J. Mod. Power Syst.

Clean Energy (2019) 7(4):899–912

https://doi.org/10.1007/s40565-019-0544-3

Exponential-function-based droop control for islanded

microgrids
Rui WANG1, Qiuye SUN1 , Yonghao GUI2, Dazhong MA1

Abstract An exponential-function-based droop control power flow. The simulation and experimental results show
strategy for the distributed energy resources (DERs) is that the reactive power sharing deviation can be reduced,
proposed to reduce the reactive power-sharing deviation, and the utilization rate of renewable energy is improved by
limit the minimum value of frequency/voltage, whilst using the proposed method. Moreover, the simulation
improving the utilization rate of renewable energy. Both results illustrate that the system can maintain stable opera-
DERs and loads are interconnected to achieve a power tion when the microgrid is switched from one supplied
exchange by converters, where the power management energy operation condition to another absorbed one.
system should accurately share the active/reactive power
demand. However, the proportional reactive power sharing Keywords Distributed energy resources (DERs), Reactive
often deteriorates due to its dependence on the line impe- power-sharing, Droop control, Bidirectional power flow
dances. Thus, an exponential-function-based droop control
is proposed to  prevent voltage and frequency from falling
to the lower restraint, ` achieve accurate reactive power 1 Introduction
sharing, ´ eliminate communication and improve the usage
ratio of renewable energy. Furthermore, its stability is Recently, more and more renewable energy is integrated
analyzed, and the application in islanded AC/DC hybrid into smart grids or microgrids. Meanwhile, the power
microgrids is investigated to achieve the bidirectional electronic converter has been widely used to improve the
utilization of the renewable source energy [1–3]. The
power electronic converter can improve not only the
CrossCheck date: 26 March 2019 renewable energy usage ratio but also the power grid
quality performance [4]. As a result, the power electronic
Received: 24 October 2018 / Accepted: 26 March 2019 / Published converter, acting as interface equipment between dis-
online: 10 June 2019
tributed energy resources (DERs) and loads, is the main
Ó The Author(s) 2019
part of the AC/DC microgrid, and allows the elasticity of
& Qiuye SUN
[email protected] the islanded operation and the grid-connected operation
[5, 6]. Nevertheless, several power-sharing and limitation
Rui WANG
[email protected] of frequency as well as voltage issues need to be
investigated.
Yonghao GUI
[email protected] Conventionally, power sharing based on droop control
strategy is broadly integrated into parallel operation of
Dazhong MA
[email protected] inverters [7], wherein, the active/reactive power sharing
should be proportional to their rated active/reactive power
1
College of Information Science and Engineering, that not surpassing any individual DER [8]. It is, never-
Northeastern University, Shenyang, China theless, not the case for reactive power sharing with the
2
Department of Electronic Systems, Aalborg University, conventional droop control strategy, where impedance
Aalborg, Denmark

123
900 Rui WANG et al.

mismatch between the DERs and bus commonly causes the curve feature, which is a method of nonlinear droop control
output voltages of inverters to be different [9]. Further- strategy. In [11], a restoration mechanism based droop
more, frequency and voltage magnitude may fall to their control strategy was first employed to improve reactive
lower limitation due to linear droop curve. Thus, the con- power sharing. However, this method was difficult to avoid
ventional droop control has intrinsic limitation so that the the output voltage deviation of the converter being unable
proportional reactive power sharing cannot be achieved, to actualize accurate reactive power sharing. Besides, the
and frequency/voltage magnitude may fall to their lower method also added the difficulty of the secondary adjust-
limitation [10–12]. ment of converter output voltage. In [20], a control strat-
From the viewpoint of the limitation of frequency and egy, the second layer control embedded in the droop
voltage, the second level control was widely employed to controller, was first presented to realize reactive power
achieve the frequency and voltage regulation [13–16]. A sharing. Whilst, this method was equivalent to increasing
novel distributed coordinated control strategy based on the droop coefficient, impacting the stability of the system,
droop coefficient via a dynamic multi-agent consensus and this method ignored the effect of mismatched line
algorithm in our previous work [13] was applied to restore impedances. Thus, the improved droop control strategy
the system frequency. Furthermore, in order to deal with without communication, which is designed to actualize
the voltage deviation caused by the droop controller, the accurate reactive power sharing, still also remains a
dynamic consensus-based distributed controller in our problem.
previous work [14] was presented to regulate the output Additionally, the bidirectional power flow based on
voltages of all the DERs. Nevertheless, the communication droop characteristics has become an attractive problem
delay may impact the performance of the system. The [10]. In [8], an AC/DC consensus power management
novel secondary voltage and frequency restoration con- strategy was first presented to realize power interaction via
troller was employed in [15] to assess the impact of the the normalized AC frequency and DC voltage. Reference
communication delay. However, these methods need [21] researched the problem of active power sharing among
additional communication networks, which lose the a serious of microgrids formed by numerous AC/DC
advantage of droop control strategy. Thus, the local droop microgrids network-interconnected by interface converters.
controller without communication networks was proposed In order to deal with the aforementioned problems, this
to achieve the voltage regulation by combining with an paper proposes a novel droop control strategy based on
event-driven operator of the parameter-varying filter [16]. exponential function. The main characteristics and advan-
Meanwhile, the control strategy without using sparse tages of this paper are shown as follows:
communication networks, which can achieve both fre- 1) An exponential-function-based droop control strategy
quency and voltage regulations, still remains a without additional communication networks is proposed to
challenge. achieve accurate reactive power sharing and restrict the
From the viewpoint of the reactive power sharing, there minimum value of frequency/voltage. Furthermore, when
are two main approaches, i.e., droop controller with/with- output voltage and frequency are located far to the lower
out communication. On the side of the droop controller limitation in the initial stage, they can be quickly reduced
with communication, in [17], an adaptive virtual impe- to improve the utilization rate of DER by the features of
dance droop control strategy based on multi-agent was exponential function.
proposed to make equivalent output impedance of each 2) According to the relationship between the frequency
converter be adjusted inversely to its rated capacity. In and the square of the voltage, this paper presents a novel
[18], an adaptive droop coefficient control strategy was AC/DC droop control strategy to realize a bidirectional
presented to realize accurate reactive power sharing. power flow without additional communication networks.
Thereinto, the excessive droop coefficients would cause the This paper first proposes an exponential-function-based
system instability, whereas low droop coefficients would droop control strategy to achieve reactive power sharing
affect the response speed of the system. Moreover, these accurately and restrict the frequency and amplitude of
above methods also need additional communication net- voltage in Section 2. Subsequently, the stability of the
works. Therefore, some methods without communication proposed droop control strategy is proved in Section 3.
were developed one after another. The static virtual Moreover, a bidirectional power flow application is
impedance concept was first presented to achieve reactive demonstrated to show the feasibility of the proposed droop
power sharing in the high R/X ratio low-voltage microgrids control strategy in Section 4. Section 5 and Section 6
[19]. Nevertheless, the performance of the reactive power provide simulation and experimental results that highlight
sharing was not ensured on account of uncertain line the performance of the proposed control strategy. Finally,
impedance parameters. Therefore, some scholars put for- the conclusion is obtained in Section 7.
ward some improved droop controllers by adjusting droop

