Reduced order model based neural network control of a squirrel cage induction motor drive

In recent years, much attention has been focused upon neural networks which are generally used to solve highly nonlinear control problems. The implementation of such a control strategy on machine drives has greatly improved their performances. The paper deals with the neural network control of a squirrel cage induction motor drive where the training data base has been obtained using a reduced order model of the controlled system. As a result, the learning rules are found to be easier yielding a reduced structure of the neural net compared to those given by the complete model. Furthermore, a new torque feedback control loop has been introduced in an attempt to improve the dynamic response of the drive. Considering the reduced order model based neural network control and the complete model based neural network control, simulation results show that the training data base given by the reduced order model is sufficient to reach high dynamic responses which are better than those yielded by the complete model training data base. Moreover, it has been found that the robustness of the implemented control system is not affected by measurement perturbations.