Influence of Aging on Thermal Runaway Behavior of Lithium-Ion Batteries: Experiments and Simulations for Engineering Education

Abstract

This study investigates the impact of aging on the thermal runaway behavior of lithium-ion batteries. By combining external heating tests, cone calorimetry experiments, and numerical simulations, the thermal runaway characteristics of LFP and NMC batteries at different SOH levels (100%, 90%, 80%) were systematically evaluated. Experimental results show a non-monotonic effect of aging on thermal runaway: mildly aged batteries (90% SOH) exhibited the earliest TR trigger and highest risk due to unstable SEI film growth, while new batteries (100% SOH) released the most energy. Significant differences were observed between battery chemistries: LFP batteries displayed fluctuating temperature curves indicating a staged buffering mechanism, whereas NMC batteries had smooth heating but abrupt energy release. Cone calorimeter tests revealed that aged LFP batteries had multi-stage HRR curves, while NMC batteries showed consistent HRR profiles; mass loss data confirmed reduced active material consumption with aging. Numerical simulations integrating SEI decomposition and other reactions validated the impact of aging on internal processes. The study recommends prioritizing monitoring of moderately aged batteries, optimizing early-warning systems for NMC batteries, and preventing secondary explosions, providing support for safety assessments of aged batteries.

1. Introduction

The versatility and strategic importance of lithium-ion batteries (LIBs) have profoundly shaped modern energy utilization, industrial structures, and the global competitive landscape [1]. Spanning applications from smartphones to space exploration and from urban transportation to national energy security, LIBs have emerged as a core technology underpinning three global transitions: digitalization, electrification, and decarbonization [2]. Widespread adoption began in the 2010s, entering a period of rapid development post-2015. The total cumulative shipments are estimated to exceed 1000 GWh, with existing stock approximated at 3500 GWh. Given an average lifespan of about 8–10 years for power batteries and 3–5 years for consumer batteries, retired batteries currently account for about 3% of the total volume, a figure projected to increase rapidly in the near future.

Aging exacerbates the safety concerns associated with LIBs. The primary causes include (1) mechanical abuse, typically referring to the destruction of the internal structure due to external forces like excessive compression or collision [3,4]; (2) electrical abuse, which generally involves overcharging or short circuits during charging/discharging and may cause damage [5,6]; and (3) thermal abuse, referring to exposure to temperatures far beyond the optimal operating range, where excessively low temperatures can also induce irreversible damage [7,8]. These three causes are often interrelated; for instance, compression or collision can damage the internal structure, potentially resulting in short circuits, excessive heat generation, and ultimately, thermal runaway (TR) [9,10].

Thermal runaway (TR) involves a series of exothermic reactions that generate heat, toxic gases, and explosive particulate matter [11,12,13,14]. The primary reactions include the initial decomposition of the solid electrolyte interphase (SEI), reactions between the graphite anode and electrolyte, reactions between the cathode and electrolyte, decomposition and combustion of the electrolyte, and separator melting, among others. The Accelerating Rate Calorimeter (ARC) is considered the most prevalent method for TR testing, as it maintains near-adiabatic conditions [15,16]. Additionally, TR is induced by many researchers using methods such as heater coils or hot plates [17,18]. A TR event in a single cell can potentially propagate throughout a battery module, leading to a large-scale failure [19,20].

As the application scope of LIBs expands, the number of safety incidents has increased annually, leading to significant economic losses and posing serious risks to human life [21,22,23,24]. Recent incidents involving thermal runaway of aged LIBs are listed in

Table 1. Recent incidents involving thermal runaway of aged lithium-ion batteries.

