This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2017.2656062, IEEE
Transactions on Power Delivery

Stability Analysis of Two Parallel Converters with

Voltage-Current Droop Control
Yin Li, Student Member, IEEE and Lingling Fan, Senior Member, IEEE

Abstract—Voltage-current (V-I) droop control has been pro- achieved through dq-axis current sharing. The current sharing
posed for real and reactive power sharing. Compared with is inversely proportional to the ratio of droop coefficients.
the conventional droop control using real power/frequency and Larger droop coefficients, mk or nk , may lead DERk ’s voltage
reactive power/voltage, it enhances stability since adding a V-I
droop is equivalent to providing a resistor. However, for a system blow the range under heavy loading conditions. On the other
with parallel converters with V-I droops, oscillations are observed hand, small droop coefficients may lead to inaccurate power
when small droop coefficients are adopted. In this paper, multi- sharing according to [8].
input multi-output (MIMO) model of the system is derived in Although V-I droop control achieves faster dynamics, for a
frequency domain. This model is based on dq-reference frame and system with multiple converters, there are possible oscillation
consists of converter control and power network model, which
is represented by an admittance matrix. Linear system analysis issues, which have not been identified in the literature. [8] has
is carried out to identify the root causes of oscillations in both multiple converters with V-I droop, but the oscillation issue
grid-connected mode and autonomous mode. Analysis results are is not observed. In this paper, stability investigation will be
validated by simulation results of detailed model-based systems conducted to identify scenarios when oscillations may occur.
built in Matlab/Simpowersystems. Stability issue is very common in multiple parallel-
Index Terms—V-I droop, current sharing, MIMO model, droop connected converters with droop control and have been studied
coefficient, oscillations in the previous literature [11], [13]–[16]. Eigenvalue analysis
based on linearized systems is a popular approach to analyze
I. I NTRODUCTION the stability issues in multiple converters systems [14]–[19].
In such system models, converters are modeled as voltage

M ORE and more distributed energy resources (DERs)

are integrated in microgrids. The conventional droop
control achieves real and reactive power sharing by introducing
sources with control loop dynamics and output filter dynamics
while the switching dynamics are ignored [20]–[23]. Based
on the stability analysis, several methods have been proposed
additional feedback loops, e.g., power-frequency and VAr- in the previous literature to enhance system stability such
voltage droop. Such type of droops (ω = ω ∗ −m(P −P ∗ ), E = as a controller based on the second derivative of the out-
E ∗ − n(Q − Q∗ )) is adopted in [1], [2] where the converter put capacitor voltage [16], arctan gradient algorithm concept
works as a current source. For converters operating as a [19], an adaptive decentralized droop controller [11], virtual
voltage source (in voltage/frequency control mode), frequency- complex impedance [14], and high gain angle droop control
power droop (P = P ∗ − m(ω − ω ∗ )) is adopted [3], [4]. [15]. However, for parallel converters with V-I droops, stability
As a variation of power-frequency (P − f ) droop, power- issues have not been identified and studied. In our research,
angle (P − δ) droop can eliminate the frequency deviation oscillation issues are observed for a system with parallel
[5], [6]. For the pure resistive network, P − V droop is used converters with different V-I droop coefficients. Different from
normally [7]. Nonetheless, the power control block in the the conventional droop control, the stability issue of V-I droop
conventional droop control has a low bandwidth to cause the is caused by the smaller droop coefficients.
slow dynamic problem regardless increasing or decreasing the The objective of this paper is to derive the linear model
droop coefficients [8]–[11]. of a system with parallel converters with V-I droop and find
While power-frequency droop can achieve power sharing the root causes of oscillations. The paper is organized as
according to the droop coefficients, it is known that E − Q follows. The topology of the microgrid system with V-I droop
droop based reactive power sharing is inaccurate due to line and the steady-state analysis will be introduced in Section II.
impedance [12]. Section III presents the MIMO model in frequency domain
To achieve fast dynamics and accurate real power and and linear analysis of the MIMO model. Section IV validates
reactive power sharing, V-I droop is proposed in [8], which is the linear analysis using detailed model simulation results
suitable for microgrids with small inertia DERs and frequent via MATLAB/Simpowersystem. Section V gives simulation
load variation. It utilizes the function of output current (Ed∗ = results for the CIGRE microgrid benchmark test system.
E0 − miLd , Eq∗ = 0 − niLq ) to adjust a DER’s output voltage Section VI concludes this paper.
references. In a microgrid consisting of multiple converters
equipped with V-I droops, real and reactive power sharing is II. T HE M ICROGRID WITH V-I DROOP
A. Topology of the circuit
Y. Li and L. Fan are with Dept. of Electrical Engineering, Univer-
sity of South Florida, Tampa FL 33620. Emails: [email protected], In the microgrid with V-I droop, the global positioning sys-
[email protected]. tem (GPS) signals are required to keep the system frequency

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2017.2656062, IEEE
Transactions on Power Delivery

