节点图和边图切换卷积驱动的快速静态安全分析方法

Switching Convolution of Node Graph and Line Graph-driven Method for Fast Static Security Analysis

针对预想事故在线分析中,实时性要求和时域方程计算复杂性之间的矛盾,提出了一种基于图卷积神经网络的潮流计算模型,进而实现快速预想事故分析。该方法包含了一个基于节点图和边图切换卷积的快速潮流计算模型,能够实现多场景的新能源出力、负荷数据与支路潮流、节点电压之间的非线性关系的快速拟合。针对不同预想事故导致的网络结构变化设计了反映拓扑结构变化的矩阵,避免了支路开断导致的边图节点消失,保证了模型的鲁棒性;同时,将荷源数据分离,构造了考虑荷源变化的输入特征向量。IEEE 39、118节点系统算例测试表明,所提出的模型能够适应N–1故障引起网络拓扑结构的变化和新能源的波动性,实现潮流快速拟合,可为预想事故在线分析提供新工具。

Aiming at the contradiction between the real-time requirement and the computational complexity of the time-domain equation,a power flow calculation model based on the graph convolutional neural network was proposed to realize the fast credible contingency analysis.The method includes a fast power flow calculation model based on switching convolution of node graph and line graph,which can quickly fit the nonlinear relationship between new energy resources,load data,branch power flow and node voltage in multiple scenes.According to the changes of network structure caused by different credible contingency,a matrix reflecting the changes of topology structure was designed to avoid the disappearance of edge graph nodes caused by branch breaking and ensure the robustness of the model.In addition,the load and power data were separated,and the input feature vector considering the load and power changes was constructed.The IEEE 30 and 118 bus test results show that the proposed model can adapt to the fluctuation of new energy and the changes of network topology caused by the N–1 fault,realize the rapid power flow fitting,and provide a new tool for the fast credible contingency analysis.