Journal of Energy Storage 43 (2021) 103252

Contents lists available at ScienceDirect

Journal of Energy Storage

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

A comprehensive equivalent circuit model for lithium-ion batteries,

incorporating the effects of state of health, state of charge, and temperature
on model parameters
Manh-Kien Tran a, *, Manoj Mathew a, b, Stefan Janhunen b, Satyam Panchal c,
Kaamran Raahemifar d, e, Roydon Fraser c, Michael Fowler a
a
Department of Chemical Engineering, University of Waterloo, 200 University Ave W, Waterloo, Ontario, N2L 3G1, Canada
b
Nuvation Energy, 40 Bathurst Dr, Waterloo, Ontario, N2V 1V6, Canada
c
Department of Mechanical and Mechatronics Engineering, University of Waterloo, 200 University Ave W, Waterloo, Ontario, N2L 3G1, Canada
d
College of Information Sciences and Technology (IST), Pennsylvania State University, State College, Pennsylvania PA 16801, USA
e
Department of Electrical and Computer Engineering, Sultan Qaboos University, Al-Khoud 123, Muscat, Oman

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

Keywords: The equivalent circuit model (ECM) is a battery model often used in the battery management system (BMS) to
Lithium-ion batteries monitor and control lithium-ion batteries (LIBs). The accuracy and complexity of the ECM, hence, are very
Battery modeling important. State of charge (SOC) and temperature are known to affect the parameters of the ECM and have been
Equivalent circuit model
integrated into the model effectively. However, the effect of the state of health (SOH) on these parameters has
State of health
State of charge
not been widely investigated. Without a good understanding of the effect of SOH on ECM parameters, parameter
Battery management system identification would have to be done manually through calibration, which is inefficient. In this work, experi­
ments were performed to investigate the effect of SOH on Thevenin ECM parameters, in addition to the effect of
SOC and temperature. The results indicated that with decreasing SOH, the ohmic resistance and the polarization
resistance increase while the polarization capacitance decreases. An empirical model was also proposed to
represent the effect of SOH, SOC, and temperature on the ECM parameters. The model was then validated
experimentally, yielding good results, and found to improve the accuracy of the Thevenin model significantly.
With low complexity and high accuracy, this model can be easily integrated into real-world BMS applications.

1. Introduction
The demand for rechargeable and high-performance batteries has
soared in recent years. Lithium-ion batteries (LIBs) have gathered the
most interest out of all battery types. In 2018, over 90% of large-scale
battery storage power capacity was provided by LIBs in the United
States [1]. The exponential growth of power capacity was also reported,
with 125 energy storage systems storing a total of 869 MW by the end of
2018, doubling the value reported in 2015. LIBs are used in numerous
applications, ranging from large-scale products like electric vehicles
(EVs) to small portable devices such as smartphones [2]. With the effects
of climate change worsening, the push for EVs has drastically increased
over the past 10 years to reduce carbon dioxide emissions [3]. In addi­
tion, clean energy systems have also emerged from obscurity, which can
produce excess electricity in some circumstances. In such cases, batteries
will be required to store additional clean energy and release it back onto
the grid on demand [4]. LIBs play an important role in the future of
energy storage systems as they have numerous advantages such as high
energy density, high power density, long cycle life, low self-discharge
rate, small size, light weight, rapid charging capabilities, and wide
temperature range [5]. With an increasing interest in LIBs, some re­
searchers have been able to use different materials in the anode and
cathode to further increase energy density and lower costs, allowing the
battery technology to advance further [6–9].
In real-world applications, battery models are used to predict
working voltage, power, and energy capability to ensure reliable per­
formance and safety [10]. Above all, battery models are critical in
developing accurate algorithms for battery state estimation, such as
state of health (SOH) and state of charge (SOC), of working battery packs
[11]. Model predictions are commonly executed within an embedded
battery management system (BMS) to provide a high-efficiency

* Corresponding author.
E-mail address: kmtran@uwaterloo.ca (M.-K. Tran).

