An adaptive observer with signal injection for interior permanent magnet synchronous motors

Abstract

This thesis deals with methods for the sensorless control of permanent magnet synchronous motors. The methods are investigated by means of theoretical analysis, simulations, and laboratory experiments with a 2.2-kW interior-magnet motor. To expand the operating range and to improve the estimation dynamics, two hybrid speed and position estimation methods are proposed. In the methods, high-frequency signal injection is used for stabilizing a modified voltage model and an adaptive observer at low speeds. The hybrid methods allow sustained operation at low speeds, yet preserve good dynamic properties. For the adaptive observer, an observer gain is selected such that the estimation dynamics are improved. To reduce the sensitivity of the adaptive observer to the motor parameter errors, a method is developed where two electrical parameters are estimated on-line. A model for the spatial harmonics in the permanent magnet motors is included in the adaptive observer for the reduction of estimation errors, and the signal injection method is modified to improve position estimation accuracy when inductance harmonics are present. In addition, a method is proposed for suppressing the electromagnetic torque ripple caused by the spatial harmonics. A control system and a full-order observer are also developed for drives equipped with an LC filter at the inverter output. It is also shown that signal injection can be used with the filter if the injection frequency is suitably chosen. For drives equipped with DC-link current measurement, a method is proposed that also uses high-frequency signal injection at low speeds.