Reinforcement Learning Approach with Deep Deterministic Policy Gradient DDPG-Controlled Virtual Synchronous Generator for an Islanded Microgrid

As the integration of renewable energy sources (RES) into the grid continues to rise, replacing the conventional synchronous generators, the total inertia of the grid correspondingly decreases. This trend presents a potential risk to the stability of the grid’s frequency. This paper presents a novel virtual synchronous generator (VSG) controller based on the Deep Deterministic Policy Gradient (DDPG) reinforcement learning algorithm. The controller’s dynamic performance is evaluated using Matlab/Simulink software and compared with a conventional PI controller under various load scenarios. The results revealed that the proposed controller offers an enhanced inertial response. Two scenarios are considered: In the first, the VSG equipped with the proposed controller served as the sole power source for the microgrid and demonstrated superior performance under static and dynamic load conditions. The second scenario integrated the controller into a microgrid, supplemented by a synchronous generator and a renewable energy source. The controller successfully provided inertia support and functioned effectively as a secondary controller working on restoring the frequency to the nominal values.