Professional Documents
Culture Documents
Single-Particle Model For A Lithium-Ion Cell: Thermal Behavior
Single-Particle Model For A Lithium-Ion Cell: Thermal Behavior
The single-particle model presented by Santhanagopalan et al. 关J. Power Sources, 156, 620 共2006兲兴 is extended to include an
energy balance. The temperature dependence of the solid phase diffusion coefficient of the lithium in the intercalation particles, the
electrochemical reaction rate constants, and the open circuit potentials 共OCPs兲 of the positive and negative electrodes are included
in the model. The solution phase polarization is approximated using a nonlinear resistance, which is a function of current and
temperature. The model is used to predict the temperature and voltage profiles in a lithium-ion cell during galvanostatic operations.
The single-particle thermal model is validated by comparing the simulated voltage and temperature profiles to the results obtained
using a distributed porous electrode model. The simulation results from the single-particle thermal model also show good
agreement with experimental voltage data obtained from lithium-ion pouch cells under different discharge rates 共C/33, C/2 and C兲
at different temperatures 共15, 25, 35, and 45°C兲.
© 2010 The Electrochemical Society. 关DOI: 10.1149/1.3521314兴 All rights reserved.
Manuscript submitted January 25, 2010; revised manuscript received July 22, 2010. Published December 15, 2010;
publisher error corrected February 3, 2011.
Several thermal models for lithium-ion cells have been presented this work, the single-particle model is extended to include thermal
in the literature; for example, Newman and Pals1,2 presented cell and effects by adding the energy balance equation to the SP model.
battery stack thermal models incorporating the pseudo two- Instead of using a two term polynomial approximation, the solid
dimensional porous electrode model3,4 共P2D model兲 with an energy phase diffusion equations are solved by the eigenfunction expansion
balance,5,6 in which the isothermal and adiabatic discharge behav- method, which improved the accuracy of the model. Parameters in
iors were studied. In 1999, Botte et al.4 developed a pseudo two- this SP thermal model are estimated by fitting the simulated dis-
dimensional 共2D兲 thermal model for LixC6 /LiNiO2 cells, in which charge curves 共up to 1C rate兲 with the experimental data obtained on
the temperature dependence of parameters7-10 were considered and lithium-ion pouch cells. Also, good agreement between the SP ther-
the effects of parameters on the predicted cell temperature were mal model and the PP thermal model 共presented in Ref. 15兲 is ob-
included. In 2003, Srinivasan and Wang11 developed a 2D thermal tained.
model for large lithium-ion cells, in which the energy balance equa-
tion and the porous electrode model with polynomial approxima-
tions 共PP model兲 for the solid phases were integrated. In this model, Mathematical Model
experimentally measured entropy data12-14 for electrochemical reac-
tions were also incorporated. In 2008, Kumaresan et al.15 developed The solid phase diffusion.— In the SP model, electrodes are as-
a thermal model for the LiCoO2-mesocarbon microbead 共MCMB兲 sumed to consist of spherical intercalation particles with identical
pouch cells based on the PP model and obtained good agreement size. The key assumption in the single-particle model is that the
between model predictions and experimental data. current distribution is taken to be uniform along the thickness of the
A disadvantage common to the P2D model and the PP model is porous electrode. Newman and Tobias 17 derived expressions to
the long simulation time due to the large number of nonlinear equa- calculate the reaction current distribution along the thickness of po-
tions, so these models become computationally inefficient for simu- rous electrode. To validate the assumption of uniform current den-
sity distribution, the reaction current distribution was evaluated for
lating conditions such as cycling behavior and series/parallel con-
the case of ohmically limited system 关using Eq. 18 in Ref. 17兴 using
figuration of stacked cells in battery packs. To improve
the parameter values for this system. The calculations showed that
computational run time without compromising accuracy, the single-
the reaction current distribution due to the electrochemical reaction
particle model 共SP model兲 共Ref. 3 and 16兲 was proposed. The SP
was fairly uniform. Therefore, each entire porous electrode
model ignores the detailed distribution of local concentration and
共positive/negative兲 can be represented by a single intercalation par-
potential in the solution phase and instead accounts for a lumped
ticle. The mass balance of lithium ions in an intercalation particle of
solution resistance term. Furthermore, the local reaction currents electrode active material is described by Fick’s second law in a
across the porous electrode are assumed to be constant, which al- spherical coordinate system16
冉 冊
lows treatment of a porous electrode as a large number of single
particles, all of which are subjected to the same conditions. These cs,j Ds,j 2 cs,j
assumptions are reasonable for low applied current densities, thin = 2 r 关1兴
t r r r
electrodes, and highly conductive electrodes. In such cases the over-
potential is primarily affected by the diffusion in the solid state. At where cs,j is the concentration of lithium ions in the intercalation
high current densities, the concentration gradients in the electrolyte particle, t is time, r is the radius direction coordinate, Ds,j is the solid
become important. The model presented here does not include these phase diffusion coefficient, which is a function of temperature, and
concentration gradients and is consequently limited to low to mod- the subscript j = p/n represents the positive/negative electrode.
