Journal of Power Sources 345 (2017) 21e29

Contents lists available at ScienceDirect

Journal of Power Sources

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

A novel method for identification of lithium-ion battery equivalent

circuit model parameters considering electrochemical properties
Xi Zhang a, *, Jinling Lu a, Shifei Yuan a, Jun Yang a, Xuan Zhou b
a
School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai, China
b
Department of Electrical and Computer Engineering, Kettering University, Flint, USA

h i g h l i g h t s

Propose a combination method of Li-ion battery ECM and electrochemical properties.

Establish a unified transfer-function-style expression between cell voltage and current.

Obtain a relationship between the ECM electrical and electrochemical parameters.

Present a novel method to identify ECM parameters.

Improve battery modeling precision for further application, e.g. SOC estimation.

a r t i c l e i n f o a b s t r a c t

Article history: This paper proposes a novel parameter identification method for the lithium-ion (Li-ion) battery
Received 27 September 2016 equivalent circuit model (ECM) considering the electrochemical properties. An improved pseudo two-
Received in revised form dimension (P2D) model is established on basis of partial differential equations (PDEs), since the elec-
15 January 2017
trolyte potential is simplified from the nonlinear to linear expression while terminal voltage can be
Accepted 28 January 2017
Available online 6 February 2017
divided into the electrolyte potential, open circuit voltage (OCV), overpotential of electrodes, internal
resistance drop, and so on. The model order reduction process is implemented by the simplification of
the PDEs using the Laplace transform, inverse Laplace transform, Pade approximation, etc. A unified
Keywords:
Lithium-ion battery
second order transfer function between cell voltage and current is obtained for the comparability with
Electrochemical properties that of ECM. The final objective is to obtain the relationship between the ECM resistances/capacitances
Equivalent circuit model and electrochemical parameters such that in various conditions, ECM precision could be improved
Pade approximation regarding integration of battery interior properties for further applications, e.g., SOC estimation. Finally
SOC estimation simulation and experimental results prove the correctness and validity of the proposed methodology.
© 2017 Elsevier B.V. All rights reserved.

1. Introduction with precise battery modeling methods is necessary. Batteries can

be modeled using the PDE model-based, equivalent circuit or
Lithium-ion (Li-ion) batteries have been commanding a high empirical approaches. In an empirical approach, a cell model is built
market share replacing lead-acid and nickel-metal hydride (NiMH) and tested iteratively to make the modeling error approach to the
batteries, owing to their better specific energy, specific power and minimum gradually, which is time-consuming and expensive.
relatively longer lifespan [1e4]. Li-ion batteries have been suc- In an equivalent circuit model (ECM), resistances, capacitances
cessfully applied into the pure electric vehicle (EV) and hybrid and voltage sources are used to describe charging and discharging
electric vehicle (HEV) in terms of reducing the environmental processes and the model is built in frequency- or time-domain
pollution or overcoming the energy shortage. [5e7]. The Thevenin model is the most popular time-domain
To ensure safety and accuracy of state estimation and avoid model among regular models such as the Rint, PNGV, and GNL
energy and time wasting, the battery management system (BMS) models, and meanwhile the electrochemical impedance spectros-
copy (EIS) is the most widely used in the frequency-domain [8]. A
relatively low-order equivalent circuit model (e.g., second order
* Corresponding author. circuit model) is an optimal choice due to its low computational
E-mail address: [email protected] (X. Zhang). complexity and high calculation accuracy whose parameters can be

http://dx.doi.org/10.1016/j.jpowsour.2017.01.126
0378-7753/© 2017 Elsevier B.V. All rights reserved.
22 X. Zhang et al. / Journal of Power Sources 345 (2017) 21e29

