Multi-microgrid Optimization Based on Hierarchical Reinforcement Learning

Abstract

A hierarchical reinforcement learning optimization method is proposed for multi-microgrid system. Decompose the multi-microgrid optimization problem into upper and lower layers for solving. The upper layer determines the energy storage optimization strategy of each microgrid and the power interaction strategy between microgrids. In lower layer each microgrid autonomously optimizes the power of distributed generation within the microgrid based on the upper layer policy, and feeds back reward signals to the upper layer to guide the update the policy of upper layer. The decoupling of multi-microgrid optimization problem in time and space is realized, and each microgrid can complete the optimization of multi-microgrid system based on its own state information, which effectively protects the internal privacy of each microgrid. And the reinforcement learning method and mathematical programming method are combined to solve the model, which can not only improve the training speed, but also effectively improve the accuracy of the solution. Finally, a numerical example is given to verify the effectiveness of the proposed method, and compared with the traditional method, it is proved that the solution speed and accuracy of the proposed method are greatly improved.