Research On Control Strategies Of Permanent Magnet Synchronous Motor Based Servo Drive System With Model Prediction

This thesis is presented for the application for a position servo drive system,which constructed by a two-level inverter fed three-phase permanent magnet synchronous motor(PMSM).The model prediction based control schemes are researched and demonstrated for the servo drive system in detail.The main work includes the following parts:According to the mathematical model of PMSM in synchronous coordinate frame and the deadbeat control principle,the current loop and the speed loop in the servo three-loop control system are designed to avoid the use of cascading PI controller.When designing the current loop,the one-step delay of the digital controller in the actual system is compensated by substituting the predictive current into the voltage equation,in such a way that more precise reference voltages are obtained.The reference currents are obtained by the maximum torque per ampere(MTPA)control method and the mechanical equation when designing the speed control loop.A proportional controller is used to design the position loop.It provides the basic control framework for predictive control of servo drive system.A hybrid model predictive direct speed control(MPDSC)strategy is proposed for the surface-mount PMSM(SPMSM)position servo system,which combines the finite control set MPDSC(FCS-MPDSC)strategy and the double vectors MPDSC(DV-MPDSC)strategy.The FCS-MPDSC strategy is adopted in the dynamic system while the DV-MPDSC strategy is adopted in the steady state.The current ripple in the steady state is optimized while the fast dynamic response of the system is ensured.In order to improve the robustness of the system,a parameter disturbance and load observer is designed based on sliding mode control theory.The parameter mismatch and load of the system are observed and compensated as a total set of disturbance.The robust hybrid MPDSC strategy combined with the sliding mode observer(SMO)can achieve relatively good current control performance and accurate positioning under inaccurate motor parameters and load disturbance conditions.In order to avoid the use of PI controller in traditional deadbeat predictive current control(DPCC),a deadbeat predictive speed control(DPSC)strategy is designed for the SPMSM position servo system in this thesis.According to the discrete-state model of the motor,the reference voltages are predicted and then converted into the switching signal of inverter by space vector pulse-width-modulation(SVPWM).In this thesis,the incremental model is adopted to predict the current without the specific value of permanent-magnet flux,and an extended state observer(ESO)is designed with the incremental model to estimate and compensate the prediction errors caused by inaccurate parameter and load disturbance.The robust DPSC strategy combined with ESO can achieve accurate tracking of the desired position in the case of inaccurate parameter and external disturbance in load.In order to reduce the oscillation caused by the addition of the observer when the system approaches the given position reference,a prediction error compensation method is proposed,which adopts different compensation quantities to compensate the system according to the different working states of the system.The proposed compensation method can improve the robustness of the system and maintain the good dynamic performance of the DPSC strategy.Thus,the system can achieve a better position tracking performance.Finally,the control performance of the SPMSM position servo system based on different control strategies are analyzed and compared,including the comparison between hybrid MPDSC strategy and DPSC strategy and the comparison among DPSC strategy,DPCC strategy and PI control.Both the simulation and experimental results show that the hybrid MPDSC strategy and the DPSC strategy can achieve satisfied position tracking performance.The robustness of the system against motor parameter variations and the anti-disturbance performance against load disturbances can be improved by using disturbance observer.