Abstract: The performance difference between cells is one of the main factors affecting the safety of power battery pack. In order to identify the batteries with poor consistency, that is, batteries that are more prone to risk. This paper improves the conventional DBSCAN algorithm on the basis of its sensitivity to input parameters. An improved DBSCAN algorithm based on data distribution characteristics is proposed to determine the parameters of MinPts and Eps adaptively. In this paper, the improved DBSCAN algorithm is used for battery safety diagnosis, and related experiments are designed and verified by using battery charge and discharge data. The verification results show that the improved algorithm proposed is effective in identifying dangerous batteries.

Key words: Power battery pack; DBSCAN algorithm; Outliers detection; Consistency; Self-adaptive

摘要： 单体电池间的性能差异是影响动力电池组使用安全性的主要因素之一，为识别电池组中一 致性差的电池，即更容易发生危险的异常电池。文章在常规 DBSCAN 算法的基础上进行改进，针 对其对输入参数敏感的缺陷，提出一种基于数据分布特性自适应确定 MinPts 参数和 Eps 参数的改 进 DBSCAN 算法。将改进 DBSCAN 算法用于电池一致性分析，并利用电池充放电数据设计相关 实验进行验证，验证结果表明所提出的算法异常电池识别效果良好。

关键词: 动力电池组；DBSCAN 算法；离群检测；一致性；自适应