See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/235602059

A Cell Level Model for Battery Simulation

Conference Paper · November 2012

CITATIONS READS

19 50,907

5 authors, including:

Suguna Thanagasundram Raghavendra Arunachala

Temasek Polytechnic Continental TAEM GmbH
21 PUBLICATIONS   193 CITATIONS    14 PUBLICATIONS   173 CITATIONS   

SEE PROFILE SEE PROFILE

Karl Löffler Tanja Teutsch

Technische Universität München Universität Stuttgart
13 PUBLICATIONS   104 CITATIONS    12 PUBLICATIONS   100 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

ESPEN - Potentials of electrochemical storages in power grids in competition to further technologies and system solutions View project

Energy Storage Systems View project

All content following this page was uploaded by Suguna Thanagasundram on 17 May 2014.

The user has requested enhancement of the downloaded file.

 European Electric Vehicle Congress
Brussels, Belgium, 20th – 22nd November 2012

A Cell Level Model for Battery Simulation

Suguna Thanagasundram1, Raghavendra Arunachala1, Kamyar Makinejad1, Tanja Teutsch2,
Andreas Jossen2
1
TUM CREATE Centre of Electromobility, 1 CREATE Way, #10-02, CREATE Tower, Singapore 138602
suguna.thanagasundram@tum-create.edu.sg
2
Institute for Electrical Energy Storage Technology, Technische Universität München, Arcisstrasse 21, 80333 Munich,
Germany

Abstract
An equivalent circuit model is the most common and straight-forward way of representing the dynamic
behaviour of a lithium-ion battery. In literature, many examples of circuits are proposed and various
techniques are also discussed for parameterisation of the models. In this paper a second order equivalent
circuit is proposed and the parameters identification method by Hybrid Power Pulse Characterization
(HPPC) testing is described. The modelling and parameter identification process is done in the
Matlab/Simscape environment. This non-linear model encapsulates the dynamic electrical behaviour of a
typical automotive cell. A validation process is carried out to benchmark the voltage errors between
estimated voltage profile of the battery cell model and actual cell measurements. The comparison between
measurement and simulation shows a good accordance. The current pulse technique is also presented to
verify the ohmic resistance values obtained by the optimisation process in the parameters identification
method. A study is also conducted to investigate how cell chemistry affects the proposed model.

Keywords: Battery Electric Vehicle (BEV), lithium battery, modelling, battery model, internal resistance

optimize the energy usage and prolong its useful

1 Introduction life.
Researchers around the world have developed a
The ever increasing demand for limited fossil wide variety of techniques with varying degrees of
fuels and growing concerns over the CO2 complexity to model the battery. Battery models
emission have spurred worldwide interests in can be classified into electrochemical models
developing alternative energy and storage (chemistry-based), mathematical models and
systems, particularly for Electric Vehicles (EV) electrical models (circuit-based). Chemistry-based
and Hybrid Electric Vehicles (HEV) application. models are derived from porous electrode theory
Lithium-ion battery is a promising power source and concentrated solution theory proposed by
for EVs and HEVs due to its high energy density, Newman and Tiedemann [1] and Doyle et al. [2]
long cycle life and low self-discharge. The which mathematically describe the
integration of lithium-ion batteries in EV and charge/discharge processes and transport kinetics
HEV applications require a dynamic model in the solid and electrolyte phases in the battery as
which predicts the performance of the battery at simplified 1D spatial structures. These models
different operating conditions in order to have the ability to accurately predict the
performance of battery. However they require a

EEVC European Electric Vehicle Congress 1

detailed understanding of complex impedance-based models are developed by fitting
electrochemical processes, often involving a with impedance spectra. Circuit-based models are
large number of algebraic equations and state recommended for thermal studies within the range
variables to model the battery behaviour. of fitting. Circuit-based models are real-time
Chemistry-based models are commonly implementable [10] and can be run on Hardware-
developed using Finite Element Analysis in-Loop (HIL) platforms. Hence they are useful for
Software such as Ansys and Comsol [3, 4]. In BMS development and serve as a part of vehicle
cell design studies, chemistry-based models are level simulation studies for battery pack sizing and
used to determine optimal parameters of the cell range estimation calculations. The accuracy of
such as the form factor or electrode thickness. circuit-based models lies between the chemistry-
They can also be used at the pack level to aid in based and mathematical-based models.
the design of the battery system or to determine
rate of coolant flow inside the modules to 2 Second Order Equivalent
maintain the cells at a safe operating temperature
within the limits of a certain thermal band as Circuit Model
required by the vehicle application. The One example of circuit based battery model is the
advantage of chemistry-based models is that they Second Order Equivalent Circuit (SOEC) model.
can be used to predict the behaviour of the cells The schematic of the model is shown in Fig.1.
beyond the range of experimental data e.g. There are 6 components in the model: VOC, Ri
thermal runaway conditions. The main and the two parallel polarization RC network
drawbacks of chemistry-based models are that combinations, namely, R1, R2, C1 and C2. Each
they are computationally time-consuming and component is representing a different aspect of the
very complex hence not applicable in EV battery.
application due to the real-time demands of
Battery Management Systems (BMS). C1(SOC,T,I) C2(SOC,T,I)

