Design and Performance Optimization of Battery Pack with AI-Driven Thermal Runaway Prediction ^†

Abstract

Battery thermal management is a critical factor in ensuring the performance, safety, and longevity of electric vehicle (EV) battery packs. This study investigates the effectiveness of a forced air convection cooling system, optimized cell spacing and suitable configuration in maintaining optimal battery cell temperatures. A 3D computational model was developed to analyze the temperature distribution of a battery pack under varying airflow velocities, cell spacings and configurations. The numerical simulations were validated through experimental testing, demonstrating a strong correlation between simulated and measured results. The findings reveal that with a 2 m/s velocity of the fan, the battery’s maximum temperature is reduced by 7% compared to the case of natural convection, while the fan consumed only 4% of the battery pack available capacity. An AI algorithm was trained on the experimental data obtained to perform data-driven predictions of failures. The results provide valuable insights for optimizing air cooling systems in EV applications. Future work will explore the effect of non-uniform air flow distribution in reducing the risk of thermal runaway and avoiding hot spots in the battery pack for optimal performance.

1. Introduction

Thermal condition management of batteries is critical to lithium-ion (Li-ion) battery performance, safety, and longevity, particularly in high-power applications like electric vehicles (EVs). Due to their high energy density, extended lifespan, and minimal natural decay characteristics, Li-ion batteries function as the main choice for EV power storage systems. Nevertheless, their performance and reliable operation rely heavily on the heat evolved in charge and discharge modes. The literature shows that capacity degradation of Li-ion batteries is accelerated once temperatures exceed 40 °C, with the risk of thermal runaway—a condition where runaway reactions are initiated by overheat—increasing the battery pack temperatures dramatically above 60 °C [1]. Furthermore, even when thermal runaway is prevented, temperature cycling in a battery module leads to non-uniform aging rates in individual cells, ultimately endangering both performance and longevity [2]. Thus, a highly efficient battery thermal management system (BTMS) is essential in minimizing risk of temperature variations and maintaining uniform heat distribution across the battery cells. Different cooling approaches have been explored in battery systems, including liquid cooling, phase-change materials (PCMs), and air cooling, each with strengths and weaknesses. While liquid cooling is very efficient, it is linked with high costs, complexity, and significant weight additions to battery modules, which makes it less attractive for compact electric vehicle (EV) battery designs. Phase-change materials (PCMs) are effective in passive heat absorption; however, they are prone to encounter low thermal conductivity and long-term stability issues [3]. Due to its cost, simplicity, and low energy consumption, air cooling is one of the most widely used BTMS solutions in EV applications [4].

Air cooling is advantageous but also has some major issues, particularly when batteries are being heavily used, resulting in a substantial amount of heating. Natural or forced air is used by air-cooled systems to shed heat, but natural air alone is generally insufficient for high-power battery packs. This can lead to overheating of parts of the batteries and uneven temperatures [5]. Forced air cooling utilizes fans or blowers to facilitate the removal of heat better, and its improvement has been widely explored by research all over the world. Researchers have examined the air flow rates, optimized design of cooling channels, and explored positioning of the fan to improve the performance of an air-cooled BTMS. An uneven airflow can generate different temperatures in battery packs, which can influence its performance [6]. Research indicates that making the air flow can quickly reduce the peak temperatures by as much as 23%, but excessive airflow can consume more power and reduce cooling efficiency, so the optimization of BTMS is required for this trade-off [7].

In addition to the air flow approach, the battery pack design itself also plays a critical role in cooling. The distance between cells, arrangement of cells being stacked in series or in parallel, and material used for the exterior casing all influence the dissipation of the heat. Earlier research indicates that creating more distance between cells can increase cooling by 6.5%, but excess distance can cause battery modules to become larger and more difficult to fit into EVs [3,7]. Also, using outer casing materials which are better suited for heat-conductivity can allow heat to naturally escape, which reduces the need for active cooling strategies. Among the key influences on heat generation within batteries is the discharge rate, which dictates the rate of energy extraction from the battery. Increasing discharge rates promote increased internal resistance losses, resulting in excess heat that needs to be efficiently eliminated to prevent overheating. In previous research, it has been shown that a battery’s surface temperature rises as much as 45% with a change in discharge rate from 1 C to 2 C, corresponding to the excessive thermal stress brought about by high-power demands [8]. While most prior research has highlighted the thermal behavior of batteries for moderate discharge rates (typically less than 1.5 C), relatively limited research has explored the performance of forced air cooling under heavy discharge conditions (≥2 C), where conventional air-cooling techniques may prove inadequate in maintaining safe operating temperatures [1].

