Applied Thermal Engineering 120 (2017) 484–495

Contents lists available at ScienceDirect

Applied Thermal Engineering

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

Research Paper

3-D CFD modeling and experimental testing of thermal behavior

of a Li-Ion battery
Emre Gümüsßsu, Özgür Ekici ⇑, Murat Köksal
Department of Mechanical Engineering, Hacettepe University, Ankara, Turkey

h i g h l i g h t s

A thermally fully predictive 3-D CFD model is developed for Li-Ion batteries.

Complete flow field around the battery and conduction inside the battery are solved.

Macro-scale thermophysical properties and the entropic term are investigated.

Discharge rate and usage history of the battery are systematically investigated.

Reliability of the model was tested through experimental measurements.

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

Article history: In this study, a 3-D computational fluid dynamics model was developed for investigating the thermal
Received 10 January 2017 behavior of lithium ion batteries under natural convection. The model solves the complete flow field
Revised 19 March 2017 around the battery as well as conduction inside the battery using the well-known heat generation model
Accepted 4 April 2017
of Bernardi et al. (1985). The model is thermally fully predictive so it requires only electrical performance
Available online 6 April 2017
parameters of the battery to calculate its temperature during discharging. Using the model, detailed
investigation of the effects of the variation of the macro-scale thermophysical properties and the entropic
Keywords:
term of the heat generation model was carried out. Results show that specific heat is a critical property
Li-ion battery
CFD
that has a significant impact on the simulation results whereas thermal conductivity has relatively minor
Natural convection importance. Moreover, the experimental data can be successfully predicted without taking the entropic
Thermophysical properties term into account in the calculation of the heat generation. The difference between the experimental and
Heat generation predicted battery surface temperature was less than 3 °C for all discharge rates and regardless of the
Thermal behavior usage history of the battery. The developed model has the potential to be used for the investigation of
the thermal behavior of Li-Ion batteries in different packaging configurations under natural and forced
convection.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction
Lithium ion (Li-Ion) batteries are one of the most potent sources
of energy, which are used to produce and back up electrical energy.
Their high energy and power densities make them ideal to power
high performance machines and devices. Li-Ion batteries are
already used in modern electrical vehicles, cellphones, laptop
computers and many other devices [2]. However, substantial safety
issues were faced during the operations of systems using Li-Ion
batteries [3]. These issues were generally related to thermal
behavior of these batteries [4,5].
⇑ Corresponding author.
E-mail address: ozgur.ekici@hacettepe.edu.tr (Ö. Ekici).
In the near future it can be expected that these batteries will
have minimum five times as much higher energy and power den-
sity as a result of rigorous research aiming to make them sufficient
enough to satisfy the power requirements of the future machines
and equipment [6]. On the other hand, while even today’s Li-Ion
batteries cannot be used as completely safe equipment, the higher
performance batteries of the future will possibly create more
safety problems due to their higher energy and power densities.
Therefore, understanding and modeling of thermal behavior of
Li-Ion batteries is essential.
Chen and Evans, Newman and Tiedemann, Evans and White,
and Al-Hallaj and his research group carried out extensive research
on thermal behavior of Li-Ion batteries [7–17]. These studies com-
prise a large area of research from cell level to cell stack (battery)
and even battery pack levels. Effect of material properties [8,9],

http://dx.doi.org/10.1016/j.applthermaleng.2017.04.017
1359-4311/Ó 2017 Elsevier Ltd. All rights reserved.
 E. Gümüsßsu et al. / Applied Thermal Engineering 120 (2017) 484–495
Nomenclature
cp specific heat capacity (J kg1 K1)
DoD depth of discharge
g gravity acceleration (m2 s1)
I current (A)
k thermal conductivity (W m1 K1)
OCV open circuit voltage (V)
p pressure (Pa)
q heat generation (W)
q_ volumetric heat generation (W m3)
T temperature (K)
t time (s)
V air velocity (m s1)
geometrical assumptions [12] and dynamic power profiles [11]
were investigated in some of these studies. In addition, thermal
management and cooling methods have been among the topics
under investigation especially in recent years [13–16,18–28]
including forced convection in a specified layout or in a battery
package [18,20,21,23], variable patterns of forced convection
[19,28] and usage of phase change materials for cooling instead
of air [13–16,22,26].
Broadly, thermal modeling of a Li-Ion battery includes the heat
generation problem inside the battery due to electrochemical reac-
tions and heat transfer problem as conduction inside the battery
and free or forced convection between the battery surface and
the surrounding medium. Therefore, the thermal models of Li-Ion
batteries can be classified based on the methods to handle the heat
generation and solution approaches used for the heat transfer
problem. In the literature, there are various approaches for han-
dling the heat generation inside the battery and the heat transfer
problem during the course of thermal modeling.
As seen in Fig. 1(a), the heat generation in the battery can be
specified either based on direct experimental measurements or
based on calculations, which in turn may use empirical data. Mod-
els using direct measurement techniques [13–16,22,23,26,27] usu-
ally employ accelerated rate calorimeter or a similar experimental
set up to measure heat generation of the battery. Models calculat-
ing heat generation can be generally discussed in two major
groups. In the first group [29–33] heat generation is calculated
via electrochemical reactions. These reactions can be utilized for
calculation of heat generation [29–31,33] or for calculation of elec-
trical properties of the battery [32], as preferred. In the second
a b
Methods to handle heat generation
Direct measurement Calculation of
of heat generation heat generation
[13-16, 22, 23, 26, 27]
by means of measured
electrical properties
and special formulas
[7-12, 17-21, 24, 25, 28, 35-39]
Fig. 1. Classification of literature about modeling of thermal behavior of Li-Ion batteries.
485
dOCV
dT
entropic term (V K1)
q density (kg m3)
b thermal expansion coefficient (K1)
l dynamic viscosity (N s m2)
Subscripts
1 ambient
a air
b battery
s battery surface
group [32], specified electrical properties are used in a model cal-
culating heat generation separately. Electrochemical models are
generally based on concentrated solution theory proposed by New-
man [34]. These models have intrinsically high accuracy obtained
at the expense of simplicity; hence, they are not suitable to inves-
tigate characteristics of more complex systems such as a battery
(cell stack) or a battery pack from practical point of view. They
require quite a high number of parameters (>50) as they consider
all the details of a battery operation such as anode and cathode
reactions, transport of Li ions, etc. Therefore, these models are gen-
erally used to analyze electrical performance and thermal behavior
of a single cell.
Models based on experimentally measured electrical properties
of the battery [7–12,17–21,24,25,28,35–39] calculate the heat gen-
eration using the equation of Bernardi et al. [1] or ohmic law. These
models are relatively simple compared to electrochemical models
and they are suitable for more complex systems like battery packs.
The equation of Bernardi et al. [1] is one of the most popular equa-
tions used in these kinds of studies. It consists of two parts; irre-
versible heat generation and reversible heat generation. In the
literature there are different approaches about the usage of this
equation. Evans and White [12] completely neglected reversible
heat generation by assuming zero entropic term, while Chen and
Evans [8–11] used constant entropic term which indeed demon-
strates significant change with respect to depth of discharge
(DoD) [40]. In addition, open circuit voltage (OCV) is assumed con-
stant in some studies [8,9,11,12,17] during the calculation of irre-
versible heat generation, although it is known that OCV has a
substantial variation during discharge similar to the entropic term
Solution approaches for heat transfer problem
Models solving Models solving
conduction and convection
only conduction
[7-12] [13]
[14-17] [18-20]
[21] [22, 23, 28, 33]
by means of [24-27]
electrochemical [29-32]
calculations [35-39]
[29-33]
486 E. Gümüsßsu et al. / Applied Thermal Engineering 120 (2017) 484–495