VPCC (rated)=110V If we assume that mk  Rk , nk  Rk , mk  Xk , and

L-L
VSC PCC nk  Xk , then we will have the following relationship:
VDC=400V L=10mH R=100mΩ LL=50uH
1
i1 C=500uF
+ iL1 m1 iLd1 = m2 iLd2 (6a)
m1=0.2 E1 (rated)=110V iPCC 1.61+j0.81Ω
n1=1.5 - n1 iLq1 = n2 iLq2 (6b)
RL
VDC=400V
VSC Closed at 0.7s Therefore, the real power sharing is proportional to 1/mk
L=10mH R=100mΩ LL=50uH
2 +
Open at 1.5s
and the reactive power sharing is proportional to 1/nk . The
m2=0.4 i2 C=500uF iL2 R 4.03Ω
n2=3
E1 (rated)=110V assumptions are that mk  Rk , nk  Rk , mk  Xk , and
-
nk  Xk . If we have a resistive network or Rk  Xk , then
the real power sharing is according to 1/(Rk + mk ) and the
Fig. 1: Two DERs support one load through parallel VSCs.
reactive power sharing is according to 1/(Rk + nk ).
If the droop coefficients mk and nk are comparable with
Rk or Xk , then we cannot obtain accurate real/reactive power
as constant. The constant frequency, ω0 , is used to synchronize
sharing. On the other hand, large droop coefficients may lead
all DER voltages and currents in the same dq-frame (d + jq).
converter voltages drop below the range during heavy load
It is assumed that for the initial condition, VP CC is aligned
conditions.
at the d-axis (VP CCd = VP CC , VP CCq = 0). In this paper,
a microgrid consisting of two DERs and one load is used to
analyze the characteristics of V-I droop control. Each DER
C. Control loop design
contains a voltage source converter (VSC) and an RLC filter.
Two VSCs are controlled in coordination by V-I droop control. V-I droop control is based on voltage/frequency (V /F )
There is a pure inductive transmission line (LL ) between DER control where the output voltage references are sent to the
and the point of common coupling (PCC). The reactance XL outer loops and frequency is fixed as shown in Fig. 2.
of the line is negligible compared with droop coefficients. 1) Inner current loop: For the voltage source converter, the
block diagram shown in Fig. 3 is normally used to design the
PI controllers in the inner current loop [24].
B. Steady State Analysis 1
The plant model is represented by Ls+R where ud = Vtd −
Assume that the two converters (including the RLC filter) Ed − ωLiq and uq = Vtq − Eq + ωLid (Vtd and Vtq are
are connected to the PCC bus through an RL impedance Rk + the converter output voltages. Ed and Eq are the capacitor
jXk , k = 1, 2. The complex power injected to the load or the voltage.). Note that in Fig. 2, the feedforward components are
grid through the PCC bus is notated as Sk , k = 1, 2. Let the added back to uid and uiq to generate the references for the
d-axis be aligned with the PCC voltage space vector (vd = converter voltages. Kip + Ksii is the controller. Kip and Kii
|−
→
v P CC |). Then the complex power expression is are the proportional gain and integrator gain which can be
3 3 designed using (7) [24].
Sk = vd (iLdk − jiLqk ) = (vd iLdk − jvd iLqk ) (1)
2 2
Therefore, L R
Kip = , Kii = (7)
3 τi τi
Pk = vd iLdk (2a)
2 where τi is the inner loop time constant.
3
Qk = − vd iLqk (2b) Based on Fig. 3 and Eq. (7), the closed-loop transfer
2
function is as follows.
Therefore, the d-axis current sharing determines the real power
sharing while the q-axis current sharing determines the reactive id Gio 1
Gic = = = (8)
power sharing. i∗d 1 + Gio τi s + 1
At steady-state, the converter voltage, PCC voltage and the
current have the following relationship. 2) Outer voltage loop: Fig. 4 shows the block diagram for
       the outer loop to design the gains of PI controller in the outer
Rk −Xk iLdk Ekd v 1
loop. τi s+1 , the closed-loop transfer function from the current
= − d (3)
Xk Rk iLqk Ekq 0 reference to the current measurement, is included in the plant
model.
If we apply the V-I droop and assume the droop coefficients are
mk for d-axis and nk for q-axis, then we have the following The gains Kvp and Kvi can be calculated using the follow-
relationship. ing equations [24], [25] .

Ekd
     
v E0 mk 0

iLdk
  
v C Kvp
− d = − − d (4) Kvp = Cωc = √ , Kvi = (9)
Ekq 0 0 0 nk iLqk 0 τi τv τv
      
Rk + mk −Xk iLdk E0 vd where ωc is the cut-off frequency and τv is the outer loop time
⇒ = − (5)
Xk Rk + nk iLqk 0 0 constant.

