Scholarlink

<p>The widespread use of lithium-ion batteries in electric vehicles, drones, and energy storage systems highlights the need for efficient battery thermal management systems (BTMS). This study optimizes air-cooled BTMS by maximizing cooling performance and minimizing energy consumption. A Bézier curve parameterizes flow channel reshaping, reducing design complexity to five parameters, with ten additional battery spacing variables, totaling 15 design variables. Multi-objective Bayesian optimization (MBO) is performed using Gaussian process regression (GPR) and additive Gaussian process regression (AGP). Finite element method (FEM) simulations in COMSOL Multiphysics validate thermal and fluid dynamic performance. Results show that while GPR converges faster, AGP identifies superior designs by exploring non-local interactions. Notably, AGP reveals optimal cooling can often be achieved without extensive flow channel reshaping, providing practical insights for BTMS development. This research advances energy-efficient and high-performance thermal management solutions for modern battery systems.</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

Comparative study of Gaussian process and additive Gaussian process for multi-objective optimization in air-cooled battery thermal management systems

  • 期刊论文

  • 发表日期：

    2025年12月6日

查看全文
  • Hanbin Cho,
  • Minjik Kim,
  • Youngseo Park,
  • Hugon Lee,
  • Ikjin Lee,
  • Seunghwa Ryu

摘要

The widespread use of lithium-ion batteries in electric vehicles, drones, and energy storage systems highlights the need for efficient battery thermal management systems (BTMS). This study optimizes air-cooled BTMS by maximizing cooling performance and minimizing energy consumption. A Bézier curve parameterizes flow channel reshaping, reducing design complexity to five parameters, with ten additional battery spacing variables, totaling 15 design variables. Multi-objective Bayesian optimization (MBO) is performed using Gaussian process regression (GPR) and additive Gaussian process regression (AGP). Finite element method (FEM) simulations in COMSOL Multiphysics validate thermal and fluid dynamic performance. Results show that while GPR converges faster, AGP identifies superior designs by exploring non-local interactions. Notably, AGP reveals optimal cooling can often be achieved without extensive flow channel reshaping, providing practical insights for BTMS development. This research advances energy-efficient and high-performance thermal management solutions for modern battery systems.