[7]. On the other hand, Al Hallaj et al. and Rad et al. considered
variable OCV and entropic term with respect to DoD in their stud-
ies [7,38].
The classification of thermal models of Li-Ion batteries based on
the solution approach is given in Fig. 1(b) and can be discussed in
two groups. One of these groups includes models focusing only on
conduction inside the battery by introducing convection just as a
boundary condition with a specified convection coefficient [7–12,
14–17,21,24–27,29–32,35–39], while the other group of models
solves both conduction throughout the battery and convection
around the battery [13,18–20,22,23,28,33]. The convection solu-
tion inherently involves in the solution of the flow field around
the battery under free or forced convection conditions.
This study falls into the same categories of literature calculating
heat generation while solving conduction and convection. How-
ever, to the best of authors’ knowledge, the effects of thermophys-
ical properties and heat generation calculation models have not
been addressed in the context of 3D CFD approach. In this study,
a 3D computational model was developed for thermal behavior Fig. 2. Schematic definition of the thermal problem.
of lithium ion batteries under natural convection. The model solves
complete flow field around the battery as well as conduction inside
the battery. In addition, macro-scale thermophysical properties,
qa ffi q1 ½1  bðT a  T 1 Þ ð4Þ
entropic term and usage history of the battery as well as discharge
@T b
rate were analyzed to quantify their respective effects in the com- qb cp;b ¼ kb r:ðrT b Þ þ q_ ð5Þ
putational solution. The model is thermally fully predictive, so only @t
power profile from the demand side is required to calculate battery Eq. (5) shows the energy transport equation for the battery.
temperature. Electrical performance parameters of the battery due Source term for heat generation in this equation is modeled using
to the required power profile were obtained through in-house the well-known equation of Bernardi et al. [1]:
experimental studies and heat generation was modeled using the  
equation of Bernardi et al. [1] with these parameters. In the calcu- dOCV
q ¼ I ðOCV  VÞ  T b ð6Þ
lation of the heat generation, variations of OCV and entropic term dT
with respect to DoD were taken into account in accordance with
In Eq. (6), the first group of terms in parenthesis represents irre-
experimental observations. Using the model developed in this
versible heat generation (Joule heating) while the second term
study, effects of different operational and environmental condi-
stands for reversible heat generation in the battery. In this study,
tions on thermal behavior of Li-Ion batteries can be further ana-
all of the terms (OCV, V, I and dOCV dT
) in Eq. (6) with the exception
lyzed and investigated. Furthermore, the model can be used as a
of temperature are experimentally determined as a function of
tool for designing novel packaging configurations of Li-Ion batter-
DoD specifically for the battery used in this work, which can be
ies under free or forced convection.
also expressed as a function of time as required. Although the bat-
tery temperature is measured and recorded during discharge tests,
2. Theory and modeling predicted temperature values by the model are preferred to be
used in the simulations for the calculation of reversible heat gener-
The thermal behavior of Li-ion batteries during discharge is ation term in order to provide a self-consistent model.
modeled using a 3D CFD approach. The model, composed of a Based on initial tests that were performed, the OCV variation
cylindrically shaped battery with internal heat generation and with respect to temperature was found to affect the irreversible
the surrounding air, solves the fundamental mass, momentum heat generation at most 2%, so this dependency on temperature
and energy equations as required. Internal heat generation occurs is neglected for the calculation of the irreversible heat. In the equa-
as a result of electrochemical reactions taking place during charge tion of Bernardi et al. [1], any change in OCV is multiplied only by
or discharge of batteries. Schematic definition of the thermal prob- current for the calculation of the irreversible heat. Consequently,
lem is shown in Fig. 2. contrary to the reversible heat – temperature dependent OCV vari-
ation doesn’t have an important impact for the irreversible heat.
r:V a ¼ 0 ð1Þ Neglecting this variation, the irreversible heat is determined only
as a function of DoD.
DV a In this study, galvanostatic discharge is applied so that DoD var-
qa ¼ qa g  rpa þ la ðr2 V a Þ ð2Þ ies linearly as a function of time. Therefore, for each discharge rate,
Dt
Eq. (6) can be written in terms of time and this time dependent
@T a equation can be used to calculate the heat generation within the
qa cp;a þ qa cp;a r:ðT a V a Þ ¼ r:ðka rT a Þ ð3Þ
@t battery in the model.
Due to low velocities, flow is assumed incompressible and as a
result of that continuity equation can be written in the form of Eq. 3. Experimental set up and methodology
(1). Conservation of momentum for an incompressible flow is rep-
resented with Eq. (2) whereas conservation of energy is shown in The experimental part of the study involves in the measure-
Eq. (3). Buoyancy forces due to temperature difference in the air ment of the performance of Li-ion batteries (Panasonic NCR18650B
should be incorporated in the momentum equation and this is type) with a battery testing system (MACCOR model 4300) as well
accomplished by employing Boussinesq approximation as shown as the measurement of battery surface temperature during dis-
in Eq. (4). charge. A natural convection oven (Nuve FN300) was used to carry
 E. Gümüsßsu et al. / Applied Thermal Engineering 120 (2017) 484–495 487

out the measurements at higher temperatures than the ambient. A Table 1

platform was employed to hold the battery from its tips and the Corresponding currents at each discharge rate for Panasonic NCR18650B type battery.

