Enhanced Induction Motor Control Using Deep Reinforcement Learning and PWM with DTC

Abstract

This article presents an enhanced approach for controlling an induction motor, combining Deep Reinforcement Learning (DRL), Pulse Width Modulation (PWM), and a Speed Controller. Flux, torque, and ripples in the stator current are problems with the traditional Direct Torque Control (DTC). A 12-sector DTC technique is suggested to overcome these problems and get rid of torque ripple. In order to reduce the ripples in the stator current vectors brought on by white noise, an enhanced Q-learning-based Kalman filter (IQL-KF) is also employed. The rotation speed and flux control are achieved using fuzzy VIKOR. Simulation data with various load conditions, processing, and noise power covariances are utilized to estimate the effectiveness of the suggested technique. Comparison with the typical DTC approach demonstrates that the suggested method outperforms in transient situations, as assessed by the integrated time absolute error performance index, undershoot, and ripple of crucial parameters.