数字化变电站继电保护装置故障智能检测识别方法研究  

Research on intelligent fault detection and identification method of relay protection device in digital substation

作　　者：于洋 王同文 汪伟 杨瑞金 YU Yang;WANG Tongwen;WANG Wei;YANG Ruijin(Power Dispatching and Control Center,State Grid Anhui Electric Power Co.,Ltd.,Hefei 230022,China;Super High Voltage Branch,State Grid Anhui Electric Power Co.,Ltd.,Hefei 230009,China)

机构地区：[1]国网安徽省电力有限公司电力调度控制中心,安徽合肥230022 [2]国网安徽省电力有限公司超高压分公司,安徽合肥230009

出　　处：《电子设计工程》2024年第5期113-117,共5页Electronic Design Engineering

基　　金：国网安徽省电力有限公司科技项目(52120021N00L)。

摘　　要：为了提升变电站的数字化水平、提高故障的识别精度,文中对继电保护装置故障的智能检测算法进行了研究,并设计了一种基于灰色关联分析的堆栈编码器网络(GRA-SAE)。该网络以自动编码器为基本单元,可实现数据的自动化特征提取。通过支持向量机(SVM)对提取的特征进行分类识别,并引入灰色关联分析法优化了传统SAE算法中隐藏节点的结构,从而提升了算法的拟合能力。在自建数据集上进行的仿真结果表明,相较于SAE-SVM算法,该算法对异常数据与线路故障的识别精度分别提升了4%和10.27%,证明了GRA-SAE-SVM算法在复杂分类场景下具有更显著的优势。In order to improve the digitization level of the substation and improve the accuracy of fault identification,the intelligent detection algorithm of relay protection device fault is studied in this paper.A stacked autoencoder network based on(GRA-SAE)is designed.The network takes the automatic encoder as the basic unit and can realize the automatic feature extraction of data.The Support Vector Machine(SVM)algorithm is used to classify and recognize the extracted features.The structure of hidden nodes in the traditional SAE algorithm is optimized by introducing the grey correlation analysis algorithm,which improves the fitting ability of the algorithm.The simulation results on the self built data set show that,compared with SAE-SVM algorithm,the recognition accuracy of this algorithm for abnormal data is improved by 4%,and that for line fault is improved by 10.27%,which proves that GRA-SAE-SVM algorithm has more obvious advantages in complex classification scenarios.

关 键 词：堆栈编码器网络 支持向量机 数字化 继电保护 故障识别 PSCAD 

分 类 号：TN-9[电子电信]