You are on page 1of 10

Journal of Energy Chemistry 91 (2024) 609–618

Contents lists available at ScienceDirect

Journal of Energy Chemistry


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

Understanding the combustion behavior of electric bicycle batteries


and unveiling its relationship with fire extinguishing
Zhanglong Yu a,b,1, Xueling Shen a,b,1, Ran Xu a,b, Zheng Wang c, Zengming Wan a,b, Mingyang Chen d,
Yi Cui a,b, Yanyan Fang a,b,⇑, Xiaoli Ma a,b,⇑
a
China Automotive Battery Research Institute Co., Ltd., Beijing 101407, China
b
China Grinm Group Co., Ltd., Beijing 100088, China
c
Ministry of Industry and Information Technology, Equipment Industry Development Center, Beijing 100846, China
d
Hefei Guoxuan High-tech Power energy Co., Ltd., Hefei 230000, Anhui, China

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

Article history: In this study, a detailed analysis of the combustion behaviors of the lithium iron phosphate (LFP) and
Received 14 December 2023 lithium manganese oxide (LMO) batteries used in electric bicycles was conducted. This research included
Revised 31 December 2023 quantitative measurements of the combustion duration, flame height, combustion temperature, heat
Accepted 1 January 2024
release rate, and total heat release. The results indicated that LMO batteries exhibited higher combustion
Available online 18 January 2024
temperatures of 600–700 °C, flame heights of 70–75 cm, a significantly higher heat release rate of
40.1 kW (12 Ah), and a total heat release of 1.04 MJ (12 Ah) compared to LFP batteries with the same
Keywords:
capacity. Based on these experimental results, a normalized total heat release (NORTHR) parameter was
Combustion behavior
Electric bicycle
proposed, demonstrating good universality for batteries with different capacities. Utilizing this parame-
Lithium-ion battery ter, quantitative calculations and optimization of the extinguishing agent dosage were conducted for fires
Fire extinguishing involving these two types of batteries, and the method was validated by extinguishing fires for these two
Normalized total heat release types of battery packs with water-based extinguishing fluids.
Ó 2024 Science Press and Dalian Institute of Chemical Physics, Chinese Academy of Sciences. Published
by ELSEVIER B.V. and Science Press. All rights reserved.

1. Introduction [12–17]. In ordinary life, during the charging and discharging pro-
cess of a battery, needle-like lithium is formed due to the uneven
Lithium-ion batteries, known for their portability and environ- deposition of lithium, which reduces the thermal stability of the
mental friendliness, have become integral in various industries, battery [18]. Furthermore, overcharging is a typical form of electri-
including the consumer electronics, electric vehicle, and electric cal abuse [17,19–21]. During overcharging, internal side reactions
bicycle (E-bicycle) industries [1–3]. E-bicycles, in particular, have in the battery are intensified, accumulating heat and potentially
become the preferred mode of short-distance travel in many coun- leading to thermal runaway and the production of jet flames,
tries [4–6]. As of 2022, E-bicycle ownership in China reached 350 which can ultimately lead to further fires [22–25]. Moreover, the
million, with the global market valued at 17.83 billion USD [7]. amount of heat and the types and quantities of chemical products
However, in just the first half of the same year, China reported released during the combustion of lithium-ion batteries are key
8370 cases of fires involving E-bicycles [8], and in October, the Uni- indicators in electrochemical safety research. This includes the
ted States recalled approximately 22,000 E-bicycles due to poten- heat release rate (HRR) and the total heat release (THR), which
tial fire and explosion risks from their lithium batteries [9]. are critical for understanding the chemical and physical behaviors
Serious E-bicycle fires occur during the thermal runaway pro- of batteries under extreme conditions and reflecting the intensity
cess in lithium-ion batteries [10,11]. Thermal runaway and subse- relationship with the battery materials [26–28]. Wang et al. [29]
quent ignition of lithium-ion batteries can result from multiple summarized the impact of reactions between the cathode, anode,
factors, including mechanical, electrical, and thermal abuse and electrolyte of lithium-ion batteries on the thermal runaway,
fires, and explosions of these batteries. Feng et al. [30] proposed
⇑ Corresponding authors. a new energy release diagram summarizing the reactions of
E-mail addresses: fangyy@glabat.com (Y. Fang), maxl@glabat.com (X. Ma). materials in batteries during thermal runaway, which could
1
These authors contributed equally to this work. quantify the reaction kinetics for all the battery component

https://doi.org/10.1016/j.jechem.2024.01.005
2095-4956/Ó 2024 Science Press and Dalian Institute of Chemical Physics, Chinese Academy of Sciences. Published by ELSEVIER B.V. and Science Press. All rights reserved.
Z. Yu, X. Shen, R. Xu et al. Journal of Energy Chemistry 91 (2024) 609–618

