Decision-oriented wind power scenario generation considering meteorological uncertainty

High renewable penetration introduces substantial uncertainty into day-ahead power system scheduling, motivating scenario generation for uncertainty-aware dispatch. Although data-driven generative models have improved scenario realism and diversity, most existing methods (i) are only weakly aligned with downstream operational objectives and (ii) treat numerical weather prediction (NWP) inputs as deterministic despite their inherent uncertainty. This paper proposes the decision-oriented and meteorological uncertainty-aware diffusion model (DO-MUDiff), a diffusion-based framework for day-ahead wind power scenario generation that jointly incorporates approximate decision-cost information and meteorological uncertainty. A denoiser tailored to renewable time series is trained with a differentiable economic dispatch (ED) proxy network to inject cost signals, while an adaptive perturbation mechanism and a sampling guidance strategy jointly model NWP uncertainty and balance trend consistency with diversity. Experiments evaluate both scenario quality and dispatch performance on the IEEE RTS-96 system. The results show that DO-MUDiff improves cost-effectiveness without sacrificing statistical accuracy compared with representative benchmarks.