See discussions, stats, and author profiles for this publication at: [Link]

net/publication/272020603

Modeling and simulation of the thermal behavior of Li-ion cells

Conference Paper · September 2014

DOI: 10.13140/2.1.1867.3444

CITATIONS READS

0 3,337

5 authors, including:

Andreas Melcher Carlos Ziebert

Karlsruhe Institute of Technology Karlsruhe Institute of Technology
37 PUBLICATIONS 250 CITATIONS 141 PUBLICATIONS 2,276 CITATIONS

SEE PROFILE SEE PROFILE

Boxia Lei Magnus Rohde

Karlsruhe Institute of Technology Karlsruhe Institute of Technology
11 PUBLICATIONS 195 CITATIONS 100 PUBLICATIONS 1,458 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Andreas Melcher on 09 February 2015.

The user has requested enhancement of the downloaded file.

Modeling and simulation of the thermal behavior of
Li-ion cells
Andreas Melcher, Carlos Ziebert, Boxia Lei, Magnus Rohde, Hans Jürgen Seifert
Institute of Applied Materials - Applied Materials Physics (IAM-AWP)
Karlsruhe Institute of Technology

Abstract:
The thermal behavior of Li-ion cells is an important issue with regard to safety considerations. Therefore, in order to
construct a safe Lithium based battery system for use in electrical cars or large stationary electrochemical storage the
behavior of the Li-ion cell has to be known under varying conditions. At present the control of the charge and discharge
process, respectively, as well as the temperature monitoring and control is achieved by so called battery management
systems (BMS) which mostly rely on relative simple charging and discharging algorithms due to the lack of a complete
data base.
One of the objectives of this work is to gain a better understanding of the temperature increase within the cell consider-
ing different heat sources under normal working conditions but also under abuse condition with regard to the so called
“thermal runaway”. Therefore a mathematical multi-scale multi-domain model (MSMD) has been developed based on a
three-dimensional thermal model coupled with a one dimensional electrochemical model. This model was realized using
the commercial finite element (FEM) package COMSOL Multiphysics. First simulations were performed to calculate the
generated heat and temperature increase under various charge and discharge rates.
In order to study the so called “thermal runaway” additional heat sources were implemented into the model which de-
scribes the exponential temperature increase using a phenomenological approach in which an adapted Arrhenius equation
was used as a heat source term in the heat transport equation.1 .

1 Introduction
Lithium-ion batteries (LIB) have been found a wide application in many consumer products in the last 25 years. The
range is spread from cell phones over Tablet and laptop computers until the usage in electric (EV) or hybrid electric
(HEV) driven vehicles. The size and the capacity of the batteries differs with the application. A battery used in an EV or
HEV is much bigger than one used for a cell phone. But the scale up of the batteries leads to a complexity of the physical
phenomenons that do not play any significant role in single or small battery systems.
Physical phenomena can occur over a wide range of time and length scales, starting with atomic variations up to the heat
transfer of the whole battery. It is of imminent importance how these different phenomenons are related to each other
especially in the context of the safety of the battery systems.
The usage of model-based investigations promises a theoretical understanding of the battery physics even beyond the
point what is possible in experiment. Due to their general structure the behavior of the batteries is strongly affected by the
interaction of physics on varying length and time scales. Many attempts have been made in the last 25 years beginning
with electrochemical considerations in the work of Newman [DO93, FU94, DO96]. This model is based on the theory
of porous electrode theory [NE75] describes the diffusion dynamics and charge transfer kinetics of a battery and is able
to predict the electric response of a cell in an intercalation electrode system. This model is adequate in small battery
systems. In large formatted battery cells however the uniformity of the electrical potential along the current collectors in
the cell composites are not longer given as well as the temperature distribution must also be taken into account due to its
inhomogeneity along the cell geometry.
Therefore the electrochemical model must be extended with thermal and multi-scale contributions. First thermal models
for Lithium batteries are given by Newman and Pals [NE95, PA95-1, PA95-2] using a heat equation with the heat genera-
tion as inhomogeneity, where the coupling of the thermal and electrochemical model is realized with the heat generation
terms. This idea is used for single battery cells as well as for whole battery packs [HA99, CH05, IN07, KU08, ZH13,
HU14]. Parallel the multi-scale and multi-domain character of the electrochemical model is developed [WA98-1, WA98-2]
by using the mathematical theory of homogenization resp. the spatial averaging theorem [HO85]. Similar newer attempts
can be found for example in [LA11, RI12] and [LA13].
First coupling between the multi-scale, multi-domain electrochemical-thermal model is introduced by [GU00, WA02]
and are under development until now. A general short overview can be found in [CA11] and actual some results in

1 Typesetting of this document by LATEX

2 The multi-scale multi-domain electrochemical-thermal model

[KI11, GU13, LE13]. The purpose of this modeling ansatzes are to gain a better understanding of the behavior of large
LIB systems because the transport of electrical current and heat must be considered on the length scales of the electrode
particles up to the length scale of the cell composite which is in general an anisotropic medium. One central point in the
modeling of LIBs are the aspects of cell safety which correspond to the thermal stability of a LIB. Several exothermic
reactions can occur inside a cell during operation as its inner temperature is increasing if the heat creation is bigger than
the dissipated heat in the surrounding space. This may cause heat to accumulate inside the cell and the chemical reactions
will be accelerated which yields in a further temperature increase until a thermal runaway is reached.
This phenomenon is not captured with the common mathematical models describing the dynamics of a LIB. A first at-
tempt to overcome this disadvantage is given in [HA01]. The heat equation is coupled with ordinary differential equations
(ODEs) describing the temporal evolution of the concentration of the exothermic reactions based on an Arrhenius-type
law. In [KI07, PE14] this model is extended with additional contributions. But until now no framework is given describing
the abuse of LIB in a general way.
In the next section 2 we give an short overview of the multi-scale, multi-domain electrochemical-thermal modeling of
LIBs [KI11, CA11]. In section 3 the corresponding partial differential equations (PDEs) with respect to the cell geometry
are discussed. These equations are extended with additional equations coming from mathematical combustion theory
[BE89, VO14] in section 4 to describe abuse behavior and thermal runaway. In the section 5 some simulation result from
Comsol Multiphysics are shown. In the final section an outlook of possible future work is given.

