IEEE TRANSACTIONS ON SMART GRID, VOL. 13, NO.

6, NOVEMBER 2022 4837

Probability-Density-Dependent Load Frequency

Control of Power Systems With Random Delays
and Cyber-Attacks via Circuital Implementation
Shen Yan , Zhou Gu , Member, IEEE, Ju H. Park , Senior Member, IEEE, Xiangpeng Xie , Member, IEEE,
and Chunxia Dou , Senior Member, IEEE

Abstract—This paper studies the decentralized H∞ secure load I. I NTRODUCTION

frequency control issue and circuital realization of multi-area net-
S WE known, there are basically three layers of the
worked power systems subject to random transmission delays and
deception attacks. To make full use of the stochastic feature of the
network-induced transmission delay, its distribution described by
A control framework of power systems, which are primary
control (conventional droop control), secondary control (volt-
the probability density function is utilized. Due to this feature, age/frequency regulation), and tertiary control (economic dis-
the normal control signal and the injected deceptive attack signal patch and power flow optimization) [1]. As a representative
transmitted over the network can be formed as two distributed
delay terms. Then, the ith load frequency control area is estab- secondary control, load frequency control (LFC) is a critical
lished as a new distributed delay system, in which the delay strategy to maintain the frequency stable for power systems
probability density is treated as the distributed kernel. By utiliz- and much effort has been devoted to studying it [2]–[8].
ing an integral inequality dependent on the kernel, new sufficient In a multi-area power system, the centralized strategy of
controller design conditions are derived to guarantee the system
stability with given H∞ performance. Moreover, a physical execu-
LFC utilizing global system information and the distributed
tion approach is addressed to transfer the load frequency control control method using the neighbor subsystem information
systems into electrical analogy circuits. The effectiveness of the are effective strategies to regulate the system frequency.
proposed approach is illustrated via the professional simcape Nevertheless, the utilization of power interchange through
toolbox in Simulink/MATLAB for circuit simulations. tie-lines among subsystems increases the complexity and dif-
Index Terms—Networked power systems, random transmis- ficultly of frequency regulation. Compared with them, the
sion delays, cyber attacks, load frequency control, circuital decentralized LFC scheme investigated in [9], [10] is more
implementation.
practical and easier for implementation because only the local
area information is utilized to attenuate the frequency devia-
tion. In recent years, as the trend of the integration of physical
power systems and communication network, networked power
systems (NPSs) have received much attention from a lot of
Manuscript received 28 January 2022; revised 8 April 2022; accepted researchers.
24 May 2022. Date of publication 30 May 2022; date of current version With the introduction of communication network, the
21 October 2022. This work was supported in part by the National Natural
Science Foundation of China under Grant 62103193 and Grant 62022044; in
network-induced delay is a common phenomenon, which
part by the Natural Science Foundation of Jiangsu Province of China under could degrade the system performance, even make system
Grant BK20200769; in part by the Natural Science Foundation of Jiangsu unstable [11]. However, some published results of LFC
Provincial Universities under Grant 20KJB510045; in part by the Project
funded by China Postdoctoral Science Foundation under Grant 2021TQ0155; systems using proportional-integral (PI) controller have not
in part by the Jiangsu Natural Science Foundation for Distinguished Young considered the network communication delay [12]. To tackle
Scholars under Grant BK20190039. The work of Ju H. Park was supported this problem, a decentralized LFC method for power systems
by the National Research Foundation of Korea (NRF) grant funded by the
Korea Government (MSIT) under Grant 2020R1A2B5B02002002. Paper no. with multi constant communication delays is developed
TSG-00141-2022. (Corresponding authors: Zhou Gu; Ju H. Park.) in [13], where the constant delays are treated as multiplicative
Shen Yan and Zhou Gu are with the College of Mechanical and Electronic uncertainties. Reference [14] addresses a new decentralized
Engineering, Nanjing Forestry University, Nanjing 210037, China (e-mail:
[email protected]; [email protected]). sliding mode LFC scheme for multi-area NPSs with con-
Ju H. Park is with the Department of Electrical Engineering, Yeungnam stant delay and wind power integration. Since the situation
University, Gyeongsan 38541, South Korea (e-mail: [email protected]). of communication network is usually uncertain and varying,
Xiangpeng Xie is with the Institute of Advanced Technology, Nanjing
University of Posts and Telecommunications, Nanjing 210023, China, and also a time-varying delay model is more practical than the con-
with the School of Information Science and Engineering, Chengdu University, stant one. Due to this fact, some delay-dependent controller
Chengdu 610106, China (e-mail: [email protected]).
Chunxia Dou is with the Institute of Advanced Technology for
design conditions are derived for LFC systems with time-
Carbon Neutrality, Nanjing University of Posts and Telecommunications, varying communication delays [15]. In [16], a power system
Nanjing 210023, China (e-mail: [email protected]). model with multiple time-varying communication delays is
Color versions of one or more figures in this article are available at
https://doi.org/10.1109/TSG.2022.3178976.
established and delay-dependent conditions for designing H∞
Digital Object Identifier 10.1109/TSG.2022.3178976 controller are obtained. Most of the existing outcomes, for
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
4838 IEEE TRANSACTIONS ON SMART GRID, VOL. 13, NO. 6, NOVEMBER 2022