contact between battery and battery holder was small enough that Discharge rate Current (A)
the battery was treated as being fully exposed to air without any 0.5C 1.625
contact. 1.0C 3.25
The battery was placed horizontally on the holder. In order to 1.5C 4.875
measure the surface temperature of the battery a capillary J type
thermocouple was used. The acquisition and processing of temper-
ature data were carried out with a data acquisition system (NI USB 4.2
6341). Surface temperature values were collected from the top sur-
face of the battery.
4
3.1. Determination of OCV, V and entropic term

3.1.1. Operating voltage measurements 3.8

OCV (V)
Voltage measurements were performed during galvanostatic
discharge (constant current) via Maccor battery testing system
showing operating voltage of the battery. Before the galvanostatic 3.6
discharge, the battery was fully charged and rested for 14 h to
become electrochemically stable. This duration was determined
experimentally by recording the variation of the OCV after charge 3.4
and discharge processes and observing the electrochemical behav-
ior of the battery during this time interval. During operating volt-
age tests, battery and ambient temperatures were also measured. 3.2
Since the operating voltage changes with respect to discharge rate, 0 0.2 0.4 0.6 0.8 1
tests were conducted separately for each discharge rate modeled in DoD
this study, as seen in Fig. 3. Considering Panasonic NCR18650B
Fig. 4. Variation of OCV with DoD.
type battery’s nominal capacity, current values at galvanostatic
discharge process are specified for each discharge rate as shown


in Table 1. 3.1.3. Entropic term dOCV
dT
measurements
In order to calculate the reversible heat generation in the bat-
tery, the entropic term should be determined experimentally as a
3.1.2. OCV measurements
function of DoD. During entropic term experiments the battery
The variation of the OCV with respect to DoD was determined
was discharged at 1.0C rate to a specified DoD value defined by
with a set of experiments in which battery’s OCV was measured
0.1 DoD intervals. Following stabilization after discharge, the bat-
in 0.1 DoD intervals. After each measurement the battery was
tery was placed in an oven kept at 55 °C. The oven temperature
again fully charged in order to relieve it from any possible effects
was set considering the operating range of the battery in this study.
of the self-discharge. Fully charged battery was discharged at
The battery was kept inside the oven for five hours while being
1.0C rate to a specified DoD value and rested to become electro-
connected to the battery tester in open circuit state. In this dura-
chemically stable. The recorded value after the rest was taken as
tion, the variation of OCV was recorded and compared with the val-
the OCV value corresponding to the specified DoD. All tests were
ues measured at the ambient temperature. Using these data, the
performed in ambient conditions. The results of the tests show
entropic term was calculated by employing a finite difference
an almost linear relationship between OCV and DoD as depicted
approximation for the first order derivative as shown below.
in Fig. 4.
dOCV OCVðT 2 Þ  OCVðT 1 Þ
Entropic term ¼ ffi ð7Þ
dT ðT 2  T 1 Þ
0.5C
4 The variation of the entropic term with respect to DoD is non-
1.0C
1.5C linear and mostly positive in the entire range of the discharge as
3.8
shown in Fig. 5. A similar entropic term behavior was also observed
3.6 by Bandhauer et al. [41]. Considering the nonlinear behavior of the
entropic term, a piecewise linear approximation was initially spec-
Voltage (V)

3.4 ified to approximate the reversible heat generation in this study.

This approximation is discussed in Section 5.3 in detail.
3.2

3 3.2. Repeatability of surface temperature measurements

2.8
Repeatability tests were conducted at 1.0C discharge rate under
2.6 almost identical ambient temperature. All of these tests were per-
formed with a fully charged battery rested for stabilization. In
2.4 total, five tests were completed in consecutive days. Results of
0 20 40 60 80 100 120
these tests are shown in Fig. 6. It is seen that results of repeatability
time (minutes)
tests exhibit a set of curves which are quite similar to each other,
Fig. 3. Variation of the voltage with time during discharge for different discharge qualitatively and quantitatively. The mean of the standard devia-
rates. tion of all temperature measurements was found to be 0.2 °C with
488 E. Gümüsßsu et al. / Applied Thermal Engineering 120 (2017) 484–495

0.3 4.2. Domain, mesh, boundary and initial conditions

In order to obtain a reliable mesh structure, mesh dependency

0.2
analysis was performed with four different meshes. In Fig. 7, along
with temperature variations, calculated convection coefficients on
the top surface of the battery are shown graphically in order to
dOCV/dT (mV/K)

0.1
examine their mesh dependency. The convection heat transfer
coefficient was determined from the surface energy balance equa-
0 tion given in Eq. (8) using the battery surface temperature obtained
from the CFD solution. A first order approximation was used to
-0.1 evaluate the temperature gradient at the battery surface.