https://doi.org/10.1016/j.est.2021.103252
Received 18 May 2021; Received in revised form 30 July 2021; Accepted 20 September 2021
Available online 1 October 2021
2352-152X/© 2021 Elsevier Ltd. All rights reserved.
M.-K. Tran et al.
Nomenclature
C1 RC pair capacitance [F]
I Battery current [A]
OCV Open circuit voltage
R0 Battery internal resistance [Ω]
R1 RC pair resistance [Ω]
Δt Time difference [s]
V Battery terminal voltage [V]
U1 Voltage across RC pair [V]
Acronyms
BMS Battery management system
CCDC Constant charge/discharge cycle
EV Electric vehicle
estimation method, which ensures stable battery operation. In applica­
tions such as EVs, the BMS ensures safety and improves performance
during runtime based on the modeling of application-related loads and
stress factors [12]. In addition, health diagnosis models are employed to
predict the lifetime of LIBs, to improve the reliability of vehicle usage,
and to avoid malfunctions and catastrophic failures [13].
The determination of battery states is not trivial. In recent years, a
large variety of methods for battery state estimation have been proposed
in the literature, ranging from electrochemical models and equivalent
circuit models to neural network-based approaches. The equivalent
circuit model (ECM) is used in many battery applications due to its fast
execution time, simplicity, and relatively high accuracy [14]. However,
the ECM has the problem of poor model extrapolation under a wider
range of operating conditions if the battery is pushed towards its oper­
ating limits, and hence, it is not a battery model used often for appli­
cations that demand high current rates or are run at very low
temperatures [15]. The electrochemical battery models are often more
accurate, but they demand a great amount of computing power, which
makes them not suitable for many real-time applications. Besides the
electrochemical model and the ECM, there is a recent trend of combining
both models to develop more advanced battery management algorithms,
which are computationally efficient with the capability to signify the
battery states and fault conditions [16,17]. Currently, due to the limi­
tations in data storage and computing capability, among the existing
battery models used in the BMS, the equivalent circuit model (ECM)
remains the most prominent. Gandolfo et al. [18] derived a dynamic
estimation model to rapidly identify electrical parameters adopted by a
simple 1-resistor ECM and used it to predict the SOC of lithium polymer
batteries. Dey et al. [19] proposed a nonlinear adaptive observer scheme
that estimates both electrochemical model parameters and temperature
state inside the electrode simultaneously to predict the SOC and SOH of
a battery cell. Tran et al. [20] utilized the Thevenin ECM to derive a
real-time diagnosis algorithm to detect sensor faults in the BMS. Gao
et al. [21] proposed a hybrid model that consisted of an ECM and a
convective thermal model to determine the states of LIBs. Plett et al.
[22] developed an approach to estimate the SOC of a LIB cell through an
ECM and a Kalman filter observer.
When modeling the dynamic behavior of a battery, accuracy and
complexity are two important factors. The ECM has the potential to
satisfy both factors effectively. The model consists of three major com­
ponents: a component representing the thermodynamic properties of the
battery chemistry, such as the open-circuit voltage (OCV) as a function
of SOC; another representing the kinetic aspects of the cell internal
impedance behavior; and a source or load to complete the circuit for the
charge or discharge procedures [23]. ECM parameters typically include
an internal ohmic resistance, followed by one or more resistor-capacitor
(RC) pairs. The simplest model only takes the internal ohmic resistance
2
Journal of Energy Storage 43 (2021) 103252
ECM Equivalent circuit model
FUDS Federal urban dynamic schedule
HPPC Hybrid pulse power characterization
LFP Lithium iron phosphate (LiFePO4)
LIB Lithium-ion battery
MATLAB Matrix laboratory
MACCOR Battery cell cycler
MAPE Mean absolute percentage error
OCV Open circuit potential
RC Resistance-capacitance
RMSE Root-mean-squared error
SOC State of charge
SOH State of health
UDDS Dynamometer driving schedule
into account which does not accurately represent the battery dynamics
during operation [24,25]. Thus, the Thevenin ECM, which has one
additional RC pair to combine with the internal ohmic resistance, is
widely used since it has a good balance between accuracy and
simplicity. The ECM parameters are often estimated using a hybrid pulse
power characterization (HPPC) test at different SOC values [26]. Some
factors such as temperature and SOC can affect these parameters. For
instance, an increase in temperature can lead to higher OCV in certain
LIB chemistries, and a decrease in SOC can lead to lower charge transfer
resistance, which ultimately affects the ECM parameters [11].
Continuous usage of a LIB can lead to degradation, caused by loss of
active materials or solid electrolyte interphase formation, resulting in
the decrease of its SOH over time [27–29]. By understanding and
including the possible effects of SOH on ECM parameters in BMS algo­
rithms, the BMS can improve the performance, reliability, and safety of
the battery [30]. There are several methods that enable the online
estimation of ECM parameters [31–33]. However, these methods do not
consider either battery degradation or sensor faults, which can affect the
