IEEE TRANSACTIONS ON SMART GRID, VOL. 12, NO.

1, JANUARY 2021 157

Distributed Observer-Based Finite-Time Control of

AC Microgrid Under Attack
Rentao Lu , Jie Wang , and Ziqiang Wang

Abstract—In this article, a novel distributed observer-based the advantages of the first two controls and overcome their
finite-time control scheme is proposed to improve the resilience shortcomings. Consequently, network-based distributed control
of islanded AC microgrids under attack. Existing research on is widely implemented in islanded microgrids [6]–[8]. Recently,
microgrids under attack mainly focuses on detection technology
and attenuation technology, but rarely considers the key indicator multi-agent systems (MAS) based distributed consensus control
of convergence speed. The proposed control strategy can accel- methods have become increasingly popular and many related
erate the convergence process of microgrids under attacks. The methods have been proposed [9]–[12]. Through the information
designed controller can converge the frequencies and voltages of exchange among neighbor agents which depends on the com-
the attacked microgrid to the reference values within a finite time, munication topology, the parameters of all agents will converge
while achieving accurate power sharing. The proposed strategy
has an improved performance compared with the conventional to their reference values by consensus control. Some issues,
control methods against attack by applying finite-time theory. such as switching topology [12], communication delay [13],
The effectiveness of the control in case of different types of noise-resilient [14], and battery energy storage system [15],
attack is verified by the simulation of an islanded AC microgrid which affect the stability of microgrid, are well addressed in
in MATLAB/Simulink. corresponding distributed control strategy. Moreover, finite-
Index Terms—Distributed observer-based finite-time control, time theory has been implemented in consensus control to
secondary control, frequency stability, AC microgrid, attack. accelerate synchronization process.
Recently, many distributed finite-time control methods
have been proposed to improve the control performance of
I. I NTRODUCTION microgrid [16]–[24]. In [16], a finite-time consensus proto-
col under switching topology is proposed to guarantee the
ECAUSE the power system vigorously promotes energy
B transition, the proportion of renewable energy has
increased rapidly in recent years. Microgrid gains much atten-
convergence of frequency and voltage in a finite time. A dis-
tributed finite-time convergence algorithm based on feedback
linearization theory for restoration of frequency and voltage
tion as it expedites the integration of renewable energy sources is proposed in [17]. The average voltage regulation and cur-
with high -efficiency, reliability, and scalability [1]. In the rent sharing are studied in [18] and developed a finite-time
microgrid, multiple distributed generations (DGs) operate in control scheme with reduced transient overshoot and chat-
parallel to supply power for loads to enhance the reliabil- tering. However, the coupling effect of current sharing and
ity and capacity of power supply. Therefore, ensuring proper voltage regulation is not discussed in the theoretical analysis.
power sharing among all DGs and maintaining frequency and In order to eliminate the coupling effect, a finite-time sec-
voltage stability are prerequisites for stable operation of the ondary control scheme which consists of four separate control
microgrid [2]. To achieve proper coordination and improve loops is proposed in [19]. A robust finite-time controller is
efficiency of microgrid, many control strategies are proposed designed in [20] which irrespective of parameters uncertain-
to optimize microgrid performance [3]. ties, unmodeled dynamics, and disturbances. Moreover, a con-
Traditional centralized control utilizes the Microgrid Central sensus control architecture for heterogeneous battery energy
Controller (MGCC) to collect information and solve the storage systems is presented in [21], where frequency/voltage
optimal value, then send the reference command to the local convergence and active/reactive power sharing in a finite
controller [4]. Decentralized control relies on local information time period. The distributed energy storage system (DESS)
to make decisions and obtain control instructions [5]. is utilized in [22] to enhance the resilience of microgrid and
Distributed control realizes global observation and control by ensuring finite-time convergence. Furthermore, for additional
exchanging information among neighboring units, combining application scenarios including low voltage microgrid [23] and
Manuscript received December 26, 2019; revised May 1, 2020 and July autonomous microgrid [24], corresponding finite-time control
20, 2020; accepted August 11, 2020. Date of publication August 19, 2020; methods have been proposed respectively.
date of current version December 21, 2020. Paper no. TSG-01925-2019. Although distributed finite-time control has been widely
(Corresponding author: Jie Wang.)
The authors are with the School of Electronic, Information and Electrical used in microgrids, there is no relevant finite-time control
Engineering, Shanghai Jiao Tong University, Shanghai 200240, China (e-mail: method for the vulnerability of microgrid to attack caused
[email protected]). by distributed control. Existing research on microgrid attacks
Color versions of one or more of the figures in this article are available
online at https://ieeexplore.ieee.org. can be divided into two classes. The first class focuses on
Digital Object Identifier 10.1109/TSG.2020.3017793 attack detection technologies. For example, a signal temporal
1949-3053 
c 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:28:33 UTC from IEEE Xplore. Restrictions apply.
 158 IEEE TRANSACTIONS ON SMART GRID, VOL. 12, NO. 1, JANUARY 2021

