摘要
分布式能源入网规模的日益增大对传统过流继电器的故障检测及分类产生了重大影响。通过新兴的图学习技术构建了能有效检测分布式电网故障的时空递归图神经网络模型。该神经网络结构可以通过检测母线电压单元数据来提取时空特征,并根据数据的时空特征进行故障事件检测、故障类型分类、故障相位识别及故障定位。在IEEE 123节点系统上进行仿真模拟。结果表明,所提的基于电压测量的故障诊断策略与已有的传统方案相比具有较高的精度。所提策略仅需要提取电压信号而非电流信号,不受继电器安装数量的限制,因此所提策略更具有实操性与通用性。
The increasing scale of distributed energy network has had a significant impact on the fault detection and classification of traditional overcurrent relays.A spatiotemporal recursive graph neural network model that can effectively detect faults in distributed power grids was constructed through emerging graph learning techniques.The neural network structure can extract spatiotemporal features by detecting bus voltage unit data,and perform fault event detection,fault type classification,fault phase recognition,and fault localization based on the spatiotemporal features of the data.The simulation was performed on an IEEE 123 node system.The results indicate that the proposed fault diagnosis strategy based on voltage measurement has higher accuracy compared to existing traditional schemes.The proposed strategy only needs to extract voltage signals rather than current signals,and is not limited by the number of relays installed,therefore,the proposed strategy is more practical and versatile.
作者
刘伟
蒙永苹
张明媚
欧睿
许懿
Liu Wei;Meng Yongping;Zhang Mingmei;Ou Rui;Xu Yi(State Grid Chongqing Electric Power Co.,Ltd.,Chongqing 400014,China)
出处
《电气自动化》
2024年第3期104-107,112,共5页
Electrical Automation
基金
国家电网有限公司科技项目(SGCQWZOOHTMM2001367)。
关键词
故障检测
故障分类
数据挖掘
神经网络
分布式电网
fault detection
fault classification
data mining
neural network
distributed grid
