CAJ | 학술논문

针对绝缘子污秽放电模式识别过程中声发射信号的特征参量维数过高的问题，采用主成分分析法对特征参量降维，利用提取到的绝缘子污秽放电声发射信号的特征参数构成原始特征参量矩阵，通过对原始特征参量矩阵进行K-L正交变换，产生了包含原始特征参量矩阵主要信息的K个主成分，最后利用小波神经网络进行绝缘子污秽放电的模式识别。结果表明：利用主成分分析法降低特征参量的维数，使分类器的结构更简单，小波神经网络比传统的BP神经网络具有更高的识别率和更优的识别效果。
침대절연자오예방전모식식별과정중성발사신호적특정삼량유수과고적문제，채용주성분분석법대특정삼량강유，이용제취도적절연자오예방전성발사신호적특정삼수구성원시특정삼량구진，통과대원시특정삼량구진진행K-L정교변환，산생료포함원시특정삼량구진주요신식적K개주성분，최후이용소파신경망락진행절연자오예방전적모식식별。결과표명：이용주성분분석법강저특정삼량적유수，사분류기적결구경간단，소파신경망락비전통적BP신경망락구유경고적식별솔화경우적식별효과。
In view of the problem that the dimension of characteristic parameters of acoustic emission signal is high in the process of pattern recognition for contamination discharge of insulators, we adopted the principal component analysis to reduce the dimension of characteristic parameters. A original character-istic parameter matrix was constructed by the characteristic parameters of acoustic emission signals extract-ed from the contamination discharge of insulator, and the original characteristic parameter matrix informa-tion was conducted K-L orthogonal transformation, then K principal components conaining main informa-tion of original characteristic parameters matrix were generated. Finally, the wavelet neural network was used to recognize the contamination discharge pattern of insulators. The results show that using the princi-pal component analysis method to reduce the characteristic parameters’ dimension can make the structure of classifier become simple, and the wavelet neural network has higher recognition rate and better recogni-tion result than that of the traditional BP neural network.