Mathematical-based models are developed R1(SOC,T,I) R2(SOC,T,I) Ri(SOC,T)

primarily using the Shepherd relation and I
modifications to the Shepherd model [5, 6]. Fast Time Slow Time Ohmic
These modifications usually consist of adding Constant Constant Resistance V
and altering terms of the original model to relax VOC(SOC,T)
assumptions behind the Shepherd model. In
mathematical oriented models, the parameters are
extracted by curve fitting the manufacturer’s Figure 1: Second Order Equivalent Circuit (SOEC)
discharge curves. A classic example of the model of a battery cell.
mathematical model of a battery is the battery VOC is the Open Circuit Voltage and one of the
blockset in SimPowerSystems Toolbox in Matlab most important parameter of a battery. Ri is the
[7]. Mathematical-based models are simple but ohmic or DC resistance. Ri is representative of the
application dependent and not accurate enough to internal resistance in the cell, similar to terminal
reproduce fast varying battery voltage dynamics. resistance and electrode resistance and is
responsible for the immediate voltage drop or rise
However, circuit-based models are very useful when the battery is being discharged or charged.
and simple, because the complex electrochemical R1, C1, R2 and C2 are the two RC parallel
processes can be transformed into electrical polarization elements, responsible for the transient
circuit elements which describe the battery response of the battery [11, 12]. R1 and C1 describe
kinetics [8, 9]. Circuit-based models use the fast dynamics in the battery depicting surface
electrical components such as voltage sources, effects on the electrodes and reaction kinetics. R1
resistors, capacitors and inductors to encapsulate is the charge transfer resistance and C1 represents
the battery behaviour. Circuit-based models can the electrochemical double layer capacitance. R2
be further classed into two main categories: and C2 represent the slower dynamics of the cells
Thevenin-based models and impedance-based in order of hours. They are more representative of
models [10]. The Thevenin-based models are the diffusion processes in the electrolyte and active
constructed by curve-fitting a set of experimental material. All these parameters are functions of the
voltage and current measurements over a range State-of-Charge (SOC) of the battery, temperature
of battery operating conditions whereas the (T) and current (I).

EEVC European Electric Vehicle Congress 2

 Figure 2: Implementation of the SOEC model in Simscape.

