Smart Energy Management System for Distributed

Generations in AC Microgrid
Sertac Bayhan1 , Haitham Abu-Rub2
1 Qatar
Environment and Energy Research Institute, Hamad Bin Khalifa University, Doha, Qatar
2 Department of Electrical and Computer Engineering, Texas A&M University at Qatar, Doha, Qatar

sbayhan@hbku.edu.qa

Abstract—This paper proposes a novel smart energy manage- to control their energy consumption to save energy and cost
ment system (EMS) for distributed generations (DGs) in AC while maintaining their comfort level at the optimal value [10],
microgrid. The considered microgrid consists of two DGs, energy [11].
storage system, and loads. The solar PV system and the wind
energy conversion system (WECS) are selected as an input power The control techniques in EMS can be divided into two
source of DGs. Model predictive control technique is used to major categories: classical and smart techniques. The classical
regulate each DG’s output current whereas adaptive neuro-fuzzy control techniques that include proportional-integral (PI) con-
inference system (ANFIS) based EMS is developed to manage trollers based on the system model. For that reason, the perfor-
the available energy in AC microgrid. The proposed method is
validated analytically and through Hardware-in-the-Loop (HIL) mance of such control techniques is very sensitive to parameter
experiments. The carried out HIL experiments prove the merit variations . On the other hand, smart control techniques such
of the proposed method. as fuzzy logic, artificial neural networks, and neuro-fuzzy
Index Terms—Demand response, renewable energy, local en- are independent from the mathematical model of the system.
ergy market, distributed generation. This brings considerable benefits to the system including fast
dynamic response and robustness against parameter variations.
I. I NTRODUCTION
One of the smart control techniques is adaptive neuro-fuzzy
Electrical grids have the vision and trends to become more inference system (ANFIS) that is faster in convergence when
distributed, smart, and flexible in order to meet the huge compared to the other neuro-fuzzy models [2], [12]–[14].
increase of electricity demand. The trend of new grids is to This paper proposes a novel smart EMS for DGs in AC
design energy generation and consumption areas together, in microgrid. To regulate the output power of the DGs, MPC
distributed form. For that, especially power electronic devices based control technique is used. Furthermore, the ANFIS
play the crucial role to regulate distributed generation (DG)s based EMS is developed to manage the available energy
and dispersed energy-storage devices together and into the among DGs in AC microgrid. This paper is organized as
grid [1]. The intensive use of power electronic converters in follows. The considered microgrid architecture is described
microgrid brings their control methods to the forefront. It is in Section II. The developed ANFIS based EMS is detailed
generally expected that these control methods should meet in Section III. The carried out hardware-in-the-loop (HIL)
good dynamic response and high reference tracking charac- experiments are presented in Section IV. Finally, conclusions
teristics. To fulfill these control objectives, many advanced are presented in Section V.
control methods have been proposed recently [2], [3]. One of
these control methods is model predictive control (MPC) that
has appeared as an attractive solution for the control of power
II. OVERVIEW OF THE M ICROGRID A RCHITECTURE
converters due to its fast dynamic response and increased
control accuracy [4]–[7].
To integrate DGs and energy storage systems, a microgrid The simplified single-line diagram of the considered MG
concept is becoming a key element in the new electricity with the developed EMS is depicted in Fig. 1. The considered
paradigm. The microgrid not only handle the renewable energy MG consists of three subsystems: (1) PV system, (2) wind
penetration, but also allow the end user to be an important energy conversion system (WECS), and (3) hybrid energy
player in the electricity grid while being able to produce and storage devices (HESD). In addition to this, power electronic
consume energy (prosumer). However, using traditional energy converters are used in these subsystems to ensure the smooth
generation systems with renewable energy sources (RES) power transfer between load and power sources. In the con-
brings significant challenges in the management of microgrids sidered microgrid, the primary power source is RES, which
due to the stochastic and intermittent nature of RESs. A energy is composed of solar PV system and WECS. Furthermore, to
management system (EMS) is one of the important elements of store the excess energy generated by RES and to improve the
the microgrid for ensuring a harmony among DGs, loads, and stability of microgrid, the HESD is employed.
the utility grid [8], [9]. The EMS provides significant benefits Each subsystem is implemented in a Matlab/Simulink, as
for both customers and energy providers. It allows customers desribed in the following subsections.
 Fig. 1. The simplified single-line diagram of considered microgrid (MG) with developed EMS technique