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2017.2656062, IEEE
Transactions on Power Delivery

E0
iLd Ed
+ * *
iLd -m
+ Ed + +
+
id + ud +
+
Vtd*
PI PI
-
-
- - dq Vta*
Ed ωC id ωL
Vtb*
Eq ωC iq ωL

+ Eq * - + iq * -
uq + Vtq* abc Vtc*
iLq -n + PI + + PI +

+ + +
iLq Eq
0 θ=ω0t
(GPS)
Fig. 2: Control block diagram for V-I droop.

+ ud with V /F control. Compared with Fig. 2, Fig. 5 includes the

1
id*(A) Kip+Kii/s
Ls+R
id (A) RLC filter between VSC and the transmission line as a part
- (V)
of dynamics of VSC. The dynamics of RCL filter is presented
1
a) in both the inner loop and outer loop plant models, Ls+R
1 1
and Cs . Ls+R has been combined with the current feedback
+ uq 1 control and feedforward elements. The resulting block from
iq*(A) Kip+Kii/s
Ls+R
iq (A)
-
(V) i∗d to id is a simple first-order unit (8).
There are totally four inputs (voltage references: Ed∗ , Eq∗ ,
b)
and load current measurements: iLd , iLq ) and two outputs (the
Fig. 3: The block diagram for the inner loop transfer function. capacitor voltages: Ed and Eq ) for each VSC with an RLC
filter. For the system with two converters, there are 8 inputs
and 4 outputs.
+ uvd* 1 1
Ed*(V) Kvp +Kvi /s
τi s+1 Cs Ed (V)
- (A)
iLd (A)
a) * + -
Ed + Kvp s+Kvi uvd +
*
id* 1 id + 1 Ed (V)
* + uvq* 1 1 (V) s (A) (A) τis+1 (A) Cs
Eq (V) Kvp +Kvi /s Eq (V) - - +
(A) τi s+1 Cs
- Ed (V) ωC ωC
b)
Eq (V) ωC ωC
Fig. 4: The block diagram for the outer loop transfer function. - + -
* * *
Eq + Kvp s+Kvi uvq + iq 1 iq + 1
(V) s (A) (A) τis+1 (A) Cs Eq (V)
+ -
3) Droop control loop: The voltage/current droop generates iLq (A)
a voltage reference based on current feedback. After com-
bining three layers of control loops, the whole control block Fig. 5: The block diagram of a VSC with an RLC filter.
diagram of V-I droop can be presented by Fig. 2. The droop
control is different from the one implemented in [8] in two 2) Transmission network: The dynamics of the transmis-
aspects. 1) Transformers in the system are not considered sion network can be represented by a multiple-input-multiple-
and in turn the droop formulation (XT and RT ) are also not output (MIMO) matrix. VSCs are considered as two voltage
considered; 2) the considered droop functions are linear, while sources. The transmission network is shown in Fig. 6. A pure
[8] considers piecewise linear ones. resistive load is assumed for the load. Based on the circuit
shown in Fig. 6, the relations between the voltages E1 , E2 and
III. S TABILITY A NALYSIS the currents in abc frame can be derived using superposition:
A. Linear Model Derivation 
1

1 Rlo

iL1 = E1
Rlo kLL s+LL s − E2 Rlo kLL s+LL s Rlo +LL s
1) Linear model for converter control: To conduct linear  
1 1 Rlo
analysis, a linear model of the study system needs to be iL2 = E2
Rlo kLL s+LL s − E1 Rlo kLL s+LL s Rlo +LL s
derived. The system can be separated into three parts, VSC (10)
with V /F control, the transmission network, and the droop Since the control loop is modeled in dq frame, (10) should
control. also be converted to the dq frame by replacing s using s + jω,
The block diagram of the controlled-frequency VSC shown where ω = 377 rad/s. This is due to the fact that a space
in Fig. 5 can be used to represent the dynamics of the VSC vector in the abc-frame is related to a vector in the dq-frame

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2017.2656062, IEEE
Transactions on Power Delivery

