Coordinated Energy and Reserve Sharing of Isolated Microgrid Cluster using Deep Reinforcement Learning

The integration of distributed energy resources (DER) into the power grid is facilitated by the micro-grids paradigm. However, the uncertainties associated with DERs affect the operational performance of Mgs. Considering the uncertainties of these resources, the effective/optimal scheduling of energy and reserves in the day-ahead timeframe is vital. Moreover, various micro-grids can be operated together as a micro-grid cluster which works in isolation from the power grid. In this paper, the concept of energy and reserve scheduling of isolated micro-grid clusters using deep reinforcement learning is presented to improve the economic performance of MGs via jointed energy and reserve sharing among them. The Markov Decision Process (MDP) is used to model micro-grid energy management to minimize the MG operating cost. To solve the MDP process deep reinforcement learning (DRL) approach is used. The contribution of each MG is quantified using a cooperative game approach. Simulations are carried out on a test system with 4 MGs, and results show the merits of the presented model.