4270 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 36, NO.

5, SEPTEMBER 2021

Inertia-Enhanced Distributed Voltage and Frequency

Control of Low-Inertia Microgrids
Congyue Zhang , Student Member, IEEE, Xiaobo Dou , Member, IEEE, Zhang Zhang,
Guannan Lou , Member, IEEE, Fan Yang, and Guixin Li

Abstract—This paper proposes a novel inertia-enhanced dis- is developed from consensus algorithms [7], [8]. It provides
tributed control method to complement the inertia of microgrids. a novel way to eliminate the deviations. Distributed control
The rate of change of frequency (RoCoF) and the rate of change uses only neighboring rather than global information and has
of voltage (RoCoV) are employed in this paper to quantify the
frequency inertia and voltage inertia, respectively. Then, a fully better reliability and lower sensitivity to communication failure
distributed algorithm with constrained changing rates is proposed. than centralized control. The convergence rates of voltage and
By bounding the changing rates of frequency and voltage during frequency are the research focus of distributed control in mi-
the consensus control, the algorithm can address the consensus crogrids. Many fix-time and finite-time control algorithms are
problem while enhancing the inertia of microgrids. Compared with proposed to achieve the rapid convergence of DGs in microgrids
most inertia control methods, the proposed method can utilize
the reserve power of scattered DGs to supply inertia. Besides, it [9]–[11].
performs better under disturbances and delays than conventional In recent years, other factors that may degrade the distributed
distributed control methods. The effectiveness of the proposed control performance also attract the attention of researchers.
method is validated by several cases in MATLAB/Simulation and The factors of noise [12], communication delay [12]–[15], event
a hardware experiment. trigger [16], [17], model uncertainties [18], and actuator/sensor
Index Terms—Distributed control, inertia control, microgrids. faults [19], [20] have been concerned in the distributed control
design. However, it is worth noticing that few distributed studies
pay attention to the inertia problem of microgrids. The control
I. INTRODUCTION
object of most distributed methods is the droop controlled DG,
Y MIMICKING the behavior of the conventional syn-
B chronous generator (SG), droop control provides a solution
for the control of high penetration of distributed generators
which is mathematically equivalent to a low inertia system [21].
Even though low inertia ensures fast response, it also leads to
poor response characteristics. The control system is prone to
(DGs) in microgrids [1]–[3]. However, droop controlled DGs instability under disturbances.
are difficult to address the accurate power sharing issue. There Existing studies of inertia control mostly focus on local con-
are always deviations from the nominal values of voltage and trol like the virtual synchronous generator (VSG) concept [22],
frequency. These limitations drive researchers to concern about [23] and the centralized control of microgrids [24], [25]. But in
the control strategies in higher layers. practice, it is not economical to upgrade all droop controlled
In the hierarchical control framework of microgrids, sec- DGs to VSGs or install extra inertia supplement devices to
ondary control determines the correction terms, commonly via supply inertia. Moreover, centralized inertia supplement shares
a centralized method or a distributed method [3], [4]. In the the limitations of centralized control. Therefore, compensating
centralized control framework, the information of all DGs is inertia through a distributed control may be another alternative.
gathered and calculated in a control center. Due to the complex Different from operation states like voltage and frequency,
communication networks among the center and DGs, the con- inertia is an estimated value of the system, making it hard to
trol efficiency is hard to guarantee [5], [6]. Distributed control be directly controlled by algorithms. According to the results
of [26] and [27], the RoCoF and the frequency nadir are the
Manuscript received July 2, 2020; revised November 16, 2020 and January two critical metrics of inertia. Most distributed methods can
19, 2021; accepted January 24, 2021. Date of publication February 4, 2021; date decrease the nadir of frequency by rapidly eliminating the de-
of current version August 19, 2021. This work was supported in part by the
National Natural Science Foundation of China under Grant 51777031, in part viation. Suppressing the RoCoF is the remaining problem to be
by the Scientific Project of Tianjin Electric Power Company KJ20-1-08. Paper addressed for inertia enhancement. The control of the changing
no. TPWRS-01100-2020. (Corresponding author: Xiaobo Dou.) rate attracts little attention in distributed control, but velocity
Congyue Zhang, Xiaobo Dou, and Guannan Lou are with the School
of Electrical Engineering, Southeast University, Nanjing 210096, China control is widely investigated in the multi-agent system. For
(e-mail: [email protected]; [email protected]; bingzhi0828 instance, to avoid collision between unmanned aerial vehicles,
@163.com). studies in [28]–[32] investigate serval consensus-based methods
Zhang Zhang, Fan Yang, and Guixin Li are with the State Grid Tianjin Electric
Power Company, Tianjin 300010, China (e-mail: [email protected]; with constrained velocities. Studies in [28] and [29] discuss the
[email protected]; [email protected]). consensus problem with the nonconvex constraint of velocity.
Color versions of one or more of the figures in this article are available online The formation and consensus control of second-order multi-
at https://ieeexplore.ieee.org/10.1109/TPWRS.2021.3057078.
Digital Object Identifier 10.1109/TPWRS.2021.3057078 agent systems are discussed in [30]–[32]. Since a microgrid

0885-8950 © 2021 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: Politecnico di Bari. Downloaded on February 19,2024 at 10:36:23 UTC from IEEE Xplore. Restrictions apply.
ZHANG et al.: INERTIA-ENHANCED DISTRIBUTED VOLTAGE AND FREQUENCY CONTROL OF LOW-INERTIA MICROGRIDS 4271