iL1 iL2 the system is always stable. However, if both of mk and nk

are selected as very small values, the system will be unstable
LL LL because of the locations of the dominant poles as shown in Fig.
AC E1 E2 AC
Rlo 10a. The only exception is that when m1 = m2 and n1 = n2 ,
the system is stable as shown in Fig. 10b.
2) Grid-connected mode: Fig. 8-Fig. 10a show the move-
Fig. 6: The system without the VSC dynamics. ment of the eigenvalues for the system in the autonomous
mode. To check if there are stability issues for a system in
grid-connected mode, the following test system shown in Fig.
as follows. 11 is examined, where the two converters not only serve a load
−
→ but also connected to the grid through a line with inductance
f abc = f dq ejωt . (11)
LG .
After the conversion, the transfer function will be separated The same method is used to derive a new G2 , the dynamics
into real term and imaginary terms. of the grid and load. Because the oscillation only happens with
     very small droop coefficients, n1 and n2 are selected as 0.01
iL1d G21r −G21i G22r −G22i E1d
iL1q   G21i G21r G22i G22r  E1q  and 0.02 to make the system have stability issue. Short circuit
iL2d  = G22r −G22i G21r −G21i  E2d  (12)
     ratio (SCR) is a good factor to decide if the grid is strong or
weak. The SCR can be changed by changing the inductance of
iL2q G22i G22r G21i G21r E2q
| {z } the transmission line connected to the grid, LG . Fig. 12 shows
G2 the movement of the dominant eigenvalues corresponding to
where ω is the nominal frequency, 377rad/s. The transfer the variation of SCR (from 0.5 to 5). According to Fig. 12,
functions, G21r,i and G22r,i , are expressed in (13). the damping ratio of the dominant eigenvalue becomes larger
(12) illustrates that the inputs are four voltages, Eid,q , and with larger SCR.
output are four currents, iLid,q , which will be sent back to the Remarks: Based on the above analysis, we identified the
first part, VSC with control loop. root causes of oscillations are due to the following aspects.
3) Overall Block Diagram: When the output currents are Small values for m and n, generally, cause poles to be located
obtained, the references of the output voltages can be calcu- closer to the imaginary axis, therefore decreasing the stability,
lated using the outmost droop control loop shown in Fig. 2. and showing poorer damping ratios (see Figs. 8 and 9); 2) If
For two VSCs, the no load voltage, E0 , is the same while the the mk and/or nk values are different from each other, some
droop coefficients are different. Therefore, the inputs of power poles are in the RHP (see Fig. 10); 3) Weaker grids also show
sharing part are four currents and outputs are four references poorer damping ratios; stronger grids show better damping
of voltages. Finally, Fig. 7 shows the combination of three ratios (see Fig. 12).
dynamic parts of the microgrid system with V-I droop. G1
presents the dynamic of VSCs with control loop, G2 is the IV. S IMULATION VALIDATION : T HE T WO
matrix to show the dynamic of load and transmission lines, PARALLEL -C ONVERTER C ASE
and G3 illustrates the power sharing or the droop control loop.
In this section, a detailed model was designed to verify the
capability of V-I droop control’s power sharing as well as the
B. Linear Analysis effect of parameters on stability.
1) Autonomous mode: Based on Fig. 7, the map of poles The test system is the autonomous system in Fig. 1. The
and zeros of the microgrid with V-I droop can be plotted using detailed model including power electronic switching details
Control Design Toolbox in MATLAB. The locations of the and controls shown in Fig. 2 was built and simulated in MAT-
dominant poles can be found if specific droop coefficients are LAB/Simpowersystems. The parameters are listed in Table I.
provided. The m and n values are all based on physical unit Note that the reactance from the converter to the PCC voltage
(V/A). If we keep m1 = 0.4 and m2 = 0.8 as constants is 0.0038 Ω. Therefore, m and n should be  0.0038 to have
and increase n1 (from 0.01 to 1) and n2 (from 0.02 to 2), the accurate power sharing. The gains of the PI controllers are
movement of dominant poles can be found easily after merging listed in Table II. The gains are based on physical units.
them in one plot (shown in Fig. 8). With an increasing nk , In Section III, the effect of droop coefficients on the
the dominant poles are further away from the imaginary-axis, movement of the dominant poles has been presented in Fig. 9
so the dynamic response becomes faster and damping ratio and Fig. 8. Both mk and nk can affect the system stability. As
becomes larger. mentioned in Section III, the load was considered as a pure
Fig. 9 shows the movement of dominant poles with increas- resistive load (Rlo = 1.34Ω) when studying the stability issue.
ing m1 (from 0.004 to 0.4) and m2 (from 0.008 to 0.8) while Three cases with different sets of droop parameters were
n1 = 1 and n2 = 2. simulated to verify this stability issue based on the detailed
In the above cases, although the dominant poles will be model. The values of mk and nk are shown in the caption of
more closed to the imaginary-axis with decreasing mk or nk , Fig. 13.
they are still in the left half plane (LHP). In another word, The dynamic event is designed as follows. At the initial
either of small mk or nk may cause the oscillation issue, but condition, two DERs support one resistive load (rated power:

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2017.2656062, IEEE
Transactions on Power Delivery

L3L s3 + 3L2 Rlo s2 + (L3L + ω 2 + 2LL Rlo

2
)s + L2L Rlo ω 2
G21r =
G21D
L3L ωs2 + 2L2L Rlo ωs + (L3L ω 3 + 2LL Rlo
2
ω)
G21i =
G21D
L4 Rlo s4 + 4L3 Rlo2 3
s + 5L2L Rlo3 2
s + 2LRlo 4
s − (L2L Rlo
3 2
ω + L4L Rlo ω 4 )
G22r = L
G22D
L4L Rlo s4 + 4L3L Rlo
2 3
s + 5L2L Rlo
3 2
s + 2LRlo4
s − (L2L Rlo
3 2
ω + L4L Rlo ω 4 )
G22i =
G22D
where
G21D = L4L s4 + 4L3 Rlo s3 + (2L4 ω 2 + 4L2L Rlo
2
)s2 + 4L3L Rlo ω 2 s + (L4L ω 4 + 4L2L Rlo
2 2
ω )
G22D = L6L s6 + 6L5L Rlo s5 + (3L6L ω 2 + 13L4L Rlo
2
)s4 + (12L5L Rlo ω 2 + 12L3L Rlo
3
)s3
+ (3L6L ω 4 + 18L4L Rlo
2 2
ω + 4L2L Rlo
4
)s2 + (6L5L Rlo ω 4 + 12L3L Rlo
3 2
ω )s + (L6L ω 6 + 5L4L Rlo
2 4
ω + 4L2L Rlo
4 2
ω ) (13)