123
Exponential-function-based droop control for islanded microgrids 901

2 Exponential-function-based droop control fx Vx

strategy DER1 DER1
f* DER2 V* DER2

Although the droop control method enables the con-

verters to share the active and reactive power, there are Steady-state points Steady-state points
some intrinsic limitations in this area so that proportional Px Qx
reactive power sharing has deviation [8]. As seen in Fig. 1, (a) (b)
the active/reactive power flow through the lines can be Fig. 2 Active and reactive power droop features for an AC microgrid
represented as follows: (AMG)
8
> 3 X x Vx U 3 Rx Vx ðVx  U cos /x Þ "
>
<Px ¼ sin /x þ
2 Zx Zx 2 Zx Zx 2 Zx
: ð1Þ Vx ¼ U cos /x þ  ðU cos /x Þ2 :
>
> 3 Rx V x U 3 Xx Vx ðVx  U cos /x Þ 3 nx
:Qx ¼  sin /x þ
2 Z x Zx 2 Zx Zx   #12 ð5Þ
2 Zx 2 4 Zx 8 Zx
where x ¼ 1; 2; Px and Qx are the output active and reactive þ þ U cos /x  U
3 nx 3 nx 3 nx
power flows of the interface converter, respectively; Zx ¼
Rx þ jXx is the equivalent impedance of DERx; /x is the Furthermore, the accurate reactive power sharing is
power angle of DERx; Vx is the voltage of DERx; and U is achieved if and only if (6) is satisfied.
the voltage at the point of common coupling. When the line 8
impedance is pure inductive, (1) can be switched to (2). < /1 ¼ /2
8 Z Z ð6Þ
3 Vx U : 1¼ 2
>
> n1 n2
<Px ¼ sin /x
2 Zx
: ð2Þ On account of the impact of line impedances, it is hard
>
> 3 Vx ðVx  U cos /x Þ
:Qx ¼ to satisfy (6) in practical system. According to the inherent
2 Zx
characteristics of output voltage change rate of inverters,
Thus, as shown in Fig. 2, the conventional droop control an exponential-function-based droop control strategy is
can be presented with (3) to achieve the frequency fx and proposed, as shown in Fig. 3 to overcome the effect of line
the voltage magnitude Vx regulation. impedances on reactive power sharing, where f0min and
(
fx ¼ f   mx Px
ð3Þ
Vx ¼ V   nx Qx
fx Initial
where f  and V  are the frequency and voltage of the droop f0 stage
Droop stage
controller without load demand; and mx and nx are the
conventional droop controller coefficients. In light of (3),
the accurate reactive power sharing is actualized if and f0min
only if (4) is satisfied.
n1 Q1  n2 Q2 ¼ ðV   V1 Þ  ðV   V2 Þ ¼ V2  V1 ¼ 0
ð4Þ 0
Px
(a)
Substitute (2) into (3), we have
Vx Initial
Droop stage
V0 stage

R1 jX1 U 0° jX2 R2
V1 1 V2 2
V0min
S1=P1+jQ1 S2=P2+jQ2

DER1 ZL=RL+jXL DER2

0
Qx
(b)
Fig. 3 A multifunctional and wireless droop control strategy based
Fig. 1 Simplified circuit of the system on exponential function

123
902 Rui WANG et al.

V0min are the minimum values of the system, f0 and V0 are dVx
Q2(t1)
the no-load frequency and voltage of the converter, Q2(t0) Q1(t0) Q2(t2) Q1(t2) Q1(t1)
respectively. Compared with conventional droop control, Qx
the proposed exponential-function-based droop control
strategy is a nonlinear control strategy. The controllers dV1(t0), dV2(t0)
are described as: dV2(t1)
  dV1(t2), dV2(t2)
 Px
fx ¼ f0  Dfx 1  e KP0x ð7Þ dV1(t1)
  (a)
 Qx
Vx ¼ V0  DVx 1  e KQ0x ð8Þ
Qx
Dfx Q1
Pref ¼ KP0x ln ð9Þ
fx þ Dfx  f0
DVx
Qref ¼ KQ0x ln ð10Þ
Vx þ DVx  V0 Q2

