1
Accurate Model of a Supercapacitor Bank for
Power System Dynamics Studies
Matej Krpan, Student Member, IEEE, Igor Kuzle, Senior Member, IEEE,
Abstract—In this paper, an accurate and simple dynamic model
of a supercapacitor bank system for power system dynamics stud-
ies is presented. It is shown through comprehensive simulations
arXiv:1908.00919v2 [eess.SY] 23 Mar 2020
in MATLAB-Simulink that an ideal capacitor representation
is not always adequate. The proposed model is derived from
a detailed RC circuit representation. Furthermore, a complete
control system of the supercapacitor bank is also presented.
The proposed model is easy to integrate in any power system
simulation software and consists of maximally 4 easy-to-obtain
parameters. The performance of the proposed model in grid
frequency control and low-voltage ride through was tested in
an IEEE 14-bus test system in DIgSILENT PowerFactory.
Index Terms—power system dynamics, power system simula-
tion, power system modelling, power generation control
I. I NTRODUCTION
T HE trend of increasing inverter-interfaced generation
(IIG) in power systems throughout the world and subse-
quent reduction of synchronous inertia has motivated many re-
search efforts to understanding stability of low-inertia systems
as well as developing new algorithms which enable the IIG
participation in system frequency control and other ancillary
services [1]–[5].
Supercapacitor energy storage system (ESS) [6] is an al-
ternative to fast frequency response services provided by
batteries and it can be used in coordination with batteries, e.g.
supercapacitor can provide fast power injection or absorption
to/from the grid immediately after an active power disturbance
while batteries can take over in the longer time scale. The
benefits of a supercapacitor system are as follows [7], [8]:
• supercapacitor can be charged and discharged at full
power in the time scale of several tens of seconds or
faster;
• supercapacitors are much smaller than batteries and can
withstand significantly more charging/discharging cycles.
Relatively speaking, their high power / low energy density
make them complementary with batteries which are low power
/ high energy density.
There are many papers utilizing supercapacitor technology
in power system or industrial applications. However, as we will
show in the following paragraphs, there were no detailed stud-
ies on the adequate complexity level of a supercapacitor model
for power system dynamics studies, no rigorous derivation of
the supercapacitor bank model and no complete representation
of a supercapacitor bank system with associated control that
M. Krpan and I. Kuzle are with the Department of Energy and Power
Systems, Faculty of Electrical Engineering and Computing, University of
Zagreb, 10000 Zagreb, Croatia, e-mail: matej.krpan@fer.hr; igor.kuzle@fer.hr
can be easily integrated in power system simulation software.
Furthermore, many papers use overly simplified supercapacitor
models which do not adequately capture the supercapacitor
dynamics. Hence we attempt to bridge this gap with this paper.
A. Literature survey
In [9]–[11], authors based their supercapacitor model on an
ideal capacitor model with a constant capacitance. However,
supercapacitor model is nonlinear and its capacitance varies
with the applied voltage. Li et al. [12] showed real-time sim-
ulation of a wind turbine and a battery-supercapacitor hybrid
system. They use a very accurate model of a supercapacitor
from [13], but the detailed representation of the system and
the controls are not shown. Arani and Saadany [14] used a
supercapacitor cell as the DFIG DC link for virtual inertial
response application, but they also based their supercapacitor
model on an ordinary ideal capacitor. Rahim and Nowicki [15]
use the supercapacitor system for DFIG fault-ride through, but
their model is also based on ideal capacitor. Saw et al. [16]
use a supercapacitor for electric vehicle applications, but the
used model is numerical and not applicable to power system
applications. Fang et al. [17] presented a hybrid ultracapacitor-
battery system for implementation of a virtual synchronous
generator and Garcia et al. [18] proposed a control strategy for
a battery-ultracapacitor hybrid system. Both groups of authors
based their supercapacitor model on an ideal capacitor as well.
Molina and Mercado [19] presented a model of a DSTATCOM
with ultracapacitor storage for power distribution system ap-
plications. Their model is very detailed, but it is represented
in the RLC form which is not useful for implementation in
power system simulation software. Furthermore, they did not
show charging/discharging control. Many other papers [20]–
[28] have the same shortcomings: too simple model and/or
model not applicable for power system dynamic simulations
application.
B. Contribution
As shown in the literature survey, all of the papers have the
same shortcomings:
• supercapacitor is modelled as an ideal capacitor which
is not always appropriate because the capacitance varies
with the applied voltage and therefore the stored energy
as well;
• supercapacitor energy storage system model is not ap-
plicable for power system dynamics because the model
is either given in its RC/RLC form or the complete

