Deep Reinforcement Leaning-based Data Driven Method for Fast Solution of Stability Constrained Optimal Power Flow in Isolated Microgrids

An increasing penetration of intermittent renewable energy sources has introduced significant uncertainties and instability issues into microgrids. To rapidly and economically respond to the changes in isolated microgrids operating state, this manuscript introduces a deep reinforcement leaning (DRL)-based data-driven approach for quickly solving the stability-constrained optimal power flow (SC-OPF) in microgrids. The method in question frames the SC-OPF as a task of constraint adherence within the training regimen of a DRL agent, with the agent’s action-value function being enhanced by the incorporation of small signal stability constraints. Numerical evaluations were conducted on modified IEEE 33-bus systems. The outcomes demonstrate that the proposed agent is capable of swiftly (in real-time) securing high-quality SC-OPF solutions that adhere to stability constraints.