accuracy of the estimated values and ultimately the reliability of the
BMS. Lai et al. [34] performed a sensitivity analysis of ECM parameters
for LIBs using the one-factor-at-a-time method. The analysis was con­
ducted for different SOC and SOH ranges while excluding temperature
analysis. Gomez et al. [35] examined the effect of SOC and temperature
on the ECM parameters in the temperature range of 20 to 50ºC at various
SOC values. A simple model was proposed to integrate these effects into
the ECM. The predicted model parameters showed a low variance of 5%
when compared to the experimental data, and therefore indicated a
good statistical agreement of the proposed model to experimental
values. However, the relationship between ECM parameters and all
three metrics together (SOH, SOC, and temperature) has not been fully
explored in the literature. Also, a form of ECM that includes the effect of
battery aging has never been proposed in the literature either. The
contribution of this paper is the investigation into the effects of SOH,
SOC, and temperature on the ECM parameters. Another contribution is
the proposal of an empirical model, which integrates these effects into
the Thevenin ECM to improve its accuracy and reliability to be used in
the BMS for real-world applications. Another important advantage of the
proposed model is that it will allow for the simulation of LIBs at various
SOH levels and temperatures, which can be used to develop and test
other advanced BMS algorithms or to identify energy storage system
design considerations. The rest of the paper is organized as follows.
Section 2 describes the experimental setup and procedures. Section 3
analyzes the ECM parameters characterization results and introduces
the proposed empirical model. Section 4 presents the validation of the
proposed model, while Section 5 provides some concluding remarks.
M.-K. Tran et al. Journal of Energy Storage 43 (2021) 103252

Table 1 temperature environment for at least an hour to allow the cell internal
Specifications of the lithium-ion cells used in the experiments. temperature to match the environment temperature. The voltage and
Cell Specifications Cell 1 Cell 2 Cell 3 Cell 4 Cell 5 current data were collected at a frequency of 1 Hz and then stored in a
computer that has a software program that controls the Maccor, shown
Capacity (Ah) 18.40 15.70 13.00 14.58 17.10
Cathode Material LiFePO4 (all 5 cells) in Fig. 1d.
Anode Material Graphite (all 5 cells) Characterization tests of cells 1, 2, and 3 included a capacity test, a
Nominal Voltage (V) 3.3 (all 5 cells) SOC–OCV test, and multiple HPPC tests at varying temperatures. The
Cell Weight (g) 496 (all 5 cells) capacity test measured cell capacity in Ah through fully discharging and
Dimensions (mm) 7 × 160×227 (all 5 cells)
charging the cells at a C-rate of 1C. The SOC–OCV test established the
SOC–OCV relationship through a lookup table from the experimental
2. Experimental procedures results. The SOC–OCV test was conducted by fully discharging and
charging the cells at a C-rate of C/25. The HPPC test was used to
The experiments involved five lithium iron phosphate (LFP) pouch determine the ECM parameters of each cell at different temperatures and
cells at different nominal capacities with specifications shown in SOC values [26]. For the HPPC test, at 0.9 SOC, a pulse was run, which
Table 1. All cells have similar electrical properties except for their ca­ consisted of a 10-second discharge at 1C, a 40-second rest period, and a
pacities, in order to minimize any unwanted variables and ensure the 10-second charge at 0.75C. The cell was then rested for 1 hour and
reliability of the experimental results. Cells 1, 2, and 3 were used for subsequently discharged for 6 min at 1C to decrease the SOH level by
characterization and investigation of the effect of SOH, SOC, and tem­ 0.1. It should be noted that in this paper, SOH is defined using battery
perature on the ECM parameters. Cells 4 and 5 were used to validate the capacity (total remaining capacity/total initial capacity). These steps
proposed model. The original nominal capacity of these cells was 20 Ah. were followed by a 1-hour rest period and repeated until the battery
The lowered capacities were achieved by degrading the battery by reached a SOC of 0.1. The HPPC tests were run for cells 1, 2, and 3, at
cycling the cells over time, where the cells were discharged and charged 5ºC, 15ºC, 25ºC, 35ºC, and 45ºC, for a SOC range going from 0.9 to 0.1
at C-rates of 2C and 1C, respectively. with an interval of 0.1.
The completed experimental setup can be seen in Fig. 1. The Maccor To validate the performance of the proposed empirical model, an
4200, shown in Fig. 1a, is the battery tester equipment used in the ex­ Urban Dynamic Drive Schedule (UDDS) current profile and a non-
periments. The current collectors and voltage sensors from the Maccor dynamic constant charge/discharge cycle (CCDC) current profile were
were connected to the cells in a thermal chamber, as seen in Fig. 1b. To run on cells 4 and 5, at temperatures of 10ºC and 40ºC. The UDDS drive
simulate several different temperatures for the battery testing environ­ cycle simulates the dynamic charge and discharge of the battery to
ments, a CSZ MicroClimate thermal chamber was used as shown in recreate the city driving conditions for light-duty vehicles. The CCDC
Fig. 1c. Before a test was conducted, the cells rested in the new profile was conducted by discharging the battery at 1C and charging the