erate current densities. These assumptions simplify the model equa- The maximum concentration is determined by and m. The
tions significantly. The SP model 共using a two term polynomial ap- maximum concentration for lithium in the particles is calculated
proximation兲 shows good agreement with the detailed PP model for based on the density, molecular mass, and the maximum lithium
charge/discharge below 1C, where C denotes the cell capacity.3 In stoichiometry of the electrode
Downloaded 22 Feb 2011 to 129.252.25.182. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp
Journal of The Electrochemical Society, 158 共2兲 A122-A132 共2011兲 A123
冉 Ds,j
cs,j
r
冊 r=0
=0 关3兴
共j = p,n兲
where 1,j is the solid phase potential, 2,j is the solution phase
关12兴
and the other boundary condition at particle surface 共r = Rj兲 is potential, and Uj is the OCP depending on both the surface SoC and
冉 冊
temperature. In this work, the solution phase was simplified as a
cs,j nonlinear resistor, so the potential drop in the solution phase be-
Ds,j = − Jj 关4兴 tween the positive and negative electrodes can be expressed as
r r=Rj
2,p − 2,n = IRcell 关13兴
where Ji is the molar flux of lithium ions at the surface and can be
expressed as The electrolyte resistance, Rcell, is actually determined by a complex
mass and charge transfer processes in the solution phase;18 the value
I I of this parameter is assumed to depend on cell temperature and
Jp = Jn = − 关5兴
FSp FSn current; the approximation of Rcell is discussed later. The cell voltage
where I is the total current passing through the cell tabs and is equals to the difference of solid phase potential between the positive
defined in this work as positive for charge process and negative for and negative electrodes
discharge process and Sj is the total electroactive surface area of Vcell = 1,p − 1,n 关14兴
electrode j
By using the inverse hyperbolic function on Eq. 11, a more explicit
3jVj expression of cell voltage is
冉冑 冊
Sj = 关6兴
Rj 2RT m2p + 4 + mp
Vcell = Up共xp,surf,T兲 − Un共xn,surf,T兲 + ln
where j is the volume fraction of solid phase active material in F 2
冉冑 冊
electrode j and Vj is the total volume of that electrode.
The state of charge 共SoC兲 for positive and negative electrodes is 2RT m2n + 4 + mn
+ ln + IRcell 关15兴
defined as the local lithium-ion concentration divided by the maxi- F 2
mum intercalable lithium-ion concentration
where
cs,j
xj = I
cs,j,max mp =
FkpSpcs,p,maxc0.5
e 共1 − xp,surf兲0.5xp,surf
0.5
共j = p,n兲 关7兴 I
The initial and surface SoCs for positive and negative electrodes mn =
FknSncs,n,maxc0.5
e 共1 − xn,surf兲0.5xn,surf
0.5
are defined, respectively, as
The format of Eq. 15 enables us to directly obtain the analytical
cs,ini,j Jacobian matrix of Vcell with respect to parameters, and the compu-
xini,j =
cs,j,max tation efficiency in the parameter estimation is greatly improved.
共j = p,n兲 关8兴 The energy balance.— In this model, the spatial temperature dis-
tribution in cell is neglected, so the cell temperature T is a function
兩cs,j兩r=Rj of time only. The general energy balance equation is given as Eq. 24
xj,surf = in Ref. 2, and in this work, it is written as
冋 册
cs,j,max
dT Up Un
vC p = IT 共xp,surf兲 − 共xn,surf兲 + I共p − n + IRcell兲 − q
共j = p,n兲 关9兴 dt T T
As shown in Eq. 8 and 9, the initial SoC xini,j is a constant while the 关16兴
surface SoC xj,surf is only a function of time. The initial cell temperature is assumed to be same with the ambient
temperature
Electrochemical reaction kinetics.— The electrochemical reac- 兩T兩t=0 = Tamb 关17兴
tions for lithium-ion intercalation/deintercalation at the solid/
solution interface can be expressed as where T is the cell temperature, v is the volume of cell, is the
density of cell, C p is the specific heat capacity of cell, and q is the
Li − s Li+ + e− + s 关10兴 rate of heat transfer between cell and surroundings. In this work,
where s represents a vacant host on the solid particle surface. The heat flux at cell surface are assumed to follow Newton’s law of
rate of such a reaction is related to the surface SoC xj,surf and the cooling, so the term q is expressed as
overpotential j and is expressed through the Butler–Volmer equa- q = hA共T − Tamb兲 关18兴
tion 共Appendix A兲
冋 冉 冊
where h is the heat transfer coefficient and A is the cell surface area.
0.5F On the right hand side of Eq. 16, the first term IT关 Up /T共xp,surf兲
e 共1
kjcs,j,maxc0.5 xj,surf兲0.5xj,surf j
0.5
Jj = − exp
RT − Un /T共xn,surf兲兴 is the reversible heat caused by the reaction en-
冉 冊册
tropy change. The value of the reversible heat can be either positive
0.5F or negative depending on the direction of electrode reactions. The
− exp − j
RT second term on the right hand side of Eq. 16, I共p − n + IRcell兲, is
the irreversible heat generated from electrode polarization. The irre-
共j = p,n兲 关11兴 versible heat generated is always positive for both forward and re-
where kj is the temperature-dependent reaction rate constant and the verse reactions. The third term q on the right hand side of Eq. 16
electrolyte concentration in solution phase ce is assumed to be a represents the heat exchanged between the cell and the surround-
constant in the SP model. The overpotential j is defined as ings. The variation of cell temperature depends on the combined
contribution of reversible and irreversible heats as well as the heat
j = 1,j − 2,j − Uj exchange between the cell and the surroundings.
