Special Issue Reprint

Advanced Electric
Power System 2023

Edited by
Ying-Yi Hong

mdpi.com/journal/energies
Advanced Electric Power System 2023
Advanced Electric Power System 2023

Editor
Ying-Yi Hong

Basel • Beijing • Wuhan • Barcelona • Belgrade • Novi Sad • Cluj • Manchester

Editor
Ying-Yi Hong
Chung Yuan Christian University
Taoyuan
Taiwan

Editorial Office
MDPI AG
Grosspeteranlage 5
4052 Basel, Switzerland

This is a reprint of articles from the Special Issue published online in the open access journal
Energies (ISSN 1996-1073) (available at: https://www.mdpi.com/journal/energies/special issues/
TVD92H72EC).

For citation purposes, cite each article independently as indicated on the article page online and as
indicated below:

Lastname, A.A.; Lastname, B.B. Article Title. Journal Name Year, Volume Number, Page Range.

ISBN 978-3-7258-1615-6 (Hbk)

ISBN 978-3-7258-1616-3 (PDF)
doi.org/10.3390/books978-3-7258-1616-3

© 2024 by the authors. Articles in this book are Open Access and distributed under the Creative
Commons Attribution (CC BY) license. The book as a whole is distributed by MDPI under the terms
and conditions of the Creative Commons Attribution-NonCommercial-NoDerivs (CC BY-NC-ND)
license.
Contents

About the Editor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

Yuan-Kang Wu, Deng-Yue Gau and Trinh-Duc Tung

Overview of Various Voltage Control Technologies for Wind Turbines and AC/DC Connection
Systems
Reprinted from: Energies 2023, 16, 4128, doi:10.3390/en16104128 . . . . . . . . . . . . . . . . . . . 1

Hakim Bennani, Ahmed Chebak and Abderrazak El Ouafi

Unique Symbolic Factorization for Fast Contingency Analysis Using Full Newton–Raphson
Method
Reprinted from: Energies 2023, 16, 4279, doi:10.3390/en16114279 . . . . . . . . . . . . . . . . . . . 18

Joaquı́n Pedra, Luis Sainz and Lluı́s Monjo

DC Admittance Model of VSCs for Stability Studies in VSC-HVDC Systems
Reprinted from: Energies 2023, 16, 5457, doi:10.3390/en16145457 . . . . . . . . . . . . . . . . . . . 35

Stefano Quaia, Alessandro Mauri, Alessio Marchesin and Paolo Manià

Actions to Handle Increasing Distributed Generation and Rising Voltage during Low-Demand
Periods: Tap Staggering Practices and Field Tests in the Italian Transmission Network
Reprinted from: Energies 2023, 16, 6258, doi:10.3390/en16176258 . . . . . . . . . . . . . . . . . . . 52

Oluwafemi Emmanuel Oni and Omowunmi Mary Longe

Analysis of Secondary Controller on MTDC Link with Solar PV Integration for Inter-Area Power
Oscillation Damping
Reprinted from: Energies 2023, 16, 6295, doi:10.3390/en16176295 . . . . . . . . . . . . . . . . . . . 65

Sheng Li, Can Zhang and Jili Zuo

Long-Term Voltage Stability Bifurcation Analysis and Control Considering OLTC Adjustment
and Photovoltaic Power Station
Reprinted from: Energies 2023, 16, 6383, doi:10.3390/en16176383 . . . . . . . . . . . . . . . . . . . 83

Chandu Valuva, Subramani Chinnamuthu, Tahir Khurshaid and Ki-Chai Kim

A Comprehensive Review on the Modelling and Significance of Stability Indices in Power
System Instability Problems
Reprinted from: Energies 2023, 16, 6718, doi:10.3390/en16186718 . . . . . . . . . . . . . . . . . . . 100

Prottay M. Adhikari, Luigi Vanfretti, Hao Chang and Koushik Kar

Real-Time Control of a Battery Energy Storage System Using a Reconfigurable Synchrophasor-
Based Control System
Reprinted from: Energies 2023, 16, 6909, doi:10.3390/en16196909 . . . . . . . . . . . . . . . . . . . 145

Guoqing Zhang, Wengen Gao, Yunfei Li, Xinxin Guo, Pengfei Hu and Jiaming Zhu
Detection of False Data Injection Attacks in a Smart Grid Based on WLS and an Adaptive
Interpolation Extended Kalman Filter
Reprinted from: Energies 2023, 16, 7203, doi:10.3390/en16207203 . . . . . . . . . . . . . . . . . . . 165

Eduardo Tavares Silvério and Jose Rubens Macedo Junior

Measuring and Modeling the Skin Effect for Harmonic Power Flow Studies
Reprinted from: Energies 2023, 16, 7913, doi:10.3390/en16237913 . . . . . . . . . . . . . . . . . . . 185

Tianmeng Yang, Suhua Lou, Meng Zhang, Yanchun Li, Wei Feng and Jicheng Liu
Research on a Two-Layer Optimal Dispatching Method Considering the Mutual Aid of Peak
Regulating Resources among Regional Power Grids
Reprinted from: Energies 2024, 17, 667, doi:10.3390/en17030667 . . . . . . . . . . . . . . . . . . . . 204

v
About the Editor
Ying-Yi Hong
Ying-Yi Hong (Senior Member, IEEE) received a B.S.E.E. degree from Chung Yuan Christian
University (CYCU), Taiwan, in 1984, an M.S.E.E. degree from National Cheng Kung University
(NCKU), Taiwan, in 1986, and a Ph.D. degree from the Department of Electrical Engineering,
National Tsing Hua University (NTHU), Taiwan, in December 1990. Sponsored by the Ministry of
Education, Taiwan, he conducted research in the Department of Electrical Engineering, University of
Washington, Seattle, from August 1989 to August 1990. He has been with CYCU since 1991, where he
was the Dean of the College of Electrical Engineering and Computer Science from 2006 to 2012. He
was promoted to Chair Professor in 2024 due to his exceptional research, leadership, teamwork, and
international collaboration. From 2012 to 2018, he was the General Secretary of CYCU, where he is
currently the Vice President. He received the Outstanding Professor of Electrical Engineering Award
from the Chinese Institute of Electrical Engineering (CIEE), Taiwan, in 2006 and the Outstanding
Professor of Engineering Award from the Chinese Institute of Engineers (CIE), Taiwan, in 2024. He
was the Chair of the IEEE PES Taipei Chapter in 2001. His research interests include power system
analysis and artificial intelligence applications.

vii
 energies
Article
Overview of Various Voltage Control Technologies for Wind
Turbines and AC/DC Connection Systems
Yuan-Kang Wu *, Deng-Yue Gau and Trinh-Duc Tung

Department Electrical Engineering, National Chung-Cheng University, No. 168, University Rd.,
Chiayi 62102, Taiwan; willie890486@yahoo.com.tw (D.-Y.G.); tungtd@alum.ccu.edu.tw (T.-D.T.)
* Correspondence: allenwu@ccu.edu.tw

Abstract: Wind power generation is one of the mainstream renewable energy resources. Voltage
stability is as important as the frequency stability of a power system with a high penetration of wind
power generation. The advantages of high-voltage direct current (HVDC) transmission systems
become more significant with the increase of both installed capacity and transmission distance in
offshore wind farms. Therefore, this study discusses various voltage control methods for wind
turbines and HVDC transmission systems. First, various voltage control methods of a wind farm
were introduced, and they include QV control and voltage droop control. The reactive power of
a wind turbine varies with active power, while the active power from each wind turbine may be
different owing to wake effects. Thus, QV and voltage droop control with varying gain values
are also discussed in this paper. Next, the voltage control methods for an HVDC transmission
system, such as power factor control, voltage control, and Vac -Q control, are also summarized and
tested in this study. When a three-phase short circuit fault occurs or a sudden reactive power load
increases, the system voltage would drop immediately. Thus, various voltage control methods for
wind turbines or HVDC can make the system’s transient response more stable. Therefore, this study
implemented the simulation scenarios, including a three-phase short circuit fault at the point of
common coupling (PCC) or a sudden increase of reactive power load, and adopted various voltage
control methods, which aim to verify whether additional voltage control methods are effective to
improve the performance of transient voltage. The voltage control method has been implemented
in PSCAD/EMTDC, and the simulation results show that the QV control performs better than the
Citation: Wu, Y.-K.; Gau, D.-Y.; Tung,
droop control. In addition, when applying the voltage control technique during a three-phase fault,
T.-D. Overview of Various Voltage
transient voltage nadir can be improved through either an HVDC transmission system or an AC
Control Technologies for Wind
transmission system.
Turbines and AC/DC Connection
Systems. Energies 2023, 16, 4128.
Keywords: renewable energy; voltage stability; high-voltage direct current (HVDC); voltage control;
https://doi.org/10.3390/en16104128
voltage droop control
Academic Editor: Juan-José
González de la Rosa

Received: 15 April 2023

Revised: 13 May 2023 1. Introduction
Accepted: 15 May 2023 With global warming and sustainable development, renewable energy has developed
Published: 16 May 2023 rapidly during the last few decades. Wind energy is one of the primary types of renewable
energy, and the installation is more concentrated than solar energy. The high penetration
of offshore wind farms causes concern about frequency and voltage stability. Thus, many
countries have requested wind farms to provide voltage support at the point of connection
Copyright: © 2023 by the authors.
in their grid codes [1,2]. Using the power electronic converter, the doubly-fed induction
Licensee MDPI, Basel, Switzerland.
generator (DFIG) and full converter wind turbine (FCWT) can provide reactive power sup-
This article is an open access article
port to the system’s voltage [3]. To connect an offshore wind farm, the HVDC transmission
distributed under the terms and
conditions of the Creative Commons
system is much more suitable than the HVAC transmission system, with a transmission
Attribution (CC BY) license (https://
distance of over 100 km and a capacity larger than 100 MVA [4]. Besides, voltage source
creativecommons.org/licenses/by/ converter-based high voltage direct current (VSC-HVDC) can control active and reactive
4.0/).

Energies 2023, 16, 4128. https://doi.org/10.3390/en16104128 1 https://www.mdpi.com/journal/energies

Energies 2023, 16, 4128

power independently [5], which makes it control the voltage easily. Without the considera-
tion of reactive power compensations like Static Synchronous Compensators (STATCOM)
or Static Var Compensators (SVC), this study first tested various control methods for wind
turbines and VSC-HVDC systems.
With a high penetration of wind power generation, the voltage issues become impor-
tant. The Power-Voltage analysis [6] about static voltage stability indicated that the voltage
would collapse without any contingency if a transfer of wind power generation increases.
Meanwhile, it only raises a small transfer limit by installing more reactive compensation
resources. In the aspect of dynamic voltage stability, compared to a stiff system, a weak grid
is generally concerned about overvoltage after a grid fault owing to a higher sensitivity of
dV/dQ [7]. The voltage variation to reactive power injection is defined as the grid strength,
presented by the Short Circuit Ratio (SCR). The voltage at a strong grid with a high SCR
fluctuates less than at a weak grid with a small SCR [8]. Ref. [9] provides a study about the
influence of SCR on the voltage control of wind power plants. It shows that it is easier to
impact the voltage at the point of common coupling (PCC) with a lower SCR value.
In Ref. [10], it divided the voltage control method of a large-scale wind farm into three
categories: decentralized, centralized and hierarchical controls. The reactive power capa-
bility of different wind turbines varies according to wake effects; thus, Ref. [11] proposed
an adaptive Q-V method that allows wind turbines with more reactive power capability to
provide more reactive power. Then, the adaptive and fixed Q-V schemes were simulated
by considering different grid stiffness and disturbance types. Ref. [12] presented both
variable voltage droop control and constant droop control to reduce the voltage fluctua-
tion caused by varying loads at PCC. A reactive power coordination control strategy was
proposed in [13] to optimize voltage quality and minimize power loss using a genetic
algorithm; moreover, the control method was confirmed better than the unit power factor
by investigating the maximum delta voltage and the voltage characteristic coefficient in
a test system integrated with three wind farms. To sum up, two factors that influence
the voltage regulation in an offshore wind farm include the capability of reactive power
supported by wind turbines and the grid strength at PCC.
Ref. [14] proposed a coordinated control scheme of active and reactive power oscil-
lation damping controllers for a utility-scale wind power plant to support voltage sta-
bility and provide fast voltage recovery after grid faults. In [15], the work presented
a communication-free coordinated fault ride-through (FRT) control between the wind farm-
side converters and the wind turbine inverters based on the injected sequence harmonics;
the proposed method enables the DC voltage to be recovered quickly within an accepted
range, which helps improve the decline of AC voltage. In [16], a reactive voltage control
technique for PMSG-based wind farms was suggested. It considers sufficient reactive
power and balanced terminal voltage. Thus, the method can regulate the PCC voltage
of wind farms while still maintaining the balance of the terminal voltage of a wind farm.
In [17], a multi-objective Q-V coordinated control technique for wind farms was suggested,
which considers voltage variation, converter junction temperature, and power loss. In [18],
a coordinated voltage regulation technique based on model predictive control was sug-
gested for enabling wind farms to support black start. Additionally, the coordinated voltage
regulation can precisely consider the effect of active power on voltage variation, and the
reactive and active power from wind turbines and energy storage can be coordinated to
increase the capability of wind farms. To increase the voltage stability of power systems
at the PCC, Ref. [19] suggested a synchrophasor data-based Q-V droop (SQVD) control
technique for wind farms. The suggested method enhances the Q-V droop performance of
wind farms by coupling the corresponding wide-area signals with the local signals at each
wind farm. A distributed reactive power control (DVC) scheme based on the consensus
alternating direction method of multipliers (ADMM) was introduced in [20]. This scheme
was applied to wind farms to optimize reactive power utilization. To reduce voltage devia-
tion and active power losses, a decentralized voltage control based on the standard ADMM
was also created in [20] for wind farm operation.

2
Energies 2023, 16, 4128

Besides wind turbines, reactive power compensation devices also play a vital role in
voltage regulation. Ref. [21] utilized STATCOM with CMC-based topology, and the effect
was examined to resolve voltage fluctuation. Ref. [22] suggested potential methods that
can improve the voltage stability of wind farms: one is to install a static var compensator
(SVC) to provide dynamic reactive power support, and the other is to select a doubly-fed
induction generator (DFIG) that can control reactive power flexibly without installing
reactive power compensation devices. After a three-phase short circuit fault, the problem
of overvoltage could cause wind turbines to trip off. Thus, a coordinated control between
SVC and DFIG can decrease the magnitude of overvoltage compared to the use of SVC
only [23].
Compared to traditional AC transmission, VSC-HVDC can enhance voltage stability
by providing extra reactive power [24]. In [25], the stability analysis was carried out for
a 100 MW solar plant through a connection of HVDC. Ref. [26] implemented a system
simulation and showed the voltage at PCC during a three-phase short circuit fault. The
simulation considered different levels of wind power penetration with a VSC-HVDC
connection. The results showed that the voltage at the PCC is proportional to the wind
power integration. In Ref. [27], the control strategy that considers wind farm, STATCOM
and HVDC was tested by load switching and three-phase short circuit fault, where the
offshore wind farm is connected by the line commutated converter high-voltage direct
current (LCC-HVDC) transmission system. Ref. [28] showed that the control scheme of VSC-
HVDC can influence the maximum transfer of active power. In Refs. [29–31], some voltage
control methods for a VSC-HVDC-connected weak grid were proposed. Two control modes
were proposed in [32,33] by coordinating wind generators with VSC-HVDC based on model
predictive control. The normal operation mode can maintain a stable voltage and reduce
power loss, while the corrective mode can help the voltage reach 1.0 p.u. rapidly when
wind turbines were reconnected to the grid after a storm. The control method presented
in [34] can minimize power loss of offshore wind farms, increase the amount of active
power transfer, and improve voltage stability during system transients. From the above
literature reviews, the HVDC-connected system can provide better performance than the
AC-connected system.
In addition to reviewing various up-to-date voltage control methods, this paper also
compared the performances of transient voltage responses based on different voltage
control techniques. To demonstrate the superior performance of the QV control, this paper
also carried out the control methods for the voltage droop control and the QV control,
and it compared all of these control methods with AC or DC transmission systems. The
simulation results reveal that the voltage nadir can be improved by the above control
methods. Additionally, the performance of voltage responses with HVDC transmission
systems is better than that with AC transmission systems.
The paper is organized as follows: Section 2 introduces various voltage control meth-
ods utilized for wind turbines and VSC-HVDC. Section 3 presents the used AC and HVDC
test systems and demonstrates the simulation results under various operating scenarios.
Finally, the conclusions are drawn in Section 4.

2. Various Voltage Control Methods

In this study, the control methods are separated into two parts. The first part is about
the control methods for wind turbines, which is introduced in Section 2.1, while the second
part is about the control methods for VSC-HVDC, which is presented in Section 2.2.

2.1. Wind Farm Control

In this study, the DFIG-based wind turbine was used based on the generic model,
and its block diagram is listed in Figure 1. There are four main blocks inside the generic
model: the converter control model, the generator/converter model, the pitch control
model, and the drive train model. The reactive power reference (Qre f ) generated by QV,

3
Energies 2023, 16, 4128

voltage droop or other controls is sent to the converter control model to obtain the required
voltage. A detailed description of each model can be found in [35,36].

Figure 1. The block diagram of the generic model of a wind turbine.

2.1.1. PQ Diagram
The main restriction of the reactive power supported by a wind turbine is its stator
current and rotor current. Figure 2 shows a PQ diagram from the stator current of a DFIG [9],
which indicates the maximum reactive power absorbed or provided by a DFIG is based on
its active power. For instance, Table 1 shows the relationship between active and reactive
power for a DFIG-based wind turbine. As the active power is maximal, the reactive power
can be supported or absorbed by 20% and 30% of capacity, respectively. However, if active
power is below the maximum, the supported or absorbed reactive power can be increased.
That is, the controller of a wind turbine can control its reactive power according to its
PQ diagram.

Figure 2. The PQ diagram of a DFIG-based wind turbine.

4
Energies 2023, 16, 4128

Table 1. The PQ chart of a DFIG.

Point P (%) Q (%)

0 0 10
1 10 60
2 70 60
3 100 20
4 100 −30
5 85 −60
6 10 −60
7 0 −15

2.1.2. QV Control
The adaptive QV control enables a wind turbine to provide more reactive power.
Figure 3 shows the block diagram to obtain the gain k q_i used in Figure 4. First, the
upper Qmax_i and lower limit Qmin_i of a wind turbine are determined in the adaptive
QV controller based on the desired PQ diagram. Next, the gain k q_i at each wind turbine
is obtained by considering its reactive power limit. This control method can control the
voltage at the PCC and the terminal of each wind turbine. For example, in Figure 4, the limit
of ±0.1 (i.e., the block of “limit” in Figure 4) restricts the difference between the reference
voltage (Vre f _pcc ) and the measured voltage (Vmeas_pcc ) at the PCC [11]. Next, the difference
between the reference voltage and the measured voltage at the wind turbine is added to
the output signal from the limit to obtain the error signals. Finally, the required Qre f during
the fault can be computed. In contrast, the QV gain of the fixed QV control is set to 5 in
this study.

Figure 3. The block diagram to obtain the kq_i.

Figure 4. The control block of the adaptive QV control.

2.1.3. Voltage Droop Control

Voltage droop control is a kind of reactive control method. The functional theory of
voltage droop control is similar to the frequency control of a synchronous generator. It can
stabilize the voltage by providing or absorbing reactive power from wind turbines. The
conventional voltage droop is expressed as
1 
Qre f − Duringthe f ault = Qre f Pre− f ault − Vmeaspcc − Vre f pcc (1)
Ri

5
Energies 2023, 16, 4128

where Vmeaspcc is the voltage that measures at the PCC; Vre f pcc is the voltage reference at
the PCC; Qre f Pre− f ault is the original reference value of reactive power from a wind turbine;
Qre f − Duringthe f ault is the new reference value of reactive power from a wind turbine. The
coefficient 1/Ri is the reciprocal of k q_i . The structure of voltage droop control is shown in
Figure 5.

Figure 5. The control block of the voltage droop control.

2.2. VSC-HVDC Control

Figure 6 shows the control strategy of a VSC-HVDC-connected offshore wind farm
(OWF). The Idre f of inner control is derived from either VDC control or P control at the
GSVSC or WFVSC. The Iqre f of inner control is derived from the proposed control method
to control the voltage at PCC. A detailed description of the inner control loop, VDC control
and P control can be obtained from [37,38].

Figure 6. The control strategy of VSC-HVDC connected OWF.

2.2.1. Power Factor Control

Figure 7 shows the block diagram of power factor control, where both power factor
and active power are used to obtain the required reactive power. The sign of the power
factor can identify whether the reactive power is provided or absorbed. Finally, Qre f is
sent to the Q controller to compare with the measured reactive power Qmeas at the PCC,
and the error between them is sent to the PI controller to obtain Iqre f . Figure 8 shows the
relationship between reactive power and power factor. To identify the leading or lagging
power factor, Figure 8 uses different signs to present them. For instance, a negative sign
indicates a leading power factor that a wind turbine absorbs reactive power; the sign is
positive for a lagging power factor.

6
Energies 2023, 16, 4128

Figure 7. Power factor control.

Figure 8. Presentation of power factor.

2.2.2. Vac -Q Droop Control

Figure 9 shows the block diagram for the Vac -Q droop control. In this figure, the
deadband can set an acceptable voltage margin that the controller does not trigger, which
only provides the desired reactive power. While the voltage exceeds the range, the controller
can provide or absorb reactive power for voltage support. The Qre f − Duringthe f ault is sent to
Qre f and Qmeas to get Iqre f .

Figure 9. Vac -Q droop control.

2.2.3. Voltage Control

Voltage control uses the error between the reference voltage Vre f and the measured
voltage Vmeas ; then the error between them is sent to the PI controller. The output of the
PI controller is Iqre f that can adjust reactive power to maintain the voltage at the reference
value. The topology of voltage control is shown in Figure 10.

7
Energies 2023, 16, 4128

Figure 10. Voltage control.

3. Results and Discussions

3.1. Simulation Topology of the Used AC System and HVDC System
To investigate the effect of the proposed methods on the system voltage, the simula-
tions have been done in an AC test system, which consisted of equivalent wind farms and
an onshore power system, as shown in Figure 11; in addition, Figure 12 shows the HVDC
test system. Each equivalent wind farm includes four wind turbines with a rated power of
8 MW each. The total rated power of the offshore wind farm is 160 MW. On the onshore
grid, there are six synchronous generators with a rated power of 125 MW each. Table 2
shows the assumed wind speeds for WTs. It is assumed to give a steady-state (constant)
wind-speed scenario for each wind farm, and different wind farms would have different
wind speeds because of wake loss or other reasons. For example, the wind speed to the first
and fifth wind farms are assumed as 11.6 m/s and 9 m/s, respectively. Power Generation
from wind turbines will be affected by different wind speeds.

Figure 11. AC Test system.

8
Energies 2023, 16, 4128

Figure 12. HVDC Test system.

Table 2. Wind speed of equivalent WFs.

No. of Wind Farm Wind Speed (m/s)

1 11.6
2 11.4
3 10.6
4 10
5 9

Table 3 indicates the parameter of the cable. The turns ratio of the transformer in
TR1 and TR2 is 0.69 KV/66 KV and 66 KV/161 KV, respectively. In a large-scale wind
farm with multiple wind turbines, the equivalent wind-farm modeling technique from
multiple wind turbines with many connection lines to an equivalent wind turbine with
an equivalent connection line is significant, which greatly reduces simulation time. Thus,
this work established an equivalent wind farm modeling based on [39,40]. In Table 3, the
cable parameters before equivalence indicate the parameters of an original cable from
a wind turbine to the connection point, as shown in the red line of Figure 11 or Figure 12.
However, the parameters after equivalence mean the equivalent impedance of the cable
from an equivalent wind turbine to the connection point, as shown in the blue line of
Figure 11 or Figure 12.

Table 3. Parameters of cable.

Parameters Before Equivalent After Equivalent

R (Ohms) 0.045 0.084375
L (mH) 0.371362 0.696303
C (uF) 0.24 0.96
Rate voltage (kv) 66 66

9
Energies 2023, 16, 4128

The duration of the experiment is 12 s. However, most of the figures (Figures 13–24)
only show the runtime within 10 s because the time length is enough. In Sections 3.2 and 3.3,
an additional reactive power load of 180 MVar is assumed to be increased into the onshore
grid at 8 s. In Sections 3.4 and 3.5, a three-phase short-circuit fault occurs at 8 s at point
A of Figures 11 and 12, and the fault lasts for 100 ms.

3.2. A Reactive Power Load Is Added to the System with an AC Transmission System
To investigate whether the proposed methods can work adequately in the AC system
or not, an increase of reactive load (180 MVar) is added into the system in 8 s.
Figure 13 shows the transient voltage curve at the PCC using the Adaptive QV, Fixed
QV, Variable droop and Constant droop controls. Figure 14 shows the transient reactive
power that inserts into the PCC with the corresponding control methods. The voltage
nadir using the fixed QV is higher than that using the variable droop control, although
the peak reactive power using the fixed QV is lower than that using the variable droop
control because the fixed QV method has the characteristics of rapid response. That is, the
response time of reactive power is quick.

Figure 13. Transient voltage curve with an AC transmission system.

Figure 14. Transient Reactive power with an AC transmission system.

3.3. A Reactive Power Load Is Added to the System with a VSC-HVDC Transmission System
To investigate whether the proposed methods can work appropriately in the HVDC-
connected wind power system, a reactive load (180 MVar) is assumed to be added into
the system in 8 s. The power factor of power factor control and Vac -Q droop control is set
to 0.97 leading. Figure 15 shows the transient voltage curve at the PCC using the Voltage
control, Vac -Q Droop and Power factor controls. The voltage by using the voltage control

10
Energies 2023, 16, 4128

recovers quickly because it provides a rapid reactive power response after the reactive
power load is added. The transient voltage using the power factor control drops the most
because it maintains a constant power factor without providing extra reactive power.

Figure 15. Transient voltage curve with a VSC-HVDC transmission system.

Figure 16. Transient reactive power with a VSC-HVDC transmission system.

3.4. Three-Phase Short Circuit Fault at the System with an AC Transmission System
In this study, a three-phase short circuit fault for 100 ms occurs at 8 s in point A
(Figures 11 and 12). Figure 17 shows the transient voltage curve at the PCC using Adaptive
QV, Fixed QV, Variable droop and Constant droop controls. Figure 18 shows the maximum
reactive power that inserts into the PCC using Variable droop control, Adaptive QV, Fixed
QV, and Constant droop control. Like the case in Section 3.2, the characteristics of the QV
method with a rapid response can cause a higher voltage nadir than the droop control. The
transient voltage using adaptive QV or fixed QV is similar, while the voltage using the
variable droop or the constant droop control is similar. The voltage nadir at the PCC can
be improved slightly using adaptive QV or fixed QV, which is like the simulation results
in [11].

11
Energies 2023, 16, 4128

Figure 17. The transient voltage at the PCC using AC transmission.

Figure 18. The transient reactive power from the wind farm using AC transmission.

3.5. Three-Phase Short Circuit Fault at the System with a VSC-HVDC Transmission System
A three-phase short circuit fault for 100 ms occurs at 8 s in point A. Figure 19 shows
the transient voltage curves at the PCC using Voltage control, Vac -Q Droop and Power
factor control. The voltage at the PCC using Vac -Q droop control recovers quicker than
that using the power factor control because it provides extra reactive power after the fault.
A large kd value in the Vac -Q droop control can provide more reactive power. The voltage
at the PCC using the voltage control becomes slightly higher after the fault because this
control method can provide much reactive power.

3.6. Comparison of an Increase of Reactive Power Load at Both Systems with Different
Transmission Types
Figure 21 summarizes the transient voltage curves at PCC using Adaptive QV, Voltage
control, Power factor control, and no control. The red line and black line represent the
transient voltage based on the AC and HVDC transmission systems, respectively. Notably,
there is no control for the power factor-based method in the HVDC transmission system.
The voltage drop is less if any control method is implemented at both AC and HVDC
transmission systems because they can provide extra reactive power after the fault. Without
a doubt, the voltage nadir with controls is higher than those without any control. The
voltage nadir using the adaptive QV is higher than that using the voltage control because

12
Energies 2023, 16, 4128

of the the parameters of PI controller. The PI parameter can be adjusted to avoid providing
much reactive power during a three-phase short circuit fault, which causes overvoltage
after a fault.

Figure 19. The transient voltage at the PCC using DC transmission.

Figure 20. The transient reactive power from the wind farm using DC transmission.

Figure 21. Comparison of transient voltage curves.

13
Energies 2023, 16, 4128

Figure 22. Comparison of reactive power curves.

3.7. Comparison of Three-Phase Short Circuit Fault at Both Transmission Systems

Figure 23 shows the transient voltage curves at the PCC using the Voltage control,
Adaptive QV, Power factor control, and no control. The voltage nadir using the voltage
control is higher than that using the adaptive QV control because the reactive power using
the Voltage control can be increased quickly.

Figure 23. Comparison of transient voltage curves.

Figure 24. Comparison of reactive power curves.

14
Energies 2023, 16, 4128

3.8. Comparison of Tripping of Synchronous Generators at Both Transmission Systems

It is assumed that a synchronous generator is tripped at 8 s, and Figure 25 shows the
simulation results, which reveals that the influence on voltage is less.

Figure 25. Transient voltage in Section 3.8.

4. Conclusions
With a high penetration of wind power generation in a power system, wind turbines
should provide more ancillary services like traditional synchronous generators. Thus,
some voltage control methods, such as voltage droop control and QV control, have been
proposed recently. Besides, many offshore wind farms will be installed in the world, and
high voltage direct current-based transmission systems should be utilized to connect wind
farms and the main power grid, especially when the installed capacity or the transmission
distance of offshore wind farms is increased. Therefore, this study compared both voltage
droop control and QV control for wind turbines, and both voltage control and Vac -Q control
for HVDC transmission systems. In addition, equivalent wind farms are also established.
From the simulation results based on the test system, the transient voltage nadir using
any control is higher than that without any control, and the QV control can achieve a better
result compared to the droop control in both scenarios. Additionally, the voltage nadir
using the voltage control with an HVDC transmission system is higher than that with
an AC transmission system during a three-phase fault.
The fluctuation of voltage at the connected point varies based on the grid strength.
That is, the voltage fluctuation is high if the wind farm is connected to a weak grid,
and vice versa. As a result, applying an appropriate voltage control method should
consider the grid strength. For instance, if the grid strength at the PCC is weak, additional
reactive power compensation devices could be needed to maintain the voltage stability.
Additionally, coordinated control schemes combining wind turbines with other devices
should be considered in the future.

Author Contributions: Conceptualization, Y.-K.W. and D.-Y.G.; methodology, Y.-K.W. and D.-Y.G.;
validation, Y.-K.W.; formal analysis, D.-Y.G. and T.-D.T.; investigation, Y.-K.W.; resources, Y.-K.W.;
data curation, Y.-K.W.; writing—original draft preparation, D.-Y.G. and T.-D.T.; writing—review and
editing, Y.-K.W.; visualization, Y.-K.W.; supervision, Y.-K.W.; project administration, Y.-K.W.; funding
acquisition, Y.-K.W. All authors have read and agreed to the published version of the manuscript.
Funding: This research was funded by the Ministry of Science and Technology (MOST) of Taiwan,
grant number MOST 110-2221-E-194-029-MY2.
Conflicts of Interest: The authors declare no conflict of interest.

15
Energies 2023, 16, 4128

References
1. Wu, Y.-K.; Chang, S.-M.; Mandal, P. Grid-connected wind power plants: A survey on the integration requirements in modern grid
codes. IEEE Trans. Ind. Appl. 2019, 55, 5584–5593. [CrossRef]
2. Ellis, A.; Nelson, R.; Von Engeln, E.; MacDowell, J.; Casey, L.; Seymour, E.; Peter, W.; Williams, J.R.; Walling, R.; Barker, C.; et al.
Review of existing reactive power requirements for variable generation. In Proceedings of the 2012 IEEE Power and Energy
Society General Meeting, San Diego, CA, USA, 22–26 July 2012; pp. 1–7. [CrossRef]
3. Martin, J.A.; Hiskens, I.A. Reactive power limitation due to wind-farm collector networks. In Proceedings of the 2015 IEEE
Eindhoven PowerTech 2015, Eindhoven, The Netherlands, 29 June–2 July 2015; pp. 1–6. [CrossRef]
4. Sharma, R. Electrical Structure of Future Off-Shore Wind Power Plant with a High Voltage Direct Current Power Transmission; Technical
University of Denmark: Kongens Lyngby, Denmark, 2012.
5. Silva, B.; Moreira, C.L.; Seca, L.; Phulpin, Y.; Lopes, J.A.P. Provision of Inertial and Primary Frequency Control Services Using
Offshore Multiterminal HVDC Networks. IEEE Trans. Sustain. Energy 2012, 3, 800–808. [CrossRef]
6. Nawir, M.; Adeuyi, O.; Wu, G.; Liang, J. Voltage stability analysis and control of wind farms connected to weak grids.
In Proceedings of the 13th IET International Conference on AC and DC Power Transmission (ACDC 2017), Manchester, UK,
14–16 February 2017. [CrossRef]
7. Schmall, J.; Huang, S.-H.; Li, Y.; Billo, J.; Conto, J.; Zhang, Y. Voltage stability of large-scale wind plants integrated in weak
networks: An ERCOT case study. In Proceedings of the 2015 IEEE Power & Energy Society General Meeting, Denver, CO, USA,
26–30 July 2015; pp. 1–5.
8. Huang, S.-H.; Schmall, J.; Conto, J.; Adams, J.; Zhang, Y.; Carter, C. Voltage control challenges on weak grids with high penetration
of wind generation: ERCOT experience. In Proceedings of the 2012 IEEE Power and Energy Society General Meeting, San Diego,
CA, USA, 22–26 July 2012; pp. 1–7. [CrossRef]
9. Martínez, J.; Kjær, P.C.; Rodriguez, P.; Teodorescu, R. Design and Analysis of a Slope Voltage Control for a DFIG Wind Power
Plant. IEEE Trans. Energy Convers. 2011, 27, 11–20. [CrossRef]
10. Li, Q.; Zhang, Y.; Ji, T.; Lin, X.; Cai, Z. Volt/Var Control for Power Grids with Connections of Large-Scale Wind Farms: A Review.
IEEE Access 2018, 6, 26675–26692. [CrossRef]
11. Kim, J.; Seok, J.-K.; Muljadi, E.; Kang, Y.C. Adaptive Q–V Scheme for the Voltage Control of a DFIG-Based Wind Power Plant.
IEEE Trans. Power Electron. 2015, 31, 3586–3599. [CrossRef]
12. Li, Y.; Xu, Z.; Zhang, J.; Meng, K. Variable Droop Voltage Control for Wind Farm. IEEE Trans. Sustain. Energy 2017, 9, 491–493.
[CrossRef]
13. Liu, B.; Zhang, X.; Kong, D.; Liu, X.; Chang, X. Research on Reactive Power and Voltage Coordination Control Strategy for
Multi-wind Farm Connecting to Power Grid. In Proceedings of the 2019 IEEE Innovative Smart Grid Technologies-Asia (ISGT
Asia), Chengdu, China, 21–24 May 2019; pp. 1633–1637. [CrossRef]
14. Basu, M.; Kim, J.; Nelms, R.M.; Muljadi, E. Coordination of Utility-scale PV Plant and Wind Power Plant in In-ter-area-Oscillation
Damping. In Proceedings of the IEEE Transactions on Industry Applications 2023, Nashville, TN, USA, 29 October–2 November
2023; pp. 1–9.
15. Jia, K.; Dong, X.; Wen, Z.; Wu, W.; Bi, T. Harmonic Injection Based Fault Ride-through Control of MMC-HVDC Connected
Offshore Wind Farms. IEEE Trans. Sustain. Energy 2023, 1–11. [CrossRef]
16. Dai, J.; Wan, L.; Chang, P.; Liu, L.; Zhou, X. Reactive Voltage Control Strategy for PMSG-Based Wind Farm Considering Reactive
Power Adequacy and Terminal Voltage Balance. Electronics 2022, 11, 1766. [CrossRef]
17. Tan, H.; Li, H.; Yao, R.; Zhou, Z.; Liu, R.; Wang, X.; Zheng, J. Reactive-voltage coordinated control of offshore wind farm
considering multiple optimization objectives. Int. J. Electr. Power Energy Syst. 2021, 136, 107602. [CrossRef]
18. Liu, W.; Liu, Y.; Wu, L. Model Predictive Control Based Voltage Regulation Strategy Using Wind Farm as Black-Start Source. IEEE
Trans. Sustain. Energy 2023, 14, 1122–1134. [CrossRef]
19. Mahish, P.; Mishra, S. Synchrophasor Data Based Q-V Droop Control of Wind Farm Integrated Power Systems. IEEE Trans. Power
Syst. 2022, 38, 358–370. [CrossRef]
20. Huang, S.; Wu, Q.; Guo, Y.; Chen, X.; Zhou, B.; Li, C. Distributed Voltage Control Based on ADMM for Large-Scale Wind Farm
Cluster Connected to VSC-HVDC. IEEE Trans. Sustain. Energy 2019, 11, 584–594. [CrossRef]
21. Han, C.; Huang, A.Q.; Baran, M.E.; Bhattacharya, S.; Litzenberger, W.; Anderson, L.; Johnson, A.L.; Edris, A.-A. STATCOM impact
study on the integration of a large wind farm into a weak loop power system. IEEE Trans. Energy Convers. 2008, 23, 226–233.
22. Mi, Z.; Tian, H.; Yu, Y.; Su, X.; Fan, X.; Feng, J. Study on voltage stability of power grid with large scale wind farm inter-connected.
In Proceedings of the 2009 International Conference on Sustainable Power Generation and Supply, Nanjing, China, 6–7 April 2009;
pp. 1–6.
23. Ding, R.; Meng, C.; Qiao, Y. The coordinating control of voltage and reactive power between SVC and DFIG after LVRT.
In Proceedings of the 2015 IEEE Eindhoven PowerTech 2015, Eindhoven, The Netherlands, 29 June–2 July 2015; pp. 1–5.
[CrossRef]
24. Latorre, H.F.; Ghandhari, M. Improvement of voltage stability by using VSC-HVdc. In Proceedings of the 2009 Transmission &
Distribution Conference & Exposition: Asia and Pacific, Seoul, Republic of Korea, 26–30 October 2009; pp. 1–4.
25. Hussein, I.I.; Essallah, S.; Khedher, A. Improvement of the Iraqi Super Grid Performance Using HVDC/HVAC Links by the
Integration of Large-Scale Renewable Energy Sources. Energies 2022, 15, 1142. [CrossRef]

16
Energies 2023, 16, 4128

26. Liu, H.; Chen, Z. Impacts of large-scale offshore wind farm integration on power systems through VSC-HVDC. In Proceedings of
the 2013 IEEE Grenoble Conference, Grenoble, France, 16–20 June 2013; pp. 1–5. [CrossRef]
27. Xu, J.; Liu, B.; Torres-Olguin, R.E.; Undeland, T. Grid integration of large offshore wind energy and oil & gas installations using
LCC HVDC transmission system. In Proceedings of the SPEEDAM 2010, Pisa, Italy, 14–16 June 2010; pp. 784–791. [CrossRef]
28. Bernat, J.O.; Preece, R. Impact of VSC-HVDC Reactive Power Control Schemes on Voltage Stability. In Proceedings of the 2019
IEEE PowerTech, Milan, Italy, 23–27 June 2019; pp. 1–6. [CrossRef]
29. Arani, M.F.M.; Mohamed, Y.A.-R.I. Analysis and Performance Enhancement of Vector-Controlled VSC in HVDC Links Connected
to Very Weak Grids. IEEE Trans. Power Syst. 2016, 32, 684–693. [CrossRef]
30. Egea-Alvarez, A.; Fekriasl, S.; Hassan, F.; Gomis-Bellmunt, O. Advanced vector control for voltage source converters connected to
weak grids. IEEE Trans. Power Syst. 2015, 30, 3072–3081. [CrossRef]
31. Zhao, M.; Yuan, X.; Hu, J.; Yan, Y. Voltage dynamics of current control time-scale in a VSC-connected weak grid. IEEE Trans.
Power Syst. 2015, 31, 2925–2937. [CrossRef]
32. Guo, Y.; Gao, H.; Wu, Q.; Zhao, H.; Østergaard, J. Coordinated voltage control scheme for VSC-HVDC connected wind power
plants. IET Renew. Power Gener. 2018, 12, 198–206. [CrossRef]
33. Guo, Y.; Gao, H.; Wu, Q.; Zhao, H.; Ostergaard, J.; Shahidehpour, M. Enhanced Voltage Control of VSC-HVDC-Connected
Offshore Wind Farms Based on Model Predictive Control. IEEE Trans. Sustain. Energy 2017, 9, 474–487. [CrossRef]
34. Sakamuri, J.N.; Rather, Z.H.; Rimez, J.; Altin, M.; Göksu, Ö.; Cutululis, N.A. Coordinated voltage control in offshore HVDC
connected cluster of wind power plants. IEEE Trans. Sustain. Energy 2016, 7, 1592–1601. [CrossRef]
35. Pourbeik, P. Proposed Changes to the WECC WT3 Generic Model for Type 3 Wind Turbine Generators; Electric Power Research Institute,
Inc.: Palo Alto, CA, USA, 2013; Volume 12, p. 11.
36. Ellis, A.; Kazachkov, Y.; Muljadi, E.; Pourbeik, P.; Sanchez-Gasca, J.J. Description and technical specifications for generic WTG
models—A status report. In Proceedings of the 2011 IEEE/PES Power Systems Conference and Exposition, Phoenix, AZ, USA,
20–23 March 2011; pp. 1–8. [CrossRef]
37. Lu, G.-L.; Lin, C.-H.; Wu, Y.-K. Comparison of Communication-Based and Coordination-Based Frequency Control Schemes for
HVdc-Connected Offshore Wind Farms. IEEE Trans. Ind. Appl. 2021, 57, 3352–3365. [CrossRef]
38. Lin, C.-H.; Wu, Y.-K. Overview of Frequency-Control Technologies for a VSC-HVDC-Integrated Wind Farm. IEEE Access 2021,
9, 112893–112921. [CrossRef]
39. Muljadi, E.; Butterfield, C.; Ellis, A.; Mechenbier, J.; Hochheimer, J.; Young, R.; Miller, N.; Delmerico, R.; Zavadil, R.; Smith, J.
Equivalencing the collector system of a large wind power plant. In Proceedings of the IEEE Power Engineering Society General
Meeting 2006, Montreal, QC, Canada, 18–22 June 2006. [CrossRef]
40. Wu, Y.-K.; Zeng, J.-J.; Lu, G.-L.; Chau, S.-W.; Chiang, Y.-C. Development of an equivalent wind farm model for frequency
regulation. IEEE Trans. Ind. Appl. 2020, 56, 2360–2374. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

17
 energies
Article
Unique Symbolic Factorization for Fast Contingency Analysis
Using Full Newton–Raphson Method
Hakim Bennani 1, *, Ahmed Chebak 2 and Abderrazak El Ouafi 1

1 Mathematic Computer and Engineering Department, University of Quebec at Rimouski,

Rimouski, QC G5L 3A1, Canada; abderrazak_elouafi@uqar.ca
2 Green Teck Institute, Mohammed VI Polytechnic University, Benguerir 43150, Morocco;
ahmed.chebak@um6p.ma
* Correspondence: hakim1201@hotmail.com

Abstract: Contingency analysis plays an important role in assessing the static security of a network.
Its purpose is to check whether a system can operate safely when some elements are out of service.
In a real-time application, the computational time required to perform the calculation is paramount
for operators to take immediate actions to prevent cascading outages. Therefore, the numerical
performance of the contingency analysis is the main focus of this current research. In power flow
calculation, when solving the network equations with a sparse matrix solver, most of the time is
spent factorizing the Jacobian matrix. In terms of computation time, the symbolic factorization is the
costliest operation in the LU (Lower-upper) factorization process. This paper proposes a novel method
to perform the calculation with only one symbolic factorization using a full Newton–Raphson-based
generic formulation and modular approach (GFMA). The symbolic factorization retained can be used
during the iterations of any power flow contingency scenario. A computer study demonstrates that
reusing the same symbolic factorization greatly reduces computation time and improves numerical
performance. Power system security assessment under N-1 and N-2 contingency conditions is
performed for the IEEE standard 54-bus and 108-bus to evaluate the numerical performance of the
proposed method. A comparison with the conventional power flow method shows that the time
required for the analysis is shortened considerably, with a minimum gain of 228%. The comparative
Citation: Bennani, H.; Chebak, A.; El analysis demonstrates that the proposed solution has better numerical performance for large-scale
Ouafi, A. Unique Symbolic networks.
Factorization for Fast Contingency
Analysis Using Full Keywords: contingency analysis; symbolic factorization; generic formulation; KLU solver
Newton–Raphson Method. Energies
2023, 16, 4279. https://doi.org/
10.3390/en16114279

Academic Editors: Ying-Yi Hong 1. Introduction

and Sérgio Cruz Contingency analysis is regarded as an integral part of power system security anal-
Received: 3 April 2023
ysis as it determines the integrity of the power system and verifies the post-contingency
Revised: 3 May 2023 equilibrium state in terms of overloads and voltage deviations [1,2]. Its main purpose is to
Accepted: 17 May 2023 check whether the system can operate safely without any limit violation (i.e., equipment
Published: 23 May 2023 overloading, under- or over-voltage) after the occurrence of a contingency [3].
In N-1 contingency analysis, the main focus is to evaluate the effects of a single
equipment outage on power system operating conditions once the single component (such
as the loss of a generator, line, and transformer) has been retrieved from the network.
Copyright: © 2023 by the authors. Traditional N-1 contingency analysis performs numerous power-flow runs with a total
Licensee MDPI, Basel, Switzerland. number of N scenarios. After solving the power flow problem for each contingency scenario,
This article is an open access article the limits are verified for branches and nodes. The system is N-1 secure if, for all scenarios,
distributed under the terms and the following constraints are fulfilled:
conditions of the Creative Commons
Attribution (CC BY) license (https:// I ≤ Imax
creativecommons.org/licenses/by/ Umin ≤ U ≤ Umax
4.0/).

Energies 2023, 16, 4279. https://doi.org/10.3390/en16114279 18 https://www.mdpi.com/journal/energies

Energies 2023, 16, 4279

The extension of N-1 contingency analysis considers simultaneous component outages

(N-k contingencies). N-k security refers to the ability of the system to maintain secure
operation after k components have failed in a system with N components [4,5]. For a
large-scale network, strict N-k checking requires a large amount of power flow calculation.
However, in real-time security assessment, the time window for system operators to analyze
the problem and take corrective actions is quite limited. Therefore, the main challenge is to
find an efficient method to significantly reduce the resolution time without compromising
solution accuracy [6].
In contingency analysis, the power flow method used to perform the contingency
analysis has a direct impact on the solution time. The Newton–Raphson power flow-based
approach is generally preferred for its superior convergence characteristics and accuracy.
The most computationally demanding part of this method is solving the linear equations
with LU (Lower-upper) factorization at each iteration. In terms of computation time, the
symbolic factorization is the costliest operation in the LU factorization process. This paper
focuses on using the Newton–Raphson method to solve contingency analysis. The central
question is whether it is possible to avoid the repetitive symbolic factorizations to effectively
shorten the analysis time.
We will now highlight the strengths and weaknesses of the most important methods
used to reduce the computational time of contingency analysis. This critical literature
review will make evident the effectiveness of the proposed solution in simulating the
contingency analysis of large-scale networks (Transmission and distribution systems).
The issue of improving computational speed in contingency analysis has been dis-
cussed in many works. In [7,8], the compensation theorem is applied to simulate the
changes in the passive elements of the network without changing the Jacobian matrix
structure. However, this method is well suited to applications involving linear network
equations, which is not the case for analysis based fully on Newton’s method.
The main alternative to the compensation method for network matrix modification is
to perform partial refactorization [9]. This method consists of updating the LDU (Lower-
diagonal-upper) factors to reflect changes in some elements of the Jacobian matrix. The
weakness of this method is that the refactorization effort is affected by the position of the
modified matrix elements; thus, little or no savings can be obtained if one of the modified
elements is near the top of the matrix.
The DC power flow method has been widely used to accelerate the computation of
contingency analysis and improve its numerical performance [10]. The DC power flow
simplifies the AC power flow to a linear circuit problem. This method is convergent and
non-iterative but less accurate than AC power flow solutions.
A reduction in computational time can be achieved by reducing the number of contin-
gency cases to be tried [11,12]. This approach involves ranking contingencies in descending
order of severity index. The Contingencies can then be simulated, starting with the most
severe and continuing until there is no overloading caused by the outage of branches.
The main shortcoming of this method is that the contingency ranking by the index may
introduce some errors due to the screening effect.
In [13], the graph theory was used to demonstrate that the base case symbolic factor-
ization can be reused to solve contingency cases. The key assumption of this method is
that there is no branch to be put in service during a contingency. This limitation prevents
practical use in distribution systems where the introduction of new lines to reroute flows
occurs in the most interesting contingencies.
Recent research in the area includes optimization-based methods [14–20]. These meth-
ods do not enumerate every contingency but attempt to find the most severe contingency
to reduce the computational burden.
Table 1 demonstrates a summary of the reviewed classical methods for managing
contingency analysis, with the major strengths and weaknesses of each one.

19
Energies 2023, 16, 4279

Table 1. Summary of the classical methods of contingency analysis.

Method Strengths Weaknesses Reference

• Fast linearized power flow outage
analysis. • Nonlinear networks are not
• Requires less computer storage. supported.
Compensation method • Simulate only the outage of [7,8]
• Include the assessment of
simultaneous outages of passive elements.
k-branches (N-k contingencies).
• Numerical performance depends
on the position of the modified
• Factorization is performed only elements.
Partial refactorization on the affected part of the matrix. [9]
• Numerical performance decreases
if the number of changes in the
matrix is small.
• Very fast. • Risk of divergence.
DC power flow [10]
• Non-iterative approach. • Less accurate.

• Number of scenarios to be • Errors due to the screening effect.

Severity index evaluated is greatly reduced. • Do not enumerate every [11,12]
contingency.

• Unsuited to simulate remedial

• Possibility of re-using the base actions.
Graph theory case symbolic factorization. • Efficient for contingencies [13]
• Fast. involving only a few branch
elements.
• Reduction in CPU time.
• Remedial actions can be included • Do not enumerate every
in the optimization function. contingency.
Optimization methods [14–20]
• Include the assessment of • High Complexity.
simultaneous outages of
k-branches.

Additionally, the aforementioned methods have only addressed cases where the
change in the sparsity pattern of the Jacobian matrix is caused by the change in network
topology. However, in an automatic adjusting solution [21,22] a new symbolic factorization
is not only required for a change involving network topology but also during Newton’s
iterative process (due to control adjustment); hence, the symbolic factorization of the base
case Jacobian matrix cannot be systematically reused in contingency cases.
In this paper, an original modeling technique has been proposed to avoid the occur-
rence of change in the sparsity pattern within the same power flow run (due to control
adjustments) and after evaluating a new contingency scenario (due to a change in network
topology). The proposed solution will overcome the limitations identified in classical meth-
ods (Table 1) and allow for the systematic and fast solving of contingency analysis. More
specifically, we propose to perform the contingency analysis using only one symbolic fac-
torization. The proposed approach is completely generic and can easily accommodate the
simultaneous outages of k-branches in the network (i.e., N-k contingency analysis) with the
possibility to include any remedial actions that consist of putting in new service elements.
This paper uses the power flow solver based on GFMA formulation to solve the power
flow problem [23]. The algorithm of GFMA is based on the rigorous full Newton–Raphson
approach and relies on the automatic adjustment technique, which provides an accurate
and fast power flow solution. The high flexibility of this formulation offers the possibility
of re-using the base case symbolic factorization of the Jacobian matrix to recompute the
power flow of any other scenario. The basic idea behind the proposed approach is to imbue
the GFMA power-flow algorithm with the concept of the dynamic parameters to preserve

20
Energies 2023, 16, 4279

the same sparsity pattern of the Jacobian matrix. The dynamic parameters concept is the
main contribution of this research paper.
This paper starts with a brief presentation of the GFMA method for solving power flow
problems. The second part presents the modeling concept of dynamic parameters. In the
third part, the dynamic parameters are introduced into the component models to perform a
unique symbolic factorization in the contingency analysis. The last part presents illustrative
simulation examples to evaluate saved time for relatively large transmission networks.

2. GFMA Formulation
The formulation of GFMA is recalled here to establish the basis for the following
algorithms. GFMA provides a generic power flow formulation capable of handling complex
component models and arbitrary network topology. In this method, each component of
a system is modeled autonomously as a subsystem using a non-causal mathematical
representation. The implicit representation of the system and the modular approach avoid
many theoretical modeling complications, as the variables and load flow constraints can be
arbitrarily defined. The GFMA formulation is generic, extensible, and can handle arbitrary
component models without any known limitation.
Each component is described using the model representation in the form of Equation (1).
For the convenience of the reader, matrices and vectors are indicated in bold type characters.

gk (uk , yk , xk , λk ) = 0 (1)
where:
• uk are the model input variables.
• yk are the model output variables.
• xk are the model internal variables.
• λk are the model variables.
Each model introduces m inputs, n outputs, and h internal state variables. These
variables are expressed in real units and can be arbitrarily defined. The voltages are defined
as inputs to establish the electrical connection with the connected bus and the currents
as outputs to formulate Kirchhoff’s Current Law. The inputs and outputs represent the
interface between the model and the system. The readers can refer to [23] for more details.
The component models are aligned along the diagonal to form the equipment Jacobian
matrix. This block diagonal matrix is augmented with the linking equations to establish the
connection between components. These extra-equations are external to the components
model and have a linear form as follows:

∑ (ci,k .yi,k + d j,l .u j,l ) = 0 (2)

i,j

where yi,k is the ith output variable of the kth model; u j,l is the jth input variable of the lth
model; ci,k and d j,n are the coefficients associated with yi,k and u j,l , respectively.
Herein, the system is initialized using the single-iteration FP solution [24]. The estima-
tion of the initial guess of the state vector is obtained by converting (1) to the linearized
version and solving the resulting linear system of equations in the form of

A.x = b (3)

A comprehensive example is shown in Figure 1 to explore the structure of the Jacobian

matrix. As can be seen from Figure 2, the upper and lower block of the Jacobian matrix
represent the network equipment and linking equations, respectively. The obtained Jaco-
bian matrix is sparse and unsymmetrical. It is important to notice that the system buses
do not contribute to the Jacobian matrix of the system with additional nonzero elements
(Figure 2).

21
Energies 2023, 16, 4279

Figure 1. Test system example.

Figure 2. Matrix structure.

3. Dynamic Parameters Approach

A sparse matrix solver provides efficient numerical solutions to solve linear equations.
The KLU solver is suitable for solving unsymmetrical and sparse linear systems [25] and
is considered one of the most efficient packages designed for electrical circuits. Roughly
speaking, the KLU solver performs two steps in sequence:
1. Symbolic factorization: performs optimal permutations and pre-ordering to reduce
the fill-ins (number of non-zero elements).
2. Numerical factorization: Computes the factorization sub-matrices L and U.
The change in the Jacobian matrix structure is due to the presence of conditional
statements (“if-else”) in the code corresponding to different operating conditions. These
conditions are evaluated, and the appropriate constraint equation is executed (see Equation (4)).

g = f1 if Condition(1)
.. .. ..
. . .
g = fi else i f Condition(i ) (4)
.. .. ..
. . .
g = fn else Condition(n)

22
Energies 2023, 16, 4279

The equation can be formulated as one compact mismatch equation.

i=n
g = λ0 f 0 + ∑ λi f i (5)
i =1

where:
• fi and λi are the non-linear constraint equation and the dynamic parameter for the ith
condition, respectively.
• f 0 and λ0 are the linear constraint equation and the dynamic parameter used for the
ini-tialization. It is worth noting that the term λ0 f 0 can be dropped from Equation (5)
if the model is linear.
The dynamic parameters approach consists of updating the parameters λi from itera-
tion to iteration to enforce the appropriate mismatch equation. Specifically, when a new
condition is flagged, the parameter associated with the enforced constraint equation is
set to one, while the other parameters are all set to zero. This modeling technique pre-
vents a change in the sparsity pattern of the Jacobian matrix since the same mathematical
expression holds for all operating conditions.
Equations (4) and (5) are mathematically equivalent but numerically different. The
main advantage of formulation (5) resides in its high computational performance since the
repetitive symbolic factorizations are systematically avoided.
The dynamic parameters approach has been successfully applied to the power flow
problem and will be extended here for contingency analysis. It will be demonstrated in the
next section that the use of this modeling technique finds useful applications in network
problems involving a change in network topology.

4. Outage Modeling
In this section, the dynamic parameters approach presented in [23] is extended to
avoid the symbolic factorizations related to the change in network topology. Herein, the
constraint equation enforcing the outage of the equipment ( f c ) and its associated parameter
(λc ) are incorporated in the original formulation as follows:

i=n
g = λ c f c + (1 − λ c ).( λ0 f 0 + ∑ λi f i ) (6)
i =1

The first term of Equation (6) prevents change in the nonzero pattern of the Jacobian
matrix caused by the change in the network topology. To simulate a new contingency
scenario, the parameter λc associated with the localized equipment outage must be set
to one. For the base case scenario, the same parameter is set to zero for all elements in
the network.
A flowchart depicting the updates in the dynamic parameters in the contingency
algorithm is shown in Figure 3.
From a computational point of view, the constraint equation f c is considered a judicious
choice if the following conditions are fulfilled:
• Condition 1: Ideally, the constraint equation fc should not introduce additional non-
zero elements in the Jacobian matrix other than those initially generated from fi .
• Condition 2: A linear constraint equation is preferred, as the Newton–Raphson con-
verges more robustly due to its more linear formulation.
In the following sections, the equations for each model are developed using a compact
mismatch equation according to the dynamic parameter approach. For the sake of clarity,
the model equations are expressed in complex form, and the partial derivatives are not
detailed here, as they can be obtained using the symbolical computation [26].

23
Energies 2023, 16, 4279

Figure 3. Updates in the dynamic parameters in the contingency case power flow solution.

4.1. Line Outage

The outage of the line is represented by a linear constraint equation forcing the entering
currents to zero. For the sake of clarity, the line model of Figure 4 is presented as an example.

24
Energies 2023, 16, 4279

Figure 4. Line model.

The vector mismatch equation of the line formulated with the dynamic parameters is
given by      
I I V
g = λc . 1 + (1 − λc ).( 1 − Y. 1 ) (7)
I2 I2 V2
  
‘ fc ‘ f1

It should be noted that the linear constraint equation f 0 used for the initialization is
not required here since the line model is inherently linear.

4.2. Generator Outage

The synchronous generator (SG) is modeled as variable susceptance and conductance
(G and B) [23]. The latter two (G and B) are defined in the model as internal state variables
which are solved to satisfy the power flow constraints of the machine. This model is
numerically more efficient and converges faster than the conventionnel internal voltage
behind reactance [27,28].
The SG state variables added to the system’s variables are as follows:

VkR,a , VkR,b , VkR.c I R,a , I R,b , I R.c

XSG = [ I,a I.b
G, B , k I,a kI.b kI,c ]
I,c ,  (8)
V , Vk , Vk Ik , Ik , Ik
 k  Internal  
inputs outputs

This model introduces eight unknown variables (six for output currents and two for
internal variables) required for the solvability of the same number of equations.
For a machine with PV constraint, the mismatch equation written with dynamic
parameters is given by
         
G G − G0 P − Pdes 0 0
g = λc . + (1 − λ c ).( λ0 . + λ1 . G + λ2 . + λ3 . ) (9)
B B − B0 VG − Vdes QG − Qmin QG − Qmax
        
fc f0 ‘ f1 ‘ f2 ‘ f3

where the active and reactive power is expressed as a function of the state variables
∗
PG + jQG = Iabc Vabc (10)

The sign ∗ indicates the complex-conjugate operator.

This mismatch equation can capture all possible operating conditions of the machine,
including the initialization and equipment outage.
In the initialization stage, the parameter λ0 is set to one while all other parameters
are set to zero, and the non-linear constraint equations of the generator are reverted to
the linearized version. The linear constraint equation f 0 is specifically introduced in the
mismatch equation to estimate the initial value of the susceptance and conductance. For a

25
Energies 2023, 16, 4279

machine with PV constraint, G0 and B0 are estimated, assuming a nominal power factor of
the machine and the desired active power.

conj( Pdes + jQdes )

G0 + jB0 = (11)
|VLLnom |2

The parameters λ1 ,λ2 ,λ3 are dynamically updated during the iterative power flow
process to flag all operating conditions of the generator. This allows the machine to switch
from PV to PQ when its reactive power limit is violated.
The outage of the generator is represented by an equation forcing the susceptance
and conductance to zero. This constraint equation ( f c ) is linear and does not introduce any
additional nonzero elements in the mismatch equation other than those generated from
f 0 , f 1 , f 2 , and f 3 . The generator is considered effectively out of service if the parameter is set
to one.
Six equations are also introduced from the equation relating the terminal voltages
and currents.
Iabc = (AY012 A−1 ).Vabc (12)
where A is the Fortescue’s transformation matrix; Vabc and Iabc are the vector of generator
voltages and current in phase domain, respectively.
⎡ ⎤
Y0 0 0
Y012 =⎣0 G + jB 0⎦ (13)
0 0 Y2

With the introduction of the dynamic parameters, the constraint Equation (13) becomes
⎡ ⎤
Y0 0 0
Iabc = A⎣ 0 λc .( G0 + jB0 ) + (1 − λc ).(λ0 .( G0 + jB0 ) + (λ1 + λ2 + λ3 ).( G + jB)) 0 ⎦A−1 Vabc (14)
0 0 Y2

4.3. Transformer Outage

In this section, the transformer outage model is implemented in a three-phase power-
flow solution method based on the GFMA formulation. Therein, an ideal transformer
connected between nodes k and n is modeled by describing the voltage and current equa-
tions relating the primary and secondary windings. The node m is inserted to include
the three-phase impedance transformer. This solution is generic and can be used to han-
dle arbitrary transformer connections. The transformer model depicted in Figure 5 is
used to derive the constraint equations in the phase domain for an arbitrary three-phase
transformer connection.
The constraint equations for the transformer are given by
⎧       
⎪ −Imabc  −Im Ytr −Ytr Vmabc 
⎪
⎪ λ · + ( 1 − λ ) · abc
−
⎪
⎪
c
Inabc c
Inabc −Ytr Ytr Vnabc
⎨    
g= (15)
⎪
⎪
‘ fc f1
⎪
⎪ Vmabc − D( ratio ).Vkabc
⎪
⎩
Ikabc + DT ( ratio ) · Imabc

where Ytr is the three-phase transformer admittance matrix; ratio is the turn ratio of the
master transformer; D is the dependency that is expressed as a function of turn ratio; and
the sign T indicates the transpose matrix operator.
The transformer outage is simulated by forcing the output currents at both sides of
the transformer to zero. This is achieved by setting the parameter λc to one.

26
Energies 2023, 16, 4279

Figure 5. Three-phase transformer.

4.4. Switching Device Outage

For an ideal switch connected between nodes k and m, the model equation of the
switch device depicted in Figure 6 is given by

g = λc . Ikm + (1 − λc ).(Vk − Vm ) (16)

  
‘ fc ‘ f1

Figure 6. Switch device.

In this equation, the switch is considered in-service if λc is set to zero (i.e., closed
position). The same equation can be used to simulate the outage of the ideal switch by
setting λc to one (i.e., opened position).

4.5. Load Shedding

An optimal load-shedding scheme is necessary under contingency conditions to
prevent cascade outages and complete black-out [29]. The load is formulated using the
current-mismatch equations [30,31]. The load mismatch equation is given by

P − jQdes
g = λc . Ikm + (1 − λc ).(λ0 .( Ikm − YL (Vk − Vm )) + λ1 .( Ikm − des∗ )) (17)
   Vk − Vm∗
‘ fc f0  
‘ f1

where Vk and Vm are the voltages at nodes k and m, respectively; Ikm is the injected load
current; and YL is the estimated load admittance used for the initialization.

5. General Aspects of the Algorithm

With the introduction of the dynamic parameters in the formulation of mismatch
equations, the sparsity pattern of the Jacobian matrix corresponding to the base case sce-
nario becomes identical to that of any contingency case. The unique symbolic factorization
retained in all scenarios of contingency analysis corresponds to that performed for the
system initialization of the base case scenario.
In the base case scenario, all the network equipment is initially in service. However,
any equipment that was initially out of service but expected to be connected as part of
remedial action must be included in the base case scenario and modeled as an out-of-service

27
Energies 2023, 16, 4279

element. From a modeling point of view, adding and removing an element from the system
are treated similarly.
Simultaneous outages (N-2) are treated as single outages (N-1). Both can reuse the
symbolic factorization of the base case scenario.
The main drawback of the proposed approach is that the nonzero elements and the
size of the Jacobian matrix in the contingency cases are slightly overestimated. To illustrate
this point, let us consider an outage of a transmission line that is modeled using dynamic
parameters. The Jacobian elements are obtained by replacing the dynamic parameter λc in
Equation (7) with one and then deriving the partial derivatives with respect to the state
variables. Expressed in real form, the Jacobian elements of the positive sequence model of
the line are given by
∂g ∂g
∂I1R,I = 1 ∂I2R,I = 1
∂g ∂g (18)
∂V1R,I
= 0 ∂V2 R,I
=0
As can be seen from Equation (18), eight nonzero elements contribute to removing a
branch element from the system. These elements will exist only if the symbolic factorization
of the base case scenario is reused for the contingency scenario. This is evident from the fact
that all elements in the base case are initially considered in service. It is worth noting that
the KLU solver does not drop numerically zero entries from its sparse matrix; therefore,
all the entries that are present in the data structure of the Jacobian matrix are not dropped,
even though they are numerically zero.
The contingency analysis is performed with the algorithm depicted in Figure 7.

Figure 7. Contingency algorithm.

28
Energies 2023, 16, 4279

6. Validation
The proposed solution has been tested for the IEEE 57-bus and the IEEE 118-bus test
systems. These networks are used to evaluate the numerical performance of the proposed
solution. The objective is to compare the computational cost of the proposed solution
(GFMA) with the contingency analysis method implemented in CYME power engineering
software. This conventional method consists of performing extensive simulations and
the strict checking of all possible contingencies using a MANA (Modified-augmented-
nodal-analysis)-based power flow solution [32,33]. MANA and GFMA are two different
formulations, but both are implemented in a-b-c reference frame and solved using the
full-Newton algorithm and KLU solver.
Herein, the simulations are performed for the N-1 and N-2 security criteria. The N-1
criterion considers a unique contingency that is cleared by the primary protection, in this
case, the faulted element is disconnected from the system without any further impact on
the system. However, the N-2 criterion considers that the first outage in the system results
in a second component failure. It is worth noting that the N-k security criterion leads to CkN
possible contingencies (CkN = k!( NN!−k)! ).
To evaluate the resolution time of the contingency analysis, timers were inserted into
the code sections, where the symbolic and numerical factorizations are performed. The
normalized CPU time with respect to the proposed solution is presented at the end of this
section, and a comparison between the numerical efficiency of these two methods serves as
a closing statement.
All algorithms have been programmed and executed using the MATLAB Platform,
which runs on a dedicated machine (3.4 GHz i7-2600 CPU with 16 GB of RAM). A tolerance
of 0.01% on the voltage mismatch is used as the stopping criterion for each power flow run.

6.1. Case-1: IEEE 57-Bus Test System

The IEEE 57-bus system shown in Figure 8 is used to test the numerical performance of
the proposed method and to assess the gain in computation time. The network contains 57
buses, 63 lines, 7 generators, 15 two-winding transformers, and 42 loads. The total number
of components that can potentially fail in the system is 85 (N = 63 + 7 + 15). Contingency
is carried out using the GFMA formulation proposed in this paper and MANA formulation
implemented in CYME software.
The strict N-l of the IEEE 57-bus system needs 85 (C185 ) outage analyses. Table 2
shows that both methods require the same number of numerical factorizations; however,
the number of symbolic factorizations required in the MANA is 190, while the proposed
solution performs the same calculation with only one symbolic factorization. Moreover,
when the nonzero pattern is already known, the depth-first search used in Gilbert/Peierls
method can be skipped. This means that numerical factorization becomes computationally
less expensive when the sparsity pattern is not changed. Therefore, the cost of the numerical
factorization is smaller for GFMA than MANA. In this case, the GFMA is about 2.30 faster
than MANA (Table 2).
For N-2 contingency analysis, the IEEE 57-bus system needs 3570 scenarios (C285 ).
As expected, the CPU time is significantly reduced since one symbolic factorization is
performed to evaluate this large number of outage scenarios. In this case, the GFMA is about
2.28 faster than MANA (Table 3). Although the performance ratio depends on the system
size and the number of controller devices, this case study clearly demonstrates that the time
required for the analysis is shortened considerably with the use of the proposed method.

29
Energies 2023, 16, 4279

Table 2. CPU Timings for the N-1 contingency analysis using MANA and GFMA.

IEEE 57-Bus System IEEE 108-Bus System

Proposed Solution CYME Proposed Solution CYME
(GFMA) (MANA) (GFMA) (MANA)
Number of Scenarios 85 240
Reading and Treatment of network file Not included Not included
Build and Update Jacobian Matrix 1.84 s 1.76 s 5.57 s 5.13 s
Symbolic Factorizations Symbolic Factorizations
Number of Symbolic Factorizations 1 190 1 531
Timing of Symbolic Factorizations 0.010 s 2.09 s 0.013 s 6.90 s
Numerical Factorizations Numerical Factorizations
Number of Numerical Factorizations 425 * 425 1177 * 1177
Timing of Numerical Factorizations 1.06 * s 2.97 s 3.01 s * 8.43 s
Solve 0.084 s 0.082 s 0.12 s 0.10 s
2.99 s 6.91 s 4.24 s 20.56 s
Total CPU Time
(100%) (230%) (100%) (484%)
* Numerical Refactorization.

Table 3. CPU Timings for the N-2 contingency analysis using the MANA and GFMA.

IEEE 57-Bus System IEEE 108-Bus System

Proposed Solution CYME Proposed Solution CYME
(GFMA) (MANA) (GFMA) (MANA)
Number of Scenarios 3570 28,680
Reading and Treatment of network file Not included Not included
Build and Update Jacobian Matrix 76.28 s 72.46 s 687.40 s 651.43 s
Symbolic Factorizations Symbolic Factorizations
Number of Symbolic Factorizations 1 8925 1 71,690
Timing of Symbolic Factorizations 0.011 s 93.71 s 0.013 s 931.97 s
Numerical Factorizations Numerical Factorizations
Number of Numerical Factorizations 16,950 * 16,950 136,170 * 136,170
Timing of Numerical Factorizations 42.37 s * 110.17 s * 267.00 s * 695.00 s
Solve 3.65 s 3.54 s 31.86 s 29.45 s
122.31 s 279.89 s 986.27 s 2308.00 s
Total CPU Time
(100%) (228%) (100%) (234%)
* Numerical Refactorization.

The sparsity pattern of the Jacobian matrix of the base case scenario has a dimension
of R3078X3078 with 16,242 nonzero elements (see Figure 9). The matrix structure of the base
case scenario is identical to any contingency case.

30
Energies 2023, 16, 4279

Figure 8. IEEE 57-bus System.

Figure 9. Sparsity pattern of the Jacobian matrix of the base case scenario (IEEE 57-bus System).

31
Energies 2023, 16, 4279

6.2. Case-2: IEEE 108-Bus Test System

This IEEE 108-bus test system was used in various studies for contingency evaluation
and voltage security assessment. This system contains 19 generators, 35 synchronous
condensers, 177 lines, 9 transformers, and 91 loads (Figure 10). The total number of
components is 240 (N = 19 + 35 + 177 + 9). The strict N-l and N-2 checking of the IEEE
108-bus system needs 240 and 3570 outage analyses, respectively.
The comparative analysis presented in Tables 2 and 3 leads to the same findings as
those observed in the IEEE 54-bus system. Once more, the proposed solution proves to be
faster than MANA, with a significant reduction in resolution time.
In addition, the proposed method performs better for larger systems and its effi-
ciency increases as the system increases in size. This can be observed from the results of
Tables 2 and 3, where the detailed CPU time is provided for both networks (IEEE-54 and
IEEE-108 systems). This is explained by the fact that the number of symbolic factorizations
required to cover all contingencies increases drastically with the size of the network.

Figure 10. IEEE 118-bus test system.

