Research On Predictive Control For Three-Phase Four-Leg Grid-Tied Inverters

In recent years,the renewable power generation filed,including photovoltaic power generation,wind power generation and so on,has attracted more and more attentions.A new worldwide energy technology revolution is about to begin.The grid-tied inverter is critical interface between the renewable power and the grid.Control method is an important link to ensure the quality of power supply and energy conversion efficiency.This paper focuses on improving the power quality,and conducts a research on predictive control for three-phase four-leg grid-tied inverters.The main contents are as follows:A three-phase grid-connected inverter model is established.Starting from the inverter topology,the main circuit current and voltage equation is analyzed,and the mathematical description model is established,and the state equation of the vector form is deduced.In order to facilitate the digitization of the control method,the inverter mathematical model is discretized.Transform the discrete model to predictive model with two parameters,prediction horizon and control horizon,which are analyzed in detail,to lay the foundation for the design of predictive controller.The unconstrained MPC strategy of grid-tied inverter is proposed.Based on predictive model,a cost function is defined by a main taget,which is the minimized error between predicted trajectory and desired trajectory,and a sub taget,which is smooth control variable.Obtain the optimal formulation by minimize the cost function to complete the design of predictive controller.Then,select the parameters of controller,include prediction horizon,control horizon and wighting factors.A simulation platform is builted to imitate the model predictive control.Meanwhile,the simulation results of FCS-MPC and PI control are used as contrasts.It demonstrate that the predictive controller is feasible and effective.The constrained MPC strategy of grid-tied inverter is proposed.The derivative of duty cycle is added in the cost function as a constraint to enhance curbing strength of control variable.Tramsformed the cost function with a constraint to a quadratic form,and use the neural network optimal algorithm to get the optimal duty cycle quickly and accurately.Based on the simulation platform,the simulation results verify that the constraint MPC can further improve the performance of the grid-tied inverter.An implementation is designed for predictive controller,which is based on a FPGA plus DSP hybrid architecture.In order to overcome the weakness of MPC,poor real-time capability and large amount of calculation,an FPGA plus DSP hybrid architecture is proposed.Taking advantage of FPGA to construct predictive controller,taking advantage of DSP to achieve sampling,driving and communication,finish the design of control circuit and software.The causes of the harmonic generation of the grid are analyzed,and the state estimation is carried out in the software for the characteristics of the dead zone and the sampling delay,thus offsetting the influence of the dead zone and the sampling delay.In this paper,the unconstrained MPC and the constrained MPC are proposed,which are further verified experimentally to verify the effectiveness.