Pole−Zero Map
G3
E1d iL1d 300
E1d*

Imaginary Axis (seconds−1)

E1d*
E1q* G1 E1q
G2 iL1q -m1 0 0 0
0 -n1 0 0 E1q* 200
E2d* E2d iL2d 0 0 -m2 0 E2d*
E2q* E2q iL2q 0 0 0 -n2 100
8x4 4x4 E2q*
0 m
increase
−100
Fig. 7: The block diagram of the linear model of the system
with two VSCs with V-I droop. −200

−300
Pole−Zero Map −400 −300 −200 −100 0
Real Axis (seconds−1)
300
Imaginary Axis (seconds−1)

Fig. 9: The values of m1 and m2 affect the continuous

200
movement of dominant eigenvalues.
100
TABLE II: Gains of PI Controllers
0 n increase
Loop Time constant Kp Ki
−100 Inner 0.5 ms 20 200
Outer 5 ms 0.316 63.25
−200

−300

−400 −300 −200 −100 0 6 kW). At 1.5 s, a pure resistive load (rated power: 3 kW)
Real Axis (seconds−1)
was added.
Fig. 8: Increasing n1 and n2 leads to the dominant poles In the case study, load is modeled as impedances. The
moving to the left half plane (LHP). impedance of the load will be reduced to have a power
increase. Due to the effect of the current controller, the currents
are kept constant at the moment when the step change is
TABLE I: Circuit Parameters
enforced. This will cause a sudden reduction in the PCC
voltage or load bus voltage.
Parameter Values
DC voltage 400 V
No-load voltage, E0 102 V
A. Effect of the droop coefficients
Rated voltage on load 110 V (rms, L-L)
Nominal frequency, ω0 377 rad/s The effect of droop coefficients are shown in Fig. 13. In
Switching frequency 3060 Hz
case (a), there is no oscillation issue with the chosen droop
R of RLC filter 100 mΩ
L of RLC filter 10 mH coefficients. The voltages are within the ±4% range. If m1
C of RLC filter 0.5 mF and m2 are reduced to 0.004 and 0.008, real power shows
LL of transmission line 0.05 mH (0.0038 Ω) oscillations as shown in (b). If n1 and n2 are reduced to
0.01 and 0.02, then the reactive power shows oscillations.

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2017.2656062, IEEE
Transactions on Power Delivery

Pole−Zero Map Pole−Zero Map

300
300

Imaginary Axis (seconds )

Imaginary Axis (seconds−1)

−1
200
200
100
100
0
SCR
0
−100
increase
−100
−200
−200
−300

−350 −300 −250 −200 −150 −100 −50 0 50 −300

Real Axis (seconds−1)
−400 −300 −200 −100 0
(a) Real Axis (seconds−1)
Pole−Zero Map Fig. 12: When a grid is connected, the value of SCR affects
300
the continuous movement of dominant eigenvalues.
Imaginary Axis (seconds−1)

200

the network resistance plays a big role in oscillations. With a

100
high R/X ratio at 26, oscillations will be suppressed. For a R/X
0 ratio at 2.6, oscillations still exist. For a resistive network, even
−100
with very small droop coefficients, oscillations are not issues.

−200

−300 C. Sensitivity of SCR

−350 −300 −250 −200 −150 −100 −50 0 50
As mentioned above, a strong grid can enhance the system
Real Axis (seconds−1)
stability. This case was designed to check how the strength
(b) of the grid affects the system stability. The strength of the
Fig. 10: (a) shows poles and zeros when m1 = 0.004, m2 = grid is dependent on SCR. Fig. 15(a) and Fig. 15(b) show
0.008, n1 = 0.01, and n2 = 0.02; (b) shows poles and zeros the effect of SCR on oscillations. A grid with SCR=1 is
when m1 = m2 = 0.04, n1 = n2 = 0.04. normally considered as a weak grid while SCR=5 means the
very strong grid. It can be concluded that the strong grid can
reduce oscillations even with small droop coefficients, while
iL1 iL2 the weak grid case still shows obvious oscillations.

