Energy Conversion and Management 133 (2017) 204–215

Contents lists available at ScienceDirect

Energy Conversion and Management

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

Investigation of phase change material based battery thermal

management at cold temperature using lattice Boltzmann method
Yutao Huo, Zhonghao Rao ⇑
School of Electric Power Engineering, China University of Mining and Technology, Xuzhou 221116, China

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

Article history: The power and energy densities of battery will decrease at cold temperature and the heat preservation is
Received 9 September 2016 necessary to keep the working temperature. In this paper, the lattice Boltzmann model for phase change
Received in revised form 5 December 2016 material based battery thermal management cold temperature is constructed and the effects of thermal
Accepted 6 December 2016
conductivity, latent heat and environmental temperature have been considered. The results show that
Available online 18 December 2016
with the latent heat of phase change material (PCM), the temperature of battery decreases slowly and
the temperature distribution can be guaranteed. However, once the PCM adjacent to the interface of bat-
Keywords:
tery and PCM, the temperature of battery drops faster and the temperature standard deviation rises shar-
Battery thermal management
Phase change material
ply. Besides, the lower thermal conductivity, greater latent heat and higher environmental temperature
Solidification are able to slow down the solidification process of PCM and keeps the temperature of battery. However,
Lattice Boltzmann with larger latent heat of PCM, the battery temperature distribution gets more non-uniform, which is able
to decrease the cycle life of battery.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction
The battery thermal management (BTM) is of significant impor-
tance to power battery, especially for lithium-ion battery [1]. In
recent years, most of the research only focused on the performance
of BTM’s heat dissipation at the excessive ambient temperature,
since the thermal runaway will occur under extreme condition
[2]. However, when the ambient temperature is low, the energy
density and power densities of battery decrease sharply [3]. Conse-
quently, the demand of temperature maintaining in low-
temperature environment must be reached for BTM.
The thermal performance of BTM in low-temperature environ-
ment is divided into two parts, the heat preservation and heating.
Heating is an effective way to keep the temperature of battery. The
simplest way of battery warming is using high-temperature air or
other liquid medium to transfer heat to battery [4]. Zhang et al. [4]
used warm air (obtained from cabin) to heat the battery. However,
the results showed that the heating method would increase the
load of air-conditioning system. Another heating method is the
electrical heating, using the ohmic heat or the Peltier effect [5,6].
Stuart and Hande [6] designed an alternating current (AC) inter-
nally heating for hybrid electrical vehicle (HEV) at cold tempera-
ture. The results showed that the heating process was
⇑ Corresponding author.
E-mail address: raozhonghao@cumt.edu.cn (Z. Rao).
accelerated when the amplitudes of AC was larger. Using the Pel-
tier effect, Troxler et al. [5] have proposed a BTM using Peltier ele-
ment to capture temperature difference of battery. The
temperature controlling system was able to be used to warm bat-
tery. However, the efficiency of all the heating methods is not great
enough at the extreme environment condition. Consequently, the
heat preservation of BTM is essential as well.
The phase change material (PCM) has received more and more
attention in temperature maintaining. During the phase change
process, the temperature of PCM varies a little since a large amount
of energy is consumed by latent heat. PCM has been successfully
applied to cool down battery [7–9] in recent years. In the same
way, the PCM is able to be used in heat preservation for battery.
The mass, momentum and heat transfer behaviors at solid-
liquid interface are extremely complex, with which the analytical
results are difficult to be obtained. Consequently, several numeri-
cal methods have been proposed to solve the solid-liquid phase
change problem, including the computational fluid dynamics
(CFD) method [10], dissipative particle dynamics (DPD) method
[11,12] and molecular dynamics (MD) method [13]. Unlike the
conventional CFD method, lattice Boltzmann (LB) is one of the
methods derived from Boltzmann transport equation [14]. The LB
method is simpler than conventional CFD method in programming.
Besides, with the easy boundary treatment and inherently paral-
lelizable computation, the LB method has been used to solve the
solid-liquid phase change problem successfully [15]. Jiaung et al.

http://dx.doi.org/10.1016/j.enconman.2016.12.009
0196-8904/Ó 2016 Elsevier Ltd. All rights reserved.
 Y. Huo, Z. Rao / Energy Conversion and Management 133 (2017) 204–215 205

Nomenclature

B weighting function t time [s]

c lattice speed [m
s1] t dimensionless time
cs lattice sound speed [m
s1] T temperature [°C]
Cp specific heat at constant pressure [kJ
kg1
K1] T0 initial temperature [°C]
ei discrete velocity in direction i [m
s1] T low temperature at wall [°C]
F force [N] T dimensionless temperature
Fi discrete force in direction i [N] u velocity [m
s1]
fi distribution function for density in direction i [kg
m3] us velocity of solid [m
s1]
eq
fi equilibrium distribution function for density in direc- V volume [m3]
tion i [kg
m3] x; y coordinates [m]
Fi discrete force in direction i [kg
m3
s1] x ; y dimensionless coordinates
g acceleration due to gravity [m
s2]
gi distribution function for total enthalpy in direction i Greek symbols
[kJ
kg1] a thermal diffusivity [m2
s1]
g eq
i
equilibrium distribution function for total enthalpy in b thermal expansion coefficient [1
K]
direction i [kJ
kg1] l dynamic viscosity [kg
m1
s1]
H total enthalpy [kJ
kg1] m kinetic viscosity [m2
s1]
Hl total enthalpy corresponding to the liquids temperature xi weight coefficient in direction i
[kJ
kg1] Xsi additional collision term
Hs total enthalpy corresponding to the solidus temperature q density [kg
m3]
[kJ
kg1] sf dimensionless relaxation time for density distribution
hsl latent heat [J
kg1] function
k thermal conductivity [W
m1
K1] sg dimensionless relaxation time for total enthalpy distri-
L characteristic length [m] bution function
mi distribution function in direction i sn dimensionless relaxation time for temperature distribu-
meq i
equilibrium distribution function in direction i tion function
mneqi
non-equilibrium distribution function in direction i w volume fraction of liquid
Nu Nusselt number r standard deviation of temperature
ni distribution function for temperature in direction i [K]
neq
i
equilibrium distribution function for temperature in Subscripts
direction i [K] ave average
p pressure [Pa] b battery
Pr Prandtl number i opposite direction of direction i
Q heat source [J] PCM phase change material
Ra Rayleigh number
Ste Stefan number

