Fault diagnosis of real-scenario battery systems based on modified entropy algorithms in electric vehicles

Abstract

Entropy algorithm is an efficient fault diagnosis technology gradually gaining popularity, which is highly promising in terms of battery safety protection in electric vehicles. This paper compares two entropy algorithms commonly used for battery fault diagnosis and their ability to diagnose abnormal fluctuations implied in the pre-fault phase. By exploring the influence of key calculation factors in the entropy algorithm on the fault diagnosis effect, this study pioneered the discovery of a significant normal distribution pattern and the logarithmic relationship between the diagnostic effect and the size of the Shannon entropy's calculation window and multi-scale sample entropy's scale factor, respectively. The relationship between computation time and computation factor was found to provide a theoretical basis for improving diagnosis efficiency through the rational selection of computation factors. Furthermore, based on actual vehicle fault data verification, a multi-level diagnosis strategy with strong robustness and generality is proposed using statistical methods. More importantly, this study continues to explore the importance of deeper mining of different entropy features, and the proposed control strategies are of high practical significance for battery fault diagnosis/prognosis in real-world vehicle applications.