ka dT a

-0.2 h¼ dn s
ð8Þ
ðT s  T 1 Þ
As seen in Fig. 7, all meshes result in similar temperature pro-
0 0.2 0.4 0.6 0.8 1 files. However, calculated convection coefficients exhibit substan-
DoD tial variations depending on the mesh structure. Only after mesh
3, the convection coefficient becomes nearly invariant with respect
Fig. 5. Variation of the entropic term with DoD. to mesh, so that any further increase in the number of cells does
not result in an important difference. On the other hand, despite
of giving similar results, solution time of mesh 3 was considerably
55
less than mesh 4. Therefore, Mesh 3 was selected as the mesh
structure in this study. Mesh structure of the model was efficiently
designed by employing a mesh grading to have a higher mesh den-
50 sity in the proximity of the battery. Other parts of the domain have
a coarser mesh so that any unnecessary computational effort was
Temperature ( C)

prevented during simulations. As a result of that, a mesh structure

45 having 326,708 cells was obtained with the smallest cell height of
1e4 m just above the battery and with the maximum cell volume
of 3.002438e6 m3. Using this mesh structure the model runs
40 approximately 2 days on a HP Z640 workstation to obtain a con-
verged solution. The mesh structure of the battery surface at the
symmetry plane can be seen in Fig. 8.
35 The specified boundary and initial conditions are shown in
Table 2. The total pressure and temperature at the domain bound-
aries are always equal to atmospheric pressure and temperature.
30 No slip condition is valid on the solid surface. As an initial condi-
0 10 20 30 40 50 60
tion, temperature of the whole domain, including the battery, is
time (minutes) equal to ambient temperature. At the beginning of the simulation,
air pressure also equals to atmospheric pressure and velocity of the
Fig. 6. Variation of surface temperature in different tests during discharge at 1.0C.
air is zero.

a maximum of 0.5 °C. Therefore the test procedure and results are
confirmed to be repeatable. 4.3. Evaluation of the thermophysical properties

Temperature dependent air properties are incorporated in the

4. Implementation of the numerical model with computational
fluid dynamics (CFD) approach
4.1. Solution parameters
The model is implemented with CFD approach employing a
commercial CFD software, ANSYS Fluent 14.5. A pressure based,
laminar, incompressible, transient solver is used in the simula-
tions. The effect of buoyancy forces as a result of density differ-
ences due to varying temperature of the air is modeled using
Boussinesq approximation in the incompressible solver. Second
order upwind scheme is preferred for the momentum and energy
equations. The SIMPLE algorithm is selected for pressure and veloc-
ity coupling.
Heat generation in the battery – determined as a function of
time – is integrated into the solver via a user-defined function
(UDF) developed for this study. This UDF takes the battery temper-
ature from the main solver and employs Eq. (6) to calculate the
heat generation during the discharge process.
simulations using their corresponding values at standard atmo-
spheric pressure. Physical properties of the battery were deter-
mined using the values from the literature.
In the literature, thermal conductivity of the Li-ion batteries
varies substantially. In some studies it was specified assuming
the battery as an isotropic material [7,14,42,43], while in some
other studies anisotropy was taken into account [13,44–47]. In
studies with anisotropic battery approximation, thermal conduc-
tivity values range from 0.219 W/mK to 3.40 W/mK in radial direc-
tion while the range is in between 20 W/mK and 30 W/mK in the
axial one [13,44–47]. For isotropic batteries, thermal conductivity
varies between 0.219 W/mK and 3 W/mK in the literature
[7,14,42,43]. Specific heat is also a controversial property in the lit-
erature. It was presented in the range of 814 J/kg K and 2400 J/kg K
[7,13,14,42–47].
The density of the battery was assumed to be uniform and cal-
culated by dividing its mass to its volume. All thermophysical
properties of the battery are presumed to be independent of tem-
perature. Properties of the battery are shown in Table 3.
 E. Gümüsßsu et al. / Applied Thermal Engineering 120 (2017) 484–495 489

60 3.5
mesh 1 - 28150 cells mesh 1
mesh 2
55 mesh 2 - 206023 cells 3 mesh 3
mesh 3 - 326708 cells mesh 4
50
mesh 4 - 577423 cells 2.5
Temperature ( C)

45

htop (W/m2K)
2
40
1.5
35

1
30

0.5
25

20 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
time (minutes) time (minutes)

Fig. 7. Mesh dependency analysis. k = 28.05 W/mK, cp = 1300 J/kg K, discharge rate = 1.0C.

Fig. 8. Mesh structure. (a) 2D view of the region, in the vicinity of the battery, at the middle plane (b) 3D view of the region close to the battery.

Table 2 5. Results and discussions

Boundary and initial conditions.

Boundary conditions
5.1. Effect of thermophysical properties
Battery surface No slip condition
Flow domain boundary ptotal;a ¼ p1 , T total;a ¼ T 1 5.1.1. Specific heat
Initial conditions Thermophysical properties of a battery are one of the most
Battery Tb ¼ T1 important parameters for the development of a computational
Flow domain T a ¼ T 1 ; V a ¼ 0, pa ¼ p1 , model to predict the thermal behavior. Due to a wide range of mate-
ptotal;a ¼ pstatic;a þ 12 qa V 2a rials used in the composition of Li-ion batteries, these properties are
T total;a ¼ T static;a (based on incompressible flow assumption) found to be highly diversified in the literature so that the determi-
p1 = 90.8 kPa, T 1 ¼ 25  C
nation of properties of a specific battery is a difficult task. Moreover,
their effects on a computational solution have not been systemati-
Table 3 cally investigated. In order to elaborate the problem systematically,
Thermophysical properties. a range of values from the literature for each thermophysical prop-
erty is specified, and their effects on the computational solution are
Thermophysical properties Battery Shell
examined. Additionally, comparing the results against the experi-
q (kg/m3) 2939 2059
mental data provides a means to quantify their effects.
cp (J/kg K) 814–2400 875
k (W/mK) 0.219–28.05 0.638 As has been shown in Table 3, specific heat values range from
814 J/kg K to 2400 J/kg K for Li-ion batteries in the literature. Four
different values of specific heat were chosen within this interval
The surface of the battery includes an aluminum sheet and a and used in the model with 3 W/mK thermal conductivity. The
protective plastic coverage. Thermophysical properties of PVC model was run at various discharge rates of 0.5C , 1.0C and 1.5C
were used as the material properties of the protective plastic. and results are compared with the corresponding experimental
These aluminum sheet and plastic coverage are assumed as a data.
unique composite material having 0.4 mm thickness and its prop- All model results are comparable with experimental data in
erties were derived by averaging the properties of aluminum and terms of error and trend as shown in Fig. 9 and Table 4. However
PVC. It is modeled with a shell conduction approach in the simula- as seen in Table 4, the lower two specific heat values (1300 J/
tion without creating any mesh for the shell itself. Properties of the kg K and 1600 J/kg K) result in higher errors for all discharge rates
resultant composite material are also shown in Table 3. compared to 1900 J/kg K and 2400 J/kg K. Among the higher two
490 E. Gümüsßsu et al. / Applied Thermal Engineering 120 (2017) 484–495

70 70
0.5C 1.0C

Temperature ( C)

Temperature ( C)
60 60

50 50

40 40

30 30

0 50 100 150 0 20 40 60
time (minutes) time (minutes)

70 c = 1300 J/kgK
1.5C p
Temperature ( C)

60 c = 1600 J/kgK
p

50
cp = 1900 J/kgK

c = 2400 J/kgK
40 p
experimental
30

0 10 20 30 40
time (minutes)

Fig. 9. Variation of surface temperature with time for different specific heats at various discharge rates. k = 3 W/mK.