Downloaded 22 Feb 2011 to 129.252.25.182. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp
A124 Journal of The Electrochemical Society, 158 共2兲 A122-A132 共2011兲
冋 冉 冊册
As a result, the model is reduced to a system of ODEs shown in Eq.
Ead,j 1 1 16 and 25, from which the dependant variables T and uk,j are solved,
Ds,j共t兲 = Ds,j,ref exp − 关19兴 while other variables involved in the model, such as the SoC, OCP,
R T Tref
and cell voltage can all be expressed as the functions of T and uk,j.
冋 冉 冊册
R T Tref
Ds,j 2k Ds,j
共j = p,n兲 关20兴 uk,j = 2 2 ␦j 1 − exp − t 关26兴
Rj Rj2
where Tref is the reference temperature, Ds,j,ref is the solid phase Substituting Eq. 26 into Eq. 22 yields the analytical solution for
diffusion coefficient of electrode j at the reference temperature, kj,ref solid phase diffusion
冋
is the reaction rate constant of electrode j at the reference tempera-
⬁
ture, Ead,j is the diffusion coefficient activation energy of electrode sin共kr̄j兲
兺
Ds,j 1 2
j, Ear,j is the reaction rate activation energy of electrode j, and R is xj = xini,j + ␦j 3 t+ 共5r̄2 − 3兲 −
Rj2 10 j r̄j k=1 2k sin共k兲
冊册
the ideal gas constant. As mentioned above, the cell temperature
冉
varies only with time, so Ds,j and kj,ref are all functions of time only.
The OCP Up and Un are the functions of both the SoC and tempera- 2k Ds,j
⫻exp − t 关27兴
ture, and the correlation can be approximated by Taylor’s expansion Rj2
冋 冊册
Uj共xj,surf,T兲 = Uj共xj,surf,Tref兲 + 共T − Tref兲
冉
T ⬁
xj,surf,Tref
2k Ds,j
兺
Ds,j 1 2
xj,surf = xini,j + ␦j 3 2 t + − 2 exp − t
共j = p,n兲 关21兴 Rj 5 k k=1 Rj2
共j = p,n兲 关28兴
Solution procedure.— Analytical techniques to solve for concen-
Eq. 27 and 28 are consistent with the analytical solutions derived by
tration in the single-particle model.— As shown above, this ther-
Subramanian and White using the separation of variables method.19
mal model includes two partial differential equations describing the
solid phase diffusion for each electrode and an ordinary differential The solution accuracy.— The mathematical model described by Eq.
equation 共ODE兲 for the overall cell energy balance. These equations 1-28 was solved using the backward differentiation formula method
are coupled together by the nonlinear dependence of transport and with MATLAB. The first issue regarding this model is the accuracy
kinetic parameters on cell temperature; so, no analytical solution can and efficiency in simulations. According to Eq. 24, the solution for
possibly be obtained for this equation system when the variation of solid phase diffusion theoretically includes an infinite series of
the cell temperature is taken into account. However, Eq. 1 can be
eigenfunctions uk,j, but this infinite series must be truncated in run-
solved using the eigenfunction expansion method 共see Appendix B兲
ning a practical model to save simulation time and computer
to obtain a solution with the following form
memory. Obviously, keeping a large number of terms in the series
冕冋 册
t will ensure high accuracy for the solution; however, the more the
r̄j2 3 Ds,j共兲
xj = xini,j + ␦j共t兲 − ␦j共t兲 + 3 ␦j共兲 d eigenfunctions are kept, the longer it will take to converge; so, there
2 10 0
Rj2 has to be a compromise between the accuracy and efficiency. In this
冋 册
⬁ work, the solution for an equation system including 2000 eigenfunc-
sin共kr̄j兲 2␦j共t兲
+ 兺 r̄
k=1 j sin k
uj,k共t兲 −
2k
关22兴 tions is regarded as the complete solution for the model, and the
truncated solutions with 1, 5, 10, and 20 eigenfunctions are com-
pared with the complete solution. The summary of equations and
where r̄j = r/Rj is the dimensionless radius coordinate, is the input variable values are presented in Table I. The profiles of first
dummy variable of integration, and j共t兲 is the dimensionless five eigenfunctions for both positive and negative electrodes in a 1C
lithium flux. The SoC at the surface of intercalation particle is discharge starting with 25°C are presented in Fig. 1. As shown in
冕冋 册
t the plots, there are no steady states for the eigenfunctions because
1 Ds,j共兲 the solid phase diffusion coefficients keep changing with the cell
xj,surf = 兩xj兩r̄j=1 = xini,j + ␦j共t兲 + 3 ␦j共兲 dt
5 0
Rj2 temperature. In such a case, it often requires more terms in the series
冋 册
for an unstable model than for a stable one to ensure the accuracy.
