Thermal safety and risk assessment of lithium-ion batteries aged in low-temperature and low-pressure environments

Abstract

The safety of lithium-ion batteries (LIBs) is a key factor leading to frequent fire accidents in new energy vehicles. The special environmental conditions of high-altitude regions with wide temperature ranges and low pressure present even greater challenges to the cycle performance and safety of LIBs, increasing the risk of thermal runaway. Therefore, to effectively reduce the occurrence of battery safety incidents, there is an urgent need to develop a thermal safety (TS) evaluation model for LIBs that is suitable for high-altitude environments, to enhance the monitoring and management of battery safety. This study proposes a TS evaluation model for LIBs in high-altitude environments and uses machine learning techniques to assess the safety risk of aged batteries. A dataset comprising 15 different aging conditions is constructed based on experimental data, considering ambient temperature, pressure, and cycle number. Key TS characteristic parameters, including the turning point of expansion force (F_tp), sudden drop point of expansion force (F_sr), and surface temperature difference (ΔT), are selected as critical indicators for evaluating the TS of LIBs. Pearson correlation analysis is conducted to identify aging characteristic parameters correlated with TS parameters. A support vector regression algorithm is then used to establish the TS evaluation model. After cross-validation and grid search optimization, the model’s prediction accuracy significantly improves, with R2 values ranging from 0.853 to 0.987, demonstrating good accuracy and stability. Based on the predicted TS parameters, a battery risk assessment method is proposed. The results show that the accuracy of the risk assessment using the predicted values exceeds 99%. This study provides theoretical and methodological support for the safety management and early warning of battery systems in low-temperature and low-pressure environments.