基于改进GRU模型的配电网故障线路区段检测  

Detection of Fault Line Sections in Distribution Network Based on Improved GRU Model

作　　者：李世明 顾东健 余志文 赵瑞锋 黎皓彬 LI Shiming;GU Dongjian;YU Zhiwen;ZHAO Ruifeng;LI Haobin(Power Control Center,Guangdong Power Grid Co.,Ltd.,Guangzhou 510600,China;Guodian NARI Nanjing Control System Co.,Ltd.,Nanjing 211106,China)

机构地区：[1]广东电网有限责任公司,电力调控控制中心,广东广州510600 [2]国电南瑞南京控制系统有限公司,江苏南京211106

出　　处：《微型电脑应用》2023年第10期105-109,共5页Microcomputer Applications

摘　　要：为了实时快速准确检测配电网故障线路区段,提出了一种改进门控循环单元(GRU)模型的配电网故障线路区段检测方法。通过智能馈线表(SFM)采集线路区段两侧的电压和电流之间的角度来实时检测故障区段。当发生网络故障时,保护继电器触发断路器的跳闸命令,利用从配电网节点处所有SFM获取故障前样本和故障后样本角度数据,运用预训练改进GRU模型来检测故障线路区段。通过对IEEE-33总线系统中各种故障类型的仿真,结果表明:与循环神经网络(RNN)方法和长短期记忆人工神经网络(LSTM)方法相比,提出的方法检测精度高且速度更快,可用于实时检测配电网中的故障线路区段。In order to quickly and accurately detect the fault line section of distribution network in real time,a fault line section detection method of distribution network based on improved gating cycle unit(GRU)model is proposed in this paper.The angle between voltage and current on both sides of the line section is collected by intelligent feeder meter(SFM)to detect the fault section in real time.When a network fault occurs,the protection relay triggers the tripping command of the circuit breaker,obtains the angle data of pre fault samples and post fault samples from all SFMs at the distribution network nodes,and uses the pre training improved GRU model to detect the fault line section.Through the simulation of various fault types in IEEE-33 bus system,the results show that compared with cyclic neural network(RNN)method and short-term memory artificial neural network(LSTM)method,this method has higher detection accuracy and faster speed,and can be used to detect the fault line section in distribution network in real time.

关 键 词：配电网 门控循环单元 故障检测 线路区段 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TM715[自动化与计算机技术—控制科学与工程]