Fig. 1. Experimental setup. a) Battery tester Maccor 4200. b) Connected and tested cell in the thermal chamber. c) CSZ MicroClimate thermal chamber. d) Computer
software controlling the Maccor 4200.

3
M.-K. Tran et al.
Fig. 2. Current profiles used to validate the proposed model. a) UDDS. b) CCDC.
Fig. 3. Schematic of the Thevenin equivalent circuit model.
Fig. 4. Experimentally established SOC–OCV curve of the LFP cells.
battery at 0.75C several times. The current profiles are shown in Fig. 2.
The battery was first charged to 0.8 SOC, and each of the current profiles
was then run to reach 0.3 SOC. Using the proposed model, the param­
eters of the Thevenin ECM would be calculated based on the SOH, SOC,
and temperature of the cells, and used to estimate the voltage to
compare to the experimental voltage measurements. The accuracy of the
model would then be evaluated and compared to the previous ECM
method that does not consider the effect of SOH on the ECM parameters.
4
Journal of Energy Storage 43 (2021) 103252
3. Trend analysis and development of the proposed model
3.1. Equivalent circuit model
The Thevenin ECM is shown in Fig. 3. The model is used to calculate
the battery voltage in response to the current. The OCV is represented by
an ideal voltage source, which is correlated with the battery SOC. R0
accounts for the ohmic resistance of the battery, and the parallel RC
network (R1 and C1) represents the transient behavior of the battery
resulting from the interfacial charge-transfer reactions at the electrode.
The product of R1 and C1 represents the time constant for the RC pair.
The Thevenin ECM is shown in Eqs. (1) and (2), expressed in discrete
form.
The Thevenin ECM is as follows,
Vj = OCV − R0 Ij − U1,j (1)
( ) [ ( )]
Δt Δt
U1,j+1 = exp − U1,j + R1 1 − exp − Ij (2)
R1 C1 R1 C 1
where V is the battery terminal voltage, I is the battery current, OCV is
the battery open circuit voltage, R0 is the internal ohmic resistance, U1 is
the voltage of the RC network, R1 and C1 are the polarization resistance
and capacitance with the product R1C1 being the time constant of the RC
network, and Δt is the sampling time, with the subscript j being the
discrete index. The model parameters OCV, R0, R1, and C1, are known to
be functions of SOC and temperature.
3.2. Battery characterization results and trend analysis
The OCV-SOC relationship was established experimentally by dis­
charging and charging the cells slowly at a C-rate of C/25 and measuring
the voltage through time. The OCV-SOC curve was constructed by
averaging the results from all 5 tested cells at the beginning of their
lifespan. It is noted that the OCV-SOC does not often change significantly
throughout the lifespan of a Li-ion battery, especially within the studied
SOC range of 0.3–0.8 [36], and therefore, the initial established
OCV-SOC curve was used for all SOH values in this study. Another
assumption that was made in this study for the OCV-SOC curve con­
struction is the exclusion of the entropic coefficient, which represents
the relationship between the OCV-SOC curve and temperatures. This
was done to simplify the model, since the entropic coefficient is often
only significant at more extreme SOC levels, and the SOC range studied
is 0.3–0.8 where the entropic coefficient values would be close to zero
[37]. A look-up table was built, which was needed to estimate the cell
OCV which is a parameter in the ECM. The OCV-SOC curve is shown in
M.-K. Tran et al.
Fig. 5. Effect of SOH, SOC, and temperature on R0.
Fig. 6. Effect of SOH, SOC, and temperature on R1.
Fig. 7. Effect of SOH, SOC, and temperature on C1.
5
Journal of Energy Storage 43 (2021) 103252
Table 2
Parameters of the proposed model after fitting.
X b0 b1 b2 b3
R0 10,424.73 − 48.2181 − 114.74 − 1.40433
R1 13,615.54 − 68.0889 − 87.527 − 37.1084
C1 − 11,116.7 180.4576 237.4219 40.14711
Fig. 4.
The other parameters of the ECM (R0, R1, and C1) are known to be
functions of SOC and temperature. The model parameters were deter­
mined by fitting the experimental results from the HPPC test to the ECM
using the “nlinfit” built-in function in MATLAB, based on the Lev­
enberg–Marquardt algorithm, which is often used to solve least-squares
curve-fitting problems. The model was fitted in MATLAB using data
from the 1-minute discharge-rest-charge pulses at different SOC levels
ranging from 0.1 to 0.9 with an interval of 0.1 as well as at different
temperatures ranging from 5 ◦ C to 45 ◦ C with an interval of 10 ◦ C. The
results for the model parameters are expressed as lookup tables with
SOC and temperature, and provided in Tables A.1, A.2, and A.3 in the
Appendix for SOH of 92%, 78.5%, and 65%, respectively. Figs. 5, 6, and
7, using values from the tables in the Appendix, show the relationship
between the ECM parameters (R0, R1, and C1) and SOH, SOC, and
temperature.
From Fig. 5, it can be seen that the ohmic resistance R0 increases
significantly with decreasing temperature at all SOH values. Also, the
effect of SOH on R0 is dependent on temperature, as the interaction ef­
fect between SOH and T is shown to be significant. Another observation
indicates that R0 increases as the battery ages, but this increase is only
seen significantly at colder temperatures. At warmer temperatures, an
increase in resistance is not observed at different SOH values. Finally, R0
does not seem to vary significantly with SOC.
Fig. 6 shows that the polarization resistance R1 increases with SOH,
and unlike the ohmic resistance R0, the effect is observed at both high
and low temperatures. R1 also increases with decreasing temperature,
showing a relatively linear and inversely proportional relationship.
Finally, the polarization resistance R1 varies significantly with SOC, with
the resistance increasing when approaching both high and low SOC
values. A higher level of variability is observed in the data for this
parameter.
It can be seen from Fig. 7 that the polarization capacitance C1
changes very significantly with SOH at all SOC and temperature values,
as the three surfaces are relatively parallel. This shows that C1 can be a
good indicator of degradation, which agrees with the results found in
[20]. Overall, C1 decreases with SOH and increases with temperature.
The effects of temperature and SOC on the polarization capacitance do
not change with SOH, meaning that the interaction effects between
temperature and SOH, and between SOC and SOH are not significant.
Overall, all three ECM parameters R0, R1, and C1 are affected by SOH,
as well as SOC and temperature. Therefore, it is possible to represent
each of these parameters as a function of SOH, SOC, and temperature,
which will be discussed in the next section.
3.3. Proposed empirical model for the ECM parameters
We use an empirical model to represent the relationship between the
ECM parameters and the three factors SOH, SOC, and temperature,
shown in Figs. 5, 6, and 7. It should be noted that the SOC range
considered in this study was 0.3 to 0.8 which is a normal operating range
for many real-world applications. In this range, the relationship between
the three parameters and the three factors can be observed from the
figures to be acceptably linear, and no interaction effect between the
three factors was prominent. Also, within the studied SOC range, it is
observed that the relationships between the ECM parameters (R0, R1,
and C1) and the three factors SOH, SOC, and temperature are relatively
monotonic functions. This explains an observation that, when second-
M.-K. Tran et al. Journal of Energy Storage 43 (2021) 103252