In this study, we aim to bridge these research gaps by improving the performance of a fan-cooled air-cooled BTMS under a 1.5 C discharge rate. Our investigation experimentally compares the impact of battery pack configuration, airflow optimization, and casing material selection on cooling performance, particularly in high-power applications where effective heat removal is critical. Unlike previous studies focused primarily on moderate discharge conditions, this study provides real-world assessment of forced air cooling under demanding operating conditions. The findings of this study justify further development of compact, efficient, and cost-effective air-cooled BTMS solutions for future EV battery systems, with improved thermal stability and battery life in high-performance applications.

2. Literature Review

The optimal performance, protection and durability of Li-ion batteries in electric vehicles (EVs) demands effective thermal condition management systems. High temperatures can result in a loss of performance, increased internal resistance, and severe issues like thermal runaway. Lightning and cycling through energy outside the optimum temperature range of 15 °C to 40 °C yields unfavorable results. Research indicates that a temperature shift within a battery pack exceeding 10°C has the potency to accelerate aging processes and cause a decrease in cycle life by close to 50%, which highlights the significance of advanced cooling systems [9].

The designed BTMS utilizes air cooling as a distinctive method because it provides effective cooling while requiring minimal energy expenditure and construction complexity compared to liquid cooling systems. The forced air cooling performs exceedingly well while uniformly dispersing and equalizing the battery pack temperature, with routinely maintaining high discharge rates. Various criteria, for example, the air flow rate, the position of the fan, the configuration of the cooling ditch, and the structure of the battery pack, set adiabatic boundaries of cooling effectiveness. This section covers major developments in air cooling methods, the effect of battery pack design on thermal management, and the effect of high discharge rates on cooling effectiveness.

2.1. Air Cooling Methods for Battery Thermal Control

Air cooling methods can either operate via natural or forced convection. Forced convection cooling is suitable for ease and can be energy efficient; however, it has a low heat transfer coefficient which does not provide sufficient energy for high power output scenarios. Studies have shown that when relying on natural convection in isolation, the battery components become systematically overheated after enduring high discharge currents due to overheating and insufficient temperature diversity [10]. Unlike conventional methods, forced air cooling does not employ blowers or fans; instead, it passively ventilates the rechargeable batteries by actively blowing air over the battery cells to improve overall convection cooling. Forced air cooling design relies on multiple features such as air flow patterns, directions, velocities, and target zones. Several researchers have focused their interests on the perpendicular diverging and converging air stream zones that increase the thermal energy outflow from the component. In battery modules, the fan position and airflow direction of vertical battery cooling modules can have a significant effect [11]. Top–down airflow provides the best results due to a large increase in temperature uniformity. Likewise, analysis of different air inlet and outlet locations confirms that staggered inlets increases the cooling efficiency at the same time when airflow resistance was reduced [12]. Through CFD simulations, an increasing airflow velocity from 2 m/s to 5 m/s resulted in a 23% decrease in peak battery temperatures, thus underlining the significance of optimizing steam and airflow velocity for load thermal management [13].

While forced air cooling has been effective, its design can impact the performance of a BTMS. Improper airflow can cause hotspots and individual battery cells may become much hotter than neighboring cells, which may lead to uneven aging and capacity loss. Thus, the need is to ensure that all battery cells are capable of dissipating heat equally through the integration of optimized ventilation channels and cooling paths within the battery modules.

2.2. Effect of Battery Pack Design on Thermal Management Effectiveness

Battery pack configuration is a critical thermal management aspect as the position and spacing of the cells affect the ease of heat dissipation. Cell-to-cell thermal conduction, air flow circulation, and contact resistance all affect cooling performance, and therefore battery configuration is a key design parameter in air-cooled BTMS design. It has been demonstrated that thermal performance of cylindrical battery module improves by 6.5% when cell spacing is increased from 0 to 4 mm because of better airflow circulation through the cells [14]. However, excessive spacing increments may result in larger module sizes, which could be inadmissible for compact EV battery design. In addition, comparison of different battery pack configurations like honeycomb, rectangular, and staggered arrangements showed that honeycomb arrangement provided the most even temperature distribution due to optimized airflow channels [15]. A study on impact of battery connections in series and in parallel on thermal properties revealed that parallel arrangements provided lower temperature gradients, thereby improving thermal balance within the module [16].