2.1 Non-linear dynamic SOEC Model 3.2 Battery Cells used for the study
The SOEC model of the cell is implanted in Tab.1 shows the specification of Lithium-ion cells
Matlab/Simscape [13] as shown in Fig.2. of three different chemistries considered for the
Simscape, which is one of the toolboxes in the study. They are Lithium Cobalt Oxide (LCO),
Mathworks suite of products, is a flexible, Lithium Manganese Oxide (LMO) and Lithium
acausal and an object-oriented tool for modelling Iron Phosphate (LFP). LCO and LMO have higher
and simulating complex integrated multi-physics, theoretical capacity [18], compared to LFP due to
multi-engineering systems. The Simscape the higher operating voltage. LFP cells are
platform was chosen because it is a physical characterized by high intrinsic safety and that
modelling tool with multi-domain capability. makes them suitable for high power applications.
This was a useful feature as one of the intended
objectives of the work is to couple the electrical
and thermal properties of the cell behaviour. The 3.3 Platform of Implementation
SOEC model in Simscape is intuitive and easy to The tests were carried out on BaSyTec CTS
understand as it looks just like a circuit in real- battery tester at TUM CREATE Centre for
life and the solution to the model is found by Electromobility in Singapore. The battery tester
symbolic reduction techniques. Additional provides an automatic and dynamic current range
components like resistors or capacitors can easily selection allowing non-interrupted current flow in
be added to the circuit. The preceding sections the constant voltage mode when current decreases
describe how the values of VOC, Ri, R1, R2, C1 with charging time.
and C2 were determined for the battery SOEC Ideally the power pulse characterisation should be
model. Fig.2 shows purely the electrical model of done at 3C charge/discharge rate but since the
the SOEC cell. The thermal behaviour of the cell battery cyclers that were available for testing were
has also been modelled in a similar way using the limited to 5A, the maximum charging and
thermal library in Simscape but it is not shown in discharging was done at 5A. The cell rests for 3
this paper. minutes and then discharged for 6 minutes at 1 C
rate (which is 2.3A in this case) to achieve a 10%
3 Experiment Description decrease in SOC. Afterwards the cell rests for 1
hour to reach equilibrium before the next HPPC
3.1 Current Pulse Techniques test profile begins. The partial discharge-HPPC-
The parameterization method used in the study is rest phase cycle is repeated at decrements of 10%
similar to the one described in [14, 15]. It is the of SOC till the cell reaches 10% SOC. The process
current pulse technique, commonly known as the is repeated in the charging direction as well. Fig.4
HPPC test and is described in detail in the shows typical current, voltage and Ah plots of cell
FreedomCAR Battery Test Manual [16, 17]. This type 1 obtained from the HPPC test profile. The
method is used to calculate the dynamic nominal voltage and capacity of the cell are 3.3V
properties of a battery. The battery is charged and nominal capacity of 2.3 Ah as stated in Tab.1.
and discharged under a controlled condition and The conventions used for charging current are
the terminal voltage, current and temperature positive and discharging current is negative as
measurements are monitored. A HPPC test shown in Fig.4.
profile is shown in Fig.3.

EEVC European Electric Vehicle Congress 3

 Figure 3: Hybrid Pulse Power Characterization (HPPC) test profile.

Table 1: Specification of Lithium-ion cells.

Cell Geometry Size Cell Weight Nominal Nominal

(mm) Chemistry (g) Voltage Capacity
(V) (m Ah)
Cell Type 1 Cylindrical 26 (D) Lithium Iron 70 3.3 2300
65 (H) Phosphate

Cell Type 2 Cylindrical 18 (D) Lithium Cobalt 45 3.6 2250

65(H) Oxide

Cell Type 3 Cylindrical 18(D) Lithium 44 3.6 2200

65(H) Manganese
Oxide

The Ah curve was recorded for balance counting. voltage equilibrium. The measurement starts with
Tab.2 summarizes the specifications of the a complete charged cell. The HPPC test profile
BaSyTec CTS battery tester. Four wire consists of 10s pulse discharge at -5A, followed by
measurement method is used to connect the test 10s charge pulse. After each pulse
cells to the battery tester. All experiments on the discharge/charge, there is a pause for 3 minutes.
cells are conducted at room temperature. The Then the cell is discharged at 1 C-Rate to remove
cells are preconditioned by conducting two 10% of its capacity. The cell is rested for 1 hour
complete cycles of charge-discharge followed by before the next HPPC test profile begins.
a 4 hour rest period to attain temperature and

EEVC European Electric Vehicle Congress 4

 Table 2 Specifications of the BaSyTec CTS battery the battery cell when there is no current flow and
tester. the cell is in an equilibrium state. The VOC is
Channels per unit 32 determined at different SOCs, normally at the end
Four wire Yes of each rest/relaxation phase (in this case, 1 hour)
measurement when the voltage is at steady-state and the
Maximum Charging 5 temperature of the cell has stabilised.
Current (A)
Maximum -5 The voltage of a cell under current flow is different
Discharging Current from the VOC. The cell shows an increased
(A) voltage during charge and a decreased voltage
Voltage range during +5/0 during discharge. The difference occurs due to the
charging/discharging polarization voltage which can be attributed to two
(V) mechanisms. One mechanism is the overpotential
Time resolution (ms) 20 at the electrodes caused by electro chemical
Current Range 5 A/300 mA/15 mA/1 mA reactions and concentration deviations due to the
(charge/discharge) (Automatic and dynamic transport phenomena in the cell. The other
range switching) mechanism is an ohmic voltage drop across the
current collectors and electrolyte when there is a
current flow. VOC was measured on the candidate
3.4 SOEC Parameters Determination cell type 1 by giving a relaxation of 1 hour after
The VOC is defined as the potential difference each charge/discharge for SOC
between the positive and negative electrodes of increment/decrement (Fig.5).