Date	Incident Description	Casualties/Losses
23 February 2024	Major fire accident in Jiangsu Province in China caused by thermal runaway of large-format LIBs.	15 deaths, 2 serious injuries
April 2023	Fire triggered by LIB thermal runaway due to overcharging, involving exothermic reactions on electrodes and internal short circuits by Li dendrites.	No casualties
16 April 2022	Explosion at an electrochemical energy storage station in Beijing South Fourth Ring Road of China, involving 25 MWh LFP batteries, potentially due to aging, environmental, installation, and product factors.	3 deaths, 1 injury
13 June 2023	Fire at a warehouse in Lanzhou City, Gansu Province, China, storing approximately 200 tons of waste LIBs; fire area about 200 m2	No casualties
16 April 2021	Fire in a battery cabinet in Sichuan, China; direct cause was internal short circuit leading to thermal runaway in LFP batteries.	No casualties

.

This study aims to provide a comprehensive understanding of the impact of aging on thermal runaway in lithium-ion batteries. To this end, two different types of lithium-ion batteries were aged, and thermal runaway experiments were conducted on three batteries at State of Health (SOH) levels of 100%, 90%, and 80%. Their safety behaviors were compared through external heating tests [25,26]. Subsequently, cone calorimetry was employed to analyze in depth the fire-induced heat release rate, to identify the influence of both aging and battery chemistry on thermal runaway. Finally, simulation analysis was conducted to elucidate the effect of aging on the progression of internal reactions during thermal runaway [27].

2. Experimental and Numerical Model Design

2.1. Physical Experiment Design

The test samples are 26,650 LFP- and 18,650 NMC-type LIBs (

Table 2. Basic battery parameters.

Parameter	LFP Battery	NMC Battery
Capacity (Ah)	2.3	3.3
Weight (g)	77 ± 1	49 ± 1
Size (mm)	26 × 65	18 × 65

). A capacity tester, heating mantle, potentiostat, K-type thermocouples, vent valve, data acquisition system, and camera were utilized in the experiment.

2.1.1. Aged LIB Module for Experimental Design

To comprehensively assess thermal runaway risk, a common method is direct heating using a high-temperature oven. A simpler experimental setup involves wrapping the battery with a heating mantle or bottom heater. The heater power is adjusted during the heating process; appropriate power settings for observing TR can be guided by reference values from the literature [28].

After the heating power is determined and the setup is assembled, aging is induced in LIBs through standard charge–discharge cycles. The capacity loss, monitored using a capacity tester, serves as the aging criterion.

The heating mantle or bottom heater are used to gradually raise the temperature while the power data curve is recorded. The heating power can be regulated by a potentiostat. Both sudden voltage drop and rapid temperature rise (dT/dt > 2 K/s) are recognized as criteria for thermal runaway onset [29]. After a new battery is successfully tested, the same procedure is applied to aged batteries, with the observation that the thermal runaway point typically occurs at an earlier stage in aged batteries.

2.1.2. Heat Transfer Module for Aged LIB Experimental Design

The second module is focused on thermal measurement, as shown in Figure 1. Bottom-side heating was employed in physical experiments, while heating collars were employed for model validation experiments. K-type thermocouples were placed on the positive and negative terminals and surrounding areas to monitor the temperature distribution in real time through a data acquisition system (Smacq M2101), with a temperature error within ±1 K. A multi-channel data acquisition system was used to record temperature and power curves simultaneously, employing low-pass filtering to reduce noise. All equipment was placed inside a sealed explosion-proof chamber equipped with a Gore-Tex vent valve to release gases while blocking flames and solid ejecta. A copper drain pan at the bottom directs ejecta outside the chamber. An infrared camera monitors battery deformation and rupture throughout the experiment. A nitrogen injection system is activated upon thermal runaway to suppress secondary combustion risks.

A cone calorimeter (MOTIS CCT) was employed to analyze lithium-ion batteries with varying degrees of aging; Figure 2 shows a photograph of the actual equipment. The heat release rate (HRR) of the aged batteries was determined based on the oxygen consumption principle. This methodology provides a deeper insight into the impact of aging on battery thermal runaway. Furthermore, the results obtained can be cross-validated with temperature profiles from simpler experimental setups. This approach not only enhances the reproducibility of the findings but also enables a more comprehensive assessment of how aging affects lithium-ion battery thermal runaway.