where Dfx and DVx are the rated frequency and voltage
deviations of the converter, respectively; P0x and Q0x are
the rated active power and reactive power, respectively; 0 t1 t2 t
(b)
Pref and Qref are the rated active power and reactive power
of the inverter, respectively; and K is the proportion dVx
constant. Furthermore, (11) needs to be satisfied.
dV2
8 
< DVx ¼ C dV1
1
KQ0x ð11Þ
: 
Dfx ¼ C2
where C1 and C2 are the constant.
0 t1 t2 t
Remark 1 Although the linear droop control characteris-
(c)
tic seems to be simply replaced with the exponential-
function-based droop control characteristic, the finding Fig. 4 Operations of proposed exponential-function-based droop
process of the droop controller based on exponential- controller
function is very difficult. Meanwhile, this replacement can
bring many benefits. Additionally, the stability assessment in relatively higher V2 ðt1 Þ. Eventually, dV1 ðt1 Þ and dV2 ðt1 Þ
of the proposed exponential-function-based droop control are same, and the resulting V1 and V2 varies at the same
strategy should be studied. pace, thus their output reactive power stabilize at Q1 ðt2 Þ
To actualize the accurate reactive power sharing, the and Q2 ðt2 Þ, respectively [11].
derivative of (8) is given by: According to the above discussion, all the DERs will
operate at the identical voltage change rate in the steady
DVx KQQx
V_ x ¼  e 0x ð12Þ state, i.e.,
KQ0x
DV1 KQQ1 DV2 KQQ2
Figure 4 explains the managements of proposed V_ 1 ¼ V_ 2 )  e 01 ¼  e 02 ð13Þ
KQ01 KQ02
controller in the simplified circuit shown in Fig. 1.
Supposing the load is located closer to DER1, and its The output reactive power of each DER meets the
reactive power demand Qload consumption is provided by following relationship as:
DER1 ðQ1 ðt0 ÞÞ and DER2 ðQ2 ðt0 ÞÞ. When Q0load increases Q
KQ1
Q
KQ2 Q1 Q2
Ce 01 ¼ Ce 02 ) ¼ ð14Þ
to Q1load at certain point, the output reactive power of DER1 Q01 Q02
adds at express speed to Q1 ðt1 Þ in order to meet the load where C ¼ DV1 =ðKQ01 Þ ¼ DV2 =ðKQ02 Þ.
change demand, and the reactive power of DER2 increases Thus, the output reactive power of DERs is in coinci-
marginally to Q2 ðt1 Þ, since it is away from the load. Thus, dence with the rated reactive power capacities of DERs. In
the more negative dV1 ðt1 Þ drives V1 ðt1 Þ lower to reduce the addition, the frequency of DERs is identical at the steady
output reactive power of the DER1, and DER2 adds its state shown as follows:
output reactive power with its less negative dV2 ðt1 Þ results

123
Exponential-function-based droop control for islanded microgrids 903

f1 ¼ f2 ) [25], (20) could be given to reflect the dynamics of the

 P   P 
 1  2 ð15Þ entire system, which can be shown as follows:
f0  Df1 1  e KP01 ¼ f0  Df2 1  e KP02
A ¼ M s þ Cs ðI s þ Es Y s ÞK s ð20Þ
Then, from (15), we can conclude that the output active
power of each DER satisfies: where
 
P1 P2 M1 0
¼ ð16Þ Ms ¼ ð21Þ
P01 P02 0 M2
 
So, the output active power of DERs is also in C1 0
Cs ¼ ð22Þ
coincidence with the rated active power capacities of 0 C2
DERs. When the output voltage and frequency are located 2 3
id1 iq1 0 0
far to their lower limitation values in the initial stage, they 6 iq1 id1 0 0 7
can be quickly reduced to improve the utilization rate of Is ¼ 64 0
7 ð23Þ
0 id2 iq2 5
renewable energy by the properties of exponential function. 0 0 iq2 id2
When the output voltage and frequency are located close to 2 3
their lower limitation values, they can never fall below the kp x 0
lower limitation value so as to limit the voltage and 6 kq mqx x 7
6 0 7
frequency of the system. Thus, the proposed droop control Cx ¼ 6 6 m n
dx qx  m n 7
qx dx 7 ð24Þ
4 kq mdx x 5
strategy can satisfy the relevant standards better [22–24]. 0
With conventional droop control, the frequency and mqx ndx  mdx nqx
2 3
voltage magnitude are always regarded as the cut-off 0 1 0 0 0 0
condition. When the frequency or voltage magnitude is 60 0 1 0 0 07
inferior to the minimum value, the droop control will be Ks ¼ 6 40 0 0 0 1 05
7 ð25Þ
shifted to active/reactive power control. On account of the 0 0 0 0 0 1
volatility of renewable energy, it is advisable that active 2 3
ed1 eq1 0 0
power or reactive power is regarded as the cut-off 6 eq1 ed1 0 0 7
condition, which can utilize the renewable energy better. Es Y s ¼ 6
4 0
7 ð26Þ
0 ed2 eq2 5
0 0 eq2 ed2
2 3
3 Stability analysis of proposed control strategy x 0 0
6 nqx mqx ndx x mqx ndx x 7
6 mdx nqx  mqx ndx mdx nqx  mqx ndx mdx nqx  mqx ndx 7
Mx ¼ 6 7
The small-signal model of the conventional-function- 4 ndx mdx ndx x mdx nqx x 5
based droop control system has been widely investigated mqx ndx  mdx nqx mqx ndx  mdx nqx mqx ndx  mdx nqx
[25]. In this section, the steady-state function of the pro- ð27Þ
posed exponential-function-based droop control is built by
small-signal perturbation method. The dynamic of the where edx ¼ Vx cos dx ; eqx ¼ Vx sin dx ; dx ¼ arctanðeqx =
power flow can be shown as follows: edx Þ; mdx ¼ eqx =ðe2dx þ e2qx Þ; mqx ¼ edx =ðe2dx þ e2qx Þ; ndx ¼
8 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
> oPx oPx
>
< DPx ¼ oV DVx þ o/ D/x edx =ð e2dx þ e2qx Þ; nqx ¼ eqx =ð e2dx þ e2qx Þ; and idx , iqx are
x x
: ð17Þ the d-q axis output currents of the inverter. In order to
>
> oQx oQx
: DQx ¼ DVx þ D/x guarantee the stability of the proposed exponential-
oVx o/x
function-based droop control strategy, an eigen-
The dynamics using the small-signal approximation to participation analysis is finished by using the linearized
linearize the equations can be expressed as: model to assess the small-signal stability of the system. It is
 clear that the system has positive eigenvalues that means
Dx_ ¼ xDx  kp xDPx
ð18Þ the system is unstable and vice versa. The root locus is
DV_ x ¼ xDVx  kq xDQx
8 shown in Figs. 5 and 6 by changing kp and kq , respectively.
>
> Dfx KPPx System parameters are as follows: input voltage is 400 V;
< kP ¼ e 0x
KP0x ð19Þ filter inductance and capacitance are 0.02 H and 0.185 mF,
>
> DVx KQQx respectively; capacitance of DC capacitor is 5 mF; the
: kq ¼ e 0x
KQ0x proportional and integral parameters of current inner loop
where x is angle speed. Following the model approach in Kpi and Kii are 10.5 and 16000, respectively; the

123
904 Rui WANG et al.