4 6
U[V] I[A] Ah
3.5
4

2
2.5
I (A)
U (V)

2 0

1.5
-2

-4
0.5

0 -6
0 5 10 15 20 25 30 35 40
Time (hr)
Figure 4: HPPC test profiling for Cell Type 1. Cell is discharged and charged in steps of 10% SOC. Charge and
Discharge pulses are applied at each step of SOC.

EEVC European Electric Vehicle Congress 5

Depending on whether the cell has recently 3.38
undergone a charge or discharge event, the VOC 3.36 VOC when charged
is slightly different for this cell at each SOC VOC when discharged
3.34 Average VOC
indicating that there is some hysteresis and that
3.32

Cell Voltage (V)

the cell has not reached true steady state.
3.3
Although the time period of one hour was not
sufficient to truly reach equilibrium state, the 1 3.28

hour rest period was chosen as a trade-off 3.26

between testing time and measurement accuracy. 3.24

The true VOC is assumed to be in the area 3.22

between the charged and discharged regions and 3.2

10 20 30 40 50 60 70 80 90
for the purpose of the SOEC model, the average SOC (%)
value of VOC is chosen by taking the mean value
Figure 5: Measured VOC vs. SOC during charging and
of charge and discharge VOC. discharging for Cell Type 1.

Other parameters of the SOEC model are

evaluated by analysing the voltage response of 3.4
HPPC pulse discharge and charge. Fig.6 shows 3.35 u0

the voltage response of the cell when -5A pulse 3.3

is applied for a duration of 10s. 3.25

3.2 Ri at Pulse End

Ri at Pulse Start
Voltage (V)

To automate the parameterization process, once a 3.15

profile such as shown in Fig.4 has been obtained, 3.1

3.05
a Matlab script is used to separate the u2
3 u1
measurements of HPPC test done at different 2.95 u3
t1 t2
SOC during charging and discharging. The 10 2.9
second pulse is further split into the short time 150 155 160 165 170 175 180
Time (s)
constant t1, representing the fast dynamics, and
the long time constant t2, representing the slow
dynamics. From each of these measurements, the Figure 6: Voltage response to a 10s -5A discharge pulse
values of R1, C1, R2 and C2 are calculated by in the HPPC test for Cell Type 1.
applying Eq.(1)-(5) to obtain initial estimates of The difference between the measured voltage and
the parameters from the zoomed curves at each simulated voltage is the modelling error and is
steps of the SOC. These values are then fed into defined using Eq.(7). The unconstrained
Eq.(6) to simulate the voltage of the battery cell. optimisation algorithm fminsearch (Nelder-Mead)
provided by Matlab is used to minimise the error
R i  (u0  u1 ) i (1) for each of the analysed pulses and the optimized
R1  (u1  u2 ) i (2) parameters Ri, R1, C1, R2 and C2 at each state of
R 2  (u2  u3 ) i (3) SOC are obtained. These values are fed in the
original SOEC model in Simscape to mimic the
t1  R1C1 (4) cell behaviour. The time constant of the RC
t2  R2C2 (5) parallel circuits in Fig.1 are given by t1 and t2.

t
t1
They differ by an order of magnitude and therefore
Vs (t )  VOC  I (t ) Ri  I (t ) R1 (1  e )  they are representative of the fast and slow
t
(6) transient dynamics of the battery cell.

t2
I (t ) R2 (1  e )
2
3.5 Validation of SOEC model
LSE  (V (t )  Vs (t )) (7)
3.5.1 Validation of internal ohmic resistance
Let Vs(t) be the simulated battery terminal Ri
voltage obtained from Eq.(6) and V(t) be the Besides the automatic calculation of the internal
battery terminal voltage obtained from the ohmic resistance based on the developed SOEC
measurement as shown in Fig.6. cell model and the HPPC test results with the use
of the Matlab script in this paper, other methods

