Wind Power Scenario Generation for Microgrid Day-Ahead Scheduling Using Sequential Generative Adversarial Networks

Abstract

With the rapid increase in the distributed wind generation, considerable efforts have been devoted to the microgrid day-ahead scheduling. The effectiveness of these methods will highly depend on the selection of the uncertainty set. In this work, we propose a distribution-free approach for wind power scenario generation using sequential generative adversarial networks. To capture the temporal correlation, the proposed model adopts the long short-term memory architecture and uses the concept of generative adversarial networks coupled with reinforcement learning to guide the learning process. In contrast to the existing methods, the proposed model avoids manual labeling and captures the complex dynamics of the weather. The proposed scenario generation method is applied to the wind power dataset of Bonneville Power Administration. The results indicate that the scenarios generated by our model can characterize the variability of wind power in a better manner. The generated scenarios are compared with those produced by Gaussian distribution and kernel density estimation, in terms of two statistical scores.