Knowledge-data driven sampling diagnosis algorithm for lithium batteries on electric vehicles

The voltage is one of limited reliable information for battery management system, and the faults of voltage sampling will result in adverse effects and lead to potential risks for operation, which emphasize the importance for investigating the failure modes of voltage sampling and diagnosis algorithm. In this article, a knowledge-data driven sampling diagnosis algorithm is established and an online intelligent diagnosis algorithm is proposed accordingly based on outlier detection with fuzzy entropy. The fault diagnosis algorithm is established and evaluated under positive exploitation, where the knowledge-base of failure mode based on equivalent simulating models is firstly constructed. 6 kinds of potential failure modes from battery management system are simulated to investigate the performances, and the symmetrical voltage distribution or near-zero voltage can be extracted as the feature for determining the failure mode. Then, a diagnosis algorithm is established based on outlier detection method, and results are validated according to fault matrix method. The batter-in-loop experiments confirm the symmetrical voltage performances when there is a sampling line cut down, and furthermore, we use the dataset from the cloud monitoring platform to verify the applicational results. The article contributes to design the fault diagnosis algorithm from failure modes, which can be further promoted for other systems or for cloud-platform.