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

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2017.2740163, IEEE
Transactions on Power Systems
1

Battery Energy Storage System Models for

Microgrid Stability Analysis and Dynamic
Simulation
Mostafa Farrokhabadi, Student Member, IEEE, Sebastian König, Claudio Cañizares, Fellow, IEEE,
Kankar Bhattacharya, Fellow, IEEE, and Thomas Leibfried, Member, IEEE

Abstract—With the increasing importance of battery energy For several interesting technical features, BESS have re-
storage systems (BESS) in microgrids, accurate modeling plays a ceived considerable attention recently, particularly as a solu-
key role in understanding their behaviour. This paper investigates tion to the challenges facing modern active distribution net-
and compares the performance of BESS models with different
depths of detail. Specifically, several models are examined: an works and microgrids. BESS can provide several key ancillary
average model represented by voltage sources; an ideal dc source services, such as load shifting, dynamic local voltage support,
behind a voltage source converter; a back-to-back buck/boost short-term frequency smoothing, grid contingency support,
and bi-directional three phase converter, with all models sharing and reduce the need for fossil-fuel-based generation [4], [5].
the same control system and parameters; and two additional Therefore, BESS are considered a key enabling element of
proposed models where the switches are replaced by depen-
dent sources to help analyze the differences observed in the modern smart grids and microgrids.
performance of the models. All these models are developed in Since many utilities and researchers use simulation software
PSCAD and their performances are simulated and compared packages to model and investigate various issues in microgrids,
considering various issues such as voltage and frequency stability grid components need to be adequately modeled to properly
and total harmonic distortion, in a benchmark test microgrid. It reflect the behaviour and performance of the system. Several
is shown through simulation results and eigenvalue studies that
the proposed models can exhibit different performance, especially components of microgrids have been extensively studied, and a
when the system is heavily loaded, highlighting the need for more variety of models have been developed and reported in the lit-
accurate modeling under certain microgrid conditions. erature. However, since inverter-based BESS are relatively new
Index Terms—Energy storage systems, dynamic simulation, elements of microgrids, fewer studies have been conducted on
microgrids, modeling, stability. their modeling and control. In [6], a generalized mathematical
model of ESS is presented for voltage and angle stability
analysis based on the balanced fundamental-frequency model
I. I NTRODUCTION
of the Voltage Source Converter (VSC) and the dynamics of

M ICROGRIDS are defined as a cluster of interconnected

distributed energy resources (DERs), energy storage
systems (ESS), and loads which can operate in parallel with
the dc link, based on a set of linear differential algebraic
equations. However, the impact of inverter switching on the
system performance is not modeled, and the performance of
the grid or in an islanded mode [1]. Under the smart grid the proposed model under different microgrid conditions such
paradigm, microgrids are considered a critical link in the evo- as heavy and/or unbalanced loading are not studied. Detailed
lution from vertically integrated bulk power systems to smart ESS models for transient analysis in microgrids are presented
decentralized distribution networks [2], with high penetration in [5] and [7]. However, the focus of these papers is on ESS
of renewables, easily scalable structures, and increased reli- applications in microgrids, without considering the impact of
ability levels [3]. However, such active distribution networks ESS modeling on the system dynamic performance. Simplified
present various challenges, such as the integration of plug- models of ESS are presented in [8] and [9], but similar
in electric vehicles and the proliferation of renewable energy to the aforementioned papers, they dwell on applications of
sources (RES), which have to be addressed when significant ESS in power systems rather than the modeling of ESS; the
penetration levels are reached [4]. In this context, BESS can be ESS models consist of only active and reactive power loops,
practically helpful to deal with such challenges, as explained neglecting VSC and dc link circuits, and the ESS dynamic
next. performance.
This work has been partially supported by NSERC-Canada through Dis- Considering the variety of BESS components, such as dc-dc
covery Grants. and dc-ac converters, studies on the impact of level of detail in
M. Farrokhabadi, C. A. Cañizares, and K. Bhattacharya are with the BESS modeling on the stability and dynamic performance of
Department of Electrical and Computer Engineering, University of Wa-
terloo, Waterloo, ON N2L 3G1, Canada (e-mail: [email protected]; microgrids have not yet been reported in the literature. Hence,
[email protected]; [email protected]). considering the importance of BESS in active distribution
S. Koenig and T. Leibfried are with Institute of Electric Energy Systems and networks and microgrids, this paper investigates and compares
High-Voltage Technology (IEH), Karlsruhe Institute of Technology, Karlsruhe,
Germany (e-mail: [email protected]; [email protected]). microgrid dynamic performance using BESS models with dif-
ferent depth of detail. Specifically, several models are studied:
an average model represented by voltage sources [10]; an ideal

0885-8950 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2017.2740163, IEEE
Transactions on Power Systems
2