2 The multi-scale multi-domain electrochemical-thermal model

Following [KI11, LE13] a survey of the multi-scale multi-domain (MSMD) electrochemical-thermal model depending on
the geometry of a single cell is given in this section. The corresponding partial differential equations are depending on
the geometry of the LIB. The most common used single cells consist of cylindrical geometry and of rectangular geometry
for pouch cells. We consider the different models on the different length scales, derive coupling conditions between the
models with respect to the electrochemical nature of the problem for a single cell. Finally the electrochemical model is
coupled with the thermal model.
The MSMD model possesses a hierarchical domain structure, consisting of several model parts on his own length scale
for every subdomain. Variables defined on a lower hierarchical domain have a finer spatial resolution than variables on
higher hierarchical domain levels. Each domain level is connected with the next finer and next coarser domain level due
to coupling conditions.
This approach provides a modular framework which is well suited for multiphysics investigations arising in the research of
LIB. It can be applied for single cells as well as for whole battery stacks and packs with rising computational complexity.
In the sequel the considerations are restricted to a single cell, but the extending to whole battery packs and stacks is
straightforward.

2.1 The multi-scale multi-domain model for Lithium-ion batteries

The modeling of the dynamic phenomena of LIBs with the MSMD approach is an extension of existing simpler models
from several authors [DO93, FU94, NE95, HA99, WA02, CH05, IN07, KU08]. The general scheme of the approach can
be seen in Figure 1. The three domains of interest and their inclusion in the LIB are identified as the particle domain (a),
electrode domain (b) and (c) cell domain. The coordinate system in every domain are associated to the corresponding
length scale in each domain. Charge transfer kinetics have to be solved on the electrode-electrolyte interface. The transport
of the Li-ions is modeled with a diffusion mechanism and the migration and diffusion of the Li-ions through the liquid
electrolyte can be evaluated. The charge balances in the solid cathode and anode as well as the liquid electrolyte are
resolved in the corresponding matrices too. The temperature is solved over the whole cell geometry. The model is able to
predict the electrical and thermal behavior of a LIB cell for given electrical load with respect to given thermal boundary
conditions.

2.2 The particle domain model

In LIB design the electrode material that include the Lithium ions are particulates of given size and shape. The Lithium
transport in solid phase is the important physical process that governs the LIB performance. This process depends on the
particle geometry i.e. shape, size and on the thermodynamic and kinetic properties of the solid phase materials. Using an
own coordinate system for the solid state diffusion separate this process from the other variables. In the MSMD model the
process of solid-phase Lithium diffusion and the charge transfer kinetics are solved in the particle domain submodel. In
Figure 2(a) and Table 1(a) the state variables, input and output variables are shown. The input of the particle domain are
the state variables of the electrode domain submodel, in detail the liquid phase concentration ce , the electrical potential
at positive and negative electrode φs,{c,a} , Temperature T and potential in liquid phase φe . These variables are treated

2
2 The multi-scale multi-domain electrochemical-thermal model

(c) Cell Domain

(b) Electrode Domain
(a) Particle Domain

ξ2

x2

ξ1 X2
ξ3
x1
x3
X1
X3

Figure 1: Scheme of the multi-scale multi-domain approach.

as averaged lumped values in the particle domain. Since the transport in the liquid phase is much faster than in the solid
phase this approach in reasonable. The state space variables of the particle domain are the concentration of Lithium cs and
the volumetric heat source in the solid electrode particle qξ000 . Further variables are the transfer current density at particle
surface i00ξ and surface heat flux at electrode-electrolyte interface qξ00 . Output variables to the electrode domain model are
spatial averaged volumetric heat source in the solid electrode particle q̄ξ000 , the surface heat flux at the electrode-electrolyte
interface q̄ξ00 and the transfer current density ī00ξ .

2.3 The electrode domain model

The structure of the electrode is a porous one, to form a structure possessing a good electrical network and mechanical
properties due to the solid matrix. The active particles are mixed with a conductive agent and a polymer binder and
coated on a thin metal current collector. In a conventional battery the cathode and anode are separated with a thin porous
polymeric layer called separator to prevent an electronic short between the two electrodes. The transport of the Lithium
ions through the liquid electrolyte is strongly influenced by the pore structure of the composite electrodes and separator
and the wetting of the pore surfaces. The charge balance across the composite electrodes, their corresponding matrices,
the liquid electrolyte as well as the Lithium-ion transport in the liquid electrolyte is resolved in the electrode subdomain
model. In Figure 2(b) and Table 1(b) the state, input and output variables are shown. The static electric potential φ± along
the current collectors and the temperature T are the input from the cell domain submodel and are again treated as lumped
averaged values in this submodel. This will be justified by the argument that lumped values of the electrode potential at
the interface with the current collectors are without spatial dependence in the electrode domain coordinate system. This is
reasonable because the current collectors on the boundary with positive and negative electrode have an electrically higher
conductance than the composite electrodes. The state variables in the electrode subdomain model are the concentration of
Lithium ions in the liquid electrolyte ce , the electric potential in the positive, negative solid matrices φs,{c,a} and electric
potential in the electrolyte φe . Additional variables are the volumetric reaction current at the electrode composite volume
jx000 , current density at the current collector interface i00x , the volumetric heat generation in electrode domain qx000 including
Joule heat due to the current flow in the solid composite electrodes and electrolyte and the surface heat generation qx00 due
to the contact resistance between electrode and metal current collector surface.

3
2 The multi-scale multi-domain electrochemical-thermal model

(a) Particle Domain (b) Electrode Domain (c) Cell Domain

i00ξ (ξ) cs (ξ) ī00ξ , q̄ξ00 , ī00x , q̄x00 , φ+ (X), φ− (X)
qξ00 (ξ) qξ000 (ξ) q̄ξ000 q̄x000 T (X)
000
i00x (x) φs,c (x) jX (X)
000
qx00 (x) φs,a (x) qX (X)
ξ2 x2 ce (x) X2
φs,a , φs,c , φe (x)
ξ1 φe , c e , T x1 φ+ , φ − ,
ξ3 x3 jx000 (x) X1
T X3
qx000 (x)

Figure 2: Scheme of model variables in each domain and corresponding coupling.

