Data-Driven TD3 Control of IM Considering Magnetic Saturation and Temperature Effect

Abstract

Induction machines (IM) are still widely used in the industry due to their advantages, such as low maintenance requirements and improved robustness. The field-oriented control (FOC), direct torque control (DTC), and model predictive control (MPC) techniques are used to control IM in high-performance control applications. The common disadvantage of these control techniques is that the control performances are negatively affected by changes in machine parameters, and machine parameters vary non-linearly depending on the magnetic saturation and temperature. To solve this negative affect, the control technique can be optimized by using a parameter estimation methods. Another solution to eliminate these negative effects is to design a reinforcement learning (RL)-based controller that regulates the control variables without the knowledge of machine parameters. In this study, IM speed control is performed using a twin-delayed deep deterministic policy gradient (TD3) agent. The dynamic and steady-state performance of the designed controller are compared with the traditional control techniques. Extensive simulation results have shown that the dynamic and steady-state performance of the designed controller is better than other control techniques.