LL LL
LG V. S IMULATION VALIDATION : T HE CIGRE B ENCHMARK
AC E1 E2 AC T EST C ASE
Rlo
Egrid AC Simulation case study is also conducted for the CIGRE
benchmark test case [26]. The system topology is shown in
Fig. 16. The ratings of DERs are shown in the figure. Rated
Fig. 11: The grid-connected system without the VSC dynam- load real power levels are also shown. All loads are assumed
ics. to have 0.7 lagging power factor. The loads are modeled as
constant impedances. The network consists of cables with
large R/X ratio. In the simulation model, the network dynamics
The oscillations in reactive power are caused by circulating are ignored. Converters are modeled in the dq reference frame.
currents which are basically q-axis currents going back and The simulation results are shown in Fig. 17. A set of small
forth between the DER units, and therefore increasing the droop coefficients is used in Case (a) while a set of large droop
stress on and losses in transmission lines. coefficients is used in case (b). At t = 1 second, the load at
DER2 is doubled by increasing its admittance to twice of its
B. Effect of the network resistance original value. The simulation results show that even with very
Sensitivity analysis of the network parameters is conducted. small droop coefficients, there is no oscillation issue. This is
The network parameters are varied. For the same droop due to the resistive network of the CIGRE benchmark system.
coefficients employed in case (c), instead of using a pure The droop coefficients were designed based on the power
inductor LL to connect the converter after the RLC filter to ratings. For example, the rating real power of DER1 10 times
the PCC bus, the inductor was replace by an RL circuit where of that of DER4, so m4 = 10m1 . The same method was
r = 0.1Ω or r = 0.01Ω. It can be found from Fig. 14 that used to designed n. Fig. 17 illustrated the real and reactive

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2017.2656062, IEEE
Transactions on Power Delivery

10 10 10
Load
Load Load

P (kW)
5 DER1 5 DER1 5 DER1

DER2 DER2 DER2

0 0 0
1 1.5 2 2.5 1 1.5 2 2.5 1 1.5 2 2.5

4 4 4
DER1
2 2 2
Q (kVAr)

0 0 0 Load

−2 −2 −2
DER2
−4 −4 −4
1 1.5 2 2.5 1 1.5 2 2.5 1 1.5 2 2.5

70 70 70
(V)
RMS

65 65 65
V

60 60 60
1 1.5 2 2.5 1 1.5 2 2.5 1 1.5 2 2.5
Time (s) Time (s) Time (s)

Fig. 13: (a) m1 = 0.09, m2 = 0.18, n1 = 1, n2 = 2; (b) m1 = 0.004, m2 = 0.008, n1 = 1, n2 = 2; (c) m1 = 0.09, m2 =
0.18, n1 = 0.01, n2 = 0.02. The two straight lines in voltage plots indicate ±4% range.

Q (kVAr) (a)
4 3

2 with 0.1 Ω resistance SCR=1

2
(a)

0 1
−2
(a)

0
−4
1 1.5 2 2.5 −1

4 −2
with 0.01 Ω resistance
2 −3
1 1.5 2 2.5
(b)

0
−2 (b)
2
−4 SCR=5
1 1.5 2 2.5 1

4 0
without resistance −1
2
(b)
(c)

0 −2

−3
−2
−4
−4
1 1.5 2 2.5 −5
Time (s) 1 1.5 2 2.5
Time (s)

Fig. 14: Effect of the network resistance. (a) with 0.1 Ω

resistance; (b) with 0.01 Ω resistance; (c) without resistance. Fig. 15: (a) Effect of SCR = 1 on reactive power; (b) Effect
of SCR = 5 on reactive power. In both cases, m1 = 0.4,
m2 = 0.8, n1 = 0.01 and n2 = 0.02.
power shared by five DERs based on the relative droop
coefficients. Compared to case (b), case (a) shows a more
obvious oscillation mode with sufficient damping. For this power measured at the DERs. Fig. 17(b) shows that the
network, due to the resistive network, there is no oscillation reactive power sharing is not accurate. This may due to the
issue concerned. assumption of shunt constant impedances as load models. The
We would like to point out that small droop coefficients shunt impedance reduces the equivalent Thevenin network
lead to inaccurate power sharing as observed in case (a). impedance and causes inaccurate sharing when the droop
With increased droop coefficients, the power sharing are more coefficients are close to the network resistance/reactance.
aligned towards the desired power sharing ratio. Note that the Remarks: It is possible that oscillations occur when converters
P and Q presented in Fig. 17 are real power and reactive are connected to PCC through inductive devices. Oscillations

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2017.2656062, IEEE
Transactions on Power Delivery

