Sub-hourly Validation of the Direct Normal Solar Irradiance Clear Sky Models

The Direct Normal Irradiance (DNI) is the fuel for all concentrating solar energy systems. It could either be measured by means of highly sophisticated instruments or computed using DNI models. These models utilize meteorological and atmospheric parameters as inputs to predict instantaneously the DNI in regions where no measurements are available. This paper presents a performance analysis of the Ineichen, Iqbal and Solis models to predict the direct normal irradiance when we present low-quality input parameters. The models have been used to predict DNI in a southern Tunisian region. The prediction results have been compared to measurements. The sub-hourly statistical evaluation of these models shows the following errors: when the DNI is higher than 700 W/m² (at noon), the Iqbal model is the most accurate prediction model with a relative root mean square error (rmse%) equal to 9.82%. The same model has recorded the best accuracy when DNI is lower than 700 (the sun is close to the horizon), the rmse% reaches 27.88%. Such results clearly show that presenting a limited quality inputs to the models, at the limit, can be accepted to perform DNI prediction at noon but they are not suitable to predict effectively the DNI when the sun is close to the horizon.