identified via various methods, e.g., the recursive least square (RLS) the single electrode, which makes modeling more realistic. With
method [9e13]. When loads change, the RLS model might be overall consideration of the calculation difficulties and accuracy,
imprecise enough without relationship of interior electrochemical the proposed improved P2D model is a better choice.
properties. Therefore, further applications (e.g., SOC estimation) This paper is organized as follows. In Section 2, the improved
using the ECM parameters will be influenced negatively. P2D model is established. In Section 3, the ECM resistances/ca-
In a PDE model-based approach, modeling starts with electro- pacitances are derived taking into account the comparability be-
chemical properties based on the partial differential equations tween the ECM transfer function and that of the improved P2D
considering the electrochemical kinetics, the charge diffusion and model. In Sections 4 and 5, the simulation and experiment results
transferring process, and the conservation of mass and charge. The are given along with detailed analysis. The conclusions are drawn
basic pseudo two-dimensional (P2D) model is based on the as- finally in Section 6.
sumptions of fixed particle radius of the electrodes, considering the
diffusion and immigration only. The electrochemical models are the
processes of simplifying or reforming the equations based on the 2. Model establishment
P2D model [14e16]. P2D model is actually an equation set with 5
partial differential equations, and the equation contains about 20 or As shown in Fig. 1, the Li-ion battery is assumed to be con-
more parameters closely related to the microscopic changes inside structed by two electrodes composed of a large number of uniform
the battery, such as electrolyte diffusion coefficient, electrode re- radius balls and the separator. In the positive and negative elec-
action coefficient, etc. All these parameters may change when the trodes, the active materials are oxidized or reduced, and meanwhile
reactions happen under various cycles, so the P2D model involving the lithium ions are de-intercalated from or intercalated to the
these parameters could reflect the electrochemical mechanism, electrode and migrated through the electrolyte. The active mate-
resulting in better modeling precision due to the integration of rials in the electrodes, are both bonded to the current collector and
realistic and interior factors. Thus some further applications (e.g., electric isolated by the separate film. The particle radius of the
SOC estimation) could benefit from this. Many efforts have been active materials is defined as R. The parameters are assumed to
made to find the relationship between the output cell voltage and change only in the direction of x, and L stands for the distance
the input current [17e25]. To simplify the calculation and reduce between the negative and positive electrodes in the battery. The
the number of related parameters, a single particle model (SPM) potentials or concentrations are related to the x position and time
consist of some ordinary differential equations (ODEs) was pro- while the ambient temperature is assumed to be constant.
posed [26]. The SPM is based on the assumption that both the two
electrodes are regarded as a single sphere with a certain radius. The
electrochemical model reflects the initial changes of the electrodes 2.1. Base model
and electrolytes well, but the parameters are difficult to acquire,
and in the meantime, the calculation process is time-consuming Four PDEs based on the conservation of species and charge in
and complicated [27]. the electrolyte and electrodes govern the dynamics of battery. The
Considerable researchers have been focusing on the ECM and PDEs are linked with the Butler-Volmer equation. In this study, the
electrochemical models respectively. A novel ECM was proposed model is regarded as a P2D model since the adjacent particles are
using a two-level state-of-power prediction algorithm in Ref. [28]. not coupled directly. One dimension is the radius r, and the other is
In Ref. [29], a fractional-order model was presented with verifica- x. The base model is established as follows.
tion in different drive cycles. A thermal-electrochemical P2D model
was raised up whose state estimation was based on orthogonal
2.1.1. Conservation of species
collocation and a modified extended Kalman filter in Ref. [30]. In
According to the porous electrode theory, the electrodes parti-
Ref. [31], an improved theoretical electrochemical-thermal
cles, namely the solid phase, are assumed to distribute uniformly.
modeling strategy was studied considering the battery packs in
The expression of conservation of the lithium ions at the electrodes
electric vehicles. However, the combination of the ECM and elec-
based on Fick's second law and the boundary conditions are
trochemical models is seldom seen.
delineated as
In this paper, an improved pseudo two-dimension (P2D) model
is proposed to create a close connection between the ECM and
electrochemical properties in terms of transfer function, due to
simplification of electrolyte potential from the nonlinear to linear
LixC6 Li1-xCoO2
expression. This improved modeling method is easier to be realized Negative Separator Positive
and meanwhile guarantee high precision. The basic equations are
derived from the PDE-based P2D model. Some mathematical
Current Collecter

