基于Hebb与BP神经网络组合的故障电弧辨识方法

Fault Arc Identification Method Based on Combination of Hebb and BP Neural Networks

针对低压交流故障电弧发生时造成的波形畸变,提出一种基于Hebb与BP神经网络结合且适用于多负载的故障电弧辨识方法。提取电流波形的傅里叶变换特征值,通过Hebb神经网络进行计算以实现负载分类,从而确定该负载所用的BP神经网络的权参数,再提取电流波形小波变换特征值,通过BP神经网络计算得出故障与否的辨识结果。仿真结果表明,算法解决了负载类型过多造成BP神经网络识别度下降的问题,使得故障电弧辨识方法的准确率和通用性得到了大幅度提升。

In view of waveform distortion caused by low voltage AC fault arc,an arc fault identification method applicable to multiple loads was proposed on the basis of combination of Hebb and BP neural networks. Fourier transform eigenvalues of the current waveform were extracted,and computation was made through the Hebb neural network so as to achieve load classification and determine weight parameters of the BP neural network used by the loads. Then,wavelet transform eigenvalues of the current waveform were extracted and computed through the BP neural network to identify whether there was a fault or not. Simulation results showed that this method could solve the problem of decrease of recognition rate of the BP neural network due to excessive load types,thus greatly improving the accuracy and versatility of the fault arc identification method.