1000 the proposed control strategy is stable. Furthermore, the

0.4 0.28 0.2 0.14 0.09 0.04
800 800 stability margin of the proposed control strategy can be
0.56 600 improved by adding K.
600
8
> Dfx KPPx
400 0.8 400 >
< kp ¼ e 0 2 ð1:353352832366  106 ; 1  105 Þ
Imaginary axis (s-1)

200 KP0
200 :
>
> k ¼ DVx eKQ Qx
5 4
0
: q 0 2 ð2:7067056  10 ; 2  10 Þ
KQ0
-200 200 ð28Þ
0.8
-400 400
-600 0.56 4 Application of proposed control strategy
600
-800 800
0.4 0.28 0.2 0.14 0.09 0.04
-1000 With the development of the islanded AC/DC hybrid
-400 -350 -300 -250 -200 -150 -100 -50 0 50 100 microgrids, the AC/DC droop control strategy should be
Real axis (s-1)
proposed to achieve the bidirectional power flow. Mean-
Fig. 5 Root locus of system dynamic model with kp changing from while, the main application of the proposed control strategy
105 to 102 is applied to AMGs. Furthermore, the application of the
proposed control strategy is also extended in AC/DC
800 0.83 0.72 0.58 0.4 0.2 hybrid microgrids. Thus, the application of the proposed
0.91
control is discussed in this section, which is the bidirec-
600
tional power flow. Normally, the AMG and DC microgrid
0.96 (DMG) are connected through an island interconnection
400
device (IID). The AMG is interfaced with its neighboring
Imaginary axis (s-1)

0.99
200 DMG through an interlinking converter (IC), as shown in
0
1400 1200 1000 800 600 400 200 Fig. 7, which can be regarded as a voltage source converter
(VSC) [26].
-200 0.99
In contrast to the AMG or DMG, the power management
0.96 of the IC needs to bring forward a novel control strategy to
-400
manage the bidirectional power flow between AMGs and
-600 0.91 DMGs. Besides, to eliminate fast communication modes,
0.83 0.72 0.58 0.4 0.2 the proposed droop control is applied to the novel AC/DC
-800
-1600 -1400 -1200 -1000 -800 -600 -400 -200 0 200 400 droop control strategy to achieve bidirectional power flow.
Real axis (s-1) The frequency of the AMG and the DC bus voltage of the
IC should be measured as input signal to power manage-
Fig. 6 Root locus of system dynamic model with kq changing from
ment strategy to provide the power reference value [10]. As
105 to 102
shown in Fig. 8, the stored energy in the capacitor is:
proportional and integral parameters of voltage inner loop
Kpv and Kiv are 0.05 and 390, respectively; P0 is 20 kW; Q0
is 10 kvar; active and reactive droop coefficients kx and kdc DC microgrid
are 11.45 and 2, respectively; system frequency is 50 Hz; Biomass
load resistance, inductance and capacitance are 50 X, 0.01 energy Photovoltaic DC
EMS
H, 0.1 mF, respectively; frequency and voltage deviations DC/DC DC/DC load
of the converter Dfx and DVx are 0.2 and 2, respectively; VDMG
DC bus
and K is 1. As shown in Fig. 5, the proposed control
strategy is stable if and only if the number of the right-hand IC IC
AC bus
planes (RHPs) of the minor loop gain is equal to zero.
Thus, the proposed control strategy is stable with kp
DC/AC AC/AC DC/AC
changing from 105 to 0.0083, while the value kq is 105 .
Photovoltaic Wind Biomass Wind Building Electric
As shown in Fig. 6, the proposed droop control strategy is energy vehicle
stable if and only if kq changes from 105 to 0.0013, while AC microgrid
the value kp is 105 . Since kp and kq are in stable interval,
Fig. 7 A typical hybrid AC/DC microgrid

123
Exponential-function-based droop control for islanded microgrids 905

8
PDMG iabc PAMG < xAMG0  xAMG ¼ kx VDMG0
2 2
 VDMG
ua 1 Cdc 1 ð34Þ
: kx ¼ kx
Cdc 2 Ts KPAMG0x
DMG VDGM ub AMG
where Ts is the sampling period; and kx is the coefficient of
uc
AC/DC droop control.
The equation of AC/DC droop control is:
Fig. 8 Structure of IC interfacing AMG and DMG xAMG ¼ kx VDMG
2
þ xAMG0  kx VDMG0
2
ð35Þ

The AC/DC droop control strategy is exhibited in Fig. 9,

1 2
Wdc ¼ Cdc VDMG ð29Þ where xmin is the minimum value of the angle speed in
2 AMG. A system block diagram of the proposed
where Wdc is the stored energy in the capacitor; Cdc is DC interlinking bidirectional power converter control is
capacitor; and VDMG is the DC voltage of the DMG. depicted in Fig. 10. The responses of IC in different
Neglecting the switching losses converter, the dynamics in islanded modes are shown as follows.
the capacitor energy is the difference of power transfer 2 2
1) Mode 1 (xAMG \xmin and VDMG \VDMG;min ): Pref ¼ 0
between AMG and DMG. Thus, and IC transfers no power. Both DMG and AMG are
d 1 d 2 overloading. An overloading blocking logic shown in
Wdc ¼ Cdc VDMG ¼ PDMG  PAMG ¼ DPH ð30Þ Fig. 11a is adopted to satisfy mode 1.
dt 2 dt
2 2
2) Mode 2 (xAMG [ xmin and VDMG \ VDMG;min ): Pref ¼
where PDMG and PAMG are output active power of the DMG
and AMG, respectively; and DPH is the active power PAMG;ref and IC supplies power to the DMG, where PAMG;ref
deviation of the microgrid. In consideration of the is the rated active power of the AMG. An IC transfer
characteristic of droop control strategy in the AMG and opening logic shown in Fig. 11b is taken to satisfy
DMG, the related formulations are shown as follows: mode 2.
2 2
 PAMG0 P  3) Mode 3 (xAMG \xmin and VDMG [ VDMG;min ): Pref ¼
  AMGx
DxAMG ¼ xAMG0  xAMG ¼ kx e KPAMG0x  e KPAMG0x PDMG;ref and IC supplies power to the AMG. An IC
ð31Þ transfer opening logic shown in Fig. 11b is employed to
satisfy mode 3.
1 2 2
PDMG;ref ¼ ðVDMG  VDMG0 Þ þ PDMG0 ð32Þ 4) Mode 4 (xAMG [ xmin and VDMG [ VDMG;min ): Pref ¼
kdc
0 and IC transfers no power. A light load blocking logic
VDMG;max  VDMG;min exhibited in Fig. 11c is utilized to satisfy mode 4.
kdc ¼ ð33Þ
PDMG;max The rated reactive power control of the IC is simple as
the reactive power does not exist in DMG, and the IC is
where xAMG is angular velocity of the AMG; DxAMG is the
studied to share the constant reactive power during IC
angular velocity deviation of the AMG; xAMG0 is no-load
angular velocity of the AMG; PAMG0 is the rated output
active power of AMG; PAMG0x is the rated output active
power of AMGx; PAMGx is the output active power of
AMGx; PDMG;max and PDMG;ref are the maximum and rated
active power in the DMG, respectively; PDMG0 is no-load
0
active power in the DMG; VDMG0 is no-load voltage of the
DMG; and VDMG;max and VDMG;min are the maximum and
minimum permitted voltage of the DMG, respectively.
In accordance with (30) and (31), a novel AC/DC droop
control strategy for the IC is proposed through using the Forbidden
Taylor series expansion up to ðq þ cÞ and assuming that the region
frequency in the microgrid is constant, where c denotes the min