Current Collecter

methods including the Laplace transform, inverse Laplace trans-

form and Pade approximation have been employed. After PDEs'
transformation and corresponding term combination and elimi-
Discharge
nation, a second-order transfer-function-style expression is then
achieved such that it is comparable with that derived from an ECM.
Li+
Then the ECM resistances and capacitances are achieved with ex-
pressions by electrochemical parameters. This is probably benefi-
Charge
cial for improvement of SOC evaluation accuracy. Finally
verification is made via comparison between the numerical, x δ- δsep δ+
COMSOL and experimental results. Compared to the basic P2D R
model, the improved one simplifies the equations by linearization Active Materials
i
and Pade approximation for further transfer function form L
expression acquisition. With comparison to single particle model
(SPM), it additionally considers the active particle interaction inside Fig. 1. Schematic diagram of the Li-ion battery.
 X. Zhang et al. / Journal of Power Sources 345 (2017) 21e29 23

 2.1.3. Reaction kinetics

vcs Ds v vcs The Butler-Volmer equation is the bridge of the four afore-
¼ 2 r2 (1)
vt r vr vr mentioned PDEs which can be described by
 



vCs  vCs  j j ¼ i0 exp
aa F aF
h  exp  c h (10)
 ¼ 0 ; Ds ¼ (2)
vr r¼0 vr r¼Rs F RT RT

where cs and Cs are the concentration of electrode in time and where aa and ac are the transfer coefficients, and
frequency domains, respectively, r is the radial coordinate, Ds is the i0 ¼ FkðCs;max  Cs;surf Þaa Cs;surf
aa
Ceac is the exchange current density
diffusion coefficient of the solid phase, j is the intercalation current and h is overpotential in time domain.
density, and F (96487C/mol) is the Faraday constant. The overpotential generated at the interface of the electrode and
Conservation of the lithium ions in electrolyte yields electrolyte is

  h ¼ fs  fe  U (11)
vce v2 ce as 1  t0þ
εe ¼ Deff
e þ j (3)
vt vx2 F where U is the equilibrium potential.
At the current collectors, the boundary conditions are
  2.2. Derivation of improved P2D model
vCe  vCe 
 ¼ ¼0 (4)
vx x¼0 vx x¼L The main objective is to obtain the expression of the battery
voltage vs various input currents. Based on the P2D model, the
where ce and Ce are the concentration of electrolyte in time and output voltage can be calculated using the open circuit voltage, the
frequency domain, εe is the electrolyte volume fraction, overpotential, the electrolyte potential and the initial voltage drop.
Deff
e ¼ De εe
1:5 is the effective electrolyte diffusion coefficient,

as ¼ 3εs =Rs is the specific interfacial surface area, and t0þ is the li- 2.2.1. Open circuit voltage
ion transference number. The subscript “s” and “e” indicate the Referring to Equation (1) and utilizing the Laplace transform and
solid and electrolyte phase. the variable substitution, the relationship of the solid phase con-
centration and the current density can be given by
0  pffiffiffiffiffiffiffiffiffiffi 1
2.1.2. Conservation of charge Cs ðsÞ Rs @ tanh Rs s=Ds
Conservation of charge in the electrodes according to Ohm's law ¼  pffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiA (12)
JðsÞ FDs tanh Rs s=Ds  Rs s=Ds
yields

v2 f s where JðsÞ is the Laplace transform of the intercalation current

seff  as j ¼ 0 (5) density j .
vx2
The intercalation current density is proportional to the volume
The boundary conditions at the current collectors and the
current density expressed by
separator are
    Jf ;s ðsÞ ¼ as $JðsÞ (13)
vfs  
eff vfs  I vfs  vfs 
seff ¼ s ¼ ; ¼ ¼0