logic (STL) detection algorithm is presented in [25], which V = {v1 , v2 , . . . , vN } represents the set of nodes in graph G,
can identify two major types of cyber-attacks. In [26], the that is the collection of all agents; E ⊆ V × V represents
attacked agent(s) can be accurately identified by a novel the collection of edges in graph G, that is the collection of
cooperative vulnerability factor (CVF) framework. False- communication paths between agents, and A = {aij }n×n is the
data injection attacks (FDIAs) can be detected by com- non-negative weighted adjacency matrix of graph which repre-
paring the candidate invariants and the actual invariants sent the weight of communication lines between agents. Define
in [27]. The second class is to strengthen the resilience of the edges of node vj to node vi as: (vj , vi ), which represent
microgrid to attacks and it can be achieved by secure con- node vj can receive information from node vi , aij is the weight
trol design or enhanced communication architecture. In [28], of edge (vj , vi ), aij > 0 if (vj , vi ) ∈ E, otherwise aij = 0. For all
a software-defined networking (SDN)-based network archi- nodes vi in this article, we have aii = 0. Node vi is called the
tecture is proposed to improve the resilience and security of neighbor of node vj if (vj , vi ) ∈ E. The set of neighbors of the
microgrid operations against cyber-attacks. It is shown in [29] node vi is given by Ni = {j|(vj , vi ) ∈ E}. The Laplace matrix of
that designing a recursive systematic convolutional (RSC) code the oriented graph G can be represented as L = D − A, where
and Kalman Filter (KF) based method is effective to eliminate  matrix is defined as D = diag{d1 , d2 , . . . , dN } with
the degree
the effect of attacks. An observer-based controller with confi- di = j∈Ni aij . Moreover, because some nodes can be con-
dence factors and trust factors is proposed for synchronization nected to a virtual leader node by unidirectional edges, define
of microgrid in [30]. Due to the lack of distributed finite-time the pinning gain matrix as A = diag{ai0 }, where ai0 > 0 if
control considering attacks, this article proposes an observer- follower node vi is connected to the leader node across the
based finite-time controller based on the observer proposed communication link (vj0 , vi ), otherwise ai0 = 0.
in [30]. The designed controller can converge the frequencies
and voltages of DGs in finite time and achieve accurate power B. Corresponding Mathematical Lemmas
sharing after the microgrid is attacked. Moreover, the effect of Lemma 1 [31]: Consider a double integrator system
attacks can be limited in the attacked agent(s) without spread-
ing in the communication network by the designed controller. ẋi (t) = yi (t) (1)
The stability and resilience of islanded AC microgrids can ẏi (t) = Aui (t) (2)
been greatly enhanced by the proposed controller.
The main contributions in this article can be summarized as where A is a positive-definite matrix. To make the system glob-
follows. ally asymptotically stable in a finite time, the control input is
1) A distributed observer-based finite-time control strategy designed as
against attack is proposed to provide a reliable control frame- ui (t) = −k1 sig(xi )α − k2 sig(yi )β (3)
work for islanded AC microgrid, where the frequency and
voltage of islanded AC microgrid under attack can converge where k1 , k2 > 1, sig(·)α = sign(·)| · |α , 0 < α < 1, β = 1+α 2α
,
within settling finite time. the state variable xi (t) will achieve converge to the initial state
2) An improved attack detection and localization method in a finite settling time under the designed control input.
is proposed based on the confidence factor and trust factor Lemma 2 [32]: For continuous control system ẋ = f (x, t),
in [30], which can eliminate the defect of indistinguishability where f (0, t) = 0, if there is a function V(t) satisfy:
between attacks and real disturbances. 1) V(t) is positive definite and differentiable in the neigh-
3) Unlike existing finite-time control methods, this article borhood of the origin;
considers the vulnerability of distributed control to malicious 2) Existing real numbers c > 0 and 0 < d < 1 such that
attacks and developed an observer-based finite-time control
V̇(t) + cV d (t) ≤ 0 (4)
based on the observer introduced from [30].
V 1−d (t
0)
4) In comparison with conventional attack attenuation meth- Let T ≤ c(1−d) . One has
ods and detection methods, the proposed method has fast 
convergence speed and shows a superior performance for limt→T V(t) = 0
(5)
a stealthy attack. V(t) = 0, ∀t ≥ T
The rest of this article is organized as follows: Section II (t0 )
1−d
provides preliminaries of graph theory and corresponding i.e., V(t) reaches zero within finite time T ≤ Vc(1−d) .
mathematical lemmas. In Section III, a cooperation control Assumption 1 [30]: The communication link of each DG is
of AC microgrids are discussed. The distributed observer- at least (2z + 1), where z is the number of attacked neighbors,
based finite-time control scheme under attack is described in which means at least half of the information received by any
Section IV. Digital simulation verification of the proposed con- DG is intact.
trol strategy performance is shown in Section V. Conclusions Assumption 2: The attacker is assumed not to generate
are presented in Section VI. a switch command to the executor, which is used to ensure
exist device to execute control command.
II. P ROBLEM F ORMULATION
A. Preliminaries on Graph Theory C. Cooperative Control of AC Microgrids
For a multi-agent system with N agents, the communication Consider an islanded AC microgrid with multiple distributed
network can be represented by graph G = (V, E, A), where generation resources, the block diagram is depicted in Fig. 1.

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:28:33 UTC from IEEE Xplore. Restrictions apply.
 LU et al.: DISTRIBUTED OBSERVER-BASED FINITE-TIME CONTROL OF AC MICROGRID UNDER ATTACK 159

III. ATTACK M ODELS AND C OUNTERMEASURES

There are two main types of cyber-attacks in microgrids,
FDIA and DoS. FDIA produce adverse impact by injecting
false data to change the state variables of the attacked DG,
DoS destroy the stable operation of the system via interrupt-
ing data exchange between distributed controller by attacking
communication links. The DoS attack can be ignored because
the lack of certain information in the steady state has little
effect on the state of the observer. FDIA attacks are the main
objects in this article.
Fig. 1. The block diagram of an inverter-based AC microgrid. The AC microgrid is more susceptible to attacks that
become invisible due to communication restrictions between
DGs. The signals of attacks are mainly constant or arbi-
Hierarchical control scheme is implemented in the microgrid trary, and the arbitrary attack signal can be eliminated by
to alleviate system performance. noise filtration techniques [33]. The constant signal attacks are
In this article, DGs in the microgrid adopt SPWM modula- considered here, FDIA attacks are mainly divided into two
tion mode, and the output voltage often has a large harmonic categories based on the consensus tracking protocol which are
component. The LC filter is used to filter out the harmonic studied in this article.
components and the stable voltage output is obtained by setting
the filter parameters reasonably. The inverter adopts double-
A. Link Attack
loop control, which is composed of outer-loop voltage control
and inner-loop current control. The output of the outer-loop The link attack injects the false data into the transmission
voltage control is used as the input reference signal of the data of the communication link, where the data receiving agent
inner-loop current control. The outer-loop control has a longer of the communication link will receive the destroyed neighbor
dynamic response, so the output voltage control is achieved information. The link attack applied on the communication
through a PI controller during the control process. In order to link between two agents which can be modeled by
speed up the response speed of the control system, the current yi = aij xi − xj + τj xja + ai0 xi − ωref (9)
output by the filter inductor is used as the control variable of j∈Ni
the current inner loop controller. The primary droop control
uses ω − P and v − Q droop characteristics to keep system where xia is the injected false data of the ith DG, τi is the coef-
voltage and frequency near nominal values. ficient of xia , where τi = 1 if there exists a false data injection,
The primary control which is based on droop technique is otherwise τi = 0, yi is the measurements of ith DG.
given by
 B. Node Attack
ωi = ωni − mi Pi
(6) The node attack directly injects the false data into the
Vi = Vni − ni Qi
agent and destroys the node information, the damaged node
where ωi and Vi are the output angular frequency and volt- information is sent to the neighbor node which thus causing
age magnitude of the ith DG, respectively; ωni and Vni are the global instability. Attack on the controller can lead to entire
frequency and voltage amplitudes of the nominal set point, controller to be hijacked and this attack can be modeled by
respectively, which will be regulated by the secondary con-  
trol; mi and ni are the droop coefficients of active power and yi = aij xi + τi xia − xj + τj xja
reactive power, respectively; Pi and Qi are the measured active j∈Ni
 
