1

An FDI Resilient Dynamic Average Distributed

2023 IEEE International Conference on Power Electronics, Smart Grid, and Renewable Energy (PESGRE) | 979-8-3503-1057-3/23/$31.00 ©2023 IEEE | DOI: 10.1109/PESGRE58662.2023.10404728

Control for DC Microgrids

Phani Swecha Tadepalli*, Student Member, IEEE, Deepak Pullaguram, Member, IEEE
M. N. Alam, Member, IEEE.

Abstract—The distributed control DC microgrids with inte- also increases the exposure and vulnerabilities to cyber attacks
grated communication technologies form cyber-physical systems. in the DC microgrid [4]. FDI is considered one of the most
Cyber-physical DC microgrids rely on peer-to-peer (P2P) com- challenging types of cyberattacks to detect compared to other
munication to regulate average voltage and share proportional
power. These P2P communications have greater exposure and forms of cyberattacks. In an FDI attack, intruder adds mali-
increase the possibility of cyber intrusions. This paper aims to cious data to actual data to disrupt the system objectives. FDI
develop a cyber-resilient dynamic average consensus control for can occur in different forms, one can devise an FDI to either
an autonomous DC microgrid. The proposed dynamic average destabilize the microgrid [5] or to create deception, which
resilient controller is designed to accurately track the DC micro- deviates operation points without compromising regulation [6].
grid network average voltage despite false data injections (FDI)
attacks and also restores average voltage to nominal value while The main objective of this paper is to design a distributed
ensuring proportional current sharing. The proposed approach dynamic averaging control that provides resiliency against
eliminates the need for full network connectivity and without any FDI attacks on DC microgrids while satisfying the control
restrictions on the number of agents attacked. The convergence objectives.
studies are carried out using input-to-state stability analysis. In the literature, many strategies have been proposed for
Further, the simulation results validate the proposed controllers
efficacy against actuator and communication layer FDI attacks. detecting and mitigating attacks, which are mainly classified
as model-based and data-driven. The state-of-art literature on
Index Terms—Cyber-Physical Systems, Cyber-attacks, Dis- these techniques are detailed in [7]. In model-based detection,
tributed control, DC Microgrids, Dynamic-average consensus
Consensus, False data injection, Resilience. Observer-based methods [8], [9] are largely adopted. The
design complexity of such methods depends on the model’s
size and may fail under parameter uncertainties. In [10],
I. I NTRODUCTION
model- free distributed sliding observer is proposed, which

T HE emerging DC distributed generators (DG) (like solar,

fuel cells) and storage units require fewer conversion
stages and suit modern loads, which makes DC microgrids
involves the transmission of additional distributed terms, and
steady-state chattering may be observed due to the sliding
surface. Data-based techniques like Artificial neural networks
more advantageous at distribution levels [1]. In the microgrid, [11] and reinforcement learning [12] use historical data to
each DG is expected to share the load changes proportionally. detect anomalies in the output. The availability of a high-
Virtual resistance-based droop control is extensively adopted quality training data set decides the performance of these
for improving per-unit current sharing among all DG in the DC methods. These detection and mitigation are usually more of
microgrids. The virtual resistance and differing line resistances an after-event approach, but it is always better to have design-
in the network may lead to poor voltage regulation and for-security approaches. Thus, many researchers are exploring
may not guarantee proportional current sharing. To achieve resilient techniques to offer such a comprehensive solution.
better voltage regulation and provide improved current sharing, [13], uses an event-driven signal constructed from actual mea-
a secondary controller is used in the DC microgrids [2]. surements to replace attack signals to ensure normal operation
These secondary control methods are mostly communication- during stealth FDI attacks. A trust-based strategy is proposed
based and implemented using either centralized or distributed in [14] to hinder the propagation of the attack to the rest
control. The distributed control algorithms can be more robust network by reducing the communication weight of the attacked
and adaptable than centralized systems, as they can continue DG. This has limitations on the number of DGs compromised.
functioning even if one or more DGs fail and require less Multi-layer resiliency in DC microgrid is addressed in [15]
communication bandwidth [3]. using authentication signal for all neighboring agents. These
The distributed controllers are designed to coordinate among methods may not always guarantee the accurate tracking of
different agents using peer-to-peer communication to achieve average voltage during the attack.
the microgrid objectives: i) average voltage regulation, ii) pro- In this paper, an FDI resilient control approach for DC
portional current sharing. But this distributed communication microgrids is developed, to ensure accurate average voltage
Phani Swecha Tadepalli and M. N. Alam, are with the Department of regulation and sharing proportional power during FDI attacks
Electrical Engineering, National Institute of Technology Warangal, Telangana, on both actuator and communication channel. The main con-
India 506004 Email: [email protected], [email protected]. tributions of this paper are:
Deepak Pullaguram, is with the Department of Electrical Engineering,
Indian Institute of Technology Kharagpur, West Bengal, India 721302. Email: • The proposed control technique effectively estimates and
[email protected]. restores the average voltage to nominal value while en-