vx x¼0 þ
vx x¼L A vx x¼d vx x¼d þdsep The relationship of the volume current density and current at
the negative and positive electrodes are
(6)
Zd
where fs is the potential of electrode, seff ¼ sεs is the effective IðsÞ
electrode conductivity. Jf ;n ðsÞ ¼ (14)
A
Conservation of electrolyte is expressed as 0

v2 f e v2 lnCe ZL
keff 2
þ keff
d þ as j ¼ 0 (7) IðsÞ
vx vx2 Jf ;p ðsÞ ¼  (15)
A
To simplify the calculation process, Equation (7) is linearized as d þdsep
below
Then the intercalation current density at the electrodes are
2
v fe keff 2
v Ce described as
keff þ d
þ as j ¼ 0 (8)
vx2 Ce;0 vx2 IðsÞ
Jn ðsÞ ¼ (16)
The boundary conditions are Ad as
 
vfe  vfe  IðsÞ
¼ ¼0; Jp ðsÞ ¼  (17)
vx x¼0 vx x¼L Adþ as


 (9)
vf vCe  eff vfe

eff vCe 
keff e þ keffd vx  ¼ k þ kd vx 
vx x vx xþ
2.2.2. Overpotential
where fe is the potential of electrolyte, k is the ionic conductivity, With the Laplace transform of Equation (10), we can obtain the
keff ¼ kε1:5
e is the effective ionic conductivity, and following expression between the intercalation current density and
keff ¼ 2RTkeff ðt0  1Þ is effective diffusion ionic conductivity. overpotential
d F þ
24 X. Zhang et al. / Journal of Power Sources 345 (2017) 21e29

Table 1
Transfer function between the electrolyte concentration and the input current.

Positive (x ¼ L) Negative (x ¼ 0)

1st 
0:1240 DLe
 
0:1170 DLe

2 2
AF 1þ0:1052 DL e s AF 1þ0:0997 DL e s
2nd 3
0:1240 DLe þ0:0012 L 2 s
3
0:1170 DLe þ0:0010 L 2 s
 D
e  D
e
2 4 2 4
AF 1þ0:1148 DL e sþ0:0012 L 3 s2 AF 1þ0:1086 DL e sþ0:0011 L 3 s2
D D
e e

standard transfer function format. The approximation error de-

RT creases with the increasing order as shown in Fig. 2.
NðsÞ ¼ JðsÞ (18)
Fi0 ðaa þ ac Þ From this figure, we can see that when frequency is low, the
approximation fits the primitive curve well. As the Li-ion battery
works at a relatively low frequency state, a first or second order
approximation of the electrolyte concentration is sufficient. For a
2.2.3. Electrolyte potential
first order expression, the final cell voltage expression is transferred
Derive the Laplace transform of Equation (8) by integrating x
to second order since there are two electrodes in a battery. How-
from 0 to L with the assumption that the concentration and po-
ever, for a second order expression, the order of the cell voltage
tential of the electrolyte are independent. If the concentration is
expression reaches four which is too complicated. Thus we choose
obtained, the potential can be achieved as follows.
the first order electrolyte concentration expression and the verifi-
  cation later to prove the accuracy acceptable.
~ e ðx; sÞ ¼ 2RT 1  t0þ x
F ðCe ðx; sÞ  Ce ð0; sÞÞ þ ∬ 0 as JðsÞdx (19)
Ce;0 F
2.2.7. Transfer function expression
The difference of electrolyte potential from the negative to The cell voltage of the electrochemical model is based on the
positive electrode is given as follows. block diagram shown in Fig. 3. The cell voltage consists of the open
  circuit voltage part and impedance part which represents the
~ e ðL; sÞ ¼ 2RT 1  t0þ IðsÞ d 2dsep transient properties of the battery. In conclusion, the second order
F ðCe ðL; sÞ  Ce ð0; sÞÞ  þ eff
Ce;0 F 2A keff
 ksep transfer function expression of the voltage-current regardless of
! the open circuit voltage part is expressed by
dþ
þ !
keff
þ Ud1 ðsÞ RT 1 1 1 d
¼ Rf þ þ þ
(20) IðsÞ AFðaa þ ac Þ as;p i0;p dþ as;n i0;n d 2A keff

