Scenario Generation and Reduction Method for Clustered Wind-Photovoltaic Joint Power Output

Multi-site joint wind and photovoltaic outputshave complex spatiotemporal correlations and cross-resource dependencies.Accurately characterizing these features is a key issue in large-scale renewable energy planning.To address this issue,this study proposes a joint output scenario generation and reduction framework that couples Least Squares Generative Adversarial Network (LSGAN),Vine Copula,and Fast Forward Selection (FFS).First,LSGAN constructed multi-site wind and photovoltaic output models to characterize single-resource spatiotemporal correlation features.Then,Vine Copula established a wind-photovoltaic joint distribution model to describe nonlinear dependency relationships between different resources.Finally,FFS reduced the generated scenarios and extracted representative operating modes while preserving statistical characteristics.A case study based on the clean energy base in the middle reaches of the Yalong River basin demonstrates the effectiveness of the method.The results show that the generated scenarios maintain consistency with real data in temporal correlations,spatial correlation structures,and cross-resource dependency relationships.The mean absolute error of the correlation matrix is 0.035,which is about 67.6% lower than that of the LSGAN method.The reduced representative operating modes retain joint distribution characteristics and extreme output features while effectively compressing massive scenarios.The proposed method provides statistically representative operating scenario inputs for renewable energy base capacity planning and regulation resource allocation analysis.