Weather-Informed Load Scenario Generation Based on Optical Flow Denoising Diffusion Transformer

With distributed energy resources (DER) continuing to grow, power systems are encountering increased challenges for resource inference and optimization in the distribution network. Given the intrinsic uncertainty and variability associated with various connected resources as well as their aggregated representation, generating diverse load scenarios is of great significance for system planning, operation, and control. However, traditional time series statistical methods, in the absence of supplementary meteorological data, have limited capabilities in accurately generating load scenarios. To address such issues, this paper proposes an Optical Flow Denoising Diffusion Transformer model (OFDDT) to generate representative load scenarios. Specifically, the optical flow denoising diffusion module is adopted to capture load fluctuation characteristics influenced by the temperature, while the Transformer module integrates variability and historical load features to generate load profile scenarios. Case studies using real data demonstrate that the proposed method is effective in such load scenario generation tasks, outperforming traditional generative artificial intelligence models.