A. PV System In recent years, various types of battery technologies are

The total rating of PV system is 90 kW, which is modeled used for energy storage systems. In spite of their maturity
based on Sun-Power SPR-305-WHT PV panel. The DC/AC and variety, batteries still have limited life-cycle and poor
converter topology used in the PV system is five-level neutral- power density, which is an important feature for balancing the
point-clamped (NPC)/H-bridge quasi-impedance source in- RES based power generations. To improve the energy store
verter (qZSI) due to its considerable benefits that include system performance, ultra-capacitors are becoming key player
single stage topology, lower dv/dt and THD [15]. in the HESD. Ultra-capacitor has a number of advantages
The traditional Perturb and Observe (P&O) technique is em- that include high efficiency (95%), high power density (up
ployed as a maximum power point tracking (MPPT) algorithm to 10000 W/kg), tolerance for deep discharges, and long life-
due to its easy implementation and high tracking efficiency cycle (500000 cycles at 100% depth-of-discharge) [17].
capabilities. Furthermore, MPC based primary controller is
used to regulate the output power of the PV system. The III. ANFIS BASED E NERGY M ANAGEMENT S YSTEM
reference output power is obtained from ANFIS based EMS.
The developed EMS is based on ANFIS, which is one of the
B. Wind Energy Conversion System most attractive neuro-fuzzy systems. The working principle of
To convert kinetic energy from wind into electrical en- ANFIS based on the actual data set. In this system, ANFIS
ergy, the WECS is employed in the microgrid. A double employees neural learning rules so as to determine and adjust
fed induction generator (DFIG) is selected in this study for the parameters of a fuzzy inference system. ANFIS brings
WECS. The DFIG is one of the most used generators for considerable benefits that includes easy implementation, fast
wind energy conversion because of its unique features [16]. and precise learning, and strong generalization skills [18].
The installed capacity of WECS is 40 kW and a three-phase The main aim of the proposed ANFIS based EMS is to
back-to-back inverter is employed as a converter topology. utilize the power obtained from PV and WECS with efficiently
Furthermore, a torque reference-based maximum power point and control the HESD according to the load demand in the
tracking (MPPT) control technique is used to obtain maximum considered microgrid. The training data set, which is a key
power whereas the MPC technique is used to regulate the element of ANFIS, is required to manage the available energy
DFIG output power. in the microgrid. The training data set is obtained according to
the available power from PV, WECS, HESD, and load demand
C. Energy Storage System at individual instant.
The main power source in microgrids is RES based power The total available power in the microgrid is
generation, especially the WECS and photovoltaic (PV) based
power generation systems have become more dominant than Ptotal = PP V + PW ECS + PHESD (1)
other RESs. Furthermore, an HESD plays a critical role
to maintain stability and robustness as well as to improve where PP V is the available PV power at time t, PW ECS is
the power quality of microgrids against the intermittent and the available WECS power at time is t, PHESD is the power
stochastic nature of the RES. An effective HESD should supplied by the HESD, and the Ptotal is the total power at
characterize high power density as well as high energy density. time t in the microgrid.
 Fig. 2. Typical architecture of ANFIS [18].

The training dataset of ANFIS can be determined as follows

PRES (0), Pload (1) Pref (0)
   
 P
RES (1), Pload (2)
  P (1) 
   ref 

.
 
= .
 Fig. 3. The simplified structure of the developed EMS.

  

 (2)

 . 
 
 . 
 the equation (2). In the following stage, training data set is
PRES (t − 1), Pload (t) Pref (t) used in the ANFIS so as to obtain reference power value for
the subsystems. Finally, the subsystems operates according to
where Pref is the reference available power in the microgrid. the data obtained from ANFIS.
The obtained dataset is employed to train the ANFIS and to
manage the available energy at the time of testing. IV. H ARDWARE -I N - THE -L OOP T EST R ESULTS
Fig. 2 shows the simplified architecture of ANFIS. In this Theoretical analysis and control structures have been ver-
scheme, a fixed node is illustrated as a circle, whereas an ified in HIL experiments. The structure of HIL setup is
adaptive node is illustrated as a square [18]. In this study, illustrated in Fig. 4. The HIL setup can be divided into two
the inputs of ANFIS are represented as x (the previous instant parts. The MPC based primary control algorithms for the
power of RES, PRES (t−1)) and y (the load demand at time t, subsystems are implemented on DE0-Nano FPGA boards. On
PL (t)), whereas the output is represented as z (the reference the other hand, the considered MG model given in Fig. 1
available power, Pref ). It is shown that ANFIS consists of and ANFIS based EMS are modeled in MATLAB/Simulink
five layers. Each layer is used to ensure particular function environment and execute in OPAL-RT real-time simulator. The
and these layers connected each other as a feed-forward parameters used in the tests are given in Table I.
network. The first layer contains adaptive nodes (Ai , Bi ) that Three HIL tests were carried out to verify the developed
are formed by membership functions (MF). The identified EMS technique. In the first case, the islanded mode opera-
rules are performed in the second layer. The rule set for the tion is studied, while in the second case the grid-connected
interference system with two fuzzy layers is defined as follows operation is studied. Furthermore, to verify the dynamic per-
Rule 1 : If PRES (t − 1) is A1 and PL (t) is B1 
 formance of the MPC, transient-state analysis were performed.