EEVC European Electric Vehicle Congress 6

and techniques are also investigated to calculate with decreasing temperature. In short, with
the internal resistance of the cell. The most lithium-ion batteries, the internal resistance shows
common method of measuring the internal a significant temperature dependency.
resistance of the cell is to use the conventional
current step pulse technique. When using this The Ri values in Fig.7 are also compared with the
method, some important consideration should be Ri values obtained from the optimization scripts
taken into account. In this method the mentioned in section 3.4. Refer to Fig.8 and Fig.9.
determination of the internal resistance is based Both set of values obtained from the different
on the analysis of the immediate jump in the techniques show similar trend with higher
terminal cell voltage due to the current pulses. resistances at very small SOCs, plateau in the
Input current can be in the form of short duration region of 20-80% SOC and slight increase in
charging or discharging pulses. It is also possible resistance at the end at 90-100% SOC. This
to investigate the change in the cell terminal indicates that the optimisation script does work
voltage during switching off the input current to well and does give comparable results. The only
measure the internal resistance. This method is difference was that the optimisation script gives
commonly used for homogeneity tests of the slightly bigger resistances (about 3 mΩ higher) but
lithium-ion cells in in EV and HEV applications. this is a reasonable and acceptable limitation of the
method considering the bigger changes in the
With the BaSyTec battery tester system we have resistance values that can occur if there are
the possibility of measuring the internal measurement inaccuracies or unintended contact
resistance of the cell directly during the test as it resistance at the cell current collector terminals.
is mounted on a 4 wire measuring kelvin probe
cell holder. For the proper recording of these In [21], other methods of calculating the internal
voltage jumps, a sampling rate of 1ms to 5ms resistance of the cell are suggested.
[19] is necessary to increase the voltage Electrochemical Impedance Spectroscopy (EIS),
resolution to capture minute changes in the energy loss method and Joule’s law technique with
voltage. The internal resistance can then be the aid of the Accelerated Rate Calorimeter (ARC)
calculated from Eq.(1). This method was used production of Thermal Hazard Technology (THT)
and the results for 3 cell samples of cell type 1 company are used and compared with the results
(Cell 1, Cell 2 and Cell 3) LFP cells are from the current pulse technique. The
presented in Fig.7. comprehensive tests and experiments are under
investigation and out of the scope of this paper.
For cell type 1, the maximum discharge pulse
can be as high as 120A for a short duration pulse 3.5.2 Comparison of the simulated and
test but because of the limitation of our battery measured voltages
testers, we used +5/-5A to apply To validate the SOEC model, the parameters
charge/discharge pulses to the cell. In Fig.7, Ri extracted from the optimisation script are used to
results are based on the discharge and charge simulate the cell voltage response for cell type 1
current pulses during discharging section of the fed with another HPPC test profile similar to the
HPPC test as well as the discharge and charge one used for the training data. Fig.10 shows the
pulses during the charging section of the HPPC simulated voltage response and the measured
test. Ri was calculated for both the switching on voltage measurement. Fig.11 shows the modelling
and switch off of the charge/discharge pulses. It error. It can be seen that SOEC model can capture
can be seen from the results of Fig.7 that there is the dynamic voltage response of the cell very well
a general increase in Ri values as SOC reduces with a modelling error of approximately less than
for all three samples of the cell type 1. 3% from 10% SOC to 90% SOC. However the
model response is poor at SOCs below 10%
These tests can be conducted at different producing a big modelling error.
temperatures, but in this paper we only present
the results at 20°C. To give an overview about
the approximate values of the internal resistance
at different temperatures, the internal resistance
of a similar tested cell can be as high as 30mΩ at
-20°C to 10mΩ at 35°C [20]. It is understood that
the internal resistance of the battery increases

EEVC European Electric Vehicle Congress 7

 Cell 1 Cell 2
18 18

16 16
R i (m  )

R (m  )
i
14 14

12 12
0 50 100 0 50 100
SOC (%) SOC (%)
Cell 3
18
Discharge Pulse during Discharging
Charge Pulse during Discharging
16
R i (m  )

Discharge Pulse during Charging

Charge Pulse during Charging
14 R at Pulse Start
i

R at Pulse End
i

12
0 50 100
SOC (%)

Figure 7: Internal resistance Ri measurement for Cell Type 1 based on the current charge/discharge pulses at both
current switch on and switch off.

Optimisation Script Optimisation Script

Current Pulse Technique Current Pulse Technique
25 25

20 20

15
Ri (mΩ)

15
Ri (mΩ)

10 10

5 5

0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
SOC (%) SOC (%)

Figure 8: Measured internal resistance Ri values Figure 9: Measured internal resistance Ri values versus
versus Ri values obtained from the optimization script Ri values obtained from the optimization script at 10%
at 10% step SOC for charge pulses. step SOC for discharge pulses.

