Journal of Energy Storage 15 (2018) 23–31

Contents lists available at ScienceDirect

Journal of Energy Storage

journal homepage: www.elsevier.com/locate/est

Equivalent circuit model parameters extraction for lithium ion

batteries using electrochemical impedance spectroscopy
Petr Vyroubal* , Tomáš Kazda
Department of Electrical and Electronic Technology, Faculty of Electrical Engineering and Communication, Technická 10, 616 00 Brno, Czech Republic

A R T I C L E I N F O A B S T R A C T

Article history:
Received 1 September 2017 Numerical modelling is the method by virtually testing and verifying the functionality of a specific
Received in revised form 4 October 2017 product or component. The primary goal is to get approximate results of how the system behaves in a
Accepted 11 October 2017 given time and environment. We are able to accept a certain numerical error from a real experiment, thus
Available online 8 November 2017 significantly speeding up part of the development of the device. In the field of electrochemistry of
lithium-ion accumulators, several variants of numerical models have been proposed that yield
satisfactory results in modelling certain physical fields of these batteries (electric field, temperature field,
current field).
© 2017 Elsevier Ltd. All rights reserved.

One model is the battery model with its equivalent circuit,
which is represented by passive components, namely a resistor and
a capacitor. This is a model that describes the internal behaviour of
a battery by RC circuit. However, the problem is to find or measure
these parameters. One way to get the parameters of this spare
circuit is by means of electrochemical impedance spectroscopy
(EIS).
This article deals with the application this method to obtain the
parameters of the equivalent battery circuit and comparing the
charging and discharging curves of a specific battery with the ECM
model.
1. Introduction
Electrochemical Impedance Spectroscopy (EIS) allows the
observation of the electrochemical and physical processes occur-
ring in the monitored system. This is a very sensitive method in
which the interpretation of results is rather complex due to its high
sensitivity. The total measurement result consists of a mixture of
responses from all the events that occur in the measured system,
which often leads to ambiguity of the result [1].
For a good interpretation of the results, it is necessary to
understand the physical and chemical essence of the measured
system. Then compare the results with other analytical methods,
both electrochemical and physical. The basic principle of this
method is to set up a small sinusoidal AC voltage of a given
frequency with offset of a certain value of the polarization voltage
(mostly between 1 and 10 mV) [2].
2. EIS basics
The impedance value of the system is calculated using the ratio
of alternating voltage and alternating current. This is composed of
amplitude and phase shift values that are dependent on the set
frequency. Thanks to this, the dependence of the system
impedance on the frequency is obtained. The individual chemical
elements that make up the measured system differ from each other
by the time constant of the polarization, which causes the
individual components of the impedance to vary with frequency
variations [3] and [4].
00
These components are: Z0 which is the real impedance, jZ the
imaginary impedance and |Z|  the total impedance. The ratio of
impedances "Z0 /Z00 / Z is called the loss factor tgd. EIS requires stable
system, which is difficult in practice due to the possibility of
influencing the measurement by external noise, the temperature
change and contamination. It is necessary to perform the
measurement in the moment when the electrochemical response
of the system is stable. The impedance is described by the formula
according to [5–7]:
00
Z ¼ Z0 þ jZ ð1Þ
Alternatively, using polar coordinates:

Z ¼ jZjej’ ¼ jZjcos’ þ jjZjsin’ ð2Þ

* Corresponding author. Thanks to these parameters, we are able to describe the

E-mail address: vyroubal@feec.vutbr.cz (P. Vyroubal). electrochemical process with the help of spare electrical circuits.

https://doi.org/10.1016/j.est.2017.10.019
2352-152X/© 2017 Elsevier Ltd. All rights reserved.
24 P. Vyroubal, T. Kazda / Journal of Energy Storage 15 (2018) 23–31

These circuits consist of several basic elements, passive or active,

which can be connected serial, parallel or mixed. These type are
resistor R, capacitor C and several others. In these circuits, the
capacitor with polarization processes and a resistor with
conductivity of the sample are related.