(a) State Variables of particle domain Symbol Input to Symbol

Lithium concentration within solid electrode particle cs
Transfer current density at particle surface i00ξ Electrode domain ī00ξ
Surface heat flux at electrode-electrolyte interface qξ00 Electrode domain q̄ξ00
Volumetric heat source within solid electrode particle qξ000 Electrode domain q̄ξ000

(b) State Variables of electrode domain Symbol Input to Symbol

Lithium concentration in electrolyte ce Particle domain ce
Potential in liquid phase φe Particle domain φe
Potential at positive/negative electrode φs,{c,a} Particle domain φs,{c,a}
Current density at electrode boundary i00x Cell domain ī00x
Surface heat source in electrode domain qx00 Cell domain q̄x00
Volumetric current source at electrode domain jx000 - -
Volumetric heat source in electrode domain qx000 Cell domain q̄x000

(c) State Variables of cell domain Symbol Input to Symbol

Electric potential at positive/negative current collector Φ± Electrode domain φ±
Temperature T Particle/Electrode domain T
000
Volumetric current source in cell domain jX - -
000
Volumetric heat source in cell domain qX - -

Table 1: Variables used in MSMD model.

2.4 The cell domain model

To build prismatic or cylindrical LIBs, unit assemblies of the electrode layers are stacked or wound. The cell domain model
describes the battery cell as a complete device. This model is able to resolve the spatial inhomogeneities in temperature
and potential with respect to a given current load and ambient temperatures conditions, and has a big influence on the
overall performance of LIBs. In Figure 2(c) and Table 1(c) the corresponding state, input and output variables are shown.
The spatial distribution of the temperature T and the electric potential φ± at the current collectors must be resolved with
respect to the battery geometry and suitable initial and boundary conditions. The cell domain model is a macroscopic
approach to describe the electrical and thermal transport properties of a LIB. Furthermore the volumetric current source
000 000
jX and the volumetric heat generation qX is described in this model depending on corresponding variable from the lower
domain models.

2.5 Domain coupling

In the MSMD approach the underlying multiphysics model for temperature T , electric potential φ and concentration c of
Lithium ions is solved in the corresponding subdomain simultaneously. This needs a communication between the different
subdomain levels as in Figure 2 described. It is a two-way inter-domain coupling. The information flow from a higher
domain level to a lower domain level is realized in terms of average lumped variables as input for the lower level model.

4
2 The multi-scale multi-domain electrochemical-thermal model

The flow of information from a lower domain level to higher one is realized due to source terms which will be averaged in
a suitable way using the spatial average theorem [HO85]. This approach eliminates the dependence on coordinate system
of lower domain level before passed to the higher domain level.
In the sequel ξ, x, X are the coordinate systems of the particle, electrode and cell domain respectively and ξ = (ξ1 , ξ2 , ξ3 )T ,
x = (x1 , x2 , x3 )T , X = (X1 , X2 , X3 )T denotes the corresponding cartesian coordinates. With X = (r, θ, z)T cylindri-
cal coordinates are denoted for cylindrical cell.
Starting from the bottom up on particle domain level the transfer current density is averaged over the electrode-electrolyte
interface Aξ
Z
i00ξ (ξ, x, X)dAξ
Aξ
ī00ξ (x, X) = , (2.1)
Aξ
and delivered to the electrode domain level. On this level this variable will be transferred into a volumetric transfer
reaction current at the electrode composite volume using

jx000 (x, X) = ī00ξ (x, X)as,x , (2.2)

with the specific area of active interface between electrode particles and the electrolyte in the electrode volume as,x .
On the electrode domain level the current density of electrode boundary is averaged in the same way as in equation 2.1
over the area of interface between current collector and electrode domain Ax
Z
i00x (x, X)dAx
ī00x (X) = Ax , (2.3)
Āx
where Āx is the electrode plate area. This averaged value is delivered to the cell domain level and transferred in a
volumetric current source at the current collector
000
jX (X) = ī00x (X)as,X , (2.4)
where as,X is the specific electrode plate area in the cell volume. In a cylindrical geometry the volumetric current
sources on cell domain level are given as
000
j± (X) = ± ((ī00x (X)as,X )iep + (ī00x (X)as,X )oep ) , (2.5)
where as,X is again the specific electrode plate area in the cell volume and iep, oep denotes the inner and outer electrode
pair in the spiral wound inner geometry of the cylindrical cell. For more details we refer to [LE13].
The coupling of the heat sources is done in a similar way and occur on all length scales. The surface heat sources at the
electrolyte-electrode interface and the heat source within a volumetric particles are averaged with respect to the particle
domain boundary and delivered to the electrode domain submodel

Z
qξ00 (ξ, x, X)dAξ
Aξ
q̄ξ00 (x, X) = , (2.6)
Aξ
Z
qξ000 (ξ, x, X)dVξ
Vξ
q̄ξ000 (x, X) = , (2.7)
Vξ

where Aξ , Vξ are the electrode-electrolyte interface area and the volume of particle domain geometry respectively.
In the electrode domain a volumetric heat source originate from the averaged heat sources of particle domain
000
qx,ξ (x, X) = q̄ξ00 (x, X)as,x + q̄ξ000 (x, X)s , (2.8)
where as,x is the specific area of active interface between electrode particle and electrolyte and s the volume fraction
of active particles in electrode volume. The volumetric heat source from equation 2.8 is incooperated with other heat
contributions on this domain level like Joule heat due to electronic current flow in the electrode matrix and ionic current
flow in the electrolyte.
X
qx000 (x, X) = qx,ξ
000
(x, X) + 000
qx,k (x, X). (2.9)
k

5
3 The electrochemical and thermal partial differential equations for rectangular and cylindrical geometry

(a) Particle Domain (b) Electrode Domain (c) Cell Domain

x X3
r X1
X2

Rs
la ls lc

1D spherical particle model 1D porous electrode model 3D continuum model

Figure 3: The submodel in each domain for a prismatic cell.

In electrode domain level surface and volumetric heat sources are averaged over the corresponding geometries and
delivered to cell domain

Z
qx00 (x, X)dAx
Ax
q̄x00 (X) = , (2.10)
Āx
Z
qx000 (x, X)dVx
Vx
q̄x000 (X) = , (2.11)
Vx
where Vx the volume of electrode domain geometry. In the cell domain level equations 2.10, 2.11 are used to form the
heat source
000
q̄X,x (X) = q̄x00 (X)as,X + q̄x000 (X)asc , (2.12)
where as,X is the specific electrode plate area, asc is the volume fraction occupied by electrode and separator in the
cell volume. Finally the volumetric heat source in the cell domain is given by
X
000 000 000
qX (X) = qX,x (X) + qX,k (X), (2.13)
k

where several other heat sources like electric heating at current collector and other passive conductors are also taking
into account.

