Research On T-type Three-level With Finite Set Model-based Predictive Control

Performance and demand of three-level power converters such as the T-type triplet are increasing,and traditional linear control schemes are unable to meet the non-linear demands of power converters.Benefit from the advancement of digital technology,the Finite Control Set-Model Predictive Control(FCS-MPC)strategy has been widely used in the control technology of three-level converters as a non-linear control scheme.FCS-MPC has the advantages of fast dynamic response and high robustness.However,the model predictive control itself still suffers from problems such as large computational effort,optimisation of weighting coefficients and irregular switching frequency.To address these shortcomings,a T-type three-level inverter system based on FCS-MPC is designed and an improvement strategy is proposed.The main work of this paper is as follows.The working mechanism of the T-type three-level inverter is analysed,its mathematical model in the rotating coordinate system is established,on this basis the model prediction controller is designed,the cost function with the AC voltage and the DC bus voltage as the control target is designed,and then the 27 space vectors are traversed to find the optimum according to the characteristics of the T-type three-level space vector distribution,the optimum control vector for a single cycle is determined and the switching signal is output.The influencing factors of the system model mismatch problem are also analysed,and the influence of different resistive parameters on the model predictive controller is analysed.An FCS-MPC method for optimising the computational effort of T-type three-level grid-connected inverters is proposed to address the problem of large computational effort of the model predictive controller in traversing the optimisation search.By constructing a single-objective cost function based on voltage vectors,the problem of designing weight coefficients is avoided and the steps of single-searching are reduced.Then,the influence of redundant small vectors on the DC bus potential is exploited to optimise the finite control set,so that the number of predictions per control cycle is reduced from 27 to 3,which improves the efficiency of the search while ensuring the balance of the neutral point potential.In addition,to improve the prediction accuracy,a control set division basis using the minimum weighted error square of the vector is proposed,and the vector angle is used to compensate for system delays,resulting in improved grid-connected current quality.A two-stage FCS-MPC method for T-type three-level inverters is proposed for the problem of variable switching frequency of the model prediction controller.In the two-stage model prediction controller,a double cost function is established for sequential execution,including a current prediction control objective function and a neutral point potential prediction control objective function.The former is constructed based on the inverter AC measurement mathematical model to control the inverter output current by predicting the reference voltage vector,and the latter is based on the zero-sequence injection principle,which averages the additional modulated signal for traversal seeking,so as to effectively balance the grid by outputting a voltage with a suitable duty cycle The latter is based on the zero-sequence injection principle,which equalises the additional modulation signal and carries out an iterative search for optimisation,so as to effectively balance the potential at the neutral point of the DC bus by generating a voltage vector with the appropriate duty cycle.