Authorized licensed use limited to: University Town Library of Shenzhen. Downloaded on August 28,2024 at 07:57:41 UTC from IEEE Xplore. Restrictions apply.
 2

suring proportional power sharing despite of FDI attacks B. Control layer

on DC microgrid. In DC microgrid, the controller is designed to ensure
• The designed resilient distributed controller is system- proportional load sharing between DGs and maintain the
independent, so individual DG has plug-and-play capa- average voltage at PCC. These objectives are achieved by
bilities. implementing cooperative distributed hierarchical control al-
• The proposed method is resilient towards all bounded gorithms [16]. In this, primary and secondary control provides
FDI attacks irrespective of the number of DGs attacked the voltage set point (v ∗ ) to a well-tuned local cascaded
simultaneously. voltage and current controller. The voltage set point (vi∗ ) of
The subsequent sections of the paper are arranged as fol- ith DG is given by,
lows: In Section II, the cyber and physical aspects of the DC
microgrid and problem formulation are detailed. Section III, vi∗ = vnom − Rdi ii + δvi + δii , (5)
discusses the proposed resilient strategy against FDI attacks where, vnom is nominal voltage of microgrid, Rdi is the virtual
in the DC microgrid. Section IV provides simulation case droop impedance, and ii , is the output current of DGi . δvi
studies using MATLAB/SIMULINK. Section V provides the and δii are correction terms of voltage and current mismatch
conclusion of the paper. from secondary control. The local inner current and outer
II. C YBER P HYSICAL S YSTEM M ODELLING voltage controls are implemented using PI controllers [17],
whose dynamics are,
The DC microgrid forms a CPS by integrating physical
components (energy generation, power converter, storage, and ϕ̇vi = vi∗ − vi ; i∗f i = kpv (vi∗ − vi ) + kiv , ϕvi (6)
distribution network) with digital control systems and com- ϕ̇ci = i∗f i − if i ; di = kpc (i∗fi − if i ) + kic , ϕci (7)
munication networks. This integration of cyber and physical
elements creates a dynamic and interconnected system that where ϕvi and ϕci are the states associated with voltage and
can respond to real-time changes in energy demands and other current controller, respectively. kpv , kiv and kpc , kic are the
disturbances. proportional and integral gains of the voltage and current
controller, respectively.
A. Physical Layer The secondary controller is designed to restore the voltage
The physical layer of a DC microgrid comprises n DG deviation caused by droop and ensure load sharing using
units interconnected through m distribution lines and feeding correction terms δvi and δii in (5). These are realized by
multiple distributed loads. PI controllers, whose inputs are obtained from the resilient
Each DG unit consists of DC voltage source, identifying a distributed controller detailed in the subsequent section. The
dispatchable source such as a fuel cell or renewable sources dynamics of the secondary controllers are given by
with storage units having DC output, and a boost converter φ̇vi = vnom − v̄i ; δvi = kpvs (vnom − v̄i ) + kivs φvi (8)
connected to the point of Common Coupling (PCC) and a
φ̇ci = īi − ii ; δii = kpcs (īi − ii ) + kics φci , (9)
local DC load as shown in Fig. 1(a). By applying Kirchhoff’s
laws, the averaged dynamic model of ith DG boost converter where v̄i and īi , are average voltage and current estimation
is obtained as, obtained from the resilient distributed controller. kpvs , kivs
and kpcs , kics are the proportional and integral gains of the
(
Lf i i̇f i = vdci − Rf i if i − (1 − di )vi ,
DGi (1) secondary voltage and current correction controllers, respec-
Cf i v̇i = if i (1 − di ) + iij − iLi ,
tively. From (4)-(9), the complete state space representation of
where, Lf i , Rf i , and Cf i are the boosting inductor, resistance, a DG, can be formulated as,
and filter capacitor of the DC-DC converter of DGi , respec-
ẋDG = AxDG + B1 uDG + B2 ūDG . (10)
tively. vdci , if i , di , indicates input dc voltage, filter current,
and duty cycle of converter i, respectively. vi is PCC voltage of Here, xDG = [if v ϕv ϕc φv φc ] uDG = [vdc i] , and T T
DGi , iij representing current in the distribution line between ūDG = [v̄ ī]T . The ūDG , is obtained from the distributed
DGi and DGj . The dynamics of the distribution line between control and is realized using the cyber graph.
DGi and DGj are modelled as:
n
Lineij Lij i̇ij = vj − vi − Rij iij . (2) C. Cyber Layer
In distributed control, each converter exchanges local vari-
Here, Rij , Lij indicates resistance and inductance of the line
ables, ū = [v̄, ī], with their neighbors via a communication
between DGi and DGj . For the distribution lines, Rij >>
network, as shown in Fig. 1(a). The cyber communication
Lij , this implies to i̇ij =−i̇ji =0 and (2) is modified as
network of a DC microgrid can be modeled using graph theory.
iij = −iji = (vj − vi )/Rij , (3) Based on the data flow direction, the graph G(V, E, Aa ) can
Now, replacing variable iij in (1) with (3) results in quasi- be either directed or undirected. The converters are considered
stationary line (QSL) approximated model of (1): as a set of nodes V = (1, 2, ...n), and communication links
( between them are depicted by set of edges E ⊂ V × V.
Lf i i̇f i = vdci − Rf i if i − (1 − di )vi , The adjacency matrix is indicated by Aa = [aij ] ∈ Rn×n .
DGi (4)
Cf i v̇i = if i (1 − di ) + (vj /Rij − vi /Rij ) − iLi , (i, j) ∈ E are called adjacents and the entire neighboring set