2.2. Simulation Model for Aged LIBs

Establishing a physical model based on thermal runaway theory can significantly enhance understanding of this phenomenon in lithium-ion batteries. Although discrepancies are observed between simulations and experiments, analysis of the sequence of internal reaction processes remains highly beneficial for elucidating the mechanisms of thermal runaway. Given the current lack of models focusing on thermal runaway in aged batteries, a comparison of different models is presented in

Table 3. Comparison of thermal runaway models for fresh and aged lithium-ion batteries.

Model	Highlight	Limitations
Lumped heat generation model [30]	The model does a good job of simulating the complete thermal runaway process using a three-stage approach based on experiments	It fails to accurately describe the impact of internal reactions. Additionally, because the three-stage model is uncommon, parameters must be re-determined for batteries with different degrees of aging
SOH model [31]	This model uses internal reactions to simulate thermal runaway in batteries with varying State of Health (SOH)	It requires experimentally measuring parameters for each specific aging state
SEI growth model [32]	This model uses internal reactions based on aging mechanisms to simulate thermal runaway in aged batteries, showing innovative potential	Its poorer performance for severely aged batteries (SOH < 80%)

. Based on this analysis, the third model was consequently selected as an auxiliary tool.

The model is based on the A123 26,650 cylindrical cell. Neglecting external factors such as crush or penetration, thermal runaway typically results from a continuous increase in the heat generation rate within a single battery during storage, while heat dissipation is insufficient to maintain balance due to material and spatial constraints [33,34]. The rising temperature accelerates internal reactions, such as reactions between the electrolyte and active materials, and electrolyte decomposition, leading to uncontrolled temperature increase and thermal runaway.

This model incorporates the exothermic reactions of the battery at high temperatures, including the decomposition of the solid electrolyte interphase (SEI), the reaction between the negative electrode and electrolyte, the reaction between the positive electrode and electrolyte, and decomposition of the electrolyte. The total heat generation rate $Q$ (W/m³) is given by

$$Q=Q_{\mathrm{SEI}}+Q_{\mathrm{ne}}+Q_{\mathrm{pe}}+Q_{\mathrm{e}}$$

(1)

where $Q_{\mathrm{SEI}}$ is the heat generation rate per unit volume for SEI decomposition (W/m³), $Q_{ne}$ is that for the negative electrode reaction (W/m³), $Q_{\mathrm{pe}}$ is that for the positive electrode reaction (W/m³), and $Q_e$ is that for electrolyte decomposition (W/m³). The heat generation model is constructed in

Table 4. Side reaction mechanisms in lithium-ion batteries.

Mechanism	Equation	No.
SEI Decomposition	$Q_{\mathrm{SEI}}=H_{\mathrm{SEI}}\cdot W_{\mathrm{ne}}\cdot R_{\mathrm{SEI}}$	(2)
	$R_{\mathrm{SEI}}=A_{\mathrm{SEI}}\cdot \exp (-E_{\mathrm{SEI}}/R/T_1)\cdot c_{\mathrm{SEI}}$	(3)
Negative Electrode Reaction	$Q_{\mathrm{ne}}=H_{\mathrm{ne}}\cdot W_{\mathrm{ne}}\cdot R_{\mathrm{ne}}$	(4)
	$R_{\mathrm{ne}}=A_{\mathrm{ne}}\cdot \exp (-E_{\mathrm{ne}}/R/T_1)\cdot \exp (-t_{\mathrm{SEI}1}/t_{\mathrm{SEIref}})\cdot c_{\mathrm{ne}}$	(5)
Positive Electrode Reaction	$Q_{\mathrm{pe}}=H_{\mathrm{pe}}\cdot W_{\mathrm{pe}}\cdot R_{\mathrm{pe}}$	(6)
	$R_{\mathrm{pe}}=A_{\mathrm{pe}}\cdot c_{\mathrm{pe}}\cdot (1-c_{\mathrm{pe}})\cdot \exp (-E_{\mathrm{pe}}/R/T_1)\cdot c_{\mathrm{pe}}$	(7)
Electrolyte Decomposition	$Q_{\mathrm{e}}=H_{\mathrm{e}}\cdot W_{\mathrm{e}}\cdot R_{\mathrm{e}}$	(8)
	$R_{\mathrm{e}}=A_{\mathrm{e}}\cdot \exp (-E_{\mathrm{e}}/R/T_1)\cdot c_{\mathrm{e}}$	(9)
Aging Reaction	$t_{\mathrm{SEI}1}=t_{\mathrm{SEI}}{\delta }_{\mathrm{SEI}}/{\delta }_{\mathrm{SEI}0}$	(10)
	${\delta }_{SEI}={\delta }_{SEI0}+C_{loss}{M}_{SEI}/{\rho }_{SEI}/S_n/2F$	(11)
	$S_n=3A\epsilon {\delta }_n/R_n$	(12)

