Energy Management of DC building based on Reinforcement Learning

The new distribution system transforms DC buildings from a single consumer in the power system to a new role of "production, consumption, and storage", which can effectively utilize flexible resources and promote the consumption of photovoltaics. We firstly use the difference between photovoltaic output and load demand power to adjust the DC bus voltage, and uses this voltage as the adjustment signal for flexible load, energy storage, and electric vehicle charging pile power, establishing an energy management model for DC flexible system; then we propose to use the reinforcement learning algorithm considering pulse experience replay for offline training and online optimization of the energy management model. Finally, the superiority of the proposed optimization strategy was verified through numerical simulations.