dc source behind a voltage source converter [11]; and a back-

1
to-back buck/boost and bi-directional three phase converter EB = E0 − K + A−BQ(1−SOC) (1)
[12], with all models sharing the same control systems and SOC
parameters; and two additional proposed models where the where E0 is the battery constant voltage in V, K is the
switches are replaced by dependent sources to help analyzing polarization voltage in V, Q is the battery capacity in Ah,
the differences observed in the performance of the models. and A and B are parameters determining the charge and
The models are developed in PSCAD and their performance discharge characteristics of the battery. The parameters A, B,
is compared considering various variables such as voltages and and K can be tuned to mimic a specific battery type discharge
currents at the point-of-common-coupling, active and reactive characteristic.
power injection, and total harmonic distortion, in the context Note that the time frame of the test scenarios discussed in
of the impact on the the stability of the system. Thus, the main this paper is in the order of a few seconds; hence, since the
objectives and contributions of the paper are the following: BESS designed and simulated in this paper has a rated power
• Develop and study various BESS models for microgrid
of 1 MW and a rated capacity of 1 MWh, which is a typical
simulation and analysis, including a new and efficient power to capacity ratio for BESS in microgrids, the battery
model where the switches are replaced by dependent voltage and SOC relation cannot be observed in the presented
sources, identifying the conditions in which each model simulation results.
can be used. A systematic mechanism is applied to isolate
and evaluate the impact of various parameters. B. Buck/Boost Converter
• For the first time, study the impact of BESS con- The buck/boost converter is in charge of controlling the dc
verter switching on microgrid stability, including high- link capacitor voltage by properly charging and discharging
frequency signals and internal resistances of switches. the battery. A cascaded PI controller is used here to generate
• Study the impact of BESS dc link voltage dynamics the duty cycle for the switches based on the difference of the
on microgrid stability, through eigenvalue studies and dc link voltage and its set-point, as shown in Fig. 2. Note that
dynamic simulations. the second PI controller reacts to changes in the active power
• Study and demonstrate the impact of unbalanced loading of the system first, thus improving the ability of the buck/boost
on microgrids stability. controller to maintain the dc link voltage.
• Compare the performance of the proposed models in a When the dc link voltage is lower than the pre-defined
realistic test microgrid, determining their advantages and set-point, the converter works in the boost mode, discharging
limitations, and their computational efficiency. the battery. When the dc link voltage is higher than the set-
The rest of this paper is organized as follows: Section II de- point, the converter operates in the buck mode, charging the
scribes the various components and parameters used to model battery. The current ripple is bounded by a proper choice of
BESS in detail, and Section III presents three approximated the inductor Lchopf , as follows [14]:
models to approximately represent BESS in dynamic system
studies. Case studies and simulation results are presented in EB EB 1
∆I = ∆t = (2)
Section IV, comparing the performance of BESS models with Lchopf Lchopf 2fs
different depth of detail. Section V analyzes in detail the where fs is the switching frequency.
reasons for the differences observed in the performance of
the various models presented. Finally, the main findings and
C. DC-AC Converter
conclusions of the studies presented in the paper are provided
in Section VI. Figure 8 illustrates how the dc-ac converter connects the
battery and buck/boost converter to the grid through the
II. BESS D ETAILED M ODEL ac filter. The converter control system provides the voltage
magnitude and phase set-points to create sinusoidal reference
The BESS models used in this paper mainly include a
signals for the Pulse Width Modulation (PWM) scheme of the
buck/boost converter, a dc link capacitor, a three-phase bi-
converter.
directional dc-ac converter, an ac filter, and a transformer
As shown in Fig. 8, when the switches are in State 1, the
connecting the system to the microgrid. In this section, these
voltage magnitude and phase set-points are obtained based
models and the corresponding parameters are discussed, and
on the reference voltage and frequency; thus, the BESS in
the control techniques used for each BESS converter are also
this case is the master voltage and frequency controller of
described. Figure 8 depicts in detail the BESS components [5].
the system, which is referred to as a grid-forming control
technique. If the switches are in State 2, the voltage magnitude
A. Battery and phase set-points are obtained based on the reference active
The battery model described here is based on the generic and reactive powers; hence, the battery is operated in constant
model proposed in [13], and is modeled as a controllable ideal P Q mode, injecting/absorbing constant active and reactive
dc source in series with an internal resistance RB . The no-load power, which is referred to as grid-feeding mode. Grid-
voltage of the battery EB is calculated based on the state-of- supporting and grid-following control modes are based on
charge (SOC) of the battery using a nonlinear equation, as these two fundamental control strategies, and are accomplished
follows: by changing the reference voltage, frequency, active power,

0885-8950 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2017.2740163, IEEE
Transactions on Power Systems
3

Battery Buck/Boost DC Link Inverter AC Filter Grid

Battery Energy Storage System Scheme

d2 Ichop Idc
1 3 5 7
L f Rf I ga
a
Lchopf Rchopf I gb
d1 Vdc b
VOoa I gc Vggaa
RB c a

VOob VVggbb
2 b

VOoc Rd
4 6 8 Vggcc
EB c

Cf N
n
PLL
Modulation Technique Modulation Technique
Measurement and Control Scheme

VVref_abc
fref Vref  PLL
ref abc
VVdc_meas
measdc 1 VV
abc ref_dq
ref dq V-F Reference abc
2
dq0
 ref Generator dq0
Pref Vref dc Vref_dq
V ref dq Qref Pref IIg_dq
g dq VVg_dq
g dq

DC Voltage DC Voltage
2 1 Current
Current Reference P Active and
closed-loop closed-loop closed-loop Reactive Power
Generator
control control  PLL  ref control Q Calculation

P
Fig. 1. Schematic of a battery energy storage system components and its controls [5].

Vrateddc In grid-feeding control mode, the injected active and reac-

+
K p  Ki s tive power are calculated first, as follows:
-
Vdc + Vrefd 3

P K p '  Ki ' s p= Vgd Igd − Vgq Igq (3)
+ 2
-
Vdc Idc 3

q= Vgd Igq − Vgq Igd (4)
2
Fig. 2. Buck/boost dc link voltage controller.
To obtain the corresponding fundamental P and Q compo-
Vratedd nents, the instantaneous active p and reactive q powers are
+
K p  Ki s Vrefd passed through low-pass filters. The fundamental active and
- reactive powers are then passed through the current reference
Vgd Vrefq generator block to obtain the current dq-axes reference set-
0
points, as follows:
ref 1  ref 2 Pref Vgd + Qref Vgq
s Irefd = (5)
3 Vg2d + Vg2q
Fig. 3. Grid-forming voltage and phase reference generator.
2 Pref Vgq + Qref Vgd
Irefq = (6)
3 Vg2d + Vg2q
and/or reactive power [15]. Observe that these controls are
based on Park’s dq-axes transformations. These current references are then passed through the current
In the grid-forming mode, the voltage dq-axes reference closed-loop control to obtain the final voltage dq-axes refer-
set-points would be directly used to create the abc-reference ences; feed-forward terms should be used to decouple the two
signals, as shown in Fig. 3. The reference angle is obtained by axes, and should be considered for the difference between the
integrating the reference angular frequency. Observe that a PI voltages after and before the ac filter. Neglecting Rd , a single
controller is used to maintain the point of common coupling line diagram can be used to derive such a relation, as shown
(PCC) voltage at its rated value. in Fig. 4, resulting in the following equations:

0885-8950 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2017.2740163, IEEE
Transactions on Power Systems
4

Lf Rf I ga Inverter AC Filter Grid

Battery Energy Storage System Scheme

Vo Cf Vg
L f Rf I ga
a
I gb
Fig. 4. Single line diagram of ac filter. Vdc b
VOoa I gc Vggaa
c a

VOob VVggbb
 2Lf C f
b
Vgd VOoc Rd Vggcc
c

I refd - Vrefd
K p  Ki s + Cf
+ + -
- + -
PLL

Measurement and Control Scheme

I gd Rf Lf Rf C f Vgq Modulation Technique

VVref_abc
fref Vref  PLL
I gq ref abc
Rf Lf Rf C f Vgd 1 V
abc ref_dq
ref dq
V-F Reference abc
2
I refq - + + dq0
 ref Generator dq0
K p  Ki s + +
+
+ Vrefq VVref_dq
ref dq Qref Pref IIg_dq
g dq VVg_dq
g dq
-
Vgq  Lf C f2
2 1 Current P Active and
Current Reference
closed-loop Reactive Power
Generator
 PLL  ref control Q Calculation
Fig. 5. Current closed-loop control.
Fig. 6. Ideal dc link model of the BESS.

