基于混沌系统和离散小波变换的卷积神经网络的电力电缆故障诊断

Power Cable Fault Diagnosis Based on Chaotic System and Discrete Wavelet Transform Convolutional Neural Network

针对传统电力电缆特征提取方法存在信息冗余及故障模型诊断不精准的问题,提出了一种基于混沌系统和离散小波变换的卷积神经网络的故障诊断算法,即采用离散小波变换对采集的局部放电信号进行滤波,采用洛伦兹混沌系统建立动态误差散布图以提取故障特征,最后通过卷积神经网络(CNN)进行故障识别。结合4种典型电力电缆绝缘故障及测试平台进行验证,结果表明,所提算法能够快速准确地识别电力电缆的故障状态,识别准确率达到97.5%,证明了所提算法的可行性和有效性,其能够为电力电缆的故障诊断提供一定的参考价值。

Aiming at the problems of information redundancy and inaccurate fault model diagnosis in traditional power cable feature extraction methods,this paper proposes a fault diagnosis algorithm of convolutional neural network based on chaotic system and discrete wavelet transform,that is,using discrete wavelet transform to filter the collected partial discharge signal,and using Lorentz chaotic system to establish dynamic error scatter diagram extraction.The fault features are finally identified through a convolutional neural network(CNN).Combining four typical power cable insulation faults and test platforms for verification,the results show that the proposed algorithm can quickly and accurately identify the fault status of power cables,and the recognition accuracy rate reaches 97.5%,which proves the feasibility and effectiveness of the proposed algorithm.It provides a certain reference value for the fault diagnosis of power cables.