control system was not shown which makes it difficult
to integrate in power system simulation software;
• there were no studies which compared different levels of
supercapacitor model accuracy.
To the best of our knowledge, there were no papers that
developed an accurate, simple and complete model of a
supercapacitor/ultracapacitor bank for power system stability
studies. A complete model should include: accurate dynamics
of a supercapacitor cell, supercapacitor DC current calculation,
charge/discharge control, active power and voltage/reactive
power inverter control as well as frequency control loop.
Therefore, the contributions of this paper are as follows:
• detailed analysis and comparison of different supercapac-
itor models with varying levels of detail;
• derivation of the adequate level of detail of the super-
capacitor model for power system dynamics applications
(voltage and frequency control, transient stability stud-
ies);
• accurate dynamic model of the supercapacitor bank with
all the necessary controls.
The presented model is easy to integrate in any power
system simulation software. Rest of the paper is structured as
follows: methodology is described in section II; supercapacitor
bank model is derived in section III; complete supercapacitor
bank system is presented in section IV. Performance of the
model and simulation results are given in section V. Section
VI concludes the paper.
II. M ETHODOLOGY
We will start by reviewing the relevant literature on su-
percapacitor modelling and showing the detailed state-of-the-
art supercapacitor cell model. The model structure is identi-
fied from experimental measurements, most often impedance
spectroscopy. Starting from the full model, we will start
simplifying it by gradually reducing the number of parameters
describing the model with the final goal of arriving to an
accurate and as simple as possible model which captures the
relevant supercapacitor/ultracapacitor dynamics. The model is
intended to be used in power system short-term dynamics stud-
ies (transient stability, voltage control, inertial response and
primary frequency control), i.e. the time scale of observation
is up to 60 seconds after a disturbance. With every step of the
way, we will compare different models with varying levels
of detail to prove the validity of our simplification. Once
the simplified model is derived, it will be scaled to form a
supercapacitor bank of a higher rated power (MW order of
magnitude). Then, the complete supercapacitor bank system
with controls will be developed. The performance of the model
will be shown in a standard IEEE 14-bus test system modelled
in DIgSILENT PowerFactory 2019 software package. The
type of simulations conducted are stability simulations (RMS,
integration step 0.01 s), i.e. power electronic converters are
represented by their average models.
III. S UPERCAPACITOR BANK MODEL DERIVATION
A. Supercapacitor theory
Core of the supercapacitor bank model is the supercapacitor
cell. Overview of different supercapacitor models can be found
2
in [29] while the state-of-the-art supercapacitor models can
be found in [13], [29], [30]. Basically, these models are
all similar and are based on RC circuit identification using
impedance spectroscopy. All these models can be described
with the same type of RC circuit consisting of three parallel
sections as shown in Fig. 1. The first branch (blue area) models
fast dynamics, parallel branches (green area) model slower
recombination phenomena after a fast charge or discharge
and the last branch (orange area) models the long-term self-
discharge phenomena [30]. We will gradually reduce the
number of elements of this model in order to arrive to a
model which is suitable for power system electromechanical
dynamics time scale.
A few characteristics of a supercapacitor must be noted
before we continue:
• majority of the ultracapacitor capacitance comes from
Csc ;
s s s s
• series combination of parallel branches R1 C1 –Rn Cn
is actually an infinite series of these parallel groups.
However, 5 elements are enough to obtain an accurate
model according to [30];
s s
• capacitance Csc as well as infinite sum elements Rk , Ck
are dependent on the ultracapacitor voltage uC (t). This
model is nonlinear with time-varying parameters. That is
why ideal capacitor representation used in many papers
is not always appropriate;
• the number of parallel branches in the green group is
also theoretically infinite, but between two branches and
4 branches are sufficient to achieve accurate results [30].
A few assumptions are made to simplify the model:
• Rs is the series resistance determined at very high
frequency and is also voltage dependent [30]. However,
since this resistance is small (< 10 mΩ) and the impact
on the model performance is insignificant, we consider
it as a constant parameter which is also usually done in
reviewed literature on accurate supercapacitor modelling;
• temperature dependence of the parameters is neglected.
Temperature is considered constant. The assumption is
that the cooling of the system is adequate and that the
system operates at room temperature. This effect can be
included in a future version of the model, but we consider
it not important for the initial derivation of the model for
power system dynamics.
Parameters of the first branch are calculated according to
(1)–(3) [30].
Csc (uC ) = C0 + kv uC (t) (1)
1
Cks = Csc , k ∈ {1...n} (2)
2
2τ (uC )
Rks = 2 2 (3)
k π Csc
C0 is the ultracapacitor capacitance at 0 V and kv is a constant
expressed in F/V. τ (uC ) is another experimentally determined
parameter (it has a dimension of time) that can be expressed
as function changing linearly with the voltage uC : τ (uC ) =
τ0 + kτ uC (t) [30]. However, it can also be approximated by
(4) [30]:
τ (uC ) ≈ 3Csc (Rdc − Rs ), (4)
 3

{M3}:Leakage current branch

Rleak TABLE I
C OMMERCIAL SUPERCAPACITORS PARAMETERS OF THE FIRST BRANCH
FROM [29], [30]

p
p
Rm Cm Cell Rs [mΩ] Rdc [mΩ] C0 [F] kv [F/V]
Maxwell BCAP0150 6.3 7.1 108.4 10.8
Epcos 110 F 10.0 11.0* 89.0 29.1
{M2}:m parallel branches
*data not available, arbitrarily chosen

R1p C1p
TABLE II
E PCOS 110 F 3 BRANCH MODEL PARALLEL BRANCHES DATA [29]
C1s (uC ) Cns (uC ) Parameter R1p [Ω] C1p [F] Rleak [kΩ]

