Aalborg Universitet

Distributed Secondary Control for Islanded MicroGrids - A Novel Approach

Shafiee, Qobad; Guerrero, Josep M.; Quintero, Juan Carlos Vasquez

Published in:
I E E E Transactions on Power Electronics

DOI (link to publication from Publisher):

10.1109/TPEL.2013.2259506

Publication date:
2014

Document Version
Early version, also known as pre-print

Link to publication from Aalborg University

Citation for published version (APA):

Shafiee, Q., Guerrero, J. M., & Vasquez, J. C. (2014). Distributed Secondary Control for Islanded MicroGrids - A
Novel Approach. I E E E Transactions on Power Electronics, 29(2), 1018-1031. DOI:
10.1109/TPEL.2013.2259506

General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

? Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
? You may not further distribute the material or use it for any profit-making activity or commercial gain
? You may freely distribute the URL identifying the publication in the public portal ?
Take down policy
If you believe that this document breaches copyright please contact us at [email protected] providing details, and we will remove access to
the work immediately and investigate your claim.

Downloaded from vbn.aau.dk on: maj 03, 2018

IEEE TRANSACTIONS ON POWER ELECTRONIC 1

Distributed Secondary Control for Islanded

MicroGrids - A Novel Approach
Qobad Shafiee, Student Member, IEEE, Josep M. Guerrero, Senior Member, IEEE, and Juan C. Vasquez,
Member, IEEE

 frequency droop control and the reactive power–voltage droop

Abstract— This paper presents a novel approach to conceive are adopted as the decentralized control strategies in the power
the secondary control in droop-controlled MicroGrids. The electronic based MGs for the autonomous power sharing
conventional approach is based on restoring the frequency and operations. Although the primary level does not require for
amplitude deviations produced by the local droop controllers by
using a MicroGrid Central Controller (MGCC). A distributed communications, in order to achieve global controllability of
networked control system is used in order to implement a the MG, secondary control is often used.
distributed secondary control (DSC) thus avoiding its The conventional secondary control approach relays on
implementation in MGCC. The proposed approach is not only using a MicroGrid Central Controller (MGCC), which includes
able to restore frequency and voltage of the MicroGrid but also slow controls loops and low bandwidth communication
ensures reactive power sharing. The distributed secondary systems in order to measure some parameters in certain points
control does not rely on a central control, so that the failure of a
single unit will not produce the fail down of the whole system. of the MG, and to send back the control output information to
Experimental results are presented to show the feasibility of the each DG unit [1], [2]. On the other hand, this MGCC also can
DSC. The time latency and data drop-out limits of the include tertiary control, which is more related to economic
communication systems are studied as well. optimization, based on energy prices and electricity market [1].
Tertiary control exchanges information with the distribution
Keywords – Secondary control, Distributed Control, Networked system operator (DSO) in order to make feasible and to
Control Systems, Droop Control, Cooperative Control.
optimize the MG operation within the utility grid.
Secondary control is conceived to compensate frequency and
I. INTRODUCTION voltage deviations produced inside the MG by the virtual
inertias and output virtual impedances of primary control. This
concept was used in large utility power systems for decades in
M icroGrids (MGs) are local grids comprise different
technologies such as power electronics converters,
distributed generations (DGs), energy storage systems, and
order to control the frequency of a large area electrical network
[14], [15] and it has been applied to MGs to restore frequency
and voltage deviations [1], [2], [9]-[13]. Furthermore, global
telecommunications which can operate connected to the
objectives regarding voltage control and power quality of the
traditional centralized grid (macrogrid) but also could operate
MG, such as voltage unbalance and harmonic compensation
autonomously in islanded mode.
have been proposed recently in additional secondary control
Control structures are essential to proper control of MGs
loops [16], [17]. In all of these literatures, a central secondary
providing stability and efficient operation. The important roles
control (CSC) has been used in order to manage the MG.
that can be achieved using these control structures are
On the other hand, the reactive power sharing of the Q–V
frequency and voltage regulation, active and reactive power
droop control is hard to achieve, since the voltage is not
control between DG units and with the main grid,
constant along the MG power line, as opposed to the frequency
synchronization of MG with the main grid, energy management
[18]. Consequently, reactive power sharing can be achieved by
and economic optimization [1]-[13]. Recently, hierarchical
implementing an external loop in the secondary level [19].
control for MGs has been proposed in order to standardize their
Significant efforts have been done in order to improve the
operation and functionalities [1]. In such a hierarchical
primary control method for power sharing in the recent years.
approach, three main control levels have been defined. The
In [20], a power controller was proposed, which contains a
primary control is the first level which is independent, dealing
virtual inductor loop for both active and reactive power
with the local control loops of the DG units. This can be
decoupling, and an accurate reactive power sharing algorithm
performed by voltage and current loops, droop functions, and
with an online impedance voltage drop effect estimation
virtual impedances. Conventionally, the active power–
considering different location of the different local loads in a
MG. This strategy, which is an improvement of the
Q. Shafiee, J. M. Guerrero, and J. C. Vasquez are with the Institute of conventional droop method, operates in the primary control
Energy Technology, Aalborg University, Aalborg East DK-9220, Denmark level therefore it does not need physical communications
(e-mails: [email protected], [email protected], [email protected]). among DG units. Alternatively, a reactive power-sharing
IEEE TRANSACTIONS ON POWER ELECTRONIC 2