instance [15], [16], are accomplished based on the interval introduced in developing the model and obtaining the stability
time-varying delay method and boundary information of the and synthesis conditions.
communication delay. It is worthy noting that the time-varying 1) Since the existing methods [15], [25], [26] based on an
communication delays in [15], [16] are assumed to be with a interval time-varying delay model do not take into account
uniform distribution. Practically, for a real TCP/IP-based com- the probability distribution of random delay, how to model the
munication network, communication delay often has stochastic random delay by making use of the distribution information
properties. Generally speaking, it is usually distributed with is the first challenging problem.
some probabilistic features [17]–[19]. Thus, the uniform dis- 2) Considering the first challenging problem, a new dis-
tribution assumption without using these features may yield tributed delay model with the kernel representing the delay
conservative results. Considering the inter features of the distribution will be proposed in this paper. However, the exist-
communication delay in system analysis/synthesis, less con- ing method [29] based on Legendre polynomials to handle this
servative conditions are expected to be obtained. As a fact, the new model will cause approximation error and further lead to
probability density is more specific and precise for depicting conservative results. Thus, how to remove such approximation
the random delay. error and reduce conservativeness is the second challenging
On the other hand, the security of communication network problem.
that is vulnerable to be wrecked by hackers is very impor- 3) There exist stochastic transmission delays when the
tant for NPSs. If the remote control signals transmitted over attackers inject attack signals into network, therefore, how to
the network are injected by some false data, the NPS will model and deal with the deception attack will also bring the
be difficult to be maintained at the nominal state. Over the challenging problem similar with handling the random delay.
past decade, much attention has been paid to the control To solve these challenging issues, the main contributions of
problem against cyber-attacks, such as Denial-of-service (DoS) this work are given below:
attack [21], [22], deception attack [23], [24]. Concretely speak- 1) To make efficient use of the stochastic feature of the
ing, a resilient event-triggered LFC approach for NPSs subject network-induced transmission delay, a novel distributed delay
to DoS attack with finite energy is addressed in [20]. Different model with kernel is established, where the kernel is used
from DoS attack by blocking the connections of nets, decep- to represent the delay probability density. To model the ran-
tion attack aims to change the real value of control signal dom deception attacks, not only the stochastic feature induced
by injecting a false data, which makes it more deceptive. by attacker, but also the stochastic feature caused by random
Considering multi-area NPSs with time-varying communica- delays are considered in this paper. Then, a new representa-
tion delay and deception attack, a resilient LFC approach is tion of random delay and deception attack is proposed for the
addressed in [25], in which the deception attack is described first time. This model is more general than the traditional one
as a time-varying delay-dependent function. Reference [26] based on interval time-varying delay method [15], [25], [26].
studies the H∞ LFC issue of multi-area NPSs with DoS 2) In the existing result [29], the distributed delay with
attack and deception attack, which is also modeled as a kernel is approximated by Legendre polynomials and han-
time-varying delay-dependent function. In [25], [26], the time- dled by Bessel-Legendre inequality, which could lead to
varying delay-dependent deception attack model is established approximation error and design conservativeness. To solve this
only by using the bounds of delays. Compared to the con- problem, the distributed kernel is utilized directly to construct
ventional interval time-varying delay method, using not only a new Lyapunov-Krasovskii functional (LKF) in this paper.
the upper and lower bounds, but also the delay probability An integral inequality based on the kernel is used to deal with
distribution is potential for obtaining less conservative results. the distributed delay term directly without any approximation
It is known that testing the effectiveness of control strategies error. Then, new sufficient and less conservative conditions
in some practical systems is costly and difficult. Therefore, it formed by linear matrix inequalities (LMIs) are derived to
is worthy to develop suitable hardware simulators for some design the decentralized H∞ secure controller.
real control systems. Due the fact that many control devices The organization of this paper is provided as follows. The
of practical systems are executed by electrical equipments, it is preliminaries of modeling the closed-loop NPS is given in
natural and reasonable to transform the physical systems into Section II. Then the stability and controller design conditions
analogy circuits to verify the validity of control method and are obtained in Section III. Some simulation results are imple-
save the testing cost. For instance, a circuital implementation mented in Section IV. Section V summaries the conclusions
method is studied to convert the robust adaptive fault tolerant and presents some future investigations.
control schemes into electrical circuits in [27]. For uncertain Notation: In this paper, we define He(A) = A + AT . ⊗
nonlinear multi-agent systems with actuator faults, the adaptive a
means the Kronecker product. Sy(A, B)  BT AB. denotes
control issue is developed and its circuital realization is studied b
in [28]. a!
.
Based on the above motivations, this article addresses the (a − b)!b!
H∞ load frequency regulation of NPSs with random delays
and cyber-attacks in a decentralized method. A circuital real-
ization approach is proposed to simulate the dynamics of II. P RELIMINARIES
power system components and the controller via equivalent In the LFC problem of power systems, there exist some
analogy circuits. The following new challenging problems are complicate nonlinear dynamics. Due to the fact that the load
YAN et al.: PROBABILITY-DENSITY-DEPENDENT LFC OF POWER SYSTEMS 4839

in which zi (t) is the performance output and

⎡ Di ⎤
−M − M1 i 0 1
Mi 0 ⎡ ⎤
⎢  n
i
⎥ 0
⎢ Tij 0 0 0 0⎥ ⎢ 0 ⎥
⎢ ⎥ ⎢ ⎥
⎢j=1,j=i ⎥ ⎢ 1 ⎥
Ai = ⎢ ⎥, Bi = ⎢ Tgi ⎥,
⎢ − Rgi1Tgi 0 − T1gi 0 0⎥ ⎢ ⎥
⎢ ⎥ ⎣ 0 ⎦
⎣ 0 0 1
− T1ti 0⎦
Tti 0
βi 1 0 0 0
Fig. 1. The ith control area of a multi-area LFC NPS.    
βi 1 0 0 0 − M1 i 0 0 0 0
Ci = , FiT = .
0 0 0 0 1 0 −1 0 0 0
variations of power system are usually small, it is permissi- In order to keep the frequency fi (t) at the prescribed
ble and useful to linearize them around the operating point value, the following decentralized LFC scheme based on the
[30], [31]. In many existing results [6]–[8], [31], the dynam- conventional PI control is utilized:

ics of governor, turbine and power generator are linearized as
the following first-order transfer functions: ui (t) = KPi ACEi (t) + KIi ACEi (t)dt = Ki yi (t), (4)

⎧ where Ki = [KPi KIi ] and KPi , KIi are the PI controller gains
⎪
⎨ Governor : Ggi (s) = 1+sTgi ,
1
to be designed.
Turbine : Gti (s) = 1+sTti ,
1 When the control signal is transmitted over the communi-
⎪
⎩ Power generator : cation network, it is highly possible that the transmitted signal
Gpi (s) = Di +sMi .
1
is delayed and damaged by sudden congestion and malicious
attackers. Considering these factors, the output of controller
where Tgi , Tti , Di , Mi and Rgi are the governor and tur-
with delay and deception attack is represented as
bine time constants, damping coefficient, inertia moment
of generator and droop coefficient of the ith control area, uai (t) = Ki Ci xi (t − τ (t)) + λ(t)Ki ϕi (t − τ (t)), (5)
respectively.
where τ (t) ∈ [0, h] means the random transmission delay; h
In this paper, following the above linearization method,
is the upper bound of the delay; λ(t) ∈ {0, 1} is a Bernoulli
the structure of the linear dynamic model of the ith control
variable to describe the occurrence of deception attack by
area in a multi-area NPS is illustrated in Fig. 1. Let fi (t),
λ(t) = 1 and no attack λ(t) = 0; the mathematical expec-
Ptie−i (t), Pgi (t), Pti (t) and Pdi (t) represent the devi-
tations of λ(t) = 1 and λ(t) = 0 are E{λ(t) = 1} = λ̄ and
ations of system frequency, tie-line power exchange, valve
E{λ(t) = 0} = 1 − λ̄, respectively.
position, mechanical output of turbine and load disturbance
In some existing results [8], [26], the deception attack is
of the ith control area, respectively. Then, the differential
usually bounded and modeled by a Lipschitz nonlinear func-
equations to describe the system dynamics of the ith control
tion related with system state or output. Following the same
area [6], [25] are given as:
way in [8], [26], the deception signal ϕi (t) is assumed to satisfy
⎧
⎪ ḟi (t) = M1 i (Pti (t) − Pdi (t) ϕi (t)2 ≤ Hi Ci xi (t)2 , (6)
⎪
⎪
⎪
⎪
⎨ −Ptie−i (t) −
⎪ D f (t))
 i i
where Hi is a given constant matrix used to describe the upper
Ṗtie−i (t) =  nj=1,j=i Tij fi (t) − fj (t)  . (1) bound of deception attack.
⎪
⎪ It is noted that the random transmission delay τ (t) usu-
⎪
⎪ Ṗgi (t) = T1gi ui (t) − R1gi fi (t) − Pgi (t)
⎪
⎩ Ṗ (t) = 1 P (t) − P (t)

