Sensorless Control And Paramster Identification Of PM Motors

The major contents of this thesis are on the position sensorless algorithms of permanent magnet(PM)motors and the simultaneous parameter identifications during sensorless operations.The sensorless algorithms depend on the model of the motors.Therefore,the position estimations of the sensorless algorithms vary with the parameters.Because of the limitations of the observability of the PM motor model,only one parameter can be identified during the sensorless operation.However,under many circumstances,the sensorless algorithm is sensitive to more than one parameters.To solve this problem,this thesis has contributed the following works:Firstly,a high-gain sliding mode electromotive force(EMF)observer with a frequency adap-tive band-pass filter is proposed to serve as a basic sensorless algorithm.Existing methods to sup-press the chattering in the sliding mode observers(low-pass filters or switch the high-gain signum function to low-gain functions)will introduce extra phase displacements or make the currents es-timations unable to converge.The proposed method preserves the signum function of traditional sliding mode observers(high-gain),and adopts a frequency adaptive band-pass filter to suppress the chattering problem.The center frequency of the adaptive filter locates at the fundamental fre-quency of the EMF vector,and will not introduce extra phase displacements.Because the input of the adaptive filter is heavily distorted,the influence of the heavily distorted input on the frequenc adaptive scheme is analyzed.Based on the analysis,a method to suppress the speed estimation error by band-width designing is proposed.Secondly,a harmonic model based on the harmonic EMFs of the PM motors is proposed to enhance the degree of freedom(DOF)for parameter identifications.Existing methods require signal injections or transient operations to enhance the DOF for parameter identifications due to limitations of the fundamental model.To solve this problem,the harmonic characteristics of the PM motor are adopted.Since the harmonic model are inherent properties of the motor,and independant of the fundamental model,it is able to provide extra information for parameter identification under steady state without signal injections.A series of transformations are proposed to convert the 5th and 7th order harmonic EMFs to a virtual fundametal EMF vector.From this virtual fundamental vector the rotor position can be extracted.During the extraction of the 5th and 7th order harmonic EMFs,the corresponding harmonic currents are suppressed by harmonic current controllers.Therefore,the rotor position calculated from the 5th and 7th order harmonic EMFs are robust to motor parameters.Thirdly,a stator resistance and rotor fluxlinkage adaptive sensorless control algorithm is pro-posed;a novel inverter nonlinearity compensation method based on sigmoid curve fitting is also proposed to better compensate the harmonic voltage components caused by the inverter nonlin-earities.Through parameter sensitivity analysis of the sensorless control algorithm,it is pointed out that the position estimation of the sensorless algorithm is sensitive to both stator resistance and rotor fluxlinkage in a large operation area.Hence,it is necessary to identify both stator resis-tance and rotor fluxlinkage simultaneously during sensorless operation.In the proposed method,the fundamental model is the adaptive model,and the harmonic model is the reference model,the position estimation error between the two models are used to regulate the stator resistance through an adaptive scheme.Because the proposed method adopted the harmonic model,a novel inverter nonlinearity compensation method based on sigmoid curve fitting is proposed because the sigmoid curve and the actual voltage error curve are more alike.