scheme has been presented in [21], which introduces an integral of the primary control in MGs is described. Then, details of
control of the load bus voltage, combined with a reference that centralized secondary control for MGs are discussed in Section
is drooped against reactive power output. Further, active power III. Section IV is dedicated to the proposed secondary control
sharing has improved by computing and setting the phase angle strategy, which includes frequency control, voltage control and
of the DGs instead of its frequency in conventional frequency reactive power sharing. Experimental results and discussion are
droop control. In [22], a control strategy which increases the presented in Section V. Furthermore, the proposed secondary
droop gain to improve the accuracy of reactive power sharing is control is applied on a two paralleled 2.2kW-inverter system as
proposed by making a feedback reactive power injection loop a case study. Finally, the paper is concluded in Section VI.
around the conventional droop loop of each DG, while
maintaining the system stability. Additionally, secondary
control loops implemented in the MGCC has proposed to share II. PRIMARY CONTROL FOR MICROGRIDS
reactive power between DG units and also to restore the voltage Power electronics based MG consists of a number of elements
deviations in [19]. In all those techniques, reactive power that can operate in parallel either in islanded mode or connected
sharing cannot be achieved completely since voltage is a local to the main grid. Fig. 1 shows a general structure of MG, which
variable, as a contrary of frequency. composes n DG units. The MG is connected to the utility
Moreover, primary and tertiary controls are decentralized system through a static transfer switch (STS) at the point of
and centralized control levels respectively, since while one is common coupling (PCC). As depicted in Fig.1, each DG
taking care of the DG units, the other concerns about the MG system comprises a renewable energy source (RES), an energy
global optimization. However, although secondary control storage system (ESS), and a power electronic interface, which
systems conventionally have been implemented in the MGCC, normally consist of a dc-ac inverter. Each DG can be connected
in this paper we propose to implement it in a distributed way to a predefined load or to the AC common bus directly in order
along the local control with communication systems. In this to supply power.
sense, a local secondary control is determined for each DG to The dc/ac inverters are classified as voltage source inverters
generate set-points of the droop control to restore of the (VSIs) and current source inverters (CSIs) which the former is
deviations produced by the primary control. commonly used to inject current in grid connected modes and
This kind of distributed control strategies, which are also the latter to keep the frequency and voltage stable in
named networked control systems (NCS), have been reported autonomous operation. Both can operate in parallel in a MG.
recently in some literatures [9], [23]-[24]. In [9], technical However, VSIs are convenient since they can enhance power
aspects of providing frequency control reserves (FCRs) and the quality and ride-through capability for DGs in a MG [1], [25].
potential economic profitability of participating in FCR The primary control of VSIs based MG includes voltage and
markets for both decentralized and centralized coordination current control loops, virtual impedance loop and droop control
approach based on a setup of multiple MGs are investigated. In strategy as shown in Fig. 3. Linear and nonlinear control
[23], a pseudo-decentralized control strategy has been strategies are designed and performed in order to regulate the
presented for distributed generation networks which operate in output voltage and to control the current while maintaining the
distributed manner using a Global Supervisory Controller system stable. Normally, inner control loops include
(GSC) and local controllers with some intelligence. In the other proportional-resonant (PR) controller when they use stationary
hand, a master-slave control by using networked control framework (αβ), and proportional-integral (PI) controller when
strategy for the parallel operation of inverters has been they use the dq framework. The reference of the voltage control
introduced in [24]. The method is employed to achieve the loop will be generated, together with the droop controller and a
superior load-sharing accuracy compared to conventional virtual impedance loop.
droop scheme with low-bandwidth communication. Further, Droop control is responsible for adjusting the frequency and
the system robustness has been considered in the case of the amplitude of the voltage reference according to the active
communication failure as well. Distributed control strategies and reactive powers (P and Q), by using the well-known P/Q
have been used in all these literatures, however, the application droop method [1], [25]- [29]. Furthermore, a virtual impedance
of these control strategies to secondary control of MGs still has loop is also added to the voltage reference in order to fix the
not been proposed. output impedance of the VSI which will determine the P/Q
In this paper, a distributed secondary control strategy is power angle/amplitude relationships based on the droop
proposed for power electronics-based MGs, including method control law. In contrast with physical impedance, this
frequency, voltage and reactive power sharing controllers. This virtual output impedance has no power losses, and it is possible
way, every DG has its own local secondary control which can to implement resistance without efficiency losses [13]. More
produce appropriate control signal for the primary control level details about the primary control can be found in [1], [13],
by using the measurements of other DGs in each sample time. being out of scope of this paper.
In order to investigate the impact of communication on this new
control strategy, the communication latency is considered when
sending/receiving information to/from other DG units and the
results are compared with the conventional MGCC.
The paper is organized as follows. In section II, the structure
IEEE TRANSACTIONS ON POWER ELECTRONIC 3

Islanded MicroGrid
DG1 DG2 DGn

Renewable Renewable Renewable

Energy source Energy source Energy source
. . .
Energy Storage Energy Storage Energy Storage
System System System

Load

Load

Load
Grid

Fig. 1. General structure of MG.

being kpf and kif the control parameters of the secondary control
III. CENTRALIZED SECONDARY CONTROL FOR MICROGRIDS PI compensator. The frequency levels in the MG ( ) are
Since the primary control is local and does not have measured and compared to the references ( ) and the errors
intercommunications with other DG units, in order to achieve processed through the compensators (δf ) are sent to all the DG
global controllability of the MicroGrid, secondary control is units in order to restore the frequency of MG.
often used. Conventional centralized secondary control loop is
implemented in MGCC [2]. Fig. 2 shows MG secondary B. Voltage control
control architecture consists of a number of DG units locally The voltage also can be controlled by using similar
controlled by a primary control and a secondary control, which procedure as the frequency secondary control [1]. When the
measures from a remote sensing block a number of parameters voltage in the MG is out from a certain range of nominal rms
to be sent back to the controller by means of a low bandwidth values, a slow PI control that compensates the voltage
communication system. Hence, those variables are compared amplitude in the MG, pass the error through a dead band, and
with the references in order to be compensated by the send the voltage information by using low bandwidth
secondary control, which will send the output signal through communications to each DG unit. Thus, it can be implemented
the communications channel to each DG unit primary control. together with the frequency restoration control loop at the
The advantage of this architecture is that the communication MGCC. The voltage restoration control loop can be expressed
system is not too busy, since only unidirectional messages are as follows:
sent in only one direction (from the remote sensing platform to
the MGCC and from the MGCC to each DG unit). The
 E  kPE  EMG