Fig. 8. Model validation results at 72.9% SOH. a) Using UDDS profile at 40 ◦ C. b) Using UDDS profile at 10 ◦ C. c) Using CCDC profile at 40 ◦ C. d) Using CCDC profile
at 10 ◦ C.

order terms and interaction effects were added in the model develop­ validation. Two SOH values of 72.9% and 85.5%, and two temperature
ment process, it was found that they did not improve the fitting results settings of 10 ◦ C and 40 ◦ C were selected, which were different than the
significantly while demanding more computing capability from adding values used for characterization and fitting the proposed model, to
more parameters to fit. Therefore, the terms in the proposed model are ensure the validation results are unbiased and reliable. The validation
all of first order (linear); this would also allow the model to have low results for the 4 runs at SOH of 72.9% are shown in Fig. 8, and the results
computational complexity, increasing its suitability for practical appli­ for the 4 runs at SOH of 85.5% are shown in Fig. 9.
cations like the BMS. The equation for the empirical model for the ECM Figs. 8 and 9 show good agreement between the predicted voltage
parameters is shown as, using the proposed model and the actual voltage obtained from the
experiments. To report the error quantitatively, the root-mean-squared
X = b0 + b1 SOH + b2 T + b3 SOC (3)
error (RMSE) and the mean absolute percentage error (MAPE) can be
where X represents any of the ECM parameters R0, R1, and C1. calculated. Table 3 shows the summary of errors for all the validation
Using the results from Tables A.1, A.2, and A.3 to fit Eq. (3) for each runs, including RMSE and MAPE. The equations to calculate the RMSE
ECM parameter, along with the “nlinfit” built-in function in MATLAB, and MAPE for N samples are shown as,
the values for b0, b1, b2, and b3 in the proposed model were obtained, as √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
∑N ( )2
shown in Table 2. The proposed model parameters of C1 are inverse in i=1 Vmod,i − Vexp,i
RMSE = (4)
sign compared to the ones of R0 and R1, which agrees with the obser­ N
vations of the trends from Figs. 5, 6, and 7. With this empirical model,
N ⃒ ⃒
we should be able to estimate the values of R0, R1, and C1 at any SOC, 1 ∑ ⃒Vmod,i − Vexp,i ⃒
MAPE = ⃒ ⃒ (5)
temperature, and SOH. N i=1 ⃒ Vexp,i ⃒

4. Validation of the proposed model
In order to validate the proposed model, we conducted 8 validation
runs as described in Section 2. A dynamic UDDS drive cycle current
profile and a non-dynamic CCDC current profile were used for
where Vmod,i and Vexp,i are used to represent the predicted and experi­
mental voltage of the validation runs.
From Figs. 8a, 8b, 9a, and 9b, there was not any significant trend of
deviation between each run, an indication to show that the model fitting
was sufficient. When examining Figs. 8c, 8d, 9c, and 9d, during the

6
M.-K. Tran et al. Journal of Energy Storage 43 (2021) 103252

Fig. 9. Model validation results at 85.5% SOH. a) Using UDDS profile at 40 ◦ C. b) Using UDDS profile at 10 ◦ C. c) Using CCDC profile at 40 ◦ C. d) Using CCDC profile
at 10 ◦ C.