Authorized licensed use limited to: University Town Library of Shenzhen. Downloaded on August 28,2024 at 07:57:41 UTC from IEEE Xplore. Restrictions apply.
 3

Physical Layer ifi vi ii iij DG1 R14 L14 DG4

PCCi PCCj

Source

Source
DC DC

DC

DC
Rfi Lfi D Lij vj DC DC
Rij iL1 iL4
vdci iLi Cyber
C fi

L34 R34
Si

R12 L12
Attack

DG2 DG3
Cyber 4 Control layer

Source
DC

Source
DC

DC

DC
Layer ifi DC R23 L23 DC
1 3
ii vi DG Controller vi iL2 iL3
ii Cyber Attack
2
ij v j
(a) Schematic of a ith DG unit with control and cyber layer (b) DC microgrid test system with 4 DG

Fig. 1: Schematic of a microgrid test system indicating DG units, physical network, cyber communication network and control
unit

of DGi is represented as Ni = j|(i, j) ∈ E. If there is data This paper proposes a control law, denoted as ui in (11) ,
exchange between DGi and DGj , then aij = 1, otherwise that is both distributed and resilient. The objective is to ensure
aij = 0. The Laplacian matrix L = D P − Aa , analyzes that the state variable, x̄i (t), satisfies (12) and (13), even in
the distributed graph dynamics. D = diag{ j∈Ni aij } is a the presence of cyber-attacks.
degree matrix. The eigenvalues of the L matrix determine the
condition for convergence and stability of the system and it III. P ROPOSED FALSE -DATA I NJECTION R ESILIENT
mandates that graph G must be a spanning tree for achieving CONSENSUS P ROTOCOL
consensus.
This section proposes the detailed design dynamic average
consensus protocol to withstand FDI attacks in DC microgrid
D. Problem Statement systems. Each agent tracks the average of a time-varying
In a DC microgrid consisting of n DG units, each DG unit exogenous input signal xi of ith DG using the following
is modeled in the form of (10), and its control input is obtained consensus protocol [18]:
from the dynamic average consensus algorithm given by:
X

