Improved Estimation of Non-Sinusoidal Back-EMF with Kalman Filter Applied to Surface-Mounted Permanent Magnet Synchronous Machines

The back-electromotive force (back-EMF) estimation with the Kalman filter can be applied to the sensorless control of surface-mounted permanent magnet synchronous machines. The adoption of different back-EMF waveform dynamics modeling in the system model affects the estimation performance. This work proposes an improved estimation of non-sinusoidal back-EMF using the Kalman filter algorithm, which considers the actual back-EMF dynamics on the system model; hence, this choice relies on the previous knowledge of the back-EMF waveform. The proposed approach is compared to two other methods present in the literature. Simulation results show a more than 90% reduction in the estimation integral absolute error (IAE) compared to the case when the back-EMF dynamics are neglected and up to 80% when the back-EMF is considered sinusoidal. Accordingly, an improvement in the orientation angle estimation is also noticeable, especially for the lower evaluated speed operating point whose IAE reduction was higher than 60%.