Articles | Volume 20
02 Jan 2024
 | 02 Jan 2024

Probabilistic end-to-end irradiance forecasting through pre-trained deep learning models using all-sky-images

Samer Chaaraoui, Sebastian Houben, and Stefanie Meilinger

Data sets

A comprehensive dataset for the accelerated development and benchmarking of solar forecasting methods (Version V1) [Data set] H. Carreira Pedro, D. Larson, and C. Coimbra

Short summary
This method generates probabilistic solar power forecasts of up to 30 min using artificial intelligence and fish eye images of the sky. We investigated the impact of parameters describing the variability of the solar power on the forecast performance and divided the solar power to its direct and diffuse components. The method achieves overall good performance, while variability parameters and dividing the solar power don't show any performance gain. However, training time can be decreased.