[16] firstly proposed a LB model for solid-liquid phase change
based on the enthalpy method and solved the phase change prob-
lem successfully. Based on this, Huber et al. [17] combined the flow
field and temperature field, with which the heat and mass transfer
behaviors of phase change in a cavity were revealed. However,
Huber’s phase change LB model needed large amount of iterations
in each time step to obtain the temperature and liquid fraction. To
overcome this, Eshraghi and Felicelli [18] established an implicit LB
model for phase change problem. They used two kinds of evolution
equation. One was the original evolution equation to obtain the
temperature field without phase change. Another was used to
solve the region undergoing phase change. Feng et al. [19]
extended Eshraghi and Felicelli’s model to the problem of phase
change by convection. Nonetheless, the implicit LB model can’t
solve the case that the liquidus temperature is equal to the solidus
temperature. As a result, Huang et al. [15] proposed a phase change
model based on total enthalpy, and temperature was obtained
from total enthalpy directly. Based on this, Huang et al. [20,21]
extended the total enthalpy-based LB model to the multiple-
relaxation-time (MRT) scheme and adaptive mesh. In our previous
work, based on Huang’s LB model, the heat flux boundary condi-
tion with three kinds of distribution has been considered and the
temperature field has been revealed [22]. Moreover, the phase
change LB model is applied to solve the phase change problem in
porous media as well, including the representative elementary vol-
ume (REV)-scale [23,24] and pore-scale [25,26]. Except the
enthalpy-based method, the phase-field method [27] and
immersed boundary method [20] for LB have been established as
well, which are more complex.
In this paper, in order to keep the temperature of battery in
cold-temperature environment, a LB model for BTM has been
established. The rest of this paper is constructed as follows. In Sec-
tion 2, the macroscopic conservation equations and corresponding
LB model are developed. The effects of heat conductivity, latent
heat and environmental temperature on thermal performance of
BTM have been investigated in Section 3. The summary is pre-
sented in Section 4.
2. Numerical model
The schematic of PCM based BTM is shown in Fig. 1(a), where
ABEF represents half part of battery and AF is the symmetry condi-
tion. BCDE is filled with PCM, where BC and ED are adiabatic. The
length of AB is 13 mm, which is equal to that of BC. The height of
the battery, AF, is 65 mm. At t > 0, the temperature at CD suddenly
decreases to T low , which is lower than the initial temperature of
battery, T 0 . The thermal contact resistance between PCM and bat-
tery is ignored in this paper. The numerical region shown in Fig. 1
(a) is divided into two parts, the battery region and PCM region. For
the PCM region, the macroscopic quantities are obtained from the
total-enthalpy based LB model, and the temperature field is
revealed using another thermal LB model. The heat transfer
206 Y. Huo, Z. Rao / Energy Conversion and Management 133 (2017) 204–215

1

A B C g i ðx þ ei Dt; t þ Dt Þ ¼ g i ðx; tÞ  g i ðx; tÞ  g eq
i ðx; tÞ ð2Þ
sg
where f i denotes distribution function of density in direction i and
g i is the distribution function of total enthalpy. sf and sg are the
eq
dimensionless relaxation times. f i and g eq
i are the equilibrium func-
Symmetry tions, as follows:
 
ei
u ðei
uÞ u2
f i ¼ xi q 1 þ 2 þ
eq
 ð3Þ
cs 2c4s 2c2s
Battery PCM 8  
>
< H  C p T þ xi C p T 1  2c u2
i¼0
Tlow g eq ¼ h i
2
s
ð4Þ
i
>
: xi q 1 þ ei
u þ ðe2ci
uÞ
2
4  2c2 i–0
u
c2s s s

where q, H, C p , T, u denote the density, total enthalpy, heat capacity,

temperature and velocity. cs is lattice sound speed and equal to
pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi
1=3c ¼ 1=3Dx=Dt. ei and xi are the discrete velocity in direction
i and corresponding weight coefficient for D2Q9 model used in this
paper:
84
>
<9 i¼0
F E D xi ¼ 1
i ¼ 1; 2; 3; 4 ð5Þ
> 91
:
i ¼ 5; 6; 7; 8
(a) 36

8
< ð0;
>
0Þ

p 
p 

i¼0
ei ¼ c cos ði  1Þ ; sin ði  1Þ i ¼ 1; 2; 3; 4 ð6Þ
Interface >
: pffiffiffi
2

p 
2

p 

2c cos 4 ð2i  1Þ ; sin 4 ð2i  1Þ i ¼ 5; 6; 7; 8

j+1
From Eqs. (3) and (4), the macroscopic quantities can be derived

X
8
a e b q¼ fi ð7Þ
0

j X
8
Dt
qu ¼ ei f i þ F ð8Þ
0
2

d f c X
8
H¼ gi ð9Þ
0

j-1 where denotes the body force, which is the buoyancy in this paper:

F ¼ qgbðT  T ref Þ ð10Þ

i-1 i i+1 where g is the acceleration due to gravity and b represents the ther-
mal expansion coefficient. The subscript ‘‘ref” denotes the reference
(b) temperature. The discrete force term in velocity space F i is obtained
from:
Fig. 1. Schematics of (a) BTM based on PCM and (b) FVM at the interface of battery
  
and PCM. 1 ei  u ei
u
F i ¼ xi 1  þ e
F
2sf
i
c2 c4s
through the interface between battery and PCM is solved by a   s  ð11Þ
1 ei  u ei
u
finite-volume method (FVM), which is introduced in the following. ¼ xi 1  þ 4 ei
qgbðT  T ref Þ
2sf c2s cs
2.1. Lattice Boltzmann model for PCM Eq. (1) is the modified evolution equation proposed by Nobel and
Torczynski [28] to deal with the immersed boundary condition,
The lattice Boltzmann model for PCM is constructed in this sec-
where B is the weighting function and Xsi is the additional collision
tion. All the thermal properties presented in Section 2.1 are that of
term based on the concept of ‘‘bounce-back” [28,29]
PCM. For convenience, the subscript ‘‘PCM” has been omitted in all
the equations in this section. For the flow and heat transfer behav- ð1  wÞðsf  0:5Þ
B¼ ð12Þ
ior of PCM, the double-distribution LB model has been applied in w þ sf  0:5
this paper. Two distribution functions are solved by [15,28]:
Xsi ¼ f i ðx; tÞ  f i ðx; tÞ þ f i ðq; us Þ  f i ðq; uÞ
eq eq
1 B
eq  ð13Þ
f i ðx þ ei Dt; t þ Dt Þ ¼ f i ðx; tÞ  f i ðx; tÞ  f i ðx; tÞ þ BXsi þ DtF i
sf
where w denotes the liquid fraction and ‘‘‘‘ denotes the opposite
ð1Þ direction.
 Y. Huo, Z. Rao / Energy Conversion and Management 133 (2017) 204–215 207

Table 1
The macroscopic quantities of materials.

k0 (W
m1
K1) Cp (J
kg1
K1) q (kg
m3) l (kg
m1
s1) hsl0 (J
kg1) b (K1)

Battery 3 1000 2500 – – –
PCM 0.3 1630 930 9.20  104 195,000 104

From the equations presented above, the total enthalpy can be @q

þ r
ðquÞ ¼ 0 ð16Þ
obtained in each time step, and then the temperature and liquid @t
fraction are determined by:
@u @ðuuÞ p
F
8 H
H < Hs þ ¼ r þ r
ðsf  0:5Þc2s Dt ru þ ð17Þ
>
>
< C ps
@t @x q q
T ¼ T s þ Hl Hs ðT l  T s Þ Hs 6 H < Hl
HHs
ð14Þ
>
> @H

: T þ ðHHl Þ H P Hl þ r
C p Tu ¼ r
ðsg  0:5Þc2s Dt rðC p TÞ ð18Þ

l C pl @t
8 with the relationships between dimensionless relaxation time and
>
<0 H < Hs kinetic viscosity and thermal diffusivity:
w¼ HHs
Hl Hs
Hs 6 H < Hl ð15Þ
>
:
1 H P Hl
Table 2
where the subscripts ‘‘l” and ‘‘s” denote the liquidus phase and soli- The mesh independent test (n = 1, c = 1, h = 1).
dus phase. The corresponding macroscopic conservation equations
XY 52  130 78  195 104  260 130  325
can be derived by Chapman-Enskog expansion from Eqs. (1) and
(2) [15]: Nuave 19.490 19.579 19.598 19.595

4.0

0.9 With latent heat

3.5
Liquid phase without latent heat
0.6 Solid phase without latent heat
0.3
Nuave

3.0
T*ave

0.0

-0.3
2.5 Mencinger [32]
Prensent LB model -0.6

-0.9
2.0
0 3 6 9 12 15 18 21 24 27 30
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
t*
t*
(a) (a)
1.0

0.1
0.8

0.6
log ( )

0.01
ψ

0.4
Mencinger [32]
Prensent LB model With latent heat
0.2 Liquid phase without latent heat
1E-3
Solid phase without latent heat
0.0
0 3 6 9 12 15 18 21 24 27 30 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
t*
t*
(b) (b)
Fig. 2. The comparisons of (a) average Nusselt number and (b) total liquid fraction Fig. 3. Comparisons of (a) average temperature and (b) temperature standard
between present LB model and Mencinger’s results. deviation between material with latent heat and that without latent heat.
208 Y. Huo, Z. Rao / Energy Conversion and Management 133 (2017) 204–215

l
ðsf  0:5Þc2s Dt ¼ m ¼ ð19Þ The discrete velocity and weight coefficient are the same as Eqs.
q (5) and (6). With the Chapman-Enskog expansion, the following
equation can be obtained [30]:
k
ðsg  0:5Þc2s ¼ a ¼ ð20Þ @T

qC p ¼ r
ðsn  0:5Þc2s Dt rT ð23Þ

@t
where l, k are dynamic viscosity and thermal conductivity.
ðsn  0:5Þc2s Dt ¼ a ð24Þ
2.2. Lattice Boltzmann model for battery

The thermal lattice Boltzmann model for battery is presented in
this section. Once again, the subscript ‘‘b” which denotes battery
has been omitted for convenience. All thermal properties pre-
sented in this section is of battery.
In this paper, the evolution of temperature distribution function
is showed as follows:
2.3. Boundary condition
In this paper, for all the boundaries, the non-equilibrium
extrapolation scheme proposed by Guo et al. [31] have been used.
The distribution function is divided into two parts, the equilibrium
part and the non-equilibrium part, as follows:

1
 mi ðx; tÞ ¼ meq neq

i ðx; tÞ þ mi ðx; tÞ ð25Þ
ni ðx þ ei Dt; t þ Dt Þ ¼ ni ðx; tÞ  ni ðx; tÞ  neq
i ðx; tÞ ð21Þ
sn where mi denotes the distribution function, including f i , hi and ni .
where hi denotes the distribution function for temperature in direc- The superscript ‘‘neq” represents ‘‘non-equilibrium”. The non-
tion i. sn is the dimensionless relaxation time for temperature. The equilibrium part at the wall is approximately equal to that of the
eq
equilibrium distribution function hi is derived from: node adjacent to the wall from Eq. (25). The equilibrium part is
computed from Eqs. (3), (4) and (22) after the macroscopic quanti-
i ¼ xi T
neq ð22Þ ties are determined.

