Supply–demand scenario generation of active distribution systems using conditional generative adversarial networks

The intermittency of distributed renewable energy resources (DRERs) and the fluctuation of load demand make the supply–demand scenario of active distribution systems significantly different from that of traditional radial distribution systems. Scenario generation is an effective tool for describing supply–demand uncertainty and is therefore of great significance for the planning and operation of active distribution systems with a high penetration of DRERs. This paper proposes a supply–demand scenario generation method for active distribution systems using conditional generation adversarial networks (CGANs). The proposed method is data-driven and utilizes neural networks to learn and represent distribution laws from data without revealing the expression probability distribution. Specifically, it first extracts the features information of historical scenarios as the conditional control vectors (CCVs), then embeds the CCVs to the input of the generator network and discriminator network to achieve conditional control. The proposed approach is verified by the actual time series wind power data and load profiles. Results show that the proposed method can efficiently generate supply–demand scenarios satisfying various statistical characteristics and capture complex spatial–temporal correlations.