as follows:

Unlike traditional side reaction models, the negative electrode reaction incorporates the variables $t_{SEI1}$ and $t_{SEIref}$, which represent the increase in SEI film thickness. Those are derived by comparing SEI thickness ${\delta }_{SEI}$ to its initial thickness ${\delta }_{SEI0}$ combined with $t_{SEI0}$. The variable $t_{SEIref}$ imposes a limit on SEI growth, typically set to 1 under normal conditions.

3. Experimental Implementation of the Module

The experiment utilized 18,650-type lithium nickel manganese oxide (NMC) batteries and 26,650-type lithium iron phosphate (LFP) batteries. Thermal abuse at elevated temperatures is one of the primary triggers of lithium-ion battery thermal runaway. The heater power was set to 1000 W. The heater was activated to heat the battery until thermal runaway occurred, at which point the heater was promptly switched off. During the experiment, the battery temperature was recorded in real time, and physical changes on the battery surface were observed. Batteries with State of Health (SOH) of 100%, 90%, and 80% for both battery types were heated to a target temperature of 700 K. The specific experimental procedure was as follows:

(1)

Attach a temperature sensor to the lithium-ion battery and place it in the test chamber.

(2)

Activate the ventilation system and start the temperature data acquisition unit.

(3)

Set the heater to the target temperature and begin heating until battery failure occurs.

(4)

Record the battery temperature and observe external physical changes throughout the thermal runaway process.

(5)

Upon finishing the experiment, implement appropriate fire safety measures and clean the test platform with alcohol.

(6)

Analyze the collected data to determine the battery temperature profile.

A heat flux of 40 kW/m² was applied in the cone calorimeter tests. The test samples were, again, 18,650 NMC (capacity is 3.3 Ah) and 26,650 LFP batteries (capacity is 2.3 Ah). The samples did not require special fixtures to minimize the impact of any external pressure on the experimental results. Prior to testing, the data acquisition system was calibrated to establish a baseline. The experimental steps were as follows:

(1)

Initiate preheating in advance and activate the cone heater to reach 40 kW/m².

(2)

Engage the pump, fan and load cell, and then adjust oxygen concentration to 20.95%.

(3)

Weigh the battery.

(4)

Place the battery, open the insulating shutter beneath the heater, and commence heating until flame extinction.

The simulation experiments employed COMSOL Multiphysics 6.2 to develop a thermal runaway model. The objective was to investigate the thermal behavior of thermal runaway in lithium-ion batteries at different aging states. The model was validated against experimental data, enabling a comparison of temperature variation patterns, internal reaction progression, and heat generation across differently aged batteries. The model construction teaching steps are as follows:

(1)

Preparation: Set the necessary geometric and physical parameters.

(2)

Geometry: Construct a cylinder representing the LIB and add a cylindrical coordinate system.

(3)

Model Definition: Add domain probes to detect changes in physical quantities. Add a nonlocal coupling average to treat the battery as a whole with uniform temperature. Define variables for side reactions and the aging reaction.