then f1 = m1 PRES (t − 1) + n1 PL (t) + k1 




(3)
Rule 2 : If PRES (t − 1) is A2 and PL (t) is B2 




(t − 1) + n P (t) + k 

then f = m P
2 2 RES 2 L 2

where m1 , m2 , n1 , n2 , k1 , and k2 represents the linear

parameters, A1 , A2 , B1 , and B2 represents the non-linear
parameters. The third layer is normalization layer which com-
putes the normalized firing strength of each rule (ω1 , ω2 ). After
this layer, the linear functions are obtained by evaluating the
normalized strength of each rule together with the input signals
in the fourth layer. This layer is also known as defuzzification
layer. In the final layer, the output of ANFIS is illustrated as
z, which is the sum of the output of fourth layer.
Fig. 3 illustrates the overall structure of the ANFIS-based
EMS. First, the available RES based power source, state
of charge (SOC), and load demand are received from the
subsystems. Then, the training dataset is obtained based on
Fig. 4. The structure of the HIL setup.
 TABLE I
S YSTEM PARAMETERS

Main Grid
Frequency, fs 50 Hz
Source Voltage, Vs 380 V rms
Feeder impedance, (RS + jXs ) 0.8+j1.3 Ω

Microgrid

PV Rated Power 90 kW
WECS Rated Power 40 kW
HESD Rated Capacity 25 Ah

Fig. 6. The HIL results in grid-connected mode.

A. Islanded Mode
In the following HIL studies, the proposed EMS perfor-
mance in islanded mode operation were investigated and
evaluated. During this mode, the load demand is met by DGs.
In other words, the load demand in the microgrid is lower than
the generated power. The excess power is stored in the HESD.
The HIL results for this mode operation are given in Fig. 5.
As mentioned before, the renewable sources are generating
the maximum available power. It can be seen that the PV
power and the WECS power vary according to irradiance
level, temperature and wind speed during the simulation.
Furthermore, the selected load profile is not constant and varies
during the simulation. It is clear that the load is supplied by
the PV and WECS since the generated power is higher than
the demanded power during the simulation as shown in Fig.
5.
B. Grid Connected Mode
In the following HIL studies, the proposed EMS perfor-
mance in grid-connected mode operation was investigated and Fig. 7. The HIL results of dynamic response of primary controller; (a) NPC
Inverter output current; (c) Back-to-Back converter output current.
evaluated. During this mode, the load demand is met by DGs
and the main grid in case of insufficient energy generated excess power is injected to the main grid. The HIL results for
by DGs. Furthermore, the load demand in the microgrid is this mode operation are given in Fig. 6.
sometimes lower than the generated power. In this case the
C. Transient-State Analysis
The HIL results of transient-state analysis are shown in Figs.
7 (a) and (b). For NPC inverter, the reference current was set at
5 A at the beginning then the reference current was increased
at 10 A at 0.15 s (see Fig. 7 (a)). It can be seen that the
output currents track to the references with fast rise time and
no overshoot. Furthermore, the transient-response of back-to-
back converter is given in Fig. 7 (b). In this HIL study, the
DFIG speed was 1050 rpm so the output frequency of the
back-to-back converter is 15 Hz. It means that the proposed
MPC based control technique tracks the wind speed with high
accuracy. It can be seen that the reference output current was
decreased from 20 A to 10 A at 0.15 s (see Fig. 7 (b)). It can
be clearly seen from the results that the transient time is very
short, and there is no overshoot.