materials. Liu et al. [31] suggested that the combustion character- illustrated in Fig. 1(a), the experimental apparatus comprised a
istics of overcharged 228 Ah lithium iron phosphate batteries are combustion chamber, a thermal runaway triggering device, an
more intense than those triggered by heating, as reflected in the electronic balance, analyzers and data recorders, cameras, and
peak HRR values. Therefore, these studies demonstrated that the other components. The HRR testing device is manufactured by
combustion characteristics of lithium-ion batteries are not only a Suzhou Phoenix Instrument Co. Ltd. (model PX 07-001). The
crucial aspect of energy chemistry research but also essential for combustion gas was collected by a high-powered fan with a flows
understanding and preventing E-bicycle fires and their impacts speed of 0.5–0.6 m3/h to analyze the O2 consumption, and the HRR
on the environment and safety. was calculated by the HRR testing device. Other data were
Researchers have begun exploring the effectiveness of various recorded using a HIOKI LR8450 data logger.
fire extinguishants at the chemical level on battery fires, including An extinguishing fluid spraying device designed in-house was
understanding the interactions between different extinguishants used for the extinguishing tests. The extinguishing fluid was circu-
and the chemical reactions of the battery [32]. Sun et al. [33] calcu- lated through a high-pressure pump into a simulated fire extin-
lated the design mass of the gaseous extinguishant C6F12O based guishing box of E-bicycle battery. The internal dimensions of the
on the formula in the DB37/T 3642-2019 standard, considering fac- fire extinguishing box were 500 mm  356 mm  410 mm. Addi-
tors like the volume of the protected area, design extinguishing tionally, an electronic igniter was installed inside the extinguishing
concentration, specific volume of superheated steam, minimum box. Once the box was filled with smoke, the electronic ignition
ambient temperature, and altitude correction factor. Zhang et al. device was activated for electronic ignition. This was primarily
[34] discussed the effectiveness of water sprays in controlling bat- due to the difficulty in igniting LFP battery packs during overcharg-
tery thermal runaway, mentioning that the design consumption of ing. Two rows of sprinklers were evenly arranged at the top of the
water sprays is related to the efficiency of water use, and they con- fire extinguishing box, and each row contained five heads with
ducted experiments to determine the final adjustments to obtain 100 mm spacing between the adjacent sprinklers, shown in
precise design dosages. These studies, primarily covering the com- Fig. 1(b).
bustion characteristics of batteries, can provide significant insights
for the prevention and mitigation of E-bicycles fires, particularly in 2.3. Test methods
which equipped with retired electric vehicle batteries [35,36].
However, research on predicting and devising firefighting strate- Single-cell battery pretreatment method: All single cells were
gies specifically for E-bicycles remains relatively scarce, especially charged to 100% SOC before the experiments. The charging process
concerning the combustion characteristics of battery materials like involved 1/3 C constant-current charging to the maximum voltage,
lithium manganese oxide (LMO), which is not commonly used in followed by constant-voltage charging until the current dropped
power batteries [37]. Therefore, there is an urgent need for com- to 0.05 C.
prehensive research dedicated to fire prevention and suppression Battery pack pretreatment method: Similar to the single cells,
strategies for E-bicycle batteries, with a focus on understanding battery packs were charged to 100% SOC before the experiments.
and addressing the unique risks posed by these specific battery This involved a 1/3 C constant-current charging to the maximum
systems. voltage of the battery pack, followed by constant-voltage charging
This study focused on LMO and lithium iron phosphate (LFP) until the current was 0.05 C.
batteries and their battery packs commonly used in E-bicycles. Combustion characteristics testing method: The pretreated
The hazards posed by the heat generated in battery fires were samples were installed in the battery combustion HRR testing
assessed by studying their thermal characteristics. Furthermore, device. Samples were triggered to thermal runaway by overcharg-
a normalized total heat release (NORTHR) method, which is univer- ing at a 1 C constant current. The smoke extraction system,
sal for different kinds of lithium-ion batteries, was further pro- explosion-proof electronic balance, gas analyzers, cameras, and
posed to enable quantitative calculations and optimization of the thermal analysis systems were activated synchronously to record
amount of fire extinguishing agent required for fires involving data throughout the test. Moreover, several rulers were placed at
these batteries packs with a spraying device designed in house. different heights above the samples, allowing measurement of
This method can provide theoretical and technical guidance for the flame width and height after photography. The flame photos
surveying the fire accidents of E-bicycle batteries. are shown in Fig. S1. After testing, recording was stopped once
the sample temperature dropped below 60 °C, and the post-
experiment samples were photographed. In these tests, a type-K
2. Experimental
thermocouple sensor was placed at the center of the largest face
of the sample, and all flames were ignited by an electronic igniter
2.1. Battery samples
that was activated when the sample began to emit smoke.
Extinguishing test method: The pretreated battery packs were
The study examined E-bicycle batteries, specifically pouch cells
placed in an extinguishing box, simulating the layout in a charging
and their assembled battery packs, provided by CHIWEE Group Co.
cabinet. The box dimensions are presented in Section 2.2. Thermo-
Ltd. The batteries and battery packs used in this experiment were
couple sensors were positioned at the center of the left side, the
of two systems: LMO-graphite and LFP-graphite. To compare com-
top center, and the center of the right side of the inner-side of fire
bustion characteristics, single cells with capacity of 12, 24, and 36
extinguishing box, each at the closest distance of 100 mm from the
Ah and battery packs with capacity of 30 Ah were used in this
sample, and they were respectively labeled as T1, T2, and T3. Prior to
work. Other parameters are shown in Table 1. The batteries were
testing the battery pack samples, protective devices within the
fully charged using a constant current to constant voltage method
samples were removed and the battery pack casings were restored
at a rate of 1 C until the current dropped to 0.05 C, ensuring that
to their initial condition. Then, the entire battery pack was over-
each battery reached a 100% state of charge (SOC).
charged at a 1 C constant current to trigger thermal runaway and
ignition. The extinguishing spray system was activated when T2
2.2. Experimental setup reached 400 °C and was maintained until the fire was completely
extinguished. Based on GB 50219-2014 ‘‘Technical Specifications
The combustion characteristics tests were conducted in a setup for Water Spray Extinguishing Systems”, the extinguishing param-
compliant with ISO9705 combustion chamber requirements. As eters were a spray pressure of approximately 0.25 MPa and a flow
610
Z. Yu, X. Shen, R. Xu et al. Journal of Energy Chemistry 91 (2024) 609–618

Table 1
The basic parameter of lithium-ion battery.

