Modeling of Single Phase Inverter of Photovoltaic System Using System Identification

Abstract

This paper proposes a new method to modeling a power inverter of grid-connected photovoltaic system by using a system identification approach. In this method, the system is considered as a black box of which it is not necessary to know structures and parameters inside. Modeling of one type of grid connected single phase inverter is carried out. Four linear models have been compared, i.e. an Autoregressive with Exogenous (ARX) model, an Autoregressive Moving Average with Exogenous (ARMAX) model, a Box-Jenkins (BJ) model an Output Error (OE) model. Four nonlinear models are studied, i.e, a Nonlinear Autoregressive with Exogenous (ARX) model, a Hammerstein model, a Wiener Model and a Hammerstein-Wiener Model. The best linear model is an Output Error model whereas the best nonlinear model is a Hammerstein-Wiener model with wavelet network estimators. Comparing modeling of the inverter by an Output-Error (OE) model and a Hammerstein-Wiener model, a Hammerstein-Wiener model is better because of its lower order and higher percentage of best fit.