3 The electrochemical and thermal partial differential equations for

rectangular and cylindrical geometry
The macroscopic design of LIB features the thermal and electric behavior concerning the overall cell performance with
respect to the geometry, the underlying electrochemistry, thickness of current collector foils, particle size in the elec-
trode and many other parameters. To gain a better knowledge and understanding of temperature and potential distribution
analytical models like the MSMD model and their numerical implementation are of vital importance. In this section
we consider the governing partial differential equations, algebraic equations and the corresponding initial, transition and
boundary conditions with respect to cylindrical and rectangular geometry. In general on every domain level a coupled
system of elliptic generalized Poisson equations for the electrical potentials and parabolic diffusion equations of cor-
responding concentrations of Lithium ions must be solved [DO93, FU94]. On macroscopic scale these equations are
coupled with a heat equation, which represent also an parabolic differential equation [NE95, HA99, CH05, IN07, KU08].
For some newer publications we refer to [KI11, LE13]. The modeling in the sequel is implemented in a similar way
in Comsol Multiphysics in the Batteries and Fuel Cells Module. The governing equation and boundary conditions for
prismatic and spiral wound cylindrical geometry can be found in Table 2.

6
3 The electrochemical and thermal partial differential equations for rectangular and cylindrical geometry

3.1 The particle domain submodel

In the particle domain the geometrical shape of the particles is assumed to be perfect spheres with a radius r = Rs (Figure
3(a)). This leads to a 1D spherical particle model. The solution variables of this model are the concentration cs of the
Lithium ions inside the particle and the exchange current density i00ξ on the surface of the particle. Following Fick’s law
of diffusion one has to solve a radial-symmetric diffusion equation for the concentration cs in the spherical particle. The
transfer current density is governed by a Butler-Volmer kinetic expression, where the electrochemical reaction rate is
driven by the overpotential η. The overpotential is defined as the deviation of the potential difference between solid and
liquid phase φs , φe and the thermodynamic equilibrium potential Ueq . This two state variables are coupled via the flux
condition for the concentration on the surface of the particle. The diffusion coefficient Ds is assumed to be constant with
respect to the radial coordinate r. The temperature T is a given lumped value in particle domain as well as the electrolyte
concentration ce , solid and liquid phase potential φs , φe . The heat source at the surface of a particle is modeled by
considering heat from the charge transfer reaction and ohmic losses.

3.2 The electrode domain submodel

On the electrode domain level a simplified porous electrode model in cartesian coordinates is used similar to the 1D
spherical particle on particle domain level by assuming lumped potentials and liquid concentrations. It resolves the
charge balance in the liquid and solid phase and Lithium ions will be conserved in the electrolyte [SA06, KI09]. One has
to take into account three volumes for positive, negative electrode and separator as shown in Figure 3(b). In these volumes
the solid and liquid phase are assumed as superimposed continua. The Lithium concentration ce and the potential φe are
evaluated in the liquid phase in these three volumes. The potential for the anode φs,a and cathode φs,c are solved in the
solid phase of cathode and anode matrix. φ− , φ+ , T are provided by the cell domain model and are treated as spatial
invariants. The electronic current density at electrode boundary is equal to the average electrode plate current density i.e.
ī00x = i00x while jx000 is coming from equation 2.2. The averaged heat sources from the particle domain are converted in a
000
volumetric heat source qr,x following equation 2.8. Furthermore the Joule heat qx,Ω for the electric current flow in the
presence of potential gradients in solid and liquid has also being taking into account. These last two quantities are needed
for the volumetric heat source q̄x000 for passing to the cell domain model. The assumption of statistical homogeneity is also
valid in electrode domain.

3.3 The cell domain submodel

On the cell domain level a model was developed to solve for the temperature T and the pair of electrical potentials φ±
in the three dimensional cell geometry. Such a model is called Single-Potential-Pair Continuum (SPPC) Model. Using
the statistical homogeneity assumption the cell volume is assumed to be a continuum having orthotropic thermal and
electrical conductivities. The layered structure inside the cell is not necessarily resolved for numerical computations. In
typical battery designs the conduction transport in the cell composite is anisotropic and one has to distinguish between
in-plane and transverse diffusivity. Therefore one has to introduce a conductivity tensor σ ef f from which the effective
material properties can be computed. Depending on the cell geometry more information can be found in [KI11, LE13].
The effective material properties depends also on the volume fractions ± of negative and positive current collector and the
volume fraction of anode, separator and cathode asc in the cell volume. The boundary conditions for the electric potential
000
are governed via the electric load of the cell. The volumetric electric charge source jX is computed from equation 2.4.
000
The energy conservation yield a inhomogeneous heat equation, where the volumetric heat source qX on cell domain level
includes the heat from electrode domain submodel and Joule heat at all passive components carrying electric current in
000
the cell domain. The cell output current Io is then the volumetric average of qX .

(a) Particle domain

Variables Governing equations Boundary conditions/Notations

 
∂cs Ds ∂ 2 ∂cs ∂cs
cs (r, x, X) = 2 r =0
∂t r ∂r ∂r ∂r r=0
α ∂cs 1 00
i00ξ (r = Rs , x, X) i00ξ = ki (ce )αa (cs,max − cs,e ) a (ce,s )αc =− i
     ∂r r=Rs Ds F ξ
αa F αc F
· exp η − exp η
 RT RT
∂U eq
qξ00 (r = Rs , x, X) qξ00 = i00ξ φs − φe − Ueq − T
∂T
η((r, x, X) η = φs − φe − i00ξ R − Ueq

7
3 The electrochemical and thermal partial differential equations for rectangular and cylindrical geometry

(b) Electrode domain

1 − t0+ 000 i00e ∂t0+
 
∂e ce ∂ ∂ce ∂ce ∂ce
ce (x, X) = Deef f + jx − = 0, =0
∂t ∂x  ∂x F  F ∂x ∂x x=0 ∂x x=l
∂ ef f ∂φe ∂ ef f ∂ ∂φe ∂φe
φe (x, X) κ + κD ln ce + jx000 = 0 = 0, =0
∂x  ∂x ∂x
 ∂x ∂x x=0 ∂x x=l
∂ ef f ∂φs,{a,c} ∂φs,a
φs,{a,c} (x, X) σ{a,c} = jx000 φs,a = Φ− , =0
∂x ∂x x=0 ∂x x=la
∂φs,c
= 0, φs,a = Φ+ ,
∂x x=la +ls x=l
 Z l
ef f ∂φs,c
−σc

 =− jx000 dx l = la + ls + lc
00 00 ∂x x=l
ī00x (x, X) īx = ix = Z lal a +l s 
∂ ∂ ∂
T
∂φ ∇x = , ,
−σaef f s,a jx000 dx