Sample name Material system Capacity (Ah) Voltage range (V) Weight (g) Size: Length  Width  Height (mm  mm  mm)
Cell LFP/graphite 12 2.5–3.65 218.5 125  120  10
24 469.4 225  135  10
36 743.7 225  155  15
LMO/graphite 12 2.8–4.2 273.5 125  120  10
24 554.8 225  135  10
36 923.3 225  155  15
Pack LFP/graphite (1P15S) 30 37.5–54.75 9168.8 270.4  170.2  165.2
LMO/graphite (1P13S) 30 36.4–54.6 8048.1 250.5  165.5  155.5

Fig. 1. Schematic diagram of the experimental research device platform. (a) The experimental device of combustion characteristics. (b) The experimental device of water-
based fire extinguisher.

rate of about 10 L/min in total. The model of the extinguishing red during the combustion process, while that of the LMO battery
agent used was S-3-AB-(1 °C) (Beijing Jiu-jiu Shenlong Fire was lighter red, which indicated that the LMO flame temperature
Equipment Co., Ltd.). was higher than the LFP battery flame temperature.
Differential scanning calorimetry (DSC) tests were performed The widths of the flames for both types of batteries showed lin-
on a thermal analysis system (DSC3, Mettler-Toledo International ear increases with the capacity, with the maximum flame width of
Inc.) under an air atmosphere at a heating rate of 5 °C/min. Sam- the LMO cell being about 10 cm wider than that of the LFP cell. This
ples were prepared in a glove box by placing about 2 mg of dried approximate linear relationship was mainly due to the increase in
powder removed from the cathodes of fully charged LMO or LFP the battery capacity, and the amount of fuel for the flame (gas
batteries, and they were heated with 2 drops (0.1 mL) of conven- produced during overcharge-induced thermal runaway and elec-
tional lithium-ion battery electrolyte (1.0 M LiPF6 dissolved in trolyte) also increased almost linearly. In the range of the battery
mixed ethylene carbonate, diethyl carbonate, and ethyl methyl capacity in this work, the width of the flame increased with the
carbonate at a ratio of 1:1:1 (vol%)). capacity during thermal runaway because punch cells mostly had
cracks first appearing on the edges of the sides, causing gas and
electrolyte to escape from the sides. This led to an increase in the
3. Results and discussion flame width as the capacity increased. The timestamps in the bot-
tom left corner of each image in Fig. 2(a) represent the duration of
3.1. Combustion characteristics of E-bicycle batteries the burning for each battery. As the capacity increased, the burning
durations for both types of batteries also increased. Notably, the
The most common E-bicycle batteries are LMO and LFP types. In burning durations of the LFP batteries were generally three times
the event of thermal runaway, these batteries can cause intense longer than those of LMO batteries with the same capacity, possi-
combustion in E-bicycles. The combustion characteristics of the bly due to the greater gas production during thermal runaway in
batteries are closely related to their SOCs, and to simulate the most the LFP batteries, which prolonged the combustion process. This
extreme thermal runaway scenarios, all tests were conducted on also indicated that the combustion intensity during thermal run-
lithium-ion batteries at 100% SOCs triggered at 1 C. Fig. 2(a) shows away was greater in the LMO batteries than that in the LFP batter-
the analysis of the flame morphologies of different-capacity LFP ies. For batteries with the same specific active material loading per
and LMO batteries after thermal runaway. For LMO batteries, the unit area, there is a proportional relationship between the capacity
flame heights for the batteries with different capacities were con- and battery size. The observed correlation between the flame
sistently within the range of 70–75 cm, whereas for LFP batteries, status and the battery’s planar dimensions exhibited a trend anal-
they remained within 55–60 cm. This indicated that during the ogous to that noted for the capacity, which indicates a proportional
thermal runaway and diffusion process for 12 to 36 Ah battery relationship linking these physical parameters to the combustion
packs, the flame height was not significantly affected by the cell behavior of the battery.
capacity but was greatly influenced by the type of electrode mate- Furthermore, data analysis of the temperature collection during
rial. LFP batteries generally had flames that were about 15 cm the combustion process of the samples, as shown in Fig. 2(b),
lower than those of the LMO batteries with the same capacity. In indicated that for batteries with the same capacity, the maximum
addition, according to Fig. S1, the flame of LFP battery was mainly temperatures after thermal runaway for the LFP batteries were
611
Z. Yu, X. Shen, R. Xu et al. Journal of Energy Chemistry 91 (2024) 609–618

Fig. 2. Lithium-ion battery combustion process and temperature change. (a) Maximum flame photographs of different lithium-ion batteries. (b) Combustion temperature
diagrams of different lithium-ion batteries. (c) Relationship between flame width, maximum temperature, and battery capacity.

