Renewable Scenario Generation Under Extreme Meteorological Conditions via Improved Conditional Generative Adversarial Networks

Scenario generation is a crucial method for characterizing the uncertainty of renewable energy sources (RESs). Driven by climate change, more frequent extreme meteorological events complicate the uncertainty of RESs and challenge the safe operation of high-penetrated renewable power systems. In response, this paper proposes an improved Conditional Generative Adversarial Networks (CGANs) approach to address generate renewable scenarios under extreme meteorological conditions. Firstly, the problem of renewable scenario generation under extreme meteorological conditions and the encoding rules for meteorological conditions are established. Subsequently, the Wasserstein distance and gradient penalty are employed to refine the CGAN, and a novel class and weight assignment method is devised to tackle the scenario class imbalance and capture fluctuations in renewable power generation. Case studies conducted on an open-source dataset demonstrate that the proposed approach can generate corresponding scenarios under given conditions and effectively imitate the distribution of historical scenarios.