(4)

Physics Setup: Select the Solid Heat Transfer physics interface. Add nodes for solid materials, initial values, thermal insulation, heat sources, and heat flux. Select the Global ODE and DAE interface to add reactant concentrations for each side reaction. Parameters are sourced from the literature [32,34], as listed in

Table 5. LIB model parameters [32,34,35].

Parameter	Description	Value	Units
$E_{SEI}$	SEI decomposition activation energy	1.38 × 105	J/mol
$E_{ne}$	Negative-solvent activation energy	1.32 × 105	J/mol
$E_{pe}$	Positive-solvent activation energy	0.99 × 105	J/mol
$E_e$	Electrolyte decomposition activation energy	2.70 × 105	J/mol
$A_{SEI}$	SEI decomposition frequency factor	1.66 × 1015	1/s
$A_{ne}$	Negative-solvent frequency factor	2.50 × 1013	1/s
$A_{pe}$	Positive-solvent frequency factor	2.00 × 108	1/s
$A_e$	Electrolyte decomposition frequency factor	5.14 × 1025	1/s
$m_{SEI}$	Reaction order for SEI decomposition	1	/
$m_{ne}$	Reaction order for electrolyte decomposition	1	/
$H_{SEI}$	SEI decomposition heat release	2.57 × 105	J/kg
$H_{ne}$	Negative-solvent heat release	1.714 × 105	J/kg
$H_{pe}$	Positive-solvent heat release	1.947 × 105	J/kg
$H_e$	Electrolyte decomposition heat release	6.20 × 105	J/kg
$W_{ne}$	Specific negative active content	220	kg/m3
$W_{pe}$	Specific positive active content	520.74	kg/m3
$W_e$	Specific electrolyte content	334.68	kg/m3
$c_{SEI}$	Initial value	0.15	/
$c_{ne}$	Initial value	0.75	/
$c_e$	Initial value	1	/
$M_{SEI}$	The molar mass of the film	0.162	kg/mol
${\rho }_{SEI}$	The molar density of the film	1690	kg/m3
$\epsilon$	The volume fraction of carbon	0.58	/
${\delta }_n$	The negative electrode thickness	3.45 × 10−5	m
${\delta }_{SEI0}$	The thickness of the SEI layer	5 × 10−9	m
A	The geometric area of the negative electrode	0.18	m2
$R_n$	The graphite particle modeling the negative electrode	5 × 10−6	m

.

(5)

Computation Settings: The time unit is set to seconds, with output times defined as range (0, 20, 10,000).

For this model, it was found through further experiments that the temperature data of the batteries with 90% State of Health (SOH) demonstrated a good fit during the process (Figure 3). Unlike the experimental section (which used 1000 W heating), the temperature in this figure was obtained from a 500 W heating test. This is because 1000 W heating significantly reduces experimental time, while the extended thermal runaway duration in the model enables a clearer observation of aging effects on internal reactions. During the experiment, a heating power of 500 W was applied until the opening of the valve occurred, at which point heating was immediately terminated.

4. Simulation Results Discussion

4.1. Thermal Runaway Experiments on Aged Batteries

To comprehensively investigate the impact of aging on the thermal runaway behavior of lithium-ion batteries, thermal runaway tests were conducted on a set of batteries aged through 1C charge–discharge cycling. The experiments employed an external heating method at a temperature of 700 K. Figure 4a indicates a trend: the 90% SOH battery experienced thermal runaway first, reaching the maximum temperature. This was followed by the fresh battery (100% SOH) with a second highest peak temperature, and lastly the 80% SOH battery with the minimum peak temperature. The TR onset temperatures for these batteries, ranked in descending order, were 80% SOH, 90% SOH, and 100% SOH. The aged batteries exhibited a lower onset temperature for thermal runaway compared to the fresh battery. It is noteworthy that the slightly aged (90% SOH) battery underwent thermal runaway at a faster rate than both the fresh and the severely aged (80% SOH) batteries. Furthermore, qualitative observations during the thermal runaway (TR) process of 80% SOH LFP batteries revealed more intense smoke generation, suggesting potentially higher hazards that necessitate quantitative investigation in future work.