III. ESS A PPROXIMATE M ODELS

Vod = Vgd (1−ω 2 Lf Cf )+Igd Rf −Igq ωLf −Vgq ωRf Cf (7)
In the previous section, the different BESS components and
Voq = Vgq (1−ω 2 Lf Cf )+Igd Rf +Igd ωLf +Vgd ωRf Cf (8) their corresponding controls are discussed in detail. Mdels
The final current closed-loop control block shown in Fig. 5 have been proposed in the literature, in which some compo-
can be obtained from these equations. nents of the detailed BESS model are eliminated to achieve
The output of the current controller are the voltage refer- faster simulation speed and/or reduce control complexity, as
ences Vrefd and Vrefq , which are transformed back to the explained next.
abc-reference frame to obtain the sinusoidal control signals
for PWM scheme of the converter. More information on A. Ideal DC Link Model
alternative converter controls such as grid-supporting and grid- The schematic of this model is shown in Fig. 6, where the
following can be found in [15]. battery model along with the buck-boost converter is replaced
by an ideal dc source, thus fixing the dc link voltage. Hence,
the controls related to the dc link are eliminated. The rest of
D. AC Filter
the system remains the same, i.e. the inverter and the ac filter
The ac filter should be designed properly to limit output and the corresponding control are the same as explained in
current ripple and achieve an acceptable damping rate. The Section II.
ac filter inductor Lf in a dc-ac converter can be determined
based on the maximum acceptable current ripple, the dc link B. Average Model
voltage, and the switching frequency as follows [16]:
The schematic of this model is shown Fig. 7, where the
Vdc /3 entire switching system is replaced by ideal voltage sources,
∆Imax = (9) i.e. the battery, buck-boost converter, and dc-ac inverter are
Lf 4fs
eliminated; this increases the simulation speed significantly.
The ac filter capacitor Cf reactive power injection has The control system for the battery remains the same as in
to be less than 5% of the converter rated power [16]. A Section II; the only change is that abc-reference signals are
series damping resistance is considered to prevent harmonic directly fed to the ideal voltage sources, and the PWM scheme
oscillations, and is assumed to consume 0.2% of the rated controls are eliminated.
power as per:
C. Dependent Source Model (DSM)
s
VI2 VI2

1
Rd = − − (10) To identify and isolate the impact of switches on the
0.002PI 0.002PI (ωCf )2
performance of the system, a model is proposed here in which
Also, a resistance Rf is added in series with the filter the switches are replaced by dependent voltage and current
inductance to represent its parasitic resistance losses. sources, as shown in Fig. 8. The control system remains
the same as in Fig. 1. Two models are developed based on
the value of the dependent voltage and current sources, as
explained next.

0885-8950 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2017.2740163, IEEE
Transactions on Power Systems
5

Ideal Voltage Sources AC Filter Grid A ∆-Y transformer usually connects the the inverter to the
Battery Energy Storage System

Lf Rf I g rest of the grid, thus:

a a

b I gb Vga + Vgb + Vgc = 0

(16)
Scheme

c VOoa a
I gc Vggaa Iga + Igb + Igc = 0
VOob b
VVggbb Based on (15) and (16) the following equation can be obtained:
VOoc Vggcc
c
Van + Vbn + Vcn
VaN = Van + VnN = Van − (17)
3
VVref_abc
ref abc fref Vref PLL
Similarly for the voltage on the other phases. In addition, DS3
Measurement and Control Scheme

in (13) and (14) can replaced by its low-frequency contents,

abc
1 VV
ref_dq
ref dq V-F Reference  PLL i.e., its local average, as follows [17]:
2
dq0
 ref Generator
1 1
abc DS3avg = + ma (18)
dq0 2 2
VVref_dq
ref dq
Qref Pref IIg_dq
g dq VVg_dq
g dq
where ma is the modulating signal used to control DS3 ,
normalized to the peak value of the carrier signal. Thus, from
2 1 Current
Current Reference P Active and (17) and (18), the dependent source in phase a can be defined
closed-loop Reactive Power
Generator as follows:
 PLL  ref control Q Calculation
 
1 1
Fig. 7. Average model of the BESS. VaNavg = + ma Vdc −
2 2
3 1
(19)
2 + 2 (ma + mb + mc ) 1
Vdc = ma Vdc
1) Switching DSM: To emulate the performance of the 3 2
model with switches, the currents through and voltages across And similarly for the dependent voltage sources in the other
the switches are analyzed in each switching state. For the phases.
buck/boost converter, when Switch 1 is on and Switch 2 is Finally, using KCL, the following equation can be obtained:
off in Fig. 1, Vd1 n = Vdc , and the current through Switch 1 is
Ichop . Hence, VB and IB in Fig. 8, can be defined as follows: Idc = Ia + Ib + Ic (20)
Thus, based on (20), (18), and (16), the dependent current
Vd1 n = VB = DS1 Vdc (11) source in phase a can be defined as follows:
IB = DS1 Ichop (12) 1
Iaavg = ma Iga (21)
2
where DS1 is the duty cycle of Switch 1.
For the dc-ac inverter, when Switch 3 is on and Switch 4 And similarly for the current source in the other phases.
is off, Van = Vdc , and the current through Switch 3 is Iga .
Thus, Va and Ia in Fig. 8 can be defined as follows: IV. R ESULTS
To effectively compare the performance of the three dif-
Van = DS3 Vdc (13) ferent modeling approaches, a test system based on the
CIGRE benchmark for medium voltage distribution network
Ia = DS3 Iga (14)
introduced in [18] is implemented in PSCAD/EMTDC, as
where DS3 is the duty cycle of Switch 3. Similarly for the shown in Fig. 9. The test system has a 1.3 MVA diesel-based
other two legs of the inverter. Note that the internal resistances synchronous machine, a 1 MW BESS, and a 1 MW wind
of switches are modelled by Rin . turbine, with the latter being modeled using an average model
2) Average DSM: To eliminate the impact of high- similar to Fig. 7, and operated as a constant power source
frequency switiching, an averaging technique based on the for the short duration of time it is connected to the grid,
fundamental frequency component is used here. Thus, by since the disconnection of the wind turbine is used here as
assuming that no current goes through Cf in Fig. 1 and using a disturbance. The diesel-based synchronous machine and its
KVL, the following equations can be derived: exciter and governor are tuned and validated according the
actual measurements from a commercial grade synchronous
dIga machine in [19]. The BESS design parameters are shown
Van = Rf Iga + Lf + Vga + VnN in Table I. The loads are modeled using a voltage-sensitive
dt
dIgb exponential model with a 1.5 exponent, which is a reasonable
Vbn = Rf Igb + Lf + Vgb + VnN (15) value for typical isolated microgrids [20]. The load demand
dt
dIgc and unbalance levels are different for each test scenario
Vcn = Rf Igc + Lf + Vgc + VnN discussed next.
dt