⎪ ally meets some probability distributions. For convenience,

ti Tti gi ti we abbreviate τ (t) as τ . For example, based on the described
scenarios in [18], [32] to simulate the network channel, the
The ith area control error signal used to maintain zero communication delay with upper bound h = 0.2s and its
steady-state error for frequency deviation is defined as normalized distribution histogram are drawn in Fig. 2.
To utilize the random information of delay τ , its probabil-
h
ACEi (t) = βi fi (t) + Ptie−i (t), (2) ity density described by the kernel g(·) with 0 g(s)ds = 1
is considered, which can be accessible by statistical method.
where βi denotes the frequency bias factor. In Fig. 2, a probability density function g(τ ) is used to
Selecting the same state, output and disturbance approximate the probability distribution, which is chosen as
in [25] as
xiT (t) = [fi (t), Ptie−i (t), Pgi (t), Pti (t), ACEi (t)dt]T , g(τ ) = 2500τ e−50τ , τ ∈ [0, h].
yi (t) = [ACE
ni (t) ACEi (t)dt]T and ωi (t) = [Pdi (t) i (t)]T
(i (t)  j=1,j=i Tij fj (t)), one can represent the ith LFC Then, we define g0 (s) = g(s) and construct a new vector
system in the form of state-space by: g(s) = [g0 (s), g1 (s), . . . , gi (s), . . . , g (s)]T with ∈ N, where
all the elements gi (s) are independent. According to [33], the
⎧ basic principle of choosing gi (s) is that the chosen gi (s) should
⎨ ẋi (t) = Ai xi (t) + Fi ωi (t) + Bi ui (t)
make the vector g(s) satisfy the property:
yi (t) = Ci xi (t) , (3)
⎩
zi (t) = Ci xi (t) dg(s)/ds = Gg(s), (7)
4840 IEEE TRANSACTIONS ON SMART GRID, VOL. 13, NO. 6, NOVEMBER 2022

Lemma 1 [33]: For a symmetric matrix Y > 0 ∈ Rn×n and

the vector g(s) given in (7), it gives
 a2   a2 
Sy(Y, x(s))ds ≥ Sy  ⊗ Y, G(s)x(s)ds (10)
a1 a1
a2
with −1 = a1 g(s)gT (s)ds
> 0, G(s) = g(s) ⊗ In .
Remark 3: The Bessel-Legendre inequality in [29] uses
Legendre polynomials to approximate the probability density,
which leads to approximation error and further yields more
conservative results. Different from this method, our method
based on (10) can handle the probability density directly with-
out any approximation error. Thus, less conservative results are
obtained by the integral inequality (10) in Lemma 1 than the
Bessel-Legendre inequality in [29].
Fig. 2. Random delay τ (t) and its normalized histogram.
III. M AIN R ESULTS
To derive the controller design conditions, the Lyapunov
which means the elements gi (s) are the solutions of linear method is employed and the following Lyapunov-Krasovskii
homogeneous differential equations with constant coefficients functional (LKF) is constructed as
in the matrix G ∈ R × .

3
By utilizing the delay distribution information, the stochas- V(t) = Vk (t) (11)
tic delayed signal xi (t − τ ) is modeled by a deterministic, k=1
0
distributed delay model −h g(s)xi (t + s)ds, which can be
where
viewed as the mathematical expectation of xi (t − τ ). Then,  T
uai (t) with the probability density of stochastic delays is V1 (t) = ξiT (t)Pi ξi (t), ξi (t) = xiT (t) GTxi (t) GTϕi (t) ,
expressed as:  t
 0  0 V2 (t) = xiT (s)[S1i + (s − t + h)Y1i ]xi (s)ds,
uai (t) = Ki Ci g(s)xi (t + s)ds + λ(t)Ki g(s)ϕi (t + s)ds. t−h
−h −h t
(8) V3 (t) = diT (s)[S2i + (s − t + h)Y2i ]ϕi (s)ds,
t−h
 0  0
Based on (8), the ith closed-loop LFC cyber-physical power
system becomes Gxi (t)  G(s)xi (t + s)ds, Gϕi (t)  G(s)ϕi (t + s)ds,
−h −h
⎧ 0
⎪
⎨ ẋi (t) = Ai xi (t) + Bi Ki Ci −h g(s)xi (t + s)ds G(s)  g(s) ⊗ In , g0 (s) = g(s), g(s) = [g0 (s), . . . , g (s)]T .
0
⎪ +λ(t)Bi Ki −h g(s)ϕi (t + s)ds + Fi ωi (t) (9) Based on G(s), Gxi (t) and Gϕi (t) defined in (11), one has
⎩ z (t) = C x (t).
i i i  0
Remark 1: In this work, a new model of LFC system with g(s)xi (t + s)ds = I1 Gxi (t), (12)
−h
random communication delays is established as a distributed  0
delay system with kernel. In contrast to the conventional model g(s)ϕi (t + s)ds = I2 Gϕi (t), (13)
based on an interval time-varying delay system only using −h
the upper and lower bounds of delay, the delay probability with I1 = [ I5 05,5 ] and I2 = [ I2 02,2 ].
density is taken into consideration in system modeling. Then, By substituting (12) and (13) into (9), the ith closed-loop
the presented distributed delay modeling approach has the power system is further expressed as
potential to obtain less conservative results. ⎧
Remark 2: To describe the feature induced by random ⎨ ẋi (t) = Ai xi (t) + Bi Ki Ci I1 Gxi (t)
behaviors of adversaries, a Bernoulli variable λ(t) is adopted +λ(t)Bi Ki I2 Gϕi (t) + Fi ωi (t) (14)
⎩
to represent the control signal is attacked or not. Besides, the zi (t) = Ci xi (t)
random feature caused by the delayed deceptive signal is also To simplify the derivation, ζi (t) and Ia are defined as
utilized in (8), which is not considered in [25], [26]. Thus,
ζi (t) = [ζ1i (t)ζ2i (t)]T , (15)
the presented model of deception attacks is more general than
these existing results. where
This article aims to design the controller (8) such that  
1) The system (9) is mean-square asymptotically stable for ζ1i (t) = ẋiT (t)xiT (t)xiT (t − h)ϕiT (t)ϕiT (t − h) ,
 
ωi (t) = 0; ∞ ζ2i (t) = GTxi (t)GTϕi (t)ωiT (t)zTi (t) .
∞
2) E{ 0 zTi (t)zi (t)dt} < γi2 E{ 0 ωiT (t)ωi (t)dt} holds for
ωi (t) = 0 and zero initial condition with a given γi > 0. Then, some sufficient H∞ stability analysis criterions for
Before further proceeding, a useful lemma is given as below. the system (14) are provided in the following theorem.
YAN et al.: PROBABILITY-DENSITY-DEPENDENT LFC OF POWER SYSTEMS 4841

Theorem 1: For given parameters h, ρ and controller gain + Sy(S1i + hY1i , xi (t)) − Sy(S1i , xi (t − h))
Ki , the system (14) is mean-square asymptotically stable with + Sy(S2i + hY2i , ϕi (t)) − Sy(S2i , ϕi (t − h))
required H∞ index γi , if there exist symmetric matrices Pi ,  0  0
S1i > 0, Y1i > 0, S2i > 0, Y2i > 0, (i = 1, 2, . . . , n), − Sy(Y1i , xi (t + s))ds − Sy(Y2i , ϕi (t + s))ds.
0 < Zi < νI and matrix Xi such that −h −h
(24)
Pi > 0, (17)
Using Lemma 1 to handle the integral terms in (24) gives
i + i < 0, (18)
 0
0
for i = 1, 2, . . . , n, where  = ( −h g(s)gT (s)ds)−1 , − Sy(Y1i , xi (t + s))ds ≤ −Sy( ⊗ Y1i , Gxi (t)), (25)
−h
Pi = Pi + diag{0,  ⊗ S1i ,  ⊗ S2i )},  0
 

