基于孤立森林算法的动力电池不一致单体识别与预警方法

Inconsistent Monomer Identification and Warning Method for Power Batteries Based on Isolated Forests

动力电池组各单体之间由于生产工艺和使用环境等因素会产生不一致性故障,从而影响电池使用寿命和行车安全。针对动力电池组不一致性故障下的异常单体诊断需求,依托车联网平台数据,基于孤立森林法提出一种动力电池单体识别与预警方法,通过统计正常样本与异常样本的得分情况,以确定故障阈值T=0.75;同时,结合滑动窗口实时更新流入诊断模型的数据,实现对不一致单体的识别与预警。结果表明,该方法可以有效识别出不一致性单体,查全率和查准率分别为0.91和0.95,当滑动窗口大小为15时对实时故障的预警效果最好。研究成果有利于减少或避免因动力电池不一致造成电动汽车起火和爆炸事故,对推动电动汽车进一步普及具有重要意义。

The inconsistencies between the monomers of the power battery pack will occur due to the factors such as production process and service environment,which will affect the service life and driving safety of the battery.To meet the requirement of abnormal monomer diagnosis under inconsistent fault of power battery pack,based on data of vehicle network platform,a method of power battery monomer identification and early warning is proposed based on isolated forest method.The fault threshold T=0.75 is determined by counting the scores of normal and abnormal samples.At the same time,the data flowing into the diagnosis model is updated in real time by combining the sliding window to realize the recognition and warning of inconsistent monomers.The results show that the method can effectively identify inconsistent monomers,with the recall and accuracy of 0.91 and 0.95 respectively.When the size of sliding window is 15,the warning effect of real-time fault is the best.The research obtained in this paper is beneficial to reduce or avoid electric vehicle fire and explosion accidents caused by inconsistent power batteries,and is of great significance to promote the further popularization of electric vehicles.