一种基于SVM的直流电弧故障检测方法

A DC Arc Fault Detection Method Based on SVM

直流电弧故障检测方法能够在低压直流电源系统中实时准确地检测出电弧故障的发生。针对电动自行车电池系统,提出一种基于支持向量机SVM的直流电弧故障检测方法,该方法包括电流电压时域和频域特征量提取、特征量数据预处理和直流电弧故障SVM分类器三部分,可根据特征量通过SVM分类器来判别电弧故障。在SVM分类器设计中还采用粒子群优化算法结合K折交叉验证寻找最优SVM参数,以优化分类器的分类性能。实验结果表明,该方法能够准确有效地检测出发生的直流故障电弧。

This paper presents a DC arc fault detection method based on support vector machine （SVM） for the battery power supply system of electric bicycle.The method includes characteristics extraction in time and frequency domain,preprocess of the characteristic data,and a SVM classifier,which can detect the DC arc fault according to the input characteristic data.To find the optimal parameters of SVM,the particle swarm optimization algorithm is used with K-fold Cross Valida- tion（K-CV）.