 EMG   kiE   EMG

 EMG  dt (2)
drawback is that the MGCC is not highly reliable since a failure
of this controller is enough to stop the secondary control action.
being kPE and kiE the PI controller parameters of the voltage
A. Frequency control secondary control. The control signal ( ) is sent to the
primary control level of each DG in order to remove the steady
Traditionally, secondary controllers for large power systems
state errors produced by droop control.
are based on frequency restoration, since the frequency of the
generator-dominated grids is highly dependent on the active
power. This fact is an advantage since frequency is a control
variable that provides information related to the
consumption/generation balance of the grid. This central
controller, named Load Frequency Control (LFC) in Europe or
Automatic Generation Control (AGC) in USA, is based on a
slow PI control with a dead band that restores the frequency of
the grid when the error is higher than a certain value, e.g. +/-50
mHz in the north of Europe.
Similar concept has been implemented in MGCC in order to
restore the frequency of P–f droop controlled MG [4]. The
frequency restoration compensator can be derived as follows.

 f  kPf  f MG

 f MG   kif   f MG

 f MG  dt (1)
Fig. 2. Centralized secondary control.
IEEE TRANSACTIONS ON POWER ELECTRONIC 4

Central Secondary Control

Frequency Control (Eq. 1)

Voltage Control (Eq. 2)

f MG EMG
   
 
f MG EMG
 ki f   ki E 
 kP f    kP E  
 s   s 
f E

Communication Link

f E Droop Control
 E Q
Voltage Reference Power
Primary Control

Generator f P Calculation

E sin(t ) 

vref
Current

Voltage  Virtual
Control Loop  Control Loop  Impedance Loop

PWM

iL vo io
L Lo
DC Link

Vdc
DGK

MicroGrid bus

Fig. 3. Scheme of the central secondary control for a DG unit in a MG.

process and among other management functionalities of MG.
This approach can be also extended to more resistive MGs by The initial idea is to implement primary and secondary
using P–V droops in the primary control, and restoring the controllers together as a local controller. Fig. 4 shows the
voltage of the MG by sending the voltage correction diagram of a fully distributed control system. Primary and
information to adjust the voltage reference. Thus, voltage and secondary controls are implemented in each DG unit. The
frequency restoration controllers can be used in any R/X secondary control is placed between the communication system
condition by means of the park transformation in the primary and the primary control. Frequency control, voltage control,
control. Consequently, the secondary control is transparent to and reactive power sharing will also be reviewed by using this
the R/X nature of the power lines, as opposed to the primary control approach. However, this control strategy can be used to
control. share active power in high R/X MGs as well.
Fig. 3 depicts details of centralized secondary control In this case, secondary control in each DG collects all the
structure for an individual DG unit (DG k) in an islanded MG measurements (frequency, voltage amplitude, and reactive
based on equations (1) and (2). As seen, The frequency and power) of other DG units by using the communication system,
voltage levels in the MG are measured and compared to the average them and produce appropriate control signal to send to
their references, then errors processed through the the primary level removing the steady state errors.
compensators are sent to primary control level of all DG units Fig. 5 illustrates details of the proposed distributed secondary
in order to restore the deviations in the MG. control for an individual DG (DGk) in a MG.

A. Frequency control
IV. PROPOSED DISTRIBUTED SECONDARY CONTROL
Taking the idea from large electrical power systems, in order
The problem of using the MGCC for implementing
to compensate the frequency deviation produced by the local
secondary control is that a failure can result in a bad function of
the whole system. In order to avoid a single centralized P- droop controllers, secondary frequency controllers have
been proposed [26]. However, the approach needs
controller, a distributed control system approach is proposed in
communications in order to avoid instability in the MG system
this paper. However, even with this new control strategy there
is need of MGCC for coordination of units during black start caused probably by different stories of each local inverter.
IEEE TRANSACTIONS ON POWER ELECTRONIC 5

, N is the number of packages (frequency

measurements) arrived through communication system and n is
number of DG units.
Fig. 6 shows how secondary control removes frequency and
voltage deviation caused by primary level in the MG units. In
the Fig. 6 (a), behavior of primary and secondary control for
two DGs with different droop coefficient has been depicted.
This figure demonstrates that secondary control just shifts up
the primary response so that frequency reaches to the nominal
value, even for the DGs with different power rates. It is worth
noting that power change requirement for the proposed DSC
using the average method depends on the power rates of the
MG units.
Fig. 4. Networked controlled MG system.
In order to analyze the system and to adjust the parameters of
DSC for frequency restoration, a small signal model has been
developed for low R/X MGs [1], [30], according to (3) and P-
droop control law.

DG  DG
k

k

 GP ( s)( PDG  PDG )
k k
(4)

The active power of DGk in a low R/X islanded MG can be

presented as follows [30]

EDGk Ecom cos( DGk  com )

PDGk  (5)
Xk
(a) where is voltage at the point of common coupling (PCC),
is the phase between DGk and PCC, and is inductance
between DGk and PCC, respectively. The small signal dynamic
of P- droop control can be obtained by linearizing equations
(4) and (5) at an operating point and :

DGk  DG

k

 GP ( s)(PDGk
 PDGk ) (6)

PDGk ( s)  G   DGk ( s) (7)

(b)
where G 
Ek 0 Ecom cos( k 0   com )
Fig. 6. Secondary control response vs primary control response. (a) frequency Xk
restoration (b) Voltage amplitude restoration.
The block diagram of small signal model for frequency
In the proposed secondary control strategy, each DG measures control is shown in Fig. 7, which includes droop control model
the frequency level in every sample time, sends it to others, and distributed secondary control model. For droop control
averages the frequency measured by other DGs, and then model a low pass filter with cutting frequency of 0.2 Hz has
restores the frequency internally as been considered for power calculation ( ) [30]. The
secondary control has been modeled by means of a simplified
 f DG  kPf  f MG
k

 f DG   kif   f MG
k

 f DG  dt k
phase locked loop (PLL) first-order transfer function ( )
N
used to extract the frequency of the DG [13], a proportional
f
(3) gain ( ) to make frequency average with frequency
DGi
f DGk  i 1 measurements of other DGs ( ), and a PI controller
N ( ).
being kPf and kif the PI controller parameters, is the MG
The characteristic equation can be obtained from Fig.7 as
frequency reference, ̅ is the frequency average for all DG follows
units and is the control signal produced by the secondary
1
control of DGk in every sample time. Here, ,  f  1  GLPF ( s)  GP ( s)   G  GPLL ( s)  ka  G f sec ( s) (8)
s
IEEE TRANSACTIONS ON POWER ELECTRONIC 6