control order, and q denotes the relative degree of the

system. For the system under investigation, it is enough
obvious that c is equal to one, whilst q is equal to zero [27].
2 2 2
The droop control of the IC is given by: VDMG,min VDMG0 VDMG

Fig. 9 A novel AC/DC droop control strategy for IC

123
906 Rui WANG et al.

Mode 3 V
2
VDMG0 of DER2 with the proposed control strategy. Therefore, the
(DC to AC)
AMG0
AC/DC DMG0
droop + VDMG VDMG0 PDMG PDMG,ref proposed control strategy is able to reduce the reactive
+ + + + +
AMG control droop power sharing deviation among multi-DERs. Furthermore,
(35) control (32) the output voltages of DERs are also similar while the rated
Pref reactive power of DERs are identical.
Mode 2 +
2
(AC to DC) VDMG0 AC/DC AMG0 +
Exponential 5.2 Case 2
+ + droop + AMG
2 + droop
VDMG control
control (7) PAMG,ref
(35)
DER1 and DER2 are both DC sources, and their
capacities are set as P01 ¼ 20 kW; Q01 ¼ 10 kW,
Fig. 10 Proposed interlinking bidirectional power converter control
strategy P02 ¼ 10 kW; Q02 ¼ 5 kW. The load is located closer to
DER1. Moreover, the line impedances are set as Z1 ¼
AMG min AMG min AMG min
ð17:5 þ j0:2Þ mX and Z2 ¼ ð26:25 þ j0:3Þ mX, the line
& XOR NOR
2
VDMG
2
VDMG,min
2
VDMG
2
VDMG,min
2
VDMG
2
VDMG,min load is set as Pload ¼ 25 kW; Qload ¼ 12 kvar and
K ¼ 1:5. As shown in Fig. 13, the output reactive power of
(a) (b) (c) DER1 with the proposed control strategy is twice than that
of DER2 with the proposed control strategy, and the
Fig. 11 Mode discrimination logic sharing errors of the DERs can be reduced. Thus, the
proposed control strategy is able to reduce the reactive
supplying power to the AMG, the rated reactive limitation power sharing deviation among multi-DERs. The output
is defined as: reactive power proportion of DERs is similar to their rated
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
reactive power proportion.
QIC;max ¼ S2IC;max  P2IC ð36Þ

where QIC;max is the maximum reactive power transfer 5.3 Case 3

between AMG and DMG through the IC; SIC;max is the
maximum rating of the IC; and PIC is the power transfer The capacities of DER1 and DER2 are set as
between AMG and DMG through the IC. P01 ¼ 2 kW; Q01 ¼ 1 kvar; P02 ¼ 20 kW; Q02 ¼ 10 kvar.
The load is located closer to DER1, and the line load is set
as Pload ¼ 15 kW; Qload ¼ 6 kvar. The other parameters
5 Simulation results are the same as those in case 2. The above-mentioned case
is the strictest. As shown in Fig. 14a, b, the stability of the
To verify the proposed droop control strategy, a hybrid microgrid can be satisfied in the steady state, whereas the
AC/DC microgrid system with multi-DERs, as shown in transient output active power and reactive power of DER1
Figs. 1 and 8, is simulated in Simulink. System parameters are both beyond the rated capacities. Thus, this case can
are the same as those in Section 3. Different capacities of also be considered that the microgrid is instantaneously
the DERs and configuration of the line impedances are unstable. Since the capacitor and inductance can store
designed with identical parameters to accelerate the energy, the instantaneous overload energy can be provided
observation of the reactive power sharing accuracy and by the capacitor and inductance. In real systems, the ability
improvement of the utilization rate of renewable energy. of overcurrent/overvoltage for insulated gate bipolar tran-
Eventually, the bidirectional power flow is verified. sistor (IGBT) is better than the rated capacity. Thus, the
microgrid cannot be stable in the steady state, when tran-
5.1 Case 1 sient output active power or reactive power of DER is
beyond the rated capacity. As shown in Fig. 14c, d, the
DER1 and DER2 are both DC sources with the same accurate reactive power sharing with the conventional
capacities, and both of them provide energy to the load, and droop control cannot be achieved, and the output reactive
the load is located closer to DER1. Furthermore, the line power of DER1 is beyond the rated capacity in the steady
impedances are set as Z1 ¼ ð17:5 þ j0:2Þ mX and state. In order to ensure the stability of the system, the
Z2 ¼ ð26:25 þ j0:3Þ mX, the line load is set as Pload ¼ droop control strategy should be shifted to the active/
30 kW; Qload ¼ 16 kvar and K ¼ 1:5. As shown in Fig. 12, reactive power control in this extreme case.
compared with the sharing errors of the DERs with the
conventional control strategy, the output reactive power of
DER1 with the proposed control strategy is similar to that

123
Exponential-function-based droop control for islanded microgrids 907

120.10 121.4
120.05 121.2

Voltage (V)

Voltage (V)
121.0
120.00
120.8
119.95
120.6
119.90 120.4
119.85 120.2
119.80 120.0
0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16
Time (s) Time (s)
(a) Voltage with proposed exponential droop control (b) Voltage with conventional droop control
Reactive power (kvar)