These studies emphasize that battery pack structure needs to be well-engineered in a manner where airflow distribution is maximized, thus improving the cooling performance. Modules that are not well designed tend to have trapped heat between adjacent cells, potentially elevating the risk of thermal runaway. Ventilation paths, cell orientation, and cell-to-cell distance are thus critical components in designing a high-performance air-cooled BTMS.

2.3. Effect of High Discharge Rates on Thermal Regulations

The battery temperature increases directly with the discharge rate because high discharge rates lead to higher internal resistance losses and increased heat generation. At high discharge, the battery cells experience high levels of energy conversion, and this raises thermal stress and the possibility of overheating. The relationship between the temperature increases and the rate of discharge has been studied regarding the effectiveness of air cooling for high-power applications. The surface temperature of the battery rises 45% when transitioning from 1 C to 2 C discharge rate but exhibits a non-linear temperature increase during discharge. Above a discharge rate of 2 C, a simple air cooling design on its own is not adequate due to natural convection [17]. The excessive temperature increases from discharge rates in excess of 2 C need to be controlled by using specially designed air-cooling channels with high flow rates to avoid overheating the system. Furthermore, an assessment of the effect of high discharge rates on the thermal uniformity of battery modules found that patterned higher discharge rates lead to higher imbalance in temperatures of the module cells, generating a higher risk for localized overheating. From the results, an increased air flow speed, better positioned fans, more advanced case structures with high thermal conductivity for heat sinks, and improved cooling systems were suggested to be mandatory for high-discharge applications [18].

Since this work is performed at 1.5 C discharge rate, it is a real-world assessment of forced air cooling under heavy duty operating conditions. Although earlier research has examined each component of air cooling, fewer have examined combined effects of battery pack configuration, airflow strategies, and discharge rate under such severe thermal loads, which is the main contribution of this work. Previously reported works have mostly been performed at lower discharge rates, whereas this research is performed at a 1.5 C discharge rate per cell, thus being one of the few examining forced air-cooling effectiveness under such high-energy demands. By examining the interaction of airflow optimization, battery pack design, and thermal control, this research helps to improve air-cooled BTMS performance in high-power EV applications and provides useful insight into optimizing compact and efficient BTMS.

2.4. Effect of Busbars Material on Battery Thermal Performance

In an EV battery, busbars play a critical role in ensuring efficient power distribution within the battery pack, directly impacting electrical resistance, heat generation, and thermal management. The material, thickness, and length of the busbar significantly affect energy losses, particularly during high-current discharges. Copper (Cu) and aluminum (Al) are widely used in EV battery busbars due to their low resistivity and high conductivity. However, alternative materials such as nickel (Ni), titanium (Ti), stainless steel, and graphene-based conductors are also being explored to enhance performance, reduce weight, and minimize connection resistance issues in high-power battery packs. Busbar thickness and length influence resistance and power dissipation. A longer busbar increases resistance, leading to higher voltage drop and greater heat generation. Conversely, a thicker busbar reduces resistance, improving current-carrying capacity and lowering energy losses. However, in EV applications, increasing thickness adds weight, which affects vehicle efficiency and range. Therefore, the cross-sectional area of the busbar requires optimization because it directly affects conductivity and both weight and heat dissipation rates. The choice of material also affects connection resistance, which arises at the interface between busbars, battery terminals, and interconnects [19]. Poor electrical contact leads to localized heating, accelerating thermal degradation and increasing battery aging. Figure 1 summarizes the key considerations in design of busbars as a current collector for an EV battery.

Figure 2 shows a comparative analysis of copper and aluminum busbars in a 3s3p battery pack (consisting of MKEPA 18650 3.7 V 2550 mAh Lithium-ion ICR cells from China) configuration conducted using MATLAB (R2024a) simulations to evaluate voltage drop, power loss, and temperature rise. The results indicate that aluminum busbars exhibit higher voltage drop, power loss, and temperature rise compared to copper due to their higher electrical resistivity and lower thermal conductivity. As seen in the simulation results, copper busbars effectively reduce resistive losses and maintain a lower operating temperature, making them a superior choice for efficient thermal and electrical management in high-power applications. This study underscores the importance of material selection in optimizing the performance and longevity of battery packs in electric vehicles.