 = ∂x1 ∂x2 ∂x3
∂x x=0 0
X ef f 3
X
000
qx,Ω (x, X) 000
qx,Ω = σi k∇x φs,i k2 + κef f k∇x φe k22 k • k2 = •2i
i∈{a,c} i=1

+κef
D ∇
f
ln ce ∇x φe
R l  000 x
000
0
qξ,x + qx,Ω dx
q̄x000 (x, X) q̄x000 =
l
(c) Cell domain (Cartesian coordinates)
(
000 1 if i = j
φ± (X) ∇X̃ · (± σ± ∇X̃ φ± ) = jX δij =
0 else
X 3
et (X) = eti îi
i=1
∂ρcp T 000
T (X) = ∇X · (k∇X T ) + qX kij = (kt − kp )eti etj + kp δij
∂t  T
∂ ∂
∇X̃ = , î: Basis in X space
∂X1 ∂X2
 T
∂ ∂ ∂
∇X = , , kt,p transversal/planar componenent
∂X1 ∂X2 ∂X3
ef f
X σi,± ef f
000
qX (X) 000
qX 000
= qx,X 000
+ qΩ,X 000
= q̄X asc + k∇X Φi k2 σ±,ij = δij − eti etj ± σ±
i σi
Z i∈{±}

I0 I0 = jx000 dVX
VX

(d) Cell domain(Cylindrical coordinates)

± σ ± ∂ 2 φ ± ∂ 2 φ± 000 l±
φ± (r, θ, z) 2 +  ± σ ± = j± r± = r + ± (ls + la )
r± ∂θ2 ∂z 2 2
kp ∂ 2 T ∂2T ± σ ± ∂ 2 φ ∂2φ
     
∂ρcp T 1 ∂ ∂T 000
T (r, θ, z) = kt + 2 + k p + qX 2 +  σ
± ±
∂t r ∂r ∂r r ∂θ2 ∂z 2 r± ∂θ2 ∂z 2
Θ1 =0,π
± σ± ∂ 2 φ ∂2φ
 
000 000
qX (r, θ, z) qX = qx000 + qΩ
000
= 2 + ± σ ± 2
r± ∂θ2 ∂z
Θ2 =π,0
qx000 = (q̄x000 iep )ep + (q̄ 000 ep )oep
 2  2 !
000
X
000
X 1 ∂φi ∂φi
qΩ = qi = i σi +
ri ∂θ ∂z
i∈{±} i∈{±}

Table 2: Governing equations for 3D model in cartesian and cylindrical coordinates.

8
4 Modeling thermal abuse and thermal runaway

Internal energizing events Thermal runaway

External abuse conditions and exothermric reactions

External heating Heating rate

Over-charging > Leak
Over-discharging Electrode-electrolyte reactions Dissipation rate Smoke
High current charging Decompostions Gas venting
Nail penetration Electrochemical Reactions Flames
Crush Rapid disassembly
External short

Figure 4: Thermal abuse for LIBs.

4 Modeling thermal abuse and thermal runaway

The model introduced in the last sections is able to resolve the normal use of Lithium-ion batteries. It is not able to
resolve phenomena resulting in an abusive usage of LIBs. The worst case scenario in abusive usage of LIBs is the thermal
runaway which leads to a complete destruction of the battery. In this chapter abuse mechanisms are investigated and a
first mathematical model will be derived for the contributions coming from abuse mechanisms. These mechanisms will
be described with an system of reaction-diffusion equations coming from mathematical combustion theory [BE89, VO14]
and first abuse simulations are performed in Comsol Multiphysics.

4.1 Thermal abuse mechanisms

The thermal stability of LIBs is a very important issue for cell safety. Several exothermic chemical reactions can occur
inside a battery as the temperature rise. This may generate heat that accumulates inside the cell and accelerates the
chemical reaction between the cell components, if the heat transfer to the surroundings is not sufficient [GU10], i.e. if the
heat generation rate exceeds the dissipation rate. In Figure 4 some abuse mechanisms are shown and the corresponding
chemical answer. Several authors have given models to describe abuse behavior and thermal runaway. First attempts
to describe the exothermic reaction kinetics are given by [HA01]. A simplified reaction-diffusion model is given in the
works of [KI07, PE14, PE14a]. The case of nail penetration is given in [CH14]. A catastrophe theoretic approach is given
in [WA10] which based on a simplified ODE model and is extended in [WA12]. To describe the abuse behavior resp. the
thermal runaway one has to identify the main exothermic chemical reactions. Following [LI11] the general mechanism
that lead to a thermal runaway can be described with respect to rising temperature in four steps which are given in Table
3.

Temperature Reaction

T > 90◦ C: Start of anodic reaction Rate limiting step

T > 120◦ C: Decomposition of solid electrolyte interface (SEI) Reduction of electrolyte at negative electrode
T > 140◦ C: Exothermic reaction starts at positive electrode Oxygen rapidly evolves
◦
T > 180 C: Positive electrode decompose High rate exothermic process
◦
C
Electrolyte decomposition due to oxidation Temperature rise > 100
min
Table 3: Abuse mechanisms depending on cell temperature.

4.2 The reaction-diffusion model

Since the abuse of LIBs could lead to thermal runaway, this can result in fire and explosion of the battery. The mathemat-
ical theory of choice for a proper modeling of this phenomena is the combustion theory[BE89] resp. a solid fuel model
which represents a description in form of reaction diffusion equations for the temperature T and the n concentrations

9
4 Modeling thermal abuse and thermal runaway

c = (c1 , . . . , cn )T of the chemical species on the corresponding domains in the battery[VO14]. Under the assumption
that the density ρ and the heat capacity cp are constant one yields

n
∂T X
ρcp = ∇ · (κ∇T ) + Qi , (4.1)
∂t i=1
∂c
= ∇ · (D∇c) + ΓR, (4.2)
∂t
where κ is the thermal conductivity, Qi = qi Ri , qi, Ri , i = 1, . . . , n are the heat sources, the heat release and reaction
rate respectively. D is the diffusion matrix, which represents a diagonal matrix since the chemical reactions considered
here are independent from each other. Γ is the matrix of the stoichiometric coefficients which is also a diagonal matrix.
Since we consider exothermic chemical reactions in the abuse of LIBs the reaction rates are governed by an Arrhenius
law, which have the general form
 
mi Ea,i
Ri = Ai ci exp − , i = 1, . . . , n, (4.3)
RT
where Ai is a frequency factor, ci the concentration, mi the reaction order, Ea,i the activation energy and R the universal
gas constant. Following [WA10] the different contribution can be found in Table 4.

