Study On Speed Sensorless Direct Torque Control Strategies

Replacing the DC speed adjustment, the AC speed adjustment hasbecome the main method in the fields of electrical drives. The speedsensorless direct torque control (DTC) technique has a significantdevelopment in its ability to improve the availability, cost, and reliabilityof the AC drive system. Using advanced control theory, the strategies ofthe speed sensorless DTC are studied in depth in the dissertation. Themain research works are as follows:An AC motor model in the 2-phase staticαβreference frame isestablished, taking stator current and rotor flux as the state variables.Based on the motor model, a simulation model of DTC system is builtconsequently. The performance of DTC at low speed, the influence on thesystem performance due to the variation of stator resistance, and theinitial value and DC drift existing in flux estimation are analyzed indetails.The accurate estimation of the stator flux is important to the choiceof voltage space vector and the torque estimation. It is the key part for thehigh-performance DTC. The open loop stator flux estimation method forclassical DTC, modified U-I integrator method, and stator flux observermethod are analyzed in depth. Because of the limitations of thosemethods, a stator flux observer based on the modified integrator withvariable cutoff frequency is proposed in this dissertation. Utilizing afuzzy adaptive filter, the cutoff frequency of low-pass filter can beadjusted in real time according to the online operating states of inductionmotor. Consequently, it can eliminate the errors caused by the DC driftand initial value and improve the precision of stator flux estimationefficiently.The adaptive speed observer is the most commonly used method inthe speed sensorless technology. The performance of speed observer isdecided by its gain matrix G and the choice of the speed adaptive law.Using the pole-placement technique to obtain the gain matrix G appearsin almost all the current speed adaptive observer schemes. After studyingin depth the existence of unstable regions in the pole-placement technique for the design of adaptive observer, it is known that the resulting gainmatrix G may lead to system instability, since the induction motor is ahigh-order nonlinear strong-coupling system of multi-variables. On theother hand, the rotor flux error is ignored in the current schemes whenchoosing the speed adaptive law, because it is difficult to get the accuraterotor flux value. In the research, it is shown that the rotor flux error islarger than the current error when the motor operates in the low-speedre-generating mode, so the rotor flux error item should not be ignored.Otherwise, it will degrade the celerity and stability of observer.The adaptive speed observer that can remain stable over its fullspeed region is proposed in the dissertation. On one hand, the gain matrixG of the observer is obtained according to the Lyapunov stable theory.When defining the Lyapunov inequality, an undetermined matrix variableP is introduced to transform the Lyapunov inequality into two bilinearmatrix inequalities of G and P. The gain G and P can be obtained byusing the LMI Toolbox of Matlab. On the other hand, both stator currenterror and rotor flux error terms are included in the speed adaptive law.The accurate rotor flux value can be calculated indirectly by utilizing thestator flux obtained from the stator flux observer, which is based on themodified integrator with variable cutoff frequency. With the appropriatechoice of gain matrix G and speed adaptive law, the Lyapunov matrixinequality of the speed adaptive observer can hold over its full speedregion consequently. It overcomes the disadvantage of the existence ofthe unstable regions and the inaccurate speed identification in theclassical observer.As the stator resistance changes with temperature in motor operation,if the adoptive value of stator resistance for DTC is different from itsactual value, it will cause a big flux and speed estimation error and evenlead to system instability. Thus, it is necessary to estimate the statorresistance online. Based on the analysis of the limitation of DTC statorresistance estimation with the pole-placement technique, the statorresistance identification law on condition that the Lyapunov inequalitystrictly holds is proposed. Then, the online stator resistance estimationcan be achieved by adding the law to the adaptive speed observer based on LMI. The system performance and the estimation precision of flux andspeed are improved by feeding back the estimation value of statorresistance to DTC.The speed sensorless induction DTC simulation platform is built inthe Matlab6.5/Simulink environment, which introduces the modulatespace vector modulate (SVM) technique to replace the traditional switchscheme of DTC for improving the low-speed performance and the fuzzyPI speed controller to replace the traditional PI controller for improvingthe celerity and robustness of system. Simulations and comparisons aremade for the strategies and schemes proposed in this dissertation.Simulation results show that the speed sensorless DTC strategiesproposed in this dissertation can ensure the speed adjustment systemstable over its full speed region.