Research On Torque And Magnetization Manipulation Control Of Variable Flux Memory Machines

Permanent magnet synchronous machines(PMSM)have been widely used in traction systems such as electric vehicles.However,the traditional PMSM has to be controlled in flux-weakening mode when DC-link voltage is not sufficient.Continuous flux-weakening current leads to extra copper losses,which reduces efficiency at high speeds.Recently,variable flux memory machines(VFMM)have been proposed as a new type of PMSM whose air gap flux density can be altered by a more flexible approach.VFMM uses low coercive force(LCF)magnets and these magnets can be magnetized or demagnetized by pulse currents,so that back electro-motive force(EMF)of VFMM can be controlled.Because the flux-weakening currents are eliminated,machine losses are significantly reduced.These merits of VFMM show promising applications.However,because this machine has new characteristics and principles,there are still problems to be solved.One of the major concerns is how to control the machine.Based on the state-of-art of VFMM studies,few papers have focused on the VFMM drive.So,this thesis mainly works on the control of VFMM.This thesis begins with the machine parameter calculation and measurement,which are the basis of the machine control.Then,this thesis concerns three major aspects of the VFMM control: how the torque is controlled within a single magnetization state(MS),how the magnetization or demagnetization decisions are determined,and how the current is controlled during the magnetization or demagnetization manipulations.Because the magnets exhibit different properties at different states,analysis of VFMM is difficult.This thesis proposes a magnetization-frozen finite element method(MF-FEM).Based on the combination of the AlNiCo magnet material properties and physics of the MS manipulation process,distinctive BH curves are assigned to the magnets in the finite element analysis to calculate the machine magnetization and demagnetization properties.Frozen permeability method is usually used to calculate on-load inductances;however,the permeability of the AlNiCo is variable at different MSs.Thus,magnetization-frozen frozenpermeability FEM(MFP-FEM)is proposed.The permeability of the machine core is frozen in the method;besides,the MFP-FEM freezes distinctive AlNiCo magnet permeability at different MSs,which increases the calculation accuracy.The MFP-FEM has to be verified by measurement.However,the traditional inductance measurement can cause machine demagnetization,which deteriorates the measurement accuracy.A DC-biased inductance measurement is proposed.The protection of the LCF magnets is achieved by injecting the DC current,such that no negative current appears during the measurement.Additional DC current is injected to evaluate cross-magnetization effects.The DC-biased method can eliminate influences from resistance measurement error.According to the deviation patterns of the flux linkage-current curves in the measurement,a resistance tuning algorithm is proposed to minimize the resistance error influences.Maximum torque per ampere(MTPA)control can be implemented to the VFMM with saliency.The MTPA control fully utilizes reluctance torque to reduce losses and increase torque output.Parameter non-linearity and variable flux properties of VFMM increase the control difficulties.An artificial neural network(ANN)based MTPA control is proposed.Featuring the excellent non-linear fitting capabilities of the ANN,the controller reduces the on-line control complexities.A double-loop gradient descent based MTPA current search algorithm is proposed.The MTPA search algorithm generates the MTPA currents from the torque commands and the MS,and then these MTPA current samples are used in the training process of the ANN.Verified by experiments,the ANN based MTPA control increases machine efficiency and torque.The magnetization or demagnetization manipulations should be properly triggered according to the relation between machine voltage and DC-link voltage.An MS control system is proposed and it generates the manipulation triggers by calculating the DC-link voltage margin ratio.The DC-link voltage margin ratio presents the remaining voltage that can be used for machine operations.By comparing the margin ratio to its criterion,either demagnetization or magnetization are manipulated.As results,either the machine speed range are enlarged or the torque is increased.Because the DC-link voltage is considered in the calculation of the margin ratio,the proposed MS controller can trigger the demagnetization manipulation at lower speeds when the DC-link voltage drops.The magnetization and demagnetization current control accuracy is closely related to the control accuracy of the machine EMF.Increasing manipulation speed can reduce losses and mechanical impact.A feed-forward control based MS manipulation current trajectory control is proposed.The feed-forward controller compensates the induction voltage,the rotating coupling voltage,and the resistive voltage during the magnetization and demagnetization manipulations.Fast current response and high control accuracy are achieved by the proposed controller.The flux linkage and the gradient of the flux linkage are obtained from the MFP-FEM.The flux linkages and their gradients are used to derive the voltages during the manipulations,and then the voltages are implemented for the feedforward control.Moreover,the controller reduces difficulties of tuning the controller gains.The contributions of this thesis will help further study of the VFMM.The methods,theories,and practices shown in this thesis can be extended to other machine with the memory machine principles.