摘要
故障诊断对于确定故障隔离和系统恢复非常重要,基于逆变器分布式能源的日益集成对传统过电流继电器的故障检测影响很大。利用新兴图像学习技术建立了用于故障诊断的时空递归图神经网络模型,其神经网络结构可从安装在关键节点的电压测量单元数据中提取时空特征,根据这些特征进行故障检测、故障类型/相位分类和故障定位等操作。与以往研究成果相比,时空递归图神经网络对故障诊断具有更好的泛化能力。此外,该方案提取电压信号,而不是电流信号,因此不需要在配电系统的所有线路上安装继电器。因此,所提出的方案具有通用性,不受安装继电器数量限制。在IEEE 33节点配电系统上对该方法的有效性进行综合评价,并与其他神经网络结构进行了比较,可为相关配电网故障提供借鉴。
Fault diagnosis is important to determine fault isolation and system recovery.The increasing integration of inverter-based distributed energy sources has a impact on the fault detection of traditional overcurrent relays.A temporal recurrent graph neural network model for fault diagnosis is developed using the emerging image learning technique.It's neural network architecture can extract spatio-temporal features from voltage measurement unit data installed at critical nodes.Based on these features,operations such as fault event detection,fault type/phase classification,and fault localization are performed.Compared with previous works,the spatio-temporal recurrent graph neural network has better generalization ability for fault diagnosis.In addition,the proposed scheme extracts voltage signals instead of current signals,so there is no need to install relays on all lines of the distribution system.Therefore,the proposed scheme is generic and not limited by the number of installed relays.The effectiveness of this method is evaluated on IEEE 33-node system,and compares with other neural network structures,which can provide reference for related distribution network faults.
作者
闫林凤
杨贵营
龙洁
王海港
YAN Linfeng;YANG Guiying;LONG Jie;WANG Haigang(State Grid Jiaozuo Power Supply Company,Jiaozuo,Henan 454150,China)
出处
《东北电力技术》
2024年第12期20-25,共6页
Northeast Electric Power Technology
基金
国网河南省电力公司科技项目(5217C020000F)。
关键词
故障检测
故障定位
微电网保护
深度神经网络
图像学习
fault detection
fault localization
microgrid protection
deep neural network
image learning
作者简介
闫林凤(1989),女,硕士,工程师,研究方向为配电线路开关保护运行管理。
