Research And Optimization Of MMC Model Predictive Control

After proposed for the first time in 2002,Modular Multilevel Converter(MMC)has been widely applied in many power electronics applications due to its outstanding operation conditions.With the highly modular sub-module structure,MMC has good scalability and could be better extended to higher voltage and higher power applications.Large numbers of output voltage levels lead to its high voltage and current output characteristic.Large numbers of sub-modules and unique modulation methods reduce its switching frequency and switching loss.Therefore,MMC is very suitable in high voltage and high power applications such as High Voltage Direct Current.Because of complicated topology,the mathematical model of MMC has the multi-input multioutput nonlinear characteristic.What's more,several control targets need to be met such as output voltage or output current control,circulating current depression and voltage balance of sub-modules and there will be more targets in the case of unbalance grids.Consequently,a proper control strategy is of great importance for the steady and dynamic performance of MMC.Model predictive control(MPC)has the characteristics of dealing with multi-variable,multi-system constraints and non-linear systems directly.Therefore,model predictive control is very suitable for the application of MMC.Based on the background of application of MMC in HVDC and the basic work of establishing a discrete mathematical model and analyzing modulation strategy for MMC,this paper mainly researches the design process of cost function,the calculation amount of algorithm and the optimization of dynamic control effect in MMC model predictive control.Firstly,the topological structure and basic operation principle of MMC are discussed.On this basis,the single equivalent continuous-domain mathematical model of MMC is deduced.Combining with its operation principle,two most common modulation modes applied to MMC-HVDC are discussed.The model predictive control method used in power electronic converter is briefly introduced,and the discrete-domain mathematical model of MMC is deduced.The basic method and design process of model predictive control applied in MMC are discussed.Based on the formation mechanism of internal circulating current in MMC and the relationship between harmonics in arm current and the amplitude and phase of fundamental part of output current,an optimized design method of cost function weight for MMC-MPC is proposed.Its effectiveness is verified by simulation.Several typical Finite Control Set MPC methods used in MMC are compared,and an N+1 level MPC with saving in computation is proposed.This method is extended to the application of 2N+1 level and compared with other methods.For its dynamic performance,two optimization methods are proposed.In the environment of MATLAB/Simulink,a three-phase MMC model with multiple submodules is built,and the above theory is verified and analyzed by simulation.A single-phase MMC experimental platform was built to verify and analyze the theory mentioned above.The effectiveness of the proposed MPC and two dynamic performance optimization methods are verified.