Communication Link
f DG1 , f DG2 ,..., f DGn EDG1 , EDG2 ,..., EDGn QDG1 , QDG2 ,..., QDGn

N N N

Frequency Control (Eq. 3)

E Q

Distributed Secondary Control

Reactive Power sharing (Eq. 14)

f DGi

Voltage Control (Eq. 13)

DGi DGi
i 1 i 1 i 1

1 1 1
N N N
  
  
 
f MG EMG QDGk
 ki f   ki E   ki Q 
 kP f    kP E    kPQ  
 s   s   s 
 f DG k
 EDG  QDG k
k



Droop Control
 E Q
Voltage Reference 
Power
Primary Control

Generator
 f P Calculation
E sin( t ) 

vref
Current

Voltage  Virtual
Control Loop  Control Loop  Impedance Loop

PWM

iL vo io
L Lo
DC Link

Vdc
DGK

MicroGrid bus

Fig. 5. Scheme of the proposed distributed secondary control for a DG unit in a MG.

where is a parameter to obtain the average of error, and tries to compensate the voltage deviation caused by
the Q–E droop. The advantage of this method in front of the
frequency. Other transfer functions can be express as
conventional one is that the remote sensing used by the
1 secondary control is not necessary, so that just each DG
GLPF ( s )  (9)
1 p s terminal voltage, which can be substantially different one from
the other, is required. In this case, the voltage restoration is
1
GPLL ( s )  (10) obtained as follows:
1 s
k s  kiP
GP ( s )  pP (11)  EDG  kPE  EMG

 EDG   kiE   EMG

 EDG  dt
s k k k

k s  kif N

G f sec ( s )  pf (12) E DGi

(13)
s EDGk  i 1

being kiP the droop coefficients, while kpP can be considered as a N

virtual inertia of the system. By analyzing eigenvalues obtained where is the restoration voltage of DG k is produced by
from (8), we can adjust properly the control parameters of using the PI control of the error between voltage reference of
droop and secondary control [25]. MG ( ) and voltage average of DG units ( ̅ ) in every
sample time.
B. Voltage control According to the proposed average method, secondary control
Similar approach can be used as in the distributed frequency is able to remove voltage deviations caused by primary control
control one, in which each inverter will measure the voltage level in every DG unit as shown in Fig. 6 (b).
IEEE TRANSACTIONS ON POWER ELECTRONIC 7

Secondary Control
GPLL ( s )

 
MG
DG

k
Droop Control

avg   sec  DG PDGk

ka  G f sec ( s ) k 1
G
 
s


GP ( s ) GLPF ( s )


P DGk

Fig. 7. Small signal model of distributed frequency control for a DG unit in a (a)
low R/X islanded MGs.
E

C. Line impedance independent power equalization

It is well-known that in a low R/X MicroGrid the reactive E1

power is difficult to be accurately shared, and the same effect E2

occurs when trying to share active power in high R/X

Q
MicroGrids. The reason is that as opposed to the frequency, the
voltage is not common in the whole MG as well as the Q1 Q2
impedance between the DG units and common point is not the (b)
same. Therefore, by using the voltage as a variable is hard to Fig. 8. Q-E droop control problem in reactive power sharing.
control Q flow (or P in case of resistive line MG). As a result,
reactive power is not precisely controlled by using the E-Q all DG units which act as a reactive power reference, and
droop control. Fig. 7 demonstrates this concept. In Fig. 8(a) a is the control signal produced by the secondary control
simple example has been displayed which consist of two units.
in every sample time, to share the reactive power between the
As seen voltage and phase of DG units as well as impedance
DG units. This way, reactive power sharing can be obtained
between DGs can be different, so that Q cannot be shared
independently from voltage sensing mismatches or line
between DG units. Fig. 8(b) depicts that by using E-Q droop,
impedances in the MG.
reactive power is not perfectly shared because voltage is not
It is noteworthy that, the outputs of secondary control must
common in DGs.
be limited, as shown in Fig. 5, in order to not exceed the
As aforementioned, several methods have been reported to
maximum allowed frequency and amplitude deviations as well
improve the reactive power sharing by using only primary
as maximum reactive power that each unit can inject or absorb.
control loop. In all those techniques reactive power sharing
Similar small signal model as in the frequency control one
cannot be precisely achieved since the voltage is a local
can be obtained for voltage control and reactive power sharing
variable. Moreover, Tuladhar et Al. have proposed the use of a
by using equations (13), (14) and Q-E droop control law.
small ripple between converters in order to compensate the
errors due to the different voltage drops along the electrical  
EDGk  EDG  GQ ( s)(QDG  QDGk ) (15)
network of a MG [18]. However this method is difficult to be k k

applied with MGs that contains more than two DG units.

Alternatively, a possible solution is to implement a The reactive power of DGk in a low R/X islanded MG can be
secondary control for power sharing locally, so that each DG presented as follows [30]
unit sends the measured Q (or P in high X/R MicroGrids) to the
other DG units in order to be averaged. This way, as the
2
EDG  EDGk Ecom cos( DGk   com )
QDGk  k
(16)
information is common, all of them will have the same Xk
reference. Therefore, the reactive power sharing by the
By linearizing equations (15) and (16) at an operating point
secondary control can be expressed as
, and the small signal dynamic of Q-E droop
control can be obtained.
 QDG  kPQ QDG  QDG   kiQ  QDG  QDG  dt
k k k k k

 
N
EDGk  EDG  GQ ( s)(QDG  QDGk ) (17)
Q DGi
(14) k k

QDGk  i 1
N QDGk ( s)  H  EDGk ( s)  F  Ecom ( s) (18)

being kPQ as the proportional term, kiQ is the integral term,

is reactive power of DGk, ̅ is average of reactive power for
IEEE TRANSACTIONS ON POWER ELECTRONIC 8