Rs Csc (uC ) +
− +
− Value 17.5 13.7 5
uC1s uCns
uRs − −
+ +

uC TABLE III
isc M AXWELL 150 4 BRANCH MODEL PARALLEL BRANCHES DATA [30]
{M1}:First branch with
R1s (uC ) Rns (uC )
n parallel groups
Parameter R1p [Ω] C1p [F] R2p [kΩ] C2p [F] Rleak [kΩ]
+
−
usc (t) Value 80.2 28.0 3.73 27.1 66.6

Fig. 1. Detailed RC circuit of a supercapacitor cell

TABLE IV
M AXWELL 150 6 BRANCH MODEL PARALLEL BRANCHES DATA [30]

where Rdc is the resistance experimentally obtained at very
low frequencies (essentially DC). Naturally, Rdc > Rs .
All the parameters of the {M1} branch can be identified
using manufacturer’s data sheet.
All the parameters of the first branch can be identified using
manufacturer’s data sheet. Parameters of the parallel branches
are more difficult to obtain since they must be obtained ex-
perimentally. Furthermore, these parameters are not universal
and they depend on the time scale of the phenomena to be
observed (described by the RC time constant τRC = RC).
The time scales are arbitrary, however they usually imply a
range from several minutes to several weeks or even more
[29], [30].
B. Simplification of the supercapacitor cell model
In this section, we will show that the supercapacitor model
for power system dynamics studies can be described with
only the first branch and with none of the parallel groups
Rks Cks . Real experimental data from [29], [30] will be used
in simulations for model simplification. This experimental
data concerns two commercial supercapacitors which will be
simulated with different models with varying levels of detail to
show that our simplification is valid. Parameters of the parallel
branches (green and orange section in Fig. 1) are determined
experimentally through constant current charge test (Table I –
Table IV).
Simulations of different model responses are conducted in
MATLAB-Simulink using Simscape Electrical toolbox. The
number of parallel branches is being sequentially reduced and
the different model responses to the charge/discharge test are
compared. Input to the model is the current isc (t) and output
of the model is the supercapacitor voltage usc (t). Results are
shown in Fig. 2. For clarification, 6 branch model represents
the total number of branches (first branche, 4 parallel branches
and a self-discharge branch). Results for both supercapacitors
with different levels of detail (Fig. 2b and Fig. 2c) show that
Parameter R1p [Ω] C1p [F] R2p [Ω] C2p [F]
Value 32.7 15.3 275 18.1
Parameter R3p [kΩ] C3p [F] R4p [kΩ] C4p [F]
Value 3.5 14.4 18.4 27.1
Parameter Rleak [kΩ]
Value 111
all the parallel branches as well as the self-discharge branch
do not have an impact on the model accuracy for the time
scale of interest. Therefore, all the branches except the first
branch can be neglected.
In the next step, the adequate number of parallel RC groups
in the first branch is determined. Results are shown in Fig
3. Here, the models are also compared to the ideal capacitor
representation used in many papers to show the difference.
Firstly, it can be seen that the parallel groups do not play
a significant impact in the voltage response, although this
depends on the model parameters since it can be seen that
the effect is more pronounced for the Epcos model (Fig.
3b). Generally, accuracy is not lost if the parallel groups are
neglected, although at least one should be included if greater
accuracy is to be achieved.
Fig. 3a and Fig. 3b show that using the ideal capacitor
representation will yield inaccurate voltage response. The
value of the capacitance was chosen as C0 of the respective
supercapacitors model which is the worst case scenario. More
accurate profile could be obtained by choosing a value which is
much closer to the capacitance at rated voltage (Fig. 3c). Nev-
ertheless, the ideal representation will not reflect the voltage
transient effect which occurs when the charging or discharging
current is discontinued. Fig. 3c shows the difference between
stored energy for a detailed model and an ideal capacitor. If
ideal capacitor representation has to be used, than it is better to
use a capacitance value which is closer to the supercapacitor
capacitance at rated voltage as the error is significantly smaller.
 5 RC g
1 RC g
2 4

0 RC g

usc [V]
5 RC groups
20
2
1 RC group ideal capaci
isc [A]
0 RC group
0

usc [V]
ideal capacitor 108 F
−20 31 1 1 RC g
0 10 20 30
Time [s]
40 50 60
0 0 RC g
0 10 20 30 40 50 60
(a) Test current 30 A
Time [s] ideal capac
2

usc [V]
6 branch model (a) Maxwell BCAP0150 voltage
2 4 branch model

0
usc [V]

1 branch model
3 1 RC groups
1 0 RC group
ideal capacitor 89 F
2

usc [V]
1 0 10 20 30 40
0
0 10 20 30 40 50 60
1
Time [s]
(b) Maxwell BCAP0150 voltage 0
0 10 20 30
Time [s]
40 50 60
3 branch model Time [s]
2
0
1 branch model
usc [V]

(b) Epcos 110F voltage

1
10,000 Detailed model

Stored energy [J]

