An Early Fault Diagnostic Method Based on Anomaly Cumulation for Power Batteries

Early detection of faults in power batteries is essential for the safety of electric vehicles (EVs). However, the character of early faults is similar to that of noise, which makes it challenging to accurately diagnose early faults without misdiagnosis. To address this problem, a method based on the cumulation of anomalies is proposed in this article, which can quantify the anomalies of the battery and accumulate them through biased integration. Consequently, the anomalies induced by faults are progressively amplified during the fault duration, whereas the noise-induced anomalies diminish once the noise ceases. The method consists of three steps: prediagnosis, fusion, and cumulation. The quantitative battery anomaly degree can be obtained through steps 1 and 2, which are then further processed and accumulated in step 3 to differentiate between faults and noise. Experimental results reveal that the presented approach simultaneously guarantees a high fault detection rate of 95% and a low misdiagnosis rate of 20%, significantly surpassing traditional methods by approximately 35% and 30%, respectively. Thus, the proposed method proves effective for the detection of early faults in EV power batteries.