V. EXPERIMENTAL RESULTS AND DISCUSSION

Secondary Control Droop Control Ecom
 
An experimental MG setup as shown in Fig. 10 was used to
EMG E DGk
F test the performance of the proposed approach, consisted on
Eavg  
 Esec   EDGk   QDGk two DG inverters forming as an islanded MG. Fig. 11 shows an
ka GE sec ( s ) H
  experimental setup with the two Danfoss 2.2 kW inverters, the
dSPASE1103 control board, LCL filters, and measurement

GQ ( s ) GLPF ( s ) LEM sensors. A diode rectifier is used as nonlinear load, loaded

Q 
DGk
by a capacitor, and a 200 ohms linear load. The switching
(a) frequency was 10 kHz. The electrical setup and control system
parameters are listed in Table I. All the parameters are the same
Secondary Control Droop Control for both DG units. All parameters have been adjusted based on
Ecom
QDGk

EDGk
the developed model. The secondary control parameters have
F been selected so that its response at least six times is slower
Qavg  
 Qsec   EDG   QDGk

ka
Qref
GQ sec ( s ) k
H than primary control [25].

Inverter 1

GQ ( s ) GLPF ( s ) Nonlinear Load
 DC

QDG Power C NL
k LCL Filter
Supply
(b) 650 V
L R NL
Fig. 9. Small signal model of distributed control for a DG unit in a low R/X NL

islanded MicroGrids a) voltage control b) reactive power sharing. iL1 v1 io1

PC-Simulink
EDGk cos( k 0   com ) RTW & dSPACE
where F   , Control Desk
Xk
iL2 v2 io2
2 Ek 0  Ecom cos( k 0   com )
H 
Xk
DC
Power LCL Filter
Supply
Taking in to account a low pass filter to reactive power 650 V

calculation, block diagram of Q-E droop control for an

Inverter 2
individual DG unit in a low R/X MicroGrid is shown in Fig. 9.
Fig. 10. Schematic of Experimental setup.
The small signal model of the secondary control for voltage
restoration and reactive power sharing has been derived by
using equations (13) and (14), and has been depicted in Figs.
9(a) and 9(b) respectively. The characteristic equations for
voltage control and Q sharing is presented as (19) and (20)

 E  1  (GLPF ( s)  GQ ( s)  H )  (ka  GE sec ( s)) (19)

Q  1  (GLPF ( s)  GQ ( s)  H )  (ka  H  GQ sec ( s)) (20)

where transfer functions can be express as

GQ ( s)  k pQ (21)
k pE s  kiE
GE sec ( s )  (22)
s
k PQ s  kiQ
GQ sec ( s )  (23)
s
being the droop coefficient, and are
transfer function of PI controller for voltage restoration and Q
sharing. These models allow us to set the control parameters of
secondary control properly.

Fig. 11. Experimental setup.

IEEE TRANSACTIONS ON POWER ELECTRONIC 9

TABLE I
ELECTRICAL SETUP AND CONTROL SYSTEM PARAMETERS
Type Parameters Value
Symbol Quantity
Vdc DC Voltage 650 V
VMG MG Voltage 311 V
F MG Frequency 50 Hz
Electrical setup

C Filter Capacitance 25 μF
L Filter Inductance 1.8 mH
Lo Output Impedance 1.8 mH
RL Resistance Load 200 Ω /400Ω
LNL Nonlinear load inductance mH
RNL Nonlinear load resistance Ω (a)
CNL Nonlinear load capacitance μF
kpI Current proportional term 0.35
Loops
Inner

kiI Current integral term 200

kpV Voltage proportional term 0.35
kiV Voltage integral term 400
kpP Active power droop coefficient 0.00001Ws/rd
kiP Active power droop integral term 0.0008 Ws/rd
Control
Droop

kpQ Reactive power droop coefficient 0.16 VAr/V

Rv Virtual Resistance 1Ω
Lv Virtual Inductance 4 mH
kpf Frequency proportional term 0.001
Secondary Control

kif Frequency Integral term 4 s-1

kpE Amplitude proportional term 0.001
kiE Amplitude Integral term 0.6 s-1 (b)
KPQ Reactive power proportional term 0.0001 VAr/V
kiQ Reactive power integral term 0.3 VAr/Vs
PLL time constant 50 ms

Four different sections have been considered to present the

experimental results. In the first section, procedure of black
start for the MG setup is illustrated. Then, performance of the
new secondary control strategy in restoring frequency and
voltage variations as well as reactive power sharing for
different scenarios is depicted in the subsection B. In the (c)
subsections C and D, the effects of communication latency
delay and data drop-out on the proposed secondary control is
investigated and the results are compared with the conventional
secondary control. In this comparison, all the electrical and
control parameters are the same for both distributed and central
controllers as listed in Table I.

A. Black Start Process for the Proposed DSC

If a blackout occurs in a MG, a sequence of actions and
conditions must be checked during the restoration procedure (d)
which called black start process. Conventionally, the MG black
start will be performed centrally by the MGCC based on the Fig. 12. Black start process for the islanded MG setup based on the proposed
DSC a) frequency b) voltage amplitude c) active power d) reactive power.
information stored in a database about the last MG load
scenario. This central controller detects the occurrence of a
Fig. 12 shows the black start process for the islanded MG
blackout and decides when to trigger the MG black start
setup. As can be observed in this figure, DG units 1 and 2 start
procedure. Local controllers and the communication
to act at t=5s and t=10s respectively while primary control
infrastructure are important for the success of the restoration
(inner loops and droop control loop) is running. DG1 is in no
scheme in the MG. The main steps to be considered include
load operation at the time, while DG2 is connected to 400 Ω
building the islanded MG, connecting distributed generations
load feeding around 700 W and 50 VAr to the line impedance.
(DGs) which feed their own protected loads, controlling
A large amount of frequency deviation is seen as a result of load
voltage and frequency, synchronizing DG units inside islanded
connection to the DG2. After activating synchronization
MG, connecting controllable loads and MG synchronization
process (t=20s), DG units are connected at t=25s and then they
with the LV network [31].
works as an islanded MG. As seen, active power is shared after
IEEE TRANSACTIONS ON POWER ELECTRONIC 10

this point, however primary control is not able to share reactive No load Adding Activating Frequent Load Disconection of
power between DG units. Then, a load was connected to the Operation Loads DSC switching DG1

built islanded MG at t=35s, which produce more frequency and

voltage deviation. Finally, DSC is activated at t=40s, which
remove deviations and shares reactive power between two
DGs.