0
0 10 20 30 40 50 60 0 10 5,000
20 30 40
Ideal capacitor C = C0
Ideal capacitor C = C0 + kv Umax

Time [s]
(c) Epcos 110F voltage
Fig. 2. Comparison of model response for different number of branches 0
Time [s]
0 0.5 1 1.5 2 2.5
usc [V]

(c) Energy stored in a supercapacitor with respect to

C. Building a supercapacitor bank model voltage
Fig. 3. Comparison of model response for different number of parallel RC
groups in the first branch
In section III-B it was shown that the supercapacitor dy-
namics can be accurately represented using first branch only
(gray section in Fig. 1) with at least one parallel RC group. To
build a bank of a higher power rating, a certain number of cells n
X
ns can be connected in series to form a string and a certain usc (t) = isc (t)Rs + uC (t) + uCks = y(t) (7)
number of strings np can be connected in parallel to form a k=1
module. Modules can then be connected in parallel to form a isc (t) = u(t) (8)
bank. Assuming completely identical cells, it is easily shown ussc (t) = ns usc (t) = ns y(t) (9)
using Kirchoff’s voltage and current laws that the voltage of
the string ussc and the current of the module im im
sc (t)= np isc (t) = np u(t) (10)
sc are equal to
(5) and (6), respectively. duC isc (t)
= (11)
dt C0 + kv uC (t)
duCks uC s isc (t)
ussc (t) = ns usc (t) (5) = − s ks + (12)
dt Rk Ck Ck
Complete block diagram of the supercapacitor bank model
described by (7)–(12) is shown in Fig. 4.
im
sc (t) = np isc (t) (6)
IV. C OMPLETE CONTROL SYSTEM
The complete control system consists of inverter PQ con-
Finally, the dynamic model of the bank can be built using trol, charge/discharge control, DC current calculation and
circuit analysis in the time domain for the first branch only frequency control loop. The block diagram of the complete
by setting usc (t) as an output y(t), isc (t) as an input u(t). supercapacitor bank energy storage system system is shown
Capacitor voltages are chosen as state variables. Complete in Fig. 5. P and Q are active and reactive power injected or
nonlinear model of the supercapacitor bank in the analytic absorbed by the inverter to or from the grid, while asterisk
form is described by (7)–(12) where Rks and Cks are defined grid
(*) denotes a set-point value. Vac is the AC voltage of the
by (2) and (3), respectively. bus the inverter is connected to. id and iq are the direct and
 5

im 1/np isc(t) uRs (t)

sc (t) Rs

C0 + ÷
R uC (t)
R uC1s (t)
2 −
uC (t) kv
÷
3(Rdc −Rs ) 1
π2 12
R uC2s (t) P usc(t) n
− s ussc(t)

÷
1
22
−
R uCns (t)

÷
1
n2

Fig. 4. Block diagram of the nonlinear supercapacitor module model

P ∗; P
quadrature axis currents of the inverter. Inverter is controlled i∗d
Q∗ ; Q
i∗q
in the grid voltage reference frame. PLL estimates the grid i0d
Charge control P
Grid frequency δP PQ
f & Inverter
voltage angle as well as the frequency for frequency control control control i0q
LVRT Q

block. DC current calculation block calculates the supercapac-

V ∗ ; Vac
grid

itor current for charging or discharging. Individual blocks are cos θ∗ ;

sin θ∗
further elaborated in the following subsections.

im Ultracapacitor/ usc ; ussc f PLL

DC current
A. Charge control P
calculation
sc
supercapacitor
model

Fig. 6 shows the structure of this block. State-of-Voltage

(SoV) measurement is used to control the charging and dis- grid
Vac
imax
dch
charging process. Charging is stopped if the supercapacitor imax
ch

bank is charged to nominal voltage, while discharging is

stopped when the supercapacitor voltage falls below a user
defined low voltage threshold. Charging/discharging is enabled
again when the voltage reaches a user defined minimum volt-
age level for charging/discharging. The input to the block are
the d and q axis currents i0d and i0q from the PQ control, while
the final inverter current set-points i∗d and i∗q are determined by
this block. Simple low-voltage ride through logic and current
limitation is also implemented in this block which won’t
be shown since this something that can be found in many
literature.
B. DC current calculation
Input to the supercapacitor model is the current, but in
power system applications we usually deal with controlling
the power, not current. This block calculates the charging or
discharging DC current based on the actual inverter power
output. Block diagram of this subsystem is shown in Fig. 7. It
should be noted that this module as well as the supercapacitor
model work with SI units, while other subsystems work in
max max
p.u. Ich and Idch are the maximum single cell charging and
discharging current in A (e.g. ±100 A).
Fig. 5. Complete model of the supercapacitor bank system
C. PQ control
Fig. 8 shows the PQ control structure of the supercapacitor
bank inverter. In this case, the inverter is modelled as a
controlled current source and the d and q axis currents are
obtained from the active and reactive power control error,
respectively. Measurement/control lag is also included in this
block diagram. The term i∗d − i0d is a compensation term for
active power during low-voltage ride through when the active
power should be low and reactive power high. Reactive power
or terminal voltage control can be both chosen. However,
if reactive power control is chosen it will be overridden by
terminal voltage control during low-voltage ride through.
D. Grid frequency control
This block is shown in Fig. 9. The input to this block is
the grid frequency signal estimated by the PLL and the output
is the requested change in power. The type of implemented
algorithm for frequency response can be arbitrary. However,
 6