[4] J. M. Guerrero, J. C. Vasquez, J. Matas, M. Castilla, and L. G. de Vicuna,

3φ, “Control strategy for flexible microgrid based on parallel line-interactive
6 ups systems,” IEEE Transactions on Industrial Electronics, vol. 56,
VPCC (rated)=400V
4.5kW no. 3, pp. 726–736, Mar 2009.
[5] R. Majumder, A. Ghosh, G. Ledwich, and F. Zare, “Angle droop
versus frequency droop in a voltage source converter based autonomous
microgrid,” in 2009 IEEE Power Energy Society General Meeting, Jul
2009, pp. 1–8.
DER1 7 [6] R. Majumder, G. Ledwich, A. Ghosh, S. Chakrabarti, and F. Zare,
30kW “Droop control of converter-interfaced microsources in rural distributed
1 generation,” IEEE Transactions on Power Delivery, vol. 25, no. 4, pp.
2768–2778, Oct 2010.
2 [7] A. Samui and S. R. Samantaray, “Assessment of p-v droop characteristics
18kW in islanding detection with frequency-dependent zip-exponential static
DER2 load,” in Energy, Automation, and Signal (ICEAS), 2011 International
8
DER3 45kW 30kW Conference on, Dec 2011, pp. 1–6.
10kW [8] M. S. Golsorkhi and D. D. C. Lu, “A control method for inverter-
3 4 based islanded microgrids based on V-I droop characteristics,” IEEE
4.5kW Transactions on Power Delivery, vol. 30, no. 3, pp. 1196–1204, Jun
2015.
9 DER4 [9] E. Rokrok and M. E. H. Golshan, “Adaptive voltage droop scheme
3kW for voltage source converters in an islanded multibus microgrid,” IET
5
Generation, Transmission Distribution, vol. 4, no. 5, pp. 562–578, May
20kW 2010.
[10] E. A. A. Coelho, P. C. Cortizo, and P. F. D. Garcia, “Small-signal sta-
DER5 bility for parallel-connected inverters in stand-alone ac supply systems,”
10kW IEEE Transactions on Industry Applications, vol. 38, no. 2, pp. 533–542,
Mar 2002.
Fig. 16: CIGRE microgrid test case. The load power factor is [11] Y. A. R. I. Mohamed and E. F. El-Saadany, “Adaptive decentralized
0.7. droop controller to preserve power sharing stability of paralleled invert-
ers in distributed generation microgrids,” IEEE Transactions on Power
Electronics, vol. 23, no. 6, pp. 2806–2816, Nov 2008.
[12] Y. W. Li and C.-N. Kao, “An accurate power control strategy for power-
are possible when multiple batteries or DERs are connected electronics-interfaced distributed generation units operating in a low-
in parallel to a PCC with small droop coefficients. voltage multibus microgrid,” IEEE Transactions on Power Electronics,
vol. 24, no. 12, pp. 2977–2988, 2009.
VI. C ONCLUSION [13] Z. Miao, A. Domijan, and L. Fan, “Investigation of microgrids with both
inverter interfaced and direct ac-connected distributed energy resources,”
This paper investigates the stability issues for a microgrid IEEE Transactions on Power Delivery, vol. 26, no. 3, pp. 1634–1642,
with parallel converters equipped with V-I droop control. Jul 2011.
[14] W. Yao, M. Chen, J. Matas, J. M. Guerrero, and Z. M. Qian, “Design and
Our paper’s investigation shows that oscillations are possible analysis of the droop control method for parallel inverters considering
with small droop coefficients. To have accurate power shar- the impact of the complex impedance on the power sharing,” IEEE
ing among converters, large droop coefficients are desired. Transactions on Industrial Electronics, vol. 58, no. 2, pp. 576–588, Feb
2011.
However, large droop coefficients may lead voltages below [15] R. Majumder, B. Chaudhuri, A. Ghosh, R. Majumder, G. Ledwich, and
the range under heavy load conditions. With small droop F. Zare, “Improvement of stability and load sharing in an autonomous
coefficients, not only power sharing may not be accurate microgrid using supplementary droop control loop,” in IEEE PES
General Meeting, July 2010, pp. 1–1.
but also there is possibility of oscillations depending on the [16] W. R. Issa, M. A. Abusara, and S. M. Sharkh, “Impedance interaction
network characteristics. A linear frequency-domain model is between islanded parallel voltage source inverters and the distribution
derived in this paper to show the effect of droop coefficients on network,” in Power Electronics, Machines and Drives (PEMD 2014),
7th IET International Conference on, Apr 2014, pp. 1–6.
stability. The analysis results are validated by detailed model-
[17] C. C. Chang, D. Gorinevsky, and S. Lall, “Stability analysis of dis-
based simulation conducted in Matlab/SimPowersystems. tributed power generation with droop inverters,” IEEE Transactions on
Power Systems, vol. 30, no. 6, pp. 3295–3303, Nov 2015.
ACKNOWLEDGEMENT [18] F. Cavazzana, P. Mattavelli, M. Corradin, and I. Toigo, “Grid sensitivity
considerations on multiple parallel inverters systems,” in 2016 IEEE
The authors wish to thank anonymous reviewers for provid- 8th International Power Electronics and Motion Control Conference
ing thorough reviews and detailed comments. (IPEMC-ECCE Asia), May 2016, pp. 993–999.
[19] C. N. Rowe, T. J. Summers, R. E. Betz, D. J. Cornforth, and T. G.
Moore, “Arctan power x2013;frequency droop for improved microgrid
R EFERENCES stability,” IEEE Transactions on Power Electronics, vol. 28, no. 8, pp.
[1] J. Quesada, J. A. Sainz, R. Sebastin, and M. Castro, “Decoupled 3747–3759, Aug 2013.
droop control techniques for inverters in low-voltage ac microgrids,” [20] H. J. Avelar, W. A. Parreira, J. B. Vieira, L. C. G. de Freitas, and
in Systems, Signals Devices (SSD), 2014 11th International Multi- E. A. A. Coelho, “A state equation model of a single-phase grid-
Conference on, Feb 2014, pp. 1–6. connected inverter using a droop control scheme with extra phase shift
[2] T. Goya, E. Omine, Y. Kinjyo, T. Senjyu, A. Yona, N. Urasaki, and control action,” IEEE Transactions on Industrial Electronics, vol. 59,
T. Funabashi, “Frequency control in isolated island by using parallel no. 3, pp. 1527–1537, Mar 2012.
operated battery systems applying H∞ control theory based on droop [21] J. M. Guerrero, L. G. de Vicuna, J. Matas, M. Castilla, and J. Miret,
characteristics,” IET Renewable Power Generation, vol. 5, no. 2, pp. “A wireless controller to enhance dynamic performance of parallel
160–166, Mar 2011. inverters in distributed generation systems,” IEEE Transactions on Power
[3] H. Xin, L. Huang, L. Zhang, Z. Wang, and J. Hu, “Synchronous insta- Electronics, vol. 19, no. 5, pp. 1205–1213, Sep 2004.
bility mechanism of P − f droop-controlled voltage source converter [22] C. K. Sao and P. W. Lehn, “Autonomous load sharing of voltage source
caused by current saturation,” IEEE Transactions on Power Systems, converters,” IEEE Transactions on Power Delivery, vol. 20, no. 2, pp.
vol. PP, no. 99, pp. 1–2, 2016. 1009–1016, Apr 2005.