approximately 400–500 °C, which were lower than the peak tem- combustion heat release was more intense in the LMO batteries.
peratures (600–700 °C) of the LMO batteries. This suggested that As shown in Fig. 3(b), the THR values for both battery types
the combustion intensity was less for the LFP batteries than that increased with the capacity. For the same capacity, the THR of
for the LMO batteries. Furthermore, the combustion temperatures the LMO batteries was about 1.2–1.3 times that of LFP batteries,
of both the LFP and LMO battery samples increased with the indicating that LMO batteries released more heat upon combus-
battery capacity. However, the highest temperature rose to some tion. Fig. 3(c) shows the linear fits of the peak values of the HRR
extent as the battery capacity increased. corresponding to Fig. 3(a). The values of the square of the linear
The reason that the temperature differences of the battery com- correlation coefficients (R2) in Fig. 3(c) show that the LFP batteries
bustion flames found in many studies were more than 600 °C, or had better linear relationships between the battery capacities and
even more than 1000 °C, needs to be determined. The temperature the peak values of HRR than the LMO batteries. Meanwhile, the
has a strong relationship with the capacities, material systems, and fitted curves of the two batteries were significantly different,
combustion spaces of lithium-ion batteries [38–40]. The capacity which indicated that the peak values of THR and HRR in the
of the single batteries used in this test was only 12–36 Ah. The combustion process were closely related to the battery material
combustion space was also relatively empty in the explosion- system. These differences were likely due to variations in the
proof room, and thus the actual temperature obtained was also battery systems, as illustrated by the DSC curves in Fig. 3(d). The
relatively low. Based on Fig. 2(a and b), if we establish functional thermal release of the cathode active material of the fully charged
relationships between the flame width W, the maximum tempera- LMO battery was approximately twice that of the corresponding
ture Tmax, and the battery capacity Q, as illustrated in Fig. 2(c), we LFP battery, releasing more heat during combustion, which
can observe that both the flame width and maximum temperature resulted in larger THR values of the LMO batteries than those of
increased almost linearly with the battery capacity. However, the the LFP batteries with the same capacity. The primary cause of this
intercepts differed for the LFP and LMO systems, which provided issue was that the chemical stability of the LMO cathode material
quantitative guidance for extinguishing fires in E-bicycles. at 100% SOC was poorer than that of the LFP. The LMO cathode
Fig. 3 presents the measurement results of the HRR and THR material underwent multiple-step reactions when heated in
during the combustion processes of the two types of lithium-ion conjunction with the electrolyte, resulting in the release of various
batteries. Fig. 3(a) shows that the peak values of the heat release combustible gases and oxygen as well as the heat release of
for the single-cell batteries of the same type increased with the approximately 252.95 J/g. In contrast, the 100% SOC LFP cathode
capacity. To clearly observe the peak HRR values of each sample, material exhibited fewer reactions and released less heat,
the curves in Fig. 2(a) were shifted, and the x-axis coordinates approximately 98.43 J/g.
are not provided. For LFP batteries, the peak HRR raised from Additionally, the mass losses and mass loss rates during com-
8.9 kW for the 12 Ah battery to 23.4 kW for the 36 Ah battery, bustion were measured, and the results are shown in Table 2.
while for LMO batteries, it increased from 40.1 kW for the 12 Ah The peak mass loss rates increased with the battery capacity,
battery to 58.3 kW for the 36 Ah battery. The shapes of the HRR suggesting that larger batteries incurred greater combustion
peaks for the LFP batteries were broader, but the peak HRRs for losses. The peak mass loss rates of the LMO batteries were not
the LMO batteries were significantly higher, indicating that the significantly higher than those of the LFP batteries. This was

612
Z. Yu, X. Shen, R. Xu et al. Journal of Energy Chemistry 91 (2024) 609–618

Fig. 3. Combustion characteristics data of different lithium-ion batteries. (a) HRR. (b) THR. (c) Linear fitting curves of peaks in (a). (d) DSC curves of fully charged LMO and LFP
cathode active materials.

Table 2 dation for the suppression of fires involving lithium-ion batteries


Combustion process characteristics parameter of LFP and LMO battery. and battery packs.
Battery type Peak of HRR (kW) THR (MJ) Peak of mass loss rate (g/s) Drawing on the results of the battery combustion characteris-
12 Ah LFP 8.9 0.77 3.33
tics tests, we performed an energy normalization analysis on the
12 Ah LMO 40.1 1.04 2.67 HRR and THR of each battery relative to the rated energy of the
24 Ah LFP 18.5 1.66 9.65 battery, designated as NORHRR and NORTHR, which are defined
24 Ah LMO 53.7 2.25 9.33 respectively as follows.
36 Ah LFP 23.4 2.57 12.67
36 Ah LMO 58.3 3.08 13.33 HRR
NORHRR ¼ ð1Þ
U  Q  3600

primarily because the LFP batteries could experience case rupture THR
NORTHR ¼ ð2Þ
and venting during overcharging-induced thermal runaway, lead- U  Q  3600
ing to electrolyte decomposition and vaporization without igni- where U represents the rated voltage of the sample, with calculated
tion. In contrast, LMO batteries typically ignited upon thermal values of 3.2 and 3.7 V for the LFP and LMO batteries, respectively,
runaway during overcharging. Consequently, the THRs of the LFP and Q denotes the sample capacity in Ah. The normalized NORHRR
batteries were less than those of the LMO batteries, although the represents the HRR per unit energy of a lithium-ion battery, which
peak mass loss rates did not differ substantially. can be used to assess the ignition intensities of different types of
lithium-ion batteries. Similarly, NORTHR represents the amount of
3.2. Energy normalization analysis of combustion characteristics energy released per unit of energy in a lithium-ion battery fire,
which can be used to evaluate the fundamental characteristics of
Based on the preliminary discussions of the combustion charac- such fires.
teristics, we have identified significant differences in the HRR and In Fig. 4(a), the results of the normalized HRR (NORHRR) energy
THR between the different types of lithium-ion batteries. Notably, for the LFP and LMO batteries are shown, and to observe the
the HRRs of the LFP batteries were distinctly lower than those of variation of the NORHRR values over time, a shift similar to that
the LMO batteries, and the THR was correlated with the battery in Fig. 2(a) was performed. The data revealed that the NORHRR
capacity. To a certain extent, the combustion characteristics can curve peaks for the LFP batteries, irrespective of the capacity, were
serve as a basis for fire extinguishing strategies, providing a foun- remarkably consistent, with an average value of approximately
613
Z. Yu, X. Shen, R. Xu et al. Journal of Energy Chemistry 91 (2024) 609–618

Fig. 4. Energy normalization of LFP and LMO battery. (a) Normalized HRR energy. (b) Normalized THR energy.

