Research On Key Techniques Of Predictive Control For Permanent Magnet Synchronous Machine Drives

Permanent magnet synchronous machine(PMSM)has been widely used in many industrial areas due to its advantages of high efficiency and stability,etc.Model predictive control(MPC)has been widely used in PMSM drive systems in recent years because of its simple structure and strong flexibility.However,the traditional MPC has some problems that need to be improved.First of all,since a traditional MPC imposes a voltage vector within a fixed control period,it produces a high steady-state error,and the steady-state control performance is not satisfactory.Secondly,as a finite set MPC needs to predict the controlled variables for all voltage vectors,the computational is heavy and the calculation efficiency is low.In addition,a traditional MPC relies on the discrete mathematical model of the system and thus has poor parameter robustness.In this thesis,the effect of two-vector MPC on improving the steady-state control performance is studied,and a model predictive torque control method based on torque boundary is proposed.This method uses preset torque boundaries to calculate the vector switching instant of the candidate vector combinations,and then the torque ripple is controlled within a permissible range.By defining valid candidate solutions and monitoring their number,the torque boundaries can be adjusted online so as to minimize the torque ripple and avoid manual setting of the torque boundaries.According to the torque v ariation trend in each period and the torque slope of each voltage vector,the candidate vector combinations can be pre-selected so as to effectively reduce the calculation amount.In addition,due to the introduction of the torque boundaries,this strategy can eliminate the weight coefficient in the objective function and thus avoid the adjustment work.The effectiveness of this strategy is verified by experiments and compared with traditional MPC strategies.In this thesis,reference voltage vector based MPC is studied.Such strategy can reduce the computational load of the algorithm by means of equivalent transformation.However,since this strategy only considers a fixed reference point,the number of corresponding candidate solutions is quite limited,which makes it difficult to include additional system constraints.Therefore,a reference-variant-based MPC strategy is proposed in this thesis.This strategy builds a reference-variant region based on an original reference point,and samples a finite number of points to introduce more optimal candidate solutions.Based on the well-constructed candidate solutions,additional system constraints can be directly included within the objective function to further improve the control performance.The effectiveness of this strategy is verified by simulations and experiments and compared with the traditional reference voltage vector based MPC strategy.In this thesis,to solve the parameter dependence problem of traditional MPC,a modelfree predictive current control(MFPCC)is proposed and applied to PMSM drive system.By sampling the current variation information of two successive control periods,the current variants related to all voltage vectors can be calculated,and the current predictions can be then realized based on these information.This strategy does not rely on the motor model,but is implemented based on online sampling data,thus completely avoiding the influence of the motor parameter uncertainties and system nonlinearities.In addition,compared with the traditional MFPCC.this strategy can solve the problem of update stagnation of current variation information and gurantee the reliability of current prediction.The effectiveness of the proposed strategy is verified by simulations and experiments,and compared with the traditional model predictive current control and the traditional MFPCC.