usc max
≥ Uch S the output power will slowly diminish while the conventional
disable charging flag imax ∈ {0,1}
units pick up. Therefore, the difference between the inertia
Q ch
start
control and quasi-droop control is in the washout filter time
≤ Uch R constant (τwd  τwi ).
enable charging flag
SR flip-flops
V. S IMULATION AND RESULTS
min
≤ Udch S
The performance of the proposed model is implemented and
disable discharging flag −Q̄ imax
dch ∈ {−1,0} tested on a standard IEEE 14-bus test system shown in Fig.
start
≥ Udch R 10 in DIgSILENT PowerFactory. Three scenarios are tested:
enable discharging flag underfrequency event, overfrequency event and low-voltage
ride through event. Base case is without the supercapacitor
grid
Vac
bank contribution and a response with an ideal capacitor
i0d i∗d representation is also tested. Supercapacitor bank is connected
LVRT logic
i0q and
to bus 06. Parameters of the supercapacitor bank system and
max
ich current limitation i∗q the test grid are given in Appendix A and Appendix B,
imax respectively.
dch

Fig. 6. Charge control, LVRT and current limitation block A. Underfrequency / Overfrequency event
At t = 1 s, 13.5 MW load at bus 13 is connected to
imax max
ch Ich np trigger an underfrequency event. Supercapacitor cell is initially
charged to 1 V. Results are shown in Fig. 11.
1
P ÷ τc +1 im
sc It can be seen that the ideal capacitor representation (with
constant capacitance equal to the supercapacitor capacitance at
imax max
dch Idch np
Control and
ussc measurement lag rated voltage) accurately describes the nonlinear model up to
the first nadir (Fig. 11). However, the ideal representation gives
Fig. 7. DC current calculation block overly optimistic results regarding the stored energy which
can be seen by the prolonged discharge time in Fig. 11b–Fig.
Control and 11d. On the other hand, detailed model is much more accurate
measurement lag and gives an accurate behaviour regarding discharge power
P − + − 1
τc +1 Kpd +
Kid
i0d and available energy (notice the bigger nadir of the secondary
s
P∗ frequency drop around 7 second mark in Fig. 11a).
δP Similar behavior can be observed for an overfrequency event
− i∗d
(when the same load is disconnected from the grid) in Fig.12.
Q − Initial supercapacitor cell voltage is set to 2.3 V. In this case,
q
1 the supercapacitor bank is quickly charging to compensate
Q∗ τc +1 Kpq +
Ki
s
i0q
V − for the temporary surplus of generation. In this scenario,
V∗ the ideal representation describes the nonlinear model much
Low-voltage
flag more accurately. This is because the ideal representation with
maximum capacitance much more accurately describes the
Fig. 8. Supercapacitor bank inverter PQ control nonlinear model near rated voltage as shown in Fig. 3c.
However, if the initial voltage was not near rated voltage or if
the ideal capacitor capacitance was lower, then the difference
based on the supercapacitor characteristics, in this paper we between the two models would be greater.
decided for two control loops which look identical. The bottom
loop is a standard virtual inertial response with a washout
filter to make the output signal more smooth since the time B. Low-voltage ride through
derivative operation inherently amplifies noise. The upper loop Here, the performance of the low-voltage ride through
is more akin to a standard droop control, but it also has a algorithm is tested. Initial supercapacitor cell voltage is 2 V
washout filter which means this contribution will diminish in (i.e. string voltage is 200 V). Supercapacitor bank inverter is
steady-state, hence the name quasi-droop. set to control unity power factor with the grid (Q = 0). At
The reasoning for this choice is the following: the superca- t = 1 s, 13.5 MW load at bus 13 is connected to trigger
pacitor does not have a lot of stored energy—if the standard an underfrequency event and the supercapacitor bank starts
droop control is employed then the supercapacitor output discharging. At t = 2.6 s, a three-phase short circuit is applied
power is initially proportional to the frequency deviation. to bus 6 which is cleared after 400 ms at t = 3 s. Results are
However, once the supercapacitor is discharged, the output shown in Fig. 13.
power will fall to zero which will cause a bigger secondary It can be seen in Fig. 13d and Fig. 13e that the current, thus
frequency drop. By setting a large washout filter time constant, output active power are reduced when the voltage dips at t =
 7

Quasi-droop control 60 No supercapacitor

Frequency [Hz]
Ideal capacitor C = 3000 F
59.95 Nonlinear model
sτwd
∆f Kd + δP
sτwd + 1
59.9

