Modeling the irradiance dependency of photovoltaic modules using Artificial Neural Network

This paper suggests an efficient approach to accurately model the irradiance dependency of photovoltaic (PV) modules by using the artificial neural network (ANN). In fact, a Feed-Forward Back Propagation (FFBP) neural network has been developed in this paper to deal with this concern. This network uses the Levenberg-Marquardt (LM) as the back-propagation training algorithm to tuning the network bias and weights. Furthermore, in order to show the merits of this strategy, a set of real experimental current-voltage (I-V) characteristics, measured from a commercialized PV module (Multi-crystalline Shell S75 PV module) are employed as the training and test datasets. These data are measured at nine irradiances (200, 300, 400, 500, 600, 700, 800, 900 and 1000 W/m
²
). Finally, the real measured I-V characteristics and the estimated ones (using the trained network) are compared. The obtained results demonstrate clearly the high accuracy of the developed model to mimic the irradiance dependency of PV modules.