Multi-Timescale Optimization Scheduling of Distribution Networks Based on the Uncertainty Intervals in Source-Load Forecasting

With the increasing integration of large-scale distributed energy resources into the grid, traditional distribution network optimization and dispatch methods struggle to address the challenges posed by both generation and load. Accounting for these issues, this paper proposes a multi-timescale coordinated optimization dispatch method for distribution networks. First, the probability box theory was employed to determine the uncertainty intervals of generation and load forecasts, based on which, the requirements for flexibility dispatch and capacity constraints of the grid were calculated and analyzed. Subsequently, a multi-timescale optimization framework was constructed, incorporating the generation and load forecast uncertainties. This framework included optimization models for day-ahead scheduling, intra-day optimization, and real-time adjustments, aiming to meet flexibility needs across different timescales and improve the economic efficiency of the grid. Furthermore, an improved soft actor-critic algorithm was introduced to enhance the uncertainty exploration capability. Utilizing a centralized training and decentralized execution framework, a multi-agent SAC network model was developed to improve the decision-making efficiency of the agents. Finally, the effectiveness and superiority of the proposed method were validated using a modified IEEE-33 bus test system.