An Indirect Adaptive Vector Control of the Induction Motor Velocity Using Neural Networks

The paper proposes a neural network solution to the indirect vector control of three phase induction motor including a real-time trained neural controller for the IM angular velocity, which permitted the speed up reaction to the variable load. The basic equations and elements of the indirect field oriented control scheme are given. The control scheme is realized by one recurrent and two feedforward neural networks. The first one is learned in real-time by the dynamic BP method and the two FFNNs are learned off-line by the Levenberg-Marquardt algorithm with data taken by PIcontrol simulations. The final set up MSE of the LM algorithm is of 10-10. The graphical results of modeling show a better performance of the adaptive NN control system with respect to the PI controlled system realizing the same computational control scheme with variable load.