!
When calculating the electrolyte concentration, according to the 2dsep dþ
different values of the parameters in the negative and positive þ eff þ eff
ksep kþ
electrodes and the separator, we can divide the equation into three
  2
0 2
parts and each is an ordinary differential equation. The intercala- 2RT 1  tþ 0:1240 DLe
þ 4 
tion current density j in Equation (3) can be replaced by the input Ce;0 F 2

current IðsÞ based on Equation (16) or (17). Then the differential

εe;p AF 1 þ 0:1052 L s De
3
equation can be solved with the boundary conditions (4) and the
0:1170 DLe
continuity of the concentration curve. Table 1 shows the expression þ  5
2
with a fixed proportion of the length of the positive electrode, εe;n AF 1 þ 0:0997 DL e s
separator and negative electrode.
(23)

2.2.4. Resistance voltage drop

When the resistance of the battery is equal to Rf , the resistance
voltage drop is depicted as
3. Acquisition of ECM parameters
Vf ðsÞ ¼ Rf IðsÞ (21)
For a certain battery, the cell voltage change with time is
consistent. This relationship is used to calculate the electrical pa-
rameters of ECM considering the electrochemical parameters
2.2.5. Terminal voltage which have close connections with the internal changes of the
With respect to the above equations, the final terminal voltage battery solid and electrolyte phase.
in frequency domain is given by



 3.1. Transfer function of ECM
Cs;pos ðsÞ Cs;neg ðsÞ
Vcell ðsÞ ¼ Up  Un  ðNþ ðsÞ þ N ðsÞÞ
cs;p;max cs;n;max To ensure the comparability between the ECM and the elec-
 
 Fe;þ ðsÞ  Fe; ðsÞ  Rf IðsÞ trochemical model, a classified 2nd order Thevenin model is chosen
(22) as shown in Fig. 4. According to the electrochemical reactions, the
diffusion and migration at the electrodes are assumed to be realized
by paralleled resistance and capacitance parts respectively, and the
added resistance compensates the changes of the electrolyte.
2.2.6. Order selection Based on the schematic, the cell voltage V and the transient
Pade approximation is used to simplify the expressions into voltage Ud2 are expressed by
 X. Zhang et al. / Journal of Power Sources 345 (2017) 21e29 25

Fig. 2. Bode diagram of electrolyte concentration.

V ¼ Voc þ Up þ Un þ R0 I (24) Ud1 ðsÞ Ud2 ðsÞ

¼ (27)
IðsÞ IðsÞ
Ud2 ¼ V  Voc ¼ Up þ Un þ R0 I (25)
 2 L
Then the transfer function expression is obtained as follows 0
0:2480RT 1  tþ
De
Ud2 ðsÞ Rp Rn Rp ¼
¼ R0 þ þ (26) εe;p Ce;0 AF 2
IðsÞ 1 þ Rp Cp s 1 þ Rn Cn s
0:1052εe;p LCe;0 AF 2
To obtain the complete model, we need to identify five param- Cp ¼  2 (28)
0
eters contained in the equation before. The most commonly 0:2480RT 1  tþ
method is call as Recursive Least Square (RLS) method. This method 
0
2 L
is based on the calculation of some matrix and the measured data of 0:2340RT 1  tþ
De
the input current and output voltage [9e13]. The disadvantage is Rn ¼
that a large scale of data is needed to pursue the accuracy of the εe;n Ce;0 AF 2
identification.
0:0997LCe;0 AF 2
Cn ¼  
3.2. Parameters identification of ECM 0:2340RT 1  tþ 0 2

In another way, we raised up a new method to do the identifi- !

cation. Combining the electrochemical transfer function and the RT 1 1 1 d 2dsep
equivalent circuit forms, the relationship between the resistance/ R0 ¼ þ þ þ eff
AFðaa þ ac Þ as;p i0;p dþ as;n i0;n d 2A keff
 ksep
capacitance and the electrochemical parameters can be established !
to better reflect interior mechanism. dþ
þ þ Rf
keff
þ