0.065 s1. This consistency suggested the uniformity in the HRR tal in determining the necessary quantity of the fire extinguishing
per unit energy across the studied capacity spectrum for the LFP agent for a lithium-ion battery fire per unit capacity. NORTHR can be
lithium-ion batteries. In contrast, the NORHRR curves for the LMO used to calculate the total heat released during the combustion of a
batteries displayed distinct patterns. For instance, the NORHRR peak lithium-ion battery pack as follows.
for the 12 Ah LMO battery was notably higher, reaching 0.25 s1,
Eexo ¼ NORTHR  n  U  Q  3600 ð3Þ
compared to around 0.17 s1 for the 36 Ah battery. This variability
in the NORHRR peaks between the LMO batteries was likely attrib- where n represents the number of individual burning cells in the
uted to the intrinsic characteristics of the LMO cells, particularly pack, U is the rated voltage of each cell, and Q is the capacity in
the lower stability of the positive electrode active materials. This Ah of each cell. The heat absorption calculation formula is as
instability resulted in more intense heat release during thermal follows.
runaway due to the accelerated decomposition. Moreover, the
Eendo ¼ C  q  V  ðT  T 0 Þ  1000 ð4Þ
lower-capacity LMO batteries underwent more extensive combus-
tion and decomposition in thermal runaway scenarios, leading to where C is specific heat capacity (J/(g °C)), q is the density of the
higher NORHRR peaks in their larger-capacity counterparts. In con- absorbing substance, V is the required volume (L) of the substance,
trast, the LFP batteries, with decomposition temperatures exceed- and T0 and T are the temperatures before and after heat absorption
ing 700 °C, exhibited superior stability compared to the LMO (°C), respectively. The required volume of fire extinguishing fluid
batteries. Their combustion primarily involved electrolyte decom- can be calculated using the following formula based on the type
position, which was relatively consistent in composition and quan- and number of lithium-ion batteries undergoing thermal runaway,
tity across the different capacities, thus resulting in uniform ensuring that the exothermic energy (Eexo) equals the endothermic
NORHRR peaks for LFP batteries. energy (Eendo).
Fig. 4(b) shows a comparative bar graph of the NORTHR values
NORTHR  n  U  Q  3:6
for various capacities of the LFP and LMO batteries. The graph illus- V¼ : ð5Þ
C  q  ðT  T 0 Þ
trates that, within the examined battery capacity range, the LFP
batteries exhibited marginally lower NORTHR values than the The NORTHR methodology is proved to be an effective tool for
LMO batteries, yet the values fluctuated within a narrow band. This calculating the required volume of fire extinguishing agents for
observation implied that the NORTHR values for lithium-ion batter- cooling lithium-ion batteries during thermal runaway scenarios.
ies, regardless of their type or capacity, can be approximately Specifically, this study focused on water-based extinguishing
regarded as constant. This approximation is significant for formu- agents, considering their primary component, water. The specific
lating quantitatively fire extinguishing strategies for E-bicycle heat values of these agents were set at 4.1868 J/(g °C), equal to
battery assemblies. the standard specific heat of water, and the density was close to
water density of 1 g/cm3 at room temperature. The initial temper-
3.3. Establishment and calculation of NORTHR ature was set to 25 °C, and Eq. (5) facilitated the determination of
the necessary extinguishing fluid volume. The fluid volume was
The incidence of E-bicycle fires is primarily attributed to the calculated for various values of the parameters, such as battery
ignition of lithium-ion batteries, and these fires emit significant capacity (Q), the number of burning cells (n) within a battery pack,
heat. Notably, certain lithium-ion battery materials, such as and the post-heat-absorption temperature (T) of the extinguishing
ternary materials, release oxygen at elevated temperatures. Thus, fluid. It should be noted that the volume of the extinguishing agent
effective fire extinguishment of lithium-ion batteries and their obtained in this calculation is only the minimum value for practical
assemblies requires rapid cooling. This cooling serves a dual pur- applications. In addition, since the calculation is based on the nor-
pose: first, to quickly absorb the heat generated by a malfunction- malization of the combustion THR of a single battery, we can also
ing battery, and second, to prevent heat from spreading to other inversely derive the number of burning cells from the fire extin-
cells within the battery pack due to high temperatures. This study guishing results.
delved into the use of aqueous fire extinguishing agents for Through practical application and adjustment of Q, n, and T, the
lithium-ion battery packs, informed by the detailed study of their amount of extinguishing agent needed under various burning
combustion characteristics. conditions was efficiently quantified. Fig. 5(a and b) shows three-
The analysis of the combustion properties of lithium-ion batter- dimensional representations of the volume of extinguishing fluid
ies discussed above indicated that the NORTHR metric is instrumen- (different colored dots) required for LMO and LFP battery packs
614
Z. Yu, X. Shen, R. Xu et al. Journal of Energy Chemistry 91 (2024) 609–618

Fig. 5. Calculation of fire extinguishing agent dosage using NORTHR method. (a) LMO battery and battery pack. (b) LFP battery and battery pack (n indicates the number of
fully burned batteries in the battery pack).