can be noticed as a multi-agent system, the velocity-constrained B. Basic Control Model of DGs
methods provide feasible solutions for inertia enhancement in
Droop function is the key component in the power control
microgrids. loop of DGs It describes the relationship between frequency,
To supplement the inertia of microgrids through a distributed
output voltage magnitude and output powers:
control scheme, this paper provides a novel inertia-enhanced
distributed control method. The proposed method is inspired ωi = ωni − mP i Pi ∗ (1)
by velocity-constrained methods in [29]–[33] and addresses the ∗ ∗ ∗
vo,magi = vi,odi = vni − nQi Qi (2)
insufficient inertia problem by a consensus-based distributed
∗
algorithm. Key contributions can be summarized as follows: vi,oqi =0 (3)
1) A novel distributed control method is proposed to enhance
inertia and address consensus problems at the same time. where ω i and v∗ o,magi are the frequency and voltage magnitude;
The introduced inertia compensation in the algorithm ω ni and vni are the reference value of the frequency and the
makes the proposed distributed control perform better voltage that are generated from the secondary control; Pi ∗ and
against uncertainties, delays, and disturbances than most Qi ∗ are the measured real and reactive power; mpi and nqi are the
common ones [10]–[19]. droop coefficients; v∗ i,odi and v∗ i,oqi are the d-axis and q-axis
2) The rate of change is employed in this paper to quantify components of v∗ o,magi . The relationship of v∗ o,magi = v∗ i,odi is
inertia. This quantification makes the inertia can be indi- due to the fact that v∗ o,magi often aligns itself on the d-axis [11].
rectly controlled by the constrained changing rate in the The measured value Pi ∗ and Qi ∗ are extracted from the
algorithm. It also makes the inertia control extend from fundamental component of the output power Pi and Qi via a
frequency to voltage. first-order low pass filter, which can be expressed as follows:
3) This paper employs a distributed framework to supplement 1 1
Pi ∗ = Pi = (vodi iodi + voqi ioqi ) (4)
inertia. Compared with most local and centralized inertia τi s + 1 τi s + 1
supplement methods [22]–[24], the proposed inertia sup- 1 1
plement method is more flexible and stable. Qi ∗ = Qi = (voqi iodi − vodi ioqi ) (5)
τi s + 1 τi s + 1
4) The proposed method is more economical for inertia com-
pensation in practices. It uses the reserve power of existing where τi represents the time constant of the filter. vodi , voqi , iodi ;
DGs instead of installing extra devices to supplement ioqi are the d-axis and q-axis components of output voltage voi
inertia. and output current ioi .
The rest of this paper is organized as follows: Section II intro- With the discussion of the inner voltage and current control
duces the preliminaries and the problem formulation. Then, the loops of the droop-controlled DG in [9], the dynamic model for
relationship between inertia and the changing rate is discussed in the ith DG can be represented as
Section III . Section IV presents the proposed inertia-enhanced ẋi = fi (xi ) + di (xi )Di + gi (xi )ui
distributed control algorithm. Finally, several cases and an ex-
periment test are discussed and analyzed in Section V, and = Fi (xi ) + gi (xi )ui (6)
Section VI concludes.
where xi = [δi Pi Qi φdi φqi γdi γqi ildi ilqi vodi voqi iodi ioqi ]T ,
δi is the angle of the DG reference frame with respect to the
common reference frame. ildi and ilqi are the d-axis and q-axis
II. PRELIMINARIES AND PROBLEM FORMULATION
components of the inner current ili . ui = [vni ωni ]T is the
A. Preliminaries of Graph Theory system input and Di = [ωcom vbdi vbqi ]T is considered as
disturbances; ω com is the rating frequency in the common
A microgrid is usually recognized as a multi-agent system. In
reference frame.
this paper, we consider such a microgrid composed of N DGs,
In the control model (6), the fi (xi ), gi (xi ), di (xi ) are similar
and DGs play the role of agents. The communication among
to those in [9]–[12], [18], [20], so we omitted the detailed
these DGs is modeled as the undirected graph G = {V, E, A}
expressions of them in this paper.
with the agents set V = {1, 2, …, N}, the edge set E = {(i, j), i,
jࢠV}, and the adjacency matrix A = [aij ]N × N . An undirected
C. Problem Formulation
graph is connected if there exists an undirected path between
any two distinct agents in the graph. If the DGi and the DGj can Fig. 1 shows a distributed control framework for a microgrid
exchange data, the edge (j, i) will exist and aij = aji ,aij >0, aii containing N DGs. Each DG connects to the microgrid through
= 0. The Laplacian matrix of G is denoted  by L = [lij ]N × N a DC/AC converter, an inductance-capacitance (LC) filter, and
with elements lij = −aij if jࣔi, else lii = j=i aij . For the a resistance-inductance (RL) output connection. Based on this
leader-following microgrid, including one leader and N DGs, its control framework, two issues are considered in this paper:
communication topology between the leader and DGs is defined Insufficient inertia in the microgrid: Increased penetration of
by a diagonal matrix B = diag{b1 , …bN }. If the ith DG can inverter interfaced DGs reduces the inertia of microgrids [34].
receive information from the leader, bi = 1; otherwise, bi = 0 Low inertia increases the RoCoF and reduces the time reaching
[10]. frequency nadir, leading the microgrids prone to instability
under disturbances.

Authorized licensed use limited to: Politecnico di Bari. Downloaded on February 19,2024 at 10:36:23 UTC from IEEE Xplore. Restrictions apply.
4272 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 36, NO. 5, SEPTEMBER 2021

expressed as [34]
 
