1238 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 7, NO.

6, NOVEMBER 2021

A Distributed Reactive Power Sharing Approach in

Microgrids with Improved Droop Control
Xinsheng Wang, Jiancheng Zhang, Mingli Zheng, and Lingyu Ma

Abstract—Conventional droop control causes frequency and often leads to the problem of dynamic performance [8]. In
voltage deviations (from rated value) in a inverter-intensive addition, the parallel operation of the DG in the MG will make
microgrid (MG), and the reactive power sharing cannot be the reactive power sharing unavailable [9], and the deviations
obtained when the communication structure of the MG or load
suddenly changes. Compared with a centralized control and of frequency and voltage from their rated values will also occur
droop control scheme, a distributed hierarchical control structure when the MG is in switching topology or disturbed by load.
of the MG can overcome the limitation of communication and To make up for the deficiency of droop control, J. He
realize reactive power sharing. In this paper, an improved droop proposed a MG control method without using droop con-
control is adopted, which is based on the hierarchical control trol [10], A. Micallef proposed a reactive power sharing
structure. The hierarchical control structure consists of zero-
level control, primary control and a proposed secondary control. approach in a droop-controlled MG [11]. These methods of
First, the secondary controller is modeled, and the MG system using central controllers make the system of the MG dependent
composed of distributed generators (DGs) is considered as a on communication networks. In this way, there is a single
multi-agent system. The secondary controller can make up for the fault point in the system, which makes the system unreliable.
shortcomings of the droop controller and adjust the frequency In order to avoid the use of a central controller and reduce
and voltage to their rated values. Secondly, the reference voltage
and frequency of the zero-level control are calculated, and the dependence on the communication network, a completely
combined with the primary control. The zero-level control and decentralized hierarchical control method is adopted to realize
primary control can make the voltage and frequency of the frequency adjustment and power optimization [12], [13]. Using
MG run stably and provide reference voltage for the inverter. this method, DGs do not communicate with each other, which
Finally, the stability of the system is proved by the theory makes the coordinated control of the system impossible. With
of multi-agent consistency. A simulation system is established
in the Matlab/Simulink environment, and the results show the the intention of making a network of agents with switching
effectiveness of the proposed controller. topology control more flexible, a distributed control strategy
is proposed in [14], which considers the DGs of the MG as a
Index Terms—Consensus, distributed generator, hierarchical multi-agent system. For instance, N. M. Dehkordi proposed a
control, microgrid, reactive power sharing. fully distributed cooperative control, which is based on the
multi-agent consensus theory [15]. This method effectively
controls the frequency and voltage. W. Liu proposed a dis-
tributed cooperative frequency control. This method considers
I. I NTRODUCTION
the MG composed of DGs as a multi-agent system considering

M ICROGRID (MG) is a distributed system, which is

composed of several distributed generators (DGs), en-
ergy storage equipment and loads [1], [2]. It can operate in
communication constraints [16], and X. Wang presented a
distributed finite-time cooperative control under switching
topology [17]. However, using the above methods, a reactive
two modes, grid-connected mode and islanded mode [3]. A power sharing cannot be obtained.
reasonable and effective control strategy is the basis of the MG In practice, the DGs are widely distributed, and the distance
operation. Conventional droop control regulates voltage and between them is different. So the link impedances of the DGs
frequency by power, and makes them stable when the system are not equal. To obtain power sharing, the link impedances
is disturbed [4]. No communication between the devices is between DGs should be proportional. Z. Zhang [18] proposed
required in the MG with conventional droop control [5]–[7]. a distributed cooperative control with coordinate rotational
However, this control method is easy to be affected by the load virtual impedance, which needs a leader in the system and
change and disproportionate line impedance of the MG, which will fail when the leader fails. Y. Zhu [19] presented an
Manuscript received October 15, 2019; revised December 25, 2019; ac- enhanced optimization method with virtual impedance. Using
cepted February 17, 2020. Date of online publication April 6, 2020; date of this method, the reactive power sharing in the MG is obtained.
current version August 14, 2020. This work was supported by the Natural H. Xu [20] used a virtual capacitor method. However, the
Science Foundation of Shandong Province, China (No. ZR2017MEE053).
X. S. Wang, J. C. Zhang (corresponding author, e-mail: zhangjianchenghit@ flexibility of the system using these two methods is poor
126.com), and M. L. Zheng are with the Department of Control Science under switching topology due to no communication between
and Engineering, Harbin Institute of Technology, No. 2 Culture West Road, the DGs. The methods mentioned above realize reactive power
Weihai, China.
L. Y. Ma is with the Department of Information Engineering, Shandong sharing, but each has its own shortcomings. Moreover, the MG
Management University, No. 3500 Dingxiang Road, Jinan, China. should work normally when the communication structure of
DOI: 10.17775/CSEEJPES.2019.00530 the MG changes, i.e., it is very important to improve the plug-
2096-0042 © 2019 CSEE
WANG et al.: A DISTRIBUTED REACTIVE POWER SHARING APPROACH IN MICROGRIDS WITH IMPROVED DROOP CONTROL 1239

