Evaluation of Platform Oriented Battery Fault Diagnosis Algorithms

With the development of the research on battery fault diagnosis, more and more algorithms have been proposed, but how to compare the effectiveness of different algorithms and whether they are suitable for the current on-board battery management system (BMS) has not been discussed enough. This paper discusses and summarizes the evaluation indicators of fault diagnosis algorithm in cloud platform integrated application environment: algorithm accuracy, warning time and computational complexity, and puts forward the calculation method of each evaluation indicator. Based on the operation data of electric vehicles (EVs) providing public services collected by cloud platform, the fault segments of thermal runaway EVs and the normal segments of normal EVs were extracted as the test inputs of the Shannon entropy method (SEM), correlation coefficient method (CCM) and 30 multi-level screening strategy (3$\sigma$-MSS). By comparing and analyzing the diagnostic results of different segments, the characteristics of each method were summarized.