Identification of feasible regions for virtual power plants based on AST-GCN

Virtual Power Plants (VPPs) integrate diverse distributed energy resources (DERs) such as solar panels and wind turbines to optimize performance and enhance grid stability. The inherent variability and complexity of these resources make scheduling a challenging task. This study proposes the use of an Attribute-Augmented Spatiotemporal Graph Convolutional Network (AST-GCN) to accurately predict power generation and identify feasible scheduling regions within a VPP. The AST-GCN model captures both spatial and temporal dependencies of DERs, enabling precise and dynamic scheduling. By incorporating load prediction errors into the feasible region determination using the Minkowski sum method, this approach can characterize the feasible region of VPP. A case study employing a 100-day dataset demonstrates the model's effectiveness. The results highlight the potential of AST-GCN in supporting flexible and reliable resource scheduling in VPPs.