The rapid growth of wind and solar energy poses new challenges to safe and reliable system operation. Effectively characterizing and generating high-risk wind and photovoltaic (PV) power output scenarios is therefore essential for system risk assessment and preventive dispatch and control. However, existing scenario generation methods either rely on predefined probability distributions or focus narrowly on extreme output levels, failing to comprehensively reflect system-level operational risk induced by renewable energy. To this end, a power system optimal dispatch model and a flexibility indicator system mainly incorporating system ramping and transmission margins are established. Thereafter, analytic hierarchy process (AHP) and the entropy weight method (EWM) are used to fuse indicators into a quantitative operational risk index. Historical wind and PV scenarios are evaluated through the dispatch model to generate risk-labeled samples, based on which a conditional generative adversarial network (cGAN) is trained to produce wind and PV power output scenarios with specified risk levels. Case studies verify that the risk labels constructed can effectively guide the subsequent conditional generation model and scenarios corresponding to a given risk level can be effectively generated by the model.
Zhou et al. (Tue,) studied this question.