Comparison of Off Line Neural Network Training Methods for Sensorless Induction Motor Drive

Author

Naghdinezhad, A. ; Mohamadian, M. ; Dastfan, A.

Author_Institution

Sharood Univ.

Volume

2

fYear

2006

fDate

6-8 Sept. 2006

Firstpage

709

Lastpage

713

Abstract

Four back-propagation training method for an offline neural network is compared in this paper. The neural network is designed to estimate the speed of an induction motor in a sensorless vector control system. In first step, four different training algorithms are used to train the neural network. Each training algorithm is tested with up to 35 neurons with fixed number of training iterations. The number of neurons that results in least training error is specified. Next, the number of training iterations is increased and the training algorithm with least amount of error is specified. The training algorithms are also examined with same number of neurons and learning iterations. Finally the training algorithm with least error and best convergence is used to estimate an induction motor vector control rotor speed

Keywords

backpropagation; induction motor drives; machine vector control; neural nets; back-propagation training; convergence; induction motor drive; learning iterations; off line neural network; rotor speed control; sensorless vector control; speed estimation; Artificial neural networks; Biological neural networks; Equations; Feedforward neural networks; Induction motor drives; Induction motors; Machine vector control; Neural networks; Neurons; Rotors;

fLanguage

English

Publisher

ieee

Conference_Titel

Universities Power Engineering Conference, 2006. UPEC '06. Proceedings of the 41st International

Conference_Location

Newcastle-upon-Tyne

Print_ISBN

978-186135-342-9

Type

conf

DOI

10.1109/UPEC.2006.367571

Filename

4218778