You are on page 1of 11

Applied Energy 362 (2024) 123004

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Modeling the propagation of internal thermal runaway in


lithium-ion battery
Yue Zhang , Laifeng Song , Jiamin Tian , Wenxin Mei *, Lihua Jiang , Jinhua Sun ,
Qingsong Wang *
State Key Laboratory of Fire Science, University of Science and Technology of China, Hefei 230026, China

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• A thermal runaway propagation model


is built to investigate internal thermal Figure Illustration of LiFePO4 battery internal thermal runaway mechanism
propagation features.
• The triggering energy concept is pro­
posed to characterize the thermal safety
performance of batteries.
• The effects of battery configuration on
thermal runaway behavior are revealed.

A R T I C L E I N F O A B S T R A C T

Keywords: The trend toward high capacity and huge size in lithium-ion batteries has made it necessary to investigate the
Lithium-ion battery internal thermal characteristics. In this study, a thermal runaway model was developed to describe lithium-ion
Internal thermal runaway batteries’ internal thermal characteristics. Moreover, triggering energy was proposed as a critical feature for
Numerical modeling
evaluating and characterizing the thermal runaway under diverse thermal abuse situations, with large differ­
Triggering energy
Battery configuration
ences among characteristic temperatures. Finally, the effects of battery configuration on thermal runaway be­
haviors were investigated. The modeling results showed that internal temperature distribution can be divided
into four characteristic stages with two jelly rolls, and the application of more numerous and thinner cells inside
a battery can accelerate the propagation of thermal runaway. The experimental results showed that the ratio of
triggering energy of self-heat onset to total self-heat generation remained consistent in an adiabatic environment.
The mean value of the ratio was 24.5%, indicating that lithium iron phosphate batteries obtain most of the
energy (generally 80%) from internal exothermic reactions during adiabatic thermal abuse. The triggering en­
ergy of thermal runaway remained constant when various heating powers were applied to one of the batteries’

* Corresponding authors.
E-mail addresses: heart@ustc.edu.cn (W. Mei), pinew@ustc.edu.cn (Q. Wang).

https://doi.org/10.1016/j.apenergy.2024.123004
Received 26 November 2023; Received in revised form 5 March 2024; Accepted 8 March 2024
Available online 20 March 2024
0306-2619/© 2024 Elsevier Ltd. All rights reserved.
Y. Zhang et al. Applied Energy 362 (2024) 123004

laterals (about 20.8% of theoretical energy contained inside lithium iron phosphate batteries). Triggering energy
can provide new insights into the modeling of thermal runaway mechanisms and propagation.

