A new method of PWM pulses generation in induction motor drive using artificial neural networks

Due to simplicity and low cost, induction motors are more useful than direct current motors. Hence the control of these motors is important. The pervious methods are fitted normally for a limited speed range and could not be used for both low speed and high speed. The voltage model is suitable for high speed because the voltage drop of stator resistance is not small in low speed voltage. The current model is suitable for low speed because of the problems of current saturation at high speed. This research present a new method of PWM pulse generating in induction motors based on artificial neural networks that in which, the switching pulses are generated by a multilayer feed-forward neural network that trained with the voltage and current references. Also, for estimation of required torque and flux information a multilayer perceptron is used. By application of this new method, there is no problem of stability at low and high speeds. The simulation results by matlab-simulink verify the proposed method in transient and steady-state operating modes.