Comparative study of à ngström's and artificial neural networks' methodologies in estimating global solar radiation

The aim of the present research is the comparative development of a variety of models for the estimation of solar radiation on a horizontal surface. By using two different methodologies, models of various complexities have been developed and tested. The first methodology refers to the traditional and long-utilized Ångström's linear approach which is based on measurements of sunshine duration. The second methodology refers to the relatively new approach based on artificial neural networks (ANN) and it can be based on sunshine duration measurements but also on other climatological parameters. Three Ångström-type models and seven ANN-type models are presented. All of these models are verified against independent data and compared. Lack of sunshine duration measurements renders Ångström's approach inapplicable; hence the feasibility of applying the ANN models for the calculation of solar radiation in places where there is a lack of sunshine duration measurements is investigated.