Table 3 Table 4
Summary of errors for the model validation runs. Comparison of errors between validation results using the proposed model and
Figure Run# Run description RMSE (mV) MAPE (%)
not using the proposed model.
Run# RMSE (mV) RMSE (mV) MAPE (%) MAPE (%)
8a 1 UDDS – 72.9% SOH – 40 ◦ C 11.99 0.32
using not using using not using
8b 2 UDDS – 72.9% SOH – 10 ◦ C 16.74 0.48
proposed proposed model proposed proposed model
8c 3 CCDC – 72.9% SOH – 40 ◦ C 14.78 0.39
model model
8d 4 CCDC – 72.9% SOH – 10 ◦ C 18.13 0.49
9a 5 UDDS – 85.5% SOH – 40 ◦ C 7.63 0.2 1 11.99 21.76 0.30 0.63
9b 6 UDDS – 85.5% SOH – 10 ◦ C 11.14 0.31 2 16.74 30.12 0.48 0.92
9c 7 CCDC – 85.5% SOH – 40 ◦ C 8.05 0.20 3 14.78 29.09 0.39 0.84
9d 8 CCDC – 85.5% SOH – 10 ◦ C 10.38 0.27 4 18.13 38.22 0.49 0.96
5 7.63 14.07 0.20 0.37
6 11.14 23.46 0.31 0.49
CCDC runs, it is apparent that the modeled voltage is higher during 7 8.05 15.13 0.20 0.40
charge and lower during discharge compared to the experimental 8 10.38 25.41 0.27 0.63
voltage. This trend is less apparent in the UDDS runs, and the error
values are also lower in UDDS runs compared to their corresponding model inaccuracy, including experimental data noise, equipment error,
CCDC runs. This observation agrees with the findings in [38] and is due and the capacity of the battery not being modeled as a function of
to the lack of representation of hysteresis in the Thevenin ECM, which temperature.
will be a focus in our future research. Overall, in all cases, a strong Some comparison between the errors found using this proposed
agreement was observed between modeled and experimental results at model and typical error values found in the literature can be conducted
all temperatures and SOHs, as shown by the low RMSE and MAPE values to evaluate the performance of the model. Results from He et al. [39]
in Table 3. The largest deviation between experimental and modeled showed an RMSE of around 13 mV when using a Thevenin ECM for the
results was at 10 ◦ C and 72.9% SOH using the CCDC current profile, Federal Urban Dynamic Schedule (FUDS) test, while Hu et al. [40] ob­
where an RMSE value of 18.13 mV and a MAPE of 0.49% were observed. tained an RMSE of around 12 mV using the same model. In [38], the
There are a few other factors that could cause errors aside from the

7
M.-K. Tran et al. Journal of Energy Storage 43 (2021) 103252

Table A.1 Table A.2

Lookup table of ECM parameters for various SOC and temperature values at 92% Lookup table of ECM parameters for various SOC and temperature values at
SOH. 78.5% SOH.
Temperature (⁰C) SOC (%) R0 (µΩ) R1 (µΩ) C1 (F) Temperature (⁰C) SOC (%) R0 (µΩ) R1 (µΩ) C1 (F)

5 10 5910 7850 5075 5 10 6240 10,560 4206

5 20 5810 6360 5981 5 20 6100 7570 4975
5 30 5710 5390 6665 5 30 6020 6500 5515
5 40 5610 4480 7207 5 40 5960 5950 6018
5 50 5520 3830 7781 5 50 5890 5070 6396
5 60 5440 3680 8293 5 60 5870 5250 6835
5 70 5360 3770 8476 5 70 5850 5450 7048
5 80 5280 4370 8025 5 80 5780 6520 6508
5 90 5170 4070 8942 5 90 5700 5130 7100
15 10 3490 6220 7823 15 10 3960 8410 6312
15 20 3440 4740 9129 15 20 3860 5550 7496
15 30 3380 3800 10,123 15 30 3850 5010 8204
15 40 3310 3190 10,653 15 40 3860 5110 8607
15 50 3280 3170 11,140 15 50 3820 4110 9013
15 60 3230 2780 11,686 15 60 3820 3810 9389
15 70 3220 3520 11,726 15 70 3840 4330 9430
15 80 3160 3670 10,933 15 80 3870 5240 8421
15 90 3110 2980 12,001 15 90 3860 4400 9374
25 10 2880 5360 10,126 25 10 2870 6134 8350
25 20 2840 3830 11,778 25 20 2849 4702 9693
25 30 2820 3190 12,992 25 30 2834 3438 10,543
25 40 2760 2400 13,689 25 40 2842 3266 11,006
25 50 2760 2430 14,049 25 50 2873 3529 11,246
25 60 2720 2080 14,361 25 60 2886 3383 11,682
25 70 2700 2810 14,406 25 70 2907 3792 11,577
25 80 2660 2400 13,780 25 80 2910 3868 10,541
25 90 2610 2150 15,140 25 90 2908 3286 11,770
35 10 2466 4613 12,573 35 10 2065 5861 9925
35 20 2445 3136 14,746 35 20 2033 3543 11,925
35 30 2412 2711 16,276 35 30 2042 3314 13,101
35 40 2347 1886 17,081 35 40 2044 3122 13,513
35 50 2337 1913 17,221 35 50 2048 2766 13,924
35 60 2321 2011 17,762 35 60 2030 2214 14,185
35 70 2318 2368 17,727 35 70 2062 2800 14,134
35 80 2301 2399 16,837 35 80 2086 3318 13,148
35 90 2270 1654 18,445 35 90 2076 2368 14,536
45 10 2340 4010 14,846 45 10 1834 4711 11,914
45 20 2330 3220 17,382 45 20 1849 3728 13,869
45 30 2290 1850 19,457 45 30 1845 2778 15,770
45 40 2290 2140 20,348 45 40 1855 2901 16,178
45 50 2250 1520 20,685 45 50 1839 2022 16,568
45 60 2270 1720 21,595 45 60 1875 2429 16,839
45 70 2230 1600 21,373 45 70 1905 2737 16,783
45 80 2240 1880 20,064 45 80 1884 2152 15,771
45 90 2230 1900 21,977 45 90 1901 1678 17,107

