基于IT降噪与CNN的电缆终端故障诊断方法

Fault Diagnosis Method of Cable Terminal Based on IT Noise Reduction and CNN

在电力电缆终端的制作过程中常因操作不当而产生各种缺陷,进而导致局部放电乃至绝缘击穿故障,严重影响了电缆的可靠运行,需要提出高效准确的缺陷诊断方法。基于实际故障工况设计制作了含有典型缺陷的电缆样品,获得了不同缺陷下的局部放电数据;提出了基于间隙阈值(intervalthreshold,IT)降噪与变尺度卷积神经网络(convolutional neural network,CNN)的电缆终端故障诊断方法;通过分解时域波形生成本征模态函数实现IT降噪,设计变尺寸整合层优化CNN特征提取过程。试验结果表明:所提方法的识别准确率达到了94.1%,在同等条件下的网络稳定性和迭代效率均有明细提升。方法消除了诊断过程中无效特征的干扰,避免了直接截取原始数据造成的特征损失。

In the production process of power cable terminals,various defects are often generated due to improper operation,which in turn leads to partial discharge and even insulation breakdown faults,which seriously affects the reliable operation of the cables.It is necessary to put forward an efficient and accurate defect diagnosis method.The cable samples with typical defects were designed and manufactured based on the actual fault conditions,and the partial discharge data under different defects were obtained;a cable terminal fault diagnosis method based on interval threshold(IT)noise reduction and variable scale convolutional neural network(CNN)was proposed;the IT noise reduction was realized by decomposing the time-domain waveform to generate the eigenmode function,and the variable size integration layer was designed to optimize the CNN feature extraction process.The experimental results show that the recognition accuracy of the proposed method is 94.1%.Under the same conditions,the network stability and iterative efficiency are improved.This method eliminates the interference of invalid features in the diagnosis process and avoids the feature loss caused by directly intercepting the original data.