Aged NMC batteries exhibited a similar trend to LFP batteries in the thermal runaway tests, with the occurrence sequence being 90% SOH, 100% SOH, and finally 80% SOH batteries. The peak thermal runaway temperatures differed between LFP and NMC, which is attributed to the dominant influence of the battery material itself on the maximum temperature. However, for both battery types across various aging states, the fresh batteries (100% SOH) consistently reached the highest temperatures. This can be explained by the aging process itself, which increases the proportion of thermally stable inorganic compounds and reduces the amount of active material available for exothermic reactions, thereby lowering the maximum temperature.

A comparison of the curves in Figure 4a,b reveals distinct characteristics. The pre-thermal runaway temperature curve for the aged NMC batteries is notably smoother but exhibits a much steeper ascent once the TR onset temperature is reached, indicating a larger instantaneous energy release. This violent energy release explains the severe rupture and complete loss of structural integrity observed in the NMC batteries, whereas the LFP batteries remained relatively intact. The pronounced fluctuations in the LFP battery’s pre-TR curve signify its superior thermal stability. These fluctuations result from a periodic buffering mechanism, where continuous heat generation from SEI decomposition and reactions between the anode and electrolyte is intermittently alleviated by pressure release through the safety vent and electrolyte evaporation. This “accumulation–release” mechanism endows LFP batteries with a longer time to thermal runaway. In contrast, the smooth yet sharply escalating curve of the NMC battery suggests that the onset of SEI decomposition rapidly triggers other exothermic reactions. This chain reaction proceeds irreversibly and accelerates, leading to an abrupt thermal runaway. This implies that a more critical early warning system for thermal runaway is needed for NMC batteries compared to LFP batteries.

Additionally, Figure 5a displays the external structure of the three LFP batteries post-test, arranged from left to right in the order of 90% SOH, 100% SOH, and 80% SOH batteries. Clear observation reveals that the 90% SOH battery exhibited more violent ejection of internal components. In contrast, the negative vent of the 80% SOH battery remained intact, indicating a relatively milder thermal runaway process.

The underlying mechanisms for these observations are as follows. For the 80% SOH battery, the high degree of internal aging led to substantial consumption of active material. Furthermore, a substantial portion of its organic components had been converted into inorganic species with higher thermal stability. This finding was supported by Ref. [36], which analyzed the SEI structure of batteries at different aging stages and found that the SEI layer that formed during early aging stages was structurally unstable, whereas it became more stable as aging progressed.

The fresh battery possessed a highly stable internal structure. However, once thermal runaway was triggered, the internal reactions in the fresh battery were more violent than those in the 80% SOH battery, which caused its internal components to be ejected violently from the negative terminal.

The 90% SOH battery exhibited the poorest thermal stability. This was likely attributable to the loose and disordered structure of the SEI layer formed during the mild aging phase, which compromised the overall mechanical integrity of the internal structure.

Figure 5b shows the severe structural damage and rupture of the NMC battery after testing. This condition provided further evidence for the differences in thermal runaway behavior between the two battery types, as illustrated in Figure 4. Consequently, the specific influence of aging degree on the post-TR structure of NMC batteries could not be observed. Beyond battery chemistry, other factors must be considered. The internal energy and external structure also differed; specifically, the LFP battery used had a capacity 1 Ah smaller than the NMC battery, and its 26,650 format offered a larger surface area for heat dissipation.

4.2. Influence of Aging on the Heat Release Rate of LIBs

Figure 6 illustrates variations in the heat release rate (HRR) for both NMC and LFP batteries during calorimetry experiments. The HRR curves of NMC batteries correlated well with the temperature trends observed from external heating tests. Furthermore, HRR curves remained similar across different aging states of NMC batteries. In contrast, the LFP batteries exhibited distinctly different HRR curves depending on their state of health. The HRR curve of the new LFP battery closely resembled that of the NMC battery, both showing a concentrated heat release peak. The timing of the HRR peaks aligned with the external heating tests, and the sequence of thermal runaway remained the same for both chemistries: 90% SOH, followed by 100% SOH, and then 80% SOH.

