Bi-Level Flexible Planning of Source-Grid-Load-Energy Storage Based on GAN
摘要
To address the high penetration of renewable energy, it is necessary to advance the planning and operation of both renewable and coal-fired power, and to develop a source-grid-load-storage flexible planning method that considers uncertainty scenario generation. First, generative adversarial networks (GANs) are employed to generate seasonal (spring, summer, autumn, and winter) scenarios of wind and solar uncertainties. Based on a further analysis of renewable energy uncertainty and coal-fired power flexibility retrofit characteristics, a two-stage stochastic planning model with both long-term and short-term time scales is established, incorporating renewable energy and coal-fired power. This model explores flexible source-grid-load-storage planning schemes that account for renewable energy decisions and coal-fired power flexibility retrofit. Scenario accuracy indicators and insufficient flexibility indicators are used to investigate the correlation between uncertainty scenario generation methods and planning outcomes. Finally, the proposed method is validated via a practical case system. The proposed scheme exhibits a remarkable deep peak-regulation effect during severe noon peak-regulation periods in summer and autumn. It can accommodate large-scale renewable energy integration in the future and achieves superior economic performance for the source-grid-load-storage system.