10

Reactive power (kvar)

10
8
8
6 6
4 4
2 2

0 0.02 0.04 0.06

0.08 0.10 0.12 0.14 0.16 0 0.02 0.04 0.060.08 0.10 0.12 0.14 0.16
Time (s) Time (s)
(c) Reactive power with proposed exponential droop control (d) Reactive power with conventional droop control
DER1; DER2

Fig. 12 Simulation results of voltage and reactive power sharing in case 1

120.5 121.2
120.0 121.0
Voltage (V)

Voltage (V)

120.8
119.5 120.6
119.0 120.4
120.2
118.5
120.0
118.0 119.8
0 0.02 0.04 0.06
0.08 0.10 0.12 0.14 0.16 0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16
Time (s) Time (s)
(a) Voltage with proposed exponential droop control (b) Voltage with conventional droop control
10 10
Reactive power (kvar)
Reactive power (kvar)

8 8
6 6
4 4
2 2

0 0.02 0.04 0.06

0.08 0.10 0.12 0.14 0.16 0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16
Time (s) Time (s)
(c) Reactive power with proposed exponential droop control (d) Reactive power with conventional droop control
DER1; DER2

Fig. 13 Simulation results of voltage and reactive power sharing in case 2

5.4 Case 4 improved, when renewable energy adopts conventional

droop control strategy and other energy employs proposed
DER1 adopting conventional droop control and DER2 droop control strategy.
adopting proposed droop control are both DERs with same
capacities, and both of them provide energy to the load. The 5.5 Case 5
line impedances are set as Z1 ¼ Z2 ¼ ð17:5 þ j0:2Þ mX, and
the line load is set as Pload ¼ 30 kW; Qload ¼ 16 kvar. As The classic microgrid test system consisting of three
shown in Fig. 15, the DER employing exponential-function DERs is provided to verify the performance of the pro-
based droop control strategy can be integrated into the posed droop control strategy, which is shown in Fig. 16.
DERs which adopt conventional droop control strategy. Meanwhile, the rated ratio of the reactive power of the
Furthermore, the utilization rate of renewable energy is three DERs is 4:3:2. The proposed control strategy is

123
908 Rui WANG et al.

14 7

Reactive power (kvar)

12

Active power (kW)

6
10 5
8 4
6 3
4 2
2 1
0 0
-2 -1
0 0.02 0.04 0.06
0.08 0.10 0.12 0.14 0.16 0 0.02 0.04 0.06
0.08 0.10 0.12 0.14 0.16
Time (s) Time (s)
(a) Active power with the proposed droop control (b) Reactive power with the proposed droop control

Reactive power (kvar)

6
5
Active power (kW)

15
4
10 3
2
5 1
0
-1
0 0.02 0.04 0.06
0.08 0.10 0.12 0.14 0.16 0 0.02 0.04 0.06
0.08 0.10 0.12 0.14 0.16
Time (s) Time (s)
(c) Active power with the conventional droop control (d) Reactive power with the conventional droop control
DER1; DER2

Fig. 14 Simulation results of active power and reactive power in case 3 where one of the DERs is with low power

60.001
120.10 60.000
Frequency (Hz)

120.05
Voltage (V)

59.999
120.00
119.95 59.998
119.90
59.997
119.85
119.80 59.996
0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16
Time (s) Time (s)
(a) Voltage (b) Frequency
10 20
Reactive power (kvar)

Active power (kW)

8 15
6
10
4
5
2
0 0
-2 -5
0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16
Time (s) Time (s)
(c) Reactive power (d) Active power
DER1; DER2

Fig. 15 Simulation results of voltage, frequency, reactive power, and active power in case 4 when DER1 adopts proposed exponential droop
control and DER2 adopts conventional droop control

simulated with the parameters exhibited in Section 3.

Moreover, the line impedances are set as Zline1 ¼
Zline1 Zline2 Zline3
Load 1 Load 2 ð17:5 þ j0:2Þ mX, Zline2 ¼ ð20 þ j0:25Þ mX, Zline3 ¼
ð26:25 þ j0:3Þ mX, and the line load is set as Pload ¼
25 kW; Qload ¼ 18 kvar and K ¼ 1:5. As shown in
Fig. 17, the accurate reactive power sharing can be
DER1 DER2 DER3
achieved by the dynamic virtual impedance [17] and the
Fig. 16 Classic microgrid test system with high capacity of DERs proposed droop control strategy. The performance of four
droop controllers can be summarized as shown in Table 1.
In the viewpoint of the droop controller without the extra

123
Exponential-function-based droop control for islanded microgrids 909

10

Reactive power (kvar)

7

Reactive power (kvar)

6 8
5 6
4
4
3
2 2
1
0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16
Time (s) Time (s)
(a) Conventional droop control (b) Static virtual impedance

10 9

Reactive power (kvar)

Reactive power (kvar)

8
8 7
6
6 5
4
4 3
2
2 1
0
-1
0 0.02 0.04 0.06
0.08 0.10 0.12 0.14 0.16 0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16
Time (s) Time (s)
(c) Dynamic virtual impedance (d) Proposed droop control
DER1; DER2; DER3

Fig. 17 Reactive power simulation results in case 5

Table 1 Performance comparison analysis load demands. The IC switches to the inverter mode, and
DMG feeds the redundant power to the AMG. The system
Method Communication Performance
response is exhibited in Fig. 18. It can be seen that a certain
Conventional droop control No Low percentage of the load power requirement in the AMG is
Static virtual impedance [19] No Middle provided via the IC at first. Then, a fixed percentage of the
Dynamic virtual impedance [17] Yes High load power requirement in the DMG is supplied through
The proposed droop control No High the IC at 0.1 s. The power flow, current injection by the IC
and the DC bus voltage reflect a stable system operation.
Thus, this proposed control strategy is able to achieve
bidirectional power flow.

communication, the performance of the proposed droop

control method is better than that in [19]. In the viewpoint
of the droop controller with the extra communication, the
Current (A) Voltage (V)

ABC
performance of the proposed droop control method is 200
0
similar to that in [17]. Nevertheless, since loss of packet, -200
communication delay and failures of communication
modes will also impact the performance of the system, the 200 ABC
100
reliability of the droop controller with the dynamic virtual 0
impedance is reduced due to the extra communication -100
-200
networks. Thus, the effectiveness of the proposed droop
power (kvar) power (kW)

20
Active

control strategy is verified.