and-play capability. So we will consider the reactive power the rated values by a conventional droop control. Once the
sharing problem based on a distributed cooperative algorithm sum of the nominal power is not equal to the power at the
under switching topology. output of the inverter, the voltage and frequency of each DG
In this paper, distributed hierarchical control is used to will deviate from the reference values respectively. According
obtain power sharing. Dual loop control is used as a zero- to the droop characteristics of P/ω and Q/e, the active and
level control. The primary control consists of droop control, reactive power sharing can be achieved if the voltage and
and the secondary control is introduced to eliminate deviations frequency are synchronized.
and share active and reactive power. Being different form [21]– (
ωi∗ = ωrated − mi Pi
[23], the system does not require a leader, and transmits the (1)
voltage and measured power values as information between e∗i = erated − ni Qi
two DGs. The stability of the control strategy is proved by The conventional droop controllers’ theory is (1), where
multi-agent theory. The proposed control method is verified ωi∗ is the reference angular frequency output by the primary
in a MATLAB/SIMULINK environment. controller [21], e∗i is the reference voltage value output by
This paper is organized as follows. The proposed hi- the primary controller, Pi and Qi are the measured active
erarchical control structure is modeled in Section II. The and reactive power of i th DG, mi and ni are the droop
power sharing and stability of the MG are demonstrated in coefficients, ωrated and erated are the rated voltage and rated
Section III. The proposed control method is verified using frequency respectively.
MATLAB/SIMULINK in Section IV. The conclusion of this In practice, the DGs are widely distributed, and their
paper is presented in Section V. distance from the load is different. So the link impedances
of the DGs are not equal. Although the conventional droop
II. D ISTRIBUTED C ONTROL F RAMEWORK control can stabilize the frequency and voltage of the MG, the
deviations of frequency and voltage from their rated values
To improve the reliability of the MG, achieve the power will occur when the system operates stably. As a result, the
sharing and compensate for the deviations, this paper presents active and reactive power sharing will not be obtained under
a distributed hierarchical control method. We will discuss it in complex line impedance or unbalanced loads. At this time, it
detail next. is necessary to introduce a secondary control.
A. Primary Control B. Secondary Control
The conventional droop control is usually used as the To eliminate the deviations of frequency and voltage from
primary control, which can stabilize the voltage and frequency their rated values and obtain power sharing, the secondary
of the system [24]. The structure diagram of the DG based on control is indispensable. Figure 2 is the proposed distributed
an inverter is shown in Fig. 1. hierarchical control framework for the MG.
When the line impedance of the MG is inductive, the Inspired by the properties of multi-agent consistency, this
frequency and voltage of each DG will be synchronized to paper establishes a secondary controller, as shown in Fig. 2.

Physical system of DG

R Lf Lc Grid
ea
eb
ec

Cf PCC
LC filter

ilabc iabc uabc

θ
abc θ iodq
dq abc
dq
PWM ildq vodq Voltage
Power
Measurement Measurement

Current P Q
Controller ω*
Primary
Control
e*
Voltage Secondary e
Controller Control
Dual Loop Control Pjnorm ej Q norm
j

Proposed hierarchical control structure

Fig. 1. Internal structure diagram of a DG.

1240 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 7, NO. 6, NOVEMBER 2021

Secondary control erated Q1

n1
Voltage regulator _
+
t _ m1P1 P1
e (t ) δe11 v1* +
e1 (t ) + ∫ ∑ aa1 j (e j ( τ ) − e1 ( τ ))dτ 1
G(s) ∑ m1
0 j∈ N i
PI Control Primary control
δQ1 δe12
∑ ba
j∈ N i
1j (Q norm
j −Q1
norm
) H(s) DG1
P1 Q1 e1*
PI Control
Reactive power regulator
m1 n1 ωrated m1P1 Dual Loop
Control PWM
Active power regulator + _
mopt1 nopt1 δω1 + +
∑ ca ω *
1j ( Pjnorm − P1norm ) 1

Popt1 Qopt1 j∈ N i zero-level Physical

control system of DG
Tertiary
control e1 P1norm Q1norm Q norm
j Pjnorm ej
composed
of Communication graph PCC
a central
en Pnnorm Qnnorm Q norm Pjnorm ej
controller j
erated Qn
Poptn Qoptn
nn
+ _
moptn noptn t
en _ δe1n vn* +
en (t ) + ∫ ∑ aanj (e j ( τ ) − en ( τ ))dτ G(s) ∑
mn nn 0 j∈ N i

δQn δen2
Pn Qn
∑ ba
j∈ N i
nj (Q norm
j − Qnnorm ) H(s) en*
Pn Dual Loop
mn PWM
ωrated Control
+ _ ωn*
δωn + + DGn
∑ ca
j∈ N i
nj ( Pjnorm − Pnnorm )

Fig. 2. The proposed structure of a distributed hierarchical control.

