Worldwide performance assessment of 95 direct and diffuse clear-sky irradiance models using principal component analysis

Abstract

Abstract Accurate estimations of clear-sky direct normal irradiance (DNIcs) and diffuse horizontal irradiance (DIFcs) are crucial in solar resources assessment. This study examines 95 and 88 popular clear-sky irradiance models for their worldwide estimation capability of DNIcs and DIFcs, respectively. Atmospheric inputs are from MERRA-2 reanalysis and irradiance observations for validation are extracted from 100 ground stations across five major Koppen-Geiger climate zones. They consist of 24 million 1-min measurements of DINcs and 18.7 million 1-min measurements of DIFcs after quality control and clear-sky detection, during the 5-year period 2015–2019. Using principal component analysis, the performance of each clear-sky irradiance model is ranked separately among the five climate zones, as well as given a global rank. For the Equatorial, Arid, Temperate, Cold and Polar climates, it is found that Heliosat1-I , ESRA-I , REST2v9.1 , REST2v9.1 and CLS , respectively, are the best DNIcs models, while Modified Iqbal-C , Heliosat1-R , Calinoiu , Modified Iqbal-C and PSI-REST are the best DIFcs models. On a global worldwide basis, the three top-ranking models are REST2v9.1 , REST2v5 and MMAC-V2 for DNIcs, and Modified Iqbal-C , PSI-REST and MRMv5 for DIFcs. The results and rankings presented are strictly relative to MERRA-2 input data, and should not be extrapolated to results from alternate sources of atmospheric data. This detailed validation exercise revealed inconsistent performance of many models across different climates, possibly due to insufficient training and/or over-fitting of empirical relationships. Codes of all clear-sky irradiance models in the public domain are available online on the Github repository JamieMBright/clear-sky-models .