基于离群值检测的铅酸电池故障诊断方法

单体电池故障是导致铅酸电池组在运行过程中突发失效的主要原因，传统的识别方法需要依靠高精度的测试设备和复杂的电池机理模型，部署成本高、使用范围有限。考虑到故障电池和正常电池之间的等效电阻、等效电容等内部参数的差异，会由于浮充电流在时间尺度上通过电压凸显出来，本文设计了一种基于离群值检测的故障诊断方法，该方法采用时间序列聚类分析技术，对各个单体电池运行过程中产生的电压时间序列进行相似性分析，通过判断相异度较高离群值对故障电池进行定位。为了减小长跨度的时间序列造成的计算爆炸风险，采用分段聚合近似表示方法对时间序列进行降维处理，加快了计算速度。该方法可直接应用于微控制器，有较强的实用性。

A Fault diagnostics method for lead-acid battery pack based on outlier detection

The cell failure is the main reason to make lead-acid battery pack lose efficacy suddenly during the process, the traditional identify methods should depend on high precision detection equipment and complicated battery mechanism model, which needs much production costs and limited usage. Considering the differences of internal parameters of equivalent resistance and equivalent capacitance between failure battery cell and normal one; due to float charging flow will show on the time scale by the voltage, there is one failure detection methods based on outlier detection in this article, it takes time series clustering analysis technology to make similar analysis of voltage time series that is produced by each single battery, and positions the failure battery with judging the larger differences of outlier. In order to reduce the calculating explosion risk that produced by long span time series, it will take approximate representation of piecewise polymerization to make dimension reduction operation for time series, and also improve the speed of accelerating. This method has strong practicability that could apply for micro-controller directly.