Search

All sky imager (ASI) based nowcasting system can provide spatially and temporally highly resolved solar irradiance information for the next minutes ahead. Nowcasts, which capture the intra-hour variability of the incoming downward shortwave solar irradiance, have the potential to optimize the operation of solar power plants as well as electrical grids. Such automatized optimizations require a deep understanding of the accuracy in the nowcasting system at any given moment. State of the art validation procedures of ASI based nowcasting systems use scalar error metrics without regards to the actual weather conditions. Yet, the performance of nowcasting systems varies strongly with the prevailing weather conditions. Deviations increase for more complex and variable conditions, for which it is more challenging to detect and model the clouds in the sky. Thus, depending on the used data set such validation results may not be meaningful to describe the expected accuracy in realistic and individual optimization situations. A novel validation procedure is presented in this work, which discretizes the validation data set in distinct temporal DNI variability classes. Individual error metrics are determined as function of the lead time and DNI variability class. This approach is applied for a two ASI based nowcasting system as operated on five distinct sites distributed in Spain, Portugal and Germany, over a combined period of more than 4.5 years. The obtained validation results emphasize that the novel classification method enables a comparison in nowcast performance between the sites despite of distinct local meteorological conditions. The presented method allows the estimation of the overall accuracy of nowcasting systems at a new site if DNI data in 1 min resolution are available. [ABSTRACT FROM AUTHOR]