Model-Based Reinforcement Learning Method for Microgrid Optimization Scheduling

Due to the uncertainty and randomness of clean energy, microgrid operation is often prone instability, which requires implementation a robust adaptive optimization scheduling method. In this paper, model-based reinforcement learning algorithm applied optimal problem microgrids. During training process, current learned networks are used assist Monte Carlo Tree Search (MCTS) in completing game history accumulation, updating network parameters obtain strategies simulated environmental dynamics model. We establish environment simulator that includes Heating Ventilation Air Conditioning (HVAC) systems, Photovoltaic (PV) Energy Storage (ES) systems for simulation. The simulation results show microgrids both islanded connected modes does not affect effectiveness algorithm. After 200 steps, can avoid punishment exceeding red line bus voltage, after 800 result converges loss values value reward converge 0, showing good effectiveness. This proves proposed paper be