Decision-Driven Identification of Planning Scenarios for Large-Scale Renewable Energy Bases

In stochastic planning for large-scale renewable energy bases, scenario reduction is widely employed to mitigate computational burden without sacrificing planning optimality. Existing data-driven techniques predominantly minimize statistical discrepancy between original and reduced scenario sets, yet such fidelity offers no theoretical assurance of decision quality in the underlying planning problem. Therefore, this paper proposes a decision-driven three-layer framework for scenario reduction of renewable energy bases. In the upper layer, candidate representative scenario subsets are generated and screened through an optimization-driven search. The middle layer embeds a refined planning model for large-scale renewable bases, incorporating the operational characteristics of transmission corridors. The lower layer evaluates the obtained configuration against all original scenarios to assess the operational reliability. Finally, a hybrid solution strategy that combines an enhanced simulated annealing algorithm with the Gurobi solver is developed to coordinate these layers. Case studies based on real data of western China show that the proposed method attains a total cost of 6.141 billion yuan, reducing costs by 14.9 % and 19.5 % compared with K-means and KDE, respectively.