RT) markets)
Multi-domain and cooperative/co-simulation, seller and buyer, 205
619 See also Networked microgrid clusters
Multi-input-multi-output (MIMO), 383 (NMCs)
NMCs, see Networked microgrid clusters
(NMCs)
N NMGs, see Networked microgrids (NMGs)
Networked microgrid clusters (NMCs) Novel/hybrid testing approaches, 624–627
ancillary services improvement, 22 Nuisance tripping of protective devices
best utilization, DERs, 21 fuse—re-closer, 596–598
bilateral and out of market transactions, 23 relay-relay, 594–596
challenges
disallowed transactions, 25
privacy of MGs, 24 O
protection coordination, 24 Objective functions formulation
stability of the system, 23–24 and constraints (see Constraints)
threat of cyberattack, 24–25 cost operation modeling, 47
compare EMS structures, 39–44 defined, 235
control strategy DRP, 50–51
DCS, 35–36 economic–environmental, 116
hierarchical, 33–35 MGDC, 107–109
master-slave, 33 MGs, 51–53
peer-to-peer, 32–33 numerical result, 53–57
and DS, 18 optimization algorithm, 111–113
DS structure, 19 pollution emission modeling, 47–48, 109
EMS, 37–39 problem constraints
energy management modeling and solution charge and discharge rate limit, 48
methods, 40, 45 generation capacity limit, 48
improvement power balance constraint, 48
640 Index

Objective functions formulation (cont.) stability in, 622

reserve, 49–50 stability issues, 623
self-sufficiency, 49 virtual impedance implementation, 623
Off-line accelerated simulation, 620 P2P trading, see Peer-to-peer (P2P) trading
Off-line digital simulation, 616 Price- based DR (PDR), 191–192
Off-line simulation, 617 costs, 198, 199
OMS, see Outage management system (OMS) PDR-based cost, 191–192
Online simulators, 616 strategies, 181–183
Operation management Processor in the loop (PIL), 616
distribution systems, 205 Protection based on voltage and directional
NMCs/NMGs (see Networked microgrid overcurrent, 561
clusters (NMCs)) Protection coordination in a grid-connected
objective functions (see Objective functions mode
formulation) with DERs, 548–551
Optimal dispatch, 124, 144 in islanded mode with DERs and BESS,
Optimization algorithm, 471–476 552–554
fuzzy method, 112–113 without DERs, 546–548
GOA, 230–232 Protection of LV AC microgrids, 561–564
multi-objective grey wolf, 111–112 Protection operation principles, 532–535
NSIDE, 332 Protection scheme requirements, MG,
PSO, 223 535–543
WOA, 232–233 adaptivity, 539
Outage detectors, 584–585 circuit breaker (CB) technology, 542–543
Outage management system (OMS), 68, 583, reliability, 538–539
584, 589, 611 re-synchronization, 540–542
Overcurrent (OC) protection, 524, 525 selectivity or coordination, 537–538
definite-time vs. inverse time, 543–553 sensitivity, 536–537
with multi-function characteristic, 606–608 Protection schemes of MG, 592–610
multi-inverse characteristic, 604–606 Protection time selectivity issues, 531, 532
PSO, see Particle swarm optimization (PSO)

P
Parallel simulation, 624 R
Particle swarm optimization (PSO), 28, 42, Rapid controller prototyping, 616
58, 98, 221, 223, 224, 234, 237, RBTS-bus4 test system, 591
239–242, 244, 306 Real-time (RT)
PCC, see Point of common coupling (PCC) constraints, 615
PDR, see Price- based DR (PDR) emulation, 623–624
Peer-to-peer (P2P) trading, 152, 160–162 markets
Phase-locked loop (PLL), 343, 347–350, 356, EMS, 69
407, 408, 410, 432, 542 energy management, 208–210
Piece-wise linear characteristic, 610 integration platform, 170
Point of common coupling (PCC), 4, 18, 30, load shedding computation, 306
31, 34, 39, 290, 376, 413, 461, 548, RTS, 72
552 power level emulation, 616
Pool-based trading approaches, 159–161 simulators, 621
Power electronic (PE) converters and testing
equipment, 615 centralized control, 628
applications, 624 digital, 620–621
Power hardware in the loop (PHIL), 616, 618 distributed HIL, 628
actual power transfer, 622 emulation, 623–624
ideal interface, 622 HILS, 621–623
modeling and simulation aspects, 622 methods, 616–619
power level exchange, 622 MRAC, 628
Index 641

