Stochastic Scenario Generation for Wind Power and Photovoltaic System Based on CGAN

Abstract

With the high proportion accession of renewable energy, the uncertainty of the power system gradually increases. Scenario generation is an important method to describe the uncertainty of a high proportion of renewable power system, and plays an important role in the operation planning and scheduling of power systems. In this work, we proposed a wind power and photovoltaic random scenario generation method based on Conditional Generative Adversarial Networks (CGAN), Combined with the k-means clustering method to cluster and label the historical output power data, the mapping relationship between the noise input and the real output power data under the constraint of the clustering label is learned through the game mechanism of the generative adversarial network, different types of possible scenarios that retain features such as spatiotemporal correlations of real data are generated, the effectiveness and applicability of the proposed method are verified by experimental analysis of actual data. The simulation results show that CGAN can efficiently generate scenarios that conform to the characteristics of real data according to the conditions.