Scholarlink

<p>The present study compares different battery thermal management system (BTMS) designs for lithium-ion batteries using computational fluid dynamics (CFD) simulations, with battery packs modeled in SolidWorks considering realistic heat dissipation characteristics and flow behavior. Effective thermal regulation of lithium-ion batteries (LIBs) is essential to ensure the performance, reliability, and safety of electric vehicles (EVs). Radiator-based BTMS, which operates without a compressor, has recently gained prominence due to its lower energy consumption and efficient temperature control. LIB performance is highly sensitive to temperature-dependent parameters such as state of charge (SOC), self-discharge rate, and state of health (SOH), necessitating the maintenance of an optimal thermal window to enhance overall battery longevity. Three BTMS configurations were evaluated: Design 1, a basic flow arrangement; Design 2, an improved airflow design with enhanced vent placement; and Design 3, an optimized gradient cooling-channel configuration. All three designs successfully reduced the battery temperature from an initial 60&#xa0;℃, with Design 3 exhibiting the highest temperature drop of approximately 45.27&#xa0;℃ and maintaining a low spatial temperature variation (&lt; 2&#xa0;℃) across the pack. Its balanced airflow distribution and improved thermal uniformity make it particularly suitable for high-ambient-temperature environments. To further evaluate predictive capabilities, machine-learning models including MLP-NN, XGBoost, SVR, Random Forest, Gradient Boosting, and a Stacked Hybrid model were developed. The Stacked Hybrid model achieved the highest accuracy with an R<sup>2</sup> of 0.96 and the lowest mean square error, while XGBoost and Gradient Boosting showed stable convergence. In contrast, SVR and Random Forest models underperformed due to their reduced sensitivity to nonlinear thermal variations.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Thermal Performance Optimization of Air-Cooled Battery Packs in Electric Vehicles: An Integrated CFD and Machine-Learning Approach

  • 期刊论文

  • 发表日期：

    2026年2月24日

查看全文
  • Madhan Kumar Subramani,
  • Sadees Mohan,
  • Thiyagarajan Subramani,
  • Anbalagan Sathishkumar,
  • Sung Chul Kim

摘要

The present study compares different battery thermal management system (BTMS) designs for lithium-ion batteries using computational fluid dynamics (CFD) simulations, with battery packs modeled in SolidWorks considering realistic heat dissipation characteristics and flow behavior. Effective thermal regulation of lithium-ion batteries (LIBs) is essential to ensure the performance, reliability, and safety of electric vehicles (EVs). Radiator-based BTMS, which operates without a compressor, has recently gained prominence due to its lower energy consumption and efficient temperature control. LIB performance is highly sensitive to temperature-dependent parameters such as state of charge (SOC), self-discharge rate, and state of health (SOH), necessitating the maintenance of an optimal thermal window to enhance overall battery longevity. Three BTMS configurations were evaluated: Design 1, a basic flow arrangement; Design 2, an improved airflow design with enhanced vent placement; and Design 3, an optimized gradient cooling-channel configuration. All three designs successfully reduced the battery temperature from an initial 60 ℃, with Design 3 exhibiting the highest temperature drop of approximately 45.27 ℃ and maintaining a low spatial temperature variation (< 2 ℃) across the pack. Its balanced airflow distribution and improved thermal uniformity make it particularly suitable for high-ambient-temperature environments. To further evaluate predictive capabilities, machine-learning models including MLP-NN, XGBoost, SVR, Random Forest, Gradient Boosting, and a Stacked Hybrid model were developed. The Stacked Hybrid model achieved the highest accuracy with an R2 of 0.96 and the lowest mean square error, while XGBoost and Gradient Boosting showed stable convergence. In contrast, SVR and Random Forest models underperformed due to their reduced sensitivity to nonlinear thermal variations.