Scenario Generation of Renewable Energy Based on Improved Diffusion Model
- Resource Type
- Conference
- Authors
- Li, Sheng; Xu, Chuanyu; Wei, Lishen; Li, Ruijie; Ai, Xiaomeng
- Source
- 2023 IEEE Sustainable Power and Energy Conference (iSPEC) Sustainable Power and Energy Conference (iSPEC), 2023 IEEE. :1-7 Nov, 2023
- Subject
- Power, Energy and Industry Applications
Renewable energy sources
Correlation
Uncertainty
Probability distribution
Spatiotemporal phenomena
Complexity theory
Scenario generation
renewable energy
diffusion Model
scenario generation
unsupervised learning
- Language
- ISSN
- 2837-522X
The large-scale integration of renewable energy into the power system has resulted in increased complexity of its uncertainty. This complexity arises from the rapidly increased renewable energy sites and the intricate spatial-temporal correlations involved. To offer a more precise description of the uncertainty of renewable energy, a scenario generation method based on an improved diffusion model is proposed. This approach can autonomously learn the probability distribution and spatiotemporal correlation of the training data. It utilizes a U-Net model to progressively denoise noise and iteratively generate renewable energy scenarios. Experiments using historical power data from five wind farms and five $PV$ plants in Jiangsu Province show that the training process of the proposed method is more stable compared to existing methods. Moreover, the proposed method achieves higher accuracy in fitting probability distribution and captures the spatial correlation between multiple sites more accurately.