Research On Control Strategy Of Active Power Filter Based On Model Predictive Control

With a large number of power electronic equipment and nonlinear loads are widely used in power systems,the problem of harmonic pollution becomes more and more serious.Active Power Filter(APF)has attracted much attention because of its large compensation capacity,flexible configuration,and good compensation effect.In this dissertation,three-phase three wire three-level APF is taken as the research object,and its current tracking control strategy and neutral point potential balance strategy are studied.Firstly,the mathematical model of the three-phase three-wire Neutral Point Clamped(NPC)three-level APF is established,and the problem of the neutral point potential balance is analyzed.The predictive model and the link of constructing value function of traditional model predictive control strategy are analyzed,its working principle,advantages and disadvantages are described in detail;Aiming at the delay problem in the model prediction algorithm,the influence of delay is analyzed,and a simple two-step prediction method is proposed to solve the delay problem.Aiming at the neutral point balance problem,this paper proposes a neutral point potential balance algorithm based on the prediction principle,uses the feature that the redundant small vectors have the opposite effect on the neutral point potential,and combines the current neutral point potential and the redundant small vectors to predict the neutral point potential at the next moment.Calculating the action time of the redundant small vectors that makes the neutral point potential of the next moment zero,and readjusting the action time of the redundant small vectors according to the calculated time,and the purpose of neutral point potential balance is finally achieved.Aiming at the problems of many switch states to be optimized and a large amount of calculation in the rolling optimization of traditional model prediction algorithm,although a literature has proposed a simplified algorithm based on sector judgment,the number of switch states to be optimized is reduced from 27 to 8,but its calculation amount is still too large.On this basis,this paper proposes a simplified model prediction algorithm based on two-level partition judgment.two-level partition judgment reduces the switch states to be optimized from 27 to 3,which greatly reduces the calculation amount of the algorithm,and the correctness is verified through simulation and experiment.A dynamic weight coefficient model prediction algorithm is proposed.This algorithm improves the dynamic performance of the system by configuring a constant weight coefficient with a dynamic coefficient that changes according to the error.Firstly,a new objective function construction method is given to determine the constant weight coefficient,and then the dynamic coefficient is adjusted according to whether the existing control target tracking error is within the expected range.This algorithm can not only quickly determine the constant weight coefficient,but also improve the antiinterference performance of the control target and the ability to adapt to dynamic changes.Combined with the small calculation amount of the model prediction algorithm based on the two-level partition judgment,a compound predictive control algorithm based on dynamic weight coefficients is proposed,when the control target error is within the expected range,the model prediction algorithm based on the two-level partition judgment is selected.When any control target error exceeds the expected range,select the model prediction algorithm of dynamic weight coefficient,the algorithm takes into account both dynamic and steady-state performance.The hardware and software design of the APF experimental platform is introduced in detail,and the correctness and feasibility of the proposed model prediction algorithm and neutral point balance algorithm are verified by experiments.This dissertation includes 79 figures,7 tables and 60 references.