Research on Thermal Field Prediction Method of Lithium Battery Module Based on PINN

Qiang-Sheng Xiao, Yan-Zuo Chang, Lei Huang

battery thermal management, lithium battery module, multi-physics constraints, PINN, thermal field prediction

Accurate thermal field prediction is essential for optimizing battery thermal management systems and preventing thermal runaway of lithium battery modules. Traditional thermal prediction methods, including experimental measurement, numerical simulation, and pure data-driven modeling, suffer from obvious limitations such as heavy data dependence, high computational consumption, and poor physical interpretability. To address these issues, this paper develops a multi-physics constrained three-dimensional transient thermal field prediction framework for lithium battery modules based on physics-informed neural networks. The framework integrates transient heat conduction principles and couples Joule irreversible heat and entropy reversible heat to characterize internal heat generation. On the basis of conventional thermal boundary and initial constraints, charge consistency and material interface continuity constraints are innovatively introduced to optimize the loss function, combined with efficient training optimization strategies. The results indicate that the proposed mesh-free framework can achieve high-precision full-domain thermal field reconstruction using only sparse temperature data. It strictly complies with fundamental heat transfer and charge conservation laws, adapts to complex battery module structures, and provides an effective solution for fast, accurate thermal prediction of lithium batteries under sparse sensor working conditions.

Received: 03 Jul 2026, Received in revised form: 01 Aug 2026, Accepted: 05 Aug 2026, Available online: 12 Aug 2026

10.22161/ijaers.138.4