1. Introduction 50% using the cell-to-pack (CTP) strategy compared to conventional


lithium iron phosphate(LFP) batteries. The new Tesla 4680 “tabless”
The need for lithium-ion batteries has been rising, with the spike in battery significantly reduces the ohmic resistance and heat [3], indi­
demand for commercial electronics products and electric vehicles. cating that we are still searching for better battery configurations and
Additionally, electrochemical energy storage systems have caused structure. One conventional battery design includes layering anode,
another sharp increasing demanding for lithium-ion batteries, which are separator, and cathode. Multiple layers of these materials can then be
designed with high energy density and long cycle life. Therefore, the rolled up and inserted in a cylinder or rolled up and flattened before
expansion of the global lithium-ion battery market has been anticipated being inserted in a corresponding housing. Multiple jelly rolls are often
to accelerate. However, self-accelerating heating features prevent inserted into one pouch or prismatic container. Currently, the number of
lithium-ion battery expansion at a quick rate in safety-emphasis fields state-of-the-art battery internal jelly rolls sometimes exceeds four
[1,2]. Lithium-ion battery performance is sensitive to certain factors in because the curvature of a jelly roll can be larger when preparing a
the operating environment, such as temperature [3]. Non-normal envi­ thicker jelly roll, and the manufacturing process is in high demand.
ronments can induce battery failure and even safety incidents, like However, safety concerns have not been well addressed and under­
thermal runaway (TR), fire, and even explosions. The state-of-the-art standing of the impacts of battery size on TR is limited [26]. Chen et al.
trend of multiple cells, large capacity, and high-level integrations of [27] investigated the effects of battery thickness on thermal runaway
lithium batteries will exacerbate incident consequences and also high­ temperature characteristics under different thermal abuse conditions
light the significance of the thermal runaway progress [4], especially in using a one-dimensional thermal resistance network model, and the
the case of lithium iron phosphate (LFP) batteries characterized by results demonstrated that surface temperature remained consistent
prolonged thermal runaway development. across various conditions tested. This work investigates the effect of cell
Most investigations emphasized thermal runaway propagation be­ number and battery thickness on battery safety performance, encom­
tween batteries [5–8] and even packs [9–11]. Understanding internal passing temperature characteristics as well as crucial time elements.
thermal runaway in real-world applications is becoming crucial for This work focuses on the thermal runaway characteristics inside the
ensuring the safety and reliability of lithium-ion batteries. Some pio­ battery and the uniformities between the LFP batteries under different
neers employed innovative and advanced experimental apparatus to scenarios. The first section focuses on the development and validation of
investigate internal thermal runaway in recent years. In order to better the thermal runaway model. This is followed by the proposal of a
understand internal temperature during thermal runaway, Ouyang’s method to characterize the thermal features of batteries through the
group [7,12–14] employed thermocouples implanted within the battery concept of triggering energy. The last section investigates the impact of
between jelly rolls to investigate in-cell temperature rather than the battery thickness and the number of cells on the thermal runaway
lateral surface. Fransson et al. [15] used high-speed X-ray imaging to characteristics.
capture for the first time an internal sidewall breach process caused by
nail penetration, as well as to analyze the speed of layers delamination. 2. Model description and validation
Due to the expensive and demanding experimental apparatus for
internal thermal runaway investigation, researchers usually employed 2.1. Model development
numerical methods and validated by conventional abuse tests. The
evolution of lithium-ion battery thermal runaway modeling has pro­ Simulation settings are illustrated in Fig. 1. To make it clear, the
gressed from simple conceptual frameworks to sophisticated multi- terms used in this work are clarified in Fig. 1(a). A cell, which is the most
physics models, incorporating detailed material properties and muiti- essential component of a battery, is a jelly roll that can operate relatively
physics considerations. In the early stages, advances in independently. Typically, one prismatic battery contains two jelly rolls,
electrochemical-thermal coupling emerged in the 2000s. Models started and multiple batteries are tabbed together to form a module. The details
integrating electrochemical reactions with heat generation, recognizing of boundary conditions in the thermal runaway model are visualized in
the complex interplay between chemical processes and temperature Fig. 1(b). The simplified orange rectangle symbolizes the heating device,
changes [16,17]. Zhang et al. [18] investigates the characteristics of while the light blue rectangle represents the battery. For the LFP battery,
thermal runaway front of long, large-format Li-ion batteries through cascading failure in the open space rarely catches fire, so conduction and
experiments and simulations. Special long pouch cells were used to convection are the principal heat transfer pathways. The thermal
facilitate the quantification of the internal thermal runaway front. Zhao runaway model is built based on the internal structure, as the computed
et al. [19] developed a coupled reaction-heat conduction model to tomography (CT) image reveals in Fig. 1(c). Because a thin metallic layer
describe battery thermal runaway propagation within a cell, with which can precede heat conduction and the battery’s perpendicular direction
propagation speed can be estimated in a typical condition. Kong et al. has low thermal conductivity, the geometric model refines the
[20–23] developed a comprehensive model that can potentially help aluminum shell and internal air domain between the jelly rolls and the
improve understanding of the TR mechanisms considering internal prismatic container, providing more precise and accurate solutions.
multiphase process of battery venting. Recent trends include the appli­ Assuming that the battery capacity and density are constant, the
cation of detailed reaction mechanisms to thermal runaway modeling three-dimensional thermal runaway model is based on the energy con­
[24]. However, few research focuses on the internal thermal runaway servation Eq. (1).
propagation behaviors and key uniformities of large-format batteries.
This paper focuses on understanding how energy is transferred within ∂T
ρC P = − ∇(k∇T) + q (1)
the cell during thermal runaway in response to thermal abuse. ∂t
Internal battery structure can also influence the development of where ρ is battery density. Cp is the specific heat capacity of the battery.
battery internal thermal runaway. To the best knowledge of the authors, T is the temperature of the battery. k is the heat conductivity of the
major participants in the worldwide lithium-ion battery industry are battery. q is the heat source inside the battery.
experimenting with new configurations. For example, BYD launched the Description of the heat source inside the battery is the most crucial
blade battery [25], and the space utilization of the battery pack is over part of thermal runaway model development. Researchers principally

