With the continuous increase in the penetration of renewable energy

the accurate characterization and effective management of its inherent uncertainty have become crucial for the planning and secure operation of modern power systems. To address this challenge

this paper proposes a renewable energy scenario generation method based on a Wasserstein generative adversarial network with gradient penalty (WGAN-GP)

integrated with multi-scale convolution and a multi-head attention mechanism. The method efficiently extracts features of renewable energy output at different time scales through a parallel multi-scale convolutional network. It incorporates a multi-head attention mechanism to capture long-term dependencies

thereby significantly enhancing the model's capability to model complex dynamics. Furthermore

a comprehensive interpretability analysis framework is established

combining SHapley additive exPlanations (SHAP) attribution analysis with attention weight visualization to gain deep insights into the model’s decision-making process. Experimental results demonstrate that the proposed method provides an effective and interpretable tool for the uncertainty analysis of renewable energy output.