The controller consists of three parts: voltage part, reactive the voltage of the DG are adjusted towards the rated value.
power part and active power part. Similar to the voltage regulator, the deviation generated by
The voltage regulator generates correction δe1i , and the the reactive power regulator is shown as:
reactive power regulator generates correction δe2i . These two X
corrections eliminate the deviation of voltage from the rated δQi = baij (Qnorm
j − Qnorm
i ) (4)
j∈Ni
voltage. The equation is shown in (2).
vi∗ (t) = erated + δe1i + δe2i (2) where Qnorm
i = Qi /Qrated
i , b is a designed parameter. The
purpose of this regulator is to make the nominal reactive power
The MG system composed of DGs is considered as a multi- of each DG tend to be equal, so as to obtain reactive power
agent system (MAS). In the MAS, the agent interacts with sharing.
adjacent agents through distributed communication systems, Then the deviation δQi is fed back to the PI controller H(s),
and then judges its own operational status. It adjusts the and the PI controller produces another correction term δe2i
output according to this information, so that it is identical with after eliminating static error.
the target features of adjacent agents. Overall consistency is The function of the active power regulator is to adjust the
finally achieved. Using this property, the output of the voltage frequency of the DG to the rated value, and make the nominal
estimator is: active power of each DG tend to be equal, so as to obtain
Z t X
active power sharing. The deviation δωi generated by the
ēi (t) = ei (t) + aaij (ēj (τ ) − ēi (τ ))dτ (3) active power regulator based on MAS is shown as:
0 j∈N
i X
where aij represents the communication weight a is a designed δωi = caij (Pjnorm − Pinorm ) (5)
j∈Ni
parameter. The function of the voltage estimator is to estimate
the average voltage value of the MG. where Pinorm = Pi /Pirated , c is a designed parameter. By the
The deviation between predicted voltage and rated voltage is above equations the angular frequency set-point is:
fed back to the PI controller G(s), and then the PI controller
produces the first correction term δe1i after eliminating the ωi∗ (t) = ωrated + δωi (t) − mi Pi (6)
static error. Through the function of a voltage regulator, the So the voltage setting is:
average value will converge to the nominal voltage of the DG, √
but there may be a deviation between the voltage value of Ei∗ = e∗i 2 sin(θi ) (7)
the DG and the rated voltage value. So it is necessary to
The voltage set-points and phase angle are as follows:
introduce another correction term δe2i . After the adjustment
of the secondary control, the average voltage of the MG and e∗i = erated + δe1i + δe2i − ni Qi (8)
WANG et al.: A DISTRIBUTED REACTIVE POWER SHARING APPROACH IN MICROGRIDS WITH IMPROVED DROOP CONTROL 1241

Z t
θi = ωi∗ (t)dτ (9) two levels is decoupled. The correction terms δe1i and δe2i are
0 written as:
where δe1iand δe2iare described in detail in the next section. δe1i = G(erated − ēi ) (13)
When the primary controller is used to adjust the voltage,
δe2i = H · δQi (14)
the voltage is usually transferred to a d / q coordinate system
through coordinate transformation. The controller adjusts the Considering the whole MG, the equations (13) and (14) can
component of the output voltage on the d-axis to make it equal be written in matrix form, and expressed as:
to the voltage set-point [21]. Therefore, the second equation
of (1) can be written as: δe1 = δe10 + (GP + GI (t − t0 ))(erated − ē) (15)
2
( δe = δe20 + (Hp + HI (t − t0 ))δQ (16)
Udrefi = erated − ni Qi
(10)
Uqrefi = 0 where GI and GP are the diagonal matrices carrying the
integral and proportional gains of the voltage-regulator matrix
where Udrefi and Uqrefi are the d and q axis components of e∗i , G such that GP + GI /s = G. HI and HP are similar
respectively. to GI and GP respectively. ē = [ē1 , ē2 , · · · , ēn ]T denotes
From the above equations, we can see that the function of the voltage estimation vector, δe10 and δe20 are the initial
the voltage regulator and reactive power regulator is to elimi- values of the integrators GI and HI at t = t0 , respectively.
nate the voltage deviation and maintain the voltage amplitude δQ = [δQ1 , δQ2 , · · · , δQn ]T , all elements of erated ∈ RN ×1
at the rated value, so as to realize reactive power sharing. are made up of erated .
The active power regulator maintains the frequency at the Next, we take the differential on both sides of (3), and there
rated value, and eliminates the active power deviation from is:
the adjacent DG by adjusting the set-points of the phase angle X
to achieve active power sharing. ē˙ i (t) = ėi (t) + aij (ēj (t) − ēi (t))
j∈Ni
C. Dual Loop Control X
= ėi (t) + aij ēj (t) − din
i ēi (17)
The dual loop controller consists of a voltage loop controller j∈Ni
and a current loop controller. The outer loop is the voltage loop
and the inner loop is the current loop. The dual loop controller considering the whole MG, the equation (17) can be written
and the droop controller work together to stabilize the voltage in matrix form, expressed as:
and frequency of the DG. The location of this controller in ˙ = Ė + A Ē − D in Ē
Ē G G
the system is shown in Fig. 2. The model of the outer loop is in
shown in (11) [21], = Ė + (AG − DG )Ē
 ∗ = Ė − LĒ (18)
i = F iod − ωref Cf voq + KPV (Udref − vod )
 ld


