Application Study Of Predictive Control In Four-leg Active Power Filter

At present,with the rapid development of power electronics,the problems in the grid,like harmonics and three-phase unbalance,are more and more serious.Four-leg active power filter(APF)is an effective measure to solve the two common power quality problems above in low voltage distribution network.The research of APF current loop control algorithm is an important link affecting APF performance.Model predictive control(MPC)is simple to be implemented and has excellent dynamic performance,which can realize the tracking control of APF current loop.In this paper,the application of model predictive control in current loop of four-leg active power filter is studied.In this paper,the discrete mathematical model of four-leg APF is established.All the possible switching states are analyzed,and the model predictive current control(MPCC)algorithm is applied to control current loop of four-leg APF by utilizing the inherent characteristics of four-leg inverters with finite switching states.The single-objective and multi-objective control are realized.Aiming at the problem that the calculation amount of MPCC algorithm is large,the influence of time delay on the compensation performance of four-leg APF is analyzed,and the delay compensation is carried out to improve the control effect of MPCC.Although MPCC algorithm is simple,flexible,fast dynamic response and easy to implement in hardware,it still has some shortcomings,such as large amount of calculation,unsteady switching frequency,which lead to difficulties in filter parameter design.In order to solve the above problems,model predictive vector control is adopted to control four-leg APF in this paper.Aiming at the problem of dead time that must be added in the APF hardware implementation process,the influence of dead time on the control error of MPVC algorithm is analyzed.The advantages and disadvantages of MPCC and MPVC are analyzed and compared in several different aspects.Real vector based model predictive control algorithm has some limitations: big tracking error,high sampling frequency,big inductance value,which cauce large volume and high cost of APF system.To solve the above problems,the virtual vector based model predictive control is proposed in this paper.The real states of four-leg converter are projected onto the ?? plane and the ? axis respectively.The positions of the maximum voltage error on ?? plane and ? axis are considered,and the feasibility of the virtual vectors in these positions is determined.The virtual vectors are synthesized in the feasible position,so that the converter control region is fully used.To reduce the calculation of the algorithm,the optimization algorithm based on regional judgment is taken.The simulation of model predictive control strategy adopted in the four-leg APF is carried out.The simulation results show that the harmonic compensation performance of the four-leg APF with MPCC is good and the dynamic response is fast.By further analyzing,the validity of delay compensation and the correctness of dead-time effect analysis are verified.The feasibility of predictive control based on virtual vector model is verified by simulation.The steady and dynamic performances of real vector based MPC and virtual vector bases MPC are analyzed and compared.The results show that the virtual vector based MPC has higher control accuracy and better compensation performance.When the same compensation performance is achieved,the virtue vector based model predictive control could reduce the demand for sampling frequency or filter inductance.Based on the two-level four-leg APF experimental platform,the experiment of model predictive control is carried out.The validity of model predictive current control and delay compensation is verified.The steady and dynamic performance of the real vector based MPC and virtual vector based MPC are compared and analyzed.The conclusions are consistent with the simulation results.