2.1. Resistor R

This is an element that is reflected in the impedance chart, the

so-called Nyquist graph, only by the real axis shift (see Fig. 1). This
shift corresponds to the value of the real resistance. The real
resistance may by dependent on the conductivity of used
Fig. 2. Nyquist graph for capacitor.
electrolyte. The overall impedance of the model represented by
the resistance is:
Z¼R ð3Þ

2.2. Capacitor C

Unlike resistivity, the capacity is frequency dependent and the

phase change is 90 (see Fig. 2). This is an element that appears in
the Nyquist graph in the negative values of the imaginary axis as a
half-line. With an increasing frequency, the value decreases
towards zero at an infinite frequency. The capacitor corresponds
to two layers at the electrode-electrolyte interface and its capacity
is inversely proportional to the thickness of this double layer. The
overall impedance of the model represented by the capacitor is: Fig. 3. Nyquist graph for inductor.

1
Z¼ ð4Þ
jvC

2.3. Inductor L

Inductance, like capacitance, is frequency dependent and also

changes phase, but unlike capacity + 90 (see Fig. 3). This element,
like the capacitor, will appear in the Nyquist graph as a half-line,
but in the positive values of the imaginary axis. Inductive
behaviour may be due to a different current distribution and a
potentiostatic error, indicating an error in the EIS measurement.
The total impedance of the model represented by the inductor is:
Fig. 4. Nyquist graph for CPE.
Z ¼ jvL ð5Þ
information on the surface structure. The overall impedance of
2.4. Constant phase element (CPE) the CPE model is therefore:
1
It is an element that is similar to a capacitor but, unlike it, the ¼ Y ¼ Q 0  ðjvÞn ð6Þ
Z
phase is changed at a different angle than 90 (see Fig. 4). In the
Nyquist chart, it is again represented as a semi-line which forms where Q is numerically equal to admittance at v = 1 rad/s refers to
the angle a with the real axis. As with a capacitor with increasing the unit [S  sn ], n is the slope of the semi-line from 0 to 1. For n = 1 it
frequency, the value decreases towards zero at infinite frequency. is a capacitor and for n = 0 it is a resistor.
CPE therefore behaves as a non-ideal capacity, it provides
2.5. Warburg W

Warburg impedance is a special case of CPE, occurring when the

Fig. 1. Nyquist graph for resistor.
angle a in the Nyquist chart is equal to 45 (see Fig. 5). This element
is used to model the diffusion of ions. It consists of frequency-
dependent CW capacity and RW resistance in a serial combination.
The total impedance of model W corresponds to CPE when n = 0.5:
1
¼ Y ¼ Q 0  ðjvÞ0:5 ð7Þ
Z
A simplified overview of the processes taking place in the
lithium-ion battery captured by EIS is shown in Fig. 6. These
storages are divided into three parts according to the speed at
which they proceed. Each part is assigned a corresponding portion
of the progress in the Nyquist chart [8] and [9].
 P. Vyroubal, T. Kazda / Journal of Energy Storage 15 (2018) 23–31 25

Fig. 7. The Equivalent circuit model (ECM) corresponding to the representation of

Li-ion mass transfer phenomena which occur in lithium ion battery electrodes and
their respective Nyquist plots (Fig. 6).

equivalent circuit, by the constant phase element CPE [12]. At

lower frequencies, the impedance becomes purely capacitive,
because the lithium ions can no longer diffuse into the material and
Fig. 5. Nyquist graph for Warburg.
the capacity is thus formed by the ions surrounding the electrode
material [13].
The first action that occurs (see Fig. 6) is the diffusion within the
electrolyte. The frequency at which this happening is greater than
3. Mathematical physical model
105 Hz, which implies that this action is very fast and takes place in
the order of microseconds. In the Nyquist chart, this plot is simply
We are used the MSMD model (Multi-Scale Multi-Domain),
displayed as a displacement along the axis-x, which implies that it
which comprises for each layer different physical properties.
has no capacitive part and the impedance is pure resistance. This
MSMD model is fully implemented in ANSYS Fluent battery
resistance is then represented in the model by the R1 resistance
module. This MSMD model simulate the electrochemical behav-
(see Fig. 7 below). The value of this resistance corresponds to the
iour at the particle, electrode, and cell domains, with appropriate
distance of the first half-circle point from the y-axis [10].
coupling between the scales [10], [11] and [12]. This allows us
The charge transfer between the electrolyte and the electrode is
multiple cell design tested. We are able to simulate the
represented in the Nyquist graph by a semicircle. This part
temperature field in 3D, thermal and electrical fields, which are
corresponds to the parallel combination of R2 and C1 [11]. The
calculated in the active core the battery according to the following
charge transfer occurs approximately in the frequency range of 100
differential equations:
to 106 Hz. This process is slower and takes place in the time range of
10 ms to 1 s. The distance between the first and the last point of the @rC p T
semicircle corresponds to the diffusion occurring in the electrode  r  ðkrTÞ ¼ q ð8Þ
@t
material.
resp.
The last part is represented by small frequencies below 100 Hz.
At these low frequencies, the impedance is formed by the @rC p T
capacitance, in this case the non-ideal, represented in the  r  ðkrTÞ ¼ s þ jrFþ j2 þ s  jrF j2 þ q_ ECh þ q_ short ð9Þ
@t