B. Frequency/Voltage Restoration and Q Sharing

The performance of DSC applied to a MG has been depicted
in Fig. 13. Fig 13(a) and Fig 13(b) showing how the new
secondary control strategy restores frequency and voltage (a)
deviation of the DGs. Frequency and voltage deviations are
seen at t=3s and t=5s when loads suddenly are connected to the
MG. At t=10s, the restoration process starts to act by activating
the DSC for both DG units at the same time. It can be seen that
frequency and voltage values are slowly and successfully
regulated inside the islanded MG, removing the static
deviations produced by the droop control. Frequent load
changes has been considered at t=20s (from 200 to 400 )
and t=27s (from 400 to 200 ) respectively. As seen, DSC
restores frequency and voltage amplitude properly after
changing the load. In the last scenario impact of disconnection (b)
of one DG on the whole system has been investigated. At t=35
DG1 is disconnected from the MG setup, however DSC is still
active for that DG as well. As seen in the results, DSC restores
voltage and frequency successfully even after disconnection of
a unit from the MG. Results show restoration process of
frequency and amplitude for DG1 as result of its own local
secondary control effort.
Fig 13(c) shows active power changes in the DGs for each
scenario. This figure shows that active power can be shared
sufficiently between DGs even before activating the DSC by
(c)
means of droop control. These results illustrates that the P-f
droop control is sufficient to share the active power accurately
since the frequency is a global variable in a MG. Notice that
there is a small increase in active power to restore the frequency
deviation when secondary control is activated.
In Fig. 13(d), reactive power sharing has been illustrated.
This figure demonstrates the effectiveness of the proposed
secondary control method when reactive power is shared. As
seen, while there is a big difference between reactive power of
DGs as a result of the droop control, the DSC is able to share
properly the reactive power between the DGs. The proposed
distributed secondary control is able to keep the reactive power (d)
shared between DG units when the load changes frequently as
Fig. 13. Performance of DSC in a) frequency restoration b) voltage
well. After disconnection of DG1 from the MG system in the amplitude restoration c) active power sharing d) reactive power sharing.
last scenario, DG2 feeds the entire load by injecting double DG1 (blue), DG2 (red).
active power.
In order to evaluate the effectiveness of the proposed DSC It can be seen that even with different power rate, the DSC
for islanded MGs which have units with different power rates, with the proposed averaged method is still able to regulate the
another experiment was done when power rate of unit 2 was system frequency successfully. This figure verifies the concept
double of unit 1. Fig. 14 illustrates frequency response of the of Fig. 6 that primary control determines the power rate of MG
system to a frequent step load changes as well as corresponding units, and secondary control is responsible for recovering the
active power of the units. deviations of the units. It is worth to mention that restoration
process requires different amount of power according to the
power rate of the units.
IEEE TRANSACTIONS ON POWER ELECTRONIC 11

different amount of pocket losses, 50% and 95%, considering

100 ms communication delay. It can be seen that both
controllers has an acceptable performance in restoring
frequency and voltage deviation for 50% of data drop-out.
When data drop-out is up to 95%, the central controller is not
able to control the system and system goes to instability after a
while. However, the proposed distributed controller is still
stable and restores deviations properly.

(a) VI. CONCLUSION

This paper has introduced a distributed control strategy for
droop controlled MGs. In this method, a decentralized
secondary control encompasses every DG unit local controller
and the communication system. Thus producing an appropriate
control signal to be locally sent to the local primary controller.
In this sense, the failure of a DG unit will fail down only that
individual unit and other DGs can work independent. Thus,
adding more DG units is easy, making the system expandable.
However, still having a MGCC is mandatory to achieve some
other purposes like coordination of the MG units in black start
(b)
process or energy management.
Fig. 14. Performance of the DSC for an islanded MG consists of two DGs with The concept is evaluated based on the system performance in
different power rates. a) Frequency b) Active power. DG1 (blue), DG2 (red). a laboratory case study with the goal of regulating voltage and
frequency, and at the same time properly sharing reactive
power between DG units. Furthermore, the impact of
C. Impact of Communication Latency
communication system delay as well as data drop-out over the
Communication has a predominant role in providing the MG has been compared between the proposed decentralized
infrastructure that enables data to be exchange among the secondary control system and the conventional centralized one.
different elements of the MG. This importance increases when The results experimental showed that the proposed control
DSC is used for the secondary level of the MGs. strategy has a good performance in removing frequency and
In this section, the impact of communication latency on the voltage steady state errors and can share reactive power
proposed control approach is presented, and then compared between DG units perfectly. Even though the proposed
with those in the conventional centralized approach. secondary control needs more information interchange
Performance of the distributed secondary control has been capability, however, it shown higher robustness in front large
compared with the central one for three amounts of fixed communication latency delays and date drop-out.
communication latency, 200ms, 1sec. and 2sec. For sake of
simplicity, only frequency and voltage responses are depicted.
Table II illustrates the effects of the communication delay on
the control strategies performance, when they remove
frequency and voltage deviations.
As can be seen, both controllers have good performance for
the time delay of 200ms. However, the central one is not able to
restore the frequency and voltage well in the MG when
communication delay is up to 1 second. For a communication
delay of 2 second, as presented in Table II, the central controller
cannot make the system stable, becoming unstable after a
while. However, the proposed control strategy is able to be
stable with a delay of 4 second.

