Virtual vector-based neural network DTC scheme for dynamic performance improvement of dual-star induction motor drive

Abstract

Recently, direct torque control (DTC) of the dual-star induction motor (DSIM) has been widely appreciated over other conventional control techniques due to its numerous advantages, notably its simple structure, good dynamic performance, and excellent robustness. However, despite these qualities, it is often confronted with torque ripples and harmonic currents that limit its operational efficiency. To overcome these challenges and improve the global control of the drive system, this paper proposes a novel study to improve the performance of DTC for DSIM based on a set of three techniques. Firstly, by appropriately selecting two voltage vectors at each sampling period, the impact of current harmonics is considerably reduced, but torque and flux ripples remain significant. Secondly, the method above is combined with a switching table featuring three virtual voltage groups, significantly reducing torque ripples and harmonic losses. Finally, an intelligent control based on artificial neural networks (ANNs) will replace the speed regulator, the above switching table, the two-level hysteresis flux regulator, and the seven-level hysteresis torque regulator to select an optimal virtual voltage vector. The performance of the final technique shows the following advantages: further reduction of torque and stator flux ripples, less overshoot in speed and torque, and almost complete suppression of harmonic currents. The simulation results presented in this article confirm the effectiveness of the proposed technique.