摘要
现有变压器故障诊断方法忽视了原始输入信号中噪声对诊断准确率的影响。针对上述问题,提出了一种基于经验小波变换(Empirical Wavelet Transform,EWT)和改进卷积神经网络(Im⁃proved Convolution Neural Network,ICNN)的智能故障诊断方法。基于经验小波变换,将原始输入信号在傅里叶频谱进行分割,然后利用小波滤波器对分割后的信号进行滤波,大幅降低原始信号中的噪声比例,进而输出具有高信噪比的信号集合;将信号集合输入改进卷积神经网络进行训练,以快速获得变压器故障诊断模型。通过算例对所提模型进行验证,结果表明:所提诊断模型能够有效识别变压器故障状态,110 kV变压器5种典型故障的平均诊断准确率高达94%。
The existing transformer fault diagnosis methods ignore the influence of noise in the original input signal on the diagnosis accuracy.To solve these problems,an intelligent fault diagnosis method based on empirical wavelet transform and improved convolution neural network is proposed.Based on the empirical wavelet transform,the original input signal is segmented in the Fourier spectrum,and then the wavelet filter is used to filter the segmented signal,greatly reducing the noise proportion in the original signal,and then the signal set with high signal⁃to⁃noise ratio is output;then,the signal set is input into the improved convolution neural network for training,so as to obtain the transformer fault diagnosis model.The test results show that the proposed model can effectively diagnose the fault state of transformer,and the average diagnostic accuracy of five typical faults of 110 kV transformer is 94%.
作者
李阳
路鹏
朱伯涛
李会彬
LI Yang;LU Peng;ZHU Botao;LI Huibin(State Grid Xingtai Power Supply Company,Xingtai 054000,China)
出处
《电子设计工程》
2021年第8期140-144,共5页
Electronic Design Engineering
关键词
变压器
故障诊断
经验小波变换
卷积神经网络
transformer
fault diagnosis
empirical wavelet transform
convolutional neural network