1 
m 
n
Hω,M G = Hω,SGi · SSGi + Hω,i · Si (11)
SM G i=1 i=1

where Hω,SGi and Hω,i are the inertia constant the ith SG and
the ith DGs, respectively. SSGi and Si are the nominal powers
of the ith SG and the ith DG, respectively. SMG is the nominal
apparent power of the whole system.
Assumption 1: The initial values of RoCoF and RoCoV of
Fig. 1. Block diagram of an inverter-based microgrid. DGs in this paper are considered to be in finite scopes (−vω ,max ,
vω ,max ) and (−vv,max , vv,max ), respectively.
Lemma 1: Under Assumption 1, if the RoCoF of each DG
Secondary control of DGs: Traditional secondary control can be bounded with
commonly adopts centralized mode, which suffers high com-
munication dependency, low fault tolerance, and poor plug-and- −vω,max ≤ ω̇i (t) ≤ vω,max ∀i ∈ N (12)
play capability [11]. the frequency inertia of each DG and the whole microgrid can
In view of those issues, the control objective is to design a obtain minimum values. The same applies to the RoCoV and the
novel inertia-enhanced distributed control algorithm to achieve voltage inertia.
the inertia complement and states consensus, such that: Proof: For common generators in microgrids, the well-
1) The output voltages of all DGs can converge to the global known swing equation for frequency response is
tracking voltage reference vn given by the leader:
Hω dω
lim [vodi (t) − vn ] = 0 ∀i ∈ N (7) 2S · =Pm − Pe − D(ω − ωn ) (13)
t→∞ ωn dt
and the voltage inertia can be enhanced by constraining the where S is the nominal power of the generator, D is the damping
operation scope of the RoCoV of each DG: coefficient, Pe and Pm are the output real power and mechanical
power, respectively, ω and ω n are the frequency and its reference
−vv,max ≤ v̇odi (t) ≤ vv,max ∀i ∈ N (8) value, respectively. dω/dt is the RoCoF.
where vv,max is the maximum changing rate of voltage. If we consider inertia constant H is variable, according to
2) The output frequencies of all DGs can asymptotically (13), the frequency inertia constant is inversely proportional to
reach the global frequency reference ωn given by the RoCoF. With the constraint of (12), the minimum inertia constant
leader, and the accurate active power sharing can be ad- of a DG can be expressed as
dressed: ωn
 Hω,i,min = (ΔP̄i + DΔω̄i ) (14)
limt→∞ [ωi (t) − ωn ] = 0 ∀i ∈ N 2vω,max Si
(9)
limt→∞ [mi Pi (t) − mj Pj (t)] = 0 ∀i, j ∈ N
where ΔP̄i and Δω̄i are the maximum positive value of the real
Also, the frequency inertia can be enhanced by constraining power and frequency deviations of DGs, respectively.
the scope of the RoCoF of each DG: Then, the inertia constant of the ith DG satisfies
−vω,max ≤ ω̇i (t) ≤ vω,max ∀i ∈ N (10) Hω,i ≥ Hω,min ∀i ∈ N (15)
where vω,max is the maximum changing rate of frequency. and the inertia constant of the whole microgrid is bounded with
Remark 1: In most studies, inertia commonly represents the
ωn m
 
frequency inertia. The concept of voltage inertia is rarely men- Hω,M G ≥ ΔP̄i + DΔω̄i (16)
tioned. Same with the frequency inertia, the appropriate intro- 2SM G vω,max i=1
duction of inertia for voltage control can improve the stability
Thus, the Lemma 1 holds.
of the system.
Remark 3: In practice, the value of ΔP̄i and Δω̄i are often
Remark 2: As mentioned in [10] and [15], the line impedance
hard to be accurately obtained, making the minimum value of
effect makes it difficult to simultaneously address the voltage
inertia constants cannot be calculated accurately. Therefore, the
consensus and the reactive power sharing, especially in a micro-
results in this paper only illustrate the ensured constraint of
grid system with a complicated topology. Therefore, like most
RoCoF and RoCoV.
distributed voltage control studies, this paper focuses on the
voltage consensus problem and does not discuss the reactive
IV. DISTRIBUTED SECONDARY CONTROL WITH CONSTRAINED
power sharing problem.
RATE OF CHANGE
III. DISCUSSION OF INERTIA AND RATE OF CHANGE A. Distributed Voltage Control
Microgrids commonly consist of loads, several DGs, and a Motivated by [9], we adopt the method of input-output feed-
few SGs. The inertia constant of a microgrid can commonly be back linearization to render the dynamic model (6). Then, the

Authorized licensed use limited to: Politecnico di Bari. Downloaded on February 19,2024 at 10:36:23 UTC from IEEE Xplore. Restrictions apply.
ZHANG et al.: INERTIA-ENHANCED DISTRIBUTED VOLTAGE AND FREQUENCY CONTROL OF LOW-INERTIA MICROGRIDS 4273

voltage state-space model of the ith DG with the output states Lemma 2: (see [30]): Combine the system (19) and the algo-
[yvi,1 , yvi,2 ]T = [vodi , v̇odi ]T is represented as rithm (20) as

ẏvi,1 = yvi,2 ŷ˙ vi,2 (t) = −kv ŷvi,2 (t) − kv γiv (t) (24)
(17)
ẏvi,2 = fvi (xi , wvi ) + bvi uvi
Under Assumption 1, if the constraint
where fvi (xi , wvi ) = L2Fi hi (xi ) + wvi . The input gain is de-
fined as bvi = Lgi LFi hi (xi ). L2Fi hi (xi ) is the second Lie vv,min ≤ γiv (t) ≤ vv,max (25)
derivative [35] of hi (xi ) along Fi (xi ), Lgi LFi hi (xi ) is the Lie is always satisfied, the constraint
derivative of LFi hi (xi ) along gi (xi ). The detailed expressions
of them can be found in [11]. wvi represents model uncertainties −vv,max ≤ ŷvi,2 (t) ≤ vv,max (26)
and noise, which cannot be precisely established in the control can always be guaranteed in (24).
mode [10], [18]. uvi = vni is the secondary voltage control input. We now state and establish the following theorem.
In practical terms, the variable v̇odi is full of noise and hard to Theorem 1: Under Assumption 1, for the distributed system
measure, and the expression fvi (xi , wvi ) is difficult to calculate described by (17), using the inertia-enhanced distributed control
[18]. Thus, we employ a three order ESO to estimate these scheme (23), the voltage consensus is addressed, and the voltage
variables : inertia is enhanced with bounded RoCoV.
ŷvi,1 → yvi,1 , ŷvi,2 → yvi,2 , ŷvi,3 → f (xi , wvi ) (18) Proof: Since γiv (t) is designed with the activation function
tanh(·), which satisfies the scop (−1,1), and μv is no greater than
where ŷvi,1 , ŷvi,2 , ŷvi,3 are estimated values.
vv,max , the constraint (25) and (26) in Lemma 2 can always be
Then, we establish the second-order model for distributed
satisfied. The guaranteed (26) is equivalent to (12), making the
voltage control based on the observation:
minimum inertia of voltage can be ensured by Lemma 1.

