Application of Unscented Kalman filter to sensorless permanent-magnet synchronous motor drive

This paper studies the application of the Unscented Kalman filter (UKF) for a sensorless permanent-magnet synchronous motor (PMSM) drive. The principle of the UKF is reviewed and its adaptation to the PMSM model in the rotor reference frame is explained. For evaluation of drive performance a comparison with the classical Extended Kalman filter (EKF) based estimation method is made. The simulation results indicate that the UKF in general gives more accurate tracking of the rotor position than the EKF for constant torque operation. Besides, the UKF drive retains control over a longer standstill period. With the UKF-based sensorless PMSM drive, however, the proportional gain of the speed controller needs to be carefully chosen to avoid rotor hunting subsequent to step load changes.