7. Conclusions
The proposed modeling technique presented in this paper exploits the flexibility of
the GFMA formulation to avoid the repetitive symbolic factorizations required during the
iterative power flow process and after the modification of network topology. Herein, the
contingency analysis is performed with only one symbolic factorization using a Newton-
like method and KLU sparse matrix solver.
The proposed approach is completely generic and can be applied to simulate simulta-
neous outages of k-branches elements in the network and any remedial actions that consist
of putting in service new elements. The proposed solution is generic and can be applied to
the strict N-k checking without any known limitation.
The new method has been tested on IEEE 57-bus and IEEE 108-bus test systems, with
all the details represented and simulated. The power system security assessment under N-1
and N-2 contingency conditions demonstrates that a significant computational gain has
been achieved by applying the concept of dynamic parameters. The results also showed

32
Energies 2023, 16, 4279

that the method performs better for larger systems and that its efficiency increases as the
system increases in size.

Author Contributions: Conceptualization, H.B.; methodology, H.B.; software, H.B.; validation, H.B.;
formal analysis, H.B.; investigation, H.B.; resources, H.B.; data curation, H.B.; writing—original draft
preparation, H.B.; writing—review and editing, H.B. and A.C.; visualization, H.B.; supervision, A.C.
and A.E.O.; project administration, A.C. and A.E.O.; funding acquisition, A.E.O. All authors have
read and agreed to the published version of the manuscript.
Funding: Financial support from the University of Quebec at Rimouski under grant AEO-760700.
Data Availability Statement: Not applicable.
Acknowledgments: The financial support from the University of Quebec at Rimouski under grant
AEO-760700 is gratefully acknowledged.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Kip, M.; Lei, W.; Prabha, K. Power system security assessment. IEEE Power Energy Mag. 2004, 2, 30–39.
2. Mohammed, S.; William, T.; Yong, F. Impact of Security on Power Systems Operation. Proc. IEEE 2005, 93, 2013–2025.
3. Bulat, H.; Franković, D.; Vlahinić, S. Enhanced Contingency Analysis—A Power System Operator Tool. Energies 2021, 14, 923.
[CrossRef]
4. Liana, C.; Xuan, L.; Zuyi, L. Preventive Mitigation Strategy for the Hidden N-k Line Contingencies in Power Systems. IEEE Trans.
Reliab. 2018, 67, 1060–1070.
5. Babalola, A.; Belkacemi, R.; Zarrabian, S. Real-time cascading failures prevention for multiple contingencies in smart grids
through a multi-agent system. IEEE Trans. Smart Grid 2018, 9, 373–385. [CrossRef]
6. Hassan, H.; Mehrdad, T.; Pedram, S. Accurate, simultaneous and Real-Time screening of N-1, N-k, and N-1-1 contingencies. Int. J.
Electr. Power Energy Syst. 2022, 136, 107592. [CrossRef]
7. William, F.T. Compensation Methods for Network Solutions by Optimally Ordered Triangular Factorization. IEEE Trans. Power
Appar. Syst. 1972, PAS-91, 123–127.
8. Ongun, A.; Brian, S.; William, F.T. Sparsity-Oriented Compensation Methods for Modified Network Solutions. IEEE Trans. Power
Appar. Syst. 1983, PAS-102, 1050–1060.
9. Sherman, M.C.; Vladimir, B. Partial Matrix Refactorization. IEEE Trans. Power Syst. 1986, 1, 193–199.
10. John, K.; Felix, W. Analysis of linearized decoupled power flow approximations for steady-state security assessment. IEEE Trans.
Circuits Syst. 1984, 31, 623–636.
11. Mikolinnas, T.A.; Bruce, W. An Advanced Contingency Selection Algorithm. IEEE Power Eng. Rev. 1981, 2, 25–26. [CrossRef]
12. Halpin, T.F.; Robert, F.; Fink, R. Analysis of automatic contingency selection algorithms. IEEE Power Eng. Rev. 1984, 5, 29–30.
13. Elliott, M.C.; Jingjin, W.; Yiting, Z.; Renchang, D.; Guangyi, L. Symbolic Factorization Re-utilization for Contingency Analysis. In
Proceedings of the IEEE Power & Energy Society General Meeting (PESGM), Portland, OR, USA, 5–10 August 2018; pp. 1–5.
14. Pudi, S.; Sanjeeb, M. Power system contingency ranking using Newton Raphson load flow method. In Proceedings of the Annual
IEEE India Conference (INDICON), Mumbai, India, 13–15 December 2013; pp. 1–4.
15. Rocco, C.M.; Ramirez-Marquez, J.E.; Salazar, D.E.; Yajure, C. Assessing the Vulnerability of a Power System Through a Multiple
Objective Contingency Screening Approach. IEEE Trans. Reliab. 2011, 60, 394–403. [CrossRef]
16. Ding, T.; Li, C.; Yan, C.; Li, F.; Bie, Z. Bilevel Optimization Model for Risk Assessment and Contingency Ranking in Transmission
System Reliability Evaluation. IEEE Trans. Power Syst. 2017, 32, 3803–3813. [CrossRef]
17. Street, A.; Oliveira, F.; Arroyo, J.M. Contingency-Constrained Unit Commitment with n–K Security Criterion: A Robust
Optimization Approach. IEEE Trans. Power Syst. 2011, 26, 1581–1590. [CrossRef]
18. Wang, Q.; Watson, J.; Guan, Y. Two-stage robust optimization for N-k contingency-constrained unit commitment. IEEE Trans.
Power Syst. 2013, 28, 2366–2375. [CrossRef]
19. Levitin, G. Optimal Defense Strategy Against Intentional Attacks. IEEE Trans. Reliab. 2007, 56, 148–157. [CrossRef]
20. Levitin, G.; Hausken, K. Redundancy vs. Protection vs. False Targets for Systems Under Attack. IEEE Trans. Reliab. 2009, 58,
58–68. [CrossRef]
21. Stott, B. Review of load-flow calculation methods. Proc. IEEE 1974, 62, 916–929. [CrossRef]
22. Norris, M.P.; Scott, M. Automatic Adjustment of Transformer and Phase-Shifter Taps in the Newton Power Flow. IEEE Trans.
Power Appar. Syst. 1971, PAS-90, 103–108.
23. Hakim, B.; Ahmed, C.; Abderrazak, E. A generic three-phase power flow formulation for flexible modeling and fast solving for
large-scale unbalanced networks. Int. J. Electr. Power Energy Syst. 2023, 148.
24. Wilsun, X.; Hermann, W.D.; José, R.M. Generalized three-phase power flow method for the initialization of EMTP simulations. In
Proceedings of the POWERCON ’98. 1998 International Conference on Power System Technology. Proceedings (Cat. No.98EX151),
Beijing, China, 18–21 August 1998; Volume 2, pp. 875–879.

33
Energies 2023, 16, 4279

25. Timothy, A.D.; Eknathan, P.N. Algorithm 907: KLU, a direct sparse solver for circuit simulation problems. ACM Trans. Math.
Softw. 2010, 37, 3.
26. Canizares, C.A. Applications of symbolic computation to power system analysis and teaching. In Proceedings of the 2006 PES
Power Systems Conference and Exposition, Atlanta, GA, USA, 29 October–1 November 2006; p. 361.
27. Chen, T.H.; Chen, M.S.; Inoue, T.; Kotas, P.; Chebli, E.A. Three-phase cogenerator and transformer models for distribution system
analysis. IEEE Trans. Power Deliv. 1991, 6, 1671–1681. [CrossRef]
28. Tatianna, B.; Penido, D.R.R.; Araujo, L.R. A new synchronous DG model for unbalanced multiphase power flow studies. IEEE
Trans. Power Syst. 2020, 35, 803–813.
29. Bhuvanagiri, R.; Mohan, K.; Surya, V. Priority Based Optimal Load Shedding in a Power System Network under Contingency
Conditions. In Proceedings of the International Conference for Advancement in Technology (ICONAT), Goa, India, 21–22 January
2022; pp. 1–5.
30. Sereeter, B.; Vuik, K.; Witteveen, C. Newton Power Flow Methods for Unbalanced Three-Phase Distribution Networks. Energies
2017, 10, 1658. [CrossRef]
31. Garcia, P.; Pereira, J.; Carneiro, S.; Da Costa, V.; Martins, N. Three-phase power flow calculations using the current injection
method. IEEE Trans. Power Syst. 2000, 15, 508–514. [CrossRef]
32. Ilhan, K.; Jean, M.; Ulas, K.; Gurkan, S.; Omar, S. Multiphase Load-Flow Solution for Large-Scale Distribution Systems Using
MANA. IEEE Trans. Power Deliv. 2014, 29, 908–915.
33. Jean, M. Régimes transitoires électromagnétiques: Simulation. In Réseaux Électriques et Applications; TECHNIQUES DE
L’INGENIEUR: Saint-Denis, France, 2015.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

34
 energies
Article
DC Admittance Model of VSCs for Stability Studies in
VSC-HVDC Systems
Joaquín Pedra 1, *, Luis Sainz 1 and Lluís Monjo 2

1 Department of Electrical Engineering (ETSEIB—UPC), Universitat Politècnica de Catalunya, Av. Diagonal 647,
08028 Barcelona, Spain; luis.sainz@upc.edu
2 Department of Electrical Engineering (EPSEVG—UPC), Universitat Politècnica de Catalunya,
Av. Victor Balaguer s/n, 08800 Vilanova i La Geltrú, Spain; lluis.monjo@upc.edu
* Correspondence: joaquin.pedra@upc.edu

Abstract: High-voltage direct current (HVDC) systems linked to AC grids with converters are
promising energy transmission systems. These systems present complex AC- and DC-side dynamic
interactions. Impedance-based stability studies have recently been proposed to assess DC-side
dynamics from DC-side characterization of voltage source converters (VSCs) considering AC-side
dynamics. However, the existing approaches used for stability studies in VSC-HVDC systems do
not completely model VSCs because they do not consider together the VSC delay, the grid voltage
feedforward filter, and all the d- and q-reference current controls. Moreover, these approaches are
analytically characterized from dq-real space vectors (less related to circuit theory than dq-complex
space vectors), and some work with simple AC grids. The main contribution of this paper is a detailed
and complete DC admittance model of VSCs from dq-complex space vectors, which considers the
VSC delay, feedforward filter, and d- and q-reference current controls, and also a general AC grid.
The proposed model can be used for DC-side stability studies in VSC-HVDC systems considering AC
grid dynamics. The capabilities and drawbacks of impedance-based stability methods for DC-side
stability assessment were analyzed, and the positive-net-damping criterion was validated as a robust
approach. The model was validated by PSCAD/EMTDC simulations and applied to a stability study
in a VSC-HVDC system.

Keywords: impedance modeling; voltage source converters; HVDC transmission

Citation: Pedra, J.; Sainz, L.; Monjo,
L. DC Admittance Model of VSCs for
Stability Studies in VSC-HVDC
Systems. Energies 2023, 16, 5457. 1. Introduction
https://doi.org/10.3390/ VSC-HVDC systems are emerging as a future transmission energy technology [1].
en16145457 Near-synchronous [1–3] and harmonic [4,5] oscillatory instabilities due to low-damped
Academic Editor: Ying-Yi Hong resonances are a usual concern in these systems, and their accurate assessment is challeng-
ing for researchers. The state-space and frequency-domain methods are the most common
Received: 26 June 2023 approaches used to analyze these oscillatory phenomena [1–10]. The state-space method [3]
Revised: 13 July 2023
is a global stability approach (i.e., stability is determined regardless of location [2]) but uses
Accepted: 16 July 2023
high-order dynamic models and requires specific information about all system elements,
Published: 18 July 2023
which is not always accessible. On the other hand, the frequency-domain methods assess
system stability with less computing effort than the state-space method and are used either
with simulations or system measurements. Broadly speaking, frequency-domain methods
Copyright: © 2023 by the authors.
are applied using impedance-based stability criteria, such as the Nyquist criterion [4–6],
Licensee MDPI, Basel, Switzerland. impedance-based analysis [1], and the positive-net-damping stability criterion [4–9] for
This article is an open access article load–source equivalent systems, and the generalized Nyquist criterion (GNC) [2,6,10] and
distributed under the terms and impedance matrix determinant analysis [3] for interconnected systems. Most of these
conditions of the Creative Commons criteria are local stability approaches (i.e., they are locally applied) sensitive to system
Attribution (CC BY) license (https:// partitions, which can lead to inaccurate stability predictions.
creativecommons.org/licenses/by/ This weakness has been discussed in the literature to understand the proper applica-
4.0/). tion of the different criteria [2].

Energies 2023, 16, 5457. https://doi.org/10.3390/en16145457 35 https://www.mdpi.com/journal/energies

Energies 2023, 16, 5457

Stability studies in HVDC systems require DC-side equivalent modeling of converters.

Early studies on VSC-HVDC systems considered weak and simple R–L AC grids and a
few VSC controls to obtain the DC-side model of VSCs [4,9]. This, in general, leads to
inaccurate stability predictions because AC-side and VSC control dynamics are overlooked.
Improved DC-side models of VSCs that consider AC-side dynamics have recently been
proposed [11–15]. The first DC-side equivalent model for VSC rectifier and inverter stations
of VSC-HVDC systems with active power and DC voltage controls, respectively, and
grid alternating voltage control is derived from dq-real space vectors in [11]. The model
disregards the VSC delay and phase-locked loop (PLL) control. Moreover, the study
considers a simple R–L impedance on the AC-side and wrongly concludes that AC-side
instabilities are not detected on the DC side. A DC-side equivalent admittance model
of a VSC-HVDC station in an offshore wind farm with DC voltage and PLL controls
using dq-real space vectors is presented in [12]. Additionally, the modified sequence and
phasor-domain models are derived from the previous model. The model disregards the
VSC delay, grid voltage feedforward filter, and q-reference current controls. A general
AC grid characterization to correctly analyze the influence of AC-side dynamics on DC-
side instabilities (e.g., the mirror frequency effect [6,16]) is also considered in [12]. A DC
impedance model of VSC rectifier and inverter stations of VSC-HVDC systems with active
power and DC voltage controls, respectively, and PLL control is derived by using dq-real
space vectors in [13–15]. The VSC delay and a general AC grid are also considered, but the
grid voltage feedforward filter and the q-reference current controls are disregarded.
The above comments are summarized in Table 1. The first row (labeled “Phase-locked
loop control”) shows the models in which the PLL control is included. The second row
(labeled “VSC delay”) shows the models in which the VSC time delay transfer function
is included. The third and fourth rows show the models which use voltage feedforward
without and with a low-pass filter. The fifth and sixth rows show the models which include
the outer control loops on channels d and q.

Table 1. Comparison of models.

References
Proposed
Model Feature [11] [12] [13] [14] [15] Model

Phase-locked loop control

VSC delay
Grid voltage feedforward
Grid voltage feedforward filter
D-outer control loop
Q-outer control loop

The influence of AC-side dynamics on stability and the stability assessment using
the impedance-based stability criteria have been discussed. To better explain frequency
coupling dynamics between the AC and DC sides (i.e., the mirror frequency effect), modi-
fied sequence [16] and dq-complex-domain [17] approaches have been proposed from the
dq-real-domain model based on the Park (dq-real) and Ku (dq-complex) transformations [18].
Both dq-real- and dq-complex-domain approaches can be used for VSC modeling, but they
have advantages and disadvantages for Laplace and frequency studies of grid-connected
VSC stability:
• dq-real-domain approach:
– Laplace stability studies: It allows system poles and participation factors to be
easily obtained but is not appropriate for integrating components of large power

36
Energies 2023, 16, 5457

systems. Moreover, frequency black-box models obtained from measurements

cannot be used with this approach.
– Frequency stability studies: It is the most used in the literature [11–15] but reduc-
ing AC networks to their equivalent Thevenin circuit in dq-real variables is more
laborious than in dq-complex variables. This is because AC-side impedances are 2
× 2 matrices with dq-real variables.
• dq-complex-domain approach:
– Laplace stability studies: The characterization of system poles poses numerical
problems because of rounding errors produced by the complex coefficients of
dq-complex-domain transfer functions. Polynomials in the s-domain with only
real coefficients are required in the DC impedance to correctly obtain the system
poles.
– Frequency stability studies: It has recently been used in the literature [19,20]. In
the dq-complex domain, the AC-side elements use the impedances ZTh (s + jω1 )
and ZTh (s − jω1 ) obtained by shifting the operational Thevenin impedance ZTh (s).
The dq-complex domain is more appropriate for integrating VSC models in large
power systems.
According to the above comments, the benefits of the dq-complex-domain approach
over the dq-real-domain approach for grid impedance modeling and stability studies
are greater.
This paper contributes to filling the above gaps in the characterization of the DC
admittance model of VSCs. Moreover, it studies the inaccuracies of impedance-based
stability criteria due to partitions in VSC-HVDC systems in two ways. The first contribution
is a novel DC admittance model for VSCs based on dq-complex space vectors. The model
extends existing models by considering the VSC delay, grid voltage feedforward filter, PLL,
and main d-reference (active power and DC voltage controls) and q-reference (reactive
power and grid alternating voltage controls) current outer loops, and also a general AC grid.
The second contribution involves the study of VSC-HVDC system stability. The sensitivity
of impedance-based stability criteria to VSC-HVDC system partitions is discussed, and
the positive-net-damping stability criterion is proposed to overcome this issue. Finally,
the influence of AC-side dynamics on the DC-side stability of VSC-HVDC systems was
analyzed. The stability issues in a VSC-HVDC system were studied with the proposed DC
admittance model to show the paper’s contribution. PSCAD/EMTDC simulations were
performed to validate the study.

2. Model of VSCs in DQ-Complex Domain

Complex transfer functions are used in this section to model VSCs with dq-complex
space vectors [6,16,17]. This approach is used in the next section to obtain the DC admittance
of the VSCs.
The Laplace variable s will be omitted in the dq-complex transfer functions hereinafter
√
to simplify notation, and the dq-complex space vectors are defined as Udq = (Ud + jUq )/ 2
√
and Idq = (Id + jIq )/ 2.

2.1. AC Grid Relations

The relation between voltage and current at the point of common coupling in Figure 1
is expressed as
   +  
ΔEdq −  ΔI
ZTh ZTh ±
= −  − ∗
   ∗
dq
⇒ ΔEm dq = −ZTh ΔIdq ,
m
(1)
ΔE*dq ZTh Z+Th ΔI*dq

37
Energies 2023, 16, 5457

where Z+Th = ZTh (s + jω 1 ), (Z+Th )∗ = ZTh (s − jω 1 ) and ZTh

−
= 0. The voltage balance across
the converter Rc − Lc filter [20] is expressed as
   +    
ΔVdq
 Zc  Zc− ΔIdq ΔEdq ± m
=− ∗ + ⇒ ΔVm dq = −Zc ΔIdq + ΔEdq ,
m
(2)
ΔV*dq Zc− (Z+c )∗ ΔI*dq ΔE*dq

where Z+c = Rc + Lc (s + jω 1 ), (Z+c )∗ = Rc + Lc (s − jω 1 ), and Zc− = 0.

Figure 1. Grid-connected VSC system for DC impedance characterization.

−
Note that ZTh = Zc− = 0 because the grid and the Rc − Lc filter are three-phase symmetric
systems whose impedance matrices in the dq-domain verify the relations Zdd (s) = Zqq (s)
and Zdq (s) = −Zqd (s). Considering the above, the terms Z+Th and Z+c can also be expressed
by using the coordinate transformation properties Z+Th = ZTh (s + jω 1 ) and Z+c = Zc (s + jω 1 )
= Rc + Lc (s + jω 1 ) [6].
The relation between VSC output voltage and AC current is derived from (1) and (2):
± m ± ± ±
dq = −Zt ΔIdq Zt = (Zc + ZTh ),
ΔVm (3)

where Z+t = Zt (s + jω 1 ) and Zt (s) = ZTh (s) + Rc + Lc s is the AC impedance.

2.2. DC-AC Converter Relations

The relation between the VSC output voltage and the DC voltage and the relation
between the DC and AC currents are expressed [20] as
     
ΔVdq Δmdq mdq0
= V + ΔVdc ⇒ ΔVmdq = Δmdq Vdc0 + mdq0 ΔVdc ,
m m
(4)
ΔV*dq Δm*dq dc0 m*dq0
   
  ΔIdq   Δmdq
ΔIdc = − m*dq0 mdq0 − I*dq0 Idq0
ΔI*dq Δm*dq
(5)
⇒ ΔIdc = −(mm dq0 ) ΔIdq − (Idq0 ) Δmdq ,
H m m H m

√ √
where mdq = (md + jmq )/ 2 and Idq0 = (Id0 + jIq0 )/ 2, and the superscript H indicates the
transpose and complex conjugate (Hermitian conjugate). Note that the three-phase set of
modulation functions {ma , mb , mc } = M{cos(ω 1 t + φm ), cos(ω 1 t + φm − 2π/3), cos(ω 1 t + φm
√
+ 2π/3)} is transformed into the dq-complex space vector mdq = ( 3/2)M∠φm by applying
the normalized Ku (dq-complex) transformation [6,18].

38
Energies 2023, 16, 5457

2.3. Simplified Model #1: Case without Current Control

∗ = 0.
In the case without current control, the modulation function is constant, Δmdq
Then, (4) and (5) result in

dq = mdq0 ΔVdc ΔIdc = −(mdq0 ) ΔIdq .

H
ΔVm m m m
(6)

By considering (3) and (6), the DC admittance is

± −1
ΔIdc = Ydc ΔVdc Ydc = (mm
dq0 ) (Zt )
H
mm
dq0 , (7)

where
   Z+  −1    
0 mdq0 3M2
Ydc = m*dq0 mdq0 t
= 1
+ 1
0 (Z+t )* m*dq0 4 Z+t (Z+t )*
(8)
Z+t = Zt (s + jω1 ) = ZTh (s + jω1 ) + Rc + Lc (s + jω1 )
(Z+t )* = Zt∗ (s + jω1 ) = ZTh (s − jω1 ) + Rc + Lc (s − jω1 ),

and the DC impedance is the inverse of the DC admittance, Zdc = 1/Ydc .

The AC grid is represented by the circuit in Figure 2 [12], with the Thevenin transfer
function written as
1 1 1
ZTh (s) = ; Z f (s) = + . (9)
1
Rg + Lg s + 1
Z f (s)
Cf s 1
Lf s + 1
Rf

Figure 2. AC grid connected at the VSC terminals [12].

Note that the frequency response of a conjugate transfer function such as (Z+ t )* (s) in
(8) can be calculated as [19].

Z ∗ (s) = [ Z (s∗ )]∗ ⇒ Z ∗ ( jω ) = [ Z (− jω )]∗ . (10)

2.4. Simplified Model #2: Case with Current Control

Only the modulation function and current control of the VSC with delay are considered.
Then, the current control loop [20] is expressed as
± ± ±
dq, r = Fcc ΔIdq − Zω ΔIdq + Hf ΔEdq ,
ΔVm m m m
(11)

− = Z− = H− = 0, with
where F+cc = D(s)Fcc (s), Z+ω = D(s)·jLc ω 1 , H+f = D(s)Hf (s), and Fcc ω f
D(s) = e − sTd being the VSC time delay transfer function, Fscc (s) = ksp, cc + ksi, cc /s, and
Hf (s) = αf /(s + αf ), with ksp, cc, and ksi, cc being the proportional and integral gains of the
current control, αf being the bandwidth of the voltage feedforward low-pass filter, and Td
being the VSC time delay. The VSC reference voltage, ΔVm dq, r = Vdc0 Δmm dq (11), is the
dq-complex space vector determining the modulation function:

39
Energies 2023, 16, 5457

1  ± m 
dq =
Δmm Zpi ΔIdq + Hf± ΔEm ± ± ±
dq ; Zpi = Fcc − Zω . (12)
Vdc0

By considering (1) and (12), (5) becomes

⎡ ⎤
(Im )H  ± ± ±

ΔIdc = ⎣−(mdq0 ) − Zpi − Hf ZTh ⎦ΔIm
m H dq0
dq , (13)
Vdc0

and (4) can be expressed from (1), (2), and (12) as

 ±  m  
± m ± ± ± ±
dq = −Zc ΔIdq + ΔEdq = − Zc + ZTh ΔIdq = Zpi − Hf ZTh ΔIdq + mdq0 ΔVdc , (14)
ΔVm m m m

which can be rewritten as

  −1
± ± ± ± ±
dq = −Zc − ZTh − Zpi + Hf ZTh
ΔIm dq0 ΔVdc .
mm (15)

By replacing (15) in (13), the DC admittance function is

± ±
ΔIdc = Ydc ΔVdc Zcf = Zpi − Hf± ZTh±
 H

T ∗
(Idq0 )
m
± (16)
Ydc = (mm dq0 ) + V dc0
Zcf (Zt± + Zcf
± −1 m
) mdq0 .

The DC admittance in (16) can finally be expressed as

 
2
Ydc = 3M 4
1
Zt +Zcf
+ + + 1
(Z+ +Z+ )
*
 ( I − jI )(m + jm t )Z+ cf ( I + jI )(m − jm )(Z+ )∗  (17)
+ 2Vdc0
1 d0 q0 d0
Z+t +Z+cf
q0 cf
+ d0 q0 + ∗ d0 + q0∗ cf ,
(Zt ) +(Zcf )

where  
Z+cf = D (s) Fcc (s) − jLc ω1 − H f (s) ZTh (s + jω1 )
  (18)
(Z+cf )* = D (s) Fcc (s) + jLc ω1 − H f (s) ZTh (s − jω1 ) ,

and the DC impedance is calculated with the inverse of the DC admittance, ZDC = 1/YDC .
The DC impedance in the previous simplified models, (8) and (17), is numerically validated
by PSCAD/EMTDC simulation.
Figure 3 shows these DC impedances (see data in Table 2). The case without current
control, ZWC , calculated with (8), is plotted with a solid line whereas the frequency response
obtained by PSCAD/EMTDC simulation is plotted with crosses. The DC impedance case
with current control, ZCC , calculated with (17), is represented in the same way. The
accurate results of the DC-side equivalent admittance obtained with the proposed model
are worth noting.
In the case without current control, the coupling between the DC and AC sides is
revealed by the two peak values of ZWC . The DC-impedance ZWC is affected by the mirror
frequency effect. This is reflected in the expression of YDC (8) (see (1) in [12]), which depends
on the impedances Zt (s + jω 1 ) and Zt (s − jω 1 ). As can be seen in the DC-impedance ZCC in
Figure 3, the control damps this coupling effect between the DC and AC sides [13].

40
Energies 2023, 16, 5457

Figure 3. DC-impedance in the case without control, ZWC , and with current control, ZCC . (Lines:
inverse of (8) and (17). Cross: PSCAD/EMTDC simulations.)

Table 2. AC Grid and VSC Parameters.

√
Vga 12.25/ 2 kV Pvsc 30 MW Rf , Lf 2.3 Ω, 0.18 mH
V dc0 30 kV Rg , Lg 0.52 Ω, 7.8 mH Rc , Lc 0, 4 mH
Cf 35 μF fsw , Td 10 kHz, 0.3 ms Id0 , Iq0 −1996.3 A, 0 A
αf 106.18 s−1 md0 0.5168 pu mq0 0.0836 pu
kp,cc 2 pu kp,dc 0.8662 pu kp,pll 0.2757 pu
ki,cc 0.096 pu ki,dc 0.0334 pu ki,pll 0.2203 pu
kp,ac 0.9 pu kp,P 0.13 pu kp,Q 0.13 pu
ki,ac 0.008 pu ki,P 0.0083 pu ki,Q 0.0083 pu

3. Detailed Model of DC Admittance of VSCs

The PLL and the outer loop relations are included in the detailed model, where the
current control loop is modified to account for the PLL effect. Now, the current control loop
relations must consider the relationship between the variables in the converter dq-domain,
which has the superscript “c”, and the variables in the grid dq-domain, where no superscript
is used. The current control loop [20] is expressed as
 
± ± ±
dq, r = −Fcc ΔIdq, r − ΔIdq − Zω ΔIdq + Hf ΔEdq .
ΔVcm cm cm cm cm
(19)

The dq-complex space vector of the VSC reference voltage is written as

ΔVcm dq, r = Vdc0 Δmcm dq . The PLL relations [9,20] are expressed as

±
dq = ΔIdq + GPLLi ΔEdq
ΔIcm m m
− (20)
G+PLLi = −GPLLi = − GPLL (s)( Id0 + jIq0 )/2,

±
dq = GPLLv ΔEdq
ΔEcm m

GPLLv = 1 − GPLL (s)( Ed0 + jEq0 )/2

+ (21)
−
GPLLv = 1 − G+PLLv = GPLL (s)( Ed0 + jEq0 )/2,

41
Energies 2023, 16, 5457

±
Δmcmdq = Δmdq + GPLLm ΔEdq
m m
− (22)
G+PLLm = −GPLLm = − GPLL (s)(md0 + jmq0 )/2,
where Gpll (s) = Fpll (s)/(s + Vd0 Fpll (s)), Fpll (s) = kp, pll + ki, pll /s, and kp, pll and ki, pll are the
proportional and integral gains of the PLL control.
The reference currents generated by the outer control loops are related to the DC
voltage and the input AC voltage and current of the VSC with dq-complex space vectors as
the following general expression:
± cm ±
dq, r = −Fo ΔVdc − Go ΔIdq − Yo ΔEdq ,
ΔIcm m cm
(23)

where the complex transfer matrices Fm o , G± o , and Y± o are the complex transfer matrices
depending on the VSC operation mode (see Appendix A).
The outer control loop (23) can be introduced into the inner current control loop (19) as
   ±  cm
± m ± ± ± ± ±
dq = ΔVdq, r = Fcc Fo ΔVdc + Zpi + Fcc Go ΔIdq + Hf + Fcc Yo ΔEdq , (24)
Vdc0 Δmcm cm cm

where (24) is expressed in the converter dq-complex domain. By considering the PLL rela-
tions (20), (21), and (22) in the grid dq-domain (where no superscript is used), (24) becomes
 
± m ± ± ± ±
dq = Vdc0 Fcc Fo ΔVdc + (Zpi + Fcc Go ) ΔIdq + Ho ΔEdq
1
Δmm m m
± ± G± )G± + ( H± + F± Y± )G± − V G± ,
(25)
Ho± = (Zpi + Fcc o plli f cc o pllv dc0 pllm

and introducing (1) into (25) results in

1  ± m ± ± ±

dq =
Δmm Fcc Fo ΔVdc + (Zpi + Fcc Go − Ho± ZTh
±
)ΔIm
dq . (26)
Vdc0

The DC voltage in (4) can be related to the AC current from (26) and the AC grid
relations (1) and (2) as follows:
 ± ±
 m ± m ± ± ± ± ±
ΔVm dq = − Zc + ZTh ΔIdq = (mdq0 + Fcc Fo ) ΔVdc + (Zpi + Fcc Go − Ho ZTh ) ΔIdq . (27)
m m

By using (1), (25) is rewritten as

GV ΔVdc = −ZV ΔIm GV = mm ± m

dq dq0 + Fcc Fo
(28)
ZV = Zt± + Za± Za± ± ± ±
= Zpi + Fcc Go − Ho± ZTh
± .

The DC current in (5) can be related to the AC current using (25) and the AC grid
relations (1) and (2) as follows:
H 
∗ dq0 )
( Im ± m
ΔIdc = −(mm
dq0 ) ΔIdq −
T m
Fcc Fo ΔVdc + Za± ΔIm
dq , (29)
Vdc0

which can be rewritten as

H H
dq0 )
( Im ± m ∗ dq0 )
( Im
ΔIdc = −YI ΔVdc − GI ΔIm
dq YI = Fcc Fo GI = (mm
dq0 ) +
T
Za± . (30)
Vdc0 Vdc0

The DC-admittance Ydc = ΔIdc /ΔVdc is obtained from the DC voltage and current
relations (28) and (30), respectively, resulting in
⎛ ⎞
H H
ΔIdc −1 dq0 )
( Im ± m dq0 )
( Im −1
Ydc = = −YI + GI ZV GV = − ⎝
F F + (mdq0 ) +
m H
Za (Zt + Za± ) (mm
±⎠ ± ± m
dq0 + Fcc F0 ). (31)
ΔVdc Vdc0 cc 0 Vdc0

42
Energies 2023, 16, 5457

PSCAD/EMTDC simulations were performed to study the DC impedance calculation

accuracy of different control modes. The parameters of each case are listed in Table 2.
Figure 4 shows the DC impedance in the case of the outer control mode P − Q, ZPQ and P
− Eac , ZPE . Figure 5 shows the DC impedance in the case of the outer control mode Vdc − Q,
ZUQ and Vdc − Eac , ZUE . The simulation data are shown with x crosses in Figures 4 and 5.
The continuous curves are calculated from the inverse of (31) and they fit well with the
simulation results.

Figure 4. DC-impedance of P-Q and P-E control modes. (Lines: inverse of (31). Cross: PSCAD/
EMTDC simulations.)

Figure 5. DC-impedance of Vdc -Q and Vdc -E control modes. (Lines: inverse of (31). Cross:
PSCAD/EMTDC simulations.)

43
Energies 2023, 16, 5457

4. Impedance-Based Stability Criteria

Once the DC-side equivalent admittance of VSCs is obtained, the general HVDC
interconnected system in Figure 6a can be considered for stability studies. The incremental
symbol Δ of the small-signal variables is omitted for the sake of simplicity. The external
components connected to the DC grid buses are represented by their equivalent circuits as
follows:
• Grid-connected VSCs are characterized by the current balance at their Norton equiva-
lent circuit buses:

Ii = Idc, i − Ydc, i (s)Vi (i = 1, . . . a), (32)

where the expression of Ydc, i (s) is the DC admittance in (31).
• DC loads are characterized by their admittance transfer function Yi (s) (i = 1, . . . b).
• Terminals without any connected external component are represented as open-circuit
buses characterized by zero values of the current source and admittance of the Norton
equivalent circuit, Ii = Yi (s) = 0 (i = 1, . . . c).
Several impedance-based stability criteria are proposed in the literature [1–10]. Some
of the differences between these criteria are discussed in [2]. This discussion is extended
in the sections below considering the general DC system in Figure 6a where Ydc, i (s) (31)
is used to assess DC interconnected system stability. The discussion is also supported by
Section 5.

Figure 6. Stability assessment in HVDC transmission grids: (a) HVDC transmission grid. (b) SISO
equivalent system partitioned at VSC terminals.

4.1. Norton Admittance Matrix-Based Stability Assessment

The relations between voltages and currents at the HVDC transmission grid terminals
in Figure 6a are expressed as

i = YG ( s ) v
⇒ v = (YG (s) + YE (s))−1 iE , (33)
i = iE − YE ( s ) v

where YG (s) is the admittance matrix of the HVDC transmission grid; iE and YE (s) are the
current source and the diagonal admittance matrix of the external components, respectively;
and v and i are the voltage and the injected current at the grid terminals, respectively.
Stability can be analyzed in the frequency domain by rewriting (33) as

v = (I + ZG (s)YE (s))−1 ZG (s)iE , (34)

where I is the identity matrix, and ZG (s) = YG −1 (s) is the grid impedance matrix. The
stability of the closed-loop system in (34) can be assessed by the GNC if and only if L(s)
does not have a right-half-plane (RHP) pole. This criterion (called SC1 in this paper)
extends the traditional Nyquist criterion to the Nyquist curves of the eigenvalues of the
loop transfer function L(s) = ZG (s)YE (s) [2,6,10,12,13]. Note that the RHP pole condition

44
Energies 2023, 16, 5457

fails when system instability comes from the AC side of any VSC because the corresponding
VSC subsystem in the diagonal of YE (s) is unstable. This instability was not detected in this
case and, obviously, was also not detected when AC grid dynamics were disregarded in
the DC-side equivalent admittance modeling of the VSCs.

4.2. SISO-Based Stability Assessment

The SISO system in Figure 6b can be derived at the VSC terminals, where Ieq, i and
Yeq, i (s) represent the current source and the admittance of the DC grid equivalent circuit
observed from any VSC. The relation between the voltage Vi and the current sources is
expressed as
1
Vi = Zt (s)( Idc, i + Ieq, i ) ; Zt (s) = . (35)
Yeq, i (s) + Ydc, i (s)
Two approaches are used to assess the stability of the SISO closed-loop system in (35):
• Impedance-ratio stability criterion
Stability can be analyzed in the frequency domain by rewriting (35) as

Zeq, i (s)
Vi = (I + Ieq, i ) = ZT, i (s) IT, i . (36)
1 + Zeq, i (s)Ydc, i (s) dc, i

If and only if the impedance ratio Zeq, i (s)·Ydc, i (s) does not have a right-half-plane
(RHP) pole, the stability of the closed-loop system in (36) can be assessed by the Nyquist
stability criterion (denoted as SC in this paper), which evaluates the Nyquist curve of the
impedance ratio [4–6,9,11,15]. This criterion can only be applied if instability comes from
the interaction between the VSC and the DC grid equivalent circuit because the RHP pole
condition fails when system instability comes from the AC side of the VSC or the DC grid
equivalent circuit observed from the VSC.
• Positive-net-damping stability criterion
The positive-net-damping stability criterion (denoted as SC3 in this paper) evaluates
system transfer impedance damping in (36) at resonance frequencies, i.e., the value of
Re{ZT, i (jω)} at the frequencies of the peak values of ZT, i (jω) [4–9]. This criterion assesses
the closed-loop transfer function of the system, which avoids inaccuracies derived from
local instabilities in the stability conclusion.

5. Application
The stability in the VSC-HVDC system of Figure 7a was studied. Two converter
stations, VSC1 and VSC2 , were connected through a 1000 MW 440 kV HVDC transmission
grid. VSC1 operated in mode Vdc − Q, controlling the DC voltage of the HVDC transmission
grid, whereas VSC2 operated in mode P − Q, controlling the transmitted active power. Both
converters worked at unity power factor, typical of normal operating conditions [8,12–15].
According to Figure 7b, VSC1 is characterized by the DC-side equivalent admittance in (31),
while VSC2 is considered as constant power consumption P20 characterized by the fictive
resistance R20 = −(Vdc2 , 0 )2 /P20 [4,9]. This converter could also be characterized with the
DC-side equivalent admittance in (31), but a simpler model was used in the stability study
to focus the attention on VSC1 . VSC-HVDC system data are in Table 3. Three cases were
studied in the application:
• Case #1 (stable reference case): this corresponds to the data in Table 3, with the power
consumed by VSC2 being half the nominal power (i.e., P20 = 500 MW).
• Cases #2 and #3: the influence of the short-circuit ratio SCR and the VSC2 -consumed
power P20 , respectively, on system stability were analyzed.

45
Energies 2023, 16, 5457

Figure 7. HVDC application: (a) HVDC transmission system. (b) HVDC transmission grid with VSC
equivalent admittances. (c) Equivalent HVDC circuit.

Table 3. VSC-HVDC System Data (UB, AC = 220 kV, UB, DC = 440 kV, SB = 1000 MVA).

U0 (f 1 ), Xg /Rg 220 kV (50 Hz), 20 pu

AC main grid
SCR = Scc /(Pvsc1, N ) 15 pu
VSC AC filter Cf 6 MF
Rdc , Ldc 10.8 mΩ/km, 0.149 mH/km
DC cable
Cdc , Length 0.145 MF/km, 25 km
VSC DC filters Ccvi (i = 1 to 2) 30 MF
VSC1 Pvsc1, N , Vdc1, 0 1000 MW, 440 kV
fs , T d 2 kHz, 0.3 ms
Rc , Lc 0.025 pu, 0.25 pu
kp,cc , ki,cc 1.21 pu, 0.121 pu
VSC1 control
kp,pll , ki,pll 0.48 pu, 0.031pu
kp,dc , ki,dc 0.15 pu, 0.058 pu
kp,p , ki,p 0.15 pu, 0.058 pu

The system stability was verified by PSCAD/EMTDC time-domain simulations, and

the predictions obtained by the stability criteria SC1 , SC2 , and SC3 in Section 4 were
investigated. The results are shown in Figures 8 and 9.

Figure 8. Stability study: Case #2 (top) and Case #3 (bottom).

46
Energies 2023, 16, 5457

Figure 9. Impedance-based stability study.

Case #1: The PSCAD/EMTDC simulations before SCR and P20 changes in Figure 8
show the system stability. The results obtained from the stability criteria are as follows (see
Figure 9, left):
• SC1 : the GNC verifies system stability because the curves of the eigenvalues of L(s) do
not encircle the −1 point.
• SC2 : the Nyquist criterion verifies system stability because the Nyquist curve of
Zeq,i Ydc, i does not encircle the −1 point.
• SC3 : The positive-net-damping stability criterion verifies system instability because
the damping Re{ZT (jω)} at resonance frequencies fr ≈ 537.5, 712.3 and 810.3 Hz is
positive. The first resonance is due to the HVDC grid capacitors and inductors whereas
the second and third resonances are at the DC mirror resonance frequencies of the

47
Energies 2023, 16, 5457

AC-side resonance between the AC grid inductance Lg and the VSC1 filter capacitor
Cf at fr = 760 Hz. It is observed that the VSC control damps (but does not eliminate)
the AC-side resonance at the DC side. This means that AC-side dynamics are only
damped at the DC side by the VSC control and should therefore be considered in the
DC-side equivalent admittance modeling of the VSCs [12,13].
Case #2: The short-circuit ratio SCR increases from 15 pu to 30 pu at 0.1 s. The
PSCAD/EMTDC simulation in Figure 8 shows that the system becomes unstable after
the SCR value steps up. It was numerically verified that this is because the AC system
resonance frequency is shifted to a negative damping frequency range and the AC-side
voltage oscillations at f ≈ 980 Hz are amplified. The results obtained by the stability criteria
are as follows (see Figure 9, middle):
• SC1 : the GNC leads to an inaccurate stability assessment (i.e., the curves of the
eigenvalues of L(s) do not encircle the −1 point) because the instability comes from
the AC system resonance, causing the DC-admittance Ydc, 1 (s) in the diagonal of YE (s)
to be unstable.
• SC2 : the Nyquist criterion also leads to an inaccurate stability assessment because
Ydc, 1 (s) is unstable [13].
• SC3 : The positive-net-damping stability criterion verifies system stability (i.e., Re{ZT (jω)}
is negative at mirror resonance frequencies fr ≈ 935.5 and 1045 Hz). Closed-loop trans-
fer function assessment by the positive-net-damping stability criterion avoids the SC1
and SC2 stability inaccuracy derived from the instability of Ydc, 1 (s). If AC-side dynam-
ics are disregarded in the DC-side equivalent admittance model, AC-side resonances
are overlooked, which leads to inaccurate stability assessments [12,13].
Case #3: The VSC2 -consumed active power increases from P20 = 500 MW to the nomi-
nal value P20 = 1000 MW at 0.3 s. The PSCAD/EMTDC simulation in Figure 8 shows that
the system becomes unstable after the P20 value ramps up. It was numerically verified
that this is because the damping becomes negative at the DC resonance frequency, and
the DC-side voltage oscillations at f ≈ 535 Hz are amplified. The results obtained by the
stability criteria are as follows (see Figure 9, right):
• SC1 : the GNC verifies system instability because the curve of one eigenvalue intersects
the unit circle at approximately 527 Hz, enclosing the −1 point in a clockwise direction.
• SC2 : the Nyquist criterion leads to an inaccurate stability assessment because of
instability in Yeq, 1 (s).
• SC3 : The positive-net-damping stability criterion verifies system instability (i.e.,
Re{ZT (jω)} is negative at fr ≈ 537 Hz). The closed-loop transfer function assessment in
the positive-net-damping stability criterion avoids the SC2 stability inaccuracy derived
from the instability of Yeq, 1 (s).

6. Comparison of DC Admittance Models

The influence of the different simplifications that are usually carried out in DC admit-
tance models is shown in Table 4. The results obtained with the full model in Section 3 for
the three cases in Section 4 were compared with those obtained considering five different
simplifications. It can be observed that the simplification of the d-outer control loop and the
PLL do not affect the results of the full model, while this is not true for the simplifications
related to the inner control loop (i.e., the VSC delay and the bandwidth of the grid voltage
feedforward simplifications). Nevertheless, it must be noted that the above conclusions
correspond to the harmonic instabilities of the studied cases, and they cannot be extrapo-
lated to other types of instabilities. In general, harmonic instabilities are mainly affected by
the inner control loop parameters, while synchronous and subsynchronous instabilities are
affected by the outer control loops and the PLL.

48
Energies 2023, 16, 5457

Table 4. Comparison of simplifications.

Cases
Model Feature Case #1 Case #2 Case #3
Full model Stable Instable Instable
No phase-locked loop control Stable Instable Instable
No VSC delay Stable Stable Stable
No grid voltage feedforward (αf = 0) Stable Stable Instable
No grid voltage feedforward filter (αf = ∞) Stable Instable Instable
No d-outer control loop Stable Instable Instable

7. Conclusions
This paper presents a complete and detailed DC admittance model for VSC rectifier
and inverter stations of VSC-HVDC systems with active power, DC voltage, grid alternating
voltage, and reactive power controls. The model also considers the VSC delay, grid voltage
feedforward filter, PLL, and several outer control loops. It is derived from dq-complex space
vectors commonly used in circuit theory. A comparison between the different models in the
literature and a study of the impact on the instability of the different model simplifications
were carried out to illustrate the contribution of the paper. The proposed model was applied
to discuss VSC-HVDC system stability and the inaccuracies of impedance-based stability
criteria due to system partitions. It is concluded that the stability assessment of VSC-HVDC
systems from VSC and DC grid equivalent circuits could lead to inaccurate conclusions by
the Nyquist criterion (e.g., when instability comes from AC-side dynamics) whereas the
positive-net-damping criterion overcomes this issue because it is a global stability criterion
which evaluates system transfer impedance damping at resonance frequencies.

Author Contributions: Conceptualization and methodology, J.P. and L.S.; software development and
validation, L.M. and J.P.; writing—original draft preparation, J.P. and L.S.; visualization, L.M. All
authors have read and agreed to the published version of the manuscript.
Funding: This research was funded by the MCIN (Proyectos de Generación de Conocimiento) under
grant number PID2021-123633OB-C33 (J-02895).
Data Availability Statement: Data sharing not applicable.
Conflicts of Interest: The authors declare no conflict of interest.

Appendix A. Outer Loop Controls

The VSC control is characterized by the current controller (CC) and the outer control
loops, which make it possible to characterize the main VSC operation modes (i.e., mode P
− Q, mode Vdc − Q, mode Vdc − Eac , and mode P − Eac ) [8].
Vdc − Q control: the direct voltage control (DVC) and reactive power control (QPC)
of Figure 1 are expressed with dq-complex space vectors [20] using the notation of (23) as
     
1 Fdc (s) ± Fq (s) Ed0 1 −1 ± Fq (s) −I*dq0 Idq0
Fm = √ ; Go, = ; Y = √ , (A1)
2 Fdc (s) −1 1 I*dq0 −Idq0
o, UQ UQ o, UQ
2 2
where Fdc (s) = kp, dc + ki, dc /s, with kp, dc and ki, dc being the proportional and integral gains
of the DVC, and Fp (s) = kp, q + ki, q /s, with kp, q , and ki, q being the proportional and integral
gains of the QPC.
P − Q control: the alternating power control (APC) and reactive power control of
Figure 1 are expressed with dq-complex space vectors [20] using the notation of (23) as
   
± 1 0 ±
√ 0 Idq0
Go, PQ = Fp (s) Ed0 ; Yo, PQ = 2Fp (s) * , (A2)
0 1 Idq0 0

49
Energies 2023, 16, 5457

where Fp (s) = kp, p + ki, p /s, with kp, p = kp, q and ki, p = ki, q being the proportional and integral
gains of the APC.
Vdc − Eac control: the direct voltage control (DVC) and alternating voltage control
(AVC) control law of Figure 1 are expressed with dq-complex space vectors using the
notation of (23) as
   
1 Fdc (s) ± Fv (s) j j
Fmo, UE = √ ; Yo, = , (A3)
2 Fdc (s) −j −j
UE 2

where Fv (s) = kp, v + ki, v /s, with kp, v and ki, v being the proportional and integral gains of
the AVC.
P − Eac control: the alternating power control (APC) and alternating voltage control
(AVC) control law of Figure 1 are expressed with dq-complex space vectors using the
notation of (23) as
     
± Fp (s) Ed0 1 1 ± Fp (s) I*dq0 Idq0 Fv (s) j j
Go, PE = ; Yo, PE = √ + . (A4)
2 1 1 *
2 Idq0 Idq0 2 −j −j

References
1. Sun, J.; Li, M.; Zhang, Z.; Xu, T.; He, J.; Wang, H.; Li, G. Renewable energy transmission by HVDC across the continent: System
challenges and opportunities. CSEE J. Power Energy Syst. 2017, 3, 353–364. [CrossRef]
2. Zhang, C.; Molinas, M.; Rygg, A.; Cai, X. Impedance-Based Analysis of Interconnected Power Electronics Systems: Impedance
Network Modeling and Comparative Studies of Stability Criteria. IEEE J. Emerg. Sel. Top. Power Electron. 2020, 8, 2520–2533.
[CrossRef]
3. Liu, H.; Xie, X.; Liu, W. An Oscillatory Stability Criterion Based on the Unified dq-Frame Impedance Network Model for Power
Systems with High-Penetration Renewables. IEEE Trans. Power Syst. 2018, 33, 3472–3485. [CrossRef]
4. Stamatiou, G.; Bongiorno, M. Stability Analysis of Two-Terminal VSC-HVDC Systems Using Net-Damping Criterion. IEEE Trans.
Power Deliv. 2016, 31, 1748–1756. [CrossRef]
5. Sainz, L.; Cheah-Mane, M.; Monjo, L.; Liang, J.; Gomis-Bellmunt, O. Positive-net-damping Stability Criterion in Grid-connected
VSC Systems. IEEE J. Emerg. Sel. Top. Power Electron. 2017, 4, 1499–1512. [CrossRef]
6. Harnefors, L. Modeling of Three-Phase Dynamic Systems Using Complex Transfer Functions and Transfer Matrices. IEEE Trans.
Ind. Electron. 2007, 54, 2239–2248. [CrossRef]
7. Harnefors, L. Proof and Application of the Positive-net-damping Stability Criterion. IEEE Trans. Power Syst. 2011, 26, 481–482.
[CrossRef]
8. Harnefors, L.; Bongiorno, M.; Lundberg, S. Input-admittance Calculation and Shaping for Controlled Voltage-source converters.
IEEE Trans. Ind. Electron. 2007, 54, 3323–3334. [CrossRef]
9. Wen, B.; Boroyevich, D.; Burgos, R.; Mattavelli, P.; Shen, Z. Analysis of D-Q Small-Signal Impedance of Grid-Tied Inverters. IEEE
Trans. Power Electron. 2016, 31, 675–687. [CrossRef]
10. Pinares, G.; Bongiorno, M. Analysis and Mitigation of Instabilities Originated from DC-side Resonances in VSC-HVDC Systems.
IEEE Trans. Ind. Appl. 2016, 52, 2807–2815. [CrossRef]
11. Xu, L.; Fan, L.; Miao, Z. DC Impedance-model Based Resonance Analysis of a VSC-HVDC System. IEEE Trans. Power Deliv. 2015,
30, 1221–1230. [CrossRef]
12. Shah, S.; Parsa, L. Impedance Modeling of Three-phase Voltage Source Converters in dq, Sequence and Phasor Domains. IEEE
Trans. Energy Convers. 2017, 32, 1139–1150. [CrossRef]
13. Amin, M.; Molinas, M. Small-signal Stability Assessment of Power Electronics Based Power Systems: A Discussion of Impedance-
and Eigenvalue-based Methods. IEEE Trans. Ind. Appl. 2017, 53, 5014–5030. [CrossRef]
14. Amin, M.; Molinas, M.; Lyu, J.; Cai, X. Impact of Power Flow Direction on the Stability of VSC-HVDC Seen from the Impedance
Nyquist Plot. IEEE Trans. Power Electron. 2017, 32, 8204–8217. [CrossRef]
15. Liu, J.; Tao, X.; Yu, M.; Xia, Y.; Wei, W. Impedance Modeling and Analysis of Three-Phase Voltage-Source Converters Viewing
from DC Side. IEEE J. Emerg. Sel. Top. Power Electron. 2020, 8, 3906–3916. [CrossRef]
16. Rygg, A.; Molinas, M.; Zhang, C.; Cai, X. A Modified Sequence-domain Impedance Definition and its Equivalence to the
dq-domain Impedance Definition for the Stability Analysis of AC Power Electronic Systems. IEEE J. Emerg. Sel. Top. Power
Electron. 2016, 4, 1383–1396. [CrossRef]
17. Wang, X.; Harnefors, L.; Blaabjerg, F. Unified Impedance Model of Grid-connected Voltage-source Converters. IEEE Trans. Power
Electron. 2018, 33, 1775–1785. [CrossRef]
18. Paap, G.C. Symmetrical Components in the Time Domain and their Application to Power Network Calculations. IEEE Trans.
Power Syst. 2000, 15, 522–528. [CrossRef]

50
Energies 2023, 16, 5457

19. Harnefors, L.; Wang, X.; Chou, S.; Bongiorno, M.; Hinkkanen, M.; Routimo, M. Asymmetric Complex-Vector Models With
Application to VSC–Grid Interaction. IEEE J. Emerg. Sel. Top. Power Electron. 2020, 8, 1911–1921. [CrossRef]
20. Pedra, J.; Sainz, L.; Monjo, L. Three-port small-signal admittance-based model of VSC for studies of multi-terminal HVDC hybrid
AC/DC transmission grids. IEEE Trans. Power Syst. 2021, 36, 732–743. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

51
 energies
Article
Actions to Handle Increasing Distributed Generation and
Rising Voltage during Low-Demand Periods: Tap Staggering
Practices and Field Tests in the Italian Transmission Network
Stefano Quaia 1, *, Alessandro Mauri 1 , Alessio Marchesin 2 and Paolo Manià 3

1 Department of Engineering and Architecture, University of Trieste, Via Valerio 10, 34127 Trieste, Italy;
alessandro.mauri@phd.units.it
2 Terna Rete Italia S.p.A., Via Botticelli 139, 10154 Torino, Italy; alessio.marchesin@terna.it
3 AcegasApsAmga S.p.A., Via del Teatro 5, 34121 Trieste, Italy; pmania@acegasapsamga.it
* Correspondence: quaias@units.it

Abstract: This paper deals with the voltage regulation strategies implemented in the Italian transmis-
sion network to face the increasing problem of high voltages during time periods characterized by
low demand. After an introduction in which this very actual problem is discussed, the focus is on tap
staggering practices. Although tap staggering is not a new idea, it is gaining practical importance
only in the very last few years as a means of enhancing the inductive power drawn from the grid
and, therefore, limiting the voltage rise. Accordingly, tap staggering contributes to the mitigation of
the problems caused by the increasing penetration of renewable energy sources and thus can allow
an increase in the share of renewable energy sources. The paper presents the different tap staggering
practices that are being defined by the Italian transmission system operator and reports some tap
staggering tests recently performed on large autotransformers as well as phase-shifter transformers.

Keywords: voltage regulation; transformer tap staggering; autotransformers; phase-shifting transformers

Citation: Quaia, S.; Mauri, A.;

Marchesin, A.; Manià, P. Actions to 1. Introduction
Handle Increasing Distributed
In the last decade, an increasing upward trend in voltages in some parts of the transmis-
Generation and Rising Voltage
sion systems has been taking place, mainly in low-demand periods like nights, weekends,
during Low-Demand Periods: Tap
and holidays [1–3]. Two important factors are behind this trend. First, in operating condi-
Staggering Practices and Field Tests
in the Italian Transmission Network.
tions characterized by low demand and high production from nonprogrammable renewable
Energies 2023, 16, 6258. https://
sources, the committed sets of production units (PUs) with regulation capability are partic-
doi.org/10.3390/en16176258
ularly small. Second, the reactive energy drawn/introduced from/into the transmission
systems is changing, with a clear progressive reduction in the total annual reactive energy
Academic Editor: Ying-Yi Hong
drawn (“inductive energy”) and a clear progressive increase in the total annual reactive
Received: 13 July 2023 energy introduced (“capacitive energy”). In Italy, the total amount of capacitive energy
Revised: 17 August 2023 introduced in HV networks roughly doubled in the period 2014–2020, approaching the
Accepted: 24 August 2023 total amount of inductive energy drawn [3]. Table 1 shows that, in this period, the total
Published: 28 August 2023 amount of reactive energy exchanged between HV networks and the sum of distribution
networks and HV users has remained approximately constant at nearly 50 TVArh/year,
but there has been a clear progressive decrease in inductive energy and a clear progressive
increase in capacitive energy [4]. In 2020, the latter almost matched the former. In the
Copyright: © 2023 by the authors. figures reported, distribution systems are by far the main part.
Licensee MDPI, Basel, Switzerland. Transmission system operators (TSOs) control voltages and manage reactive energy in
This article is an open access article
their HV networks by means of proper devices/techniques (which include synchronous
distributed under the terms and
compensators, capacitor banks, shunt reactors, STATCOMs, and transformer tap staggering)
conditions of the Creative Commons
and stocking up resources in the dispatching services market (MSD). In order to meet
Attribution (CC BY) license (https://
voltage regulation requirements, the current upward trend in voltages in some parts of the
creativecommons.org/licenses/by/
transmission systems forces TSOs to commit additional PU compared to those committed
4.0/).

Energies 2023, 16, 6258. https://doi.org/10.3390/en16176258 52 https://www.mdpi.com/journal/energies

Energies 2023, 16, 6258

as a result of the electricity markets. On the one hand, this leads to increased dispatching
costs and, on the other hand, constitutes an obstacle to the increasing penetration of
nonprogrammable renewable sources required by the ongoing energy transition process.
This latter aspect is made even more crucial and urgent by the recent tensions in energy
prices and dramatic geopolitical events.

Table 1. Total amounts of reactive energy exchanged between the HV grids and distribution sys-
tems/HV users.

Total Reactive
Inductive Energy Capacitive Energy
Year Energy Exchanged
[TVArh/year] [TVArh/year]
[TVArh/year]
2014 36.1 13.4 49.5
2018 32.5 17.9 50.4
2019 29.9 20.3 50.4
2020 25.3 23.1 48.4

As for dispatching costs, in 2014, the cost incurred by Terna, the Italian TSO, on the
MSD for reactive energy management was estimated at approximately 150 M€ [5]. But, in
the two years 2018–2019, this cost rose to more than 1000 M€/year [4].
In addition, the increasing voltages in some parts of the transmission system have
caused a significant increase in TSO investments for devices for voltage regulation and
reactive energy control (in the period 2020–2026, the total investment costs, made and
planned, are estimated at approximately 1 G€ [4]).
As for nonprogrammable renewable sources, their increasing penetration has im-
portant effects on the generators in service, as they lead to small sets of regulating PUs
committed as a result of the electricity markets and, consequently, to the need to commit
additional PUs in order to provide adequate system security and network services (like
voltage regulation).
Aiming to allow an increasing share of nonprogrammable renewable generation, Terna
has started, since the second half of 2021, a revision of its real-time management processes.
In particular, special effort has been focused on reducing the commitment of PUs not
resulting from the electricity markets but required to provide adequate system security
and services.
The TSO has several means to counteract the rising trend in voltage values during
low-demand periods. In detail, the actions based on already existing network components
are listed below. Note that the actions performed on compensation devices generally have
higher priority than configuration changes or power flow changes. For example, shunt
reactor management has higher priority than PST control, tap change on autotransformers
(ATRs), disconnection of ATRs, disconnection of lines, and so on.
• Synchronous generator excitation control on the primary transmission network (which
includes the 380–220 kV voltage levels);
• Shunt reactor management;
• Active power flow control on the primary kV transmission network by means of
phase-shifting transformers (PSTs) and set point changes in HVDC links;
• Changes in HVDC filter arrangement;
• Network configuration changes by acting on parallel busbar switch breakers;
• Control of ATRs on-load tap changers (OLTCs) and disconnection of ATRs;
• Distributed tap staggering (TS) of ATRs with limited tap unbalance (see Section 3)
and/or localized TS of ATRs (and, when available, of PSTs).
In addition, the following actions can also be performed, but note that they are costly
for the system:
• Active power flow changes by means of PU re-dispatching;
• Early PU commitment and/or delayed PU decommitment;

53
Energies 2023, 16, 6258

• Commitment of PUs not programmed as a result of the electricity markets.

In agreement with what is reported above, the target of Terna is to minimize the
impact of the three latter actions while trying to maximize the benefit obtained through the
implementation of the previous “not costly” actions. The benefit consists of a higher share
of nonprogrammable renewable source generation allowed, reduced dispatching costs,
or both.
From the viewpoint of organization, different actions have been put in place:
• A push/acceleration for the installation of reactive power compensators like shunt
reactors, synchronous compensators, and STATCOMs;
• Close coordination between the National and Regional Control Centers for manag-
ing the primary transmission and subtransmission networks with a view to voltage
regulation in the overall system;
• Creation of a list of maneuvers on the network intended to limit system voltages.
For example, the list indicates the lines that can be disconnected, the effect of each
disconnection, and information concerning the possibility of performing TS;
• Creation of teams of experts supporting 24 h the control rooms in the decisions
concerning the acquisition of PUs not committed as a result of the electricity markets.
Although TS is a means for reactive power management that has been known for a
long time, it has been gaining increasing importance only in the very last few years to help
counteract the actual problem of high voltages during periods of low demand. As TS does
not require the connection of new components to the power system, it is cheaper than
other solutions like the installation of new shunt reactors (in Italy, the investment costs
for the installation of new shunt reactors in the 4 years 2021–2024 will be approximately
48 M€ [4]). Moreover, operative field data suggest that the reliability of shunt reactors
could be lower than that of TRSs, ATRs, and PSTs involved in TS. However, TS stresses the
involved machines more, and it is still early to evaluate any reliability reductions.
The following sections specifically concern TS and describe the application rules
that are being defined by the Italian TSO. Section 2 briefly reports the principles of TS
performed on TRSs/ATRs and PSTs and includes a short comparison among different
available techniques for reactive power management.
Section 3 analyzes the different modalities adopted by the Italian TSO for TS implemen-
tation/management. Section 4 reports the results of recent field tests performed on both
ATRs and PSTs in the Italian primary transmission network. Sections 3 and 4 present some
original rules/solutions for what concerns the TS performed using PSTs and distributed or
localized TS.

2. Principles of Tap Staggering

The theory of TS is briefly described in this section. TS consists of operating in
parallel two tap-change transformers (TRSs) or ATRs on a dedicated secondary busbar with
different tap positions [6–8]. This configuration provides a means of absorbing reactive
power and thus contributes to regulating voltage in the substation area in low-load periods.
The reactive power absorbed at the input node and therefore the voltage variation on the
primary side depends on its short-circuit power and steady-state operating conditions
before TS implementation.
Consider two tap-change TRSs connected in parallel. The number of turns of the
secondary windings depends on the selected taps. Assume the upper TRS has (1 + n a )
turns, and the bottom TRS has (1 + nb ) turns. By neglecting the no-load currents and the
series (short-circuit) resistances, this arrangement is represented by the circuit reported in
Figure 1 [9]. This circuit can be transformed in the “equivalent transformer” illustrated in
Figure 2.

54
Energies 2023, 16, 6258

Figure 1. Electrical scheme of two tap-change transformers connected in parallel.

Figure 2. Equivalent transformer.

The parameters of the resulting equivalent transformer (i.e., series reactance Xeq , per
unit tap-change setting neq and transversal admittance Yeq ) can be expressed as follows
as a function of the per unit tap-change settings n a and nb and the series (or short circuit)
reactances Xa and Xb :
 
( n b − n a )2
Yeq = − j (1)
X a (1 + n a )2 + Xb (1 + n b )2

X a (1 + n a )2 ( n a − n b )
neq = n a − (2)
X a (1 + n a )2 + Xb (1 + n b )2
 
X a Xb X a (1 + n a )2 + Xb (1 + n b )2
Xeq =  2 (3)
X a (1 + n a )2 + Xb (1 + n b )2

Equation (1) shows that unequal tap-change settings (n a = nb ) give rise to an equiva-
lent transversal reactive admittance, with reactive power absorption from the upstream
power system. Therefore, this arrangement can be used to improve voltage regulation dur-
ing light-load conditions. If TS is performed with TRSs in no-load operation (see Section 3),
according to Figure 2, the reactive power absorbed is only due to the resulting transversal
admittance Yeq .
Note that, in principle, this action has the same qualitative effect of other devices
capable of absorbing reactive power, like shunt reactors, TCRs (i.e., continuously controlled

55
Energies 2023, 16, 6258

reactors [10]), and synchronous compensators. However, the costs are very different. Unlike
the existing TRSs or ATRs of an electrical substation, shunt reactors, TCRs, and synchronous
compensators must be regarded as new devices to be installed by the TSO. Accordingly,
if TRSs or ATRs can be operated, during light-load conditions, in TS configuration (this
possibility must be verified case-by-case by the TSO), this practice can allow the installation
of new devices to be avoided.
Aiming to compare these different techniques, first we note that the high and very
high dynamic responses of respectively synchronous compensators and TCRs are not
required to face the problem of rising voltages in low-demand periods. As is known, in
most cases, these techniques are used in order to obtain other technical benefits: system
inertia and short-circuit power increase (synchronous compensators), fast dynamic voltage
regulation, and/or system stability increase (TCRs). It follows that, in practice, the most
significant comparison can be performed between transformer TS and shunt reactors. Shunt
reactors are relatively simple but allow just an on/off operation (reactive power cannot be
regulated). On the other hand, their connection/disconnection is not subject to the network
checks required by TS (see Section 3). Table 2 briefly summarizes the main differences
concerning both costs and technical aspects of these four techniques.

Table 2. Comparison among different techniques capable of facing the problem of rising voltage
during low-demand periods.

Technique Cost Main Advantages Main Disadvantages

Connection of new devices
Operation must be
Transformer Tap is not needed;
Low verified case-by-case by
Staggering several available machines
the TSO.
can be used.
Relatively simple;
operation is largely
Shunt reactors High Only on/off operation.
independent of
network conditions.
Continuous reactive High dynamics are not
TCR Very high power control; needed for this voltage
high dynamics. regulation problem.
Continuous reactive
Inertia is not involved
Synchronous power control;
Very high in this voltage
compensators increase in system inertia
regulation problem.
and short-circuit power.

Aiming to increase the circulation current between the two TRSs and thus the reactive
power absorbed from the upstream network, the taps selected on the two OLTCs must be
changed in opposite directions. Assuming that:
• The same OLTC tap-change settings (with opposite sign) are used for the two TRSs:
n a = Δn and nb = −Δn;
• The series reactances Xa and Xb are equal (in practice, often the two involved TRSs/
ATRs are twins and the limited difference between the two reactances due to the
different number of turns n a and nb can be neglected).
From Equation (1), denoting by X the series reactances Xa and Xb , the resulting
transversal admittance Yeq is:

. 4Δn2 2Δn2
Y eq ∼
= −j ∼
= −j (4)
2X (1 + Δn2 ) X

More generally, given the nominal data of the involved machines and the tap position
settings, the reactive power absorbed in TS operation can be estimated.
The typical values of the ATRs used in Italian HV substations are as follows: nominal
power A is 250 MVA, short-circuit impedance Xsc goes from 11% to 12%, and OLTC ±10%

56
Energies 2023, 16, 6258

with ±5 taps and 11 total positions. According to Equation (4), the maximum reactive
power that two typical ATRs can absorb in a no-load configuration is approximately
40 MVAr.
TS can also be performed using two TRSs under normal load. In this case, the currents
caused by the TS add to the load currents, and therefore, the two TRSs carry different
currents. It follows that the maximum allowed unbalance in the selected taps is generally
lower than the maximum one. At the rated current, the reactive power that can be absorbed
by two TRSs operated in TS in under-load mode depends on (1) the load level and (2) the
power factor. Obviously, the reactive power that can be absorbed is lower than that in the
no-load mode. In addition, the secondary voltage of the two TRSs in parallel is subject to
more stringent constraints.
In principle, TS can also be performed by exploiting the normal layout of the primary
substations (owned by the distribution system operators, DSO), which include two HV/MV
TRSs in parallel [7,11,12]. However, in this case, the benefits mainly concern the upstream
TSO HV network, whereas TS is performed by DSO machines.
In practice, until today, Italian DSOs have not performed TS on their TRSs. However,
as already mentioned in Section 1, DSOs are responsible for most of the capacitive energy
introduced in HV networks. Therefore, for DSOs (and HV users), recent tariff regulation
concerning the reactive energy exchanged with HV networks, in force since 1 April 2023,
states that capacitive energy cannot be introduced into HV networks in any time slot and
sets the unit fees to be applied in case of noncompliance with this prescription [13]. In this
regard, DSOs are developing network models to predict grid behavior. In addition, they
are utilizing the provisions of the Italian compulsory standards for connections to improve
their capability of managing reactive energy.
A general assumption for Terna is that the available assets already installed on the grid
and able to provide multi-service performance should do so. This idea can also be applied
to the case of PSTs. It is known that these machines are used to control the active power
flow on the electric transmission lines they are connected to, by acting on the voltage phase
angle difference between the two nodes at the extremities of the lines. However, exploiting
a proper arrangement, a PST can also be used to absorb reactive power, and therefore, in
“TS mode” [9].
The TS mode is obtained by closing the PST bypass circuit (see the field test described
in Section 4.2 below). Accordingly, unlike TS performed with two TRSs/ATRs, one PST is
sufficient. The current absorbed depends on the equivalent reactances of the PST, which
depend in turn on the selected tap. However, PSTs are generally connected on interconnec-
tion lines between neighboring countries or on important transmission lines. Therefore, the
possibility of operating a PST in TS mode must be verified case-by-case, since in this case,
the machine is subtracted to its primary task of active power flow regulation.
For a PST with a symmetrical structure (as in the case of the field test illustrated in
Section 4.2), the input and output voltages have equal magnitudes, and the voltage differ-
ence between the input and output, ΔV, depends on the selected tap (which determines
the phase shift angle introduced, α) and on the input voltage magnitude V:

ΔV = 2V sin(α/2) (5)

The TS mode is obtained by closing the PST bypass circuit. The current absorbed
depends on the equivalent reactances of the PST, which depend in turn on the selected tap.
For a two magnetic core machine (depicted in Figure 3), as the involved PST is in the
field test reported below, reference [14] provides the equivalent shunt and series reactances
as a function of the phase shift angle, α.

57
Energies 2023, 16, 6258

Figure 3. Circuit diagram of a two magnetic core symmetrical PST.

The shunt reactance Xshunt (α) is defined as the equivalent reactance, which crossed by
the series input current I of the PST produces the reactive losses of the shunt (exciting) unit.
The series reactance Xseries (α) is defined as the equivalent reactance, which crossed by the
same series input current I produces the reactive losses of the series unit. The equations
are as follows: !  2 "
  α 2 n3
Xshunt (α) = 4 sin X3 + X4max (6)
2 n4max
 2  2   α 2
n1 n1
Xseries (α) = X1 + X2 − X2 sin (7)
n2 n2 2
where:
• n1 (fixed) is the number of turns of the series windings (series unit);
• n2 (fixed) is the number of turns of the delta-connected “booster” windings
(series unit);
• n3 (fixed) is the number of turns of the star-connected exciting windings
(exciting unit);
• n4 (variable and selected by the OLTC) is the number of turns of the star-connected
regulation windings (exciting unit).
The total equivalent reactance of the PST X (α), which crossed by the series input
current I produces the total PST reactive power absorption, is then the sum of Xshunt (α)
and Xseries (α):
!  2 " !  2   "
n1   α 2 n3 n1 2
X ( α ) = X1 + X2 + 4 sin X3 + X4max − X2 (8)
n2 2 n4max 2n2

where each reactance Xi refers to the i-th winding made up of ni turns.

Finally, the total reactive power absorbed Q is:

Q = 3 X (α) I 2 (9)

Using these models, the reactive power that can be absorbed by any given PST can be
theoretically calculated.

58
Energies 2023, 16, 6258

3. Tap Staggering Application Rules of the Italian TSO

As already reported, TS is one of the actions that can be used to counteract rising
voltage values in low-demand time periods. TS can be either programmed to define a time
for its actuation or implemented in real time. In the Italian primary transmission network,
TS can be performed using the following:
• Two ATRs connected in the same electrical substation;
• One PST (using a PST in “TS mode” requires giving up the active power control of the
PST: if this is acceptable, the machine can be used during a limited time interval for
TS purposes. The TS test with PST described below exploits a (not frequent) situation
where two PSTs are installed in the same substation).
From a power system perspective, TS can be performed involving a higher or lower
number of nearby electrical substations. The former case is referred to as “distributed” TS.
It is performed using ATRs with a limited tap unbalance, typically not exceeding 50% of the
maximum unbalance allowed by OLTCs. The latter case is referred to as “localized” TS and
can be performed using ATRs but also PSTs (when available), setting the tap unbalance to
the maximum (provided that the rules listed below are satisfied). Of course, distributed TS
causes less stress to the involved machines but, on the other hand, involves more machines
than localized TS. In all cases, once the critical conditions time is over, best practice says to
stop TS in order to limit wear on the involved machines.
As already reported in Section 2, TS performed with two ATRs of a substation can
be performed either at no-load (more effective case) or under normal-load conditions.
Performing TS at no-load requires the real-time control rooms to check the possibility of
excluding two ATRs from their normal operation. If the network conditions do not allow
this exclusion, TS can be performed with the machines under load.