2
Y. Zhang et al. Applied Energy 362 (2024) 123004

employ two approaches to describe the internal heat source. First and results, as well as literature with similar experimental settings [37,38],
foremost, the Arrhenius equation has been used to calculate the reaction all reached the same conclusions about this vertical temperature
rate from the material level kinetics to safety forcasts at the scale of cells gradient, which will be discussed further in Section 3.1.
in 2000 [17,28–33]. The utilization of cell-scale experiments, particu­
larly accelerating rate calorimeter (ARC) tests, enables the extraction of 3. Results and discussion
heat source and the characteristic temperature of thermal runaway
[7,13,34,35], in addition to standard reaction kinetics approaches. This 3.1. Internal thermal runaway propagation analysis
approach facilitates the description of the internal heat source pro­
gression over time. Accelerating rate calorimeter (ARC) test is one Temperature distribution in thermal abuse scenarios can be regarded
standard test that can provide an adiabatic environment for thermal as a mathematical transient propagation problem and a convection-
runaway experiments. In this work, a model is developed based on the diffusion problem in physics. Thermal diffusion disturbs the homoge­
comprehensive kinetic analysis from the material level to explore the neity of mixture composition and has a dissipative feature. It is helpful to
heat generation behavior [36].⋅Representative adiabatic and non- introduce the thermal diffusion constant α, which can be determined
adiabatic experiments were conducted to investigate the internal ther­ from the expression α = k/ρCP . The rate of heat propagation increases as
mal runaway of LFP batteries. the quantity α increases. Due to its lower density and higher heat ca­
The initial and boundary conditions for the lateral overheating abuse pacity, the air zone exhibits a greater thermal diffusion constant
scenarios are described as follows. compared to cells. As shown in Fig. 3, heat propagates faster in the air
zone near tabs at the beginning of the heating period, resulting in a
t = 0, T(x, t) = T0
higher temperature near the tab than on the bottom side. Moreover, the
x→Hbatt , T→T0 (2) internal cell-to-cell thermal runaway progress is depicted in the top view
∂T of the battery temperature distribution in Fig. 3(b). Parallel electrodes
x = 0, − λ = q0 have a temperature gradient due to the large thermal conductivity of the
∂t
in-plane and relatively small thermal conductivity of the cross-plane.
where the Hbatt is the thickness of the battery and q0 is the heat flux from However, the thermal runaway area is not strictly parallel to the elec­
the lateral overheating source. trode surface because of the different dissipation conditions of every
battery surface and quite different thermal conductivity of battery
components. Typically, thermal runaway begins in a localized region
2.2. Model validation near the heater where the area reaches the critical temperature. In other
words, the internal thermal runaway process can be characterized as the
Based on our previous experimental work [36], modeling validation occurrence of a localized thermal runaway, which subsequently spreads
is provided in Fig. 2, which delineates the temperature profile and shows throughout the entire cell due to the temperature gradient and internal
the accuracy in predicting the maximum temperature and process of TR exothermic reactions.
events by comparing simulation results with the lateral overheating To further explore the internal thermal runaway progress, temper­
experiments. The largest relatively error in estimating the TR onset ature at the centerline of the battery at a symmetrical section is depicted
temperature is 3.2%, while the maximum relatively error in predicting in Fig. 4. The x-axis of Fig. 4(b) is in the descending arrangement cor­
the TR onset time is 2.9%, which is within a reasonable error range. responding to the geometry model settings, as the origin of the left-
Therefore, the numerical modeling is adopted for further internal ther­ handed coordinate system is located on the battery side opposite the
mal runaway propagation analysis. heater. Internal thermal propagation can be divided into four featured
There is a temperature gradient across the vertical direction. Before stages. At the very beginning of the heating stage (stage I), the so-called
the thermal runaway, the temperature near the tabs is higher than that non-regular regime because this stage lasts very short, the temperature
of the bottom side, but this is reversed after the thermal runaway. distribution is affected by the initial temperature distribution. Temper­
Generally speaking, the temperature of lateral surfaces is considered to ature distribution will be influenced by boundary conditions and in­
be homogeneous in the engineering field since the Biot number is less ternal heat sources as heat propagates deeper. This stage (stage II-IV) is
than 0.1. The Biot number describes temperature distribution unifor­ the regular regime and takes up the majority of the time. When the
mity under unsteady heat conduction. Our experimental and simulation

Fig. 1. Simulation settings. (a) Geometry distinction; (b) Boundary conditions; (c) Geometry, meshing, and jelly roll setting of the battery.

3
Y. Zhang et al. Applied Energy 362 (2024) 123004

Fig. 2. Validation of simulation results. (a) lateral overheating experiments results; (b) simulation results of three temperature probe points [36].

Fig. 3. Surface plot of battery temperature distribution. (a) side view; (b) top view.

4
Y. Zhang et al. Applied Energy 362 (2024) 123004

Fig. 4. Internal temperature distribution at different stages of TR. (a) Temperature variance in the centerline of the battery; (b) Internal temperature distribution of
four characteristic stages.

temperature of one cell increases exponentially, the temperature of the cascading thermal runaway from cell to cell. It is concluded that heat
other increases rapidly (stage II). Once the cell has accumulated suffi­ conduction is the dominant heat transfer pathway before batteries go
cient energy, chemical reactions will occur between the battery’s active into thermal runaway. Heat conduction remains relatively stable for the
materials. If the heat continues to accumulate, cascading thermal first cell (cell 1), whereas reaction heat dominates the heat transfer
runaway (stage III) will be aroused. A large temperature gradient along pathway at stage II. In contrast, there was a marginal increase in the
the thickness direction can be observed from 2700 s to 2800 s, which is conduction value seen in the second cell before stage III. And thereafter,
attributed to lateral overheating. Resulting in a longer-than-predicted the reaction heat surpasses conduction as the predominant source of
thermal runaway detection delay. It is noteworthy that the battery’s heat. The power of reaction heat of cell 1 increases sharply over 15,000
maximum temperature during thermal runaway is located in the second W, leading to cell temperature soaring past 400 ◦ C. Owing to the poor
cell rather than the geometric center in stage III because the second cell heat dissipation and excessive heat accumulation in a short time, the
has a heat accumulation stage in stage II. other goes into thermal runaway so fast that process only lasts less than
Quantitative heat transfer and generation analysis are illustrated to 100 s, as shown in Fig. 5(c).
comprehend the detailed internal thermal runaway process, as shown in
Fig. 5. Conduction refers to the internal heat conduction within the
battery, while dissipation is the heat dissipation of the heated battery to 3.2. Triggering energy of onset and thermal runaway
surrounding environment. The energy conducted from the heater acti­
vates the chemical reactions and releases energy, which propels High-quality data is essential to the accurate assessment and com­
parison of battery technologies. Standardized testing procedures and

Fig. 5. Quantitative analysis of heat transfer and generation. (a-b) Heat power of heat transfer and generation of two cells in four stages (cell 1 refers to the cell close
to the heater); (c) Percentage stacked bar graph of heat transfer and generation.