0
-20
5.6 Case 6
10
Reactive

DC renewable sources provide a fixed power in DMG. 0

One part of the required load is provided via the local -10
0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
sources in DMG and another inadequate power is supplied Time (s)
via the IC at first. At 0.1 s, a large part of the DC loads is
switched off, and the DC power sources are more than the Fig. 18 System response in case 6

123
910 Rui WANG et al.

6 Experiments

In this section, an experiment platform with two paral-

leled DERs as exhibited in Fig. 19 is tested to verify the
proposed exponential-function-based droop control strat-
egy. The experimental parameters are similar to those
parameters in the above simulation, and K ¼ 10. The
paralleled hard-switched PWM converters are digitally
controlled by TMS28335 fixed-point DSP to implement the
power and voltage control algorithm where switching fre-
Fig. 20 Reactive power sharing with proposed droop control in
quency is 20 kHz. experiment case 1 when DC source capacity is same

6.1 Experiment case 1

The two DERs have the same power capacities

(S1 ¼ S2 ¼ 5 kVA), and both of them provide power for
the load. It is considered that the load is located closer to
DER1. At t ¼ 0 ms, the load is set as
Pload ¼ 2 kW; Qload ¼ 1:2 kvar. Figures 20 and 21 show
the reactive power sharing with different droop controllers
in this experimental test condition. With the proposed
exponential-function-based droop control, the output
reactive power of DER1 and DER2 both is 0:6 kvar, as
shown in Fig. 20. Nevertheless, with the conventional Fig. 21 Reactive power sharing with conventional droop control in
droop control, the output reactive power of DER1 is dif- experiment case 1 when DC source capacity is same
ferent from that of DER2. Further, at t ¼ 120 ms, the
reactive power of the load increases to 2:4 kvar, and at 0:9 kvar. Figures 22 and 23 show the reactive power
t ¼ 350 ms, the reactive power of the load decreases to sharing with different droop controllers in this experi-
1:2 kvar. Wherein, the output reactive power with the mental test condition. With the proposed exponential-
proposed droop control of DER1 and DER2 is equal, function-based droop control, the output reactive power of
whereas the output reactive power with the conventional DER1 is 0:6 kvar and output reactive power of DER2 is
droop control of DER1 and DER2 is different. Thus, the 0:3 kvar, as shown in Fig. 22. Nevertheless, the sharing of
proposed exponential-function-based droop control strat- the reactive power has obvious errors with the conventional
egy can decrease the reactive power-sharing deviation droop control. Further, at t ¼ 120 ms, the reactive power of
among multiple DERs. the load increases to 1:8 kvar, and at t ¼ 350 ms, the
reactive power of the load decreases to 0:9 kvar. From
6.2 Experiment case 2 Fig. 22 to Fig. 23, it is obvious that the reactive power
sharing with the proposed droop control is always accurate,
In this case, the two DERs have different power but the sharing of the reactive power with the conventional
capacities (S1 ¼ 5 kVA; S2 ¼ 2:5 kVA), and both of them
provide power for the load. It is considered that the load is
located closer to DER1. Therefore, the line impedances are
set as Z1 ¼ 0:01 X and Z2 ¼ ð1000 þ j20Þ mX. At
t ¼ 0 ms, the load is set as Pload ¼ 2 kW; Qload ¼

Z1
DER1

Z2 Load
DER2

Fig. 22 Reactive power sharing with proposed droop control in

Fig. 19 Experiment platform experiment case 2 when DC source capacity is different