Therefore the above equation is expressed in frequency
R
 +KIV (Udref − vod )dt
(11) domain as:


 i∗lq = F ioq + ωref Cf vod + KPV (Uqref − voq )
Ē = s(sIN + L)−1 E
R
+KIV (Uqref − voq )dt

(19)
where KPV is the gain coefficient and KIV is the integral where IN ∈ RN ×N , e = [e1 , e2 , · · · , en ]T is the voltage
coefficient. Cf is the capacitance of the LC filter. The model measurement vector, E and Ē are the Laplace transforms
of the inner loop is shown in (12): of e and ē, respectively. AG represents the communication
in
 Z weightPmatrix, DG = diag{dini } is a diagonal matrix with
 vPWMd = −ωref Lf ilq +KPI (i∗ld − ild )+KII (i∗ld − ild )dt
 in
di = j∈Ni aij , L is a Laplacian matrix in Graph Theory.
According to reference [14], a communication structure with
Z
 vPWMq = ωref Lf ild + KPI (i∗lq − ilq ) + KII (i∗lq − ilq )dt

a spanning tree has the following relationships,
(12)
lim s(sIN + L)−1 = M
s→0
where KPI is the gain coefficient and KII is the integral
coefficient. Lf is the inductor of the LC filter. where M ∈ RN ×N is a matrix in which all elements are 1/N .
Hierarchical control also includes tertiary control used as the Using this property and the final value theorem, equation (3)
scheduling layer, as shown in Fig. 2. Tertiary layer composed can be written as:
of a central controller usually controls the power flow, and is
often used when considering the cost of generating electricity, lim ē(t) = lim sĒ = lim s(sIN + L)−1 (sE)
t→∞ s→0 s→0
which is not the focus of this paper. = lim s(sIN + L)−1 × lim (sE)
s→0 s→0
= M × lim (e) = ⟨ess ⟩1
III. S TEADY- STATE A NALYSIS t→∞
ss N ×1
The primary control and the secondary control operates where e ∈ R is the steady state value of vector e. ⟨e⟩
with different timeframes, thus the controller design of the is the average of all elements of vector e. All elements of
1242 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 7, NO. 6, NOVEMBER 2021

Load1 Load4
1 ∈ RN ×1 are 1. For the whole system, (4) can be written in vo1
Lc1 Lc4
vo4
matrix form, expressed as: DG1 DG4

