Online diagnosis and prediction of power battery voltage comprehensive faults for electric vehicles based on multi-parameter characterization and improved K-means method

The battery terminal voltage in the power battery system is a comprehensive indicator of its internal resistance, capacity, state of charge (SoC) and other parameters, which can more comprehensively assess the safety performance of the battery system, so it is of great significance to accurately diagnose and predict the voltage faults of individual cells. Based on this, two-dimensional fault characteristics that can effectively recognize the voltage fluctuation are first extracted. And then based on the improved K-means method to carry out the identification of fault cells. In order to achieve online applications and comprehensive detection of different forms of voltage faults, this paper proposes for the first time a double sliding time window simultaneous implementation strategy and optimizes the window length and evaluation coefficient (EC) threshold based on a data-driven approach, and the proposed algorithm can be performed in real time without any significant delay. Finally, the necessity, reliability and stability of the method are verified and compared with Shannon entropy method and correlation coefficient method. Results indicate the method in this article is capable to recognize the various data patterns of the potential threat and can accurately identify anomalies prior to thermal runaway (TR) or failure of the vehicle.