0.1

1.0 1.0
5/6
0.8
1.0 0.01 2/3
0.6 5/6
log( )

2/3
0.4
T*ave

0.08
1E-3
0.2

0.0
0.06
0.16 0.20 0.24

-0.2 1E-4
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
t *
t*
(a) (a)

0.04

0.02
0.01
t*=1.5
0.00 t*=1.6
t*=1.7
log( )

-0.02
t*=1.5
Tave
*

-0.04
t*=1.6 1E-3
t*=1.7
-0.06

-0.08
1E-4
0.70 0.75 0.80 0.85 0.90 0.95 1.00 0.70 0.75 0.80 0.85 0.90 0.95 1.00

(b) (b)
Fig. 4. The average temperature of battery via (a) dimensionless time and (b) Fig. 5. The temperature standard deviation of battery via (a) dimensionless time
thermal conductivity. and (b) thermal conductivity.
 Y. Huo, Z. Rao / Energy Conversion and Management 133 (2017) 204–215 209

As mentioned above, two kinds of distribution function are where Q denotes the heat source from aefd to ebcf. Using the impli-
applied to temperature updating, the total enthalpy distribution cit scheme and central difference, Eqs. (26) and (27) become:
for PCM and temperature distribution function for battery. The
heat transfer between battery and material is not able to be treated Dx2 tþDt
by LB directly. Consequently, the simulation region is divided into ðqC p Þb ðT i;j  T ti;j Þ
2 !!
two parts, the PCM region and battery region. For each region, the
T ti;j  T tiDx;j 1 T ti;j  T ti;jþDx T ti;j  T ti;jDx
interface is regarded as boundary, and the temperature at each ¼ kb DxDt þ þ Q
time step should be defined. The FVM has been used to obtain Dx 2 Dx Dx
the temperature at (i, j) in this paper. As shown in Fig. 1(b), the ð28Þ
node located at interface (i, j) is divided into two parts, where
the left region denotes battery and right region represents PCM.
In order words, abcd represents the control volume of node (i, j).

Dx2 tþDt
qC p ðT i;j  T ti;j Þ
PCM
aefd is in the battery region and ebcf is in the PCM region. Integrat- 2 !!
ing the control volume aefd and ebcf respectively, ones obtain [22]: T ti;j  T tiþDx;j 1 T ti;j  T ti;jþDx T ti;j  T ti;jDx
Z Z Z ¼ kPCM DxDt þ þ þQ
i
@T jþ12Dx tþDt Dx 2 Dx Dx
ðqC p Þb dtdydx
i12Dx t @t j12Dx ð29Þ
Z tþDt       
@T 1 @T @T
¼ k b Dx  þ Q ð26Þ Combining Eqs. (28) and (29), the temperature at interface can be
t @x ad 2 @y ae @y df
determined at each time step. Furthermore, the distribution func-
Z iþ12Dx Z jþ12Dx Z tþDt tion is obtained from Eq. (25).
@T
ðqC p ÞPCM dtdydx Moreover, the solid-liquid interface is treated by modified evo-
i @t j12Dx t
lution equation, Eq. (1), with the concept of ‘‘bounce-back”.
Z tþDt        
@T 1 @T @T
¼ DxkPCM   þ þQ ð27Þ
t @x bc 2 @y eb @y cf

8
1.0
7 1.0
0.8
6 5/6
1.0
2/3
5 5/6 0.6
2/3
4
Nu

0.4
3
0.2
2

1 0.0

0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
t * t*
(a) (a)

0.24
1
t =1.5
* t*=1.5
t*=1.6 0.20 t*=1.6
t*=1.7 t*=1.7
0.16
log(Nu)

0.1 0.12

0.08

0.04

0.01 0.00
0.68 0.72 0.76 0.80 0.84 0.88 0.92 0.96 1.00 0.68 0.72 0.76 0.80 0.84 0.88 0.92 0.96 1.00

(b) (b)
Fig. 6. The average Nusselt number via (a) dimensionless time and (b) thermal Fig. 7. The total liquid fraction via (a) dimensionless time and (b) thermal
conductivity. conductivity.
210 Y. Huo, Z. Rao / Energy Conversion and Management 133 (2017) 204–215