Table 4
Root mean square errors (°C) for different specific heats at different discharge rates.

Discharge rates Specific heats (J/kg K)

1300 1600 1900 2400
0.5C 3.4 2.8 2.3 1.6
1.0C 5.1 3.4 2.2 1.1
1.5C 6.0 3.2 1.8 3.0
Average 4.8 3.1 2.1 1.9

specific heat values, 2400 J/kg K gives the most accurate result for 50
all discharge rates, so it is preferred to be used as the nominal k = 0.3 W/mK
k = 1 W/mK
specific heat value of the battery for the rest of the study.
k = 3 W/mK
45 k = 15 W/mK
5.1.2. Thermal conductivity k = 28.05 W/mK
Following the selection of the specific heat value a similar pro- kaxial = 28.05 W/mK, kradial = 3 W/mK
Temperature ( C)

cedure is applied to determine the effects of the thermal conduc-

40
tivity. Six different thermal conductivity values, ranging between
the minimum and the maximum values given in Table 3, are cho-
sen from the literature to investigate. The computational model is
run separately with each of these thermal conductivity values at 35
1.0C discharge rate. In addition to different thermal conductivities,
anisotropic battery assumption is also investigated by defining
axial and radial thermal conductivities separately in one of the 30
simulations.
Fig. 10 shows that the choice of a thermal conductivity value
has a minor impact on the predicted temperature values. It is seen
25
that an increase in the thermal conductivity causes a slight 0 10 20 30 40 50 60
decrease, which is less than 2 °C, in the calculated temperatures. time (minutes)
It is also important to note that a non-isotropic approach by assign-
Fig. 10. Variation of surface temperature with time for different thermal conduc-
ing different thermal conductivity values in axial and radial direc-
tivities. cp = 2400 J/kg K, discharge rate = 1.0C.
tions has negligible effect on the results. Similar results are
obtained for 0.5C and 1.5C discharge rates and consequently, it is
found that an isotropic material approach in thermal modeling The relatively small effect of thermal conductivity values in the
seems to be accurate enough for the regular discharge rates inves- thermal behavior can be further investigated by considering both
tigated in this study. the convection and conduction heat transfer processes during a
 E. Gümüsßsu et al. / Applied Thermal Engineering 120 (2017) 484–495 491

discharge. During a typical discharge, the average convection coef- x 10

5

ficient around the battery is numerically calculated and plotted 2.5

0.5C rev.
along with the results of the well-known free convection empirical
1.0C rev.
correlations proposed by Churchill and Morgan as cited in Incrop- 2 1.5C rev.
era et al. [48] in Fig. 11. Except the very early times of discharge, a 0.5C irrev.

heat generation (W/m )

good agreement is obtained especially with the correlation of Mor-

3
1.0C irrev.
gan. It is important to note that, the main reason of discrepancy at 1.5 1.5C irrev.
early times is due to different initial conditions, i.e., the battery or
cylinder is assumed to be heated from the beginning of the process
1
in the correlations whereas its temperature is the same as the envi-
ronment temperature at the beginning in the numerical calculation
as in the case of experiments. 0.5
In order to quantify the relative importance of the conduction
heat transfer compared to the convective heat transfer the Biot
0
number is also calculated. According to Fig. 12, Biot number is
below 0.1 for all of the thermal conductivity values except for
0.3 W/mK. It means that majority of the suggested thermal con- -0.5
0 20 40 60 80 100 120
ductivity values in the literature result in a nearly uniform temper-
time (minutes)
ature distribution inside the battery. As a result, while most of the
Fig. 13. Variation of heat generation in the battery with time for different discharge
rates.

8
numerical
7 Churchill thermal conductivity values give almost identical results, only val-
Morgan ues close to the lowest limit of the literature create a slightly dif-
6 ferent temperature profile, still less than 2 °C compared to the
others, during a regular discharge. Therefore, the thermal conduc-
5 tivity value of 3 W/mK, a widely used value in the literature, is cho-
haverage (W/m2K)

sen as the nominal value for the simulations conducted in this

4
study.
3
5.2. Effect of discharge rate
2
Electrical response of a battery changes with a discharge rate as
1
shown in Fig. 3 before. This variation in electrical response affects
0 thermal behavior of the battery as well. Discharge rate affects both
the irreversible and reversible heat generations, while the variation
-1 of the irreversible part constitutes a larger portion in magnitude
0 10 20 30 40 50 60
and dominates the overall increase. As it is presented in Fig. 13,
time (minutes)
the irreversible heat generation increases with an increasing dis-
Fig. 11. Variation of the calculated convection coefficient with time in comparison charge rate, whereas the reversible one decreases. Heat generation
with correlations in the literature. k = 3 W/mK, cp = 2400 J/kg K, discharge values shown in Fig. 13 are consistent with studies of Al Hallaj
rate = 1.0C. et al. and Bandhauer et al. [7,41].
In Fig. 14, calculated transient surface temperature results for
three different discharge rates are compared with the experimen-
tal results. Calculated and measured temperature profiles are con-
0.25 sistent with the experimental results in the literature [7].
k = 0.3 W/mK
k = 1 W/mK
According to the experimental data and model results the surface
k = 3 W/mK temperature increases substantially with discharge rate. In addi-
0.2 k = 15 W/mK tion to temperature increase, the tendency of the temperature pro-
k = 28.05 W/mK
file also changes. At 0.5 °C discharge rate the temperature profile is
nearly constant in most of the discharge process. Then the profile
Biot number

0.15 becomes steadily incremental at 1.0C discharge rate with a vari-

able slope. In the case of 1.5C discharge rate, the temperature pro-
file shows an almost linear increase. These experimental
0.1 tendencies are well predicted and can be seen in the model results
for all of the three discharge rates. The maximum average error of
the model results is obtained at 1.5C discharge rate as 3.0 °C as pre-
0.05 sented in Table 4.