− Sy(Y2i , ϕi (t + s))ds ≤ −Sy  ⊗ Y2i , Gϕi (t) . (26)

i = i + He(X i G i ), i = Sy ν(Hi Ci )T Hi Ci , I2 , −h
 
i = He JT Pi Q + Sy(S1i + hY1i , I2 ) − Sy(S1i , I3 ) Based on Assumption (6) and 0 < Zi < νI, one gets

− Sy( ⊗ Y1i , Ix ) + Sy(S2i + hY2i − Zi , I4 ) νxiT (t)(Hi Ci )T Hi Ci xi (t) − ϕiT (t)Zi ϕi (t) ≥ 0. (27)

− Sy(S2i , I5 ) − Sy  ⊗ Y2i , Iϕ − γi2 Sy I, I8+2 According to (15), (22) and (23), it results in

 
− Sy I, I9+2 + He IT9+2 Ci I2 ,
ξi (t) = Jζi (t), ξ̇i (t) = Qζi (t). (28)
G i = −I1 + Ai I2 + Bi Ki Ci I1 Ix + λ̄Bi Ki I2 Iϕ + Fi I8+2 ,
Combining (24), (25), (26) and (27), we obtain
1 = G ⊗ I( +1)4 , G
X i = (μ1 Xi I1 +μ2 Xi I2 )T , G 2 = G ⊗ I( +1)2 ,
⎡ ⎤ ⎡ ⎤ V̇(t) − γi2 ωiT (t)ωi (t) + zTi (t)zi (t) ≤ Sy(i + i , ζi (t)). (29)
I2 I1
J = ⎣ Ix ⎦, Q = ⎣ G(0)I2 − G(−h)I3 − G 1 Ix ⎦,
From the definition of ζi (t) in (15), the system (14) can be
Iϕ G(0)I4 − G(−h)I5 − G 2 Iϕ expressed as
 T  T

Ix = IT6 · · · IT6+ , Iϕ = IT7+ · · · IT7+2 . −I1 + Ai I2 + Bi Ki Ci Ix + λ(t)Bi Ki Iϕ + Fi I8+2 ζi (t) = 0.

Proof: With the chosen LKF (11) for ith power system, by (30)
adopting Lemma 1, it leads to Taking the mathematical expectation of (30) yields
 0 

Sy(Y1i , xi (t + s))ds ≥ Sy( ⊗ Y1i , Gxi (t)), (19) E −I1 + Ai I2 + Bi Ki Ci Ix + λ̄Bi Ki Iϕ + Fi I8+2 ζi (t)
−h
 0 = E{G i ζi (t)} = 0. (31)

Sy(Y2i , ϕi (t + s))ds ≥ Sy  ⊗ Y2i , Gϕi (t) . (20)

−h
Based on X i = (μ1 Xi I1 + μ2 Xi I2 )T and (31), it leads to

In terms of (11), (19) and (20), one has E ζiT (t)X i G i ζi (t) = 0. (32)
 0
V(t) ≥ Sy(Pi , ξi (t)) + Sy(Y1i , xi (t + s))ds According to (29), (32) and i + i < 0 from (18), it gives
−h 
 0
E V̇(t) − γi2 ωiT (t)ωi (t) + zTi (t)zi (t)
+ Sy(Y2i , ϕi (t + s))ds. (21) 
−h
≤ E ζiT (t)(i + i )ζi (t) < 0. (33)
From Y1i > 0, Y2i > 0 and Pi > 0, V(t) > 0 is ensured.
According to (7), it gives By integrating (33) over [0, ∞), one has
! ∞   "
 xi (t), (22)
Ġxi (t) = G(0)xi (t) − G(−h)xi (t − h) − GG 2 T T
E{V(∞) − V(0)} < E γi ωi (t)ωi (t) − zi (t)zi (t) dt .
 ϕi (t). (23)
Ġϕi (t) = G(0)ϕi (t) − G(−h)ϕi (t − h) − GG 0
(34)
Calculating the time derivative of V(t) and then subtracting
γi2 ωiT (t)ωi (t) − zTi (t)zi (t) yield If ωi (t) = 0, one has E{V̇(t)} < 0 from (33). Hence the
system (9) is mean-square asymptotically stable. Moreover,
V̇(t) − γi2 ωiT (t)ωi (t) + zTi (t)zi (t) ∞
for ωi (t) = 0 and x(0) = 0, one gets E{ 0 zTi (t)zi (t))dt} ≤
∞
= 2ξiT (t)Pi ξ̇i (t) − γi2 ωiT (t)ωi (t) + zTi (t)zi (t) E{γi2 0 ωiT (t)ωi (t)dt} from (34).

⎧# $
⎪
⎪ 05,5(a−1) I5 05,5(3−a) 05,4 05,5(1+ ) 05,2(1+ ) 05,4 , a = 1, 2, 3
⎪
⎪ # $
⎪
⎨ #02,15 02,5(a−4) I2 02,5(5−a) 02,5(1+ ) 02,2(1+ ) 02,4 , a = 4, 5
$
Ia = 0 05,4 05,5(a−6) I5 05,5(6+ −a) 05,2(1+ ) 05,4 , a = 6, . . . , 6 + (16)
⎪
⎪ # 5,15 $
⎪
⎪ 0 02,4 02,5(1+ ) 02,2(a−7− ) I2 02,2(7+2 −a) 02,4 , a = 7 + , · · · , 7 + 2
⎪
⎩#
2,15
$
02,15 02,4 02,5(1+ ) 02,2(1+ ) 02,2(8+2 −a) I2 02,2(9+2 −a) , a = 8 + 2 , . . . , 9 + 2
4842 IEEE TRANSACTIONS ON SMART GRID, VOL. 13, NO. 6, NOVEMBER 2022

Remark 4: In terms of the condition (17), Pi > 0 is not

necessary to make the selected LKF (11) positive. Thus, this
relaxation could lead to less conservative stability conditions
than some existing conditions requiring Pi > 0.
According to the stability results in Theorem 1, the con-
troller design conditions of system (14) are derived in
Theorem 2.
Theorem 2: For given scalars μ1 , μ2 , h, ρ, the system (14)
is asymptotically stable in mean-square with required H∞
index γi , if there exist symmetric matrices P̃i , S̃1i > 0, Ỹ1i > 0,
S̃2i > 0, Ỹ2i > 0, Zi > ν −1 I and matrices Ni , Mi , Li such that
P̃i > 0, (35)
  Fig. 3. Circuital implementation of (a) governor; (b) turbine; (c) power
˜ i νIT Mi (Hi Ci )T
 generator and (d) droop constant for Area 1.
2 < 0, (36)
∗ −νI
 
−I (Ni Ci − Ci Mi )T
< 0, (37)
∗ −I
for i = 1, 2, . . . , n, where

P̃i = P̃i + diag 0,  ⊗ S̃1i ,  ⊗ S̃2i ) ,
     
˜ i = He JT P̃i Q + Sy S̃1i + hỸ1i , I2 − Sy S̃1i , I3

 
− Sy  ⊗ Ỹ1i , Ix + Sy S̃2i + hỸ2i + ρ 2 Zi − 2ρNi , I4 Fig. 4. Circuital implementation of the controller for Area 1.
 