2.4. Initial condition T  Tm T low  T m ab t x

T ast ¼ ; h¼ ; t ¼ ; x ¼ ;
T0  Tm T0  Tm L2 L
As mentioned before, the temperature at CD is kept at T low y mPCM gbðT 0  T m ÞL3
y ¼ ; Pr ¼ ; Ra ¼ ;
when time is larger than zero. Based on the physical truth, the tem- L aPCM mPCM aPCM
perature field of both battery and PCM is at steady state at the ini- C p ðT 0  T m Þ k hsl
tial time in this paper, since the battery works in a long time before Ste ¼ PCM ; n¼ ; c¼ ð31Þ
hsl k0 hsl0
simulations. In order words, the initial condition needs to be con-
sidered. At the initial time, the temperature of battery is uniform where Pr, Ra, Ste denote the Prandtl number, Rayleigh number and
and equal to T 0 . On the contrary, the distributions of density and Stefan number. T  and t are the dimensionless temperature and
temperature in the PCM region are computed from the steady- time. h represents the dimensionless environmental temperature
state calculation, in which the temperature of interface is kept at and L is the characteristic length. The subscript ‘‘0” represents the
T 0 and temperature of CD is kept at melting temperature T m (in reference value. The reference thermal conductivity and latent heat
this paper, T m ¼ T l ¼ T s ). The following criteria has to be ensured are 0.3 W
m1
K1 and 195,000 J
kg1, as shown in Table 1. The
during the calculations: effects of thermal conductivity, latent heat and environmental tem-
P
jCtþDt  Ct j perature on thermal performance of BTM have been considered in
P t < 109 ð30Þ the following section. The thermal conductivity varies from 0.2 to
C
0.3 W
m1
K1, and the latent heat is in the range of 185,000–
where C denotes temperature and velocity. 205,000 J
kg1. Obviously, from Eq. (31), the corresponding dimen-
The paraffin is chosen as the PCM in this paper. The thermal sionless parameters are changed with different macroscopic quan-
properties of battery and PCM are presented in Table 1. In order tities. The initial temperature T 0 is 10 K higher than phase change
to obtain more accurate results, some dimensionless parameters temperature T m . Namely, the dimensionless temperature of battery
have been defined as follows: is set as 1.0 initially. When h is 1.0, the environmental tempera-
ture T low is 10 K lower than T m and 20 K lower than T 0 . Moreover,
the dimensionless environment temperature considered in this

0.10
1.2
0.08

0.9 =37/39
=37/39 0.06
=1.0 =1.0
0.6 =41/39 =41/39
0.04
Tave
*

0.3
0.02

0.0
0.00

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
t
t
(a) (a)

0.01

0.00 0.01

-0.01
t*=1.4
-0.02 t*=1.5
log( )
Tave

t*=1.6
*

-0.03 t*=1.4 1E-3

t*=1.5
-0.04
t*=1.6
-0.05

-0.06 1E-4
0.95 0.96 0.97 0.98 0.99 1.00 1.01 1.02 1.03 1.04 1.05 0.95 0.96 0.97 0.98 0.99 1.00 1.01 1.02 1.03 1.04 1.05

(b) (b)
Fig. 8. The average temperature of battery via (a) dimensionless time and (b) latent Fig. 9. The temperature standard deviation of battery via (a) dimensionless time
heat. and (b) latent heat.
 Y. Huo, Z. Rao / Energy Conversion and Management 133 (2017) 204–215 211

paper varies from 1.5 to 0.5. The liquidus temperature is 2.5. Model verification
assumed to be equal to solidus temperature.
Besides, the local Nusselt number is determined by: The problem of phase change by convection have been solved to
 verify the present numerical model. The cavity is filled with PCM,
@T
Nu ¼ ð32Þ whose temperature is equal to the melting temperature at the ini-
@x interface tial time. The left wall is kept at hot temperature and the right wall
is kept at the cold temperature. The detailed schematic can be
and the average Nusselt number at the interface is obtained from:
found at Ref. [32]. The Stefan number, Prandtl number and Ray-
Z
1 @T   leigh number are set as 0.01, 0.02 and 2.5  104, respectively.
Nuave ¼ dy The simulation results have been presented in Fig. 2. The results
Linterface interface @x
ð33Þ of present numerical model agrees well with that of Mencinger’s
1X M
T  T interfaceþDx
¼ interface
[32], which demonstrates the validity of the model.
M 1
Dx The accuracy of this work is guaranteed by the mesh indepen-
dent test, as shown in Table 2. As can be seen, the variation of aver-
where M is the number of lattices on the interface. Furthermore, the
age Nusselt number reduces with the increase of mesh number.
average temperature and temperature standard deviation of battery
The difference of grids system between 104  260 and 130  325
are defined as:
is less than 1%. Consequently, the mesh of 104  260 has been
R
ðT b ÞdV adopted in the following investigations.
T ave
Vb
¼ R ð34Þ First of all, the comparisons of thermal performance between
Vb
dV
BTM with PCM and without PCM have been conducted. Three kinds
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi of materials has been considered, as follows:
uR 
2
u T b  T ave dV
r¼ t Vb
R ð35Þ Case 1: PCM (c ¼ 1:0; n ¼ 1:0)
Vb
dV Case 2: liquidus non-PCM without latent heat (n ¼ 1:0)
Case 3: solidus non-PCM without latent heat (n ¼ 1:0)
where V b denotes the volume of battery, namely ABEF in Fig. 1(a).

1.0
6

0.8
=37/39
4
0.6 =1.0
=37/39
=41/39
=1.0 0.4
2
Nuave

=41/39
0.2

0
0.0

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

-2
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 t*

t* (a)
(a)
0.12

1 0.10

0.08

t*=1.4
0.06
t*=1.5
Nuave

0.1 t*=1.6
0.04
t*=1.4
0.02 t*=1.5
t*=1.6
0.00
0.01 0.95 0.96 0.97 0.98 0.99 1.00 1.01 1.02 1.03 1.04 1.05
0.95 0.96 0.97 0.98 0.99 1.00 1.01 1.02 1.03 1.04 1.05

(b) (b)
Fig. 10. The average Nusselt number via (a) dimensionless time and (b) latent heat. Fig. 11. The total liquid fraction via (a) dimensionless time and (b) latent heat.
212 Y. Huo, Z. Rao / Energy Conversion and Management 133 (2017) 204–215

