energies

Article
Power Decoupling Method Based on the Diagonal
Compensating Matrix for VSG-Controlled Parallel
Inverters in the Microgrid
Bin Li * ID
and Lin Zhou
State Key Laboratory of Power Transmission Equipment & System Security and New Technology,
Chongqing University, Shapingba District, Chongqing 400044, China; zhoulin@cqu.edu.cn
* Correspondence: lbsas@163.com; Tel.: +86-185-1232-5075

Received: 19 November 2017; Accepted: 13 December 2017; Published: 17 December 2017

Abstract: The thought of the virtual synchronous generator (VSG) for controlling the grid-connected
inverters and providing virtual inertia to the microgrid is emerging as a wide extension of the droop
control, power coupling that always exists in the low-voltage microgrid, which may deteriorate
the dynamic response and the stability of the system. In this paper, the principle of VSG control is
introduced first. As an important issue of VSG control, the mechanism of the power coupling in the
low-voltage microgrid is analyzed and the small-signal equivalent model of the power transmission
loop is established. Subsequently, a power decoupling method based on the diagonal compensating
matrix for VSG is proposed, which can realize the power decoupling with no impact on the original
control channel. Meanwhile, the feasibility analysis of the decoupling method and the improved
approach for reactive power sharing are also discussed. Simulation results verify the effectiveness of
the decoupling strategy for VSGs.

Keywords: virtual synchronous generator (VSG); power decoupling; diagonal compensating matrix;
small signal model

1. Introduction
A microgrid is an organic system composed of loads, distributed generations (DGs), and
energy storage devices. Relative to the main network, a microgrid is an intelligent controlled unit.
For customers, a microgrid can realize self-control, self-protection, and self-management in a flexible
manner. It can meet the demands of a variety of load and specific power quality requirements [1,2].
In a microgrid, the inverter is one of the most important components. Yet it presents some drawbacks,
such as the small output impedance, weak ability of overload, and lack of inertia [3,4]. The virtual
synchronous generator (VSG) algorithm has been proposed to alleviate the above drawbacks by
simulating the inertia element of the synchronous generator. As the extension of droop control, VSG
enables it to respond to the abnormal frequency of the grid side and improves the stability of the power
system [5–8].
Low-voltage microgrid lines present resistive or resistance-inductance properties, which can
lead to strong coupling between active and reactive power outputs. This phenomenon may cause
considerable impacts on the normal operation of microgrid, grid-connected control, and the power
allocation among the DGs. Hence, it is important to investigate the power coupling effect for the
improvement of operational stability and economic efficiency of the microgrid [9].
In engineering practice, large inductance has been introduced to reduce the coupling extent.
However, this effort increases the line voltage drop and the total cost. Traditional VSG control has not
paid much attention to power decoupling. Research on the independent control between active and
reactive powers has mostly been based on droop control. The specific decoupling methods can be

Energies 2017, 10, 2159; doi:10.3390/en10122159 www.mdpi.com/journal/energies

Energies 2017, 10, 2159 2 of 13

divided into three categories. The first one is based on the coordinate transformation. A coordinate
rotation matrix can be established according to the impedance angle of the line, and is used to
transform the real power flows [9–11] into decoupled virtual power flows. Similar results have also
been developed to transform the output voltage into decoupled virtual voltage [12,13]. This approach
can only realize the decoupling of the virtual power output, but not the real power output. Since
the equivalent droop equations have been changed, the microgrid cannot be directly accessed by
the distribution network that contains the traditional power generation, such as a steam turbine or
diesel-driven generator. Moreover, complexity increases for the power control.
The second category is the virtual impedance-based decoupling control [14–18]. Through
introducing a virtual inductance and subtracting its voltage drop at the input side of the voltage
loop, the system impedance can be adjusted for power decoupling. Although this approach is simple,
a large virtual inductance is required in the resistive environment, which can aggravate the line voltage
drop and causes harmonic pollution.
The inversion of droop control is categorized as the third type [19,20]. Traditional droop control
contains an active power channel (P–f ) and a reactive power channel (Q–V), which exists strong
coupling between the power outputs in the microgrid. Hence, some scholars have proposed the P–V
and Q–f droop control. This creates a good power decoupling effect in the resistive environment.
However, it is only appropriate for the case in which the line resistive value is far greater than its
inductance value. Moreover, the microgrid cannot be directly accessed by the distribution network
that contains the traditional power generation, such as a steam turbine or diesel-driven generator,
because the equivalent droop equations have been changed.
All three approaches do not realize the decoupling control by acting on the coupling component
directly. Thus, they can be considered as the indirect decoupling method. This paper proposes a
power decoupling method, which directly compensates for the coupling component of the power loop.
The decoupling is attained by use of a decoupling matrix in the VSG control unit and making the
power transfer matrix a diagonal matrix. This method is based on the linearized power transfer model.
It possesses a simple structure and better stability compared to the existing decoupling methods.
In addition, the proposed power decoupling method is suitable for power decoupling in the microgrid,
which has variable loads and frequent small power fluctuations.
The rest of this paper is organized as follows. The principle of the VSG control method is described
in Section 2. The power/frequency and excitation controller are also introduced. In Section 3, the
mechanism of power coupling is studied and the small signal model of the power loop is established.
The proposed decoupling method based on the diagonal decoupling matrix for VSG is analyzed in
Section 4. The feasibility analysis and improvement for power sharing are given. Simulation results
are presented in Section 5 to demonstrate the effectiveness of the proposed method, followed by the
conclusion in Section 6.

