A PROPOSED ESTIMATOR FOR A DIRECT TORQUE CONTROL OF INDUCTION MOTOR BY USING RADIAL BASIS FUNCTION NEURAL NETWORK TECHNIQUE

Abstract

The purpose of this paper is to use Radial Basis Function Neural Network (RBFNN) as an estimator for stator flux and electromagnetic torque in Direct Torque Control (DTC) systems used as a driver of a 3-phase induction motor, in order to reduce the ripples in the output torque. This paper includes design, construction and training for three different modes of operation of RBFNN, in which the spread constant has a different value for each estimated parameter during the network training. Then, the network, which has independent outputs, gives the best results choused as an estimator in the proposed DTC system. Matlab/neural network toolbox used for training the proposed estimator at different load torques.

The Simulation results are obtained using program of Matlab/Simulink. The coincidence of the values of the output data obtained from the proposed estimator and that from the conventional one proves the proposed system accuracy.