Application of Reinforcement Learning in Optimal Control of Three‐Phase Voltage Source Inverter

ABSTRACT This work presents a model‐free reinforcement learning (MFRL) method to design a controller for a three‐phase voltage source inverter (VSI). The suggested controller consists of the actor and the critic elements. The actor component is responsible for the selection of actions, while the critic component evaluates the chosen action and provides feedback to enhance the learning process. The proposed controller has the ability to ensure an adaptive response in various scenarios of working such as changes in loads, unbalances in loads, and the presence of nonlinear loads or external disturbances. Moreover, the system's stability and convergence are mathematically determined via the application of Lyapunov stability theory. Finally, the effectiveness of the MFRL‐based controller, which incorporates an actor‐critic architecture, is proven through comprehensive and comparative simulations and analysis. The results demonstrate the potential of this solution as an effective method for obtaining robust and optimal performance in controlling three‐phase VSI.