ŷ˙ vi,1 = ŷvi,2 Then, to prove the stable consensus of the proposed method,
(19)
ŷ˙ vi,2 = αiv we write the tracking errors by combining (20)–(22):
 y
where αiv = ŷvi,3 + bvi uvi is an intermediate control state. ė v,1 (t) = eyv,2 (t)
To address the voltage consensus and guarantee the con- ėyv,2 (t) = −kv eyv,2 (t) − kv μv [(L + B) ⊗ IN ]η(t − d)
straints (−vv,max , vv,max ), we consider the following algorithm (27)
where eyv,1 (t) and eyv,2 (t) are the compacted forms of the
αiv (t) = −kv ŷvi,2 (t) − kv γiv (t) (20) tracking errors, which are
y y
γiv (t) = μv tanh εv [kv ei v,1 (t) + ei v,2 (t)] (21) eyv,1 (t)
y
where kv and εv are positive constants; μv ≤ vv,max . ei v,1 (t) y y y y
= [e1v,1 (t), . . . env,1 (t)]T , eyv,2 (t) = [e1v,2 (t), . . . env,2 (t)]T
y
and ei v,2 (t) are tracking errors from neighbors, which can be
expressed as and⎧
⎧ ⎨η(t − d) = tanh[εv sv (t − d)]
⎪ yv,1 N
= [tanh[εv sv1 (t − d)], . . . , tanh[εv svn (t − d)]]T
⎪
⎨ ei (t) = aij [ŷvi,1 (t) − ŷvj,1 (t)] + bi [ŷvi,1 (t) − vn ] ⎩ y y
.
j=1 sv (t) = [sv1 (t), . . . svn (t)]T; svi (t) = kv eviv,1 (t) + eviv,2 (t)
⎪
⎪ y 
N
The derivative of sv (t) results in
⎩ ei v,2 (t) = aij [ŷvi,2 (t) − ŷvj,2 (t)] + bi ŷvi,2 (t)
j=1
ṡv (t) = −kv μv [(L + B) ⊗ IN ] tanh[εv sv (t − d)] (28)
(22)
Besides, considering the inherent delays in communication Consider the following Lyapunov-Krasovskii function:
channels, the proposed distributed algorithm can be written as  εv siv  t
N
αiv (t) = − kv {ŷvi,2 (t) V = pi tanh(x)dx + η̇(x)T Qη̇(x)(x − t + d)dx
 y y  i=1 0 t−d
+ μv tanh {εv kv ei v,1 (t − d) + ei v,2 (t − d) }} (29)
(23) where Q = diag{q1 , · · · qN } is a positive-definite matrix.
where d represents the boundness delay with 0 ≤ d ≤ dmax . The derivate of (29) along the (28) results in
Remark 4: The algorithm (20) is composed of two compo- V̇ = − η(t)T P Rη(t) + η(t)T P Rκ(t)
nents. The first component −kv ŷvi,2 (t) is similar to the dif-  t
ferential feedback of VSG [22]–[24] and can provide inertia + dη̇(t)T Qη̇(t) − η̇(x)T Qη̇(x)T dx (30)
support. The second part −kv γiv (t) is the consensus part for t−d
voltage consensus tracking.
where κ(t) = η(t) − η(t − d);R = kv μv [(L + B) ⊗ IN ]; and
Remark 5: The controller (23) can provide instant feedback
P = diag{p1 , · · · pN }is a positive-definite matrix.
based on the local RoCoV state. This instant feedback can
From the result in [38], we can conclude:
rapidly suppress overshoot before the information of neighbors  t
arrives. Thus, the proposed method has better control stability
dη̇(x)T Qη̇(x)T dx ≥ κ(t)T Qκ(t) (31)
under delays than common distributed methods. t−d

Authorized licensed use limited to: Politecnico di Bari. Downloaded on February 19,2024 at 10:36:23 UTC from IEEE Xplore. Restrictions apply.
4274 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 36, NO. 5, SEPTEMBER 2021

This estimation helps us get the following model:


ŷ˙ ωi,1 = ŷωi,2
(36)
ŷ˙ ωi,2 = αiω
where αiω = ŷωi,3 + bωi u̇ω i − bωi mP i Ṗi .
Remark 6: In reality, the RoCoF is usually averaged within
a given time window (such as 100 ms) to eliminate noises. In
this paper, we use ESO to estimate the RoCoF. The noise or
Fig. 2. Block diagram of the proposed secondary voltage control.
unmodeled dynamics can be extended in the extra order ŷωi,3
and separated by the ESO [37].
Same as the voltage tracking errors, the frequency tracking
Also, as η(t) = tanh[εv sv (t)] ≤ εv sv (t), we have errors are defined with
⎧
V̇ ≤ −η(t)T P Rη(t) + η(t)T P Rκ(t) + dε2v ṡv (t)T Qṡv (t) ⎪ yω,1 
N
⎪
⎨ ei (t) = aij [ŷωi,1 (t) − ŷωj,1 (t)] + bi [ŷωi,1 (t) − ωn ]
≤ −η(t)T P Rη(t) + η(t)T P Rκ(t) j=1
⎪
⎪ y 
N
⎩ ei ω,2 (t) = aij [yωi,2 (t) − yωj,2 (t)] + bi yωi,2 (t)
κ(t)T Zκ(t)
+ dε2v R[−η(t) + κ(t)]T Q[−η(t) + κ(t)] − j=1
d (37)
To achieve the frequency consensus, we design the frequency
= −[η(t) κ(t) ] A(d)[η(t) κ(t) ]
T T T T T
(32)
consensus control algorithm:
where  ω
αi (t) = −kω [yωi,2 (t)+γiω (t − d)]
y y
A(d) γiω (t − d) = μω tanh εω [kω ei ω,1 (t − d) + ei ω,2 (t − d)]
  (38)
1 2d(P R − M ) −d(P H + M ) where μω ≤ vω,max , εω and kω are positive constants.
= , M = dε2v RT QR
2d ∗ 2(Q − M ) Then, to address the power sharing, we define an auxiliary
If parameters εv , kv are determined in positive, we can have state ump
i (t) = mP i Ṗi (t)[10] and design the consensus algo-
d → 0, A(0) → ∞, and the matrix A(d) is monotonic decreas- rithm:
ing with respect to d [31]. This means there will exist a maximum 
N
αmp
value dmax of the time delay to guarantee A(d) > 0 and V̇ ≤ 0. ump
i (t) = kmp aij [emp
ij (t − d)] tanh[emp
ij (t − d)]
Finally, since ėyv,1 (t) = eyv,2 (t) and kv is positive, the result j=1
(39)
of sv (t) → 0 equals to vodi − vodj → 0, v̇odi − v̇odj → 0. The
where empij (t − d) = m P i P i (t − d) − m P j P j (t − d) is the
stable consensus of the proposed method is achieved.
power sharing errors from neighbors; kmp and αmp are positive
To conclude, according to (19), the actual control input for
gains.
voltage inertia enhancement and voltage consensus is
Similar to the previous section, the following theorem sum-
1 marizes the results of this section.
uvi (t) = − {kv [ŷvi,2 (t) + γiv (t − d)] + ŷvi,3 (t)} (33)
bvi Theorem 2: Under Assumption 1, for the distributed control
The schematic diagram of this distributed voltage control is system described by (35), using the inertia-enhanced control
shown in Fig. 2. scheme (38) and (39), the frequency consensus and active power
sharing are addressed, and the frequency inertia is enhanced by
bounding the RoCoF.
B. Distributed Frequency Control and Active Power Sharing
Proof: The proof is similar to Theorem 1.
According to (1) and (4), the control model for distributed According to the above (36)-(39), the actual control input can
frequency control is established as [36] be expressed as
ωi = ωni − τi ω̇i − mP i Pi (34) i (t)
uω