and reactive power, respectively. + ai0 xi + τi xia − ωref (10)
Transform (6) into d-q reference frame form as
⎧ where xia is the injected false data of the ith DG, τi is the coef-
⎨ ωi = ωni − mi Pi ficient of xia , where τi = 1 if there exists a false data injection,
Viod = Vni − ni Qi (7) otherwise τi = 0, yi is the measurements of ith DG.
⎩ oq
Vi = 0 It is clear that the system will reach different operating
oq
where Viod and Vi are the d-axis and q-axis output voltage states under different attacks. The information of each agents
references, respectively. will maintain a special deviation from the reference value
The measured active and reactive power of DG i is given by which represents the instability of the system. Accordingly,
⎧ c  
the information of the neighbor node needs to be evaluated to
⎨ Pi = ωc i V od I od + V oq I oq detect and locate the attack.
ωi +s i i i i
c   (8)
⎩ Qi = ωc i V od I oq − V oq I od
ω +s i i
i
i i C. Detection/Localization Strategies
where ωic
is the cutoff frequency of the low pass filter. Distributed cooperative control is more prone to cyber-
The design of secondary controller will be introduced in attacks, the information exchange among distributed con-
Section IV. trollers make the attack propagate across communication

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:28:33 UTC from IEEE Xplore. Restrictions apply.
 160 IEEE TRANSACTIONS ON SMART GRID, VOL. 12, NO. 1, JANUARY 2021

networks and lead to a network-wide instability. In order to information with false data injection. The agent needs to make
reduce adverse effects and prevent the propagate of attacks, an decisions based on the local information and neighbor agents
improved attack detection and localization method is proposed information to determine the destroyed data. The location of
based on the confidence factor and trust factor in [30]. The attack cannot be detected only depending on the value of
observer with confidence factors and trust factors is given by confidence factor. Therefore, it is necessary to design a cor-
  responding attack detection and localization method for the
x̂˙ i (t) = aij Cj (t)Tij (t) x̂i (t) − x̂j (t) + ai0 x̂i (t) − xref (11) information transmitted by the communication link.
j∈Ni To achieve this goal, define
where x̂i is the observation of DG i; xref is the reference value ρij (t) = nxj (t) − xk (t) (16)
of virtual leader node, Cj is the confidence factor, Tij is the k∈Ni
trust factor.
where n is the number of neighbors of DG i.
Remark 1: The distributed observer is designed to track the
During normal operation, ρij = 0 for all communication
reference information of the leader node, and the secondary
links; in the presence of attack, DG i may receive false data by
control can then be realized by using the local information
the communication link from DG j which lead to ρij = 0. The
and the local estimation of the reference set point. The confi-
propagate of the attack in communication link may lead to
dence factor and trust factor are used to evaluate the fidelity
abnormal information of multiple agents. The proximity of
of the interactive information to achieve attack detection and
DG i to the attacked DG determine the magnitude of ρij , the
localization. The attack can be prevented from spreading
greater the magnitude of ρij , the more likely DG i is the attack
in the communication network by discarding the information
source or close to the attack.
of the attacked node. The communication data is replaced from
εi
the system information to the observations, the attack directly sij (t) = (17)
acting on the system frequency or voltage is limited to the εi + ρij (t)
attacked DG. where εi is a threshold which is settled by considering
An attack detection and localization method is designed microgrid system.
based on the introduced observer and attack model. According
to the local neighborhood error define ḟij (t) = χ sij (t) − χ fij (t) (18)
where X > 0 is a weighing factor of fij .
x̂i (t) = xref (t) − x̂i (t) (12)
  The trust factor of DG j for DG i is given by
bi (t) = hi aij x̂j (t) − x̂i (t)  
j∈Ni Tij = max Ci (t), fij (t) (19)
 
+ aij x̂j (t) + x̂i (t) (13) The trust factor of DG j for DG i is determined by ρij if
j∈Ni DG i is attacked. The greater the ρij , the more likely DG j is
the attacked node. Moreover, the information transmitted by
where hi is a positive constant.
communication link from DG j to DG i would be discarded
In the presence of attack, bi (t) will change from zero to
once the trust value drops below the threshold i .
a positive value which can be used to detect the existence of
attacks. Therefore, we can check the value of bi (t) which pro-
vide essential information of attacks. Because of the presence IV. D ESIGNING OF O BSERVER -BASED
of attack, we have bi (t) = 0 for the attacked node and the F INITE -T IME C ONTROL
affected nodes. Based on the introduced observer, the design of secondary
Although attack detection is possible based on checking controller can effectively prevent attack from spreading in the
the value of bi (t), locating the attack sources is crucial for network. However, the convergence speed of microgrid under
restoring microgrid to normal operations. attack is important for the performance of controller, it is nec-
To achieve this goal, define essary to speed up the stabilization process. In this section,
the observer-based finite-time control strategy is proposed to
Ċi (t) = μdi (t) − μCi (t) (14) quickly converge the frequency and voltage of the microgrid
where μ > 0 is a weighing factor of Ci ; 0 ≤ Ci ≤ 1 with under attack, while achieving accurate power sharing.
σi
di (t) = (15) A. Finite-Time Frequency and Active Power Control Design
σi + bi (t)
To make the frequency converge to the reference value
where σi is a set threshold for the confidence factor. in settling time after microgrid being attacked, a finite-time
After attack occurs, bi σi for DG i which can be affected observation control algorithm is developed as follow.
by an attack. For a node attack, the value of Ci can be used Based on the observer introduced above, the local tracking
to locate the attack which can indicate the attacked node by error of frequency observation is given by
finding the minimum value of Ci , and the value of Ci depends  
on the proximity of agent i to the attacked node. eω̂i (t) = aij Cj (t)Tij (t) ω̂i (t) − ω̂j (t) + ai0 ω̂i (t) − ωref
For a link attack, the agent that receives information in j∈Ni
the communication link will obtain the damaged neighbor (20)

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:28:33 UTC from IEEE Xplore. Restrictions apply.
 LU et al.: DISTRIBUTED OBSERVER-BASED FINITE-TIME CONTROL OF AC MICROGRID UNDER ATTACK 161

A double integrator system for synchronization error is e2ω̂i = aio ω̂i − ωref (30)
represented as j∈Ni