Temperature Reaction

T > 90 − 120◦ C: SEI decomposition reaction: qsei = Hsei Wc

 
Ea,sei msei
SEI layer is metastable Rsei (T, csei ) = Asei exp − csei
RT
∂csei
Decompose exothermic = ∇ · (dsei ∇csei ) − Rsei
∂t
T > 120◦ C: Negative solvent reaction: qne = Hne Wc
 
Ea,ne mne
Exothermic reaction between Rne (T, cne ) = Ane exp − cne
RT
∂cne
intercalated Lithium and electrolyte = ∇ · (dne ∇cne ) − Rne
∂t
T > 140◦ C: Positive solvent reaction qpe = Hpe Wp
 
Ea,pe mpe
Exothermic reaction between Rpe (T, cpe ) = Ape exp − cpe
RT
∂cpe
positive cathode material and electrolyte = ∇ · (dpe ∇cpe ) − Rpe
∂t
T > 180◦ C: Electrolyte decomposition reaction: q e = He W e
 
Ea,e me
Exothermic reaction Re (T, ce ) = Ae exp − ce
RT
∂ce
= ∇ · (de ∇ce ) − Re
∂t
Table 4: Abuse mechanisms depending on cell temperature.

In the last Table qx , x ∈ {sei, ne, pe, e} represent the volumetric reaction heat, i.e. Hx , x ∈ {sei, ne, pe, e} the
exothermic reaction heat and Wx , x ∈ {c, p, e} a specific volume content.
In [HA01, KI07, PE14, PE14a, CH14] anode-electrolyte and cathode-electrolyte are modeled in a sightly different way.
Additional anode and cathode decomposition reactions can also be taken into account [WA10]. Additional heat sources
are heat due to entropy change Qs , Ohmic and Joule heat Qp resp. Qj [WA10, WA12]. This gives the additional heat
source
 
dUeq
Qadd = I Ueq + T , (4.4)
dT
where I is the current and Ueq the equilibrium voltage of the cell. The heat exchange between the battery cell and the
environment is modeled due to heat convection and radiation.

10
5 Simulations

Figure 5: Applied current with an amplitude |Imax | = 8 A.

Qconv = hA (Tsurf − Tamb ) , (4.5)

4 4


Qrad = σA Tsurf − Tamb , (4.6)

where h is the convection heat transfer coefficient, A the area of the battery case,  the emissivity of cell surface and σ
Stefan-Boltzmann constant.

5 Simulations
In this section we present some first simulation results done in Comsol Multiphysics. We consider a cylindrical 18650
LMO cell. The internal spiral wounds is not resolved in this simulation model as well as radiation boundary conditions.
The other physical parameters can be found in Table 5 and 6. The main parameters have been taken from a Comsol
tutorial about Li-ion batteries and the parameter for the abuse contributions are taken from [KI07, PE14]. The simulations
in Comsol where performed with two models, in the first model the conventional model from the sections 2 and 3 is
considered, while for the second model we used additional exothermic terms for the electrolyte and the reaction at positive
and negative electrode. The environmental temperature is chosen as Tenv = 343.15 K. In the simulation the heat transfer
coefficient h is varied, i.e. h ∈ [−20, 4] W/(m2 · K) and the mean temperature is evaluated. The important boundary
condition in the simulation is

n · (κ ∇T ) = h (Tenv − T ) , (5.1)
which describes the inward flux condition. For vanishing h we have adiabatic boundary conditions, positive h corre-
spond to isoperibolic boundary conditions where heat flow to the environment occur, while for negative h a heat flow
from the enviroment in the battery cell occur. The simulations were performed with the current load profile as seen in
Figure 5, between the charging and discharging pulses with duration of 250 s we have insert relaxation periods of 250 s.
A complete current cycle lasts 1000 s and maximal 10 cycles, tsim = 104 s are simulated.

Parameter Value Parameter Value

2
Neg. solid phase Li-diffusivity Ds,neg = 3.9 · 10−14 ms Neg. particle radius rp,neg = 12.5 · 10−6 m
2
Pos. solid phase Li-diffusivity Ds,pos = 1 · 10−13 ms Pos. particle radius rp,pos = 8 · 10−6 m

Pos. solid phase vol.-fraction s,pos = 0.297 Neg. solid phase vol.-fraction s,neg = 0.471
Pos. Electrolyte vol.-fraction e,pos = 0.444 Neg. electrolyte vol.-fraction e,neg = 0.357
mol m
Neg. max. solid phase concentration cs,max,neg = 26390 m3
Neg. reaction rate coefficient kneg = 2 · 10−11 s

11
5 Simulations

mol m
Pos. max. solid phase concentration c,smax,pos = 22860 m3
Pos. reaction rate coefficient kpos = 2 · 10−11 s
mol mol
Initial neg. State of Charge cs,neg,0 = 7917 m3
Initial pos. State of Charge cs,pos,0 = 16002 m3
mol
Initial electrolyte salt concentration cl,0 = 2000 m3
Initial temperature T0 = 343.15 K

Battery radius rbatt = 9 mm Battery height hbatt = 65 mm

Mandrel radius rmand = 2, mm Thickness battery canister dcan = 0.25 mm
Length pos./neg. electrode Lpos/neg = 55 µm Length separator Lsep = 30 µm
Thickness neg. current collector Lneg,cc = 7 µm Thickness pos. current collector Lpos,cc = 10 µm
Cell thickness Lbatt = 157 µm
W W
Pos. electrode thermal conductivity κT,pos = 1.58 m·K
Neg. electrode thermal conductivity κT,neg = 1.04 m·K
W W
Pos. current collector thermal cond. κT,cc,pos = 170 m·K Neg. current collector thermal cond. κT,cc,neg = 298.15 m·K
W
Separator collector thermal cond. κT,sep = 0.344 m·K
kg kg
Pos. electrode density ρpos = 2328.5 m3
Neg. electrode density ρneg = 1347.33 m3
kg kg
Pos. current collector density ρpos,cc = 2770 m 3 Neg. current collector density ρneg,cc = 8933 m3
kg
Separator density ρsep = 1008.98 m3
J J
Pos. electrode heat capacity cp,pos = 1269.21 kg·K
Pos. electrode heat capacity cp,neg = 1437.4 kg·K
J J
Pos. current collector heat capacity cp,cc,pos = 875 kg·K Pos. current collector cp,cc,neg = 385 kg·K
J
Separator heat capacity cp,sep = 1978.16 kg·K

Table 5: Simulation parameters.

