A Self-Discharge Fault Diagnosis Method for Electric Vehicle Batteries

This paper proposes a method to diagnose the selfdischarge faults of lithium-ion batteries using the field data of 20 electric vehicles (EVs) collected by the State Administration for Market Regulation, China. After preprocessing, the cleaned field data is screened based on a few items (i.e., battery pack current, battery pack state of charge, vehicle charging status, and vehicle speed) and certain pieces are identified as “valid segments”. Then, three voltage-related features are extracted: average voltage, voltage standard deviation, and voltage change rate. Based on these features, two sets of thresholds (i.e., one for the first feature and the other for the second/third features) are defined according to the three-sigma rule, which are utilized to identify the battery cells with self-discharge faults. Evaluation results show that cell \#57 in EV \#A and cell \#24 in EV \#B are associated with self-discharge faults.