under different scenarios. The graph forms a three-dimensional LMO battery pack closely aligned with or was marginally less than
coordinate system with variables n, T, and Q, representing the vol- the combustion duration of a single 36 Ah cell, as shown in Fig. 2.
ume of extinguishing liquid required to extinguish a battery pack Fig. 7 illustrates the results of the temperature monitoring
fire, which is illustrated in the graph by the sizes and colors of 100 mm from the battery samples during their overcharging and
the circular points. These graphs show a rapidly increasing trend subsequent ignition. The data from Fig. 7(a and c) reveal that the
in the required extinguishing fluid volume with the rise in the LMO battery pack commenced smoking at 1268 s and ignited at
number of burning batteries and battery capacity. Additionally, 1322 s, resulting in a rapid spike in all sensor temperatures above
an inverse relationship existed between the required fluid volume 100 °C. Following the initiation of the water-based extinguishing
and the post-absorption temperature of the extinguishing fluid. process at 1329 s, the flames were extinguished within 130 s, with
Based on the comparison of the results shown in Fig. 5(a and b), a significant drop in all monitored temperatures to below 100 °C,
LMO battery packs demanded approximately 20%–30% more extin- consuming approximately 21.7 L of extinguishing fluid.
guishing agent than LFP battery packs under the same conditions. Conversely, as depicted in Fig. 7(b and d), the LFP battery pack
started smoking at 1213 s and ignited at 1233 s. The extinguishing
3.4. Verification and analysis of NORTHR fire extinguishing technology agent was deployed as the temperatures approached their peak
(1244 s), successfully extinguishing the flames within 112 s using
In an effort to replicate the ignition process of E-bicycle battery 18.7 L of fluid. Notably, both battery types reached peak flame tem-
packs during charging scenarios, a specialized fire extinguishing peratures of about 600 °C. However, for the LMO battery pack, the
box was developed, mirroring the dimensions of a charging com- monitored temperatures continued to rise, even after the extin-
partment. This setup was utilized for conducting fire extinguishing guishing spray was activated, only beginning to decline after
experiments on both LFP and LMO battery packs. It is important to reaching peak values for several seconds. This observation under-
emphasize that high-volume fire extinguishing liquid spray signif- scored the extreme severity of the thermal runaway and combus-
icantly enhances the extinguishing effect, but it also results in con- tion of the LMO battery pack, with a limited amount of
siderable waste. Therefore, a spray rate of 10 L/min was adopted in extinguishing fluid proving insufficient for immediate fire suppres-
the experiment, as recommended by GB 50219-2014 ‘‘Technical sion. In contrast, after the extinguishing spray application, the LFP
Specifications for Water Spray Extinguishing Systems”, which also battery pack exhibited a rapid temperature decrease. While the
facilitated the analysis of the extinguishing process. Fig. 6 shows right-side temperature continued to rise due to the substantial
photographs of the instances of thermal runaway in these battery gas generation and ongoing combustion of the LFP battery, temper-
packs induced by overcharging within the fire extinguishing box. atures on the other sides dropped below 100 °C. This behavior indi-
As detailed in Fig. 6, the LMO battery pack commenced smoking cated a relatively milder and more controlled combustion process
after overcharging for at 1268 s, 54 s longer than the time recorded in the LFP battery pack, effectively contained by the extinguishing
for the LFP battery pack. This was primarily because, for the same fluid and preventing further spread. These findings are consistent
capacity, the LFP pouch cell produced a larger amount of gas than with the stark differences in the NORHRR values between LMO
the other batteries, which made them more likely to cause rupture and LFP batteries, as previously shown in Fig. 4(a).
and gas escape during thermal runaway. The LMO battery pack In this experimental approach, both the LMO and LFP battery
subsequently ignited at 1322 s, exhibiting a lower threshold for packs employed had individual cell capacity of 30 Ah. Drawing
thermal instability than the LFP pack, which did not ignite sponta- upon the extinguishing fluid volume predictions for LMO and LFP
neously but manually ignited at 1233 s. Both battery types rapidly battery packs from Fig. 5, we estimated the potential extinguishing
escalated into intense flames, engulfing the entire fire extinguish- agent requirements for these battery pack configurations. As
ing box and leaking flames externally. shown in Fig. 7, when extinguishing efforts ceased, the tempera-
Upon the onset of ignition, immediate extinguishing efforts tures around the sample were mostly in the range of 60–80 °C.
were launched. The LMO battery pack presented a more formidable As depicted in Fig. 8, extinguishing a single 30 Ah LFP battery fire
challenge in fire suppression, requiring 137 s to extinguish, 14 s to a temperature below 60 °C would require approximately
more than the LFP battery pack. This finding indicated the compar- 14.15 L of water-based extinguishing fluid, while an LMO battery
atively more difficult nature for extinguishing LMO battery fires. would need 17.73 L. In scenarios involving thermal spread across
From a numerical perspective, the flame-burning duration for the 15-cell battery packs, the LFP pack would require 212.26 L, and
615
Z. Yu, X. Shen, R. Xu et al. Journal of Energy Chemistry 91 (2024) 609–618

Fig. 6. E-bicycle battery fire extinguishing process. (a) LMO battery pack. (b) LFP battery pack.