0885-8950 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2017.2740163, IEEE
Transactions on Power Systems
6

Battery Buck/Boost DC Link Inverter AC Filter Grid

Rin Ichop Idc

Ia
IB Va Rin L f Rf I ga
Ib

+
-
Lchopf Rchopf Rin Vb I gb
Ic

+
Vdc

-
Vc VOoa I gc Vggaa
RB Rin a

+
-
VB + VOob b
VVggbb
- VOoc Rd Vggcc
EB c

n Cf N
Fig. 8. Schematic of a battery energy storage system with switches being replaced by dependent sources.

TABLE I instantaneous voltage of phase a at the PCC for average, ideal

ESS D ESIGN PARAMETERS dc link, and detailed model. Note in Fig. 10(a), before t =
ESS Parameters 11.5s, the battery performance is satisfactory for all the three
Lf Rf Cf Rd Vrateddc modeling techniques. Due to switching in the system, there are
0.166 mH 4.2 mΩ 626.8 µF 84.7 mΩ 750 v some fluctuations in the active and reactive power in the ideal
Cdclink RB Lchopf fs VRM SL−L dc link and detailed models of the BESS. The Total Harmonic
20 mF 0.2 Ω 3.3 mH 3 kHz 460 v Distortion (THD) of the instantaneous voltage is 0.25% for
average model, 3.92% for ideal dc link model, and 3.63% for
the detailed model.
1 12 After t = 11.5s, the system remains stable for the average
and ideal dc link models, whereas it shows unsustained oscil-
ESS Dies. Wind 4.9 km
1.0 km
lations for the detailed model. This is due to the fact that as the
2 13 apparent power demand increases, the charging/discharging
current of the dc link capacitor also increases, and beyond
1.2 km
a certain value, the dc link capacitor voltage ripple becomes
3
3.0 km significant, leading to system instability, as illustrated in
0.6 km
Fig. 11. Note that the chosen dc link capacitor is relatively
4 large for a 1 MW BESS, and the active power is within the
0.6 km 0.5 km range for which the system is designed. Also, observe that
11 14 the voltage magnitudes at the PCC are close to their nominal
5
values in the average and ideal dc link models, as compared
10 8
to the detailed model of the BESS.
2.0 km Figure 12 shows the BESS active power, reactive power,
1.7 km phase a RMS voltage at PCC, and the dc link voltage for the
7
Average and Switching DSMs. Observe that the performance
9 Bus of the Average DSM is exactly the same as the performance
0.2 km
6 of the average model shown in Fig. 10, and the performance
Load
of the Switching DSM is exactly the same as the performance
of the detailed model. These models are used in Section V to
Fig. 9. Modified version of CIGRE benchmark microgrid. isolate and evaluate the impact of high frequency switching
on the stability of the microgrid.
A. Case A: Grid-forming BESS in Balanced Operation
B. Case B: Grid-forming BESS in Unbalanced Operation
In this case, the diesel and wind generators are not con-
nected. The system load is 950 kW of active power and 100 In this case, the diesel and wind generators are not con-
kVar of reactive power, balanced among the three phases; at nected. The system load is 950 kW of active power and 100
t = 0.5s, the load is increased by 100 kVar, and at t = 11.5s, kVar of reactive power, with the load on phase c being twice
the reactive power load is again increased by 100 kVar; note the load on phase a and b.; at t = 0.5s, the load is increased
that the active power load is near the BESS rated power. by 100 kVar.
Figure 10 illustrates the active power, reactive power, and Figure 13 presents the active power, reactive power, and
RMS phase voltages at the PCC. As seen from Fig. 13, the

0885-8950 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2017.2740163, IEEE
Transactions on Power Systems
7

DSM Switch.
1 Average
P(MW)
1

P(MW)
0.9 0.9
0.8 0.8

1 2 4 6 8 10 12Ideal
14DC 16
Link 2 4 6 8 10 12 14 DSM Av.
P(MW)

P(MW)
0.9 0.9
0.8 0.8

1 2 4 6 8 10 12 14Detailed
16 0 2 4 6 8 10 12 14 16
P(MW)

t(s)
0.9
(a)
0.8
0.3 DSM Switch.
0 2 4 6 8 10 12 14 16

Q(MVAR)
t(s) 0.2
0.1
(a)

0.3 0.3 DSM2 Av. 4 6 8 10 12 14

Q(MAVR) Q(MAVR) Q(MVAR)

Average

Q(MVAR)
0.2 0.2

0.1 0.1

0.3 2 DC Link
4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Ideal t(s)
0.2
(b)
0.1
Volt(pu) 1 DSM Switch.
0.3 2
Detailed 4 6 8 10 12 14 16
0.95
0.2
0.1
1 2 4 6 8 10 12 14 DSM Av.
0 2 4 6 8 10 12 14 16
Volt(pu)

t(s) 0.95
(b)
0 2 4 6 8 10 12 14 16
1 Average t(s)
Volt(pu)

(c)
0.95
1000 DSM Switch.
Volt(pu)

1 2 4 6 8 10 12 14 DC16Link
Ideal 800
Volt(pu)

600
0.95
1000 DSM Av.
Volt(pu)

1 2 4 6 8 10 12 14 Detailed
16 800
Volt(pu)

600
0.95
400
0 2 4 6 8 10 12 14 16
t(s)
0 2 4 6 8 10 12 14 16
t(s) (d)

(c) Fig. 12. DSM Case A: (a) Active power; (b) reactive power; (c) phase a
RMS voltage at PCC, and (d) dc link voltage.
Fig. 10. Case A: (a) Active power, (b) reactive power, and (c) phase a RMS
voltage at the PCC.
system remains stable after the disturbance for the average
1000 and ideal dc link BESS models; however, the system shows
unsustained oscillations for the detailed model, since similar
Volt(v)

800
to Case A, the dc link capacitor voltage ripples cause the
600
system instability, showing a similar behaviour as in Fig. 11.
Comparing Case A and Case B for the same loading condition,
400
0 2 4 6 8 10 12 14 16
it is noted that the same disturbance makes the unbalanced
t(s) system unstable, whereas the balanced system remains stable.