ėω̂i = kω̂i Hence it follows that
(21)
k̇ω̂i = uω̂i 1+α 1+α
V̇1 ≤ −(2λ) 2 V1 2 (31)
ωi = ė
where k ωi , the local error for k
ωi is given by
For 0 < α < 1 , by Lemma 2 [32], we have
ėω̂i (t) = aij Ċj (t)Ṫij (t) ω̂˙ i (t) − ω̂˙ j (t) + ai0 ω̂˙ i (t) (22)
2 1+α
j∈Ni
T1 ≤ 1+α
V1 (0) 2 (32)
From the Lemma 1 [31], considering the double integrator (2λ) (1 − α)
2

system (22) to achieve observation finite-time global stability, limt→T1 V1 (t) = 0
(33)
the control input is given by V1 (t) = 0, ∀t ≥ T1
 α  β
uω̂i = −cω1 sig eω̂i − cω2 sig ėω̂i (23) Then we obtain that V1 (t) converges to zero within finite
time T1 and e ωi = k ωi = 0, ∀t ≥ T1 , which means that the
where cω1 , cω2 > 1 is the controller gain, 0 < α < 1, β = 1+α
2α
. local neighbor tracking error would reduce to zero in settling
Theorem 1: The distributed finite-time controller (23) can time. Then the observations of all DGs can synchronize to
pin the observation of all DGs to the reference frequency ωref their reference ωref in finite time. This completes the proof.
asymptotically, the settling time T1 is shown in (32). The observations of all DGs have been regulated to converge
Proof: Rewrite the double integrator system (21) as to their reference frequency by the finite-time frequency obser-

ėω̂ = kω̂ vation controller (23). Then, a finite-time frequency control
(24)
k̇ω̂ = uω̂ = −cω1 sig(eω̂ )α − cω2 sig(kω̂ )β algorithm which based on the observation after stabilization is
developed as follow.
Take the Lyapunov function as
The frequency finite-time controller can be designed as:
1  λ
V1 = (eω̂ )T eω̂ (25) uωi (t) = −cω3 sig ωi (t) − ω̂i (t) (34)
2
It can be realized that where  ωi = ωref is a constant when t ≥ T1 ; c3 > 1 is the
 α  α
cω1 sig eω̂i = cω1 aij Cj (t)Tij (t)sig ω̂i (t) − ω̂j (t) controller gain; λ < 1 is the controller coefficient.
j∈Ni The frequency deviation generated by the primary control
β can be eliminated by the designed finite-time frequency con-
+ cω2 ai0 sig ω̂i (t) − ωref (26) troller (34). However, secondary frequency control may result
in the inability to accurately share active power. Therefore, it
Differentiating (25) yields
is necessary to ensure the accurate active power sharing while
V̇1 = (eω̂ )T ėω̂ achieving frequency stability:
⎛
N
 α mi Pi = mj Pj , ∀i, j (35)
= eω̂i ⎝cω1 aij Cj (t)Tij (t)sig ω̂i (t) − ω̂j (t)
i=1 j∈Ni where mi is the droop coefficient of active power.
⎞
 β To realize accurate active power sharing and fast stable
+ cω2 ai0 sig ω̂i (t) ⎠ (27) performance, the active power controller is designed as
 p
uPi (t) = −cPi aij sig mi Pi (t) − mj Pj (t) (36)
Following Lemma 2 [24] for an undirected network, (27) can
j∈Ni
be further written as
N
 1+α  1+β where cPi > 1 is the controller gain, 0 < p < 1.
1
V̇1 = − cω1 aij ω̂i − ω̂j  + 2cω2 ai0 ω̂i  Denotethe tracking error eωi = ωi −  ωi and ePi =
2 mi Pi − N1 N m P . For simplicity, the convergence of angular
i=1 i=1 i i
N N
  2  2 N
2  2 frequency tracking error eωi is discussed, and similar results
1
≤− cω1 aij 1+α ω̂i − ω̂j  − (cω2 ai0 ) 1+β ω̂i  can be obtained for the tracking error ePi .
2 Theorem 2: The distributed finite-time frequency con-
i=1 j=1 i=1
= −(e1ω̂ ) (L)e1ω̂ − 2(e2ω̂ ) (G)e2ω̂
T T troller (34) can pin the frequencies of all DGs to their
observation  ωi which is actually the reference frequency ωref
≤ −λ1 (L)(e1ω̂ )T e1ω̂ − 2λ2 (G)(e2ω̂ )T e2ω̂
1+α 1+β
when t ≥ T1 . The settling time Tω is shown in (40).
≤ − 2λ1 (L)V11
2
− 4λ2 (G)V12
2
(28) Proof: Differentiating eωi yields
 λ
where V1 is divided into two terms, V11 = (e1 ėωi = ω̇i − ω̂˙ i = ω̇i = uωi = −c3 sig ωi (t) − ω̂i (t) (37)
ωi ) e1
ωi , V1 =
T 2

(e2
ωi ) e2
T
ωi with Take the Lyapunov function as
 
e1ω̂i = aij ω̂i − ω̂j (29) 1 T
V2 (eω ) = e eω (38)
j∈Ni 2 ω
Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:28:33 UTC from IEEE Xplore. Restrictions apply.
 162 IEEE TRANSACTIONS ON SMART GRID, VOL. 12, NO. 1, JANUARY 2021

Differentiating (41) yields where v̂i = vref is a constant when t ≥ T1 ; cv3 > 1 is the
N controller gain; γ < 1 is the controller coefficient.
 λ The voltage-reactive power droop control law is not suffi-
V̇2 = eTω ėω = eωi −c3 sig ωi (t) − ω̂i (t)
i=1 cient to achieve the goal of realize accurate reactive power
sharing, which can be given by
N
 λ N
 1+λ
= eωi −c3 sig eωi = −c3 eω 
i
ni Qi = nj Qj , ∀i, j (49)
i=1 i=1
  1+λ
where ni is the droop coefficient of reactive power.
N
 2 2
1+λ
≤ −c3 eω  = −c3 2
1+λ
2 V2 2 (39) To realize accurate reactive power sharing and fast stable
i
i=1 performance, the reactive power controller is designed as
1+λ  q
Due to c3 2 2 > 0 and 0 < 1+λ
2 < 1, one has uQi (t) = −cQi aij sig ni Qi (t) − nj Qj (t) (50)
j∈Ni
2 1−λ
Tω ≤ V2 (0) 2 (40)
1+λ
c3 (2) 2 (1 − λ) where cQi > 1 is the controller gain, 0 < q < 1.
 According to (48) and (50), we can write vni as follows:
limt→Tω V2 (t) = 0
(41) 
V2 (t) = 0, ∀t ≥ Tω  
vni = uvi + uQi dt (51)
Then we obtain that V2 (t) converges to zero within finite time
Tω and eωi = ωi − ωi = 0, ∀t ≥ Tω , which means that the local Therefore, the control inputs vni can realize voltage finite-
neighbor tracking error would reduce to zero in settling time. time stabilization and achieve active power sharing:
Therefore, the actual frequency of all DGs can synchronize to  
 
