Optimal speed and torque estimations for improving the DTC dynamic performance of induction machines

Abstract

High-performance AC drives require accurate speed, flux, and torque estimations to provide a proper system operation. Thus, this thesis proposes a robust observer, i.e. Extended Kalman Filter (EKF), to offer optimal estimations of these components in order to improve the dynamic performance of Direct Torque Control (DTC) of induction motor drives. The selection and quality of EKF covariance elements have a considerable bearing on the effectiveness of motor drives. Many EKF-based optimization techniques involve only a single objective for the optimal estimation of speed without giving concern to the other variables. In addition, the optimization is performed on a complicated EKF structure. Nevertheless, in this study, both speed and torque are concurrently estimated. The work presents a new method to investigate the selection of EKF filters by using a Non-Dominated Sorting Genetic Algorithm-II (NSGA-II) developed for resolving problems with multiobjectives. Filter element selection is the process of improving the concurrent estimation of speed and torque in order to increase EKF accuracy and allow higher drive efficiency. The proposed multi-optimal EKF-based estimation observer is used in combination with the sensorless direct torque control of induction motor. The investigated results for the multi-objective optimization indicate that the speed optimization gives superior performance when compared to the optimal torque. Owing to the large computation time of EKF algorithm, it increases the sampling time of DTC which leads to an increase in the motor torque ripples. The thesis proposes a Constant Frequency Torque Controller (CFTC) to replace the hysteresis torque controller that offers constant switching frequency and reduces torque ripples. Moreover, the CFTC has the capability of continuous switching regardless of speed variation; hence, leading to a consistent rotation of flux. Consequently, improvement on speed estimation, particularly at low and zero speed regions is accomplished and enhancement on the dynamic performance of torque is achieved when the reference speed change is applied from 0 rad/s, on the condition that the EKF observer is accurately optimized. To verify the improvements of the proposed methods, simulation and experimentation as well as comparison with the EKF-based DTC with the hysteresis controller are carried out.