Lithium-ion batteries fault diagnosis based on multi-dimensional indicator

Abstract

Since lithium-ion batteries are the core components and main sources of failures in electric vehicles and energy storage systems, fault diagnosis plays a crucial role in the stable operation of lithium-ion batteries. In this paper, a multidimensional indicator-based lithium-ion battery fault diagnosis algorithm is proposed to obtain the weights of different dimensional indicators in the battery fault evaluation system, which applies entropy weight method to calculate the risk coefficients of each individual battery for battery fault diagnosis. The algorithm validation work in this paper is completed in the MIT-Stanford public experimental data set and the actual operation data set of the energy storage system. Firstly, the data feature extraction method suitable for engineering application scenarios is selected. Then, the entropy weight method is used to calculate the weights of each indicator, and then the cause of the battery failure is analysed based on the weight information of each indicator. The method with the entropy of each dimensional indicator to calculate its weight can locate the faulty battery and reduce the subjectivity in the fault analysis process. Meanwhile, the method is suitable for real-time fault diagnosis of lithiumion battery systems without complicated training models and hyper-parameter adjustment processes.