2. VSG Control Principle

The main structure of one grid-connected inverter using the VSG control strategy is shown in
Figure 1. A distributed generator (DG) with stored energy is located at the direct current (DC) side.
For simplicity of analysis, a DC source is used here instead. The three-phase voltage-source inverter is
adopted in the main circuit. V dc is the voltage of the DC source; Zline is the line impedance from DG
to the point of common coupling (PCC); Rin , L, and C denote the inductance internal resistance, the
smoothing inductance, and the smoothing capacitance of the filter, respectively; vo and io are the filter
capacitor voltage and the inverter’s output current, respectively; ic denotes the filter capacitor current;
Pset and Qset represent the input mechanical power and reactive power reference of the inverter,
Energies 2017, 10, 2159 3 of 13

respectively; and Pe and Qe represent the active and reactive power output of the inverter, respectively.
In this article, according to the instantaneous power theory, Pe and Qe can be calculated as:
Energies 2017, 10, 2159 3 of 13
Pe = vo_α io_α + vo_β io_β
(1)
Q = vo_β io_α − vo_α io_β
The outer power loop generates theecommand voltage, which is used as the input signal of the
voltage loop
The by power
outer controlling
loopthe instantaneous
generates active and
the command reactive
voltage, power
which of the
is used asinverter.
the inputTosignal
ensure
ofthe
the
accuracy, the voltage loop adopts the quasi-proportional-resonant (quasi-PR) controller
voltage loop by controlling the instantaneous active and reactive power of the inverter. To ensure which is
characterized by an excellent tracking performance.
the accuracy, the voltage loop adopts the quasi-proportional-resonant (quasi-PR) controller which is
Figure 1 also
characterized shows
by an blocktracking
excellent diagrams of the active power loop and the reactive power loop. The
performance.
activeFigure
power1loop mainly simulates
also shows block diagrams the rotor motion
of the equation
active of synchronous
power loop generator
and the reactive (SG).
power A
loop.
second-order model of SGs can be expressed as [8]:
The active power loop mainly simulates the rotor motion equation of synchronous generator (SG).
dω Pset asPe[8]:
 expressed
A second-order model of SGs can be
 J = − − Dp (ω − ωref )
(  dt ωN ωN
J dt = ω − ω − Dp (ω − ωref )
dω Pset Pe (2)
dδdδ = ω N N (2)
dtdt= ω
where δδ is
where is the
the power
power angle
angle ofof the
the SG,
SG, ωωisisthe
theelectric
electricangular
angularfrequency
frequencyof ofthe SG,JJisisthe
theSG, themoment
momentof of
inertia of the SG, D is the damping coefficient representing the function of damping
inertia of the SG, Dppis the damping coefficient representing the function of damping winding, ωref is winding, ω ref
is the synchronization angular frequency of the power grid, and ω is the
the synchronization angular frequency of the power grid, and ωN is the rated angular frequency.
N rated angular frequency.
Duringnormal
During normal operation,
operation, the
the angular
angular frequency
frequency reference
reference ω
ωref is equal
ref is equal to
to ω
ωNN, ,which
whichcan canimplement
implement
theprimary
the primaryfrequency
frequencymodulation
modulation(PFM)(PFM)ofofVSG.VSG.ForForthe
thedesign
designof ofthe
theactive
activepower
powerloop loopofofVSG,
VSG,
most existing researches on the design of the VSG active power loop are based
most existing researches on the design of the VSG active power loop are based on the PFM function. on the PFM function.

1 1 1
s Js ωN

1
Ks
2Vref

2Vo

Figure
Figure1.1.Control
Controldiagram
diagramof
ofthe
theinverter
inverterbased
basedon
onvirtual
virtualsynchronous
synchronousgenerator
generator(VSG).
(VSG).

The reactive power loop of VSG contains the floating voltage and the reactive power controlled
unit. Its control operation can be expressed as:

dEm
K = Qset + 2Dq (Vref − Vo ) − Qe (3)
dt
where Dq is the droop coefficient of voltage, K is the reactive power regulation coefficient, and Vo and
Energies 2017, 10, 2159 4 of 13

The reactive power loop of VSG contains the floating voltage and the reactive power controlled
unit. Its control operation can be expressed as:

dEm √
Energies 2017, 10, 2159 K = Qset + 2Dq (Vref − Vo ) − Qe 4 of(3)
13
dt

signal D
where is the
emq of the droop coefficient
voltage loop can of
bevoltage,
obtainedK by
is the reactive power
combining regulation
the amplitude coefficient,Eand
information V o and
m generated
Vby arereactive
refthe effectivepower
values of the
loop andoutput voltage
the phase and rated
information voltage,
δ by respectively.
the active power loopThe
as:input command
signal em of the voltage loop can be obtained by combining the amplitude information Em generated
by the reactive power loop and the phase em_a = 2Em sinδδby the active power loop as:
information

 e =√ 2E sin δ − 2 π
 em_a
m_b= m ( 3 ) (4)
 √2Em sin δ

sin(δδ − 22 π
π )


e
em_b ==√ 2
2EEm sin + (4)
 e = 2E sin δ + 2 π
 m_c m 33

m_c m 3

3.
3. Output
Output Power
Power Model
Model of
of Microgrid
Microgrid
To
To facilitate the
the analysis
analysisand
andensure
ensure
thethe generality,
generality, the simplified
the simplified equivalent
equivalent circuitcircuit of a
of a micro-
micro-source is illustrated
source is illustrated in Figure
in Figure 2. 2.

Figure 2.
Figure 2. Equivalent
Equivalent circuit
circuit of
of an
an inverter
inverter under
under the
thegrid-connected
grid-connectedmode.
mode.

In Figure 2, the micro-source is equivalent to a controlled voltage source. S = P + jQ is the line

In Figure 2, the micro-source is equivalent to a controlled voltage source. S = P + jQ is the line
power of the microgrid. e is the output voltage of the micro-source and v is the voltage of PCC. The
power of the microgrid. e is the output voltage of the micro-source and v is the voltage of PCC.
phasor form can be written as E∠δ and V∠0. The grid impedance is represented by Z, which consists of
The phasor form can be written as E∠δ and V ∠0. The grid impedance is represented by Z, which
the output impedance Zout and the line impedance Zline. Suppose that the grid impedance is referred to
consists of the output impedance Zout and the line impedance Zline . Suppose that the grid impedance
as Z∠θ = R + jX. A single phase study is given below. The output power equation can be written as:
is referred to as Z∠θ = R + jX. A single phase study is given below. The output power equation can be
written as: (  P = V E cos(θ − δ ) − V cos θ
P = V
[ ]
ZZ[ E cos( θ − δ ) − V cos θ ]
 V (5)
(5)
Q Q== ZV[ E [ E
sin(θ − δ) − V sin θ ]
sin(θ − δ ) − V sin θ ]
 Z
In a high-voltage or mid-voltage power grid, the line impedance is generally considered to be
inductive (θ = 90◦ ) because
In a high-voltage the impedance
or mid-voltage duegrid,
power to linetheinductance is much
line impedance is larger thanconsidered
generally that due totoline
be
resistance.
inductive (θMoreover, the phase
= 90°) because angle difference
the impedance due toδ lineis normally
inductanceunder ◦
30 forlarger
is much a synchronous generator
than that due to line
and in this case,
resistance. sin δ ≈the
Moreover, δ and
phase δ ≈ 1.
cosangle Then, (5)δcan
difference be simplified
is normally underas:30° for a synchronous generator
and in this case, sin δ ≈ δ and cos δ ≈ 1. Then, ( (5) can be simplified as:
P = EV Z δ
 EVV ( E −V ) (6)
 PQ== Z δZ
 (6)
Consequently, the active power can be regulated V ( E −byV )adjusting the phase angle δ and the reactive
Q=
 Z the inverter output voltage, which is the basis
power can be regulated by adjusting the amplitude of
of theConsequently,
droop characteristics of power
the active VSG. However, in the low-voltage
can be regulated by adjustingmicrogrid,
the phase there
angleisδ large
and theresistance
reactive
power can be regulated by adjusting the amplitude of the inverter output voltage, which is the basis
of the droop characteristics of VSG. However, in the low-voltage microgrid, there is large resistance
in the connecting line. The approximate decoupling cannot be obtained and strong coupling exists
between active and reactive power controls. In this paper, to solve this problem, the linearized model
of output power is established first.
Energies 2017, 10, 2159 5 of 13

