基于支持向量机的低压串联故障电弧识别方法研究

Series Arc Fault Recognition Method Based on Support Vector Machine Approach

针对串联故障电弧发生时线路电流幅值较小传统线路保护装置不能进行有效检测的情况,提出一种基于支持向量机(SVM)的串联故障电弧识别方法。该方法利用自制的电弧发生装置模拟串联故障电弧,采集典型负载在正常回路和故障电弧回路中的电流数据,采用该数据训练基于支持向量机的串联故障电弧辨识模型。经实验证明,该辨识模型可以实现对典型线性负载和非线性负载回路中串联故障电弧的特征识别,最高识别准确率可达96%。该方法对硬件电路要求低,识别效率高,并且可以实现故障电弧波形的存储和处理,具有一定参考价值。

When arc fault occurs in the circuit the traditional circuit interrupters cannot detect series arc fault because of the low current value. This paper introduces a new recognition method of series arc fault which is based on Support Vector Machine （SVM） to solve this problem. First, current data of different kinds of loads are collected by a self-made are generator, based on which, an arc fault SVM elassifier is trained, the accuracy of whieh is then tested by experiments carried out in linear and non-linear loads circuits collectively. It turns out that the SVM approach is an effective way to distinguish the series arc fault with the highest accuracy of 96%. The SVM approach is useful to deteet arc fault with a high efficiency and low requirement of hardware, meanwhile it can also save and process the current waveforms.