5
Y. Zhang et al. Applied Energy 362 (2024) 123004

even test fixtures are recommended to minimize error, and outside in­ Table 2
fluence also accounting for remaining factors that may impact the Triggering energy of self-heat onset for various LFP batteries.
experiment [39]. The thermal runaway performance and characteristic Reference Battery Nominal TE of onset/ Tonset/ Tmax/
temperatures of a battery vary under different heating abuse scenarios. Format Capacity total self-heat ◦
C ◦
C
The influencing factors include whether the environment is adiabatic or (Ah) generation
not, whether energy is introduced, the amount of energy input, and the Bugryniec
18,650 1.5 25.4% 95 315
approach employed to initiate thermal failure. Hence, we are wondering [40]
if there is a consensus feature independent of abuse scenarios and trig­ Yang [41] Pouch 3 19.9% 97.02 408.14
Abada [42] 26,650 2.3 21.9% 79.7 275
gering conditions. Perea [43] 26,650 3 17.5% 100.2 455
Here we propose a variable, “triggering energy” (TE for short below), Lei [44] 18,650 1.1 27.8% 90 259
which refers to the minimum input energy capable of triggering status Prismatic 22 24.6% 88 328.6
change in the battery. In this work, we analyze the triggering energy of Wei [45]
Prismatic 22 52.2% 125.8 318.9
Prismatic 162 22.1% 85 350
self-heat onset (TE of onset for short below) and triggering energy of
Prismatic 195 15.0% 90 524.3
thermal runaway(TE of thermal runaway for short below). The defini­ Sun [12] Prismatic 50 23.9% 135.9 619.9
tion of triggering energy of self-heat onset is the net input energy for the Zhang [36] Prismatic 280 18.1% 70.6 340.7
battery at the start of side reactions occurrence, which corresponds to This work
26,700 4 23.1% 118 428
the battery characteristic temperature Tonset in ARC tests. Meanwhile, Prismatic 40 27.4% 96.6 339.9

the triggering energy of thermal runaway denotes the minimum net


energy required by a battery to enter thermal runaway. total self-generation heat is consistent with the range of 15%–52%.
Triggering energy can be obtained from both abuse experiments and Although a wide variance of onset temperature and maximum temper­
modelings. TE of onset can be obtained from the ARC tests in "H-S-W" ature exists among LFP batteries, consistency exists in the ratio of TE of
mode and can be calculated by the integral of the difference between the onset to total self-generation heat. The mean value is 24.5%, and the
input and output heat flux of the battery based on the simulation. To the median is 23.1%. The statistics of the ratio of TE of onset to total self-
best knowledge of the authors, there are no sensitive calorimeters generation heat reflect the thermal safety of the materials inside the
capable of identifying the triggering energy of thermal runaway, cell. Under the adiabatic condition, LFP battery gets most of the energy
whereas simulation can. The threshold for the beginning of a battery sources from internal exothermic reactions, which generally respond for
triggering the internal chemical reactions is determined to be 1.87 W, 80%, and the input energy accounts for approximately 20%. Addition­
which is benchmarking the 0.02 ◦ C/min temperature rate in deter­ ally, it indicates that a large capacity is not always related to poor
mining onset temperature in ARC tests. Correspondingly 1 ◦ C/s is thermal stability because the larger LFP battery the more triggering
applied to benchmark the battery thermal runaway. energy it requires. Besides, the approach implemented in LFP batteries
ARC experiment with its particular operation scheme can easily can be used for any other chemistries, for instance, NCM batteries, which
capture the critical temperature of self-generation heat from the ex­ is about 5% - 8% [13,14,46] far less than LFP batteries. That is to say,
periments’ perspective. The onset temperature of the thermal runaway internal chemical interactions account for about 95% of NCM battery
run determined by the "H-W-S" mode is the most vital parameter to heat sources during thermal runaway under the extreme adiabatic
calculate the triggering energy of onset. The onset temperature is environment.
70.6 ◦ C, and the maximum temperature is 340.7 ◦ C. The heat capacity of Another important characteristic is the triggering energy of thermal
the battery is 1029.49 J/kg/K, and the mass of the battery is 5.45 kg runaway. TE of onset can be acquired both from ARC experiments and
[36]. Therefore, TE of onset and its ratio to the total self-generation heat numerical modeling. However, it is time-consuming and expensive for
are summarized in Table 1. abuse tests to find out the critical TE of thermal runaway. In the nu­
Qtr = CP m(Tonset − T0 ) (3) merical modeling, a 10-s step of heating time is used to determine the
critical triggering energy of thermal runaway at a certain heating power.
Qtot = CP m(Tmax − Tonset ) (4) The triggering energy denotes the net input energy of the battery, which
can be described as Eq. (5).
To further explore the TE of onset of LFP batteries, more ARC tests ∑ ∑
with different types and capacities are summarized in Table 2. Qtr = Qin − Qout = Qheater − Qdis
According to Table 2, a wide variance of onset temperature and ∫ttr ∑∫
ttr
maximum temperature exists among LFP batteries. Since the synthesis of = ϕ1 Aside dt − heq Ai (Ti − Tamb )dt (5)
lithium iron phosphate is a complex and multiphase reaction, it is i
0 0
difficult for lithium iron phosphate battery manufacturers to maintain
consistent raw materials. Let alone the electrolyte and additives will where Qtr is the triggering energy of thermal runaway. Qin and Qout
have a great impact on the characteristic temperatures. Therefore, an represent the input heat (heat from the heater) and output heat (heat
unconditional and typical feature is vital in characterizing thermal dissipation), respectively. ∅1 is input heat flux, A is the surface area of
runaway. the battery, and heq is the equivalent heat dissipation coefficient.
Although the characteristic temperatures vary between batteries, TE Based on the modeling, triggering energy and self-heat generation of
of onset proportion to total heat generation shows good consistency and different heating power is summarized in Table 3.
stability among the references. Fig. 6 provides a quick overview of the When heating power is reduced, the TE of onset increases because
battery TR features. Based on the statistics of characteristic and typical slow heating can cause reaction onset in both jelly rolls at the same time
temperature Tonset and Tmax in the references, the ratio of TE of onset to and requires additional heating time. Heat generation occurs irregularly
both before and after TR because it depends on many aspects, including
heating and cooling conditions. TE of thermal runaway remains constant
Table 1
Critical parameters of adiabatic abuse scenario.
when multiple heating powers are used to activate thermal runaway and
the inaccuracy of triggering energy is only 2.5%. From the perspective of
Critical parameters Value
energy, a thermal runaway occurs when a battery accumulates enough
Triggering energy of self-heat onset 283.6 kJ momentum from both internal and external sources to release heat that
Total self-heat generation 1566.2 kJ
has been converted from the electrochemical energy that has been
TE for onset/total self-generation heat 18.1%

