Machine learning assisted multi-objective optimization of a thermal management and barrier integration structure for battery thermal safety

To ensure both thermal safety and performance in high-energy-density lithium-ion battery systems, this study proposes an integrated modelling and optimization framework for a novel Thermal Management and Barrier Integration Structure (TMBIS), which couples phase change materials (PCM) with flame-retardant (FR) insulation layers. A hybrid modelling strategy is established by combining a computational fluid dynamic (CFD) based electrochemical–thermal model with a reduced-order thermal resistance network (TRN) model, capturing both internal heat generation and inter-module heat transfer mechanisms. To evaluate and optimize the thermal runaway (TR) mitigation and thermal management (BTMS) performance of the structure, Latin hypercube sampling (LHS) is first employed to generate representative design points. An artificial neural network (ANN) surrogate model is then trained to predict key performance indicators, including the maximum battery temperature and TR delay times. Finally, a multi-objective optimization is conducted using the NSGA-II algorithm to balance competing objectives, and representative trade-off solutions are identified via KMeans clustering. The proposed framework efficiently identifies optimal PCM–FR design configurations, achieving a favourable compromise between TR suppression and thermal management performance.