Reduced-order square-root EKF for sensorless control of PMSM drives

Performance of square-root extended Kalman filter (EKF) based on reduced order models for sensorless control of permanent magnet synchronous motor (PMSM) drives is studied. The reduced order model of PMSM has two-dimensional state vector comprising of: (i) electrical rotor speed, and (ii) electrical rotor position. These state quantities are estimated by the EKF without either speed or position sensor on the motor shaft. The reduction of the model order results in dramatic speed-up of calculation of the estimator which takes only a few tens of microseconds on a conventional fixed-point digital signal processor. Accuracy of the estimator is improved using square-root representation of the covariance matrix. Due to its low computational requirements, the proposed square-root EKF estimator is eligible for sophisticated diagnostics as well for sensorless control of PMSM drive in a wide range of industrial applications. Presented theoretical conclusions are verified by both simulations and experiments carried out on developed PMSM drive prototype of rated power of 10.7kW.