Energy management optimization strategy of virtual power plant based on deep reinforcement learning

Abstract

The large-scale integration of new energy into the grid in the future will have an impact on the dispatching operation of the power grid. As a potential controllable resource, a controllable load is gradually being tapped, and the application of virtual power plants and artificial intelligence provides a solution. It can effectively perform resource aggregation scheduling and energy management optimization. Therefore, this paper proposes an energy management optimization strategy for virtual power plants based on deep reinforcement learning. The strategy first establishes three types of controllable load models and energy storage models and then combines the deep reinforcement learning Double-DQN algorithm with the internal model of the virtual power plant to construct an environment, action, and reward functions, and finally, conducts simulation and result analysis. The calculation example shows that this strategy can realize the optimal scheduling of the virtual power plant and improve the demand side response, and the experimental model in this paper has a higher yield curve compared with other learning algorithms, which verifies the effectiveness and rationality of the strategy. It is significant to inspire and promote the high-quality development of a green energy economy with new ideas of energy management models.