Comparison of shaft position estimation and correction techniques for sensorless control of surface mounted PM synchrononous motors

This thesis is a detailed study of how two error correction schemes affect the precision of shaft position estimation in state-observer techniques for sensorless control surface-mounted Permanent Magnet Synchronous Motors (PMSM), variance correction and variable PI regulation. A novel sensorless estimation technique based on Linear Kalman Filter (LKF) through constant variance correction is proposed and compared with the conventional Flux Linkage Observer (FLO) method and other state-estimation sensorless control techniques namely, Extended Kalman Filter (EKF), variable variance correction, Single Dimension Luenberger (SDL) observer and Full-Order Luenberger (FOLU) observer both through variable PI regulation. These five sensorless control techniques for PMSM are successfully implemented in the same lab-based hardware platform, i.e. full digital float-point-type DSP control inverter-fed PMSM system. Experiments are reported on each sensorless method covering position estimation, speed response, self-startup and load behaviour. Intensive analysis has also been carried out on the impact of error correction of estimated position on the steady/dynamic PMSM characteristics with different sensorless approaches. The experiment demonstrates that the novel Linear Kalman Filter can achieve the minimum average position estimation error throughout the electrical cycle of the five sensorless estimation techniques during no load operation at rated speed and also makes PMSM capable of self-startup for any initial rotor position except the dead area. A speed response experiment for LKF shows that individual speed estimation can be extracted directly from LKF state estimation for sensorless control PMSM. Experiments on the five sensorless methods proves that position error correction scheme is the dominating factor for state estimation sensorless control PMSM and better dynamic/steady control performance can be achieved using a variance correction scheme applied in EKF/LKF than with variable PI regulation applied in SDL/FOLU. The thesis also concludes that the novel Linear Kalman Filter is an optimised cost-effective sensorless estimation method for the PMSM drive industry compared with classic and Flux Linkage observers/Extended Kalman Filters