Artificial Neural Network Based Prediction of the Effect of Temperature and Irradiance on Photovoltaic Current-Voltage Curves

Abstract

An accurate method associated with an optimization procedure for studying the current-voltage (I-V) characteristics prediction of solar modules for irradiance and temperature is carried out. The method is based on Artificial Neural Networks (ANN). In the literature, different ANN architectures have been used to translate the solar module I-V curve to any temperature and irradiance conditions. We don’t find in these works an optimization study of the data used in the ANN training. In this work, we describe a new procedure to optimize the temperature and the irradiance values number used in the network training in order to design an ANN model with a few data used in the ANN training and with a high prediction accuracy. In order to validate this procedure, we have compared the I-V curves predicted by ANN with those obtained by simulations using analytical expressions. Results show a prediction accuracy between 99.3% and 99.9%.