Fig. 6. Schematic representation of Li-ion mass transfer phenomena which occur in Liion battery electrodes and their respective Nyquist plots. [13].
26 P. Vyroubal, T. Kazda / Journal of Energy Storage 15 (2018) 23–31

and
r  ðs þ r’þ Þ ¼ j ð10Þ dðsocÞ IðtÞ
¼ ð17Þ
dt 3600Q Ah
resp.
For a given battery, the open circuit voltage, resistors
r  ðs þ r’þ Þ ¼ ðjECh  jshort Þ ð11Þ
resistances, and capacitors capacitances are functions of the
and battery state of charge (SOC) (it was measured by 0, 25, 50, 75 and
100 % SOC). These functions could be expressed in The fifth order
r  ðs þ r’þ Þ ¼ j ð12Þ Polynomial form:
resp.
Rs ¼ rs0 þ rs1 ðsocÞ þ rs2 ðsocÞ2 þrs3 ðsocÞ3 þrs4 ðsocÞ4
r  ðs  r’ Þ ¼ ðjECh  jshort Þ ð13Þ þ rs4 ðsocÞ5 ð18Þ
Cp is specific heat capacity, T is thermodynamic temperature, k is
thermal conductivity, F+ and F is the cathode- and anode-side
electric potentials, t is time, s+ a s are conductivities of positive R1 ¼ r10 þ r11 ðsocÞ þ r12 ðsocÞ2 þr13 ðsocÞ3 þr14 ðsocÞ4
and negative electrode, s þ a s  are potentials of positive and þ r14 ðsocÞ5 ð19Þ
negative electrode, jECh a q_ ECh volumetric current and heat
generated by electrochemical reactions, respective jshort a q_ short
volumetric current and heat generated by internal short circuit (in R2 ¼ r20 þ r21 ðsocÞ þ r22 ðsocÞ2 þr23 ðsocÞ3 þr24 ðsocÞ4
normal conditions are these variable zero values). þ r24 ðsocÞ5 ð20Þ
Electrochemical sub-models relate the local current density to
the potential. The cell model couples sub-models to thermal and
electrical fields within the cell, integrates over multiple electrode- C1 ¼ c10 þ c11 ðsocÞ þ c12 ðsocÞ2 þc13 ðsocÞ3 þc14 ðsocÞ4
pairs [16]. þ c14 ðsocÞ5 ð21Þ
The internal model is calculated by means of the Newman,
Tiedemann, Gu, and Kim model, which is simple semi-empirical
electrochemical model. It was proposed by Kwon [17] and has been
used by [14] and [15]. C2 ¼ c20 þ c21 ðsocÞ þ c22 ðsocÞ2 þc23 ðsocÞ3 þc24 ðsocÞ4
According to these equations are the potential, temperature and þ c24 ðsocÞ5 ð22Þ
current fields calculated by means finite volume method (FVM).
In the Equivalent Circuit Model (ECM), battery electric
behaviour is represented by an electrical circuit. The numerical Voc ¼ voc0 þ voc1 ðsocÞ þ voc2 ðsocÞ2 þvoc3 ðsocÞ3 þvoc4 ðsocÞ4
model was prepared using ANSYS FLUENT system. It has adopted þ voc4 ðsocÞ5 ð23Þ
the six parameter ECM model following the work of Chen [5]. In
this model, the circuit consists of three resistors and two capacitors The energy source term comes from the contribution of Joule
(see Fig. 8). The voltage-current relation can be obtained by solving heating, electrochemical reaction heating, and the entropic
the electric circuit equations: heating:

