Deep reinforcement learning tuned type-3 fuzzy PID controller: AC microgrid case study

This paper proposes an adaptive type-3 fuzzy controller for controlling uncertain power systems. The controller, named type-3 fuzzy PID (T3FPID), has an input-output (I/O) relationship similar to the traditional PID controller but can better handle uncertainty and nonlinearity in systems. In the proposed design, controller parameters, such as I/O scaling factors (SFs) that affect transient and steady-state performance, are adjusted using the deep deterministic policy gradient (DDPG) reinforcement learning (RL) algorithm. The RL agent is initially trained offline under different operating conditions, then used online to tune the T3FPID controller's parameters. The proposed RL-tuned T3FPID controller's effectiveness is demonstrated by applying it to the load-frequency control problem of a microgrid system in simulations across various operating points. Results show that this approach outperforms other controllers.