in the connecting line. The approximate decoupling cannot be obtained and strong coupling exists
between active and reactive power controls. In this paper, to solve this problem, the linearized model
of output power is established first.
The output power deviation matrix caused by the respective deviations Ê and δ̂ at the quiescent
Energies 2017, 10, 2159 5 of 13
point (Es , δs ) can be expressed as:

sin((θθ−−δ sδ)s ) cos

P̂  Pˆ  VV  EEs ssin ( δ s )δs )δˆ  δ̂
" # " #" #
cos θ(θ− −
==
Q̂ Qˆ  ZZ  −−EEscoscos(θ(θ−−δ sδ)s ) sinsin
(θ(θ− −  
δ s )δs ) Eˆ  Ê
(7)(7)
   s
Equation(7)
Equation (7)represents
representsthe theoutput
outputpower
powermodel
modeland
and the associated 22 ××22matrix
the associated matrixisiscalled
calledthe
the
powertransfer
power transfermatrix
matrix(PTM).
(PTM).Then,
Then,thethesmall-signal
small-signalVSG
VSGcontrol
controlstructure
structureofofthe
thepower
powerlooploopcan
canbebe
derived from (2), (3), and (7) and is illustrated in Figure 3. The power coupling
derived from (2), (3), and (7) and is illustrated in Figure 3. The power coupling components in thecomponents in the
dashed boxes
dashed boxes cancanbebe observed
observedclearly. WhenWhen
clearly. the impedance angle θ decrease,
the impedance angle θ the coupling
decrease, thecomponents
coupling
between the active and reactive power will increase. Moreover, the coupling component
components between the active and reactive power will increase. Moreover, the coupling component between the
active power and the voltage amplitude is much smaller than that between the reactive
between the active power and the voltage amplitude is much smaller than that between the reactive power and the
angular
power andfrequency.
the angular frequency.

ˆ
P̂set 1 ω̂ 1 δ VEs
sin (θ − δ s )
J ωN s + DpωN s Z P̂e
ω̂ref
VEs
- cos (θ − δ s )
Z
V
cos (θ − δ s )
Z
Q̂set 1 Ê V Q̂e
sin (θ − δ s )
Ks + 2 Dq Z

Figure
Figure3.3.Small-signal
Small-signalequivalent
equivalentmodel
modelofofVSG.
VSG.

4.4.Decoupling
DecouplingControl
ControlMethod
Methodfor forVSG
VSGControl
Control
This
This paper intends to compensate forthe
paper intends to compensate for the coupling
couplingcomponent
componentby byadopting
adoptingaadirectdirect power
power
decoupling method. Direct decoupling control methods have two popular
decoupling method. Direct decoupling control methods have two popular types: dynamic decoupling types: dynamic
decoupling and static decoupling.
and static decoupling. Dynamic decoupling
Dynamic decoupling requires thatrequires that the
the channels are channels
decoupled areindecoupled in
the transition
the transition process and in the steady-state process. This method deeply relies
process and in the steady-state process. This method deeply relies on the mathematical model of theon the mathematical
model
systemofandthe cannot
systemattain
and cannot attaineffect
the desired the desired effect if
if the model the model errors
experiences experiences errors orperturbation.
or parameter parameter
perturbation. Static decoupling only requires the channels to be decoupled in the
Static decoupling only requires the channels to be decoupled in the steady-state process. Although steady-state process. it
Although it may degrade the decoupling performance and incurs dynamic deviation,
may degrade the decoupling performance and incurs dynamic deviation, it has a lower sensitivity to it has a lower
sensitivity to model error and parameter perturbation. Thus, this paper adopts the static decoupling
model error and parameter perturbation. Thus, this paper adopts the static decoupling method.
method.
4.1. Control Principle
4.1. Control Principle
A feed-forward decoupling compensator for a two-input and two-output (TITO) system was
A feed-forward
introduced decoupling
to obtain the approximate compensator
decouplingfor a two-input
[21]. However, and two-output
the influences on (TITO) system
the original was
coupling
introduced to obtain
channel in series withthetheapproximate
compensatordecoupling [21]. However,
are not considered. This may thechange
influences on the original
the proportional gain
coupling
of the original channel and affect the stability of the power control loop. In the previouschange
channel in series with the compensator are not considered. This may section,the
the
proportional
power independentgain of control
the original
can bechannel
realizedand
whenaffect
the the
PTM stability of the matrix.
is a diagonal power control
Borrowing loop. In the
from this
previous
thought andsection, the power
the modern independent
control theory, acontrol
diagonalcandecoupling
be realizedmatrix
when the PTM is a diagonal matrix.
is introduced.
Borrowing from this thought and the modern control theory, a diagonal decoupling matrix is
introduced.
It follows from (6) and Figure 3 that the coupling channels of the output power can be regarded
as a TITO system. Inputs are and and outputs are and in the complex frequency domain.
domain. Static decoupling requires that the transfer function matrix be a nonsingular diagonal
constant matrix. The PTM of Equation (7) can be expressed as:
Energies 2017, 10, 2159 6 of 13