VðtÞ ¼ VOCV ðsocÞ þ V1 þ V2  Rs ðsocÞIðtÞ ð14Þ I

jechem ¼ ð24Þ
Volume
and
dV1 1 1  
¼ V1  IðtÞ ð15Þ I dVOCV
dt R1 ðsocÞC1 ðsocÞ C1 ðsocÞ qechem ¼ VOCV  ð’þ  ’ Þ  T ð25Þ
Volume dT

where Volume denotes the battery volume, I is the current, and

dV2 1 1 VOCV is the open circuit voltage.
¼ V2  IðtÞ ð16Þ
dt R2 ðsocÞC2 ðsocÞ C2 ðsocÞ The global parameters of thermal conductivity, specific heat
capacity, electrical conductivity, density and length parameters of
the battery were determined [18]. For calculation of effective
values of material density, thermal conductivity and specific heat
capacity, the following equations are used:
p p n n
0; 5xpc dc þ xpe de þ xs ds þ xne de þ xnc dc
xeff ¼ ð26Þ
dtotal
and

dtotal ¼ 0; 5dpc þ dpe þ ds þ dne þ 0; 5dnc ð27Þ

where xeff is the effective value of the given characterization of the

Fig. 8. Equivalent circuit model (ECM) which is used for numerical modelling the
lithium ion battery [16].
material (density, specific heat capacity, thermal conductivity) and
it is function of the material thickness. Indexes c, e and s denote
current collector, electrode and separator and indexes p and n
denote positive or negative electrode. dtot is the total thickness of
the specimen (see Fig. 9 and Table 1).
 P. Vyroubal, T. Kazda / Journal of Energy Storage 15 (2018) 23–31 27

Fig. 9. The process of simplifying the layers of a real battery into a homogeneous MSMD model [20].

Table 1
Dimensions and properties of the measured battery.

Zone Pc Pe S Ne Nc Total
d [m] 8.40E-04 2.44E-03 9.24E-04 4.28E-04 1.05E-04 8.52E-03
Density [kg m3] 2700 1500 492 2660 8900 8.04E+02
Specific Heat capacity [J (kg1 K1] 903 1260 1978 1437 385 6.94E+02
Thermal conductivity [W (m1 K1)] 238 1.48 0,334 1,04 398 1.47E+01
Electrical conductivity [S m1] 3.89E+07 1.00E-06 1.00E+04 6.33E+07 r_p 1.92E+06
r_n 3.90E+05

These equations are valid for electrical conductivity:

p p
0; 5s pc dc þ s pe de
sp ¼ ð28Þ
dtotal
and
n n
0; 5s nc dc þ s ne de
sn ¼ ð29Þ
dtotal
The simulation model accepts the Multiscale physics from sub-
micro scale to battery dimension scales [19].
Table 1 contains the measured values of the thicknesses of the Fig. 10. Dismantled KOKAM battery.
individual battery layers after disassembling and the values of
density, specific heat capacity, thermal conductivity and electrical
conductivity calculated using the equations 26–29. Materials,
which was used for calculation of battery physical properties was:
anode  Aluminium, cathode  Copper, separator  polyethylene,
cathode material  LiNi0.33Mn0.33Co0.33O2 and anode material 
Carbon.
Where Pc is positive current collector, Pe is positive electrode, S
is the separator, Ne is negative electrode, Nc is negative current
collector.

4. Experimental measurement of charging and discharging

curves Fig. 11. Block diagram of measuring cell.

Two 4 Ah batteries SLPB9543140H5 made by KOKAM company

were used for electrochemical testing and preparation of physical-
mathematical model.
The first battery was disassembled and the dimensions of its
parts have been measured and used for the model preparation (see
Fig. 10).
The second battery was connected to the potentiostat Biologic
Science Instruments1 VMP31 with a booster (Block diagram is
represented in Fig. 11).
CCCV was always used as a battery charging method. The
charging current was set to 0.2C (800 mA); cut-off voltage to 4.2 V.
The potentiostat was switched to the CV mode upon reaching 4.2 V
and charging continued until the current dropped below 200 mA.
Subsequently, the tested battery was discharged by currents of
0.2C (0.8 A), 0.5C (2 A), 1C (4 A), 2C (8 A) and 4C (16 A). Discharge
characteristics were always recorded during discharging.
28 P. Vyroubal, T. Kazda / Journal of Energy Storage 15 (2018) 23–31

Fig. 12. Nyquist diagram of impedance measuring the KOKAM battery by different SOC.