their reference ωref in finite time. This completes the proof. lim vi (t) − vref  = 0, vi (t) = vref , ∀t ≥ Tv (52)
t→Tv
According to (34) and (36), we can write ωni as follows:  
  N  N
 1  1
ωni = (uωi + uPi )dt (42) lim ni Qi − ni Qi  = 0, ni Qi = ni Qi , ∀t ≥ TQ .
t→TQ  N  N
i=1 i=1
Therefore, the control inputs ωni can realize frequency (53)
finite-time stabilization and achieve active power sharing:
 
  V. S IMULATION R ESULTS
lim ωi (t) − ωref  = 0, ωi (t) = ωref , ∀t ≥ Tω (43)
t→Tω The proposed observer-based finite-time control scheme will
 
 N  N be applied to an islanded AC microgrid with five DGs to verify
 1  1
lim mi Pi − mi Pi  = 0, mi Pi = mi Pi , ∀t ≥ TP . its effectiveness. Fig. 2 depicts the control flow of observer-
t→TP  N  N
i=1 i=1 based finite-time control scheme. The basic block diagram of
(44) the test system is shown in Fig. 3. Two different communica-
tion network topologies are presented in Fig. 4. The parameters
B. Finite-Time Voltage and Reactive Power Control Design of DGs, loads, and lines are shown in Table I. The controller
Similarly, the observer-based finite-time voltage controller gains are c1 = c2 = 2, c3 = 2, the coefficients for the sec-
and reactive power controller are given in this section. ondary controller are α = 0.8, β = 89 , λ = 0.8, the low
The finite-time voltage controller based on attack detection threshold of trust factor is set as Tij > 0.5.
is developed as follows:
Based on the observer introduced above and the defined A. Case I: Load Variations
tracking error system, the voltage controller is given by:
1) Case I-A (Stability Analysis): The performance of
 α  β
uv̂i = −cv1 sig ev̂i − cv2 sig ėv̂i (45) the designed controller under load variations is given in
Fig. 5(a)-(d). As shown in Fig. 5(a) - (b), the frequency of all
where cv1 , cv2 > 1 is the controller gain, 0 < α < 1, β = 1+α 2α
, DGs kept relatively synchronous and produced a small drop.
  It can be seen that the global synchronization of the frequency
ev̂i (t) = aij Cjv (t)Tijv (t) v̂i (t) − v̂j (t) + ai0 v̂i (t) − vref
has not been destroyed, and the voltage drop range at the load
j∈Ni
increasing node is larger than the voltage of other DGs. The
(46) frequency and voltage of each DG can be quickly restored to
ėv̂i (t) = aij Ċjv (t)Ṫijv (t) v̂˙ i (t) − v̂˙ j (t) + ai0 v̂˙ i (t) (47) the reference value. Fig. 5(c) - (d) show the changes of active
j∈Ni power and reactive power of each DG in the microgrid under
load changes. The power vacancy caused by the increased load
where v̂i is the observation of DG i; vref is the reference volt-
can be allocated to each DG according to the power sharing
age of virtual leader node, Cjv is the confidence factor, Tijv is the
protocol. Under the severe disturbance of load changes, the
trust factor.
proposed distributed controller is still capable of achieving
The finite-time voltage controller can be designed as:
 γ zero frequency deviation and voltage deviation while realizing
v̇i = uvi (t) = −cv3 sig vi (t) − v̂i (t) (48) accurate power sharing within a few seconds.

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:28:33 UTC from IEEE Xplore. Restrictions apply.
 LU et al.: DISTRIBUTED OBSERVER-BASED FINITE-TIME CONTROL OF AC MICROGRID UNDER ATTACK 163

Fig. 2. The control flow of the observer-based finite-time control for ith DG.

Fig. 3. The test islanded AC microgrid with five DGs.

Fig. 5. Case I-A: (a) Frequencies of DGs, (b) Voltages of DGs, (c) Active
power of DGs, (d) Reactive power of DGs.

Fig. 4. The topology diagram of communication network.

change as an attack may lead to decision failure and destabilize
TABLE I
T HE PARAMETERS OF THE T EST S YSTEM
the system. Fig. 6(c) - (d) show the change of detection factors
of the improved attack detection and localization method under
load variations, it can be seen that neither of the two vari-
ables has changed, which indicates that the improved attack
detection method can distinguish the real disturbance from the
attack.

B. Case II: Link Attack

Both the frequency link attack and the voltage link attack
are considered.
1) Case II-A (Frequency Link Attack): The communica-
tion link between DG 2 and DG 3 was being attacked and
a false data was injected into the information which trans-
2) Case I-B (Identify Load Change Analysis): The load mitted to DG 3. The attack is modeled by (9) to produce
variations disturbance identification under distributed attack- a corrupted frequency information of f2 = 50.1Hz at t = 5s.
resilient cooperative control and distributed observer-based Fig. 7(a)-(d) show the performance of the proposed dis-
finite-time cooperative control is compared. The confidence tributed observer-based finite-time cooperative control. The
levels and trust levels under load variations is given in frequencies and voltages can be regulated to converge to their
Fig. 6(a)-(b), it can be seen that both factors will be affected reference value in finite time which is about 10s. The active
and reduced, which means that the information will be given power of different DGs has different variation trends after the
corresponding weights. This misidentification of the load microgrid was attacked, but all of them can return to the stable

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:28:33 UTC from IEEE Xplore. Restrictions apply.
 164 IEEE TRANSACTIONS ON SMART GRID, VOL. 12, NO. 1, JANUARY 2021

Fig. 6. Case I-B with the distributed attack-resilient cooperative control: Fig. 8. Case II-B: (a) Frequencies of DGs, (b) Voltages of DGs, (c) Active
(a) Confidence levels, (b) Trust levels; Case I-B with the distributed observer- power of DGs, (d) Reactive power of DGs.
based finite-time cooperative control: (c) Confidence levels, (d) Trust levels.

under the frequency link attack will be affected. The voltage

link attack has almost no effect on the frequency and active
power of the system, which shows to some extent that the
frequency link attack is even worse for the microgrid.

