基于时域脉冲特征量的神经网络方法在变压器局放模式识别中的应用

对电力变压器特高频局放单个时域脉冲信号做小波包分解,以其能量前三位和熵值前三位加上能量和熵值共8个特征量作为神经网络的输入特征量,分别研究了BP神经网络、Elman神经网络和PNN神经网络对4种典型变压器局部放电信号的模式识别效果。通过对实验室实测信号的识别,证明了采用此特征量的神经网络识别方法简单有效实用,从而为电力变压器局放信号的识别提供了有效的参考。

Application in PD Signal Pattern Recognition of Power Transformer of Neural Network Based on The Character of PD Pulse of Time Field

Based on the message of energy and entropy of the result of wavelet transformer of single UHF pulse signal,8 characters which including the top three serial number of energy and entropy, respective researches are brought forward as input of neural network, and BP network,Elman network and PNN network are used to do the pattern recognition of the four types of PD signals of power transformers so as to get good pattern recognition result. At the same time, the parameters of neural network using above 8 characters can be set more loosely than traditional neural network using statistical messages as characters.So it can be an effective way to the engineering use in pattern recognition of PD signals of power transformers.