Study on thermal runaway characteristics of lithium batteries under high-rate charge/discharge and development of a deep learning-based early warning model

This study addresses the challenging issue of predicting thermal runaway (TR) in lithium-ion batteries under high charge-discharge rate conditions. The thermal runaway behavior characteristics of lithium-ion batteries are revealed, As the charge and discharge rate increases, the thermal stability of the lithium battery decreases, and the time to valve opening during thermal runaway is shortened. In this study, we propose an integrated framework that combines a multi-head self-attention mechanism with convolutional operations for the prediction of thermal runaway in lithium-ion batteries. A deep learning model based on the Transformer framework—ConvTransformer—was proposed to accurately predict the remaining time (RT) before thermal runaway. This model innovatively integrates convolutional neural networks (CNN), the Transformer architecture, and a multi-layer perceptron (MLP) module. By embedding convolutional layers within the Transformer encoder, the model optimizes the extraction of spatial features and the modeling of temporal dependencies from multi-variable time series data. This approach significantly enhances the representational power and predictive accuracy of the model. The ConvTransformer proposed in this study provides a real-time, accurate thermal runaway prediction solution for battery management systems (BMS), offering significant engineering application value in enhancing the safety and reliability of lithium battery operations.