Dynamic Paths: towards high frequency direct normal irradiance forecasts

Abstract

Direct normal solar irradiance (DNI) series of high-frequency time resolution permit an accurate modeling and analysis of transient processes in concentrating solar thermal power (CSTP) technologies. Numerical weather prediction (NWP) models provide an overall understanding of solar forecasting, but they are unlikely to cover a local statistical representativeness of the DNI high frequency dynamics. On the contrary, local statistical information derived from site measurements can provide statistical behavior, but may not necessarily yield an explicit model for all of the physical relationships involved. In this work, we propose a novel locally-adapted procedure for high-frequency DNI forecasting that connects these two extremes, proposing a hybrid approach in which low frequency (3-h) NWP outcomes act as boundary conditions (assuring a physical consistency with site climatic behavior) and are supplemented with Dynamic Paths of local high frequency (1-min) DNI series (assuring a statistical reproduction of site high frequency dynamics). This methodology is tested with ground measurements in 4 locations situated in different climates, and compared with a forecast base case. The analyses are carried out by classifying each measured time series into 6 categories according to its daily clearness index. Finally, metrics for adequately compare high frequency DNI forecasts are discussed.