11th IEEE International Conference on Renewable Energy Research and Applications September 18-21, 2022, Istanbul, TURKEY

A Decentralized Multiagent-Based Robust

Backstepping Control for Restoring Secondary
2022 11th International Conference on Renewable Energy Research and Application (ICRERA) | 978-1-6654-7140-4/22/$31.00 ©2022 IEEE | DOI: 10.1109/ICRERA55966.2022.9922805

Voltage and Frequency of Autonomous Microgrids

Muhammad Maaruf1,3 , Sami El-Ferik1 , Fahad Saleh Al-Ismail2 , and Muhammad Khalid2,3
1
Control and Instrumentation Engineering Department & Center for Smart Mobility and Logistics
2
Electrical Engineering Department & Center for Renewable Energy and Power Systems
3
SDAIA-KFUPM Joint Research Center for Artificial Intelligence
King Fahd University of Petroleum & Minerals, Dhahran 31261, Saudi Arabia
{g201705070, selferik, fsalismail, mkhalid}@kfupm.edu.sa

Abstract—The droop characteristics of primary control of a promising solution to compensate for the deviation and
autonomous microgrids cause voltages and frequencies of the brings the voltage close to the nominal value [5]. Several
distribution generations (DGs) to deviate from their reference types of centralized coordinated control based on multi-agent
values. Various secondary control approaches have been proposed
to tackle this problem. This paper proposes a decentralized systems (MAS) theory have been proposed for restoring the
synchronization control to restore the voltages and frequencies voltages and frequencies of DGs to their normal values [6]–
to their initial states. The control problem is formulated using [12]. However, centralized control scheme needs a complex
the robust backstepping strategy and takes into account the communication network that is susceptible to link failures,
impacts of dynamic uncertainties and external disturbances. data drop-out, and time delays [13], [14]. In addition, the
The network of the interconnected DGs is modeled with a
decentralized multi-agent system (MAS) such that each generator centralized controller requires communication with each agent
is monitored by its own local controller and then takes proper to exchange information, which makes the the communication
action to maintain the voltage and frequency synchronizations. network complex.
By employing the Lyapunov stability theory, it is shown that In order to avoid the limitations of centralized MAS, and
the closed-loop tracking errors are asymptotically stable. Finally, improve the reliability and security of the microgrid, several
simulation results demonstrated that the presented secondary
control technique is effective in synchronizing the voltages and decentralized MAS-based control methods have been studied
frequencies of the microgrids. for enhancing the voltage and frequency stability of the DGs.
Index Terms—Multi-agent systems, decentralized control, dis- In [15], the authors developed an autonomous optimal decen-
tribution generation, autonomous microgrid, secondary voltage tralized control to mitigate voltage fluctuations in an islanded
control, frequency control. microgrid with multiple solar-photovoltaic energy sources con-
nected over a directed graph. In [16], a decentralized MAS-
I. I NTRODUCTION based voltage and reactive power control is investigated. In
Clean renewable energy is fast becoming an alternative [17], a Fuzzy logic control based on MAS theory is developed
source of energy world wide due to depletion of fossil fu- for power management of the components of a stand-alone
els and environmental pollution [1], [2]. Technological ad- microgrid. In [18], a MAS-based decentralized control is
vancements of power electronics converters give room for formulated using linear matrix inequality (LMI) to enhance the
connecting large scale energy storage systems, local loads, transient performance of a large-scale power system. In [19],
and distributed energy resources to the microgrids. Microgrids a decentralized robust MAS control is implemented for micro-
can function in autonomous mode as well as grid-connected grids governed by energy internet. The aim of the controller
mode [3]. In the latter, the control objective of the microgrids is regulating the frequencies and voltages of the microgrid.
is transferring active and reactive power to the main grind. In [20], a MAS-based decentralized fractional proportion-
For the former, the microgrid is operating autonomously as integral-derivative (PID) control is used to increase the voltage
a result of expected or unexpected events. In this case, the stability of a microgrid with multiple energy sources. In [21],
control objective of the microgrids need not only to stabilize a decentralized H∞ control using MAS theory is proposed to
the voltage but also to maintain it at a certain set value. stabilize a multi-area power system.
The voltage control architecture of microgrids is divided It is worth noting that the above MAS-based decentral-
into primary and secondary voltage control [4]. The primary ized control approaches only consider voltage regulation and
control keeps the voltage close to the reference value as load stability enhancement but ignored the frequency regulation
demand and power flow vary. However, primary control may of the DGs. Moreover, they are only limited to the linear
suffer some setbacks such as frequency and voltage deviation dynamic model of the DGs. Therefore, this paper proposes
due to droop control. Secondary voltage control has become a decentralized robust backstepping control based on MAS