− Sy S̃2i , I5 − Sy  ⊗ Ỹ2i , Iϕ − γi2 Sy I, I8+2

nonlinear item XiT Bi Ki Ci in Theorem 1 can be transformed as
− Sy I, I9+2 + He(IT9+2 Ci Mi I2 ), a W-problem [35].
G̃ i = −Mi I1 + Ai Mi I2 + Bi Li Ci Ix + λ̄Bi Li Iϕ + Fi I8+2 , In terms of Ni Ci = Ci Mi , it yields
 
˜ i = ˜ i + He X̃ i G˜ i .
X̃ i = (μ1 I1 + μ2 I2 )T ,  (Ni Ci − Ci Mi )T (Ni Ci − Ci Mi ) = 0. (40)

Moreover, the controller gains are computed by Ki = Li Ni−1 . By using Schur complement to (40), one can get
Proof: Adopting Schur complement to (18) leads to  
−I (Ni Ci − Ci Mi )T
  < 0, (41)
i νIT2 (Hi Ci )T ∗ −I
< 0. (38)
∗ −νI which equals the condition (37) with a sufficient small constant
Define Mi = Xi−1 , P̃i = (I2( +1)+1 ⊗ Mi )Pi (I2( +1)+1 ⊗ Mi ),  > 0. Then, the proof is fulfilled.
S̃1i = Mi S1i Mi , Ỹ1i = Mi Y1i Mi , S̃2i = Ni S2i Ni , Ỹ2i = Ni Y2i Ni ,
Ni Ci = Ci Mi and Li Ci = Ki Ci Mi . IV. C IRCUITAL I MPLEMENTATION
Left- and right-multiplying (38) with M i = In this section, a specific circuital execution method is used
diag{Mi , Mi , Mi , Ni , Ni , I +1 ⊗ Mi , I +1 ⊗ Ni , I, I} and to imitate the considered LFC power system, where the system
its transpose, one has components and controller are realized by some equivalent
  analogy circuits based on resistors, capacitors and operational
ˆ i νIT Mi (Hi Ci )T
2 < 0, (39) amplifiers. For simplification of presentation, the first control
∗ −νI
area (i = 1) of a multi-area NPS with three control areas is
where  ˆ i = ˆ i + He(X̃ i G˜ i ), discussed below.
      It is noted that the dynamics of system components are
ˆ i = He JT P̃i Q + Sy S̃1i + hỸ1i , I2 − Sy S̃1i , I3 described via some first-order inertial elements in Fig. 1. By

  defining the parameters as: r1 = r2 , Tgi = r1 c1 , r3 = r4 , Tti =
− Sy  ⊗ Ỹ1i , Ix + Sy S̃2i + hỸ2i − NiT Zi Ni , I4 r3 c2 , Di = rr65 , Mi = r5 c3 , the equivalent analogy circuits of
 

− Sy S̃2i , I5 − Sy  ⊗ Ỹ2i , Iϕ − γi2 Sy I, I8+2 them are shown in Fig. 3 (a), (b) and (c), respectively. The
  droop constant can be viewed as a proportion element, and its

− Sy I, I9+2 + He IT9+2 Ci Mi I2 . equivalent analogy circuit is shown in Fig. 3 (d) by choosing

r8
Rgi = r7 .
1

Utilizing the inequality −NiT Zi Ni ≤ ρ 2 Zi−1 − 2ρNi to (39) The circuital realization of the PI controller is drawn in
yields (36) by introducing a novel variable matrix Zi = Zi−1 . Fig. 4, where a proportion circuit, an integration circuit and
It is infeasible to solve the equation Ni Ci = Ci Mi since it an adding
circuit are used to represent ui (t) = KPi ACEi (t) +
is not a strict inequality. Then the problem of tackling the KIi ACEi (t)dt with KPi = rr109 and KIi = r111c4 .
YAN et al.: PROBABILITY-DENSITY-DEPENDENT LFC OF POWER SYSTEMS 4843

TABLE II
T HE VALUES OF C IRCUIT E LEMENTS

Fig. 5. Circuital implementation of ACEi (t) for Area 1. For the given matrix Hi = I2 , and considering  the
tanh(yi1 (t))
TABLE I deception attack signal ϕi (t) = , we have
T HE VALUES OF PARAMETERS
tanh(yi2 (t))
ϕi (t)2 ≤ Hi Ci xi (t)2 based on |tanh(yi1 (t))| ≤ 1 and
|tanh(yi2 (t))| ≤ 1. Under the given H∞ index γi = 20, by
choosing  = 0.01, μ1 = 0.1, μ2 = 0.5, ρ = 0.3 and
ν = 10, the controller gains of three power systems solved by
Theorem 2 for different probabilities of the attack occurrence
λ̄ = 0.5 are derived as:
# $
K1 = 0.6718, −0.2956 ,
# $
K2 = 0.7781, −0.3324 ,
# $
K3 = 0.8797, −0.5440 .
The signal ACEi (t) of (2) shown in Fig. 5 is the sum of the The control system is simulated via circuital realization
signal βi fi (t) and the signal Ptie−i (t), which can be realized method in Simulink/MATLAB. The system components and
by adding circuit with OA and some resistors with the same the PI controller can be realized by equivalent electrical cir-
resistance values. The signal βi fi (t), the product of the bias βi cuits. The professional Simcape toolbox of Simulink is used
and fi (t), can be achieved by proportion circuit with βi = rr19 18
. to realize the governor, turbine and power generator of Area 1,
The signal P tie−i (t) equals to the sum of T 12 (f 1 (t) − which is given in Fig. 6. Due to page limitation, the other simu-
f2 (t)) and T13 (f1 (t) − f3 (t)), which can be derived by lation diagrams are omitted here. The communication network
integration circuits with r14 c5 = 1 and r17 c6 = 1. Moreover, is executed by a variable delay module and a random num-
the signal T12 (f1 (t) − f2 (t)) is obtained by subtraction cir- ber module generating the random delay given by Simulink.
cuit and proportion circuit with T12 = rr13 12
. The same process According to Section IV, the values of resistors and capacitors
of deriving the signal T13 (f1 (t) − f3 (t)) is executed with of analogy circuits for the Area 1 are given in Table II. Due
T13 = rr16
15
. to page limitation, the other two areas are omitted here.
Remark 5: The equivalent analogy circuits for power In the simulation, the sampling period is 0.01s and the initial
systems can be viewed as low-cost simulators, which are used condition is zero. The following two different cases of load
to test the effectiveness of the proposed control method. This variations are considered.
is helpful for avoiding the serious hazard induced by poten- Case 1: The load variations for three control areas are
tial drawbacks of designed controller to expensive devices of Pd1 (t) = 0.1sin(0.5t), Pd2 (t) = 0.1sin(π t) and Pd3 (t) =
practical systems. Meanwhile, the cost of testing can be saved 0.1sin(1.5t) for 0 < t < 5s; otherwise, Pdi (t) = 0 for t ≥ 5s
dramatically. and i = 1, 2, 3.
Case 2: The load variations for three control areas are
V. E XAMPLE Pdi (t) = 0.1 for 0 < t < 2s; Pdi (t) = −0.1 for 2 ≤ t < 4s;
The parameters of the considered NPS are borrowed otherwise, Pdi (t) = 0 for t ≥ 4s and i = 1, 2, 3.
from [34], which are presented in Table I. For the considered deception attacks and the load variations
In this example, the random transmission delays shown in in Case 1, the frequency trajectories by using the designed
Fig. controller gains are drawn in Fig. 7. Meanwhile, the corre-
0 2 are considered. In order to correspond to the form of sponding curves for Case 2 are drawn in Fig. 8. These figures
−h g(s)x(t + s)ds in the derivations, g(τ ) is written as g(s) =
denote that for the above two different situations of load vari-
g0 (s) = −2500se50s , s ∈ [−h, 0]. For = 1, to construct
ations, the designed controllers are effective for maintaining
the vector g(s) satisfying (7), another term g1 (s) = −50e50s
the system frequencies to be stable even when control signals
is chosen. Then, the vector g(s) and corresponding parameters
are attacked by deceptive data.
are given as
      To show the advantages of the proposed method over some
−2500se50s 50 50 0.16 0.08 existing ones, three groups of comparison results are provided
g(s) = , G = ,  = .
−50e50s 0 50 0.08 0.08 as below.
4844 IEEE TRANSACTIONS ON SMART GRID, VOL. 13, NO. 6, NOVEMBER 2022

