A data-driven early warning method for thermal runaway during charging of lithium-ion battery packs in electric vehicles

Abstract In recent years, thermal runaway during charging of lithium-ion batteries has become a critical issue. This problem has emerged as a significant barrier to the development of power batteries for electric vehicles (EVs). This paper addresses this challenge from a data-driven perspective by proposing a temperature prediction model for thermal runaway during charging of EV lithium-ion batteries. The model leverages both long short-term memory and Transformer algorithms to account for the time-series characteristics of batteries charging. The charging data under varying capacities and ambient temperatures are extracted using the Newman–Tiedemann–Gaines–Kim model for lithium-ion batteries, which is then used to optimize the accuracy of the hybrid algorithm through training. Additionally, real-world EV charging data is collected to further validate the temperature prediction model. Experimental results demonstrate that the proposed model achieves superior prediction accuracy compared to both single models and convolutional neural network hybrid models. Based on this model, a residual-based early warning method incorporating a sliding window approach is proposed. The experimental findings indicate that when the residual of the predicted charging temperature for EVs lithium-ion batteries exceeds the warning threshold, preemptive termination of charging effectively prevents thermal runaway.