⬁
2␦j共t兲 The Vcell versus time curves from solutions with different numbers
+ 兺 k=1
uj,k共t兲 −
2k
关23兴 of eigenfunctions 共ambient temperature is 25°C兲 are presented in
Fig. 2a, and the plots of the relative error between the truncated
where the eigenvalues k are roots of the following algebraic equa- solutions and the complete solution are presented in Fig. 2b, where
tion the relative error is defined as
sin k − k cos k = 0 关24兴 兩Vcell,k − Vcell,complete兩
Relative error = 关29兴
Due to the periodicity, the value of k becomes infinitely large as the Vcell,compete
subscript number k increases. The eigenfunctions uj,k are determined where Vcell,k is the cell voltage calculated from k eigenfunctions and
by the following ordinary differential equations Vcell,complete is the cell voltage calculated from the complete solution.
duk,j共t兲 2k Ds,j共t兲 Ds,j共t兲 We can find from Fig. 1 that when the number of eigenfunctions is
=− 2 uk,j + 2 2 ␦j共t兲 more than 共or equal to兲 10, the relative error will be less than 10−5
dt Rj Rj except for the first point; therefore, the number of eigenfunctions
Downloaded 22 Feb 2011 to 129.252.25.182. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp
Journal of The Electrochemical Society, 158 共2兲 A122-A132 共2011兲 A125
xj,surf = xini,j +
1
␦ 共t兲 + 3兰t0
5 j
冋
Ds,j共兲
␦j共兲 dt 册
冋 册
R2j
N 2␦j共t兲
+ 兺 uj,k共t兲 −
k=1
k2
IR p IRn
␦p共t兲 = − ␦ 共t兲 =
SpFDs,p共t兲cs,p,max n SnFDs,n共t兲cs,n,max
2RT
Vcell = Up共xp,surf,T兲 − Un共xn,surf,T兲 +
F
⫻ln 共 2
+
I
F 兲
冑m2p + 4 + mp 2RT 冑m2n + 4 + mn
ln
2共 + IRcell 兲
Expressions mp =
FkpSpcs,p,maxce0.5共1 − xp,surf兲0.5xp,surf
0.5
冏 冏 冏 冏
model with the experimental data obtained on the lithium-ion pouch
cell, the values of certain parameters in this model need to be esti- Rcell 2Rcell
2 = + I 关33兴
mated. The first step of this procedure is to fit the C/33 discharge T I=0,T=Tamb IT I=0,T=Tamb
curves at four ambient temperatures 共15, 25, 35, and 45°C兲; at the
very low discharge rate, the cell temperature can be regarded as the In fitting each of the C/2 and C/1 discharge curves 共at 15, 25, 35,
same with the constant ambient temperature all the time, so the and 45°C兲, 1 and 2 were estimated as parameters using the Gauss
analytical solution presented in Eq. 27 can be used directly; further- Newton method. In fitting 1C discharge curves, the heat transfer
more, with constant temperature and current, the solution phase re- coefficient h was adjusted to obtain agreement of the predicted cell
sistance can also be regarded as a constant. Therefore, by fitting the temperature profiles with those from the PP model.15
C/33 discharge curves, three parameters will be estimated simulta-
neously: the initial SoCs for the positive and negative electrodes Experimental
共xini,p and xini,n兲 and the solution phase resistance 共Rcell兲. All the experimental data used in this work were collected from
After the parameter estimation on low rates, the single-particle Mine Safety Appliances 共nameplate capacity C = 1.656 Ah兲 pouch
thermal model was then used to fit the discharge curves for higher cells and the testings were done in-house. Prior to cycling, batches
rates. The initial SoC of the individual electrodes obtained from of cells were equilibrated in the environmental chambers at the pre-
C/33 rate discharge data can also be used as initial conditions in the determined temperatures 共15, 25, 35, and 45°C兲. To obtain the initial
Downloaded 22 Feb 2011 to 129.252.25.182. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp
A126 Journal of The Electrochemical Society, 158 共2兲 A122-A132 共2011兲
Lj 70 73.5 µm
Sj 1.1167a 0.7824a m2
cs,j,max 51,410a 31,833a mol/m3
Rj 8.5 ⫻ 10 −6a
12.5 ⫻ 10−6a m
kj,ref 6.6667 ⫻ 10−11b 1.764 ⫻ 10−11b m2.5 mol−0.5 s−1
Eare,j 58c 20d kJ/mol
Ds,j,ref 1.0 ⫻ 10 −14e
3.9 ⫻ 10−14e m2 /s
Eadi,j 29f 35g kJ/mol
␣a,j 0.5 0.5
␣c,j 0.5 0.5
ce 1000a mol/m3
F 96,487 C/mol
1626a kg/m3
v 0.199 ⫻ 0.08499 ⫻ 0.002a m3
Cp 750h J/共K·kg兲
Tref 25 °C
R 8.3143 J/共mol·K兲
hA 0.085 J/共s·K兲
Ls 25 µm
a
Calculated or obtained directly from cell design parameters.
b
Ref. 15.
c
Ref. 10.
d
Ref. 8.
e
Ref. 3.
f
Ref. 7.
g
Ref. 9.
h
Assumed values
Figure 2. 共Color online兲 共a兲 The discharge curves from solutions with dif-
ferent number of eigenfunctions 共1C discharge at 25 °C兲. 共b兲. The relative
error between truncated solutions and complete solution.
