Identification of power battery voltage inconsistency faults in electric vehicles based on K-means++ clustering with dynamic <italic>k</italic>-values

Abstract

<p indent="0mm">The incompetency of power batteries is one of the most important factors in terms of their durability, reliability, and safety. Therefore, this study proposes a method for identifying power battery voltage inconsistency faults using the historical voltage data collected by onboard sensors from the monitoring platforms of vehicle enterprises and optimizing the number of clusters and the initial center (cluster center) selection based on dynamic <italic>k</italic>-value K-means++ clustering. First, according to this method, the historical operation data of 105 vehicles (of the same model) driven up to 10000 km without any abnormal alarms were analyzed, and the OF threshold for alarm-free vehicles was determined as 0.02 via the box-plot method. Finally, based on this method and the K-means clustering method with a fixed <italic>k</italic>-value, we analyzed the only vehicle with a “poor battery unit consistency” alarm (the same model as the 105 vehicles) on the vehicle monitoring platform. The analysis shows that the K-means clustering method can identify the abnormal cells two days earlier than the alarm time of the vehicle monitoring platform, and the proposed method can identify the abnormal cells six days earlier than the alarm time, demonstrating the early detection of power battery voltage inconsistency faults by the proposed method. The research results of this study are highly significant in reducing or avoiding electric vehicle (EV) fires and explosions and are conducive to developing the EV industry and achieving the strategic goal of “carbon peaking” and “carbon neutrality” in the automotive field.