Gain Washout filter

0 2 4 6 8 10 12 14
Inertia control
Time [s]
sτwi
Ki (a) Grid frequency
sτwi + 1
100 No supercapacitor
Ideal capacitor C = 3000 F
Fig. 9. Supercapacitor bank grid frequency control module

ussc [V]
Nonlinear model
80

60
Bus 13

Bus 12 Bus 14 0 2 4 6 8 10 12 14
L6-13

Bus 11
Time [s]
Bus 10
G01
~
SG

Bus 09
(b) Supercapacitor bank string voltage
Bus 01 Bus 06 T7-8
T7-9

G08
G06

SG
~
Bus 08
T4-9
T5-6

0
SG
~

Supercapacitor Bus 07
T4-7

Bus 05
−1,000

sc [A]
Bus 04

−2,000

im
No supercapacitor
Ideal capacitor C = 3000 F
Bus 02 −3,000
L3-4

Nonlinear model
SG
~

G02 0 2 4 6 8 10 12 14
Bus 03
Time [s]
SG
~

G03 (c) Supercapacitor bank module current

No supercapacitor
Ideal capacitor C = 3000 F
Fig. 10. IEEE 14-bus test system 2
P [MW]

Nonlinear model

2.6 s (Fig. 13a), i.e. voltage control takes priority over active 0
power control, supplying the grid with reactive power from 0 2 4 6 8 10 12 14
t = 2.6 s to t = 3 s (Fig. 13f). Once the fault is cleared, the Time [s]
reactive power quickly returns to the initial set-point (Q = 0). (d) Supercapacitor bank total power
Now, the whole system is speeding up and the supercapacitor
Fig. 11. Grid frequency and supercapacitor voltage, current and power profile
bank system acts as a brake reducing the change in frequency for an underfrequency event
as shown in Fig. 13b. Therefore, this simulation has shown that
the low-voltage ride through algorithm performs adequately.
performance of the presented model has been tested in an
VI. C ONCLUSION IEEE 14-bus test system to show that the model behaves
correctly. The proposed model is easy to implement in any
In this paper, an accurate and complete supercapacitor bank
power system simulation software (e.g. PSS/E, PowerFactory,
model has been presented for use in power system dynamics
etc.) since it consists of basic elements only (e.g. integrators,
simulations. Starting from the most detailed RC model of a
gains, etc.). The model structure can be easily reduced to an
supercapacitor cell, the model has been gradually reduced until
ideal capacitor representation by neglecting certain parameters
arriving to the most simple representation which adequately
(i.e. setting them to 0). The main drawback of the proposed
describes the supercapacitor dynamics, confirmed by simula-
model is that it wasn’t validated against a real supercapacitor
tion experiments. The proposed model is described with only
bank, which will be done in future research.
4 parameters which are easy to obtain from manufacturer’s
data sheet: capacitance at zero voltage, voltage-dependent
capacitance part, DC resistance and high-frequency resistance.
Furthermore, the presented model was compared to an ideal
capacitor representation to show that such representation is not
always accurate. Then, a supercapacitor bank was built using
supercapacitor cells to form strings and modules based on the
assumption of identical cells.
A complete control system is presented including DC cur-
rent calculation, charge and discharge control, PQ control,
grid frequency control and low-voltage ride through. The
 8

60.3

Frequency [Hz]
260
60.2

ussc [V]
60.1 No supercapacitor No supercapacitor
Ideal capacitor C = 3000 F 240 Ideal capacitor C = 3000 F
Nonlinear model Nonlinear model
60
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
Time [s] Time [s]

(a) Grid frequency (b) Supercapacitor bank string voltage

0
No supercapacitor
1,500
Ideal C = 3000 F

P [MW]
Nonlinear model
sc [A]

1,000 −2
im

No supercapacitor
500 Ideal C = 3000 F
−4 Nonlinear model
0
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
Time [s] Time [s]

(c) Supercapacitor bank module current (d) Supercapacitor bank total power
Fig. 12. Grid frequency and supercapacitor voltage, current and power profile for an overfrequency event

60.5 210
Without SC bank
1
Frequency [Hz]
Voltage [p.u.]

Witch SC bank
200
60

ussc [V]
0.5 190
59.5
Without SC bank Without SC bank
With SC bank 180 With SC bank
59
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14 0 1 2 3 4 5 6
Time [s] Time [s] Time [s]

(a) Bus 6 line-to-line voltage (b) Grid frequency (c) Supercapacitor bank string voltage

10 10
5,000 Without SC bank
With SC bank
Q [MVAr]

5
P [MW]
sc [A]

0
0
0
im

Without SC bank Without SC bank

−5,000 −10 With SC bank −5 With SC bank

0 1 2 3 4 5 6 0 1 2 3 4 5 6 0 1 2 3 4 5 6
Time [s] Time [s] Time [s]

(d) Supercapacitor bank module current (e) Supercapacitor bank total active power (f) Supercapacitor bank total reactive power
Fig. 13. Supercapacitor bank low-voltage ride through performance

A PPENDIX A G08: GAST turbine governor and IEEET1S excitation system.