Then, to achieve the control of the RoCoF, we extend the 1
=− {kω [yωi,2 (t) + γiω (t − d)] + ŷωi,3 (t) + ump
i (t)}
model (34) as bωi
(40)
ω̈i = bωi (u̇ω
i − ω̇i − mP i Ṗi ) (35)
Fig. 3 shows the schematic diagram of the proposed inertia-
where bωi = 1/τi is the input gain; uω i = ωni is the secondary enhanced distributed frequency control algorithm.
frequency control input. The extended control state ω̇i is used
for differential feedback in the following control design. V. CASE STUDIES
Considering the measurement noise wωi of the extended ω̇i ,
we also adopt ESO to estimate the states in (35): To verify the effectiveness of the proposed method, several
simulation cases are built in the MATLAB/Simulation, and a
ŷωi,1 → ωi , ŷωi,2 → ω̇i , ŷωi,3 → wωi − bωi ω̇i hardware experiment is designed in this section.

Authorized licensed use limited to: Politecnico di Bari. Downloaded on February 19,2024 at 10:36:23 UTC from IEEE Xplore. Restrictions apply.
ZHANG et al.: INERTIA-ENHANCED DISTRIBUTED VOLTAGE AND FREQUENCY CONTROL OF LOW-INERTIA MICROGRIDS 4275

Fig. 3. Block diagram of the proposed secondary frequency control.

Fig. 5. Performance evaluation: (a) voltage, (b) RoCoV.

Fig. 4. Diagram of the autonomous microgrid with four DGs.

TABLE I
PARAMETERS OF THE SECONDARY CONTROLLER

Fig. 6. Performance evaluations: (a) frequency, (b) RoCoF.

A. Case 1: Basic Control Performance Evaluation

As Fig. 4 shows, a 380 V, 50 Hz island microgrid is built under
the MATLAB environment. Table I provides the parameters
of the proposed controllers. The droop coefficients are set as
mp1 = 4 × 10−5 , mp2 = 5.5 × 10−5 , mp3 = mp4 = 4 × 10−5 .
Other used parameters of line impedance are referred to [18].
Several simulations scenarios are set to evaluate the performance
of the proposed method in this and several following subsections:
1) t = 0.0s: only the primary control is activated.
2) t = 1.0s: the proposed distributed control is activated. Fig. 7. Performance evaluations: (a) active power ratio, (b) active power,
(c) reactive power.
3) t = 2.0s: DG3 is disconnected (plugged out).
4) t = 3.0s: DG3 is reconnected (plugged in)
5) t = 4.0s: Load2 is cut out. robust performance against the uncertainties of loads, Load2 is
6) t = 5.0s: Load3 decreased by 50% of the load. cut out at t = 4 s, and Load3 decreased by 50% of the load at
This subsection evaluates the basic performance of the pro- t = 5 s. Fig. 5(a), Fig. 6(a), and Fig. 7(a) show that during those
posed inertia-enhanced distributed method, including voltage time periods, the frequencies, voltages, and active power ratios
and frequency consensus, plug and play capacity, and the con- of all DGs can still achieve stable consensus tracking.
straints of the rate of change. In addition, the RoCoV and RoCoF are depicted in Fig. 5(b)
At the beginning of the simulation (t = 0 s), we intentionally and Fig. 6(b). It is worth noticing that these two states are
deactivate the proposed controllers. As Fig. 5(a) and Fig. 6(a) bounded in the given scopes, making the voltage inertia and
show, the voltages and frequencies derivate from the reference frequency inertia are also restricted. Besides, the output powers
values (311 V, 50 Hz). Then, when the proposed controllers of each DGs are depicted in Fig. 7(b) and Fig. 7(c).
are applied at t = 1 s, the voltages and frequencies are rapidly
restored to the given references, and the power sharing among
B. Case 2: Inertia Enhanced Performance Evaluation
DGs is achieved.
To evaluate the plug and play capacity of the proposed method, To highlight the inertia enhancement performance of the
DG3 is plugged out at t = 2 s and in at t = 3 s. To evaluate the proposed method, we compare it with a common distributed
Authorized licensed use limited to: Politecnico di Bari. Downloaded on February 19,2024 at 10:36:23 UTC from IEEE Xplore. Restrictions apply.
4276 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 36, NO. 5, SEPTEMBER 2021

Fig. 10. Voltage performance under delay d = 100ms: (a) the method in [12];
(b) the proposed method; (c) comparisons of DG1 and DG4.
Fig. 8. Voltage performance comparison of DG4.

Fig. 11. Frequency performance under delay d = 100ms: (a) the method in
[12]; (b) the proposed method; (c) comparisons of DG1 and DG4.

Fig. 9. Frequency performance comparison of DG4.

method in [10]. For the sake of simplicity, only the performance

of DG4 is depicted in this section.
As Fig. 8 shows, compared with the common method in [10],
the voltage performance of our method has better robustness
when changes happen in the microgrid. The RoCoV of our
Fig. 12. Performance of RoCoV: (a) the method in [12]; (b) the proposed
method is lower, and the voltage nadir is much smaller. Besides, method.
the performance of the frequency and RoCoF in Fig. 9 have the
same result. Thus, we can conclude that the proposed inertia
enhanced distributed control method is more robust and more
resilient against disturbances and uncertainties than common
distributed control methods.

