New types of simple non-linear models to compute solar global irradiance from cloud cover amount

Most simple solar radiation models are built with reference to a basic model calibrated for clear sky conditions. New models are built in this paper with reference to two basic models, calibrated for clear sky and overcast sky conditions, correspondingly. The models are illustrated with measurements data from five meteorological stations in Romania (South-Eastern Europe), where the ratio between the solar irradiance on overcast sky and clear sky, respectively, ranges between 0.27 and 0.55. A three-parameter model is used for both basic models. Three new types of regression models were developed from the two basic models. They are non-linear generalizations of the Angstrom-Savinov model. Their accuracy decreases by increasing the cloud cover amount. One model has been further tested. A set of regression coefficients has been obtained by fitting that model to all available data, for all stations. The model based on this set of regression coefficients has good accuracy in any particular station, when low and intermediate cloudy skies are considered. The model based on a set of regression coefficients obtained in a given station was tested in other stations. Its accuracy is good (or marginally, good enough) on skies with low and intermediate cloudy skies. Its accuracy is good for intermediate zenith angles, ranging between 30° and 70°.