S UPERCAPACITOR BANK SYSTEM PARAMETERS All grid and element parameters have default values from
ns = 100, np = 10, number of modules: 10, module/bank DIgSILENT PowerFactory unless specified otherwise.
rated power: 1 MW / 10 MW, C0 = 1800 F, kv = 444 ACKNOWLEDGMENT
max max
F/V, Rdc = 0.29 mΩ, Rs = 0.2 mΩ, Ich /Idch = ±500 A,
max start min start The work of the authors is a part of the H2020 project
Uch = 2.7 V, Uch = 2.5 V, Udch = 0.5 V, Udch = 1.5 V,
CROSSBOW — CROSS BOrder management of variable
τc = 50 ms, Ki = Kv = 100 p.u., τwi = 1 s, τwd = 30 s,
renewable energies and storage units enabling a transnational
Kpd = Kpq = 1 p.u., Kid = Kiq = 400 p.u.
Wholesale market (Grant No. 773430). This document has
been produced with the financial assistance of the European
A PPENDIX B Union. The contents of this document are the sole responsi-
IEEE 14- BUS TEST SYSTEM DATA bility of authors and can under no circumstances be regarded
G01: HYGOV turbine governor and IEEET1S excitation sys- as reflecting the position of the European Union.
tem.
R EFERENCES
G02: TGOV1 turbine governor (T2 = 2.4 s, T3 = 8 s) and
IEEET1S excitation system. [1] F. Milano, F. Dörfler, G. Hug, D. J. Hill, and G. Verbič, “Foundations
and challenges of low-inertia systems (invited paper),” in 2018 Power
G03: GAST turbine governor and IEEET1S excitation system. Systems Computation Conference (PSCC), June 2018, pp. 1–25, doi:
G06: GAST turbine governor and IEEET1S excitation system. 10.23919/PSCC.2018.8450880.