x̄˙ i (t) = ẋi (t) + ui (t) (11) x̄˙ i = α(xi − x̄i ) + β α(x̄j − x̄i ) − (σxj − σxi ) + ẋi
j∈Ni
(16)
where x̄i ∈ {v̄i , īi } ∈ Rn , ui is the state and control input of
X
σ̇i = α(xi − σi ) + β(α(σj − σi ) + (x̄j − x̄i )) + ẋi
ith DG at cyber-layer and xi = {vi , ii } ∈ Rn is the exogenous j∈Ni
input from the physical system. Here, the control protocol ui
shall be designed to fulfill the objectives, Here, xi (t) and σi are time-varying reference and state of
ith DG, α and β are positive gains that are obatined us-
lim (x̄i (t) − x̄j (t)) = 0, i, j ∈ V, (12) ing the eigenvalues of L. Considering bounded FDI attack
t→∞ ′

1
n
X (d(t), d (t)), (16) can be reformulated in vector form as,
lim x̄i (t) = xavg = xk (13)
t→∞ n x̄˙ = α(x − x̄) − αβLx̄ + βLσ + ẋ + d(t)
k=1

Further, the communication links at the cyber layer are ex- σ̇ = α(x − σ) − αβLσ − βLx̄ + ẋ + d′ (t) (17)
posed and may be prone to cyber-attacks, disrupting system In order to demonstrate that (16) achieves dynamic average
operation. Especially the FDI attacks are more tricky to handle consensus, two error vectors are defined,
as they may destabilize the entire system while remaining
1 1
undetected if constructed in a stealthy manner. The FDI attacks ex̄ = x̄ −1n 1⊤
n x; eσ = σ − 1n 1⊤
n x. (18)
can be targeted at the actuator and/or the communication n n
channels of ith DG and can be formulated as, From the property of L, we have Lex̄ = Lx̄ and Leσ = Lσ.
Differentiating (18), one can obtain,
ũi = ui + ηi δui , actuator attack (14)
˜ij = x̄ij + ηij δx̄ij , communication attack
x̄ (15) ėx̄ = −(αIn + αβL)ex̄ + βLeσ + Πn (αx + ẋ),
ėσ = −(αIn + αβL)eσ − βLex̄ + Πn (αx + ẋ). (19)
where, ũi indicates corrupted control input of ith agent with
˜ij represents corrupted data received
bounded false data δui , x̄ Here Πn =In − n1 1n 1⊤
n ẋ. Further, (19) is written as,
at DGi from DGj . η, ηij ∈ 0, 1 is the activation function,
which is 1 in case of attack and 0 otherwise. ϱ̇ = Eϱ + Γ(αx, ẋ), (20)

Authorized licensed use limited to: University Town Library of Shenzhen. Downloaded on August 28,2024 at 07:57:41 UTC from IEEE Xplore. Restrictions apply.
 4

FDI resilient dynamic average control

vnom Secondary Voltage vnom vi ifi
Controller
vi =  (vi − vi ) +   ( (v − v ) − ( −  ) ) + v
j i vj vi i vi k ivs kiv - kic di
-+ k pvs + +-- k pv + + k pc +
j
 v =  (vi −  v ) +   ( ( −  ) + (v − v )) + v
i
Σ
i i
j
vj vi j i i s s s
i
Voltage Current
rd
  ( (i − i ) − ( −  ) ) + i i
Controller Controller
ii =  (ii − ii ) + k
+ ics
i
j i ij ii i
+ k pcs
 i =  (ii −  i
j
) +   ( ( −  ) + (i − i )) + i
i
- s ii
ij ii j i i
ii
i i
j i Secondary Current
Controller