ICRERA 2022 licensed use limited to: Universidad Industrial

Authorized 978-1-6654-7140-4/22/$31.00 ©2022
de Santander. Downloaded on March IEEE
23,2024 at 17:40:49 UTC from IEEE Xplore. Restrictions apply. 430
11th IEEE International Conference on Renewable Energy Research and Applications September 18-21, 2022, Istanbul, TURKEY

voltage control gain, voltage droop gain, and frequency droop

gain, respectively.
The primary droop control may prompt frequency and volt-
age to diverge from the desired levels. Therefore, secondary
frequency controller Uωi and voltage controller UVi can be
augmented to (1) and (2), respectively to restore the frequency
and voltage of the DGs to their desired levels.
 
τPi ω̇i + ωi − ω d + kPi Pi −Pi d + Uωi = 0 (3)
d
τQi kVi V̈i +(τQi +kVi )V̇i +Vi −V +kQi (Qi −Qdi ) + UVi = 0
(4)
B. Network model
The autonomous microgrid can be viewed as an MAS where
the DGs are the agents. The communication among the DGs
is modelled by a directed graph. Let G = (V, E, A) be a
directed graph with a set of M nodes V = {v1 , . . . , vM },
Fig. 1. The block diagram of the autonomous microgrid. an edge set E ⊆ V × V, and A = [aij ] as a an associated
adjacency matrix. The elements of A are defined as aij = 1
if (vi , vj ) ∈ E; otherwise aij = 0. The Laplacian matrix of
theory to achieve voltage and frequency restoration simultane- G is given by L = D − A with D being the in-degree matrix
ously during contingencies. This controller not only takes into [23], [24].
account the nonlinear dynamics of the DGs but also dynamic If there exists a transmission line between DGi and DGj
uncertainties and external disturbances. with an admittance Yjk = Gjk + iBjk or Yjk = 0 if there is
no link between DGi and DGj , the active and reactive power
II. S YSTEM MODELLING injected by ith DG inverter into the microgrid is given by:
A. Mathematical model of the autonomous microgrid Pi = P1i Vi 2 + P2i Vi + P3i
X
The autonomous microgrid is illustrated in Fig. 1. It consists + |Bij |Vi Vj sin(δi −δj ) (5)
of N DGs with corresponding local loads. The DGs are j∈Ni
coordinated to keep common reference voltage V d and ref- X
Qi = Q1i Vi2 + Q2i Vi + Q3i + Vi2 |Bij |
erence frequency ω d . Primary control alone cannot guarantee
j∈Ni
the elimination of voltage and frequency deviations when the X
microgrid disconnect from the main grid due to line faults, − |Bij |Vi Vj cos(δi − δj ) (6)
disturbances, power outage, etc. As such, secondary control j∈Ni
is critical in autonomous mode to return the voltage Vi and where P1i , P2i , P3i , Q1i , Q2i and Q3i are constant of the
frequency ωi of each DG to the nominal value. The mathe- ZIP load mode, δi is the power angle of the ith DG.
matical model of each of the DGs is obtained considering the Substituting (5) and (6) into (3) and and (4), respective
following assumptions [11], [12], [22]: gives:
1) Power losses in transmission lines are neglected.
2) The dynamics of the inverter is ignored.
ω̇i = fωi + gωi Uωi + δωi (7)
The primary droop frequency and voltage control of ith
DG are obtained from the active power and reactive power
measured locally, respectively: V̈i = fVi + gVi UVi + δVi (8)
  where δωi and δVi are the lumped disturbances and they are
τPi ω̇i + ωi − ω d + kPi Pi −Pi d = 0 (1) assumed to be bounded as in [25], [26],
τQi kVi V̈i +(τQi +kVi )V̇i +Vi −V d +kQi (Qi −Qdi ) = 0 (2) kP
fωi = − i P1i Vi 2 + P2i Vi + P3i


