Two-Stage Distributionally Robust Planning of Integrated Energy Systems with Scenario-Specific Wasserstein Ambiguity Sets

Uncertainty in renewable generation creates significant challenges for the planning and operation of integrated energy systems (IES). Insufficient allocation may reduce reliability, while overly conservative planning may increase costs. To address this problem, this paper develops a two-stage distributionally robust optimization (DRO) method with scenario-specific Wasserstein ambiguity sets. Typical wind and solar scenarios are first generated using the Kernel Density Estimation–Copula–Fuzzy C-Means(KDE-Copula-FCM) method. Based on these scenarios, first-moment and support constraints are introduced to construct multiple local ambiguity sets, which provide a more accurate representation of forecast error distributions compared with a single global set. In the proposed framework, the first stage determines capacity allocation of storage and controllable resources, while the second stage optimizes short-term operational scheduling under uncertainty. Case studies show that the method can reduce peak-period electricity purchases and renewable curtailment, while increasing renewable energy utilization. The results highlight a configuration mode of medium-scale storage combined with small controllable units. Compared with deterministic, robust, and traditional Wasserstein-DRO models, the proposed approach achieves a superior balance between economic efficiency and system reliability, demonstrating its effectiveness for IES planning and operation.