The analysis of busbar materials in a 3s3p square battery configuration reveals significant differences between copper and aluminum in terms of electrical and thermal performance. The voltage drop across the aluminum busbar is noticeably higher compared to copper due to its greater resistivity, leading to increased power losses and reduced efficiency in the battery pack. Additionally, aluminum exhibits a significantly higher temperature rise under the same discharge conditions, as seen in the comparison graph, which can lead to thermal management challenges. While aluminum remains a cost-effective and lightweight alternative, copper is the superior choice for minimizing energy losses and maintaining thermal stability in EV battery systems, making it the preferred option where performance outweighs cost concerns. The detailed properties of different materials to be used as busbars are mentioned in

Table 1. Comparison of different busbar materials.

	Electrical Conductivity (MS/m)	Resistivity (Ω·m × 10−8)	Density (g/cm3)	Thermal Conductivity (W/m·K)	Weight Considerations for EVs	Impact on Heat Dissipation	Suitability for High-Current EVs
Copper (Cu)	58.0	1.68	8.96	401	Heavy, adds weight	Excellent, low heat accumulation	Ideal for high-power EVs
Aluminum (Al)	37.7	2.65	2.70	237	Lightweight, reduces weight	Moderate, needs better cooling	Good for moderate-power EVs
Nickel (Ni)	14.3	6.84	8.90	90	Heavy, corrosion-resistant	Poor, increases battery temperature	Not ideal for high discharge
Titanium (Ti)	2.4	42.0	4.51	21.9	Moderate weight	Poor, increases energy loss	Not suitable for EVs
Stainless Steel	1.4	69.0	7.80	16	Heavy, robust	Very poor, requires advanced cooling	Not recommended for EVs
Carbon-based (Graphene, CNTs)	Variable (10–30)	Variable (3–10)	Low-Moderate	Variable (100–200)	Very light, flexible	Good, reduces hot spots	Promising for next-gen EVs

.

2.5. Thermal Runaway in Lithium-Ion Batteries

The main challenge in lithium-ion battery systems emerges from thermal runaway, which creates a self-sustaining failure process that activates fast exothermic reactions leading to gas venting and fire and explosion risks. This occurs when a battery cell’s internal temperature exceeds a critical threshold, typically between 120 and 150 °C, leading to the uncontrolled breakdown of key battery components. Studies show that once a single cell enters thermal runaway, the heat spreads within milliseconds, increasing the risk of a chain reaction across battery packs [20,21].

AI and machine learning have proven to become essential tools which detect thermal runaway along with its prevention mechanisms. By analyzing real-time sensor data, ML models can identify early warning signs such as unusual temperature spikes, voltage fluctuations, and rising internal resistance. Advanced algorithms like XGBoost and neural networks have demonstrated high accuracy in predicting thermal anomalies before traditional sensors can detect them [22]. AI-powered fault detection systems can trigger pre-emptive cooling measures or isolate failing cells, reducing the chances of failure spreading. Additionally, AI-based state-of-health (SOH) models continuously track battery degradation, allowing for dynamic cooling adjustments to minimize risks [23]. Integrating AI-driven thermal prediction models with real-time BTMS can significantly improve battery safety and reliability. Future advancements, such as embedded AI chips for onboard fault detection and automated response mechanisms, could revolutionize EV battery safety by reducing failure rates and enhancing thermal stability.

3. Methodology

This research implements both battery pack simulation and experimental setup validation to test its findings. The experimental platform intends to evaluate how well air-cooling controls temperature levels of 18650 lithium-ion battery cells.

3.1. Simulation of the Battery Module

This study utilized ANSYS (2024 R1) software to conduct battery module simulation and perform surface temperature distribution analysis. Both simulation results underwent comparison against experimental data that include natural and forced convection measurements. The 3s3p battery pack design along with its casing was designed in SolidWorks (2024 version SP5 release). The CAD design was then imported to ANSYS, and a mesh was generated. Figure 3 shows the isometric view of the battery pack casing and the mesh generated for simulations.