6
Y. Zhang et al. Applied Energy 362 (2024) 123004

Fig. 6. Statistics of characteristic parameters based on adiabatic TR tests. (a) Statistics of Tonset, (b) Statistics of Tmax, (c) The proportion of TE of onset to total self-
heat generation.

3.3. Thermal runaway characteristics with different cell configurations


Table 3
Triggering energy and self-heat generation of different heating power.
Internal battery structure can also influence internal thermal
Heating TE of TE of Self-heat Total self- Self-heat runaway behaviors of batteries. In this section, battery thickness and cell
power/ onset thermal generation heat generation
number are investigated to see how they effect the internal thermal
W /kJ runaway/ before TR/ generation/ before TR/
kJ kJ kJ Total self- performance.
heat Modeling analysis has been conducted to investigate the impact of
generation battery thickness on thermal runaway characteristics, as shown in Fig. 7.
600 281.1 673.8 270.2 793.4 34.0% It can be found that the battery thickness has little impact on the
500 325.1 679.3 362.0 1052.3 34.4% maximum temperature but changes the thermal runaway process in
400 343.7 700.0 289.0 906.5 31.9% terms of triggering time for thermal runaway Tonset. The triggering time
300 363.1 626.9 461.0 1224.0 37.7%
for thermal runaway and the time required to achieve the maximum
temperature grow linearly with battery thickness.
stored inside of it. For the 280 Ah LFP battery investigated in this paper, According to the control equation, the maximum temperature is
670 kJ is the minimum triggering energy of thermal runaway, ac­ constant because the thermophysical properties and kinetic parameters
counting for 20.8% theoretical energy stored inside. In addition, the remain the same. Based on the proposed triggering energy in Section
proportion of self-heat generation before TR to total self-heat generation 3.2, the ratio of triggering energy of onset to total energy remains
remains constant, although self-heat generation differs from multi- constant. When the total heat generation increases with battery thick­
heating power. The average percentage of self-heat generation before ness increase, the triggering energy proportionally goes up as well. As
and after TR is 34.5%. This means almost 65.5% of the self-heat will be mentioned before, the flat heater can be regarded as a steady-state heat
released in a short time, leading to a steep rise in the temperature. source, so the input heat has a linear relationship with time. Another
Moreover, the critical TE of thermal runaway can also be applied in interesting finding is that capacity has no evident correlation between
the development thermal runaway propagation model. TE of thermal characteristic temperatures and capacity, as capacity is closely related to
runaway can be used as a criterion to evaluate or even initiate thermal thickness in our model. This conclusion is consistent with the literature
runaway in a battery, considering temperature is a volume- and time- research [27].
dependent variable. Only a few thermal abuse experiments are needed Multiple jelly rolls (cells) are often inserted into a single container to
to calibrate the numerical model. The application of TE of thermal reduce dead space between the edge of a cell and a prismatic can.
runaway in propagation modeling will be explored in future However, the thinner the cells, the faster thermal runaway propagates,
investigations. as revealed in Fig. 8.
Thin cells, or more cells in a single battery, can accelerate the ther­
mal runaway process. In terms of characteristic battery temperature,
Tmax exhibits little variation. But cells located at greater distances from

7
Y. Zhang et al. Applied Energy 362 (2024) 123004

Fig. 7. TR characteristic variations under different battery thicknesses. (a) The average temperature of two cells. (b) The temperature of the lateral side opposite to
the heater. (c) The maximum value of average cell temperature and side temperature. (d) The time of side temperature and average cell temperature reaches
the maximum.