Table 2 Table 3
The parameters of equivalent circuit model by discharging. The parameters of equivalent circuit model by charging.

SOC [%] Voc [V] Rs [V] R1 [V] C1 [F] R2 [V] C2 [F] SOC [%] Voc [V] Rs [V] R1 [V] C1 [F] R2 [V] C2 [F]
1 4.195754 0.01889 0.001540 1.241 0.002314 4.626 1 4.19785 0.01879 0.002278 4.6990 0.001542 1.266
0.75 3.920000 0.01889 0.002425 4.378 0.001645 1.193 0.75 3.94147 0.01878 0.002328 3.9170 0.001466 1.259
0.5 3.790000 0.01889 0.003286 4.734 0.001850 1.181 0.5 3.80990 0.01885 0.003221 4.4160 0.001751 1.189
0.25 3.678890 0.01895 0.002222 1.281 0.007209 5.312 0.25 3.72448 0.01889 0.002096 1.2520 0.006020 5.118
0 3.008480 0.01902 0.002970 1.591 0.021300 4.857 0 3.02384 0.01906 0.003355 1.5710 0.022690 4.880

Fig. 13. The polynomial functions obtained by discharging (dotted line  fitted, full line  measured).
 P. Vyroubal, T. Kazda / Journal of Energy Storage 15 (2018) 23–31 29

Fig. 14. The polynomial functions obtained by charging (dotted line  fitted, full line  measured).

5. Parameters extraction
The values of the equivalent circuit parameters were extracted
from EIS measurements using the fitting program that is included
in the Biologic booster. The fitted circuit was chosen according to
the mentioned equivalent circuit in Fig. 8. For the results of these
measurements, it is important in which state of charge the battery
is at the beginning of the measurement.
The Nyquist diagram (see Fig. 12) shows the impedance
measurement at different battery SOC using the EIS method in
laboratories.
The obtained curves were then fitted to an equivalent circuit
identical to the circuit shown in Fig. 8 above. Thanks to this
imitation, the parameters R and C (Tables 2 and 3) were obtained
and then were used in Eqs. (18)–(23) above. These dependencies
were then inserted into the graphs and interleaved by the 5th order
polynomial function. In this way, we have obtained equations
describing the behaviour of the battery in different charging/
discharging state (Fig. 13).
The battery circuit parameters obtained in this way were
inserted into the numerical model. The charging and discharge
curves of the battery were modelled and compared with the curves
measured at the laboratory measurement.
6. Results and discussion
The actual comparison of the discharge characteristics is shown
in Fig. 14. It is obvious that the curves have the same course and are
slightly offset. This 0.1 V spacing is not essential to affect too much
the properties of the simulated battery and can be considered

Fig. 15. Comparison of measured and simulated discharge curves of the battery.
30 P. Vyroubal, T. Kazda / Journal of Energy Storage 15 (2018) 23–31

Fig. 16. Comparison of measured and simulated charge curve 0.2C of the battery.