The HRR curve for the 90% SOH LFP battery displayed a stepwise decline after reaching its initial peak, indicating phased internal reactions and the intermittent activation of the pressure relief valve. This phenomenon was more pronounced in the 80% SOH LFP battery, which exhibited two distinct HRR peaks. This suggested that the increased proportion of inorganic compounds in aged LFP batteries led to a sequential reaction process during thermal runaway: less thermally stable organic materials reacted first, followed by inorganic compounds that required higher temperatures to decompose.

The mass loss measurements during the calorimetry experiments were 24 g, 20.2 g, and 19.4 g for LFP batteries at 100%, 90%, and 80% SOH, respectively. For the NMC batteries, the mass losses were 21.1 g, 20.4 g, and 20.2 g respectively. Both battery types exhibited a consistent trend of mass loss decreasing with the increasing aging degree. This observation further confirmed the reduction of reactive substances in aged batteries during thermal runaway. The higher peak heat release rate of new batteries following ignition necessitated the development of safer firefighting strategies. Additionally, the secondary explosion observed in 80% SOH LFP batteries (Figure 7) posed significant challenges during fire suppression operations.

4.3. Influence of Aging Degree on LIB Thermal Runaway

In simulations, capacity loss rates of 0–20%, corresponding to 100–80% SOH (100, 95, 90, 85, and 80% SOH), were considered. Simulations were conducted for LIBs with capacity loss rates of 0% (new), 5%, 10%, 15%, and 20%. The results are shown in Figure 8. The simulated battery temperature under external heating was obtained with an initial battery temperature of 293.15 K and a heat transfer coefficient h = 50 W/(m²·K). The figure clearly demonstrates the impact of aging on LIB thermal runaway. Both the time to thermal runaway and the maximum temperature indicated that the battery with 10% capacity loss was the most hazardous, followed by the new battery, while the 80% SOH battery was the safest. This suggests a non-proportional relationship between the degree of aging and the hazard level of LIB thermal runaway.

Figure 9 presents a comparison of the heat generation rates and the degree of reaction for batteries at different aging states. The analysis reveals that the initial rapid temperature rise preceding thermal runaway can be divided into two distinct phases. The first phase was primarily driven by external heating. This transitioned into a second phase where the temperature increase was governed by the combined effects of external heating and internal exothermic reactions, namely solid electrolyte interphase (SEI) decomposition and the negative electrode reaction. During this phase, the battery entered a quasi-steady state with a moderate temperature rise rate, rather than undergoing immediate thermal runaway, which was attributed to the inherent chemical and thermal stability of the lithium-ion battery system. Once the temperature exceeded a critical threshold, the vigorous decomposition of the positive electrode was triggered. The substantial heat released from this reaction drove a rapid temperature escalation, ultimately culminating in thermal runaway.

The heat generation characteristics varied significantly with the degree of aging. From 100% to 90% SOH, the peak heat release rate exhibited a monotonic increase, accompanied by a consistent temporal advancement of the peak. However, a non-monotonic trend emerged between 90% and 80% SOH. Specifically, the battery at 85% SOH demonstrated the latest occurrence of the heat release peak and the lowest magnitude among all the tested cells. The 80% SOH battery exhibited a peak heat release and timing comparable to, though slightly lower and marginally later than, those of the fresh (100% SOH) battery.

