Precise Battery Resistance Increase Fault Diagnosis Method with Improved Deep Neural Network

Le résumé fourni par la source

Abnormal battery resistance increase is a dangerous fault for energy storage and transportation electrification applications, which increases battery temperature, fastens battery aging, and even causes fatal accidents such as fire or explosion. To achieve safe and reliable system operation, there is an urgent need for a low-cost and highly efficient monitoring method to detect and identify battery resistance increase fault. However, current monitoring methods still suffer from slow response speed, high measurement cost, and difficult field deployment. To address these issues, this article proposes a precise and highly adaptable battery resistance increase fault diagnosis method. The sliding window method and extended Kalman filter are adopted to extract key features from battery charging profiles comprehensively. With the extracted features, the fault diagnosis is conducted based on an optimized deep feedforward neural network (FNN). Then, the accuracy and generalization of the proposed FNN is further enhanced by calibrated with a delayed data check (DDC) method. The proposed model can be applied to diagnose the resistance increase fault for any battery charging voltage segment, and successfully achieves a high fault diagnostic success rate of 96.5%.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.