3.2. Governing Equations

While taking u as the air velocity, ρ as the density of air, p as the pressure, μ as the air viscosity, and g_i as the considered body force in natural convection, the continuity and momentum equations can be stated as follows [11]

$${\displaystyle \frac{\partial {\mathrm{u}}_{\mathrm{i}}}{\partial {\mathrm{x}}_{\mathrm{i}}}}=0$$

(1)

$$\mathsf{\rho }{\displaystyle \frac{\partial {\mathrm{u}}_{\mathrm{i}}}{\partial \mathrm{t}}}+\mathsf{\rho }{\mathrm{u}}_{\mathrm{j}}{\displaystyle \frac{\partial {\mathrm{u}}_{\mathrm{i}}}{\partial {\mathrm{x}}_{\mathrm{j}}}}=-{\displaystyle \frac{\partial \mathrm{p}}{\partial {\mathrm{x}}_{\mathrm{i}}}}+\mathsf{\mu }{\displaystyle \frac{{\partial }^2{\mathrm{u}}_{\mathrm{i}}}{\partial {\mathrm{x}}_{\mathrm{j}}^2}}+\mathsf{\rho }{\mathrm{g}}_{\mathrm{i}}$$

(2)

Similarly, the energy equation can be given as follows

$$\mathsf{\rho }{\mathrm{C}}_{\mathrm{p}}{\displaystyle \frac{\partial \mathrm{T}}{\partial \mathrm{t}}}+\mathsf{\rho }{\mathrm{u}}_{\mathrm{j}}{\displaystyle \frac{\partial \mathrm{T}}{\partial {\mathrm{x}}_{\mathrm{j}}}}=\mathsf{\lambda }{\displaystyle \frac{{\partial }^2\mathrm{T}}{\partial {\mathrm{x}}_{\mathrm{j}}^2}}+\dot{\mathrm{q}}$$

(3)

where C_p is the specific heat capacity, λ is the heat generation and the term $\dot{q}$ represents the heat generation rate per unit volume in battery cells, expressed as follows:

$$\dot{\mathrm{q}}={\displaystyle \frac{\mathrm{R}{\mathrm{I}}^2}{\mathrm{V}}}$$

(4)

where R is the equivalent internal resistance of the battery module, I represents the electrical current, and V is the volume of the battery cells.

3.3. Heat Dissipation Model

An increase in airflow velocity enhances heat dissipation efficiency, and the variations in airflow speeds between cooling channels correspond to temperature difference among battery units. Following assumptions have been used in our model:

• Each battery cell follows a simplified assumption of uniform temperature distribution, which allows calculation of average temperatures for each unit [12].
• The physical parameters of battery components are calculated through weighted average methodology [13].
• Heat exchange occurs solely within the cooling channels, simplifying the model by excluding other interactions between the air and the battery [12].

As illustrated in Figure 4, the cooling channel directs incoming air to transport heat produced by the left and right battery unit regions and lowers the battery unit’s temperature.

The governing equations for heat dissipation are as follows [14]:

$${{\varphi }_i - h}_{i}A\left(\Delta T_{left,i}+\Delta T_{right,i}\right)=0$$

(5)

$$C_{p,air} {\rho }_{air} Q_{\mathrm{cc},i}\left(T_{air,i}-T_0\right)=h_{i}A\left(\Delta T_{left,i}+\Delta T_{right,i}\right)$$

(6)

where ${\varphi }_i$ is the thermal power production by the battery unit, $h_i$ is the convective heat transfer coefficient between the cooling air and the surface of the battery unit, $A$ is the contact area between the air and the battery unit, $Q_{\mathrm{c}\mathrm{c},i}$ is the airflow rate, and $\Delta T_{\mathrm{l}\mathrm{e}\mathrm{f}\mathrm{t},i}$ and $\Delta T_{\mathrm{r}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t},i}$ are the temperature difference between the battery unit and both sides of the air. The following logarithmic formulas determine the cooling air temperatures on both sides of the battery unit by measuring the temperature gap between the air and battery unit [14].

$$T_{left,i}={\displaystyle \frac{T_{air,i}-T_0}{ln\left(\frac{T_u,i-T_0}{T_u,i-T_{air,i}}\right)}}$$

(7)

$${ T}_{right,i}={\displaystyle \frac{T_{air,i+1}-T_0}{ln\left(\frac{T_u,i-T_0}{T_u,i-T_{air,i+1}}\right)}}$$

(8)

where $T_{u,i}$, is the volume averaged temperature of the battery unit. The heat transfer coefficient for convection, $h$, is key to relating the flow resistance model to the heat dissipation model. The Reynolds number (Re) and Prandtl number (Pr) of the air is utilized to calculate the convective heat transfer coefficient, $h$. The equation for $h$ is as follows [14]:

$$h={\displaystyle \frac{k_{\mathrm{a}\mathrm{i}\mathrm{r}}}{d_c}}Y{Re}^y{\mathrm{P}\mathrm{r}}^{1/3}$$

(9)

where $k_{\mathrm{a}\mathrm{i}\mathrm{r}}$ is the thermal conductivity of the air, $d_c$ is the diameter of the battery cell, Pr is the Prandtl Number of air, and $Y$ and $y$ are empirical parameters.