3.1. Tap Staggering with ATRs in No-Load Operation

In this operative mode, the two ATRs run at no-load on a dedicated (insulated) busbar.
This arrangement is possible when:
• The load supplied by the underlying local network allows the exclusion from service
of two ATRs in the considered substation;
• The ATRs in normal operation (i.e., those not dedicated to TS) and/or the actual
network conditions allow the safe operation of the underlying network;
• In TS operation, the following application rules are currently established:
• The real-time room controls that the currents circulating in the two machines remain
within their limits;
• In the case that one ATR opens, the secondary of the other (still in service) ATR must
be immediately opened as well.
As to the stress caused by TS in the involved machines, Terna is performing charac-
terization campaigns to define the usage limits of OLTCs as a function of ATR secondary
voltage. At present, for not-yet characterized families of devices, conservative limits are
used (i.e., 50% of the OLTC regulation range).

3.2. Tap Staggering with Normally Loaded ATRs

When the operating conditions of the network and/or the number of ATRs available
in the area do not allow for the dedication of two ATRs for TS in no-load operation, TS in
the involved substation can be performed using two ATRs in normal under-load operation
with the secondary sides connected to the same busbar. The two ATRs are connected in
parallel, and TS is performed, changing in opposite directions the operating taps of the
two OLTCs.
The current caused by the unbalance in OLTCs adds vectorially to the load currents
and thus different currents are obtained in the two ATRs. As the taps are changed, one
ATR takes more load, and the other ATR takes less load. This implies that the unbalance
in OLTC positions must be limited and, consequently, a lower reactive power can be ab-

59
Energies 2023, 16, 6258

sorbed compared with the maximum reactive power absorbed in the no-load mode [12].
Also, in the under-load mode, the currents circulating in the two machines must be
continuously controlled.
If one ATR is opened during the operation in TS, the OLTC of the other ATR must
be adjusted in order to keep the secondary voltage at a proper value within normal op-
eration limits. A correct secondary voltage will be restored in the times required by
OLTC maneuvering.
As explained in the previous paragraph with regard to the no-load mode, also in
the under-load mode, the control room controls that the currents circulating in the two
machines remain within their limits.

4. Field Tests
4.1. Tap Staggering with ATRs in No-Load Operation
This paragraph briefly reports a TS field test with ATRs in no-load operation recently
performed in the 380 kV substation Parma Vigheffio, in Northern Italy. The two ATRs
involved are both rated 250 MVA and have the following data: short-circuit impedance
Xsc = 0.12 and OLTC ±10% with ±5 taps.
As a first step (at approximately 11:30 AM in Figure 4), the two ATRs were set in
parallel with the secondary side closed on the isolated 132 kV busbar dedicated to the test.
Of course, this action resulted in the cancellation of the previous active power flow, as
indicated by the dotted curve in Figure 4.

Figure 4. Active and reactive power of two ATRs in TS.

Next, the two transformers were operated at different (not symmetrical) tap positions:
+5 taps for one ATR (which means Δn = 0.1 p.u.) and −2 taps for the second ATR (which
means Δn = −0.04 p.u.). The reactive power measurements show opposite flows on the
two transformers, with a total balance of approximately 22 MVAr absorbed at the network
connection point. The reactive power absorbed is actually approximately 8% of the rated
power of the machines, as expected on the basis of the theoretical models reported in
Section 2. This result confirms experimentally very well what was foreseen analytically.
Overall, the TS test lasted approximately 8 h.

4.2. Tap Staggering with PST

This paragraph reports a TS field test performed with a PST in the 380 kV substation
Foggia, in Southern Italy. Under the supervision of the TSO National Control Center, the
test started after checking the network operative conditions and power flows involved.
Figure 5 shows the operational configuration adopted for the test in the involved
substation in which two PSTs are installed.

60
Energies 2023, 16, 6258

Figure 5. Substation layout, where PST A is in TS and PST B is in pushing mode (C and O stand,
respectively, for close and open).

PST A is the machine under test. TS operation is obtained by closing the bypass
disconnector and selecting a proper tap (angle shift) (in order to allow PST A operation
with the bypass closed, some block signals of the PST automatic control must be turned
off). Meanwhile, PST B is kept in normal operation, pushing active power on the 380 kV
line Foggia-Troia, on which it is connected.
This configuration (evidenced by the thick lines in Figure 4) allowed the normal oper-
ation of the 380 kV line Foggia-Troia, on which the active power flow was approximately
400 MW.
The main technical data of the two identical PSTs are as follows: symmetrical structure
with two magnetic cores (see Figure 3), buck and boost regulation, throughput rating PT is
1800 MVA, nominal rating of the two units PD is 543.2 MVA, 33 taps, maximum phase shift
at no-load α0 is 17.5◦ .
The theoretical analysis based on the considerations reported in Section 2 provides a
maximum expected reactive power absorption of approximately 150 MVAr for the PST in
TS operation. The test had the following main targets:
• To check the maximum reactive power absorption estimated (approximately
150 MVAr);
• To check the effect produced on the voltages in the area.
We briefly report the main results of the test. The first result, shown in Figure 6, regards
the maximum reactive power absorbed, which was approximately 140 MVAr, broadly in
line with the expected theoretical value.
As to the effect of TS on voltage regulation, the reactive power absorbed during the test
led to a voltage reduction of approximately 2.5 kV at the Foggia 380 kV node, as Figure 7
illustrates. The other voltages shown in Figure 7 are those recorded at two neighboring
nodes of the 380 kV network.

61
Energies 2023, 16, 6258

Figure 6. Reactive power absorbed during the TS test (changing with the selected taps).

Figure 7. Voltages measured at the 380 kV nodes of Foggia, Deliceto, and Rotello.

For reasons of safe operation during the TS test, the automatic voltage regulations
(AVRs) of the nearby rotating PUs were not suppressed/modified. The AVR operation
reduces the TS effect on the voltages in the area (in other words, the TS effect is weak-
ened). Measurements performed during the test showed that the voltage reduction caused
an overall reactive power change of approximately 75 MVAr coming from the four syn-
chronous generators in operation in the area (see Figure 8). This reactive power amount
would have resulted in a further voltage reduction at the Foggia 380 kV substation that
could be evaluated at roughly 1–1.5 kV, for a total of almost 4 kV (approximately 1% of the
nominal voltage). Figure 8 clearly shows the reduction in the reactive power absorbed by
the generators in correspondence with the peak of the reactive power absorbed by the PST
under the test.

Figure 8. Reactive power absorbed by four rotating PUs in the area of the test.

62
Energies 2023, 16, 6258

5. Conclusions
In the very last few years, TS has been gaining practical importance as a possible
action aimed at counteracting the rising voltage trend that occurs in low-demand time
periods in some areas of HV networks.
This paper describes the TS application rules that are being defined by the Italian TSO
and reports some recent field tests performed on large ATRs and PSTs. Considering the
very limited literature available today on this subject, these contents can be considered, at
least to a certain extent, original.
TS can be performed using couples of ATRs located in the same substation, either in
no-load operation or in normal under-load operation. The former case provides the most
effective results, but this operative mode is not always allowed, as it requires the exclusion
of two ATRs from their normal operation.
A further possibility is given by TS performed with PSTs. Even though ATRs are far
more widespread than PSTs, PSTs usually have a larger rated power and regulation range
and can absorb large reactive power in TS operation.
Finally, this paper reports the main results of recent TS field tests performed by the
Italian TSO using both ATRs and PSTs. The test results confirm the effectiveness of TS and
the correctness of the theoretical models used.
From an overall power system perspective, an important technical advantage con-
nected with TS implementation is an increase in the nonprogrammable renewable gen-
eration allowed without violation of voltage constraints. Of course, this advantage can
also be obtained by installing various types of voltage-controlling devices (for example,
shunt reactors). A first qualitative comparison between TS and shunt reactors can be
performed, considering both the costs and reliability of the two solutions. As for the
costs, the comparison is clearly in favor of TS. Concerning reliability, some faults that
happened on shunt reactors in Italy could lead to the same conclusion, even though more
time is still needed to obtain the necessary operative experience for TS. A further ad-
vantage of TS is the possibility of controlling, step-by-step, the reactive power absorbed,
whereas shunt reactors usually allow just on–off operation. On the other hand, shunt reac-
tors do not require careful network checks for the connection/disconnection required by
transformer TS.

Author Contributions: All the authors contributed to design the research approach. S.Q. and A.M.
(Alessio Marchesin) dealt with conceptualization and methodology; S.Q. and A.M. (Alessandro
Mauri) wrote and edited the original draft, including formal analysis and graphing; P.M. and A.M.
(Alessio Marchesin) took care of resources and data. All the authors revised the manuscript. All
authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Data Availability Statement: Additional data and measurements not explicitly provided in the
manuscript may be subject to censorship by Terna S.p.A. as they are of a sensitive nature.
Acknowledgments: The authors acknowledge Davide Rampazzo for his precious technical support.
Conflicts of Interest: The authors declare no conflict of interest.

Nomenclature
PU production unit
HV high voltage
MV medium voltage
TSO transmission system operator
DSO distribution system operator
MSD dispatching services market
PST phase-shifting transformer
ATR autotransformer

63
Energies 2023, 16, 6258

TRS transformer
OLTC on-load tap changer
TS tap staggering
TCR thyristor controlled reactor
AVR automatic voltage regulator

References
1. National Grid. Electricity Ten Year Statement; National Grid: London, UK, 2012; pp. 221–223.
2. Kaloudas, C.G.; Ochoa, L.F.; Fletcher, I.; Marshall, B.; Majithia, S. Investigating the declining reactive power demand of UK
distribution networks. In Proceedings of the IEEE Power & Energy Society General Meeting (PESGM), Denver, CO, USA, 26–30
July 2015; pp. 1–5. [CrossRef]
3. ARERA. The Italian Regulatory Authority for Energy, Networks and Environment. DCO 515/21, Appendix 2, Analisi Dei Flussi
Di Energia Reattiva Nelle Reti in Alta e Altissima Tensione (Analysis of Reactive Energy Flows in HV Networks). Available
online: https://www.arera.it/allegati/docs/21/515-21appendice2.pdf (accessed on 17 August 2023).
4. ARERA. The Italian Regulatory Authority for Energy, Networks and Environment. DCO 515/21, Completamento Della
Regolazione Tariffaria Dell’energia Reattiva (Completion of Reactive Energy Tariff Regulation). Available online: https://www.
arera.it/allegati/docs/21/515-21.pdf (accessed on 17 August 2023).
5. ARERA. The Italian Regulatory Authority for Energy, Networks and Environment. DCO 420/16, Regolazione Tariffaria
Dell’energia Reattiva per le Reti in Alta e Altissima Tensione e per le Reti di Distribuzione (Reactive Energy Tariff Regulation for
HV Networks and Distribution Networks). Available online: https://www.autorita.energia.it/allegati/docs/dc/420-16.pdf
(accessed on 17 August 2023).
6. Chen, L.; Li, H. Optimised Reactive Power Supports using Transformer Tap Stagger in Distribution Networks. IEEE Trans. Smart
Grid 2017, 8, 1987–1996. [CrossRef]
7. Chen, L.; Li, H.; Cox, S.; Bailey, K. Ancillary Service for Transmission Systems by Tap Stagger Operation in Distribution Networks.
IEEE Trans. Power Deliv. 2016, 31, 1701–1709. [CrossRef]
8. Weedy, B.M.; Cory, B.J.; Jenkins, N.; Ekanayake, J.B.; Strbac, G. Electric Power Systems, 5th ed.; John Wiley & Sons: Chichester, UK,
2012; pp. 180–181. ISBN 978-0-470-68268-5.
9. O’Kelly, D. The Representation of Tap-Change Transformers in Power System Networks. Int. J. Electr. Eng. Educ. 1970, 8, 187–192.
[CrossRef]
10. Constantin, M.; Eremia, M.; Toma, L. Comparative analysis between conventional voltage control using reactors and continuous
voltage control using TCR in the Romanian transmission grid. In Proceedings of the IEEE PowerTech, Grenoble, France, 16–20
June 2013; pp. 1–6. [CrossRef]
11. Chen, L.; Li, H.; Turnham, V.; Brooke, S. Distribution network supports for reactive power management in transmission systems.
In Proceedings of the IEEE PES Innovative Smart Grid Technologies Conference Europe (ISGT Europe), Istanbul, Turkey, 15–16
October 2014; pp. 1–6. [CrossRef]
12. Wang, D.; Chen, L.; Li, H.; Wang, Z.; Turnham, V. Impacts of tap stagger on currents of power transformers. In Proceedings of the
IEEE PES Innovative Smart Grid Technologies Conference Europe (ISGT-Europe), Turin, Italy, 26–29 September 2017. [CrossRef]
13. ARERA. The Italian Regulatory Authority for Energy, Networks and Environment. Testo Integrato Delle Disposizioni per
L’erogazione Dei Servizi di Trasmissione e Distribuzione Dell’energia Elettrica. Available online: https://www.arera.it/allegati/
docs/19/568-19TIT.pdf (accessed on 17 August 2023).
14. European Network of Transmission System Operators for Electricity—ENTSO-E. Phase Shift Transformers Modeling, Version
1.0.0. 2014. Available online: https://eepublicdownloads.entsoe.eu/clean-documents/CIM_documents/Grid_Model_CIM/
ENTSOE_CGMES_v2.4_28May2014_PSTmodelling.pdf (accessed on 17 August 2023).

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

64
 energies
Article
Analysis of Secondary Controller on MTDC Link with Solar PV
Integration for Inter-Area Power Oscillation Damping
Oluwafemi Emmanuel Oni * and Omowunmi Mary Longe

Department of Electrical and Electronic Engineering Science, University of Johannesburg,

Johannesburg 2006, South Africa; omowunmil@uj.ac.za
* Correspondence: 223248712@student.uj.ac.za; Tel.: +27-11-678984406

Abstract: Integration of renewable energy sources is important in limiting the continuous environ-
mental degradation and emissions caused by energy generation from fossil fuels and thus becoming
a better alternative for a large-scale power mix. However, an adequate analysis of the interaction
with the alternating current (AC) network during network disturbance, especially during inter-area
power (IAP) oscillations is needed. Insufficient damping of oscillations can significantly impact the
reliability and effective operation of a whole power system. Therefore, this paper focuses on the
stability of the modified Kundur two-area four-machine (MKTAFM) system. A robust secondary
controller is proposed and implemented on a line commutated converter (LCC)-based multi-terminal
high voltage direct current (MTDC) system. The solution consists of a local generator controller and
the LCC MTDC (LMTDC) system, voltage-dependent current order limiter, and extinction angle
controller. The proposed robust controller is designed for the LMTDC systems to further dampen the
inter-area power oscillations. Three operational scenarios were implemented in this study, which are
the local generator controller and double circuits AC line, local generator controller with LMTDC
controllers, and local generator controller with LMTDC controllers and secondary controller. The
simulation result carried out on PSCAD/EMTDC recorded better damping of the inter-area power
oscillation with LMTDC. A considerable improvement of 100% damping of the IAP oscillations was
observed when a secondary controller was implemented on the LMTDC.

Keywords: inter-area power oscillation; two-area four-machine network; solar PV; high voltage
Citation: Oni, O.E.; Longe, O.M. direct current; PSCAD; secondary controller
Analysis of Secondary Controller on
MTDC Link with Solar PV
Integration for Inter-Area Power
Oscillation Damping. Energies 2023,
1. Introduction
16, 6295. https://doi.org/10.3390/
en16176295 An increase in the number of power system interconnections has resulted in inter-area
power oscillations, a phenomenon involving groups of generators oscillating relatively to
Academic Editor: Ying-Yi Hong
each other that are sometimes hard to control due to their scale and complexity [1]. Being
Received: 22 July 2023 often of low frequency, these oscillations are becoming a big challenge to the safe operation
Revised: 26 August 2023 of modern power networks. When poorly damped, inter-area oscillation can further lead to
Accepted: 27 August 2023 voltage or rotor angle instability, or system collapse. Inter-area oscillations have long been
Published: 29 August 2023 a point of discussion among researchers, power utilities, and industries. The frequency is
always in the range of 0.1 to 1 Hz [2]. These oscillations lack damping or are unstable, and
when these disturbances take place due to crucial faults on the line, the network becomes
less stable.
Copyright: © 2023 by the authors. Engineers and researchers in the power and control systems have worked extremely
Licensee MDPI, Basel, Switzerland.
hard over the past 20 years to increase the stability of power systems [3–6]. The standard
This article is an open access article
controllers, which include the power system stabiliser (PSS), automatic voltage regulator
distributed under the terms and
(AVR), and speed governor control, are single-input single-output non-coordinated linear
conditions of the Creative Commons
controllers that are mostly not capable of ensuring stability when significant perturbations
Attribution (CC BY) license (https://
or contingencies arise. In addition, the AVRs, which are generally employed to maintain
creativecommons.org/licenses/by/
4.0/).
the terminal voltage magnitude of the synchronous generator, introduce negative damping

Energies 2023, 16, 6295. https://doi.org/10.3390/en16176295 65 https://www.mdpi.com/journal/energies

Energies 2023, 16, 6295

torques, which have a detrimental effect on stability. Because of operating point variations
and short circuit disruptions, power systems exhibit electromechanical oscillations. These
frequency oscillations must be suppressed to the amount that is acceptable; otherwise,
instability might develop as their amplitude grows. In order to address these problems,
power system stabilisers (PSSs) are further utilised to produce an extra stabilising signal to
the excitation system to dampen these oscillations [7].
Other controllers like the Static Var Compensators (SVC) and flexible AC transmission
systems (FACTS) have also been deployed for inter-area oscillations damping; however,
under specific operating circumstances, an inter-area mode may be viewable from one area
and controllable from another [8,9]. In this situation, a failure or system disturbance in one
area will tend to have a greater effect than before in another.
Researchers have developed several efficient techniques to examine the low-frequency
oscillation issue [10–12]. Among these, the linearised state matrix’s modal analysis offers
a fundamental technique for analysing low-frequency oscillations. This approach may be
used to identify the system’s weakly damped oscillation modes, whether they are local or
inter-area. Using linear participation factors, it is possible to pinpoint the high association
between a few chosen generators and the weak-damped mode. A fuzzy controller can
also be used to provide a fast response and coordinated control of a wide area network
for quick damping of inter-area oscillations [13]. In [14], a Factional Order Proportional
Integral (FOPI) was used against the conventional PI controller with an adaptive differential
evolution algorithm to further provide an optimal tunning of the FOPI for performance
enhancements. Another work also carried out a data-enabled predictive control algorithm
on a voltage source converter-based high-voltage DC (HVDC) for optimal damping of wide-
area power oscillations [15]. A wide area damping controller (WADC) was proposed for
energy storage systems such as capacitors for the damping of low-frequency oscillations [16].
The authors used an accelerated particle swarm optimization technique for the turning
of the wide-area controllers. In the related research [17], the authors use dual power
oscillation dampers on the active and reactive power of a double-fed induction generator
for the damping of the inter-area power (IAP) oscillations.
However, the recent trend and transition towards renewable energy and the dereg-
ulation of the electricity market point towards the usage of different DC converter links,
which has led to different changes in generation and power transmission patterns as well
as grid topology [18]. Therefore, a change is required for a better controller to adequately
dampen inter-area oscillations before the generator control goes out of step, especially
when operating a thyristor-based multi-terminal HVDC system.
Therefore, this paper presents a novel method to model a multi-terminal line com-
mutated converter (LCC) transmission link to provide a robust, flexible, and expandable
oscillation damping for a modified Kundur two-area four-machine (MKTAFM) network.
The method used in this paper is to build and validate a secondary controller for one of the
rectifying stations of the multi-terminal direct current (MTDC) system. The topology of
the secondary controller is first tuned and adjusted precisely for the MTDC network. The
LCC-MTDC (LMTDC) grid is equipped with an overall power controller that optimises the
power flow and current sharing between the two rectifier stations. Therefore, this paper
focused on the control strategy of LMTDC grids in reducing the inter-area oscillation of
two-area networks. This control aims to minimise the ripple effect of faults on the AC grid
voltage and maintain a power balance in case of a fault in one of the areas.
In the first contribution of this paper, it shows how the stability of two asynchronous
grids can be improved. The second contribution of the paper shows how a coordinated
control of three-terminal LCC MTDC links can be used to avoid inter-area oscillations
interaction between two inter-area modes. The investigation carried out in this study
evaluates the performance of an MKTAFM and proffers a better means of inter-area oscil-
lation reduction with the use of some multivariable as well as robust controller designs.
Furthermore, this research design handles voltage control and power oscillation damping
issues concurrently. Our method is distinguished by the fact that the second level of control

66
Energies 2023, 16, 6295

action optimises the performance of the existing control of the rectifier stations and thus
improves the performance of the substation. The controller configuration and the parame-
ters required for this implementation are derived methodically from the MKTAFM system
model. The suggested secondary control is robust and merges an input–output linearisation
control method with a nonlinear control approach. This secondary controller constantly
adjusts its parameters in response to changes in working circumstances or structure, thus
significantly improving the MTDC controller’s effectiveness.
This paper is structured as follows: the authors first gave an insight into the theoretical
analysis of inter-area oscillation utilising a two-machine network. The MTDC model
used was then presented alongside the control architectures, which are the main MTDC
controllers as well as the secondary controller. The test system modelling and the schematic
diagram were then explained. Furthermore, the methodology based on the fault analysis
on the test network was given, and the results were then presented. Then, the conclusion
section follows, which summarises all the work carried out in this paper.

2. Inter-Area Oscillations
Inter-area oscillations, which are oscillations occurring between different regions
within a synchronously connected power system, represent the most critical threat to
the reliable and stable operation of such a power system [19]. In situations where the
conventional control mechanisms, due to either weak tie-lines or inadequate grid strength,
are incapable of effectively dampening the system’s oscillations following a disturbance,
inter-area oscillation phenomena may arise [20]. The ability of an electrical power system
to maintain stability in the face of these minor disturbances is referred to as small-signal
stability. In contrast to small-signal perturbations, the nonlinear equations governing power
systems can be linearised around a specific steady-state operating condition. Due to the
inherent weak damping characteristics of inter-area oscillations, there exists a significant
probability of irreversible and extensive blackouts, which can lead to severe economic
losses and potential loss of human life, thereby resulting in catastrophic consequences.
The small-signal stability is evaluated using eigenvalue analysis, sometimes referred
to as modal analysis. This study is primarily intended to reduce inter-area power oscilla-
tions. The theoretical background of the small-signal stability of a power system is further
provided in Equations (1) to (5). The synchronous machine’s inertia fluctuations are initially
analysed to determine the damping coefficient needed to reduce power system oscillations
as effectively as possible. The coefficient of damping signifies the adequate torque needed
to control the synchronous speed. Figure 1 [20] gives a schematic of a basic two-area system
consisting of two incoherent machines with inertias of the generator denoted by Ha and
Hb while δa and δb indicate the rotor angle of areas A and B oscillating at two distinct
frequencies and amplitudes. Equations (1) and (2) give the swing equation for individual
machines, which, when subtracted, gives (3). Equation (4) is generated by dividing (3) with
the inertia coefficients shortly before the second derivation. Further simplification gives (5).
To avoid the complexities of swing equations for numerous synchronised machines during
the stability analysis of a large system, all swing equations must be combined to minimise
the total amount of iterations needed to produce the final swing equations. The total system
equivalent rotating inertia is obtained by adding the rotating inertia H of each generating
unit [21].
2 ∂2 δa
Pm a − Pe a = Ha · · 2 (1)
ω ∂t

2 ∂2 δb
Pmb − Peb = Hb · · (2)
ω ∂t2
 2 
Pm a − Pe a Pmb − Peb 2 ∂ δa ∂2 δ
− = · − 2b (3)
Ha Hb ω ∂t 2 ∂t

67
Energies 2023, 16, 6295

   2 
2 Ha Hb ∂ (δa − δb ) H Pm a − Ha Pmb − Hb Pe a − Hb Peb
· · = b − (4)
ω Ha + Hb ∂t2 Ha + Hb Ha + Hb

2 ∂2 δab
· Hab · = Pabm − Pabe (5)
ω ∂t2

Figure 1. Schematic diagram of a two-area network with two machines [20].

The difference in rotor tilt angle between the two generators and the corresponding
inertia constant is given as Hab , and δab as shown in (6) and (7), respectively. Pmab and
Peab in (8) and (9) are the respective electrical and mechanical power between the two
generators. The damping constant KD denotes the swing characteristics of the synchronous
machine, and it is described using the second-order differential equation in (10). The
undamped natural frequency (ω n ) and the damping coefficient (ζ) required to fully reduce
the oscillation amplitude are given in (11) and (12), respectively. Since the inertia constant
and the oscillation frequency are indirectly proportionate, a power network that is equipped
with a high constant of inertia is less prone to faults and vice versa. This demonstrates that
the output supplied power has a direct impact on the system’s inertia constant. As a result,
increasing the stability condition of the power system necessitates an inertia constant value
that is greater than 3.5 pu.
Ha Hb
Hab = (6)
Ha + Hb

δab = δa − δb (7)

Hb Pm a − Ha Pmb
Pabm = (8)
Ha + Hb
Hb Pe a − Hb Peb
Pabe = (9)
Ha + Hb

2 ∂2 δab ∂δab
· Hab · + KD = − P cos(δo ) abmax (10)
ω ∂t2 ∂t
#
ωP cos(δo ) abmax
ωη = (11)
2Hab
$
1 % ω
ζ = KD % &  (12)
2 2P cos δo ) ababmax

68
Energies 2023, 16, 6295

3. System Model
3.1. Test Network Modelling
The system under investigation in this research paper is a network, depicted in Figure 2.
It comprises four synchronous generators situated in two distinct areas, a 50 MW grid-tied
solar PV connected at BB2, and a three-terminal line-commutated multi-terminal direct
current (LMTDC). The synchronous generators employed in this study are equipped with
an IEEE AC4A exciter for automatic voltage regulation (AVR) and an IEEE single-input
power system stabiliser (PSS). To introduce a significant line impedance, the transmission
line linking BB10 to BB11 was modified, increasing its distance from 25 km to 80 km.
Consequently, a double-circuit transmission line was utilised to enhance the power transfer
between area 1 and area 2. Subsequently, a real-time domain modelling and simulation
approach is employed to analyse the complete generator’s inter-area power and frequency
oscillations. Two system faults are simulated to represent a worst-case scenario. The power
output, synchronous speed, and inter-area power transmission of each generating unit are
plotted during these system disturbances. Bus BB7, BB8, and BB9 are selected for power
transfer quality analysis at these locations and to verify the voltage profile. This mode is
classified as inter-area since it involves two coherent sets of generators oscillating at a low
frequency in opposition to each other.

Figure 2. The Kundur two-area four-machine network.

Power Systems Computer Aided Design (PSCAD) was used to model the synchronous
machines, which included the IEEE controller model for the AVR and the PSS. All generators
were evaluated in terms of the true inertial constant and were linked together by a step-
up transformer of 20/230 kV voltage range and an apparent power rating of 900 MVA.
During the evaluation time, a constant value was pre-selected for the governor control
for all four machines to obtain an in-depth impact of the DC controller, AVR, and PSS
effects on minimising the disturbances caused by the fault on the complete model. The
authors in [2,20,22] gave an explanation about the modelling of these machines and further
provided the parameters for both the generators and the transformers and the line models.
In the dynamic simulation of this network, three operational scenarios were considered:
(1) usage of only the AC line on the MKTAFM network; (2) usage of LMTDC link to
interconnect bus BB6 (rectifier 1) or BB11 (rectifier 2) to the inverter station at bus BB9; and
(3) implementation of a secondary controller on the rectifier 1 station. During the second
and third scenarios, a single circuit was used rather than the double circuits shown in the
MKTAFM network. This measure was taken to avoid excessive power transfer across the
bus and to also show the scenario of replacing an existing AC line with DC lines, thereby
showing the impact of it on the overall system.

69
Energies 2023, 16, 6295

The power system’s dynamic characteristics are captured by the collection of nonlinear
differential algebraic equations, which serve as the mathematical representation of the system.
Equations (13)–(15) express these dynamics’ analysis of the power system interconnections.
.
x = f ( x, y, u) (13)

0 = g( x, y, u) (14)

z = h( x, y, u) (15)
The vector x in these equations represents the state variables, the vector y represents the
algebraic variables, the vector u represents the input variables, and the vector z represents
the output variables. These equations regulate the dynamic behaviour of the many types
of models used in this study. The specific equations relating to the dynamic behaviour of
these models are discussed in the following sections.

3.2. Grid-Tied Solar PV Analysis

A single solar PV source is a proportionate equivalent of several strings of PV modules
connected in parallel with each of the string modules connected in series. With identical
arrays of PV modules, a general analysis can be given for a single model solar cell as shown
in Figure 3 using the electrical equivalent circuit [23,24]:

Figure 3. Solar PV cell model.

The generated photocurrent, denoted as Is, in a photovoltaic (PV) cell exhibits a linear
relationship with the solar irradiance it is exposed to from sunlight. However, the presence
of the anti-parallel diode results in non-linear current-voltage (I-V) characteristics of the
PV cell. By applying Kirchhoff’s current law to the circuit, (16) is obtained. Equation (18)
is derived by substituting the relevant expressions for the diode current (ID ) and the
shunt branch current (I1 ). The modified non-ideal factor, expressed in (18), is directly
proportional to the temperature of the cell, denoted as Tc. The photocurrent, as described
in (19), is dependent on the solar radiation (G) incident on the PV cell’s surface and the
cell temperature (Tc ). The short circuit current (ISCR ) is defined with respect to a reference
solar radiation (GR ) and a reference cell temperature (TR ). The parameter αT represents the
temperature coefficient of the photocurrent, with a value of αT = 0.0017 A/K for silicon solar
cells. The dark saturation current (Io ), given by (20), is influenced by the cell temperature,
where IoR represents the dark current at the reference temperature. In these equations,
q denotes the charge of an electron, k represents the Boltzmann constant, e.g., corresponds
to the band-gap energy of the solar cell material, and n is the ideality factor when kTc = q.
The instantaneous values of solar radiation (G) and the operating temperature (T) of the
cell are measured in watts per square meter (W/m2 ) and degrees Celsius (◦ C), respectively.
Considering that a PV module comprises numerous cells connected in series, and a
PV array consists of both series and parallel combinations of modules, the circuit model of
a single cell can be scaled up to represent any desired series/parallel configuration.

Is = ID + I1 + I2 (16)

70
Energies 2023, 16, 6295

   
Vout + I2 R2 Vout + I2 R2
I2 = Is − Io e m −1 − (17)
R1

nkTc
m= (18)
q
 
G
Is = ISCR · · [1 + α T ( TC − TR )] (19)
GR
 3   
TC qe g 1 1
Io = IoR · ·e − (20)
TR nk TR TC

3.3. PV Modelling
The schematic diagram presented in Figure 4 illustrates the comprehensive config-
uration of a grid-connected photovoltaic (PV) model consisting of DC–DC and DC–AC
converters block diagram within it (parameter details in Appendix A). The components
encompassed in this scheme include PV arrays, a DC–DC boost converter, a voltage source
converter for DC–AC conversion, and an AC filter. The PV subsystems are capable of
functioning in two primary modes, namely, continuous conduction mode (CCM) and
discontinuous conduction mode (DCM). In solar PV applications, the CCM is the preferred
mode of operation [25].

Figure 4. Grid-tied PV model.

The PV system in this study utilised a DC–DC boost converter model as shown in
Figure 4. It serves as a step-up converter extensively employed for integrating low-voltage
PV modules with the utility grid. Furthermore, it fulfills the role of maximum power point
tracking (MPPT) under nominal utility conditions. In the boost converter, the voltage
is increased by adjusting the duty cycle. The DC–DC converter plays a vital role in a
solar PV application by transforming the input direct current from one voltage level to
the required level. It serves as a crucial component for maximum power point (MPP)
tracking. The operation of the DC–DC converter serves as the foundation for detecting the
MPP, as outlined in the proposed global MPPT control algorithm. In practical scenarios,
despite the implementation of MPPT, the output voltage of a PV string remains significantly
low. Therefore, a front-end DC–DC converter with step-up capability is necessary for grid
connection. The fundamental principle behind the MPPT algorithm relies on harnessing
the voltage and current variations resulting from instantaneous power fluctuations. By
analysing these variations, it becomes possible to determine the power gradient and assess
whether the solar PV system is operating near the maximum power point. The maximum
power delivered by the solar PV array is the product of the optimal operating voltage and
current of the PV array at the maximum power output condition.
To achieve grid synchronisation, an inverter is necessary. The control of the inverter
governs the power and frequency on the AC side while minimising system harmonics.

71
Energies 2023, 16, 6295

The switches within the inverter are controlled through the implementation of an inverter
control algorithm [26].

3.4. MTDC Design

The standard control approach of LMTDC converters requires rectifier stations to be
in DC current control (DCC) mode. Voltage-dependent current order limiter (VDCOL)
and current control are further used to provide a better current order control. As shown
in Figure 5, the DCC method is used at the rectifier 1 substation. The current order is
generated by either selecting the reference input current deployed by the overall power
controller or by selecting the current order from the VDCOL. The VDCOL gets its input
from the intended filtered DC current (Idcrect1 ) measured from the DC line. The lower value
of any of the two is chosen as the standard current order (Iord ) for the generation of the firing
angle for the rectifier. The same process is used at the rectifier 2 substation as illustrated
in Figure 6. The difference between the rectifier 1 and the rectifier 2 controller is that the
rectifier 1 substation has another current order input from the secondary controller (SC).
Also, both rectifiers have different power-carrying capacities and are thus rated differently
in the parameters that are used for their control signals.

Figure 5. LMTDC rectifier 1 control diagram.

Figure 6. LMTDC rectifier 2 control diagram.

Various control approaches can be used to regulate the inverter, including extinction
angle control (EAC), DCC, and DC voltage control (DVC). The reference signal is acquired in
the same manner as in the rectifier control mode as shown in Figure 7. However, unlike the
rectifier’s DCC, the inverter’s DC current is limited to the difference between the amount
of the standard current order and that of the current margin (Imarg ). In other to positively
influence the power reversal process, the current margin is programmed into the inverter
current control and normally set on a fixed range of 0.1–0.15 pu. A current error control
(I-error) indication is commonly used between the DCC and the DVC. This generated
I-error can also be inserted between the DCC and the EAC to smoothen changes between
the control modes. The smallest of these generated parameters is chosen as the inverter
control parameters because it is preferable to run the inverter at the smallest extinction
angle possible to reduce the power losses, reduce the amount of reactive power needed to
compensate for the thyristor valve, and, ultimately, reduce the harmonic distortion content.

72
Energies 2023, 16, 6295

Figure 7. LMTDC inverter control diagram.

For every input of this control signal, being in per unit, it must undergo a first-order
inertia transfer function to make the control switching process a smooth transition. The
equation is expressed as follows:  
1
G1 (21)
1 + sT1
The LMTDC system’s fault vulnerability is entirely dependent on its general coordina-
tion; thus, each converter’s controllers were preset and adjusted to fit the case study. The
overall power controller regulates and oversees the LMTDC link’s converter operations.
It accepts a varying power order measured from the systems. This controller yields the
current order after being divided by the measured DC voltage by guaranteeing that the
total DC current of all converters is zero (Idc = 0). The master control, seen in Figure 8,
is responsible for balancing the combined converter’s power and current order. This
controller generates the current order for each converter system based on the measured
voltage at each converter station and a preset power order. It also provides allowance and
compensation in case of DC line losses.

Figure 8. Overall power controller.

The LMTDC used in this study comprises a three-terminal thyristor converter. The
firing angle at rectifiers and converter units is used to regulate the DC voltages. The overall
power controllers generate the input value of the DC current for each of the converter’s
stations. The DC voltage for the rectifier and inverter stations, as well as the rectifier firing
angle, are expressed below [27–29]:

Vdr = 3.12BTU cos α − 0.96BXc Idc (22)

Vdi = 3.12BTU cos β + 0.96BXc Idc (23)

73
Energies 2023, 16, 6295

 
Udc + Idcre f Rdc + 0.955BXIdcre f
αc = cos−1 (24)
3.1213 · BTU
The subscripts r and i in the (22) to (23) stand for rectifier and inverter. The rectifier
and inverter DC voltage are denoted as Vdr and Vdi , while the RMS L-L converter voltage
is shown as U. The number of series-connected bridges of the thyristors is denoted by B.
Therefore, for a 12-pulse thyristor converter, B = 2; meanwhile, a bipolar HVDC system
will have a B-value of four. The firing angle (α) in (24) is the required angle at which the
converter gets turned on and begins to conduct; X denotes the commutation reactance, and
Rdc is the total resistance at the DC side of the converter. Also, T stands for the transformer
ratio, and the AC voltage as each of the converter stations is depicted as U.
The classic secondary controller (SC) adopted in this study is like the power oscillation
damper that has the same mode of construction as a generator’s PSS (power system
stabiliser), which includes a washout filter, a gain G, and ‘p’ phase lead-lag blocks (Figure 9).
The transfer function provided in (25) defines the secondary controller equations with the
parameter details in Table 1. The washout filter is typically used as a link between the
measured signal and the used signal in the control loop to washout (or reject) the steady-
state component of the measurements while passing the transient ones. The traditional
parametric tuning method comprises two steps: (i) the calculation of T4 and T5 to adjust
the phase of the mode’s residue to 180◦ and (ii) the adjusting of the gain G to achieve the
required damping. The output control signal from each of the transfer functions during the
simulation of the whole system is plotted in Figure 10.
       
1 sT2 1 + sT4 n 1
Hsoc = G1 · G2 · G3 · G4 (25)
1 + sT1 1 + sT3 1 + sT5 1 + sT6

Figure 9. Secondary controller block diagram of the proposed model.

Table 1. Optimised parameter values.

Model Parameter Value

SC Gain G1, G2, G3, G4, [0.006; 0.31; 1.01; 0.24]
SC Time (s) T1, T2, T3, T4, T5, T6 [0.012; 1.5; 1.5; 0.514; 0.172; 0.055]

Figure 10. Measure control input during the simulation of the secondary controller.

4. Methodology and Scenarios

To study the small-signal characteristics of an MKTAFM network during the integra-
tion of solar PV systems, a dynamic RMS simulation was performed on PSCAD/EMTDC, a

74
Energies 2023, 16, 6295

power systems simulation software designed for the simulation of electromagnetic transient
analysis of DC networks. The base case study, which is termed the first scenario of using
only the AC lines and solar PV, was first analyzed. The simulation included a fault analysis
where a three-phase-to-ground (3Ph-G) fault occurred on the L7 line connecting BB7 and
BB8, with a fault clearing time of 210 ms and a disturbance of 1.2 s on GenC. Both events
happened concurrently at the 2.0 s mark time during the simulation, thus affecting different
areas of the system to truly test the limits of the power system operation. The chosen
disturbances simulated signify a worst-case scenario. The second scenario recorded a few
changes, as three of the transmission lines, L4, L7, and L9, were replaced with DC lines (see
Figure 2), but the same fault impedance was applied to the system when it was connected
to the LMTDC link. The same situation applies to scenario 3 with LMTDC lines; however,
it differs in the inclusion of the secondary controller for the quick damping of the inter-area
power oscillation.
In all the scenarios covered in this study, the power generated by GenB and the solar
PV link never changes, as shown in Figure 11, with GenB and the solar PV recording an
active power of 622 MW and 50 MW, respectively. The goal of this study is to evaluate the
controller contribution of LMTDC in reducing inter-area power oscillations even during a
high penetration of renewable energy.

Figure 11. Active power for (a) solar PV, and (b) GenB.

5. Results
The results of the dynamic analysis using PSCAD/EMTDC simulation tools consisted
of observations of the generator’s active power, oscillation speed, and converter data. The
magnitude and damping rate of the oscillation were studied for both situations for the
inter-area power oscillation. The generator speed also provides information about the
contributing element for each generator in the oscillatory phases. Since the focus is on the
inter-area oscillation, the system short circuit fault time was limited to a short period of
210 ms. After these system disruptions, a positive damping coefficient was recorded during
all scenarios, i.e., the systems maintained their pre-fault steady state condition, albeit with
distinct amplitudes and waveform distortion.
The plot shown in Figure 12 is the active power for all the generators in the three
scenarios. In the first scenario, the synchronous machine 2 (GenB) power dips lowest to a
201 MW value at the instant of the 3Ph-G fault occurrence; GenA follows thereafter. This
dip in GenB is due to the near proximity to the transmission line fault following a decline
in the active power produced by GenC due to a protracted generator disruption nearing 1 s
simulation time on GenC. The active power recorded on this generator reached 916 MW,
nearly 1.4 pu of its operating state value, due to the increase in the accelerating power after
being subjected to a fault. Another observation in this plot is the inaccuracy in the power
generated in each generator due to the high line impedance and losses on each of the AC
lines used. Furthermore, GenD generated the largest share of the load power, followed by
GenB, GenA, and then GenC. This disparity is caused by the high line impedance between

75
Energies 2023, 16, 6295

GenC and the load area. This points to the fact that even the inclusion of the solar PV link
is still not sufficient to provide an adequate damping torque that can sufficiently dampen
the power oscillations.

Figure 12. Active power and synchronous speed plot.

The second scenario involves the replacement of all the double transmission circuits in
the system with a single transmission line and DC lines, with the DC lines forming a three-
terminal DC (LMTDC) link. A performance comparison was carried out with regards to
the first case study to determine which links achieved better inter-area oscillation reduction
and to investigate if the LMTDC link can damp oscillation generated by solar PV during
system disturbance. In Figure 12, the active power shown in this plot exhibited a close-to-
average power generation. This close range of power transfer is due to the benefits offered
by the LMTDC system in loss reduction and enhancement of power transfer. Therefore,
after the 3Ph-G fault, due to a continuous generator fault, GenD and GenC registered the
greatest power amplitude of 916 MW. These amplitudes resulted in greater inertia value
in recovering the normal operating point of the complete network. GenB still recorded
the lowest dip of 175 MW. However, a better power profile can be seen in this plot as
all four generators recorded a close range of power. The functionality of this was made
possible with the inclusion of the LMTDC link, which further helped in the active power
dispatch. However, the simulation plots at this stage still contain a number of oscillations.
The inclusion of the secondary controller to the rectifier 1 station totally eradicated these
oscillations at 5.8 s simulation time.
Figure 12 shows the frequency swings of the generators that were observed in the
system. This result comprises incoherent machines, with area A having a different inertia
constant compared to area 2. This results in the generators in area A swinging with the
machines in area 2. The highest oscillatory modes, up to 1.007 pu, were observed in the
GenA and GenB subplots. The machine’s accelerating power played a role in meeting
the expected load demand in area B, which led to the maintenance of a stable operating
state, albeit with a high oscillating frequency. During the first two scenarios, the generators’

76
Energies 2023, 16, 6295

power system stabiliser (PSS) and automatic voltage regulator (AVR) controllers had
already provided damping torque to maintain stable operating points. However, this
was insufficient in quickly generating a damping coefficient that would adequately and
effectively return the entire test network to its steady-state operational state. An increase in
the fault clearance time (tc = 300 ms) would result in system collapse, where the systems
can no longer maintain a stable operating point. In Figure 12, the synchronous speed for
GenC and GenD both followed the same pattern of system oscillation being similar in
area 2; the same applies to GenA and GenB in area 1.
In the third scenario, although initial oscillation is the same as the previous scenarios,
the inclusion of the secondary controller into the KMTAFM network generated a quick
damping torque for the system to quickly return to its steady state condition.
Following the fault, BB7, BB8, and BB9 in Figure 13 illustrate the bus voltages during
the three scenarios. In order to fully understand the comparison between the three different
scenarios based on the voltage profile, the busbar voltage was grouped together, indicating
a particular bus voltage for the three different scenarios. The observation of these voltage
profiles shows that BB8 is the weakest bus in the network across all three scenarios, which
can be attributed to its distance from the generating plants in both areas. BB7 followed as
the second weakest bus. However, the second scenario recorded a better voltage profile
than when only HVAC and solar PV were used. The plot also indicates that an increase in
transmission distance between BB10 and BB11 did not have a significant impact on BB9’s
voltage profile in contrast to the other buses examined in both scenarios. During the third
scenario, the oscillations observed in the system voltage profile were completely improved,
and the system’s steady conditions were stabilised after approximately two cycles (5.8 s of
simulation time).

Figure 13. Bus voltage profile.

Figure 14 shows the impact of the LMTDC in reducing the oscillation after the system’s
disturbance. The plot was categorised into two parts, depicting the scenario with and
without the secondary controller. The voltage-dependent current order limiter has a look-
up graph based on the V-I characteristics of the converter that helps the converter in
reducing the amount of current that flows through the link when the voltage reduces. By
doing this, the converter reduces the amount of power that flows through the link during
this disturbance and thus reduces the amount of reactive power consumption. The current
output from the VDCOL is passed through a PI controller that further generates the firing
angle for each of the converters. The scenarios without the inclusion of the secondary
controller recorded a ripple that was still very much visible at the end of the simulation
time. These oscillations are still further cumulated into the firing angle for each of the
converters. The inclusion of the secondary controller created an immediate ripple that
further provided a quick, sufficient, and robust damping torque in totally eradicating the
oscillations in the current output from the VDCOL and further provided a better firing
angle, and thus a better DC power output.

77
Energies 2023, 16, 6295

Figure 14. LMTDC converter plots.

Figure 15 demonstrates the inter-area power transfer between BB7 and BB9. The
study revealed that without the presence of MTDC links, the inter-area power oscillations
had a slow damping rate during the initial scenario, where only AC transmission lines
were used. During the post-fault transient condition, the two areas in the study system
contributed more to the oscillations, which delayed the system from reaching a steady-state
condition. However, the addition of the MTDC link showed that the damping rate of the
inter-area power oscillations increased significantly during the same disturbance scenario.
The amplitude of the IAP transfer likewise showed a considerable positive damping rate,
leading to a more stable operative condition. However, not all the inter-area oscillation
was fully damped at the 8 s simulation time. Therefore, a secondary controller was further
used. The result from this analysis shows a better oscillation damping, whereby all the
oscillation was completely damped at exactly 5.8 s simulation time. During this scenario,
the active power transfer comprised the power transmitted from the AC line L6 to line L8
and the DC power transmitted through the rectifier 1 converter link. The plot indicates that
the damping ratio resulted in a complete suppression of the oscillation amplitude’s value
during the second period of simulation.

78
Energies 2023, 16, 6295

Figure 15. Inter-area power oscillation.

Performance Comparison
In this section, a comparison of various schemes, presented in Table 2, is conducted to
assess the contributions of the proposed model and controllers. The highlighted reviews are
based on authors who carried out inter-area power oscillation studies using the two-area
four-machine (TAFM) network.

Table 2. Performance comparison of the proposed model.

Ref Method Performance in Comparison to This Study

The results based on the 3Ph-G faults show little
Adaptive online control design for a power system
impact as both the classic and the online tuned PSS
[30] stabiliser (PSS) and comparing the classic PSS
recorded some amount of oscillation even at 10 s
controller and an online tuned PSS.
simulation time.
They used weighted signals representing different
The TLBO with the PSS provided a quick settling
areas to optimize a PSS. They used two
time of the bus voltage oscillation and recorded a
[31] optimization methods: the Harmony Search (HS)
settling time of 4.8 s, but fails to include the
algorithm and the Teaching Learned Based
dynamics of HVDC/MTDC lines on the systems.
Optimization (TLBO) technique.
They used VSC HVDC power oscillation dampers
The result utilising LQG POD provided a better
(POD), PSS, and wide area measurement
[32,33] oscillation reduction technique. However, the time
signals-based controllers designed with a novel
taken to reach a total settling phase is around 12 s.
linear quadratic gaussian (LQG) method.
These literatures used an enhanced static The result shows an improved power oscillation;
synchronous compensator to damp inter-area however, at a longer simulation time of ∼
= 10 s. It
[34,35]
power oscillation and also compared different also differs from the present study as no MTDC
FACTs devices. link was used.
They used a single thyristor controlled braking
A reduced oscillation was recorded, but the 3-phase
[36–38] resistor and dual fuzzy logic for the minimization
fault on the line generated a prolong oscillation.
of inter-area power oscillation.
They used a clustering distributed generator to They recorded a quick settling time, but the DGs in
[39,40] provide more reserves for generators. They also the load bus provided a quick supply to the load.
used a phase wide area measurement for the control. An MTDC link was not included.

6. Conclusions
The small-signal stability of two incoherent machines is one of the biggest problems
facing power utilities. Further problems also arise during the integration of renewable
energies, especially solar PV with high levels of intermittencies. Therefore, in this study,
a secondary control technique that utilises a hierarchical control structure to enhance the
damping of inter-area oscillation stability was introduced. The solution is multi-faceted,
consisting primarily of the local controllers, voltage and rotor speed regulator, a centralised

79
Energies 2023, 16, 6295

power controller for the LMTDC, and a current and extinction angle controller for the
LMTDC model. These joint controllers were added to maximally improve the system
performance. Firstly, a dynamic simulation model of an MKTAFM network was developed
(first scenario). In the second scenario, a generator disturbance at GenD and a 210 ms 3Ph-G
fault on the transmission line (L6) linking BB7 to BB8 were considered, with both faults
occurring at 2 s simulation time. In the third scenario, the system response consisting of
the details of the generators was monitored on a plot, and the times at which the system
experienced a highly imbalanced waveform were obtained, as well as the time taken for the
generator control to sufficiently dampen the system’s oscillations. Afterwards, a study case
involving the replacement of the double transmission lines on the network with an LMTDC
system (second scenario) was implemented. A secondary control was further added to the
LMTDC (third scenario). Finally, a nonlinear time-domain dynamics analysis was carried
out on the LMTDC network, and the generator and the bus voltage profile were monitored.
Following the 3Ph-G fault on the test network, the poorest damped modes in each grid
scenario being observable confirmed that the implementation of the secondary controller
on the LMTDC provided a 100% damping of the oscillation at exactly 5.8 s simulation time
with zero rates of change of the amplitudes compared to the first and second scenarios.
The same is seen in the inter-area power transfer between BB7 and BB9, which initially
recorded a continuous oscillation due to insufficient damping torque. The findings thus
show that the proposed novel controller is durable, significantly raises the stability limit of
the entire system, and improves the performance of primary controllers.
Further research will be conducted to reduce the sensitive response of the controller
to large and non-uniform transmission delays and its impact if added to other LMTDC
substations. Another point of focus is also to reduce the transient increase in the firing
angle of the LMTDC converters during faults.

Author Contributions: Conceptualization, O.E.O. and O.M.L.; methodology, O.E.O. and O.M.L.;
software, O.E.O.; validation, O.E.O. and O.M.L., formal analysis, O.E.O. and O.M.L.; investigation,
O.E.O. and O.M.L.; resources, O.E.O. and O.M.L.; data curation, O.E.O. and O.M.L.; writing—original
draft preparation, O.E.O. and O.M.L.; writing—review and editing, O.E.O. and O.M.L.; visualization,
O.E.O. and O.M.L.; supervision, O.M.L.; project administration, O.M.L.; funding acquisition, O.M.L.
All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Data Availability Statement: No data was created.
Conflicts of Interest: The authors declare no conflict of interest.

Appendix A

Table A1. Solar PV system specifications.

Parameter Value
Maximum power 250 W
Short-circuit current per cell 2.5 A
Current at maximum power point 2.0 A
Temperature coefficient of Is 0.001
Modules connected in series per 35
Modules connected parallel pers 11,000
Cells connected in series per module 35
Reference radiation 1010
Reference cell temperature 29
Diode Ideality factor (m) 1.5
PV system rating/system base 50 MW

80
Energies 2023, 16, 6295

References
1. Sitompul, S.; Fujita, G. In Impact of State-of-Charge Control Integrated with Load-Frequency Control on Battery Energy Storage
System in Islanded Microgrid System. In Proceedings of the 2021 IEEE 12th Energy Conversion Congress & Exposition-Asia
(ECCE-Asia), Singapore, 24–27 May 2021; pp. 199–203.
2. Klein, M.; Rogers, G.J.; Kundur, P. A fundamental study of inter-area oscillations in power systems. IEEE Trans. Power Syst. 1991,
6, 914–921. [CrossRef]
3. Machowski, J.; Lubosny, Z.; Bialek, J.W.; Bumby, J.R. Power System Dynamics: Stability and Control; John Wiley & Sons: Hoboken,
NJ, USA, 2020.
4. Ufa, R.; Rudnik, V.; Malkova, Y.; Bay, Y.; Kosmynina, N. Impact of renewable generation unit on stability of power systems. Int. J.
Hydrogen Energy 2022, 47, 19947–19954. [CrossRef]
5. Tina, G.M.; Maione, G.; Licciardello, S. Evaluation of Technical Solutions to Improve Transient Stability in Power Systems with
Wind Power Generation. Energies 2022, 15, 7055. [CrossRef]
6. He, X.; Geng, H. Transient stability of power systems integrated with inverter-based generation. IEEE Trans. Power Syst. 2020,
36, 553–556. [CrossRef]
7. Shair, J.; Li, H.; Hu, J.; Xie, X. Power system stability issues, classifications and research prospects in the context of high-penetration
of renewables and power electronics. Renew. Sustain. Energy Rev. 2021, 145, 111111. [CrossRef]
8. Huang, R.; Gao, W.; Fan, R.; Huang, Q. A guided evolutionary strategy based static var compensator control approach for
inter-area oscillation damping. IEEE Trans. Ind. Inform. 2022, 19, 2596–2607. [CrossRef]
9. Huang, R.; Gao, W.; Fan, R.; Huang, Q. Damping inter-area oscillation using reinforcement learning controlled TCSC. IET Gener.
Transm. Distrib. 2022, 16, 2265–2275. [CrossRef]
10. Hashmy, Y.; Yu, Z.; Shi, D.; Weng, Y. Wide-area measurement system-based low frequency oscillation damping control through
reinforcement learning. IEEE Trans. Smart Grid 2020, 11, 5072–5083. [CrossRef]
11. Saadatmand, M.; Gharehpetian, G.B.; Moghassemi, A.; Guerrero, J.M.; Siano, P.; Alhelou, H.H. Damping of low-frequency
oscillations in power systems by large-scale PV farms: A comprehensive review of control methods. IEEE Access 2021,
9, 72183–72206. [CrossRef]
12. Solomon, E.; Khan, B.; Boulkaibet, I.; Neji, B.; Khezami, N.; Ali, A.; Mahela, O.P.; Pascual Barrera, A.E. Mitigating Low-Frequency
Oscillations and Enhancing the Dynamic Stability of Power System Using Optimal Coordination of Power System Stabilizer and
Unified Power Flow Controller. Sustainability 2023, 15, 6980. [CrossRef]
13. Sreedivya, K.; Jeyanthy, P.A.; Devaraj, D. Improved design of interval type-2 fuzzy based wide area power system stabilizer for
inter-area oscillation damping. Microprocess. Microsyst. 2021, 83, 103957. [CrossRef]
14. Biswal, S.S.; Swain, D.R.; Rout, P.K. Inter-area and intra-area oscillation damping for UPFC in a multi-machine power system
based on tuned fractional PI controllers. Int. J. Dyn. Control 2022, 10, 1594–1612. [CrossRef]
15. Huang, L.; Coulson, J.; Lygeros, J.; Dörfler, F. Decentralized data-enabled predictive control for power system oscillation damping.
IEEE Trans. Control Syst. Technol. 2021, 30, 1065–1077. [CrossRef]
16. Prakash, A.; Kumar, K.; Parida, S. Energy capacitor system based wide-area damping controller for multiple inter-area modes.
IEEE Trans. Ind. Appl. 2022, 58, 1543–1553. [CrossRef]
17. Alalwani, S.; Isik, S.; Bhattacharya, S. In Inter-area Oscillation Damping Controller for DFIG based Wind Power Plants. In
Proceedings of the 2022 IEEE Energy Conversion Congress and Exposition (ECCE), Detroit, MI, USA, 9–13 October 2022; pp. 1–6.
18. Ahmed, F.; Foley, A.; McLoone, S.; Best, R.J.; Cameron, C.; Al Kez, D. Dynamic grid stability in low carbon power systems with
minimum inertia. Renew. Energy 2023, 210, 486–506. [CrossRef]
19. Eberlein, S. Small-Signal Stability Modelling and Optimization of Microgrids; BoD–Books on Demand: Norderstedt, Germany, 2021;
Volume 32.
20. Oni, O.E.; Swanson, A.G.; Carpanen, R.P. Small signal stability analysis of a four-machine system with placement of multi-terminal
high voltage direct current link. J. Energy S. Afr. 2020, 31, 73–87. [CrossRef]
21. Zou, X. Frequency and Damping Characteristics of Generators in Power Systems; Virginia Tech: Blacksburg, VA, USA, 2018.
22. Oni, O.E.; Swanson, A.G.; Carpanen, R.P. In Small signal stability analysis of a four machine system with strategic placement of
monopolar LCC-HVDC link. In Proceedings of the 2019 Southern African Universities Power Engineering Conference/Robotics
and Mechatronics/Pattern Recognition Association of South Africa (SAUPEC/RobMech/PRASA), Bloemfontein, South Africa,
29–31 January 2019; pp. 437–443.
23. Rajapakse, A.D.; Muthumuni, D. In Simulation tools for photovoltaic system grid integration studies. In Proceedings of the 2009
IEEE Electrical Power & Energy Conference (EPEC), Montreal, QC, Canada, 22–23 October 2009; pp. 1–5.
24. Liu, S.; Dougal, R.A. Dynamic multiphysics model for solar array. IEEE Trans. Energy Convers. 2002, 17, 285–294.
25. Owais, R.; Iqbal, S.J. An Intelligent Two-Level Control of Solar Photovoltaic Power Plant for Electromechanical Oscillation
Damping in Power Systems. Arab. J. Sci. Eng. 2023, 48, 6271–6292. [CrossRef]
26. Gupta, R.; Gupta, G.; Kastwar, D.; Hussain, A.; Ranjan, H. In Modeling and design of MPPT controller for a PV module using
PSCAD/EMTDC. In Proceedings of the 2010 IEEE PES Innovative Smart Grid Technologies Conference Europe (ISGT Europe),
Gothenburg, Sweden, 11–13 October 2010; pp. 1–6.
27. Faruque, M.; Zhang, Y.; Dinavahi, V. Detailed modeling of CIGRE HVDC benchmark system using PSCAD/EMTDC and
PSB/SIMULINK. IEEE Trans. Power Deliv. 2005, 21, 378–387. [CrossRef]

81
Energies 2023, 16, 6295

28. Oni, O.E.; Swanson, A.G.; Carpanen, R.P. In Modelling and control of multiterminal LCC HVDC. In Proceedings of the 2018 IEEE
PES/IAS PowerAfrica, Cape Town, South Africa, 28–29 June 2018; pp. 274–279.
29. Mankour, M.; Sami, B.S. Mitigation of commutation failure method in LCC converter based on HVDC systems by mean of
modeling and simulation. J. Ambient Intell. Humaniz. Comput. 2020, 14, 5837–5852. [CrossRef]
30. Peng, J.-H.; Nair, N.-K.; Maryani, A.; Ahmad, A. Adaptive Power System Stabilizer tuning technique for damping inter-area
oscillations. In Proceedings of the IEEE PES General Meeting, Minneapolis, MN, USA, 25–29 July 2010; pp. 1–6.
31. Yakout, A.H.; Attia, M.A. In Damping Inter-Area Oscillations Via Weighted Area Signals to PSSs Using TLBO & HS Algorithm.
In Proceedings of the 2019 21st International Middle East Power Systems Conference (MEPCON), Cairo, Egypt, 17–19 December
2019; pp. 496–501.
32. Preece, R.; Milanović, J.V.; Almutairi, A.M.; Marjanovic, O. Damping of inter-area oscillations in mixed AC/DC networks using
WAMS based supplementary controller. IEEE Trans. Power Syst. 2012, 28, 1160–1169. [CrossRef]
33. Patidar, N.; Kolhe, M.; Tripathy, N.; Sahu, B.; Sharma, A.; Nagar, L.; Azmi, A. In Optimized design of wide-area PSS for damping
of inter-area oscillations. In Proceedings of the 2015 IEEE 11th International Conference on Power Electronics and Drive Systems,
Sydney, Australia, 9–12 June 2015; pp. 1172–1177.
34. Bolzoni, A.; Johansson, N.; Hasler, J. In Modeling the impact of grid-forming E-STATCOMs on inter-area system oscillations. In
Proceedings of the 2022 24th European Conference on Power Electronics and Applications (EPE’22 ECCE Europe), Hanover,
Germany, 5–9 September 2022; pp. 1–10.
35. Fadhil, S.T.; Vural, A.M. In Comparison of dynamic performances of TCSC, Statcom, SSSC on inter-area oscillations. In
Proceedings of the 2018 5th International Conference on Electrical and Electronic Engineering (ICEEE), Istanbul, Turkey, 3–5 May
2018; pp. 138–142.
36. Fayez, M.; Mandor, M.; El-Hadidy, M.; Bendary, F. Fuzzy logic based dynamic braking scheme for stabilization of inter-area. Acta
Electron. Malays. (AEM) 2019, 3, 16–22. [CrossRef]
37. Fayez, M.; Mandor, M.; El-Hadidy, M.; Bendary, F. Stabilization of inter-area oscillations in two-area test system via centralized
interval type-2 fuzzy-based dynamic brake control. J. Electr. Syst. Inf. Technol. 2021, 8, 1–19. [CrossRef]
38. Selim, F.; Attia, A.-F. In Power System Stabilizer with Self-Tuning Based on Hierarchical Fuzzy Logic Controller. In Proceedings
of the 2022 23rd International Middle East Power Systems Conference (MEPCON), Cairo, Egypt, 13–15 December 2022; pp. 1–6.
39. Sharafi, A.; Vahidnia, A.; Jalili, M. In Application of Consensus Algorithm in Clustering of Distributed Generation Systems for
Wide-Area Management and Control. In Proceedings of the 2023 IEEE International Conference on Environment and Electrical
Engineering and 2023 IEEE Industrial and Commercial Power Systems Europe (EEEIC/I&CPS Europe), Madrid, Spain, 6–9 June
2023; pp. 1–6.
40. Sharafi, A.; Vahidnia, A.; Jalili, M. In Wide-Area Stabilizing Control using Distributed Generation Systems. In Proceedings of the
2023 11th International Conference on Smart Grid (icSmartGrid), Paris, France, 5–7 June 2023; pp. 1–6.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

82
 energies
Article
Long-Term Voltage Stability Bifurcation Analysis and Control
Considering OLTC Adjustment and Photovoltaic Power Station
Sheng Li *, Can Zhang and Jili Zuo

School of Electric Power Engineering, Nanjing Institute of Technology, Nanjing 211167, China;
zhangcan8211@163.com (C.Z.); jilizuo0324@163.com (J.Z.)
* Correspondence: lisheng_njit@126.com

Abstract: The influence of photovoltaic (PV) output with stochasticity and uncertainty on the grid-
connected system’s voltage stability is worth further exploration. The long-term voltage stability
of a 3-bus system with a large-scale PV power station considering the adjustment of an on-load tap
changer (OLTC) was studied. In this typical system, two supercritical Hopf bifurcation (SHB) points
are found using the bifurcation calculation. At the SHB point that appears first, a small sudden
increase in reactive load power or a sudden increase in PV active power Ppv can eventually cause
a voltage collapse after a long increasing oscillation. The long-term collapse phenomenon shows
that SHB cannot be ignored in the PV grid-connected system. Meanwhile, the time constant of
OLTC can affect the progress of long-term voltage collapse, but it has different effects under different
disturbances. When Ppv drops suddenly at the SHB point, due to the adjustment of OLTC, the load
bus voltage can recover to near the target value of OLTC after a long period of time. Similarly, the
time constant of OLTC can affect the progress of long-term voltage recovery. To prevent the long-term
voltage collapse when Ppv increases suddenly at the SHB point, a new locking-OLTC index Ilock ,
depending on the value of Ppv corresponding to the SHB point, and a locking OLTC method are
proposed, and the voltage can be recovered to an acceptable stable value quickly. Compared with
the system without OLTC, OLTC adjustment can effectively prevent long-term voltage oscillation
instability and collapse, so that PV power can play a bigger role in power systems.

Keywords: photovoltaic (PV); long-term voltage stability; voltage collapse; OLTC; supercritical Hopf
Citation: Li, S.; Zhang, C.; Zuo, J. bifurcation (SHB)
Long-Term Voltage Stability
Bifurcation Analysis and Control
Considering OLTC Adjustment and
Photovoltaic Power Station. Energies
1. Introduction
2023, 16, 6383. https://doi.org/
10.3390/en16176383
From the perspective of protecting the environment and reducing smog pollution, the
development of renewable energy and clean energy generation has become a mainstream
Academic Editor: Ying-Yi Hong
trend, and vigorously developing the new energy is an important strategy for achieving the
Received: 29 June 2023 goal of “carbon peaking and carbon neutrality” [1,2]. Photovoltaic (PV) power generation
Revised: 29 August 2023 is an important method of solving the problem of carbon emissions [3]. At present, PV
Accepted: 1 September 2023 power generation has been vigorously promoted and widely used around the world.
Published: 3 September 2023 PV power generation has the characteristics of volatility and intermittency. When
large-scale PV stations are connected to the power grid, it will inevitably have an adverse
impact on the stability of the power system, increasing the complexity and uncertainty of
grid operation [4,5]. Therefore, the influence of PV output on the grid-connected system’s
Copyright: © 2023 by the authors. voltage stability must be studied [6–9].
Licensee MDPI, Basel, Switzerland. Long-term voltage stability belongs to large-disturbance voltage stability. The large
This article is an open access article
disturbances in the power system refer to a short circuit, a disconnection fault, or a large-
distributed under the terms and
scale new energy off-grid, etc. The large-disturbance voltage stability includes two aspects:
conditions of the Creative Commons
the transient process and the long-term process [10]. The time frame for the transient
Attribution (CC BY) license (https://
voltage stability investigations is usually 10–15 s, while the time frame for the long-term
creativecommons.org/licenses/by/
voltage stability investigations is usually from a few minutes to tens of minutes with
4.0/).

Energies 2023, 16, 6383. https://doi.org/10.3390/en16176383 83 https://www.mdpi.com/journal/energies

Energies 2023, 16, 6383

the role of the slow-regulating devices, such as the on-load tap changer (OLTC) and the
over-excitation limiter (OEL), etc.
Currently, several studies on long-term voltage stability have been reported. The
literature [11] proposed a remedial measure system to prevent medium-term and long-
term voltage instabilities in power systems. When the operating point deviates from the
normal operating state, the system will select a required procedure, power generation
rescheduling or load shedding, based on the concept of electrical distance. According to
the selected procedure, the system implements the corresponding remedial measure to
bring the operating point back to a normal operating state. The literature [12] proposed
a coordinated decentralized emergency control strategy for long-term voltage instability
by dividing the power system into multiple local areas based on the concept of electrical
distance; and designing a performance indicator based on the load bus voltage and the
generator reactive power to assess the severity of disturbances and emergency risks in each
local area.
The literature [13] proposed the concepts of long-term voltage stability PDR (Reactive
Power Reserve) and short-term voltage stability PDR and established a dual-objective
reactive power reserve optimization model to coordinate long-term and short-term voltage
stability simultaneously. The literature [14] proposed a long-term voltage stability index
that can identify the critical load power of voltage collapse, and a transmission-distribution
distinguishing index (TDDI) is also proposed to identify whether the voltage stability
limit is due to the transmission or distribution network. Both VSI-index and TDDI-index
indicators are calculated based on the measured value of the phasor measurement unit,
which is convenient for online applications. The literature [15] designed a long-term
voltage stability monitoring index based on the measured value of voltage amplitude
and the calculation value of the Thevenin equivalent angle using a voltage trajectory
method. The literature [16] studied the coordinated scheduling problem of gas-electricity
integrated energy systems including wind power, taking the long-term voltage stability
as a constraint condition.
In recent years, machine learning and data mining technologies have developed
rapidly, and machine learning algorithms have also been applied to long-term voltage
stability research. The literature [17] used the random forest algorithm to predict the load
margin to monitor long-term voltage stability in real-time, and a variety of different voltage
stability indexes were used as the input variables of the machine learning integration model.
The literature [18] proposed a data-based learning and control method based on offline
knowledge accumulation and feature acquisition to solve the long-term voltage stability
problem caused by emergency online events. The literature [19] proposed an auto-encoder
constructed by long short-term memory networks combined with a fully connected layer,
which only needs to train the data from a safe operating state to evaluate the long-term
voltage stability of power systems.
OLTC is a kind of typical slow-adjustment device and it completes one-tap adjustments
with a required time of 10 s to 100 s. When studying long-term voltage stability, the
dynamic adjustment process of OLTC cannot be neglected. On the other hand, the dynamic
load also has a greater impact on voltage stability, and the adjustment effect of OLTC
on voltage stability is related to the dynamic load characteristics [20]. Generally, when
the system has sufficient reactive power, OLTC adjustment can contribute to the system’s
voltage stability; if the system reactive power is insufficient, OLTC will have a negative
voltage adjustment effect [21].
The literature [22] studied the impact of a variable-speed wind turbine on long-term
voltage stability, considering the actions of OEL and OLTC. When the doubly fed induction
generator adopts the grid-side converter to control the reactive power, the long-term voltage
stability can be effectively improved. The literature [23] took the Nordic 32-bus test system
including AVR (Automatic Voltage Regulator), OEL, and OLTC as an example to investigate
the impact of PV power generation on long-term voltage stability.

84
Energies 2023, 16, 6383

With the large-scale access to new energy power stations, the impact of wind power
and PV power generation on long-term voltage stability has also begun to receive attention.
In a large-scale PV grid-connected system, it is worth discussing the mechanisms of sudden
changes in PV power (including sudden increases and sudden drops) on long-term voltage
instability, especially under extreme weather conditions. The impacts of OLTC adjustment
and dynamic load characteristics on long-term voltage stability under sudden PV power
changes also deserve further study.
Because bifurcation parameters correspond to uncontrollable parameters such as load
power and PV power variations, bifurcation theory and its analysis method are suitable for
studying voltage stability in the PV grid-connected system. The literature [24] analyzed
in detail the bifurcation behaviors of a typical 3-bus system with a PV station. When
the bifurcation parameter changes gradually, once the behavior of the grid-connected
system changes, the system will bifurcate, which may be saddle-node bifurcation (SNB)
or other forms of bifurcation, such as Hopf bifurcation (HB), including supercritical Hopf
bifurcation (SHB) and subcritical Hopf bifurcation (alternatively termed unstable Hopf
bifurcation (UHB)), etc. These bifurcations may have adverse impacts on the system’s
voltage stability. For example, at the UHB point, a small increase in load power or a sudden
change in PV active power can cause a long-term voltage collapse or a long-term voltage
oscillation [24,25]. In the literature [25], an index-based predictive control approach for
UHB was proposed. However, the impact of OLTC adjustment is not considered in the
literature [24,25]. This may cause a one-sided bifurcation analysis result, as well as a poor
understanding of the mechanism of long-term voltage instability.
The biggest factor affecting PV active power is the intensity of solar light. When
encountering extreme weather conditions, such as solar eclipses, snowstorms, and sand-
storms, etc., it is highly likely that a sudden and significant drop or even complete loss
of PV power will occur, but whether it leads to long-term voltage instability should be
further investigated.
This paper mainly investigates, in detail, the long-term voltage instability mechanism
of the PV grid-connected system, considering OLTC adjustments when the PV output has
a sudden change by using bifurcation theory, explores OLTC’s key influencing factors on
long-term voltage instability, and considers the influence of load dynamic characteristics at
the same time. In addition, this paper attempts to design a novel long-term voltage stability
index that can predict the long-term voltage instability phenomena, so as to provide an
adaptive control scheme.

2. Long-Term Voltage Stability Bifurcation Analysis Based on Matcont

2.1. Brief Introduction of the Hopf Bifurcation Theory
The PV grid-connected system can be described by the single-parameter family form
of ordinary differential equations (ODEs).
.
x = f ( x, μ) (1)

where x represents the state variables; μ represents the bifurcation parameter such as the
load power or the PV power, etc., and f is a continuously differentiable function.
The eigenvalues of the state matrix A of Equation (1) at the system’s equilibrium point
(x0 , μ) can describe the grid-connected system’s dynamic stability. With the gradual change
of μ, if a pair of conjugate eigenvalues α(μ) ± jβ(μ) of A pass through the imaginary axis in
the complex plane, and satisfy the next equation at μ = μ0 [26]:
⎧
⎨ α ( μ0 ) = 0
(2)
⎩ dα(μ0 ) = 0
dμ

Then a HB happens at the equilibrium point (x0 , μ0 ) in the system, and the point
corresponding to the pure imaginary eigenvalues ±jβ(μ0 ) is namely the HB point [26].

