New combined models for estimating daily global solar radiation from measured air temperature in semi-arid climates: Application in Ghardaïa, Algeria

• New combined models been introduced for estimating daily global solar radiation.
• The BNN is the most accurate model compared with combined models.
• Results indicate that the new calibrated models are able to estimate the global solar.

In this paper, combined empirical models and a Bayesian neural network (BNN) model have been developed to estimate daily global solar radiation (GSR) on a horizontal surface in Ghardaïa, Algeria. An experimental database of daily GSR, maximum and minimum air temperatures of the year 2006 has been used to estimate the coefficients of the empirical models, as well as to train the BNN model. Six months of the year 2007 (summer period: May, June, July, and winter period: October, November, December) have been used to test the calibrated models, while six months of the year 2012 (from 1st February to 31th July) have been used to check generalisation capability of the developed models as well as the BNN model. Results indicate that the new calibrated models are able to estimate the global solar radiation with an excellent accuracy in this location. Calibrated models are also compared with the developed BNN model to show their effectiveness.