EEVC European Electric Vehicle Congress 8

At very low SOCs, the cell behaves differently 10
and the voltage decreases exponentially when
there is a discharge current. A detailed study at 5

very low SOCs is required. Either the structure of 0

Percentage Error (%)

the SOEC model has to be changed at very low
-5
SOCs or the method of obtaining the parameters
needs to be further verified. This is scope of -10
future work. Also in the context of the work the
-15
discharge VOC was used in the discharging
direction of the HPPC profile and the charge -20
VOC was used in the charging direction.
-25
This could be done as there was apriori 0 5 10 15 20 25
Time (hr)
knowledge of the charging direction and voltage
simulation was done off-line. However for on- Figure 11: SOEC Modelling Error.
line real-time simulation of the voltage response
This section describes the dynamic behaviour of
when the current measurements are captured as
different cell chemistry Lithium-ion cells at
in a BMS application, this is not possible. In
various SOC as stated in Tab.2. A study was made
most cases, other researchers have resorted to
on VOC behaviour of these chemistries. The VOC
using the average VOC method. But it has been
data is obtained from HPPC tests at every 10%
observed that the use of average VOC can
SOC step when the cells are discharged, followed
produce suboptimal results due to the hysteresis
by a 1 hour complete pause to attain voltage and
effect. A more sophisticated method such as the
temperature equilibrium. Fig.12 shows the
use of enhanced self-correcting cell model
comparison of VOC plotted vs. SOC at 10% SOC
method where there is a hysteresis rate constant
interval step.
as stated in [22] is needed and this is also the
scope of future work
4.3
Cell Type 1
3.6 Investigation on cell chemistry 4.1
Cell Type 2
The cell characterization is strongly dependent 3.9 Cell Type 3
on its chemistry. The criterion for choice of cell 3.7
chemistry depends on the EV application, 3.5
specific energy and specific power, safety, cost
VOC (V)

3.3
and its lifespan. Some chemistry have high
specific energy, low specific and viz., different 3.1
discharge characteristics, different charging 2.9
protocol and their performance is influenced by 2.7
state of charge (SOC) and temperature. A 2.5
universal battery model with same set of 0 10 20 30 40 50 60 70 80 90 100
parameters is not possible to describe the
dynamic behaviour of different cell chemistries. SOC(%)
3.8
Figure 12: VOC vs. SOC for different cell chemistry.
3.6

3.4 LCO and LMO chemistries exhibit sloping

3.2 characteristics of SOC vs. open circuit voltage.
3
However the open circuit voltage behaviour of
Voltage (V)

2.8
LFP shows complete contrast to other two
2.6
chemistries. The open circuit voltage increases
between SOC 0 and 20%. Then it is goes into a
2.4
plateau region between 20 and 90% SOC, again it
2.2
Simulated Voltage increases above 90% SOC. LFP exhibits two phase
2 Measured Voltage
transition during charging and discharging. During
1.8
0 5 10 15 20 25 this process, the electrode potential shows a minor
Time (hr)
change in VOC with respect to SOC change due to
Figure 10: Measured and simulated voltage response constant lithium concentration within the phase
the SOEC model for Cell Type 1. regions [23, 24]. However in LCO and LMO

EEVC European Electric Vehicle Congress 9

chemistries, lithium insertion and de-insertion 4 Conclusions
take place without two phase transition.
An extensive testing method for the development
A SOEC model is developed for LCO and LMO and validation of a SOEC model has been
chemistries by repeating the HPPC test. Three presented. A method to identify the parameters for
sample cells were tested from each of the the SOEC model from characteristic HPPC power
chemistry for statistical reliability of the test measurements is also described. The
measured data. The circuit parameters are fitted performance of the SOEC model is good
in Matlab as described in section 3.4. The results producing very low modelling errors as low as less
of the fitting parameters are then plotted for the than 3% in the SOC range from 20-80%. This is
chemistries described in Tab.2. Fig.13(a)-(e) sufficient to capture the dynamics of most lithium-
show the plot of circuit parameters against SOC. ion cells for most BMS applications as most EVs
The internal resistance Ri is very low (~20 mΩ) and HEVs typically operate in that range.
for cell type 1 when compared to cell type 2 and
cell type 3. This is a typical characteristic of high The current pulse step technique was also
power Lithium-ion cells. The resistance Ri for described. This method is useful for verifying Ri
the three chemistries shows similar behaviour obtained from the parameter characterization
throughout the SOC window. It is almost flat method. It is also verified that the SOEC model is
from 100% - 20 % SOC and increases below valid for cells from different chemistries but the
20% SOC, maximum at completely discharged model parameters have to be determined
state [25]. The resistance R1 which contributes to individually for each of the cell chemistry as this is
charge transfer reaction has the least value an intrinsic property of the cell.
compared to other resistance Ri and R2. It also
shows a similar trend as Ri. The resistance R2 Acknowledgments
increases abruptly when the cells are discharged This work has been done in the framework of
beyond 20%. SOC. The diffusion process CREATE research programme funded by the
becomes a limitation factor when the cell is near Singapore National Research Foundation (NRF).
complete discharge. The values of Capacitance
C1 and C2 generally increase with increase in
SOC, but at some SOC it does not change
significantly or its value decreases.