The effective thermal conductivity (radial, angular direction), density and heat capacity are given as follows [KI07]:

κT,pos Lpos + κT,neg Lneg + κT,cc,pos Lpos,cc + κT,cc,neg Lneg,cc + κT,sep Lsep
κT,ang,batt = (5.2)
Lbatt
Lbatt
κT,rad,batt = (5.3)
Lpos Lneg Lpos,cc Lneg,cc Lsep
+ + + +
κT,pos κT,neg κT,cc,pos κT,cc,neg κT,sep
ρpos Lpos + ρneg Lneg + ρpos,cc Lpos,cc + ρneg,cc Lneg,cc + ρsep Lsep
ρbatt = (5.4)
Lbatt
cp,pos Lpos + cp,neg Lneg + cp,cc,pos Lpos,cc + cp,cc,neg Lneg,cc + cp,sep Lsep
Cp,batt = (5.5)
Lbatt
(5.6)

The parameter for the exothermic reaction are taken from [PE14]:

     
J 1 kg
x ∈ {pe, ne, e}, y ∈ {c, p, e} Reaction heat Hx Reaction frequency factor Ax Init value Volume content Wy
kg s m3
Negative solvent reaction 1.714 · 106 2.5 · 1013 0.75 1.39 · 103
Positive solvent reaction 3.14 · 105 6.667 · 1013 0.04 1.3 · 103
Electrolyte decomp. reaction 1.55 · 105 5.14 · 1025 1 5 · 102

Table 6: Exothermic simulation parameters.

h i
W
Heat transfer coefficient h m2 ·K
: −20 −15 −10 −5 −4 −3 −2 −1 0 1 2 3 4
Thermal runaway time tT R [s]: 1207 1454 1755 2386 2621 2767 3089 3489 3802 4707 5826 8669 −

Table 7: Thermal runaway time tT R with respect to heat transfer coefficient h.

12
5 Simulations

In Table 7 and Figure 6 the time tT R where a thermal runaway occurs is given with respect to the heat transfer coefficient
h. It shows that if the heat inside the battery can dissipate to the environment the time where thermal runaway occurs
increases. The corresponding temperature profiles shows that for h > 0 in the relaxation intervals, i.e. Iload = 0 the
temperature decreases while for h = 0 the temperature remains almost constant and for h < 0 the temperature increase
in this intervals due heat dissipate from the exterior. But within the simulation time there exists for the most considered
K
values of h a time where the temperature gradient with respect to the time becomes very high ≈ 90 − 100 min , if the
W W
exothermic contributions are present in the model. In Figure 7 for h = 0 m2 ·K and h = 2 m2 ·K the differences in the
model with and without exothermic terms are shown. The temperature in the model with exothermic contribution is
always slightly higher than without these terms and of the order of 10−2 . In the area where the thermal runaway is starting
this difference is growing very fast.

Figure 6: Thermal runaway time and corresponding thermal profiles.

Figure 7: Differences in the temperature profile with and without exothermic terms.

13
6 Conclusion and outlook

6 Conclusion and outlook

In this work we have introduced the MSMD model for the simulation of LIBs with respect to the important state variables
temperature T , the potential φ± and the corresponding concentration of the Lithium ions in the electrolyte ce and in the
particles cs,{a,c} of positive and negative electrode for a given current profile. We have taken into account the length
scales where these dynamics occurs, resulting on a coupled system of elliptic and parabolic partial differential equation
on each length scale. These equations are coupled with the equations on the other length scales via homogenization
procedures basing on the assumption of spatial homogeneity and the spatial averaging theorem. Furthermore we have
done a first step to describe the thermal runaway of LIBs due to extending the MSMD model with additional contributions
coming from various exothermic reaction inside the LIB during abuse and thermal runaway. In this simulation done in
Comsol Multiphysics we have shown that these additional exothermic terms in the numerical FEM model can resolve this
phenomena for a specific current load profile.
For future work this first model for describing the thermal runaway of LIBs will be improved. Detailed consideration
of the underlying battery chemistry and corresponding exothermic reaction kinetics are planned. Furthermore simplified
models for the abuse mechanisms will be considered with respect to the application in equivalent circuits models (ECM)
of LIBs.

Acknowledgement
This R&D project is partially funded by the Federal Ministry for Education and Research (BMBF) within the framework
“IKT 2020 Research for Innovations” under the grant 16N12515 and is supervised by the Project Management Agency
VDI | VDE | IT.

Special thanks to Anna Ossipova for creating the first Comsol model.

References [GU00] Gu, W.B., Wang, C.Y., Thermal-Electrochemical

Modeling of Battery Systems, J. Electrochem. Soc.,
[BE89] Bebernes J., Eberly D., Mathematical Problems 147 (8), 2910-2922, 2000.
from Combustion Theory, Applied Mathematical Sci-
ences Volume 83, Springer, 1989. [GU10] Guo,G., Long, B., Cheng, B., Zhou, S., Cao,.B.,
Three-dimensional thermal finite element modeling of
[CA11] Cai, L., White, R.E., Mathematical modeling of lithium-ion battery in thermal abuse application, J.
lithium ion battery with thermal effects in COMSOL Power Sources 195, 2393-2398, 2010.
Inc. Multiphysics (MP) software, J. Power Sources [GU13] Guo, M., Kim, G.H., White, R.E., A three-
196,5985-5989, 2011. dimensional multi-physics model for a Li-ion battery,
J. Power Sources, 240, 80-94, 2013.
[CH05] Chen, S.C., Wan, C.C., Wang, Y.Y., Thermal anal-
ysis of lithium-ion batteries, J. Power Sources 140, [HA99] Al Hallaj S., Maleki, H., Hong, J.S., Selman,
111-124, 2005. J.R., Thermal modeling and design considerations of
lithium-ion batteries, J. Power Sources 83, 1-8, 1999.
[CH14] Chiu, K.-Ch., Lin, C.-H., Yeh, S.-F., Lin, Y.-H.,
Chen, K.-C., An electrochemical modeling of lithium- [HA01] Hatchard, T.D., MacNeil, D.D., Basu, A., Dahn,
ion battery nail penetration, J. Power Sources, 251, J.R., Thermal Model of Cylindrical and Prismatic
254-263, 2014. Lithium-Ion Cells, J. Electrochem. Soc., 148 (7),
A755-A761, 2001.
[DO93] Doyle, M., Fuller, T.F, Newman J,. Model-
[HO85] Howes, F.A., Whitaker, S., The Spatial Averaging
ing Galvanostatic Charge and Discharge of the
Theorem Revisited, Chemical Engineering Science,
Lithium/Polymer/Insertion Cell, J. Electrochem. Soc.,
40, (8), 1387-1392, 1985.
140 (6),1526-1533, 1993.
[HU14] Hu, J., Tao, L., Jing, L., Temperature Field Anal-
[DO96] Doyle, M., Newman J,. Comparison of the Model- ysis and Thermal Dissipation Structure Optimization
ing Predictions with Experimental Data from Plastic of Lithium-ion Battery Pack in PEVs, Research Jour-
Lithium Ion Cells, J. Electrochem. Soc., 143 (6),1890- nal of Applied Sciences, Engineering and Technology,
1903, 1996 . 7 (1), 62-71, 2014.

