IMPROVEMENT OF DIRECT TORQUE CONTROL SYSTEM USING RADIAL BASIS FUNCTION NEURAL NETWORK AND FUZZY CONTROL TECHNIQUES

Abstract

Direct Torque Control (DTC) is one of the most effective and modern methods for speed control of three phase induction motors, but it suffers from some drawbacks, that it needs an estimator for the electromagnetic torque and stator flux, and the existence of inherent ripples in the output torque. So it needs to an improvement.

            In this work a proposed DTC system supported by Radial Basis Function Neural Network (RBFNN) and a Fuzzy Controller (FC) are constructed to avoid the above drawbacks. The (RBFNN) is used as a rapid estimator for the electromagnetic torque and stator flux and the Fuzzy Controller is used instead of the hysteresis comparator for torque and flux errors in order to organize the switching state selector in a more accurate manner. After studying the (RBFNN) it is concluded that it will be more accurate to use it during training and simulation in the independent outputs mode for the torque, stator flux and the sector. Also, accurate results can be achieved from this network for the torque, stator flux by using different values for the spread spectrum in order to let the switching state selector acts regularly. The simulation of the proposed DTC system is done by using a (Matlab/Simulink) program. The proposed DTC system shows a considerable reduction in torque ripples, and best starting performance. This improvement leads to an ability to increase the sampling period four times the conventional one.