Fig. 2: FDI Resilient distributed dynamic average control strategy for DC microgrid system

where, ϱ = [e⊤ ⊤ ⊤ ⊤ TABLE I: System parameters

x̄ , eσ ] , Γ(αx, ẋ) = [1, 1]  ⊗ Πn (αx + ẋ),
−(αIn + αβL) βL
and E = . Parameter Value
−βL −(αIn + αβL)
To prove the stability of (20), first, the system with x(t) = 0 Li 3mH
is considered, then, Ci 250µF

ϱ̇ = Eϱ (21) Load resistance R1 = R2 = 20Ω,

R3 = R4 = 30Ω
Applying transformation on Line resistance R12 = 0.6Ω
  R14 = 1Ω
iIn −iIn R34 = 3Ω
ϱ = T Rϱ̄, T = , R = blockdiag(V, V). (22)
In In R23 = 1.8Ω.

Here, ϱ̄ = [ϱ̂1 , ϱ̃1 , ϱ̃2 .....ϱ̃n−1 , ϱ̂2 , ϱ̃n , .....ϱ̃( 2n − 2)]T ,V is a Line inductance L12 = 50µH
L14 = 60µH
matrix of left eigen vectors of L. Upon differentiating, (22), L34 = 75µH
is obtained as L23 = 65µH
 
˙ϱ̄ = Λϱ̄; Λ = −αIn + αβJ + iβJ 0 Device level control gains Voltage controller:
(23) Kpv = 3, kiv = 10;
0 −αIn + αβJ − iβJ
Current controller:
Kpc = 0.1, kic = 3.5
Here, J = blockdiag(0, J̃) represents Jordan normal form of
−L, (i.e., −L = VJV−1 ) and λi (J) = λi (L)∀i ∈ V. Droop resistance 0.33Ω
Hence, the eigenvalue of E is:
λi (E) = −α − αβδi + iβδi ∀i ∈ V. (24)
where, µ(Λ̃) is the measure of matrix, and from its property,
where δi is the ith eigenvalue of L and δi > 0∀i = 2, ..n, µ(Λ̃) = µ(P + βQ) ≤ µ(P ) + βµ(Q), this implies to,
This ensures system stability if α > 0 and β > 0.
Next, considering the system (20) with x(t) ̸= 0, factoring out
dynamics associated to zero eigen value of L: ||ϱ̃(t)|| ≤ exp((−κ)t) ||ϱ̃(0)|| + 1/(κ) sup ||Γ̃(αx, ẋ)|| . (28)
0≤τ ≤t
ϱ̂˙ = Υϱ̂, Υ = diag(−α, −α), (25a)
ϱ̃˙ = Λ̃ϱ̃ + Γ̃(αx, ẋ). (25b) Here, κ = α + αβλ2 (sym(L) − βµ(Q)), µ2 (αβ J̃) =
T
where,ϱ̂ = [ϱ̂1 , ϱ̂2 ] , ϱ̃ = [ϱ̃1 , ....ϱ̃2n−2 ] , Γ̃(αx, ẋ) = T −αβλ2 (sym(L)), sym(L) = 12 (L + L⊤ ) and λ2 (sym(L)
[Γ̃1 , ..., Γ̃2n−2 ] and Λ̃=blockdiag (−αIn−1 + αβ J̃ + iβ J̃, is second eigne value of sym(L). The error corresponding to
−αIn−1 + αβ J̃ − iβ J̃)=P + βQ, where Q=blockdiag(iJ̃,-iJ̃), zero input response of (28) approaches zero exponentially if α
P =blockdiag(αIn−1 + αβ J̃, −αIn−1 + αβ J̃). and β are chosen such that α + αβλ2 (sym(L)) > βµ(Q). It
To prove that (20) is input-to-state stable (ISS), we prove can be observed from (28), that when α > µ(Q)/λ2 (sym(L)),
ISS of its equivalent system (25a)-(25b). For any α > 0, and the error decreases to zero, as the value of β increases.
the system (25a) is asymptotically stable. As we know Λ̃, is Now, by premultiplying ėx̄ and ėσ in (19) by 1Tn and using
Hurwitz, for α > 0 and β > 0, the solution of (20) is, balanced property of graph G,
Z t
n n n n
ϱ̃(t) = exp(tΛ̃)ϱ̃(0) + exp((t − τ )Λ̃)Γ̃(αx, ẋ) dτ. (26) X X X X
0 ėxi = −α ėxi , ėσi = −α ėσi (29)
i=1 i=1 i=1 i=1
Using inequality ||exp(t(Λ̄))|| ≤ exp(µ(Λ̄(t))), the solution
can be modified as: Pn
Z t As a result, xavg = 1/n i=1 xi remains within the state
||ϱ̃(t)|| ≤ exp(µΛ̃t) ||ϱ̃(0)|| + exp(µΛ̃(t − τ )) ||Γ̃|| dτ trajectories x̄i and σi ∀i ∈ 1, 2..n. This proves that dynamic
0 average consensus (16), can dynamically track the average
(27) value of exogenous input.