[2] U. Markovic, O. Stanojev, E. Vrettos, P. Aristidou, and G. Hug,
“Understanding stability of low-inertia systems,” February 2019, doi:
10.31224/osf.io/jwzrq (unpublished).
[3] P. Tielens and D. Van Hertem, “The relevance of inertia in power
systems,” Renewable and Sustainable Energy Reviews, vol. 55, pp. 999–
1009, March 2016.
[4] M. Krpan and I. Kuzle, “Introducing low-order system frequency re-
sponse modelling of a future power system with high penetration of
wind power plants with frequency support capabilities,” IET Renewable
Power Generation, vol. 12, pp. 1453–1461, October 2018.
[5] M. Krpan and I. Kuzle, “Towards the new low-order system frequency
response model of power systems with high penetration of variable-
speed wind turbine generators,” in 2018 IEEE Power Energy Society
General Meeting (PESGM), Aug 2018, pp. 1–5.
[6] A. González, E. Goikolea, J. A. Barrena, and R. Mysyk, “Review on su-
percapacitors: Technologies and materials,” Renewable and Sustainable
Energy Reviews, vol. 58, pp. 1189 – 1206, 2016.
[7] X. Luo, J. Wang, M. Dooner, and J. Clarke, “Overview of current
development in electrical energy storage technologies and the application
potential in power system operation,” Applied Energy, vol. 137, pp. 511–
536, jan 2015.
[8] Maxwell Technologies, “Ultracapacitors frequency response application
brief,” 2018, technical brochure.
[9] V. Gevorgian, E. Muljadi, Y. Luo, M. Mohanpurkar, R. Hovsapian, and
V. Koritarov, “Supercapacitor to provide ancillary services,” in 2017
IEEE Energy Conversion Congress and Exposition (ECCE), Oct 2017,
pp. 1030–1036.
[10] T. Zhou and W. Sun, “Optimization of battery–supercapacitor hybrid
energy storage station in wind/solar generation system,” IEEE Transac-
tions on Sustainable Energy, vol. 5, no. 2, pp. 408–415, April 2014.
[11] L. Qu and W. Qiao, “Constant power control of dfig wind turbines
with supercapacitor energy storage,” IEEE Transactions on Industry
Applications, vol. 47, no. 1, pp. 359–367, Jan 2011.
[12] W. Li, G. Joos, and J. Belanger, “Real-time simulation of a wind turbine
generator coupled with a battery supercapacitor energy storage system,”
IEEE Transactions on Industrial Electronics, vol. 57, no. 4, pp. 1137–
1145, April 2010.
[13] S. Buller, E. Karden, D. Kok, and R. W. De Doncker, “Modeling the
dynamic behavior of supercapacitors using impedance spectroscopy,”
IEEE Transactions on Industry Applications, vol. 38, no. 6, pp. 1622–
1626, Nov 2002.
[14] M. F. M. Arani and E. F. El-Saadany, “Implementing virtual inertia
in dfig-based wind power generation,” IEEE Transactions on Power
Systems, vol. 28, no. 2, pp. 1373–1384, May 2013.
[15] A. Rahim and E. Nowicki, “Supercapacitor energy storage system for
fault ride-through of a dfig wind generation system,” Energy Conversion
and Management, vol. 59, pp. 96 – 102, 2012.
[16] L. H. Saw, H. M. Poon, W. T. Chong, C.-T. Wang, M. C. Yew, M. K.
Yew, and T. C. Ng, “Numerical modeling of hybrid supercapacitor
battery energy storage system for electric vehicles,” Energy Procedia,
vol. 158, pp. 2750 – 2755, 2019, innovative Solutions for Energy
Transitions.
[17] J. Fang, Y. Tang, H. Li, and X. Li, “A battery/ultracapacitor hybrid
energy storage system for implementing the power management of vir-
tual synchronous generators,” IEEE Transactions on Power Electronics,
vol. 33, no. 4, pp. 2820–2824, April 2018.
[18] F. S. Garcia, A. A. Ferreira, and J. A. Pomilio, “Control strategy for
battery-ultracapacitor hybrid energy storage system,” in 2009 Twenty-
Fourth Annual IEEE Applied Power Electronics Conference and Expo-
sition, Feb 2009, pp. 826–832.
[19] M. G. Molina and P. E. Mercado, “Modeling of a DSTATCOM with
ultra-capacitor energy storage for power distribution system applica-
tions,” in XIII Eriac Décimo Tercer Encuentro Regional Iberoamericano
de CIGRÉ, May 2009, pp. 1–8.
[20] M. H. Fini and M. E. H. Golshan, “Determining optimal virtual inertia
and frequency control parameters to preserve the frequency stability in
islanded microgrids with high penetration of renewables,” Electric Power
Systems Research, vol. 154, pp. 13 – 22, 2018.
[21] B. Wang, L. Xian, U. Manandhar, J. Ye, X. Zhang, H. B. Gooi, and
A. Ukil, “Hybrid energy storage system using bidirectional single-
inductor multiple-port converter with model predictive control in dc
microgrids,” Electric Power Systems Research, vol. 173, pp. 38 – 47,
2019.
[22] J. Cao, W. Du, H. Wang, and M. McCulloch, “Optimal sizing and
control strategies for hybrid storage system as limited by grid frequency
deviations,” IEEE Transactions on Power Systems, vol. 33, no. 5, pp.
5486–5495, Sep. 2018.
9
[23] L. Sigrist, I. Egido, E. Lobato Miguélez, and L. Rouco, “Sizing and
controller setting of ultracapacitors for frequency stability enhancement
of small isolated power systems,” IEEE Transactions on Power Systems,
vol. 30, no. 4, pp. 2130–2138, July 2015.
[24] S. I. Gkavanoudis and C. S. Demoulias, “A combined fault ride-through
and power smoothing control method for full-converter wind turbines
employing supercapacitor energy storage system,” Electric Power Sys-
tems Research, vol. 106, pp. 62 – 72, 2014.
[25] L. Li, Z. Huang, H. Li, and J. Peng, “A rapid cell voltage balancing
scheme for supercapacitor based energy storage systems for urban rail
vehicles,” Electric Power Systems Research, vol. 142, pp. 329 – 340,
2017.
[26] Y. Liu, W. Du, L. Xiao, H. Wang, S. Bu, and J. Cao, “Sizing a hybrid
energy storage system for maintaining power balance of an isolated
system with high penetration of wind generation,” IEEE Transactions
on Power Systems, vol. 31, no. 4, pp. 3267–3275, July 2016.
[27] A. Tahri, H. E. Fadil, F. Belhaj, K. Gaouzi, A. Rachid, F. Giri, and
F. Chaoui, “Management of fuel cell power and supercapacitor state-
of-charge for electric vehicles,” Electric Power Systems Research, vol.
160, pp. 89 – 98, 2018.
[28] L. Yang, Z. Hu, S. Xie, S. Kong, and W. Lin, “Adjustable virtual inertia
control of supercapacitors in pv-based ac microgrid cluster,” Electric
Power Systems Research, vol. 173, pp. 71 – 85, 2019.
[29] R. Faranda, M. Gallina, and D. T. Son, “A new simplified model of
double-layer capacitors,” in 2007 International Conference on Clean
Electrical Power, May 2007, pp. 706–710.
[30] V. Musolino, L. Piegari, and E. Tironi, “New full-frequency-range super-
capacitor model with easy identification procedure,” IEEE Transactions
on Industrial Electronics, vol. 60, no. 1, pp. 112–120, Jan 2013.
Matej Krpan (S’17) received his bachelor’s and
master’s degrees in electrical power engineering
from University of Zagreb, Faculty of electrical
engineering and computing in 2014 and 2016, re-
spectively. He is currently pursuing a PhD degree in
electrical engineering at the Department of Energy
and Power Systems, University of Zagreb, Faculty
of electrical engineering and computing.
His research interest include power system dy-
namics, stability and control, integration of large-
scale renewable energy sources, grid frequency sup-
port by wind power plants and low-inertia power systems.
Igor Kuzle (S’94–M’97–SM’04) is a Full Professor
and the Head of the Department of Energy and
Power Systems at the University of Zagreb Faculty
of Electrical Engineering and Computing. His sci-
entific interests include problems in electric power
systems dynamics and control, unit commitment,
maintenance of electrical equipment, as well as
power system analysis, smart grids and integration
of renewable energy sources.
He serves in 10 international journals as an as-
sociate editor or a member of editorial board. Igor
Kuzle published three books and more than 200 journal and conference papers
including technical studies for utilities and private companies. He was the
project leader for more than 60 technical projects for industry and electric
power companies.