Thermal Runaway Fault Diagnosis for Battery Systems Based on Correlation Coefficient and Local Outlier Factor

Thermal runaway fault warning for power battery and timely detection of faulty cell can effectively extend the service life of power battery systems and further promote the popularization of electric vehicles. Accordingly, this paper proposes a power battery thermal runaway fault diagnosis method based on correlation coefficient and local outlier factor (LOF). For two thermal runaway vehicles and two normal vehicles, based on the real-vehicle running voltage data of their power batteries, after data cleaning and transformation, the correlation coefficient algorithm is first used to obtain the voltage correlation coefficients of the individual batteries in the acquired voltage data set and summed, and then the LOF algorithm is introduced to obtain the LOF value of the voltage correlation coefficients of the individual batteries based on the time window, and then the thresholds are set to realize the detection and localization of faulty individual batteries. In addition, the advantage of this paper is that the data selected is the voltage data of the last startup state of the thermal runaway vehicle and the data curve has no abnormal fluctuation or outlier characteristics. The research results show that the diagnostic method can accurately alarm and lock the location of the faulty cell much earlier than common methods, which verifies its effectiveness and compares it with the Shannon entropy algorithm in the identification of abnormal cells; at the same time, it verifies the reliability of the model, which will not give false alarms when the vehicle is in a normal state.