Alternative Implementations of Square-root Cubature Kalman Filter for Sensorless PMSM Drives with Improved Dynamic Performance

Abstract

Kalman filters have been widely applied for speed and position sensorless control of Electric drives. A new range of nonlinear Kalman filters have been successfully implemented for sensorless Permanent Magnet Synchronous Motors (PMSM) in recent times. This paper addresses the alternative implementations of Square-root Cubature Kalman filter (SCKF) for speed and position estimation of a PMSM drive. SCKF offers improved numerical stability as compared with the Cubature Kalman filter (CKF). SCKF is observed to offer the possibility of incorporating combined load torque estimation, which is shown to increase the bandwidth of the speed control loop. A simple method to infer the bandwidth of the SCKF is suggested in this work, which is based on the interaction between the bandwidths of the speed controller and the observer. In the perspective of improving the transient response of the speed control loop, a modification to the PI controller is proposed to reduce the overshoot in speed without affecting the disturbance rejection property. The proposed techniques are validated through simulation and experiment on a VSI fed 1.5 kW sensorless PMSM drive system.