摘要
随着智能电网的发展,识别和处理各种类型的过电压和故障具有越来越重要的意义。文中提出了一种基于稀疏自编码的过电压智能分类识别框架,依靠多层自编码器,实现了电力系统中实测铁磁谐振过电压波形的特征自提取,然后利用Softmax分类器完成精确分类,调节模型参数实现最优分类结果。该框架可应用到实际应用中,为建立过电压智能分类识别系统提供了一个全新的思路与方法。
With the development of the smart grid, it is more and more important to identify and deal with interference factors of various types of overvoltage, fault and power quality. In this paper, we present a significant unsupervised practical overvoltage identification system based on stacked autoencoders. The overvoltage waveforms applied in this experiment are collected from the online monitoring system in a substation. This unsupervised practical overvoltage identification system overcomes the problem that the method using manually designed overvoltage features is of low efficiency and unstable. The proposed method can applied in field and provides a new method for overvoltage classification.
作者
刘明军
张涵
熊浩
张煌竟
陈铁
司马文霞
黄敏
唐娟
LIU Mingjun;ZHANG Han;XIONG Hao;ZHANG Huangjing;CHEN Tie;SIMA Wenxia;HUANG Ming;TANG Juan(Maintenance Branch Company,State Grid Chongqing Electric Power Corporation,Chongqing 400039,China;State Key Laboratory of Power Transmission Equipment & System Security and New Technology,Chongqing University,Chongqing 400030,China)
出处
《高压电器》
CAS
CSCD
北大核心
2019年第10期70-75,共6页
High Voltage Apparatus
基金
国家重点研发计划(2017YFB0902701)~~
关键词
稀疏自动编码
特征提取
波形降维
铁磁谐振过电压
sparse automatic coding
feature extraction
waveform drop dimension
ferromagnetic resonance overvoltage