The obtained results indicate that the thermal runaway risk of lithium-ion batteries did not increase linearly with the degree of aging but instead peaked at mild aging (90% SOH). This non-monotonic phenomenon was attributed to the evolution of the composition, structure, and mechanical properties of the solid electrolyte interphase (SEI) film at different aging stages. In the fresh cell (100% SOH), the SEI film was thin and relatively stable, with an intact internal structure, which resulted in a high onset temperature for thermal decomposition. For the mildly aged cell (90% SOH), following a certain number of cycles, the SEI film continued to grow and thicken. However, at this stage, the SEI film was typically porous, loose, rich in organic components, and characterized by poor mechanical integrity [36]. This unstable SEI layer acts as an inefficient thermal barrier, which not only impeded the dissipation of internal heat but was also prone to fracture at elevated temperatures. This fracture exposed a significant amount of fresh, highly active anode material to the electrolyte, which triggered intense exothermic reactions. This mechanism explained why the heat generation rate from the anode reaction was the highest and occurred earliest in the 90% SOH cell (Figure 9), consequently leading to the shortest TR triggering time and the most severe intensity. For the deeply aged cell (80% SOH), continuous side reactions had consumed a substantial amount of active lithium and active material, which reduced the available ‘fuel’ for thermal runaway. More importantly, prolonged cycling likely promoted the ‘reconstruction and mineralization’ of the SEI film, leading to the formation of more inorganic, dense, and thermally stable components [36]. Although this more stable SEI film exhibited inferior lithium-ion conductivity, its higher thermal stability and lower propensity for fracture collectively served to mitigate the intensity of thermal runaway. This was manifested in a lower peak temperature and a delayed triggering time compared to the 90% SOH cell. Therefore, the thermal runaway risk was determined by the competition between two opposing effects: the consumption of active material (which reduced risk) and the growth of an unstable SEI film (which increased risk). During the mild aging stage, the latter effect dominated, whereas in the deep aging stage, the former began to prevail, resulting in the observed peak in risk.

5. Conclusions

Lithium-ion batteries offer advantages such as high energy density, convenience, and high efficiency, demonstrating broad application prospects. However, the benefit of high energy density is accompanied by the risk of thermal runaway. Aging from prolonged use increases the probability of thermal runaway, which is extremely hazardous. This study investigated the impact and associated hazards of battery aging on the thermal runaway of lithium-ion batteries through experimental analysis and numerical simulation. The main conclusions are listed below:

(1)

The external heating tests on NMC and LFP batteries revealed distinct thermal runaway curves, which were attributed to differences in their form factor, energy density, and chemistry. The severe rupture of the NMC batteries post-test presented a striking contrast to the structurally intact LFP batteries. Despite these differences, a notable similarity emerged across aging states: mild aging led to an earlier onset of thermal runaway in both battery types. Conversely, the new batteries reached the highest temperatures due to their higher proportion of organic compounds, which promoted more intense reactions.

(2)

The cone calorimeter tests provided deeper insights into the impact of aging on thermal runaway in LFP and NMC batteries. Aligning with the external heating experiments, the heat release rate (HRR) curves of aged LFP batteries showed low similarity, whereas those of NMC batteries maintained high consistency. This divergence arose because thermal runaway in LFP batteries involved safer, staged internal reactions, while in NMC batteries, it was an irreversible chain reaction. Both the mass loss and HRR data across different aging states confirmed an inverse correlation between the aging degree and the amount of active material consumed during thermal runaway.

(3)

To solidify the pedagogical outcomes, simulations were introduced to analyze lithium-ion battery thermal runaway from a microscopic perspective. These models vividly demonstrated the impact of SEI growth on thermal runaway in LFP batteries at different aging states, providing deeper mechanistic insights: the competition between the destabilizing effect of a thickened, unstable SEI layer at moderate aging and the consumption of active material at advanced aging stages.

This study established an integrated approach that combines macroscopic experimental techniques, such as calorimetric analysis and external heating tests, with modeling to elucidate phenomena unobservable through experiments alone. This methodology facilitated a comprehensive investigation into the thermal runaway mechanisms of lithium-ion batteries across different aging states, providing insights for fire safety assessments. Furthermore, the results yielded practical recommendations: temperature monitoring should be prioritized for moderately aged batteries; early-warning systems for NMC batteries should be more sensitive; and the potential for secondary ignition or explosion should be considered during fire suppression of severely aged batteries.