This combination gives a new method to estimate the ECM

I
parameters which makes these parameters full of physical

I Rp Rn
η
_ _
s

+ Up + Un
I ( s)
+
R0
s

Cp Cn Voc
I
2 RT I ( s) − φe V
+ − − − + + +
Ce F A k− k sep k+

Rf
Vf _
Fig. 3. Block diagram of the electrochemical model. Fig. 4. Schematic diagram of the ECM.
26 X. Zhang et al. / Journal of Power Sources 345 (2017) 21e29

meanings, and the accuracy could be improved involving integra-

tion of electrochemical changes in various conditions.

4. Simulation results

For simulation, the type of LiFePO4 battery is selected for

simulation with negative electrode of carbon. The cell nominal
voltage is 3.2 V and nominal capacity is 20 Ah. The negative and
positive open circuit voltages are known and the main parameters
are listed in Table 2.
The software COMSOL and MATLAB are used to build the model
based on the block diagram in Fig. 3. The results by COMSOL are
considered as benchmark for comparison since COMSOL has been
commonly regarded as a mature and realistic software with satis-
factory precision of electrochemical and electrical parameter re-
sults. Therefore the comparison between other models and
COMSOL model is meaningful and convincible.
The standard model “li_battery_1d” in the COMSOL library is
utilized, and the input current is constant during charging and Fig. 5. Voltage changes with the parameter.
discharging. The electrochemical parameters have visible in-
fluences on the voltage. Since the number of these parameters is
very large, the classification is essential. They are divided into three charging curve of ECM with the proposed improved P2D model fits
categories: the COMSOL model (considered as a benchmark) better than that of
1) Physical constants for all batteries such as F, R, T and t0þ . 2) ECM with RLS, although the cell voltage errors between the two
Initial value or constant parameters for a certain battery e.g., Ce;0 , ECMs and the COMSOL model are both within the maximum esti-
Cs;max , d , etc. 3) Parameters that change with direction, time or mation error of 0.1 V. This shows better estimation accuracy of the
input current e.g., Ds;neg , Ds;pos , De , etc. proposed improved P2D model.
The parameters that have significant influences on the voltage
belong to the third category. Taking Ds;neg as an example, the
voltage changes at various values of Ds;neg are shown in Fig. 5. 5. Experimental results
When Ds;neg increases, the discharging voltage increases and vice
versa. This indicates that the precision of these parameters is In the experimental setup, the cell voltage and current are
important. recorded by the MKL cell tester with a linked PC. The experiments
The simulations are done considering the following four cases: are done in the following three cases: Case I, 1C discharging; Case II,
Case I, 1C (20 A) discharging; Case II, 0.5C (10 A) discharging; Case 0.5C discharging; Case III, 1C pulse discharging for 3 cycles.
III, pulse charging by 1C rate for 3 cycles; Case IV, DST test. The ECM The verifications are completed in the dSPACE based hardware-
using the RLS method is built for comparison, since it is the most in-loop (HIL) system with the ControlDesk software and the
widely used model for BMS to evaluate battery properties in EVs, MicroAutoBox. ControlDesk is the interface of hardware and soft-
HEVs and PHEVs. The ECM parameters e.g. R0, Rp, Cp, etc, with the ware. Experimental current and voltage data of the battery are
RLS method are calculated on basis of MATLAB, which are shown in measured and imported to MicroAutobox, while the current signals
Table 3. We choose a second order ECM for the comparison be- should be transformed to voltage signals first via a current sensor.
tween the ECMs with the proposed method and with RLS. In Fig. 6 Then the results of the ECMs built in the MicroAutobox using
charge/discharge current and cell voltage using COMSOL model, electrochemical properties and RLS respectively are imported into
ECM with RLS, ECM with the proposed method are shown, along ControlDesk for the display and data recording. Thus the compar-
with the errors between the two ECMs and COMSOL model. isons can be made.
From these figures, we can obviously see that the discharging or Before the experiments, tests are done first for OCV measure-
ment. The OCV is recorded after the battery is charged or dis-
charged at 0.3C rate by every 5% capacity and takes a rest for
Table 2
60 min. Both charging and discharging OCVs are detected to
Electrochemical parameters of the LiFePO4 battery.
calculate the average value, which is more accurate.
Symbol Units Positive Separator Negative The MATLAB file with updated electrochemical parameters is
Ds m2 s1 4.27 1015 e 3.9 1014 loaded into ControlDesk, and the data when the battery is dis-
De m2 s1 2.31 1010 2.31 1010 2.31 1010 charged at various currents is recorded, which is shown in Fig. 7.
d mm 88 40 83 From these figures, we can see cell voltage absolute errors be-
R mm 14.75 e 1.15 tween the two ECMs and the test results are both within the in-
A m2 1 1 1
terval of [0, þ0.06 V]. However the ECM using the proposed
k mol1=2 m5=2 s1 3.626 1011 e 1.764 1011
improved P2D model fits the test result better than the ECM with
s Sm 1 0.1 e 100
RLS.
εs e 0.435 e 0.56
εe e 0.306 0.54 0.268
t0þ e 0.363 0.363 0.363
Table 3
Cs;max mol m3 19500 31370
ECM parameters with RLS method of the LiFePO4 battery.
Cs:0 3 600 14870
mol m
Parameter R0 Rp Cp Rn Cn
Ce:0 mol m3 1500 1500 1500
a e 0.5 e 0.5 Value 3:8mU 4 mU 4434F 1.9mU 2820F
 X. Zhang et al. / Journal of Power Sources 345 (2017) 21e29 27