Authorized licensed use limited to: University Town Library of Shenzhen. Downloaded on August 28,2024 at 07:57:41 UTC from IEEE Xplore. Restrictions apply.
 5

IV. R ESULTS AND D ISCUSSION shown in Fig. 3(c) and it appears as a load change to the
The proposed control strategy is tested on a DC microgrid operator as seen in Fig.3(a). But the actual output voltages
with 4 DG sources and communication topology, as shown and the true average value of the actual voltages were largely
in Fig.1. Each DG has a rated capacity of 5kW , and the deviated from the nominal values, which can be observed
nominal voltage is selected as 315V . The Table. I details the from Fig.3(b) and Fig.3(c), respectively. The results in Fig.3
line distribution network and system parameters used for sim- demonstrate the deception caused due to FDI attacks. This
ulation. The test system is simulated in MATLAB/SIMULINK attack is not easily detectable with traditional approaches and
platform under different operating conditions. may endanger the operation of the microgrid.

A. Microgrid Operation with Conventional Control under Ac- B. Microgrid Operation with Proposed Resilient Control un-
tuator FDI Attack der Actuator FDI Attack
The DC microgrid operation with the conventional dynamic This case study examines the operation of a microgrid
average consensus [16] under actuator FDI attacks is evaluated with proposed controller under load change conditions and
in this case study. Initially, the system is at a steady state, evaluates the resiliency against actuator FDI attacks. Initially,
all DGs share load the total load of 16.5kW proportion- the system is operating as in the previous case study. At
ally, as shown in Fig.3(a), while maintaining the average
18
voltage at a nominal value. At t = 2s, a load of 1.5kW
17 Load Change

18 16

17 15

16 FDI attack initiated 14

15 13 FDI attack initiated

14 12
0 1 2 3 4 5 6 7 8 9 10
Time(s)
13
Load Change
12
(a) Output current of DGs
1 2 3 4 5 6 7 8 9 10
325
Time(s)

(a) Output current of DGs 320

325
315

320
310
FDI attack initiated
315
305 Load Change
FDI attack initiated
310
300
0 1 2 3 4 5 6 7 8 9 10
305 Time(s)
Load Change

300
(b) Ouput voltages of DGs
0 1 2 3 4 5 6 7 8 9 10
Time(s) 360

350
(b) Ouput voltages of DGs
340 Load Change
360
330
350
320
340
Load Change 310
330
300
320 FDI attack initiated
290
310
280
300 0 1 2 3 4 5 6 7 8 9 10
FDI attack initiated Time(s)
290

280 (c) Average voltage of DGs

0 1 2 3 4 5 6 7 8 9 10
Time(s)

