Research on fault diagnosis of lithium-ion batteries based on adaptive boosting of abnormal features

The battery management system (BMS) plays a pivotal role in ensuring the safety and stability of high-capacity lithium-ion batteries used in transportation systems. Early warning of battery safety issues, a core function of the BMS, traditionally relies on direct inspection of parameters such as voltage, current, and temperature. However, this approach suffers from short warning windows and a high rate of false alarms due to the subtle changes in macroscopic signals during the early stages of battery faults. To address these limitations, this study introduces the concept of safety entropy as a quantitative metric for assessing safety risks in power battery systems. Building upon this metric, a power battery system safety early warning algorithm has been developed, which integrates fault probability weights of battery system. Compared to conventional methods, this algorithm enables early quantitative risk assessment at the incipient stage of faults. Validation using historical accident data from new energy vehicles (NEVs) as well as NEV and charging station collaborative data demonstrates its effectiveness in risk identification, achieving a warning capability of 7 days in advance. The proposed safety entropy model and adaptive dynamic weighting mechanism provide a novel solution for the safety monitoring of high-capacity lithium-ion batteries in rail transit vehicles, significantly enhancing early warning capabilities.