curve is the prediction of the analytical solution described by Eq. 28,
the dashed-and-dotted curve is the prediction of the PP model.15 the
dashed curve is the simulated open circuit voltage for each cell, and
cell capacities at various temperatures and charge/discharge rates, the dotted curve is the experimental data. The discharge capacity of
the following protocols were used. The discharge protocol consists cell is defined as
冕
of galvanostatic discharge current corresponding to the three dis-
tfinal
charge rates 共C/33, C/2 and C/1兲 until the cutoff voltage of 3.3 V 1
was reached. The capacity values at the cutoff voltage were taken as Cap = 兩I兩dt 关34兴
3600 0
the fresh cell discharge capacity at the different rates. The charge
protocols are slightly different for the different rates. For C/33 rate where tfinal is the time at the end of the discharge. Since the elec-
measurement, the cells were charged to 4.1 V, followed by a rest for trode polarization is almost negligible at the very low current den-
30 min before being discharged back to 3.3 V. For C/2 and C/1 sities, the voltage profiles of cells at C/33 rate are very similar to the
rates, the cells were charged at respective currents to 4.1 V and then open circuit voltage curves. It is also interesting to note, in Figs.
potentiostatically held at the cutoff voltage until the current tapered 3a-3d, that the single-particle thermal model fits the experimental
down to 50 mA. For C/2 and C/1 discharges, the cells were dis- data better than the PP model does. The primary reason for this is
charged at respective constant current rate to 3.3 V, then kept open that more of the experimental data points were used in this study and
circuit for 30 minutes, and then discharged at C/33 rate until the our analytical solution provided more accuracy than the PP model.
voltage reached 3.3 V to make sure that the subsequent charge starts The use of the analytical solution is the reason for this improvement
at the same state of charge. The values of design parameters for this in accuracy. In this way, the analytical Jacobian matrix of Vcell with
type of pouch cells are listed in Table II, and detailed experimental respect to parameters can be obtained directly from the algebraic
procedures involved in cycling protocols and periodic performance expressions rather than from solving the differential equations. For
tests are available in Ref. 15. According to the authors of Ref. 15, the PP model, however, if the least square method is applied, a large
because the cells were charged at the respective ambient tempera- number of coupled partial derivative equations will be added to the
tures, the extent of charging could be different, which subsequently model in finding the numerical Jacobian matrix and the computation
causes the initial states of discharge to be different for cells at dif- may get extremely slow and difficult; for this reason, the parameter
ferent temperatures. estimation in Ref. 15 was performed only with the methods that are
faster but with less accuracy, such as locating the characteristic
Results and Discussion points on the discharge curves. The C/33 discharge curves at differ-
Comparison between simulation and experimental data for dis- ent temperatures are plotted in Fig. 3e with the reaction entropy
charge profiles.— In fitting the C/33 rate discharge curves, the com- change 共 Up /T − Un /T兲 curve. As shown in Fig. 3e, the
putation was performed by the LSQNONLIN subroutine in MATLAB. 共 Up /T − Un /T兲 value is positive at the initial stage of discharge,
The simulated versus experimental C/33 discharge curves with esti- so the cell at high temperature has higher voltage. As discharge goes
mated parameter values 共xini,p, xini,n, and constant Rcell兲 are presented on, the 共 Up /T − Un /T兲 value reduces to zero and then negative
in Figs. 3a-3d. As stated in Ref. 15, the cells were charged at their and the difference between the curves gets smaller. At the end of
respective ambient temperatures 共i.e., at 15, 25, 35, and 45°C兲. discharge, the deviation between the discharge capacities is deter-
Therefore, the initial states of charge vary. In each figure, the solid mined by the difference in the initial states of charge.
Downloaded 22 Feb 2011 to 129.252.25.182. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp
Journal of The Electrochemical Society, 158 共2兲 A122-A132 共2011兲 A127
Downloaded 22 Feb 2011 to 129.252.25.182. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp
A128 Journal of The Electrochemical Society, 158 共2兲 A122-A132 共2011兲
Table III. The parameter values estimated by fitting the C/2 and
1C discharge curves at different temperatures.
C/2a 1Ca
Figure 6. 共Color online兲 Variation of heat generation and overpotential for Figure 8. 共Color online兲 Variation of cell temperature in 1C discharge with
cell discharged with 1C rate and starting temperature of 25 °C. different hA values and a starting temperatures of 25 °C.
Downloaded 22 Feb 2011 to 129.252.25.182. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp
Journal of The Electrochemical Society, 158 共2兲 A122-A132 共2011兲 A129
Figure 9. 共Color online兲 The OCP profiles for 共a兲 positive and 共b兲 negative
electrodes.
Figure 10. 共Color online兲 The entropy change coefficient profiles for 共a兲
positive and 共b兲 negative electrodes.
6兲 and hence causes the cell temperature to drop. Such a drop in the
cell temperature was also observed experimentally in Ref. 20. As the heat transfer coefficient times the cell surface area hA
discharge continues, the surface SoC for the positive electrode xp,surf = 0.085 W/K. For the single-particle model, the temperature rise
gets larger and passes the peak shown in Fig. 10a and the Up /T during the entire discharge 兩T兩t=tfinal − 兩T兩t=0 with an initial tempera-
turns to negative; accordingly, the reversible heat term becomes
ture of 15°C is 9.3°C whereas that with an initial temperature of
positive 共see Fig. 6兲 and the cell temperature increases significantly
with the joint effects of the both the irreversible heat and the posi- 45°C is only 7.4°C. The PP model results from Ref. 14 also show
tive reversible heat. While the cell temperature goes up, the increas- that the difference 兩T兩t=tfinal − 兩T兩t=0 decreases as the initial tempera-
ing temperature difference between the cell and surrounding accel- ture gets higher.