It 2017,
Energies follows from
10, 2159 (6) and Figure 3 that the coupling channels of the output power can be regarded 6 of 13
as a TITO system. Inputs are Ê and δ̂ and outputs are P̂e and Q̂e in the complex frequency domain.
Static decoupling requires that the transfer function matrix be a nonsingular diagonal constant matrix.
V  Es sin (θ − δ s ) cos (θ − δ s ) 
The PTM of Equation (7) can be expressed Gs =  as:  (8)
Z − E cos (θ − δ s ) sin (θ − δ s ) 
" s #
V Es sin(θ − δs ) cos(θ − δs )
The decoupled transfer matrix Gs = can be expressed as a diagonal matrix (8)
Z − Es cos(θ − δs ) sin(θ − δs )
V  Es sin (θ − δ s ) 0 
= be
W can  expressed as a diagonalmatrix (9)
The decoupled transfer matrix Z 0 sin (θ − δ s ) 
" #
Then, the transfer function of controlled V Es sin (θ − δin
object s ) series with 0 the decoupling link Gc becomes
W= (9)
Z 0 sin(θ − δs )
 Pˆ  δˆ  δˆ 
  = GsGc   = W   (10)
ˆ
Then, the transfer function of controlled Q  object  Eˆin series  Eˆ with
 the decoupling link Gc becomes
" # " # " #
Gc can be calculated as: P̂ δ̂ δ̂
= Gs Gc =W (10)
Q̂ Ê −1 Ê
 VEs V  VEs 
−1
 Z sin (θ − δ s ) Z cos (θ − δ s )   Z sin (θ − δ s ) 0 
= Gs W =  as:
Gc calculated
Gc can be   
 − VEs cos (θ − δ ) V sin (θ − δ )   0
V
sin (θ − δ ) 
s Z s
  VEs
s # −1 " s
 #
VE V Z  Z 0 
"
− 1 Z sin(θ − δs ) Z cos(θ − δs ) Z sin(θ − δs )
Gc = Gs W  = − VE −Vsin (θ − δ ) cos (θ − δ s )  V
sin 2 Z("θ cos
− δ s()θ − δs ) Z sin(θ −s δs ) Z sin( θ − δs )
s
0 (11)
 
= E
#
− − −
2
sin (θ − δs ) s sin ( θ δ

s) cos ( θ δs)

 Es sin (θ =
Es (11)
− δs ) E cossin(θ (−θδ−
s ) δ ) cos ( sin
θ −
2
(
δ θ) − δ s ) sin2 (θ − δ )
s s " s # s
 Gc11 Gc12  Gc11 Gc12
=  =
Gc 21 Gc 22  Gc21 Gc22

It
It is
is observed
observed from
from (11)
(11) that
that only
only the
the impedance
impedance angle
angle is
is needed
needed and
and the
the definite
definite length
length of
of the
the
distributed
distributed line
line is
is not
not required.
required. This
This method
method is
is based
based on
on compensating
compensating forfor the
the coupling
coupling component.
component.
Figure
Figure 44 shows
shows the
the control
control diagram
diagram ofof the
the power
power loop
loop in
in series
series with
with the
the decoupling
decoupling link.
link.

P̂set 1 ω̂ 1 δˆ VEs
sin (θ − δ s )
J ωN s + DpωN s Z P̂e
ω̂ref
VEs
- cos (θ − δ s )
Z

V
cos (θ − δ s )
Z
Q̂set 1 Ê Q̂e
V
sin (θ − δ s )
Ks + 2 Dq Z

Figure
Figure 4.
4. Control
Control diagram
diagram of
of the
the power
power loop
loop in
in series
series with
with the
the decoupling
decoupling link.
link.

4.2. Feasibility Analysis and Improvement Approach for Power Sharing

A diagonal compensated matrix has been introduced to attain power decoupling based on the
quiescent operating point. According to the relative gain array (RGA) method of the decoupling
control [21], the RGA of the Gc can be derived as:
Energies 2017, 10, 2159 7 of 13

4.2. Feasibility Analysis and Improvement Approach for Power Sharing

A diagonal compensated matrix has been introduced to attain power decoupling based on the
Energies 2017,operating
quiescent 10, 2159 point. According to the relative gain array (RGA) method of the decoupling 7 of 13
control [21], the RGA of the Gc can be derived as:
 Gc11Gc 22  # "2
λ11 =λ11G= GGc22 λ =1 − λ  (θ 2δ
) (θ 2δ( θ) − δ )
" #
2
c11 Gc22
λ 12 =121 − λ 11
11  sin sin − ( θ − δcos
s ) cos− s
Λ= Λ= G G−Gc12−GGc21 G == s s
λ21 = 1c11− cλ2211 c12 c 21 λ22 = λ11  cos2 cos 
c11 (12)
(12)
λ21 =1 − λ11 λ22 =λ11   ( θ −2 (δθs )− δsin
s ) (θ
2
−2δ(sθ) − δs )
sin

From (12), the coefficients of RGA always vary in (0, 1), which means that the introduction of
From (12), the coefficients of RGA always vary in (0, 1), which means that the introduction of
the decoupling link cannot affect the stability of the system. On the other hand, since (7) denotes
the decoupling link cannot affect the stability of the system. On the other hand, since (7) denotes the
the average values of the instantaneous power flows over half the fundamental frequency cycle,
average values of the instantaneous power flows over half the fundamental frequency cycle,
discussions can only reflect the characteristics of the system in the frequency range that is lower
discussions can only reflect the characteristics of the system in the frequency range that is lower than
than two times the power frequency. The decoupling method based on the diagonal matrix is only
two times the power frequency. The decoupling method based on the diagonal matrix is only
applicable to the power regulation under small disturbance near the quiescent operating point. If the
applicable to the power regulation under small disturbance near the quiescent operating point. If the
output voltage amplitude and power angle change over a wide range, the decoupling effect of the
output voltage amplitude and power angle change over a wide range, the decoupling effect of the
system may be reduced and it is necessary to dynamically adjust the coefficient of power decoupling
system may be reduced and it is necessary to dynamically adjust the coefficient of power decoupling
link. Otherwise, the decoupling may be a failure. Nonetheless, due to its simple structure and better
link. Otherwise, the decoupling may be a failure. Nonetheless, due to its simple structure and better
stability, this method is suitable for power decoupling in a microgrid which has a diversity of loads
stability, this method is suitable for power decoupling in a microgrid which has a diversity of loads
and frequent small power fluctuations.
and frequent small power fluctuations.
The other issue considered by power decoupling control is power sharing. To guarantee the
The other issue considered by power decoupling control is power sharing. To guarantee the
sharing of actual power, improved virtual power-based control with a unified rotation angle is
sharing of actual power, improved virtual power-based control with a unified rotation angle is
proposed [9]. Because the transformation matrix is nonsingular, the actual power will be shared
proposed [9]. Because the transformation matrix is nonsingular, the actual power will be shared
equally as long as the virtual power is evenly shared among parallel inverters. Although the limitation
equally as long as the virtual power is evenly shared among parallel inverters. Although the
of the actual power sharing performance heavily relies on the consistence of line impedance angles of
limitation of the actual power sharing performance heavily relies on the consistence of line
the inverters is overcome, when the transmission lines among parallel inverters are different, the power
impedance angles of the inverters is overcome, when the transmission lines among parallel inverters
decoupling effect will still be diminished.
are different, the power decoupling effect will still be diminished.
Introduction of the decoupling link does not change the gain of the original channel. When the
Introduction of the decoupling link does not change the gain of the original channel. When the
system reaches the steady state, a simplified equation can be derived:
system reaches the steady state, a simplified equation can be derived:
PsetP− −eP=
set P e =D ωNN((ωω−−ωref
Dppω ωref) ) (13)
(13)