C. Case 3: Performance Evaluation Under Delay.

The control performance under larger communication delay
(d = 100ms) is investigated in this subsection. The theoretical Fig. 13. Performance of RoVoF: (a) the method in [12]; (b) the proposed
result in Section IV concludes that the maximum delay dmax method.
correlates with specific control parameters. Thus, we reset the
parameters εv = 2 × 10−5 , kv = 100, εω = 0.02, kω = 6 in
this section to guarantee the control stability (dmax > 150 ms).
The comparisons of our method and the method in [12] are
illustrated in Fig. 10 and Fig. 11. Without inertia enhancement,
the voltages and frequencies of the method [12] appear unstable
oscillations. In contrast to it, our method can rapidly suppress
the changing rates of voltage and frequency based on the local
feedback, avoiding the system from losing stability before the
information of neighbors arrives. Fig. 10(c) and Fig. 11(c) dis-
tinctly compare the better performance of our proposed method.
Besides, the results in Fig. 12 and Fig. 13 illustrate that the
RoCoF and RoCoV also meet the given constraints under the
large delay. Since more conservative parameters are configured Fig. 14. Modified IEEE 34 bust test feeder [38].

Authorized licensed use limited to: Politecnico di Bari. Downloaded on February 19,2024 at 10:36:23 UTC from IEEE Xplore. Restrictions apply.
ZHANG et al.: INERTIA-ENHANCED DISTRIBUTED VOLTAGE AND FREQUENCY CONTROL OF LOW-INERTIA MICROGRIDS 4277

Fig. 19. Frequency performance evaluation under PV changes.

Fig. 15. Performance evaluations: (a) frequency, (b) RoCoF.

Fig. 20. Voltage performance evaluation under PV changes.

Fig. 16. Performance evaluation: (a) voltage, (b) RoCoV.

Fig. 21. Performance evaluation of the acturate power sharing under PV

changes.

scopes, and the consensus problems of voltage, frequency, and

Fig. 17. Performance evaluation of the active power ratio. active power action are all well addressed.
The long time scale simulation results are shown in Fig. 19,
Fig. 20, and Fig. 21. We employ the actual photovoltaic (PV)
curve of one day in EPRI PV Data and scale down the time scale
to 1/100 for simulation simplification and data storage.
The actual PV curves in the microgrids are shown in Fig. 18,
and the regulation performance is illustrated in Fig. 19, Fig. 20,
Fig. 18. Actual PV changes in the microgrid. and Fig. 21. After applying the proposed controllers (t = 5 s), the
voltage and frequency of each DG are well controlled around
the references. The rapid fluctuation caused by PV changes,
to tolerate larger delay, the control system spends more time plugging out of DG3 (t = 300s–400s), and communication
reaching the consensus, which can be noticed from Fig. 10(b) failure of DG4 (t = 800 s) are all suppressed by the proposed
and Fig. 5(a) (or form Fig. 11(b) and Fig. 6(a)). controllers. This shows that the proposed method can adapt to
large-scale microgrids with complicated scenarios.
D. Case 4: Testing on Large-Scale Microgrids
This case investigated the scalability of the proposed method E. Case 5: Experimental Verification
with a modified standard IEEE 34 bus system [38]. The param- To validate the effectiveness of the proposed method in a
eters of the added DG5 and DG6 are the same with DG4 and practical scenario, a hardware-in-the-loop MG testbed, as shown
DG1, respectively. in Fig. 22 is developed. The control algorithm of each DG
Fig. 15 to Fig. 17 show the simulation results under a short is programmed using a TMS320F28377D DSP, and the main
time scale. We implement the proposed controllers as t = 1 s, circuit is built by the RT-LAB system. The topology of the
plug out the DG5 during t = 2 s to t = 3 s, and cut out the experimental MG is shown in Fig. 23, and the parameters are
Load3 at t = 4 s. As the results show, the proposed method also the same as the Fig. 4.
can maintain similar regulation results in a larger and extended Several experimental cases are designed in this hardware
microgrid. The RoCoV and RoCoF are bounded in the given experiment, including plug-and-play capability test and different

Authorized licensed use limited to: Politecnico di Bari. Downloaded on February 19,2024 at 10:36:23 UTC from IEEE Xplore. Restrictions apply.
4278 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 36, NO. 5, SEPTEMBER 2021

Fig. 22. Schematic diagram of the hardware platform.

Fig. 26. Active power ratio performance of the hardware experiment.

Fig. 23. Topology of the experiment.

Fig. 27. Performance evaluation of the hardware experiment: (a) RoCoF,

(b) RoCoV.

Fig. 24. Voltage control performance of the hardware experiment.

VI. CONCLUSION
This study proposes a novel distributed control method to
compensate inertia for low inertia microgrids. We employ the
RoCoV and RoCoF to quantify the voltage and frequency in-
ertia, making these inertias can be controlled as variables of
the microgrid. The inertia enhancement of this study is ad-
dressed through a distributed algorithm with the constrained
changing rates. Thus, by restricting the RoCoV and the RoCoF
of each DG, the voltage inertia and frequency inertia can be
bunded above certain minimum values. Compared with existing
decentralized and centralized inertia supplement methods, the
Fig. 25. Frequency control performance of the hardware experiment. proposed distributed based inertia supplement method provides
a more flexible, reliable, and economical way to compensate
inertia for microgrids. The scattered resources and reserve power
loads change. The voltage and frequency control performance of can be further utilized to enhance inertia. The simulation and
this hardware experiment are shown in Fig. 24 and Fig. 25. These experiment results demonstrate that the proposed method has
hardware results show similar trends to the above simulation better performance against uncertainties, disturbances, and com-
results. Both the voltage and frequency are rapidly and stably munication delays than most existing distributed methods. The
restored to the given references under different scenarios. The operation RoCoV and RoCoF of all DGs are restricted in the
active power sharing is achieved as Fig. 26 shows. We set the given scopes. The fundamental functions of the distributed con-
constraints (−3,3) and (−200, 200) for the RoCoF and the trol, including voltage and frequency consensus, active power
RoCoV in the algorithm, respectively. As Fig. 27 illustrates, both sharing, and plug-and-play capacity, are all achieved.
the RoCoF and the RoCoV are well bounded in the given scopes, In the future, more complicated problems of communication
which further proves the effectiveness of our inertia-enhanced failure, switching topologies, and stochastic delay will be inves-
distributed control method. tigated in our method.