Fig. 6. Cell voltage comparison for simulation: a) Case I, b) Case II, c) Case III, d) Case IV.
28 X. Zhang et al. / Journal of Power Sources 345 (2017) 21e29

Fig. 7. Cell voltage comparison for experiments: a) Case I, b) Case II, c) Case III.

6. Conclusions properties, since it has lower error compared to ECM with RLS and
provides a precise methodology to describe and analyze the
In this paper, a novel battery model is presented combining the microscopic changes of the battery during charging and discharg-
ECM transfer function format and electrochemical properties for ing processes. This model combines the electrochemical properties
equivalent electrical parameter identification and update. This and equivalent circuit to study lithium-ion battery performance in
model contains several parts whose simplification and classifica- different conditions, which better reflects the battery interior
tion are obtained on basis of mathematical schemes including the mechanism. Moreover the proposed model as an alternative of
Laplace transform, inverse Laplace transform and Pade approxi- previous models provides a possibility to improve BMS perfor-
mation. The final transfer function with the battery voltage vs mance in some extent.
different current inputs is achieved for comparability with that of
an ECM. Then the ECM capacitances and resistances are achieved
based on the electrochemical parameters, and the ECM precision is References
improved for further applications, e.g. SOC estimation. From the
[1] M.U. Cuma, T. Koroglu, J. Power Sources 42 (2015) 517e531.
simulation and experimental results, we can conclude that the [2] K. Snyder, X. Yang, T. Miller, SAE Technical Paper 2009-01-1385, 2009.
proposed model is suitable for accurately evaluating the battery [3] F. Lima, J. Ramalho, D. Tavares, J. Duarte, In: Proceedings of the European 2003
 X. Zhang et al. / Journal of Power Sources 345 (2017) 21e29 29

