Fault diagnosis method for lithium-ion batteries in electric vehicles based on isolated forest algorithm

Electric vehicle safety accidents caused by Lithium-ion (Li-ion) batteries failures are numerous in recent years. The voltage data of a faulty battery will have abnormal changes before a safety accident occurs. The voltage variation of a progressive failure is more obvious, while the voltage change of a sudden failure is concealed. This paper proposes a fault diagnosis method for power lithium battery based on isolated forest algorithm. Firstly, the original voltage data are signal processed and decomposed into static components highly correlated with aging inconsistency and dynamic components reflecting anomalous information, and then the characteristic parameters of static and dynamic components are extracted and fed into the isolated forest algorithm for anomaly detection to identify anomalous cells. Finally, the proposed method is tested with voltage data from four faulty vehicles. The tests prove that the method has good advance detection ability for both progressive and sudden failures, which confirms its advance detection effect in power lithium battery fault diagnosis and its feasibility of real-time application in real vehicles.