Fig. 11. DC link voltage for the detailed model for Case A. The performance of the Average DSM has found to be
exactly the same as the performance of the average model, and

0885-8950 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2017.2740163, IEEE
Transactions on Power Systems
8

1 Average
P(MW)
1

P(MW)
0.9
0.8 Average 0.8

1 2 4 6 8 10 1 Link
Ideal DC 2 3 4 5
P(MW)

P(MW)
0.9
0.8 Ideal DC Link 0.8

1 2 4 6 8 10 1
Detailed 2 3 4 5
P(MW)

P(MW)
0.9
0.8 Detailed 0.8
0 2 4 6 8 10 0 1 2 3 4 5 6
t(s) t(s)

(a) (a)

0.4 Average

Q(MVAR) Q(MVAR) Q(MVAR)

Average
Q(MAVR) Q(MAVR) Q(MVAR)

0.2 0.3
0.1
0 0.2
-0.1
2 4 6 8 Ideal 10
DC Link 0.4 1 2 3 4 5 DC Link
Ideal
0.2 0.3
0.1
0 0.2
-0.1
2 4 6 8 10
Detailed 0.4 1 2 3 4 5 Detailed
0.2 0.3
0.1
0 0.2
-0.1
0 2 4 6 8 10 0 1 2 3 4 5 6
t(s) t(s)

(b) (b)

1.15
VaAverage VbAverage VcAverage
VaAverage VbAverage VcAverage
Volt(pu)

1.1
Volt(pu)

1.05
1.1 1
0.95
0.9
1 0.85
0.8
0.9
1.15
1 VaIdeal DC
2
Link
3Vb 4
Ideal DC Link
VcIdeal
5
DC Link
2 Va 4 Vb6Ideal DC Link
8 10
VcIdeal
Volt(pu)

1.1
Ideal DC Link DC Link
Volt(pu)

1.05
1.1 1
0.95
0.9
1 0.85
0.8
0.9 1 2 Va
3 4 Vb 5 VcDetailed
1.15 Detailed Detialed
2 4 6
VaDetailed 8
VbDetailed 10
Vc
Volt(pu)

1.1
Detailed
Volt(pu)

1.05
1.1 1
0.95
0.9
1 0.85
0.8
0.9 0 1 2 3 4 5 6
2 4 6 8 10 t(s)
t(s)
(c)
(c)
Fig. 14. Case C: (a) Active power, (b) reactive power, and (c) phase RMS
Fig. 13. Case B: (a) Active power, (b) reactive power, and (c) RMS phase voltages at the PCC.
voltages at the PCC.

Figure 14 illustrates the active power, reactive power, and

the same for the performance of the Switching DSM compared
RMS phase voltages at the PCC. Note that the system remains
to the detailed model. Hence, these plots are not shown here.
stable for the average and ideal dc link models, but shows
unsustained oscillations for the detailed model. Similar to Case
C. Case C: Grid-forming BESS with Wind Generator Outage B, the dc link capacitor voltage ripple becomes greater than
In this case, the diesel generator is not connected, and the what the system is designed for, after the wind generator is
wind turbine is generating a constant 200 kW of active power disconnected.
and 200 kVar of reactive power; at t = 1s, the wind generator The performance of the Average DSM has found to be
is disconnected. The system active power load is 1 MW and exactly the same as the performance of the average model, and
reactive power load is 350 kVar, with the load on phase c the same for the performance of the Switching DSM compared
being twice the load on phase a and b.

0885-8950 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2017.2740163, IEEE
Transactions on Power Systems
9

to the detailed model. Hence, these plots are not shown here. 1

P(MW)
0.8
0.6
D. Case D: Grid-supporting BESS With Wind Generator Out- 0.4 Average
0.2
age
1 1 2 12 4 5 6 7 8 9 10 11 12 13 14

P(MW)
In this case, the diesel generator is connected and is the 0.8
master voltage and frequency controller, with the governor and 0.6
0.4 Ideal DC Link
excitation controls, and the BESS is providing 1 MW of active 0.2
power and half of the system reactive power demand. The wind 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
turbine is generating 500 kW of active power and 500 kVar of 1

P(MW)
reactive power, and the load active power is 2.1 MW and the 0.8
0.6
reactive power is 1 MVar, balanced among the three phases. 0.4 Detailed
At t = 1s, the wind generator is tripped, and at t = 7s, the 0.2
load reactive power is increased by 500 kVar. 0 1 2 3 4 5 6 7
t(s)
Figure 15 shows the BESS active power, reactive power,
and RMS voltages at the PCC. The diesel generator active (a)
and reactive power, and the system frequency are illustrated

Q(MVAR) Q(MVAR)
1
in Fig. 17. As seen from these figures, the system is able
0.5
to retain its stability for the average and ideal dc link BESS 0 Average
models after the wind generator is disconnected, i.e between -0.5
t = 1s to t = 7s; however, the system loses its stability for the 1 2 3 4 5 6 7 8 9 10 11 12 13 14
1
detailed BESS model. In fact, since the battery has to increase 0.5
its reactive power generation after the disturbance, it passes 0 Ideal DC Link
a point where the dc link voltage cannot be maintained any -0.5
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
longer; as a result, the battery active power generation also
Q(MVAR)

shows large unsustained oscillations. Consequently, the diesel 1

generator is not able to meet the demand level and the system 0.5
frequency, and voltage collapses. The battery dc link voltage 0 Detailed
for the detailed BESS model is shown in Fig. 16. -0.5
1 2 3 4 5 6 7
After the load reactive power is increased at t = 7s, the t(s)
system continues to remain stable for the average model (b)
of the BESS, but the system loses stability for the ideal
dc link model. This is an important observation, since the 1
Volt(pu)