85
Energies 2023, 16, 6383

According to the stability of the limit cycle arising from the HB point, HB can be divided
into SHB (the corresponding first Lyapunov coefficient (FLC) is less than 0) and UHB (the
corresponding FLC is greater than 0) [27].

2.2. Model of the 3-Bus System with OLTC and PV

As shown in Figure 1a, the traditional 3-bus system (not including the OLTC and PV
stations), has been widely used in traditional voltage stability bifurcation analysis [28,29].
Now a PV station and an OLTC station are introduced into the system. n is the transforma-
tion ratio of OLTC, in general, 0.8 pu ≤ n ≤ 1.2 pu. The PV station is connected to the load
bus to investigate the impact of PV power on the load bus’s voltage stability. For facilitating
the bifurcation calculation, the PV transmission line is omitted. Ppv + jQpv is the PV power.
Em ∠δm is the equivalent generator potential, E0 ∠0 is the infinite bus voltage, and V ∠δ is
the load bus voltage. Ym and Y0 are the grid equivalent admittance modulus, θ m and θ 0
are the corresponding admittance angles, and Ym includes the equivalent admittance of
OLTC. The load PD + jQD adopts the Walve dynamic load model [28,29]. The load P1 + jQ1
is a constant power load. C is the parallel capacitor bank. In the traditional 3-bus system
without OLTC and PV, C is used to raise the load bus voltage. Due to the intermittency of
Ppv , this setting is retained in this case.

EP ∠δ P V ∠δ E ∠
ʌ n ʌ
YP ∠ − θ P − Y ∠ − θ  −
 

2/7&
(TXLYDOHQWJHQHUDWRU ,QILQLWHEXVV\VWHP
PSY + MQSY
C
&DSDFLWRUEDQNV
39VWDWLRQ
P + MQ P' + MQ'
(a)
nEP ∠δ P Y V ∠δ E  ∠
ʌ ʌ
P
∠ − θP − Y  ∠ −θ  −
n  

P + MQ
(TXLYDOHQWJHQHUDWRU ,QILQLWHEXVV\VWHP
PSY + MQSY

39VWDWLRQ
P + MQ P' + MQ'
(b)
Figure 1. The 3-bus system with OLTC and PV. (a) The initial model. (b) The simplified model.

To simplify Figure 1a, the generator branch is converted to the load bus side, and
the capacitor banks and the infinite system branch are subjected to a Thevenin equivalent
transformation, as shown in Figure 1b.
The inverter of the PV station adopts a 1-order dynamic model [30], and its dynamic
characteristic equation is as follows [31]:
⎧.  
⎪
⎨id = T1p V1 ( Qpv cos δ − Ppv sin δ) − id
.   (3)
⎪
⎩ iq = 1 1 ( Ppv cos δ + Qpv sin δ) − iq
Tq V

86
Energies 2023, 16, 6383

where id , iq are the d-axis and q-axis components of the PV inverter’s output current, and
Tp , Tq are the responsive time constants of the PV inverter.
The PV station is set as a PQ-type station operating in a constant power factor mode [9],
so we can assume:
Qpv = kPpv (4)
where k is the tangent of the PV station’s power factor angle. For example, if k is 0.2, the PV
power factor is 0.98.
The discrete model of OLTC is as follows [32]:

nk+1 − nk = d f (Vref − V ) (5)

where d is the tap step length, and V ref is the load bus reference voltage (namely the
adjustment target value of OLTC).
The function f (V ref − V) can be expressed as follows:
⎧
⎨ +1 Vref − V ≥ ΔV
f (Vref − V ) = 0 |Vref − V | ≤ ΔV (6)
⎩
−1 Vref − V ≤ −ΔV

where ΔV is the adjustment dead zone.

In order to facilitate the bifurcation calculation with the numerical bifurcation analysis
software, in this paper, OLTC adopts a continuous adjustment dynamic model, ignoring
the adjustment dead zone and the tap time delay [33]:

. 1
n= (V − V ) (7)
Tn ref

where Tn is the time constant of OLTC (s). To approach the discrete model, Tn can be
expressed as Tn = τ/d, τ is the tap time delay.
To simplify the analysis process, the equivalent generator uses the 2-order model:
' .
δm = ω
. (8)
ω= M 1
( Pm − PG − Dω )

where δm and ω are the generator power angle and the rotor angular velocity, respectively,
M is the inertia, D is the damping coefficient, and Pm and PG are the mechanical power
and output electric power of the generator, respectively.
Derived from Figure 1b, the output active power of the generator can be given below:

Em VYm
PG = − Em
2
Ym sin θm − sin(δ − δm − θm ) (9)
n
The power supplied by the grid to all loads is:
'
P = − E0 VY0 sin(δ + θ0 ) − Em VY
n
m
sin(δ − δm + θm ) + (Y0 sin θ0 + Ynm2 sin θm )V 2
(10)
Q = E0 VY0 cos(δ + θ0 ) + n cos(δ − δm + θm ) − (Y0 cos θ0 + Ynm2 cos θm )V 2
Em VYm

The dynamic load model adopts the Walve load model, then the ODEs of the system
shown in Figure 1 can be given below:

87
Energies 2023, 16, 6383

⎧ .
⎪
⎪ δm = ω
⎪
⎪ .
⎪
⎪ ω= M 1
( Pm − PG − Dω )
⎪
⎪
⎪
⎪ .  
⎪ = k qω − k qV V − k qV2 V + Q − Q0 − Q1 + kPpv
1
⎪
⎪ δ 2
⎪
⎪   
⎪
⎪
⎨ . k pω k qV2 V 2 + k pω k qV − k qω k pV V
V = Tkqω1 k    ( (11)
⎪ pV +k pω Q0 + Q1 − Q − kPpv −k qω P0 + P1 − P − Ppv
⎪
⎪
⎪ n = 1 (V − V )
⎪
.
⎪
⎪ Tn ref
⎪
⎪  
⎪
⎪
.
1 Ppv
⎪ id = Tp V (k cos δ − sin δ) − id
⎪
⎪
⎪  
⎪
⎪ .
⎩ iq = 1 Ppv (cos δ + k sin δ) − iq
Tq V

where T is the dynamic load time constant (s), kpω , kqω , kpV , kqV, and kqV2 are the load
coefficients of the Walve load, and P0 and Q0 are the constant powers of the Walve load.
The values of the power supply and network parameters are shown in Table 1, and the
load parameters are shown in Table 2 (part of the data from the literature [24,29]). In this
paper, all electrical quantities are in per-unit value and the phase angle’s unit is rad, the
time’s unit is s. We set τ = 10 s and d = 0.625%, then Tn = τ/d = 1600 s, as listed in Table 1.

Table 1. The power supply and network parameters.

Em /pu Pm /pu M/pu D/pu E0 /pu Y m /pu θm /rad Y 0 /pu θ0 /rad C/pu Tn /s V ref /pu Sbase
1 1 0.3 0.05 1 5 −0.0872 20 −0.0872 12 1600 1 100 MVA

Table 2. The load and PV parameters.

T/s kpω kqω kpV kqV kqV2 P0 /pu Q0 /pu P1 /pu Tp /s Tq /s k

8.5 0.4 −0.03 0.3 −2.8 2.1 0.6 1.3 1.2 10 10 0.2

2.3. The Equilibrium Point Curves

Keep Ppv = 1 pu, and take Q1 as the bifurcation parameter, the numerical bifurcation
calculation software Matcont (Version 6.2, A. Dhooge, W. Govaerts, etc., Universiteit Gent,
Belgium; Utrecht University, The Netherlands) was used to calculate the equilibrium point
curves and bifurcation points of the ODEs shown in Equation (11). We can obtain the
equilibrium point curves of the system, as shown in Figure 2. In the Q1 –n curve shown in
Figure 2a, the lower half of the curve is a stable equilibrium point curve. Before Q1 reaches
the SNB point named “LP” (corresponding to Q1 = 12.6347 pu), two Hopf bifurcation points
appear: H1 (corresponding to Q1 = 10.3418 pu) and H2 (corresponding to Q1 = 11.9544 pu),
FLC’s values are both less than 0 (FLCH1 = −0.00876 and FLCH2 = −2.21), so the two Hopf
bifurcation points are both SHB points.
Now keep Q1 = 10.3418 pu and take Ppv as the bifurcation parameter. The Ppv –n
curve is shown in Figure 2b, and the lower half of the curve is the stable equilibrium point
curve. H1 (corresponding to Ppv = 1 pu) and H’2 (corresponding to Ppv = 16.7645 pu) are
both SHB points. The installed capacity of a large-scale PV station can reach several pu
(namely several hundred MW) currently, so it is possible that the system operates at the
SHB point H1 , and the impact of this SHB point on the system’s long-term voltage stability
is noteworthy.

88
Energies 2023, 16, 6383




6+%


nSX

 /3 61%
+ 6+%
+ 6+% 



   
Q SX


(a)


 6+%


nSX


1HXWUDOVDGGOH
1HXWUDOVDGGOH


/3 61%
/3

+  6+% 
+ 6+% 1HXWUDOVDGGOH



          

P SX
SY

(b)

Figure 2. The equilibrium point curves. (a) Q1 –n curve (Ppv = 1 pu). (b) Ppv –n curve (Q1 = 10.3418 pu).

2.4. Impact of Load Power Disturbance on Long-Term Voltage Stability

2.4.1. Sudden Increase in the Reactive Load Q1 —Long-Term Voltage Collapse
Take the SHB point H1 as the initial operating point, and the parameters of this
bifurcation point are: (δm ω δ V n id iq Q1 ) = (0.3057 0 0.1227 1 0.8478 0.0761 1.0170 10.3418).
Assume that Q1 has a forward small disturbance (a small increase) at time t = 50 s suddenly,
and it increases from 10.3418 pu to 10.3918 pu. That is, the increment ΔQ1 is 0.05 pu. The
time-domain change curves are shown in Figure 3. The load bus voltage V eventually
collapses at about t = 975 s after a long increasing oscillation, as shown in Figure 3a.

89
Energies 2023, 16, 6383





ΔQ SX ΔQ SX

VSX

 ΔQ SX

ΔQ SX



ΔQ SX




          
tV

(a)









nSX












          
tV

(b)
Figure 3. The time-domain change curves when Q1 increases suddenly from the SHB point H1
(Tn = 1600 s). (a) t–V curves at different increments of Q1 . (b) t–n curve (ΔQ1 = 0.05 pu).

Figure 3b is the time-domain change curve of the OLTC transformation ratio n. After
the disturbance occurs, the transformation ratio n keeps increasing to increase V, but the
voltage collapse is unavoidable, and n no longer increases.
Now assume that starting from the SHB point H1 , the load power increment ΔQ1
is 0.1 pu~0.4 pu, respectively. It can be seen from Figure 3a that the larger ΔQ1 is, the
faster V collapses, and the time frame becomes smaller as the long-term voltage collapse
gradually evolves into the short-term voltage collapse. Therefore, it is recommended that
the maximum value of Q1 should not exceed the value of Q1 corresponding to the SHB
point H1 .
In general, it is believed that if there is a UHB point in power systems, it will
pose a threat to the voltage stability and lead to oscillation-type voltage instability and
collapse [24,25]. For this example, it can be seen from the above analyses that at the SHB
point, a forward small or large disturbance of the load power can also cause long-term

90
Energies 2023, 16, 6383

voltage oscillation instability and collapse. This phenomenon regarding SHB deserves
more attention.

2.4.2. Impact of OLTC’s Time Constant—Key Influencing Factor Investigation

The impact of OLTC’s time constant Tn is investigated on long-term voltage collapse
when the system operates at the SHB point H1 . Changing the tap step length d or the tap
time delay τ can change the time constant of OLTC. Now set τ = 10 s and d = 1.25%, then
Tn = τ/d = 800 s. Set ΔQ1 = 0.05 pu, the time to reach the voltage collapse increases, as
shown in Figure 4. Moreover, set τ = 20 s and d = 0.625%, Tn = τ/d = 3200 s, it can be
seen from Figure 4 that the time taken to reach voltage collapse gets shorter. Therefore,
increasing the time constant Tn can shorten the process of the long-term voltage collapse,
that is, a larger Tn is not conducive to voltage stability.





Tn V Tn V

VSX

 Tn V






   
tV

Figure 4. t–V curves at different time constants of OLTC when Q1 increases suddenly from the SHB
point H1 (ΔQ1 = 0.05 pu).

2.5. Impact of PV Power Disturbance on Long-Term Voltage Stability

2.5.1. Sudden Drop of PV Power—Long-Term Voltage Recovery
Assume that the system shown in Figure 1 is operating at the SHB point H1 , the
initial Ppv is 1 pu, and Q1 is maintained at 10.3418 pu. Considering extremely bad weather
conditions, such as a total solar eclipse, at time t = 50 s, Ppv suddenly drops from 1 pu
to 0 pu, 0.5 pu, and 0.8 pu, respectively. The drop depth dPpv is 1 pu, 0.5 pu, and 0.2 pu,
respectively. It can be seen from Figure 5a that the load bus voltage V drops after the
disturbance occurs, but an OLTC adjustment can finally recover V to near the target value
V ref (1 pu) after a long period of time. As shown in Figure 5b, the drop depth dPpv is 1 pu
and it is big enough. When the disturbance occurs, the transformation ratio n continues to
increase to raise the load bus voltage and eventually stabilizes at a value of 0.884 pu after a
very long time.
It can be determined by calculation that when Ppv drops suddenly, OLTC can all adjust
V to near the target value, and V can be recovered faster under the smaller drop depth dPpv .

2.5.2. Sudden Increase in PV Power—Long-Term Voltage Collapse

Now assume that the system shown in Figure 1 is operating at the SHB point H1 , Ppv
increases suddenly from 1 pu to 1.1 pu, 1.3 pu, and 1.5 pu, respectively. The increment
ΔPpv is 0.1 pu, 0.3 pu, and 0.5 pu, respectively. It can be seen from Figure 6 that the voltage
V collapses after a long period of time. The larger ΔPpv is, the faster V collapses, but the
time frame is still in the category of long-term voltage stability.

91
Energies 2023, 16, 6383



VUHI SX
GPSY SX


VSX

GPSY SX



GPSY SX


          
tV

(a)





V UHI  SX
ΔPSY  SX



nSX




Δ PSY  SX


ΔPSY  SX






          
tV

(b)
Figure 5. The time-domain change curves when Ppv drops suddenly from the SHB point H1
(Tn = 1600 s). (a) t–V curves at different drop depths of Ppv . (b) t–n curve (dPpv = 1 pu).

According to the above analyses, in the absence of any control measures taken, Ppv is
preferably not greater than the PV active power corresponding to the SHB point H1 (Ppv.H1 ).
In practice, Ppv is generally less than the PV-installed capacity PpvN , hence a solution is to
ensure that PpvN ≤ Ppv.H1 . However, this solution limits the use of PV power generation.

2.5.3. Impact of OLTC’s Time Constant—Key Influencing Factor Investigation

Different time constants Tn can make different impacts on long-term voltage stability
when Ppv has a sudden change. In Figure 7, when Ppv drops from 1 pu to 0 pu, a smaller
Tn is conducive to the recovery of the voltage V. However, in Figure 8, when Ppv increases
from 1 pu to 1.3 pu, a smaller Tn is not conducive to voltage stability.

92
Energies 2023, 16, 6383





ΔPSY SX
ΔPSY SX
VSX



ΔPSY SX






      
tV

Figure 6. t–V curves when Ppv increases suddenly from the SHB point H1 (Tn = 1600 s).



VUHI SX

Tn V
Tn V
VSX



Tn V


          
tV

Figure 7. t–V curves at different time constants of OLTC when Ppv drops suddenly from the SHB
point H1 (dPpv = 1 pu).





Tn V
VSX



Tn V Tn V






   
tV

Figure 8. t–V curves at different time constants of OLTC when Ppv increases suddenly from the SHB
point H1 (ΔPpv = 0.3 pu).

93
Energies 2023, 16, 6383

3. Comparison with the PV Grid-Connected 3-Bus System without OLTC

As shown in Figure 1, the PV grid-connected 3-bus system without OLTC is equivalent
to the system with OLTC being locked in n = 1 pu, and the system’s ODEs take the first
four equations of Equation (11) [24]. Keeping Ppv = 1 pu, the bifurcation calculation results
show that on the stable half of the Q1 –V equilibrium point curve, there is a UHB point H1
(corresponding to Q1 = 11.1309 pu), a SHB point H2 (corresponding to Q1 = 11.4999 pu),
and a SNB point LP (corresponding to Q1 = 11.5069 pu) [24].
Taking the UHB point H1 as the initial operating point, since it is a subcritical Hopf
bifurcation point, a small sudden increase in the reactive load Q1 can cause long-term
voltage collapse. As shown in Figure 9, assuming that Ppv is maintained at 1 pu, Q1
increases suddenly from 11.1309 pu to 11.1809 pu at time t = 50 s (namely ΔQ1 = 0.05 pu),
and the load bus voltage V collapses at about t = 708 s.



VSX








          
tV

Figure 9. t–V curve when Q1 increases suddenly from the UHB point H1 (ΔQ1 = 0.05 pu).

Compared with Figure 3a, due to the lack of OLTC adjustment, the time to reach voltage
collapse has been reduced, but it still belongs in the category of long-term voltage stability.
Assume that Q1 remains unchanged (keeping 11.1309 pu), and the impact of the
sudden change of Ppv on long-term voltage stability is investigated. As shown in Figure 10,
Ppv drops suddenly from 1 pu to 0 pu (namely dPpv = 1 pu) at time t = 50 s. After V drops,
the increasing oscillation is carried out firstly and the continuous oscillation (namely the
constant-amplitude oscillation) begins at about t = 350 s after the disturbance occurs, and
the oscillation amplitude of V is (0.9942 − 0.9344) × 100% = 5.98% > 5%. For the continuous
oscillation, if the voltage’s oscillation amplitude is greater than or equal to 5%, it can be
considered that the load bus is experiencing long-term voltage oscillation instability.
The smaller the drop depth dPpv is, the longer the time for the voltage V to begin the
continuous oscillation is, but the larger the oscillation amplitude of V will be [24]. For exam-
ple, when Ppv suddenly drops from 1 pu to 0.8 pu (namely dPpv = 0.2 pu), the time for the
load voltage V to begin the continuous oscillation is about t = 4800 s after the disturbance oc-
curs, and the oscillation amplitude of V reaches (1.072 − 0.8562) × 100% = 21.58% >> 5%.
On the other hand, as shown in Figure 10, when Ppv suddenly increases from 1 pu to
1.5 pu (namely ΔPpv = 0.5 pu), after the voltage V increases, it quickly enters a continuous
oscillation, and the oscillation amplitude is (1.105 − 1.099) × 100% = 0.6%, far less than 5%,
which does not pose a threat to voltage stability.

94
Energies 2023, 16, 6383



ΔPSY SX




VSX

GPSY SX




          
tV

Figure 10. t–V curves when Ppv changes suddenly from the UHB point H1 (dPpv = 1 pu and
ΔPpv = 0.5 pu).

The operation comparison results of the system with OLTC and without OLTC are
listed in Table 3. According to the comparative analyses, in order to avoid the long-term
voltage oscillation instability when Ppv has a sudden drop, it is advisable to adopt the
scheme of installing OLTC. For the sudden increase in Ppv , measures such as limiting the in-
stalled capacity of the PV station or locking OLTC can be used to solve the potential threat.

Table 3. Comparison of the PV grid-connected 3-bus system with OLTC and without OLTC, taking
the HB point H1 as the initial operating point.

The Content of Comparison With OLTC Without OLTC

Bifurcation type of point H1 SHB UHB
Reactive load Q1 of H1 (pu) 10.3418 pu 11.1309 pu
Voltage amplitude of H1 (pu) 1 pu (V ref ) 1.0744 pu
Sudden small increase of Q1 Long-term voltage collapse Long-term voltage collapse
Long-term voltage oscillation instability
Sudden drop of Ppv Long-term voltage recovery
(Oscillation amplitude ≥ 5%)
Long-term voltage oscillation without
Sudden increase of Ppv Long-term voltage collapse
instability (Oscillation amplitude << 5%)

4. Design of the Index and Control Methods for Preventing Long-Term

Voltage Collapse
4.1. Prevention of Voltage Collapse Caused by a Sudden Increase in PV Power
For the 3-bus system with OLTC and PV shown in Figure 1, the installed capacity PpvN
can be limited according to the value of Ppv at the SHB point H1 (Ppv.H1 ) so that Ppv does
not exceed Ppv.H1 . However, in practice, PpvN may be much larger than Ppv.H1 , so limiting
the installed capacity is not an economic method.
In Figure 10 and Table 3, at the UHB point, when Ppv increases suddenly from 1 pu to
1.5 pu (ΔPpv = 0.5 pu), long-term voltage oscillation with minimal amplitude will occur
in the system without OLTC. Therefore, the locking OLTC method (namely maintaining a
constant transformation ratio) can be used when Ppv increases suddenly in the system with
OLTC, in order to prevent long-term voltage collapse.
For a better implementation of this locking OLTC method, a locking-OLTC index
is designed:

95
Energies 2023, 16, 6383

Ppv
Ilock = 1 − ≤ 5 ∼ 10% (12)
Ppv.H1
Under the premise of a constant Q1 , where the value of the Ilock index is less than or
equal to 0, OLTC should be locked in time. To keep a certain margin, the threshold can
even be raised to 5~10%.
For example, when the system with OLTC operates at the SHB point H1 , assuming
that Q1 is maintained at Q1.H1 and Ppv fluctuates, once the Ilock index value is less than or
equal to 5~10%, the locking OLTC method can be used to avoid long-term voltage collapse.
Assume that the system with OLTC operates at the SHB point H1 , and Q1 is maintained
at 10.3418 pu. When Ppv increases suddenly from 1 pu to 1.1 pu, 1.3 pu, and 1.5 pu,
respectively, Ilock is −0.1, −0.3, and −0.5, respectively. Now OLTC is locked in n = 0.9 pu,
it can be seen from Figure 11 that V can quickly recover to an acceptable stable value.




ΔPSY SX
ΔPSY SX



ΔPSY SX
VSX






          
tV

Figure 11. t–V curves when Ppv changes suddenly from the SHB point H1 (OLTC is locked in n = 0.9 pu).

The flow chart of the locking OLTC method is shown in Figure 12. Since the voltage
collapse caused by the sudden increase in Ppv is a long process (up to hundreds of seconds
or more), the method of locking OLTC is completely feasible.
Considering the fluctuation of Ppv , the Ilock index should be calculated and judged
twice or even more consecutively, with a delay of 0.5~1 min between the two consecu-
tive calculations and judgments. Once the two (or more) consecutive judgment results
are the same, the command of locking OLTC should be issued, and n will be locked
at a constant value.

4.2. Prevention of Voltage Collapse Caused by a Sudden Increase in Load Power

The traditional load-margin index [7] can be modified as follows:

Q1
ILMHB = 1 − (13)
Q1.H1

where Q1.H1 is the reactive power of the SHB point H1 .

Once the ILMHB index value is less than 0, the load-shedding method can be taken to
improve the load margin. A detailed load-shedding scheme is referred to in the literature [7].
Since the impact of the sudden increase in Ppv at the SHB point can be solved by using the
locking OLTC method, the value of ILMHB can be 0.
Load shedding is certainly a conservative method, and it may also cause a small-
amplitude oscillation in the load bus voltage.

96
Energies 2023, 16, 6383

Start

Calculate
Ipv.H1

Sample Ppv

Calculate
Ilock

N
Ilock ǂ 5~10% ?

Y
Ilock.t Ă Ilock.t+i
ǂ 5~10%?

Delay 0.5~1 min

Lock OLTC

Figure 12. The flow chart of the locking OLTC method (t and i are the sampling time and times).

5. Conclusions
The typical 3-bus system containing an OLTC and a PV station is investigated for
long-term voltage stability using bifurcation theory, and a new index and a locking OLTC
method are used to prevent long-term voltage collapse. Using the bifurcation calculation
and the time-domain calculation, and by comparison with the system without OLTC, the
innovation conclusions are stated as follows:
(1) There are two SHB points and an SNB point on the stable half of the equilibrium
point curve in the 3-bus system with OLTC and PV, and at the SHB point, a small sudden
increase in reactive load power Q1 , or a sudden increase in PV active power Ppv can all
cause long-term voltage collapse. This phenomenon suggests that SHB is also potentially
threatening to the voltage stability.
(2) Aiming to change of Ppv at the SHB point suddenly, when Ppv drops suddenly,
an OLTC adjustment can recover the load bus voltage to near the target value after a long
period of time; when Ppv increases suddenly, locking OLTC and taking a suitable ration n
can recover the voltage to a stable value. The impact of PV power disturbance at the SHB
point on long-term voltage stability can be eliminated through the OLTC adjustment, and
the system’s long-term voltage stability will focus on the impact of load power disturbance.
(3) The time constant Tn of OLTC is an important influencing factor in long-term
voltage stability, as it can affect the process of long-term voltage collapse and long-term
voltage recovery. However, under different disturbances, the magnitude of Tn can cause
different effects.
(4) A new locking OLTC index Ilock based on PV active power was proposed. It can be
taken as the benchmark of locking OLTC when Ppv increases suddenly at the SHB point.

97
Energies 2023, 16, 6383

Moreover, a modified load-margin index ILMHB is proposed to prevent long-term voltage

collapse caused by the small sudden increase in load power at the SHB point.
Obviously, the discrete model of OLTC and the continuous changes of Ppv are not
considered in the bifurcation analysis of the present case, and the calculation workload of
bifurcation is very heavy for the multi-bus grid-connected system. The above topics require
further study.

Author Contributions: Conceptualization, S.L.; methodology, S.L. and C.Z.; software, S.L., C.Z.
and J.Z.; validation, S.L. and C.Z.; formal analysis, S.L. and C.Z.; investigation, S.L., C.Z. and J.Z.;
resources, S.L. and J.Z.; data curation, S.L. and C.Z.; writing—original draft preparation, S.L. and
C.Z.; Writing—review & editing, S.L. All authors have read and agreed to the published version of
the manuscript.
Funding: This work was supported by the University Student Innovation & Entrepreneurship
Training Program Project of Jiangsu Province (202211276009Z), and the Scientific Research Foundation
of Nanjing Institute of Technology (ZKJ202102).
Data Availability Statement: The data presented in this study are available on request from the
corresponding author.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Zhou, X.; Zhao, Q.; Zhang, Y.; Sun, L. Integrated energy production unit: An innovative concept and design for energy transition
toward low-carbon development. CSEE J. Power Energy Syst. 2021, 7, 1133–1139.
2. Guo, J.; Ma, S.; Wang, T.; Jing, Y.; Hou, W.; Xu, H. Challenges of developing a power system with a high renewable energy
proportion under China’s carbon targets. iEnergy 2022, 1, 12–18.
3. Landera, Y.G.; Zevallos, O.C.; Neves, F.A.S.; Neto, R.C.; Prada, R.B. Control of PV systems for multimachine power system
stability improvement. IEEE Access 2022, 10, 45061–45072.
4. Sajadi, A.; Rañola, J.A.; Kenyon, R.W.; Hodge, B.M.; Mather, B. Dynamics and stability of power systems with high shares of
grid-following inverter-based resources: A tutorial. IEEE Access 2023, 11, 29591–29613.
5. Campanhol, L.B.G.; Silva, S.A.O.D.; Oliveira, A.A.D.; Bacon, V.D. Power flow and stability analyses of a multifunctional
distributed generation system integrating a photovoltaic system with unified power quality conditioner. IEEE Trans. Power Electr.
2018, 34, 6241–6256.
6. Kawabe, K.; Ota, Y.; Yokoyama, A.; Tanaka, K. Novel dynamic voltage support capability of photovoltaic systems for improvement
of short-term voltage stability in power systems. IEEE Trans. Power Syst. 2017, 32, 1796–1804.
7. Li, S.; Wei, Z.; Ma, Y. Fuzzy load-shedding strategy considering photovoltaic output fluctuation characteristics and static voltage
stability. Energies 2018, 11, 779.
8. Rahman, S.; Saha, S.; Islam, S.N.; Arif, M.T.; Mosadeghy, M.; Haque, M.E.; Oo, A.M.T. Analysis of power grid voltage stability
with high penetration of solar PV systems. IEEE Trans. Ind. Appl. 2021, 57, 2245–2257.
9. Li, S.; Lu, Y.; Ge, Y. Static voltage stability zoning analysis based on a sensitivity index reflecting the influence degree of
photovoltaic power output on voltage stability. Energies 2023, 16, 2808.
10. General Administration of Quality Supervision, Inspection and Quarantine of the People’s Republic of China; Standardization
Administration of the People’s Republic of China. Code on Security and Stability for Power System (GB 38755–2019); Standards Press
of China: Beijing, China, 2019; p. 13.
11. Khoshkhoo, H.; Yari, S.; Pouryekta, A.; Ramachandaramurthy, V.K.; Guerrero, J.M. A remedial action scheme to prevent
mid/long-term voltage instabilities. IEEE Syst. J. 2021, 15, 923–934.
12. Islam, S.R.; Sutanto, D.; Muttaqi, K.M. Coordinated decentralized emergency voltage and reactive power control to prevent
long-term voltage instability in a power system. IEEE Trans. Power Syst. 2015, 30, 2591–2603.
13. Sun, Q.; Cheng, H.; Song, Y. Bi-Objective reactive power reserve optimization to coordinate long- and short-term voltage stability.
IEEE Access 2018, 6, 13057–13065.
14. Matavalam, A.R.R.; Singhal, A.; Ajjarapu, V. Monitoring long term voltage instability due to distribution & transmission
interaction using unbalanced μPMU & PMU measurements. IEEE Trans. Smart Grid. 2020, 11, 873–883.
15. Matavalam, A.R.R.; Ajjarapu, V. Long term voltage stability Thevenin index using voltage locus method. In Proceedings of the
2014 IEEE PES General Meeting, National Harbor, MD, USA, 27–31 July 2014; pp. 1–5.
16. Wang, C.; Ju, P.; Wu, F.; Lei, S.; Pan, X. Long-term voltage stability-constrained coordinated scheduling for gas and power grids
with uncertain wind power. IEEE Trans. Sustain. Energy 2022, 13, 363–377.
17. Dharmapala, K.D.; Rajapakse, A.; Narendra, K.; Zhang, Y. Machine learning based real-time monitoring of long-term voltage
stability using voltage stability indices. IEEE Access 2020, 8, 222544–222555.

98
Energies 2023, 16, 6383

18. Cai, H.; Ma, H.; Hill, D.J. A data-based learning and control method for long-term voltage stability. IEEE Trans. Power Syst. 2020,
35, 3203–3212.
19. Yang, H.; Qiu, R.C.; Tong, H. Reconstruction residuals based long-term voltage stability assessment using autoencoders. J. Mod.
Power Syst. Clean Energy 2020, 8, 1092–1093.
20. Peng, Z.; Hu, G.; Han, Z. Dynamic analysis of the power system voltage stability affected by OLTC. Proc. CSEE 1999, 19, 61–65, 78.
21. Sun, Y.; Wang, Z. Modeling of OLTC and its impact on the voltage stability. Autom. Electr. Power Syst. 1998, 22, 10–13.
22. Londero, R.R.; Affonso, C.d.M.; Vieira, J.P.A. Long-term voltage stability analysis of variable speed wind generators. IEEE Trans.
Power Syst. 2015, 30, 439–447.
23. Munkhchuluun, E.; Meegahapola, L. Impact of the solar photovoltaic (PV) generation on long-term voltage stability of a power
network. In Proceedings of the 2017 IEEE Innovative Smart Grid Technologies-Asia (ISGT-Asia), Auckland, New Zealand,
4–7 December 2017; pp. 1–6.
24. Li, S.; Wei, Z.; Sun, G.; Gao, P.; Xiao, J. Voltage stability bifurcation of large-scale grid-connected PV system. Electr. Power Autom.
Equip. 2016, 36, 17–23.
25. Li, S.; Wei, Z.; Ma, Y.; Cheng, J. Prediction and control of Hopf bifurcation in a large-scale PV grid-connected system based on an
optimised support vector machine. J. Eng. 2017, 14, 2666–2671. [CrossRef]
26. Canizares, C.A.; Mithulananthan, N.; Milano, F.; Reeve, J. Linear performance indices to predict oscillatory stability problems in
power systems. IEEE Trans. Power Syst. 2004, 19, 1104–1114. [CrossRef]
27. Cutsem, T.V.; Vournas, C. Bifurcations. In Voltage Stability of Electric Power Systems; Springer: New York, NY, USA, 1998;
pp. 153–160.
28. Wang, Y.; Chen, H.; Zhou, R. A nonlinear controller design for SVC to improve power system voltage stability. Int. J. Electr. Power
Energy Syst. 2000, 22, 463–470.
29. Jing, Z.; Xu, D.; Chang, Y.; Chen, L. Bifurcations, chaos, and system collapse in a three node power system. Int. J. Electr. Power
Energy Syst. 2003, 25, 443–461.
30. Tamimi, B.; Cañizares, C.; Bhattacharya, K. System stability impact of large-scale and distributed solar photovoltaic generation:
The case of Ontario, Canada. IEEE Trans. Sustain. Energy 2013, 4, 680–688. [CrossRef]
31. Li, S.; Wei, Z.; Sun, G.; Wang, Z. Stability research of transient voltage for multi-machine power systems integrated large-scale PV
power plant. Acta Energiae Solaris Sin. 2018, 39, 3356–3362.
32. Bao, L.; Duan, X.; He, Y. Analysis of the effect of OLTC on the voltage stability through time domain simulation. J. Huazhong Univ.
Sci. Tech. 2000, 28, 61–63.
33. Duan, X.; He, Y.; Chen, D. Dynamic analysis of the relation between on-load tap changer and voltage stability. Autom. Electr.
Power Syst. 1995, 19, 14–19.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

99
 energies
Review
A Comprehensive Review on the Modelling and Significance of
Stability Indices in Power System Instability Problems
Chandu Valuva 1,† , Subramani Chinnamuthu 1, *, Tahir Khurshaid 2,† and Ki-Chai Kim 2, *

1 Department of Electrical and Electronics Engineering, SRM Institute of Science and Technology,
Kattankulathur 603203, India; chanduv43@gmail.com
2 Department of Electrical Engineering, Yeungnam University, Gyeongsan 38541, Republic of Korea;
tahir@ynu.ac.kr
* Correspondence: csmsrm@gmail.com (S.C.); kckim@ynu.ac.kr (K.-C.K.)
† These authors equally contributed to this work.

Abstract: Many technological advancements in the modern era have made actual use of electrical
power and the constrained operating of power systems within stability limits. Some expeditious
load variations and rising power demands initiate complications in voltage stability and can put
stress on performance, leading to voltage instability. Voltage Stability Indices can be used to perform
voltage stability assessment. This review evaluates various VSIs based on mathematical derivations,
assumptions, critical values, and methodology. VSIs determine the maximum loadability, voltage
collapse proximity, stability margin, weak areas, and contingency ranking. Stability indices can also
specify the optimal placing and sizing of Distributed Generators. Thus, VSIs play a vital role in power
system voltage stability. This review is a comprehensive survey of various indices and analyses their
accuracy in determining the instability of power systems. Voltage stability is a crucial concern in
operating a reliable power system, and the systematic evaluation of voltage stability is essential in a
power system. This review considered and analyzed 34 indices from 138 articles from the literature
for their significant performance in various power system stability problems. Of 33 indices, were
22 derived from transmission line parameters, referred to as line indices, and 12 from bus and line
parameters, referred to as bus indices.
Citation: Valuva, C.; Chinnamuthu,
S.; Khurshaid, T.; Kim, K.-C. A
Keywords: power system stability; voltage stability analysis; stability indices; power losses; optimal
Comprehensive Review on the placement; weakest bus
Modelling and Significance of
Stability Indices in Power System
Instability Problems. Energies 2023,
16, 6718. https://doi.org/ 1. Introduction
10.3390/en16186718 Voltage instability is a crucial phenomenon that effects power systems because it is the
Academic Editor: Ying-Yi Hong main reason for blackouts and voltage collapse. Some unexpected consequences like an
unexpected increase in load, line outage, and generator tripping may provoke an excessive
Received: 28 July 2023 load demand in the power system, leading to voltage instability. Consequently, a power
Revised: 6 September 2023
system may be disrupted without appropriate action, resulting in cascading failures and
Accepted: 14 September 2023
large-scale blackouts. The notable blackouts around the world have primarily been caused
Published: 20 September 2023
by voltage instability. From [1], From 1965 to 2005, it can be seen that 22 significant backouts
occurred; out of these 22 blackouts, 11 had voltage instability as a leading cause. Voltage
instability is also known as voltage collapse; moreover, when instability is developing, the
Copyright: © 2023 by the authors.
continuity of such events affects the system, resulting in low voltage and even more power
Licensee MDPI, Basel, Switzerland. outages [2].
This article is an open access article According to an IEEE Power System Engineering Committee, “Voltage stability is
distributed under the terms and the ability of a system to maintain voltage so that, when increasing load admittance, load
conditions of the Creative Commons power increases, and so that both power & voltage are manageable”. The consolidation
Attribution (CC BY) license (https:// definition of the IEEE and CIGRE is: “Voltage stability refers to the ability of a power
creativecommons.org/licenses/by/ system to maintain steady voltages at all buses in the system after being subjected to a
4.0/). disturbance from a given initial operating condition”.

Energies 2023, 16, 6718. https://doi.org/10.3390/en16186718 100 https://www.mdpi.com/journal/energies

Energies 2023, 16, 6718

Power system voltage instability may be due to the inadequacy of the source to deliver
sufficient reactive power or to faulty power transmission lines not transmitting the desired
reactive power. Voltage instability may affect the stability of a wide area or local area,
resulting in the voltage collapse of a whole power system—mainly reliable active power
delivered from the generators and reactive compensators like shunt capacitors. Voltage
instability in a network is most probably the cause of unexpected load demand, transformer
tripping, being over the limit of the generators’ reactive power, or defects in the on-load tap
changing transformers. Consequently, the regular tracking and forecasting of instability
are essential in the power system. In this respect, definite methods for suppressing voltage
instability are:
1. Strengthen the weak bus while planning the power system, like accurately selecting
distributed generators and balancing voltage.
2. Temporarily defending the supply by load shedding or improving the power factor
by shunt capacitor switching.
3. Enhancing the Voltage Stability Margin (VSM) with FACTS devices.
4. Blocking on-load tap changers (OLTCs) to operate the transformer.
5. Build the generation station near to the load centre and reduce the transmission
line length.
6. For executive controllers, make use of system voltage fluctuation controllers, Shunt Com-
pensation, step-up transformer controllers, and Automatic Voltage Regulators (AVRs).
7. For Real-time Performance: Scheduling the generation according to load demand,
evaluating voltage stability, and protecting the load shedding.
8. For safeguarding systems: Instant Load Tap Changer (LTC) control, Contingency of
load demand, and High Voltage Direct Current (HVDC).
Indeed, over the past few decades, the reliability and stability of power systems have
gone through considerable improvements. Stability analysis has evolved from simple static
and transient stability analyses to more sophisticated and comprehensive methods.
Voltage Stability Indices (VSIs) generally measure voltage stability. The stability in-
dices’ performance is essential in analyzing and illustrating the consequences of power
system operation. Furthermore, the indices assist in predicting eventual improvement
and determining long-term advancement. Especially in the last three decades, numerous
techniques and methodologies have been developed for easy stability analysis. The per-
formances of various indices for determining voltage stability have been explained in this
review. For assessing a system’s stability, several indices consider the system impedance.
However, few indices are unconstrained by impedance and performed using the current
and voltage of the system. In real-time operation, it is impossible to determine the accurate
impedance of a system because of troposphere conditions and inadequate knowledge
of the system. Therefore, the accuracy of the impedance-operated indices is repeatedly
accompanied by errors.
A taxonomy of comprehensive characteristics, variations in accuracy, and some pa-
rameters being inaccurate is presented below, as well as the terminology concerning VSIs,
including mathematical derivation, instability conditions, and assumptions of specific pa-
rameters. This taxonomy can serve to make researchers aware of the advancements in the
VSIs, and the respective operations including voltage stability enhancement, strengthening
the weak bus, locating the reactive compensation devices, reducing the cost-effectiveness
and location of FACTS devices, and counteracting the prevention of voltage collapse.
Determining the VSI values for a system is carried out in three modes of operation.
One is the online mode: In this mode of operation, the engineer observes the voltage
stability and operates accordingly to maintain the system’s stability. The statistics used for
this in this paper were composed of mathematical models and power system measurements.
The second mode of the stability indices is the offline mode: In this mode of operation, the
designer gathers the simulated or historical data. The preserved data identify the voltage
instability intensity and voltage collapse and do not contain real-time data on voltage

101
Energies 2023, 16, 6718

instability. The real-time method is the third mode of stability indices, and is obtained
using a Phasor Measurement Unit, Power Flow Analysis, and Artificial Intelligence.
This article aims to thoroughly review and scrutinize numerous Voltage Stability
Indices. This review indicates the limitations, strengths, underlying principles, and ap-
plicability of these indices by sharply examining them. This article provides clarity to
researchers, engineers, and power system operators in choosing suitable stability indices
depending on the particular characteristics of a system and functional needs. Researchers
will gain synthesized knowledge of Voltage Stability Indices, as this article presents the
diverse range of stability indices with mathematical formulations, highlighting their pecu-
liar characteristics and underlying principles. This review supports decision-making for
system operators with the knowledge needed to estimate instabilities and ensure voltage
stability, sustainable energy supply, and grid reliability.

2. Voltage Stability Indices

Voltage Stability Indices are values of an arithmetic sequence adopted for evaluating
voltage stability. The indices specify the accessible and equitable regulation of the sys-
tem’s stability so that voltage levels are within sustainable limits. Indices predominantly
assist power engineers in determining voltage stability, predicting voltage collapse, and
evaluating the significance of system characteristics. This review elaborates on the numer-
ous methodologies for assessing voltage stability. Furthermore, categorizing the different
VSIs in [3–5] as line VSIs, Bus VSIs, and overall VSIs. Jacobian matrix-based indices and
PMU-based indices fall under overall VSIs.
The overall VSIs’ precision is superior compared to the other stability indices. The
overall VSIs are challenging to determine and take more computation time. Line VSIs
obtain easy and simple indices that are reasonable in determining all optimization issues
through the proficiency of these indices and identifying the critical line and weak bus.
These VSIs are derived from a mathematical equation by considering the standard two-bus
concept. With these mathematical derivations, the system examines the stability of various
stability indices. Similar aspects of voltage collapse points determine the VSIs’ formation.
Figure 1 represents an illustration of the interconnected network with a single line in a
standard two-bus model.

Figure 1. Standard two-bus model.

Jacobian matrix-based VSIs specify the scope of system stability, determine the voltage
collapse point, and evaluate the VSM. However, they consume more time for computation,
and in general, mathematical topology variations compel the transformation of the Jacobian
matrix, causing the matrix to need to be again; therefore, they are inconvenient in real-time
applications. Additionally, Jacobian matrix indices tend to increase the operational time of
DG placing and sizing issues. However, the VSIs established by system variables claim
lesser computational time and are acceptable for real-time operations. The detriment of

102
Energies 2023, 16, 6718

these stability indices is that they are less accurate in evaluating the VSM. Consequently,
they are not as accurate in determining the weak line and critical bus. In most applications,
like placement of the DG and sizing issues, VSIs prevail for identifying the weak bus and
critical line or stimulating the deterrents for voltage instability.

Sr = Pr + jQr = Vr I ∗ (1)

 ∗
Sr ( Pr − jQr )
I= = (2)
Vr Vr ∠ − δr
Vs , Vr : magnitude of bus voltage at sending and receiving bus; Ss , Ps , Qs : apparent,
real, and reactive power at sending bus; Sr , Pr , Qr : apparent, real, and reactive power at
receiving bus; δs , δr : voltage angle at sending and receiving bus; R, X, Z, θ = line resistance,
line reactance, line impedance, and line impedance angle.
Considering the Kirchhoff Voltage Law (KVL), the current equation for two buses is,

Vs ∠δs − Vr ∠δr
I= (3)
R + jX

Substituting Equation (3) in Equation (2), we obtain:

Vs ∠δs − Vr ∠δr ( Pr − jQr )

= (4)
R + jX Vr ∠ − δr

Vs Vr ∠(δ s − δr ) − Vr2 = Pr R − jQr R + jPr X + Qr X (5)

Let δs − δr = δ and simplify Equation (5) as:

Vs Vr ∠δ − Vr2 = Pr R − jQr R + jPr X + Qr X (6)

Transforming Vs Vr ∠δ into a rectangular form, we obtain:

Vs Vr cos δ + jVs Vr sin δ − Vr2 = Pr R − jQr R + jPr X + Qr X (7)

Segmenting Equation (7) into real and imaginary parts, we obtain:

Vs Vr cos δ − Vr2 = Pr R + Qr X (8)

Vs Vr sin δ = − Qr R + Pr X (9)
Readjusting Equations (8) and (9), we obtain:

− Qr X + Vs Vr cos δ − Vr2
Pr = (10)
R

Pr X − Vs Vr sin δ
Qr = (11)
R
Substituting Equation (11) in Equation (10) and vice versa, we obtain:
 
− Pr X −VRs Vr sin δ X + Vs Vr cos δ − Vr2
Pr = (12)
R
 
− Qr X +Vs Vr cos δ−Vr2
R X − Vs Vr sin δ
Qr = (13)
R

103
Energies 2023, 16, 6718

Rearrange Equations (12) and (13) allows us to obtain:

   
X X2
Vr2 − Vs Vr cos δ + sin δ + Pr R + =0 (14)
R R
   2 
R R
Vr2 + Vs Vr sin δ − cos δ + Qr +X =0 (15)
X X
For a few VSIs, the line resistance (R) and reactance (X) increase with the impedance
(Z) and impedance angle (θ); therefore, reducing the current equation, Equation (3) is
redefined as:
Vs ∠δs − Vr ∠δr
I= (16)
Z ∠θ
Substituting Equation (16) into Equation (1), the receiving end’s apparent power
concerning the real and reactive power is:
 ∗
Vs ∠δs − Vr ∠δr
Pr + jQr = Vr (17)
Z ∠θ

Assuming δs − δr = δ then Equation (17) is simplified as:

Vs Vr V2
Pr + jQr = ∠(θ − δ) − r ∠θ (18)
Z Z
Equating the real and imaginary parts of Equation (18), we obtain:

Vs Vr V2
Pr = cos(θ − δ) − r cos θ (19)
Z Z

Vs Vr V2
Qr = sin(θ − δ) − r sin θ (20)
Z Z
Simplifying Equations (19) and (20), we obtain:

Vr2 cosθ − Vs Vr cos(θ − δ) + Pr Z = 0 (21)

Vr2 sinθ − Vs Vr sin(θ − δ) + Qr Z = 0 (22)

Most of the stability indices were derived mathematically; the quadratic equation’s
voltage discriminant is greater than or equal to zero. The characteristics vary for the index,
so different assumptions are considered for deriving the stability indices.

2.1. Line Voltage Stability Indices

2.1.1. Voltage Stability Load Index (VLSI)
The Voltage Stability Load Index (VLSI) derives from the mathematical derivation [6]
of a standard two-bus system and Thevenin equivalent circuit. Figure 1 presents the voltage
equation considered in this index, obtained from the network. Moreover, we determine
the power flow by performing a power flow analysis directed toward speedy converging.
Considering the square of the sending-end current, receiving-end active and reactive power
with losses, we obtain:
P2 + Q2
| Is |2 = s 2 s (23)
Vs

Pr = Ps − Ploss (24)

Qr = Qs − Qloss (25)

104
Energies 2023, 16, 6718

 
Pr2 + Q2r
Ploss = Rs (26)
Vr2
 2 
Pr + Q2r
Qloss = Xs (27)
Vr2
We then substitute Equations (24)–(27) in Equation (23) to simplify it, resulting in:
   2   2 2  2
Pr2 + Q2r
Pr + Vr2
Rs + Qr + Pr V+2Qr Xs
2
|I| = r
(28)
Vs2

The analyzed equation is further reduced by equating it to zero and obtaining the
sending end voltage equation:
 
Pr2 + Q2r  2 
Vs2 = Vr2 + 2( Pr Rs + Qr Xs ) + Rs + Xs2 (29)
Vr2
    
Vr4 + Vs2 2( Pr Rs + Qr Xs ) − Vs2 + Pr2 + Q2r R2s + Xs2 = 0 (30)

If the obtained receiving end voltage quadratic equation contains real roots, then the
discriminant is greater than or equal to zero; hence:
 
8Pr Qr Rs Xs − 4Vr2 ( Pr Rs + Qr Xs ) + Vs4 − 4 Pr2 Xs2 + Q2r R2s ≥ 0 (31)

 
4 Vs2 ( Pr Rs + Qr Xs ) + ( Pr Xs − Qr Rs )2
≤1 (32)
Vs4
Accordingly, L represents the voltage stability load index,
 
4 Vs2 ( Pr Rs + Qr Xs ) + ( Pr Xs − Qr Rs )2
L= (33)
Vs4

since Vs Vr cos(δs − δr ) − Vr2 = Pr Rs + Qr Xs & Vs Vr sin(δs − δr ) = Pr Xs − Qr Rs

 
4 Vs Vr cos(δs − δr ) − Vr2 cos(δs − δr )2
Therefore L = (34)
Vs2
 
4 Vs Vr cos(δ) − Vr2 cos(δ)2
L= (35)
Vs2
Hence, the value L should be less than 1.0 for the system’s stability. Whenever the
index value is more significant than 1.0, corresponding to Equation (32), the voltage emerges
as imaginary and collapses. To defend the load instability, the load index value must be Li
≤ 0, and if Li is nearer to 1.0, then bus ‘i’ approaches the stability limit.

2.1.2. Line Stability Index (Lmn )

The Line Stability Index is determined in [7] and derived from the voltage quadratic
equations discriminant. Considering this index and Equation (20), we can derive Vr :
 0.5
Vs sin(θ − δ) ± [Vs sin(θ − δ)]2 − 4ZQr sinθ
Vr = (36)
2sinθ

105
Energies 2023, 16, 6718

Assuming Zsinθ = x,
 0.5
Vs sin(θ − δ) ± [Vs sin(θ − δ)]2 − 4xQr
Vr = (37)
2sinθ
The above voltage equation should have real roots to obtain absolute values of Vr in
terms of Qr . Therefore, the following condition is to manipulate the stability of the system.
 
[Vs sin(θ − δ)]2 − 4xQr ≥ 0
4xQr (38)
2 = L mn ≤ 1
[Vs sin(θ −δ ]

Lmn represents the stability index of the line. Voltage collapse is predicted rigorously
depending on the stability index’s value. If the obtained value of Lmn is lesser than the
unity, then the system is in a stable condition; otherwise, if the value is more significant
than the unity, then the system exhausts the stability; moreover, the voltage collapses.

2.1.3. Line Stability Factor (LQP)

The Line Stability Factor (LQP) developed in [8] makes use of a similar theory of the
discriminant of the voltage quadratic equation, primarily from receiving end real power:
 
Ps2 − Q2s
Pr = Ps − R (39)
Vs2
 
P2 Q2
Reorganizing the power equation as R Vs2 − Ps + Pr + Pr + R V 2s = 0 allows us to
s s
obtain the roots of Ps :   
R Q2s
1−4 Pr + R ≥0 (40)
Vs2 Vs2
From Equation (40), it is clear that the system loses its stability when it violates this
condition. Further, deriving the reactive power is similar to the derivation of active power:
  
X Ps2
1−4 Qr + X ≥0 (41)
Vs2 Vs2
  
X Ps2
LQP = 4 Qr + X (42)
Vs2 Vs2
Since LQP < 1, the system is stable. Here, for deriving the index, it is assumed that the
lines are lossless (R/X << 1) and neglect the shunt admittance.

2.1.4. Voltage Collapse Proximity Index (VCPI)

The VCPI [9] evaluates the line voltage stability depending on the theory of maximum
power. The analysis assumes a frequent occurrence and varying of the absolute impedance
value by keeping φ constant. With this inference, the performance may not be accurate, but
it is easy to solve. During the operation, the power factor remained constant.
Along with the load demand increase, the current increases with a decrease in Zr ,
resulting in a receiving-end voltage drop:

Vs
I = )  (43)
( Zs cosθ + Zr cos∅) + ( Zs sinθ + Zr sin∅)2
2

Vs
Vr = Zr I = )  (44)
2
1 + ( Zr /Zs ) + 2( Zr /Zs )cos(θ − ∅)

106
Energies 2023, 16, 6718

Hence, the receiving-end power is Pr = Vr Icosφ and Qr = Vr Isinφ

Simplifying Equation (44) with that for power, we obtain:

(Vs )2 /Zs Zr
Pr = 2
cos∅ (45)
1 + ( Zr /Zs ) + 2( Zr /Zs )cos(θ − ∅) Zs

(Vs )2 /Zs Zr
Qr = 2
sin∅ (46)
1 + ( Zr /Zs ) + 2( Zr /Zs )cos(θ − ∅) Zs
Correspondingly, the power loss is Pl = I 2 Zs cosθ and Ql = I 2 Zs sinθ.
Substituting the current equation, we then obtain:

(Vs )2 /Zs
Pl = 2
cosθ (47)
1 + ( Zr /Zs ) + 2( Zr /Zs )cos(θ − ∅)

(Vs )2 /Zs
Ql = 2
sinθ (48)
1 + ( Zr /Zs ) + 2( Zr /Zs )cos(θ − ∅)
Considering the boundary condition ∂Pr /∂Zr = 0 allows us to calculate the maxi-
mum active power delivered at the receiving end. Hence, this condition results in the
impedance Zr /Zs = 1. The maximum power transfer obtained by substituting this ratio in
Equation (45) is:
V2 cos∅
Pr(max) = s (49)
Zs 4cos2 (θ −∅)
2
Similarly, adopting this approach, the following power equation can be derived:

Vs2 sin∅
Maximum transferable reactive power Qr(max) = (50)
Zs 4cos2 (θ −∅)
2

Vs2 cosθ
In line with the maximum active power loss Pl (max) = (51)
Zs 4cos2 (θ −∅)
2

Vs2 sinθ
In line with the maximum reactive loss Ql (max) = (52)
Zs 4cos2 (θ −∅)
2
Considering these maximum limits, Voltage Collapse Proximity Indicators can be
designed as:

Pr Real power tran f erred to the receiving end

VCPI (1) = = (53)
Pr(max) Maximum real power that can be trans f erred

Qr Reactive power tran f erred to the receiving end

VCPI (2) = = (54)
Qr(max) Maximum reactive power that can be trans f erred

Pl Real powerin the line

VCPI (3) = = (55)
Pl (max) Maximum possible real power that loss

Qr Reactive power loss in the line

VCPI (4) = = (56)
Qr(max) Maximum possible reactive power loss
The system may collapse if the values of the VCPIs are more significant than the unity.
Therefore, VCPI (1) = VCPI (2) and VCPI (3) = VCPI (4), since only active or reactive terms
were considered instead of all proximity indices.

107
Energies 2023, 16, 6718

2.1.5. Voltage Stability Index (Lp )

The line stability index proposed in [10] is a similar mathematical formulation to earlier
line VSIs. Basically, from the active and reactive power equations, i.e., from Equations (19)
and (20), δ can be derived as:

Vs2 − ( Ps R + Qs X )
δ = cos−1 (57)
Vs Vr

Reorganizing Equation (19) and deriving the voltage equation Vr , we obtain:

 0.5
Vs cos (θ − δ) ± [Vs cos(θ − δ)]2 − 4RPr
Vr = (58)
2cosθ
Hence, the Vr equation should possess real roots for obtaining the absolute values of
Vr expressed in terms of Pr . Accordingly, it is then simplified to be enough to withstand the
following conditions:  
[Vs cos(θ − δ)]2 − 4RPr ≥0 (59)

4RPr
= Lp ≤ 1 (60)
[Vs cos(θ − δ)]2
LP is a line stability index that determines the condition of the transmission line and
indicates instability limits. If the calculated value of Lp is more significant to the unity, the
system approaches instability. The voltage collapses when the system exceeds the critical
limits; moreover, the Vr becomes imaginary.

2.1.6. Fast Voltage Stability Index (FVSI)

The Fast Voltage Stability Index is derived in [11] primarily by considering the cur-
rent through the line and then calculating the absolute roots of a receiving-end voltage
(Equation (15)):
  )  2  
R2
X sinδ + cosδ Vs ± X sinδ + cosδ −4 X+
R R
Vs X Qr
Vr = (61)
2
For determining the absolute roots of Vr , the Vr discriminant is greater than or equal
to zero.
4Z2 Qr X
≤1 (62)
Vs2 ( Rsinδ + Xcosδ)2
The angular difference δ is minute; thus, δ ≈ 0, Rsinδ ≈ 0, and Xcosδ ≈ X. As a result,
Equation (62) can be simplified as:

4Z2 Qr
FVSI = (63)
Vs2 X

The line is in-transit to the instability limit if the FVSI value is adjacent to 1. With
further increases in a stability index value, that line may encounter an unexpected voltage
drop accompanying system collapse.

2.1.7. Voltage Stability- Load Bus Index (VSLBI)

VSLBI evaluates the voltage stability by adopting the PMU [12]. Considering the
maximum power condition, this stability index is subject to a voltage-drop ΔVr over the
transmission impedance Zr that is equivalent to load bus voltage Vr:

ΔVr = Vr (64)

108
Energies 2023, 16, 6718

Accordingly, to approach the voltage collapse exposure during the constant power
loads, the Voltage Stability Load Bus Index (VLSBI) is represented as:

Vr
VSLBI = (65)
ΔV
If the obtained value of the VSLBI is more significant than the unity, then the system is
considered stable; if the VSLBI is less than 1, then the system is unstable, and the system
may collapse.

2.1.8. Voltage Stability Margin Index (VSMI)

The VSMI, proposed in [13], depends on the correlation between the power transfer
from the line and the angular difference between the sending- and receiving-end buses. To
derive the fundamental equations of real and reactive power, the receiving end voltage is:

cos(∅r + δ)
Vr = Vs (66)
cos(∅r )
 
1 Vs2 sin(∅r + 2δ)
Pr = − tan(∅r ) (67)
2 X cos(∅r )

Since tan(∅r ) = QPrr , considering Equations (66) and (77), the power Pr maximum value
can be calculated for whatever the value of Qr /Pr is. The respective angular differences of the
sending- and receiving-end buses (δ), and the receiving-end voltage (Vr ), are expressed as:
 
1 Vs2 1
Prmax = − tan(∅r ) (68)
2 X cos(∅r )
 
cos π2 + ∅r
Vrmax = Vs (69)
cos(∅r )
π −∅
δrmax = 2 2 r for any ratio of Qr /Pr
With the correlation between the voltage stability and angular difference, the voltage
stability margin that is determined depends on what degree of δ is adjacent to δrmax . There-
fore, whatever the operating constraints, the receiving end index can be represented as:

δrmax − δ
VSMI = (70)
δrmax

The calculated value of the VSMI should be higher than zero to sustain the system’s
stability and protection.

2.1.9. Voltage Collapse Proximity Index (VCPI_1)

The significant motive of VCPI_1 is that, during the voltage collapse, the voltage drop
at the Thevenin impedance equals the load voltage. The proximity index derived in [14]
is denoted as VCPI_1 to prevent confusion from the VCPI notation. An easy arithmetic
prediction indicates that, during the critical situation, the generator phasor voltage ( Vs ) is
double the load phasor voltage, Vr . Hence, for the estimation of the possibility of voltage
instability, VCPI_1 is represented as:

VCPI_1 = Vr cosδ − 0.5Vs (71)

If the VCPI is greater than or equal to zero, then the system is stable, otherwise it is
unstable. In this index, the lines are complicated. However, considering that the sending-
end bus connects to the ideal voltage source, the receiving end with the equivalent Thevenin
impedance is neglected.

109
Energies 2023, 16, 6718

2.1.10. Critical Voltage (Vcr)

The Critical Voltage is an intelligible index that is derived from a single load, an infinite
bus system utilizing the Eigen-value theorem, and load flow equations [15]. In this paper,
it is used for adapting the power factor cosφ to be constant. From the active and reactive
power flow equations, the bus voltage angle δ is eliminated, and the resulting equation is:
   
Vr4 + 2Z (Vr cosθ + Qr sinθ ) − Vs2 Vr2 + Pr2 + Q2r Z2 = 0 (72)

Corresponding to favorable circumstances, Equation (72) contains a specific solution;

moreover, the correlation between the Vs , Vr , I vectors is developed as:

Vs
Vcr = V + = V − = (73)
2cosθ
The load bus critical voltage is denoted by Vcr (P-V Curve nose point). V+ and V−
are the P-V curves’ upper and lower parts, respectively. Nearer to this, V+ and V− are
superposed values of Vcr . The indices concern the maximum power limit as a stability
limiting point. Evaluating the simple system by the constant MVA load and cosφ power
factor, Equation (72) is:

V2 V 4 − Vs2 Vr2
S2 + 2S cos(θ − ∅) + =0 (74)
Z Z2

P
Apparent Power S = (75)
cos∅
To obtain the maximum S, dS/dV is equated to zero. Accordingly:

Vs
Vcr = * (76)
2(1 + cos(θ − ∅))

After the phasor relationship, further reducing Equation (76) allows us to obtain:

E
Vcr = (77)
2cosθ

2.1.11. Power Transfer Stability Index (PTSI)

The Power Stability Index proposed in [16] is derived with the help of a standard
two-bus Thevenin-equivalent system; here, the slack bus and the load bus were connected
on a single branch, as represented in Figure 2. The current delivery to load and load
power is:
V Thev
I= (78)
Z Thev + Z L
+ +2
SL = ZL I I ∗ = ZL + I + (79)
Substituting Equation (78) in (79), we obtain:
+ +2
+ V Thev ++
S L = Z L ++ (80)
Z Thev + Z L +

Assuming that Z L = ZL ∠ ϕ and Z Thev = ZThev ∠θ, and alternating them into Equation (80),
we obtain: + +2
+ VThev +
S L = ZL ∠ ϕ++ + (81)
ZThev ∠θ + ZL ∠ ϕ +

110
Energies 2023, 16, 6718

2
VThev ZL
SL = (82)
| ZThev ∠θ + ZL ∠ ϕ|2
2
VThev ZL
SL = (83)
2
ZThev + ZL2 + 2ZThev ZL cos(θ − ϕ)
To obtain the maximum value of the apparent power of the load, the SL is differentiated
and equated to zero, ∂S L /∂ZL = 0.
2
VThev
S Lmax = (84)
2ZThev (1 + 2cos(θ − ϕ))

The power margin at the load bus distance when it is approaching voltage collapse
is SLmax – SL . If ZL = ZThev, , the power margin is equal to zero; moreover, it indicates no
power transfer and may cause voltage collapse. Conversely, a voltage may collapse if the
ratio of power is:
SL
=1 (85)
S Lmax
Substituting Equations (83) and (84) in Equation (85), the power transfer stability
index is:
2S L Z (1 + cos(θ − ϕ))
PTSI = (86)
Vs2
The value of the PTSI is determined with Equation (86), considering the load’s power,
impedance, Thevenin Voltage, Thevenin Impedances, and the Angular phase of the load.
The obtained PTSI value lies between 0 and 1. Voltage collapse occurs when the index
value reaches 1.

Figure 2. Simple two-bus Thevenin Equivalent Circuit.

2.1.12. Voltage Stability Index (VSI_1)

The stability index predicts the voltage stability at the steady state operation; this
index is proposed in [17]; however, for easy identification, the notation of the stability
index is represented as VSI_1. Initially, this index calculates the maximum active, reactive,
and apparent power transfer. To obtain the receiving-end voltage (Vr ), considering the
Equation (19) for active power (Pr ) and Equation (20) for reactive power (Qr ), we obtain:
)
Vs2 √
Vr = − ( QX + PR) ± A (87)
2
Vs2
Here, = 4 − ( QX + PR)Vs2 − ( PX − QR)2 , A ≥ 0.

111
Energies 2023, 16, 6718

Transmission lines containing a higher X/R ratio and ignoring the resistance value,
resulting in Pmax , Qmax , and Smax , can be derived as:
,
QR VS2 R | ZL |Vs VS − 4QX
2
Pmax = − 2
+ (88)
X 2X 2X 2
)
Vs4 V2
Pmax = 2
−Q s (89)
4X X
,
PX V2 X | ZL |Vs VS2 − 4PR
Qmax = − S2 + (90)
R 2R 2R2

Vs2 P2 X
Qmax = − 2 (91)
4X Vs

Vs2 [| ZL | − (sin (θ ) X + cos (θ ) R)]

Smax = (92)
2(cos (θ ) X − sin(θ ) R)2
(1 − sin(θ )) Vs2
Smax = (93)
2cos(θ )2 X
The derived maximum powers identified that the maximum power transfer increases
with an increase in the voltage magnitude (Vr ) or a decrease in the impedance (ZL ). Equiv-
alently, Pmax decreases with an increase in Q, and Qmax decreases with an increase in P.

Pmargin = Pmax − P, Qmargin = Qmax − Q, Smargin = Smax − S (94)

From the maximum powers, the corresponding load margins can be obtained with
Equation (94). Hence, the derived VSI from the obtained load margin is represented in
Equation (95). If the obtained VSI is small, the load bus is adjacent to the stability margin
and the lesser load margin that is available. The VSI equals zero if the load bus reaches its
stability margin point.
 
Pmargin Qmargin Smargin
VSI = min , , (95)
Pmax Qmax Smax

2.1.13. Novel Line Stability Index (NLSI)

The NLSI is determined with a similar power transmission approach [18], considering
the quadratic Equations (8) and (9):
*
Vs cosδ ± Vs2 cos2 δ − 4( Pr R + Qr X )
Vr = (96)
2
The discriminant should be greater than or equal to 0 to obtain the absolute value
of Vr .
PR + QX
≤1 (97)
0.25Vs2 cos2 δ
As the angular difference (δ) between receiving and sending ends is usually very small,
cosδ ≈ 1. Subsequently, the Novel Line Stability Index can be represented as:

Rsr Pr + Xsr Qr
NLSI = (98)
0.25V 2s

If the value of the NLSI of any line is adjacent to 1, that line is moving closer to the
stability limit; consequently, the system may collapse. Hence, the NLSI allows us to find
the stability limit, i.e., the index value is lesser than the unity.

112
Energies 2023, 16, 6718

2.1.14. Stability Index (SI)

For load flow analysis, the quadratic equation is preferable for calculating the sending-
end voltage of a line and can be simplified in general form as:
 
Vr4 + 2Vr2 ( PR + QX ) − Vs2 Vr2 + P2 + Q2 | Z |2 = 0 (99)

Further, the receiving-end line active and reactive power is derived as:
 , 
P = −cos(θ )Vr2 ± cos2 (θ )Vr4 − Vr4 − Z 2 Q2 − 2Vr2 QX + Vs2 Vr2 /| Z | (100)

 , 
Q = −sin(θ )Vr2 ± sin2 (θ )Vr4 − Vr4 − Z 2 P2 − 2Vr2 PR + Vs2 Vr2 /| Z | (101)

Therefore, to obtain the absolute value of the active and reactive power, the corre-
sponding equation should satisfy the following:

cos2 (θ )Vr4 − Vr4 − Z 2 Q2 − 2Vr2 QX + Vs2 Vr2 ≥ 0 (102)

sin2 (θ )Vr4 − Vr4 − Z 2 P2 − 2Vr2 PR + Vs2 Vr2 ≥ 0 (103)

Adding both equations, we obtain:
 
2Vs2 Vr2 − Vr4 − 2Vr2 ( PR + QX ) − | Z |2 P2 + Q2 ≥ 0 (104)

The above equation illustrates that equation values decrease with increased power
and impedance. Moreover, Equation (104) considers a stability index [19].
 
SI (r ) = 2Vs2 Vr2 − Vr4 − 2Vr2 ( PR + QX ) − | Z |2 P2 + Q2 (105)

Therefore, the classic stability principle determines the stability index for individual
lines of radial distribution networks. The minimum stability index value is the most
vulnerable to collapse. If the index is equal to zero, then the system collapses. Here, also,
neglecting the shunt admittance to determine the stability index will lead to collapse.

2.1.15. Voltage Stability Margin (VSM)

VSM technique [20] is when the admissible rise in load power occurs during the
current operating mode or the critical mode. Considering Figure 1, the voltage (Vs ) and
load apparent power (SL ) are derived as follows:

Vs Zr
Vr = - (1/2 (106)
Zr2 + Z02 + ZL Z0 cos(θ0 − θr )

Vr2 Zr
Sr = - ( (107)
Zr2 + Z02 + ZL Z0 cos(θ0 − θr )
Concerning the critical point, the load power attains the maximum value and can be
written as:
Vs2
Scr = (108)
2Z0 [1 + cos(θ0 − θr )]
Zr is the available load impedance, and the variation between Zr and Z0 is assumed as
an invulnerability limit. Accordingly, the VSM in terms of impedance is as follows:

Zr − Z0
VSM Z = (109)
Z0

113
Energies 2023, 16, 6718

The allowable power increases until the stability limit is contingent to a higher degree
on the angle of the load power factor with a similar VSMs . Therefore, it can be enhanced to
signify the margin concerning the load apparent power, VSMs .

Scr − S L
VSMs = (110)
Scr

Hence, the margin of the apparent load power is derived from the VSM. Two assump-
tions for the stability margin are (i) that the line shunt admittance neglected and (ii) a
constant power factor. The system is unstable when the apparent load power is more
significant than the critical one. Even if the value of the VSM is less than zero, the system
is unstable.

2.1.16. Voltage Reactive Power Index (VQI)

The line index proposed in [21] has a similar methodology to that of Lp . To insert the
relation between the voltage (Vs ) and reactive power (Qr ) comprised from Equation (19)
into Equation (18), we obtain:

|Vr Vs Ysr |.cos(θ − δ) − |Vr |2 |Ysr |.cos(θ ) − jQr = |Vr Vs Ysr |∠(θ − δ) − |Vr |2 |Ysr |∠θ (111)

sin(θ − δ) Qr
|Vr |2 − |Vr Vs | + =0 (112)
sin(θ ) |Ysr |sin(θ )
As the value of δ is negligible and minimized to zero, then the entire term of
(sin(θ − δ)/sin(θ)) is omitted:

Qr
|Vr |2 − |Vr Vs | + =0 (113)
|Ysr |sin(θ )

Replacing the term Yrs sin(θ) with Brs, , the corresponding equation is:

Qr
|Vr |2 − |Vr Vs | + =0 (114)
| Bsr |

Extracting the real roots of the quadratic equation in terms of Vr , we obtain:

,
|Vr | ∓ |Vr |2 − 4Qr
| Bsr |
Vr = (115)
2
When the discriminant of Vr equates to zero, the respective equation contains one
distinct real root, or two equal roots. Hence, the respective real roots are:

4Qr 4Qr
|Vr |2 − ≤ 0 =⇒ ≤1 (116)
| Bsr | | Bsr ||Vr |2

The value of Vr ranges from 0 to 1, illustrating the limitation of real roots as the voltage
stability limits. Moreover, real roots should be less than 1 and greater than 0, or the stability
is affected. Thus, the derived equation evaluates the system stability and identifies the
voltage collapse point as the voltage reactive power index.

4Qr
VQI = ≤1
| Bsr ||Vr |2

If the obtained value of VQI is nearer to 1, then the system is within stability limits.
Otherwise, a system beyond the stability limits leads to instability. VQI also identifies the
instability point of occurrence and voltage collapse.

114
Energies 2023, 16, 6718

2.1.17. Line Collapse Proximity Index (LCPI)

An improved LCPI was developed in [22] to overcome the limitations of specific
indices [7,8,11] which ignored the line charging reactance while deriving their values. Since
the line charging reactance is a critical factor supporting voltage stability, this proximity
index was derived considering the definite transmission line model as well as the conse-
quences of active and reactive power flows through lines. The definite transmission line
model primarily represents an equivalent circuit of a two-port network considering the
ABCD matrix. Thus, the proposed proximity index is derived from the ABCD matrix.
    
Vs A B Vr
= (117)
Is C D Ir

Hence, the transmission line parameters for a two-port circuit are A, B, C, and D.
   
