Prediction Error Analysis And Suppression Of Model Predictive Current Control For PMSM Drives

Permanent magnet synchronous motor(PMSM)has been widely investigated and applied in many industry applications due to its advantages,such as high efficiency,high power indensity,high torque-current ratio and wide speed range.At the same time,with the improvement of microprocessors,model predictive control(MPC)has been developed as a promosing high-performance control strategy of PMSM.It owns some obvious advantages,especially simple structure,fast dynamic response,easy achievement of multi-object control,and easy inclusion of nonlinearities and constraints.However,MPC is based on the prediction model of PMSM.There exist some influencing factors resulting in model mismatch,therefore prediction error is generated.It can affect the prediction accuracy and final control performance,which also limits the achievement of more high-performance control objects.Therefore,this dissertation conducts an in-depth research about prediction error of model predictive current control(MPCC)and its suppression for PMSM drives.The main contributions of this dissertation are summarized as below:(1)Due to the fact that the model mismatch can lead to prediction error of MPCC,three main influencing factors related to model mismatch,i.e.discretization method,prediction stepsize and parameter mismatch,are selected to conduct a research.Formulas of prediction error are derived considering these three factors,respectively,and then theoretical analysis is provided to reveal the variation trends of prediction error under three factors.Experimental results and analysis verify the theoretical conclusions and further quantitatively show the influence degrees of these facors to prediction error.The above research lays the foundation for the following proposed MPCC strategies which are aimed at prediction error suppression.(2)According to the different influencing factors for prediction error,two improved MPCC strategies are proposed based on the direct compensation of prediction error.The first one is aimed at the suppression of prediction error resulting from inductance mismatch.The prediction errors of previous two control periods are utilized to obtain the prediction error corresponding to every available voltage vector in present control period which is then compensated in the prediction process.The second one comprehensively includes the suppression of prediction error brought by internal and external influencing factors.Based on the previous prediction error,an extended state observer(ESO)is designed to obtain the predictive dq currents and lumped prediction error.Different from the conventional disturbance compensation method of ESO,all outputs of ESO in the proposed MPCC strategy are fully utilized to reconstruct the prediction models considering control delay compensation,which achieves the both prediction error compensation and computation burden reduction with respect to prediction process.(3)Considering the dependence of motor parameters for both prediction model and some high-performance control objects,an online inductance correction algorithm is proposed based on previous prediction errors.Then it is further combined with maximum torque per ampere(MTPA)control goal.The inductance values in prediction model and MTPA algorithm are corrected at the same time so that both prediction error suppression and high efficiency control of PMSM are achieved.Compared with the conventional MTPA strategies,two cascaded cost functions are utilized to achieve the MTPA control goal.It shows the advantage of multi-object control of MPCC and reduces the number of weighting factors as well as computation burden of evaluation process.All above proposed MPCC strategies are fully based on the prediction error,which is an inherent characteristic of MPCC.The new algorithms including prediction error compensation and online parameter correction are well combined with the conventional predition-evalution process and delay compensation structure.On the one hand,the prediction error is obviously reduced and the prediction accuracy is impoved.On the other hand,they do not bring much more computation burden compared with traditional MPCC strategy.Besides theoretical analysis and design,all proposed methods are discussed and compared with traditional MPCC strategy and typical strategy in literature by simulations and experiments.The results verify the effectiveness of the proposed methods.