As Reynolds numbers (Re) increase between 1 and 250,000 both parameters $Y$ and $y$ show a downward trend for $Y$ and upward trend for $y$. The Y value starts at 0.989 during Re = 1–4 and reaches 0.0266 when Re = 40,000–250,000, while y increases from 0.330 to 0.805 within this same range [14].

3.4. Initial and Boundary Conditions

A laminar flow model was designed for numerical simulation to evaluate how the battery module interacts thermally. The computational domain was discretized into 112,100 nodes and 271,479 elements, with an element size of 2 mm to ensure an optimal balance between accuracy and computational efficiency. The boundary conditions included a specified inlet velocity for the airflow and an outlet pressure condition to maintain flow continuity. The air velocity was set to 1 m/s, 2 m/s, and 3 m/s, and the corresponding maximum temperature values were recorded. The battery cells were modeled with a uniform volumetric heat generation rate of 48,750 W/m³ to replicate real operational conditions [15]. A simulation temperature of 298 K combined with 1 s time steps and 200 maximum iterations per step was used for numerical stability and convergence purposes. These initial and boundary conditions were defined to accurately capture the thermal and flow characteristics of the cooling system.

3.5. AI Model Development

To achieve early warning of thermal runaway, we developed a unified model by combining our in-lab temperature–voltage measurements of a 3s3p lithium-ion battery pack (we sampled at 1 Hz) with the existing mechanically induced thermal runaway dataset of Lin et al. using the XGBoost classifier [23]. Having concatenated the two sources into one CSV file, we engineered five consecutive lags for each raw signal (voltage_lag₁…_lag₅, temperature_lag₁…_lag₅) and dropped all rows with empty entries to obtain ten time-history predictors for each sample.

All features were min–max normalized before stratified splitting of the data by class (faulty versus non-faulty) into 80% train and 20% test sets. To compromise between bias and variance, we set the XGBoost (version 2.1 2024) classifier to 100 trees, a learning rate of 0.1, and maximum depth of five. A small internal validation fold was used to apply early stopping to avoid overfitting. The performing model provides a binary label and a probability score for each test sample and supports threshold tuning in real-time alarm applications.

4. Experimental Setup

The research battery pack contains three cells series-connected to three cells parallel-connected (3s3p). The setup for this experiment includes a 3s3p battery pack, PLA casing, a power supply, cooling fan, thermocouple, Arduino UNO, and a personal computer. The battery pack was initially designed with 1 mm spacing between the battery cells. The experimental system serves to examine how air cooling affects thermal management for 18,650 battery cells. Figure 5 shows the flow path for cooling air along with the cooling fan installation.

Additionally, voltage and temperature data from the 3s3p battery pack were recorded at one-second intervals during different discharge experiments to capture its normal operating conditions. To enhance fault detection capabilities, an external dataset was incorporated, containing both faulty and non-faulty battery conditions [23]. A collective dataset was then created for training an AI-based classification model. A list of electrical and thermophysical properties for these battery cells appears in

Table 2. Electrical and thermophysical properties of the battery cells.

Parameter	Value	Parameter	Value
Nominal Voltage, (V)	3.7	Nominal Capacity, (A)	2.5
Mass, (kg)	0.045	Diameter, (mm)	18
Length, (m)	65	Thermal Conductivity, ($\mathrm{W}·{\mathrm{m}}^{-1}·{\mathrm{K}}^{-1}$)	${\mathrm{K}}_{\mathrm{r}}=0.2$, ${\mathrm{K}}_{\mathrm{z}}=37.6$ [16]
Density, ($\mathrm{k}\mathrm{g} {\mathrm{m}}^{-3}$)	2722	Specific Heat Capacity, ($\mathrm{J}·{\mathrm{k}\mathrm{g}}^{-1}·{\mathrm{K}}^{-1}$)	1200

.

