Estimation of vertical global solar irradiance using artificial neural networks

Solar radiation data are essential for solar photovoltaic systems and passive energy-efficient building designs. Nevertheless, there is a lack of availability of these data, leading to its growing demand. More recently, the artificial neural network (ANN) has been discovered as a better alternative to empirical models when estimating solar irradiance data. However, unlike empirical models, fewer studies have explored the use of ANN in global irradiance modelling, especially for short time intervals. In this study, ANN is used to correlate vertical global irradiance along the four principal vertical surfaces. Hourly solar radiation and meteorological data measured between June 2019 and May 2020 were used for the study. Furthermore, a sensitivity analysis test was carried out to ascertain the contributions of the individual predictors to the developed ANN models. Finally, a simple regression model for estimating vertical global irradiance was also developed. The results from the predictions and other statistical analysis were presented.