Equation (13)
Equation (13) describes
describes the
the droop
droop characteristic
characteristic between
between the
the active
active power
power and
and frequency
frequency of
of VSG.
VSG.
The active
The active power
power sharing
sharing can
can be
be achieved
achieved directly
directly by
by making
making D Dpp of
of each
each inverter
inverter proportional
proportional toto its
its
capacity. For Q–V control, however, the output voltage of the inverter will be not same
capacity. For Q–V control, however, the output voltage of the inverter will be not same due to the due to the
existence of
existence of line
line voltage
voltage drop.
drop. This
This phenomenon affects the
phenomenon affects the reactive power sharing.
reactive power sharing. To overcome this
To overcome this
problem, the PCC voltage is introduced as a consistent variable. Figure 5 shows the improved
problem, the PCC voltage is introduced as a consistent variable. Figure 5 shows the improved reactive reactive
power loop.
power loop. The
Thereactive
reactivepower
powersharing
sharingcan
canbebeensured
ensured byby making
making DqDofq of each
each inverter
inverter proportional
proportional to
to its
its capacity.
capacity.

2Vref
1
Ks
2Vpcc

Figure 5. Improved Q–V control loop.

Figure 5. Improved Q–V control loop.

5. Simulation Verification

5.1. Verification of the Power Decoupling Performance

The proposed power decoupling method is verified by using MATLAB/Simulink in various
respects. The main circuit and control scheme of one inverter are illustrated in Figure 1. Control
parameters for the VSG controlled inverter are listed in Table 1. The line impedance is 0.8 + j0.5 Ω.
Energies 2017, 10, 2159 8 of 13

5. Simulation Verification

5.1. Verification of the Power Decoupling Performance

The
Energies proposed
2017, power decoupling method is verified by using MATLAB/Simulink in various
10, 2159 8 of 13
respects. The main circuit and control scheme of one inverter are illustrated in Figure 1. Control
Table
parameters for the VSG controlled 1. Electrical
inverter and control
are listed 1. The line impedance is 0.8 + j0.5 Ω.
parameters.
in Table
Parameters Values Parameters Values
Table 1. Electrical and control parameters.
Vdc 800 V Vpcc 220 V
Parameters L 2 mH
Values Pset
Parameters 10 kW Values
V dc C 300
800 V μF QsetVpcc 5 kVar 220 V
L Rin 2 0.05
mH Ω J Pset 0.2 10 kW
C 300 µF Qset 5 kVar
Rin ωN 314 Ω
0.05 rad/s Dp J 20 0.2
ωN Es 314235.7
rad/sV Dq Dp 500 20
Es δs 235.7 V
0.07 K Dq 50 500
δs 0.07 K 50

Two cases are considered. In Case 1, the frequency of the grid voltage is 50 Hz from 0 s < t < 1 s
Two
and is cases are
changed considered.
to 49.9 Hz at 1.0In
s. Case 1,6a
Figure theshows
frequency of the grid
the frequency voltage isof50the
waveform Hzoutput s < t < 1Ins
from 0voltage.
and is2,changed
Case a 0.02 putoincrease
49.9 Hzin
atthe
1.0 voltage
s. Figure 6a showsof
amplitude the frequency
the waveform
distribution network ofoccurs
the output
at 1 s,voltage.
as shown In
Case
in 2, a 0.02
Figure 6b. pu increase in the voltage amplitude of the distribution network occurs at 1 s, as shown
in Figure 6b.

(a) (b)
Figure
Figure 6.
6. (a)
(a) The
The frequency
frequency waveform
waveform of
of the
the output
output voltage; (b) The
voltage; (b) The PCC voltage waveform.
PCC voltage waveform.

Figure 7a,b show the waveforms of line power without and with the power decoupling link,
Figure 7a,b show the waveforms of line power without and with the power decoupling link,
respectively. Plots in the left column correspond to the frequency deviation and that in the right
respectively. Plots in the left column correspond to the frequency deviation and that in the right
column to the voltage deviation. Note that the line power is the power injected into the distribution
column to the voltage deviation. Note that the line power is the power injected into the distribution
network. The measured values are slightly smaller than the set values. After the frequency decreases,
network. The measured values are slightly smaller than the set values. After the frequency decreases,
it can be seen from Figure 7a that the active power output increases by about 3.8 kW, which means
it can be seen from Figure 7a that the active power output increases by about 3.8 kW, which means that
that VSG can change its active power output to respond to the frequency exception event of the grid-
VSG can change its active power output to respond to the frequency exception event of the grid-side.
side. To benefit from the introduction of the swing Equation (2), the frequency of VSG control changes
To benefit from the introduction of the swing Equation (2), the frequency of VSG control changes
slowly, which means that VSG provides virtual inertia to the system to allow the frequency of VSG
slowly, which means that VSG provides virtual inertia to the system to allow the frequency of VSG
control to change slowly. This is beneficial to the stable operation of the power grid. Meanwhile, the
control to change slowly. This is beneficial to the stable operation of the power grid. Meanwhile,
impact of the frequency fluctuation on the reactive power decreases with the help of the power
the impact of the frequency fluctuation on the reactive power decreases with the help of the power
decoupling link.
decoupling link.
Energies 2017, 10, 2159 99 of
of 13
13
Energies 2017, 10, 2159 9 of 13