the heater experience a decrease in thermal runaway triggering tem­ release in the form of heat. If the heat generation consistently exceeds
perature. Cell characteristic temperature simulation results are consis­ the heat dissipation, resulting in heat accumulation until it approaches
tent with thermal runaway propagation between batteries [7,13]. When the triggering energy of thermal runaway, which accounts for 20.8% of
more cells are added to a battery, the time required to initiate thermal theoretical energy held inside LFP batteries, the battery will suffer a
runaway in the first cell decreases and further shortens the triggering rapid temperature rise and a large amount of energy release (65.5% of
time gap between the first thermal runaway cell and the others. The the self-heat generation for LFP batteries) in a short period of time.
underlying mechanism of the variation of thermal runaway character­ Moreover, the triggering energy of thermal runaway is a constant value
istic time is worth exploring. Fig. 9 displays the mechanism for the under various heating abuse scenarios, which can be a critical feature for
multi-cell influence on the thermal runaway. Temperature distribution characterizing battery safety.
remains the same at the very beginning, yet the average temperature of
cells near the heater might rise as more cells are added. Besides, cell 4. Conclusions
temperature and heat generation have positive feedback. Thus, more
cells and thinner cells can accelerate the thermal runaway propagation This study investigates the internal thermal runaway characteristics
by increasing the average temperature of the thinner cells. This serves as of large-format energy storage batteries by developing a thermal
a reminder to manufacturers that to offer safer performance in multi-cell runaway model. A critical concept, triggering energy, is introduced to
setups, an additional thermal-isolation layer must be placed between describe the minimum energy required by the battery to alter its internal
two cells. An additional cooling pad might also be needed for a battery safe status. Triggering energy of onset and thermal runaway are
pack design using multiple cells in a single battery. explored to investigate common traits in battery abuse. The influence of
Thermal runaway evolution from the triggering energy standpoint is battery thickness and cell count on thermal runaway characteristics is
illustrated in Fig. 10. The front view of the battery is intended to show investigated based on the simulation.
the typical internal microstructure from functioning to malfunctioning,
and the energy flow during this process is depicted on the lateral side of (1) A three-dimensional thermal runaway model is established to
the battery. In normal operation, the lithium-ion transfers between the investigate the internal thermal runaway characteristics. The
cathode and anode during charging and discharging. When the battery dynamic process of thermal runaway propagation between cells
temperature reaches the onset temperature, internal exothermic chain is clarified. The thermal propagation pathway is quantified and
reactions will occur and result in heat generation and gas evolution visualized in a stacked chart. Thermal runaway is a self-
which may lead to thermal runaway. If the battery constantly acquires accelerating process, with only 78 s between the initiation of
momentum from external energy sources, the battery will lose its safety the thermal runaway process in two neighboring cells. This in­
balance. When the input energy reaches the critical threshold (TE of dicates that detecting lateral temperature can be delayed during
onset), internal energy stored in electrochemical forms will begin to localized heating scenarios.

8
Y. Zhang et al. Applied Energy 362 (2024) 123004

Fig. 8. TR characteristic variations under different cell numbers. (a-c) The temperature of cells, two, three, and four cells, respectively; (d) The temperature of the
lateral side opposite to the heater of different cell numbers; (e) The time to TR for different cell numbers.

Fig. 9. Illustration of the influence of multi-jelly rolls on temperature characteristics.

(2) The concept of triggering energy is proposed to identify a com­ (3) The thermal runaway time is greatly affected by battery thickness
mon feature across various heat-abuse scenarios. According to and cell number, while the maximum temperature is slightly
the simulations and experiments, the ratio of triggering energy of influenced. Thermal runaway time rises approximately linearly
onset to total self-heat generation is 24.5%. This means that LFP as battery thickness grows since the increase in volume leads to
battery gets most of the energy sources from internal exothermic an increase in triggering energy. Conversely, increasing the
reactions, which generally respond to 80%. For the investigated number of cells and the thickness of each cell in a battery can
280 Ah LFP battery, 670 kJ is the minimum triggering energy of accelerate the thermal runaway process, due to the average
thermal runaway, and 65.5% of the self-heat will be released temperature of each cell rising compared to counterpart cells at
during thermal runaway in a short period. The triggering energy that time. Besides, temperature and heat generation are mutually
of thermal runaway stays consistent for the same battery type reinforcing. This analysis could be helpful to manufacturers and
under various heat abuse conditions, indicating that the trig­ producers aiming to optimize structures. Despite advancements,
gering energy of thermal runaway has the potential application in safety performance remains a potential Achilles’ heel obstructing
thermal runaway numerical modeling. the widespread use of batteries.

9
Y. Zhang et al. Applied Energy 362 (2024) 123004

Fig. 10. Energy characteristics of thermal runaway of LIBs.

