电力系统及其自动化学报2019，Vol.31Issue(7)：70-75,6.DOI:10.19635/j.cnki.csu-epsa.000199
基于自编码器和卷积神经网络的电能质量扰动分类
Classification of Power Quality Disturbances Based on Auto-encoder and Convolutional Neural Network
摘要
Abstract
The traditional classification methods for power quality disturbances often rely on expert experiences to ex?tract features,and the corresponding classification accuracy is limited. In this paper,a deep learning based classifica?tion method for power quality disturbances is proposed,which combines the sparse auto-encoder with a strong feature extraction capability and the convolution neural network. The proposed method includes two links,i.e.,unsupervised feature extraction and supervised disturbance classification. First,the high-dimensional input data are mapped onto a low-dimensional hidden variable feature by means of the encoder,and the new feature is restored to the original input signal in the same way. Then,the hidden variable output from the encoder is used as the feature,and the convolution network will output the types of disturbance. Simulation results show that the extracted features and classifier perfor?mance obtained using the proposed method are better than those obtained using the traditional ones.关键词
电能质量/扰动分类/稀疏自动编码器/卷积神经网络Key words
power quality/disturbance classification/sparse auto-encoder/convolutional neural network分类
信息技术与安全科学引用本文复制引用
李志军,王亚楠,安平,张鸿鹏,孙乐,徐铎..基于自编码器和卷积神经网络的电能质量扰动分类[J].电力系统及其自动化学报,2019,31(7):70-75,6.基金项目
中国南方电网有限责任公司重点科技资助项目(066601(2016)030101XT198) （066601(2016）