Besides, h is set as 1.0 and 1.0. The variations of average tem-
perature and temperature standard deviation via time are pre-
sented in Fig. 3. Obviously, the temperature of Case 2 drops most
quickly and reaches to 0 at t  ¼ 0:1389. Thanks to the convection,
the heat loss of Case 1 is more than that of Case 3. However, when
the solidification occurs, the latent heat is dissipated to the envi-
ronment and the temperature of battery decreases much more
slowly. At t  ¼ 1:3, the temperature of Case 1 is 0.9623 higher than
that of Case 2 and 0.6104 higher than that of Case 2. Furthermore,
as shown in Fig. 3(b), the convection gets the temperature dis-
tribute more non-uniform. Once again, after t ¼ 0:7, r of Case 1
is kept less than 0.006, which is smaller than that of Case 2 and
3. Consequently, the PCM based BTM is able to keep the tempera-
ture of battery and make temperature uniform.
3. Results and discussions
3.1. Effects of latent heat
As mentioned before, the latent heat is one of the key impact
factor on thermal performance of BTM. In this section, the latent
heat varies from 185,000 to 205,000 J
kg1, namely, the corre-
sponding Stefan number and c are different, as demonstrated in
Eq. (31). n and h are set as 1.0 and 1.0.
The variations of battery average temperature are shown in
Fig. 4. The difference of average temperature between each case
is unapparent at first stage. However, since the latent heat of
PCM with c ¼ 37=39 is the smallest (less energy released from
PCM), the PCM transforms from liquid phase to solid phase most
quickly, which makes the difference of battery temperature larger,
as presented Fig. 4(a). Moreover, with the greater latent heat,
namely the smaller Stefan number shown in Eq. (31), the average
temperature of battery drops more slowly, as shown in Fig. 4(b).
At t  ¼ 1:6, the average temperature of c ¼ 1:0 is 0.0331 higher
than that of c ¼ 37=39. However, the larger latent heat leads to
greater temperature difference when the dimensionless time is
1.4, as shown in Fig. 5(b). When most of the PCMs solidify, the
temperature standard deviation increases suddenly. It can be con-
cluded that increasing latent heat gets the temperature distribu-
tion non-uniform. The same variation trend can be obtained from
Fig. 6 for average Nusselt number. It demonstrates that the tem-
perature difference between battery and PCM at the interface
increases with the latent heat, when the PCMs are still in liquid
phase. Once the PCM transfers from liquid phase to solid phase,
the Nusselt number increases quickly, as shown in Fig. 6.
The total liquid fraction has been presented in Fig. 7. Obviously,
with the larger latent heat, the solidification process is slowed
down, namely the total liquid fraction is greater at the same time.

0.10

0.1
0.08

0.06
0.0
T*ave

0.04
-0.1
0.02

-0.2 0.00

0.7 0.8 0.9 1.0 1.1 1.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2

t*
t*
(a) (a)

0.1

0.04

0.02 t*=1.0
t*=1.1
0.00
t*=1.2
0.01
-0.02 t*=1.0
T*ave

log(

t*=1.1
-0.04
t*=1.2
-0.06

-0.08 1E-3

-0.10
-1.5 -1.4 -1.3 -1.2 -1.1 -1.0 -0.9 -0.8 -0.7 -0.6 -0.5 -1.5 -1.4 -1.3 -1.2 -1.1 -1.0 -0.9 -0.8 -0.7 -0.6 -0.5

(b) (b)
Fig. 12. The average temperature of battery via (a) dimensionless time and (b) Fig. 13. The temperature standard deviation of battery via (a) dimensionless time
environmental temperature. and (b) environmental temperature.
 Y. Huo, Z. Rao / Energy Conversion and Management 133 (2017) 204–215 213

At t  ¼ 1:6, when c are equal to 37/39, 38/39, 1.0, 40/39 and 41/39,
the total liquid fractions are 0.009, 0.015, 0.023, 0.034 and 0.046,
respectively. Consequently, the temperature of upper region is
greater than that of bottom region for PCM. The same phenomenon
can be obtained for battery. Although increasing the latent heat
will have the temperature distributed non-uniform, PCM with lar-
ger latent heat is more suitable for BTM, considering the cooling
performance at high ambient temperature.
3.2. Effects of thermal conductivity
In this section, the effects of thermal conductivity on thermal
performance of BTM have been considered. The thermal conductiv-
ity of PCM varies from 0.2 to 0.3 W
m1
K1 and the corresponding
ratio, n, is in the range of 2/3 to 1.0. c and h are kept as 1.0 and 1.0,
respectively. The average temperature of battery is shown in Fig. 8.
In the initial phase, the average temperature drops quickly. It
demonstrates that the large amount of battery energy is trans-
ferred to PCM since the great temperature difference between bat-
tery and PCM at the initial time. From then on, the decrease trend
of average temperature becomes gentler and gentler, as shown in
Fig. 8(a). However, for n ¼ 1:0, the average temperature drops fas-
ter when time is larger than 1.6. At this moment, some PCMs adja-
cent to the battery have solidified completely, and the energy of
battery is consumed by sensible heat. Furthermore, decreasing
the thermal conductivity is able to get the average temperature
of battery dropping more slowly, as shown in Fig. 8(b). At
t ¼ 1:5, when n are 0.7, 0.8, 0.9 and 1.0, the average temperatures
are 0.0278, 0.0157, 0.0094 and 0.0071, respectively. The sudden
decrease in Fig. 8(b) results from the completely solidification of
PCM.
Fig. 9 presents the variations of temperature standard deviation
via time and thermal conductivity. Since the temperature differ-
ence between battery and PCM at the initial time is large, the tem-
perature standard deviation rises steeply at the first stage. The
temperature standard deviation reaches the maximum value firstly
when n is equal to 1.0, as shown in Fig. 9(a). From then on, the tem-
perature standard deviation starts to drop. When some PCMs adja-
cent to the interface solidify completely, as mentioned in Fig. 8, the
temperature standard deviation increases suddenly. Moreover,
with the larger thermal conductivity, the PCM is able to decrease
the temperature standard deviation of battery, as presented in
Fig. 9(b). When n are 0.7, 0.8, 0.9 and 1.0, the temperature standard
deviations of battery are 1.14  103, 8.01  104, 4.91  104 and
2.75  104 at t ¼ 1:5, respectively. In other words, the greater
thermal conductivity is able to get the temperature distribution
more uniform.