5.3. Effect of the entropic term

0
0 10 20 30 40 50 60
As mentioned in earlier sections, the entropic term is used to
time (minutes)
calculate the reversible heat generation in the battery. On account
Fig. 12. Variation of the Biot number with time for different thermal conductivities. of being a parameter requiring significant efforts to measure, the
cp = 2400 J/kg K, discharge rate = 1.0C. entropic term should be investigated to understand its effect on
492 E. Gümüsßsu et al. / Applied Thermal Engineering 120 (2017) 484–495

60 50
0.5C experimental
55 experimental 0.5C 1st approximation
1.0C 45 2nd approximation
50 experimental 1.0C 3rd approximation
1.5C
4th approximation

Temperature ( C)
Temperature ( C)

experimental 1.5C 40
45

40 35

35
30
30
25
25

20 20
0 20 40 60 80 100 120 0 10 20 30 40 50 60
time (minutes) time (minutes)

Fig. 14. Variation of the surface temperature with time for different discharge rates. Fig. 16. Variation of the surface temperature with time for different entropic term
k = 3 W/mK, cp = 2400 J/kg K. approximations. k = 3 W/mK, cp = 2400 J/kg K, discharge rate = 1.0C.

the process and on the reliability of the model. This is accom- Table 5
plished through employing four different approaches to model dis- Usage details of five different batteries.
charge test at 1.0C rate. The first approach uses piecewise linear
B1 B2 B3 B4 B5
equations to obtain an entropic term function with respect to
Usage (Charge + Discharge) >36  2 29  2 10  2 32 22
DoD. The second approach mimics the main trend of the entropic
term which increases before 0.7 DoD and then shows a decrease.
Therefore, it employs two linear equations fitted to the experimen-
tal data set. The third approach considers the whole data set as a 55
linear equation and the fourth approach assumes zero entropic B1
B2
term in the whole discharge process neglecting the reversible heat
B3
generation in the battery completely. These approximations are 50 B4
presented graphically in Fig. 15. B5
As seen in Fig. 16, the surface temperature can be accurately
Temperature ( C)

predicted with all of the four entropic term approaches. The effect 45
of the entropic term is more pronounced after 30 min.
The main conclusion drawn from these results is that the irre-
versible heat generation is the major parameter defining the bat-
40
tery temperature during discharge. Considering the tedious
process of the entropic term measurement, using only the irre-
versible heat which can be calculated easily and precisely, one
35
can have accurate results in terms of battery temperature
variation.

30
0 10 20 30 40 50 60 70
0.3 time (minutes)
experimental
1st approximation
Fig. 17. Variation of surface temperature with time using batteries of different
0.2 2nd approximation
usage history. Discharge rate = 1.0C.
rd
3 approximation
4th approximation
0.1
dOCV/dT (mV/K)

5.4. Effect of usage history

0 During experimental studies usage or cycling is noticed as

another important parameter affecting both electrical performance
and thermal behavior of the lithium ion batteries. In order to reveal
-0.1 the effects of usage, five different batteries, having various opera-
tional backgrounds depicted in Table 5, are used to discharge at
1.0C discharge rate. During these tests both electrical and thermal
-0.2
responses of these batteries are measured and corresponding
temperature variations are shown in Fig. 17.
Fig. 17 clearly shows that thermal responses of the batteries
0 0.2 0.4 0.6 0.8 1
DoD separate from each other depending on their operational
backgrounds. Generally the more the battery is used, the higher
Fig. 15. Various approximations for the entropic term. temperature increase it demonstrates. The main difference occurs
 E. Gümüsßsu et al. / Applied Thermal Engineering 120 (2017) 484–495 493

55 Table 6
model-new battery Maximum temperature variation in the battery at the end of the discharge.
model-old battery Discharge rates DTmax
50 experimental-new battery
Radial direction (°C) Axial direction (°C)
experimental-old battery
0.5 °C 0.2 0.5
1.0 °C
Temperature ( C)

45 0.5 1.1
1.5 °C 0.5 2.0

40
accurate results regardless of the battery usage history as long as
35
the corresponding electrical data for each battery are provided to
the model.

30
5.5. Temperature distribution and velocity profile at the end of the
discharge
25
0 10 20 30 40 50 60
As seen in Fig. 19(a), temperature increase of the battery affects
time (minutes)
air temperature in the proximity of the surface. In addition, above
Fig. 18. Model and experimental results for a new and old battery. k = 3 W/mK, the battery a thermal plume is observed at all discharge rates as a
Cp = 2400 J/kg K, discharge rate = 1.0C. result of natural convection. As expected, the symmetry plane at
the upper side of the battery has higher temperature compared
to lower side and end points. Temperature distribution throughout
at the beginning of the discharge. Batteries which are used more the battery is nearly uniform; only small temperature differences
face with faster temperature increase approximately during first occur in the battery, as seen in Table 6. At the lowest discharge rate
10 min, which can be deducted from the difference in the slopes. temperature variations are lower due to smaller heat generation
The remaining part of discharge has almost the same slope for all and longer discharge duration. Higher discharge rates cause higher
batteries. temperature differences in the battery especially in axial direction.
The developed model was run to simulate temperature behav- In radial direction, no temperature variation is observed at 1.0C
iors of a used and a new battery. As seen in Fig. 18, the model gives and 1.5C discharge rates. It can be argued that, temperature