(a)
(a)

(b)
(b)
Figure 7. (a) The power output response to the frequency fluctuation; (b) The power output response
Figure
Figure 7.7. (a)
(a) The
The power
power output
output response
response to
to the
the frequency
frequency fluctuation;
fluctuation; (b)
(b) The
The power
power output
output response
response
to the voltage amplitude fluctuation.
to
to the
the voltage
voltage amplitude
amplitudefluctuation.
fluctuation.
It is observed from Figure 7b that active and reactive power responses with and without power
It is
It is observed
observed fromfrom Figure
Figure 7b7b that
that active
active and
and reactive
reactive power
power responses
responses with
with and
and without
without power
power
decoupling are almost the same. This indicates that the voltage fluctuation does not affect the active
decoupling are
decoupling are almost
almost the
the same.
same. This
This indicates
indicates that
that the
the voltage
voltage fluctuation
fluctuation does
does not
not affect
affect the
the active
active
power and reactive power much. In other words, the coupling extent between the active power and
power and reactive
power is less than that power
power much. In
much. the other words,
In reactive
other words,the coupling
the couplingextent between
extent betweenthe active power
the active and
power
voltage between power and frequency, which is consistent with the
voltage
and voltageis less thanthan
is less that that
between
between the reactive power and and
the reactive frequency, which is consistent withwith
the
theoretical analysis in Section 3. It is observed from power frequency,
Figure 7b that VSG alsowhich is consistent
responds to the voltage
theoretical
the theoretical analysis in Section
analysis in 3. It is
Section 3. observed
It is from Figure
observed from 7b that7b
Figure VSG also
that VSG responds
also to the voltage
responds to the
fluctuation.
fluctuation.
voltage fluctuation.
5.2. Verification of the Power Sharing Performance
5.2.
5.2. Verification
Verification of
of the
the Power
Power Sharing
Sharing Performance
Performance
Figure 8 depicts the microgrid platform used in the simulation, which contains a 20 kVA VSG
Figure
Figure 88 depicts
depicts the
the microgrid
microgrid platform
platform used
used inin the
the simulation,
simulation, which contains a 20 kVA
kVA VSGVSG
and a 10 kVA VSG. The microgrid is running under the grid-connected mode. The parameter values
and
and aa 10
10 kVA
kVAVSG.
VSG.The Themicrogrid
microgridisisrunning
runningunder
underthethegrid-connected
grid-connectedmode.
mode. The
The parameter
parameter values
values
of respective VSGs are listed in Table 2. The line impedances of two DGs are different: one is ZLine1 =
of
of respective VSGsVSGs arearelisted
listedininTable
Table2. 2.
TheThe
lineline impedances
impedances of two
of two DGs DGs are different:
are different: one is one
ZLine1is=
0.8 + j0.5 Ω, and another is ZLine2 = 0.5 + j0.83 Ω.
Z
0.8 = 0.8
+ j0.5
Line1 Ω,+and Ω, and another
j0.5another ZLine2
is ZLine2 =is0.5 Ω. + j0.83 Ω.
= 0.5
+ j0.83

Microgrid simulation
Figure 8. Microgrid
Figure simulation platform.
Figure 8. Microgrid simulation platform.
Case 1: The frequency of grid voltage decreases by 0.1 Hz at 1.0 s and restores at 2.0 s. Figure 9
Case 1: The frequency of grid voltage decreases by 0.1 Hz at 1.0 s and restores at 2.0 s. Figure 9
shows the line powers of VSGs without and with the decoupling link. In Figure 9a, the two VSGs
shows the line powers of VSGs without and with the decoupling link. In Figure 9a, the two VSGs
Energies 2017, 10, 2159 10 of 13

EnergiesTable
2017, 10,
2. 2159
Electrical and control parameters of two virtual synchronous generators (VSGs).10 of 13

respond well to the frequency fluctuation and increase their active powers output according to their
Parameters (VSG1) Values Parameters (VSG2) Values
capacity ratios. This result verifies the active power sharing effect proposed in Section 4. It is also
V dc
observed that 800power
the change of the active V V dc of the reactive power.
results in the change 600 V On the other
L 3 mH L
hand, Figure 9b illustrates that the decoupling link reduces the influence of the 2 mHactive power
C 400 µF C 300 µF
fluctuation on the reactive power. Figure 9c shows the output current waveform of VSG1 and the
Rin 0.08 Ω Rin 0.05 Ω
total harmonic distortion (THD) values verify the accuracy of the parameters selection.
ωN 314 rad/s ωN 314 rad/s
Es 235.7 V Es 230.1 V
Table 2. Electrical and control parameters of two virtual synchronous generators (VSGs).
δs 0.07 δs 0.05
V pcc Parameters (VSG1) 220 V Values Parameters V pcc(VSG2) Values 220 V
Pset 10 kW Pset 5 kW
Qset Vdc 5 kVar 800 V V
Qdcset 600 V 5 kVar
J L 0.15 3 mH LJ 2 mH 0.2
Dp C 30 400 μF D
Cp 300 μF 15
Dq Rin 600 0.08 Ω D
Rin q 0.05 Ω 300
K 55 K 50
ωN 314 rad/s ωN 314 rad/s
Es 235.7 V Es 230.1 V
Case 1: The frequency ofδsgrid voltage decreases 0.07 by 0.1 δHz
s at 1.0 s and restores at 2.0 s. Figure 9
0.05
shows the line powers of VSGs Vpcc without and 220 with
V the decoupling
Vpcc link. In
220Figure
V 9a, the two VSGs
respond well to the frequency fluctuation
Pset and
10 kW increase their active
Pset powers output
5 kW according to their
capacity ratios. This result verifies
Qset the active
5 kVarpower sharing Qset effect proposed
5 kVarin Section 4. It is also
observed that the change of the J active power0.15 results in theJ change of the
0.2reactive power. On the
other hand, Figure 9b illustrates Dp that the decoupling
30 link reduces
Dp the influence
15 of the active power
fluctuation on the reactive power. Dq Figure 9c shows 600 the output Dqcurrent waveform 300 of VSG1 and the total
harmonic distortion (THD) values K verify the accuracy
55 of the Kparameters selection.
50

(a)

(b)

Figure 9. Cont.
Energies 2017, 10, 2159 11 of 13

EnergiesEnergies
2017, 10, 2159
2017, 10, 2159 11 of 13
11 of 13

(c)
(c)
Figure 9. Output
Figure waveforms
9. Output waveforms of of
VSGVSGcontrol
controlto
to frequency deviation.
frequency deviation. (a)(a) Without
Without decoupling
decoupling link; link;
(b) (b)
Figure 9. Output waveforms of VSG control to frequency deviation. (a) Without decoupling link;
With With decoupling
decoupling link;link;
(c) (c) Output
Output currentof
current ofVSG1
VSG1 with
with decoupling
decouplinglink.
link.
(b) With decoupling link; (c) Output current of VSG1 with decoupling link.