0.9
only difference between these two models is the presence of 0.8
switches. 0.7 Average
0.6
The response of the Average DSM has found to be exactly
1 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Volt(pu)

the same as the performance of the average model. The 0.9

BESS and diesel engine response of the Switching DSM is 0.8
0.7 Ideal DC Link
shown in Fig. 18 and Fig. 19. Observe that similarly to the 0.6
detailed model, the Switching DSM becomes unstable after the 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
disturbance at t = 1 s. However, its response is not exactly 1
Volt(pu)

the same as the detailed model, due to the different modeling 0.9
0.8
approaches; nevertheless, in both cases the system is instable, 0.7 Detailed
and thus they both lead to the same conclusions. 0.6
1 1 2 3 4 5 6 7
t(s)
V. A NALYSIS
(c)
In the previous section, it was shown that when the system
was pushed to its loading limits, different BESS models Fig. 15. Case D: BESS (a) Active power, (b) reactive power, and (c) RMS
demonstrated significantly different performances. These dif- phase voltages at the PCC.
ferences can be due to three main factors: the dc-link voltage
dynamic, the high frequency switching, and the internal re- A. Small-Signal Analysis
sistance of the switches. To effectively isolate the impact of
each factor, first, an eigenvalue study is performed based on Typically, in conventional power systems, eigenvalue anal-
a signal-processing technique using the dc-link voltage signal. ysis is carried out by linearization of the power system model
Second, the proposed DSM is used to eliminate the higher around an equilibrium point [21]. Such an approach requires
frequencies and investigate the performance of the system. considerable simplifications limited to balanced systems, or
Additionally, the proposed models are compared in terms of the development of detailed linearized models of the system
the simulation computation times. [22], which may not be feasible in practical systems. In this

0885-8950 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2017.2740163, IEEE
Transactions on Power Systems
10

1500
1.2
1

P(MW)
0.8
0.6
Volt(v)

1000 0.4
0.2 Average
500 1.2 1 2 3 4 5 6 7 8 9 10 11 12 13 14
1

P(MW)
0
0.8
0 1 2 3 4 5 6 7 0.6
t(s) 0.4 Ideal DC Link
0.2
Fig. 16. DC link voltage for the detailed model - Case D. 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
1.2
1

P(MW)
0.8
paper, a modal estimation approach is utilized to identify the 0.6
0.4 Detailed
dominant eigenvalue of the system [23], [24]. In particular, 0.2
the Steiglitz-McBride technique is utilized to identify the 0 1 2 3 4 5 6 7
t(s)
dominant eigenvalue of the system [25], [26].
To investigate the impact of dc-link voltage dynamics, the (a)
dc-link voltage signal of the detailed model in Case A is

Q(MVAR) Q(MVAR)
1.6 Average
used to identify the critical eigenvalues before and after the 1.2
second disturbance, using the MATLAB built-in Steiglitz- 0.8
0.4
McBride function [27]. The critical eigenvalues of the system
are shown in Fig. 20, where it can be observed that the 1.6 1Ideal2 DC3 Link
4 5 6 7 8 9 10 11 12 13 14
dominant eigenvalues are pushed to the right half plane after 1.2
0.8
the second disturbance, resulting in the undamped oscillations 0.4
in Fig. 11.
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Q(MVAR)

B. Dynamic Analysis 1.6 Detailed

1.2
Without loss of generality, the performance of Switching 0.8
DSM and Average DSM are compared for Case A only in 0.4
Fig. 12. Note that the behaviour of the Switching DSM is 0 1 2 3 4 5 6 7
t(s)
exactly the same as the performance of the detailed model
shown in Fig. 10; thus, by comparing the performance of (b)
the Switching and Average DSMs, it is possible to isolate 61
the impact of high-frequency switching due to the switches, 60
f(Hz)

59
since the physical components of both DSMs are all the same, 58 Average
since only the values of the coefficients of the dependent 57
sources in (13), (14), (19), and (21) are different, reflecting 60 1 2 3 4 5 6 7 8 9 10 11 12 13 14
f(Hz)

the different switching content in the two models. As seen in 50

Fig. 12, the performance of the Average DSM is the same as 40 Ideal DC Link
30
the performance of the average model, showing that including
the dc link dynamics is not necessary in average modeling 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
approaches; furthermore, it confirms that is not possible to 60
f(Hz)

capture the behaviour of the detailed model by neglecting the 50

40 Detailed
switches and/or the impact of high-frequency swithcing. In 30
addition, it can be concluded that including the dc link voltage
0 1 2 3 4 5 6 7
dynamics is necessary but not sufficient to capture the accurate t(s)
dynamic performance of the system.
(c)
With the Switching DSM, it is also possible to isolate and
investigate the impact of Rin on the system’s performance. Fig. 17. Case D: Diesel generator (a) active power; (b) reactive power; and
Thus, it was found that eliminating Rin has no considerable (c) system frequency.
impact on the performance of Switching DSM. Hence, it
can be concluded that the internal resistance of the switches performed in PSCAD/EMTDC 4.6.1.0 on a PC with an
does not significantly contribute to the differences in the Intel(R) Xeon(R) CPU consisting of four 1.86 GHz processors,
performance of the different models. and 64 GB RAM. The computation times were obtained by
averaging waiting times for multiple 1 s simulation times,
C. Computation Times under similar server conditions, for the different models in the
The simulation times in Case A are compared for the same scenario. Note that the Switching DSM is faster than the
proposed BESS models in Table II. The simulations were detailed model, while capturing its behaviour. Considering that

0885-8950 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2017.2740163, IEEE
Transactions on Power Systems
11

1.2 1.4
1 1.2
1

P(MW)
0.8
P(MW)

0.6 0.8
0.6
0.4 0.4
0.2 0.2
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
t(s) t(s)
(a) (a)
1.5 1.6

Q(MVAR)
Q(MVAR)

1 1.2
0.5 0.8
0 0.4
-0.5 0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
t(s) t(s)
(b) (b)
1.1 70
1 60
Volt(pu)

0.9 50

f(Hz)
0.8
0.7 40
0.6 30
0.5 20
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
t(s) t(s)
(c) (c)

1500 Fig. 19. Switching DSM Case D: diesel generator (a) active power; (b)
reactive power; and (c) system frequency.
Volt(v)

1000
500
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 120
t(s)
(d)

Fig. 18. Switching DSM Case D: BESS (a) active power, (b) reactive power,
Imag(Hz)

(c) RMS phase voltages at the PCC, and (d) dc-link voltage.

