Study On Model Predictive Control Algorithm Of Active Power Filter

Active Power Filter(APF) is an effective measure to solve the problem of power system harmonic pollution, and the current tracking control is one of the key links of the active power filter system, which seriously affects the compensation performance. In this thesis, two level and three-level APF used in low voltage power system were taken as research objects, and the compensation control strategy of APF was studied in details.The traditional current loop controller of APF usually takes compensation effect as the single objective, in addition, SVPWM modulation process is relatively complex. Deadbeat model predictive control(MPC) strategy was used in this thesis in order to achieve multi-objective control of APF.Firstly the design approach of MPC controller used in two-level APF was given, and the range of current tracking error was analyzed; By using graded optimization method of cost function to improve the controller, the MPC strategy can realize multi-objective control as well as limit the current tracking error within certain range. Referenced to the analysis of current tracking error range of three-level APF, deadbeat model predictive control strategy which was suitable for neutral-point clamped(NPC) three-level APF was given in this thesis: the cost function which evaluated the performance index of tracking the reference current, keeping the balance of neutral point potential and reducing the switching loss was established to realize multi-objective control. Graded optimization design method of cost function based on the judgment of reference vector sector was used to improve the traditional MPC controller, which has not only reduced the amount of calculation, but also simplified the design of cost function and the process of vector optimization.In order to ameliorate the flaw that deadbeat model predictive control effect is depended on system parameters, a composite predictive control strategy which combined repetitive deadbeat control with model predictive control was proposed. The parameters of repetitive deadbeat controller were designed and then the steady state error and the adaption to parameters’ shift were analyzed. Theoretical analysis indicates that repetitive deadbeat control has good steady-state performance and adaptability to parameters’ shift, meanwhile, MPC control has rapid dynamic response and control flexible. The composite predictive control method was given based on their advantages respectively.The adopted control strategies were simulated on the basis of theoretical analysis. The simulation results show that two-level APF based on deadbeat model predictive control strategy can compensate harmonic current effectively as well as reduce the switching loss, and it has good adaptability for load fluctuation. By using graded optimization method of cost function, the MPC strategy can limit the current tracking error within certain range and make it more easily to realize multi-objective flexible control. Compared to two-level APF, deadbeat MPC control strategy applied to three-level APF has more advantages, which can realize tracking reference current and keeping the balance of neutral point potential under lower sample frequency, meanwhile, it can achieve better control precision and lower average switching frequency, and the control system has good dynamic performance. Graded optimization design method of cost function based on the judgment of reference vector sector can achieve similar control effect with traditional model predictive control while reducing the complexity of the algorithm and more easily to regulate parameters. This algorithm has good adaptability to load fluctuation, but current tracking accuracy much depends on the main circuit parameters. The system of composite predictive control has good dynamic and static performance, moreover, it improves the flaw that deadbeat MPC control effects is depended on system parameters to a certain extent, the compensation effect is better than deadbeat MPC control when parameters are changing.Experimental study of deadbeat model predictive control has been conducted based on two-level APF experimental platform. The experiment results indicate that the control method can restrain the harmonic effectively and has rapid dynamic response. Compared with traditional deadbeat control, the MPC deadbeat control has the advantage for lower average switching frequency under the same control accuracy.