Fig. 7. Temperature monitoring 100 mm from the sample during the battery pack fire extinguishing process. (a) LMO battery pack. (b) LFP battery pack. (c and d) The
enlarged views of the fire extinguishing process corresponding to (a) and (b), respectively.

the LMO pack would need 265.88 L of extinguishing fluid to However, during the actual extinguishing process of the LMO
achieve temperatures below 60 °C. Moreover, the calculations and LFP packs, the volumes of consumed extinguishing fluid were
revealed a notable reduction in the required volume of extinguish- only 21.7 and 18.7 L, respectively, which were significantly
ing fluid as the temperature increased after absorbing heat. For different from the volumes required to extinguish the entire packs
instance, in the case of 15 thermal-runaway LMO batteries, if the calculated using the NORTHR method. The reason for this discrep-
extinguishing fluid’s temperature post-heat absorption climbed ancy was primarily that the uncontrolled burning of batteries
to 80 °C, only 169 L would be required, corresponding to a 38% within a pack did not occur simultaneously but sequentially. Once
decrease in volume compared to that of 60 °C. the flames were extinguished and the temperature dropped, the

616
Z. Yu, X. Shen, R. Xu et al. Journal of Energy Chemistry 91 (2024) 609–618

Fig. 8. Calculation results using NORTHR method for the amount of fire extinguishing agent used to extinguish thermal runaway fires with different numbers of batteries to
different temperatures. (a) LMO battery & battery pack. (b) LFP battery & battery pack.

extinguishing spray operation was promptly halted, meaning that CRediT authorship contribution statement
not all samples had experienced thermal runaway ignition at that
time. Based on the combustion times of the individual 36 Ah LMO Yanyan Fang: Conceptualization, Resources, Writing — review
and LFP battery from Fig. 2(a) and the NORTHR-based calculations and editing, Supervision, Project administration. Xiaoli Ma: Con-
in Fig. 5, it was inferred that after extinguishing, only about two ceptualization, Resources, Project administration. Ran Xu: Concep-
cells in each battery pack were actively burning in the extinguish- tualization, Resources. Zhanglong Yu: Methodology, Validation,
ing experiments. This also provides a method for calculating the Investigation, Visualization, Data curation, Writing — original draft
number of burning batteries in actual fire incidents. A significant preparation, Writing — review and editing. Xueling Shen: Method-
limitation, however, was that the extinguishing fluid was unable ology, Investigation, Resources, Data curation, Writing — original
to fully cool the samples post-flame due to the casing protection draft preparation. Yi Cui: Methodology, Validation, Visualization.
of the battery pack, resulting in ongoing thermal propagation and Zengming Wan: Validation, Writing — original draft preparation.
substantial white smoke emission after the fire was extinguished, Mingyang Chen: Validation, Supervision. Zheng Wang: Investiga-
as shown in Fig. 6. This factor hindered the precise determination tion, Visualization, Supervision.
of the number of actively burning cells at the time of flame extin-
guishment. Nonetheless, the study thoroughly demonstrated Declaration of competing interest
through NORTHR calculations that suppressing and containing the
flames in E-bicycle battery packs can substantially reduce extin- The authors declare that they have no known competing finan-
guishing fluid consumption and enhance the effectiveness of the cial interests or personal relationships that could have appeared
fire suppression process. to influence the work reported in this paper.

Acknowledgments

4. Conclusions This work was supported by the New Energy Vehicle Power Bat-
tery Life Cycle Testing and Verification Public Service Platform Pro-
This research underscores the marked disparities in the ject [2022-235-224], the Beijing Science and Technology Planning
combustion characteristics of LMO and LFP lithium-ion batteries, Project [Z221100005222004], the Key Technologies Research and
providing essential insights for mitigating and managing fires in Development Program [2021YFB2012504], and the Beijing Golden-
E-bicycle batteries. In contrast to the prolonged and relatively sub- bridge Project [ZZ2023002].
dued combustion of LFP batteries, LMO batteries exhibited more
vigorous burning, characterized by greater flames heights and tem-
Appendix A. Supplementary material
peratures. The study affirmed the efficacy of water-based extin-
guishing agents for both battery types, though it emphasized the
Supplementary material to this article can be found online at
need for optimizing the volume and application techniques of
https://doi.org/10.1016/j.jechem.2024.01.005.
the extinguishing agent in accordance with the specific battery
type and burning properties of the battery. The NORTHR method,
predicated on normalizing the THR energy of a lithium-ion battery, References
introduces an innovative approach for calculating fire extinguish- [1] T. Kim, W. Song, D.-Y. Son, L.K. Ono, Y. Qi, J. Mater. Chem. A 7 (2019)
ing requirements. This method promises to effectively inform 2942–2964.
real-world applications, facilitating more efficient and effective fire [2] J. Xie, Y.-C. Lu, Nat. Commun. 11 (2020) 2499.
[3] C.P. Grey, D.S. Hall, Nat. Commun. 11 (2020) 6279.
prevention strategies. Looking ahead, future research should delve
[4] L. Li, B. Liu, W. Zheng, X. Wu, L. Song, W. Dong, Case Stud. Therm. Eng. 37
into an expanded range of lithium-ion battery types and larger (2022) 102304.
capacity, particularly examining their combustion behaviors and [5] W. Liu, H. Liu, W. Liu, Z. Cui, Renew. Sustain. Energy Rev. 139 (2021) 110596.
extinguishing responses under extreme conditions. Such investiga- [6] L. Stilo, D. Segura-Velandia, H. Lugo, P.P. Conway, A.A. West, Transp. Res.
Interdiscip. Perspect. 10 (2021) 100347.
tions are pivotal for advancing battery technologies that are both [7] Grand View Research, E-Bikes Market Size, Share & Trends Analysis Report By
safer and more efficient. Propulsion Type (Pedal-Assisted, Throttle-Assisted), By Battery Type, By Power,

