A fault diagnosis method for electric vehicle power lithium battery based on wavelet packet decomposition. (1st December 2022)

• Record Type: Journal Article
• Authors: Jiang, Jiuchun; Zhang, Ruhang; Wu, Yutong; Chang, Chun; Jiang, Yan
• Is Part Of: Journal of energy storage. Volume 56:Part A(2022)
• Publication Date: 2022-12-01
• Subjects: Battery fault diagnosis -- Wavelet packet decomposition and reconstruction -- Sparse autoencoder -- Singular value decomposition -- Discrete Fréchet distance -- Chauvenet criterion; Energy storage -- Periodicals; Energy storage -- Research -- Periodicals; 621.3126
• DOI: 10.1016/j.est.2022.105909
• Languages: English
• ISSNs: 2352-152X

Abstract

Abstract: This paper proposes a fault diagnosis method for electric vehicle power lithium battery based on wavelet packet decomposition. Firstly, the original voltage signal is decomposed into the low-frequency part and high-frequency part based on wavelet packet decomposition. For the high-frequency part, after filtering the noise using wavelet packet energy noise reduction method, the time domain voltage signal is obtained by wavelet packet reconstruction, and the characteristic parameters reflecting the battery fault are extracted by using sparse autoencoder; for the low-frequency part, the characteristic parameters reflecting the battery inconsistency are obtained by using singular value decomposition. The similarity between each individual cell and the average cell is then measured using the discrete Fréchet distance algorithm. Finally, the outlier detection method based on the Chauvenet criterion is used to detect the faulty cells using the obtained curve similarity. The effectiveness of the proposed method is verified by the data of two thermal runaway vehicles. Highlights: The original voltage signal is processed by wavelet packet decomposition. Feature extraction using autoencoder and singular value decomposition. The similarity between feature parameters is calculated by Discrete Fréchet distance. The outlier filter is designed based on the Chauvenet criterion.