Thevenin ECM yielded a MAPE of 0.48%. The difference of the proposed
model in this paper compared to the Thevenin model used in [38–41] is
that this new model took into account the effect of SOH, whereas the
simple model used in previous research works only considered SOC and
temperature while keeping the SOH constant. Even with the increase in
the number of variables, the error values obtained using the proposed
model are comparable to the error values found in the literature, indi­
cating good model accuracy.
Further analysis was conducted to show the improvement of results
from using the proposed model. All 8 validation runs shown in Table 3
were simulated again but without considering the effect of SOH on the
ECM parameters. This was done by using the ECM parameters found at
92% SOH shown in Table A.1 for the 8 validation runs, despite them
being at lower SOH levels, to represent the lack of SOH effect consid­
eration. The resulting error values using this approach are compared to
the errors when using the proposed model, as shown in Table 4. It can be
seen that using the proposed model can reduce the errors, both RMSE
and MAPE, by approximately half. This reduction in errors shows that
the proposed model which considers the effect of SOH on ECM param­
eters can improve the Thevenin model accuracy significantly.
The results from these validation experiments suggest that the pro­
posed model can be utilized by researchers in the future to accurately
predict the voltage response of the battery at different SOH levels,
without having to make the incorrect assumption of constant ECM pa­
rameters or having to characterize the battery constantly at different
SOHs while it undergoes degradation.
5. Conclusions
This study investigated the effect of SOH, SOC, and temperature on
the Thevenin ECM parameters experimentally using LFP batteries. An
empirical model was also developed and validated to reflect the effect of
these factors on the ECM parameters. The main conclusions from this
work are as follows:
1. All three ECM parameters were highly dependent on SOH, SOC,
and temperature.
2. The ohmic resistance R0 increased significantly with decreasing
temperature at all SOH values. It also increased with decreasing
SOH, more so at colder temperatures.
3. The polarization resistance R1 increased with SOH, and unlike the
ohmic resistance R0, the effect was observed at both high and low
temperatures.
4. The polarization capacitance C1 decreased with SOH at all SOC
and temperature values at a relatively constant rate, which showed
that C1 can be a good indicator of battery degradation.

8
M.-K. Tran et al. Journal of Energy Storage 43 (2021) 103252

Table A.3 Visualization. Manoj Mathew: Conceptualization, Methodology, Soft­

Lookup table of ECM parameters for various SOC and temperature values at 65% ware, Validation, Writing – original draft, Visualization. Stefan Jan­
SOH. hunen: Methodology, Software, Writing – review & editing. Satyam
Temperature (⁰C) SOC (%) R0 (µΩ) R1 (µΩ) C1 (F) Panchal: Formal analysis, Investigation, Writing – review & editing.
5 10 9224 13,033 3641
Kaamran Raahemifar: Resources, Validation, Writing – review &
5 20 8984 8576 4413 editing, Supervision, Funding acquisition. Roydon Fraser: Resources,
5 30 8920 7866 4814 Writing – review & editing, Supervision, Project administration, Fund­
5 40 8871 6588 5246 ing acquisition. Michael Fowler: Conceptualization, Methodology,
5 50 8850 6071 5572
Writing – original draft, Supervision, Project administration, Funding
5 60 8867 6166 5879
5 70 8855 5559 6107 acquisition.
5 80 8918 7272 5593
5 90 8914 6694 5944
15 10 5692 11,605 5084 Declaration of Competing Interest
15 20 5557 7398 6257
15 30 5521 6010 6820
The authors declare that they have no known competing financial
15 40 5510 4860 7292
15 50 5539 4547 7697 interests or personal relationships that could have appeared to influence
15 60 5595 4268 7958 the work reported in this paper.
15 70 5655 4254 8103
15 80 5776 5861 7083
15 90 5835 4597 7728
Acknowledgments
25 10 3532 10,086 6220
25 20 3454 6060 7761 This work was supported by the Department of Chemical Engineer­
25 30 3453 4894 8417 ing at the University of Waterloo, Canada Research Chair Tier I - Zero-
25 40 3462 3804 9051
Emission Vehicles and Hydrogen Energy Systems Grant number: 950-
25 50 3497 3367 9262
25 60 3566 3339 9532 232215, and The Natural Sciences and Engineering Research Council of
25 70 3651 3214 9645 Canada (NSERC), Discovery Grants Program, RGPIN-2020-04149.
25 80 3822 5843 8181
25 90 3879 3609 9097
35 10 2752 7957 7527
Appendix
35 20 2664 4190 9426
35 30 2599 3211 10,161 Lookup tables of ECM parameters for various SOH, SOC, and tem­
35 40 2623 2938 10,747 perature values (Tables A.1, A.2, and A.3).
35 50 2656 3271 11,024
35 60 2604 2576 11,212
35 70 2592 2756 11,328 References
35 80 2658 4495 9875
35 90 2686 2864 11,224 [1] Battery storage in the United States: an update on market trends, U.S. energy
45 10 1962 7408 8440 information administration, 2020. https://www.eia.gov/analysis/studies/elect
45 20 1933 4444 10,807 ricity/batterystorage/pdf/battery_storage.pdf (accessed on November 8, 2020).
45 30 1949 3675 11,876 [2] X. Li, C. Wang, Engineering nanostructured anodes via electrostatic spray
45 40 1871 2588 12,585 deposition for high performance lithium ion battery application, J. Mater. Chem. A
45 50 1890 2518 12,911 1 (2) (2013) 165–182, https://doi.org/10.1039/C2TA00437B.
[3] M.-.K. Tran, S. Sherman, E. Samadani, R. Vrolyk, D. Wong, M. Lowery, M. Fowler,
45 60 1841 1988 12,944
Environmental and economic benefits of a battery electric vehicle powertrain with
45 70 1932 2491 13,521
a zinc–air range extender in the transition to electric vehicles, Vehicles 2 (2020)
45 80 1906 2918 12,021
398–412, https://doi.org/10.3390/vehicles2030021.
45 90 1935 2050 13,093 [4] Z.W. Seh, Y. Sun, Q. Zhang, Y. Cui, Designing high-energy lithium–sulfur batteries,
Chem. Soc. Rev. 45 (20) (2016) 5605–5634, https://doi.org/10.1039/
C5CS00410A.
5. The empirical model developed in this paper can be used to [5] S. Panchal, K. Gudlanarva, M.-.K. Tran, R. Fraser, M. Fowler, High Reynold’s
accurately predict the voltage profiles of an LFP battery cell at number turbulent model for micro-channel cold plate using reverse engineering
approach for water-cooled battery in electric vehicles, Energies 13 (2020) 1638,
various SOH, SOC, and temperature conditions. The RMSE values https://doi.org/10.3390/en13071638.
and MAPE values when validating the model were low, showing [6] Y. Liu, H. Liu, L. An, X. Zhao, G. Liang, Blended spherical lithium iron phosphate
good model accuracy. cathodes for high energy density lithium–ion batteries, Ionics 25 (2019) 61–69,
https://doi.org/10.1007/s11581-018-2566-7.
6. The proposed model, which considers the effect of SOH on ECM [7] D. Jugović, D. Uskoković, A review of recent developments in the synthesis
parameters, was shown to improve the accuracy of the Thevenin procedures of lithium iron phosphate powders, J. Power Sources 190 (2) (2009)
model significantly. The errors when using the model were approx­ 538–544, https://doi.org/10.1016/j.jpowsour.2009.01.074.
[8] R. Fang, K. Chen, L. Yin, Z. Sun, F. Li, H.M. Cheng, The regulating role of carbon
imately half of the errors when the model was not used. nanotubes and graphene in lithium-ion and lithium-sulfur batteries, Adv. Mater.
(2018), 1800863, https://doi.org/10.1002/adma.201800863.
The findings in this study show the significant effect of SOH, or [9] A. Fotouhi, D.J. Auger, K. Propp, S. Longo, M. Wild, A review on electric vehicle
battery modelling: from lithium-ion toward lithium–sulphur, Renewable
degradation, on the ECM parameters. The proposed empirical model
Sustainable Energy Rev. 56 (2016) 1008–1021, https://doi.org/10.1016/j.
attempting to represent this effect contributes towards a better under­ rser.2015.12.009.
standing of battery degradation as well as the future development of [10] Y. Cui, P. Zuo, C. Du, Y. Gao, J. Yang, X. Cheng, G. Yin, State of health diagnosis
model for lithium ion batteries based on real-time impedance and open circuit
more advanced and accurate battery models to be used in real-world
voltage parameters identification method, Energy 144 (2018) 647–656, https://
applications. For future research, more investigation can be done for doi.org/10.1016/j.energy.2017.12.033.
other types of battery cell such as cylindrical or prismatic, as well as for [11] H. Chaoui, H. Gualous, Online parameter and state estimation of lithium-ion
other Li-ion battery chemistry such as LMO, NMC, and NCA. batteries under temperature effects, Electr. Power Syst. Res. 145 (2017) 73–82,
https://doi.org/10.1016/j.epsr.2016.12.029.
[12] M.-.K. Tran, M. Akinsanya, S. Panchal, R. Fraser, M. Fowler, Design of a hybrid
CRediT authorship contribution statement electric vehicle powertrain for performance optimization considering various
powertrain components and configurations, Vehicles 3 (2021) 20–32, https://doi.
org/10.3390/vehicles3010002.
Manh-Kien Tran: Conceptualization, Methodology, Software, [13] Y. Cui, C. Du, G. Yin, Y. Gao, L. Zhang, T. Guan, F. Wang, Multi-stress factor model
Formal analysis, Investigation, Data curation, Writing – original draft, for cycle lifetime prediction of lithium ion batteries with shallow-depth discharge,