Fig. 19. Temperature distribution (°C) and velocity profile (m/s) at the end of discharge. k = 3 W/mK, cp = 2400 J/kg K, for 0.5C, 1.0C and 1.5C discharge rates, from left to right.
494 E. Gümüsßsu et al. / Applied Thermal Engineering 120 (2017) 484–495
variation in radial direction reaches a stable condition at higher
discharge rates but the difference in axial direction continues to
increase.
In Fig. 19(b) velocity vectors are shown graphically. Flow veloc-
ities around the battery are low and symmetrical with respect to
vertical axis of battery’s cross section as expected. At 1.5C dis-
charge rate, the highest velocity is observed due to the highest
temperature increase but the difference in velocity values between
discharge rates is not significant.
6. Conclusions
A 3D CFD model was developed to predict the temperature vari-
ation of a Li-ion battery subjected to natural convection during dis-
charge using electrical performance data of the battery. The model
solves both the conduction heat transfer in the battery and convec-
tion heat transfer between the battery surface and the surrounding
medium along with the fluid flow around the battery. Comparison
of the model results with the in-house experimental data shows
that the model can predict the battery temperature with an aver-
age error of 2 °C considering all of the three discharge rates
(0.5C, 1.0C, 1.5C) tested in this work.
The model was employed to investigate the effects of the ther-
mophysical properties of the battery on the computational solu-
tion. A wide range of specific heat values were tested and results
showed that it was a major parameter affecting battery tempera-
ture. Therefore, an accurate prediction of battery temperature
requires the usage of the correct value of specific heat in the sim-
ulations. On the other hand, thermal conductivity was found to
have a minor impact on the results. Furthermore, no considerable
difference was obtained in the results of isotropic and anisotropic
battery models.
The heat generation in the battery was modeled using the equa-
tion proposed by Bernardi et al. [1], which is comprised of irre-
versible (Joule heating) and reversible (entropic) terms. The
Entropic term is the most intractable parameter for the calculation
of the heat generation. The entropic term was calculated based on
experimental OCV measurements and it was shown that battery
temperature can be successfully predicted by only taking the irre-
versible term into account in the calculation of the heat generation.
The entropic term slightly affects the time evolution of the temper-
ature only during a limited part of the discharge process.
Using the results of this study, different operational and envi-
ronmental conditions and various packaging options for Li-Ion bat-
teries can be modeled to investigate their thermal response.
Acknowledgements
The authors would like to thank the Scientific Research Projects
Coordination Unit of Hacettepe University for their support in the
funding of this research project (project code: FBA-2015-5419).
References
[1] D. Bernardi, E. Pawlikowski, J. Newman, A general energy balance for battery
systems, J. Electrochem. Soc. 132 (1985) 5–12.
[2] S.L. Wu, W. Zhang, X. Song, A.K. Shukla, G. Liu, V. Battaglia, V. Srinivasan, High
rate capability of Li(Ni1/3Mn1/3Co1/3)O2 electrode for Li-ion batteries, J.
Electrochem. Soc. 159 (2012) A438–A444.
[3] Q. Wang, B. Jang, B. Li, Y. Yan, A critical review of thermal management models
and solutions of lithium ion batteries for the development of pure electric
vehicles, Renew. Sustain. Energy Rev. 64 (2016) 106–128.
[4] T.M. Bandhauer, S. Garimella, T.F. Fuller, A critical review of thermal issues in
lithium-ion batteries, J. Electrochem. Soc. 158 (2011) R1–R25.
[5] D. Doughty, E.P. Roth, A general discussion of li ion battery safety, Electrochem.
Soc. Interface (Summer) (2012) 37–44.
[6] M.M. Thackeray, C. Wolverton, E.D. Isaacs, Electrical energy storage for
transportation—approaching the limits of, and going beyond, lithium-ion
batteries, Energy Environ. Sci. 5 (2012) 7854–7863.
[7] S. Al-Hallaj, H. Maleki, J.S. Hong, J.R. Selman, Thermal modeling and design
considerations of lithium-ion batteries, J. Power Sources 83 (1999) 1–8.
[8] Y. Chen, J.W. Evans, Heat transfer phenomena in lithium/polymer-electrolyte
batteries for electric vehicle application, J. Electrochem. Soc. 140 (1993) 1833–
1838.
[9] Y. Chen, J.W. Evans, Thermal analysis of lithium polymer electrolyte batteries
by a two dimensional model-thermal behavior and design optimization,
Electrochem. Acta 39 (1994) 517–526.
[10] Y. Chen, J.W. Evans, Thermal analysis of lithium-ion batteries, J. Electrochem.
Soc. 143 (1996) 2708–2712.
[11] Y. Chen, J.W. Evans, Three-dimensional thermal modeling of lithium-polymer
batteries under galvanostatic discharge and dynamic power profile, J.
Electrochem. Soc. 141 (1994) 2947–2955.
[12] T.I. Evans, R.E. White, A thermal analysis of a spirally wound battery using a
simple mathematical model, J. Electrochem. Soc. 136 (1989) 2145–2152.
[13] R. Sabbah, R. Kizilel, J.R. Selman, S. Al-Hallaj, Active (air-cooled) vs. passive
(phase change material) thermal management of high power lithium-ion
packs: limitation of temperature rise and uniformity of temperature
distribution, J. Power Sources 182 (2008) 630–638.
[14] A. Mills, S. Al-Hallaj, Simulation of passive thermal management system for
lithium-ion battery packs, J. Power Sources 141 (2005) 307–315.
[15] S.A. Khateeb, M.M. Farid, J.R. Selman, S. Al-Hallaj, Design and simulation of a
lithium-ion battery with a phase change material thermal management
