基于局部离群点检测的动力电池组不一致早期故障预警OA北大核心CSTPCD

With the rapid development of new energy vehicles,the safety of power batteries has gained growing public attention.On the new energy vehicle operation monitoring platform,the existing power battery safety detection function fails to provide early warnings of battery failures.A more suitable process for early warnings of battery inconsistency in power battery packs has been designed to address the battery inconsistency warnings.First,a dynamic gradient data cleaning strategy based on box graph method is designed to effectively eliminate abnormal data.Then,the data are divided into charging stages and inconsistent characteristics of individual voltage changes are extracted.Based on this,the outlier detection algorithm is employed to obtain the outlier values of each battery cell,conduct initial warning of inconsistent faults,and identify abnormal battery cells.Our retrospective analysis of actual vehicles with inconsistent battery faults demonstrates that the preexisting alarm mechanism of the monitoring platform for this process has no less than 7 charging cycles and accurately locates abnormal cells.