For the air-cooling system, there are 36 holes (6 rows × 6 columns) for the air inlet to enhance the air-cooling effect, each with a diameter of 5 mm. The outlet of the air is a suction fan with the dimensions 40 mm × 40 mm, controlled by an adapter through Arduino. The airflow velocity of the fan is directly proportional to the voltage supplied. The experimental testing was based on the three-dimensional temperature distribution model of the battery module. The initial value of temperature and inlet velocity were 26 °C and 2 m/s, respectively. The cooling fan used for forced air cooling had the parameters shown in

Table 3. Properties of the fan used for air cooling of battery pack.

Parameter	Value	Parameter	Value
Outer diameter, (mm)	50	Module case size (Length × width × height), (mm)	62 × 62 × 65
Inlet air temperature, (°C)	26	Inlet air velocity at 12 V, (m·s−1)	2

.

The discharge rate, internal electrochemical reaction and resistances are the prime causes of the temperature rise in the battery module. The heat generated (q) can be calculated as follows [3]:

$$\mathrm{q}=\mathrm{I}\left[\left({\mathrm{V}}_{oc}-\mathrm{V}\right)-\mathrm{T}{\displaystyle \frac{\partial {\mathrm{V}}_{oc}}{\partial {\mathrm{T}}_{\mathrm{a}}}}\right]$$

(10)

where I, V_oc, T and V are the discharge current, open-circuit voltage, temperature, and operating voltage of the battery, respectively. The simulation of the battery module uses a uniform and constant heat generation of 48,750 $\mathrm{W} {\mathrm{m}}^{-3}$ because of the 1.5 C discharge rate. The test bench operation requires connecting the battery module to the load that has been calculated to achieve 1.5 C discharging of the battery module. A thermocouple is attached to the battery pack at a central location and is connected to the Arduino UNO and PC to record the temperature values at each instance. Figure 6 shows the experimental setup with an Arduino attached to a PC for measurements.

5. Results and Discussion

5.1. Simulation Results

5.1.1. Air Cooling with Different Pack Configuration

Based on the simulation results, the square configuration appears to have several advantages over the nested configuration. These include an improved level of uniform temperature distribution and better cooling efficiency. In the square configuration, the heat distribution appears more evenly spread across all cylinders, whereas in the nested configuration, some cylinders have higher temperature variations. Moreover, the square configuration allows for more even cooling as the spaces between the cylinders provide better airflow or heat dissipation pathways when the peak temperature is 313 K. The nested configuration might have localized hot spots due to restricted heat transfer in the inner cylinders, as can be seen by a peak temperature of 314 K in Figure 7. As a next step, a square configuration was selected and analyzed for different inlet velocities.

5.1.2. Air Cooling with Different Inlet Velocity

The research evaluates how different inlet velocities affect both cooling effectiveness and temperature distribution within the battery module. A parametric numerical study utilized ANSYS simulation software to analyze airflow velocity and maximum temperature effects through different inlet velocity inputs. These results indicate a reduction in local temperature with increasing airflow velocity, demonstrating an inverse correlation between inlet velocity and the module’s peak temperature. Figure 8 illustrates the velocity contours for each case. The highest battery cells’ temperature was 319 K, 313 K, and 312 K for the air velocities of 1 m/s, 2 m/s, and 3 m/s, as shown in Figure 8a, Figure 8b and Figure 8c respectively. Furthermore, a further increase in the velocity may result in diminishing returns because of the increased power consumption while not reducing the battery temperature significantly. Therefore, optimizing the balance between the inlet velocity and the cooling efficiency is essential. Additionally, if the ambient temperature exceeds the standard range, keeping effective and uniform cooling necessitates increasing the inlet air velocity.

5.1.3. Air Cooling with Different Cell Spacing

Figure 9 shows the impact of cell spacing on battery’s cooling performance. The cells spaced 2 mm apart reach a maximum temperature of 313 K as shown in Figure 9a. When the spacing between the cells is increased to 4 mm, the value of maximum temperature becomes 315 K, rising by 2 degrees as shown in Figure 9b. This shows that further increasing the cell spacing only increases the system volume with no positive effect on the heat transfer from battery cells. This is because the flowing air hardly achieves any contact with the battery cells due to large spacing. Wider gaps allow air to pass through more freely and reduce the contact time with the battery cells. Also, the increased cell spacing creates stagnant air zones in the battery module. The battery cells experience poor heat dissipation because hot air pockets form inside the battery pack. At very small spacings, thermal boundary layers from adjacent cells overlap, creating low-velocity regions and reducing effective heat transfer. Therefore, optimal spacing should avoid both excessive gaps and overly tight configurations. Thus, as per the above analysis, increasing the spacing between the cells beyond 2 mm would not be a good option.

