Research On Model Predictive Control Strategy For Permanent-Magnet Synchronous Machine

The model predictive control(MPC)can solve the optimization of multiple objectives with constraints and meanwhile obtains fast dynamic responses and perfect steady-state performance,thus it is considered one of the most talented control schemes applied for the control of high-performance permanent-magnet synchronous machine(PMSM)drive system.However,conventional MPC is parameter sensitive,and its performance of prediction and optimization will be deteriorated when disturbances exist,such as parameter mismatches,current distortion,and dead-time effects,which affect the dynamical and steady-state performance and probably lead to power fault.This dissertation concentrates on these issues of the MPC-based PMSM drive system and researches MPC strategies to improve the control accuracy in steady-state,disturbances rejection ability,and parameter robustness of the system.The modeling and solving methods for MPC-based PMSM systems are firstly researched.Moreover,a two-step prediction scheme is applied to overcome the control delay in MPC.The conventional single-vector MPC and existing duty-ratio MPC strategies lead to poor steady-state performance,to overcome this issue,an optimal two-vector combination-based MPC method is proposed.In this method,all available combinations of two vectors and corresponding dwell time are considered.A candidate collection of two-vector combinations is constructed for global optimization and the optimal combination is selected to minimize the approximating error between the output vector and the reference vector.To simplify the optimization,an equivalent formulation and sector transform are proposed to convert the optimizing target into a voltage vector within the specified sector in stationary coordinate,thus,the candidate collections are simultaneously transformed to five fixed straight lines.Moreover,an off-line algorithm is proposed to conduct parts of the complex calculations,which efficiently reduces the real-time computations.Current model-based stator flux estimation algorithm significantly relies on the PMSM parameters.To promote the parameter robustness for prediction and optimization performance of MPC,a voltage model-based vector transform and signal filtering flux estimation method is proposed.Principle of this flux method is presented,where a low-pass signal filter and a band-pass filter-based vector transforming method are separately designed.An optimal vector transforming flux estimator is proposed by combining the two filters and vector transforming methods with specified optimization function.Moreover,the structure of proposed optimal flux estimator is simplified to reduce the system complexity and computation burdens.The effects on control performance of cut-off frequency are investigated through Fourier Transform Theory,and the preferred cut-off frequencies are presented.The performance of MPC-based PMSM drive system will be deteriorated by external disturbances.The effects caused by certain kinds of disturbances such as parameter mismatches,current distortion,and dead-time effects are theoretically analyzed.The form of disturbances is presented as harmonics associated with the fundamental frequency.To suppress these harmonics,the conventional repetitive control method(RC)is introduced,but its strength in tracking harmonics will be significantly deteriorated when DC drifts exist in the disturbances,thus,a simplified repetitive control method(SRC)is proposed.The proposed SRC produces simple structures and avoids the DC drifts issues,moreover,it responds faster than conventional RC.A second-order SRC controller(SRC-2)with the first and second resonant unit is designed,and the effects of frequency quantizing error on control performance are investigated.An output filter is designed to ensure system stability and to correct the quantizing error.Finally,the promoted dynamical performance and harmonics rejection ability of the proposed SRC method are validated with injected current distortion and parameter mismatches,respectively.The system performance of cascaded MPC will be deteriorated by the speed control loop with poor-performance PID controller.To overcome this issue and to promote the control performance of the speed and current,a backstepping control method combined with the prediction and optimization of MPC(BSC-MPC)is proposed.The disturbance observer is designed based on the predictive error of MPC,and the backstepping controller is applied for speed regulating.For the proposed BSC-MPC,the system stability is firstly proved,and then,the detailed calculations for the voltage target are illustrated along with the processing flow.This method enlarges the stable area of the original backstepping method,and furthermore,its current control performance is significantly improved when the virtual voltage target is adjusted by optimization of MPC.Simulation researches for the dynamical performance,control accuracy in steady-state,and parameter robustness are respectively conducted among the proposed MPC methods including the optimal two-vector combination MPC,the optimal flux estimation method with MPC,the simplified repetitive control MPC,and the backstepping control MPC.Moreover,experiment platform for the MPC-based PMSM drive system is designed,and experimental researches are performed among all the mentioned MPCs.Detailed experiment results are presented,and the feasibility and effectiveness of all MPCs are validated through experiments.