基于经验小波变换复合熵值与特征融合的故障电弧检测

Arc Fault Detection Based on Empirical Wavelet Transform Composite Entropy and Feature Fusion

针对电力系统交流配电线路中串联故障电弧易引发电气火灾且检测精度不高的问题,提出一种基于经验小波变换(empirical wavelet transform,EWT)复合熵值与信号特征融合的故障电弧诊断方法。首先搭建故障电弧实验平台,对典型负载实测电流归一化处理,利用经验小波变换进行频谱分割并提取出具有紧支撑的模态分量,根据燃弧前后信息熵熵减分析选取特征分量。为反映时频多域细节复杂度特征,提取时频域特征分量复合熵值与时域敏感特征组成多域高维特征,利用主成分分析(principal component analysis,PCA)选取累积贡献率高于90%的主元实现特征降维融合,最后输入概率神经网络(probabilistic neural network,PNN)验证检测精度。结果表明,融合特征较单域特征检测精度更高,选用负载最低诊断率达98%验证了该方法的有效性。

Aiming at the problem that the series fault arc in the AC distribution line of the power system maycause an electrical fire and the detection has a low accuracy,a fault arc diagnosis based on the empirical wavelet transform(EWT)composite entropy and the signal feature fusion is proposed.Firstly,the fault arc experimental platform is built,normalizing the measured current of the typical load.The spectrum is segmented by the empirical wavelet transform and the closely supported modal components are extracted.The characteristic components are selected according to the information entropy subtraction analysis before and after the arcing.To reflect the time-frequency multi-domain detail complexity features,the time-frequency domain feature component composite entropy and the time-domain sensitive features are extracted to form the multi-domain high-dimensional features.The principal components with a cumulative contribution rate higher than 90%are selected by the principal component analysis(PCA)to realize the feature dimension reduction fusion.Finally,the detection accuracy is verified by inputting a probabilistic neural network(PNN).The results show that the detection accuracy of the fused features is higher than that of the single domain featureswith the minimum load diagnosis rate is 98%,which verifies the effectiveness of this method.