V. C ONCLUSION
This paper proposes a novel smart energy management sys-
tem (EMS) for distributed generations (DGs) in AC microgrid.
Fig. 5. The HIL results in islanded mode; (a) Power demanded by the loads
The considered microgrid consists of two DGs, energy storage
in the microgrid, PV power and WECS power; (b)Excess power.
system, and loads. The solar PV system and the wind energy [7] S. Bayhan, M. Trabelsi, H. Abu-Rub, and M. Malinowski, “Finite-
conversion system (WECS) are selected as an input power control-set model-predictive control for a quasi-z-source four-leg inverter
under unbalanced load condition,” IEEE Transactions on Industrial
source of DGs. Model predictive control technique is used to Electronics, vol. 64, no. 4, pp. 2560–2569, 2017.
regulate the each DG’s output current whereas adaptive neuro- [8] H. C. Chiang, K. K. Jen, and G. H. You, “Improved droop control
fuzzy inference system (ANFIS) based EMS is developed to method with precise current sharing and voltage regulation,” IET Power
Electronics, vol. 9, no. 4, pp. 789–800, 2016.
manage the available energy in AC microgrid. The proposed [9] T. L. Vandoorn, B. Meersman, J. D. M. D. Kooning, and L. Vandevelde,
method is validated analytically and through Hardware-in-the- “Transition from islanded to grid-connected mode of microgrids with
Loop (HIL) experiments. The carried out HIL experiments voltage-based droop control,” IEEE Transactions on Power Systems,
vol. 28, no. 3, pp. 2545–2553, 2013.
prove the merit of the proposed method. [10] H. Han, X. Hou, J. Yang, J. Wu, M. Su, and J. M. Guerrero, “Review
of power sharing control strategies for islanding operation of ac micro-
ACKNOWLEDGMENT grids,” IEEE Transactions on Smart Grid, vol. 7, no. 1, pp. 200–215,
This publication was made possible by NPRP 9-310-2- 2016.
[11] Z. Wang, B. Chen, J. Wang, M. M. Begovic, and C. Chen, “Coordinated
134 from the Qatar National Research Fund (a member of energy management of networked microgrids in distribution systems,”
Qatar Foundation). The statements made herein are solely the IEEE Transactions on Smart Grid, vol. 6, no. 1, pp. 45–53, 2015.
responsibility of the authors. [12] D. E. Olivares, J. D. Lara, C. A. Canizares, and M. Kazerani,
“Stochastic-predictive energy management system for isolated micro-
grids,” IEEE Transactions on Smart Grid, vol. 6, no. 6, pp. 2681–2693,
R EFERENCES 2015.
[1] H. Abu-Rub, M. Malinowski, and K. Al-Haddad, Power Electronics for [13] D. E. Olivares, C. A. Canizares, and M. Kazerani, “A centralized energy
Renewable Energy Systems, Transportation and Industrial Applications. management system for isolated microgrids,” IEEE Transactions on
A John Wiley&Sons Ltd., 2014. Smart Grid, vol. 5, no. 4, pp. 1864–1875, 2014.
[2] P. Arul, V. K. Ramachandaramurthy, and R. Rajkumar, “Control strate- [14] M. N. Arafat, S. Palle, Y. Sozer, and I. Husain, “Transition control
gies for a hybrid renewable energy system: A review,” Renewable and strategy between standalone and grid-connected operations of voltage-
Sustainable Energy Reviews, vol. 42, pp. 597 – 608, 2015. source inverters,” IEEE Transactions on Industry Applications, vol. 48,
[3] S. Bayhan and H. Abu-Rub, “Model predictive droop control of dis- no. 5, pp. 1516–1525, 2012.
tributed generation inverters in islanded ac microgrid,” in 11th IEEE In- [15] S. Bayhan, M. Trabelsi, O. Ellabban, H. Abu-Rub, and R. S. Balog,
ternational Conference on Compatibility, Power Electronics and Power “A five-level neutral-point-clamped/h-bridge quasi-impedance source
Engineering (CPE-POWERENG), 2017, pp. 247–252. inverter for grid connected pv system,” in IECON 2016 - 42nd Annual
[4] S. Bayhan, H. Abu-Rub, and R. S. Balog, “Model predictive control Conference of the IEEE Industrial Electronics Society, 2016, pp. 2502–
of quasi-z-source four-leg inverter,” IEEE Transactions on Industrial 2507.
Electronics, vol. 63, no. 7, pp. 4506–4516, 2016. [16] I. Colak, G. Fulli, S. Bayhan, S. Chondrogiannis, and S. Demirbas,
[5] S. Bayhan, H. Abu-Rub, J. I. Leon, S. Vazquez, and L. G. Franquelo, “Critical aspects of wind energy systems in smart grid applications,”
“Power electronic converters and control techniques in ac microgrids,” Renewable and Sustainable Energy Reviews, vol. 52, pp. 155 – 171,
in IECON 2017 - 43rd Annual Conference of the IEEE Industrial 2015.
Electronics Society, 2017, pp. 6179–6186. [17] A. Etxeberria, I. Vechiu, H. Camblong, J. M. Vinassa, and H. Camblong,
[6] T. Vandoorn, J. D. Kooning, B. Meersman, and L. Vandevelde, “Re- “Hybrid energy storage systems for renewable energy sources integration
view of primary control strategies for islanded microgrids with power- in microgrids: A review,” in 2010 Conference Proceedings IPEC, 2010,
electronic interfaces,” Renewable and Sustainable Energy Reviews, pp. 532–537.
vol. 19, pp. 613 – 628, 2013. [18] D. Karaboga and E. Kaya, “Adaptive network based fuzzy inference
system (anfis) training approaches: a comprehensive survey,” Artificial
Intelligence Review, Jan 2018.