Authorized licensed use limited to: Politecnico di Bari. Downloaded on February 19,2024 at 10:36:23 UTC from IEEE Xplore. Restrictions apply.
ZHANG et al.: INERTIA-ENHANCED DISTRIBUTED VOLTAGE AND FREQUENCY CONTROL OF LOW-INERTIA MICROGRIDS 4279

REFERENCES [23] X. Quan et al., “Photovoltaic synchronous generator (PVSG): Architecture

and control strategy for a grid-forming PV energy system,” IEEE J. Emerg.
[1] A. Bidram and A. Davoudi, “Hierarchical structure of microgrids control Sel. Topics Power Electron., vol. 8, no. 2, pp. 936–948, Jun. 2020.
system,” IEEE Trans. Smart Grid, vol. 3, no. 4, pp. 1963–1976, Dec. 2012. [24] A. Fathi, Q. Shafiee, and H. Bevrani, “Robust frequency control of mi-
[2] J. Peas Lopes, C. Moreira, and A. Madureira, “Defining control strategies crogrids using an extended virtual synchronous generator,” IEEE Trans.
for microgrids islanded operation,” IEEE Trans. Power Syst., vol. 21, no. 2, Power Syst., vol. 33, no. 6, pp. 6289–6297, Nov. 2018.
pp. 916–924, May 2006. [25] T. Kerdphol, F. S. Rahman, M. Watanabe, and Y. Mitani, “Robust virtual
[3] J. C. Vasquez, J. Matas, L. G. de Vicuna, and M. Castilla, “Hierarchical con- inertia control of a low inertia microgrid considering frequency measure-
trol of droop-controlled AC and DC microgrids a general approach toward ment effects,” IEEE Access, vol. 7, pp. 57550–57560, 2019.
standardization,” IEEE Trans. Ind. Electron., vol. 58, no. 1, pp. 158–172, [26] B. K. Poolla, D. Groß, and F. Dörfler, “Placement and implementation
Jan. 2011. of grid-forming and grid-following virtual inertia and fast frequency re-
[4] M. Yazdanian and A. Mehrizi-Sani, “Distributed control techniques in sponse,” IEEE Trans. Power Syst., vol. 34, no. 4, pp. 3035–3046, Jul. 2019.
microgrids,” IEEE Trans. Smart Grid, vol. 5, no. 6, pp. 2901–2909, [27] C. Phurailatpam, Z. H. Rather, B. Bahrani, and S. Doolla, “Measurement
Nov. 2014. based estimation of inertia in AC microgrids,” IEEE Trans. Sustain.
[5] T. Qian, Y. Liu, W. H. Zhang, W. H. Tang, and M. shahidehpour, “Event- Energy, vol. 11, no. 3, pp. 1975–1984, Jul. 2020.
Triggered updating method in centralized and distributed secondary con- [28] P. Lin, W. Ren, and H. Gao, “Distributed velocity-constrained consensus of
trols for islanded microgrid restoration,” IEEE Trans. Smart Grid, vol. 11, discrete-time multi-agent systems with nonconvex constraints, switching
no. 2, pp. 1387–1395, Mar. 2020. topologies, and delays,” IEEE Trans. Autom. Control, vol. 62, no. 11,
[6] Y. Khayat et al., “On the secondary control architectures of AC microgrids: pp. 5788–5794, Nov. 2017.
An overview,” IEEE Trans. Power Electron., vol. 35, no. 6, pp. 6482–6500, [29] P. Lin, W. Ren, C. Yang, and W. Gui, “Distributed consensus of second-
Jun. 2020. order multiagent systems with nonconvex velocity and control input
[7] J. P. Mishra, C. Li, M. Jalili, and X. Yu, “Robust second-order consensus constraints,” IEEE Trans. Autom. Control, vol. 63, no. 4, pp. 1171–1176,
using a fixed-time convergent sliding surface in multiagent systems,” IEEE Apr. 2018.
Trans. Cybern., vol. 50, no. 2, pp. 846–855, Feb. 2020. [30] T. A. Jesus and L. C. A. Pimenta “Consensus for double-integrator dy-
[8] W. Ren, “On consensus algorithms for double-integrator dynamics,” IEEE namics with velocity constraints,” Intern. J. Control, Autom. Syst., vol. 12,
Trans. Autom. Control, vol. 53, no. 6, pp. 1503–1509, Jul. 2008. no. 5, pp. 930–938, Oct. 2014.
[9] A. Bidram, A. Davoudi, F. L. Lewis, and J. M. Guerrero, “Distributed [31] J. Fu, G. Wen, W. Yu, T. Huang, and X. Yu, “Consensus of second-order
cooperative secondary control of microgrids using feedback linearization,” multiagent systems with both velocity and input constraints,” IEEE Trans.
IEEE Trans. Power Syst., vol. 28, no. 3, pp. 3462–3470, Aug. 2013. Ind. Electron, vol. 66, no. 10, pp. 7946–7955, Oct. 2019.
[10] V. N. M. Dehkordi, N. Sadati, and M. Hamzeh, “Distributed robust finite- [32] J. Fu, Y. Lv, G. Wen, X. Yu, and T. Huang, “Velocity and input constrained
time secondary voltage and frequency control of islanded microgrids,” coordination of second-order multi-agent systems with relative output
IEEE Trans. Power Syst., vol. 32, no. 5, pppp. 3648–3659, Sep. 2017. information,” IEEE Trans. Netw. Sci. Engin., vol. 7, no. 3, pp. 1925–1938,
[11] P. Ge, Y. Zhu, T. Green, and F. Teng, “Resilient secondary voltage control Jul.-Sep. 2020.
of islanded microgrids: An ESKBF-Based distributed fast terminal sliding [33] W. Ren, “On consensus algorithms for double-integrator dynamics,” IEEE
mode control approach,” IEEE Trans. Power Syst, vol. 36, no. 2, pp. 1059– Trans. Autom. Control, vol. 53, no. 6, pp. 1503–1509, Jul. 2008.
1070, Mar. 2021. [34] A. G. Fahad and J. Xiang, “Demand management of active distribution
[12] M. A. Shahab, B. Mozafari, S. Soleymani, N. M. Dehkordi, H. M. network using coordination of virtual synchronous generators,” IEEE
Shourkaei, and J. M. Guerrero, “Stochastic consensus-based control of Trans. Sustain. Energy, vol. no. 12, no. 1, pp. 250–261, Jan. 2021.
µGs with communication delays and noises,” IEEE Trans. Power Syst., [35] H. K. Khalil, Nonlinear Systems, 3rd ed. Upper Saddle River, NJ: Pearson,
vol. 34, no. 5, pp. 3573–3581, Sep. 2019. Dec. 2001.
[13] G. Lou, W. Gu, X. Lu, Y. Xu, and H. Hong, “Distributed secondary [36] G. Lou, W. Gu, Y. Xu, M. Cheng, and W. Liu, “Distributed MPC-Based sec-
voltage control in islanded microgrids with consideration of communi- ondary voltage control scheme for autonomous droop-controlled micro-
cation network and time delays,” IEEE Trans. Smart Grid., vol. 11, no. 5, grids,” IEEE Trans. Sustain. Energy, vol. 8, no. 2, pp. 792–804, Apr. 2017.
pp. 3702–3715, Sep. 2020. [37] L. Qu, W. Qiao, and L. Qu, “An extended-state-observer-based sliding-
[14] G. Lou, W. Gu, Y. Xu, W. Jin, and X. Du, “Stability robustness for mode speed control for permanent-magnet synchronous motors,” IEEE J.
secondary voltage control in autonomous microgrids with consideration Emerg. Sel. Topics Power Electron., to be publishsed.
of communication delays,” IEEE Trans. Power Syst., vol. 33, no. 4, [38] K. Gu, V. Kharitonov, and J. Chen, Stability of Time-Delay Systems.
pp. 4164–4178, Jul. 2018. Boston, MA, USA: Birkh¨auser, 2003.
[15] B. Ning, Q. Han, and L. Ding, “Distributed finite-time secondary frequency
and voltage control for islanded microgrids with communication delays
and switching topologies,” IEEE Trans. Cyber., vol. 51, no. 8, pp. 3988–
3999, Aug. 2021.
[16] B. Abdolmaleki, Q. Shafiee, M. M. Arefi, and T. Dragičević, “An instanta- Congyue Zhang (Student Member, IEEE) received
neous event-triggered hz–watt control for microgrids,” IEEE Trans. Power the B.S. degree in electrical engineering in 2017
Syst., vol. 34, no. 5, pp. 3616–3625, Sep. 2019. from Southeast University, Nanjing, China, where
[17] P. Ge, B. Chen, and F. Teng, “Event-triggered distributed MPC for resilient he is currently working toward the Ph.D. degree in
voltage control of an islanded microgrid,” Int. J. Robust Nonlinear Control, electrical engineering. His research interests include
Accepted. the control of distributed generations, microgrids, and
[18] P. Ge et al., “Extended-State-Observer-Based distributed robust secondary active distribution networks.
voltage and frequency control for an autonomous microgrid,” IEEE Trans.
Sustain. Energy, vol. 11, no. 1, pp. 195–205, Jan. 2020.
[19] N. M. Dehkordi and S. Z. Moussavi, “Distributed resilient adaptive control
of islanded microgrids under sensor/actuator faults,” IEEE Trans. Smart
Grid, vol. 11, no. 3, pp. 2699–2708, May 2020.
[20] M. A. Shahab, B. Mozafari, S. Soleymani, N. M. Dehkordi, H. M.
Shourkaei, and J. M. Guerrero, “Distributed consensus-based fault tolerant Xiaobo Dou (Member, IEEE) received the B.S.E.E.
control of islanded microgrids,” IEEE Trans. Smart Grid, vol. 11, no. 1, degree in power system and its automation from
pp. 37–47, Jan. 2020. Hohai University, Nanjing, China, in 2001 and the
[21] X. Meng, J. Liu, and Z. Liu, “A generalized droop control for grid- Ph.D. degree in power system and its automation
supporting inverter based on comparison between traditional droop control from Southeast University, Nanjing, in 2006. He is
and virtual synchronous generator control,” IEEE Trans. Power Electron., currently a Professor in electrical engineering with
vol. 34, no. 6, pp. 5416–5438, Jun. 2019. the School of Electrical Engineering, Southeast Uni-
[22] Q.-C. Zhong and G. Weiss, “Synchronverters: Inverters that mimic versity. His current research interests include smart
synchronous generators,” IEEE Trans. Ind. Electron., vol. 58, no. 4, grid, microgrid, and renewable energy resources.
pp. 1259–1267, Apr. 2011.