A = 1 + Z × Y2 , B = Z, C = Y × 1 + Z × Y4 , D = A
(118)
Receiving-end current : Ir = ( Pr − jQr )/Vr∗ = ( Pr − jQr )/Vr ∠ − δr

Considering the receiving-end current, the sending-end voltage Vs is derived as:

Vs ∠δs = A∠α∗ Vr ∠δr + B∠ β∗ Ir ∠00 (119)

A and B are magnitudes, and their corresponding phase angles are α and β, respectively.
Combining the current with the voltage equation, we obtain:

Vs ∠δs = A∠α∗ Vr ∠δr + B∠ β∗ ( Pr − jQr )/Vr ∠ − δr (120)

Vs Vr ∠δ = A∠α∗ Vr2 + B∠ β∗ ( Pr − jQr ) (121)

Here, δ = δs − δr . Ignoring the imaginary part and considering the real part of the
above quadratic equation, we obtain:

Vr2 ( Acosα) − Vr (Vs cosδ) + ( Pr Bcosβ + Qr Bsinβ) = 0 (122)

Obtaining the roots of the quadratic equation allows us to obtain:

,
−V s cosδ ± (Vs cosδ)2 − 4Acosα( Pr Bcosβ + Qr Bsinβ)
Vr = (123)
2Acosα
It should contain the actual non-zero values of the Voltage equation and the actual
non-zero roots obtained from the voltage equation discriminant.

(Vs cosδ)2 − 4Acosα( Pr Bcosβ + Qr Bsinβ) > 0 (124)

Therefore, to maintain the system as stable and prevent voltage collapse, the following
conditions should be satisfied:
4Acosα( Pr Bcosβ + Qr Bsinβ)
<1 (125)
(Vs cosδ)2

4Acosα( Pr Bcosβ + Qr Bsinβ)

∴ LCPI = (126)
(Vs cosδ)2
Therefore, the index value should be less than 1 to maintain the system stability, i.e.,
LCPI < 1. At no load, the LCPI value is zero, and the LCPI value is adjacent to the unity
during the system instability.

115
Energies 2023, 16, 6718

2.1.18. New Voltage Stability Index (NVSI)

The NLSI derived in [23] is a similar concept of the power transmission quadratic
equation represented in Equation (8). Determining the NLSI is similar to the process for
determining Lp . Neglecting δ, Equations (19) and (20) represented as:
 2    
Vr2 + 2Qr X − Vs2 Vr2 + X 2 Pr2 + Q2r = 0 (127)

The Voltage Equation Vr is a second-order equation. The limitation to possessing at

 2  
least one solution is: 2Qr X − Vs2 − 4X 2 Pr2 + Q2r ≥ 0
,
2X ( Pr2 +Q2r )
4Qr X −Vs2 ,
≤1
(128)
2X ( Pr2 + Q2r )
∴ NVSI = 4Qr X −Vs2

Hence, to sustain stability, the value of the NLSI must be less than 1.00 in order
toobtain the NLSI by neglecting the line resistance and line shunt admittances.

2.1.19. Integrated Transmission Line Transfer Index (ITLTI)

An integrated transmission line transfer index depends on the radial topology. The
ITLTI’s purpose is identifying the power transfer; moreover, it identifies the weak line
depending on the transmission line parameters, consequent to the unity, lagging, and
leading power factor states. The transmission line in Figure 1 depicts a constant sending-
end voltage Vs ∠δ transferring the power Ss . Hence, the functional ABCD parameters
represent the receiving-end power Sr , Voltage Vr ∠δ with the respective power factor angle
θ r . Further, the receiving-end complex power is derived as:

AVr2 Vs Vr
Sr = Vr Ir = − ∠( β − α) + ∠( β − δ) (129)
B B
Equivalently, the sending-end complex power is represented as:

AVs2 Vs Vr
Ss = Vs Is = − ∠( β − α) + ∠( β + δ) (130)
B B
Considering the radial transmission line [24], we obtain two-power circles with the
same radius and two different centers. Receiving-end complex power circle polar coordi-
nates are represented in Figure 3 with a variable power angle δ.

Figure 3. The receiving-end circle.

116
Energies 2023, 16, 6718

Considering the receiving-end circle and ΔOCR, we obtain:

∠C = ∠δ = ∠( β − α) − ( β − δ) = ∠(δ − α)
∠O = ∠θr = 180 − ∠( β − α) + ∠θr
∠ R = ∠∅r = 180 − ∠δ − ∠θr (131)
Vs Vr AVr2
Sr
sinδ = B
sinθr
= B
sin∅R

Therefore, the relation between the receiving-end and sending-end voltage can be
∅r
derived by Vr = VAs sin
sinθr
. Accordingly, an increased power angle at distinct power factors
decreases the receiving-end voltage. The receiving-end power magnitude is:

Vs2 sin(θr + δ )sinδ

Sr = (132)
AB (sinθr )2

To deliver the maximum power, Sr is partially derived using δ’ and is equated to zero,
which determines the critical value of the power angle.

θr
δr_CR = 90 − +α (133)
2
Critical voltage value:
Vs 1
Vr(CR) = (134)
2A sin θr
2
Maximum Power Transfer:

Vs2 1
Sr(max) =
4AB  2 (135)
θ
sin 22

The maximum limit of the receiving-end power for different power angles is as follows:

sin(θr + δ )sinδ
Sr = Sr(max)  2 (136)
cos θ2r

Moreover, the integrated transmission line transfer index (ITLTI) is expressed as:

sin(θr + δ ) sinδ
Sr_index =  2 (137)
cos θ2r

Hence, the ITLTI determines the system’s stability for all healthy operations. De-
termining the Sr_index with varying power angles, the maximum value should be 1; if it
exceeds 1, the system is unstable. Therefore, the index value is maintained at less than
1. The loadability concerning the critical power angle increased with an increase in the
receiving-end power factor, irrespective of leading or lagging operations. Separating the
active and reactive power of Sr , the active power Pr is expressed as follows:

AVr2 Vs Vr
Pr = − cos( β − α) + cos( β − δ) (138)
B B

2.1.20. Critical Boundary Index (CBI)

The Critical Boundary Index derived in [25] considers the active and reactive power
variations and utilizes arithmetical methods to determine the critical boundaries. This
index is preferable because of its high accuracy in predicting system performance and

117
Energies 2023, 16, 6718

analyzing and adding the actual (Equation (8)) and imaginary (Equation (9)) parts of the
power flow equation (Equation (7)).
 2
Pr Rsr + Qr Xsr + Vr2 + ( Pr Xsr − Qr Rsr )2 = Vs2 Vr2 (139)

 2     
V2
Vr2 + 2 Rsr Pr + Xsr Qr − s Vr2 + R2sr + Xsr
2
Pr2 + Q2r = 0 (140)
2
  √
V2
Vr2 = − Rsr Pr + Xsr Qr − s ± A (141)
2
Imagining that the intensity of the receiving-end and sending-end voltages is more
significant than zero, and rearranging Equation (141), we obtain the following:
#
 2
Vs2
A= Rsr Pr + Xsr Qr − − ( R2sr + Xsr
2 )( P2 + Q2 )
r r (142)
2

Hence, the above equation examines that the power transmission had limits for a
specific time duration. Those limits are acknowledged as power and voltage stability limits.
These limits are logical only when the magnitude of A is 0. Hence, determining the Qr
when A = 0 is written as follows:
,
± Xsr 2 V 4 + R2 V 4 − 4X 2 R V 2 P − 4X 3 V 2 P
s sr s sr sr s r sr s r 2Xsr Rsr Pr − Xsr Vs2
Qr = + (143)
2R2sr 2R2sr

We plot the power stability limit P-Q characteristics with Qr . The P-V characteristic
is the nose curve. The power system stability in Figure 4 includes the unstable and stable
boundaries, starting from a current stable point K (Po , Qo ) and reaching a critical boundary
point C (X, Y). Considering the Lagrange Multipliers, Equation (141) simplifies as:
 2   
Vs2
C ( X, Y ) = Rsr X + Xsr Y − − R2sr + Xsr
2
X2 + Y2 = 0 (144)
2

Figure 4. P-Q Characteristics.

The function f (X, Y) is the length between the current stable point K (Po , Qo ) and the
adjacent point of voltage collapse C (X, Y). The displacement of f (X, Y) is shown as:

f 2 = ( X − P0 )2 + (Y − Q0 )2 (145)

118
Energies 2023, 16, 6718

Here, the ΔPik , ΔQik , stability evaluation index, and displacement in the stability curve
are represented in Figure 4. Further, as shown in Figure 1, the operating point between the
sending-end and receiving-end transmission line can evaluated by:

ΔPsr = X − P,
0 & ΔQsr = Y − Q0
(146)
CBI sr = ΔP2sr + ΔQ2sr

Due to its high-accuracy prediction, this index is preferable. If the index is close to
zero, it is the worst-stability system.

2.1.21. Line Voltage Stability Index (LVSI)

A novel LVSI developed to overcome existing stability indices’ drawbacks, LVSI [26],
is derived from the quadratic voltage equation. This index is also derived considering the
transmission line ABCD parameters; moreover, the inclusion of line charging capacitance
and resistance is ignored by existing stability indices. Consequently, this index evaluates
the voltage sensitivity accurately, subject to all circumstances, and predicts the voltage
collapse point. The receiving-end active power in terms of transmission line parameters is
obtained using:
Vs Vr cos( βδ) AV 2r cos( β − α)
Pr = − (147)
B B
Rearranging the power equation according to a quadratic equation of voltage is
obtained as:
Vs Vr cos( β − δ) Pr B
Vr2 − + (148)
Acos( β − α) Acos( β − α)
The derivation of the voltage concerning the active power to obtain the sensitivity is
shown as:
dVr −B
= (149)
dPr 2V r Acos ( β − α) − Vs cos( β − δ)
Hence, the sensitivity should be negative to maintain the system’s stability.

−B
<0 (150)
2V r Acos ( β − α) − Vs cos( β − δ)

Vs cos( β − δ) − 2V r Acos( β − α) < 0 (151)

Further simplifying Equation (151), the stability index can be shown as:

2V r Acos( β − α)
LVSI = >1 (152)
Vs cos( β − δ)

It is highly stable if the index value is 2 at no load. If the index value is 1 at the
maximum load, it is a collapse point. Therefore, the system to maintain the stability index
value must be greater than the unity.

2.1.22. New Line Voltage Stability Index (BVSI)

The index BVSI derived in [27] depends on the power transmission approach. Consid-
ering Equation (14), the quadratic equation can be rearranged as follows:
 
RV 2r − Vs Vr ( Rcosδ + Xsinδ) + Pr R2 + X 2 = 0 (153)

Deriving the quadratic equation roots, we obtain:

,
Vs ( Rcosδ + Xsinδ) ± (Vs ( Rcosδ + Xsinδ))2 − 4R( Pr ( R2 + X 2 ))
Vr = (154)
2R

119
Energies 2023, 16, 6718

For determining the absolute and non-zero values of Vr , the discriminant is made
greater than or equal to 0:
  
(Vs ( Rcosδ + Xsinδ))2 − 4R Pr R2 + X 2 ≥0 (155)

4RPr Z2
BVSI = ≤1 (156)
(Vs ( Rcosδ + Xsinδ))
Therefore, the index value is less than the unity for the system to be at the stability
limit, such that the index value lies between 0 and 1. If the BVSI value exceeds 1, the
corresponding line encounters a voltage drop, and the system may collapse. The BVSI line
ignores the shunt admittance and the effects of reactive power on stability.

2.2. Bus Voltage Stability Indices

The Bus Voltage Stability Indices evaluate the transmission bus’s voltage stability and
do not derive the weak characteristics of the potential voltage issues across the system.
Therefore, Bus VSIs are unable to be utilized for determining weak performances.

2.2.1. L-Index
The L-Index [28] is derived from the power flow and standard two-bus system and
obtains a detailed analysis. The L index determines peculiar characteristics of power
system susceptibility, measuring significant active power, restructuring weak areas or
buses, predicting voltage collapse, and identifying instability. Considering the matrix Fij of
an ith column and jth row, and elements obtained from the Y matrix, αL, and αG are the
load bus and generator bus. The linear transmission system represents a Hybrid matrix (H).
+ L+ + L + + LL ++ +
+V + + + + F LG ++ ++ I L ++
+ + = H.+ I + = + Z
+V G + +V G + +K GL Y GG +.+V G + (157)

Here, VL , IL : Voltage vector, Current vector at load nodes; VG , IG : Voltage vector,

Current vector at generator nodes; ZLL , FLG , KGL , YGG : H submatrices. By partial inversion,
the H is obtained from the Y-matrix. For any load node j, j ∈ αL , and the corresponding
voltage equation Vj is:
Vj = ∑i∈α Zji .Ii + ∑i∈α Eji Vi (158)
L G

S+∗
Vj2 + V oj V ∗j =
j
(159)
Y+
jj

Substituting the equivalent voltage Voj , admittance Yjj + , and power transferred Si + :

V oj = −∑i∈α F ji V i (160)
G

1
Y+
jj = (161)
Z jj
Hence, Sj + contains two parts:

S+
j = Sj + Sj
corr
(162)

⎛ ⎞
⎜ Z ∗ji Si ⎟
Scorr
j =⎜ ∑ ⎟
⎝ i ∈ α L Z ∗ . V ⎠.V j (163)
jj i
i = j

120
Energies 2023, 16, 6718

The exceptional characteristic of Equation (159) is that its formation is similar to the
admittance equation. Dissimilarities exist between the equivalent voltage Voj and power
transferred Sj . The voltage Voj varies slightly, not remaining as a constant value, since the
generator voltage remains mostly constant at varying loads. Adopting these limitations
can evaluate and regulate local indicators, and Lj can allow us to design for every node j
corresponding to the line.
+ + + + +
+ V oj ++ ++ S j ++
+
L j = +1 + + = + +∗ 2 + = . . . .. (164)
+ V j + + Y jj Vj +

Under stable operations, the Lj ≤ l condition should never disobey any node j. There-
fore, the global indicator L represents the stability of the entire subsystem, expressed as:
0 1
L = MAX L j
j∈α L
+ +
+ ∑ F ji Vi + (165)
+ i∈α G +
L = MAX+1 − +
j∈α L + Vj +

Here, a series of load nodes is αL , and a series of generator nodes is αG . If the L index
value is 0, the system is stable; if the value is 1, the system is approaching instability.

2.2.2. Voltage Instability Proximity Index (VIPI)

The proximity index derived in [29] for estimating instability depends upon coordinat-
ing various load flow solutions and instability. The VIPI is a scalar index that analyses the
stability margin, considering the angle separating the critical and specified-value vectors.
The power flow equation in terms of rectangular coordinates is as follows:

Ys = Y ( x ) (166)

Here, Ys is the specified value; x = (e1 , f 1 , e2 , f 2 , . . . , en , f n ) T ; ei = Bus voltage real

value; fi = Bus voltage imaginary value. Assume two types of voltage vectors (Figure 5):
x is the operating solution, and x* is the fictitious solution; comply with the uniform
specified-value vector. Employing critical vector (a) and deflection vector (b), x and x* can
∗ ∗
be represented as x = a + b, x* = a − b. Further simplifying them, a = x+2x , b = x−2x .
As this proximity index is the angle separating critical vector Y(a), and specified value
vector Ys , the VIPI is as follows:

YsT Y ( a)
V IPI = θ = cos−1 (167)
Ys .Y ( a)

Here, θ is the angle separating two vectors; critical vector Y(a) is in the space of
node specification.

121
Energies 2023, 16, 6718

Figure 5. Concept of VIPI in the Node-Specification Space.

2.2.3. Voltage Collapse Proximity Index (VCPIBUS )

VCPIBUS , proposed in [30], is derived out of the standard power flow equation, and
considering a network’s voltage phasor measurement and admittance matrix determines
the VCPI at every bus. The complex power at bus k of an N bus network is represented as:
⎡ ⎤
⎢ N ++ + + + +⎥
Sk∗ = |Vk |2 − (|Vk |cosδk − j|Vk |sinδk )⎢ ++ + + + + ⎥
⎣∑m = 1 Vm cosδm + j Vm sinδm ⎦Ykk (168)
m = k

The term Vm in the above equation is expressed as:

Ykm
Vm = N
Vm (169)
∑ j=1 Ykj
j=k

Hence, the right-hand term in Equation (168) is complex in the form of a − jb. Let the
two equations contain two unknown terms (Vk , δ).
+ +
f 1 (|Vk |, δ) = |Vk |2 − ∑m = 1 +Vm +|Vk |cosδ
N
(170)
m = k
+ +
∑m = 1 +Vm +|Vk |sinδ
N
f 1 (|Vk |, δ) = (171)
m = k
A partial derivative matrix is attained by simplifying two equations for evaluating
the unknowns with the Newton-Raphson Methodology. During the voltage collapse, the
matrix determinant is equal to 0, which results as:

|Vk |cosδ 1
= (172)
N
|
∑ m =1 m V | 2
m=k

Further simplifying Equation (172), the voltage collapse prediction index (VCPI)
determined at bus k is: + +
+ +
=1 |Vm | +
N
+ ∑m
+ m=k +
VCPI kth bus = +1 − + (173)
+ Vk +
+ +

122
Energies 2023, 16, 6718

Therefore, the above equation is the voltage collapse condition for bus k. The VCPI
value fluctuates between 0 and 1, where 0 represents a stable system, and 1 represents
voltage collapse.

2.2.4. S Difference Criterion (SDC)

The S Difference Criterion (SDC) [31] needs a protective relay to perform arithmetic
derivation and utilizes the successive evaluation of apparent power (S) at the relay point.
During the voltage instability environs, the inflation of sending-end power flow delivers
the transmission losses. Consequently, the rise in the sending-end apparent power does
not afford a rise in receiving-end power. It specifies that ΔS = 0 at the relay point during
the voltage instability. Hence, the transfer of apparent power to the receiving end is:

(k) (k) (k)∗

Sj = V j I ji (174)

Considering the change in a time period Δt, the time interval changes to tk = tk + Δt. A
respective rise in apparent power is as follows:

  
( k +1) (k) ( k +1) (k) ( k +1) (k) ( k +1) ∗ (k) (k+1) (k)∗ (k) (k+1)∗ ( k +1) (k+1)∗
Sj = Sj + ΔS j = Vj + ΔV j . I ji + ΔI ji = S j + ΔV j I ji + V j ΔI ji + ΔV j ΔI ji (175)
 
≈0

( k +1) (k+1)∗
Here, the term ΔV j ΔI ji is very small and can be ignored.

( k +1) (k+1) (k)∗ (k) (k+1)∗

ΔS j = ΔV j I ji + V j ΔI ji =0 (176)

(k+1) (k)∗
ΔV j I ji
1+ (k) (k+1)∗
= 1 + ae jϕ = 1 + a(cosϕ + jsinϕ) = 0 (177)
V j ΔI ji

Thus, the SDC is a positive real value; then, considering the magnitude of Equation (177),
we obtain: + +
+ + ++ ΔV j
(k+1) (k)∗ +
I ji +
+ jϕ +
SDC = +1 + ae + = ++1 + (k) (k+1)∗ ++ (178)
+ V j ΔI ji +

Intending to achieve voltage collapse, the value of ΔS = 0, and correspondingly, the

SDC equates to 0. The relay operates if the SDC value is below a predetermined threshold.

2.2.5. Impedance Stability Index (ISI)

The ISI bus stability index depends on Tellegen’s rule, an index derived in [32]
made simple in evaluating Thevenin’s variables. Let a bus r with a complex load (Sr )
and impedance (Zr ) be linked to a power system; the remaining system is expressed as
Thevenin’s equivalent with the variables E and Zth . Let the current equation be:

Sr /Vr = Ir * = ((E − Vr )/Zth )* (179)

Vr (E − Vr )* − Sr Zth * = 0 (180)
Two voltage solutions empower the phasor of Equation (180) with a power Sr ; if the
two solutions are equal, it allows maximum power transfer.

Vr = (E − Vr )* (181)

123
Energies 2023, 16, 6718

Equation (181) is fundamental to a local voltage stability approach that conveys

predetermined consequences through circuit theory.

| Zr | = | Z th | (182)

The system delivers the maximum power when the load-impedance magnitude equals
Thevenin’s impedance magnitude. The adjacent Thevenin network N̂ possesses an identical
analysis to the base network N. Hence, the derivation is similar to putting a hat (ˆ) on every
term, which signifies the adjoint network parameters. The divergence order of Tellegen’s
theorem is equivalent to the following:
 
∗
Î r ΔV r − V̂ r ΔI r = 0 (183)

Here, ΔV r and ΔI r∗ Express the advancement of the complex voltage and current at
bus ‘r’ regarding the Vr and Ir * base values. The adjacent network load impedance Ẑr is
described as a ratio of phasor voltage V̂ r to phasor current Î r .
+ + + +
Ẑr = + Ẑr + = +V̂ r / Î r + (184)

The load impedance specifies that it is the base case; moreover, it is determined at an
instant earlier than that of the network that is regulated to power disturbances.

Ẑr = |V r /I r | (185)
∗
For determining the Thevenin’s impedance, the current equation is substituted in
Î r
Equation (183) from Equation (179).
  ∗ ∗
Ê − V̂ r / Ẑ th ΔV r − V̂ r Δ Î r = 0 (186)

 ∗  
∗ ∗
Ẑ th = Ê − V̂ r / V̂ r Δ Î r ΔV r (187)

Similarly, from Equation (181), the condition for the maximum power transfer of
adjacent network can be expressed as:
 ∗
V̂ r = Ê − V̂ r (188)

Analyzing Equations (187) and (188), during the system collapse point, the equivalent
Thevenin’s impedance is expressed as:
∗
Ẑ th = ΔV r /ΔI r∗ (189)

∗
Ẑ th = ΔV̂ r /ΔI r (190)
However, the above two equations are conjugate with each other; for discarding the
conjugate values, Thevenin’s impedance absolute value is as follows:
+ +
Ẑth = + Ẑ th + = |ΔV r /ΔI r | (191)

The contingency during the regular loading is:

Ẑr  Ẑth (192)

During the commencement of voltage instability, the dissimilarities among these

impedances converge to 0. Moreover, at that instant of collapse, the impedances are
identical, i.e., Ẑth = Ẑr .
|ΔV r /ΔI r | = |V r /I r | (193)

124
Energies 2023, 16, 6718

This index precisely determines Thevenin’s impedance by evaluating the advancement

in voltage and current following the base-case exposure to disturbances. Furthermore,
the load impedance is evaluated through the ratio of the voltage to the current. Using
Equations (192) and (193), a modest regulated impedance stability index (ISI) for identifying
the voltage stability margin is:
 
ISI = Ẑr − Ẑth / Ẑr (194)

Further simplifying by substituting the impedances in terms of voltage and current,

the stability index derives as:

ISI = 1 − | I r ΔV r |/|V r ΔI r | (195)

2.2.6. Voltage Stability Index (VSIBUS )

The VSI bus index relies on system elements to identify the distance between the
operating and collapse points. This index [33] derives from power flow equations that are
similar to the assumptions of the SDC. Let ‘i’ be one local bus of the system. The complex
load power of bus i is:
Si = Si ∠θ i = ( Pi + jQi ) = Vi Ii∗ (196)
The complex bus voltage is Vi = Vi ∠δi, and the complex load current is Ii = Ii ∠(δi − θ i ).
Employing Taylor’s theorem, the interrelation among the advancement difference between
Vi and Ii as a consequence of advancement evolution in Si is expressed as:

∂Si ∂S
ΔSi = ΔI + i ΔV i + higher order terms (197)
∂Ii i ∂Vi

Meanwhile, ignoring the higher-order terms,

ΔSi = Vi ΔIi + Ii ΔVi (198)

Usually, if the load increments, there is a corresponding increase in load current and
decrease in load voltage. This means that, if the value of ΔSi is positive, Current ΔIi is
positive and the ΔVi Voltage is negative. Although the bus reaches a critical point or
voltage collapse, adding additional load on a bus might not be able to raise the apparent
power Si of the load due to rapid decrement in voltage in contrast to the current increment.
Going through this case converges ΔSi to zero. Accordingly, the power ΔSi through the
voltage-stability limit is expressed as:
  
Ii ΔVi
0 ≤ 1+ (199)
Vi ΔI i

Considering that the power of Equation (199) is made to α (>1.0), an actual value, it
determines almost the same level of linear characteristics. Therefore, excluding the general
loss, the voltage stability index at the ‘i’ bus is as follows:
   α
Ii ΔVi
VSI i = 1 + (200)
Vi ΔI i

Hence, at no load, the VSI varies from 1, and at voltage collapse, the value varies to 0.

2.2.7. ZL /ZS Ratio

The ZL /Zs ratio index proposed in [34] depends on a similar technique to the ISI.
The correspondence index is the load bus impedance (ZL ) ratio to Thevenin’s impedance
(Zs ). The voltage stability margin was generally evaluated from different scenarios. The
evaluation of the stability margin depends on the determination of the voltage stability. For

125
Energies 2023, 16, 6718

an equivalent circuit, the derivation of the apparent load power concerning load admittance
is as follows:  
dS V 2 1 − ( ZS Y )2
= (201)
dY 1 + ( ZS Y )2 + 2ZS Ycosβ
Whereas Y = 1/ZL , dividing both sides of the equation with V2 :

dS∗ 1 − ( ZS Y )2
= (202)
dY ∗ 1 + ( ZS Y )2 + 2ZS Ycosβ

While dS*/dY* = (Y/S) (dS/dY), thenthe Equation (202) together evaluates (ZL /Zs ):

ZL M+1
=  0.5 (203)
ZS
− Mcosβ + ( Mcosβ)2 − M2 + 1

Here, M = (dS*/dY*), and the relevance between ZL /ZS and M relies upon the angle
β. The estimation of β is a phase angular difference between the source and the load,
i.e., β = ϕS − ϕL . The definite value of ϕS operates on the existent form of the system.
The parameter M is determined by evaluating the derivative of the apparent load power
to admittance:  ∗
dS (S − S1 )(Y2 + Y1 )
M= = 2 (204)
dY ∗ (S2 + S1 )(Y2 − Y1 )
Here, S and Y refer to the load power and admittance, respectively. Moreover, suffixes
1 and 2 refer to the starting of a change (time, t1 ) and completion of the change (time, t2 ),
respectively. The variation between the two-time measurements needs to be around 500
ms. If the index value remains above 1, it is a reliable condition.

2.2.8. Equivalent Node Voltage Collapse Index (ENVCI)

The index ENVCI depends on the ESM (Equivalent System Model) and local voltage
phasors [35]. The ENVCI has numerous advantages like accurate evaluation and modeling;
moreover, this index is convenient for performing real-time and online operations. The
outgoing power at node N of a single-line equivalent model (Figure 6) should perform the
consequent equation of plain power flow.
⎛→ →
⎞∗
→ E − V
Pn + jQn = V n .⎝ ⎠
k n
(205)
Zkn

Let the phasor voltage equations for two nodes of a single-line equivalent network in
→ →
terms of rectangular coordinates be E k = ek + j f k and V n = en + j f n , then the correspond-
ing line impedance is Zkn = Rkn + jXkn . Separating the real terms and imaginary terms of
the power flow equation, we obtain:

Pn Rkn + Qn Xkn = en (ek − en ) + f n ( f k − f n )

(206)
Pn Xkn − Qn Rkn = f k en − ek f n

126
Energies 2023, 16, 6718

Figure 6. Equivalent System Model.

Hence, the power flow equation, Equation (206), is crucial for determining the phasor
voltage at the receiving node, describing the system’s performance. In contrast, the phasor
voltage of the sending node is known. Whenever the respective ESMs of all equivalent
lines possess an arithmetical solution of the receiving node, signifying that they contain
every node voltage of the system, at that instant, the system should possess a complete
power flow solution; moreover, this should influence the voltage stability. The system’s
stability relies on evaluating Equation (206) at every node. The evaluation of Equation (206)
can be examined by the Jacobina matrix singularity:
 
ek − 2en fk − 2 f n
J= (207)
fk −ek
 
det( J ) = 2(ek en + f k f n ) − e2k + f k2 = 0 (208)

Therefore, Equation (208) specifies a novel stability index that is known as the Equiva-
lent Node Voltage Collapse Index:
 
ENVCI = 2(ek en + f k f n ) − e2k + f k2 (209)

The ENVCI expression in terms of polar coordinates is:

ENVCI = 2Ek Vn cosθkn − En2 (210)

Determining the ENVCI requires only the phasor voltage of two nodes. Every node
contains ESM. The system converges to a collapse point if the ENVCI of a minimum of one
node is approaching 0; the respective node is the weak node that induces system instability.

2.2.9. Power Stability Index (PSI)

The power stability index [36] was developed to identify the optimal position of the
Distributed Generator (DG), including critical buses that are nearer voltage collapse. This
index derives from a standard two-bus system, which is stable within the unity margin.
Let the power flow in a standard two-bus system without DG be:

S L = PL + jQ L = Vr Ir∗ (211)

V r = V r − Ir Z (212)

127
Energies 2023, 16, 6718

where
( PL ) − j( Q L )
Ir = (213)
Vr∗
Now, with DG, the current Ir concerning the active power and reactive power support
are expressed as:
( P − PG ) − j( Q L − QG )
Ir = L (214)
Vr∗
Combining Equations (212) and (214), and writing real and imaginary terms separately,
we obtain:
|Vr ||Vs | |Vr |2
PL − PG = ∗
cos(θ − δs + δr ) − cos(θ ) (215)
Vr Z

|Vr ||Vs | |Vr |2

Q L − QG = sin ( θ − δs + δr ) − sin(θ ) (216)
Vr∗ Z
Rearranging Equation (215), we obtain:

|Vr ||Vs |cos(θ − δ) Z ( P L − PG )

|Vr |2 − + =0 (217)
cos(θ ) cos(θ )

Hence, Equation (217) is in a quadratic form in terms of Vr . When stable, the node
voltage must contain real roots, i.e., the discriminant is more significant than zero, and the
index derives as:
4Rij ( PL − PG )
PSI = ≤1 (218)
[|Vi |cos(θ − δ)]2
If the PSI value is less than 1, the system is stable; if the value is nearer to zero, the
system is more stable. Hence, the PSI determines the optimal location of the DG as the
calculated value of the PSI for each line patterned from maximum to minimum value. The
optimal location of the DG is at the line-end with a maximum value of the PSI.

2.2.10. Voltage Deviation Index (VDI)

The Voltage Deviation Index proposed in [37] determines the voltage deviation’s real
value compared to 1 per unit. Assuming an N-bus system, the total VDI is the addition of
all N buses’ voltage deviation index.
+ +
VDI j = +1 − Vj + (219)

+ +
∑ j=1 +1 − Vj +
N
VDI T = (220)

2.2.11. Simplified Voltage Stability Index (SVSI)

An advanced SVSI proposes [38] to evaluate the system’s stability. This SVSI depends
on the Relative Electrical Distance (RED) concept, which considers identifying the conve-
nient generator to a particular load bus; moreover, the relation of electrical parameters for
enhancing the performances.
• Relative electrical distance (RED)
The concerned system, the interrelation among the current (I), Generator bus voltage
phasor (V), and the load buses (L) represented in the form of an admittance matrix can be
shown as:     
IG Y YGL VG
= GG (221)
IL YLG YLL VL

128
Energies 2023, 16, 6718

Rearranging Equation (221), we obtain:

    
VL ZLL FGL IL
= (222)
IG KGL YGG VG

FLG = −|YLL |−1 |YLG | represents a composite matrix, which indicates the association
between the voltages of the source and load buses. The REDs determined from the matrix
FLG are shown as:
 
R LG = [ A] − abs[ FLG ] = [ A] − abs |YLL |−1 |YLG | (223)

Here, the size of matrix A is (n-g) X g, the total number of buses is represented by n,
and the total generator count is denoted by g. Matrix A is a particular matrix in that all
elements are 1, i.e., a matrix of ones. The data obtained from matrix RLG are considered
rather than path algorithms for determining the electrical distance from the generator bus
to the load bus. Subsequently, after identifying the shortest distance of the generator to
a particular load with the RLG matrix, the Thevenin impedance voltage drop ΔVi can be
estimated with: + + + +
n j −1 + → → + +→ →+
∼ +V g − V i +
ΔVi = ∑b=1 ++ V b − V b+1 ++ = (224)
+ +

Vl and Vg are the phasor voltages at the load bus and closest generator. While an
increase in demand for power reaches the maximum loading point in one particular bus or
various buses in the system, some buses may experience high voltage drops. This process
illustrates the potential of voltage collapse and comprises the formation of the SVSI in
terms of a correction factor β.

β = 1 − (max (|Vm | − |Vl |))2 (225)

The derived correction factor is proportionate with higher variations between the
voltage magnitude of bus m and bus l, which can be determined directly by measuring the
PMU during the power system analysis, considering particular operating conditions. Thus,
the SVSI is as follows:
ΔVi
SVSI i = (226)
β × Vi
Suppose the obtained value of the proposed index is near the unity. In that case, the
system considers the voltage instability on these terms: the Thevenin impedance voltage
drop equals the voltage of the load bus.

2.2.12. P-Index
The P-index [39] evaluates the distance to collapse and measures the load shedding.
The P-index is derived from a standard two-bus system, shown in Figure 1. The load at
the receiving end bus is Pr + jQr , and Vr is the voltage magnitude. The equivalent load
admittance is Gr − jBr ,
Pr Qr
Gr = 2 , Br = 2 (227)
Vr Vr
Let ΔPr , ΔQr be an incremental increase in load without varying the power factor. The
equivalent increment in admittance is ΔGr , ΔBr . The voltage magnitude changed by an
amount of ΔVr , is typically negative; therefore, the updated bus voltage is Vr + ΔVr . The
corresponding active power variation is as follows:

ΔPr = (Vr + ΔV r )2 ( Gr + ΔGr ) − Vr2 Gr

(228)
= (Vr + ΔV r )2 ΔGr + (2V r + ΔV r ) Gr ΔV r

129
Energies 2023, 16, 6718

The substantial consequences of the two terms in Equation (228) are: the primary term
is positive and describes the power attained because of the additional load ΔGr , while the
secondary term is negative and is the power wasted on the original load Gr because of
the voltage drop ΔV. Because of these opposite terms, the overall active power gained
is balanced. At the stability limiting point, these two terms diminish, and the overall
increment in power is zero.
The P-index bus stability index depends on the ratio of two terms in Equation (228),
i.e., the ratio of power lost to power gained. When ΔV is negative for positive ΔGr , a minus
sign is added to make the index positive.

(2V r + ΔV r ) Gr ΔV r
P-index = − . (229)
(Vr + ΔV r )2 ΔGr

In the limiting case, such as ΔGr , ΔV → 0,

2Gr dVr
P-index = − . (230)
Vr dGr

Although dVr /dGr is often not defined in terms of network terminology, it is simple to
do so in terms of the power and voltage sensitivity of the system. Provided that dVr /dGr :

dV r dV r dPr
= . (231)
dGr dPr dGr

Thus, differentiating the power, Pr = Vr 2 Gr is as follows:

dPr = Vr2 dGr + 2V 2r Gr dV r (232)

Or:
dPr dV r
= Vr2 + 2Vr Gr (233)
dGr dGr
Substituting Equation (233) in Equation (231), we obtain:
 
dV r dV r dV r
= Vr2 + 2Vr Gr (234)
dGr dPr dGr

Further simplifying Equation (234), it is reduced as:

dV r Vr2 dV
dPr
r
= (235)
dGr 1 − 2Vr Gr dV r
dPr

Substituting Equation (235) in the P-index Equation (230), we obtain:

−2Vr Gr dV
dPr
r
P-index = (236)
1 − 2Vr Gr dV
dPr
r

Expressing Equation (236) in terms of active power, we obtain:

−2 VPrr × dV r
dPr
P-index = (237)
1 − 2V
Pr
r
× dV
dPr
r

Therefore, the P-index is expressed in terms of power and voltage sensitivities. When
dV/dPr = ∞, the theoretical value of the stability index is 1.0 at the stability limit. The
stability limit varies from 0 (no load) to 1.0 at the collapse point of the system.

130
Energies 2023, 16, 6718

3. Summary
The Voltage Stability Indices that are preferred in power system analysis are derived
in detail in this paper. Our review of stability indices examined the overall parameters of
voltage stability analysis, including the voltage stability margin, stability indices, identify-
ing voltage collapse proximity, and the maximum loadability. This comprehensive review
features the significance of definite stability assessment and specifies the effectiveness and
vulnerability of various bus and line Voltage Stability Indices.
The line stability indices, like the VSLI, Lmn , LQP, LP, FVSI, NLSI, VQI, and NVSI,
and the bus stability indices like the L-Index, Voltage Instability, and VCPI, are derived
from the power flow in a standard two-bus system. In contrast, the stability indices like
the VCPI, PTSI, VSI_1, VSM, SDC, and PSI are determined by considering the maximum
power transfer and maximum power losses. At the same time, stability indices like the SI,
LCPI, LVSI, and BVSI derive from the quadratic voltage equation—the overall objectives of
the listed stability indices are expressed in Figure 7.

Objective of Voltage Stability Indices

Line Voltage Stability Indices Bus Voltage Stability Indices

Identifies the critical lines Estimates the impact of

and buses load variations
Calculate the Voltage
Determine Weak buses
stability
Simplify the system Compute the efficacy of
planning & Operation voltage control
System Reliability System reliability
Enhancement enhancement

Figure 7. The objective of Voltage Stability Indices.

The VSLI stability index depends on the load reactive power demand ratio to the
available reactive power. It helps to determine the critical points and identifies the load bus
nearer to the collapse point. The Lmn stability index is online monitoring that predicts the
voltage collapse; the reliability test on the IEEE 24 bus system results in an accurate voltage
collapse prediction. It also identifies the stressed lines and weak areas that are exposed to
voltage collapse; moreover, it allows for flexible operation. The LQP is a static indicator
that performs faster, evaluates stability, and determines the origin of voltage collapse.
The VCPI is a proximity index that determines the range of the collapse point; this
index depends upon the systems’ generation and load characteristics and identifies ex-
posure to voltage collapse. The VCPI is a method for the real-time prediction of voltage
collapse. This proximity index illustrates the collapse position and is an alarming tool for
preventing system collapse. LP performs voltage stability assessment in a radial distribu-
tion system. It predicts instabilities easily when changing the load. The FVSI can determine
the collapse point, weakest bus, critical line, and maximum loadability. The VSLBI is used
for local monitoring of threshold voltage collapse and emergency management during
voltage-sensitive loads. The VSLBI operates within a small stability margin and maintains
reactive reserve power. The VSMI presents data about weak locations, evaluates the voltage
stability margin, and maintains stable voltage levels during load variations. Moreover, the
VSMI depends on the relation between voltage stability and angular difference between the
receiving- and sending-end buses. VCPI_1 is an online voltage stability evaluation index
that depends on power transmission lines and identifies the weakest transmission lines by
distinguishing the minor power outage buses, including the electrical distance.

131
Energies 2023, 16, 6718

The Critical Voltage (Vcr ) specifies the minimum voltage where the system performs
away from voltage collapse; the system is unstable if the Jacobian power flow matrix is
singular. In Vcr , the Jacobian power matrix is a static voltage stability index specifying the
range of the voltage stability limit. The PTSI is an essential technique that forecasts dynamic
voltage collapse, calculates the effect of adding additional equipment like generators or
transmission lines, and determines the weak points in the system. VSI_1 is a novel online
stability index for predicting steady-state voltage stability. Coordinated with the network
simplification method, it presents the stability margin of every load bus, sequentially
determining the buses which are vulnerable to voltage collapse.
The NLSI performs by varying the active and reactive power and evaluates the voltage
stability more accurately than other indices that vary only the reactive power. Moreover,
the NLSI evaluates the voltage collapse point, specifies the order of transmission lines
based on loadability, and considers the consistency of wind and solar power output and
the response time of the system. The SI identifies the most vulnerable bus that is exposed
to voltage collapse in the radial distribution system, is easily implemented for load flow
analysis, and does not need extra parameters. The VSM is the variation between the
minimal permissible voltage and absolute voltage at a particular bus; it determines the
loss of voltage contribution during the voltage collapse. The VSM can efficiently prevent
voltage collapse from voltage crisis, and can be measured by PV and QV curves. The VQI
is an effective technique that determines the critical bus, detects instability in large-scale
systems, and determines the distance of the collapse point.
The LCPI combines the influence of the proportional flow of active and reactive
power with ABCD parameters, determines the consequences of different power system
arrangements, and specifies the most sensitive lines. The NVSI evaluates the voltage
stability effectively by varying the active and reactive power, determining the weakest
bus and sensitive line, and utilizes the Reactive Power Sensitivity theory. The ITLTI
was first derived for radial transmission networks and subsequently adapted for larger
systems. The ITLTI contains extended integrated significance; moreover, it is probably
a better resolution for online stability assessment and specifies the weakest line and the
specific data about a critical voltage instability range. The CBI operates with the Lagrange
Constant Computational Method (LCM); the main motive of this index is to determine the
critical boundaries and voltage stability with lesser parameters. Varying the active and
reactive power predicts voltage stability margin precisely with better accuracy. The LVSI
determines the stability margin by considering the transmission line (ABCD) parameters
and is expressed in terms of MVA. The LVSI can determine the voltage collapse point
and critical lines, and involves only the information on the phasor bus voltage. The BVSI
forecasts the voltage security level, considering various operating conditions and identifies
the sensitive lines that are nearer to stability. This index can identify the capability of the
bus to place a single DG and size effectively; moreover, it can accurately identify the critical
lines and weakest bus.
The L-index in online testing depends on the eigenvalues of the Jacobian matrix of a
network; these eigenvalues determine the gap between the voltage collapse point and the
current operating point. Moreover, the L-index measures stability linearly, and non-linear
behavior may not be measured thoroughly. The VIPI determines very sensitive buses that
are exposed to voltage instability and finds the possible precautions for preventing voltage
collapse; the VIPI depends on the relation between the instability and various load flow
solutions. The VCPI is a real-time prediction that determines the system’s proximity to
voltage collapse, which is evaluated by analyzing the present voltage and voltage levels
at the point of collapse and utilizing the magnitude of voltage and angle; the admittance
matrix forecasts the voltage collapse. The SDC’s utilization in the protection algorithm for
voltage collapse depends on the local bus phasor of the voltage and current at every line’s
relay point. The relay operation is local to voltage collapse, and the overall rise in apparent
power is because of transmission losses to the supply.

132
Energies 2023, 16, 6718

The ISI calculates the system’s stability; this index illustrates the ratio of variation in
system impedance to load impedance. The ISI is a local stability index that is derived using
Tellegen’s theorem and adjoint network. The VSIBUS is a simple index which identifies
the distance to the collapse point with the help of local data like the magnitude of voltage
and load current, and thoroughly examines the influence of line tripping, reactive power
limits, and non-linear variation in the load. The ZL /ZS measures the derivation of apparent
power concerning admittance, quickly calculating both the power and admittance results
in switching the impedance load tap-changing of a transformer. The SVSI utilizes the data
of present operating conditions combined with phasor voltage measurement. The SVSI is
easily accessible and less computationally derived.
P-index calculation is simple. It involves the system model, nodal power, and only
the single datum of system states. The only computation performed is Jacobian matrix
inversion, which can be attained smoothly. Therefore, the P-index is appropriate for online
voltage stability analysis. Mainly, the load shedding is evaluated by the P-index. Instead of
system-wide shedding, the method makes use of single-load shedding.
The Stability index’s concept, assumptions considered during derivation, mathematical
equation, condition for index stability, objective, and some references are represented in
Tables 1 and 2 of the line Voltage Stability Indices and bus Voltage Stability Indices, respectively.
Figure 8 represents the number of articles published on the line and bus Voltage
Stability Indices in the last decade. It shows an eventual rise in research in Voltage Stability
Indices to enhance voltage stability and supply the desired power under various load
conditions. Table 3 represents some distinct characteristics of Voltage Stability Indices that
are the optimal location of the DG and sizing of the DG, as well as dependency on VSI.

200
180
Articles Published

160
140
120
100
80
60
40
20
0

Year

Figure 8. Number of voltage stability indices articles published.

133
 Table 1. Line Voltage Stability Indices.

VSI Concept Assumption Equation Condition for Stability Objective References

Identify the critical

VSLI Power flow in a transmission Shunt admittance 4[Vs Vr cos(δ)−Vr2 cos(δ)2 ] Stable: VLSI < 1
VSLI = buses near voltage [40–42]
(1995) line of a two-bus system. is ignored. Vs2 Unstable: VLSI > 1
collapse.
Energies 2023, 16, 6718

The effect of active Online monitoring that

Lmn Power flow in a transmission power is ignored, and 4xQr Stable: Lmn < 1 predicts the voltage
Lmn = [43–45]
(1998) line of a two-bus system. the shunt admittance is [Vs sin(θ −δ ]2 Unstable: Lmn > 1 collapse and identifies
approximately zero. the stressed condition.
   Stable: LQP < 1 Performs faster,
LQP Power flow in a transmission Y≈0& X P2
LQP = 4 Qr + X Vs2 Unstable: evaluates the static [46–48]
(1998) line of a two-bus system. R/Z << 1 Vs2 s
LQP > 1 voltage collapse.
Pr
VCPI (1) = Pr(max)
V2 cos∅
Pr(max) = Zss (θ −∅)
4cos2 2
VCPI (2) = Q Qr
r (max ) Determines the range of
Vs2 sin∅
Maximum Power loss & Constant Power Factor Qr(max) = the collapse point,
VCPI Zs 4cos2 (θ −∅) Stable: VCPI < 1
Maximum Power transfer in Constant Shunt 2 which depends upon the [49–52]
(1998) P Unstable: VCPI > 1
a Transmission Line. admittance is ignored. VCPI (3) = P l systems’ generation
l (max )
Vs2 cosθ load characteristics.
Pl (max) = Zs 4cos2 (θ −∅)
2
VCPI (4) = Q Qr
r (max )
Vs2 sinθ

134
Ql (max) = Zs 4cos2 (θ −∅)
2
Voltage stability
Shunt admittance is assessment in a radial
Lp Power flow in a transmission 4RPr Stable: Lp < 1
ignored, and the effect of Lp = distribution. Simple [53–56]
(2001) line of a two-bus system. [Vs cos(θ −δ)]2 UnstableLp > 1
reactive power. index with higher
accuracy.
Shunt admittance is Determines collapse
Stable: FVSI < 1
FVSI Power flow in a transmission ignored and sinδ ≈ 0, 4Z2 Qr point, weakest bus,
FVSI = Unstable: [57–60]
(2002) line of a two-bus system. cosδ ≈ 1, Rsinδ ≈ 0, Vs2 X critical line, and
FVSI > 1
Xcosδ ≈ X. maximum loadability.
The Access to voltage
During the maximum power Thevenin equivalent collapse is local
Stable: VSLBI > 1
VSLBI conditions, the voltage drop impedance connected to Vr monitoring, watchful,
VSLBI = ΔV Unstable: [61,62]
(2003) in impedance equals load the sending end bus is and emergency control
VLSBI < 1
bus voltage. ignored. during voltage-sensitive
load.
VSMI considers the relation
Stable: VSMI > 0 Evaluate stability
VSMI between the angular The shunt admittance δrmax −δ
VSMI = δrmax Unstable: margin and determine [63–66]
(2004) difference of voltage and neglected
VLMI < 0 weak locations.
maximum power transfer.
 Table 1. Cont.

VSI Concept Assumption Equation Condition for Stability Objective References

Online evaluation index.

The voltage drop across the Thevenin equivalent
Identify the weakest
VCPI_1 Thevenin impedance is impedance connected to VCPI_1 ≥ 0: Stable
VCPI_1 = Vr cosδ − 0.5Vs lines by distinguishing [67,68]
(2005) equal to the load; at the the sending end bus is VCPI_1 < 0: Unstable
Energies 2023, 16, 6718

the minor power outage

collapse point. ignored.
buses with distance.
E
Specifies the minimum
Vcr = √ The system is unstable if
Vcr The load flow equations and Constant Power factor 2(1+cos(θ −∅)) voltage where the
the Jacobian power [69,70]
(2006) Eigenvalue theorem. and load. E system performs away
Vcr = 2cosθ matrix is singular
from collapse.
Predicts the dynamic
Maximum Power loss and
Stable: PTSI < 1 voltage collapse and
PTSI maximum power Shunt admittance 2S L Z (1+cos(θ − ϕ))
PTSI = Unstable: calculates the effect of [71–74]
(2006) transferable through a line is ignored. Vs2
PTSI > 1 adding additional
are limited.
equipment.
Predicts steady-state
Maximum Power loss and
The resistance of line Stable: VSI_1 > 0 voltage stability,
VSI_1 maximum power P VSI = 
and shunt admittance is margin Q Smargin Unstable: determines the stability [75,76]
(2006) transferable through a line min Pmax , Qmargin ,
ignored. max Smax VSI_1 < 0 margin of every load
are limited.
bus.
The minimal angular By varying active and
Power flow in a transmission
difference between reactive power,
NLSI line of a two-bus system, Rsr Pr + Xsr Qr Stable: NLSI < 1
receiving and sending NLSI = evaluating the collapse [42,77,78]
(2007) “Critical Clearing Time” 0.25V 2s Unstable: NLSI > 1
voltage and the shunt point, Rank

135
(CCT).
admittance is ignored. transmission lines.
Predict the most
SI The shunt admittance is SI (r ) = 2Vs2 Vr2 − Vr4 −
  Stable: SI = 0 vulnerable bus exposed
Voltage Quadratic Equation. [79–81]
(2007) ignored. 2Vr2 ( PR + QX ) − | Z |2 P2 + Q2 Unstable: SI = 0 to collapse in the radial
distribution system.
Zr − Z0 Determining the loss of
Maximum Power loss and Constant power factor VSM Z = Stable: VSM > 0
VSM Z0 voltage, VSM can
maximum power transfer in and shunt admittance is Vs2
Unstable: [82–85]
(2009) Scr = efficiently prevent
a transmission line. ignored. 2Z0 [1+cos(θ0 −θr )] VSM < 0
voltage collapse.
Determines the critical
bus, detects the
Zero angular difference
VQI Power flow in a transmission 4Qr Stable: VQI < 1 instability in large-scale
and shut admittance are VQI = ≤1 [86–89]
(2010) line of a two-bus system. | Bsr ||Vr |2 Unstable: VQI > 1 systems, and the
ignored.
distance of the
collapse point.
Combines the influence
LCPI The transmission lines 4Acosα( Pr Bcosβ+ Qr Bsinβ) Stable: LCPI < 1 of the relative flow of
Voltage Quadratic Equation. LCPI = [90–93]
(2012) model is like π model. (Vs cosδ)2 Unstable: LCPI > 1 reactive and active
power flow with ABCD.
 Table 1. Cont.

VSI Concept Assumption Equation Condition for Stability Objective References

Evaluate the voltage

Power flow in a The Shunt admittance , stability effectively by
NVSI 2X ( Pr2 +Q2r ) Stable: NVSI < 1
transmission line. Reactive and line resistance are NVSI = varying the P & Q, [94–97]
(2013) 4Qr X −Vs2 Unstable: NVSI > 1
Energies 2023, 16, 6718

Power Sensitivity (RPS). ignored. determining the weak

bus and sensitive line.
ABCD parameters, the For radial transmission
power factor of the receiving Two power circles with networks and
ITLTI sin(θr +δ )sinδ Stable: ITLTI < 1
end, and the power angle two distinct centers but Sr_index =  2 subsequently adapted [4,98]
(2016) θ Unstable: ITLTI ≥ 1
between the receiving and identical radius. cos 2r for larger systems, the
sending ends. Weakest line.
Operates with Lagrange
Constant Computational
Negligible system , Method (LCM);
CBI Active and Reactive power Stable: CBI > 1
impedance, linearized CBI sr = ΔP2sr + ΔQ2sr Determines the critical [99–102]
(2018) changes. Unstable: CBI = 0
power flow model. boundaries and voltage
stability with lesser
parameters.
Evaluate the stability
line’s resistance and
LVSI Voltage Quadratic Equation, 2V r Acos( β−α) Stable: LVSI > 1 margin considering the
charging capacitance are LVSI = Vs cos( β−δ)
[103–105]
(2018) ABCD Parameters. Unstable: LVSI < 1 ABCD parameters
ignored.
expressed in MVA.
Line shunt admittance Optimal location and
BVSI 4RPr Z2 Stable: BVSI < 1
Voltage Quadratic Equation. and the reactive power’s sizing of distributed [27]

136
BVSI = (Vs ( Rcosδ+ Xsinδ))
(2022) Unstable: BVSI > 1
effects are ignored. generations.

Table 2. Bus Voltage Stability Indices.

VSI Concept Assumption Equation Condition for Stability Objective References

+ +
L–Index Power flow equation All generator voltages + ∑i∈αG F ji Vi + Stable: L-Index < 1 Identify the critical
+ [106–108]
(1986) solution, Eigen Values. remain constant. L = MAX++1 −
j∈α L V + j Unstable: L-Index > 1 points of the system.
The operational The potential voltage
steady-state condition, solutions estimate instability problems and
VIPI Power flow equation YsT Y ( a)
system impedance is V IPI = θ = cos−1 critical points and a the efficient control [109–112]
(1989) solution. Ys .Y ( a)
negligible. proximate fictitious approach for avoiding
solution. instability.
+ N +
VCPI It utilizes offline and + ∑m Stable: VCPI < 1 The system’s proximity
Power Flow Equation. + =1 |Vm | ++
m=k [113–116]
(2004) online measurement. VCPI kth bus = ++1 − + Unstable: VCPI > 1 to voltage collapse.
Vk +
+ +
Used to protect voltage
Minimal values, such as + + collapse depends on
SDC Maximum Power Transfer + (k+1) (k)∗ + SDC > 0: Stable
+ ΔV j I ji +
a change in voltage at SDC = +1 + + local bus phasor current [117–119]
(2004) Theorem. + (k) (k+1)∗
+ SDC < 0: Unstable
receiving, are ignored. V j ΔI ji and voltage at every
line’s relay point.
 Table 2. Cont.

VSI Concept Assumption Equation Condition for Stability Objective References

Maximum transferred when  

ISI Thevenin impedance’s Constant system ISI = Ẑr − Ẑth / Ẑr ISI > 0: Stable Calculates the stability
[120–123]
(2006) magnitude equals the load topology. ISI = 1 − | I r ΔV r |/|V r ΔI r | ISI < 0: Unstable of the system.
Energies 2023, 16, 6718

impedance’s amplitude.
Identifying the distance
The rise in sending apparent Minimal values, such as    α
VSIBUS I ΔVi The VSIBUS value lies to the collapse point
power no longer yields a rise a change in voltage at VSI i = 1 + Vi ΔI
[124,125]
(2007) i i between unity and zero with the help of local
in receiving line power. receiving, are ignored.
voltage.
Maximum transferred when
ZL M +1 ZL
ENCVI is accurate in
the magnitude of Thevenin ZS = 0.5 Stable : Ratio > 1
ZL /ZS Ratio − Mcosβ+[( Mcosβ)2 − M2 +1] ZS design and measuring
impedance is equal to the 730 ≤ φs ≤ 870  ∗ ZL [110,126,127]
(2007) dS (S2 −S1 )(Y2 +Y1 ) Unstable : Ratio < 1 and simple in real-time
amplitude of the load M = dY∗ = (S +S )(Y −Y ) ZS
2 1 2 1 implementation.
impedance.
Consider the effects of Optimal placement
Equivalent system model
ENVCI the local network and ENVCI =   Stable: ENVCI > 0 depends on ESM and
(ESM) & Equivalent local [4,128–130]
(2009) the system outside the 2(ek en + f k f n ) − e2k + f k2 Unstable: ENVCI < 0 considers only local
network model (ELNM).
local network. voltage phasors.
PSI Maximum power transfer DG depends on the most 4Rij ( PL − PG ) Stable: PSI < 1 DG and sizing for
PSI = [110,131,132]
(2012) theorem. critical bus. [|Vi |cos(θ −δ)]2 Unstable: PSI > 1 distribution networks.
0: Perfect voltage
The real value of the + + Optimal placement of
VDI Negligible phase angle + + regulation
deviation of the bus VDI T = ∑ N
j=1 1 − Vj DG and sizing for [133–135]
(2012) deviations. 1: worst case voltage
voltage. distribution network.

137
regulation.
During the maximum Power, The voltage at the Uses the data of present
SVSI the load bus voltage equals nearest generator to the ΔVi Stable: SVSI < 1 operating conditions
SVSI i = β×Vi [136–138]
(2014) the impedance voltage drop load bus equals the Unstable: SVSI > 1 combined with phasor
across the line. Thevenin load voltage. voltage measurement.
The ratio of For incremental increase, Pr dV r
At stability limit when Dynamic voltage
P-Index −2 V
power loss to the power factor is r dPr dVr /dPr = ∞, the value stability assessment and [39]
(2017) P-index = Pr dV r
1−2 V
power gained. unchanged. r dPr would be 1 load shedding purposes.

Table 3. Characteristics of Stability Indices.

Characteristic Stability Indices

Optimal placement of DG & DG sizing Line VSIs, Bus VSIs except for SDC, VSIbus , ISI, and ZL /ZS ratio
Impedance dependent VSI VSLI, L VCPI_1, VSLBI, ISI, SDC, VSIbus , ZL /ZS ratio
Independent VSI VSMI, SI, LCPI, VCPIBUS , NLSI, VCPI, NVSI, SVSI, FVSI, Lmn , LQP, Lp , VIPI
Reduce Power Losses Line VSIs, Bus VSIs
Energies 2023, 16, 6718

4. Conclusions
Voltage stability indices are essential numerical parameters for measuring and evalu-
ating stability points in the power system network. These indices’ significant roles help
detect instability points early, preventing systems from voltage collapse and power out-
ages, minimizing the actual power losses, and enhancing the system’s overall efficiency.
The evaluation of these indices indicates the power system operators’ performances of
a transmission system. Further, the stability of a system can improve through proper
design, reactive power compensation, and optimal power flow. This article discusses the
mathematical modeling of 34 line and bus stability indices. The significance of these indices
for different power system problems with suitable objectives is also discussed with the
help of suitable works from the literature, around 138. This article may be attractive to
a researcher who has started researching power system stability analysis, the placement
of reactive power compensators, and DG placement for stability enhancement. The main
intention of this review is to summarize and highlight how crucial it is to choose suitable
stability indices to preserve systems’ reliability and stability. Developing various stability
indices for effective utilization in sustaining the instabilities across various operational
scenarios has been undertaken.
The VSIs review various elements and perspectives and derive from the concepts of
maximum power transfer through a line (VCPI, PTSI, VSI_1, VSMI), solution of voltage
equation (VQIline , VSLI, L-index, SI, VCPIbus LCPI, FVSI, Lmn , Lp, LQP, NVSI, NLSI), P-V
Curve (SDC, SI, VIPI), and maximum power transfer theorem (SDC, VSIbus , ISI, ZL /ZS ,
VCPI_1, VSLBI, SVSI). Most VSIs which correspond to the voltage equation concept should
have a solution. A few VSIs neglect the sending-end bus Thevenin impedance (VCPI_1,
VSLBI), and few VSIs consider bus Thevenin impedance to be extremely sensitive for a
minute variation in parameters of two consecutive measurements (VSIbus , SDC, ZL /ZS
ratio, and SDC).
In the procedure to enhance the efficiency and minimize the operational time of DG
placement and sizing issues, the primary phase considers the simple line and bus VSIs, and
the secondary phase considers the bus and overall VSIs operating with improved efficiency.
This review article mainly focuses and presents information on the significant measurement
of line and bus stability indices for maintaining the voltage and reactive power in a power
system network.
Furthermore, this review can assist in choosing the desired stability index for a given
scenario. The main findings of this paper can be highlighted as follows:
• The power system stability analysis;
• A comprehensive review of 34 Voltage Stability Indices derived mathematically;
• Voltage Stability Indices evaluated the sizing and placement of distributed energy
sources;
• Various power system issues and their corresponding application of Voltage Stability
Indices were presented;
• The corresponding data: name, mathematical calculation, concept, assumptions, con-
dition for stability, and objective for each VSI are listed;
• This review article supports researchers, power system operators, and engineers
regarding stability indices.

Funding: This work was supported by the Basic Science Research Program through the National
Research Foundation of Korea (NRF), which is funded by the Ministry of Education under Grant
NRF-2021R111A3053429.
Data Availability Statement: Not Applicable.
Conflicts of Interest: The authors declare no conflict of interest.

138
Energies 2023, 16, 6718

References
1. Eremia, M.; Shahidehpour, M. Handbook of Electrical Power System Dynamics: Modeling, Stability, and Control; Wiley: Hoboken, NJ,
USA, 2013.
2. Kundur, P.; Paserba, J.; Ajjarapu, V.; Andersson, G.; Bose, A.; Canizares, C. Definition and classification of power system stability
IEEE/CIGRE joint task force on stability terms and definitions. IEEE Trans. Power Syst. 2004, 19, 1387–1401.
3. Modarresi, J.; Gholipour, E.; Khodabakhshian, A. A comprehensive review of the voltage stability indices. Renew. Sustain. Energy
Rev. 2016, 63, 1–12. [CrossRef]
4. Danish, M.S.S.; Senjyu, T.; Danish, S.M.S.; Sabory, N.R.; Narayanan, K.; Mandal, P. A recap of voltage stability indices in the past
three decades. Energies 2019, 12, 1544. [CrossRef]
5. Lim, Z.J.; Mustafa, M.W.; Muda, Z.B. Evaluation of the effectiveness of voltage stability indices on different loadings. In
Proceedings of the International Power Engineering and Optimization Conference, Melaka, Malaysia, 6–7 June 2012.
6. Rahman, T.K.A.; Jasmon, G.B. A new technique for voltage stability analysis in a power system and improved load flow algorithm
for the distribution network. In Proceedings of the International Conference on Energy Management and Power Delivery EMPD,
Singapore, 21–23 November 1995.
7. Moghavvemi, M.; Omar, F.M. Technique for contingency monitoring and voltage collapse prediction. IEE Proc. Gener. Transm.
Distrib. 1998, 145, 634–640. [CrossRef]
8. Mohamed, A.; Jasmon, G.B.; Yusof, S. A static voltage collapse indicator using line stability factors. J. Ind. Technol. 1998, 7, 73–85.
9. Moghavvemi, M.; Faruque, O. Real-time contingency evaluation and ranking technique. IEE Proc.—Gener. Transm. Distrib. 1998,
145, 517–524. [CrossRef]
10. Moghavvemi, M.; Faruque, M.O. Technique for assessment of voltage stability in the ill-conditioned radial distribution network.
IEEE Power Eng. Rev. 2001, 21, 58–60. [CrossRef]
11. Musirin, I.; Khawa, T.; Rahman, A. Novel fast voltage stability index (FVSI) for voltage stability analysis in a power transmission
system. In Proceedings of the Student Conference on Research and Development Proceedings, Shah Alam, Malaysia, 17 July 2002.
12. Milosevic, B.; Begovic, M. Voltage-stability protection and control using a wide-area network of phasor measurements. IEEE
Trans. Power Syst. 2003, 18, 121–127. [CrossRef]
13. He, T.; Kolluri, S.; Mandal, S.; Galvan, F.; Rastgoufard, P. Identification of weak locations in bulk transmission systems using
voltage stability margin index. In Proceedings of the IEEE Power Engineering Society General Meeting, Denver, CO, USA,
6–10 June 2004.
14. Wang, L.; Liu, Y.; Luan, Z. Power transmission paths-based voltage stability assessment. In Proceedings of the IEEE/PES
Transmission and Distribution Conference and Exhibition: Asia and Pacific, Dalian, China, 18 August 2005.
15. Sun, H.; Zhou, X.; Li, R. Accuracy analysis of static voltage stability indices based on power flow model. In Proceedings of the
2005 IEEE/PES Transmission and Distribution Conference and Exposition: Asia and Pacific, Dalian, China, 18 August 2005.
16. Nizam, M.; Mohamed, A.; Hussain, A. Performance evaluation of voltage stability indices for dynamic voltage collapse prediction.
Appl. Sci. 2006, 6, 1104–1113. [CrossRef]
17. Gong, Y.; Schulz, N.; Guzman, A. Synchro phasor-based real-time voltage stability index. In Proceedings of the PSCE Conference,
Atlanta, GA, USA, 29 October–1 November 2006.
18. Yazdanpanah-Goharrizi, A.; Asghari, R. A novel line stability index (NLSI) for voltage stability assessment of power systems. In
Proceedings of the 7th WSEAS International Conference on Power Systems, Beijing, China, 15–17 September 2007.
19. Eminoglu, U.; Hocaoglu, M.H. A voltage stability index for radial distribution networks. In Proceedings of the 42nd International
Universities Power Engineering Conference, Brighton, UK, 4–6 September 2007.
20. Guiping, D.; Yuanzhang, S.; Jian, X. A new index of voltage stability considering distribution network. In Proceedings of the
Asia-Pacific Power and Energy Engineering Conference, Wuhan, China, 27–31 March 2009.
21. Althowibi, F.A.; Mustafa, M.W. Line voltage stability calculations in power systems. In Proceedings of the IEEE International
Conference on Power and Energy, Kuala Lumpur, Malaysia, 29 November–1 December 2010.
22. Tiwari, R.; Niazi, K.R.; Gupta, V. Line collapse proximity index for prediction of voltage collapse in power systems. Int. J. Electr.
Power Energy Syst. 2012, 41, 105–111. [CrossRef]
23. Kanimozhi, R.; Selvi, K. A novel line stability index for voltage stability analysis and contingency ranking in power system using
fuzzy based load flow. J. Electr. Eng. Technol. 2013, 8, 694–703. [CrossRef]
24. Chuang, S.J.; Hong, C.M.; Chen, C.H. Improvement of integrated transmission line transfer index for power system voltage
stability. Int. J. Electr. Power Energy Syst. 2016, 78, 830–836. [CrossRef]
25. Furukakoi, M.; Adewuyi, O.B.; Danish, M.S.S.; Howlader, A.M.; Senjyu, T.; Funabashi, T. Critical Boundary Index (CBI) based on
active and reactive power deviations. Int. J. Electr. Power Energy Syst. 2018, 100, 50–57. [CrossRef]
26. Ratra, S.; Tiwari, R.; Niazi, K.R. Voltage stability assessment in power systems using line voltage stability index. Comput. Electr.
Eng. 2018, 70, 199–211. [CrossRef]
27. Ismail, B.; Wahab, N.I.A.; Othman, M.L.; Radzi, M.A.M. New Line Voltage Stability Index (BVSI) for Voltage Stability Assessment
in Power System: The Comparative Studies. IEEE Access 2022, 10, 103906–103931. [CrossRef]
28. Kessel, P.; Glavitsch, H. Estimating the voltage stability of a power system. IEEE Trans. Power Deliv. 1986, 1, 346–354. [CrossRef]

139
Energies 2023, 16, 6718

29. Tamura, Y.; Sakamoto, K.; Tayama, Y. Voltage instability proximity index (VIPI) based on multiple load flow solutions in
ill-conditioned power systems. In Proceedings of the 27th IEEE Conference on Decision and Control, IEEE, Austin, TX, USA,
7–9 December 1988.
30. Balamourougan, V.; Sidhu, T.S.; Sachdev, M.S. Technique for online prediction of voltage collapse. IEEE Gener. Transm. Distrib.
2004, 151, 453–460. [CrossRef]
31. Verbic, G.; Gubina, F. A new concept of voltage-collapse protection based on local phasors. IEEE Trans. Power Deliv. 2004, 19,
576–581. [CrossRef]
32. Smon, I.; Verbic, G.; Gubina, F. Local voltage-stability index using Tellegen’s theorem. IEEE Trans. Power Syst. 2006, 21, 1267–1275.
[CrossRef]
33. Haque, M.H. Use of local information to determine the distance to voltage collapse. In Proceedings of the International Power
Engineering Conference, Singapore, 3–6 December 2007.
34. Wiszniewski, A. New criteria of voltage stability margin for the purpose of load shedding. IEEE Trans. Power Deliv. 2007, 22,
1367–1371. [CrossRef]
35. Wang, Y.; Li, W.; Lu, J. A new node voltage stability index based on local voltage phasors. Electr. Power Syst. Res. 2009, 79, 265–271.
[CrossRef]
36. Aman, M.M.; Jasmon, G.B.; Mokhlis, H.; Bakar, A.H.A. Optimal placement and sizing of a DG based on a new power stability
index and line losses. Int. J. Electr. Power Energy Syst. 2012, 43, 1296–1304. [CrossRef]
37. Yang, C.F.; Lai, G.G.; Lee, C.H.; Su, C.T.; Chang, G.W. Optimal setting of reactive compensation devices with an improved voltage
stability index for voltage stability enhancement. Int. J. Electr. Power Energy Syst. 2012, 37, 50–57. [CrossRef]
38. Perez-Londõno, S.; Rodríguez, L.F.; Olivar, G. A Simplified Voltage Stability Index (SVSI). Int. J. Electr. Power Energy Syst. 2014, 63,
806–813. [CrossRef]
39. Kamel, M.; Karrar, A.A.; Eltom, A.H. Development and Application of a New Voltage Stability Index for On-Line Monitoring and
Shedding. IEEE Trans. Power Syst. 2018, 33, 1231–1241. [CrossRef]
40. Makasa, K.J.; Venayagamoorthy, G.K. Online voltage stability load index estimation based on PMU measurements. In Proceedings
of the 2011 IEEE Power and Energy Society General Meeting, Detroit, MI, USA, 24–28 July 2011.
41. Alias, M.R.N.M.; Verayiah, R.; Yin, L.J. Investigation of Power Flow Tracing Strategy Using an Index in a Transmission System
with Solar PV for Congestion Management. In Proceedings of the 11th International Conference on Power and Energy Systems
(ICPES), Shanghai, China, 18–20 December 2021.
42. Oukennou, A.; Sandali, A. Analysis and Comparison of Line Voltage Stability Indices. In Proceedings of the 6th International
Renewable and Sustainable Energy Conference (IRSEC), Rabat, Morocco, 5–8 December 2018.
43. Salah, M.; Wang, Y.; Chen, T. A novel collapse prediction index for voltage stability analysis and contingency ranking in power
systems. Prot. Control. Mod. Power Syst. 2023, 8, 7.
44. Shekhawat, N.; Gupta, A.K.; Sharma, A.K. Voltage Stability Assessment Using Line Stability Indices. In Proceedings of the 3rd
International Conference and Workshops on Recent Advances and Innovations in Engineering (ICRAIE), Jaipur, India, 22–25
November 2018.
45. Calma, E.R.B.; Pacis, M.C. Artificial Neural Network-based Voltage Stability Analysis of Power Transmission Networks with
Distributed Generation Using Phasor Measurement Unit Synthetic Data. In Proceedings of the IEEE 12th Control and System
Graduate Research Colloquium (ICSGRC), Shah Alam, Malaysia, 6 August 2021.
46. Singh, P.; Parida, S.K.; Chauhan, B.; Choudhary, N. Online Voltage Stability Assessment Using Artificial Neural Network
Considering Voltage Stability Indices. In Proceedings of the 21st National Power Systems Conference (NPSC), Gandhinagar,
India, 17–19 December 2020.
47. Chintakindi, R.; Mitra, A. Detecting Weak Areas in a Real-time Power System on a Static Load Model using Standard Line Voltage
Stability Indices. In Proceedings of the IEEE 2nd International Conference on Smart Technologies for Power, Energy, and Control
(STPEC), Chhattisgarh, India, 12 November 2021.
48. Verayiah, R.; Abidin, I.Z. A Study on static voltage collapse proximity indicators. In Proceedings of the IEEE 2nd International
Power and Energy Conference, Johor Bahru, Malaysia, 1–3 December 2008.
49. Isaiah, A.; Yanxia, S. New performance indices for voltage stability analysis in a power system. Energies 2017, 10, 2042.
50. Salama, M.M.; Saied, E.M.; Abou-Elsaad, M.M.; Ghariany, E.F. Estimating the voltage collapse proximity indicator using an
artificial neural network. Energy Convers. Manag. 2001, 42, 69–79. [CrossRef]
51. Chaithra, A.; Modi, S. Power System Vulnerability Assessment using Voltage Collapse Proximity Index. In Proceedings of the
2nd IEEE International Conference on Power Electronics, Intelligent Control and Energy Systems (ICPEICES), Delhi, India, 22–24
October 2018.
52. Chaithra, A.; Modi, S. Evaluation of Power System Stability using Voltage Proximity Index. In Proceedings of the International
Conference on Recent Innovations in Electrical, Electronics & Communication Engineering (ICRIEECE), Bhubaneswar, India,
27–28 July 2018.
53. Khunkitti, S.; Premrudeepreechacharn, S. Voltage Stability Improvement Using Voltage Stability Index Optimization. In
Proceedings of the 2020 International Conference on Power, Energy, and Innovations (ICPEI), Chiangmai, Thailand, 14–16
October 2020.