δQ = −bLQnorm (20) RLine1 RLine3

When the system is stable, (15) and (16) can be written as LLine1 LLine3
(21) and (22) through the above derivation, respectively. Then: Lc2 Lc3
vo2 vo3
DG2 DG3
δe1 = δe10 + (GP + GI (t − t0 ))(erated − M ess ) (21)
Load2 Load3
δe2 = δe20 + (Hp + HI (t − t0 ))(−bLQnorm, ss ) (22) RLine2 LLine2
norm, ss norm
h Q
where is the stable iT state value of vector Q , Qnorm
Fig. 3. Schematic of the MG physical system.
= QQrated
1
, QQrated
2
, · · · , QQrated
n
. As shown in Fig. 2, the reference
1 2 n
voltage value output by the proposed controller in stable state
can be expressed as: DG1 DG4
∗ss 1 2 ss
e = erated + δe + δe − N Q
= erated + δe10 + δe20 + GP (erated − ⟨ess ⟩1)
− bHP LQnorm, ss + (GI (erated − ⟨ess ⟩1)
− bHI LQnorm, ss )(t − t0 ) − N Qss (23) DG2 DG3
ss ∗ss
where N = diag{ni }, Q and e is the steady-state value
of Q = [Q1 , Q2 , · · · , Qn ]T and e∗ = [e∗1 , e∗2 , · · · , e∗n ]T , Fig. 4. Communication network.
respectively. Equation (23) holds for all t ≥ t0 . Therefore,
the time-varying part in (23) is zero. Then: TABLE I
L INE I MPEDANCE PARAMETERS OF MG
−1
b HI−1 GI (erated ss
− ⟨e ⟩1) = LQ norm, ss
(24)
Symbol Quantity Value
Multiplying (24) by 1T , (25) can be obtained. ZLine1 Line impedance 0.8 + 3.6j
ZLine2 Line impedance 0.4 + 1.8j
1T b−1 HI−1 GI (erated − ⟨ess ⟩1) = 1T LQnorm, ss (25) ZLine3 Line impedance 0.7 + 1.2j
T
Because the Laplace matrix satisfies the relationship 1 L =
0 [14], and 1T b−1 HI−1 GI ̸= 0, then: TABLE II
S IMULATION PARAMETERS
ss
erated = ⟨e ⟩1 (26)
Symbol Quantity DG1&DG2 DG3&DG4
So it can be obtained that, Udc Dc-bus voltage 650 V 650 V
τQi Reactive power filter 0.04 s 0.04 s
erated = ⟨ess ⟩ (27) time constant
τP i Active filter time 0.04 s 0.04 s
It can be seen from (27) that the average voltage of all DGs constant
Rc Output end resistance 0.12&0.03 Ω 0.06&0.09 Ω
in the MG converges to the rated value under stable operation Lc Output end inductance 1.4&0.35 mH 0.7&1.05 mH
of the MG. By substituting erated − ⟨ess ⟩1 = 0 into (24), ni Reactive droop 0.01 V/Var 0.02 V/Var
coefficient
LQnorm, ss = 0 (28) 4 × 10−4 8 × 10−4
mi Active droop
rad/(s·W) rad/(s·W)
According to [25], the only nonzero solution of the equation coefficient
of Lx = 0 is x = k1, k is a real number. So according Q Reactive power load 0&200 Var 0&300 Var
P Active power load 0&600 W 0&400 W
to (28), equation Qnorm, ss = k1 holds. This means that the P rated Rated active power 800 W 400 W
nominal reactive power of each DG is equal to each other, Qrated Rated reactive power 600 Var 300 Var
i.e., the reactive power sharing is obtained. kPQ Gi (s) proportional 0.01 0.01
term
Similar to the previous proof, (29) can be derived from (6). kiQ Gi (s) integral term 3 3
kpv Hi (s) proportional 0.005 0.005
cLP norm, ss = 0 (29) term
norm, ss kiv Hi (s) integral term 2 2
hwhere P is the isteady state value of P norm , P norm = a Designed parameter 1.5 1.5
T
P1
, P2 , · · · , PPrated
P rated P rated
n
, there is b Designed parameter 30 30
1 2 n c Designed parameter 0.1 0.1
P norm, ss = ξ1 (30) Note: the designed parameter AG as follows:
0 1 0 1
 
where ξ is a real number. This completes the proof.
1 0 1 0
AG = 1.5 × 
0 1 0 1
IV. S IMULATION VALIDATION 1 0 1 0
The proposed control method is verified in this section. The
experimental setup is shown in Fig. 3, which consists of four The communication structure is shown in Fig. 4. The line
DGs to form a MG system. impedance parameters are shown in Table I, and the simulation
WANG et al.: A DISTRIBUTED REACTIVE POWER SHARING APPROACH IN MICROGRIDS WITH IMPROVED DROOP CONTROL 1243

315 200
DG1 DG2 DG3 DG4 DG1 DG2 DG3 DG4
150
314.5
ω (rad/s)

Vo (V)
120
314.2 100
314
50

313.5 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
t (s) t (s)
(a) (b)
1 15
DG1 DG2 DG3 DG4 DG1 DG2 DG3 DG4
10
0.5
5
m×P

n×Q
0
0 2
−5 1
0
−0.5 −10 0.55 0.6 0.65 0.7
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
t (s) t (s)
(c) (d)

Fig. 5. Comparison of two control methods (a) frequency (b) voltage (c) active power ratios (d) reactive power ratios.