9
M.-K. Tran et al.
J. Power Sources 279 (2015) 123–132, https://doi.org/10.1016/j.
jpowsour.2015.01.003.
[14] M.-.K. Tran, M. Fowler, A review of lithium-ion battery fault diagnostic algorithms:
current progress and future challenges, Algorithms 13 (2020) 62, https://doi.org/
10.3390/a13030062.
[15] Y. Li, M. Vilathgamuwa, T. Farrell, S.S. Choi, N.T. Tran, J. Teague, A Physics-Based
Distributed-Parameter Equivalent Circuit Model for Lithium-Ion Batteries,
Electrochim. Acta 299 (2019) 451–469, https://doi.org/10.1016/j.
electacta.2018.12.167.
[16] Y. Li, B. Xiong, D.M. Vilathgamuwa, Z. Wei, C. Xie, C. Zou, Constrained ensemble
Kalman filter for distributed electrochemical state estimation of lithium-ion
batteries, IEEE Trans. Ind. Inf. 17 (2021) 240–250, https://doi.org/10.1109/
TII.2020.2974907.
[17] Y. Li, M. Vilathgamuwa, E. Wikner, Z. Wei, X. Zhang, T. Thiringer, T. Wik, C. Zou,
Electrochemical model-based fast charging: physical constraint-triggered PI
control, IEEE Trans. Energy Convers. (2021) 1, https://doi.org/10.1109/
TEC.2021.3065983.
[18] D. Gandolfo, A. Brandão, D. Patiño, M. Molina, Dynamic model of lithium polymer
battery—load resistor method for electric parameters identification, J. Energy Inst.
88 (4) (2015) 470–479, https://doi.org/10.1016/j.joei.2014.10.004.
[19] S. Dey, B. Ayalew, P. Pisu, Nonlinear adaptive observer for a lithium-ion battery
cell based on coupled electrochemical–thermal model, J. Dyn. Syst. Meas. Contr.
137 (11) (2015), 111005, https://doi.org/10.1115/1.4030972.
[20] M.K. Tran, M. Fowler, Sensor fault detection and isolation for degrading lithium-
ion batteries in electric vehicles using parameter estimation with recursive least
squares, Batteries 6 (2020) 1, https://doi.org/10.3390/batteries6010001.
[21] Z. Gao, C. Chin, W. Woo, J. Jia, Integrated equivalent circuit and thermal model for
simulation of temperature-dependent LiFePO4 battery in actual embedded
application, Energies 10 (1) (2017) 85, https://doi.org/10.3390/en10010085.
[22] G.L. Plett, Extended Kalman filtering for battery management systems of LiPB-
based HEV battery packs, J. Power Sources 134 (2) (2004) 252–261, https://doi.
org/10.1016/j.jpowsour.2004.02.031.
[23] B.Y. Liaw, G.;. Nagasubramanian, R.G. Jungst, D.H. Doughty, Modeling of lithium
ion cells—simple equivalent circuit model approach, Solid State Ion. 175 (2004)
835–839, https://doi.org/10.1016/j.ssi.2004.09.049.
[24] J. Jiang, Y. Liang, Q. Ju, L. Zhang, W. Zhang, C. Zhang, An equivalent circuit model
for lithium-sulfur batteries, Energy Procedia 105 (2017) 3533–3538, https://doi.
org/10.1016/j.egypro.2017.03.810.
[25] X. Hu, H. Yuan, C. Zou, Z. Li, L. Zhang, Co-estimation of state of charge and state of
health for lithium-ion batteries based on fractional-order Calculus, IEEE Trans.
Veh. Technol. 67 (11) (2018) 10319–10329, https://doi.org/10.1109/
TVT.2018.2865664.
[26] L. He, M. Hu, Y. Wei, B. Liu, Q. Shi, State of charge estimation by finite difference
extended Kalman filter with HPPC parameters identification, Sci. China Technol.
Sci. 63 (2020) 410–421, https://doi.org/10.1007/s11431-019-1467-9.
[27] J. Li, K. Adewuyi, N. Lotfi, R.G. Landers, J. Park, A single particle model with
chemical/mechanical degradation physics for lithium ion battery State of Health
(SOH) estimation, Appl. Energy 212 (2018) 1178–1190, https://doi.org/10.1016/
j.apenergy.2018.01.011.
10
Journal of Energy Storage 43 (2021) 103252
[28] J. Vetter, P. Novák, M.R. Wagner, C. Veit, K.C. Möller, J.O. Besenhard,
A. Hammouche, Ageing mechanisms in lithium-ion batteries, J. Power Sources 147
(1–2) (2005) 269–281, https://doi.org/10.1016/j.jpowsour.2005.01.006.
[29] R. Hausbrand, G. Cherkashinin, H. Ehrenberg, M. Gröting, K. Albe, C. Hess,
W. Jaegermann, Fundamental degradation mechanisms of layered oxide Li-ion
battery cathode materials: methodology, insights and novel approaches, Mater. Sci.
Eng.: B 192 (2015) 3–25, https://doi.org/10.1016/j.mseb.2014.11.014.
[30] Y. Li, P. Chattopadhyay, A. Ray, C.D. Rahn, Identification of the battery state-of-
health parameter from input–output pairs of time series data, J. Power Sources 285
(2015) 235–246, https://doi.org/10.1016/j.jpowsour.2015.03.068.
[31] F. Feng, R. Lu, G. Wei, C. Zhu, Online estimation of model parameters and state of
charge of LiFePO4 batteries using a novel open-circuit voltage at various ambient
temperatures, Energies 8 (4) (2015) 2950–2976, https://doi.org/10.3390/
en8042950.
[32] T. Feng, L. Yang, X. Zhao, H. Zhang, J. Qiang, Online identification of lithium-ion
battery parameters based on an improved equivalent-circuit model and its
implementation on battery state-of-power prediction, J. Power Sources 281 (2015)
192–203, https://doi.org/10.1016/j.jpowsour.2015.01.154.
[33] C. Zhang, W. Allafi, Q. Dinh, P. Ascencio, J. Marco, Online estimation of battery
equivalent circuit model parameters and state of charge using decoupled least
squares technique, Energy 142 (2018) 678–688, https://doi.org/10.1016/j.
energy.2017.10.043.
[34] X. Lai, S. Wang, S. Ma, J. Xie, Y. Zheng, Parameter sensitivity analysis and
simplification of equivalent circuit model for the state of charge of lithium-ion
batteries, Electrochim. Acta (2019), 135239, https://doi.org/10.1016/j.
electacta.2019.135239.
[35] J. Gomez, R. Nelson, E.E. Kalu, M.H. Weatherspoon, J.P. Zheng, Equivalent circuit
model parameters of a high-power Li-ion battery: thermal and state of charge
effects, J. Power Sources 196 (10) (2011) 4826–4831, https://doi.org/10.1016/j.
jpowsour.2010.12.107.
[36] A. Farmann, D.U. Sauer, A study on the dependency of the open-circuit voltage on
temperature and actual aging state of lithium-ion batteries, J. Power Sources 347
(2017) 1–13, https://doi.org/10.1016/j.jpowsour.2017.01.098.
[37] S.J. Bazinski, X. Wang, The influence of cell temperature on the entropic coefficient
of a lithium iron phosphate (LFP) pouch cell, J. Electrochem. Soc. 161 (1) (2013)
A168–A175, https://doi.org/10.1149/2.082401jes.
[38] M.K. Tran, A. Mevawala, S. Panchal, K. Raahemifar, M. Fowler, R. Fraser, Effect of
integrating the hysteresis component to the equivalent circuit model of lithium-ion
battery for dynamic and non-dynamic applications, J. Energy Storage 32 (2020),
101785, https://doi.org/10.1016/j.est.2020.101785.
[39] H. He, R. Xiong, H. Guo, S. Li, Comparison study on the battery models used for the
energy management of batteries in electric vehicles, Energy Convers. Manage. 64
(2012) 113–121, https://doi.org/10.1016/j.enconman.2012.04.014.
[40] X. Hu, S. Li, H. Peng, A comparative study of equivalent circuit models for Li-ion
batteries, J. Power Sources 198 (2012) 359–367, https://doi.org/10.1016/j.
jpowsour.2011.10.013.
[41] M.-.K. Tran, A. DaCosta, A. Mevawalla, S. Panchal, M. Fowler, Comparative study
of equivalent circuit models performance in four common lithium-ion batteries:
LFP, NMC, LMO, NCA, Batteries 7 (2021) 51, https://doi.org/10.3390/
batteries7030051.