ANN-based DTC scheme to improve the dynamic performance of an IM drive

In addition to simplicity, the DTC of the IM allows a good torque control in steady-state and transient operating conditions. However, high torque ripple is produced, which is reflected in speed estimation responses. It also increases acoustical noise and makes the voltage source inverter operated in high and variable switching frequency, requiring a high sampling frequency. In this paper, a simple effective ANN-based DTC of the induction machine is proposed. Neural networks with a simple architecture are designed and implemented in influential points of the direct torque control block of the three phase induction machine in order to improve its dynamic performance while preserving the DTC structure simplicity as much as possible. In particular, this paper proposes to reduce, on the one hand, the torque ripple and commutation frequency. Neural comparators are used to select the appropriate bandwidth for the torque and flux hysteresis controllers. Their aim is to optimize the ripple level in the developed torque and flux. On the other hand, a neural speed controller is designed in order to improve the system ability to respond rapidly to changes in process variables and mitigate the effects of external perturbations. The performance of the proposed control are tested in simulation and highlighted by comparing to the conventional DTC control based on conventional hysteresis comparators and conventional PI speed controller. The obtained results show the feasibility and good performances of the proposed control.