Application of neural networks for direct torque control

This paper presents the use of a multi-layer neural network to emulate the traditional switching look up table method for induction motor Direct Torque Control (DTC) obtaining optimal switching patterns. The complete system simulation of the DTC including induction motor, inverter switch, firing circuit and control unit is done using the MATLAB/SIMULINK program. After choosing the best type of neural network, which represents the switching lookup table, a suitable neural network configuration is deduced and then tested. Then the neural network is tested as a part of the DTC in order to know its stability and its reliability. Finally, the simulation results of the DTC using the switching lookup table and the DTC using the neural network are compared.