A Comparative Assessment of Diffuse Fraction Models

Abstract

Many building performance applications (energy use, solar gains, thermal comfort, renewable energy systems, daylight, etc.) require information about both direct and diffuse components of the incident solar radiation. However, most meteorological stations only monitor global horizontal irradiance. Consequently, multiple methods have been proposed in the past to derive from measured global horizontal irradiance data the diffuse fraction. Thereby, additional data regarding other parameters such as clearness index, solar altitude, air mass, and turbidity are used. Given the importance of this procedure for the down the line tools, its reliability represents a critical issue. To address this point, we pursued an empirical approach. A number of existing methods for the computation of the diffuse fraction were selected. Actual measurements of global and diffuse irradiance were obtained for seven locations in USA and one location in Austria. The measured global irradiance data for these locations were fed to the aforementioned diffuse fraction models. The calculation results were then compared with the corresponding empirical data. The comparative assessment yielded a number of findings. The relative performance ("ranking") of the models was found to be more or less consistent across the different locations. However, none of the models can be said to be performing wholly satisfactory. For instance, the best performing model displayed only in 45 to 65 percentage of the cases relative errors less than 20%. In case of the worst performing model, the percentage of the cases for which relative errors were less than 20% was even smaller, namely 30% to 60%.