123
Exponential-function-based droop control for islanded microgrids 911

References

[1] Li Y, Zhang H, Liang X et al (2019) Event-triggered based

distributed cooperative energy management for multi-energy
systems. IEEE Trans Ind Inform 15(4):2008–2022
[2] Wang Z, Huang Z, Song C et al (2018) Multiscale adaptive fault
diagnosis based on signal symmetry reconstitution preprocess-
ing for microgrid inverter under changing load condition. IEEE
Trans Smart Grid 9(2):797–806
[3] Sun Q, Zhang Y, He H et al (2017) A novel energy function-
based stability evaluation and nonlinear control approach for
energy internet. IEEE Trans Smart Grid 8(3):1195–1210
Fig. 23 Reactive power sharing with conventional droop control in [4] Ma D, Sun Q, Xie X et al (2018) Event triggering power sharing
experiment case 2 when DC source capacity is different control for AC/DC microgrids based on P-F droop curve
method. J Frankl Inst. https://doi.org/10.1016/j.jfranklin.2018.
droop control has obvious errors. Thus, the proposed 10.025
[5] Huang Z, Wang Z, Zhang H (2018) A diagnosis algorithm for
exponential-function-based droop control strategy is able
multiple open-circuited faults of microgrid inverters based on
to decrease the reactive power-sharing deviation among main fault component analysis. IEEE Trans Energy Convers
multiple DERs. 33(3):925–937
[6] Han R, Wang H, Jin Z et al (2018) Compromised controller
design for current sharing and voltage regulation in DC
microgrid. IEEE Trans Power Electron. https://doi.org/10.1109/
7 Conclusion tpel.2018.2878084
[7] Wang R, Sun Q, Ma D et al (2019) The small-signal stability
This paper proposes an exponential-function-based analysis of the droop-controlled converter in electromagnetic
timescale. IEEE Trans Sustain Energy. https://doi.org/10.1109/
droop control strategy without additional communication
tste.2019.2894633
to actualize the accurate reactive power sharing and limit [8] Loh PC, Li D, Chai YK et al (2013) Autonomous operation of
the minimum value of frequency and voltage. Furthermore, hybrid microgrid with AC and DC subgrids. IEEE Trans Power
the voltage and frequency can be quickly reduced to Electron 28(5):2214–2223
[9] Lee CT, Jiang RP, Cheng PT (2013) A grid synchronization
improve the utilization rate of renewable energy by the
method for droop-controlled distributed energy resource con-
properties of exponential function when renewable energy verters. IEEE Trans Ind Appl 49(2):954–962
adopts the conventional droop control strategy, and other [10] Eghtedarpour N, Farjah E (2014) Power control and manage-
energy employs the proposed exponential-function-based ment in a hybrid AC/DC microgrid. IEEE Trans Smart Grid
5(3):1494–1505
control strategy. Then, the output voltages of DERs are
[11] Lee CT, Chu CC, Cheng PT (2013) A new droop control
also identical when the rated reactive power of DERs is method for the autonomous operation of distributed energy
the same. Moreover, the proposed control strategy can also resource interface converters. IEEE Trans Power Electron
be applied to the bidirectional power flow control in the 28(4):1980–1993
[12] Zhang H, Zhou J, Sun Q et al (2017) Data-driven control for
AC/DC microgrids by measuring the AMG frequency as
interlinked AC/DC microgrids via model-free adaptive control
well as the DC bus voltage and using proposed droop and dual-droop control. IEEE Trans Smart Grid 8(2):557–571
characteristic. The performance of the proposed control [13] Sun Q, Han R, Zhang H et al (2015) A multiagent-based con-
strategy considering different operating cases has been sensus algorithm for distributed coordinated control of dis-
tributed generators in the energy internet. IEEE Trans Smart
demonstrated by the simulation of a hybrid AC/DC
Grid 6(6):3006–3019
microgrids in the MATLAB/Simulink and experiments. [14] Zhou J, Kim S, Zhang H et al (2018) Consensus-based dis-
tributed control for accurate reactive, harmonic, and imbalance
Acknowledgements This work was supported by National Key power sharing in microgrids. IEEE Trans Smart Grid
Research and Development Program of China (No. 9:2453–2467
2017YFF0108800), National Natural Science Foundation of China [15] Chen G, Guo Z (2018) Distributed secondary and optimal active
(Nos. 61773109, 6143304), and Major Program of National Natural power sharing control for islanded microgrids with communi-
Foundation of China (No. 61573094). cation delays. IEEE Trans Smart Grid 10(2):2002–2014
[16] Rey JM, Rosero CX, Velasco M et al (2018) Local frequency
Open Access This article is distributed under the terms of the restoration for droop-controlled parallel inverters in islanded
Creative Commons Attribution 4.0 International License (http:// microgrids. IEEE Trans Energy Convers. https://doi.org/10.
creativecommons.org/licenses/by/4.0/), which permits unrestricted 1109/tec.2018.2886267
use, distribution, and reproduction in any medium, provided you give [17] Zhang H, Kim S, Sun Q et al (2017) Distributed adaptive virtual
appropriate credit to the original author(s) and the source, provide a impedance control for accurate reactive power sharing based on
link to the Creative Commons license, and indicate if changes were consensus control in microgrids. IEEE Trans Smart Grid
made. 8(4):1749–1761

123
912 Rui WANG et al.

[18] Chen X, Wang L, Sun H et al (2017) Fuzzy logic based adaptive Rui WANG received the B.S. degree in Northeastern University,
droop control in multiterminal HVDC for wind power integra- Shenyang, China, in 2016. He is currently pursuing the Ph.D. degree
tion. IEEE Trans Energy Convers 32(3):1200–1208 in the School of Information Science and Engineering, Institute of
[19] Guerrero JM, de Vicuna LG, Matas J et al (2005) Output Automation, Northeastern University, Shenyang, China. His current
impedance design of parallel-connected UPS inverters with research interests include collaborative optimization of distributed
wireless load-sharing control. IEEE Trans Ind Electron generation and its stability analysis of electromagnetic timescale in
52(4):1126–1135 Energy Internet.
[20] Zhong QC (2013) Robust droop controller for accurate pro-
portional load sharing among inverters operated in parallel. Qiuye SUN received the Ph.D. degree in control theory and control
IEEE Trans Ind Electron 60(4):1281–1290 engineering from Northeastern University, Shenyang, China, in 2007.
[21] Zhou J, Zhang H, Sun Q et al (2018) Event-based distributed He is currently a professor with Northeastern University. His current
active power sharing control for interconnected AC and DC research interests include analysis and diagnosis of uncertain
microgrids. IEEE Trans Smart Grid 9(6):6815–6828 information in power distribution network, optimization analysis
[22] Li Y, Li YW (2011) Power management of inverter interfaced technology of power distribution network and network control of
autonomous microgrid based on virtual frequency–voltage distributed generation system.
frame. IEEE Trans Smart Grid 2(1):30–40
[23] Becker C, Braun W, Carrick K et al (1994) Proposed chapter 9 Yonghao GUI received the B.S. degree in automation from
for predicting voltage sags (dips) in revision to IEEE Std 493, Northeastern University, Shenyang, China, in 2009, and the M.S.
the Gold Book. IEEE Trans Ind Appl 30(3):805–821 and Ph.D. degrees in electrical engineering from Hanyang University,
[24] Tian J, Zhou D, Su C et al (2017) Reactive power dispatch Seoul, South Korea, in 2012 and 2017, respectively. From February
method in wind farms to improve the lifetime of power con- 2017 to November 2018, he worked with the Department of Energy
verter considering wake effect. IEEE Trans Sustain Energy Technology, Aalborg University, Denmark, as a postdoctoral
8(2):477–487 researcher. Since December 2018, he has been working with the
[25] Coelho EAA, Cortizo PC, Garcia PFD (2002) Small-signal Automation and Control Section at the Department of Electronic
stability for parallel-connected inverters in stand-alone AC Systems, Aalborg University, where he is currently an assistant
supply systems. IEEE Trans Ind Appl 38(2):533–542 professor. He has been working with power converter control and
[26] Micallef A, Apap M, Spiteri-Staines C et al (2014) Reactive control theory.
power sharing and voltage harmonic distortion compensation of
droop controlled single phase islanded microgrids. IEEE Trans Dazhong MA received the B.S. degree in automation, and the Ph.D.
Smart Grid 5(3):1149–1158 degree in control theory and control engineering from Northeastern
[27] Yang J, Zheng W, Li S et al (2015) Design of a prediction University, Shenyang, China, in 2004 and 2011, respectively. He is
accuracy enhanced continuous-time MPC for disturbed systems currently an assistant professor with Northeastern University. His
via a disturbance observer. IEEE Trans Ind Electron current research interests include fault diagnosis, fault-tolerant
62(9):5807–5816 controls, energy management systems, control and optimization of
distributed generation systems, microgrids, and Energy Internet.

123