| For the high-frequency Si C photovoltaic grid-connected inverter,it is of great significance to study the optimal design of its filter and controller to reduce the mass and loss of the inverter and improve the grid-connected performance.This paper is based on the single-phase photovoltaic grid-connected inverter controlled by current loop quasi-PR,studies its control strategy,and uses machine learning algorithm to optimize the design of output filter and controller parameters.The specific research contents are as follows:Firstly,the principle of single-phase grid-connected inverter and its closed-loop control method are analyzed,and its small-signal model is established.Based on the traditional empirical method,the parameters of the grid-connected LC filter and the parameters of the current loop quasi-PR controller are designed.At the same time,the design ranges of filter parameters and controller parameters are determined for the subsequent optimization design.On this basis,a simulation model of a singlephase grid-connected inverter is established and the simulation results are analyzed as a control group for the subsequent optimization design based on machine learning.Then,the inductance parameter model and the discrete capacitance parameter model are analyzed and designed,and the overall design mod-el of the filter is established by considering the internal resistance of the components.The principles of support vector machine(SVM)and artificial neural network(ANN)are analyzed,and a machine learning-based filter parameter optimization design scheme is established.The training data through the filter design process is collected to input support vector machine and artificial neural network to train the corresponding prediction model,and the model is used for the prediction of LC parameters.The design point of the filter is taken as the result of the filter optimization design,and the simulation analysis is carried out.Next,the parameters of the current loop quasi-PR controller are optimized and designed based on the machine learning method.Within a given range,the simulation results corresponding to different controller parameters are generated.As a data set,the artificial neural network model is trained to obtain the optimal solution of the controller parameters with the smallest current total harmonic distortion(THD).An iterative optimization method for filter parameters and controller parameters based on machine learning is established.The results of each generation determine the design scope of the next generation.After multiple generations of optimization,the optimal solution for the comprehensive design objectives of filter mass,loss,and current THD is obtained.Finally,based on the working principle of single-phase photovoltaic gridconnected inverter,its parameters are designed and a complete hardware experimental platform of single-phase grid-connected inverter is built.The comparative analysis based on the traditional empirical method,the parameter optimization design method based on machine learning,and the steady-state experimental results obtained by the iterative optimization method shows the effect of the filter mass and loss,and the current THD optimized by the machine learning method.Based on the parameters obtained by the iterative optimization design method,the dynamic response test of the system is carried out to verify the effect of the method proposed in this subject. |