5.2. Experimental Results

Experiments were performed using the test bench shown previously in Figure 6. The battery module was discharged at 1.5 C. First, the battery module was discharged with only natural air convection. The temperature values were noted down using thermocouples for two cases of 2 mm and 4 mm spacing between the battery cells. The results of change in temperature with the time are shown in Figure 10 for both natural and forced convection. The battery module was discharged for 230 s at 1.5 C discharge rate and the maximum temperature of the battery pack achieved without air cooling was 44 °C (317 K) for 2 mm spacing and approximately 42 °C (315 K) for 4 mm spacing between the cells. The battery pack module was then discharged at the same discharge but with forced air convection. The air velocity was maintained at 2 m/s. The temperature values are noted down the same way as for the first case. The highest temperature recorded among the battery module cells reached 38 °C (311 K) when the cells were spaced 2 mm apart and 40 °C for cells spaced 4 mm apart. The highest temperature value for the forced air convection case was lower than that for natural convection. This is because the forced convection offers a higher convection coefficient than natural convection. Also, it can be observed that natural convection is more effective in the case of 4 mm spacing between the cells. Because a larger surface area is in contact with the ambient air, the heat transfer from the battery cells improves. But it can also be observed that in case of the forced convection, 2 mm spacing has a lower highest temperature in the module compared to the 4 mm spacing. This is because in forced convection, the larger spacing between the cells offers lower resistance to the flow of air and air passes between the cells without significant contact area and time with the battery cells. This reduces the heat transfer from cells to the forced cooling air. However, the experimental results are in accordance with the simulation results, which show a decrease in the battery module temperature in forced air convection compared to natural convection. Figure 11 shows a comparison of the experimental results for different cell spacings, showing a great deal of agreement with the simulation results. Thus, it was found that for the optimal spacing of 2 mm, the highest battery temperature is 313 K, which is the lowest among all the other configurations used in this study. Air cooling alone is generally insufficient to maintain battery temperatures within the optimal range; however, under the identified optimal cell spacing conditions, it performs at its maximum potential.

5.3. Fault Detection and Machine Learning Analysis

To complement the thermal observations, a fault detection model was trained on battery sensor data to predict potential failures. The confusion matrix, as shown in Figure 12, demonstrates the model’s ability to classify faulty and non-faulty conditions with high accuracy. The model successfully identified 405 true negatives and 349 true positives, with minimal false misclassifications (41 false positives and 37 false negatives).

The results expressed within our confusion matrix correspond to an accuracy of 0.94, a precision of 0.89, a recall of 0.90, an F₁ score of 0.90, a Cohen’s kappa of 0.87, and a Matthews correlation coefficient of 0.88. These values confirm that the XGBoost classifier reliably captures the key electrical and thermal precursors to battery faults.

To gain further insight into how battery conditions correlate with fault predictions, a graphical visualization was generated, as shown in Figure 13. This plot displays sensor readings of voltage and temperature along with the model’s fault predictions. It can be observed that temperature fluctuations strongly influence fault occurrences, reinforcing the idea that thermal anomalies are a primary trigger for battery faults.

6. Conclusions

This study investigated the impact of the cell configuration, air flow velocity, and cell spacings on the battery temperature control using both simulations and experimental work. The results in both cases confirmed that with the proposed battery design of 3s3p size with air cooling, the battery temperature remains in the safe region for up to 1.5 C discharge rate. The findings of this study are valuable in the improvement of battery thermal management for electric vehicles. This study investigated the upper limit of the discharge rate at which the battery temperature could be kept in safe regions. Thus, this study also contributes to the safety and reliability of electric vehicles, where lithium-ion batteries are mostly used. An AI model was trained on the obtained data and projections were made for data-driven prediction of thermal runaway. Further work needs to be performed on investigating the effect of non-uniform air flow distribution on thermal runaway, in which certain battery cells are cooled better than others due to variations in air flow distribution. This leads to hotspots in the battery module, which accelerates battery degradation and increases the risk of thermal runaway. Thus, optimization is required for air flow through the battery module.