Sensorless PMSM control: Hybrid rotor position estimator using maximum likelihood model selection

This paper presents sensorless PMSM control employing a combination of the extended Kalman filter (EKF) and the high-frequency (hf) injection algorithm for estimation of rotor position and speed. Both algorithms have limited usage. The hf injection algorithm can be used in standstill and in very low speeds whilst EKF can be employed in higher speeds. The problem of the introduced “hybrid” estimator is a proper transition between the EKF and the hf injection strategy. The transition is based on calculating of probabilities for each particular estimation algorithm. We proposed two strategies for proper algorithm selection. The first one is with constant prior probability of particular algorithms and the second one has variable prior probabilities. For their changes is used transition matrix that is known from Markov system. Behavior of the proposed sensorless controls is analyzed and verified by experiments made on a developed laboratory prototype of PMSM drive of rated power 10.7kW.