Micro-fault identification for lithium-ion batteries in electric vehicles based on similarity measurement of the charging sequence

Abstract Battery fault diagnosis is critical to safeguard the safety of electric vehicles. Since lithium-ion battery cells are usually inconsistent and difficult to distinguish from faults, existing micro-fault identification methods are difficult to detect faults as early as possible based on real-world vehicle data. In contrast to the operation process, the voltage and temperature changes are more regular since the current relatively remains constant and the charging strategy is fixed, which is conducive to accurate micro-fault identification. According to the charging strategy of electric vehicles, charging data extraction and selection methods are introduced. Considering the practicality and the difficulty of threshold value determination, the micro-fault identification approach based on DBSCAN is proposed. Normalized IAVs and LIP distances are utilized as features, which come from charging sequences. The results of the thermal runaway vehicle show that the method can effectively find fault cells in voltage signals 18 days before thermal runaway occurs.