Title :
Neural network aided unscented Kalman filter for sensorless control of PMSM
Author :
Talla, Jakub ; Peroutka, Zdenek
Author_Institution :
Regional Innovation Centre for Electr. Eng., Univ. of West Bohemia, Pilsen, Czech Republic
fDate :
Aug. 30 2011-Sept. 1 2011
Abstract :
This paper introduces neural network aided unscented Kalman filter (NNUKF) for sensorless control of ac motor drives. Unscented Kalman filter (UKF) is completed by on-line trained neural network which compensates unmodeled dynamics and uncertainties of a drive model. This technique significantly improves behaviour of estimator in critical operating states, especially in low speeds.
Keywords :
Kalman filters; neurocontrollers; permanent magnet motors; sensorless machine control; synchronous motor drives; ac motor drives; critical operating states; neural network aided unscented Kalman filter; permanent magnet synchronous motor; sensorless control; Adaptation models; Artificial neural networks; Estimation; Kalman filters; Mathematical model; Rotors; Sensorless control; Estimation technique; Neural network; Permanent magnet motor; Sensorless control;
Conference_Titel :
Power Electronics and Applications (EPE 2011), Proceedings of the 2011-14th European Conference on
Conference_Location :
Birmingham
Print_ISBN :
978-1-61284-167-0
Electronic_ISBN :
978-90-75815-15-3