τPi
where τPi and τQi are the time constants of the frequency and  
voltage filters, respective, ωi and Vi stand for the frequency kPi  X
− |Bij |Vi Vj sin(δi −δj ) −Pi d 
and output voltage respectively, Pi and Qi denote the active τPi
j∈Ni
and reactive power of each DG, respectively, ω d , Vid . Pid and
Qdi stand for the desired frequency, voltage, active power, and ωi − ω d kP P d
− + i i ; gωi = 1/(τPi )
reactive power, respectively, KVi , KQi , and KPi stand for the τPi τPi

ICRERA 2022 licensed use limited to: Universidad Industrial de Santander. Downloaded on March 23,2024 at 17:40:49 UTC from IEEE Xplore.
Authorized Restrictions apply. 431
11th IEEE International Conference on Renewable Energy Research and Applications September 18-21, 2022, Istanbul, TURKEY

N
X
ei = λ1ω ω̃i + λ2ω aij (ω̃i − ω̃j )
j=1
N
X
= (λ1ω + λ2ω lii ) ω̃i + λ2ω lij ω̃j
j=1
N
X
= ζωi ω̃i + λ2ω lij ω̃j (12)
j=1
Fig. 2. Diagram of the decentralized robust backstepping controller for ith
DG where ζVi = (λ1V + λ2V lii ) and ζωi = (λ1ω + λ2ω lii )
A. Decentralized secondary voltage control
The time derivatives of (11) gives:
!
P N
kQi Q1i + |Bij | żi = ζVi Ṽ˙ i + λ2V
X
lij Ṽ˙ j
τQi + kVi j∈Ni
fVi = − V̇i − Vi2 j=1
τQi kVi τQi kVi N
lij Ṽ˙ j
 
kQi X
X
+ |Bij |Vi Vj cos(δi − δj ) = ζVi V̇i − V̇ir + λ2V (13)
τQi kVi j=1
j∈Ni
1 + kQi Q2i kQ (Q3i − Qdi − V d ) We develop the virtual controller as follows:
− Vi − i N
τQi + kVi τQi + kVi λ2V X ˙
gVi = 1/(τQi kVi ) βVi = V̇ir − lij Ṽj − K1V zi (14)
ζVi j=1

Then, the tracking error between V̇i and βVi is:

III. D ECENTRALISED SYNCHRONIZATION C ONTROL OF z2i = V̇i − βVi =⇒ V̇i = z2i + βVi (15)
DG S
Using (15), (13) can be rewritten as:
N
In this section, a robust decentralized control scheme is
lij Ṽ˙ j
  X
developed for restoring the secondary voltage and frequency żi = ζVi z2i + βVi − V̇ir + λ2V (16)
of each DG to its nominal states. The control objective is to j=1

ensure Vi and ωi asymptotically track the reference states Vir A candidate Lyapunov function is chosen as follows:
and ωir , respectively. The control block diagram is depicted 1 2
in Fig. 2. The reference tracking errors of Vi and ωi for each L1i = z (17)
ζVi i
DG are given by:
zi żi
L̇1i =
ζV i
ω̃i = ωi − ωir (9) 
N

zi 
lij Ṽ˙ j 
  X
= ζVi z2i + βVi − V̇ir + λ2V (18)
ζVi j=1

Substituting (14) into (18) yields:

Ṽi = Vi − Vir (10)

and the corresponding synchronization errors are defined as L̇1i = −K1V zi2 + zi z2i (19)
From (15), one has:
N
X
zi = λ1V Ṽi + λ2V aij (Ṽi − Ṽj ) ż2i = V̈i − β̇Vi
j=1 N
λ2V X ˙
N = V̈i − V̈ir + lij Ṽj + K1V żi
X ζVi j=1
= (λ1V + λ2V lii ) Ṽi + λ2V lij Ṽj
j=1 N
λ2V X ¨
N = fVi + gVi UVi + δVi − V̈ir + lij Ṽj + K1V żi
X ζVi j=1
= ζVi Ṽi + λ2V lij Ṽj (11)
j=1 (20)

ICRERA 2022 licensed use limited to: Universidad Industrial de Santander. Downloaded on March 23,2024 at 17:40:49 UTC from IEEE Xplore.
Authorized Restrictions apply. 432
11th IEEE International Conference on Renewable Energy Research and Applications September 18-21, 2022, Istanbul, TURKEY