C. Case III: Node Attack

Both the frequency node attack and the voltage node attack
are considered. The communication topology G1 cannot sat-
isfy the assumption 1 which will lead to failed to determine the
feasibility of the information. Therefore, the communication
topology G2 is implemented in this case to perform simulation
verification.
1) Case III-A (Frequency Node Attack): The entire
frequency controller of DG 2 was hijacked and lead to the
destruction of frequency and observation. The attack is mod-
Fig. 7. Case II-A: (a) Frequencies of DGs, (b) Voltages of DGs, (c) Active eled by (10) to generate a compromised frequency information
power of DGs, (d) Reactive power of DGs. of f2 = 50.1Hz between t = 5s and t = 15s. The effi-
ciency of the observer-based finite-time control is verified in
Fig. 9(a)-(d). In this case, the attack acts directly on the actual
state within a few seconds. It can be seen from Fig. 7(d) that frequency which cannot be eliminated by ignoring commu-
frequency attack has little effect on reactive power. nication information. Therefore, the frequency of DG 2 is
2) Case II-B (Voltage Link Attack): The voltage maintained in an abnormal state during attack, which cause
information transmitted from DG2 to DG3 through the the system voltage to fail to converge. The frequencies of the
communication link is injected with the false data which is remaining DGs quickly restored to the reference value after
modeled by (9) to produce a destroyed voltage information the destroyed information is ignored. The voltage of each DG
of v2 = 381pu at t = 5s. The resultant plot of all DICs shows a downward trend during the existence of the attack,
responses is shown in Fig. 8(a)-(d). Clearly, due to this and gradually converges to a stable state after the attack is
voltage link attack, the voltage of each DG increased slightly removed. Active and reactive power are also in an abnormal
and the frequency almost keeps unchanged. Similarly, the state due to the persistence of the attack.
active power and reactive power remain almost unchanged. 2) Case III-B (Voltage Node Attack): The entire voltage
For link attacks, according to the simulation of two different controller of DG 2 was hijacked and generate a compromised
link attacks, it can be proved that the proposed observer-based voltage information of v2 = 381pu between t = 5s and
finite-time controller can effectively detect and locate the link t = 15s. As Figs. 10 (a)-(d) shown, the voltages of other
attacks, and achieve frequency and voltage stabilization within DGs show a downward trend and the system frequency can
a few seconds, while achieving accurate power sharing. In continue to keep in synchronous. The frequency and active
addition, comparing the simulation results of the two attacks, power of each DG have almost no change, and the reactive
it can be seen that the voltage and active power of the system power of each DG shows a different trend during the attack,

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:28:33 UTC from IEEE Xplore. Restrictions apply.
 LU et al.: DISTRIBUTED OBSERVER-BASED FINITE-TIME CONTROL OF AC MICROGRID UNDER ATTACK 165

Fig. 9. Case III-A: (a) Frequencies of DGs, (b) Voltages of DGs, (c) Active
power of DGs, (d) Reactive power of DGs.

Fig. 11. Case IV-A with the distributed attack-resilient cooperative control:
(a) Frequencies of DGs, (b) Voltages of DGs, (c) Confidence levels, (d) Trust
levels; Case IV-A with the distributed observer-based finite-time cooperative
control: (e) Confidence levels, (f) Trust levels.

the proposed method and the distributed attack-resilient coop-

erative control proposed in [10] is performed here. Two
different frequency attacks are simulated and compared here
as frequency attacks have a greater impact on microgrid.
1) Case IV-A (Frequency Link Attack): The simulation set-
tings are the same as Case II-A. Fig. 11(a)-(b) show the
Fig. 10. Case III-B: (a) Frequencies of DGs, (b) Voltages of DGs, (c) Active
power of DGs, (d) Reactive power of DGs.
performance of the attack-resilient distributed control, where
the frequencies and voltages of all DGs are asymptotically
stable over infinite time which is more than 25s, while the
but it can quickly return to the normal state after the attack is proposed method can converge the system frequency and volt-
removed. age to the reference value within 10s as seen in Case II-A. The
For node attacks, it is clear that frequency node attack has levels of confidence factors and trust factors under the two
a worse impact on microgrids than voltage node attack. The methods are shown in figures Fig. 11(c)-(d) and Fig. 11(e)-(f),
agent loses its control ability because the whole controller the threshold of the trust factor was set as Tij < 0.5, it can
is hijacked. At this time, the best solution is to remove the be seen from Fig. 11(d) and Fig. 7(f) that the information
information of the damaged node from the consensus proto- transmitted by the compromised communication link would
col and conduct consensus tracking control for the remaining be ignored once the corresponding trust factor lower than the
intact agents. Although the attack cannot be eliminated and the threshold. It is clearly to see the superiority of the proposed
frequency or voltage of DG are abnormal, the proposed con- approach in convergence speed by comparing the simulation
trol protocol already provides the best solution for this attack results. The steady speed of a system under attack is an
from the distributed control perspective. important index for operating performance. Consequently, the
proposed control approach can effectively respond to such
attacks and significantly improve the resilience and stability
D. Case IV: Comparing the Proposed Method With the of islanded AC microgrid.
Method in [30] 2) Case IV-B (Frequency Node Attack): The simulation
To verify the superiority of the observer-based finite-time settings are the same as Case II-B. The responses shown
controller proposed in this article on the convergence speed of in Fig. 12(a)-(d) reflect the performance of the distributed
the microgrid under attack, a simulation comparison between attack-resilient control. The efficiency of the observer-based

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:28:33 UTC from IEEE Xplore. Restrictions apply.
 166 IEEE TRANSACTIONS ON SMART GRID, VOL. 12, NO. 1, JANUARY 2021

Fig. 13. Experimental AC Microgrid Setup. (a) The AC microgrid with four
DGs; (b) Communication topology for four DGs.
Fig. 12. Case IV-B with the distributed attack-resilient cooperative control:
(a) Frequencies of DGs, (b) Voltages of DGs, (c) Confidence levels, (d) Trust
levels; Case IV-B with the distributed observer-based finite-time cooperative
control: (e) Confidence levels, (f) Trust levels.

finite-time control is verified in Fig. 12(e)-(f). Clearly, the

observer-based finite-time control can obviously accelerate the
convergence process which is obvious by comparing the con-
fidence factor and trust factor in the results. The proposed
algorithm can ensure the frequency converges in the set-
tling time which is an important property for the microgrid
operation.