1.0

0.8

1 0.6
Nuave

0.4

0.2

0.0

0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2
0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2
t* t*
(a) (a)
0.6

t*=1.0 0.5
t*=1.0
1 t*=1.1 t*=1.1
0.4
t*=1.2 t*=1.2
Nuave

0.3

0.2

0.1 0.1

0.0
-1.5 -1.4 -1.3 -1.2 -1.1 -1.0 -0.9 -0.8 -0.7 -0.6 -0.5 -1.5 -1.4 -1.3 -1.2 -1.1 -1.0 -0.9 -0.8 -0.7 -0.6 -0.5

(b) (b)
Fig. 14. The average Nusselt number via (a) dimensionless time and (b) environ- Fig. 15. The total liquid fraction via (a) dimensionless time and (b) environmental
mental temperature. temperature.
214 Y. Huo, Z. Rao / Energy Conversion and Management 133 (2017) 204–215
The average Nusselt number is presented in Fig. 10. The temper-
ature of PCM decreases till to the phase change temperature and
the heat transfer by natural convection is weakened, which makes
the average Nusselt number smaller in the initial stage, as shown
in Fig. 10(a). When the PCMs adjacent to the interface transform
to solid phase completely, the temperature of PCM decreases more
quickly, because the energy is no longer consumed by latent heat.
From Eq. (33), it results in the steep increase of Nusselt number,
especially for the PCM with larger thermal conductivity, as shown
in Fig. 10(b). As described in Fig. 11, the total liquid fraction drops
faster when the thermal conductivity is greater, and the tempera-
ture difference between battery and PCM at the interface is small,
which gets the average Nusselt number smaller (Fig. 9(a)). This
phenomenon is more obvious for the larger thermal conductivity,
as presented in Fig. 10 (b). Since the PCM with thermal conductiv-
ity can’t reach the demand of cooling at high temperature, the BTM
with n ¼ 1:0 is more appropriate.
3.3. Effects of environmental temperature
In this section, the effects of environmental temperature have
been considered. Both n and c are defined as 1.0. h is changed from
1.5 to 0.5. The average temperature of battery with different
environmental temperature is shown in Fig. 12. Obviously, the
average temperature drops fastest with the colder environmental
temperature. At t ¼ 1:0, when h are 0.5, 0.8, 1.1 and 1.4,
the dimensionless average temperatures are 0.043, 0.037, 0.027
and 0.017, respectively. As mentioned before, a large amount of
heat is transferred to the environment in the form of latent heat,
which makes the temperature of battery drop slowly. Once the
PCM adjacent to the interface is in solid phase, the temperature
starts to drop faster, as presented in Fig. 12(a). In Fig. 12(b), the
average temperature shows a sharp decrease trend for the case
with h ¼ 1:5 at t  ¼ 1:1. With time increases, the similar variation
trend can be obtained for the cases with higher environmental
temperature. However, the lower environmental temperature
leads to smaller temperature standard deviation of battery when
the PCMs are all in solid phase, as shown in Fig. 13(b). The sharp
variation of temperature standard deviation in Fig. 13(a) results
from the solidification of PCM near the interface.
The average Nusselt number at the interface has been presented
in Fig. 14. The average Nusselt number of h ¼ 1:5 is the largest at
the first stage, as shown in Fig. 14(a). As time goes by, the Nusselt
number decreases, especially when the PCM near to the interface
transforms to solid phase. It makes the Nusselt number of
h ¼ 1:5 the smallest at t ¼ 1:2, as presented in Fig. 14(b). The
total liquid fraction is shown in Fig. 15. Obviously, the lower envi-
ronmental temperature accelerates the solidification of PCM.
When h are 0.5, 0.8, 1.1 and 1.4, the total liquid fractions
are 0.429, 0.265, 0.130 and 0.016 at t ¼ 1:2, respectively.
4. Conclusions
The heat preservation is necessary to battery on account of
power and energy densities decreases in cold environment. In this
paper, the lattice Boltzmann model of battery thermal manage-
ment using phase change material has been constructed. The
effects of thermal conductivity, latent heat and environmental
temperature on thermal performance of battery are considered.
The natural convection of PCM accelerates the temperature
decrease and have the temperature more non-uniform. With the
latent heat, these effects can be weaken. However, once the PCM
adjacent to the interface of battery and PCM, the temperature of
battery drops faster and the temperature standard deviation rises
sharply. Besides, the larger thermal conductivity and lower envi-
ronmental temperature are able to make the heat dissipation
quicker. Considering the cooling effect at high ambient tempera-
ture, increasing the latent heat is the best way to preserve heat
at low temperature generally.
Acknowledgements
This work was supported by the National Natural Science Foun-
dation of China (No. 51406223).
References
[1] Rao Z, Wang S. A review of power battery thermal energy management. Renew
Sustain Energy Rev 2011;15:4554–71.
[2] Huo Y, Rao Z, Liu X, Zhao J. Investigation of power battery thermal
management by using mini-channel cold plate. Energy Convers Manage