315 200
DG1 DG2 DG3 DG4 DG1 DG2 DG3 DG4
150
314.5
ω (rad/s)

Vo (V)

120
314.2 100
314
50

313.5 0
0 5 10 15 0 5 10 15
t (s) t (s)
(a) (b)
1
DG1 DG2 DG3 DG4
0.4
0.5
0.2 0.3
m×P

0.1 0 0.2
0 0.1
4.8 5 5.2 5.4 5.6 5.8 9.8 10.2 10.6
−0.5
0 5 10 15
t (s)
(c)
15
DG1 DG2 DG3 DG4
10 3
2.5
2 5 2.5
n×Q

1.5 2
0
1 1.5
0.5 −5 1
4.8 5 5.2 5.4 5.6 5.8 9.8 10.2 10.6 11
−10
0 5 10 15
t (s)
(d)

Fig. 6. Switching load under fixed communication structure. (a) frequency (b) voltage (c) active power ratios (d) reactive power ratios.

parameters of each DG are listed in Table II given in the communication structure. The second case is to verify the
section Table. stability of the system under variable communication structure
The reference value of the voltage is 120 V, the reference
value of the angular frequency is 100 π (rad/s), the filter A. Fixed Communication Structure
inductance Lf is 1.8 mH, the filter resistance Rf is 0.1 Ω, In order to compare with the conventional control method,
and the filter capacitor Cf is 50 µF. The remaining parameters only the primary controller works from 0 s to 1 s, and the
are in Table II. Next, we verify the proposed method through proposed controller starts to work when t = 1 s. The total
two cases. The first case is to switch the load under the fixed simulation time is 3 seconds. The calculation results of the
1244 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 7, NO. 6, NOVEMBER 2021

DG1 DG4 DG1 DG4 the system change from the old balance point to the new
balance point. In the case of load change, the performance
verification of the proposed controller is shown in Fig. 6. The
proposed controller starts to work when t = 1 s. The load is
disconnected from the MG at t = 5 s and connected to the
MG at t = 10 s. As seen, the frequency and voltage return to
DG2 DG3 DG2 DG3 their reference values after a slight fluctuation, and different
power sharing points are achieved according to the variation
Fig. 7. Communication network configuration of switching topology. of the load. This means that the method proposed in this paper
is effective when the load fluctuates.
two controllers are shown in Fig. 5.
B. Variable Communication Structure
As shown in Fig. 5, the conventional droop controller
stabilizes the voltage and frequency, but they will deviate from In this section, we verify the effectiveness of the method
their reference values. Moreover, reactive power sharing is when the communication structure of the system changes. The
not obtained. Under the influence of the proposed hierarchical communication structure of the MG at t = 0 s is shown in
control, the frequency and voltage of each DG is regulated to Fig. 7. The proposed controller starts to work when t = 1 s.
the rated values respectively at about t = 1.35 s. It can be seen We assume that DG3 is disconnected from the MG at t = 5 s
from Fig. 5(c) and (d) that mi pi and ni qi are equal to each and connected to the MG at t = 10 s. Or DG3 is broken when
other respectively, and the curves are completely coincident, t = 5 s, and resumed when t = 10 s.
i.e., the active and reactive power sharing are all obtained. As The links of every switching topology still form a connected
the voltage is synchronized to the reference value, the loads graph, thus the control method will remain functional. The
operate at rated voltage. So the total power emitted by each performance verification of the proposed controller is shown
DG increases. in Fig. 8.
The load of the MG often changes suddenly in actual use. It can be seen from Fig. 8 that the control strategy proposed
This will cause the power of each DG to change. When makes the frequency and voltage stable at their respective rates
the load fluctuates, a good control method can quickly make when the topology is switching. The active and reactive powers

315 200
DG1 DG2 DG3 DG4 DG1 DG2 DG3 DG4
150
314.5
ω (rad/s)

Vo (V)

120
314.2 100
314
50

313.5 0
0 5 10 15 0 5 10 15
t (s) t (s)
(a) (b)
1
DG1 DG2 DG3 DG4

0.2 0.5 0.2

0.15 0.15
m×P

0.1 0.1
0.05 0 0.05
0 0
4.8 5.2 5.6 9.8 10.2 10.6
−0.5
0 5 10 15
t (s)
(c)
10 DG1 DG2 DG3 DG4

1.5 5 1.5
1
n×Q

0 1
0.5
0 −5 0.5
4.9 5.1 5.3 5.5 9.9 10.1 10.3
−10
0 5 10 15
t (s)
(d)

