Research On The Design Of Model Predictive Control Weighting Factors Of Wind Turbine Systems

With the development of wind energy conversion and the increasing installed capacity,the development of wind power industry has become an important part of supporting China's renewable energy strategy.As one of the key of wind turbine system,various control strategies for wind turbine system are developing.Model predictive control(MPC)has always been a research hotspot in this field in recent years,because it can flexibly contain a variety of control objectives and nonlinear constraints of different physical scales in wind turbine system,and it has been developed and applied in industry.However,there are also some problems in MPC,including the complex design of weighting factors in cost function..At present,there are many solutions to this problem,but most of them are based on specific models or transform the problem into another problem.The solution of the weighting factors design problem can give full play to the advantages of MPC,so that it can freely add control objectives of different physical scales according to the demand,and improve the system efficiency,power quality,safety and other possible performance.Therefore,based on the per-unit system model,this paper studies the weighting factors design of the MPC.Firstly,this paper introduces four kinds of control strategies in detail:vector control,direct control,deadbeat control and MPC,and verifies the control performance of the four kinds of control strategies in the direct-drive back-to-back power converter permanent-magnet synchronous generator wind turbine systems through simulation comparison,expounds their advantages and disadvantages,and summarizes the advantages of model predictive control.Then,the weighting factors of MPC is studied in depth.At present,there are two main ways.to design the weighting factors of MPC,one is to obtain the weighting factors through various optimization methods,the other is to avoid the design problem of the weighting factors by using cascaded structure cost function and other methods.Next,according to the two kinds of design ideas,this paper proposes a weighting factors design scheme,respectively.The first solution is to introduce artificial intelligence into this field and explore the use of artificial neural network(ANN)to solve the problem of weighting factors.Firstly,the paper introduces the ANN and analyzes the theoretical feasibility of solving the problem.Then the ANN and MPC are combined to obtain the training data through repeated simulation,and then the corresponding neural network,namely the generation model of the actual simulation model,is obtained by using the training data.Based on this generation model,the corresponding weighting factors combination is selected according to each index.At the same time,the advantage of the scheme is confirmed by using the simulation software of PLECS.Another solution is to improve the existing sequential model predictive control,which avoids the weighting factors design.A dynamic sequential model predictive control is proposed.In this scheme,the cost function of cascade structure is adopted,and the switching vector is initially screened by the cost function corresponding to the high priority control target.The selected switching vector is input into the cost function of the sub priority to further screen,and finally the optimal switching voltage vector is obtained.By dynamically adjusting the threshold value of the cost function selection vector,the number of switching vector selection is automatically determined.Finally,the scheme is verified by the simulation of PLECS,and compared with traditional sequential model predictive control and traditional model predictive control.