A fast data analysis method for abnormity detecting of lithium-ion batteries in electric vehicles

In recent years, the big data platforms for electric vehicles have widespread set up by governments and enterprises. Analysts can monitor or trace anomalies through the historical operation data from electric vehicles. However, repetitive operating conditions result in a challenge for the analysis of massive historical data, and methods to extract macroscopic changes of the batteries state is still lack. This paper proposes a method for observing battery pack characteristics and variations from a macroscopic perspective, enabling rapid identification and analysis of pack abnormalities. The method involves calculating the area under the voltage curve of battery packs and extracting outlier cells and pack state changes using quartile normalization and Kullback-Leibler divergence. Using historical data from normal and incident-prone bus and van validating the method. Abnormal battery conditions such as under-voltage, aging, internal short-circuits, thermal runaway, and specific usage patterns are successfully identified. The method demonstrates high robustness and maintains effectiveness with a one-hour sampling interval. Moreover, the proposed method is computationally efficient, making it suitable for deployment in onboard units or online platforms for regular monitoring of batteries. This represents the first approach for macroscopic analysis of battery pack historical data.