A Multi-Fault Diagnosis Framework for Electric Vehicle Batteries

To enhance the safety of electric vehicles (EVs) powered by lithium-ion batteries, this paper proposes a battery multi-fault diagnosis framework involving four phases. First, the battery field data is cleaned and divided into multiple segments with approximately equal time spans. Second, to diagnose the battery faults, the segments with faults are detected by calculating the Euclidean distance between two different segments. Third, within the fault segment, the fault cell is located and the fault time is identified using a longitudinal outlier average (LOA) method. Finally, the specific fault type is analyzed. The proposed framework is evaluated using the field data of 20 on-road EVs collected by the State Administration for Market Regulation, China. Results show that three types of battery faults can be identified: connection faults, self-discharge faults, and internal short circuit (ISC) faults.