Voltage abnormality-based fault diagnosis for batteries in electric buses with a self-adapting update model

This study aims to solve the key issue for electric buses on how to improve the accuracy and reliability of battery fault diagnosis with the emerging intelligence technology on battery management. The battery fault diagnosis method needs to fuse both the physic and cyber systems, reflecting the real-time dynamic battery system in the physical-layer, as well as taking full advantage of battery historical data and outside information in the cyber-layer. Given that, this paper proposes a fault diagnosis method based on the physical-layer model updated by cyber-layer deep learning algorithms. Long short-term memory network (LSTMN) and back propagation neural network (BPNN) are used in the proposed framework to anticipate present aging conditions and update physical-layer model parameters. As a result, the battery model may self-update in response to the environment and aging circumstances, increasing the accuracy of problem detection and prediction. Finally, data gathered from electric buses in real-world conditions is utilized to validate the proposed method's accuracy and dependability in normal voltage and failure prediction. In addition, compared with the traditional model-based method, the fault warning time can be advanced by 51 s on average. And the misdiagnosis rate and the diagnosis failure rate can decrease by 15.8 % and 11.6 % respectively through the proposed system.