E. Experimental Evalution
The proposed strategy is also verified experimentally by
using an AC microgrid comprising four DGs with the commu-
nication topology are shown in Fig. 13. The primary system
was built using lithium iron phosphate battery, inverter and Fig. 14. Experimental evalution of frequency link attack: (a) Frequencies of
load. The control algorithm was written based on MATLAB- DGs, (b) Voltages of DGs, (c) mP of DGs, (d) nQ of DGs.
Simulink software and ran in the real-time digital control
system rtunit. The signal acquisition was realized through
the connection of I/O port on the junction box. The on-line stability of the system and the distributed observer-based
debugging of the program and the real-time observation of finite-time control can realize the rapid recovery of the system.
the running results can be carried out on the upper com- The frequencies and voltages of all DGs can be restored within
puter, and the output analog signals can also be observed the setting time, while realizing accurate power sharing.
through the oscilloscope. The setting of controller parameters A frequency node attack with f2 = 50.1Hz attacks the AC
and reference values are the same as used in previous stud- microgrid system between t = 5s and t = 15s. Fig. 15 show
ies. Converter parameters are L=0.6mH, C=1.5mF. The local the performance of the attack-resilient distributed control in
loads are R1 = R2 = 10, C1 = 340μF. the experimental evalution, where the frequencies and volt-
A frequency link attack with amplitude of 0.1Hz as false ages of all DGs can converge to their reference values within
data is initiated at about 5 s. The effect of frequency link 15s. Fig. 15(c)-(d) show that the active and reactive sharing of
FDIA on prototype AC microgrid system are shown in Fig. 14. the system can be quickly achieved even when the frequency
It can be seen that the emergence of attack destroys the of DG2 is abnormal under the power controller. After the

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:28:33 UTC from IEEE Xplore. Restrictions apply.
 LU et al.: DISTRIBUTED OBSERVER-BASED FINITE-TIME CONTROL OF AC MICROGRID UNDER ATTACK 167

[8] J. W. Simpson-Porco, Q. Shafiee, F. Dörfler, J. C. Vasquez,

J. M. Guerrero, and F. Bullo, “Secondary frequency and voltage con-
trol of islanded microgrids via distributed averaging,” IEEE Trans. Ind.
Electron., vol. 62, no. 11, pp. 7025–7038, Nov. 2015.
[9] Z. Wang and W. Jie, “A practical distributed finite-time control scheme
for power system transient stability,” IEEE Trans. Power Syst., vol. 35,
no. 5, pp. 3320–3331, Sep. 2020.
[10] F. Guo, C. Wen, J. Mao, and Y. Song, “Distributed secondary volt-
age and frequency restoration control of droop-controlled inverter-based
microgrids,” IEEE Trans. Ind. Electron., vol. 62, no. 7, pp. 4355–4364,
Jul. 2015.
[11] T. Zhao, Z. Li, and Z. Ding, “Consensus-based distributed optimal
energy management with less communication in a microgrid,” IEEE
Trans. Ind. Informat., vol. 15, no. 6, pp. 3356–3367, Jun. 2019.
[12] W. Liu, W. Gu, W. Sheng, X. Meng, S. Xue, and M. Chen, “Pinning-
based distributed cooperative control for autonomous microgrids under
uncertain communication topologies,” IEEE Trans. Power Syst., vol. 31,
no. 2, pp. 1320–1329, Mar. 2016.
[13] L. Ding, Q. Han, L. Y. Wang, and E. Sindi, “Distributed cooperative
optimal control of DC microgrids with communication delays,” IEEE
Trans. Ind. Informat., vol. 14, no. 9, pp. 3924–3935, Sep. 2018.
Fig. 15. Experimental evalution of frequency node attack: (a) Frequencies [14] N. M. Dehkordi, H. R. Baghaee, N. Sadati, and J. M. Guerrero,
of DGs, (b) Voltages of DGs, (c) mP of DGs, (d) nQ of DGs. “Distributed noise-resilient secondary voltage and frequency control
for islanded microgrids,” IEEE Trans. Smart Grid, vol. 10, no. 4,
pp. 3780–3790, Jul. 2019.
[15] Z. Wang and J Wang, “A delay-adaptive control scheme for enhancing
frequency node attack is removed, the change of the system smart grid stability and resilience,” Int. J. Elect. Power Energy Syst.,
state will also affect the power sharing, and the power con- vol. 110, pp. 477–486, Mar. 2019.
troller can quickly realize the accurate power sharing of the [16] X. Wang, H. Zhang, and C. Li, “Distributed finite-time cooperative
control of droop-controlled microgrids under switching topology,” IET
system in the new operating state. Renew. Power Gener., vol. 11, no. 5, pp. 707–714, Dec. 2017.
[17] S. Zuo, A. Davoudi, Y. Song, and F. L. Lewis, “Distributed finite-time
VI. C ONCLUSION voltage and frequency restoration in islanded AC microgrids,” IEEE
Trans. Ind. Electron., vol. 63, no. 10, pp. 5988–5997, Oct. 2016.
The islanded AC microgrid distributed synchronization [18] S. Sahoo and S. Mishra, “A distributed finite-time secondary average
under attack is solved in this article. Based on an observer voltage regulation and current sharing controller for DC microgrids,”
IEEE Trans. Smart Grid, vol. 10, no. 1, pp. 282–292, Jan. 2019.
with confidence factors and trust factors, an observer-based [19] Y. Hu, X. Wang, Y. Peng, J. Xiang, and W. Wei, “Distributed finite-
finite-time stability control method is proposed. The frequency time secondary control for DC microgrids with virtual impedance
and voltage convergence speed of microgrid under attack are arrangement,” IEEE Access, vol. 7, pp. 57060–57068, 2019.
greatly accelerated by the designed controller. Moreover, the [20] N. M. Dehkordi, N. Sadati, and M. Hamzeh, “Distributed robust finite-
time secondary voltage and frequency control of islanded microgrids,”
controller can limit the impact of cyber-attacks on the dynam- IEEE Trans. Power Syst., vol. 32, no. 5, pp. 3648–3659, Sep. 2017.
ics of the microgrid. We only discuss the finite-time control [21] J. Hu and A. Lanzon, “Distributed finite-time consensus control
strategy under cyber-attacks, and cannot completely eliminate for heterogeneous battery energy storage systems in droop-controlled
microgrids,” IEEE Trans. Smart Grid, vol. 10, no. 5, pp. 4751–4761,
certain types of attacks, so we would focus on more power- Sep. 2019.
ful attenuation techniques and control methods for physical [22] Z. Wang and J. Wang, “A novel finite-time control scheme for enhancing
attacks in our future research. smart grid frequency stability and resilience,” IEEE Trans. Smart Grid,
vol. 10, no. 6, pp. 6538–6551, Nov. 2019.
[23] Y. Xu and H. Sun, “Distributed finite-time convergence control of an
R EFERENCES islanded low-voltage AC microgrid,” IEEE Trans. Power Syst., vol. 33,
no. 3, pp. 2339–2348, May 2018.
[1] Y. Han, K. Zhang, H. Li, E. A. A. Coelho, and J. M. Guerrero, “MAS-
[24] Z. Deng, Y. Xu, H. Sun, and X. Shen, “Distributed, bounded and
based distributed coordinated control and optimization in microgrid and
finite-time convergence secondary frequency control in an autonomous
microgrid clusters: A comprehensive overview,” IEEE Trans. Power
microgrid,” IEEE Trans. Smart Grid, vol. 10, no. 3, pp. 2776–2788,
Electron., vol. 33, no. 8, pp. 6488–6508, Aug. 2018.
May 2019.
[2] X. Lu, X. Yu, J. Lai, Y. Wang, and J. M. Guerrero, “A novel distributed
[25] O. A. Beg, L. V. Nguyen, T. T. Johnson, and A. Davoudi, “Signal tem-
secondary coordination control approach for islanded microgrids,” IEEE
poral logic-based attack detection in DC microgrids,” IEEE Trans. Smart
Trans. Smart Grid, vol. 9, no. 4, pp. 2726–2740, Jul. 2018.
Grid, vol. 10, no. 4, pp. 3585–3595, Jul. 2019.
[3] M. Yazdanian and A. Mehrizi-Sani, “Distributed control techniques in
microgrids,” IEEE Trans. Smart Grid, vol. 5, no. 6, pp. 2901–2909, [26] S. Sahoo, S. Mishra, J. C. Peng, and T. Dragièević, “A stealth cyber-
Nov. 2014. attack detection strategy for DC microgrids,” IEEE Trans. Power
[4] A. Nisar and M. S. Thomas, “Comprehensive control for microgrid Electron., vol. 34, no. 8, pp. 8162–8174, Aug. 2019.
sutonomous operation with demand response,” IEEE Trans. Smart Grid, [27] O. A. Beg, T. T. Johnson, and A. Davoudi, “Detection of false-
vol. 8, no. 5, pp. 2081–2089, Sep. 2017. data injection attacks in cyber-physical DC microgrids,” IEEE Trans.
[5] J. M. Guerrero, M. Chandorkar, T. Lee, and P. C. Loh, “Advanced Ind. Informat., vol. 13, no. 5, pp. 2693–2703, Oct. 2017.
control architectures for intelligent microgrids—Part I: Decentralized [28] D. Jin et al., “Toward a cyber resilient and secure microgrid using
and hierarchical control,” IEEE Trans. Ind. Electron., vol. 60, no. 4, software-defined networking,” IEEE Trans. Smart Grid, vol. 8, no. 5,
pp. 1254–1262, Apr. 2013. pp. 2494–2504, Sep. 2017.
[6] Q. Shafiee, J. M. Guerrero, and J. C. Vasquez, “Distributed secondary [29] M. M. Rana, L. Li, and S. W. Su, “Cyber attack protection and control of
control for islanded microgrids—A novel approach,” IEEE Trans. Power microgrids,” IEEE/CAA J. Automatica Sinica, vol. 5, no. 2, pp. 602–609,
Electron., vol. 29, no. 2, pp. 1018–1031, Feb. 2014. Mar. 2018.
[7] J. Lai, H. Zhou, X. Lu, X. Yu, and W. Hu, “Droop-based distributed [30] S. Abhinav, H. Modares, F. L. Lewis, F. Ferrese, and A. Davoudi,
cooperative control for microgrids with time-varying delays,” IEEE “Synchrony in networked microgrids under attacks,” IEEE Trans. Smart
Trans. Smart Grid, vol. 7, no. 4, pp. 1775–1789, Jul. 2016. Grid, vol. 9, no. 6, pp. 6731–6741, Nov. 2018.

