Study On Photovoltaic Grid-connected Inverter Modeling And Identification

Fossil energy supply and demand and the environmental pollution is becomingmore and more serious, the development and utilization of new energy sources becomerepresent the general trend. Solar energy because of its infinite, clean and so oncharacteristic, make its become the darling of the new field of energy. As the mostimportance application of solar energy, photovoltaic technology has become a researchhotspot. The inverter is the solar system access network interface, is the key part of aphotovoltaic array changing DC into AC, MPPT, filter, reducing the fluctuation ofoutput power is. Grid connection control inverter is the key to meet requirements of gridsystem, the traditional control method of the inverter take more dependence on preciseinverter model, and the inverter circuit is highly nonlinear, perturbed unpredictable andso on are challenge and difficulty of inverter modeling. The traditional modelingmethod based on inverter main circuit structure and working principle, through theestablishment of the circuit equation and deduces the model. But these models are basedon certain assumed and simplified conditions, and these assumptions and simplificationin the actual system is not set up. Therefore, a practical applied inverter model hasimportant significance in the system of grid-connected inverter control system toimprove flexibility, photovoltaic grid-connected analysis and so on.In this paper the commonly used inverter models were classified and compared, themodeling of these models was induced and summarized, in the foundation of analyzingthe advantages and disadvantages of the modeling, proposed a new inverter modelingmethods based on the experimental test, the modeling process was introduced and a newmultiple input multiple output model of inverter was established.On the assumption that the measurement noise is white noise, inverter modelidentification equations based on least squares method was presented. The equationsbased on “black box” model, make the algorithm have simultaneous identification ofstructure and parameter estimation function, the validity of the model was verified byexperiment.On the assumption that the measurement noise is colored noise, for structureidentification and parameter estimation of inverter model by generalized least squaresalgorithm was presented. At the same time, the two algorithms are compared in details.Based on the above modeling ideas, Experimental test platform based on inverteridentification was built, and the experimental procedure of input signals, the sampling interval and the length of the data and so on key factors on the identification order andparameters precision are analyzed in detail. Finally, according to the different timesections and different conditions of the experimental data on the model inspection, theresults show that the correctness of the model and the validity of modeling method.