satisfactory, particularly for copying the same waveform as the
measured battery. It is necessary to realize that the battery
represents an ideal RC circuit. Some parameters represented by the
Nernst and Butler-Volmer equations are neglected.
Discharging by 4C can be definitely considered as fast
discharging for example: Maximum NCR18650B discharge limit
is 2C under continuous current load or KOKAM SLPB7570180 10.6
Ah with maximal continuous discharge 3C.
The charging characteristics were measured as well as the
accuracy of the previous measurements. To simulate battery
charging, input data was used as described above. The charging
current was set to 0.2C, which ensures a slow but still kindly charge
of the battery (see Fig. 15).
Numerical model allows us to display some physical fields.
Fig. 16 shown the potential distribution at the positive and
negative electrode at the end of discharging (Fig. 17).
7. Conclusion
The method presented in this paper, which is suitable for the
extraction of lithium ion battery parameters, is called electro-
chemical impedance spectroscopy. In assessing the suitability of
this method it can be stated that, within the tolerance of a certain
degree of inaccuracy, it has proved to be an effective tool for
obtaining a battery model at the RC circuit level. The use is
unambiguously where we do not know the exact properties of the
materials in which the batteries are composed. Conventional
methods of battery microstructure modelling are based on often
complex electrochemical models. These models are loaded with
errors already at the time of determining some of the coefficients
of the given materials.
When modelling a particular device (electric vehicle, electric
train, etc.) when looking at a battery in a microstructure is not
important, replacing the battery with the model thus obtained is
sufficient. In principle, we are able to obtain a solution for any
discharge current by approximation. However, as soon as we
exceed a certain amount of discharging current, the endothermic
reaction battery begins to run. This model is unable to capture it.
But the temperature field can be captured. Therefore, it is also
important to monitor this parameter and not exceed the
manufacturers declared permissible values.
The presented process of numerical modelling was compared
with real-time measurements, which yielded satisfactory model
match and exact results.

Fig. 17. Potential distribution at the positive (left) and negative (right) electrode.
 P. Vyroubal, T. Kazda / Journal of Energy Storage 15 (2018) 23–31
Acknowledgments
This research work has been carried out in the Centre for
Research and Utilization of Renewable Energy (CVVOZE). Authors
gratefully acknowledge financial support from the Ministry of
Education, Youth and Sports of the Czech Republic under NPU I
programme (project No. LO1210)
References
[1] P. Vyroubal, J. Maxa, T. Kazda, J. Vondrak, ECS Trans. 48 (2014) 289.
[2] P. Vyroubal, T. Kazda, et al., Int. J. Electrochem. Sci. 11 (2016) 1938–1950.
[3] P. Vyroubal, Tomáš Kazda, R. Bayer, FEM Model of the Lithium Ion Battery Nail
Test ECS Trans. 74 (1) (2016) 71–75.
[4] L. Cai, R.E. White, J. Power Sources 196 (14) (2011) 5985–5989.
[5] M. Chen, G.A. Rincon-Mora, Accurate electrical battery model capable of
predicting runtime and I-V performance, IEEE Trans. 21 (2) (2006) A154–A161.
[6] ANSYS, Inc. ANSYS Fluent Advanced Add-On Modules. 16.0. Canonsburg, PA
15317, 2015.
31
[7] A.A. Andreev, et al., Simulation of lithium battery operation under severe
temperature conditions, Procedia Eng. 129 (2015) 201–206.
[8] D. Linden, T.B. Reddy, Handbook of Batteries, McGraw-Hill, Michigan, 2002.
[9] N. Nitta, F. Wu, J.T. Lee, G. Yushin, Mater. Today 18 (2015) 252.
[10] B. Scrosati, J. Solid State Electrochem 15 (2011) 1623.
[11] M.S. Whittingham, Chem. Rev. 104 (2004) 4271.
[12] R.J. Brodd, M. Doeff, in Batteries for Sustainability, Springer, New York, p. 5.
[13] N. Williard, W. He, C. Hendricks, M. Pecht, Energies 6 (2013) 4682.
[14] Y. Fernández Pulido, C. Blanco, D. Anseán, V.M. García, F. Ferrero, M. Valledor,
Measurement 106 (2017) 1–11.
[15] M. Itagaki, K. Honda, Y. Hoshi, I. Shitanda, F. Ferrero, M. Valledor, J. Electroanal.
Chem. 737 (2015) 78–84.
[16] Tiedemann, S. Newman, Gross Editor. “Battery Design and Optimization.”
Proceedings, Electrochem. Soc. Princeton, NJ, 79-1, 39, (1979).
[17] K.H. Kwon, et al., J. Power Sources 163 (2006) 151.
[18] U.S. Kim, et al., J. Electrochem. Soc. 158 (2011) A611.
[19] U.S. Kim, et al., J. Power Sources 180 (2008) 909.
[20] H. Gu, Mathematical Analysis of a Zn/NiOOH Cell, J. Electrochemical Soc.
Princeton, NJ. 1459-1, 464, (1983).