基于小波与栈式稀疏自编码器的电力电缆早期故障定位方法研究  

Research on incipient fault location method of power cable based on wavelet neural network

作　　者：李胜辉 白雪 董鹤楠 刘宁[2] Li Shenghui;Bai Xue;Dong Henan;Liu Ning(State Grid Liaoning Electric Power Co.,Ltd. Electric Power Research Institute Shenyang 110006 ,China;College of Electrical Engineering & Information Technology,Sichuan University,Chengdu 610065 ,China)

机构地区：[1]国网辽宁省电力有限公司电力科学研究院,沈阳110006 [2]四川大学电气信息学院,成都610065

出　　处：《国外电子测量技术》2019年第5期146-151,共6页Foreign Electronic Measurement Technology

摘　　要：随着电力电缆的应用日益广泛,电力电缆早期故障的发生日趋频繁,对早期故障定位精度以及速度的要求日益提高。为实现电力电缆早期故障的定位,结合早期故障特征,提出了一种基于小波变换方法结合栈式稀疏自编码器(SAE)的电力电缆早期故障定位方法。该方法利用小波变换对故障信号进行特征提取,并将提取到的特征作为SAE的输入,依靠SAE网络强大的学习和预测能力,实现故障特征信号到故障距离的一一对应,解决电力电缆早期故障定位问题。仿真结果表明,利用小波变换提取到的故障特征训练后的神经网络可以快速准确的实现对电力电缆早期故障的定位。With the increasing application of power cable, incipient failure of power cable is becoming more frequent, the requirements for the accuracy and speed of incipient fault location are also increasing. In order to achieve the location of incipient fault for power cable, combined with the features of incipient failures, this paper proposes an incipient fault location method for power cable based on wavelet transform, combined with stacked autoencoder. This method uses wavelet transform to extract low-level features of fault signals, and uses the extracted features as input of the stacked autoencoder. By using the powerful learning and predictive capabilities of neural networks, the relationship between the fault signal and the fault distance has been found, and the problem of incipient fault location of the power cable has been solved. Simulation results showed that the neural network, trained by fault features extracted by wavelet transform, can quickly and accurately locates the incipient faults of power cable.

关 键 词：早期故障 小波变换 特征提取 稀疏自编码器 定位 

分 类 号：TM247[一般工业技术—材料科学与工程]