Consider the following Lyapunov candidate function:

L2i = zi2 + z2i
2
(21)
The time derivative of (21) is :
L̇2i = zi żi + z2i ż2i = −K1V zi2 + zi z2i + z2i ż2i
= −K1V zi2 + z2i [zi + ż2i ] (22)
The decentralized secondary voltage controller for each DG
is designed as:
Fig. 3. Directed communication graph showing the interconnection among
the DGs
N
h λ2V X ¨
UVi = 1/gVi − fVi + V̈ir − lij Ṽj − zi
ζVi j=1
i
− K1V żi − K2V z2i − K3V sign(z2i ) (23)
Substituting (20) and (23) into (22) gives:
L̇2i = −K1V zi2 − K2V z2i
2
+ z2i [δVi − K3V sign(z2i )] (24)
Noting that |z2i |δVi − K3V |z2i | ≤ 0, then, (24) is asymp-
totically stable
L̇2i ≤ −K1V zi2 − K2V z2i
2
, i = 1, 2, . . . , N (25)
B. Decentralized frequency control
The time derivative of (12) gives:
Fig. 4. Active power outputs of the four DGs
N
X
ėi = ζωi ω̃˙ i + λ2ω lij ω̃˙ j
j=1 IV. S IMULATION RESULTS
N
X
= ζωi fωi + gωi Uωi + δωi − ω̃˙ ir + λ2ω lij ω̃˙ j


(26) In order to validate the proposed two-loop distributed op-
j=1 timal secondary voltage controller for the islanded micro-
Consider the following candidate Lyapunov function: grid, simulations are performed in MATLAB/Simulink. The
autonomous microgrid under consideration comprises of four
1 2
L3i = e (27) DGs with a local load each, and three transmission lines. The
ζωi i
parameters of the DGs, loads, lines, and their rated values
The time derivative of L3i is : were adopted from [27], [11]. Parametric variation of ±40 % is
1 added to the system to highlight the robustness of the proposed
L̇3i = ei ėi controller. The directed communication network for the four
ζωi

N
 DGs is depicted in Fig. 3.
λ2ω
X
= ei fωi + gωi Uωi + δωi − ω̃˙ ir + lij ω̃˙ j  (28) The simulation results are presented in Figs. 4-7. It is
ζωi j=1 assumed that the system. The primary control is activated
alone for 0 < t ≤ 10 s to stabilize the microgrid. Beyond
The local frequency controller for each DG is designed as:
t = 10 s, the proposed decentralized MAS-based robust back-
N stepping secondary control is applied to restore the voltages
h λ2ω X ˙
Uωi = 1/gωi − fωi + ω̃˙ ir − lij ω̃j − K1ω ei and frequencies of the DGs to their respective nominal values.
ζωi j=1
i The real power and reactive power delivered by each of
− K2ω sign(ei ) (29) the DGs are depicted in Fig. 4 and Fig. 5, respectively. The
variation of the power outputs of the DGs has a significant
Using (29), (28) becomes: impact on the voltages and frequencies of the DGs, as shown
L̇3i = −K1ω e2i + ei [δωi − K2ω sign(ei )] (30) in Fig. 6 and Fig. 7, respectively. From these figures, it can be
observed that an increase in output power leads to a drop in
Noting that |ei |δωi − K2ω |ei | ≤ 0, then, (30) is asymptoti- voltages and frequencies of the DGs and vice-versa. With the
cally stable: aid of the proposed controller, the voltages and frequencies of
L̇3i ≤ −K1ω e2i , i = 1, 2, . . . , N (31) the DGs are regulated to their required levels and maintain the
microgrid stability despite the increase/decrease in load.

