基于EGAT的高比例分布式光伏配电网的脆弱性辨识OA

To address the complexities in regulating and operating distribution grids with a high proportion of distributed photovoltaic(PV)systems,which challenges conventional methods in vulnerability identification,a novel methodology utilizing an edge-feature graph attention network(EGAT)is proposed for state estimation and further identification of vulnerabilities via over-limit indicators.Initially,busbars within the distribution network are conceptualized as nodes of a graph,with transmission lines serving as edges,thus forming a topological dataset grounded in their connectivity matrix.Subsequently,the power of each transmission line and the voltages at the connected busbars are consolidated into an edge feature matrix.The employment of EGAT layers facilitates targeted feature extraction and learning,where attention weights are attributed to each node and transmission line.In response to various operational scenarios of the PV distribution network,the EGAT model undergoes training with the integration of transfer learning techniques.Ultimately,by establishing indices for node vulnerability and line vulnerability,a thorough evaluation of the network's vulnerability under different scenarios is achieved.Demonstrative case studies validate the effectiveness of this approach in recognizing vulnerabilities within distribution networks with substantial integration of distri-buted PV systems,accurately identifying vulnerable nodes and lines induced by fluctuations in PV output and topological modifications.