(c) Average voltage of DGs Fig. 4: Performance of DC microgrid under actuator FDI
attacks on all DGs
Fig. 3: DC microgrid operation with proposed consensus
algorithm under link failure conditions t = 2s, the load at DG3 has increased by 1.5kW , which is
proportionally distributed among the DGs and average voltage
is increased at DG3 , and the controller responds to load is regulated by the proposed controller, as shown in Fig. 4.
change and esures the load variation is distributed among A time-varying bounded attack vector δui similar to that of
all DGs. At t = 6s, a bounded stealth FDI attack vector the previous case is injected into all DGs, at t = 6s. This
δui = [0 5sin(100πt) 0 40] is injected into the actuator attack has led to some transient voltage drop across all the
of DGs. This attack did not induce any steady state deviations DGs, as shown in Fig. 4(b). But in less than 1s, the voltages
on the average voltage estimate (v̄) at individual DGs, as and currents were restored to their pre-attack condition by

Authorized licensed use limited to: University Town Library of Shenzhen. Downloaded on August 28,2024 at 07:57:41 UTC from IEEE Xplore. Restrictions apply.
 6

the resilient action of the proposed controller. Further, the ACKNOWLEDGMENT

average voltage estimated (v̄) at all DGs perfectly tracked the This work is partly supported by the NITW-Hitachi Energy
true average voltage as shown in Fig. 4(c) and regulated at Smart Electric Grid Laboratory and by the Science & Engi-
the nominal value, Vnom = 315V , unlike in the conventional neering Research Board (SERB), Govt. of India, under the
controller (Fig. 3(c)). head of SRG/2020/002557