Fig. 6. The diagram for governor, turbine and power generator of Area 1 via circuital realization in Simulink/MATLAB.

Case 3: The controller is designed without taking into

account the deception attacks, which is called a ‘standard’
controller. In this case, the controller design conditions can
be obtained by removing the terms related to ϕi (t). With
the above parameters and λ̄ = 0.3, the controller gains are
derived as:
K1 = [1.7601, −0.5126],
K2 = [2.1556, −0.3525],
K3 = [2.5467, −1.2538];
Case 4: The controller is designed by our method con-
sidering the deception attacks, which is called a ‘resilient’
controller. Namely, the designed resilient controller is able
Fig. 7. Responses of the frequency for three areas under λ̄ = 0.5 and load
to ensure the system stability even if the system is attacked.
variations in Case 1. In this case, with λ̄ = 0.3 and the same parameters used in
Case 3, the controller gains solved by Theorem 2 are:
K1 = [0.8119, −0.3225],
K2 = [0.9781, −0.3599],
K3 = [1.0859, −0.5885].
In this simulation, the load disturbances are taken as
Pd1 (t) = 0.1cos(π t)e−0.5t , Pd2 (t) = 0.1sin(π t)e−0.5t
and Pd3 (t) = 0.1e−0.5t for 0 < t < 8s (ω(t) = 0
for t ≥ 8s) for three areas. The responses of system
frequencies in Case 3 and Case 4 are drawn in Fig. 9 and
Fig. 10, respectively. By comparing the curves in Fig. 9
and Fig. 10, it is observed that our proposed controller con-
sidering the deception attacks (Case 4) can produce better
control performance than the controller designed without con-
sidering the deception attacks (Case 4). From Fig. 9, it is
Fig. 8. Responses of the frequency for three areas under λ̄ = 0.5 and load seen that the standard controller is not able to stabilize the
variations in Case 2. system frequencies when the transmitted signals are attacked
by injecting some deceptive data. However, Fig. 10 shows
our designed resilient controller can still guarantee the sta-
A. The First Comparison Group bility of system frequencies even deception attacks happen.
The following comparison further shows the better con- This means that our proposed control strategy is resilient to
trol performance of our designed controller against deception the deception attacks and can outperform the standard control
attacks than the existing controller. Two cases are considered: method.
YAN et al.: PROBABILITY-DENSITY-DEPENDENT LFC OF POWER SYSTEMS 4845

TABLE IV
C OMPARISON OF ℘ AND THE M INIMUM H∞ I NDEX γ
U NDER D IFFERENT M ETHODS

Fig. 9. Case 3: Responses of the frequency for three areas with ‘standard
controller.’

Fig. 11. g(τ ), ĝ(τ ) and e(τ ) = ĝ(τ ) − g(τ ) with = 1, 3, 5.

According to this table, all the optimized γi of three con-

trol areas obtained by our method are smaller than the results
derived by traditional interval time-varying delay approach,
which indicates better stability performance will be achieved.
Specifically, for h = 0.1 and i = 1, 2, 3, the values of γi are
reduced by 5.6%, 6.5% and 5.5%; for h = 0.2 and i = 1, 2, 3,
Fig. 10. Case 4: Responses of the frequency for three areas with our ‘resilient
controller.’ the reductions of the values of γi are 22.3%, 19.4% and 21.9%;
for h = 0.3 and i = 1, 2, 3, the corresponding reductions are
TABLE III 32.6%, 27.7% and 27.5%. In addition, it is also seen that the
C OMPARISON OF THE M INIMUM H∞ I NDEX γ U SING
D IFFERENT D ELAY H ANDLING M ETHODS corresponding values of γi increase as the growth of the values
of delay bound h.

C. The Third Comparison Group

To show the merit of our approach based on Lemma 1
to handle the distributed delay with kernel, the comparison
results of the H∞ index derived by our method and the
Bessel-Legendre inequality in [29] are obtained in Table IV.
The second case of random delay with h = 0.2 in the above
second comparison group is considered here. Following the
method in [29], we utilize Legendre polynomials to approx-
B. The Second Comparison Group imate the distributed kernel g(τ ). For given , g(τ ) can be
To illustrate the advantage of the delay modeling approach approximated by ĝ(τ ):
based on probability density, the comparison results of the H∞ 

 2i + 1 h τ 
index for three power systems obtained by our approach and ĝ(τ ) = g(τ )Li dτ,
the conventional interval time-varying delay approach in [15] h 0 h
i=0
are given in Table III, where three different random delay cases 
are considered. where Li ( τh ) = (−1)i ij=0 (−1)j ( ij )( i+j τ j
j )( h ) . The figures
To be specific, the first case is g(s) = −10000se100s , s ∈ of the distributed kernel g(τ ) and its approximation ĝ(τ )
[−h, 0] and h = 0.1; the second case is g(s) = −2500se50s , with different are shown in Fig. 11. Furthermore, the opti-
s ∈ [−h, 0] and h = 0.2; the third case is g(s) = −1110se33s , mized H∞ index γ and the number of decision variables
s ∈ [−h, 0] and h = 0.3. And the second case is the same (℘ = 52 (2 + 3)(1 + 5(2 + 3))) in Pi with our method
with the random delay given in Fig. 2. and the method in [29] are shown in Table IV. In terms of
4846 IEEE TRANSACTIONS ON SMART GRID, VOL. 13, NO. 6, NOVEMBER 2022

