摘要
采用PSPICE软件建立变压器频率响应等效电路,分析不同变形故障情况幅频特性曲线与正常情况的不同以及不同变形程度的同种故障情况下曲线的变化,利用小波变换法找到频率响应曲线的谐振点和幅值的变化情况,从而提取出特征向量,构造神经网络的训练样本。依据变压器绕组故障识别诊断的特点对BP神经网络的结构进行了设计和实例验证。最后,对实际变压器常见的两种变形方式(轴向变形和径向变形)进行检测和识别,并通过随机实验数据验证了该系统对变压器绕组变形故障识别诊断的实际可行性。
The paperestablishes transformer equivalent circuit of the frequency responseby PSPICE software.It analyzes the difference between different deformation fault condition and normal condition in amplitude-versus-frequency curve and the changes of the same fault condition in different deformation degree.Wavelet transform method is used to find the frequency response curve of the resonant point and changes of the amplitude to extract feature vector and make the training sample of neural network.It designs the structure of BP neural network respectively based on the characteristics of the transformer winding fault identification diagnosis and it makes case verification for the trained neural network.In this paper,finally adopts the established BP neural network identification and diagnosis system to detect and identificate two common kinds of deformation in the actual transformer(axial deformation and radial deformation).And the system is verified by two randomized experiment data to identify practical feasibility for the fault diagnosis of transformer winding deformation.
作者
邹竟成
ZOU Jingcheng(China Railway Siyuan Survey And Design Group CO..LTD,Wuhan,430000,China)
出处
《铁道勘测与设计》
2021年第4期61-67,共7页
Railway Survey and Design
关键词
变压器绕组变形
频率响应法
小波变换
神经网络
Transformer winding deformation
Frequency response method
Wavelet transform
Neural network