This model can be utilized for predicting internal battery tempera­ China National Postdoctoral Program for Innovative Talents (No.
tures, forecasting thermal runaway in overheating scenarios, validating BX20220286), China Postdoctoral Science Foundation (No.
and optimizing battery design parameters. Triggering energy can serve 2022M723040). Jiang LH is supported by the National Natural Science
as a more precise criterion for thermal runaway triggering. Embedding Foundation of China (No. 52204248). Zhang Y is supported by the
the model into battery management systems in the fields of electric Students’ Innovation and Entrepreneurship Foundation of USTC (No.
vehicles and energy storage systems enables thermal runaway predic­ CY2022C12).
tion. Meanwhile, the proposed model is aimed to illustrate the internal
thermal characteristics of batteries, drawing on principles from both References
thermal and chemical kinetics. Consequently, the applicability of the
model is limited to thermal abuse scenarios and is incapable of depicting [1] Jeevarajan JA, Joshi T, Parhizi M, Rauhala T, Juarez-Robles D. Battery Hazards for
large energy storage systems. ACS Energy Lett 2022;7(8):2725–33.
the venting process. Comprehensive analysis of internal gas generation [2] Lyu P, Liu X, Qu J, Zhao J, Huo Y, Qu Z, et al. Recent advances of thermal safety of
and fluid dynamics will be the future priorities. lithium ion battery for energy storage. Energy Storage Mater 2020;31:195–220.
[3] Chen S, Wei X, Zhang G, Wang X, Zhu J, Feng X, et al. All-temperature area battery
application mechanism, performance, and strategies. Innovation 2023;4(4):
CRediT authorship contribution statement 100465.
[4] Jia Z, Min Y, Qin P, Mei W, Meng X, Jin K, et al. Effect of safety valve types on the
Yue Zhang: Writing – original draft, Software, Methodology, gas venting behavior and thermal runaway hazard severity of large-format
prismatic lithium iron phosphate batteries. J Energy Chem 2024;89:195–207.
Investigation, Formal analysis, Data curation, Conceptualization. Lai­ [5] Huang Z, Li X, Wang Q, Duan Q, Li Y, Li L, et al. Experimental investigation on
feng Song: Investigation, Conceptualization. Jiamin Tian: Investiga­ thermal runaway propagation of large format lithium ion battery modules with two
tion, Conceptualization. Wenxin Mei: Writing – review & editing, cathodes. Int J Heat Mass Transf 2021;172:121077.
[6] Feng X, Sun J, Ouyang M, Wang F, He X, Lu L, et al. Characterization of penetration
Supervision, Software, Methodology. Lihua Jiang: Writing – review &
induced thermal runaway propagation process within a large format lithium ion
editing. Jinhua Sun: Supervision, Resources, Methodology. Qingsong battery module. J Power Sources 2015;275:261–73.
Wang: Writing – review & editing, Supervision, Resources, Project [7] Jin C, Sun Y, Wang H, Zheng Y, Wang S, Rui X, et al. Heating power and heating
energy effect on the thermal runaway propagation characteristics of lithium-ion
administration, Funding acquisition.
battery module: experiments and modeling. Appl Energy 2022;312:118760.
[8] Lamb J, Orendorff CJ, Steele LAM, Spangler SW. Failure propagation in multi-cell
Declaration of competing interest lithium ion batteries. J Power Sources 2015;283:517–23.
[9] Coman PT, Darcy EC, Veje CT, White RE. Numerical analysis of heat propagation in
a battery pack using a novel technology for triggering thermal runaway. Appl
The authors declare that they have no known competing financial Energy 2017;203:189–200.
interests or personal relationships that could have appeared to influence [10] Jia Y, Uddin M, Li Y, Xu J. Thermal runaway propagation behavior within 18,650
lithium-ion battery packs: a modeling study. J Energy Storage 2020:31.
the work reported in this paper. [11] Mishra D, Zhao P, Jain A. Thermal runaway propagation in Li-ion battery packs
due to combustion of vent gases. J Electrochem Soc 2022;169(10).
Data availability [12] Sun T, Wang L, Ren D, Shi Z, Chen J, Zheng Y, et al. Thermal runaway
characteristics and modeling of LiFePO4 power battery for electric vehicles.
Automot. Innov. 2023;6:414–24.
Data will be made available on request. [13] Wang H, Liu B, Xu C, Jin C, Li K, Du Z, et al. Dynamic thermophysical modeling of
thermal runaway propagation and parametric sensitivity analysis for large format
lithium-ion battery modules. J Power Sources 2022;520:230724.
Acknowledgments
[14] Jin C, Sun Y, Wang H, Lai X, Wang S, Chen S, et al. Model and experiments to
investigate thermal runaway characterization of lithium-ion batteries induced by
This work is supported by the National Key Research and Develop­ external heating method. J Power Sources 2021;504:230065.
ment Program of China (No. 2022YFE0207400). Mei WX is supported by