[FU94] Fuller, T.F., Doyle, M., Newman, J., Simulation [IN07] Inui, Y., Kobayashi, Y., Watanabe, Y., Watase, Y.,
and Optimization of the Dual Lithium Ion Insertion Kitamura, Y., Simulation of temperature distribution
Cell, J. Electrochem. Soc., 141 (1),1-10, 1994. in cylindrical and prismatic lithium ion secondary

14
 References

batteries, Energy Conversion and Management, 48, [PA95-2] Pals, C.R., Newman, J., Thermal modeling of
2103-2109, 2007. Lithium/Polymer Battery, II. Temperature Profiles in a
Cell Stack, J. Electrochem. Soc., 142 (10),3282-3288,
[KI07] Kim, G.H., Pesaran, A., Spotnitz, R., A three- 1995
dimensional thermal abuse model for lithium-ion
cells, J. Power Sources, 170, 476-489, 2007. [PE14a] Peng, P., Sun, Y., Jiang, F., Numerical simulations
and thermal behavior analysis for oven thermal abus-
[KI09] Kim, U.S., Shin, C.B., Kim, C.S., Modeling for the ing of LiCoO2 lithium-ion battery, CIESC Journal,
scale-up of a lithium-ion polymer battery, J. Power 65, 2, 647-657,2014.
Sources 189,841-846, 2009.
[PE14] Peng, P., Sun, Y., Jiang, F., Thermal analyses of
[KI11] Kim, G.H., Smith, K., Lee, K.J., Santhanagopalan, LiCoO2 lithium-ion battery during oven tests, Heat
[Link], Multi-Domain Modeling of Lithium-Ion Mass Transfer, April,2014.
Batteries Encompassing Multi-Phyiscs in Varied
[RI12] Richardson, G., Denault, G., Please, C.P., Mul-
Lenght Scales, J. Electrochem. Soc., 158 (8), A955-
tiscale modeling and analysis of lithium-ion battery
A969, 2011.
charge and discharge, [Link]., 72, 41-72, 2012.
[KU08] Kumaresan, K., Sikha, G., White, R.E., Thermal [SA06] Santhanagopalan, S., Guo, G., Ramadass, P.,
Model for Lion-Ion Cell, J. Electrochem. Soc., 155 White, R.E., Review of models for predicting the cy-
(2),A164-A171, 2008. cling performance of lithium ion batteries, J. Power
Sources 156,620-628, 2006.
[LA11] Latz, A., Zausch, J., Iliev, O., Modeling of Species
and Charge Transport in Li-Ion Batteries Based on [VO14] Volpert, V., Elliptic Partial Differential Equa-
Non-equilibrium Thermodynamics, in Dimov, I., Di- tions, Volume 2: Reactions-Diffusion Equations,
mova, S., Kolkovska, N. (Eds.): NMA 2010, LNCS Birkhäuser, 2014.
6046, 329-337, Springer 2011.
[WA98-1] Wang, C.Y., Gu, W.B., Liaw, B.Y., Micro-
[LA13] Landstorfer, M., Jacob, T., Mathematical modeling Macroscopic Coupled Modeling of Batteries and Fuel
of intercalation batteries at the cell level and beyond, Cells. [Link] Development, J. Electrochem. Soc.,
Chem. Soc. Rev. 42, 3234-3252, 2013. 145 (10),3407-3417, 1998 .

[LE13] Lee, K.J., Smith, K., Pesaran, A., Kim, [WA98-2] Gu, W.B., Wang, C.Y., Liaw, B.Y., Micro-
G.H., Three dimensional thermal-, electrical, and Macroscopic Coupled Modeling of Batteries and
electrochemical-coupled model for cylindrical wound Fuel Cells. II. Application to Nickel Cadmium and
large format lithium-ion batteries, J. Power Sources, Nickel-Metal Hybride Cells, J. Electrochem. Soc., 145
241, 20-32, 2013. (10),3418-3427, 1998 .
[WA02] Wang, C.Y., Venkat Srinivasan, Computational
[LI11] Lisbona, D., Snee, T., A review of hazards as-
battery dynamics (CBD)-electrochemical/thermal
sociated with lithium and lithium-ion batteries, Pro-
coupled modeling and multi-scale modeling, J. Power
cess Safty and Environmental Protection, 89, 434-442,
Sources, 110, 364-376, 2002.
2011.
[WA10] Wang, Q., Ping, P., Sun, J., Catastrophe analysis
[NE75] Newman, J., Tiedemann, W., Porous-Electrode of cylindrical lithium ion battery, Nonlinear Dyn., 61,
Theory with Battery Applications, AIChE Journal, 21, 763-772, 2010.
(1), 25-41, 1975.
[WA12] Wang, Q., Ping, P., Zhao, X., Chu, G., Sun, J.,
[NE95] Newman, J., Tiedemann, W., Temperature Rise in Chen, C., Thermal runaway caused fire and explosion
a Battery Module with Constant Heat Generation, J. of lithium ion battery, J. Power Sources, 208, 210-224,
Electrochem. Soc., 142 (4),1054-1057, 1995 . 2012.

[PA95-1] Pals, C.R., Newman, J., Thermal modeling of [ZH13] Zhu, C., Li, X., Song, L., Xiang, L., Development
Lithium/Polymer Battery, I. Discharge Behavior of of the theoretically based thermal model for lithium
a Single Cell, J. Electrochem. Soc., 142 (10),3274- ion battery pack, J. Power Sources, 223, 155-164,
3281, 1995 . 2013.

15

View publication stats