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2017.2656062, IEEE
Transactions on Power Delivery

(a)
(b)
50 25
DER2
40 20 DER2
PDER (kW)
30 15
DER1
20 10
DER3 DER5
10 5
DER4
0 0
0.9 0.95 1 1.05 1.1 1.15 0.9 0.95 1 1.05 1.1 1.15

50 30
DER2
40
QDER (kVAr)

20 DER2
30
DER1
20
10
DER3 DER5
10
DER4
0 0
0.9 0.95 1 1.05 1.1 1.15 0.9 0.95 1 1.05 1.1 1.15

1.05 1

1.045 0.95
VDER (pu)

1.04 0.9

1.035 0.85

1.03 0.8
0.9 0.95 1 1.05 1.1 1.15 0.9 0.95 1 1.05 1.1 1.15
Time (s) Time (s)

Fig. 17: (a) m = 0.001[1, 1, 3, 10, 3], n = 0.001[1, 1, 3, 10, 3]. (b)m = 0.02[1, 1, 3, 10, 3], n = 0.01[1, 1, 3, 10, 3]. The m and n
values are in per unit (Sb = 10 kW and Vb = 400 V).

[23] T. L. Vandoorn, J. D. M. D. Kooning, B. Meersman, J. M. Guerrero, and Lingling Fan received the B.S. and M.S. degrees in electrical engineering
L. Vandevelde, “Automatic power-sharing modification of p / v droop from Southeast University, Nanjing, China, in 1994 and 1997, respectively,
controllers in low-voltage resistive microgrids,” IEEE Transactions on and the Ph.D. degree in electrical engineering from West Virginia University,
Power Delivery, vol. 27, no. 4, pp. 2318–2325, Oct 2012. Morgantown, in 2001. Currently, she is an Associate Professor with the
[24] A. Yazdani and R. Iravani, Voltage-sourced converters in power systems: University of South Florida, Tampa, where she has been since 2009. She was a
modeling, control, and applications. John Wiley & Sons, 2010. Senior Engineer in the Transmission Asset Management Department, Midwest
[25] A. Tazay, Z. Miao, and L. Fan, “Blackstart of an induction motor in ISO, St. Paul, MN, form 2001 to 2007, and an Assistant Professor with North
an autonomous microgrid,” in Power Energy Society General Meeting, Dakota State University, Fargo, from 2007 to 2009. Her research interests
2015 IEEE, Jul 2015, pp. 1–5. include power systems and power electronics. Dr. Fan serves as a technical
[26] S. Papathanassiou, N. Hatziargyriou, and K. Strunz, “A benchmark program committee chair for IEEE Power System Dynamic Performance
low voltage microgrid network,” in Proc. CIGRE Symp. Power Syst. Committee and an editor for IEEE Trans. Sustainable Energy.
Dispersed Gen., Dec 2005, pp. 1–8.

Yin Li received his B.S. in electrical engineering from University of South

Florida in May 2014. He is currently a Ph.D. student at USF SPS Lab. His
research interests include microgrids and power electronics.

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.