Fig. 8. Under switching communication structure. (a) frequency. (b) voltage. (c) active power ratios. (d) reactive power ratios.
WANG et al.: A DISTRIBUTED REACTIVE POWER SHARING APPROACH IN MICROGRIDS WITH IMPROVED DROOP CONTROL 1245

of DG3 drop to zero at t = 5 s, while the powers supplied [7] Y. W. Zheng, M. Y. Chen, C. Li, R. L. Xu, and X. Xu, “A microgrid
by other DGs keep sharing. When the topology is switched control strategy based on adaptive drooping coefficient adjustment,”
Automation of Electric Power Systems, vol. 37, no. 7, pp. 6–11, Apr.
back at t = 10 s, the power sharing is again obtained after 2013.
about 0.4 s. Similar to the DG3 disconnection from the MG, if [8] Y. Gao and Q. Ai, “A distributed coordinated economic droop control
DG3 fails, the rest of the MG can work normally. These good scheme for islanded AC microgrid considering communication system,”
Electric Power Systems Research, vol. 160, pp. 109–118, Jul. 2018.
performances show that the control method can enhance the [9] Z. Y. Yu, M. Lu, Z. N. Wang, and Y. G. Zhang, “A droop control
flexibility and scalability of the MG and improve the quality strategy with impedance compensation for low voltage microgrid,”
of the power supply. The above cases show the effectiveness Applied Mechanics and Materials, vol. 441, pp. 245–248, Dec. 2013.
[10] J. W. He, Y. W. Pan, B. H. Liang, and C. S. Wang, “A simple
of the proposed method in this paper. decentralized islanding microgrid power sharing method without using
droop control,” IEEE Transactions on Smart Grid, vol. 9, no. 6, pp.
6128–6139, Nov. 2018.
V. C ONCLUSION [11] A. Micallef, M. Apap, C. S. Staines, and J. M. Guerrero Zapata,
A distributed reactive power sharing approach in the MG “Secondary control for reactive power sharing and voltage amplitude
restoration in droop-controlled islanded microgrids,” in Proceedings of
with improved droop control is proposed in this paper. This the 2012 3rd IEEE International Symposium on Power Electronics for
method adopts a hierarchical control structure, which not only Distributed Generation Systems, 2012.
retains the advantages of conventional droop control, but also [12] H. H. Xin, R. Zhao, L. Q. Zhang, Z. Wang, K. P. Wong, and W. Wei, “A
decentralized hierarchical control structure and self-optimizing control
makes up for its shortcomings. Based on the properties of strategy for F-P type DGs in islanded microgrids,” IEEE Transactions
multi-agent theory, a secondary controller is constructed. The on Smart Grid, vol. 7, no. 1, pp. 3–5, Jan. 2016.
stability and feasibility of the controller is proved by the [13] H. Z. Xu, X. Zhang, F. Liu, R. L. Shi, C. Z. Yu, and R. X. Cao,
“A reactive power sharing strategy of VSG based on virtual capacitor
final value theorem and multi-agent consistency. The proposed algorithm,” IEEE Transactions on Industrial Electronics, vol. 64, no. 9,
control strategy requires a lesser amount of communication pp. 7520–7531, Sep. 2017.
connection and information exchange, which realizes the [14] R. Olfati-Saber and R. M. Murray, “Consensus problems in networks
of agents with switching topology and time-delays,” IEEE Transactions
power sharing and adjusts the voltage and frequency of the MG on Automatic Control, vol. 49, no. 9, pp. 1520–1533, Sep. 2004.
to rated values respectively. This method saves the investment [15] N. M. Dehkordi, N. Sadati, and M. Hamzeh, “Fully distributed cooper-
of communication equipment and circuits. At the same time, ative secondary frequency and voltage control of islanded microgrids,”
IEEE Transactions on Energy Conversion, vol. 32, no. 2, pp. 675–685,
the proposed control strategy can enhance the flexibility and Jun. 2017.
reliability of the MG and improve the quality of the power sup- [16] W. Liu, W. Gu, W. X. Sheng, X. L. Meng, Z. J. Wu, and W. Chen,
ply. The simulation results of fixed communication structure “Decentralized multi-agent system-based cooperative frequency control
for autonomous microgrids with communication constraints,” IEEE
and variable communication structure show that the proposed Transactions on Sustainable Energy, vol. 5, no. 2, pp. 446–456, Apr.
control strategy can achieve accurate power sharing. 2014.
At present, there are many methods to realize reactive [17] X. S. Wang, H. Q. Zhang, and C. X. Li, “Distributed finite-time
cooperative control of droop-controlled microgrids under switching
power sharing. But these methods have not been applied in topology,” IET Renewable Power Generation, vol. 11, no. 5, pp. 707–
practice. The method proposed in this paper provides a new 714, 2017.
way of thinking and research direction for large-scale practical [18] Z. Zhang and L. Jiang, “Distributed cooperative control of microgrids
with unbalance impedances,” in Proceedings of the 35th Chinese Control
applications in the future. The application of the proposed Conference, 2016.
method in practice will be the focus of the author’s future [19] Y. X. Zhu, Q. H. Fan, B. Q. Liu, and T. Wang, “An enhanced
research study. virtual impedance optimization method for reactive power sharing in
microgrids,” IEEE Transactions on Power Electronics, vol. 33, no. 12,
pp. 10390–10402, Dec. 2018.
[20] H. Z. Xu, X. Zhang, F. Liu, R. L. Shi, C. Z. Yu, W. Zhao, Y. Yu, and
R EFERENCES W. Cao, “A reactive power sharing method based on virtual capacitor
[1] C. C. Cai, S. C. Cheng, B. Jiang, W. L. Dai, M. Wu, and J. Y. Zhang, in islanding microgrid,” in Proceedings of 2014 International Power
“Optimal operation of microgrid composed of small hydropower and Electronics Conference, 2014, pp. 567–572.
photovoltaic generation with energy storage based on multiple scenarios [21] A. Bidram, A. Davoudi, F. L. Lewis, and J. M. Guerrero, “Distributed co-
technique,” in Proceedings of the 2015 5th International Conference on operative secondary control of microgrids using feedback linearization,”
Electric Utility Deregulation and Restructuring and Power Technologies, IEEE Transactions on Power Systems, vol. 28, no. 3, pp. 3462–3470,
2015. Aug. 2013.
[2] C. X. Dou, Z. Q. Zhang, D. Yue, and Y. H. Zheng, “MAS-based [22] F. H. Guo, C. Y. Wen, J. F. Mao, and Y. D. Song, “Distributed secondary
hierarchical distributed coordinate control strategy of virtual power voltage and frequency restoration control of droop-controlled inverter-
source voltage in low-voltage microgrid,” IEEE Access, vol. 5, pp. based microgrids,” IEEE Transactions on Industrial Electronics, vol. 62,
11381–11390, 2017. no. 7, pp. 4355–4364, Jul. 2015.
[3] S. Parhizi, H. Lotfi, A. Khodaei, and S. Bahramirad, “State of the art [23] N. M. Dehkordi, N. Sadati, and M. Hamzeh, “Fully distributed cooper-
in research on microgrids: a review,” IEEE Access, vol. 3, pp. 890–925, ative secondary frequency and voltage control of islanded microgrids,”
2015. IEEE Transactions on Energy Conversion, vol. 32, no. 2, pp. 675–685,
[4] Y. X. Zhu, F. Zhuo, F. Wang, and B. Q. Liu, “Virtual impedance Jun. 2017.
optimization method for microgrid reactive power sharing control,” [24] A. Bidram and A. Davoudi, “Hierarchical structure of microgrids control
Proceedings of the CSEE, vol. 36, no. 17, pp. 4552–4563, Sep. 2016. system,” IEEE Transactions on Smart Grid, vol. 3, no. 4, pp. 1963–1976,
[5] H. Wang and G. Q. Li, “Control strategy of microgrid with different Dec. 2012.
DG types,” Electric Power Automation Equipment, vol. 32, no. 5, pp. [25] V. Nasirian, S. Moayedi, A. Davoudi, and F. L. Lewis, “Distributed
19–23, May 2012. cooperative control of DC microgrids,” IEEE Transactions on Power
[6] J. Y. Huang, C. W. Jiang, and R. Xu, “A review on distributed energy Electronics, vol. 30, no. 4, pp. 2288–2303, Apr. 2015.
resources and microgrid,” Renewable and Sustainable Energy Reviews,
vol. 12, no. 9, pp. 2472–2483, Dec. 2008.
1246 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 7, NO. 6, NOVEMBER 2021