CaseCase
2: The2: The amplitude
amplitude of of
thethe gridvoltage
grid voltage decreases
decreases by by0.02pu
0.02puatat 1s1sand
andrestores at 2s.
restores atFigure 10 10
2s. Figure
shows2:the
Case The line powers of
amplitude ofVSGs
the without
grid voltageanddecreases
with decoupling
by link.atIn1sFigure
0.02pu and 10a, theattwo
restores 2s. VSGs
Figure 10
shows the line powers of VSGs without and with decoupling link. In Figure 10a, the two VSGs
shows respond
thetolineto the voltage fluctuation by adjusting their reactive power output. However, without the
respond thepowers
voltageoffluctuation
VSGs without and with decoupling
by adjusting their reactivelink. In Figure
power 10a, However,
output. the two VSGs respond
without the
to theintroduction
voltage of the PCC
fluctuation by voltage, the
adjusting their two VSGs power
reactive does not increase
output. their reactive
However, without power
the output
introduction
introduction of the PCC voltage, the two VSGs does not increase their reactive
according to their capacity ratios. This situation has been improved in Figure 10b. Moreover, the power output
of the PCCto
according voltage,capacity
the two ratios.
VSGs does not increase
hastheir reactive power output10b.according to their
change oftheirthe reactive power willThis situation
not affect the activebeen
power improved in Figure
output, which is consistent Moreover,
with the the
capacity
change ratios.
of This situation
the reactive power has
willbeen
not improved the in
affectFigure Figure
active 10b. output,
power Moreover, the change
which of the with
is consistent reactive
theoretical analysis in Section 3. Similarly, 10c shows the output current waveform of VSG2 the
power will
theoretical not
and theanalysis affect
THD values the active
in Section power
3. accuracy
verify the output,
Similarly, which
of Figure is
10c shows
the parameters consistent with the theoretical
the output current waveform of VSG2
selection. analysis in
Section 3. Similarly, Figure 10c shows the output current
and the THD values verify the accuracy of the parameters selection. waveform of VSG2 and the THD values
verify the accuracy of the parameters selection.

(a)

(a)

(b)

(b)

Figure 10. Cont.

Energies 2017,
Energies 2017, 10,
10, 2159
2159 12
12 of
of 13
13

(c)
Figure 10. Output waveforms of VSG control to voltage deviation. (a) Without decoupling link; (b)
Figure 10. Output waveforms of VSG control to voltage deviation. (a) Without decoupling link;
With decoupling link; (c) Output current of VSG2 with decoupling link.
(b) With decoupling link; (c) Output current of VSG2 with decoupling link.

6. Conclusions
6. Conclusions
This paper has presented a power decoupling method based on the diagonal compensating
This paper has presented a power decoupling method based on the diagonal compensating matrix
matrix for VSG. By making the power transfer matrix a diagonal matrix, the nearly decoupling effect
for VSG. By making the power transfer matrix a diagonal matrix, the nearly decoupling effect can be
can be obtained. Compared with the traditional decoupling method, this new scheme requires no
obtained. Compared with the traditional decoupling method, this new scheme requires no complex
complex design and specific line parameter, which is helpful to the service for practical engineering.
design and specific line parameter, which is helpful to the service for practical engineering. Moreover,
Moreover, power sharing among parallel inverters can be implemented with the introduction of the
power sharing among parallel inverters can be implemented with the introduction of the PCC voltage.
PCC voltage. Simulation results have demonstrated the merits of the proposed power decoupling
Simulation results have demonstrated the merits of the proposed power decoupling control and the
control and the ability of VSG in the response to the grid abnormal events.
ability of VSG in the response to the grid abnormal events.
Acknowledgments: This
Acknowledgments: This work
work was
was supported
supported by the National
by the National Natural
Natural Science
Science Foundation
Foundation of
of China
China (grant
(grant #:
#:
Talent Recruitment
51477021) and the “111” State innovation & Talent RecruitmentBase
Base(B08036).
(B08036).
Author Contributions: All
Author Contributions: All authors
authors contributed
contributed substantially
substantially to
to all
all aspects
aspects of
of this
this article.
article.
Conflicts of Interest: The authors declare no conflict of interest.
Conflicts of Interest: The authors declare no conflict of interest.

