摘要
为了提高电网调度节能效率,设计了一种基于图卷积网络(GCN)方法的电网调度暂态稳定性测试方法。通过对多个网络卷积运算,得到多个网络中各结点及其邻近结点的特性参数变化规律,能够很好地适应系统结构。在IEEE39节点系统中验证该方法的准确性,研究结果表明:基于图卷积网络对于瞬态稳定预报模式达到了更好处理效果,漏判率和误判率均小于1%,正确率达到了99%以上。图卷积网络的改进效果要好于预期扫描效果。引入包含节点阻抗的图卷积网络后,每个节点特性在整个网络上进行传播,获得了更优性能。该研究显著提高电网调度过程中异常事件的判断能力,实现快速动态安全评估。
In order to improve the energy-saving efficiency of grid scheduling,a transient stability test method for grid scheduling based on graph convolution network(GCN)method is designed.Through the convolution operation of multiple networks,the change rule of characteristic parameters of each node and its neighboring nodes in multiple networks is obtained,which can be well adapted to the system structure.The accuracy of the method is verified in the IEEE39 node system,and the research results show that the graph convolution network based on the transient stability prediction mode achieves a better processing effect,with the omission rate and misjudgment rate less than 1%,and the correct rate reaches more than 99%.The improvement effect of graph convolutional network is better than the expected scanning effect.With the introduction of graph convolutional network containing node impedance,each node characteristic is propagated over the whole network and better performance is obtained.This study significantly improves the ability to judge abnormal events during grid scheduling and realizes fast dynamic security assessment.
作者
李明
李江
江波
杨俊强
章丽
王英
Li Ming;Li Jiang;Jiang Bo;Yang Junqiang;Zhang Li;Wang Ying(State Grid Changji Power Supply Company,Changji Xinjiang 831100,China)
出处
《现代工业经济和信息化》
2024年第10期105-106,113,共3页
Modern Industrial Economy and Informationization
关键词
电网调度
网络异常
图卷积网络
暂态稳定性
grid scheduling
network anomaly
graph convolutional network
transient stability