140
Energies 2023, 16, 6718

54. Xi, Z.; Kong, W.C. Comparison of voltage stability indexes considering dynamic load. In Proceedings of the 2011 IEEE Electrical
Power and Energy Conference, Winnipeg, MB, Canada, 3 October 2011.
55. Soyemi, A.; Misra, S.; Oluranti, J.; Ahuja, R. Evaluation of Voltage Stability Indices. In Information Systems and Management Science;
Springer: Cham, Switzerland, 2022; Volume 303.
56. Hosseinzadeh, N.; Aziz, A.; Mahmud, A.; Gargoom, A.; Rabbani, M. Voltage Stability of Power Systems with Renewable-Energy
Inverter-Based Generators: A Review. Electronics 2021, 10, 115. [CrossRef]
57. Jirjees, M.A.; Al-Nimma, D.A.; Al-Hafidh, M.S.M. Voltage Stability Enhancement Based on Voltage Stability Indices Using FACTS
Controllers. In Proceedings of the 2018 International Conference on Engineering Technology and their Applications (IICETA),
Al-Najaf, Iraq, 8–9 May 2018.
58. Pacis, M.C.; Antonio, I.J.M.; Bañaga, I.J.T. Under Voltage Load Shedding Algorithm using Fast Voltage Stability Index (FVSI) and
Line Stability Index (LSI). In Proceedings of the IEEE 13th International Conference on Humanoid, Nanotechnology, Information
Technology, Communication and Control, Environment, and Management, Manila, Philippines, 28–30 November 2021.
59. Ali, M.B.; Kazmi, S.A.A.; Khan, Z.A.; Altamimi, A.; Alghassab, M.A.; Alojaiman, B. Voltage Profile Improvement by Integrating
Renewable Resources with Utility Grid. Energies 2022, 15, 8561. [CrossRef]
60. Meena, G.; Verma, K.; Mathur, A. FVSI-Based Meta-heuristic Algorithm for Optimal Load Shedding to Improve Voltage Stability.
In Proceedings of the IEEE 10th Power India International Conference (PIICON), New Delhi, India, 25–27 November 2022.
61. Ramírez-Perdomo, S.L.; Lozano, C.A. Evaluation of indices for voltage stability monitoring using PMU measurements. Ing. E
Investig. 2014, 34, 44–49. [CrossRef]
62. Arbelaez Ramirez Marcela, L. Monitoring Voltage Stability Margin across Several Transmission Lines Using Synchro Phasor
Measurements. Ph.D. Thesis, Iowa State University, Ames, Iowa, 2016.
63. Sajan, K.S.; Kumar, V.; Tyagi, B. Genetic algorithm-based support vector machine for online voltage stability monitoring. Int. J.
Electr. Power Energy Syst. 2015, 73, 200–208. [CrossRef]
64. Naganathan, G.S.; Babulal, C.K. Optimization of support vector machine parameters for voltage stability margin assessment in
the deregulated power system. Soft Comput. 2019, 23, 10495–10507. [CrossRef]
65. Reddy, M.V.; Murthy, V.S.K. Improvement of voltage stability based on static and dynamic criteria. In Proceedings of the 16th
National Power Systems Conference, Hyderabad, India, 15–17 December 2010.
66. Yunhwan, L.; Hwachang, S. Decentralized Load Shedding Method Based on Voltage Stability Margin Index Using Synchrophasor
Measurement Technology. Electronics 2018, 7, 277.
67. Dharmapala, K.D.; Rajapakse, A.; Narendra, K.; Zhang, Y. Machine Learning Based Real-Time Monitoring of Long-Term Voltage
Stability Using Voltage Stability Indices. IEEE Access 2020, 8, 222544–222555. [CrossRef]
68. Mesgarnejad, H.; Shahrtash, S.M. Power system voltage stability assessment employing phasor measurement units. In Pro-
ceedings of the 2nd International Conference on Computer and Automation Engineering (ICCAE), Singapore, 26–28 February
2010.
69. Ankit, S.; Akash, S. Application of supervised learning for voltage stability assessment using GVSM. Electr. Electron. Technol.
Open Access J. 2018, 2, 281–287.
70. Hameedullah, Z.; Shah, D.M.S.; Tomonobu, S.; Mikaeel, A.; Abdul, N.; Mohebullah, W.; Mahdi, K.; Paras, M. A Contemporary
Novel Classification of Voltage Stability Indices. Appl. Sci. 2020, 10, 1639.
71. Ghassan, S.H.; Husham, I. Assessment of voltage stability based on power transfer stability index using com-putational
intelligence models. Int. J. Electr. Comput. Eng. 2021, 11, 2790–2797.
72. Izzri, N.; Mehdi, O.H.; Abdalla, A.N.; Jaber, A.S.; Shalash, N.A.; Noori, Y.L. Fast Prediction of Power Transfer Stability Index
Based on Radial Basis Function Neural Network. Int. J. Phys. Sci. 2011, 6, 7978–7984.
73. Nizam, M.; Mohamed, A.; Hussain, A. Dynamic Voltage Collapse Prediction on a Practical Power System Using Power Transfer
Stability Index. In Proceedings of the 5th Student Conference on Research and Development, Selangor, Malaysia, 11–12 December
2007; pp. 1–6. [CrossRef]
74. Woldu, T.A.; Ziegler, C.; Wolter, M. A New Method for Prediction of Static and Dynamic Voltage Collapse Using Node Parameters
in Large Power Networks. In Proceedings of the IEEE PES Innovative Smart Grid Technologies Europe (ISGT-Europe), The
Hague, The Netherlands, 26–28 October 2020.
75. Jamroen, C.; Dechanupaprittha, S. Synchrophasor-based early-warning voltage instability index. In Proceedings of the IEEE
Manchester PowerTech, Manchester, UK, 18–22 June 2017.
76. Rohikaa, R.M.; Lakshmi, R.; Sunitha, R.; Ashok, S. Assessment of voltage stability in microgrid. In Proceedings of the International
Conference on Electrical, Electronics, and Optimization Techniques (ICEEOT), Chennai, India, 3–5 March 2016.
77. Oukennou, A.; Sandali, A. Assessment and analysis of Voltage Stability Indices in electrical networks using PSAT Software. In
Proceedings of the Eighteenth International Middle East Power Systems Conference (MEPCON), Cairo, Egypt, 27–29 Decem-
ber 2016.
78. Mahiraj, R.S.; Vadhera, S. Voltage Stability Assessment Techniques for Modern Power Systems. In Novel Advancements in Electrical
Power Planning and Performance; IGI Global: Hershey, PA, USA, 2020; pp. 128–176.
79. Gupta, A.R. Effect of optimal allocation of multiple DG and D-STATCOM in the radial distribution system for minimizing losses
and THD. In Proceedings of the 7th International Symposium on Embedded Computing and System Design (ISED), Durgapur,
India, 18–20 December 2017.

141
Energies 2023, 16, 6718

80. Narayan, K.S.; Kumar, A. Probabilistic analysis of radial distribution network performance with varying wind speed levels. In
Proceedings of the 6th IEEE Power India International Conference (PIICON), Delhi, India, 5–7 December 2014.
81. Pattabhi, M.B.; Lakshmikantha, B.R.; Sundar, K.S. A Novel Method for Contingency Ranking Based on Voltage Stability Criteria
in Radial Distribution Systems. Technol. Econ. Smart Grids Sustain. Energy 2022, 7, 9. [CrossRef]
82. Sagwal, R.; Kumar, A. Voltage Stability Margin, based congestion management for a hybrid system, considering secure bilateral
trading: MINLP approach. In Proceedings of the IEEE 6th International Conference on Power Systems (ICPS), New Delhi, India,
4–6 March 2016.
83. Rao, A.R.N.; Vijaya, P.; Kowsalya, M. Voltage stability indices for stability assessment: A review. Int. J. Ambient Energy 2021, 42,
829–845.
84. Hauqe, M.H. A linear static voltage stability margin for radial distribution systems. In Proceedings of the IEEE Power Engineering
Society General Meeting, Montreal, QC, Canada, 18–22 June 2006.
85. Ahmad, N.; Mohamad, F.; Sulaiman, M.; Fazliana, A.; Kadir, A.; Bin, R.O. Classifications of voltage stability margin (VSM) and
load power margin (LPM) using probabilistic neural network (PNN). ARPN J. Eng. Appl. Sci. 2017, 12, 5591–5596.
86. Merlyn, M.; Suresh, D.; Venkataramu, P.S.; Nagaraj, M.S. Effect of TCSC on Line Voltage Stability Indices under Single Line
Outage Condition. Int. J. Innov. Res. Electr. Electron. Instrum. Control Eng. 2015, 3, 101–105.
87. Singh, B.; Sharma, P.A.K. Voltage Stability Calculations in Power Transmission Lines: Indications and Allocations (IEEE 30 BUS
SYSTEM). Int. J. Nov. Res. Electr. Mech. Eng. 2015, 2, 63–66.
88. Mathew, M.; Gosh, S.; Suresh, D.; Ansari, A.A. An Assessment of Voltage Stability Based on Line Voltage Stability Indices and its
Enhancement Using TCSC. IOSR J. Electr. Electron. Eng. 2015, 10, 81–88.
89. Zhang, Y.; Huang, L.; Chen, X. An Enhanced Voltage-Reactive Power Index Considering Power Injection Limitations for Power
System Stability Assessment. IEEE Trans. Power Syst. 2021, 36, 262–273.
90. Kumar Kavuturu, K.V.; Sai Tejaswi, K.N.V.; Janamala, V. Performance and security enhancement using generalized optimal
unified power flow controller under contingency conditions and renewable energy penetrations. J. Electr. Syst. Inf. Technol. 2022,
9, 18. [CrossRef]
91. Alluri, A.; Lanka, R.S.; Rayapudi, S.R. Optimal Location for FACTS Devices: Hybrid Algorithm to Improve Power Flow and
System Security. Cybern. Syst. 2022, 54, 938–960. [CrossRef]
92. Ismail, B.; Wahab, N.I.A.; Othman, M.L.; Radzi, M.A.M.; Vijyakumar, K.N.; Naain, M.N.M. A Comprehensive Review of
Optimal Location and Sizing of Reactive Power Compensation Using Hybrid-Based Approaches for Power Loss Reduction,
Voltage Stability Improvement, Voltage Profile Enhancement, and Loadability Enhancement. IEEE Access 2020, 8, 222733–222765.
[CrossRef]
93. Prabhakar, P. Voltage Stability Indices: Formulation and classification perspectives. Int. J. Innov. Technol. Explor. Eng. 2019, 8,
5777–5785. [CrossRef]
94. Salah, M.; Yifei, W.; Tiancong, C. Modern voltage stability index for prediction of voltage collapse and estimation of maximum
load-ability for weak buses and critical lines identification. Int. J. Electr. Power Energy Syst. 2023, 145, 1–19.
95. Gitanjali, S.; Kabir, C.; Priyanath, D. Enrichment of voltage stability in power system through novel generalized approximate
reasoning based intelligent control with African buffalo optimization approach. Soft Comput. 2022, 27, 7473–7496.
96. Kushwaha, P.K.; Bhattacharjee, C. Research on a selection of appropriate stability index under adverse system conditions for the
assessment of voltage stability of an IEEE 14 bus power system. Eur. J. Electr. Eng. 2020, 22, 435–446. [CrossRef]
97. Samy, A.; Venkadesan, A. Evaluation of Voltage Stability Margin in a Power System using an Artificial Neural Network. In
Proceedings of the IEEE 10th Power India International Conference (PIICON), New Delhi, India, 25–27 November 2022.
98. Ademola, A.S.; Sun, Y. Power system optimization approach to mitigate voltage instability issues: A review. Cogent Eng. 2023,
10, 2153416.
99. Bode, O.A.; Lotfy, M.E.; Olabisi, B.A.; Rashid, H.O.; Senjyu, H.T.; Narayanan, K. Security-constrained optimal utility-scale solar
PV investment planning for weak grids: Short reviews and techno-economic analysis. Appl. Energy 2019, 245, 16–30.
100. Adewuyi, O.B.; Danish, M.S.S.; Howlader, A.M.; Senjyu, T.; Lotfy, M.E. Network Structure-Based Critical Bus Identification for
Power System Considering Line Voltage Stability Margin. J. Power Energy Eng. 2018, 6, 97–111. [CrossRef]
101. Akito, N.; Miyara, R.; Alharbi, T.; Prabaharan, N. Dynamic Voltage Stability Assessment in Remote Island Power System with
Renewable Energy Resources and Virtual Synchronous Generator. Energies 2021, 14, 5851.
102. Bode, O.A.; Shigenobu, R.; Ooya, K.; Senjyu, T.; Howlader, M.A. Static voltage stability improvement with battery energy storage
considering optimal control of active and reactive power injection. Electr. Power Syst. Res. 2019, 172, 303–312.
103. Ratra, S.; Singh, D.; Bansal, R.C.; Naidoo, R.M. Stochastic Estimation and Enhancement of Voltage Stability Margin considering
Load and Wind Power Intermittencies. In Proceedings of the IEEE 6th International Conference on Computing, Communication,
and Automation (ICCCA), Arad, Romania, 17–19 December 2021.
104. Saifullah, M.K.; Kabir, M.M.; Mia, N.; Tanvir, A. Optimal Placement of FACTS Devices in the Power System Using ANN to
Increase the System Loadability. Electronic Systems and Intelligent Computing. In Lecture Notes in Electrical Engineering; Mallick,
P.K., Meher, P., Majumder, A., Das, S.K., Eds.; Springer: Singapore, 2015; Volume 686.
105. Kushwaha, P.K.; Bhattacharjee, C. Role of Energy Storage System for improving System Stability considering Customer Behavior
Uncertainties. In Proceedings of the International Conference on Contemporary Computing and Applications (IC3A), Lucknow,
India, 5–7 February 2020.

142
Energies 2023, 16, 6718

106. Adebayo, I.G.; Sun, Y. Voltage Stability Enhancement Capabilities of LTCT and STATCOM in a Power System. In Proceedings of
the IEEE PES/IAS PowerAfrica, Cape Town, South Africa, 28–29 June 2018.
107. Shuai, C.; Deyou, Y.; Weichun, G.; Chuang, L.; Guowei, C.; Lei, K. Global sensitivity analysis of voltage stability in the power
system with correlated renewable energy. Electr. Power Syst. Res. 2021, 192, 106916. [CrossRef]
108. Adebayo, I.G.; Jimoh, A.A.; Yusuff, A.A.; Sun, Y. An alternative method for the identification of critical nodes leading to voltage
instability in a power system. Afr. J. Sci. Technol. Innov. Dev. 2018, 10, 323–333. [CrossRef]
109. Lakshmi, C.; Manjula, S.S. Detection of Proximity to Voltage Collapse Indicator Using Artificial Neural Networks. Int. J. Sci. Eng.
Res. 2019, 10, 59–63.
110. Salama, H.S.; Vokony, I. Voltage stability indices–A comparison and a review. Comput. Electr. Eng. 2022, 98, 107743. [CrossRef]
111. Nanba, M.; Huang, Y.; Kai, T.; Iwamoto, S. Studies on VIPI-based control methods for improving voltage stability. Int. J. Electr.
Power Energy Syst. 1997, 20, 141–146. [CrossRef]
112. Takehara, A.; Tanaka, Y.; Iwamoto, S. Voltage stability preventive and emergency-preventive control using VIPI sensitivity. Electr.
Eng. Jpn. 2002, 143, 22–30. [CrossRef]
113. Gubina, F.; Strmcnik, B. Voltage collapse proximity index determination using voltage phasors approach. IEEE Trans. Power Syst.
1995, 10, 788–794. [CrossRef]
114. Adebayo, I.G.; Jimoh, A.A.; Yusuff, A.A. Prediction of voltage collapse through voltage collapse proximity index and inherent
structural characteristics of a power system. In Proceedings of the IEEE PES Asia-Pacific Power and Energy Engineering
Conference (APPEEC), Brisbane, QLD, Australia, 15–18 November 2015.
115. Venu, Y.; Kanthala, S. Optimal placement and Sizing of Compensator for Voltage Stability improvement Using Voltage Collapse
Proximity Indicator. In Proceedings of the 3rd International Conference on Electronics, Communication and Aerospace Technology
(ICECA), Coimbatore, India, 12–14 June 2019.
116. Adebayo, I.G.; Jimoh, A.A.; Yusuff, A.A. Analysis of Voltage Collapse in a Power System Using Voltage Stability Indices. J. Energy
Technol. Policy 2018, 8, 40–45.
117. Šmon, I.; Pantoš, M.; Gubina, F. An improved voltage-collapse protection algorithm based on local phasors. Electr. Power Syst.
Res. 2008, 78, 434–440. [CrossRef]
118. Maheshkumar, M. Online voltage stability analysis using synchrophasor technology. Int. J. Sci. Res. Dev. 2014, 2, 273–275.
119. Amroune, M.; Bouktir, T.; Musirin, I. Power system voltage instability risk mitigation via emergency demand response-based
whale optimization algorithm. Prot Control Mod Power Syst. 2019, 4, 1–14. [CrossRef]
120. Gümüş, T.E.; Emiroglu, S.; Yalcin, M.A. Optimal DG allocation and sizing in distribution systems with Thevenin-based impedance
stability index. Int. J. Electr. Power Energy Syst. 2023, 144, 108555. [CrossRef]
121. Cuffe, P.; Milano, F. Validating Two Novel Equivalent Impedance Estimator. IEEE Trans. Power Syst. 2018, 33, 1151–1152.
[CrossRef]
122. Sadeghi, S.E.; Akbari Foroud, A. A new approach for static voltage stability assessment in distribution networks. Int. Trans. Electr.
Energy Syst. 2019, 30, 1–21. [CrossRef]
123. Stefan, P.; Herwig, R. Voltage stability monitoring methods for distribution grids using the Thevenin impedance. CIRED—Open
Access Proc. J. 2017, 2017, 1535–1539.
124. Zhou, Q.; Guan, Y.; Liu, Y.; Zhao, Z. A novel online local voltage stability monitoring approach based on clustering analysis. Int.
J. Electr. Power Energy Syst. 2021, 132, 106807.
125. Pan, J.; Dong, A.; Fan, J.; Li, Y. Online Static Voltage Stability Monitoring for Power Systems Using PMU Data. Math. Probl. Eng.
2020, 2020, 1–8. [CrossRef]
126. Alzaareer, K.; Saad, M. Real-Time Voltage Stability Monitoring in Smart Distribution Grids. In Proceedings of the International
Conference on Renewable Energy and Power Engineering (REPE), Toronto, ON, Canada, 24–26 November 2018.
127. Lee, Y.; Han, S. Real-Time Voltage Stability Assessment Method for the Korean Power System Based on Estimation of Thévenin
Equivalent Impedance. Appl. Sci. 2019, 9, 1671. [CrossRef]
128. Wang, Y.; Wang, C.; Lin, F.; Li, W.; Wang, L.Y.; Zhao, J. A new transfer impedance-based system equivalent model for voltage
stability analysis. Int. J. Electr. Power Energy Syst. 2014, 62, 38–44. [CrossRef]
129. Akhil, P.; Atul, M.; Prabhakar, K.; Kothari, D.P. Assessment on various node voltage stability indices—A review. In Proceedings
of the International Conference on Smart Technologies for Smart Nation (SmartTechCon), Bengaluru, India, 17–19 August 2017.
130. Nagendra, P.; Datta, T.; Halder, S.; Paul, S. Power System Voltage Stability Assessment Using Network Equivalents-A Review. J.
Appl. Sci. 2010, 10, 2147–2153. [CrossRef]
131. Xinglong, W.; Zheng, X.; Zheren, Z. Power Stability Analysis and Evaluation Criteria of Dual-Infeed HVDC with LCC-HVDC
and VSC-HVDC. Appl. Sci. 2021, 11, 5847.
132. Xu, Z.; Xu, Z.; Xiao, L. Analysis and assessment standards of power stability of multi-send HVDC systems. J. Eng. 2018, 2019,
748–753. [CrossRef]
133. Alam, M. Determination of Power System Contingency Ranking Using Novel Indices. Recent Advances in Power Systems. In
Lecture Notes in Electrical Engineering; Gupta, O.H., Sood, V.K., Malik, O.P., Eds.; Springer: Singapore, 2022; Volume 812.
134. Mehebub, A.; Shubhrajyoti, K.; Sankar, T.S.; Sumit, B. Determination of Critical Contingency Based on L-Index and Impact As-
sessment on Power System. In Smart Energy and Advancement in Power Technologies; Springer: Berlin, Germany, 2022; pp. 583–596.

143
Energies 2023, 16, 6718

135. Khunkitti, S.; Boonluk, P.; Siritaratiwat, A. Optimal Location and Sizing of BESS for Performance Improvement of Distribution
Systems with High DG Penetration. Int. Trans. Electr. Energy Syst. 2022, 2022, 6361243. [CrossRef]
136. Rodríguez Lopez, D.K.; Perez-Londoño, S.; Rodriguez-Garcia, L. Optimal under voltage load shedding based on voltage stability
index. Ing. E Investig. 2016, 36, 43–50.
137. Shabir, M.; Nawaz, S.; Vijayvargiya, A. Voltage Stability Enhancement Using SVC in PSCAD Software. In Innovations in Electrical
and Electronic Engineering. Lecture Notes in Electrical Engineering; Favorskaya, M.N., Mekhilef, S., Pandey, R.K., Singh, N., Eds.;
Springer: Singapore, 2021; Volume 661.
138. Akbarzadeh Aghdam, P.; Khoshkhoo, H. Voltage stability assessment algorithm to predict power system loadability margin. IET
Gener. Transm. Distrib. 2020, 14, 1816–1828. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

144
 energies
Article
Real-Time Control of a Battery Energy Storage System Using a
Reconfigurable Synchrophasor-Based Control System †
Prottay M. Adhikari ‡ , Luigi Vanfretti *, Hao Chang and Koushik Kar

Rensselaer Polytechnic Institute, Electrical, Computer and Systems Engineering Department,

Troy, NY 12180, USA; prottaymondaladhikari@gmail.com (P.M.A.)
* Correspondence: vanfrl@rpi.edu
† This paper is an extended version of our paper published in 2022 International Conference on Smart Grid
Synchronized Measurements and Analytics (SGSMA), Split, Croatia, 24–26 May 2022; pp. 1–6.
‡ Current address: Eaton Corporation, Moon Twp, PA 15108, USA.

Abstract: Synchrophasor-driven smart grid applications aiming to orchestrate a diverse set of Dis-
tributed Energy Resources (DERs) require extensive infrastructure including substantial instrumen-
tation hardware, communication network extensions and controller installations for coordinated
operation. This can make the overall installation expensive. Additionally, due to the computational
complexity and data-intensive nature of the PDC functionality, most of the existing PDC implemen-
tations are on a purely software level, making them unsuitable for the real-time applications. To
address this, the current paper proposes an alternate architecture for the real-time synchrophasor-
based control of DER applications (e.g., microgrids) incorporating a centralized synchronization
hardware designed to replace aggregation Phasor Data Concentrators (PDCs) and supplementary
control algorithms into a singular reconfigurable hardware. This particular hardware is termed a
Synchrophasor Synchronization Gateway and Controller (SSGC). The robustness of the proposed ar-
chitecture is tested by using real-time (RT) Controller Hardware-In-the-Loop (CHIL) simulation-based
experiments by manipulating the communication network that connects the SSGC with multiple
Phasor Measurement Unit (PMU) streams broadcasting data through the IEEE C37.118.2 protocol
in real time. These PMU streams were generated by using a real-time microgrid model running on
a Typhoon HIL 604 simulator. To manipulate the communication interface between the proposed
Citation: Adhikari, P.M.; Vanfretti, L.; SSGC hardware and the PMU streams, a configurable Wide Area Network (WAN) emulator and
Chang, H.; Kar, K. Real-Time Control communication network impairment appliance deployed in the Candela Technologies CT910 external
of a Battery Energy Storage System
hardware was utilized. The real-time control system was expanded by incorporating a low-pass filter
Using a Reconfigurable
to eliminate the potential overswitching of a Battery Energy Storage System (BESS). The proposed ar-
Synchrophasor-Based Control
chitecture demonstrated a reliable performance under ideal to moderately tampered communication
System. Energies 2023, 16, 6909.
networks. However, under a significantly corrupted network, the performance of this architecture is
https://doi.org/10.3390/en16196909
acutely affected.
Academic Editor: Ying-Yi Hong

Received: 3 September 2023

Keywords: synchrophasor; Typhoon HIL 604; IEEE C37.118; DER; communication protocols; PMU;
Revised: 22 September 2023 PMU-based control; CHIL; HIL; BESS
Accepted: 26 September 2023
Published: 30 September 2023

1. Introduction
1.1. Motivation
Copyright: © 2023 by the authors.
In modern interconnected power systems deploying synchrophasor technology, Phasor
Licensee MDPI, Basel, Switzerland.
Data Concentrators (PDCs) are expected to receive, parse, align, store and publish the
This article is an open access article
measurement data received from Phasor Measurement Units (PMUs). As networked
distributed under the terms and
devices, PDCs should be compliant with synchronized data transmission standards, i.e.,
conditions of the Creative Commons
Attribution (CC BY) license (https://
IEEE C37.118.2 [1] While maintaining this compatibility, the existing PDC hardware is
creativecommons.org/licenses/by/
inadequate to comply with hard real-time control requirements. Currently, most state-of-
4.0/).
the-art real-time compliant PDC implementations are developed entirely at the software level.

Energies 2023, 16, 6909. https://doi.org/10.3390/en16196909 145 https://www.mdpi.com/journal/energies

Energies 2023, 16, 6909

On the other end, existing industrial PDCs [2–4] are not real-time compliant as they are
implemented in nondeterministic operating systems (e.g., Windows 10/11 and similar).
This gap in engineering development has been studied and reported by the authors of [5–7].
As illustrated in Figure 1, this makes the application of PDCs in the real-time net-
worked control of power systems a challenging problem, with the major drawback of
latency build-up if the current approach for real-time synchrophasor control is adopted.
In [8], a hardware platform for wide-area control system (WACS) applications that can
only function upon a single incoming PMU stream was proposed. To the best of the
authors’ knowledge, currently, there is no real-time compliant real-time PDC hardware
architecture that can operate on multiple PMU streams. This becomes a bottleneck in
implementing synchrophasor-driven real-time control systems for microgrid applications
that require measurements from diverse Distributed Energy Resources (DERs). To bridge
this gap, as illustrated in Figure 2, this paper proposes an alternative real-time control
architecture which replaces aggregator PDCs and secondary control functionalities into a
singular reconfigurable hardware and demonstrates a microgrid control system featuring
the proposed hardware.
䛶G
][jg]YYIg
W 䛶D

&RQQHFWLRQ
&RPPRQ
3RLQWRI
%DWWHU\(66
W
396\VWHP

'LHVHO*HQHUDWRU
8WLOLW\

/RDG
䛶E

6XEVWDWLRQOHYHO,QWHUPHGLDWH
W W /LQH
3'&
308
7UDQVIRUPHU
$JJUHJDWLRQ
6RIWZDUHEDVHG0RGHOOLQJ
3DWKVVWDUWLQJIURP308VDW39
V\VWHPDQGGLHVHOJHQHUDWRUDUH 3'& 3'&
WHVWHGEXWWKHLUFRQWUROSDWKVDUHQRW 䛶F
LPSOHPHQWHGLQWKHFXUUHQWZRUN 䛶DEFG 3URFHVVLQJGHOD\V
2QO\WKHFRQWUROSDWKIRUWKH%(66LV
LPSOHPHQWHGDQGUHSRUWHG 7WRWDO 䛶DW䛶EW䛶FW䛶GW W &RPPXQLFDWLRQGHOD\V

Figure 1. Typical infrastructure for real-time control using networked PMUs and PDCs.

䛶G
][jg]YYIg
W 䛶D
&RQQHFWLRQ
&RPPRQ
3RLQWRI

%DWWHU\(66
W
396\VWHP

'LHVHO*HQHUDWRU
8WLOLW\

/RDG

W
䛶F /LQH
66*&
*& 308
3DWKVVWDUWLQJIURP308VDW39 7UDQVIRUPHU
V\VWHPDQGGLHVHOJHQHUDWRUDUH 6RIWZDUHEDVHG0RGHOOLQJ
WHVWHGEXWWKHLUFRQWUROSDWKVDUHQRW 7WRWDO 䛶DW䛶FW䛶GW
LPSOHPHQWHGLQWKHFXUUHQWZRUN 3'&
2QO\WKHFRQWUROSDWKIRUWKH%(66LV
LPSOHPHQWHGDQGUHSRUWHG 䛶DFG 3URFHVVLQJGHOD\V
W &RPPXQLFDWLRQGHOD\V

Figure 2. Proposed approach for real-time DER control with networked PMUs and SSGC.

1.2. Related Works

Synchrophasor-based networked control architectures for power systems is an emerg-
ing field of research. Due to the latency build-up within the network, the computational limi-
tations of various edge devices, and the strict real-time compliance requirement required for
designing a control system, the deployment of a fully functional real-time synchrophasor-

146
Energies 2023, 16, 6909

based networked control architecture becomes extremely challenging. These challenges

have been summarized in [9].
Most of the existing research in this domain proposes and utilizes very specific edge
devices and software tools to implement such a control system, which is summarized
in Table 1. The authors of [10] developed custom PMU hardware (MDPMU) to deploy
their proposed fast-load-control system based on synchrophasor measurements retrieved
from those custom PMUs. Similar custom PMUs (both on hardware and software levels)
were developed by the authors of [11–13] while deploying their custom FNET/GridEye
wide-area synchrophasor network. However, these implementations focused on the moni-
toring and event-detection applications only, leaving open questions related to real-time
control. Meanwhile, industrial research laboratories (e.g., SEL) have published hardware
developments for PMU-based control applications in [14]. However, these developments
did not explore the utilization of concurrent PMU streams coming from different locations
and focused on local control application development from local PMU streams. Thus, this
development was not tested for robustness under corrupted network conditions, which
take into account incoming measurement data from remote PMUs.

Table 1. Summary of existing research related to PMU/PDC-based real-time control.

Hardware/ Suitable Available at

References RT Compliant Comments
Software for Control Production Scale
[10] Hardware Yes Yes No Cannot utilize existing PMUs in the network
[11–13] Both Yes No Yes Control applications not tested
[14] Hardware Yes No Yes Tested only for local control
[15,16] Software No Yes No No major hardware development
[17] Software Yes No No Proposed a PDC placement scheme for control

The decentralized nature of DERs coupled with the hierarchical character of the
microgrid control system (reported in [18–21]) makes it possible to break up the control
system into different networked hardware according to the transient requirements of the
control action. Based on the survey presented in [21], it can be hypothesized that the
secondary and tertiary control actions for microgrids can be put into networked devices
because these control actions are relatively slow in operation. The current paper only
focuses on implementing secondary control actions, more specifically the control of active
and reactive power flow. The research reported in [22,23] established the theoretical
background for controlling BESS in real time.
The authors of [24] reported the architecture for the Synchrophasor Synchronization
Gateway (SSG) and introduced the hardware and software associated with the implemen-
tation. On the software level, the SSG uses the C-based Khorjin library reported in [25] to
parse multiple and concurrent PMU data streams. The SSG’s Graphical User Interface (GUI)
was designed by using LabVIEW; it can be configured to accommodate additional incoming
PMU/PDC streams and to apply modifications in communication-network specifications.
Because some of the functionalities provided by the proposed hardware are similar to the
functionalities offered by a PDC, it is important to take note of the existing standardization
efforts in the domains of PDC hardware implementation. The authors of [6] summarized
the standard functional blocks and communication interfaces associated with the PDC
architecture. A similar study was reported in [26] that also explored the communication
protocol between the PMUs and PDCs.
Unlike a conventional PDC, the SSGC enables real-time control functions in addition
to the aforementioned PDC functions, and some of the control functions that can be imple-
mented by using synchrophasor-based control were reviewed as part of the literature survey.
In the domain of control system design for microgrids, the authors of [21,27] surveyed
and classified the existing control strategies into three different classes depending on the
priorities, time scales and required speed of the various control actions. The primary control
class consists of the fastest control actions including voltage and current control algorithms
for the individual DERs. The secondary control class evaluates the power flows to and from
the different existing DERs and helps the microgrid navigate between the islanded and

147
Energies 2023, 16, 6909

grid-tied modes. The secondary control class tackles slower dynamic responses (e.g., power
flow) compared to the primary control class. Finally, the tertiary control class consists of
supplementary control algorithms sitting on top of both primary and secondary classes of
control and enables the microgrid to operate in an economically optimized fashion. The re-
search reported in [28] demonstrated significant efforts of standardization across these three
classes of control systems in microgrids. The authors of [29] explored the utilization of syn-
chrophasor data to monitor microgrids and to increase the reliability of measurement data.
To this end, this research proposed an Advanced Phasor Data Concentrator (APDC) hard-
ware that is capable of operating under a tampered network and estimating missing data
points in the synchrophasor streams. However, this hardware was not time synchronized,
and the reported experiments were performed by a programmable voltage source, lacking
the testing rigor that can be brought by using Hardware-In-the-Loop experimental testing
techniques [30]. The present manuscript addresses both gaps, i.e., time synchronization,
and it exploits RT Controller Hardware-In-the-Loop real-time simulations with microgrid
models. The experiments reported in [31] illustrated a synchrophasor-based control archi-
tecture for microgrids, where the synchrophasor data are used to formulate reduced-order
dynamic models for the DERs within the microgrid, and used those models to seamlessly
navigate the microgrid between the islanded mode and the grid-tied mode. Meanwhile,
Ref. [32] demonstrated the utilization of adaptive network-management tools within the
PDC to compensate the network delays between the PDC and the individual PMUs.
In [24], the authors proposed a framework where multiple concurrent instances of the
same dynamically linked library (DLL) were used to parse multiple PMU streams in real
time. This procedure includes both deciphering the measurement data and extracting the
synchrophasor configuration information. The communication link requires bidirectional
capabilities as the PMUs send header information, measurement data and configuration
information; they also receive a command message to enable or disable the data transmis-
sion. The research presented in [24,33] utilizes multiple instances of this Khorjin thread to
facilitate the real-time reception and parsing of synchrophasor data streamed from multiple
sources. Finally, it is important to note that while the current research utilized the same
Khorjin-based SSG architecture reported in [24,33], the major contribution of the current
paper lies in the extension of the SSG architecture into the domains of secondary control ac-
tions and their validation by using real-time CHIL experiments by using a microgrid model
with DERs, where the secondary controller drives a Battery Energy Storage System (BESS).

1.3. Contributions
The current paper proposes the following contributions:
• The capabilities of the Khorjin library are extended to develop a Synchrophasor Syn-
chronization Gateway and Control (SSGC) architecture, consisting of a synchronization
layer and a control layer.
• A new PMU-based approach for control of the DERs is introduced and demonstrated
by using real-time Controller Hardware-In-the-Loop (CHIL) including a microgrid
model. This approach exploits the proposed SSGC architecture.
• The performance of the proposed SSGC architecture is analyzed under varying com-
munication network conditions (including randomized configurations) with multiple
PMU streams.
• The proposed control system is further enhanced by applying Low Pass Filters (LPFs)
in the real-time controller’s control error signal to mitigate the overswitching of the
primary controllers inside a Battery Energy Storage System (BESS).
• The real-time performance of the SSGC is demonstrated under the presence of faster
transients in the system when the control signal is passed through the enhanced
controller (with LPFs).
• The SSGC architecture was validated by deploying a second type of secondary con-
troller. The controller utilizes the phase-angle difference between two PMU locations
to control the active power output.

148
Energies 2023, 16, 6909

1.4. Paper Organization

The remainder of this paper is organized as follows. Section 2 gives an overview of
the real-time controller’s architecture and the experimental setup developed to validate
the proposed approach in this paper. Section 3 presents the results of the performance
of the SSGC under both ideal and impaired communication network conditions and the
enhancement of the Battery Energy Storage System (BESS) response by enhancing the
control function through LPFs, limiting unnecessary BESS switching. Finally, Section 4
outlines future work and Section 5 presents this paper’s major conclusions.

2. Architecture and Experimental Setup

2.1. Controller Architecture
The authors of [25] presented the Khorjin library, which receives, extracts and parses
a C37.118.2 data stream before feeding the information to subscriber applications. The
authors of [24,33] demonstrated the data unwrapping and time synchronization function-
alities on the real-time embedded system (host), utilizing multiple instances of Khorjin
extracting data from multiple incoming concurrent PMU streams in parallel.
In this paper, the ‘virtual instrument code’ running on the host side (in an RT OS) is
extended and improved to incorporate crucial secondary control functionalities targeted
to control DERs in a microgrid, whose one-line diagram and simplified view is shown in
Figure 3. The real-time control system utilizes PI controllers that calculate their set point
and process variables by unwrapping the current and voltage phasors received from the
concurrent Khorjin instances. A simplified view of the system is demonstrated in Figure 4,
which identifies both the electrical connections and communication interfaces of the setup.
The detailed experimental setup is illustrated in Figure 5.
A Khorjin thread requires bidirectional data transmission capabilities for the com-
munication link. After securing the TCP/IP connection, the SSG sends a ‘turn off
transmission’ command and a ‘send CFG-2’ command. The PMU responds by sending
the CFG-2 frame. At this point, the SSG has received all the required information from this
frame to initiate the data transmission. Hence, the SSG now requests synchrophasor mea-
surement data, and the PMU begins data transmission. This exact protocol is implemented
in all the concurrent Khorjin instances. This process is graphically illustrated in Figure 6.

9+]
SKDVH
၁

၁ ၁ ၁ ၁
99

၁ ၁ ၁ ၁
99

99

99

N: N:

8SWRN9$ 8SWR09$ 8SWRN9$

Figure 3. Simplified real-time microgrid model implemented by using Typhoon HIL 604.

149
Energies 2023, 16, 6909

8VHULQSXW 8VHULQSXWFRQILJ 8VHULQSXW

(OHFWULFDO&RQQHFWLRQ 7&3QHWZRUN

570LFURJULG $QDORJ
$QDORJ ),)2V
PRGHO RXW 308V
308V

1, QDORJ
1,$QDORJ 66*&
$QDORJ
$QDORJ
R WSXW
RXWSXW
LQ
P GXOH
PRGXOH

Figure 4. Simplified block diagram of the experimental setup.

&RQQHFWLRQ
&RPPRQ
3RLQWRI
%DWWHU\(66

396\VWHP

'LHVHO*HQHUDWRU
8WLOLW\
0LFURJULG0RGHO
7\SKRRQ%UHDNRXW

308/RDG /RDG
%RDUG,2

7\SKRRQ+,/
308%DWWHU\
/LQH
308
7UDQVIRUPHU
&RQWUROOHU 6RIWZDUHEDVHG0RGHOOLQJ
66*& +DUGZDUHFRQQHFWLRQV

Figure 5. Detailed architecture of the experimental setup demonstrating the software and hard-
ware components.

WUDQVPLVVLRQ
XUQRIIGDWD
PHVVDJH7
&RPPDQG
IUDPH
HQG&)*
PHVVDJH6
&RPPDQG
&)*IU
308

DPH
66*

WUDQVPLW
XUQRQGDWD
PHVVDJHW
&RPPDQG
'DWDWUDQ
VPLW

Figure 6. PMU to SSG/SSGC data transmission protocol by utilizing Khorjin instance.

As can be observed in Figures 5 and 7, the main goal of the real-time controller is to
maintain the power at the Point of Common Connection (PCC) constant from the point
of view of the utility, meaning that the microgrid self-balances by taking advantage of
measurements at the PCC and other locations (e.g., at the BEES terminals). Because the goal
of the controller is to redispatch the battery to meet load/generation changes within the
microgrid, it is important to mitigate the BESS response in specific time scales. To achieve
this goal, the real-time control system was enhanced by incorporating LPFs, as shown in
Figure 8, which would prevent the controller from responding to possible high-frequency
transients on the load side (e.g., switching transients).

150
Energies 2023, 16, 6909

Figure 7. SSGC’s utilization within hierarchical control system for microgrid.

Figure 8. Microgrid controller implementation utilized in the experimental setup.

The current work demonstrates a scenario where the proposed SSGC architecture is
configured to regulate the power output from a Battery Energy Storage System (BESS)-
based DER. As shown in Figure 8, the PI controller coded inside the SSGC hardware
(situated at the remote location) determines the set point to be utilized at the local (primary)
controller inside the BESS. The individual parameters for the PI controller were based on
the specifications provided in the model illustrated in [34]. The PI controller was followed
by a saturation block (with the upper and lower limits set at 1 and −1, respectively) to
ensure stability. The value of K p and Ki were taken from the Typhoon specifications directly
(for this particular application, K p = 0.5 and Ki = 3.0). This control algorithm computes
P and Q in real time from the voltage and current phasors it receives from the multiple
incoming PMU streams. The controller uses P and Q as the control variable. It is also
possible to utilize the voltage magnitude and phase-angle differences (Δδ) as the control
input variables [35–38]. The controller output is fed back to the real-time simulator’s
analog input channels and used as a control input for the primary controller of the Li-Ion
BESS, as shown in Figure 7. The proposed control architecture can take advantage of the
hierarchical structure of the control system for microgrids, secondary control in particular,
as summarized in [21].
It must be noted that, as shown in Figures 3 and 5, the microgrid RT simulation model
comprises a diesel generator, a PV system and an external grid alongside the BESS. All these

151
Energies 2023, 16, 6909

energy resources work in synchronism to supply a configurable load. These models were
based on the existing components available in the Typhoon HIL’s model library [34,39].

2.2. Experimental Setup

This paper reports a reconfigurable SSGC hardware that can suitably incorporate
secondary and tertiary control functionalities. As proof of concept, the authors aim to
demonstrate its functionalities through a real-time secondary controller to regulate the
power output of the BESS implemented on the SSGC in this work. This is illustrated in the
block diagram of Figure 8.

2.2.1. Scope of the Experiments Supported by the Experimental Setup

As shown in Figure 8, the proposed secondary control action utilizes a set point, which
is computed in real time from the two parameters PLm , which represents the measurement
from the load side, and Pbat , which represents the BESS output power. The experiments in
this work will assume that the PPV (active power output from the PV system), PUti (active
power dispatch from the utility) and PDSG (active power output from the diesel generator)
are constant. Under this condition, the parameter Pre f (reference power) will be a product of
the changes in the total load consumption Pload only. For this part of the experiment, the two
PMUs were placed near the load and near the BESS as the current measurements at these
locations were expected to experience changes along with changes in the load and battery
dispatch. As an extension, a further experiment was performed by taking the phase-angle
difference (δPCC − δLoad ) as the control variable to regulate the active power output from
the BESS instead of using the active powers directly. For this part of the experiment, the
phase angles were measured at the point of common coupling (PCC) and the load. Thus,
the PMUs were placed accordingly. The reprogrammable and reconfigurable nature of
the proposed SSGC hardware can be exploited to adapt the hardware to deploy different
control schemes and control of different types of DERs. However, those experiments are
considered to be beyond the scope of the current paper. For demonstration, only real-time
synchrophasor-based P-Q control of the BESS is presented in this paper. Hence, this paper
primarily focuses on the implementation of the SSGC architecture and its validation in the
context of simple control-system development targeted for DERs. We also demonstrate
how the control performance can be improved by implementing a Low Pass Filter (LPF)
(LPF (s) = 1+1τS , where τ = ω1c ) to make the BESS operate safely under a condition where
switching transients are present in the load.

2.2.2. Real-Time Communications

It has been discussed that the SSGC requires real-time synchrophasor data (i.e., follow-
ing the IEEE C37.118.2 std.) to be streamed at its input. Modern RT simulators have the
capability to stream synchrophasor data directly from the RT model being simulated, with-
out requiring external PMU hardware. However, the present work was completed by using
Typhoon HIL control center’s 2020.4 release, where the IEEE C37.118 library was primitive and
unstable (as discovered through experimentation) when connected to external hardware
such as a PDC or SSGC (note that the 2022.4 and 2023.1 releases of the Typhoon HIL control
center toolkit have stable and dedicated library components for streaming C37.118 data, and
the authors of this paper chose not to adjust the hardware configuration and keep the PMUs
as external hardware to facilitate impairment). This was chosen to facilitate tampering
between the communications network between the PMUs and the SSGC hardware. With
the introduction of external PMU hardware based on [40,41], it is simple to pick and choose
individual PMUs and corrupt the communication network between any single PMU and
the SSGC hardware.

2.3. Hardware Integration

The aforementioned external PMUs (the hardware used is that reported in [42] and
extended in [43]) were connected to the low-voltage analog outputs of the real-time simu-

152
Energies 2023, 16, 6909

lator. It is crucial to note that these PMU designs feature both current and voltage inputs
to function as PMUs whereas Typhoon HIL 604 hardware only provides analog voltage
outputs. Thus, an external burden circuit was introduced in the experimental setup to
perform the voltage-to-current conversion, as demonstrated in Figure 9c. The connection
between Typhoon HIL 604 (the hosting microgrid model) and the PMUs (streaming syn-
chrophasor data to SSGC) is shown in Figure 9a. Figure 9b demonstrates how the PMUs
receive GPS signals to obtain precision timing, and Figure 9d shows the SSGC deployed in
a remote location, communicating to the rest of the experimental setup through a standard
Ethernet-based communication network.

Figure 9. Hardware arrangement for RT-HIL-testing the SSGC hardware: (a) connection between the
microgrid and the PMUs, (b) PMUs receiving timing information, (c) conversion of RT low-voltage
signals into current signals and (d) SSGC connected remotely to the microgrid RT model.

SSGC Configuration and Integration in the Experimental Setup

Figure 10 shows how the LabVIEW GUI interacts with the physical inputs and out-
puts of the Typhoon HIL 604 and how the analog signal levels are scaled to obtain the
measurements of actual voltages and currents. This technique enables the SSGC operation
without any high external amplifiers that are typically required for HIL experiments [30],
such as [44], making the experimental environment safer.

153
Energies 2023, 16, 6909

308GDWDSURFHVVLQJ

F5,2
308GDWDSURFHVVLQJ
30

66*&FDOFXODWLQJPHDVXUHPHQWYDOXHV

6HFRQGDU\&RQWURO6\VWHP
66*&UHWULHYLQJUDZGDWD
66*&UHWULHYLQJUDZ
7\SKRRQ+,/EUHDNRXWERDUG

7\SKRRQ+,/
$QDORJ,3
$QDORJ23

ZGD
DWD
Figure 10. Details of the control system implemented inside the SSGC.

The communication network between the SSGC hardware and the PMUs was inten-
tionally impaired during experimentation in order to test the robustness of the proposed
control architecture. To perform these network impairments, the external Wide-Area Network
emulator and impairment appliance hardware CandelaTech CT910 [45] was connected inside
the network, which would tamper the communication link between PMUs and the SSGC.
Using CT910, it is possible to introduce custom delays, data drops, jitters and bandwidth
limitations within the network. The values of these parameters can be reprogrammed every
second, and this procedure can be automated.

3. Results
3.1. SSGC Performance under Ideal Network Conditions
In this experiment, the total load is increased in a step by turning on the configurable
load, as shown in Figure 3. Initially, a fixed load of 825 kW was supplied by the PV unit
(125 kW), diesel generator (500 kW) and the utility (100 kW). This makes the initial dispatch
for the BESS inverter to be fixed at 100 kW. With the system running in this “steady state”,
a step increase of 300 kW in load was triggered externally.
Under this scenario, the control system must respond to this change in load. Hence,
the BESS inverter and its dispatch increases from 100 kW to 400 kW. It is important to note
that the portion of the controller (in Figure 8) within the dashed red box is implemented
within the SSGC hardware. This portion is capable of utilizing synchrophasor measurement
data obtained from the PMUs placed at the load and at the BESS. The PMU data are utilized
to compute the active and reactive powers, which are then used to calculate a new set point
(by using the PI controller block GPI ) to operate the BESS. This set point is then utilized
by the internal control algorithm (implemented locally inside the BESS) to control the
individual current and voltage output of the inverter inside the BESS. This portion of the
control system was implemented locally within the BESS model of the real-time simulator
because it generates high-frequency switching sequences for the individual semiconductor
switches in the inverters that are modeled within the simulator.
Figure 11a shows the 300 kW manual load injection in the system. The SSGC incorpo-
rates a PI controller onboard. The output of this PI controller modulates the set point for
the local BESS controller to operate, as shown in Figure 11b. This measurement is taken
from the SSGC side. Meanwhile, in Figure 11c, the BESS power output response is shown.
It can be observed from this figure that the power output of the battery increases from

154
Energies 2023, 16, 6909

100 kW to 400 kW to cover for the step increase in the load. Figure 11a–c demonstrate the
SSGC’s performance under ideal conditions of the communication network while there are
no external communication disturbances.
Interruptible Load (kW)

300 4 400

Set-point (scaled)

P out from Battery

200 3
300
2
100 200
1

0 100
0
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
Time (s) Time (s) Time (s)

(a) (b) (c)

Figure 11. Control system performance for the BESS under ideal network conditions: (a) 300 kW load
injection by switching the interruptible load, (b) response at the output of the PI controller inside
SSGC and (c) active power output from the Battery Energy Storage System (BESS).

3.2. SSGC Performance under Nonideal Communication Network Conditions

Even though the SSGC hardware can be configured to control different DERs and
employ sophisticated control algorithms, such demonstrations are considered to be be-
yond the scope of the current paper. For experimentation purposes, only the real-time
synchrophasor-based control of the BESS is presented in this paper. Hence, the primary
focus of this paper is the implementation of the SSGC architecture and its validation in the
context of the P-Q control of the BESS-based DER.
Because this controller architecture relies on real-time synchrophasor communications,
it is important to assess the performance of the proposed architecture under nonideal com-
munication network conditions. Therefore, to test the robustness of the SSGC architecture,
the communication network between the SSGC hardware and the PMUs was tampered
with, as shown in Figure 12. As mentioned earlier, to perform experiments impairing the
communications, an additional external hardware CandelaTech CT910 [45] was connected
between the SSGC and the rest of the communication network system. This hardware
enables the user to introduce custom delays and data drops within the network through a
GUI or command line and is used for the experiments in the sequel.

308
308.XQGHUWKHWDPSHUHG1HWZRUN

&RPPXQLFDWLRQV &7
308
1HWZRUN (7+
 (
(7+
 308.

 0*7
0*
*7

/DQIRUJH
3081
)RUWKHH[SHULPHQWVSUHVHQWHG. 

6\QFKURSKDVRU6\QFKURQL]DWLRQ
6\QFKURSKDVRU6\QFKURQL]DWLRQ
*DWHZD\ &RQWUROOHU *361HWZRUN

Figure 12. Block diagram of a system comprising SSGC and N PMUs; the network between the
SSGC and the kth PMU is impaired, while all the hardware receive timing information from the same
time base.

3.2.1. Network Tampering by Introducing Fixed Delay

In this particular experiment, artificial delays were introduced in between PMU1
and the SSG prototype. To this end, a network-traffic generator within the CT910 was
reconfigured to introduce artificial delays and thereby tamper with the network between
the two ports where the PMU1 and SSGC prototype were connected.

155
Energies 2023, 16, 6909

Figure 12 shows a representative connection where the CT910 device is utilized to

tamper with the communication link between PMU1 and the SSGC. The delays are intro-
duced by CT910 (CT910 uses the LanForge 5.4.6 software library, which provides crucial
command-line utilities and a stable GUI to interact with the communication link in real
time). The delay between the time when a synchrophasor packet is generated and when that
particular packet is received at the SSGC end can be computed by sharing the respective
time stamps. This procedure is explained in detail by the authors of [24].
As a simple representative experiment, the network was impaired by injecting an
additional 100 ms delay, and the network delay was computed in real time by time-stamp
sharing. This network delay is plotted in Figure 13, which shows a window of 10 s, during
which the communication link was tampered with an additional 100 ms delay.

180

160

140

120
Delay (ms)

100

80

60

40

20 100 ms Delay injection

-20
0 5 10 15 20
Time (s)

Figure 13. Measured delay at the SSGC due to a 100 ms delay injection from the CT910.

3.2.2. Network Tampering by Introducing Varying Delay

In this experiment, the communication link between the PMU and the SSGC is sub-
jected to a network tampering that would randomize the injected delay every second. The
hardware restricts the frequency of the network reconfiguration and dictates that, at most,
the network can be reconfigured once every second. To achieve this, a custom script was
executed on the CT910 hardware that would reconfigure its network interface with a ran-
dom amount of delay every second. A pregenerated data file contained the amount of the
injected delay in milliseconds and was used by the script to reconfigure the hardware with
varying amounts of delay every second. The overall execution is shown in the flowchart of
Figure 14a.
By following the same methodology to compute the network delay as described in
Section 3.2.1, the network delay of this experiment is computed in real time. These results
are shown in Figure 14b. It can be seen that the network reconfigures itself with a new value
of the injected delay every second. The data file containing the raw values of the network
delays was generated from a random variable following a uniform distribution between 0
and 200 ms.
Additionally, for the second part of this experiment, random delays were also generated
to compute the network delay in real time by using a random variable following a normal
distribution with a fixed mean and standard deviation (all the negative values generated
were discarded because network delay cannot be negative). The experiment is carried out
for 10 min (with the network reconfiguring the delay every 1 s) and repeated 10 times. If
the network sustains itself (i.e., the controller keeps operating after 10 min of real-time
simulation), the experiment run is termed as successful.

156
Energies 2023, 16, 6909

+DUGZDUH
7R&7 7R&7
+DUGZDUH +DUGZDUH

5HFDOLEUDWH .HHSUXQQLQJDW
QHWZRUN FXUUHQWFDOLEUDWLRQ

1R
'DWDILOH
FRQWDLQLQJ
,QLWLDWH
FRQVWUDLQHG VHF"
7LPHU

5XQVLQVRIWZDUH
GHOD\LQIR

<HV
,QGH[

(a)
Measured network delay (ms)

250
Injected delay randomized
5 times in a 5 second window
200

150

100

50

0 5 10 15 20 25
Simulation time(s)

(b)

Figure 14. Randomized delay injection: (a) algorithm for random delay injection, (b) real-time delay
measurement after randomized delay injection.

This batch of trials (10 runs, each for a duration of 10 min) was performed with
different values of the mean and standard deviation of the injected delay, and whether
or not the 10 individual runs were successful is summarized in Table 2. The mean delay
was varied from 50 ms to 750 ms, and the standard deviation was varied from 5 ms to
200 ms. It can be seen that the performance progressively deteriorates at higher values of
the mean delay. It can also be seen that when the mean delay is 500 ms, a small increase in
the standard deviation leads to a significant number of unsuccessful trials. It is interesting
to note that the unsuccessful runs failed because the First-In First-Out (FIFO) instances,
which transmit data from the PMUs to the SSGC, gradually filled up as the PMUs keep
streaming data into a slow network. This shows that in real-time applications under
nonideal communication interfaces, the adequate sizing of elastic storage elements in the
embedded systems is crucial for the control architecture to function.

Table 2. SSGC performance under varying random network delays.

σ (ms)
5 10 25 50 100 200
μ (ms)
50 10/10 10/10 10/10 10/10 10/10 10/10
100 10/10 10/10 10/10 10/10 10/10 10/10
200 10/10 10/10 10/10 10/10 10/10 9/10
500 10/10 10/10 5/10 5/10 4/10 3/10
750 0/10 0/10 0/10 0/10 0/10 1/10
Note: the color scale is used to denote from green (100% successful) to red (unsuccessful) experiments.

157
Energies 2023, 16, 6909

3.2.3. Effect of Data Drop in the Communication Network

In this experiment, the quality of a controller’s output signal is analyzed for varying
network delays and varying data-drop rates injected simultaneously in the communication
network. The controller regulates the power output of the inverter inside the BESS, as
demonstrated in Section 3.1. However, its performance is expected to deteriorate under
stressed communication network conditions (i.e., higher delay and higher data-drop rates).
The results of these tests are summarized in Figure 15. It can be observed that the analog
output of the remote controller loses the control signal resolution under higher network
delays and higher data-drop rates. However, as discussed earlier in Section 3.1, the control
objective of the SSGC does not involve any management of fast system dynamics. Thus,
in the short term (up to 30 s), the SSGC-driven control architecture can sustain itself even
while operating within a tampered network under drastic delay and data-drop conditions.

(a) Delay = 0ms, Data drop = 0% (b) Delay = 0ms, Data drop = 2% (c) Delay = 0ms, Data drop = 5%
5 5 5

4.5 4.5 4.5

4 4 4

3.5 3.5 3.5

Controller Output

Controller Output

Controller Output
3 3 3

2.5 2.5 2.5

2 2 2

1.5 1.5 1.5

1 1 1

0.5 0.5 0.5

0 0 0
-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
Simulation Time (s) Simulation Time (s) Simulation Time (s)

(d) Delay = 100ms, Data drop = 0% (e) Delay = 100 ms, Data drop = 2% (f) Delay = 100ms, Data drop = 5%
5 5 5

4.5 4.5 4.5

4 4 4

3.5 3.5 3.5

Controller Output

Controller Output

Controller Output

3 3 3

2.5 2.5 2.5

2 2 2

1.5 1.5 1.5

1 1 1

0.5 0.5 0.5

0 0 0
-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
Simulation Time (s) Simulation Time (s) Simulation Time (s)

(g) Delay = 500ms, Data drop = 0% (h) Delay = 500ms, Data drop = 2% (i) Delay = 500ms, Data drop = 5%
5 5 5

4.5 4.5 4.5

4 4 4

3.5 3.5 3.5

Controller Output

Controller Output

Controller Output

3 3 3

2.5 2.5 2.5

2 2 2

1.5 1.5 1.5

1 1 1

0.5 0.5 0.5

0 0 0
-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
Simulation Time (s) Simulation Time (s) Simulation Time (s)

Figure 15. Control signal received in the Typhoon HIL microgrid model from the SSGC under varying
network delay and data-drop rates.

158
Energies 2023, 16, 6909

An additional set of trials were conducted for this experiment. In these trials, the SSGC
was set to operate freely under tampered network conditions, and experiments were run to
test whether or not it can sustain itself for longer periods of time. The network was tampered
with by introducing network delays and data drop. Under these conditions, the network was
kept running for 10 min. After 10 min, it was determined if the SSGC was still receiving all
the PMU streams successfully and if the real-time simulator was still receiving the controller’s
output. Observations were taken 10 times for each communication network condition. The
summary of these results is shown in Table 3. As can be observed, the network delay and data
drop can both adversely effect the robustness of the SSGC. In fact, in a situation where the
SSGC is subjected to both a high network delay coupled with a high data-drop rate, the SSGC
is almost certain to be unable to sustain itself for a long period of time. However, for lower
delays and lower data-drop rates, the SSGC is proven to be reliable. For instance, consider the
case for a 0.5% data drop and 50 ms delay: 10 out of 10 runs were sustained, while for a 5%
data drop and 200 ms delay, only 2 out 10 runs were sustained.

Table 3. SSGC performance under varying communication network conditions.

Delay
0% 0.5% 1% 2% 5% 10%
Drop Rate (%)
0 ms 10/10 10/10 10/10 10/10 10/10 3/10
50 ms 10/10 10/10 10/10 10/10 9/10 0/10
100 ms 10/10 10/10 10/10 9/10 7/10 0/10
200 ms 10/10 10/10 10/10 6/10 2/10 0/10
500 ms 10/10 10/10 5/10 0/10 0/10 0/10
750 ms 1/10 0/10 0/10 0/10 0/10 0/10
Note: the color scale is used to denote from green (100% successful) to red (unsuccessful) experiments.

3.2.4. Effect of Jitter on the Communication Network

Jitter is defined as the variation in the time delay between when a signal is transmitted
and when it is received over a network connection. The CT910 hardware has the provision
of incorporating two parameters: jitter amount and jitter frequency in the network config-
uration. The parameter jitter amount defines the maximum amount of delay mismatch
introduced in the network, and the jitter frequency determines what percent of data packets
would experience the jitter. For this experiment, a jitter frequency of 10% and a maximum
jitter delay of 500 ms was introduced in the network. This test was repeated with and
without the jitter present. Figure 16 shows the effect of jitter on the control signal being
transmitted by SSGC. The jitter delay was maxed at 500 ms, which meant that only a few
packets would be experiencing a jitter delay around that range. Experimentally, it was
observed that the effect of jitter is visible intermittently for some brief windows (denoted
by red boxes in Figure 16), during which the control signal slows down its update rate
significantly (thus dropping the high-frequency components). However, these windows
are short and are only observed while the applied jitter delay is close to its maximum
permissible limit of 500 ms.
5 5

4 4
Control Output

Control Output

3 3
Effect of jitter
Effect of jitter
2 2

1 1

0 0
0 5 10 15 20 25 0 5 10 15 20 25
Simulation time (s) Simulation Time (s)

(a) (b)

Figure 16. Control signal received in the Typhoon HIL microgrid model from the SSGC with (a) and
without (b) jitter present in the network.

159
Energies 2023, 16, 6909

3.3. Control Performance Enhancement through LPF Functions

It needs to be noted that only one scenario for controlling the DERs within a microgrid
is demonstrated in this paper. In that scenario, the battery is covering for a step increase in
the load. While this experiment is an important ‘proof of concept’, additional experimenta-
tion is crucial before implementing the proposed architecture to control real-world DERs.
In real-world applications, the behavior of a BESS-based DER in a power system
is more complex than what simulation models can capture, and more importantly, the
operation of the controllers are restricted by the physics of the battery. In fact, the batteries
utilized in energy storage systems are limited by their response times, which are different
during charging and discharging operations in addition to their power availability. The
authors of [46] reported that the average response time for a Battery Energy Storage System
during charging is about 2.2 s and during discharging is about 0.6 s. Keeping these numbers
in mind, the battery cannot be subjected to faster changes than those allowed by the charge
and discharge rates. Hence, when aiming to control power flows to balance load/generation
changes, any frequency variations, if kept unfiltered, would rapidly increase the switching
and would generate excessive heat, thus compromising the health of the batteries.
To address this issue, a practical approach is adopted to restrict the response band-
width of the proposed controller. To this end, a low-pass filter was introduced as an SSG
function, with the goal to eliminate the high-frequency variations in the control signal. In
addition, it may also slow down the response of the overall control response. Hence, the
LPF needs to be adequately tuned. To illustrate, experiments are conducted by varying
the cut-off frequency of this LPF from 4 Hz to 0.25 Hz. The control responses of each
of these cases are demonstrated in Figure 17. It can be seen from the results that at a
lower cut-off frequency (≤0.5 Hz), the control system is free of high-frequency components,
and the control signal can be safely employed in the BESS without exposing the BESS to
overswitching. These filters can reliably bring the response of the SSGC-based control
system within the permissible limit for battery operations, as reported in [46].

4
Control output

No filter
2
LPF, fc =4Hz
LPF, fc =1Hz
1
LPF, fc =0.25Hz
0
5 10 15 20 25
Simulation time (s)

Figure 17. Application of LPFs to reduce overswitching of the BESS unit being controlled by the
SSGC hardware.

3.4. Control System Redesign to Utilize Phase-Angle Measurements

As a ‘proof of concept’ that the SSGC can support different control functions, in
this section, the control system is redesigned to utilize the phase-angle difference as the
controller input, and the power flow is regulated the same way as before. The usage
of the phase angle to control the power flow is a tried and tested technique which has
been proposed, theorized and archived in [47–49]. Synchrophasor technology is capable
of reporting accurate phase-angle measurements in real time [35–38]. This makes this
technology suitable for phase-angle-based power-flow control. In this paper, the phase-
angle difference between the load and the point of common coupling is used as the control
variable to direct the power flow from the BESS. It is to be noted that, for distribution
systems, the angle difference between the buses can be very small. Hence, the controller
needs to be tuned with precision. Thus, even though the control system architecture is

160
Energies 2023, 16, 6909

independent of the control variable (the active power or phase angle), the internal PI
controller requires significant retuning.
Experiments were performed with the BESS power output being controlled by the
phase-angle difference δPCC − δload . The results of this experiment are summarized in
Figure 18a–c. It can be observed from Figure 18c that the active power output increases
from 100 kW to 400 kW. This response is consistent with the results demonstrated
previously in Figure 11c.

490
300 2

Pout from the battery (kW)

Interruptiple Load (kW)

- load (degrees)
390
1.6
200
290
1.2
100 190
PCC

0.8

0 90
0.4

2 4 6 8 2 4 6 8 2 4 6 8
Time (s) Time (s) Time (s)
(a) (b) (c)

Figure 18. Control system performance for the BESS under ideal network conditions with phase
angle utilized as the control signal: (a) 300 kW load injection by switching load, (b) response of the
control variable δPCC − δload to the load change and (c) active power output from the Battery Energy
Storage System (BESS) equipped with phase-angle-based active power control.

4. Future Works
The current paper introduces a new architecture for control system implementation
for DERs in microgrids by utilizing synchrophasor measurements. As a ‘proof of concept’,
one specific case study was tested where the BESS output is increased to match the step
increase in the load. Different cases featuring the control of a full range of DERs and a full
range of behaviors on the load side, microgrid conditions, etc., must be performed in order
to demonstrate the resilience and robustness of the proposed architecture. This work is
outside the scope of this paper and is left as future work.
For BESSs, the state of charge (SOC) of the battery is a crucial parameter. In all realistic
BESS control systems, based on the SOC, the battery is often put into either charging or
discharging mode. The current paper assumes the SOC to be high enough so that the
battery can reliably operate in discharging mode; i.e., it can feed active loads. For real
systems, this will not be the case. To address this, an additional control loop must be
designed to utilize the SOC of the battery and ensure the safe and reliable operation of
the BESS. This requires coordination and information exchange with Battery Management
Systems (BMSs) and will be the subject of future research.

5. Conclusions
The proposed Synchrophasor Synchronization Gateway and Control system can re-
duce the complexity and latency of the synchrophasor system and communication network
when compared to traditional WAMS/WAMPAC systems by exploiting and expanding
upon the existing Khorjin library. This architecture was utilized successfully to implement a
networked control system for a real-time microgrid model.
A scenario for controlling the DERs within the microgrid was used for experimentation
in this paper. In that scenario, the BESS is capable of using synchrophasor data from
different locations to respond to a step increase in the load. It was also demonstrated
that this architecture can be successful irrespective of the actual control algorithm being
employed within it.

161
Energies 2023, 16, 6909

Experimentally, the SSGC architecture was tested under different network conditions
to establish its validity and robustness. More specifically, the network was subjected to
delay, jitter and data drops through external hardware. The performance of the controller
was evaluated under those different network conditions. It was observed that for lower
values of the data-drop rate (≤2%) and lower values of the network delay (≤100 ms), the
SSGC-based control system is reliable. It was also observed that the introduction of jitter
makes the controller lose a significant amount of high-frequency components. However,
up to a maximum jitter occurrence of 10% and maximum jitter delay of 500 ms, the SSGC
sustains itself successfully even for longer runs.
Finally, an LPF was designed to prevent the BESS from overswitching. This filter
would ensure that the battery operations are within the physical limits of the battery and
thus increase the lifetime of the battery units within the BESS.
The main limitation of the implementation presented in this work is the cost associ-
ated with the National Instruments’ cRIO hardware. While such hardware was ideal for
prototyping and testing, it is cost prohibitive for product development. For the real-world
deployment of the proposed architecture, it would be crucial to utilize a low-cost hardware
platform to make it economically viable. Consequently, migrating the Khorjin dynamically
linked library (DLL) to a different platform is the first step towards the authors’ future work.

