On SPMSM rotor position estimation from standstill and for low speeds - Evaluation of techniques for estimating position and speed of an SPMSM

Abstract

Knowing the rotor position is essential for performing field-oriented control of permanent magnet synchronous machines. Sometimes, sensorless strategies are employed instead of using sensors in order to estimate the rotor position. This may become troublesome for surface-mounted permanent magnet synchronous machines (SPMSM) since the usual methods rely on saliency properties. In this project, two different methods for estimating the rotor position from standstill through low speeds have been evaluated for an SPMSM for a drill application on behalf of Husqvarna AB. An extended machine model taking saturation and cross-coupling effects into account was developed and several simulations of the discrete-time control system using the position estimation schemes were conducted. A signal injection technique was compared with the statically compensated voltage model (SCVM) due to difficulties in incorporating more suitable choices and also due to time constraints. A simple magnetic pole identification method was implemented and tested in simulations since the estimated angle might have an error of 180 degrees. The results indicate that the rotor position may be estimated from standstill using the signal injection method and that saturation-induced current differences may be used to find the alignment of the magnetic field, thus eliminating the 180 degrees error offset. Further, the results also indicate that the machine can become unstable for sudden load changes but may be stable for slower load changes with the current parameters when using the signal injection technique. The high frequency signal injection scheme was able to estimate the rotor position with a higher steady-state accuracy during high starting loads than the SCVM which could be useful for drill applications. This since it may develop higher starting torque than the SCVM. Usually, SPMSMs exhibit low saliency which make the usual signal injection techniques difficult to use. Therefore, one suggestion is to try to implement and test a low frequency signal injection scheme instead. Therefore, more measurements and simulations should be made before trying to implement these strategies in reality.