TABLE II
AVERAGE C OMPUTATION T IMES OF 1 S OF S IMULATION T IME IN C ASE A

Detailed Ideal dc Average Switch. DSM Av. DSM

50.6 s 43.4 s 24.8 s 39.4 s 27.0 s

the difference in computation times increases exponentially

as the number of inverters in the system increases, those can 0
-0.1 0 1.132
become more considerable for larger systems. Finally, note that Real
the Average DSM is slightly slower than the average model,
Fig. 20. Dominant eigenvalue before and after instability in Case A.
yielding essentially the same performance.

VI. C ONCLUSIONS behaviour of other models at adequate computational perfor-

In this paper, first, three different BESS models were mance.
presented and discussed: an average model that replaced the The following can be concluded from the presented studies:
switches with ac voltage sources, an ideal dc link model in • During normal operating conditions, little differences
which the battery and buck/boost converter were replaced by were observed between different models performances;
an ideal dc voltage source, and a detailed model that included the THD were higher for the fast-switching and ideal dc
all necessary modeling and control details. In addition, to link models compared to the average models, due to the
isolate and investigate the impact of high-frequency harmonics switching in the system
caused by switchings, a new model was proposed in which • When the system loading was pushed to its limits, the
the switches were replaced by dependent voltage and cur- fast-switching models lost their stability, while the ideal
rent sources, to capture the average and the fast switching

0885-8950 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2017.2740163, IEEE
Transactions on Power Systems
12

dc link model and the average models remained stable. [10] Z. Jankovic, B. Novakovic, V. Bhavaraju, and A. Nasiri, “Average
Also, it was demonstrated that the ideal dc link model modeling of a three-phase inverter for integration in a microgrid,” in
Proc. IEEE Energy Conversion Congress and Exposition, Milwaukee,
lost its stability for certain loading levels for which the WI, Nov. 2014.
average models remained stable. [11] J. W. Choi and S. K. Sul, “Inverter output voltage synthesis using novel
• Through eigenvalue studies, it was observed that dc link dead time compensation,” IEEE Trans. Power Electron., vol. 11, no. 2,
pp. 221–227, Mar. 1996.
dynamics contribute to the instability problem observed [12] Y. Wang, G. Delille, X. Guillaud, F. Colas, and B. Francois, “Real-
in the fast-switching modelings, due to the increase in time simulation: The missing link in the design process of advanced
charging/discharging current of the dc link capacitor. grid equipment,” in Proc. IEEE Power and Energy Soc. Gen. Meeting,
Minneapolis, MN, July 2010.
• By comparing the performance of the DSMs, it can be [13] O. Tremblay, L.-A. Dessaint, and A. Dekkiche, “A generic battery model
concluded that including the dc link voltage dynamics for the dynamic simulation of hybrid electric vehicles,” in Proc. IEEE
is a necessary but not sufficient condition to capture the Vehicle Power and Propulsion Conference, Arlington, TX, Sep. 2007.
[14] S. Koenig, P. Muller, and T. Leibfried, “Design and comparison of
accurate performance of the system. It was revealed that three different possibilities to connect a vanadium-redox-flow-battery to
high-frequency dynamics also contribute the instability a wind power plant,” in Proc. 13th CHLIE, Valencia, Spain, July 2013.
phenomena observed in the test scenarios. [15] S. M. Ashabani and Y. A. R. I. Mohamed, “New family of microgrid
control and management strategies in smart distribution grids - analysis,
• It was shown that the internal resistance of the switches comparison and testing,” IEEE Trans. Power Syst., vol. 29, no. 5, pp.
is not a major factor in the dynamic performance of the 2257–2269, Sep. 2014.
system. [16] M. Liserre, F. Blaabjerg, and S. Hansen, “Design and control of an lcl-
filter-based three phase active rectifier,” IEEE Trans. Ind. Appl., vol. 41,
• Comparing the computation time of the models, it was no. 5, pp. 1281–1291, Sep. 2005.
noted that during normal operating conditions, it may not [17] M. Restrepo, J. Morris, M. Kazerani, and C. Canizares, “Modeling
be necessary to include the dc link dynamics in averaging and testing of a bidirectional smart charger for distribution system ev
integration,” IEEE Trans. Smart Grid, vol. PP, no. 99, pp. 1–11, Mar.
models. Thus, it can be concluded that when the BESS is 2016.
in normal operating conditions, the average model is the [18] K. Strunz, “Developing benchmark models for studying the integration
best option, since it significantly saves simulation time of distributed energy resources,” in Proc. IEEE Power Eng. Soc. Gen.
Meeting, Montreal, QC, July 2006.
speed, exhibiting adequate performance. However, if the [19] M. Arriaga and C. A. Cañizares, “Overview and analysis of data
BESS operates in grid-forming/grid-supporting modes, for microgrid at kasabonika lake first nation (KLFN),” Hatch Project
and the system is heavily loaded or unbalanced, using Confidential Report, University of Waterloo, Tech. Rep., Sep. 2015.
[20] G. Delille, L. Capely, D. Souque, and C. Ferrouillat, “Experimental
average or ideal dc link models, or any other model that validation of a novel approach to stabilize power system frequency by
neglects the dc link voltage dynamics and/or the high taking advantage of load voltage sensitivity,” in Proc. IEEE PowerTech,
frequency switching, do not properly capture the BESS Eindhoven, June 2015.
[21] P. Kundur, Power System Stability and Control. New York, US:
behaviour during or after a disturbance. McGraw-hill Professional, 1994.
[22] R. H. Salim and R. A. Ramos, “A model-based approach for small-
signal stability assessment of unbalanced power systems,” IEEE Trans.
R EFERENCES Power Syst., vol. 27, no. 4, pp. 1184–1190, Nov. 2012.
[1] R. H. Lasseter et al., “CERTS microgrid laboratory test bed,” IEEE [23] R. H. Salim, R. A. Ramos, and N. G. Bretas, “Analysis of the small sig-
Trans. Power Del., vol. 26, pp. 325–332, Jan. 2011. nal dynamic performance of synchronous generators under unbalanced
[2] IEEE Guide for Design, Operation, and Integration of Distributed operating conditions,” in Proc. IEEE Power Eng. Soc. Gen. Meeting,
Resource Island Systems with Electric Power Systems, IEEE Std. 1547.4, Minneapolis, MN, July 2010.
2011. [24] E. Nasr-Azadani, C. A. Canizares, D. E. Olivares, and K. Bhattacharya,
[3] J. M. Guerrero, F. Blaabjerg, T. Zhelev, K. Hemmes, E. Monmasson, “Stability analysis of unbalanced distribution systems with synchronous
S. Jemei, M. P. Comech, R. Granadino, and J. I. Frau, “Distributed machine and DFIG based distributed generators,” IEEE Trans. Smart
generation: toward a new energy paradigm,” IEEE Ind. Electron. Mag., Grid, vol. 5, no. 5, pp. 2326–2338, Sep. 2014.
pp. 52–64, March 2010. [25] K. Steiglitz and L. E. McBride, “A technique for the identification of
[4] G. Delille, B. Francois, G. Malarange, and J.-L. Fraisse, “Energy storage linear systems,” IEEE Trans. Automatic Control, vol. 10, no. 4, pp. 461–
systems in distribution grids: New assets to upgrade distribution network 464, Oct. 1965.
abilities,” in Proc. 20th International Conf. and Exhibition on Electricity [26] L. Ljung, System Identification: Theory for the User. Upper Saddle
Distribution (CIRED) - Part I, Prague, Czech Republic, June 2009. River, NJ: Prentice Hall, 1998.
[5] G. Delille, B. Franois, and G. Malarange, “Dynamic frequency control [27] Matlab mathworks inc. [Online]. Available:
support by energy storage to reduce the impact of wind and solar https://www.mathworks.com/products/matlab.html
generation on isolated power system’s inertia,” IEEE Trans. Sustain.
Energy, vol. 3, no. 4, pp. 931–939, Nov. 2012.
[6] A. Ortega and F. Milano, “Generalized model of VSC-based energy
storage systems for transient stability analysis,” IEEE Trans. Power Syst.,
vol. 21, no. 5, pp. 3369–3380, Sep. 2016.
[7] F. A. Inthamoussou, J. Pegueroles-Queralt, and F. D. Bianchi, “Control
of a supercapacitor energy storage system for microgrid applications,”
IEEE Trans. Energy Convers., vol. 28, no. 3, pp. 690–697, Sep. 2013.
[8] J. Wu, J. Wen, H. Sun, and S. Cheng, “Feasibility study of segmenting
large power system interconnections with AC link using energy storage
technology,” IEEE Trans. Power Syst., vol. 27, no. 3, pp. 1245–1252, Mostafa Farrokhabadi (S’10) received his B.Sc. in Electrical Engineering
Aug. 2012. from Amirkabir University of Technology, Tehran, Iran, in 2010, his M.Sc.
[9] J. Fang, W. Yao, Z. Chen, J. Wen, and S. Cheng, “Design of anti-windup in Electric Power Engineering from KTH Royal Institute of Technology,
compensator for energy storage-based damping controller to enhance Stockholm, Sweden, in 2012, and his Ph.D. in Electrical and Computer
power system stability,” IEEE Trans. Power Syst., vol. 29, no. 3, pp. engineering from the University of Waterloo, ON, Canada, in 2017. He is
1175–1185, May 2014. currently a postdoctoral fellow at the Electrical and Computer Engineering
Department of the University of Waterloo. His research interests includes
modeling, control, and optimization in microgrids, mathematical modeling,
and state estimation.