D. Effect of Data Drop-Out

In the real communication system, there may exist data
drop-out or pocket losses which can affect the system output
performance. The performance of proposed secondary control
in the presence of data drop-out is illustrated in Table III,
comparing to the central one. Results have been shown for
IEEE TRANSACTIONS ON POWER ELECTRONIC 12

TABLE II
PERFORMANCE OF DISTRIBUTED SECONDARY CONTROL CONSIDERING COMMUNICATION LATENCY, WHEN COMPARED WITH THE CENTRAL SECONDARY CONTROL
Time
Central Secondary Control Distributed Secondary Control
Delay
Frequency Restoration
200ms

Voltage Restoration
Frequency Restoration
1sec.

Voltage Restoration
Frequency Restoration
2sec.

Voltage Restoration
IEEE TRANSACTIONS ON POWER ELECTRONIC 13

TABLE III
PERFORMANCE OF DISTRIBUTED SECONDARY CONTROL CONSIDERING DATA DROP-OUT, WHEN COMPARED WITH THE CENTRAL SECONDARY CONTROL
Data
Central Secondary Control Distributed Secondary Control
Drop-out

Frequency Restoration
50 %

Voltage Restoration
Frequency Restoration
95 %

Voltage Restoration

[6] F. Katiraei, M. R. Iravani, and P. W. Lehn, “Microgrid autonomous

REFERENCES operation during and subsequent to islanding process,” IEEE Trans.
Power Del., vol. 20, pp. 248–257, Jan. 2005.
[1] J. M. Guerrero, J. C. Vásquez, J. Matas, M. Castilla, L. G. D. Vicuña, and
[7] S. Anand, B. G. Fernandes, J. M. Guerrero, “Distributed Control to
M. Castilla, “Hierarchical Control of Droop-Controlled AC and DC
Ensure Proportional Load Sharing and Improve Voltage Regulation in
Microgrids—A General Approach Toward Standardization, ” IEEE
Low-Voltage DC Microgrids,” IEEE Trans. Power Electron., vol.28,
Trans. Ind. Electron., vol.58, no.1, pp.158-172, Jan. 2011.
no.4, pp.1900-1913, April 2013.
[2] J. A. P. Lopes, C. L. Moreira, and A. G. Madureira, “Defining control
[8] H. Nikkhajoei and R. H. Lasseter, “Distributed generation interface to the
strategies for microgrids islanded operation,” IEEE Trans. Power Syst.,
CERTS microgrid,” IEEE Trans. Power Del., vol. 24, pp 1598–1608, Jul.
vol. 21, pp. 916–924, May 2006.
2009.
[3] M. C. Chandorkar, D. M. Divan, and R. Adapa, “Control of parallel
[9] C. Yuen, A. Oudalov, and A. Timbus, “The provision of frequency
connected inverters in standalone AC supply systems,” IEEE Trans. Ind.
control reserves from multiple microgrids,” IEEE Trans. Ind. Electron.,
Appl., vol. 29, pp. 136–143, Jan./Feb. 1993.
vol. 58, pp. 173–183, Jan. 2011.
[4] Y. A. R. I. Mohamed and A. A. Radwan, “Hierarchical control system for
[10] A. Mehrizi-Sani and R. Iravani, “Potential-function based control of a
robust microgrid operation and seamless mode transfer in active
microgrid in islanded and grid-connected models,” IEEE Trans. Power
distribution systems,” IEEE Trans. Smart Grid, vol. 2, pp. 352–362, Jun.
Syst., vol. 25, pp. 1883–1891, Nov. 2010.
2011.
[11] A. Madureira, C. Moreira, and J. Peças Lopes, “Secondary
[5] K. Jaehong, J.M. Guerrero, P. Rodriguez, R. Teodorescu, N. Kwanghee,
load-frequency control for microGrids in islanded operation,” in Proc.
“Mode Adaptive Droop Control With Virtual Output Impedances for an
ICREPQ, Frankfurt, Germany, 2005, pp. 1–4.
Inverter-Based Flexible AC Microgrid,” IEEE Trans. Power Electron.,,
[12] J. M. Guerrero, P. Loh, M. Chandorkar, T. Lee, “Advanced Control
vol.26, no.3, pp.689-701, March 2011.
Architectures for Intelligent MicroGrids – Part I: Decentralized and
Hierarchical Control,” IEEE Trans. Ind. Electron., 2012, Early access.
IEEE TRANSACTIONS ON POWER ELECTRONIC 14

