Fatih Korkmaz, M.Faruk Cakır, Ismail Topaloglu and Rıza Gurbuz, Cankırı Karatekin University, Turkey

This paper investigates solution for the chronically and the biggest problem of direct torque control scheme: high torque ripple. Otherwise, another main problem faced in direct torque control method is difficulties due to complex algorithm to get high performance control for industrial motors. The purpose of this paper is to simplify the control structure by using artificial neural networks learning abilities and to investigate the affects of this structure on torque performance of motor. For this purpose, two different artificial neural networks have been suggested replacing the optimal switching vector selection and flux sector determination process of conventional direct torque control method. Matlab/Simulink based numerical simulations have been carried out to compare motor performances with conventional control structure and proposed artificial neural network based structure. It has been observed that the dynamic response of motor is faster and torque ripple and the controller complexity of the conventional control system has been reduced with the proposed technique.

Artificial neural networks, Direct torque control, PMSM control, Vector control