Battery Pack Temperature Distribution Prediction Using Physics-Informed Neural Network and Limited Sensors

Thermal runaway in large-capacity lithium-ion (Liion) battery modules, caused by uneven temperature distribution. To tackle this issue, this study introduces a Physics-Informed Neural Network (PINN) approach to predict the temperature field of a three-series 280 Ah battery module. The method integrates heat generation, heat transfer, cell temperature ranges, and boundary conditions into the loss function. By utilizing minimal temperature measurements at pressure relief valve points, the method accurately infers the entire surface temperature distribution of the module. Experimental validation confirms the method's precision, with an absolute error of only 1.3 K in predicting the temperatures of the module's surface. This high accuracy supports improved thermal management strategies, enhancing the safety and efficiency of large-scale battery systems in energy storage applications.