2015;89:387–95.
[3] Zhang SS, Xu K, Jow TR. The low temperature performance of Li-ion batteries. J
Power Sources 2003;115:137–40.
[4] Zhang X, Kong X, Li G, Li J. Thermodynamic assessment of active cooling/
heating methods for lithium-ion batteries of electric vehicles in extreme
conditions. Energy 2014;64:1092–101.
[5] Troxler Y, Wu B, Marinescu M, Yufit V, Patel Y, Marquis AJ, et al. The effect of
thermal gradients on the performance of lithium-ion batteries. J Power Sources
2014;247:1018–25.
[6] Stuart TA, Hande A. HEV battery heating using AC currents. J Power Sources
2004;129:368–78.
[7] Rao Z, Wang S, Zhang G. Simulation and experiment of thermal energy
management with phase change material for ageing LiFePO4 power battery.
Energy Convers Manage 2011;52:3408–14.
[8] Al Hallaj S, Selman JR. A novel thermal management system for electric vehicle
batteries using phase-change material. J Electrochem Soc 2000;147:3231–6.
[9] Al-Hallaj S, Kizilel R, Lateef A, Sabbah R, Farid M, Selman JR. Passive thermal
management using phase change material (PCM) for EV and HEV Li-ion
batteries. In: Vehicle power and propulsion, 2005 IEEE conference. IEEE; 2005.
p. 376–80.
[10] Yang J, Yang L, Xu C, Du X. Numerical analysis on thermal behavior of solid–
liquid phase change within copper foam with varying porosity. Int J Heat Mass
Transf 2015;84:1008–18.
[11] Rao Z, You X, Huo Y, Liu X. Dissipative particle dynamics study of nano-
encapsulated thermal energy storage phase change material. RSC Adv
2014;4:39552–7.
[12] Rao Z, Huo Y, Liu X. Dissipative particle dynamics and experimental study of
alkane-based nanoencapsulated phase change material for thermal energy
storage. RSC Adv 2014;4:20797–803.
[13] Rao Z, Wang S, Peng F. Molecular dynamics simulations of nano-encapsulated
and nanoparticle-enhanced thermal energy storage phase change materials.
Int J Heat Mass Transf 2013;66:575–84.
[14] Peng Y, Shu C, Chew YT. Simplified thermal lattice Boltzmann model for
incompressible thermal flows. Phys Rev E 2003;68.
[15] Huang R, Wu H, Cheng P. A new lattice Boltzmann model for solid–liquid
phase change. Int J Heat Mass Transf 2013;59:295–301.
[16] Jiaung W-S, Ho J-R, Kuo C-P. Lattice Boltzmann method for the heat conduction
problem with phase change. Numer Heat Transf: Part B: Fundam
2001;39:167–87.
[17] Huber C, Parmigiani A, Chopard B, Manga M, Bachmann O. Lattice Boltzmann
model for melting with natural convection. Int J Heat Fluid Flow
2008;29:1469–80.
[18] Eshraghi M, Felicelli SD. An implicit lattice Boltzmann model for heat
conduction with phase change. Int J Heat Mass Transf 2012;55:2420–8.
[19] Feng Y, Li H, Li L, Bu L, Wang T. Numerical investigation on the melting of
nanoparticle-enhanced phase change materials (NEPCM) in a bottom-heated
rectangular cavity using lattice Boltzmann method. Int J Heat Mass Transf
2015;81:415–25.
[20] Huang R, Wu H. An immersed boundary-thermal lattice Boltzmann method for
solid–liquid phase change. J Comput Phys 2014;277:305–19.
[21] Huang R, Wu H. Total enthalpy-based lattice Boltzmann method with adaptive
mesh refinement for solid-liquid phase change. J Comput Phys
2016;315:65–83.
[22] Huo Y, Rao Z. Lattice Boltzmann simulation for solid–liquid phase change
phenomenon of phase change material under constant heat flux. Int J Heat
Mass Transf 2015;86:197–206.
[23] Gao D, Chen Z. Lattice Boltzmann simulation of natural convection dominated
melting in a rectangular cavity filled with porous media. Int J Therm Sci
2011;50:493–501.
[24] Tao YB, You Y, He YL. Lattice Boltzmann simulation on phase change heat
transfer in metal foams/paraffin composite phase change material. Appl Therm
Eng 2015.
[25] Song W, Zhang Y, Li B, Fan X. A lattice Boltzmann model for heat and mass
transfer phenomena with phase transformations in unsaturated soil during
freezing process. Int J Heat Mass Transf 2016;94:29–38.
 Y. Huo, Z. Rao / Energy Conversion and Management 133 (2017) 204–215
[26] Chen Z, Gao D, Shi J. Experimental and numerical study on melting of phase
change materials in metal foams at pore scale. Int J Heat Mass Transf
2014;72:646–55.
[27] Miller W. The lattice Boltzmann method: a new tool for numerical simulation
of the interaction of growth kinetics and melt flow. J Cryst Growth
2001;230:263–9.
[28] Noble D, Torczynski J. A lattice-Boltzmann method for partially saturated
computational cells. Int J Mod Phys C 1998;9:1189–201.
215
[29] Holdych DJ. Lattice Boltzmann methods for diffuse and mobile interfaces. Ann
Arbor: University of Illinois at Urbana-Champaign; 2003. p. 165.
[30] Guo Z, Shi B, Zheng C. A coupled lattice BGK model for the Boussinesq
equations. Int J Numer Meth Fluids 2002;39:325–42.
[31] Zhao-Li G, Chu-Guang Z, Bao-Chang S. Non-equilibrium extrapolation method
for velocity and pressure boundary conditions in the lattice Boltzmann
method. Chin Phys 2002;11:366.
[32] Mencinger J. Numerical simulation of melting in two-dimensional cavity using
adaptive grid. J Comput Phys 2004;198:243–64.