| 引用本文: | 汪颖,卢宏,杨晓梅,肖先勇,张文海.堆叠自动编码器与S变换相结合的电缆早期故障识别方法[J].电力自动化设备,2018,(8): |
| WANG Ying,LU Hong,YANG Xiaomei,XIAO Xianyong,ZHANG Wenhai.Cable incipient fault identification based on stacked autoencoder and S-transform[J].Electric Power Automation Equipment,2018,(8): |
|
| 摘要: |
| 将深度学习概念应用到电缆早期故障识别中，提出结合S变换与堆叠自动编码器(SAE)的电缆早期故障识别方法。通过对故障相电流进行S变换，将获得的S变换模时频矩阵分为低、中和高频段。求取对应频段的能量熵和奇异熵等特征量，并组成特征向量后，将时频域特征向量作为SAE网络的输入，经过预训练和参数微调，得到最优训练参数。利用构建好的网络从输入数据中挖掘有用信息，从大量扰动中识别电缆早期故障。仿真结果表明，与传统模式识别方法相比，所提方法的精度更高。 |
| 关键词: 电缆 电缆早期故障 S变换 奇异熵 能量熵 深度学习 堆叠自动编码器 |
| DOI：10.16081/j.issn.1006-6047.2018.08.017 |
| 分类号:TM247 |
| 基金项目: |
|
| Cable incipient fault identification based on stacked autoencoder and S-transform |
|
WANG Ying1, LU Hong1, YANG Xiaomei1, XIAO Xianyong1, ZHANG Wenhai2
|
|
1.School of Electrical Engineering and Information, Sichuan University, Chengdu 610065, China;2.State Grid Sichuan Maintenance Company, Chengdu 610042, China
|
| Abstract: |
| A cable incipient fault identification method based on S-transform and SAE(Stacked AutoEncoder) is proposed applying the concept of deep learning. The S-transform of fault phase current is carried out to obtain S-transform time-frequency matrix in low-,medium-and high-frequency bands. The energy entropy and singular entropy in the corresponding frequency band are calculated and then combined as the eigenvector. Taking the eigenvector as the input, the SAE network are pre-trained and its parameters are adjusted, and the optimal training parameters are obtained. The useful information extracted from the input data using the constructed network is employed to identify the cable incipient fault from lots of disturbances. Simulative results show that the proposed method has higher accuracy compared with the traditional pattern recognition method. |
| Key words: electric cables cable incipient fault S-transform singular entropy energy entropy deep learning stack autoencoder |