10
Y. Zhang et al. Applied Energy 362 (2024) 123004

[15] Fransson M, Broche L, Buckwell M, Pfaff J, Reid H, Kirchner-Burles C, et al. [31] Zhang Y, Mei W, Qin P, Duan Q, Wang Q. Numerical modeling on thermal runaway
Sidewall breach during lithium-ion battery thermal runaway triggered by cell-to- triggered by local overheating for lithium iron phosphate battery. Appl Therm Eng
cell propagation visualized using high-speed X-ray imaging. J Energy Storage 2021;192:116928.
2023;71:108088. [32] Zhou H, Parmananda M, Crompton KR, Hladky MP, Dann MA, Ostanek JK, et al.
[16] Hatchard TD, MacNeil DD, Basu A, Dahn JR. Thermal model of cylindrical and Effect of electrode crosstalk on heat release in lithium-ion batteries under thermal
prismatic lithium-ion cells. J Electrochem Soc 2001;148(7):A755–61. abuse scenarios. Energy Storage Mater 2022;44:326–41.
[17] Kim G-H, Pesaran A, Spotnitz R. A three-dimensional thermal abuse model for [33] Wang Y, Ren D, Feng X, Wang L, Ouyang M. Thermal runaway modeling of large
lithium-ion cells. J Power Sources 2007;170(2):476–89. format high-nickel/silicon-graphite lithium-ion batteries based on reaction
[18] Zhang F, Feng X, Xu C, Jiang F, Ouyang M. Thermal runaway front in failure sequence and kinetics. Appl Energy 2022;306:117943.
propagation of long-shape lithium-ion battery. Int J Heat Mass Transf 2022;182: [34] Wu H, Chen S, Chen J, Jin C, Xu C, Rui X, et al. Dimensionless normalized
121928. concentration based thermal-electric regression model for the thermal runaway of
[19] Zhao P, Liu L, Chen Y, Ge H. Theoretical and numerical analysis for thermal lithium-ion batteries. J Power Sources 2022;521:230958.
runaway propagation within a single cell. Int J Heat Mass Transf 2021;181: [35] Feng X, Lu L, Ouyang M, Li J, He X. A 3D thermal runaway propagation model for a
121901. large format lithium ion battery module. Energy 2016;115:194–208.
[20] Kong D, Wang G, Ping P, Wen J. A coupled conjugate heat transfer and CFD model [36] Zhang Y, Cheng S, Mei W, Jiang L, Jia Z, Cheng Z, et al. Understanding of thermal
for the thermal runaway evolution and jet fire of 18650 lithium-ion battery under runaway mechanism of LiFePO4 battery in-depth by three-level analysis. Appl
thermal abuse. eTransportation; 2022. Energy 2023;336:120695.
[21] Wang G, Kong D, Ping P, He X, Lv H, Zhao H, et al. Modeling venting behavior of [37] Zhou Z, Zhou X, Wang D, Li M, Wang B, Yang L, et al. Experimental analysis of
lithium-ion batteries during thermal runaway propagation by coupling CFD and lengthwise/transversal thermal characteristics and jet flow of large-format
thermal resistance network. Appl Energy 2023;334:120660. prismatic lithium-ion battery. Appl Therm Eng 2021;195:117244.
[22] Wang G, Kong D, Ping P, Wen J, He X, Zhao H, et al. Revealing particle venting of [38] Jia Z, Song L, Mei W, Yu Y, Meng X, Jin K, et al. The preload force effect on the
lithium-ion batteries during thermal runaway: a multi-scale model toward thermal runaway and venting behaviors of large-format prismatic LiFePO4
multiphase process. eTransportation 2023;16:100237. batteries. Appl Energy 2022;327:120100.
[23] Wang G, Ping P, Peng R, Lv H, Zhao H, Gao W, et al. A semi reduced-order model [39] Hayman D, Dufek EJ, Mukherjee PP, Sholklapper T, Love CT. Editorial:
for multi-scale simulation of fire propagation of lithium-ion batteries in energy establishing standards for battery data and pathways towards its validation. Front
storage system. Renew Sustain Energy Rev 2023;186:113672. Energy Res 2023:11.
[24] Kim M, Jeon J, Hong J. Reaction mechanism study and modeling of thermal [40] Bugryniec PJ, Davidson JN, Cumming DJ, Brown SF. Pursuing safer batteries:
runaway inside a high nickel-based lithium-ion battery through component thermal abuse of LiFePO4 cells. J Power Sources 2019;414:557–68.
combination analysis. Chem Eng J 2023;471:144434. [41] Yang M, Ye Y, Yang A, Jiang Z, Wang X, Yuan H, et al. Comparative study on aging
[25] Shi Y. Feasibility of BYD blade batteries in electric vehicles. Highlights Sci Eng and thermal runaway of commercial LiFePO4/graphite battery undergoing slight
Technol 2023;32:193–8. overcharge cycling. J Energy Storage 2022;50:104691.
[26] Duh Y-S, Theng J-H, Chen C-C, Kao C-S. Comparative study on thermal runaway of [42] Abada S, Petit M, Lecocq A, Marlair G, Sauvant-Moynot V, Huet F. Combined
commercial 14500, 18650 and 26650 LiFePO4 batteries used in electric vehicles. experimental and modeling approaches of the thermal runaway of fresh and aged
J Energy Storage 2020:31. lithium-ion batteries. J Power Sources 2018;399:264–73.
[27] Chen J, Ren D, Hsu H, Wang L, He X, Zhang C, et al. Investigating the thermal [43] Perea A, Paolella A, Dubé J, Champagne D, Mauger A, Zaghib K. State of charge
runaway features of lithium-ion batteries using a thermal resistance network influence on thermal reactions and abuse tests in commercial lithium-ion cells.
model. Appl Energy 2021;295:117038. J Power Sources 2018;399:392–7.
[28] Richard MN, Dahn JR. Accelerating rate calorimetry study on the thermal stability [44] Lei B, Zhao W, Ziebert C, Uhlmann N, Rohde M, Seifert HJ. Experimental analysis
of lithium intercalated graphite in electrolyte II. Modeling the results and of thermal runaway in 18650 cylindrical Li-Ion Cells using an accelerating rate
predicting differential scanning calorimeter curves. J Electrochem Soc 1999;146 calorimeter 2017;3(2):14.
(6):2078–84. [45] Wei G, Huang R, Zhang G, Jiang B, Zhu J, Guo Y, et al. A comprehensive insight
[29] MacNeil DD, Dahn JR. Test of reaction kinetics using both differential scanning and into the thermal runaway issues in the view of lithium-ion battery intrinsic safety
accelerating rate calorimetries as applied to the reaction of LixCoO2 in non- performance and venting gas explosion hazards. Appl Energy 2023;349:121651.
aqueous electrolyte. J Phys Chem A 2001;105(18):4430–9. [46] Xu C, Feng X, Huang W, Duan Y, Chen T, Gao S, et al. Internal temperature
[30] Ren D, Liu X, Feng X, Lu L, Ouyang M, Li J, et al. Model-based thermal runaway detection of thermal runaway in lithium-ion cells tested by extended-volume
prediction of lithium-ion batteries from kinetics analysis of cell components. Appl accelerating rate calorimetry. J Energy Storage 2020;31:101670.
Energy 2018;228:633–44.

11

You might also like