References
References
1. Banerji, A.; Sen, D.; Bera, A.K. Microgrid: A review. In Proceedings of the Global Humanitarian Technology
1. Banerji, A.; Sen, D.; Bera, A.K. Microgrid: A review. In Proceedings of the Global Humanitarian Technology
Conference: South Asia Satellite (GHTC-SAS), Trivandrum, India, 23–24 August 2013.
Conference: South Asia Satellite (GHTC-SAS), Trivandrum, India, 23–24 August 2013.
2. Lasseter, R.H. MicroGrids. In Proceedings of the IEEE Power Engineering Society Winter Meeting, New York,
2. Lasseter, R.H. MicroGrids. In Proceedings of the IEEE Power Engineering Society Winter Meeting, New
NY, USA, 27–31 January 2002.
York, NY, USA, 27–31 January 2002.
3. Azmy, A.M.; Erlich, I. Impact of Distributed Generation on the Stability of Electrical Power Systems. In
3. Azmy, A.M.; Erlich, I. Impact of Distributed Generation on the Stability of Electrical Power Systems. In
Proceedings of the Power Engineering Society General Meeting, San Francisco, CA, USA, 16 June 2005.
Proceedings of the Power Engineering Society General Meeting, San Francisco, CA, USA, 16 June 2005.
4. Aktarujjaman, M.; Haque, M.E.; Muttaqi, K.M.; Negnevitsky, M.; Ledwich, G. Control Stabilisation of
4. Aktarujjaman, M.; Haque, M.E.; Muttaqi, K.M.; Negnevitsky, M.; Ledwich, G. Control Stabilisation of
Multiple Distributed Generation. In Proceedings of the Power Engineering Conference (AUPEC 2007), Perth,
Multiple Distributed Generation. In Proceedings of the Power Engineering Conference (AUPEC 2007),
Australia, 9–12 December 2007.
Perth, Australia, 9–12 December 2007.
5. Li, D.; Zhu, Q.; Lin, S. A Self-Adaptive Inertia and Damping Combination Control of VSG to Support
5. Li, D.; Zhu, Q.; Lin, S. A Self-Adaptive Inertia and Damping Combination Control of VSG to Support
Frequency Stability. IEEE Trans. Energy Convers. 2017, 32, 397–398. [CrossRef]
Frequency Stability. IEEE Trans. Energy Convers. 2017, 32, 397–398.
6. Suul, J.A.; Darco, S.; Guidi, G. Virtual Synchronous Machine-Based Control of a Single-Phase Bi-Directional
6. Suul, J.A.; Darco, S.; Guidi, G. Virtual Synchronous Machine-Based Control of a Single-Phase Bi-Directional
Battery Charger for Providing Vehicle-to-Grid Services. IEEE Trans. Ind. Appl. 2016, 52, 3234–3244. [CrossRef]
Battery Charger for Providing Vehicle-to-Grid Services. IEEE Trans. Ind. Appl. 2016, 52, 3234–3244.
7. Cvetkovic, I.; Boroyevich, D.; Burgos, R. Experimental Verification of a Virtual Synchronous Generator
7. Cvetkovic, I.; Boroyevich, D.; Burgos, R. Experimental Verification of a Virtual Synchronous Generator
Control Concept. In Proceedings of the Control and Modeling for Power Electronics (COMPEL), Trondheim,
Control Concept. In Proceedings of the Control and Modeling for Power Electronics (COMPEL),
Norway, 27–30 June 2016.
Trondheim, Norway, 27–30 June 2016.
8. Wu, H.; Ruan, X.; Yang, D.; Chen, X.; Zhao, W.; Lv, Z.; Zhong, Q.C. Small-Signal Modeling and Parameters
8. Wu, H.; Ruan, X.; Yang, D.; Chen, X.; Zhao, W.; Lv, Z.; Zhong, Q.C. Small-Signal Modeling and Parameters
Design for Virtual Synchronous Generators. IEEE Trans. Ind. Electron. 2016, 63, 4292–4303. [CrossRef]
Design for Virtual Synchronous Generators. IEEE Trans. Ind. Electron. 2016, 63, 4292–4303.
9. Wu, T.; Liu, Z.; Liu, J.; Wang, S. A Unified Virtual Power Decoupling Method for Droop-Controlled Parallel
9. Wu, T.; Liu, Z.; Liu, J.; Wang, S. A Unified Virtual Power Decoupling Method for Droop-Controlled Parallel
Inverters in Microgrids. IEEE Trans. Power Electron. 2016, 31, 5587–5603. [CrossRef]
Inverters in Microgrids. IEEE Trans. Power Electron. 2016, 31, 5587–5603.
Energies 2017, 10, 2159 13 of 13

10. Rowe, C.N.; Summers, T.J.; Betz, R.E.; Cornforth, D.J. Arctan Power–Frequency Droop for Improved
Microgrid Stability. IEEE Trans. Power Electron. 2013, 28, 3747–3759. [CrossRef]
11. De Brabandere, K.; Bolsens, B.; van den Keybus, J.; Woyte, A.; Driesen, J. A Voltage and Frequency Droop
Control Method for Parallel Inverters. IEEE Trans. Power Electron. 2007, 22, 1107–1115. [CrossRef]
12. Li, Y.; Li, Y.W. Power Management of Inverter Interfaced Autonomous Microgrid Based on Virtual
Frequency-Voltage Frame. IEEE Trans. Smart Grid 2011, 2, 30–40. [CrossRef]
13. Li, Y.; Li, Y.W. Virtual frequency-voltage frame control of inverter based low voltage microgrid.
In Proceedings of the Electrical Power & Energy Conference (EPEC), Montreal, QC, Canada,
22–23 October 2009.
14. Zhang, P.; Zhao, H.; Cai, H. Power decoupling strategy based on ‘virtual negative resistor’ for inverters in
low-voltage microgrids. IET Power Electron. 2016, 9, 1037–1044. [CrossRef]
15. Wu, X.; Shen, C.; Iravani, R. Feasible Range and Optimal Value of the Virtual Impedance for Droop-Based
Control of Microgrids. IEEE Trans. Smart Grid 2016, 8, 1242–1251. [CrossRef]
16. Zhu, Y.; Zhuo, F.; Wang, F.; Liu, B. A Wireless Load Sharing Strategy for Islanded Microgrid Based on Feeder
Current Sensing. IEEE Trans. Power Electron. 2015, 30, 6706–6719. [CrossRef]
17. Mahmood, H.; Michaelson, D.; Jiang, J. Accurate Reactive Power Sharing in an Islanded Microgrid Using
Adaptive Virtual Impedances. IEEE Trans. Power Electron. 2015, 30, 1605–1617. [CrossRef]
18. He, J.; Li, Y.W.; Blaabjerg, F. An Enhanced Islanding Microgrid Reactive Power, Imbalance Power, and
Harmonic Power Sharing Scheme. IEEE Trans. Power Electron. 2015, 30, 3389–3401. [CrossRef]
19. Vandoorn, T.L.; Meersman, B.; Kooning, J. Analogy between Conventional Grid Control and Islanded
Microgrid Control Based on a Global DC-Link Voltage Droop. IEEE Trans. Power Deliv. 2012, 27, 1405–1414.
[CrossRef]
20. Guerrero, J.M.; Matas, J.; de Vicuña, L.G. Decentralized Control for Parallel Operation of Distributed
Generation Inverters in Microgrids Using Resistive Output Impedance. IEEE Trans. Ind. Electron. 2007, 54,
994–1004. [CrossRef]
21. Yan, X.; Zhang, Y. Power Coupling Analysis of Inverters based on Relative Gain Method and Decoupling
Control based on Feedforward Compensation. In Proceedings of the Renewable Power Generation, Beijing,
China, 17–18 October 2015.

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).