ICRERA 2022 licensed use limited to: Universidad Industrial de Santander. Downloaded on March 23,2024 at 17:40:49 UTC from IEEE Xplore.
Authorized Restrictions apply. 433
11th IEEE International Conference on Renewable Energy Research and Applications September 18-21, 2022, Istanbul, TURKEY

porate artificial intelligence algorithms to make the controller

model-free and resilient to false data injections.
ACKNOWLEDGEMENT
The authors would like to acknowledge the support pro-
vided by the Interdisciplinary Research Center for Renewable
Energy and Power Systems (IRC-REPS) at KFUPM under
Project No. INRE2209, and SDAIA-KFUPM Joint Research
Center for Artificial Intelligence (JRC-AI).
R EFERENCES
[1] M. Maaruf and M. Khalid, “Global sliding-mode control with fractional-
Fig. 7. Frequencies of the four DGs order terms for the robust optimal operation of a hybrid renewable
microgrid with battery energy storage,” Electronics, vol. 11, no. 1, pp.
88–114, 2021.
[2] A. Babangida and P. T. Szemes, “Electric vehicle modelling and simula-
tion of a light commercial vehicle using pmsm propulsion,” Hungarian
Journal of Industry and Chemistry, vol. 49, no. 1, pp. 37–46, 2021.
[3] M. Maaruf, S. El Ferik, and M. S. Mahmoud, “Integral sliding mode
control with power exponential reaching law for dfig,” in 2020 17th
International Multi-Conference on Systems, Signals & Devices (SSD).
IEEE, 2020, pp. 1122–1127.
[4] Y. Xu, Q. Guo, H. Sun, and Z. Fei, “Distributed discrete robust secondary
cooperative control for islanded microgrids,” IEEE Transactions on
Smart Grid, vol. 10, no. 4, pp. 3620–3629, 2018.
[5] C. Sha, J. Luo, and M. T. Ziabari, “A new robust decentralized approach
based on adaptive fuzzy backstepping method for microgrid secondary
voltage control,” Sustainable Energy, Grids and Networks, vol. 27, p.
100487, 2021.
[6] B. de Nadai Nascimento, A. C. Zambroni de Souza, D. Marujo, J. E.
Sarmiento, C. A. Alvez, F. M. Portelinha Jr, and J. G. de Carvalho Costa,
“Centralised secondary control for islanded microgrids,” IET Renewable
Fig. 5. Reactive power outputs of the four DGs Power Generation, vol. 14, no. 9, pp. 1502–1511, 2020.
[7] H. Ji, C. Wang, P. Li, J. Zhao, G. Song, F. Ding, and J. Wu, “A
centralized-based method to determine the local voltage control strate-
gies of distributed generator operation in active distribution networks,”
Applied energy, vol. 228, pp. 2024–2036, 2018.
[8] e. a. Bruno, “Centralised secondary control for islanded microgrids,”
IET Renewable Power Generation, vol. 14, pp. 1502–1511, 2020.
[9] A. Dutta, S. Ganguly, and C. Kumar, “Coordinated voltage control
of active distribution networks in presence of pv and energy storage
system,” in 2021 IEEE 12th Energy Conversion Congress Exposition -
Asia (ECCE-Asia), 2021, pp. 2028–2033.
[10] T. Qian, Y. Liu, W. Zhang, W. Tang, and M. Shahidehpour, “Event-
triggered updating method in centralized and distributed secondary
controls for islanded microgrid restoration,” IEEE Transactions on Smart
Grid, vol. 11, no. 2, pp. 1387–1395, 2019.
[11] F. Guo, C. Wen, J. Mao, and Y.-D. Song, “Distributed secondary voltage
and frequency restoration control of droop-controlled inverter-based
microgrids,” IEEE Transactions on Industrial Electronics, vol. 62, no. 7,
Fig. 6. Output voltages of the four DGs pp. 4355–4364, 2015.
[12] J. Li and D. Zhang, “Backstepping and sliding-mode techniques applied
to distributed secondary control of islanded microgrids,” Asian Journal
of Control, vol. 20, no. 3, pp. 1288–1295, 2018.
V. C ONCLUSION [13] M. S. Mahmoud, M. Maaruf, and S. El-Ferik, “Neuro-adaptive fast
This article studies the problem of frequency and voltage terminal sliding mode control of the continuous polymerization reactor
in the presence of unknown disturbances,” International Journal of
synchronization in an autonomous microgrid. A decentralized Dynamics and Control, vol. 9, no. 3, pp. 1167–1176, 2021.
MAS-based robust backstepping control strategy is proposed [14] M. S. Mahmoud and M. Maaruf, “Robust adaptive multilevel control of
to address this problem. Contrary to existing decentralized a quadrotor,” IEEE Access, vol. 8, pp. 167 684–167 692, 2020.
[15] N. Yorino, Y. Zoka, M. Watanabe, and T. Kurushima, “An optimal
MAS-based secondary control methods, the proposed control autonomous decentralized control method for voltage control devices
strategy is designed considering the complete nonlinear dy- by using a multi-agent system,” IEEE Transactions on Power Systems,
namics of the DGs and parametric uncertainties. It is observed vol. 30, no. 5, pp. 2225–2233, 2014.
[16] S. R. Islam, K. M. Muttaqi, and D. Sutanto, “A decentralized multiagent-
from the simulation results that the proposed control scheme based voltage control for catastrophic disturbances in a power system,”
exhibits excellent transient and steady-state responses in the IEEE Transactions on Industry Applications, vol. 51, no. 2, pp. 1201–
presence of dynamic uncertainties and load variations. In 1214, 2014.
[17] P. Kofinas, A. Dounis, and G. Vouros, “Fuzzy q-learning for multi-agent
addition, the proposed controller can nullify the deviation in decentralized energy management in microgrids,” Applied energy, vol.
frequency and voltage of each DG. Future research will incor- 219, pp. 53–67, 2018.

