基于GRU网络的柔性直流配电线路保护方案  

Protection Scheme for Flexible DC Distribution Lines Based on GRU Network

作　　者：杨冬锋 李绍伟 刘晓军[1] 陈盛开 YANG Dongfeng;LI Shaowei;LIU Xiaojun;CHEN Shengkai(Key Laboratory of Modern Power System Simulation and Control&Renewable Energy Technology(Northeast Electric Power University),Ministry of Education,Jilin 132012,Jilin Province,China;Guangzhou Power Supply Bureau of Guangdong Power Grid Co.,Ltd.,Guangzhou 510620,Guangdong Province,China)

机构地区：[1]现代电力系统仿真控制与绿色电能新技术教育部重点实验室(东北电力大学),吉林省吉林市132012 [2]广东电网有限责任公司广州供电局,广东省广州市510620

出　　处：《电网技术》2021年第10期3885-3894,共10页Power System Technology

基　　金：国家重点研发计划(智能电网技术与装备重点专项)(2016YFB0900600);国家电网有限公司科技项目(52094017000W)。

摘　　要：针对柔直配电线路保护存在的阈值整定困难、高阻故障时保护容易拒动和噪声干扰下灵敏性低等问题,提出一种基于门控循环单元(gate recurrent unit,GRU)深度学习模型的柔直配电线路保护方案。首先,分析了基于限流电抗器的线路短路故障的幅频特性,并利用广义S变换提取频域特性下的故障特征量,将其作为GRU网络的输入数据。然后,构建并训练GRU深度学习模型,提取时间序列下故障样本的深层次特征,并结合支持向量机(support vector machine,SVM)分类器,实现故障快速识别和隔离。最后,在PSCAD/EMTDC中搭建仿真模型,验证了保护方案的可行性,仿真结果表明该方案的抗干扰能力强,灵敏度高,满足直流保护的可靠性和速动性要求。For the problems of the difficulty in setting the thresholds,the operation failures caused by high impedance faults and the insensitivity under noise interference in the protection of the flexible DC distribution lines,a protection scheme for the flexible DC distribution lines based on the GRU deep learning model is proposed in this paper.Firstly,the amplitude-frequency characteristics of the short circuit faults based on the current-limiting reactors are analyzed,and the fault characteristic quantities under the frequency domain features are extracted through the generalized S-transform,which are taken as the input data of the gate recurrent unit(GRU)network.Then,the GRU deep learning model is constructed and trained to extract the deep features of the fault samples under time series.Combined with the support vector machine(SVM)classifier,the fast fault identification and isolation are realized.Finally,the simulation model is built in the PSCAD/EMTDC to verify the feasibility of the protection scheme.The simulation results show that the scheme has strong anti-interference ability and high sensitivity,which meets the requirements of the reliability and quickness in DC line protection.

关 键 词：直流配电系统 限流电抗器 GRU深度学习模型 故障识别 故障选极 

分 类 号：TM721[电气工程—电力系统及自动化]