erates heat dissipation, which in turn prevents an accelerated rise in This single-particle thermal model was also used to simulate the
the temperature. At the final stage of discharge, the surface SoC for discharge with adiabatic conditions. Figure 8 shows the temperature
the negative electrode xn,surf drops to close to 0, and the irreversible profiles for cell discharged with different hA values at 1C rate and
heat gets large quickly due to a sharp increase in the magnitude of an initial temperature of 25°C. As shown in Fig. 8, the cell tempera-
the overpotential, so the cell temperature goes up rapidly again. A ture increases by 30°C at the adiabatic condition 共hA = 0兲, but the
useful quantity that can be defined is the ratio of reversible heat to cell temperature remains almost constant for large hA values 共e.g.,
irreversible heat more than 1 W/K兲. So the heat transfer conditions have a very sig-
冋 册
nificant effect on the cell’s thermal behavior.
Up Un
IT 共xp,surf兲 − 共xn,surf兲
Reversible Heat T T Conclusion
= 关35兴
Irreversible Heat I共p − n + IRcell兲 The single-particle model was extended to include an energy
When current increases, the magnitude of electrode overpotential balance. The model shows good agreement with the PP model as
共the denominator in Eq. 35兲 gets large, so the ratio in Eq. 35 de- well as the experimental data for various rates and temperatures. The
creases; this means that the contribution of the reversible heat be- model offers good advantage in terms of computational efficiency
comes less significant than that of the irreversible heat for higher over the PP model. Consequently, the model developed in this work
current. Using such a conclusion, we can explain why the tempera- can be readily used to simulate the cycling conditions for cells and
ture curve of the C/2 discharge has a larger drop than that of the 1C battery packs.
discharge at the initial stage. Figure 6 also shows that the total cell University of South Carolina assisted in meeting the publication costs of
over voltage is mainly determined by the negative electrode overpo- this article.
tential and potential loss on Rcell, and these two parts are also the
major source of irreversible heat. Appendix A
The simulated cell temperature profiles during 1C discharge for The OCP profiles for LixCoO2 and MCMB obtained from Ref. 14 are presented in
different ambient temperatures are shown in Fig. 7. By comparing Fig. 9. The expressions for the OCP as the functions of state of charge are obtained from
these results with those from the PP model, we obtain the value for fitting the experimental curves. For the positive electrode
Downloaded 22 Feb 2011 to 129.252.25.182. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp
A130 Journal of The Electrochemical Society, 158 共2兲 A122-A132 共2011兲
冋 册
− 78.68074兲
␣a,jF
where the surface state of charge of the positive electrode xp equals to the stoichiometric ជij = i0,j exp 共1,j − 2,j − Uj兲 关A-14兴
value x in LixCoO2; and for the negative electrode RT
value x in LixCoO2; and for the negative electrode And the total reaction current density for the surface electrochemical reaction is
Un 共0.00527 + 3.29927xn − 91.79326x2n + 1004.91101x3n − 5812.27813x4n + 19,329.75490x5n − 37,147.89470x6n + 38,379.18127x7n − 16,515.05308x8n兲
= 关A-4兴
T 共1 − 48.09287xn + 1017.23480x2n − 10,481.80419x3n + 59,431.30001x4n − 195,881.64880x5n + 374,577.31520x6n − 385,821.16070x7n + 165,705.85970x8n兲
where the surface state of charge of negative electrode xn equals to the stoichiometric
value y in LiyC6. In these two equations, Up /T and Un /T have the units of millivolts 冋 冉 冊 冉
ij = ជij − ឈij = i0,j exp
␣a,jF
RT
j − exp −
␣c,jF
RT
j 冊册 = kj共T兲Fcs,j,maxce␣c,jxj,surf
␣a,j
共1
冋 冉 冊 冉 冊册
per Kelvin.
The electrochemical reaction for the intercalation/deintercalation at a solid/solution ␣a,jF ␣c,jF
interface can be written as − xj,surf兲␣c,j exp j − exp − j 关A-19兴
RT RT
Deintercalation
Assuming that ␣a,j = ␣c,j = 0.5, Eq. A-13 becomes
Li-⌰s Li+ + e− + ⌰s 关A-5兴
Intercalation
i0,j = kj共T兲Fcs,j,maxce0.5xj,surf
0.5
共1 − xj,surf兲0.5 关A-20兴
where Li-s represents a filled intercalation site and s represents an intercalation site.
