Regresar

Vector control of induction machines using neural networks with parametric adaptation in real time

Abstract:

A vector control scheme of induction machines using neural networks in the state estimator and in the control loop has been developed. Using the dynamic model of the induction machine a neural network which reproduces the instantaneous behavior of the non-measurable variables (electric torque and magnetizing flux), has been developed. This network allows the parameters adaptation in real time during the machine operation, without additional training. The estimation errors for the non-measurable variables obtained using this neural network are not significant. The rotor time constant is dynamically adapted with an error lower than 0.3%. The proposed control system reduces the delay due to parameter variations and increases the dynamic characteristics of the vector control.