Rotor Resistance Estimator Using Support Vector Machines and Model Reference Adaptive System

Abstract

The main drawback of the indirect vector control of the squirrel cage induction motor (SCIM) is the impairing of the drive performance because of the lost of the field orientation caused by the difference between rotor resistance of the machine, varying with temperature, saturation, skin effect, and its corresponding in the model of the controller (fixed). Rotor resistance estimator of the SCIM was designed using support vector machines (SVM) together with model reference adaptive system. The drive with variable rotor resistance was simulated, and flux error records were obtained from current and voltage models. SVM was trained with those flux error records and variable rotor resistance; the great generalization potential of the SVM was confirmed during the training stage, since with a small number of support vectors, the resistance value was predicted for a great quantity of samples that were not considered in the training stage. The drive with the rotor resistance estimator in closed loop showed a well performance when was undergone to other operation conditions different to that from the training signals were obtained and nevertheless to achieve an excellent estimation of the rotor resistance