this figure and Table IV, it is indicated that ℘ increases and [11] S. Yan, S. K. Nguang, and Z. Gu, “H∞ weighted integral event-triggered
the approximation error e(τ ) decreases as the growth of . synchronization of neural networks with mixed delays,” IEEE Trans. Ind.
Informat., vol. 17, no. 4, pp. 2365–2375, Apr. 2021.
Meanwhile, one can also observe that our approach needs [12] M. H. Khooban, T. Niknam, F. Blaabjerg, and T. Dragicevic, “A new load
fewer decision variables to deal with the random delay and frequency control strategy for micro-grids with considering electrical
deception attack without any approximation error and leads to vehicles,” Elect. Power Syst. Res., vol. 143, pp. 585–598, Feb. 2017.
[13] H. Bevrani and T. Hiyama, “On load-frequency regulation with time
better H∞ performance than the existing approach in [29]. delays: Design and real-time implementation,” IEEE Trans. Energy
Based on these comparison results, it is shown that the Convers., vol. 24, no. 1, pp. 292–300, Mar. 2009.
proposed LFC method for NPSs subject to random delay and [14] Y. Mi et al., “Sliding mode load frequency control for multi-area time-
deception attacks in this manuscript is more effective than delay power system with wind power integration,” IET Generat. Transm.
Distrib., vol. 11, no. 18, pp. 4644–4653, 2017.
some existing methods. [15] C.-K. Zhang, L. Jiang, Q. H. Wu, Y. He, and M. Wu, “Delay-dependent
robust load frequency control for time delay power systems,” IEEE
Trans. Power Syst., vol. 28, no. 3, pp. 2192–2201, Aug. 2013.
VI. C ONCLUSION [16] J. Li, Z. Chen, D. Cai, W. Zhen, and Q. Huang, “Delay-dependent sta-
This work studied the decentralized H∞ secure LFC of bility control for power system with multiple time-delays,” IEEE Trans.
Power Syst., vol. 31, no. 3, pp. 2316–2326, May 2016.
NPSs with random communication delays and deception
[17] Y. Tipsuwan and M.-Y. Chow, “Gain scheduler middleware: A
attacks. A novel model of random transmission delays has methodology to enable existing controllers for networked control and
been presented by using the delay probability density. Based teleoperation-part I: Networked control,” IEEE Trans. Ind. Electron.,
on this model, the normal control signals and deception attacks vol. 51, no. 6, pp. 1218–1227, Dec. 2004.
[18] S. Yan, Z. Gu, and S. K. Nguang, “Memory-event-triggered H∞
were represented as two distributed delay terms, in which output control of neural networks with mixed delays,” IEEE
the probability density is treated as the distributed delay Trans. Neural Netw. Learn. Syst., early access, Jun. 4, 2021,
kernel. Then, new sufficient probability-density-dependent doi: 10.1109/TNNLS.2021.3083898.
[19] C. Peng, Q.-L. Han, and D. Yue, “Communication-delay-distribution-
conditions were proposed by Lyapunov method to design the dependent decentralized control for large-scale systems with IP-based
decentralized H∞ secure controller to maintain the system communication networks,” IEEE Trans. Control Syst. Technol., vol. 21,
frequencies. The dynamics of power system were realized by no. 3, pp. 820–830, May 2013.
[20] H. Sun, C. Peng, D. Yue, Y. Wang, and T. Zhang, “Resilient
electrical circuit systems, which were executed by the profes- load frequency control of cyber-physical power systems under QoS-
sional Simcape toolbox of Simulink/MATLAB. The simulation dependent event-triggered communication,” IEEE Trans. Syst., Man,
examples demonstrated the effectiveness of the developed Cybern., Syst., vol. 51, no. 4, pp. 2113–2122, Apr. 2021.
approach. In addition, how to extend the developed approach [21] Y. Tang, D. Zhang, P. Shi, W. Zhang, and F. Qian, “Event-based for-
mation control of multi-agent systems under DoS attacks,” IEEE Trans.
to handle the LFC issue of NPSs with nonlinear dynamics or Autom. Control, vol. 66, no. 1, pp. 452–459, Jan. 2021.
integrated with wind power and solar power deserves further [22] H.-T. Sun, C. Peng, and F. Ding, “Self-discipline predictive control of
investigations in the future. autonomous vehicles against denial of service attacks,” Asian J. Control,
Feb. 2022, doi: 10.1002/asjc.2749.
[23] F. Qu, E. Tian, and X. Zhao, “Chance-constrained H∞ state estima-
R EFERENCES tion for recursive neural networks under deception attacks and energy
constraints: The finite-horizon case,” IEEE Trans. Neural Netw. Learn.
[1] C. Zhao, W. Sun, J. Wang, Q. Li, D. Mu, and X. Xu, “Distributed cooper- Syst., early access, Jan. 7, 2022, doi: 10.1109/TNNLS.2021.3137426.
ative secondary control for islanded microgrid with Markov time-varying [24] T.-Y. Zhang and D. Ye, “False data injection attacks with complete
delays,” IEEE Trans. Energy Convers., vol. 34, no. 4, pp. 2235–2247, stealthiness in cyber-physical systems: A self-generated approach,”
Dec. 2019. Automatica, vol. 120, Oct. 2020, Art. no. 109117.
[2] K. Xiahou, Y. Liu, and Q. Wu, “Robust load frequency control of power [25] S. Yan, Z. Gu, and J. H. Park, “Memory-event-triggered H∞ load
systems against random time-delay attacks,” IEEE Trans. Smart Grid, frequency control of multi-area power systems with cyber-attacks and
vol. 12, no. 1, pp. 909–911, Jan. 2021. communication delays,” IEEE Trans. Netw. Sci. Eng., vol. 8, no. 2,
[3] Y. Wu, Z. Wei, J. Weng, X. Li, and R. H. Deng, “Resonance attacks on pp. 1571–1583, Apr.-Jun. 2021.
load frequency control of smart grids,” IEEE Trans. Smart Grid, vol. 9,
[26] J. Liu, Y. Gu, L. Zha, Y. Liu, and J. Cao, “Event-triggered H∞
no. 5, pp. 4490–4502, Sep. 2018.
load frequency control for multiarea power systems under hybrid
[4] F. Yang, J. He, and Q. Pan, “Further improvement on delay-dependent
cyber attacks,” IEEE Trans. Syst., Man, Cybern., Syst., vol. 49, no. 8,
load frequency control of power systems via truncated B–L inequality,”
pp. 1665–1678, Aug. 2019.
IEEE Trans. Power Syst., vol. 33, no. 5, pp. 5062–5071, Sep. 2018.
[5] J. He, Y. Liang, X. Hao, F. Yang, and Q. Pan, “A quadratic [27] X.-Z. Jin, W.-W. Che, Z.-G. Wu, and H. Wang, “Analog control
convex framework with bigger freedom for the stability analysis circuit designs for a class of continuous-time adaptive fault-tolerant
of a cyber-physical microgrid system,” Sci. China Inf. Sci., 2022, control systems,” IEEE Trans. Cybern., early access, Oct. 23, 2020,
doi: 10.1007/s11432-021-3433-8. doi: 10.1109/TCYB.2020.3024913.
[6] C. Peng, J. Zhang, and H. Yan, “Adaptive event-triggering H∞ load [28] X. Jin, S. Wang, J. Qin, W. X. Zheng, and Y. Kang, “Adaptive fault-
frequency control for network-based power systems,” IEEE Trans. Ind. tolerant consensus for a class of uncertain nonlinear second-order multi-
Electron., vol. 65, no. 2, pp. 1685–1694, Feb. 2018. agent systems with circuit implementation,” IEEE Trans. Circuits Syst.
[7] Y. Liu, Y. Chen, and M. Li, “Dynamic event-based model predictive I, Reg. Papers, vol. 65, no. 7, pp. 2243–2255, Jul. 2018.
load frequency control for power systems under cyber attacks,” IEEE [29] A. Seuret, F. Gouaisbaut, and Y. Ariba, “Complete quadratic
Trans. Smart Grid, vol. 12, no. 1, pp. 715–725, Jan. 2021. Lyapunov functionals for distributed delay systems,” Automatica,
[8] E. Tian and C. Peng, “Memory-based event-triggering H∞ load vol. 62, pp. 168–176, Dec. 2015.
frequency control for power systems under deception attacks,” IEEE [30] P. Kundur, Power System Stability and Control. 2nd ed. New York, NY,
Trans. Cybern., vol. 50, no. 11, pp. 4610–4618, Nov. 2020. USA: Springer, 2014.
[9] T. Yang, Y. Zhang, W. Li, and A. Y. Zomaya, “Decentralized net- [31] S. Saxena and Y. V. Hote, “Load frequency control in power systems via
worked load frequency control in interconnected power systems based internal model control scheme and model-order reduction,” IEEE Trans.
on stochastic jump system theory,” IEEE Trans. Smart Grid, vol. 11, Power Syst., vol. 28, no. 3, pp. 2749–2757, Aug. 2013.
no. 5, pp. 4427–4439, Sep. 2020. [32] Y. C. Tian, D. Levy, M. O. Tade, T. Gu, and C. Fidge, “Queuing packets
[10] Y. Mi, Y. Fu, C. Wang, and P. Wang, “Decentralized sliding mode load in communication networks for networked control systems,” in Proc.
frequency control for multi-area power systems,” IEEE Trans. Power 6th World Congr. Intell. Control Automat., Dalian, China, vol. 1, 2006,
Syst., vol. 28, no. 4, pp. 4301–4309, Nov. 2013. pp. 205–209.
YAN et al.: PROBABILITY-DENSITY-DEPENDENT LFC OF POWER SYSTEMS 4847