ICRERA 2022 licensed use limited to: Universidad Industrial de Santander. Downloaded on March 23,2024 at 17:40:49 UTC from IEEE Xplore.
Authorized Restrictions apply. 434
11th IEEE International Conference on Renewable Energy Research and Applications September 18-21, 2022, Istanbul, TURKEY

[18] C.-x. Dou, D.-w. Hao, B. Jin, W.-q. Wang, and N. An, “Multi-agent- [23] M. Maaruf and S. El Ferik, “Distributed control method for heteroge-
system-based decentralized coordinated control for large power sys- neous multiagent systems subjected to faults,” in 2021 18th International
tems,” International Journal of Electrical Power & Energy Systems, Multi-Conference on Systems, Signals & Devices (SSD). IEEE, 2021,
vol. 58, pp. 130–139, 2014. pp. 1328–1333.
[19] B. N. Alhasnawi, B. H. Jasim, B. E. Sedhom, E. Hossain, and J. M. [24] H. A. Hashim, S. El-Ferik, and F. L. Lewis, “Neuro-adaptive coopera-
Guerrero, “A new decentralized control strategy of microgrids in the tive tracking control with prescribed performance of unknown higher-
internet of energy paradigm,” Energies, vol. 14, no. 8, pp. 2183–2217, order nonlinear multi-agent systems,” International Journal of Control,
2021. vol. 92, no. 2, pp. 445–460, 2019.
[20] S. Xu, H. Sun, Z. Zhang, Q. Guo, B. Zhao, J. Bi, and B. Zhang, “Mas- [25] M. S. Mahmoud and M. Maaruf, “Prescribed performance output feed-
based decentralized coordinated control strategy in a micro-grid with back synchronisation control of bilateral teleoperation system with ac-
multiple microsources,” Energies, vol. 13, no. 9, p. 2141, 2020. tuator nonlinearities,” International Journal of Systems Science, vol. 52,
[21] Q. Meng, Y. Xiao, S. Zhang, and L. Zhang, “Multi-agent method based no. 15, pp. 3115–3127, 2021.
h∞ supplementary turbine/governor control to enhance power system [26] M. Maaruf, M. S. Mahmoud, and A. Ma’arif, “A survey of control
stability,” IEEJ Transactions on Electrical and Electronic Engineering, methods for quadrotor uav,” International Journal of Robotics and
vol. 15, no. 1, pp. 128–134, 2020. Control Systems, vol. 2, no. 4, pp. 652–665, 2022.
[22] M. Maaruf, K. A. Khan, and M. Khalid, “Integrated power management [27] X. Shen, H. Wang, J. Li, Q. Su, and L. Gao, “Distributed secondary
and nonlinear-control for hybrid renewable microgrid,” in 2021 IEEE voltage control of islanded microgrids based on rbf-neural-network
Green Technologies Conference (GreenTech). IEEE, 2021, pp. 176– sliding-mode technique,” IEEE Access, vol. 7, pp. 65 616–65 623, 2019.
180.

ICRERA 2022 licensed use limited to: Universidad Industrial de Santander. Downloaded on March 23,2024 at 17:40:49 UTC from IEEE Xplore.
Authorized Restrictions apply. 435