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:28:33 UTC from IEEE Xplore. Restrictions apply.
 168 IEEE TRANSACTIONS ON SMART GRID, VOL. 12, NO. 1, JANUARY 2021

[31] X. Lu, R. Lu, S. Chen, and J. Lu, “Finite-time distributed tracking con- Jie Wang received the Ph.D. degree in indus-
trol for multi-agent systems with a virtual leader,” IEEE Trans. Circuits trial automation from Northeast University, China,
Syst. I, Reg. Papers, vol. 60, no. 2, pp. 352–362, Feb. 2013. in 1998. He is a Professor of power system and
[32] X. Lin and Y. Zheng, “Finite-time consensus of switched multiagent its automation with the Department of Electrical
systems,” IEEE Trans. Syst., Man, Cybern., Syst., vol. 47, no. 7, Engineering, Shanghai Jiao Tong University, where
pp. 1535–1545, Jul. 2017. he is also with a lab for the development of Flexible
[33] M. Basseville and I. V. Nikiforov, Detection of Abrupt Changes: Theory ac/dc transmission system technology and intelligent
and Application, vol. 104. Englewood Cliffs, NJ, USA: Prentice-Hall, control. His research interests include stability anal-
1993. ysis and control of complex power system and smart
grids. His recent research interest is to establish
a series of theoretical analysis results of nonlinear
differential algebraic control systems and apply these results to the nonlinear
control of structure preserving power systems. Based on the differential alge-
braic Hamiltonian system and its periodic solution theory, the low frequency
oscillation analysis and control method of interconnected power grid are
presented. The symmetry and conserved quantity in Noether’s theory are
applied to restrict the existence of periodic solutions of power system.

Rentao Lu received the B.S. degree in electri- Ziqiang Wang received the M.S. degree from
cal engineering from Nanchang University, China, the School of Electrical Information and Electronic
in 2018. He is currently pursuing the master’s Engineering, Shanghai Jiao Tong University, in
degree with the School of Electrical Information 2020, where he is currently pursuing the Doctoral
and Electronic Engineering, Shanghai Jiao Tong degree. His research interests include power system
University, China. His research interests include nonlinear control, wide-area control, and finite-time
control and stability analysis of microgrid and finite- control.
time control.

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:28:33 UTC from IEEE Xplore. Restrictions apply.