The model is validated by simulating the voltage

behaviour for different C-Rate and comparing
them with the measurement values. Fig.13(f)
shows the plot of measured and simulated
voltage against the discharge capacity for LCO
chemistry. The simulated voltage shows good
agreement with the measured values. However
there is a deviation in the simulated value below
20% SOC. This deviation can be correlated to the
fitting parameter values below 20% SOC. From
these results, it can be understood that the circuit
parameter values are the intrinsic property of the
cell chemistry. A more detailed understanding of
chemistry is required to quantify the values of
circuit parameters. Therefore, a SOEC model is
applicable for all practical purposes to study the
dynamic behaviour of the cells having different
chemistry.

EEVC European Electric Vehicle Congress 10

 Cell Type 1 Cell Type 2 Cell Type 3

0.02
0.12

0.015 0.1

0.08
R1(Ω)

R2(Ω)
0.01 0.06

0.04
0.005
0.02

0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
SOC(%) SOC(%)

(a) (b)
800
1800
1600
600 1400
1200
1000
C1(F)

400
C2(F)

800
600
200
400
200
0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
SOC(%) SOC(%)
(c) (d)
0.08
5 2C_Discharge 1C_Discharge
0.07 0.5C_Discharge 0.2C_Discharge
2C_Simulation 1C_Simulation
0.06 4.5
0.5C_Simulation 0.2C_Simulation
0.05
4
U(V)

0.04
Ri(Ω)

0.03 3.5
0.02
3
0.01
0 2.5
0 10 20 30 40 50 60 70 80 90 100 0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25 2.5
SOC(%)
Discharge Capacity (Ah)
(e) (f)

Figure 13: Comparison of (a) Short Transient Resistance (R1), (b) Long Transient Resistance (R2), (c) Short Transient
Capacitance (C1), Long Transient Capacitance (C2) and (d) Internal Resistance (Ri) for the given chemistries as
function of SOC. (f) shows the plot of Measurement and Simulated Voltage vs. Discharge Capacity of Cell Type 2 at
different C-Rates.

