An online inconsistency evaluation and abnormal cell identification method for real-world electric vehicles

Cell inconsistency is an inevitable challenge that large-scale battery packs must confront, directly impacting their safety and performance. Accurate cell inconsistency evaluation is essential for efficient state management and second-life utilisation. This paper presents an inconsistency evaluation method based on cell voltage only and simultaneously identifies abnormal cells that perform relatively poorly compared to others. Firstly, the discrete wavelet transform (DWT) is utilised to generate features representing each cell's state of health (SOH). Then, the effectiveness of the DWT-based features in quantitatively representing SOH is validated using the NASA dataset. Leveraging charging data extracted from real-world electric vehicles (EVs), the DWT-based feature is further applied to detect the abnormal cells of the battery system. Additionally, a model-based approach, focusing on the internal resistance (IR) difference, is introduced for comparison with the proposed DWT-based method. The results show that the proposed method accurately identifies the abnormal cells of real-world EVs within a mere 10 s, markedly outperforming the traditional model-based approach, which consumes 15.6 h. Finally, a robust indicator is proposed to characterise the evolution trend of cell inconsistency in complex operating conditions.