Microgrid dispatching strategy based on deep reinforcement learning

The high proportion of intermittent energy sources such as wind and solar energy puts forward higher requirements for the flexibility of power system operation. Traditional model-based optimization algorithms rely heavily on the accuracy of the constructed model and prediction data, which is difficult to solve and has poor scalability. Based on this, this study proposes a collaborative optimization scheduling strategy for wind-solar-storage and adjustable load systems based on Double DQN deep reinforcement learning algorithm, which avoids the modeling of complex dynamic characteristics and wind-solar uncertainty factors of micro-grid systems, learns the operation strategy to maximize the long-term benefits of micro-grid systems through data-driven methods, and solves the optimal operation problem of micro-grids in real time.