Direct field oriented neural control of a three phase induction motor

Abstract

The paper proposed a complete neural solution to the direct vector control of three phase induction motor including real-time trained neural controllers for velocity, flux and torque, which permitted the speed up reaction to the variable load. The basic equations and elements of the direct field oriented control scheme are given. The control scheme is realized by nine feedforward and recurrent neural networks learned by Levenberg-Marquardt or real-time BP algorithms with data taken by PI-control simulations. The graphical results of modelling shows a better performance of the NN control system with respect to the PI controlled system realizing the same general control scheme. © 2010 Springer-Verlag.