Xinsheng Wang received the B.S., M.S. and Ph.D. Mingli Zheng received the B.S. degree from China
degrees from Harbin Institute of Technology, Harbin, Jiliang University, Hangzhou, China, in 2015, her
China, in 1992, 1995 and 2002, respectively. She M.S. degree from Harbin Institute of Technology,
is now an Associate Professor of Control Science Weihai, China, in 2018, and she is now engaged in
and Engineering at Harbin Institute of Technology, the research and development of automatic driving
Weihai, China. Her research interests include ro- decision algorithms at Zhengzhou Yutong Bus Co.,
bust and nonlinear control, power electronics system Ltd. Her research interests include multi-agent coop-
modeling and control. erative control and autonomous driving technologies.

Jiancheng Zhang received the B.S. degree from Lingyu Ma received the B.S. degree from Shandong
Hebei Polytechnic University, Tangshan, China, in Jianzhu University, Jinan, China, in 2011, and her
2010, his M.S. degree from Harbin Institute of M.S. degree from Harbin Institute of Technology,
Technology, Shenzhen, China, in 2013, and he is Shenzhen, China, in 2013. She is now a Lecturer
currently pursuing his Ph.D. degree at Harbin In- of Control Science and Engineering at Shandong
stitute of Technology, Harbin, China. His research Management University, Jinan, China. Her research
interests include multi-agent cooperative control and interests include robust and nonlinear control, devel-
grid-connected algorithm for microgrids. opment and application of embedded system.