Author Contributions: Conceptualization, P.M.A., L.V. and K.K.; Methodology, P.M.A., L.V. and
K.K.; Software, P.M.A. and H.C.; Validation, P.M.A.; Investigation, P.M.A., L.V., H.C. and K.K.;
Resources, L.V.; Writing—original draft, P.M.A.; Writing—review & editing, L.V.; Supervision, L.V.
and K.K.; Project administration, L.V.; Funding acquisition, L.V. All authors have read and agreed to
the published version of the manuscript.
Funding: This research was funded in part by the New York State Energy Research and Development
Authority (NYSERDA) under agreement number 137948.
Data Availability Statement: The The data presented in this study are available on request from the
corresponding author.
Conflicts of Interest: The funders had no role in the design of the study; in the collection, analyses,
or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

Abbreviations

BESS Battery Energy Storage System

DER Distributed Energy Resources
HIL Hardware-In-the-Loop
CHIL Controller Hardware-In-the-Loop
LPF Low Pass Filter
PCC Point of common coupling
PMU Phasor Measurement Unit
PDC Phasor Data Concentrator
PV Photovoltaic
SSG Synchrophasor Synchronization Gateway
SSGC Synchrophasor Synchronization Gateway and Controller
WACS Wide-area control system
WAMPAC Wide-Area Monitoring Protection and Control System

References
1. IEEE Std C37.118.2-2011; IEEE Standard for Synchrophasor Data Transfer for Power Systems. Revision of IEEE Std C37.118-2005.
IEEE: New York, NY, USA, 2011; pp. 1–53. [CrossRef]
2. Available online: https://www.gegridsolutions.com/multilin/catalog/p30.htm (accessed on 25 September 2023).
3. Available online: https://selinc.com/products/3573/ (accessed on 25 September 2023).
4. Available online: https://nrec.com/en/index.php/product/productInfo/571.html (accessed on 25 September 2023).
5. Castello, P.; Muscas, C.; Pegoraro, P.A.; Sulis, S. Low-cost implementation of an active phasor data concentrator for smart grid. In
Proceedings of the 2018 Workshop on Metrology for Industry 4.0 and IoT, Brescia, Italy, 16–18 April 2018; pp. 78–82. [CrossRef]

162
Energies 2023, 16, 6909

6. Adamiak, M.G.; Kanabar, M.; Rodriquez, J.; Zadeh, M.D. Design and implementation of a synchrophasor data concentrator. In
Proceedings of the 2011 IEEE PES Conference on Innovative Smart Grid Technologies—Middle East, Jeddah, Saudi Arabia, 17–20
December 2011; pp. 1–5. [CrossRef]
7. Retty, H.; Delport, J.; Centeno, V. Development of tests and procedures for evaluating phasor data concentrators. In Proceedings
of the 2013 IEEE Grenoble Conference, Grenoble, France, 16–20 June 2013; pp. 1–5. [CrossRef]
8. Vanfretti, L.; Jónsdóttir, G.M.; Almas, M.S.; Rebello, E.; Firouzi, S.R.; Baudette, M. Audur—A platform for synchrophasor-based
power system wide-area control system implementation. SoftwareX 2018, 7, 294–301. [CrossRef]
9. Danielson, C.F.M.; Vanfretti, L.; Almas, M.S.; Choompoobutrgool, Y.; Gjerde, J.O. Analysis of communication network challenges
for synchrophasor-based wide-area applications. In Proceedings of the 2013 IREP Symposium Bulk Power System Dynamics and
Control—IX Optimization, Security and Control of the Emerging Power Grid, Rethymno, Greece, 25–30 August 2013; pp. 1–13.
[CrossRef]
10. Yao, W.; You, S.; Wang, W.; Deng, X.; Li, Y.; Zhan, L.; Liu, Y. A Fast Load Control System Based on Mobile Distribution-Level
Phasor Measurement Unit. IEEE Trans. Smart Grid 2020, 11, 895–904. [CrossRef]
11. Liu, Y.; You, S.; Yao, W.; Cui, Y.; Wu, L.; Zhou, D.; Zhao, J.; Liu, H.; Liu, Y. A Distribution Level Wide Area Monitoring System for
the Electric Power Grid–FNET/GridEye. IEEE Access 2017, 5, 2329–2338. [CrossRef]
12. You, S.; Zhao, J.; Yao, W.; Liu, Y.; Cui, Y.; Wu, L.; Guo, J.; Liu, Y. FNET/GridEye for Future High Renewable Power Grids—
Applications Overview. In Proceedings of the 2018 IEEE PES Transmission & Distribution Conference and Exhibition—Latin
America (T & D-LA), Lima, Peru, 18–21 September 2018; pp. 1–5. [CrossRef]
13. Liu, Y.; Yao, W.; Zhou, D.; Wu, L.; You, S.; Liu, H.; Zhan, L.; Zhao, J.; Lu, H.; Gao, W.; et al. Recent developments of
FNET/GridEye—A situational awareness tool for smart grid. CSEE J. Power Energy Syst. 2016, 2, 19–27. [CrossRef]
14. Schweitzer, E.O.; Whitehead, D.; Zweigle, G.; Ravikumar, K.G. Synchrophasor-based power system protection and control
applications. In Proceedings of the 2010 63rd Annual Conference for Protective Relay Engineers, College Station, TX, USA, 29
March–1 April 2010; pp. 1–10.
15. Rodrigues, Y.R.; Abdelaziz, M.; Wang, L. D-PMU Based Secondary Frequency Control for Islanded Microgrids. IEEE Trans. Smart
Grid 2020, 11, 857–872. [CrossRef]
16. Zenelis, I.; Wang, X.; Kamwa, I. Online PMU-Based Wide-Area Damping Control for Multiple Inter-Area Modes. IEEE Trans.
Smart Grid 2020, 11, 5451–5461. [CrossRef]
17. Chenine, M.; Nordström, L. Investigation of communication delays and data incompleteness in multi-PMU Wide Area Monitoring
and Control Systems. In Proceedings of the 2009 International Conference on Electric Power and Energy Conversion Systems,
(EPECS), Sharjah, United Arab Emirates, 10–12 November 2009; pp. 1–6.
18. Microgrid Knowledge. Available online: https://microgridknowledge.com/microgrids-navigant/ (accessed on 25
September 2023).
19. Lasseter, R.H. MicroGrids. In Proceedings of the 2001 IEEE Power Engineering Society Winter Meeting, Conference Proceedings
(Cat. No.01CH37194), Columbus, OH, USA, 28 January–1 February 2001; Volume 1, pp. 146–149. [CrossRef]
20. Hatziargyriou, N.; Asano, H.; Iravani, R.; Marnay, C. Microgrids. IEEE Power Energy Mag. 2007, 5, 78–94. [CrossRef]
21. Bidram, A.; Davoudi, A. Hierarchical Structure of Microgrids Control System. IEEE Trans. Smart Grid 2012, 3, 1963–1976.
[CrossRef]
22. Canciello, G.; Cavallo, A.; Guida, B. Control of Energy Storage Systems for Aeronautic Applications. J. Control Sci. Eng. 2017,
2017, 2458590. [CrossRef]
23. Cavallo, A.; Canciello, G.; Guida, B. Energy Storage System Control for Energy Management in Advanced Aeronautic Applications.
Math. Probl. Eng. 2017, 2017, 4083132. [CrossRef]
24. Adhikari, P.M.; Vanfretti, L. Delay Analysis of a Real-Time Hard Reconfigurable Synchrophasor Synchronization Gateway. In
Proceedings of the International Conference on Control and Optimization of Renewable Energy Systems (CORES), Anaheim,
CA, USA, 6–7 December 2019. Available online: https://www.actapress.com/PaperInfo.aspx?paperId=456528 (accessed on 25
September 2023).
25. Firouzi, S.R.; Vanfretti, L.; Ruiz-Alvarez, A.; Hooshyar, H.; Mahmood, F. Interpreting and implementing IEC 61850-90-5 Routed-
Sampled Value and Routed-GOOSE protocols for IEEE C37.118.2 compliant wide-area synchrophasor data transfer. Electr. Power
Syst. Res. 2017, 144, 255–267. ISSN 0378-7796. [CrossRef]
26. Buttayak, S.; Wornpuen, A.; Promparn, N.; Charbkaew, N.; Bunyagul, T. Design of Phasor Data Concentrator for phasor
monitoring system. In Proceedings of the 2012 IEEE Conference on Sustainable Utilization and Development in Engineering and
Technology (STUDENT), Kuala Lumpur, Malaysia, 6–9 October 2012; pp. 102–107. [CrossRef]
27. Bihari, S.P.; Sadhu, P.K.; Sarita, K.; Khan, B.; Arya, L.D.; Saket, R.K.; Kothari, D.P. A Comprehensive Review of Microgrid Control
Mechanism and Impact Assessment for Hybrid Renewable Energy Integration. IEEE Access 2021, 9, 88942–88958. [CrossRef]
28. Guerrero, J.M.; Vasquez, J.C.; Matas, J.; de Vicuna, L.G.; Castilla, M. Hierarchical Control of Droop-Controlled AC and DC
Microgrids—A General Approach Toward Standardization. IEEE Trans. Ind. Electron. 2011, 58, 158–172. [CrossRef]
29. Pourramezan, R.; Seyedi, Y.; Karimi, H.; Zhu, G.; Mont-Briant, M. Design of an advanced phasor data concentrator for monitoring
of distributed energy resources in smart microgrids. IEEE Trans. Ind. Inform. 2017, 13, 3027–3036. [CrossRef]

163
Energies 2023, 16, 6909

30. Guillaud, X.; Faruque, M.O.; Teninge, A.; Hariri, A.H.; Vanfretti, L.; Paolone, M.; Dinavahi, V.; Mitra, P.; Lauss, G.; Dufour, C.; et
al. Applications of Real-Time Simulation Technologies in Power and Energy Systems. IEEE Power Energy Technol. Syst. J. 2015, 2,
103–115. [CrossRef]
31. Konakalla, S.A.R.; Valibeygi, A.; de Callafon, R.A. Microgrid dynamic modeling and islanding control with synchrophasor data.
IEEE Trans. Smart Grid 2020, 11, 905–915. [CrossRef]
32. Castello, P.; Muscas, C.; Pegoraro, P.A.; Sulis, S. Active phasor data concentrator performing adaptive management of latency.
Sustain. Energy Grids Netw. 2018, 16, 270–277. [CrossRef]
33. Adhikari, P.M.; Vanfretti, L.; Mishra, C.; Jones, K.D. A Reconfigurable Synchrophasor Synchronization Gateway & Controller
Architecture for DERs. In Proceedings of the 2022 International Conference on Smart Grid Synchronized Measurements and
Analytics (SGSMA), Split, Croatia, 24–26 May 2022; pp. 1–6. [CrossRef]
34. Microgrids. Available online: https://bit.ly/TyphoonHILMG (accessed on 25 September 2023).
35. Almas, M.S.; Baudette, M.; Vanfretti, L. Utilizing synchrophasor-based supplementary damping control signals in conventional
generator excitation systems. Electr. Power Syst. Res. 2018, 157, 157–167. [CrossRef]
36. Díez-Maroto, L.; Vanfretti, L.; Almas, M.S.; Jónsdóttir, G.M.; Rouco, L. A WACS exploiting generator Excitation Boosters for
power system transient stability enhancement. Electr. Power Syst. Res. 2017, 148, 245–253. [CrossRef]
37. Rebello, E.; Vanfretti, L.; Almas, M.S. Experimental Testing of a Real-Time Implementation of a PMU-Based Wide-Area Damping
Control System. IEEE Access 2020, 8, 25800–25810. [CrossRef]
38. Chompoobutrgool, Y.; Vanfretti, L. Using PMU signals from dominant paths in power system wide-area damping control. Sustain.
Energy Grids Netw. 2015, 4, 16–28. [CrossRef]
39. Plug-and-Play Microgrid Library and Testing of Microgrid Controller. Available online: https://bit.ly/TyphoonControl (accessed
on 25 September 2023).
40. NI Expertise: Power Quality Analysis. Available online: https://bit.ly/NIPQAnalysis (accessed on 25 September 2023).
41. Derviškadić, A.; Romano, P.; Pignati, M.; Paolone, M. Architecture and experimental validation of a low-latency phasor data
concentrator. IEEE Trans. Smart Grid 2018, 9, 2885–2893. [CrossRef]
42. LabVIEW Electrical Power Toolkit API Reference. Available online: https://bit.ly/NIPMUvis (accessed on 25 September 2023).
43. Adhikari, P.M.; Hooshyar, H.; Vanfretti, L. Experimental Quantification of Hardware Requirements for FPGA-Based Reconfig-
urable PMUs. IEEE Access 2019, 7, 57527–57538. [CrossRef]
44. Available online: https://www.typhoon-hil.com/products/hil-connect/ (accessed on 25 September 2023).
45. CT 910. Available online: https://www.candelatech.com/ct910_product.php (accessed on 25 September 2023).
46. Song, L.; Wang, Z.; Luo, W. Analysis on the Response Time of the Battery Energy Storage System. Mater. Sci. Energy Eng. (CMSEE
2014) 2015, 553–560. [CrossRef]
47. Edris, A.; Mehraban, A.S.; Rahman, M.; Gyugyi, L.; Arabi, S.; Reitman, T. Controlling the flow of real and reactive power. IEEE
Comput. Appl. Power 1998, 11, 20–25. [CrossRef]
48. Pierce, R.E.; Hamilton, B.W. Phase-angle control of system interconnections. Electr. Eng. 1939, 58, 83–92. [CrossRef]
49. Jolissaint, C.H.; Arvanitidis, N.V.; Luenberger, D.G. Decomposition of Real and Reactive Power Flows: A Method Suited for
On-Line Applications. IEEE Trans. Power Appar. Syst. 1972, PAS-91, 661–670. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

164
 energies
Article
Detection of False Data Injection Attacks in a Smart Grid Based
on WLS and an Adaptive Interpolation Extended Kalman Filter
Guoqing Zhang 1,2 , Wengen Gao 1,2, *, Yunfei Li 1,2 , Xinxin Guo 1,2 , Pengfei Hu 1,2 and Jiaming Zhu 1,2

1 School of Electrical Engineering, Anhui Polytechnic University, Wuhu 241000, China;

gq17556964576@163.com (G.Z.); lyf@mail.ahpu.edu.cn (Y.L.); guoxinxin@ahpu.edu.cn (X.G.);
h382024188@163.com (P.H.); m18315378156@163.com (J.Z.)
2 Key Laboratory of Advanced Perception and Intelligent Control of High-End Equipment, Chinese Ministry of
Education, Wuhu 241000, China
* Correspondence: ahpuchina@ahpu.edu.cn

Abstract: An accurate power state is the basis of the normal functioning of the smart grid. However,
false data injection attacks (FDIAs) take advantage of the vulnerability in the bad data detection
mechanism of the power system to manipulate the process of state estimation. By attacking the
measurements, then affecting the estimated state, FDIAs have become a serious hidden danger that
affects the security and stable operation of the power system. To address the bad data detection
vulnerability, in this paper, a false data attack detection method based on weighted least squares
(WLS) and an adaptive interpolation extended Kalman filter (AIEKF) is proposed. On the basis of
applying WLS and AIEKF, the Euclidean distance is used to calculate the deviation values of the
two-state estimations to determine whether the current moment is subjected to a false data injection
attack in the power system. Extensive experiments were conducted to simulate an IEEE-14-bus
power system, showing that the adaptive interpolation extended Kalman filter can compensate for
the deficiency in the bad data detection mechanism and successfully detect FDIAs.

Keywords: false data injection attacks; adaptive interpolation extended Kalman filter; state estimation;
Euclidean distance; smart grid
Citation: Zhang, G.; Gao, W.; Li, Y.;
Guo, X.; Hu, P.; Zhu, J. Detection of
False Data Injection Attacks in a
Smart Grid Based on WLS and an
Adaptive Interpolation Extended 1. Introduction
Kalman Filter. Energies 2023, 16, 7203. The modern power system is developing towards intelligence, and plenty of intelligent
https://doi.org/10.3390/en16207203 devices, such as smart meters and sensors, promote the transformation of the power system
Academic Editors: Ying-Yi Hong,
in the modes of power generation, transformation, transmission, and distribution, which
Javier Contreras and Michael
makes the smart grid a typical cyber–physical system (CPS) [1,2]. In a smart grid, the
Negnevitsky supervisory control and data acquisition (SCADA) system collects and analyzes real-time
data from field devices across the network. Finally, the SCADA reports back to the control
Received: 8 August 2023 center, which then makes adjustments to the power generation and distribution of the grid
Revised: 10 October 2023
based on this information [3].
Accepted: 20 October 2023
The susceptibility of the power cyber–physical system (CPS) to cyber attacks is a result
Published: 23 October 2023
of the unpredictable nature of sensor data in the perception layer and the unrestricted
communication channels for data exchange [4,5]. Among the many types of cyber attacks,
attacks against smart grids and industrial control systems are the most common; the
Copyright: © 2023 by the authors. damage caused to the system cannot be underestimated, seriously affecting the normal
Licensee MDPI, Basel, Switzerland. production activities of society. For example, in 2010, the “Stuxnet” virus attack on a
This article is an open access article Belarusian enterprise, which caused anomalies in uranium enrichment centrifuges and
distributed under the terms and generators at the Iranian nuclear power plant, resulted in damage to many pieces of
conditions of the Creative Commons equipment [6]. In 2015, Black Energy, a cyber virus targeting the power grid, caused power
Attribution (CC BY) license (https:// outages at some Ukrainian power plants, disrupting the power supply to many factories in
creativecommons.org/licenses/by/ the Ivano-Frankivsk region and affecting production [7]. The investigation revealed that
4.0/).

Energies 2023, 16, 7203. https://doi.org/10.3390/en16207203 165 https://www.mdpi.com/journal/energies

Energies 2023, 16, 7203

the incident resulted in the malicious deletion of historical grid measurements stored in the
SCADA, which made recovery extremely difficult.
A false data injection attack (FDIA) is a novel attack method specifically targeting the
integrity of state estimation data in the power CPS [8,9]. The attackers inject false data,
which affects the power flow calculation, control decisions, etc., through smart grid sensors,
controllers, and remote control units to tamper with the original data of the grid. This
situation can potentially result in the malfunction of grid equipment and, in severe cases,
the complete paralysis of the power network, which not only poses a significant threat to
grid security but also carries the potential for substantial economic losses. Figure 1 shows the
structure of a smart grid system and an illustration of an FDIA.

$WWDFNHU

6PDUW*ULG (TXLSPHQWV 0HDVXUHPHQWV

0LVOHDGLQJ $FTXLVLWLRQ
,QVWUXFWLRQ (OHFWULFLW\*HQHUDWLRQ 6HQVRUV

(OHFWULFLW\7UDQVPLVVLRQ
0HWHUV

(OHFWULFLW\'LVWULEXWLRQ
5HPRWH7HUPLQDOV
(OHFWULFLW\&RQVXPSWLRQ

(QHUJ\0DQDJHPHQW6\VWHP

6 2SWLPDOSRZHUIORZ 6
)DOVH
& &
$XWRPDWLF*HQHUDWLRQ&RQWURO 0HDVXUHPHQWV
$ 6WDWH $
' (VWLPDWLRQ '
(FRQRPLF'LVSDWFK
$ $
(PHUJHQF\$QDO\VLV

Figure 1. Illustration of an FDIA in a smart grid.

Liu et al. [10] first introduced the topic of FDIAs in the literature, where it was
hypothesized that an attacker could access the current configuration information of a
smart grid and manipulate meter or sensor measurements. Such an attack could insert
false data into specific state variables, avoiding detection by current bad data detection
algorithms. Yang et al. [11] delved into the challenge of determining the most effective
attack strategy. This strategy, known as an injection attack strategy, involves selecting a
specific set of meters to manipulate in a way that maximizes the resulting damage. They
not only formalized this problem but also developed efficient algorithms to pinpoint the
ideal set of meters for such attacks. It is important to highlight that even if these attacks are
isolated to specific devices, their impact on the smart grid can be catastrophic due to the
grid’s intricate interconnections. As described by He et al. [12], electricity theft by attackers
by modifying smart meter data has seriously affected utility security. Therefore, many
researchers have devoted themselves to the detection of false data injection attacks in order
to safeguard the security of the smart grid.
When the power system is subjected to malicious false data injection, the state estima-
tion result of the WLS under attack is updated in real time by incorporating an estimation
algorithm of an extended Kalman filter, which has a hysteresis in its state estimation process,
and by observing disparities in the estimation outcomes produced by the two algorithms,
making it possible to detect the FDIA. Meanwhile, to improve the accuracy and reduce the
linearization error of the EKF, the adaptive interpolation strategy is introduced. Therefore,

166
Energies 2023, 16, 7203

in this paper, we propose a detection method based on WLS and AIEKF. Considering that
the two algorithms have different degrees of correspondence to real-time information, the
FDIA can be effectively detected. The main contributions can be summarized as follows:
• Considering the linearization error of the EKF algorithm in state estimation in a power
system, the adaptive interpolation strategy is introduced. The pseudomeasurements
between two consecutive measurements are inserted by linear interpolation to improve
the estimation accuracy.
• We propose a novel FDIA detection method that combines AIEKF and WLS, marking
the first instance of their joint application in this context.
• We conduct many experiments on an IEEE-14-bus power system to demonstrate the
proposed algorithm’s performance in detecting FDIAs. The result shows that the
method can effectively detect FDIAs.
The remainder of the paper is structured as follows. Section 2 provides an overview
of relevant literature pertaining to the detection of FDIAs. Section 3 outlines the system
model employed in this study. In Section 4, we delve into the details of the proposed AIEKF
algorithm. In Section 5, the detection principle is described. The experiments and results
are presented in Section 6. Finally, in Section 7, we present our concluding remarks and
suggest directions for future work. A list of abbreviations and acronyms is provided in
Table 1.

Table 1. A list of abbreviations and acronyms.

Abbreviations and Acronyms Full Name

AIEKF Adaptive interpolation extended Kalman filter
BDD Bad data detection
CPS Cyber–physical system
EMS Energy management system
FDIAs False data injection attacks
LNR Largest normalized residual
PMUs Phase measurement units
RMSE Root mean square error
SCADA Supervisory control and data acquisition
WLS Weighted least squares

2. Related Work
Since the concept of FDIAs was introduced, the issues related to FDIAs have received
a great deal of attention in both academic research and industry. Many scholars have
studied bad data in state estimation and proposed corresponding detection methods for
FDIAs based on their research. Although FDIA detection algorithms differ from each other,
the algorithms can be classified into two categories [13]: model-based algorithms and
data-driven algorithms.

2.1. Model-Based Detection Algorithms

Nowadays, with the increasing degree of interconnection of power systems in smart
grids, the simple use of weighted least squares is no longer applicable, and many vari-
ants have emerged. Moslemi et al. [14] introduced an innovative cyber attack detection
technique based on maximum likelihood estimation. This method capitalizes on the near-
chordal sparsity (NCS) characteristics of the power grid to establish a robust framework
for addressing the corresponding maximum likelihood estimation problem. Furthermore,
they broke down this detection approach into a series of localized maximum likelihood es-
timation problems. This approach not only safeguards privacy but also mitigates potential

167
Energies 2023, 16, 7203

issues. Chen et al. [15] introduced a novel method for detecting FDIAs using kernel density
estimation. This method utilizes the concept of kernel density estimation, which utilizes
historical data to estimate the probability distributions of both measurements and control
commands. Additionally, it calculates confidence intervals for these estimates at a specified
significance level. Zhao et al. [16] proposed a method that employs short-term state predic-
tion in smart grids to detect spurious data injection attacks. This method utilizes real-time
and predicted states to pinpoint potential false data injections by analyzing the differences
between them. By continuously monitoring and comparing the accuracy of short-term state
predictions, this approach enhances the security and resilience of smart grid systems by
improving the detection of malicious attacks. Li et al. [17] proposed a detection method
based on the watermark embedding technique. The method uses a dynamic watermark
embedding technique to embed security-enhancing markers (watermarks) into the grid
measurement data to ensure the integrity and authenticity of the data. Then, the data
embedded with the watermark are processed and detected using the EKF algorithm to
identify possible faulty data insertion attacks. Through the joint application of dynamic
watermarking and EKF, this method can improve the detection capability of smart grid
systems against FDIAs and provide increased security. The above methods cannot be
used alone; they must be used with the help of some determination methods to determine
the existence of FDIAs. Some determination methods are available, such as Euclidean
distance [16,18], maximum normalized residuals [19,20], chi-square testing [18,21], Cumu-
lative sum testing [22,23], Kullback–Leibler distance [14,24], and cumulative error sum of
squares probability density curves [25,26].

2.2. Data-Driven Detection Algorithms

In contrast to detection algorithms that rely on system models, data-driven detection
algorithms operate without the need for system parameters and models. The process
of detecting FDIAs is independent of these factors, and it relies on a large amount of
historical data from the smart grid to speculate on future data [13]. Mahi-Al-Rashid
et al. [27] proposed an innovative approach for countering FDIAs. They employed a
CNN-LSTM-based self-encoder sequence-to-sequence architecture for prediction-assisted
anomaly detection. Additionally, they suggested an adaptive optimal thresholding method
based on consumption patterns to identify unusual behaviors. Yu et al. [28] proposed
an online method for detecting FDIAs that merges wavelet transform with deep neural
networks. This approach involves extracting features from grid sensor data using wavelet
transform and subsequently employing a deep neural network for real-time detection. This
methodology effectively enhances the security and reliability of the smart grid system by
identifying potential attacks as they occur. Wang et al. [29] combined Kalman filtering with
recurrent neural networks. The actual state of the grid was first estimated using Kalman
filtering. Next, the estimated state was modeled and predicted using recurrent neural
networks. By analyzing the disparity between the predicted state and the real measured
data, the system can identify the existence of FDIAs. Jorjani et al. [30] introduced a novel
algorithm designed for the detection of FDIAs in AC state estimation. This algorithm
employs outlier detection techniques to the outcomes of state estimation because the attack
may lead to data inconsistencies between buses, resulting in an abnormal graph structure.
The above methods can be proven to detect FDIAs.

3. System Model
The state estimation of the power system usually deeply relies on the system model.
The selection and establishment of the model have a substantial impact on the results of
the system state calculation, which directly lead to the accuracy of the acquired state. State
estimation in the power system is a crucial element within EMS, as it provides essential real-
time information about the grid’s operational status, and it is the basis for other high-level
applications to realize the calculation and analysis.

168
Energies 2023, 16, 7203

The measurements for power system state estimation are collected from the grid by
SCADA or phase measurement units (PMUs). PMUs are able to provide accurate and
synchronized phase measurements for geographically dispersed buses in the grid by taking
advantage of the high accuracy, sub-microsecond time synchronization, and unprecedented
reporting rate [31]. And if the system is completely observable with PMU measurements,
the state estimation process is a linear procedure. The proposed algorithm aims at solving
the linearization of EKF for state estimation. Therefore, the proposed algorithm can be
applied to the mentioned PMU-based state estimation problem by reducing the linearization
steps of the AIEKF algorithm. We can discuss a situation in which there are m measurements
and n state variables. In an AC power system, the connection between measurements and
state variables is characterized by a nonlinear relationship, which can be represented as:
z = h( x) + e (1)
where z ∈ Rm×1 is the measurement vector; x ∈ Rn×1 is the state vector, typically bus
voltage amplitude and phase; e ∈ Rm×1 is the measurement error vector that satisfies
e = (e1 , e2 , . . . , em ) T ∼ N (0m×1 , R); and h(·) represents the nonlinear relationship between
the measurement vector (z) and the state vector (x).
To analyze the correlation between the bus voltage, phase angle, and bus current of the
grid system and determine the nonlinear relationship h(·), we must streamline the power
system branch by representing it through an equivalent circuit, as illustrated in Figure 2.
Subsequently, utilizing the AC model of the power system, we establish the connection
between the state variables and measurements, which can be formulated as follows:
3LM4L 3MM4M
3LM M4LM 3ML M4 ML
L M

*LM M%LM

M\ M\

Figure 2. Power system branch equivalent circuits.

- (
Pi = Vi ∑ Vj Gij cos θij + Bij sin θij (2)
j∈ T
- (
Qi = Vi ∑ Vj Gij sin θij − Bij cos θij (3)
j∈ T
- (
Pij = Vi2 Gij − Vi Vj Gij cos θij + Bij sin θij (4)

- (
Qij = −Vi2 Bij − Vi Vj Gij sin θij − Bij cos θij (5)
where Vi and Vj are the voltage amplitudes at bus i and bus j, respectively; Pi and Qi
represent the active and reactive power injection of bus i, respectively; Pij and Qij denote
active power flow and reactive power flow from bus i to bus j, respectively; Gij and Bij
denote the conductance and susceptance of the line from bus i to bus j, respectively; θij
denotes the phase angle difference of the line voltage from bus i to bus j; and T denotes the
set buses adjacent to bus i.

3.1. State Estimation

The most commonly used state estimation in power systems is the weighted least
squares method, which is still widely used [32–34]. Under this method, the objective
function (J ( x̂)) is the weighted sum of squares of the difference between the measured and

169
Energies 2023, 16, 7203

estimated values. With the smallest objective function value, the obtained x is the closest
approximation to the true state of the system. Based on the weighted least squares method,
the objective function (J ( x̂)) can be expressed as:
J ( x̂) = [z − h( x̂)] T R−1 [z − h( x̂)]. (6)
Then,
x̂ = arg min [z − h( x̂)] T R−1 [z − h( x̂)] (7)
To solve the nonlinear WLS problem, we can linearize the measurement equation
around x, then apply the linear WLS method. The final result is expressed as:
  −1     
x̂k+1 = x̂k + G x̂k H T x̂k R−1 z − h x̂k (8)

     
G x̂k = H T x̂k R−1 H x̂k (9)

where k is the k-th iteration index, and H ∈ Rm×n is the Jacobian matrix of the measurement
equation, which can be expressed as:
⎡ ∂P ∂Q
⎤T
∂Vi ∂Pi ∂Qi ij ij
⎢ ∂V ∂V ∂V ∂V ∂V ⎥
H=⎣ ⎦ (10)
∂Vi ∂Pi ∂Qi ∂Pij ∂Qij
∂θ ∂θ ∂θ ∂θ ∂θ

3.2. Bad Data Detection

Traditional methods for detecting bad data, like the chi-square test and largest normal-
ized residual (LNR) test, rely on the results obtained from WLS estimation.
By checking the value of the objective function (J ( x̂)), we can determine whether there
are bad data in the power system or not. In particular, in the chi-square test, we need to
perform null hypothesis testing, which can be expressed as:
'
J ( x̂) > χ2(m−n),p , Reject H0
(11)
J ( x̂) ≤ χ2(m−n),p , Accept H0

where H0 represents the original hypothesis, i.e., there are no bad data, and χ2(m−n),p is the
chi-square test threshold with a confidence level of p and a degree of freedom corresponding
to (m − n).
The LNR test stands as another commonly employed approach for bad data detection.
Its core concept revolves around the normalization of measurement residuals, which can
be formulated as follows:
|z − h ( x̂ )|
ri = i √ i (12)
Ωii
   −1 
Ω= I − H ( x̂ ) H T ( x̂) R−1 H ( x̂ ) H T ( x̂) R−1 R (13)

where zi is the ith measurement, Ωii is the ith diagonal entry of Ω, and I is the identity
matrix. If there exit bad data in the power system, the largest normalized residual is larger
than the threshold (ε).
The chi-square test and LNR test are generally effective for detecting natural bad data,
which typically induce large measurement residuals [35].

3.3. FDIA Generation

If an attacker possesses precise information regarding real-time state estimation, net-
work topology, and parameters, they can achieve an elaborate FDIA without being detected.
When measurement meters are tampered with, the measurement (z) changes to z f , and the
attacked measurement (z f ) changes to:

170
Energies 2023, 16, 7203

z f = h( x) + a + e (14)
where a ∈ Rm ×1
is the attacked vector.
As an elaborate FDIA, the attacked vector requires a certain condition, which is
expressed as:
a = h( x̂ + c) − h( x̂) (15)
where c ∈ Rn ×1
is the deviation of the state variable, and x̂ is the state-estimated vector without
an FDIA.
As indicated by the equation above, FDIAs can lead to an identical measurement resid-
ual vector compared to the condition without an attack. To be specific, the measurement
residuals between the pre-attack and post-attack states can be described as follows:
r = z − h( x̂) (16)

r f = z f − h( x̂ + c) = z + a − h( x̂) − a = z − h( x̂) (17)

The measurement residual between the pre-attack and post-attack states does not
change; hence, an elaborate FDIA is stealthy and can avoid detection by the existing BDD
system based on residuals [36].

4. Dynamic State Estimation Model

4.1. Extended Kalman Filter (EKF)
The physical power information system in an AC power system is inherently complex
and highly multidimensional and nonlinear. The state and measurement equations for
state estimation can be formalized as:
x k = f ( x k − 1 ) + ωk − 1 (18)

zk = h( xk ) + ek (19)
where xk and zk denote the state vector and the measurement vector at time k, respectively;
f (.) denotes the state transfer equation from k − 1 to k; h(.) denotes measurement equation;
and ωk−1 and ek denote the process and measurement noise, respectively, which are
independent of each other.
Since the KF algorithm can only deal with linear system problems, it is not applicable
to nonlinear problems such as power systems, so the EKF algorithm is derived. The EKF
algorithm first uses Taylor’s formula to linearize the nonlinear system, then filters it using
the basic formula of the KF algorithm. Specifically, state Equation (18) carries out Taylor
series expansion at the state estimation quantity (x̂k−1 ) and ignores items at quadratic
levels and higher. Similarly, measurement Equation (19) carries out Taylor series expansion
at the state prediction quantity (x̃k ) and ignores items at quadratic levels and higher. The
linearization models are expressed as:
+
∂ f ( x̂ )+
xk ≈ f ( x̂k−1 ) + ∂ x̂ k−1 + ( xk−1 − x̂k−1 ) + ωk−1
k −1 x̂k−1 (20)
= Fk−1 xk−1 + ωk−1 + uk−1
+
∂h( x̃ ) +
zk ≈ h( x̃k ) + ∂ x̃ k + ( xk − x̃k ) + ek
k x̃k (21)
= Hk xk + ek + yk
+
∂ f ( x̂ )+
where Fk−1 = ∂ x̂ k−1 + is the Jacobian matrix of the state equation, uk−1 = f ( x̂k−1 ) −
k −1
+ x̂k−1 +
∂ f ( x̂k−1 ) + ∂h( x̃k ) +
∂ x̂k−1 x̂ + x̂ k − 1 is an externality item, H k = ∂ x̃k + x̃ is the Jacobian matrix of the mea-
k −1 + k
∂h( x̃k ) +
surement equation, and yk = h( x̃k ) − ∂ x̃ + x̃k is an externality item.
k x̃k

171
Energies 2023, 16, 7203

On the basis of Equations (18) and (19), the basic formula of the EKF algorithm is expressed
as follows:
(1) Prediction steps:
x̃k|k−1 = Fk−1 x̂k−1 (22)

P̃k|k−1 = Fk−1 P̂k−1 FkT−1 + Qk−1 (23)

(2) Update steps:
Kk = P̃k|k−1 Hk T ( Hk P̃k|k−1 Hk T + Rk )−1 (24)

 
x̂k = x̃k|k−1 + Kk zk − Hk x̃k|k−1 (25)

P̂k = ( I − Kk Hk )P̃k|k−1 (26)

where ˜ and ˆ indicate the predicted and estimated quantities, respectively; I is the identity
matrix; P is the state covariance matrix; Q and R are the covariance matrices of the process
noise and measurement noise error vectors, respectively, which are assumed to be white
Gaussian processes; and K is Kalman gain.
The EKF algorithm is extensively employed for dynamic state estimation in power
systems due to its straightforward model development and efficient computational perfor-
mance in practical engineering applications. However, since the EKF algorithm ignores the
higher-level items in the linearization process, it results in a large truncation error in power
systems with highly nonlinear characteristics, resulting in a decrease in the filtering effect.

4.2. Adaptive Interpolation Strategy

To enhance the dynamic state estimation capabilities of the EKF algorithm in the
power system, an adaptive interpolation method is proposed to strike a balance between
estimation precision and computational efficiency [37].
Based on Equation (18), we need to quantify the nonlinear index of state function f ( x)
to obtain η f , which can be expressed as:
+
∂ f ( x̂k−1 ) ++
ε f = xk − f ( x̂k−1 ) − + ( xk−1 − x̂k−1 ) (27)
∂ x̂k −1 x̂k−1

η f = ε Tf Qk−1 ε f (28)

where ε f is the difference between f ( x) and the corresponding linear approximation.

Similarly, based on Equation (19), we can obtain the nonlinear index (ηh ) of measure-
ment function h( x), which can be represented as:
+
∂h( x̃k ) ++
εh = zk − h( x̃k ) − ( xk − x̃k ) (29)
∂ x̃ + k x̃k

ηh = εhT R− 1
k εh (30)
where εh is the difference between h( x) and the corresponding linear approximation.
As shown in Equations (28) and (30), ε f and εh are normalized by Qk and Rk . Under
the process, ε f and εh are numerically non-negative. Hence, if ε f  Qk and εh  Rk , η f
and ηh are both much less than 1, and the system can be considered quasilinear. Otherwise,
according to the size of the nonlinearization index, the pseudomeasurements must be
added between two consecutive sampling points to increase the sampling rate and reduce
the degree of nonlinearity of the system.
The interpolation factor (r) is closely related to the sizes of ε f and εh . The larger
nonlinearization indices η f and ηh are, the larger the interpolation factor (r) is. Conversely,
the interpolation factor (r) is smaller. It is important to emphasize that ε f = 0 and εh = 0
in the linear system. Therefore, the system does not interpolate.

172
Energies 2023, 16, 7203

The finite state machine model is shown in Figure 3. In practical applications, we can
introduce as many states as required to the FSM model to accommodate the nonlinearity indices.
There are three parameters in each state (i): the interpolation factor (ri ), the upper threshold
(Ui ), and the lower threshold (Li ). In addition, as the state (i) changes, the interpolation factor
is set to ri+1 > ri . The selection of the interpolation factor (r) is shown in Algorithm 1.
The thresholds of each state are different, and they are set depending on different
scenarios. When selecting the thresholds, it is necessary to ensure that the upper threshold
(Ui ) is larger than the lower threshold (Li ). Furthermore, as Ui and Li become smaller, the
interpolation factor (r) and estimation accuracy increase, and the algorithm consumes more
time. It is important to highlight that the nonlinear indices can take on discrete values. To
maintain small values for both η f and ηh , here is how the process works: If either η f or ηh
exceeds Ui , r parameter is increased to minimize the nonlinear error. Conversely, if both η f
and ηh are below Li , r is reduced to lower computational complexity. The specific values of
r for each state can be found in Table 2.

6WDWH
ȘIRUȘK!8
8L8SSHU7KUHVKROG
/L/RZHU7KUHVKROG
ȘI ȘK/

6WDWH 6WDWH

ȘI ȘK/
ȘI ȘK/

ȘIRUȘK!8
ȘIRUȘK!8 6WDWH

Figure 3. Finite-state machine (four states).

Algorithm 1 Choose the interpolation factor (r).

State i r = ri
if η f i or ηhi > Ui then
go to the state i + 1
else
η f i and ηhi < Li
go to the state i − 1
end if
Stay in the state i

Table 2. The size of the interpolation factor (r).

State (i) Interpolation Factor (r) Lower Threshold Upper Threshold

i=1 1 L1 U1
i=2 3 L2 U2
i=3 7 L3 U3
i=4 15 L4 U4

4.3. Adaptive Interpolation EKF (AIEKF)

Building upon the dynamic model outlined in Section 4.1 and the adaptive interpola-
tion approach discussed in Section 4.2, we introduce the AIEKF algorithm in this section.
The AIEKF algorithm effectively strikes a balance between computation time and estima-
tion accuracy, thereby enhancing the performance of the EKF algorithm in power systems.

173
Energies 2023, 16, 7203

A flow chart illustrating the AIEKF algorithm is provided in Figure 4, and its detailed steps
are as outlined as follows:
(1) Initialization: setting the initial state variable (x̂0 ) and state error covariance (P̂0 ).
(2) Adaptive Interpolation: In order to strike a compromise between computational
efficiency and estimation precision, the algorithm incorporates an adaptive interpolation
strategy, which comprises three key steps. Initially, we calculate the nonlinearity indices
of the state transition function and the measurement function (referred to as η f and ηh ,
respectively) using Equations (28) and (30), respectively. In the next step, we ascertain
the interpolation factor (r) by utilizing a finite-state machine model. Finally, r pseudomea-
surements are introduced between two actual measurements through linear interpolation,
which is designed to mitigate the adverse impacts of nonlinearity.
(3) EKF: On the basis of determining the number of interpolation factors (r), the
power system is estimated using the EKF algorithm. Initially, leveraging the state and its
covariance matrix from time k − 1, we derive a prior estimation at time k in accordance
with Equations (22) and (23). Secondly, the correction of the a priori estimation is used
to obtain an a posteriori estimation according to Equations (24)–(26). Thirdly, filtering is
performed between two consecutive samples based on the size of the interpolation factor.
Then, the above steps are repeated until the end of the sampling time.

6WDUW

,QLWLDOL]DWLRQ

$GDSWLYH
,QWHUSRODWLRQ
&DOFXODWLQJQRQOLQHDUPHWULFVIRUVWDWH (.)
HTXDWLRQDQGPHDVXUHPHQWHTXDWLRQ
6WDWHSUHGLFWLRQ
'HWHUPLQLQJLQWHUSRODWLRQIDFWRU µU¶ XVLQJ
ILQLWHVWDWHPDFKLQH
6WDWH XSGDWH
L L
N N ,QFUHDVLQJSVHXGRPHDVXUHPHQWVE\ OLQHDU
LQWHUSRODWLRQ

<HV
1XPEHURILWHUDWLRQVLU

1R
1R
7LPHHQG

<HV

(QG

Figure 4. Flow chart of the AIEKF algorithm.

5. Detection of FDIAs
This section proposes a methodology for FDIA detection based on power system
state estimation. As a nonlinear system in the smart grid, it is difficult to guarantee the
estimation accuracy using traditional state estimation methods. Meanwhile, in order to
improve the stability of the detection algorithm, the real-time state information of the grid
buses is solved according to the system model equation and AIEKF algorithm.
Once the attacker begins to tamper with the measuring instruments, the result of
Equation (25) is different from the previous result and expressed as:
f
x̂k = x̃k|k−1 + Kk [zk + ak − Hk x̃k|k−1 ]
(31)
= x̂k + Kk ak

174
Energies 2023, 16, 7203

f f
where x̂k is the estimated state after the FDIA. To better facilitate estimation, ck = x̂k − x̂k
is introduced. Then, for the next time (k + 1), it can be represented as:
f f f f
x̂k+1 = x̃k+1|k + Kk+1 [zk+1 − Hk+1 x̃k+1|k ]
f f
= + Kk+1 [zk+1 + ak+1 − Hk+1 x̃k+1|k ]
x̃k+1|k (32)
= x̂k+1 + [ I − Kk+1 Hk+1 ] Fk ck + Kk+1 ak+1
The analysis above highlights that the injection bias is influenced by both the currently
injected false data and the bias present in the previously estimated state. Over time, this
injection bias accumulates and gradually shifts the estimated state closer to the actual
system state. When the power system is subjected to an FDIA, the altered measurements
make the WLS state estimation results swing towards the new mean. For the AIEKF
algorithm, due to the constraints of the state transfer matrix and the fact that its estimation
is jointly determined by the predicted and measured values, the state estimation has some
hysteresis, and only small oscillations occur.
Based on WLS and AIEKF estimation results, considering the influence of bus states on
the system, the Euclidean distance in multidimensional spaces is introduced. The Euclidean
distance detection threshold required in FDIA detection is obtained from historical data,
and the Euclidean distance between two points estimated by WLS and AIEKF states is
calculated online in real time and used as the basis for attack detection. The expression for
the Euclidean distance at time k is expressed as follows:
#
n  2
d(k) = ∑ x̂W i,k
LS − x̂ AIEKF
i,k (33)
i =1

where x̂W
i,k
LS AIEKF denotes the AIEKF-
denotes the WLS-based state estimation at time k, x̂i,k
based state estimation, and n denotes the system dimension.
In the n-dimensional grid system state space, the Euclidean distance is employed to
quantify the spatial separation between two points within the same state space at a given
time point. The Euclidean distance of the two state estimation algorithms stabilizes in a
certain range during regular power system operation, which provides a basis for false data
injection attack detection. The detection threshold is expressed as:
τD = max{d(1), . . . , d(n), . . .} + μ (34)
where μ is the threshold margin, which is introduced to prevent false alarms triggered by
minor data fluctuations while the detection system is operating under normal conditions.
Attack detection is performed by comparing the Euclidean distance between the
detection threshold and the two points in the state space, and when d(k ) ≥ τD , it is
considered that there exits an FDIA in the power system; otherwise, it is considered that no
attack occurs. The relation can be expressed as:
4
d(k) < τD , No FDI A
(35)
d(k ) ≥ τD , FDI A
In order to distinguish between bad data and FDIAs, bad data detection is also required
at the end of the above steps. Only if d(k ) ≥ τD and J ( x̂ ) < χ2(m−n),p hold can we conclude
that the power system is under FDIAs. The proposed FDIA detection method based on
WLS and AIEKF is shown in Algorithm 2.

175
Energies 2023, 16, 7203

Algorithm 2 FDIA detection based on WLS and the AIEKF algorithm

1: Initialize state variable x̂k−1 and state error covariance P̂k−1 ; the Euclidean distance detection
threshold τD ;
2: Obtain the measurements by SCADA at time k;
3: In traditional static state estimation, WLS is widely used to calculate an estimated state vector
  −1
x̂W LS , x̂W LS = H T R−1 H HkT R− 1
k k k k k k zk
4: AIEKF
(1) Calculate calculate the nonlinearity indices of the state transition function and the measure-
ment function (referred to as η f and ηh ) using Equations (28) and (30);
(2) Ascertain the interpolation factor r by utilizing a finite state machine model;
(3) Interpolate r pseudo-measurement between two actual measurements through linear interpo-
lation;
(4) Execute the state prediction step of the EKF by applying Equations (22) and (23);
(5) Conduct the measurement update step of EKF by  applying Equations
 (24)–(26) to calculate
estimated state vector x̂kAIEKF , x̂kAIEKF = x̃k|k−1 + Kk zk − Hk x̃k|k−1
5: Calculate the Euclidean distance between two points estimated by WLS and AIEKF states;
6: if d(k ) ≥ τD and J ( x̂ ) < χ2(m−n),p then
7: Exist FDIA and generate early warning;
8: else
9: Continue the state estimation process at time k = k + 1,
10: end if

6. Experiments and Results

This paper introduces a detection approach that relies on state estimation. Considering
the effectiveness of the method in real systems, the power standard IEEE-14-bus system
shown in Figure 5 is used for MATLAB R2021b simulation. The active and reactive power
of each bus are shown in the following Table 3. The data used in this paper come from
MATPOWER trend calculation, which is used to obtain the bus voltage magnitude and
phase-angle truth values, superimposed with zero-mean Gaussian white noise as the
measurements. Furthermore, the estimation computations occur at one-minute intervals,
which aligns with the anticipated average sampling frequency for utilities equipped with
contemporary Energy management systems (EMS).

*
* *
*
 

*
* *

   

 

*
 



 

Figure 5. The IEEE-14-bus power system.

176
Energies 2023, 16, 7203

Table 3. Active and reactive power injections into system buses.

Bus Generation Generation Load Load

Number Active (MW) Reactive (Mvar) Active (MW) Reactive (Mvar)

1 2.3824 −0.1490 0.0000 0.0000

2 0.4000 0.4904 0.2170 0.1270

3 0.0000 0.2744 0.9420 0.1900

4 0.0000 0.0000 0.4780 −0.0390

5 0.0000 0.0000 0.0760 0.0160

6 0.0000 0.2960 0.1120 0.0750

7 0.0000 0.0000 0.0000 0.0000

8 0.0000 0.3092 0.0000 0.0000

9 0.0000 0.0000 0.2950 0.1660

10 0.0000 0.0000 0.0900 0.0580

11 0.0000 0.0000 0.0350 0.0180

12 0.0000 0.0000 0.0610 0.0160

13 0.0000 0.0000 0.1350 0.0580

14 0.0000 0.0000 0.1490 0.0500

6.1. Comparison of Estimation Effects with WLS, EKF, and AIEKF

Next, the estimation effect of AIEKF proposed in this paper is compared with the
standard WLS and EKF before injecting false data. As shown in Figures 6 and 7, bus 11
was randomly selected to compare the estimation performance of the three algorithms in
60 min. It is clear to see from the figures that although the bus voltage and phase angle
fluctuate up and down with time, the AIEKF achieves superior performance relative to
WLS and EKF in state estimation. To further demonstrate the estimation capability of the
proposed algorithm, the estimation results of the voltage amplitude and phase of each bus
under the three algorithms after stabilization are shown in Figures 8 and 9.

1.6
1.02 True value
1.5 WLS estimate
1 EKF estimate
AIEKF estimate
1.4 0.98
Voltage amplitude (p.u.)

0.96
1.3
30 32 34

1.2

1.1

0.9

0.8

0.7
0 10 20 30 40 50 60
t (min)

Figure 6. Bus 11 voltage amplitude estimation.

177
Energies 2023, 16, 7203

-5 True value
-4.4 WLS estimate
-5.02
EKF estimate
-4.5 -5.04 AIEKF estimate
-5.06
Voltage phase (degree)
-4.6
38 40 42
-4.7

-4.8

-4.9

-5

-5.1

-5.2

-5.3
0 10 20 30 40 50 60
t (min)

Figure 7. Bus 11 voltage phase estimation.

1.12 True value

WLS estimate
EKF estimate
1.1 AIEKF estimate
Voltage amplitude (p.u.)

1.08

1.06

1.04

1.02

0.98
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Bus
Figure 8. Bus voltage amplitude estimation.

To validate the efficacy of the AIEKF algorithm introduced in this paper for state
estimation, we use the root mean square error (RMSE) as a metric to assess the accuracy of
the algorithm’s estimations. The RMSE calculation formula is provided below.
$
%
%1 N
RMSE = & ∑ ( xi − x̂i )2 (36)
N i =1

where xi is the ith component of the true value of the state variable, x̂i is the ith component
of the estimation of the state variable, and N is the dimension of the state variable.

178
Energies 2023, 16, 7203

0
-8.7 True value
-2 WLS estimate
-8.75 EKF estimate
AIEKF estimate
-4
-8.8
Voltage phase (degree) -6
5

-8

-10

-12

-14

-16

-18
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Bus
Figure 9. Bus voltage phase estimation.

The RMSE performance metric is calculated in the IEEE-14-bus system, and the results
are shown in Table 4. As shown in Table 4, the RMSE of the AIEKF algorithm is the smallest
of the three algorithms. Compared with WLS and EKF, the RMSE of the AIEKF algorithm
decreases by 79% and 67%, respectively.

Table 4. Comparison of the RMSE of the three algorithms.

Algorithm RMSE
WLS 0.0969
EKF 0.0616
AIEKF 0.0201

6.2. Estimation of the State Variable Before and After FDIA

To assess the viability of the false data injection attack vector strategy, in this paper,
we uses the IEEE-14-bus standard test system for simulation and analysis. An attack
on a local subnetwork, e.g., an attack vector (a), is injected into each bus measurement
value. Meanwhile, it is necessary to ensure that the internal power of the subnetwork
is conserved and that the subnetwork boundary voltage and the transmission power be-
tween the subnetwork and the external network remain unchanged. The introduction
of line blocking constraints leads to the response of the grid security analysis system
so that the attacked measurement value (z + a) is a valid attack value. Under this con-
dition, the attack vector (a = [ P3 , Q3 , P1−2 , P2−3 , P4−2 , Q1−2 , Q2−3 , Q4−2 ] T ) is selected as
a = [0.0020, −0.2029, 0.0084, −0.0073, −0.0059, −0.4874, 0.1329, −0.0723] T , and the incre-
ment of the rest of z is zero. Figure 10 shows the change in the measurement distribution
of the system before and after the attack.

179
Energies 2023, 16, 7203

2
Pre-attack measurements
False data attack values
1.5 Post-attack measurements

Measurement value 1

0.5

-0.5

-1
5 10 15 20 25 30 35 40
Measurement unit

Figure 10. Measurement distribution before and after an attack on the IEEE-14-bus power system.

Once the measurements are tampered with, the state variable (x) changes. Assuming
that the system is subjected to a false data injection attack at 75 min, bus 11 is selected to
observe the change in bus voltage magnitude and phase angle before and after the false
data injection attack occurs. State estimation of system buses using the AIEKF algorithm
is performed to improve the stability and accuracy of the detection algorithm. The state
estimation results of the two algorithms are shown in Figures 11 and 12. As shown in
the figures, in the first 75 min without an attack, AIEKF outperforms WLS in terms of
estimation. The system is attacked by false data injection in the 75th minute, and the two
algorithms converge to the state expectation at different moments. It is clear that AIEKF
converges slowly and with small fluctuations, while WLS is affected by a sudden change in
the measurements and converges quickly to the new state value.

1.45
Ture value
WLS estimate
1.4 AIEKF estimate

1.35
Voltage amplitude (p.u.)

1.3

1.25

1.2

1.15

1.1

1.05

1
0 25 50 75 100 125 150
Time (min)

Figure 11. The voltage amplitude change of bus 11 before and after an attack.

180
Energies 2023, 16, 7203

-8
Ture value
WLS estimate
-9 AIEKF estimate

-10

Voltage phase (degree)

-11

-12

-13

-14

-15

-16
0 25 50 75 100 125 150
Time (min)

Figure 12. The voltage phase change of bus 11 before and after an attack.

6.3. Detection of FDIAs

Normal operation of the power information physics system produces a certain amount
of error, but the residuals caused by measurement noise and system noise are often very
small—much smaller than the threshold allowed for the detection of undesirable data
errors—so that undesirable data can be prevented from interfering with the system. The
values of J ( x̂ ) before and after an attack determined using weighted least squares are
shown in Figure 13. Before the attack, the value of J ( x̂ ) is 4.5728. When the false data
attack is injected into the system, the value of J ( x̂ ) is 4.7041. It is not difficult to find that
the residual does not change much before and after an attack. The IEEE-14-bus system
has a total of 41 measurements, the redundancy is k = m − n = 41 − 27 = 14, and the
significance level (α) is 0.05. According to the statistical chi-square distribution table, the
threshold of bad data detection is 23.685. The residual of the injected attack is within the
threshold. However, the voltage amplitude and voltage phase are changed. The false data
attack vector successfully achieves the attack.

4.8 24

22

20
4.7

18
Residual

16

4.6
14

Pre-attack residual 12
Post-attack residual
Bad data detection residual
4.5 10
0 25 50 75 100 125 150
Time (min)

Figure 13. The residual of WLS estimation before and after an attack.

181
Energies 2023, 16, 7203

This paper proposes a detection method based on the computation of the Euclidean
distance between two points in the state space to detect FDIAs. Using Monte Carlo
simulation with 1000 independent experiments, we can obtain the normal-case Eu-
clidean distance distribution. The maximum value is taken as the detection threshold,
i.e., max{d(1), . . . , d(n), . . .} = 1.847. The detection margin (μ) is set to 0.03, and according
to Equation (30), the detection threshold can be derived as τD = 1.85. After an attack, the
Euclidean distance changes to 17.9586. Figure 14 shows the Euclidean distance distribution
based on the two algorithms before and after an attack.

20
Euclid distance
18 Detection threshold

16

14
Euclid distance

12

10

0
0 25 50 75 100 125 150
Time (min)

Figure 14. The Euclidean distance before and after an attack.

As can be seen from the figure, during the first 75 min, when the system is not under
attack, the Euclidean distance between the state points stays within a certain range below
the predefined detection threshold, which indicates that the system does not detect an attack
according to the judgment conditions. When the system is attacked after the 75th minute,
the two algorithms converge to the new state values at different moments. At this moment,
the Euclidean distance of the voltage state estimate fluctuates considerably with the attack
and exceeds the predefined detection threshold. Therefore, FDIAs can be detected, which
triggers the attack alarm system.

7. Conclusions
In this research, we introduce an approach that combines weighted least squares with
an adaptive interpolation extended Kalman filter to detect FDIAs in power systems. AIEKF
effectively reduces the nonlinear errors associated with the extended Kalman filters, leading
to enhanced accuracy in estimating the state of the power system. When a power system is
subject to false data injection attacks, the state estimation weighted least squares statistic
is characterized by a real-time nature, where changes in state variables are instantaneous,
whereas adaptive interpolation extended Kalman filtering is characterized by hysteresis,
and a change in state variables requires a process. Based on the difference between the
two algorithms, the Euclidean distance is introduced as a metric for detecting whether the
system is injected with false data or not. Additionally, the relevant detection threshold is
obtained using Monte Carlo simulation. The experiments show that the method is effective
in detecting false data injection attacks.
Subsequent research will consider the study of the localization of FDIAs and the
development of a new joint estimation algorithm that can simultaneously achieve the
detection and localization of false data injection attacks.

182
Energies 2023, 16, 7203

Author Contributions: Conceptualization, G.Z.; methodology, Y.L. and W.G.; software, G.Z. and
X.G.; validation, J.Z.; formal analysis, X.G.; resources, W.G.; data curation, P.H.; writing—original
draft preparation, G.Z.; writing—review and editing, W.G. and Y.L.; visualization, P.H. and X.G.;
supervision, W.G.; project administration, Y.L.; funding acquisition, W.G. All authors have read and
agreed to the published version of the manuscript.
Funding: This research was supported in part by the National Natural Science Foundation of China
(NSFC) (U21A20146), the Collaborative Innovation Project of Anhui Universities (GXXT-2020-070),
the Open Research Fund of Anhui Province Key Laboratory of Detection Technology and Energy
Saving Devices (JCKJ2022C02, JCKJ2022A10), and the Open Research Fund of the Key Laboratory of
Advanced Perception and Intelligent Control of High-end Equipment of the Ministry of Education
(GDSC202208).
Data Availability Statement: Not applicable.
Acknowledgments: We thank the anonymous reviewers for their valuable comments.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Marashi, K.; Sarvestani, S.S.; Hurson, A.R. Consideration of Cyber-Physical Interdependencies in Reliability Modeling of Smart
Grids. IEEE Trans. Sustain. Comput. 2018, 3, 73–83. [CrossRef]
2. Guo, H.; Pang, Z.H.; Sun, J.; Li, J. An Output-Coding-Based Detection Scheme Against Replay Attacks in Cyber-Physical Systems.
IEEE Trans. Circuits Syst. II 2021, 68, 3306–3310. [CrossRef]
3. Ghosh, S.; Sampalli, S. A Survey of Security in SCADA Networks: Current Issues and Future Challenges. IEEE Access 2019, 7,
135812–135831. [CrossRef]
4. Zhou, J.; Chen, B.; Yu, L. Intermediate-Variable-Based Estimation for FDI Attacks in Cyber-Physical Systems. IEEE Trans. Circuits
Syst. II 2020, 67, 2762–2766. [CrossRef]
5. Gao, Y.; Ma, J.; Wang, J.; Wu, Y. Event-Triggered Adaptive Fixed-Time Secure Control for Nonlinear Cyber-Physical System with
False Data-Injection Attacks. IEEE Trans. Circuits Syst. II 2023, 70, 316–320. [CrossRef]
6. Wang, Y.; Gu, D.; Peng, D.; Chen, S.; Yang, H. Stuxnet Vulnerabilities Analysis of SCADA Systems. Commun. Comput. Inf. Sci.
2012, 345, 640–646. [CrossRef]
7. Liang, G.; Weller, S.R.; Zhao, J.; Luo, F.; Dong, Z.Y. The 2015 Ukraine Blackout: Implications for False Data Injection Attacks. IEEE
Trans. Power Syst. 2016, 32, 3317–3318. [CrossRef]
8. Lu, K.D.; Wu, Z.G. Multi-Objective False Data Injection Attacks of Cyber–Physical Power Systems. IEEE Trans. Circuits Syst. 2022,
69, 3924–3928. [CrossRef]
9. Yu, W.; Bu, X.; Hou, Z. Security Data-Driven Control for Nonlinear Systems Subject to Deception and False Data Injection Attacks.
IEEE Trans. Netw. Sci. Eng. 2022, 9, 2910–2921. [CrossRef]
10. Liu, Y.; Reiter, M.K.; Ning, P. False data injection attacks against state estimation in electric power grids. In Proceedings of the
2009 ACM Conference on Computer and Communications Security (CCS), Chicago, IL, USA, 9–13 November 2009; pp. 1–33.
[CrossRef]
11. Yang, Q.; Yang, J.; Yu, W.; An, D.; Zhang, N.; Zhao, W. On False Data-Injection Attacks against Power System State Estimation:
Modeling and Countermeasures. IEEE Trans. Parallel Distrib. Syst. 2014, 25, 717–729. [CrossRef]
12. He, Y.; Mendis, G.J.; Wei, J. Real-Time Detection of False Data Injection Attacks in Smart Grid: A Deep Learning-Based Intelligent
Mechanism. IEEE Trans. Smart Grid 2017, 8, 2505–2516. [CrossRef]
13. Musleh, A.S.; Chen, G.; Dong, Z.Y. A Survey on the Detection Algorithms for False Data Injection Attacks in Smart Grids. IEEE
Trans. Smart Grid 2020, 11, 2218–2234. [CrossRef]
14. Moslemi, R.; Mesbahi, A.; Velni, J.M. A fast, decentralized covariance selection-based approach to detect cyber attacks in smart
grids. IEEE Trans. Smart Grid 2018, 9, 4930–4941. [CrossRef]
15. Chen, Y.; Huang, S.; Liu, F.; Wang, Z.; Sun, X. Evaluation of Reinforcement Learning-Based False Data Injection Attack to
Automatic Voltage Control. IEEE Trans. Smart Grid 2019, 10, 2158–2169. [CrossRef]
16. Zhao, J.; Zhang, G.; Scala, L.M.; Dong, Z.Y.; Chen, C.; Wang, J. Short-Term State Forecasting-Aided Method for Detection of Smart
Grid General False Data Injection Attacks. IEEE Trans. Smart Grid 2017, 8, 1580–1590. [CrossRef]
17. Li, X.; Wang, Z.; Zhang, C.; Du, D.; Fei, M. A Novel Dynamic Watermarking-Based EKF Detection Method for FDIAs in Smart
Grid. IEEE/CAA. J. Autom. Sinica 2022, 9, 1319–1322. [CrossRef]
18. Manandhar, K.; Cao, X.J.; Hu, F.; Liu, Y. Combating False Data Injection Attacks in Smart Grid using Kalman Filter. In Proceedings of
the 2014 International Conference on Computing, Networking and Communications (ICNC), Honolulu, HI, USA, 3–6 February 2014;
pp. 16–20. [CrossRef]
19. Shi, W.; Wang, Y.; Jin, Q.; Ma, J. PDL: An Efficient Prediction-Based False Data Injection Attack Detection and Location in Smart
Grid. In Proceedings of the 2018 IEEE 42nd Annual Computer Software and Applications Conference (COMPSAC), Tokyo, Japan,
23–27 July 2018; pp. 676–681. [CrossRef]

183
Energies 2023, 16, 7203

20. Abreu, O.A.; Messina, F.; Vega, L.R. Stealth Attacks on the SADI with Prior Information on the State Covariance Matrix. In
Proceedings of the 2022 IEEE Biennial Congress of Argentina (ARGENCON), San Juan, Argentina, 7–9 September 2022; pp. 1–7.
[CrossRef]
21. Huang, K.; Xiang, Z.; Deng, W.; Yang, C.; Wang, Z. False Data Injection Attacks Detection in Smart Grid: A Structural Sparse
Matrix Separation Method. IEEE Trans. Netw. Sci. Eng. 2021, 8, 2545–2558. [CrossRef]
22. Khalaf, M.; Youssef, A.; El-Saadany, E. Joint Detection and Mitigation of False Data Injection Attacks in AGC Systems. IEEE Trans.
Smart Grid 2019, 10, 4985–4995. [CrossRef]
23. Kurt, M.N.; Yilmaz, Y.; Wang, X. Distributed Quickest Detection of Cyber-Attacks in Smart Grid. IEEE Trans. Inf. Forensics Secur.
2018, 13, 2015–2030. [CrossRef]
24. Cheng, G.; Lin, Y.; Zhao, J.; Yan, J. A Highly Discriminative Detector Against False Data Injection Attacks in AC State Estimation.
IEEE Trans. Smart Grid 2022, 13, 2318–2330. [CrossRef]
25. Wang, Y.; Shi, W.; Jin, Q.; Ma, J. An Accurate False Data Detection in Smart Grid Based on Residual Recurrent Neural Network
and Adaptive threshold. In Proceedings of the 2019 IEEE International Conference on Energy Internet (ICEI), Nanjing, China,
27–31 May 2019; pp. 499–504. [CrossRef]
26. Mousavian, S.; Valenzuela, J.; Wang, J.H. Real-time data reassurance in electrical power systems based on artificial neural
networks. Electr. Pow. Syst. Res. 2013, 96, 285–295. [CrossRef]
27. Mahi-al-rashid, A.; Hossain, F.; Anwar, A.; Azam, S. False Data Injection Attack Detection in Smart Grid Using Energy Consump-
tion Forecasting. Energies 2022, 15, 4877. [CrossRef]
28. Yu, J.J.Q.; Hou, Y.; Li, V.O.K. Online False Data Injection Attack Detection With Wavelet Transform and Deep Neural Networks.
IEEE Trans. Industr. Inform. 2018, 14, 3271–3280. [CrossRef]
29. Wang, Y.F.; Zhang, Z.H.; Ma, J.H.; Jin, Q. KFRNN: An Effective False Data Injection Attack Detection in Smart Grid Based on Kalman
Filter and Recurrent Neural Network. IEEE Internet Things J. 2022, 9, 6893–6904. [CrossRef]
30. Jorjani, M.; Seifi, H.; Varjani, A.Y. A Graph Theory-Based Approach to Detect False Data Injection Attacks in Power System AC
State Estimation. IEEE Trans. Ind. Informat. 2021, 17, 2465–2475. [CrossRef]
31. Muscas, C.; Pegoraro, P.A.; Sulis, S.; Pau, M.; Ponci, F.; Monti, A. New Kalman Filter Approach Exploiting Frequency Knowledge
for Accurate PMU-Based Power System State Estimation. IEEE Trans. Instrum. Meas. 2020, 69, 6713–6722. [CrossRef]
32. Deng, R.L.; Zhuang, P.; Liang, H. False Data Injection Attacks Against State Estimation in Power Distribution Systems. IEEE
Trans. Smart Grid 2019, 10, 2871–2881. [CrossRef]
33. Yuan, C.; Zhuo, Y.; Liu, G.; Dai, R.; Lu, Y.; Wang, Z. Graph Computing-Based WLS Fast Decoupled State Estimation. IEEE Trans.
Smart Grid 2020, 11, 2440–2451. [CrossRef]
34. Manousakis, N.M.; Korres, G.N. Application of State Estimation in Distribution Systems with Embedded Microgrids. Energies
2021, 14, 7933. [CrossRef]
35. Radhoush, S.; Vannoy, T.; Liyanage, K.; Whitaker, B.M.; Nehrir, H. Distribution System State Estimation and False Data Injection
Attack Detection with a Multi-Output Deep Neural Network. Energies 2023, 16, 2288. [CrossRef]
36. Ganjkhani, M.; Fallah, S.N.; Badakhshan, S.; Shamshirband, S.; Chau, K.-W. A Novel Detection Algorithm to Identify False Data
Injection Attacks on Power System State Estimation. Energies 2019, 12, 2209. [CrossRef]
37. Akhlaghi, S.; Zhou, N.; Huang, Z. A Multi-Step Adaptive Interpolation Approach to Mitigating the Impact of Nonlinearity on
Dynamic State Estimation. IEEE Trans. Smart Grid 2018, 9, 3102–3111. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

184
 energies
Article
Measuring and Modeling the Skin Effect for Harmonic Power
Flow Studies
Eduardo Tavares Silvério * and Jose Rubens Macedo Junior

Faculty of Electrical Engineering, Federal University of Uberlandia, Uberlandia 38408-100, Brazil; jrubens@ufu.br
* Correspondence: eduardot.silverio@ufu.br

Abstract: This research aims to quantify the skin effect and estimate expressions that well represent
the phenomenon for harmonic power flow studies. The primary focus is to validate the behav-
ior of the skin effect at harmonic frequencies ranging from 60 Hz to 960 Hz, while considering
various amplitudes of electric current. The investigation not only examines the measurement of
the skin effect, but also considers the temperature of the tested conductors, aiming to analyze the
increase in resistance resulting from temperature rise and resistivity changes. The measurement
outcomes demonstrate notable increments in electrical resistance, with resistivity increases of up to
1.9% observed throughout the measurement process. Finally, based on the results obtained through
laboratory measurements, mathematical expressions were estimated as a function of frequency. In or-
der to evaluate the simulation time reduction by the proposed expressions, OpenDSS (version: 9.4.1.2;
Electric Power Research Institute, Knoxville, TN, USA) software was used, which aims at quantifying
the impact of the skin effect on the technical losses. The results from these simulations demonstrate
that the proposed expressions to account for the skin effect in conductors reduce the simulation time
by around 17% for harmonic power flow.

Keywords: measurement; power distribution system; power loss; skin effect

1. Introduction
The current harmonic and interharmonic components, which encompass an insepa-
rable reality of electrical systems, have long been studied and quantified by researchers
Citation: Silvério, E.T.; Macedo and engineers around the world. The presence of these components, particularly those
Junior, J.R. Measuring and Modeling operating at higher frequencies, on any electric circuit, causes an increase in the resistance
the Skin Effect for Harmonic Power of conductors due to the skin effect phenomenon.
Flow Studies. Energies 2023, 16, 7913. Subsequently, scientific publications focused on investigating the skin effect by con-
https://doi.org/10.3390/en16237913 ducting measurements to establish a relationship between the resistance of the conductor
Academic Editor: Ying-Yi Hong
at direct current (RDC) and the resistance of the conductor at alternating current (RAC),
considering variations in frequency, such as [1]. Scientific literature that mathematically
Received: 5 November 2023 covers the phenomenon [2,3] was consolidated by utilizing Bessel’s equation as a faithful
Revised: 25 November 2023 representation of the skin effect, where the particular behavior of the magnetic field for
Accepted: 26 November 2023 different conductors was later clarified, such as in hard wires and flexible cables [4].
Published: 4 December 2023
Furthermore, laboratory experiments have addressed specific issues related to the
measurement of electrical resistance in conductors at different frequencies [5]. These
experiments involved the development of an electrical configuration capable of quanti-
Copyright: © 2023 by the authors.
fying resistance under varying electric currents. In parallel, computer simulations have
Licensee MDPI, Basel, Switzerland. become essential tools for scientific advancements [6]. The use of finite element-based
This article is an open access article simulated meshes has enabled detailed analyses of the electromagnetic performance of
distributed under the terms and various conductors and electrical equipment geometries.
conditions of the Creative Commons Building upon previous research, recent scientific studies have focused on practical
Attribution (CC BY) license (https:// applications of the phenomenon, such as power losses in arc furnace buses [6]. Additionally,
creativecommons.org/licenses/by/ the scientific community has extensively examined the reliability of mathematical solutions
4.0/). provided by computer simulations [7].

Energies 2023, 16, 7913. https://doi.org/10.3390/en16237913 185 https://www.mdpi.com/journal/energies

Energies 2023, 16, 7913

However, a shared characteristic observed in the various studies conducted [6–9] is

the lack of measurements and real-world applicability of the data to analyze the impact
of the skin effect on electric power systems. This knowledge gap highlights the need for
a better understanding of quantifying the skin effect. In line with this observation, the
study conducted in [10] focused on the practical implications of the skin effect on electrical
equipment, such as transformers, resulting in increased power losses.
Similarly, the study in [11] examined the behavior of the skin effect in a 69 kV conven-
tional transmission line with ground return, offering practical insights into the phenomenon,
albeit without fully considering the complex nature of the entire electrical system. De-
spite the innovative mathematical modeling presented, the study primarily addresses
the individual impact of the skin effect on the specific electrical device, neglecting the
comprehensive analysis of electrical power distribution systems as a whole.
While previous studies have explored the behavior of the skin effect in transformers
and transmission lines, there are studies, such as [12,13], that leverage this phenomenon to
develop a control logic for protecting a railway system with direct current rails. Specifically,
this study presents a precise methodology for analyzing the skin effect in the time domain,
accurately reproducing fault currents in the rails. In contrast to earlier research, this study
satisfactorily considers the intricate complexities of the local power distribution system.
In addition to the mentioned papers, [14] also addresses the phenomenon, pointing out
that eddy currents and the skin effect play an important role in electrical machines and
power systems.
The impact of harmonic components on eddy current losses in the core of soft magnetic
materials is discussed in [13]. The study involves the creation of a 3D model based on
the magnetic behavior of the material. Furthermore, the quantification of losses resulting
from the skin effect induced by harmonic components was carried out. Ref. [15] presents
an approach for analyzing the skin effect in a wire using finite elements. Although this
study makes significant contributions to calculating conductor resistance as a function of
frequency, it lacks laboratory exploration and measurement, which is an aspect addressed
in the current study. Ref. [16] presents a singular contribution on cables losses via a
computational method based on a probabilistic approach that predicts some features on
litz cables. Despite the singular contribution that this study makes to quantifying losses on
conductors, it lacks a focus on electrical power system conductors that the current study
aims to fulfill.
The study presented herein has as its objective to present the quantification of resis-
tance increase due to the skin effect in conductors used on power distribution networks. To
undertake this study, it is crucial to have a comprehensive understanding of the physical
and mathematical aspects of the problem, as well as to conduct laboratory measurements.
This allows for the quantification of the skin effect in the electrical resistance of various
types of conductors in power distribution lines. Additionally, the expressions estimated for
representing the skin effect in conductors were applied to the IEEE 8500-node test feeder
utilizing the OpenDSS software, which assesses the impact that the suggested approach
has in simulation time and the influence of the skin effect on power distribution losses. It
is important to point out that not only are the results from these simulations displayed
on item 4.2, but also a detailed approach regarding the electrical characteristics of IEEE
8500-node test feeder. The subsequent section will delve into the physical and mathematical
aspects related to the phenomenon.

2. Theoretical Background
In essence, the skin effect consists of the non-uniformity of the electric current den-
sity in the cross-section of an electrical conductor. In other words, circular conductors,
commonly used in power distribution systems, present a greater flow of electric current
on the surfaces of the conductor than on its inner part, and this effect is amplified with
increases in electrical current. This phenomenon occurs when a current i (t) varying in time
→
runs through the conductor and, as a result, a magnetic field H (t) is generated around

186
Energies 2023, 16, 7913

it according to Ampere’s law. However, in accordance with the law of electromagnetic

induction, any time-varying magnetic field generates an internal electromotive force (EMF)
→
in the conductor that is also time-varying, but in the opposite direction to the H (t) from
which it arises.
As such, a circulating current ie (t) arises due to the potential difference that is created
→
between the innermost part of the conductor and its surface; as such, He (t) is generated by
virtue of Ampère’s law. Figure 1 illustrates the previously mentioned electrical magnitudes.
Additionally, Figure 1 displays a solid conductor exclusively for simplicity and clarity of
the physical conceptualization of the before mentioned phenomenon. All of the conductors
discussed throughout this paper are stranded conductors.

Figure 1. Skin effect on a straight and round conductor under high frequencies.

The law of electromagnetic induction guarantees energy conservation in the flow of

electric charge within a conductor. It ensures that, despite the presence of a time-varying
magnetic field surrounding the conductor, the resulting electric current remains continuous
and does not alter the quantity of charges passing through the conductor’s cross-sectional
area. On the other hand, it modifies the current density along this cross-section, thus
distinguishing the physical process that characterizes the skin effect [17]. The currents that
arise inside the conductor ie (t) are denominated as eddy currents, as shown in Figure 1.
In the skin effect, the electric current density of the conductor is not uniform. The inner
part the resulting current intensity is i (t) − ie (t), while closer to the surface, the current is
i (t) + ie (t), and this is the main cause of the non-uniform distribution of the current density
in its cross-section.

Mathematical Approach
To mathematically contemplate the skin effect, it is necessary to address the behavior
of the magnetic field generated by the electric current that runs through the conductor. In
this sense, the magnetic field is time-varying and can be expressed by (1), just as the electric
field can be represented by (2).
→  →
H (z, t) = Hy (z) Re e jωt ay (1)

→  →
E (z, t) = Ex (z) Re e jωt a x (2)

187
Energies 2023, 16, 7913

Therefore, by applying (1) and (2) to Ampere’s law and the law of electromagnetic
induction, respectively, while considering that the electric field also varies in time and is
→
perpendicular to the direction of H (z, t), one arrives at (3) and (4).
→ ∂Ex (z) → →
∇× E = ay = − jωμHy (z) ay (3)
∂z

∂Hy →
→ →
∇×H = − a x = jωεEx (z) a x (4)
∂z
Under the intent of obtaining a differential equation that relates only to the magnetic
field, expression (4) is rotated.
   
→ →
∇ × ∇ × H = jωε∇ × Ex (z) a x (5)

Substituting the Laplacian vector Equation (6) into (5), one arrives at (7).
   
→ → →
∇2 H = ∇ ∇· H − ∇ × ∇ × H (6)

 
→ → →
∇ ∇· H − ∇2 H = jωε∇ × E (7)

→
As H (z, t) is a vector field that has a zero divergence, it is possible to rewrite (7)
according to Gauss’ law, considering the calculation of the magnetic field gradient, and
replacing (3) in (7); one arrives at (8).

∂2 Hy (z)
= −ω 2 μεHy (z) (8)
∂z2
Since expression (8) is the Helmholtz equation in phasor form [18], one can specify the
Helmholtz equation as (9) and define the propagation constant of the magnetic field (10)
utilizing a conductive material.

∂2 Hy (z)
∂z2
= −λ2 = −ω 2 με (9)
Hy (z)
) 
√ σ 
λ = ω με = ω με 1 − j (10)
ωε
σ
Noteworthy is that for conducting materials ωε  1 as such, (10) is summarized as
the following: *
λ = (1 + j) π f μσ = α + jβ (11)
In this sense, the solution to the Helmholtz equation presents the behavior of the
magnetic field. It enhances the context to clarify the physical significance of the prop-
agation constant for the skin effect. The solution to the Helmholtz equation is given
by the following:
→ →
H (z, t) = H0 e−α e j(ωt− βz) ay (12)
√
In (12), the term e j(ωt− π f μσz)
represents the oscillatory
√ feature of the magnetic field,
where β proposes the phase shift. However, the term e− π f μσz presents exponential decay
α of the original magnitude of the vector field (H0 ) as the wave travels in the direction +z.
Through these equations, it is possible to define the skin effect by means of skin depth.
The skin depth δ corresponds to the distance traveled by the magnetic field in the material

188
Energies 2023, 16, 7913

medium in which its original magnitude is decreased in the order of e−1 . In other words,
skin depth elucidates the degree of penetration of the magnetic field and of the circulating
internal currents in a conductor with losses [18], as*shown in Figure 2. Therefore, for the
condition presented in (12), z should be equal to 1/ π f μσ for attenuation on the order of
e−1 , where the skin depth is defined as the following:

1
δ= * (13)
π f μσ

Figure 2. Setup for voltage probe connection.

