Cloud type and aerosol aware solar irradiance variability classification for solar resource assessment

Classifying solar irradiance variability is essential for solar applications such as forecasting services and evaluating irradiance models. This study extends the previous direct normal irradiance classification method of Schroedter-Homscheidt et al. (2018) to use only global horizontal irradiance measurements, which are more widely available. The classification method is based on a reference database formed by 277 hours of global horizontal irradiance measurements at minute resolution from the Carpentras Baseline Surface Radiation Network station, manually classified into eight variability conditions from clear to variable and overcast skies. The algorithm consists of a three-step approach (discriminant filter, probability evaluation, and median-distance method) using 13 variability indices derived from 1-minute data and the McClear clear sky model. The method was evaluated using 66 ground measurements worldwide, across different climates. Overall accuracy of the global horizontal irradiance-based classification method in the reference database is 92%, which outperforms the 77% of the previous direct normal irradiance-based method. The evaluation of the results on a global scale is also found to be consistent in the different climates. Two application cases further demonstrate the utility of the GHI classification: evaluating the CAMS Radiation Service (CRS) over France and assessing distinct CRS versions, both under different variability conditions. • Previous DNI classification method for 8 variability conditions extended for GHI • Reference database from BSRN Carpentras with 277 GHI 1-min hours manually classified • GHI and clear sky GHI are used as input for the calculation of 13 variability indices • Classification successfully applied in 274 world stations with different climates • Evaluation of CAMS Radiation Service for different variability conditions