Research on Prior Knowledge-Guided Short-Term Wind/Solar Power Output Scenario Generation Method Considering Extreme Weather

To accurately reflect the temporal characteristics of renewable energy output, particularly the impact of extreme weather on the system, this study proposes a prior knowledge-guided scenario generation method for short-term wind/solar power output considering extreme weather. First, data identification is performed based on extreme weather indicators to construct normal and extreme weather datasets. Then, mixed density network (MDN) is employed to obtain distribution information for both datasets. Finally, noise sequences are derived from prior distribution information and input into a generative adversarial network (GAN) model along with historical measured data to generate wind/solar power output scenarios. The proposed method is validated using operational data from wind and solar farms in Zhejiang Province. Results demonstrate that the generated scenarios accurately reflect the spatiotemporal correlations of actual wind/solar power outputs while capturing the stochastic characteristics of power curves under extreme weather conditions, effectively representing the volatility of regional renewable generation. This approach provides valuable references for power system dispatch planning.