Volume 65

Multi-Objective Optimization of Liquid-Cooled Serpentine Cooling Plate for Battery Packs Based on Multi-Feature Parameters and Machine Learning Changtian Xu1,Xiang Qiu2,Weicheng Xuan2,Zhuoye Wang1,Jiangping Chen2,Qiang Li2*, Junye Shi2*

Abstract

With the superior heat transfer coefficient and operational stability, liquid cooling has become the most prevalent technology in battery thermal management systems. Traditional designs of serpentine cooling plates are often developed through trial-and-error adjustments to parameters such as plate width and channel length, relying heavily on engineering experience. These designs typically halt iteration once basic technical requirements are satisfied, rather than striving to maximize heat transfer performance, thereby constraining the cooling potential of serpentine plates. This study introduces a machine learning-driven strategy within a physics-based framework that integrates heat transfer and fluid dynamics equations to optimize serpentine cooling plates for high heat generation scenarios. Leveraging the NSGA II algorithm, the framework optimizes critical design parameters while elucidating their influence on key physical characteristics during the design process. At a heat generation rate of 6 kW, the cooling plate—validated through experimental simulations—achieves a maximum temperature difference of 8.5°C in the battery cell region, showcasing outstanding heat transfer performance. Compared to conventional serpentine cooling plates, it enhances temperature uniformity by 30% at an equivalent pressure drop.

Keywords battery thermal management, liquid-cooled plate, machine learning, physics-informed machine learning, multi-objective optimization