The concentration of filled Li-s sites at the particle surface is assumed to be equal and Eq. A-19 becomes
to the lithium-ion concentration at the surface
where ជij is the current density for the forward reaction of electrode j, kជ j共T兲 is the
temperature-dependent rate constant for the forward reaction of electrode j, and ␣a,j is
Appendix B
the anodic transfer coefficient for the forward reaction of electrode j. And the current The Eigenfunction Expansion Method
density for the reverse reaction rate 共i.e., in the cathodic reaction兲 is
In the thermal model, the diffusion coefficient Ds,j is assumed to be a function of
ឈi = kឈ 共T兲Fc
j j s,j,maxce共1 − xj,surf兲exp − 冋 ␣c,jF
RT
共1,j − 2,j兲 册 关A-9兴
time only, and the governing equation for solid phase diffusion 共Eq. 1兲 can be rewritten
as
where ឈij is the current density for the reverse reaction of electrode j, kឈ j共T兲 is the
temperature-dependent rate constant for the reverse reaction of electrode j, and ␣c,j is
xj Ds,j共t兲 1 2 xj
t
=
Rj2 r̄j2 r̄j
r̄j
r̄j
冉 冊 关B-1兴
the cathodic transfer coefficient for the reverse reaction of electrode j. The exchange and the initial and boundary conditions become
current density of the electrochemical reaction is defined as the reaction current density
at OCP 共1,j − 2,j = Uj兲 t = 0, xj = xj,ini
Downloaded 22 Feb 2011 to 129.252.25.182. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp
Journal of The Electrochemical Society, 158 共2兲 A122-A132 共2011兲 A131
冕冋 册 兺 r̄ sin共 兲 冋u
⬁ t ⬁
兺 ␣ 共t兲z 共r̄ 兲
r̄j2 3 Ds,j共兲 sin共kr̄j兲
w共r̄j,t兲 = k k j 关B-3兴 xj = xj,ini + ␦j共t兲 − ␦j共t兲 + 3 ␦j共兲 d + j,k共t兲
2 10 Rj2 j k
k=1 0 k=1
冉 冊
−
1 k2
2 zk共r̄j兲
zk共r̄j兲 = r̄j
r̄j2 r̄j r̄j
duj,k共t兲 k2Ds,j共t兲 Ds,j共t兲
=− uj,k共t兲 + 2 ␦j共t兲 关B-19兴
dt Rj2 Rj2
zk共r̄j兲 zk共r̄j兲
r̄j = 0, =0 r̄j = 1, =0 关B-4兴
r̄j r̄j 兩uj,k兩t=0 = 0
Solving for Eq. B-4, we obtain
k = 1, . . . ,⬁
sin共kr̄j兲
zk共r̄j兲 = 关B-5兴
r̄j j = p,n 关B-20兴
兺 ␣ 共t兲
␦j共t兲 2 sin共kr̄j兲
xj = r̄ + C共t兲 + k 关B-7兴 + 6.453 ⫻ 10−4兲共Tamb − 273.15兲2 + 共− 2.221 ⫻ 10−3Iapp − 1.635 ⫻ 10−2兲共Tamb
2 j r̄j
k=1
− 273.15兲 + 共2.437 ⫻ 10−2Iapp + 1.428 ⫻ 10−1兲 关C-1兴
C共t兲 can be determined from the volume-average method. The volume-average of vari-
able xj is 2 = 共− 6.824 ⫻ 10−6Iapp + 1.372 ⫻ 10−5兲共Tamb − 273.15兲3 + 共6.054 ⫻ 10−4Iapp
冕
1 − 1.216 ⫻ 10−3兲共Tamb − 273.15兲2 + 共− 1.497 ⫻ 10−2Iapp + 3.025 ⫻ 10−2兲共Tamb
xj,ave = 3r̄j2xjdr̄j 关B-8兴 − 273.15兲 + 共7.179 ⫻ 10−2Iapp − 1.456 ⫻ 10−1兲 关C-2兴
0
冕
t
cs,j,max maximum concentration of lithium ions in the intercalation par-
3␦j共t兲 3␦j共0兲 Ds,j共兲 ticles of electrode j, mol/m3
C共t兲 = C共0兲 − + +3 ␦j共兲d 关B-11兴 Ds,j solid phase diffusion coefficient of lithium ions in the intercala-
10 10 Rj2
0 tion particles of electrode j, m2 /s
Ds,j,ref solid phase diffusion coefficient of lithium ions in the intercala-
At time t = 0, tion particles of electrode j at reference temperature, m2 /s
⬁ Ead,j activation energy for solid phase diffusion coefficient of electrode
兺 ␣ 共0兲
␦j共0兲 sin共kr̄j兲 j, kJ/mol
xj,ini = r̄j2 + C共0兲 + k 关B-12兴 Ear,j activation energy for reaction rate constant of electrode j, kJ/mol
2 r̄j
k=1 F Faraday’s constant, 96,487 C/mol
Multiply both sides of Eq. B-12 by 3r̄2j and integrate from 0 to 1 and obtain h heat transfer coefficient between cell and surroundings, W/m2 /K
I applied current, A
3␦j共0兲 Jj flux of lithium ions on the surface of intercalation particles,
C共0兲 = xj,ini − 关B-13兴 mol/m2 /s
10
kj rate constant for the electrochemical reaction of electrode j,
The value of ␣k共0兲 can be determined from orthogonality of the series. Multiplying both m2.5 /mol0.5 /s
sides of Eq. B-12 by r̄j sin共kr̄j兲 and integrating from 0 to 1, we get kj,ref rate constant for the electrochemical reaction of electrode j at
reference temperature, m2.5 /mol0.5 /s
2␦j共0兲 Lj thickness of active material on electrode j
␣k共0兲 = − 关B-14兴
k2 sin共k兲 Ls thickness of separator
mi ratio of applied current density to exchange current density of
Substitute Eq. B-11 and B-13 into Eq. B-7 electrode i
冕冋 册
t ⬁ Q rate of heat transfer between cell and surroundings, W
兺
r̄j2 3 Ds,j共兲 sin共kr̄j兲 r radial coordinate in the spherical intercalation particle, m
xj = xj,ini + ␦j共t兲 − ␦j共t兲 + 3 ␦j共兲 d + ␣k共t兲 r̄j
2 10 Rj2 r̄j r̄j = r/Rj, dimensionless radial coordinate in the spherical interca-
k=1
0 lation particle j
关B-15兴 R ideal gas constant, 8.3143 J/mol/K
Rcell solution phase resistance, ⍀
Substitute Eq. B-15 into Eq. B-1, then multiply both sides by r sin共kr̄兲 and integrate Rj radius of spherical intercalation for electrode j, m
from 0 to 1 to get Sj total electroactive area of electrode j, m2
t time, s
d␣k共t兲 k2Ds,j共t兲 2 d␦j共t兲
+ ␣k共t兲 + =0 关B-16兴 tfinal time at the end of discharge, s
dt Rj2 k2 sin共k兲 dt T temperature of cell, K or °C
Tamb ambient temperature, K or °C
Now take the following transformation
Tref reference temperature, 25°C
2␦j共t兲 Uj OCP of electrode j, V
uj,k共t兲 = sin共k兲␣k共t兲 + 关B-17兴 uj,k the kth eigenfunction of electrode j, dimensionless
k2
v volume of cell, m3
and substitute Eq. B-17 into Eq. B-15, B-16, and B-14 to obtain Vcell voltage of cell, V
Downloaded 22 Feb 2011 to 129.252.25.182. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp
A132 Journal of The Electrochemical Society, 158 共2兲 A122-A132 共2011兲
Vj total volume of electrode j, m3 2. C. R. Pals and J. Newman, J. Electrochem. Soc., 142, 3282 共1995兲.