617
Z. Yu, X. Shen, R. Xu et al. Journal of Energy Chemistry 91 (2024) 609–618

By Application, By Region, And Segment Forecasts, 2023–2030. https:// [22] D.P. Kong, G.Q. Wang, P. Ping, J. Wen, eTransportation 12 (2022) 100157.
www.grandviewresearch.com/industry-analysis/e-bikes-market-report, 2022 [23] Y.H. Chen, Ionics. 28 (2022) 495–514.
(accessed 8 December 2023). [24] J. Liu, Z. Wang, J. Bai, T. Gao, N. Mao, Appl. Therm. Eng. 212 (2022) 118565.
[8] He Qitong, Why China’s E-Bike Market Is Facing a Wave of Battery Fires. [25] X. Cao, J. Du, C. Qu, J. Wang, R. Tu, J. Energy Storage 75 (2024) 109661.
https://www.sixthtone.com/news/1013638, 2023 (accessed 8 December [26] J. Sun, B. Mao, Q. Wang, Fire Saf. J. 120 (2021) 103119.
2023). [27] S. Liu, T. Ma, Z. Wei, G. Bai, H. Liu, D. Xu, Z. Shan, F. Wang, J. Energy Chem. 52
[9] CPSC, E-Bikes Recalled Due to Fire, Explosion and Burn Hazards; Distributed by (2021) 20–27.
Ancheer. https://www.dhses.ny.gov/e-bikes-recalled-due-fire-explosion-and- [28] X. Meng, K. Yang, M. Zhang, F. Gao, Y. Liu, Q. Duan, Q. Wang, J. Energy Storage
burn-hazards-distributed-ancheer, 2022 (accessed 8 December 2023). 30 (2020) 101532.
[10] C. Shuai, F. Yang, W. Wang, J. Shan, Z. Chen, X. Ouyang, iScience 26 (2023) [29] Q.S. Wang, P. Ping, X.J. Zhao, G.Q. Chu, J.H. Sun, C.H. Chen, J. Power Sources 208
105786. (2012) 210–224.
[11] W. Ying, Z. Yongping, X. Fang, X. Jian, in: 22017 IEEE Int. Conf. Comput. Sci. [30] X.N. Feng, M.G. Ouyang, X. Liu, L.G. Lu, Y. Xia, X.M. He, Energy Storage Mater.
Eng. CSE IEEE Int. Conf. Embed. Ubiquitous Comput. EUC, IEEE, Guangzhou, 10 (2018) 246–267.
China, 2017, pp. 760–762. [31] P. Liu, S. Li, K. Jin, W. Fu, C. Wang, Z. Jia, L. Jiang, Q. Wang, Fire Technol. 59
[12] L. Zhang, K. Jin, J. Sun, Q. Wang, Fire Technol. 58 (2022) 1–42. (2023) 1051–1072.
[13] Y. Chen, Y. Kang, Y. Zhao, L. Wang, J. Liu, Y. Li, Z. Liang, X. He, X. Li, N. Tavajohi, [32] M. Ghiji, V. Novozhilov, K. Moinuddin, P. Joseph, I. Burch, B. Suendermann,
B. Li, J. Energy Chem. 59 (2021) 83–99. G. Gamble, Energies 13 (2020) 5117.
[14] J. Liu, Z. Wang, J. Bai, J. Energy Chem. 70 (2022) 531–541. [33] H.L. Sun, L. Zhang, Q.L. Duan, S.Y. Wang, S.J. Sun, J.H. Sun, Q.S. Wang, Process
[15] T. Shan, Z. Wang, X. Zhu, H. Wang, Y. Zhou, Y. Wang, J. Zhang, Z. Sun, J. Energy Saf. Environ. Prot. 167 (2022) 299–307.
Chem. 72 (2022) 241–257. [34] L. Zhang, Y. Li, Q. Duan, M. Chen, J. Xu, C. Zhao, J. Sun, Q. Wang, J. Energy
[16] X. Feng, D. Ren, X. He, M. Ouyang, Joule 4 (2020) 743–770. Storage 32 (2020) 101801.
[17] S. Bi, Z. Yu, S. Fang, X. Shen, Y. Cui, F. Yun, D. Shi, M. Gao, H. Zhang, L. Tang, X. [35] Y. Hua, X. Liu, S. Zhou, Y. Huang, H. Ling, S. Yang, Resour. Conserv. Recycl. 168
Zhang, Y. Fang, X. Zhang, J. Energy Storage 73 (2023) 109039. (2021) 105249.
[18] F.N. Jiang, S.J. Yang, X.B. Cheng, P. Shi, J.F. Ding, X. Chen, H. Yuan, L. Liu, J.Q. [36] M. Mirzaei Omrani, H. Jannesari, Renew., Sustain. Energy Rev. 116 (2019)
Huang, Q. Zhang, J. Energy Chem. 72 (2022) 158–165. 109413.
[19] Y. Cui, D. Shi, Z. Wang, L. Mou, M. Ou, T. Fan, S. Bi, X. Zhang, Z. Yu, Y. Fang, [37] C. Xu, Q. Dai, L. Gaines, M. Hu, A. Tukker, B. Steubing, Commun. Mater. 1 (2020)
Batteries 9 (2023) 438. 99.
[20] E. Zhitao, H. Guo, G. Yan, J. Wang, R. Feng, Z. Wang, X. Li, J. Energy Chem. 55 [38] N. Wei, M. Li, Front. Energy Res. 11 (2023) 1230429.
(2021) 524–532. [39] P. Huang, Q. Wang, K. Li, P. Ping, J. Sun, Sci. Rep. 5 (2015) 7788.
[21] Z. Wang, J. Yuan, X. Zhu, H. Wang, L. Huang, Y. Wang, S. Xu, J. Energy Chem. 55 [40] M. Chen, Y. He, C. De Zhou, Y. Richard, J. Wang, Fire Technol. 52 (2016)
(2021) 484–498. 365–385.

618

You might also like