[33] Q. Feng and S. K. Nguang, “Stabilization of uncertain linear distributed Ju H. Park (Senior Member, IEEE) received the
delay systems with dissipativity constraints,” Syst. Control Lett., vol. 96, Ph.D. degree in electronics and electrical engineer-
pp. 60–71, Oct. 2016. ing from the Pohang University of Science and
[34] D. Rerkpreedapong, A. Hasanovic, and A. Feliachi, “Robust load Technology (POSTECH), Pohang, South Korea, in
frequency control using genetic algorithms and linear matrix inequal- 1997.
ities,” IEEE Trans. Power Syst., vol. 18, no. 2, pp. 855–861, May 2003. From May 1997 to February 2000, he was
[35] C. A. R. Crusius and A. Trofino, “Sufficient LMI conditions for output a Research Associate with Engineering Research
feedback control problems,” IEEE Trans. Autom. Control, vol. 44, no. 5, Center-Automation Research Center, POSTECH.
pp. 1053–1057, May 1999. He joined Yeungnam University, Gyeongsan, South
Korea, in March 2000, where he is currently the
Chuma Chair Professor. He has coauthored the
monographs Recent Advances in Control and Filtering of Dynamic Systems
with Constrained Signals (New York, NY, USA: Springer-Nature, 2018)
and Dynamic Systems With Time Delays: Stability and Control (New York,
NY, USA: Springer-Nature, 2019) and is an Editor of an edited volume
Recent Advances in Control Problems of Dynamical Systems and Networks
(New York: Springer-Nature, 2020). His research interests include robust
control and filtering, neural/complex networks, fuzzy systems, multiagent
systems, and chaotic systems. He has published a number of articles in these
areas. Since 2015, he has been a recipient of the Highly Cited Researchers
Award by Clarivate Analytics (formerly, Thomson Reuters) and listed in
three fields, Engineering, Computer Sciences, and Mathematics, in 2019,
2020, and 2021. He also serves as an Editor for the International Journal
of Control, Automation and Systems. He is also a Subject Editor/Advisory
Editor/Associate Editor/Editorial Board Member of several international jour-
nals, including IET Control Theory & Applications, Applied Mathematics
Shen Yan received the B.E. and Ph.D. degrees and Computation, Journal of The Franklin Institute, Nonlinear Dynamics,
from the College of Electrical Engineering and Engineering Reports, Cogent Engineering, the IEEE T RANSACTION ON
Control Science of Nanjing Technology University, F UZZY S YSTEMS, the IEEE T RANSACTION ON N EURAL N ETWORKS AND
Nanjing, China. From 2017 to 2018, he was a visit- L EARNING S YSTEMS, and the IEEE T RANSACTION ON C YBERNETICS. He
ing Ph.D. student with the University of Auckland, is a Fellow of the Korean Academy of Science and Technology (KAST).
Auckland, New Zealand. From February 2022 to
August 2022, he was a Visiting Scholar with
Yeungnam University, Gyeongsan, South Korea. He
is currently an Associate Professor with the College
of Mechanical and Electronic Engineering, Nanjing Xiangpeng Xie (Member, IEEE) received the B.S.
Forestry University, Nanjing, China. His current and Ph.D. degrees in engineering from Northeastern
research interests include networked control systems, event-triggered control, University, Shenyang, China, in 2004 and 2010,
and their applications. respectively.
From 2010 to 2014, he was a Senior Engineer with
Metallurgical Corporation of China Ltd., Beijing,
China. He is currently a Professor with the Institute
of Advanced Technology, Nanjing University of
Posts and Telecommunications, Nanjing, China. His
research interests include fuzzy modeling and con-
trol synthesis, state estimations, optimization in pro-
cess industries, and intelligent optimization algorithms. He serves as an
Associate Editor for the International Journal of Control, Automation, and
Systems and the International Journal of Fuzzy Systems.

Chunxia Dou (Senior Member, IEEE) received

the B.S. and M.S. degrees in automation from
Northeast Heavy Machinery Institute, Qiqihaer,
Zhou Gu (Member, IEEE) received the B.S. China, in 1989 and 1994, respectively, and the
degree from North China Electric Power University, Ph.D. degree in electrical engineering from Yanshan
Beijing, China, in 1997, and the M.S. and Ph.D. University, Qinhuangdao, China, in 2005.
degrees in control science and engineering from the In 2010, she joined the Department of
Nanjing University of Aeronautics and Astronautics, Engineering, Peking University, Beijing, China,
Nanjing, China, in 2007 and 2010, respectively. where she was a Postdoctoral Fellow for two
From 1996 to 2013, he was with the School of years. From 2005 to 2019, she was a Professor
Power engineering, Nanjing Normal University, as with the School of Electrical Engineering, Yanshan
an Associate Professor. He was a Visiting Scholar University. Since 2019, she has been a Professor with the Institute of
with Central Queensland University, Rockhampton, Advanced Technology for Carbon Neutrality, Nanjing University of Posts
QLD, Australia, and The University of Manchester, and Telecommunications, Nanjing, China. Her current research interests
Manchester, U.K. He is currently a Professor with Nanjing Forestry University, include MAS-based control, event-triggered hybrid control, distributed
Nanjing. His current research interests include networked control systems, coordinated control, multimode switching control, and their applications in
time-delay systems, reliable control, and their applications. power systems and smart grid.