Authorized licensed use limited to: Politecnico di Bari. Downloaded on February 19,2024 at 10:36:23 UTC from IEEE Xplore. Restrictions apply.
4280 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 36, NO. 5, SEPTEMBER 2021

Zhang Zhang received the M.E. degree in elec- Fan Yang received the M.E. degree in electrical engi-
trical engineering from Tianjin University, Tianjin, neering from North China Electric Power University,
China, in 2013. Since 2013, he has been an Engineer Baoding, China, in 2014. Since 2014, she has been
with State Grid Tianjin Economic Research Institute, an Engineer with State Grid Tianjin Economic Re-
Tianjin, China. His research focuses on the planning search Institute, Tianjin, China. Her research interests
of distribution networks. include the planning of distribution networks and
technologies of energy Internet.

Guannan Lou (Member, IEEE) received the B.S.

and M.S. degrees in control science and engineering Guixin Li received the M.E. degree in electrical en-
from North China Electric Power University, Beijing, gineering from Tianjin University, Tianjin, China, in
China, in 2008 and 2011, respectively, and the Ph.D. 2005. Since 2005, he has been an Engineer with State
degree in electrical engineering from Southeast Uni- Grid Tianjin Electrical Power Company, Tianjin. His
versity, Nanjing, China, in 2018. From 2011 to 2015, research focuses on the planning of distribution net-
she was with Guodian Nanjing Automation Company works.
Ltd., Nanjing. From 2017 to 2018, she was a Joint
Ph.D. Student with Argonne National Laboratory.
She is currently a Lecturer with the School of Electri-
cal Engineering, Southeast University. Her research
interests include distributed generation integration and microgrid modeling and
control.

Authorized licensed use limited to: Politecnico di Bari. Downloaded on February 19,2024 at 10:36:23 UTC from IEEE Xplore. Restrictions apply.