Rolling unit commitment based on dual-discriminator conditional generative adversarial networks

Abstract

The growing penetration of wind power brings unprecedented randomness and uncertainty on the supply side of power systems, which makes the traditional unit commitment (UC) model uneconomical and unreliable. This paper proposes a rolling mechanism for unit commitment optimization based on wind power scenario generation. First, a novel scenario generation approach is presented using dual-discriminator conditional generative adversarial networks (DDCGAN) which is a variant of generative adversarial networks (GAN). Second, a rolling mechanism is applied to unit commitment optimization to decide the on/off state and output of thermal units according to the scenario set generated by DDCGAN. Every time period, with newly real-time data collected, scenarios are dynamically re-selected from the scenario set for adjusting the future planning. Moreover, we improve the selection strategies with time decay factors for acquiring more accurate scenario subset. Finally, a series of experiments were carried out on the modified IEEE RTS-96 and IEEE-118 bus test systems with wind generation. Experiment results prove that the proposed method is effective.