Hence, the skin effect can be mathematically defined as the skin depth in a conductor,
as represented by (13), and it physically signifies a higher concentration of current in the
cross-section of the conductor, as depicted in Figure 2. Therefore, it is crucial to investigate
how the current density varies based on the conductor’s geometry. With this in mind,
significant emphasis is placed on studying the current distribution of various types of
conductors. As a result, in (14), two different forms are used to calculate the current
through a conductor, such as the surface integral of J and Ampère’s law.

5r
Hr = Jrdr (14)
0

To understand how the geometry of the conductor influences the current density of
the conductor, one derives (14) concerning the radius of the conductor. Furthermore, it is
also pertinent to derive (14) in time, since the magnetic field is a function of this variable,
resulting in the following:
dJ d2 H 1 dH
= + . (15)
dt drdt r dt
The result of (15) provides the variation of the current density in the conductor for
these two variables, space (r) and time (t), according to the variations of the magnetic
field. In this way, the following developments will be directed to finding mathematical
expressions for the derivatives of the magnetic field in time and space in simpler forms for
→ → →
the result of (15). Noteworthy still is that E = ρ J , thus μdr dH
dt = ρ J , where one arrives at
(16) and substitutes (16) into (15), one obtains (17), which is the differential equation that
models the phenomenon presented in Figure 2, in a way that specifies the density of the
electric current along the cross-section of the conductor.
→
dH ρdJ
= (16)
dt μ dr

189
Energies 2023, 16, 7913

d2 J 1 dJ μ dJ
+ − =0 (17)
dr2 r dr ρ dt
Moreover, in the phasor form, where d/dt → jω and k2 = jωμσ, one has the following:

d2 J dJ
r2 + r − r2 k2 J = 0. (18)
dr2 dr

The differential equation shown in (18) has the term r2 k2 arranged to facilitate the math-
ematical development to reach an elegant solution. The expression (19) is the electric current
density as a function of the radius of the conductor under analysis, expressed by Bessel’s
equations with 0 and 1 indexes (I0 and I1 ) [18,19]. The differential equation shown in (18) is
a function of r0 , which is the conductor radius, and r represent the variable, which deter-
mines the intensity of the electric current in the cross-section of the conductor, as shown in
(19). It is important to note that on (18) the term r2 k2 , during the mathematical development
to reach a solution, turns into kr within the Bessel’s expressions of indexes 0 and 1.

Ik I0 (kr )
J (r ) = (19)
2πr0 I1 (kr0 )

Considering that the current density, when a direct current runs through the conductor
is JDC = I/πr02 , then J (r )/JDC is given by the following:
  r0  r 
  
J (r ) (1 + j) rδ0 I0 (1 + j) δ r0
=    (20)
JDC 2 I1 (1 + j) rδ0

Notably, the current density on the outer part of the conductor is greater in conductors
with larger cross-sections, thus underscoring the impact of the skin effect [18,19]; this is
evident from (20) as well. Therefore, the electrical resistance of conductors under such
conditions increases, which implies an increase in power losses due to the transport of
active power. Consequently, to evaluate the impact of such a phenomenon, the proposal
is put forward for the study of the behavior of R AC/RDC , in similar fashion to the approach
already used for current density ( JAC/JDC ).
When considering the resistance in alternating current ( R AC ), one uses Ohm’s law, as
seen in (21), with the ratio being R AC/RDC as presented in (22) [19,20].

ρl k I0 (kr0 )
Z= (21)
πr0 2 I1 (kr0 )
4
R AC J (r0 )
= Re (22)
R DC JDC
It is observed in (22) that as the ratio rδ0 increases, the electrical resistance of the
conductor increases proportionally for a conductor with radius r0 and with a skin depth
that decreases (increasing the ratio rδ0 ). A trivial condition concerning the increase in ratio
r0
δ occurs when harmonic components of electric current flow through the conductor. Under
such conditions, harmonic components imply a decrease in the value of δ, in accordance
with (13). Indirectly, this causes an increase in the resistance of the conductor, as it increases
the ratio rδ0 [18,21].
It is well established that harmonic components are inherent in electrical power
systems, particularly in power distribution systems. In such systems, it is common to
encounter harmonic currents (and correspondingly, voltages) with frequencies up to 960 Hz
(16th harmonic order). Beyond this frequency range, the amplitudes of these components
become insignificant, exerting minimal impact on the skin effect.

190
Energies 2023, 16, 7913

In the specific case of electrical power distribution systems, the skin effect assumes
a very applicable role in harmonic power flow studies, especially on medium- and low-
voltage power networks. In this sense, the physical–mathematical development carried out
so far aimed at establishing physical foundations for understanding the phenomenon,
concerning justifying any sensible increase in resistance of the conductor under non-
sinusoidal conditions, as well as justifying the investigation into the quantification of this
increase. From this standpoint, the objective of this study is to assess the electrical resistance
of various conductors employed in power distribution systems, in order to quantify the
rise in resistance attributed to the aforementioned phenomenon. The subsequent section
outlines the methodology employed to measure the electrical resistance of conductors
affected by the skin effect.

3. Measurement Methodology
In pursuance of measuring the impact of the skin effect on conductors, it is necessary
to understand which electrical magnitudes are verified in the measurement methodology.
Given the relatively low magnitude of the measured quantities, it is crucial to consider
the uncertainties associated with the measurements. To determine the resistance of the
conductor, an indirect measurement approach was employed. This involved passing an
electric current through the conductor under examination and using an oscilloscope to
measure the corresponding voltage drop. By applying Ohm’s Law, the resistance of the
conductor can be calculated using Equation (23).
' . 6  
1 V measured Ω
R= Re . × 1000 (23)
laverage I km

However, to obtain a reliable measurement of the voltage drop in the conductor, an

assessment is made into the possible interferences admitted by the voltage loop. That
said, voltmeters, when measuring the potential difference between two points a and b, are
known to consider the voltage arising from the conservative electric field between these
two points. This is in addition to the voltage from the non-conservative magnetic field that
varies in time and passes through the loop voltage, where (24) is the potential difference
measured by the oscilloscope, which includes both previously mentioned vector fields.

5b → → 5d 5b → →
∂
Vmeasured ≈ E ·d L + B ext ·d S
∂t
a 0 a

5b 5d 5b
∂
Vmeasured ≈ E dz + B dzdy (24)
∂t
a 0 a

In this context, the measurement methodology used assumes two origins for the poten-
→
tial difference ascertained in the oscilloscope. The first comes from E that is independent
of the integration path inside the conductor, as well as from the voltage induced by the
→
variation of B ext in time bound to the voltage loop, as shown in Figure 2. Additionally,
Figure 2 displays a solid conductor exclusively for simplicity and clarity of the physical con-
ceptualization of the before mentioned measurement methodology. All of the conductors
measured in this study are stranded conductors.
→
From Figure 2, it can be noted that B ext passes through the voltage loop S, where an
induced voltage appears between the conductor points a and b, which is measured with
the oscilloscope, and constitutes an effective value of the voltage Vmeasured , as shown in (24).
It is important to point out that the induced voltage in question greatly impacts the mea-
7b→ →
surement, due to the low order of magnitude of the voltage arising from a E ·d L between
points a and b. In this case, the smallest possible voltage loop is used to minimize the impact

191
Energies 2023, 16, 7913

→
of the voltage induced by B ext . In accordance with [16], the suggestion is that the height
of the voltage loop equals the conductor diameter. More objectively, dvoltmeter = dconductor
while considering the setup in Figure 2.
In addition to taking electromagnetic factors into account when measuring the elec-
trical resistance of conductors, it is important to consider the presence of nearby con-
ductive materials. These materials can induce eddy currents through the proximity ef-
fect, potentially interfering with the measurement procedure. To mitigate this issue, all
conductive materials were positioned at a distance of more than one meter from the
measurement setup [22,23].
In this manner, Figure 3 illustrates the measurement setup used to measure the re-
sistance of the conductors. Figure 3 shows the programmable source, CMC 256 plus,
manufactured by Omicron Electronics Corp. (Houston, TX, USA), used to operate as a cur-
rent source. In addition, there is a digital oscilloscope, RTH1004, manufactured by Rohde &
Schwarz (Teisnach, Germany), with four isolated channels and a 5 GSa/s sampling rate.

Figure 3. Measurement setup to quantify the skin effect in conductors applied to distribution
power systems.

In order to fully describe the measurement setup, the main characteristics are stated
as follows:
• The ground is composed of concrete.
• The current return path is 1 m from the conductor under test, close to ground level.
• Wooden racks were used as a non-conductive material to keep the conductor under
testing 1 m away from ground level.
• Voltage differential probes cables were placed according to dvoltmeter = dconductor for
each conductor, as displayed in Figures 2 and 3.
The sensors used for the acquisition of electrical quantities were the A622, manufac-
tured by Tektronix (Portland, OR, USA), employed for the acquisition of electric current,
and the TA041, manufactured by PicoTech (Cambridge, UK), for the acquisition of the
voltage signal. The configuration displayed in Figure 3 shows the electronic devices used,
as well as the three crucial items for reaching the objective of this study.
Additionally, the post-processing voltage signal registered on the oscilloscope corre-
sponds to the flow chart presented in Figure 4. Briefly, the oscilloscope registered a time
domain signal that was applied to a Python algorithm developed exclusively to perform

192
Energies 2023, 16, 7913

Fourier transforms. Its purpose is to obtain the voltage magnitude and phase angle of the
desired frequency, which is the current frequency flowing through the conductor under test.
It is important to highlight that the reference phase angle (0◦ electrical degrees) is related to
the current signal, once the programmable power source is a current source. Additionally,
the magnitude of the out-of-phase inductive voltage signal is larger than the resistive
voltage signal, especially regarding high frequencies. It is an expected response, since
the inductive part of the conductor responds to a linear relation to frequency; meanwhile,
the resistive part responds according to (22) in relation to frequency. In other words, the
measured voltage signals presented a phase angle higher than 80◦ , indicating that voltage
is predominantly inductive. The considered resistive voltage signal was computed (see
Figure 4) and applied to (23), obtaining the conductor’s electrical resistance.

Figure 4. Data measurement post-processing to compute conductor’s electrical resistance.

Firstly, the conductor under test in this study includes the main conductors used in
electrical power distribution systems, which are 2 AWG—AAC, 1/0 AWG—ACSR, 2/0
AWG—AAC, 4/0 AWG—AAC, 150 mm2 protected and 336.4 MCM—ACSR. Here, it is
noteworthy that for better contact between the power source conductors (red thicker lines,
Figure 3) and the conductor under test, it is necessary to use terminals both for current
injection in the conductor under test, as well as for measuring the voltage.
As shown in Figure 3, the aluminum terminal ensures that the electrical current is
distributed as expected. To this end, its geometric dimensions are shown in Figure 5, where
D represents the diameter of the conductor under test.
To ensure that the current distribution is homogeneous in the region where the voltage
in the conductor is measured, a distance of 1 m from the input to the output terminals of
→
the electric current was assumed. This implies that J is homogeneous in the cross-section
of the conductor, guaranteeing an adequate measurement [23].
Finally, there are split bolt terminals for connecting the oscilloscope voltage channels,
allowing for a better connection, as shown in Figure 6.

193
Energies 2023, 16, 7913

Figure 5. Custom aluminum terminals, considering (a) frontal and lateral views, (b) 3D view and
(c) actual terminal connection.

Figure 6. (a) Split bolt terminal and (b) actual terminal connection.

In addition to the previous recommendation regarding the voltage loop during the
measurement of conductor electrical resistance, reference [23] also mentions the potential
influence of temperature on the accurate quantification of resistances considering the
skin effect.
This aspect is of great importance, since the electric current heats up the conductor
and, consequently, increases the resistivity (ρ) of the conductor. Additionally, the resistance
of the conductor under test alters owing to its shape, with the coefficients of linear and
superficial expansion of the conductor being physical aspects to be investigated due to an
increase in temperature along the measurement process. In this sense, the coefficients of
linear expansion of aluminum and steel are of the order of 10−6 m/◦ C and 10−5 m/◦ C,
respectively [24]. As such, errors associated with linear expansion are negligible, with the
contribution arising from linear or superficial expansion being irrelevant to the order of
magnitude of the electrical resistance. Therefore, in order to assess the impact of tempera-
ture on the resistivity of the conductor, a device was developed for the acquisition of the
temperature using an Arduino UNO and temperature sensor DS18B20, in conjunction with
the acquisition of 12-bit data and a resolution from the measurement of 0.0625 ◦ C, as shown
in Figure 3. To commence the measurement process, all devices involved were certified to

194
Energies 2023, 16, 7913

be at 20 ◦ C. Thermally conductive aluminum adhesive tape was used in order to guarantee

proper heat transfer from the conductor to the sensor. Consequently, the thermal analysis
aims to determine the error linked to the skin effect at each frequency, as presented in
Table 1 and consolidated in graph form in Figure 7.
 
Table 1. Temperature variation ΔT = T f − 20 ◦ C and resistivity (ρ) increase considering current
(40 A) flowing through the conductor at several frequencies.

1/0 336.4
Harm- 2 AWG—AAC 2/0 AWG—AAC 4/0 AWG—AAC 150 mm2 —AAC
AWG—ASCR MCM—ACSR
onic ρ ρ ρ ρ ρ ρ
ΔT (◦ C) ρ20◦ C (%) ΔT (◦ C) ρ20◦ C (%) ΔT (◦ C) ρ20◦ C (%) ΔT (◦ C) ρ20◦ C (%) ΔT (◦ C) ρ20◦ C (%) ΔT (◦ C) ρ20◦ C (%)

DC 4.406 1.775 4.527 1.824 3.473 1.399 2.602 1.048 1.724 0.694 1.395 0.562
1 4.410 1.777 4.535 1.827 3.499 1.410 2.625 1.058 1.757 0.708 1.402 0.565
2 4.447 1.792 4.572 1.843 3.510 1.414 2.636 1.062 1.798 0.724 1.432 0.577
3 4.470 1.801 4.595 1.852 3.516 1.417 2.642 1.065 1.835 0.739 1.462 0.589
4 4.485 1.808 4.610 1.858 3.521 1.419 2.647 1.067 1.868 0.753 1.493 0.602
5 4.497 1.812 4.622 1.863 3.524 1.420 2.650 1.068 1.899 0.765 1.523 0.614
6 4.507 1.816 4.632 1.867 3.527 1.422 2.653 1.069 1.927 0.777 1.553 0.626
7 4.516 1.820 4.641 1.870 3.530 1.423 2.656 1.070 1.953 0.787 1.583 0.638
8 4.523 1.823 4.648 1.873 3.532 1.423 2.658 1.071 1.978 0.797 1.613 0.650
9 4.530 1.825 4.655 1.876 3.534 1.424 2.660 1.072 2.002 0.807 1.643 0.662
10 4.535 1.828 4.660 1.878 3.536 1.425 2.662 1.073 2.025 0.816 1.673 0.674
11 4.540 1.830 4.665 1.880 3.537 1.425 2.663 1.073 2.049 0.826 1.704 0.687
12 4.545 1.832 4.670 1.882 3.539 1.426 2.665 1.074 2.073 0.836 1.734 0.699
13 4.550 1.833 4.675 1.884 3.540 1.427 2.666 1.074 2.099 0.846 1.764 0.711
14 4.554 1.835 4.679 1.885 3.541 1.427 2.667 1.075 2.126 0.857 1.794 0.723
15 4.557 1.837 4.682 1.887 3.542 1.427 2.668 1.075 2.156 0.869 1.824 0.735
16 4.561 1.838 4.686 1.888 3.543 1.428 2.669 1.076 2.188 0.882 1.854 0.747

 
Figure 7. Temperature variation ΔT = T f − 20 ◦ C considering current (40 A) flowing through the
conductor at several frequencies.

To examine the impact of temperature on the resistivity of the conductor (ρ), the linear
relationship between these two quantities was adopted as in reference [24]. Furthermore,
reference [24] highlights that at lower temperatures, the correlation between temperature
and electrical resistivity in aluminum alloys remains linear. Similarly, reference [25] also
addresses this aspect. In the cases examined within this study, the maximum temperature
increase was estimated to be approximately 5 ◦ C, starting from an initial temperature of
20 ◦ C. Additionally, it was observed that for thicker conductors, a temperature rise was

195
Energies 2023, 16, 7913

noted as current frequency increased, as Figure 7 shows for 150 mm2 and 336.4 MCM
conductors. Meanwhile, the remaining conductors displayed a notable lower temperature
increase with current frequency, which suggests that the skin effect results in significant
heating for thicker conductors, considering current frequency increases.
It should be emphasized that the increase in temperature observed for harmonic com-
ponents, compared to the temperature rise for DC current flowing through the conductor,
is primarily caused by the skin effect. The recorded data regarding temperature rise for
harmonic components are presented in Table 2.

Table 2. Temperature rise due exclusively to skin effect.

ρ
Conductor ΔT=T16harmonic −TDC (◦ C) ρDC (%)

2 AWG—AAC 0.155 0.062

1/0 AWG—ASCR 0.159 0.064
2/0 AWG—AAC 0.070 0.028
4/0 AWG—AAC 0.067 0.027
150 mm2 —AAC 0.464 0.187
336.4 MCM—ASCR 0.429 0.172

To specifically quantify the error resulting from temperature rise associated with the
skin effect, Table 2 indicates that the largest error is observed for 150 mm2 —the AAC
conductor, with an error of 0.187% corresponding to a temperature rise of 0.464 ◦ C. Another
noteworthy aspect is that the skin effect is more prominent for thicker conductors, and
Table 2 as well as Figure 7 showcase this result, as expected.
Therefore, based on the analysis of aluminum alloys used in the examined conductors,
it is concluded that the relationship between temperature and electrical resistivity can be
considered linear [23]. Furthermore, it is appropriate to model the resistivity by means of
ρ
(25) to specify the percentage increment for resistivity ρ ◦ × 100%.
20 C

 
Ω
ρ T = ρ Al (20◦ C) [1 + α( T − T0 )] (25)
m

Considering that α = 0.00403 (1/◦ C) [13] and ρ Al (20◦ C) = 61% I ACS (Ω·m), as indicated
by the manufacturer of the conductors [24], one can quantify how much the resistance of
the conductors is influenced by the increase in resistivity due to temperature rise.
The results indicate that the temperature increase observed during the process of
measuring conductor resistance, in the worst-case scenario (conductor 1/0 AWG—ACSR),
leads to an error of 1.888% in the quantification of the resistance value. As expected, the
1/0 AWG—ACSR conductor has lower ampacity, and therefore experiences greater power
dissipation through Joule’s effect. Additionally, its geometry includes steel reinforcement,
which has a resistivity 20 times greater than typical aluminum alloys [24], resulting in
greater heating during the measurement process.
To summarize the measured data, Table 1 shows an increase in electrical resistivity
due to temperature rise. For DC, all of the conductors exhibited a maximum resistivity
increase of 1.9%. This percentage remained consistent for both the fundamental frequency
and all harmonic components.
Furthermore, Table 2 estimates an approximate 0.2% increase in electrical resistivity
for the skin effect measurement due to temperature rise. However, this heating effect
has a negligible impact on the conductor’s resistance measurement. In other words, the
increase in electrical resistivity caused by the RMS values of the fundamental and harmonic
components contributes to a 1.9% rise in resistivity, while the skin effect causes only a 0.2%
increase, which is considered practically negligible.

196
Energies 2023, 16, 7913

Having discussed the measurement configuration for the skin effect and the variables
that can introduce errors, the subsequent section will delve into the analysis of the measured
data and their application to reduce simulation time for harmonic power flow studies.

4. Results and Discussion

4.1. Measurement Results—Skin Effect
Through performing tests, an increase in the resistance of the studied conductors was
obtained, which considered the skin effect. Noteworthy here is that verification was sought
for the DC resistances values of the conductors made available by the manufacturers [26,27]
and as such, was confirmed by the measurement methodology proposed in this study.
Therefore, the measured reference value was used to specify the start of the measurement
process. From this value, Table 3 presents the measured resistance values in DC resistances.

Table 3. Electrical resistances of conductors in direct current at 20 ◦ C.

Rmeasured Rdatasheet Error

Conductor
(Ω/km) (Ω/km) (%)
2 AWG—AAC 0.8652 0.8541 1.31
1/0 AWG—ASCR 0.5285 0.5360 1.40
2/0 AWG—AAC 0.4517 0.4550 0.72
4/0 AWG—AAC 0.2616 0.2676 2.21
150 mm2 —AAC 0.2046 0.2060 0.64
336.4 MCM—ASCR 0.1726 0.1703 1.36

The carried-out measurements showed errors of less than 2.5%, thus ensuring that the
measurement process is adequate, these being the reference values used in Figure 8. Based
on the measurement results, the theory is corroborated, in which thicker conductors are
subject to a greater influence of the skin effect, as illustrated in Figure 8 for the 4/0 AWG,
150 mm2 and 336.4 MCM conductors. In these conductors, a greater increase in electrical
resistance was noted, which results from the skin effect. Furthermore, one notes that the 1/0
AWG conductor, although belonging to the conductors being tested with lower thickness,
exhibited a relevant impact against the skin effect. In comparison, the 2 AWG conductor
together with the 2/0 AWG, composed only of aluminum wires (AAC), exhibited the lowest
influence from the skin effect. From Figure 8, verification was made that the conductors
that presented the greatest heating, such as 2 AWG and 1/0 AWG, stand out for showing
an increasing tendency as the electrical current increases, indicating, through progressive
3D line graphs, the impact of temperature on the measurement process. Meanwhile, this
behavior is not observed in the remainder of the conductors, since the range of test currents
(20–40 A) is much lower than the ampacity of the conductors, highlighting, therefore,
significantly lower heating of these conductors.
The comparison between the measurement results and the theoretical expectations
is significant, as it provides insights into the behavior of the measured data, as displayed
in Figure 9. Additionally, the magnitude of the resistance increase is an important aspect
to consider. It was observed that for electric currents at 960 Hz, running through the
conductor, an increase of around 60% was verified in the electrical resistanc, in comparison
to DC resistance for the 1/0 AWG, 4/0 AWG, 150 mm2 and 336.4 MCM conductors.
Against this backdrop, the following topics aim to quantitatively assess the impact
that the skin effect has on power simulation studies, while considering distorted conditions
of electric current and the measurement values obtained at this stage of the study.

197
Energies 2023, 16, 7913

Figure 8. Measured electrical resistances of conductors, considering the skin effect.

198
Energies 2023, 16, 7913

Figure 9. Measured data after (26) is applied versus simulated results of conductor’s electrical
resistance, considering the skin effect.

4.2. Statistical Simplified Method to Estimate Skin Effect for Low Frequencies
To simplify the process and obtain a single resistance value per frequency, an averaging
method was employed considering the variation in current in the conductor. For each
frequency, the resistance values were synthesized to be considered, as pointed out in (26).

Rh = average( R20A , R22A , · · · , R40A ) (26)

Based on this compilation, Figure 10 provides the values that were computed as a
result of applying expression (26) to the data. Figure 10 shows that the conductors under
test are appropriately estimated through a linear equation considering low frequencies. Not
only are the expressions presented in Figure 10 simpler, but they also reduce the processing
time when computing the skin effect. It should be emphasized that for harmonic power
flow studies, especially those covering the distribution power network of an entire operator
region, they may take up to 48 h of simulation time. To compare the time processing
between Bessel’s proposed expression (22) adapted to 60 Hz resistance as a reference
against the Figure 10 expressions, an algorithm in Python (PyCharm) was implemented.
The results show that Bessel’s expression (22) required 53.85 ms to compute the skin effect
for all conductors for the before mentioned frequency range, as opposed to 41.88 ms of
simulation time for the Figure 10 expressions. Therefore, the reduction in simulation time
is approximately 20% using the proposed method.
For the purposes of an objective assessment of the study, the IEEE 8500 node feeder [28],
which is known throughout the scientific community, was used to calculate the power
losses under distorted conditions. The studied feeder is shown in Figure 11, in which the
line thickness for the segments corresponds to the regions where energy losses are more
expressive. It is important to point out that information with greater detail on this feeder is
available in [28].

199
Energies 2023, 16, 7913

Figure 10. Sampled data after (26) is applied with linear regression expressions and R2 values.

Figure 11. IEEE 8500 nodes feeder—line thicknesses relate to power loss intensities.

Simulations were performed using OpenDSS and COM Interface with Python (Py-
Charm). To clearly meet the objective of this study, it was crucial to define the harmonic
emissivity spectrum of the non-linear loads of this feeder. To this end, the use of input data
from [29] was established for the frequency range of 60 Hz to 960 Hz. Another noteworthy
aspect to mention is that additional conductors beyond the ones measured are present
at the IEEE 8500 node feeder. To address this, an assumption was made to consider the
behavior of the closest laboratory-tested conductor for these untested conductors.
To correctly estimate the time reduction for the harmonic power flow study, the two
following separated conditions were performed:
Case 1: Bessel’s expression (22),
 adapted
 to 60 Hz resistance as a reference, as a
f R
correction factor for the skin effect R60Hz applied to the resistance of line parameters,
according to cables used at the transmission line, as expressed in (27).  
f R
Case 2: Figure 10 expressions as a correction factor for the skin effect applied R60Hz
to the resistance of line parameters, according to the cables used at the transmission line, as
expressed in (27).
Rf
Rconductor f = R [Ω] (27)
R60Hz conductor
The simulation times of both conditions previously mentioned are shown in Table 4.

200
Energies 2023, 16, 7913

Table 4. Simulation times spent for harmonic power flow study case.

Case 1 Case 2 Time Reduction Percentage Reduction

279.64 s 231.25 s 48.39 s 17.30%

Another noteworthy aspect is the reliability of the results provided from the proposed
method compared to Bessel’s expressions. Also, from a power systems perspective, power
losses are segregated into several components, as illustrated in Figure 12. Skin effect losses
are responsible for 4.00% of additional losses on the IEEE 8500 node test feeder.

Figure 12. Power losses computed for IEEE 8500 nodes, considering distorted conditions and the
skin effect.

5. Conclusions
The research conducted in this study offers a new perspective on the impact of the
skin effect on conductors used in energy distribution networks, particularly in relation to
Joule losses associated with this phenomenon. The study aimed to provide a mathematical
understanding of the non-uniform current distribution in the cross-section of the conductor,
thereby explaining its physical origin. A notable aspect of this research is the methodology
developed for measuring the skin effect, which was applied to six commonly used con-
ductors in energy distribution networks. The evaluation criteria used for measuring the
electrical resistance of the conductors in direct current showed an error of less than 2.5%
compared to the manufacturer’s specified values.
Furthermore, the study examined the behavior of conductor resistance under harmonic
electric currents ranging from the 2nd to the 16th harmonic order, in order to assess the
significance of the skin effect. It was observed that the impact of temperature rise on
resistivity during the measurement process was 1.9% due to the heating effect caused by
the current. Conversely, the temperature rise caused by the skin effect was less than 0.2%,
which is considered negligible when it comes to resistance measurements.
After measuring the resistance of the conductors under each magnitude (20–40 A)
and frequency (60–960 Hz), the IEEE 8500 node feeder was used to perform a harmonic
power flow study. In this regard, it was noted that the proposed expressions provided a
reduction in simulation time of approximately 17%, and the skin effect was responsible for
a percentage of 4.00% in additional losses.
Hence, the studies conducted in this research aim to provide a practical perspective
on a well-known phenomenon within the scientific community, specifically addressing
the proposed expressions as a method to reduce simulation times for harmonic power
flow studies.

Author Contributions: Review and editing, E.T.S. and J.R.M.J.; conceptualization, E.T.S. and J.R.M.J.;
writing—original draft, E.T.S.; simulations, E.T.S.; methodology, E.T.S. and J.R.M.J.; formal analysis,
E.T.S. and J.R.M.J. All authors have read and agreed to the published version of the manuscript.
Funding: The research is supported by the Coordination of Superior Level Staff Improvement and
National Electrical Energy Agency under funding number 23117.060101/2020-80.

201
Energies 2023, 16, 7913

Data Availability Statement: All data are displayed in graphs and figures in the study.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Cabral, S.H.L.; Bertoli, S.L.; Medeiros, A.; Hillesheim, C.R.; De Souza, C.K.; Stefenon, S.F.; Nied, A.; Leithardt, V.R.Q.; Gonzalez,
G.V. Practical Aspects of the Skin Effect in Low Frequencies in Rectangular Conductors. IEEE Access 2021, 9, 49424–49433.
[CrossRef]
2. Gassab, O.; Chen, Y.; Shao, Y.; Li, J.; Wen, D.-E.; He, F.; Su, Z.; Zhong, P.; Wang, J.; Zhao, D.; et al. Accurate Formulation of The
Skin and Proximity Effects in High-Speed Cable System. IEEE Access 2022, 10, 100682–100699. [CrossRef]
3. Davis, E.W.; Dana, A.S.; Dwight, H.B. Skin Effect and Proximity Effect in Tubular Conductors. Am. Inst. Electr. Eng. 1922, 41,
1131–1140. [CrossRef]
4. Dai, D.; Zhang, X.; Wang, J. Calculation of AC Resistance for Stranded Single-Core Power Cable Conductors. IEEE Trans. Magn.
2014, 50, 6301104. [CrossRef]
5. Yin, W.; Tang, J.; Lu, M.; Xu, H.; Huang, R.; Zhao, Q.; Zhang, Z.; Peyton, A. An Equivalent-Effect Phenomenon in Eddy Current
Non-Destructive Testing of Thin Structures. IEEE Access 2019, 7, 70296–70307. [CrossRef]
6. Li, S.; Han, Y.; Liu, C. Coupled Multiphysics Field Analysis of High-Current Irregular-Shaped Busbar. IEEE Trans. Compon. Packag.
Manuf. Technol. 2019, 9, 1805–1814. [CrossRef]
7. Kusiak, D. The magnetic field and impedances in three-phase rectangular busbars with a finite length. Energies 2019, 12, 1419.
[CrossRef]
8. Cruciani, S.; Campi, T.; Maradei, F.; Feliziani, M. Finite-Element Modeling of Conductive Multilayer Shields by Artificial Material
Single-Layer Method. IEEE Trans. Magn. 2020, 56, 7502504. [CrossRef]
9. Brito, A.I.; Machado, V.M.; Almeida, M.; das Neves, M.G. Skin and proximity effects in the series-impedance of three-phase
underground cables. Electr. Power Syst. Res. 2016, 131, 30–37. [CrossRef]
10. Lopera, J.M.; Prieto, M.J.; Díaz, J.; Jorge, G. A Mathematical Expression to Determine Copper Losses in Switching-Mode Power
Supplies Transformers Including Geometry and Frequency Effects. IEEE Trans. Power Electron. 2015, 30, 5670–5681. [CrossRef]
11. Monteiro, J.H.A.; Costa, E.C.M.; Pinto, A.J.G.; Kurokawa, S.; Gatous, O.M.O.; Pizzolato, J. Simplified skin-effect formulation for
power transmission lines. IET Sci. Meas. Technol. 2014, 8, 76–83. [CrossRef]
12. He, J.; Yu, L.; Wang, X.; Song, X. Simulation of Transient Skin Effect of DC Railway System Based on MATLAB/Simulink. IEEE
Trans. Power Deliv. 2012, 27, 848–855. [CrossRef]
13. Li, J.; Yang, Q.; Li, Y.; Zhang, C.; Qu, B. Measurement and Modeling of 3-D Rotating Anomalous Loss Considering Harmonics
and Skin Effect of Soft Magnetic Materials. IEEE Trans. Magn. 2017, 53, 6100404. [CrossRef]
14. Guo, W.; Li, D.; Ye, L. A Model of Magnetic Field and Braking Torque in Liquid-Cooled Permanent-Magnet Retarder Accounting
for the Skin Effect on Permeability. IEEE Trans. Veh. Technol. 2019, 68, 10618–10626. [CrossRef]
15. González-Teodoro, J.R.; Kindl, V.; Romero-Cadaval, E.; Asensi, R. Analysis of Skin Effect in Single Wire Resistance by Finite
Element Methods. In Proceedings of the 2020 IEEE 14th International Conference on Compatibility, Power Electronics and Power
Engineering (CPE-POWERENG), Setubal, Portugal, 8–10 July 2020; pp. 1–6. [CrossRef]
16. Gyimóthy, S.; Kaya, S.; Obara, D.; Shimada, M.; Masuda, M.; Bilicz, S.; Pavo, J.; Varga, G. Loss Computation Method for Litz
Cables with Emphasis on Bundle-Level Skin Effect. IEEE Trans. Magn. 2019, 55, 6300304. [CrossRef]
17. Sim, J.-H.; Kim, D.-Y.; Kim, S.-I.; Hong, J.-P. Analytical Electromagnetic Modeling and Experimental Validation of Vehicle Horn
Considering Skin Effect in Its Solid Cores. IEEE Trans. Magn. 2017, 53, 8001704. [CrossRef]
18. Kazimierczuk, M.K. High-Frequency Magnetic Components; John Wiley & Sons: Hoboken, NJ, USA, 2014; ISBN 978-1-118-71779-0.
19. Namadmalan, A.; Jaafari, B.; Iqbal, A.; Al-Hitmi, M. Design Optimization of Inductive Power Transfer Systems Considering
Bifurcation and Equivalent AC Resistance for Spiral Coils. IEEE Access 2020, 8, 141584–141593. [CrossRef]
20. Lakhdari, A.E.; Cheriet, A.; El-Ghoul, I.N. Skin effect based technique in eddy current non-destructive testing for thickness
measurement of conductive material. IET Sci. Meas. Technol. 2019, 13, 255–259. [CrossRef]
21. Fridman, B.E.; Medvedev, M.V. Skin Effect in Coaxial Conductors of Pulse Facilities. IEEE Trans. Plasma Sci. 2019, 48, 482–490.
[CrossRef]
22. Hayt, W.; Buck, J. Engineering Electromagnetics; McGraw Hill: New York, NY, USA, 2011; ISBN 978-0073380667.
23. Suchantke, R. Alternating Current Loss Measurement of Power Cable Conductors with Large Cross Sections Using Electrical
Methods. Ph.D. Dissertation, Berlin University, Berlin, Germany, 2018.
24. Callister, W.D., Jr. Fundamentals of Materials Science and Engineering; John Wiley & Sons: Hoboken, NJ, USA, 2001; ISBN
0-471-39551-X.
25. Brandt, R.; Neuer, G. Electrical Resistivity and Thermal Conductivity of Pure Aluminum and Aluminum Alloys up to and above
the Melting Temperature. Int. J. Thermophys. 2006, 27, 805–815. [CrossRef]
26. Basic Properties for Conducting Materials. Available online: https://www.nexans.com.br/.rest/eservice/dam/v1/file/223721
/BasicProperties_br.pdf (accessed on 20 November 2022).
27. Catalog and Electrical Characteristics for Conductors. Available online: https://www.nexans.com.br/en/products/Transmission.
and.Distribution/OHL.AWG/AWG/Aluminium.528.html (accessed on 15 December 2022).

202
Energies 2023, 16, 7913

28. Arritt, R.F.; Dugan, R.C. The IEEE 8500-node test feeder. In Proceedings of the IEEE PES T&D 2010, New Orleans, LA, USA, 19–22
April 2010. [CrossRef]
29. Neves, V.R., Jr. Determination of the Typology of Load Curves at Harmonic Frequencies for Low Voltage Residential Consumers.
Master’s Thesis, Universidade Federal de Uberlândia, Uberlândia, Brazil, 2022. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

203
 energies
Article
Research on a Two-Layer Optimal Dispatching Method
Considering the Mutual Aid of Peak Regulating Resources
among Regional Power Grids
Tianmeng Yang 1 , Suhua Lou 2, *, Meng Zhang 3 , Yanchun Li 1 , Wei Feng 1 and Jicheng Liu 1

1 Northeast Branch of State Grid Corporation of China, Shenyang 110180, China; 18240369315@163.com (T.Y.);
mty18842620317@163.com (Y.L.); 13971415920@163.com (W.F.); m202271914@hust.edu.cn (J.L.)
2 State Key Laboratory of Advanced Electromagnetic Engineering and Technology, Wuhan 430074, China
3 China Energy Engineering Group Liaoning Electric Power Design Institute Co., Ltd., Shenyang 110180, China;
mzhang4947@ceec.net.cn
* Correspondence: shlou@mail.hust.edu.cn; Tel.: +86-132-9661-3265

Abstract: Since the power generation structures and load characteristics in each province in China are
quite different, the distribution of peak regulating resources and demands are extremely imbalance.
Restricted by a low power marketization degree, peak regulating resource shortages, and transmission
channel blocks, the efficient utilization of new energy is facing greater pressures. In order to improve
the mutual aid in regional power grids and to obtain more precise simulation results, this paper
proposes a two-layer optimization dispatching model, considering the mutual aid of peak regulation
resources between each province. It determines the optimal startup mode and the units’ power
output in each province and obtains the power output arrangements for all the units and the technical
and economic indicators. The model and the solution method are original and innovative. And
it effectively solved the unequal distribution problem between the peak regulating demands and
resources of each provincial power grid. Finally, taking an actual regional power grid in China as
an example, the simulation results show that the proposed model can significantly improve the
utilization rate of new energy, which verifies the effectiveness and feasibility of the proposed model
and methods presented in this paper.
Citation: Yang, T.; Lou, S.; Zhang, M.;
Li, Y.; Feng, W.; Liu, J. Research on a Keywords: regional power grids; optimal dispatching; peak regulation resource mutual assistance;
Two-Layer Optimal Dispatching peak regulation capability; two-layer optimization
Method Considering the Mutual Aid
of Peak Regulating Resources among
Regional Power Grids. Energies 2024,
17, 667. https://doi.org/10.3390/
1. Introduction
en17030667
With the proposal of “carbon peak and carbon neutral”, power generation structures
Academic Editor: Vítor Monteiro are turning to a low-carbon transformation, and renewable energy power is developing
Received: 26 October 2023 rapidly in China [1–3]. In order to make full use of renewable energy, the government has
Revised: 22 January 2024 put forward high requirements for the utilization rate [4–6]. Therefore, how to plan the
Accepted: 22 January 2024 power grid’s development and optimize the operation modes, and how to make full use
Published: 30 January 2024 of the limited, flexible regulating resources are critical problems. What is more, how to
maximize the utilization rate of renewable energy and reduce the peak regulating costs
have become critical problems, which need to be solved for the current renewable energy
development process in China [7–10].
Copyright: © 2024 by the authors. Currently, there are papers that have studied the optimal dispatching operation meth-
Licensee MDPI, Basel, Switzerland. ods based on considering the full utilization of peak regulating resources, but the research
This article is an open access article
objects have been provincial power grids, and these are not representative. In [11], a
distributed under the terms and
multi-objective optimal scheduling model, considering the economy and flexibility of deep
conditions of the Creative Commons
peak regulation, is constructed, but the power grid under study is not representative.
Attribution (CC BY) license (https://
In [12,13], by establishing the peak regulation right transaction of wind power and thermal
creativecommons.org/licenses/by/
power units, a combined thermoelectric economic dispatching model is established. It
4.0/).

Energies 2024, 17, 667. https://doi.org/10.3390/en17030667 204 https://www.mdpi.com/journal/energies

Energies 2024, 17, 667

can stimulate the predisposition of thermoelectric units to participate in peak regulation,

which can improve the wind power utilization rate. In [14,15], based on a peak regulating
energy consumption cost model for the different stages of thermal power units, an eco-
nomic dispatching model, which gives priority to wind power utilization, is established.
In [16,17], based on the Manson–Coffin formula, a thermal power unit loss cost model with
variable loads is established. And based on the full acceptance of wind power, an economic
dispatching model based on a hierarchical deep peak regulation is established, which aims
to minimize the total power generation costs.
In the studies of the optimization dispatching method in [18,19], the feasibility of peak
regulation between regional power grids is verified. In [20,21], to solve the possible short-
age problems of peak regulating resources, a transaction mechanism for peak regulation
auxiliary service between the regional power grids is proposed, and a clearing pricing
model is first constructed. In [22,23], based on an analysis of the power source and load
characteristics of multiple regions, a typical peak–valley mutual aid operation scheme
is given, and the typical operation mode and economic evaluation results are optimized.
In [24], a two-stage iterative power exchange optimization method considering the marginal
power generation costs is put forward. However, this method only aims at obtaining the
minimum operation costs and cannot fully use regional peak regulating resources. In [25],
an environmental economy dispatching model, considering the inter-provincial power
balance of a regional power grid, is established. However, only the actual power limitation
is taken into consideration, and the influence of the inter-provincial connection lines’ actual
electricity trading is not taken into consideration. In [26], by studying the different power
sources’ peak regulating characteristics, an optimal dispatching method considering peak
regulating resources is studied. However, only the deep peak regulating operation costs
of thermal power units are considered, and the influence of other peak regulating power
sources is not considered.
In summary, peak regulating resources can be exchanged through inter-provincial
transmission lines between provincial power grids in China, so as to make full use of peak
regulating resources in the regional power grid. Based on this, this paper analyzes the
peak regulation characteristics of different power sources, fully considers the differences
between the surplus and gap of peak regulation power in the different provinces, and puts
forward a two-layer optimization dispatching model with the lowest operation costs and
highest utilization rate of renewable energy for a regional power grid. The upper-layer
optimization model’s objective function is the lowest operation costs for a regional power
grid, and the lower-layer optimization model’s objective function is the highest utilization
rate of renewable energy. Based on this two-layer optimization model, the inter-provincial
transmission curve with the lowest costs and the highest renewable power utilization rate
can be re-optimized. And the optimized operation results for the whole regional power
grid can be obtained. And last, taking an actual regional power grid as an example, the
feasibility and high efficiency of the proposed model and optimization method are verified
from the simulation results.

2. System Peak Regulating Resource Analysis

2.1. Analysis of Different Power Supplies
The peak regulation scale for active power and the quantity of electricity is large in
China. Now, the main resources that can effectively participate in peak regulation are
coal-fired units, gas units, cascade hydro-power units (non-runoff), pumped storage, and
other types of large-capacity power storage. Among them, coal-fired units are non-stop
peaking units, and gas units and cascade hydro-power units can be used as stop peaking
units, thanks to their rapid start and stop characteristics [24,25]. The peak regulating
characteristics of each power supply are shown in Table 1.

205
Energies 2024, 17, 667

Table 1. Peak regulating characteristics of different power supplies.

Power Supply Peak Regulating Ability

Coal-fired units Conventional thermal units: 40~50%

After flexibility transformation: 55~70%
Gas units Have shutdown ability, peak regulation ability: 100%
Cascade hydro-power unit (non-runoff) Have shutdown ability, peak regulation ability: 100%
Pump storage units 160%
Other energy storage units 160%
Small hydro-power stations (radial flow) 0%
Wind 0%
Photovoltaic 0%

2.2. Analysis of Coal-Fired Thermal Units’ Energy Consumption Costs

Considering the rapid development of renewable energy, the function of coal-fired
units has gradually changed from the main electricity source to regulating the power supply.
When the coal-fired thermal units change to a regulating function, all the units’ operating
states change frequently. And the characteristics of the technical and economic costs also
significantly change.
Considering the operating states and the energy consumption characteristics of coal-
fired units, the coal-fired units’ peak regulating processes can be divided into three stages: a
basic peak regulation stage (RPR), a non-oil supply peak regulation stage (DPR), and an oil
supply peak regulation stage (DPRO). The characteristic curves for the energy consumption
costs during each peak regulating period are shown in Figure 1. Pmax G represents the rated
G
power output of the coal-fired unit, Pa represents the minimum technical power output of
the RPR stage, PbG represents the minimum technical power output of the DPR stage, and
PcG represents the minimum technical power output of the DPRO stage.

Figure 1. Peak regulating costs of the fuel units.

In recent years, thanks to the technical progress of thermal units, the minimum stable
combustion load of the 200 MW and 300 MW units can be reduced to 45% of rated power
after flexibility transformation. And the minimum stable combustion load of the 600 MW
and 1000 MW units can be reduced to 30~35% after flexibility transformation. The formula
of thermal power units’ energy consumption costs is shown below.
⎧
⎪
⎪ Fcoal ( Ppeak ) PaG < Ppeak ≤ Pamax
G G
⎪
⎪
⎪
⎨ F ( PG ) + F ( PG ) Pb < Ppeak ≤ PaG
G G
coal peak loss peak
G
Fpeak = (1)
⎪
⎪ Fcoal ( Ppeak ) + Floss ( Ppeak ) + Foil ( Ppeak )
G G G
⎪
⎪
⎪
⎩+ Fen ( PG )
peak cPG < P G ≤ PG
peak b

G ) represents the thermal power units’ operating coal con-

In the formula, Fcoal (Ppeak
sumption costs, which is described by the second-order consumption model in (2).
2
Fcoal ( Ppeak
G
) = kcoal · (αPpeak
G
+ βPpeak
G
+ γ) (2)

In the formula, kcoal represents the coal price coefficient, and α, β, γ represent the
consumption characteristic coefficients of thermal units, respectively.

206
Energies 2024, 17, 667

G ) represents the unit loss costs caused by the excessive thermal stress in the
Floss (Ppeak
DPR and DPRO stages. The unit loss costs are mainly determined by the cracking cycle of
the rotor, as shown in (3).
1
Floss ( Ppeak
G
)= τkunit (3)
2NT ( Ppeak
G )

G ) represents the cracking cycle of the rotor, which is interrelated

In the formula, NT (Ppeak
with the power output of units. τ represents the actual operating loss coefficient of the
thermal power units, and the value in the DPRO stage is larger than in the DPR stage. Kunit
represents the investing cost coefficient of thermal power units.
G ) represents the fuel consumption costs in the DPRO stage, which is mainly
Foil (Ppeak
determined by fuel consumption quantity and oil price in the DPRO stage, as shown in (4).

Foil ( Ppeak
G
) = koil · Eoil (4)

In the formula, koil represents oil price, and Eoil represents fuel consumption quantity
in the DPRO stage.
G ) represents additional environmental punishment fees caused by pollutants’
Fen (Ppeak
excessive discharge in the DPRO stage, which is shown in (5).

Fen ( Ppeak
G
) = kpunish · λS (5)
In the formula, kpunish represents the penalty cost coefficient of pollutant discharge,
and λS represents the pollutant emissions.

2.3. Analysis of Pumped Storage Units’ Peak Regulating Costs

As a main tool for peak regulating, the pumped storage units have functions including
peak load shifting and valley filling, frequency regulating and phase regulating, spinning
reserve, and black start. On the one hand, pumped storage units have flexible operating
modes with reversible pump turbine units; on the other hand, limited by different operating
conditions, pumped storage units have specific operating characteristics and constraints.
The pumped storage unit works the same as the conventional generator in the gener-
ating state. In the pumping state, it absorbs power from the power grid. And the operating
costs generated by pumped storage units are mainly generated by the conversion between
pumping and generating states, which consist of generators’ and motors’ start-up costs.
PH−G PH−P
PH
Fpeak,j,t = Fj,t + Fj,t (6)

PH−G
In the formula, Fj,t represents the starting costs at the tth period of the jth pumped
PH−P
storage unit working in the generating state, and Fj,t represents the costs in the
pumping state.

3. Two-Layer Optimization Dispatching Model of the Regional Power Grid

3.1. Introduction of the Model
Considering the differences of the power generation structures and load characteristics
of each provincial power grid, the distribution of peak regulating demands is extremely
imbalanced. Therefore, the regional peak regulating resources should be fully utilized,
which can improve the new energy utilization rate of the regional power grid.
Thus, based on the provincial power balance and dispatching management principles
in China, a two-layer optimization dispatching model of the regional power grids is
proposed in this paper. The model aims to divide the optimal dispatching problem of the
regional power grid into two parts: the units’ optimal dispatching of each province and
the optimization of inter-provincial power transmission plan curves. This two-layer model
is extensible and it can be applied into any regional power grid. The flow chart of the
two-layer optimization method is shown in Figure 2.

207
Energies 2024, 17, 667

Figure 2. Two-layer optimization dispatching process of regional power grids.

3.2. Units’ Optimization Dispatching Model in Each Province

3.2.1. Objective Function
The units’ optimization dispatching model is for optimizing the startup modes and
power output in each province on the basis of the given power transmission plan curves,
so that the total operation costs in each province are the lowest. Taking one province as an
example, the objective function aims to obtain the minimum operation costs, as shown in
Equation (7).
minF = min(α1 · Foper + α2 · Fdep ) (7)
In the formula, α1 and α2 represent the weight coefficients of operation costs and new
energy abandonment penalty costs, respectively. Foper represents the operating costs of
all units in the system, including the operating costs of thermal power units and pumped
storage units. Fdep represents the penalty costs of new energy abandonment.

F = Fpeak
G + Fpeak
PH
⎧oper
⎪
⎪
T NG
⎪ Fpeak
⎨
G = ∑ ∑ Fpeak,i,t
G
t =1 i =1 (8)
⎪
⎪
T NPH
⎩ Fpeak = ∑ ∑ Fpeak,j,t
⎪ PH PH
t =1 j =1

Fdep = kdep · Edep (9)

In the formula, G and

Fpeak,i,t PH
represent the peak regulating costs of the ith
Fpeak,i,t
thermal unit and jth pumped storage unit, respectively, kdep represents the penalty cost
coefficient of abandoned power, and Edep represents abandoned electricity of new energy.
As shown above, the peak regulating costs of thermal power units are mainly obtained
according to Formula (1), and the peak regulating costs of pumped storage units are mainly
obtained according to Formula (6).

208
Energies 2024, 17, 667

3.2.2. Constraints
System Power Balance Constraints

NG NPH
PtL + PtC = PtN + ∑ Ppeak,i,t
G
+ ∑ Pj,tPH − Pdep,t (10)
i =1 j =1

In the formula, PtL represents the load level at the tth period; PtC represents the reserve
capacity at the tth period; PtN represents the power output of new energy at the tth period;
G
Ppeak,i,t PH represents
represents the power output of the ith thermal unit at the tth period; Pj,t
the power output of the jth pump storage unit at the tth period, with a positive value
representing the generating state and negative value representing the pumping state; Pdep,t
represents the abandoned power output at the tth period.

Thermal Power Unit Climbing Constraints

i ≤ Ppeak.i,t − Ppeak,i,t−1 ≤ UPi

DPG G G G
(11)

In the formula, DPG

i and UPG
i represent the climbing up and down limitations of the
thermal unit, respectively.

Pumped Storage Operation Constraints

According to the actual operating restriction of pumped storage units, the pumped
storage unit pumps with rated power in the pumping state. And pumped power generation
should meet the storage capacity constraints. Its power output and storage capacity
constraints are shown below.
PH−P −P
Pj,t = uPH
j,t · Pj,max
PH
(12)

PH−G −G PH
0 ≤ Pj,t ≤ uPH
j,t Pj,max (13)

EPH,min ≤ EPH,t ≤ EPH,max (14)

! PH−G "
NPH NPH Pj,t
EPH,t = EPH,t−1 + ΔT · ∑ ηPH−P · Pj,tPH−P − ∑ ηPH−G
(15)
j =1 j =1

PH represents the rated power output of the jth pumped storage

In the formula, Pj,max
−P −G
unit; uPH
j,t and uPH
j,t represent the pumping and generating state variables, respectively,
−P −G PH−P PH−G
which satisfy uPH
j,t + uPH
j,t ≤ 1; Pj,t and Pj,t represent the pumping power and
generating power of the jth unit at the tth period, respectively. EPH,max represents the rated
storage capacity of the pumped storage station; EPH,min represents the minimum storage
capacity of the pumped storage station; ηPH−P and ηPH−G represent the pumping and
generating efficiency of pumped storage units.

Upper and Lower Limits of Thermal Power Units

G
Pc,i ≤ Ppeak,i,t
G
≤ Pi,max
G
(16)
G represents the lowest power output limit of the ith thermal power unit.
In the formula, Pc,i

3.3. Analysis of Peak Regulation Amplitude and Gap between Provinces in the Regional Power Grid
By solving the provincial unit optimization model, the startup plan and power output
arrangements can be obtained. They are calculated with the initial inter-provincial power
transmission curves. Thus, we calculate the amplitude and gap indexes of peak regulating

209
Energies 2024, 17, 667

resources in each province, which can provide the basis for optimizing the transmission
curves between provincial power grids.
⎧ NG  
⎨ margin
Pt = ∑ Ppeak,i,t
G − Pc,i
G
i =1 (17)
⎩ insuff
Pt = Pdep,t

margin
In the formula, Pt and Ptinsuff represent the amplitude and gap of peak regulation
resource at the tth period, respectively.

3.4. Optimization Scheduling Model of Inter-Provincial Transmission Curves

Based on the peak regulation amplitude and gap in each provincial power grid calcu-
lated in Section 3.3, the initial inter-provincial transmission curves should be optimized
and adjusted. Based on the optimized transmission curves, we can obtain the highest new
energy utilization rate of the whole regional power grid. And the peak regulating resources
will be called up in a more balanced manner.

3.4.1. Objective Function

The objective function of the inter-provincial transmission curve optimization model
mainly considers two aspects; these are the highest new energy utilization rate and the
lowest system peak regulation costs. The objective function is shown below.
T NL T
minEdep = min( β 1 · kdep · ∑ Pdep,t + β2 · kpeak · ∑ ∑ ΔPl,tD ) (18)
t =1 l =1 t =1

In the formula, β1 and β2 represent the weight coefficients of new energy abandonment
penalty costs and peak regulating costs; kpeak represents the coefficient of system peak
regulating costs; Pdep,t represents the abandoned new energy power output at the tth
period after optimizing the inter-provincial transmission curves; ΔPl,t D represents power

differences of the lth inter-provincial transmission lines before and after optimization at the
tth period.
Thus, the calculation formulas of the abandoned new energy power output are as follows:
⎧ NL
⎪
⎪
⎨ Pdep,t = Pdep,t + ∑ ΔPl,t
D
l =1 (19)
⎪
⎪
NL
⎩ ∑ ΔPl,t
D ≤ Pinsuff
t
l =1

3.4.2. Constraints
Transmission Limit Constraints

P D = Pl,t D + ΔPD
'l,t l,t
Pl,t ≤ Pl,N
D Pos , P D ≥ 0
l,t (20)
Neg
Pl,tD ≥ Pl,N , Pl,tD < 0
D and P D represent the transmission power of the lth channel at the
In the formula, Pl,t l,t
Pos and P Neg
tth period before and after optimization, respectively, while Pl,N l,N represent the
transmission limitation of the lth connection channel in two directions, respectively.

Constraints of Transmission Power Limitation

T NL
ElD = ∑ ∑ Pl,tD (21)
t =1 l =1

Elmin ≤ ElD ≤ Elmax (22)

210
Energies 2024, 17, 667

In the formula, ElD represents the transmission electricity of the lth transmission
channel after optimization; Elmax and Elmin represent the upper and lower transmission
electricity limitation of the lth transmission channel.

4. Case Study
In order to fully verify the effectiveness and feasibility of the proposed model and
method, two example power systems are presented for simulation and analysis. They are
the IEEE RTS-96 test system and a regional power grid in China, respectively. Since the two-
layer model is non-linear, all optimization processes mainly depend on Matlab Software
R2016a with CPLEX and the Yalmip optimization engine. Considering the uncertainty of
new energy power, the clustering algorithm is used to gather typical scenarios to represent
all historical curves.

4.1. Example System Analysis Based on IEEE RTS-96

4.1.1. Overview of the Example System
The example system is modified from the IEEE RTS-96 system, and all parameters of
the system are based on the original system. It is divided into three sub-regional power
grids. Each region is connected through DC transmission channels. The load characteristics
of the three sub-regions are shown in Figure 3. The load level of region 1 and region 2
is lower than that of region 3, and it has the smallest peak–valley difference. The peak
regulating abilities of thermal units are shown in Table 2. The wind power capacities are
1300 MW, and the wind power output characteristics in each region are shown in Figure 4.
The peak regulating characteristics of wind power output in region 1 are positive, which
can reduce the peak regulating pressure, while it is negative in region 2, which aggravates
the peak regulating pressure. The characteristics of wind power output in region 3 are
stable, which has little influence on peak regulation.

Table 2. Peak regulating capacities of thermal units in each sub-region.

Region Wind Power Peak Regulating Peak Regulating Peak Regulating

Units Capacities/MW Capacities/MW Capacities Capacities without Oil Capacities with Oil
Region 1 29,410 1300 50% 40% 35%
Region 2 29,410 1300 50% 45% 40%
Region 3 21,100 1300 50% 43% 38%







5HJLRQ 5HJLRQ



5HJLRQ





   
3HULRGKRXU

Figure 3. Load characteristics of regional power grid.

211
Energies 2024, 17, 667

Figure 4. Wind power characteristics of regional power grid.

The weight coefficients α1 and α2 are 0.7 and 0.3, and β1 and β2 are 0.8 and 0.2.

4.1.2. Analytical Calculation

Regions 1 and 2 are the main power delivery areas with surplus unit capacities and
small load level, while region 3 is the main power-receiving area with insufficient unit
capacities and large load level. The initial transmission curves between regions take the
constant power transmission mode. The results of operating costs in each region before
and after optimization are shown in Table 3, and the regional daily transmission curves
before and after optimization are shown in Figure 5.

Table 3. Operation costs of each region before and after inter-provincial transmission curve optimization.

Before Optimization After Optimization

Operation Abandon Operation Abandon

Region Total Costs/USD Costs/USD Punishment Total Costs/USD Costs/USD Punishment
100 Million 100 Million Costs/USD 100 Million 100 Million Costs/USD
100 Million 100 Million
1 0.28 0.23 0.05 0.24 0.22 0.02
2 0.27 0.23 0.04 0.25 0.23 0.02
3 0.25 0.2 0.05 0.22 0.19 0.03
Total 0.8 0.66 0.14 0.71 0.64 0.07

As can be seen in Figure 5, the power transmission curves of regions 1 and 2 after
optimization show great differences. The main reason is that there exist extreme differences
between peak regulating capacities and demands in different regions. Since the load
peak and valley differences in region 1 are small, the peak regulation demands are the
minimum. While the load peak and valley differences in region 2 are large, the peak
regulation demands are the maximum, and the peak regulating support ability to other
regions is the minimum. Therefore, the optimized transmission curves can fully call up
peak regulating resources in the three regions. During low-load periods, region 2 transfers
power to region 3, and region 1 transfers power mainly during high-load periods. Thus, it
can reduce system operating costs and power abandonment penalty costs, and promote
the operating efficiency of each regional power grid.

212
Energies 2024, 17, 667


7UDQVPLVVLRQFXUYHRIUHJLRQDIWHURSWLPL]DWLRQ






7UDQVPLVVLRQFXUYHRIUHJLRQDQGEHIRUHRSWLPL]DWLRQ

 7UDQVPLVVLRQFXUYHRIUHJLRQDIWHURSWLPL]DWLRQ


   
3HULRGKRXU
Figure 5. Daily power transmission curves between each regional power grid before and after
inter-provincial transmission curve optimization.

4.2. Example System Analysis Based on an Actual Regional Power Grid

4.2.1. Overview of the Regional System
In this part, an actual regional power grid in China is taken as an example for analysis
and calculation, and this region consists of four provincial power grids. All the parameters
of units and load curves are based on the actual power system. In this region, provinces A,
B, and C have surplus electricity and province D mainly receives electricity from B and C.
The actual power flows are shown in Figure 6. And in actual operation, the power flow
direction is mainly opposite to what is shown in Figure 6.

Figure 6. Schematic diagram of power transmission and reception between each province.

Since this region is located at high latitude in China, the load demands and the peak
regulating demands are large in winter. Therefore, taking winter load characteristics as
an example, the load characteristics of the four provinces are shown in Figure 7. The load
demands of provinces A, B, and C are small and the load peak valley differences of province
B are the smallest. The power output characteristics of each province are shown in Figure 8.

213
Energies 2024, 17, 667




 3URYLQFH'



 3URYLQFH$


3URYLQFH%

3URYLQFH&

   
3HULRGKRXU
Figure 7. Load characteristics of each province.




 3URYLQFH'
3URYLQFH&

 
3URYLQFH$ 3URYLQFH%


 3URYLQFH&  3URYLQFH'


3URYLQFH% 3URYLQFH$
 




            
      
3HULRGKRXU 3HULRGKRXU

(a) (b)

Figure 8. (a) Wind power output characteristics of each province in regional power grid. (b) PV
power output characteristics of each province in regional power grid.

4.2.2. Analytical Calculation

Before optimization, the DC power transmission curves outside the region and the
initial transmission curves between provinces in the region are based on the historical
annual transmission curves. The operating cost results before and after optimization are
shown in Table 4. The regional daily transmission curves before and after optimization are
shown in Figure 9.

Table 4. Operation costs of each regional power grid before and after inter-provincial transmission
plan curve optimization.

Before Optimization After Optimization

Total Operation Abandon New Energy Total Operation Abandon New Energy
Province Costs/USD Punishment Utilization Costs/USD Punishment Utilization
Costs/USD Costs/USD
100 Million 100 Million Costs/USD Rate% 100 Million 100 Million Costs/USD Rate%
100 Million 100 Million
A 0.46 0.396 0.064 87.9 0.419 0.377 0.042 89.9
B 0.545 0.452 0.092 84.8 0.495 0.433 0.061 88.3
C 0.455 0.335 0.121 89.6 0.423 0.315 0.108 90.7
D 0.653 0.599 0.053 93.3 0.617 0.565 0.052 93.5
Total 2.113 1.782 0.33 87.7 1.954 1.69 0.263 90.9

214
Energies 2024, 17, 667

 

 

 $WR%WUDQVPLVVLRQFXUYHDIWHURSWLLPL]DWLRQ

%WR'WUDQVPLVVLRQFXUYHDIWHURSWLPL]DWLRQ






$WR%WUDQVPLVVLRQFXUYHEHIRUHRSWLLPL]DWLRQ

 
%WR'WUDQVPLVVLRQFXUYHEHIRUHRSWLPL]DWLRQ
 
       
3HULRGKRXU 3HULRG+RXU




$WR&WUDQVPLVVLRQFXUYHDIWHURSWLPL]DWLRQ 



 
&WR'RSWLPL]DWLRQFXUYHEHIRUHRSWLPL]DWLRQ



$WR&WUDQVPLVVLRQFXUYHEHIRUHRSWLPL]DWLRQ
 
&WR'RSWLPL]DWLRQFXUYHDIWHURSWLPL]DWLRQ





    
   
3HULRG+RXU
3HULRG+RXU

Figure 9. Daily power transmission curves of each regional power grid before and after transmission
curve optimization.

As can be seen from Figure 9, the differences in the transmission curves between
provinces C and D and between provinces B and D before and after optimization are the
largest. The main reason is that the peak regulating capacities and amplitude of province D
are the largest. After optimizing the transmission curves between C and D and between
B and D, the surplus peak regulation resources can be supported to B and C, which can
maximize the new energy utilization rate of the whole regional power grid.
According to the calculation results in Table 4, after the unit startup optimization of
the whole system and transmission curve optimization between provinces, the total costs
of the regional power grid decrease by 7.6%, the operation costs decrease by 5.2%, and the
new energy power abandonment penalty costs decrease by 20.4%. After optimization, the
utilization rate of new energy in the whole region increases by 3.2%. And the utilization
rate of new energy in province B has the maximum improvement (3.5%).

4.3. Comparison and Analysis

In order to further verify the effectiveness and feasibility of the proposed optimization
model and algorithm, the traditional optimization method based on equal peak regulating
rate was used for comparison. In the traditional equal peak regulating rate method, it
is required for all thermal units in each province in the region to share peak regulating
demands equally in accordance. In comparison, the model proposed in this paper focuses
on the whole region, which can maximize the mutual aids of peak regulating resources in
the whole regional power grid.
The calculation results of the two models are shown in Table 5. As can be seen from
the comparison results, the total costs by using the equal peak regulating rate optimization
model (model 2) are higher than the proposed model (model 1) in this paper. The new
energy utilization rate is lower, and the solution time of the two models is relatively similar.
As can be seen from the above comparison results, the proposed optimization model and
method in this paper have higher calculation accuracy and solution efficiency. It can fully
consider the mutual aid of peak regulating resources and can achieve the highest new

215
Energies 2024, 17, 667

energy power utilization rate. What is more, it is more suitable for solving the actual
optimization dispatching problems of regional power grids in China.

Table 5. Comparison of calculation results between two optimization models.

Optimization Results
Model
Total Costs/USD 100 Million New Energy Utilization Rate%
1 1.953 90.9
2 2.116 87.1

5. Conclusions
Based on deeply studying the peak regulation characteristics of different power sup-
plies, an original two-layer optimal dispatching model is proposed in this paper. The
upper layer is mainly the unit startup coordination and optimization scheduling model
of each province in the region. The lower layer is mainly the optimal scheduling model
of transmission curves in the region. By solving the two-layer optimization model, the
lowest operation costs and the highest new energy utilization rate of the regional power
grid can be obtained, which can realize full aid and sharing of peak regulating resources in
the whole region. Finally, based on the simulation and analysis of the IEEE RTS-96 power
system and an actual regional power grid, the feasibility and effectiveness of the proposed
model and optimization method are verified.

Author Contributions: T.Y. established the two-layer optimization dispatching model of the regional
power grids and established the simulation model for the verification. S.L. established the two-layer
optimization dispatching model of the regional power grids. M.Z. corrected the manuscript and put
forward many suggestions for improvement. Y.L. put forward some suggestions when building the
model and analyzed the simulation results. W.F. checked and corrected the mathematical model of
the paper. J.L. corrected the manuscript and put forward many suggestions for improvement. All
authors have read and agreed to the published version of the manuscript.
Funding: The work was funded by the science and technology project guided by northeast branch of
state grid corporation of China. Project Number: (52992623000G, SGDB0000DJJS2310060).
Data Availability Statement: Data is contained within the article.
Conflicts of Interest: Authors Tianmeng Yang, Yanchun Li, Wei Feng and Jicheng Liu were employed
by the company Northeast Branch of State Grid Corporation of China; Meng Zhang was employed
by the company China Energy Engineering Group Liaoning Electric Power Design Institute Co., Ltd.
The remaining authors declare that the research was conducted in the absence of any commercial or
financial relationships that could be construed as a potential conflict of interest.

References
1. Zhang, N.; Zhou, T.; Duan, C. Impact of large-scale wind farm connecting with power grid on peak load regulation demand.
Power Syst. Technol. 2010, 34, 152–158.
2. Zhang, Z.; Zhou, L.; Sun, Y.; Gao, X.; Liu, J.; Dong, Y. Joint Optimal Dispatching Strategy of Pumped Storage and Thermal Power
Units with Large-Scale Wind Power Integration. In Proceedings of the 2021 IEEE 5th Conference on Energy Internet and Energy
System Integration, Taiyuan, China, 22–24 October 2021.
3. Zhang, Y.; Le, J.; Liao, X.; Zheng, F.; Liu, K.; An, X. Multi-objective hydro-thermal-wind coordination scheduling integrated with
large-scale electric vehicles using imopso. Renew. Energy 2018, 128, 91–107. [CrossRef]
4. Cheng, D.; Xu, B.; Li, W.; Wu, Q.; Cheng, L.; Zhang, Y.; Zhao, W.; Ma, W.; Zhao, J. Optimal Economic Dispatch Strategy of a
Hybrid Energy Storage Microgrid Based on Genetic Algorithm. In Proceedings of the 2023 IEEE 18th Conference on Industrial
Electronics and Applications, Ningbo, China, 18–22 August 2023.
5. Yang, D.; Yang, Z.; He, G.; Geng, J.; Ligao, J. Wind Power Accommodation Method for Regional Integrated Heat-Power System
Considering Generation and Load Uncertainties. Access 2021, 9, 2169–2180. [CrossRef]
6. Zhu, L.; Chen, N.; Han, H. Key problems and solutions of wind power accommodation. Autom. Electr. Power Syst. 2011, 35, 29–34.
7. Jiang, P.; Yang, S.; Huo, Y. Static security analysis of power systems considering randomness of wind farm output. Autom. Electr.
Power Syst. 2013, 37, 35–40.

216
Energies 2024, 17, 667

8. Zhang, H.; Yin, Y.; Shen, H. Peak-load regulating adequacy evaluation associated with large-scale wind power integration. Proc.
CSEE 2011, 31, 26–31.
9. Kabouris, J.; Kanellos, F.D. Impacts of large-scale wind penetration on designing and operation of electric power systems. IEEE
Trans. Sustain. Energy 2010, 1, 107–114. [CrossRef]
10. Lin, Z.; Chen, H.; Chen, J.; Huang, J.; Li, M.; Wu, Q. Risk-averse robust interval economic dispatch for power systems with
large-scale wind power integration. CSEE J. Power Energy Syst. 2021, 12, 2096–2106.
11. Ying, Y.; Wang, Z.; Wu, X. Multi-objective strategy for deep peak shaving of power grid considering uncertainty of new energy.
Power Syst. Prot. Control 2020, 48, 34–42.
12. Yang, L.; Liang, X.; Wang, X. Combined heat and power economic dispatching considering peak regulation right trading to
improve secondary accommodation capability of wind power. Power Syst. Technol. 2020, 44, 1872–1880.
13. Tan, Q.; Ding, Y.; Zhang, Y. Optimization model of an efficient collaborative power dispatching system for carbon emissions
trading in China. Energies 2017, 10, 1405. [CrossRef]
14. Lin, L.; Zou, L.; Zhou, P. Multi-angle economic analysis on deep peak regulation of thermal power units with large-scale wind
power integration. Autom. Electr. Power Syst. 2017, 41, 21–27.
15. Yang, B.; Cao, X.; Cai, Z.; Yang, T.; Chen, D.; Gao, X.; Zhang, J. Unit Commitment Comprehensive Optimal Model Considering
the Cost of Wind Power Curtailment and Deep Peak Regulation of Thermal Unit. Access 2020, 8, 505–512. [CrossRef]
16. Lin, L.; Tian, X. Analysis of deep peak regulation and its benefit of thermal units in power system with large scale wind power
integrated. Power Syst. Technol. 2017, 41, 2255–2263.
17. Tian, L.; Xie, Y.; Hu, B.; Liu, X.; Li, F. A deep peak regulation auxiliary service bidding strategy for CHP units based on a
risk-averse model and district heating network energy storage. Energies 2019, 12, 3314. [CrossRef]
18. Huang, H.; Xiong, H.; Jiang, B. Design of day-ahead trading mechanism for trans-province peak-shaving auxiliary service of
regional power grid. Smart Power 2020, 48, 119–124.
19. Lu, S.; Zhang, X.; Xu, Y. Two-stage iterative method to optimize tie-line exchange power based on marginal power generation
cost. Power Syst. Prot. Control 2021, 49, 77–88.
20. Zhang, Y.; Hou, Y.; Hu, T. Regional power network environment/economic power dispatching considering balance of power
between provinces. Eng. J. Wuhan Univ. 2017, 50, 548–555.
21. Ohtake, H.; Uno, F.; Oozeki, T.; Yamada, Y.; Takenaka, H.; Nakajima, T. Outlier events of solar forecasts for regional power grid in
Japan using jma mesoscale model. Energies 2018, 11, 2714. [CrossRef]
22. Fang, B.; Xiao, L.; Sun, C.; Chen, Y.; Xue, Y.; Zhang, D. An optimization dispatch method for regional power grid considering
peaking resource mutual support. Electr. Drive 2022, 52, 54–60.
23. Li, J. An improved quantum-behaved particle swarm optimization method for economic dispatch problems with multiple fuel
options and valve-points effects. Energies 2012, 5, 240–250.
24. Liu, D.; Huang, Y.; Wang, W. Analysis on provincial system available capability of accommodating wind power considering peak
load dispatch and transmission constraints. Autom. Electr. Power Syst. 2011, 35, 77–81.
25. Zhuo, Z.; Zhang, N.; Xie, X. Key technologies and developing challenges of power system with high proportion of renewable
energy. Autom. Electr. Power Syst. 2021, 45, 171–191.
26. Chen, J.; Wu, W.; Zhang, B.; Wang, B.; Guo, Q. A spinning reserve allocation method for power generation dispatch accommodat-
ing large-scale wind power integration. Energies 2013, 6, 5357–5381. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

217
 MDPI AG
Grosspeteranlage 5
4052 Basel
Switzerland
Tel.: +41 61 683 77 34
www.mdpi.com

Energies Editorial Office

E-mail: energies@mdpi.com
www.mdpi.com/journal/energies

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are
solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s).
MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from
any ideas, methods, instructions or products referred to in the content.
Academic Open
Access Publishing

mdpi.com ISBN 978-3-7258-1616-3