system for an electric scooter, J. Power Sources 128 (2004) 292–307.
[16] R. Kizilel, R. Sabbah, J.R. Selman, S. Al-Hallaj, An alternative cooling system to
enhance the safety of Li-ion battery packs, J. Power Sources 194 (2009) 1105–
1112.
[17] J. Newman, W. Tiedemann, Temperature rise in a battery module with
constant heat generation, J. Electrochem. Soc. 142 (1995) 1054–1057.
[18] X. Li, F. He, L. Ma, Thermal management of cylindrical batteries investigated
using wind tunnel testing and computational fluid dynamics simulation, J.
Power Sources 238 (2013) 395–402.
[19] R. Mahamud, C. Park, Reciprocating air flow for Li-ion battery thermal
management to improve temperature uniformity, J. Power Sources 196
(2011) 5685–5696.
[20] H. Sun, X. Wang, B. Tossan, R. Dixon, Three-dimensional thermal modeling of a
lithium-ion battery pack, J. Power Sources 206 (2012) 349–356.
[21] D. Chen, J. Jiang, G.H. Kim, C. Yang, A. Pesaran, Comparison of different cooling
methods for lithium ion battery cells, Appl. Therm. Eng. 94 (2016) 846–854.
[22] N.O. Moraga, J.P. Xamán, R.H. Araya, Cooling Li-ion batteries of racing solar car
by using multiple phase change materials, Appl. Therm. Eng. 108 (2016) 1041–
1054.
[23] T. Wang, K.J. Tseng, J. Zhao, Development of efficient air-cooling strategies for
lithium-ion battery module based on empirical heat source model, Appl.
Therm. Eng. 90 (2015) 521–529.
[24] T.H. Tran, S. Harmand, B. Desmet, S. Filangi, Experimental investigation on the
feasibility of heat pipe cooling for HEV/EV lithium-ion battery, Appl. Therm.
Eng. 63 (2014) 551–558.
[25] L. Martín-Martín, J. Gastelurrutia, N. Nieto, J.C. Ramos, A. Rivas, I. Gil, Modeling
based on design of thermal management systems for vertical elevation
applications powered by lithium-ion batteries, Appl. Therm. Eng. 102 (2016)
1081–1094.
[26] N. Javani, I. Dincer, G.F. Naterer, G.L. Rohrauer, Modeling of passive thermal
management for electric vehicle battery packs with PCM between cells, Appl.
Therm. Eng. 73 (2014) 307–316.
[27] C. Lan, J. Xu, Y. Qiao, Y. Ma, Thermal management for high power lithium-ion
battery by minichannel aluminum tubes, Appl. Therm. Eng. 101 (2016) 284–
292.
[28] J. Zhao, Z. Rao, Y. Huo, X. Liu, Y. Li, Thermal management of cylindrical power
battery module for extending the life of new energy electric vehicles, Appl.
Therm. Eng. 85 (2015) 33–43.
[29] G.G. Botte, B.A. Johnson, R.E. White, Influence of some design variables on the
thermal behavior of a lithium-ion cell, J. Electrochem. Soc. 146 (1999) 914–
923.
[30] C. Zhu, X. Li, L. Song, L. Xiang, Development of a theoretically based thermal
model for lithium ion battery pack, J. Power Sources 223 (2013) 155–164.
[31] S. Basu, K.S. Hariharan, S.M. Kolake, T. Song, D.K. Sohn, T. Yeo, Coupled
electrochemical thermal modelling of a novel Li-ion battery pack thermal
management system, Appl. Energy 181 (2016) 1–13.
[32] C.R. Pals, J. Newman, Thermal modeling of the lithium/polymer battery – I.
Discharge behavior of a single cell, J. Electrochem. Soc. 142 (1995) 3274–3281.
[33] W. Tong, K. Somasundaram, E. Birgersson, A.S. Mujumdar, C. Yap, Thermo-
electrochemical model for forced convection air cooling of a lithium-ion
battery module, Appl. Therm. Eng. 99 (2016) 672–682.
[34] J.S. Newman, Electrochemical Systems, Englewood Cliffs, New Jersey, 1991.
[35] D.H. Jeon, S.M. Baek, Thermal modeling of cylindrical lithium ion battery
during discharge cycle, Energy Convers. Manage. 52 (2011) 2973–2981.
[36] G. Karimi, X. Li, Thermal management of lithium-ion batteries for electric
vehicles, Int. J. Energy Res. 37 (2013) 13–24.
[37] S.K. Mohammadian, Y. Zhang, Thermal management optimization of an air-
cooled Li-ion battery module using pin-fin heat sinks for hybrid electric
vehicles, J. Power Sources 273 (2015) 431–439.
[38] M.S. Rad, D.L. Danilov, M. Baghalha, M. Kazemeini, P.H.L. Notten, Adaptive
thermal modeling of Li-ion batteries, Electrochem. Acta 102 (2013) 183–195.
[39] Y. Wang, C. Peng, W. Zhang, Thermal analysis of multifunctional structural
battery for satellite applications, Appl. Therm. Eng. 78 (2015) 209–216.
 E. Gümüsßsu et al. / Applied Thermal Engineering 120 (2017) 484–495
[40] S. Al Hallaj, J. Prakash, J.R. Selman, Characterization of commercial Li-ion
batteries using electrochemical–calorimetric measurements, J. Power Sources
87 (2000) 186–194.
[41] T.M. Bandhauer, S. Garimella, T.F. Fuller, Temperature-dependent
electrochemical heat generation in a commercial lithium-ion battery, J.
Power Sources 247 (2014) 618–628.
[42] J.S. Hong, H. Maleki, S. Al Hallaj, L. Redey, J.R. Selman, Electrochemical-
calorimetric studies of lithium-ion cells, J. Electrochem. Soc. 145 (1998)
1489–1501.
[43] E. Barsoukov, J.H. Jang, H. Lee, Thermal impedance spectroscopy for
Li-ion batteries using heat-pulse response analysis, J. Power Sources 109
(2002) 313–320.
495
[44] H. Maleki, S. Al Hallaj, J.R. Selman, R.B. Dinwiddie, H. Wang, Thermal properties of
lithium-ion battery and components, J. Electrochem. Soc. 146 (1999) 947–954.
[45] S.C. Chen, C.C. Wan, Y.Y. Wang, Thermal analysis of lithium-ion batteries, J.
Power Sources 140 (2005) 111–124.
[46] S.C. Chen, Y.Y. Wang, C.C. Wan, Thermal analysis of spirally wound lithium
batteries, J. Electrochem. Soc. 153 (2006) A637–A648.
[47] N.S. Spinner, R. Mazurick, A. Brandon, S.L. Rose-Pehrsson, S.G. Tuttle,
Analytical, numerical and experimental determination of thermophysical
properties of commercial 18650 LiCoO2 lithium-ion battery, J. Electrochem.
Soc. 162 (2015) A2789–A2795.
[48] F.P. Incropera, D.P. Dewitt, T.L. Bergman, A.S. Lavine, Principles of Heat and
Mass Transfer, seventh ed., John Wiley & Sons, 2013.