Source-load Flexible Coordination Power Dispatch Based on Learning Optimization Method

To fully exploit the dispatching potential on the load side and better cope with the operational uncertainties brought by the large-scale integration of renewable energy sources into, this study proposes a source-load flexible coordinated dispatching optimization method based on proximal policy optimization(PPO). We consider a mode where the underlying load users are aggregated by a load aggregator(LA) to participate in the real-time active power dispatching in the grid. The LA is modeled as a virtual power plant (VPP), the adjustable capacity varies with user access and changes in basic electricity demands. The dispatching center can dynamically adjust the power consumption based on the adjustable capacity information and provide certain economic compensation. We construct a dispatching agent to achieve the dynamic power dispatching of generation units, the generated dispatching policy by this agent includes power setpoints for conventional power generation units and the virtual power plant. The optimization objectives of the dispatching agent are to achieve economic operation and enhance the integration of renewable energy while ensuring safety constraints, test cases demonstrate that the proposed method performs well in terms of the aforementioned objectives.