0885-8950 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2017.2740163, IEEE
Transactions on Power Systems
13

Sebastian König (M’17) received the Dipl.-Ing. and Dr.-Ing. degrees from
the Karlsruhe Institute of Technology, Karlsruhe, Germany, in 2011 and 2017,
respectively. His current research interest is electric energy storage and its
grid integration. In his PhD thesis, he presents a model-based design and
optimization approach for the Vanadium Redox Flow Battery.

Claudio Cañizares (S’85-M’91-SM’00-F’07) is a Full Professor and the

Hydro One Endowed Chair at the Electrical and Computer Engineering
(E&CE) Department of the University of Waterloo, where he has held various
academic and administrative positions since 1993. He received the Electrical
Engineer degree from the Escuela Politécnica Nacional (EPN) in Quito-
Ecuador in 1984, where he held different teaching and administrative positions
between 1983 and 1993, and his MSc (1988) and PhD (1991) degrees in
Electrical Engineering are from the University of Wisconsin-Madison. His
research activities focus on the study of stability, modeling, simulation,
control, optimization, and computational issues in large and small girds and
energy systems in the context of competitive energy markets and smart grids.
In these areas, he has led or been an integral part of many grants and contracts
from government agencies and companies, and has collaborated with industry
and university researchers in Canada and abroad, supervising/co-supervising
many research fellows and graduate students. He has authored/co-authored a
large number of journal and conference papers, as well as various technical
reports, book chapters, disclosures and patents, and has been invited to make
multiple keynote speeches, seminars, and presentations at many institutions
and conferences world-wide. He is an IEEE Fellow, as well as a Fellow of the
Royal Society of Canada, where he is currently the Director of the Applied
Science and Engineering Division of the Academy of Science, and a Fellow
of the Canadian Academy of Engineering. He is also the recipient of the
2017 IEEE Power & Energy Society (PES) Outstanding Power Engineering
Educator Award, the 2016 IEEE Canada Electric Power Medal, and of various
IEEE PES Technical Council and Committee awards and recognitions, holding
leadership positions in several IEEE-PES Technical Committees, Working
Groups and Task Forces.

Kankar Bhattacharya (M95-SM01-F17) received the Ph.D. degree in elec-

trical engineering from the Indian Institute of Technology, New Delhi, India,
in 1993. He was in the faculty of Indira Gandhi Institute of Development
Research, Mumbai, India, from 1993 to 1998, and the Department of Electric
Power Engineering, Chalmers University of Technology, Gothenburg, Sweden,
from 1998 to 2002. In 2003, he joined the Electrical and Computer Engineer-
ing Department, University of Waterloo, Waterloo, ON, Canada, where he is
currently a Full Professor. His current research interests include power system
economics and operational aspects.

Thomas Leibfried (M’96) received the Dipl.-Ing. and Dr.-Ing. degrees from
the University of Stuttgart, Stuttgart, Germany, in 1990 and 1996, respectively.
From 1996 to 2002, he was with Siemens AG, working in the power
transformer business in various technical and management positions. In 2002,
he joined the Karlsruhe Institute of Technology, KIT as Head of the Institute
of Electric Energy Systems and High-Voltage Technology. He is the author
of various technical papers. Prof. Leibfried is a member of VDE and CIGRE.

0885-8950 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.