Hardware-in-the-loop Testing of a Deep Deterministic Policy Gradient Algorithm as a Microgrid Secondary Controller

Microgrids play an important role in distribution system modernization by coordinating distributed energy resources and local loads. Microgrids require well-designed adaptive controllers to ensure stability in all operating points, e.g., reinforcement learning algorithms. However, recently published papers have proposed reinforcement learning-based controllers only on a simulation basis, without showing results regarding the controller performance in hardware implementation. Thus, this paper focuses on a deep deterministic policy gradient-based microgrid secondary controller in a hardware-in-the-loop setup. Its performance was analyzed and compared with the grid’s standard controller, comprised by classical tools. The DDPG controller implemented in hardware reduced the steady-state error, increased the system speed response, and reduced deviations caused by the events.