EEVC European Electric Vehicle Congress 11

References

[1] J. Newman and W. Tiedemann, Porous- [13] Mathworks Inc. Matlab Simscape 3.5 (R2011a).
Electrode Theory with Battery Applications, Available from:
AIChE Journal, 1975. 21(1): p. 25-41. http://www.mathworks.com/products/simscape/.
[2] M. Doyle, T. Fuller, and J. Newman, [14] EUCAR Traction Battery Working Group,
Modeling of galvanostatic charge and Specification of Test Procedures for Hybrid
discharge of the lithium/ polymer/insertion Electric Vehicle Traction Batteries, September
cell, Journal of the Electrochemical Society, 1998.
1993. 140(6): p. 1526-1533.
[15] EUCAR Traction Battery Working Group,
[3] Xiao Hu, et al., A Foster network thermal Specification of Test Procedures for High
model for HEV/EV battery modeling, IEEE Voltage Hybrid Electric Vehicle Traction
Transactions on Industry Applications, 2011. Batteries, January 2005.
47(4): p. 1692-1699.
[16] G.L. Plett. Results of Temperature-Dependent
[4] Cai Long and R.E. White, Mathematical LiPB Cell Modeling, in Proceedings of 21st
modeling of a lithium ion battery with thermal Electric Vehicle Symposium (EVS21). April
effects in COMSOL Inc. Multiphysics (MP) 2005. Monaco.
software, Journal of Power Sources, 2011.
196(14): p. 5985-5989. [17] Idaho National Engineering & Environmental
laboratory, FreedomCAR Battery Test Manual
[5] C.M. Shepherd, Design of Primary and for Power-Assist Hybrid Electric Vehicles, Oct
Secondary cells, Journal of the 2003.
Electrochemical Society, 1965. 112: p. 657-
664. [18] S. Yang, et al., Performance of LiFePO4 as
lithium battery cathode and comparison with
[6] O. Tremblay, L.A. Dessaint, and A.I. manganese and vanadium oxides, Journal of
Dekkiche. A Generic Battery Model for the Power Sources, 2003. 119–121(0): p. 239-246.
Dynamic Simulation of Hybrid Electric
Vehicles, in IEEE Vehicle Power and [19] P. Keil, Development of a battery model for the
Propulsion Conference. 2007. analysis of energy management strategies for
electric vehicles, in Institute for Electrical
[7] Mathworks Inc. MATLAB SimPowerSystems Energy Storage Technology 2010, Technische
User's Guide, Version 5.4 (R2011a). Available Universität München.
from:
http://www.mathworks.com/access/helpdesk/h [20] K. Wolter, Auslegung einer Pufferbatterie f¨ur
elp/toolbox/physmod/powersys/. das 12-V-Bordnetz eines Elektrofahrzeugs, in
Institute for Electrical Energy Storage
[8] V.H. Johnson, Battery performance models in Technology 2011, Technische Universität
ADVISOR, Journal of Power Sources, 2002. München.
110(2): p. 321-329.
[21] H.-G. Schweiger, et al., Comparison of several
[9] S. Jiang, A Parameter Identification Method methods for determining the internal resistance
for a Battery Equivalent Circuit Model, SAE of lithium ion cells, Journal of Sensors, 2010.
Technical Paper, 2011. 2011-01-1367. 10(6): p. 5604-5625.
[10] Chen Min and G.A. Rincon-Mora, Accurate [22] R.A. Jackey, G.L. Plett, and M.J. Klein,
electrical battery model capable of predicting Parameterization of a Battery Simulation Model
runtime and I-V performance, IEEE Using Numerical Optimization Methods, 2009,
Transactions on Energy Conversion, 2006. SAE International.
21(2): p. 504-511.
[23] V. Srinivasan and J. Newman, Discharge model
[11] L. Jaemoon, et al. Modeling and Real Time for the lithium iron-phosphate electrode, The
Estimation of Lumped Equivalent Circuit Journal of Electrochemical Society, 2004.
Model of a Lithium Ion Battery, in Power 151(A1517).
Electronics and Motion Control Conference,
2006. EPE-PEMC 2006. 12th International. [24] M.A. Roscher, J. Vetter, and D.U. Sauer,
2006. Cathode material influence on the power
capability and utilizable capacity of next
[12] S. Lee, et al., State-of-charge and capacity generation lithium-ion batteries, Journal of
estimation of lithium-ion battery using a new Power Sources, 2010. 195(12): p. 3922-3927.
open-circuit voltage versus state-of-charge,
Journal of Power Sources, 2008. 185(2): p. [25] B. Yann Liaw, et al., Modeling of lithium ion
1367-1373. cells—A simple equivalent-circuit model
approach, Journal of Solid State Ionics, 2004.
175(1–4): p. 835-839.

EEVC European Electric Vehicle Congress 12

 Authors
Suguna Thanagasundram earned her Tanja Teutsch received her B.Sc. in
B.Eng. in electrical and electronic Electrical Engineering and
engineering from National University Information Technology from
of Singapore. She completed her MSc Technische Universität München
and PhD degrees from the University (TUM) in 2012. She is currently
of Leicester UK in 2003 and 2007 pursuing a Dipl.-Ing. degree at TUM
respectively. She is currently a and writing her thesis on battery
Research Fellow in TUM CREATE in modelling and battery state estimation
Singapore, working in the area if in TUM CREATE in Singapore.
electrical storage systems.

Raghavendra Arunachala earned his

B.Eng. in electrical and electronic
engineering from Visveswaraya Andreas Jossen earned his doctorate,
Technological University, India. He dealing with "Management of photo-
completed his M.Sc from Rheinische voltaic plants using energy storage
Westfälische Technische Hochschule systems" at University of Stuttgart.
(RWTH) Aachen University, Germany From 1994 he was group leader for
in 2011. He is currently a doctoral different battery related topics with
student in TUM CREATE, Singapore, ZSW, Ulm. Since 2010 he is full
working in the area of electrical professor at the Institute for Electrical
energy storage systems Energy Storage Technology, TUM.

Kamyar Makinejad received his B.Sc.

in electrical engineering in 2007 from
Shiraz University, Iran. He earned his
MSc degree from Leibniz University
of Hanover, Germany in 2011
respectively. He is currently a
Research Associate in TUM CREATE
in Singapore and pursuing his PhD
with the Technical university of
Munich. His research interest is
battery diagnostic and state
determination in EV applications.

EEVC European Electric Vehicle Congress 13

View publication stats