Advanced Machine Learning Approach of Power Flow Optimization in Community Microgrid

With the increasing penetration of distributed renewable energy (DERs), the electrical grid is experiencing, on a daily basis, rapid and massive fluctuations in power and voltage profiles. Fast and precise control strategies in real-time have played an important role to ensure that the power system operates at an optimal status. Solving real-time optimal power flow (OPF) problems while satisfying the operational constraints of the community microgrid (CMG) is considered a promising technique to control the fluctuations of renewable sources and loads. This paper adopts a new deep reinforcement learning algorithm (DRL), called Twin-Delayed Deep Deterministic Policy Gradient (TD3), to solve the real-time OPF with consideration of DERs and distributed energy storages (DESs) in the CMG. Training and testing of the algorithm are conducted on an IEEE 14-bus test system. Comparative results show the effectiveness of the proposed algorithm.