Impact of Coordinate System Selection and Model Observability on Position Sensorless State Estimation of Nonsalient-Pole PMSM

Model observability is essential for the design of state estimators and observers, since the state vector can only be fully reconstructed from the measurements, if the model is observable. However, some research articles do not perform observability analysis and use unobservable models, while the presented results show adequate estimation performance. This paper deals with the impact of observability on the position sensorless state estimation of nonsalient-pole permanent magnet synchronous machine (PMSM). To show how the observability depends on the chosen coordinate system, observability studies are presented for two widely used nonlinear state-space models given in either the stator or the rotor reference frame. For the stationary reference frame model, a necessary and sufficient observability condition is defined, and it is shown that the model using rotor-oriented variables is unobservable. To point out the limitations of using the unobservable model, comparative simulations are carried out between extended Kalman filters using the models presented.