[13] J. Vasquez, J. M. Guerrero, M. Savaghebi, J. Eloy-Garcia, R. Teodorescu, Qobad Shafiee received the B.S. degree from Razi
“Modeling, Analysis, and Design of Stationary Reference Frame Droop University, Iran, in 2004 and the M.S. degree from
Controlled Parallel Three-Phase Voltage Source Inverters,” IEEE Trans. Iran University of Science and Technology (IUST),
Ind. Electron., 2012, Early access. in 2007, both in electrical engineering. He worked
[14] B. H. Bakken, O. S. Grande, “Automatic generation control in a with department of electrical and computer
deregulated power system,” IEEE Trans. Power Syst., vol.13, no.4, engineering in University of Kurdistan from 2007 to
pp.1401-1406, Nov 1998. 2011, where he has been teaching some electrical
[15] H. Bevrani, Robust Power System Frequency Control, Springer, New engineering courses. He is now perusing the Ph.D.
York, USA, 2009. degree in the department of energy technology at
Aalborg University. His main research interests
[16] M. Savaghebi, A. Jalilian, J. C. Vasquez, J. M. Guerrero, “Secondary
include hierarchical control, networked control
Control Scheme for Voltage Unbalance Compensation in an Islanded
systems and power quality in MicroGrids.
Droop-Controlled Microgrid,” IEEE Trans. Smart Grid, vol.3, no.99,
pp.1-11, 2011.
[17] M. Savaghebi, A. Jalilian, J. C. Vasquez, and Josep M. Guerrero, Josep M. Guerrero (S’01-M’04-SM’08) received
“Secondary Control for Voltage Quality Enhancement in Microgrids”, the B.S. degree in telecommunications
IEEE Transactions on Smart Grid, vol. 3, no. 4, pp. 1893-1902, Dec. engineering, the M.S. degree in electronics
2012. engineering, and the Ph.D. degree in power
[18] A. Tuladhar, Jin Hua, T. Unger, K. Mauch, “Control of parallel inverters electronics from the Technical University of
in distributed AC power systems with consideration of line impedance Catalonia, Barcelona, in 1997, 2000 and 2003,
effect,” IEEE Trans. Ind. Appl., vol.36, no.1, pp.131-138, Jan/Feb 2000. respectively. He was an Associate Professor with
the Department of Automatic Control Systems and
[19] A. Micallef, M. Apap, C. Spiteri-Staines, J. M. Guerrero “Secondary
Computer Engineering, Technical University of
Control for Reactive Power Sharing in Droop-Controlled Islanded
Catalonia, teaching courses on digital signal
MicroGrids” IEEE ISIE, 2012. processing, field-programmable gate arrays,
[20] Y. W. Li and C. N. Kao , “An Accurate Power Control Strategy for microprocessors, and control of renewable energy.
Power-Electronics-Interfaced Distributed Generation Units Operating in In 2004, he was responsible for the Renewable Energy Laboratory, Escola
a Low-Voltage Multibus Microgrid, ” IEEE Trans. Power Electron., Industrial de Barcelona. Since 2011, he has been a Full Professor with the
vol.24, no.12, pp.2977-2988, Dec. 2009. Department of Energy Technology, Aalborg University, Aalborg East,
[21] C. K. Sao, P. W. Lehn, “Autonomous load sharing of voltage source Denmark, where he is responsible for the microgrid research program. From
converters, ” IEEE Trans. Power Del., vol.20, no.2, pp. 1009- 1016, 2012 he is also a guest Professor at the Chinese Academy of Science and the
April 2005. Nanjing University of Aeronautics and Astronautics. His research interests is
[22] A. Haddadi,; A. Shojaei, B. Boulet, “Enabling high droop gain for oriented to different microgrid aspects, including power electronics, distributed
improvement of reactive power sharing accuracy in an energy-storage systems, hierarchical and cooperative control, energy
electronically-interfaced autonomous microgrid, ” IEEE/ECCE, management systems, and optimization of microgrids and islanded minigrids.
pp.673-679, 17-22 Sept. 2011 Prof. Guerrero is an Associate Editor for the IEEE TRANSACTIONS ON
[23] S. K. Mazumder, M. Tahir, K. Acharya, “Pseudo-decentralized POWER ELECTRONICS, the IEEE TRANSACTIONS ON INDUSTRIAL
control-communication optimization framework for microgrid: A case ELECTRONICS, and the IEEE Industrial Electronics Magazine. He has been
illustration,” T&D. IEEE/PES , pp.1-8, 21-24 April 2008. Guest Editor of the IEEE TRANSACTIONS ON POWER ELECTRONICS
[24] Y. Zhang and H. Ma, “Theoretical and Experimental Investigation of Special Issues: Power Electronics for Wind Energy Conversion and Power
Networked Control for Parallel Operation of Inverters,” IEEE Trans. Ind. Electronics for Microgrids, and the IEEE TRANSACTIONS ON
Electron., vol.59, no.4, pp.1961-1970, April 2012. INDUSTRIAL ELECTRONICS Special Sections: Uninterruptible Power
Supplies systems, Renewable Energy Systems, Distributed Generation and
[25] J. M. Guerrero, J. C. Vasquez, J. Matas, M. Castilla, and L. G. de Vicuna,
Microgrids, and Industrial Applications and Implementation Issues of the
“Control strategy for flexible microgrid based on parallel line-interactive
Kalman Filter. He was the chair of the Renewable Energy Systems Technical
UPS systems,” IEEE Trans. Ind. Electron., vol. 56, no. 3, pp. 726–736, Committee of the IEEE Industrial Electronics Society.
Mar. 2009.
[26] M. C. Chandorkar, D. M. Divan, and R. Adapa, “Control of parallel
connected inverters in standalone ac supply systems,” IEEE Trans. Ind. Juan C. Vasquez (M’12) received the B.S. degree
Appl., vol. 29, no. 1, pp. 136–143, Jan./Feb. 1993. in Electronics Engineering from Autonoma
[27] J. M. Guerrero, J. Matas, L. G. D. Vicuna, M. Castilla, and J. Miret, University of Manizales, Colombia in 2004 where
“Wireless-control strategy for parallel operation of distributed generation he has been teaching courses on digital circuits,
inverters,” IEEE Trans. Ind. Electron., vol. 53, pp. 1461–1470, Oct. 2006. servo systems and flexible manufacturing systems.
In 2009, He received his Ph.D degree from the
[28] J. M. Guerrero, J. Matas, L. G. D. Vicuna, M. Castilla, and J. Miret,
Technical University of Catalonia, Barcelona, Spain
“Decentralized control for parallel operation of distributed generation
in 2009 at the Department of Automatic Control
inverters using resistive output impedance“ IEEE Trans. Ind. Electron.,
Systems and Computer Engineering, where he
vol. 54, pp. 994–1004, Apr. 2007. worked as Post-doc Assistant and also teaching
[29] F. Katiraei and M. R. Iravani, “Power management strategies for a courses based on renewable energy
microgrid with multiple distributed generation units,” IEEE Trans. Power systems. Currently, he is an Assistant Professor at
Syst., vol. 21, pp. 1821–1831, Jan. 2005. Aalborg University in Denmark. His research interests include modeling,
[30] Y. Guan, Y. Wang, Z. Yang, R. Cao, and H. Xu, “Control strategy for simulation, networked control systems and optimization for power
autonomous operation of three-phase inverters dominated microgrid management systems applied to Distributed Generation in AC/DC Microgrids.
under different line impedance,” IEEE/ICEMS, pp. 1-5, 2011.
[31] C. L. Moreira, F. O. Resende, and J. A. P. Lopes, “Using low voltage
microgrids for service restoration,” IEEE Trans. Power Syst., vol. 22, no.
1, pp. 395–403, Feb. 2007.