Speed-sensorless direct torque control system of induction motor

In the conventional direct torque control system, the stator flux can´t be identified accurately because of the influence of stator resistance and the calculation of pure integral. In addition, the speed sensor has its deficiency. This paper deals with these problems and presents a new method. The accurate stator and rotor flux of full speed range is obtained by adopting new flux observer, and based on it, the motor speed is identified by adopting neural network. By using of this method, the accurate flux target value is acquired as well as the NN becomes simpler and needs no off-line training. After the motor is operated, by means of simple and fast calculation, the NN can learn online and obtain correct identification results. In the end, the experimental results are presented.