novel/hybrid testing approaches, reactor differential protection, 509, 513,

624–627 514
PE, 615 sliding window, 503
platforms around the world, 627 thresholds criteria, 505
Reconfigurable topology of MG, 611–612 Second-order cone programming (SOCP),
Rectifier interfaced active load (RIAL), 381, 82–83
384, 386–389 Selective and fast protection schemes, 592–610
Reliability, protection scheme, 538–539 Selective protection, MGs
Remote fault indicators, 584 nuisance tripping of protective devices
Remotely controllable switches (RCSs), fuse—re-closer, 596–598
584 relay-relay, 594–596
simultaneous placement, 589–592 protective devices blinding, inverter-based
Renewable energy sources (RESs) DERs, 598–600
autonomous operation, 128 selective dead time in reclosing scheme,
DC MG’s control system, 276 593, 594
diesel generator-based systems, 188 Selectivity/coordination, protection scheme,
and energy storage systems, 523 537–538
global warming, 4 Semiconductor technology, 615
IBR, 263–265 Sensitivity, protection scheme, 536–537
MG control system, 275 SES, see Supercapacitor energy storage (SES)
non-IBR, 265 Signal transmission time delays
optimal unit commitment, 98 design procedure, 384
small-scale, 523 IIDG unit, 385, 389
Renewable energy system emulators, 624 lower-level decentralized controller, 384
RESs, see Renewable energy sources (RESs) LQG approach, 387
Rest of system (ROS), 622 noise and disturbance models, 387
Restoration capability, MG, 611–612 optimal parameters, 388
Re-synchronization, protection scheme, phase response, 386
540–542 RIAL, 386, 389
RIAL, see Rectifier interfaced active load round-trip communication, 390
(RIAL) small-signal linearized model, 385
Simulation results
active and reactive output power, 362, 363,
S 365, 366, 370, 371
SCM, see Second central moment (SCM) AFPID controller, 319
Second central moment (SCM) block diagram, control system, 358, 359
computational complexity, 505 CHP units, 318
delta filter application, 504 frequency controller, 327
differential protection (see Differential 30% load decrease, 324–325, 331, 332
protection) load shedding, 323–324, 329
events, 499, 500 nondominated responses, 319, 320
flow chart, 502 nonlinear load current, 360, 361
magnitude, 491 proposed controller, 320
misoperation, 499 studied MG, 358
operation, 499, 501 symmetrical three-phase fault, 321–323,
results 325, 328
busbar differential protection, 513, time domain characteristics, 331
515–516 unbalanced
generator protection, 506, 509–511 load current, 360
power reactor energization, 516, 519, and nonlinear load connection, 360, 361
520 voltage
power transformer energization, 516, after compensation, 362, 367
518 after nonlinear load connection, 360,
power transformer protection, 506–508 362
642 Index

Simulation results (cont.) case of, 347–349

after unbalanced load connection, 360, DSRF, 353–355
362 pq method, 345–347
controller, 326 System average interruption duration index
harmonic components, 362, 364 (SAIDI), 582–583
harmonic components after System operation strategies
compensation, 363, 369 economic aspects, 74–75
VUF and THD, 361, 363, 368 technical aspects, 75–76
Simulation step time or execution time, 617 System security
Smart relays, 584 HAS, 85–89
SMES, see Superconducting magnetics energy RTS, 87, 90–92
storage (SMES)
SMT, see Synchronized phasor measurement
technology (SMT) T
SOC, see State of charge (SOC) Techno-economic analysis of MG planning,
SOCP, see Second-order cone programming 585
(SOCP) THD, see Total harmonic distortion (THD)
Soft RT Third-part economic entity (TPEE), 206–207,
applications, 617 210, 213, 214–216
Software in the loop (SIL), 616, 618, 624 Time-current-voltage characteristics, 608–609
Source emulation, 623–624 Total harmonic distortion (THD), 337, 339,
Specific co-simulation, 624 360, 363, 462, 559, 569
SRF, see Synchronous reference frame (SRF) TPEE, see Third-part economic entity (TPEE)
Stability Transaction management platforms (TMP),
AFPID, 307, 308 626
cryogenic liquid, 139
DERs, 306
grid-connected U
and islanded, 305 UC, see Unit commitment (UC)
MG stability, 11 Undervoltage-based protection, 529
islanded MG stability, 11 Undervoltage with a fixed time delay, 530
microgrid modeling Uninterrupted power supply (UPS), 138, 291,
CHP model, 310–311 375, 376
fixed-speed wind turbine model, Unit commitment (UC), 66
309–310 and battery charge/discharge status, 53
reduced network model, 312 demand response program, 98, 100–101
synchronous reference frame, 311 DG units (see Distributed generation (DG)
motivations and contributions, 308 units)
power distribution systems, 305 numerical results
of the system, 23–24 with DR program, 120–124
voltage and frequency compensations, 260, hourly ability, 115, 116
306, 307 IBC-SOLAR Company, 114
State of charge (SOC), 63, 73, 110, 111, 266, micro turbine, 113
292–295, 298 without DR program, 116–120
Supercapacitor energy storage (SES), 132, UPS, see Uninterrupted power supply (UPS)
139–141
Superconducting magnetics energy storage
(SMES), 132, 134, 139 V
Supervisory control and data acquisition Vacuum circuit breaker (VCB), 543
(SCADA), 583, 616 Virtual test beds (VTB) based platforms, 619
Switch placement problem, 592 Virtual validation and commissioning, 626
Synchronized phasor measurement technology Voltage based protection schemes, 559–560
(SMT), 381 Voltage imbalance due to asymmetrical loads
Synchronous reference frame (SRF) and single-phase DG units, 541
Index 643

Voltage source converter (VSC) W

control, 338, 339 WAMS, see Wide-area measurement system
DC link capacitor, 407 (WAMS)
neutral-point-clamped, 572 Whale optimization algorithm (WOA),
positive and negative sequence quantities, 232–233, 244
527 Wide-area measurement system (WAMS),
Voltage unbalance factors (VUFs), 223, 240, 381–383, 385, 386, 390, 418
307, 308, 337, 360–363, 368 WOA, see Whale optimization algorithm
VUFs, see Voltage unbalance factor (VUFs) (WOA)