Low speed control for PMSM based on reduced-order adaptive Kalman filter

This paper proposes a dual reduced-order adaptive Kalman observer to estimate the rotor speed and load torque of PMSM under low speed condition. The observed speed is more precise than encoder differential signals which has more noise, and the estimated torque can be used as feedforward compensation to reduce speed ripple when load changes. Because the calculation amount of Kalman filter is proportional to its order´s cube, a dual reduced-order observer is proposed. To reduce the influence of modeling errors, adaptive algorithm with forgetting factor is proposed to calculate the noise covariance. The simulation and experiment results show the good performance of the proposed strategy.