C. Communication and actuator FDI attacks on all DGs R EFERENCES

The microgrid operation under the combination of commu- [1] N. Eghtedarpour and E. Farjah, “Power control and management in a
nication and actuator attacks is evaluated in this case study. To hybrid ac/dc microgrid,” IEEE Transactions on Smart Grid, vol. 5, no. 3,
pp. 1494–1505, 2014.
evaluate the performance of microgrid operation with multiple [2] J. M. Guerrero, J. C. Vasquez, J. Matas, L. G. de Vicuna, and M. Castilla,
attacks, simultaneous attacks were performed on all DGs. At “Hierarchical control of droop-controlled ac and dc microgrids—a gen-
t = 4s, an attack vector of δi = [10 − 50 20 30] is injected eral approach toward standardization,” IEEE Transactions on Industrial
Electronics, vol. 58, no. 1, pp. 158–172, 2011.
at all DG actuators, and the same is communicated with [3] E. Espina, J. Llanos, C. Burgos-Mellado, R. Cárdenas-Dobson,
the neighboring agents. A significant transient is observed in M. Martı́nez-Gómez, and D. Sáez, “Distributed control strategies for
average voltage and currents as shown in Fig.5, which are microgrids: An overview,” IEEE Access, vol. 8, pp. 193 412–193 448,
2020.
damped within 1s and reached to pre-attack conditions. [4] O. A. Beg, T. T. Johnson, and A. Davoudi, “Detection of false-data
injection attacks in cyber-physical dc microgrids,” IEEE Transactions
14 on Industrial Informatics, vol. 13, no. 5, pp. 2693–2703, 2017.
[5] S. Sahoo, J. C.-H. Peng, A. Devakumar, S. Mishra, and T. Dragičević,
FDI attack initiated “On detection of false data in cooperative dc microgrids—a discordant
13.5
element approach,” IEEE Transactions on Industrial Electronics, vol. 67,
no. 8, pp. 6562–6571, 2020.
13 [6] D. Shi, P. Lin, Y. Wang, C.-C. Chu, Y. Xu, and P. Wang, “Deception
attack detection of isolated dc microgrids under consensus- based
12.5 distributed voltage control architecture,” IEEE Journal on Emerging and
Selected Topics in Circuits and Systems, vol. 11, no. 1, pp. 155–167,
2021.
12
0 1 2 3 4 5 6 7 8 9 10 [7] P. S. Tadepalli and D. Pullaguram, “Distributed control microgrids:
Cyber-attack models, impacts and remedial strategies,” IEEE Transac-
(a) Output current of DGs tions on Signal and Information Processing over Networks, vol. 8, pp.
380 1008–1023, 2022.
[8] T. Wang, L. Liang, S. K. Gurumurthy, F. Ponci, A. Monti, Z. Yang,
360
and R. W. De Doncker, “Model-based fault detection and isolation in
FDI attack initiated
dc microgrids using optimal observers,” IEEE Journal of Emerging and
340
Selected Topics in Power Electronics, vol. 9, no. 5, pp. 5613–5630, 2021.
320
[9] A. J. Gallo, M. S. Turan, F. Boem, T. Parisini, and G. Ferrari-Trecate,
“A distributed cyber-attack detection scheme with application to dc
300 microgrids,” IEEE Transactions on Automatic Control, vol. 65, no. 9,
pp. 3800–3815, 2020.
280 [10] Y. Jiang, Y. Yang, S.-C. Tan, and S. Y. Hui, “Distributed sliding mode
0 1 2 3 4 5 6 7 8 9 10
Time(s) observer-based secondary control for dc microgrids under cyber-attacks,”
IEEE Journal on Emerging and Selected Topics in Circuits and Systems,
(b) Average voltage of DGs
vol. 11, no. 1, pp. 144–154, 2021.
[11] M. R. Habibi, H. R. Baghaee, T. Dragičević, and F. Blaabjerg, “False
Fig. 5: Resilient DC microgrid: attack on the actuator and data injection cyber-attacks mitigation in parallel dc/dc converters based
communication link on artificial neural networks,” IEEE Transactions on Circuits and
Systems II: Express Briefs, vol. 68, no. 2, pp. 717–721, 2021.
[12] A. J. Abianeh, Y. Wan, F. Ferdowsi, N. Mijatovic, and T. Dragičević,
“Vulnerability identification and remediation of fdi attacks in islanded dc
V. C ONCLUSION microgrids using multiagent reinforcement learning,” IEEE Transactions
on Power Electronics, vol. 37, no. 6, pp. 6359–6370, 2022.
This paper presents a distributed control DC microgrid [13] S. Sahoo, T. Dragičević, and F. Blaabjerg, “An event-driven resilient
control strategy for dc microgrids,” IEEE Transactions on Power Elec-
system that is resilient to FDI attacks. In the DC microgrid, the tronics, vol. 35, no. 12, pp. 13 714–13 724, 2020.
proposed secondary controller ensures for accurate tracking of [14] S. Abhinav, H. Modares, F. L. Lewis, and A. Davoudi, “Resilient
the dynamic average voltage of the network, which is regulated cooperative control of dc microgrids,” IEEE Transactions on Smart Grid,
vol. 10, no. 1, pp. 1083–1085, 2019.
at nominal value while ensuring proportional current sharing [15] S. Sahoo, T. Dragičević, and F. Blaabjerg, “Multilayer resilience
even during attacks. The proposed controller is independent paradigm against cyber attacks in dc microgrids,” IEEE Transactions
of the attack or system model and makes DC microgrid re- on Power Electronics, vol. 36, no. 3, pp. 2522–2532, 2021.
[16] V. Nasirian, S. Moayedi, A. avoudi, and F. L. Lewis, “Distributed
silient against any bounded attacks. Dynamic average voltage cooperative control of dc microgrids,” IEEE Transactions on Power
tracking convergence is proved using input-to-system stability Electronics, vol. 30, no. 4, pp. 2288–2303, 2015.
analysis. Through various simulation studies under different [17] D. Pullaguram, S. Mishra, and N. Senroy, “Event-triggered commu-
nication based distributed control scheme for dc microgrid,” IEEE
attack scenarios and operating conditions, the resiliency of Transactions on Power Systems, vol. 33, no. 5, pp. 5583–5593, 2018.
proposed controller is tested. It was observed from results [18] M. Iqbal, Z. Qu, and A. Gusrialdi, “Resilient dynamic average-consensus
that, even under the multiple attacks on all DGs, both at the of multiagent systems,” IEEE Control Systems Letters, vol. 6, pp. 3487–
3492, 2022.
actuator and in communication, the microgrid was resilient
and could able to fulfill the desired control objectives.

Authorized licensed use limited to: University Town Library of Shenzhen. Downloaded on August 28,2024 at 07:57:41 UTC from IEEE Xplore. Restrictions apply.