基于深度置信网络和随机森林的电力扰动检测方法  

Power Quality Disturbance Detection Method Using Deep Belief Network

作　　者：胡婧 周洋 何志强[1] 刘兵 卢凯 HU Jing;ZHOU Yang;HE Zhiqiang;LIU Bing;LU Kai(Hohhot Power Supply Bureau,Hohhot 010020,China)

机构地区：[1]呼和浩特供电局,内蒙古自治区呼和浩特010020

出　　处：《供用电》2020年第9期17-22,28,共7页Distribution & Utilization

基　　金：内蒙古电力公司2020年第一批科技计划项目(2020-20)。

摘　　要：电力扰动数据中包含着大量与设备异常运行状态有关的信息,某些设备的扰动特征很微弱,但却会对设备的运行产生一定的影响,对这些扰动数据的检测将有利于准确感知设备运行状态,防止设备发生故障。提出了一种基于深度置信网络的电力扰动数据检测方法,利用深度置信网络强大的特征提取能力,自动学习出蕴含在扰动数据中的隐藏特征,摆脱人为提取特征的依赖。最后将学习到的特征作为随机森林的输入,实现对电力扰动数据的检测。利用某变电站记录到的实测数据和PSCAD/EMTDC中的仿真数据进行了验证,结果表明所提方法能够准确检测出电力扰动数据,证实了利用深度置信网络检测电力扰动数据的可行性。Power disturbance data contain a lot of information related to the abnormal operation status of equipment.The disturbance characteristics of some equipment are very weak,but it will have a certain impact on the operation of the equipment.The detection of these disturbance data will help to accurately perceive the operation status of the equipment and prevent equipment failure.This paper proposes a detection method of power quality disturbance data using the deep belief network.By using the powerful feature extraction ability of the deep confidence network,the hidden features contained in different power disturbance data can be learned automatically,which can get rid of the dependence of the artificial feature extraction.Finally,the features learned are used as input of random forest to detect the power disturbance data.The test data recorded in a substation and the simulation data in PSCAD/EMTDC are used to verify.The results show that the proposed method can accurately detect the power disturbance data,and the feasibility of detecting the weak power disturbance data by using the deep belief network is verified.

关 键 词：电力扰动 扰动检测 特征提取 数据分析 深度置信网络 

分 类 号：TM71[电气工程—电力系统及自动化]