Solid-State Circuits Conference, pp.209e212. 2014 Emerging & Selected Topics in Power Electronics IEEE Journal, 2(3):
[4] A. Seamana, T. Dao, J. McPhee, J. Power Sources 256 (2014) 410e423. 659e677.
[5] G. Liu, L. Lu, H. Fu, J. Hua, J Li, M. Ouyang, Y. Wang, S.Xue, P. Chen, In: Pro- [19] R. Ahmed, M. E. Sayed, I. Arasaratnam, J. Tjong, S. Habibi, In: Proceddings of
ceedings of 2014 ITEC, IEEE, Asia-Pacific. 2014 Emerging & Selected Topics in Power Electronics IEEE Journal, 2(3):
[6] V. Knap, D. Stroe, R. Teodorescu, M. Swierczynski, T. Stanciu, In: Proceedings of 678e690.
2015 Energy Conversion Congress and Exposition (ECCE), IEEE. [20] S.K. Rahimian, S. Rayman, R.E. White, J. Power Sources 224 (2013) 180e194.
[7] P.L. Moss, G. Au, E.J. Plichta, J.P. Zheng, J. Electrochem. Soc. 155 (2008) [21] P. Northrop, V. Ramadesigan, S. De, V. Diwakar, J. Electrochem. Soc. 158 (2011)
A986eA994. A1461eA1477.
[8] F. Ciucci, Electrochim. Acta 87 (2013) 532e545. [22] P. Gambhire, N. Ganesan, S. Basu, K.S. Hariharan, S.M. Kolake, T. Song, D. Oh,
[9] X. Hu, F. Sun, Y. Zou, H Peng, In: Proceedings of 2011 American Control T. Yeo, S. Doo, J. Power Sources 290 (2015) 87e101.
Conference on San Francisco, CA, USA, pp.935e940. [23] J.C. Forman, S. Bashash, J.L. Stein, H.K. Fathy, J. Electrochem. Soc. 158 (2011)
[10] Y. Chiang, W. Sean, J. Ke, J. Power Sources 196 (2011) 3921e3932. A93eA101.
[11] J. V. Barreras, C. Pinto, R. De Castro, E. Schaltz, Swierczynski M., S. J. Andreasen, [24] K.A. Smith, C.D. Rahn, C. Wang, Energy Convers. Manag. 48 (2007)
R. E. Araujo, In: Proceddings of 2015 Sustainable Mobility Applications, Re- 2565e2578.
newables and Technology (SMART), IEEE. [25] S. Y. Choe, X. Li, M. Xiao, In: Proceedings of 2013 Electric Vehicle Symposium
[12] D. Sun, X. Chen, In: Proceddings of 2014 17th International Conference on and Exhibition, IEEE, pp.1e11.
Electrical Machines and Systems (ICEMS), IEEE, Hangzhou, China. [26] M. Guo, G. Sikha, R.E. White, J. Electrochem. Soc. 158 (2011) A122eA132.
[13] Y. Hu, S. Yurkovich, J. Power Sources 196 (2011) 2913e2923. [27] P. Kemper, S.E. Li, D. Kum, J. Power Sources 286 (2015) 510e525.
[14] S. J. Moura, N. A. Chaturvedi, M. Krstic, In: Proceedings of 2012 American [28] T. Feng, L. Yang, X. Zhao, H. Zhang, J. Qiang, J. Power Sources 281 (2015)
Control Conference, pp.559e565. 192e203.
[15] S. J. Moura, N. A. Chaturvedi, M. Krstic, In: Proceedings of 2012 American [29] H. Wu, S. Yuan, C. Yin, in: Proceedings of FISITA 2012 World Automotive
Control Conference, pp.566e571. Congress, 2012, pp. 795e805.
[16] J. Forman, S. Bashash, J. Stein, H. Fathy, J. Electrochem. Soc. 158 (2011) [30] A.M. Bizeray, S. Zhao, S.R. Duncan, D.A. Howey, J. Power Sources 296 (2015)
A93eA101. 400e412.
[17] Y. Hu, S. Yurkovich, Y. Guezennec, B. Yurkovich, J. Power Sources 196 (2011) [31] P. Amiribavandpour, W. Shen, D. Mu, A. Kapoor, J. Power Sources 284 (2015)
449e457. 328e338.
[18] R. Ahmed, M. E. Sayed, I. Arasaratnam, J. Tjong, S. Habibi, In: Proceddings of