Research On Direct Torque Control Strategies Of Surface Mounted Permanent Magnet Synchronous Machines

Surface mounted permanent magnet synchronous machines (SM-PMSMs) have been extensively developed due to their high power density and efficiency, good performance, easy manufacture and low cost. On the other hand, direct torque control (DTC) is often applied for the SM-PMSM drives due to its fast response and superior dynamics. However, some problems exist in the DTC of SM-PMSM drives, which will be studied in the thesis, particularly on the aspects of torque ripple, low-speed performance, initial rotor position estimation and special control method of motors with non-sinusoidal emf waveform.The basic DTC of SM-PMSM is firstly investigated with both simulations and experiments, showing that the basic DTC exhibits good dynamics but significant torque ripple. The torque ripple is mainly caused by the low wounding inductance and the small count of voltage vectors. In order to increase the vector count, a space vector modulation with novel flux error vector estimation algorithm (FEVE-SVM) is introduced into the DTC, which is proved to be able to reduce the torque ripple obviously.The voltage-drops on the stator resistance and on the power switches are not negligible when the machine runs at low speed. Algorithms of power switch voltage-drop compensation and stator resistance estimation are presented to improve the stator flux accuracy. The performances of counting the compensation in the flux estimation and directly in voltage vector reference are compared, showing that the direct compensation is easier and more effective. Further more a PI regular with the torque and stator currents as inputs is used to estimate the stator resistance on-line. To improve the low-speed performance, a novel stator flux estimation algorithm based on the stator currents and rotor position is introduced, with which a wider speed region is achieved.The rotor initial position information is essential to start a PMSM from standstill. There are two main methods to detect the rotor's initial position for a PMSM, viz. high frequency signal injection and the stator saturation effect detection. A method based on the stator saturation effect and artificial neural networks (ANNs) is presented to estimate the rotor initial position with the maximum estimation error of 20 electrical degrees. Experimental results show that the PMSM can start from standstill with such estimation accuracy.Finally, the DTC strategy is improved such that it can be used for a NS-PMSM with non-sinusoidal emf. The strategy is verified with both simulations and experiments.