xj state of charge of electrode j 3. S. Santhanagopalan, Q. Guo, P. Ramadass, and R. E. White, J. Power Sources,
xini,j initial state of charge of electrode j 156, 620 共2006兲.
xj,surf state of charge on the surface of intercalation particles of elec- 4. G. Botte, B. Johnson, and R. E. White, J. Electrochem. Soc., 146, 914 共1999兲.
trode j 5. D. Bernardi, E. Pawlikowski, and J. Newman, J. Electrochem. Soc., 132, 5 共1985兲.
Greek 6. L. Rao and J. Newman, J. Electrochem. Soc., 144, 2697 共1997兲.
7. K. Nakamura, H. Ohnoa, K. Okamurac, Y. Michihiroa, I. Nakabayashib, and T.
␣a,j anodic transfer coefficient of electrode j Kanashiro, Solid State Ionics, 135, 143 共2000兲.
␣c,j cathodic transfer coefficient of electrode j 8. O. Egorkina and A. Skundin, J. Solid State Electrochem., 2, 216 共1998兲.
␦j共t兲 ␦j共t兲 = −JjRj /Ds,j共t兲cs,j,max, dimensionless flux of lithium ions at 9. T. Kulova, A. Skundin, E. Nizhnikovskii, and A. Fesenko, Russ. J. Electrochem.,
electrode j 42, 259 共2006兲.
j volume fraction of solid phase active material in electrode j 10. H. Zheng, J. Qin, Y. Zhao, T. Abe, and Z. Ogumi, Solid State Ionics, 176, 2219
j overpotential of electrode j, V 共2005兲.
1 parameter used in approximated solution resistance, ⍀ 11. V. Srinivasan and C. Y. Wang, J. Electrochem. Soc., 150, A98 共2003兲.
2 parameter used in approximated solution resistance, ⍀/K 12. K. E. Thomas and J. Newman, J. Power Sources, 119–121, 844 共2003兲.
k the kth eigenvalue 13. K. E. Thomas, C. Bogatu, and J. Newman, J. Electrochem. Soc., 146, A570
density of cell, kg/m3 共2001兲.
dummy variable of integration, s 14. H. Yang and J. Prakash, J. Electrochem. Soc., 151, A1222 共2001兲.
1,j solid phase potential of electrode j, V 15. K. Kumaresan, G. Sikha, and R. E. White, J. Electrochem. Soc., 155, A164 共2008兲.
2,j solution phase potential of electrode j, V 16. G. Ning and B. N. Popov, J. Electrochem. Soc., 151, A1584 共2004兲.
Subscripts 17. J. Newman and C. W. Tobias, J. Electrochem. Soc., 109, 1183 共1962兲.
18. T. F. Fuller, M. Doyle, and J. Newman, J. Electrochem. Soc., 141, 1 共1994兲.
ini initial state 19. V. R. Subramanian and R. E. White, J. Power Sources, 96, 385 共2001兲.
j positive 共p兲 and negative 共n兲 electrodes 20. S. Al Hallaj, R. Venkatachalapathy, J. Prakash, and J. R. Selman, J. Electrochem.
k kth term in the solution series Soc., 147, 2432 共2000兲.
s solid phase 21. K. E. Thomas, J. Newman, R. M. Darling, in Advances in Lithium-Ion Batteries, W.
surf at the surface of an intercalation particle
A. van Schalkwijk and B. Scrosati, Editors, pp. 345–392, Kluwer Academic/
References Plenum Publishers, New York, NY 共2002兲.
22. M. Doyle and J. Newman, J. Electrochem. Soc., 143, 1890 共1996兲.
1. C. R. Pals and J. Newman, J. Electrochem. Soc., 142, 3274 共1995兲.
Downloaded 22 Feb 2011 to 129.252.25.182. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp