Exploring possibilities for solar irradiance prediction from solar photosphere images using recurrent neural networks

Abstract

Studies of the Sun and the Earth’s atmosphere and climate consider solar variability as an important driver, and its constant monitoring is essential for climate models. Solar total and spectral irradiance are among the main relevant parameters. Physical semi-empirical and empirical models have been developed and made available, and they are crucial for the reconstruction of irradiance during periods of data failure or their absence. However, ionospheric and climate models would also benefit from solar irradiance prediction through prior knowledge of irradiance values hours or days ahead. This paper presents a neural network-based approach, which uses images of the solar photosphere to extract sunspot and active region information and thus generate inputs for recurrent neural networks to perform the irradiance prediction. Experiments were performed with two recurrent neural network architectures for short- and long-term predictions of total and spectral solar irradiance at three wavelengths. The results show good quality of prediction for total solar irradiance (TSI) and motivate further effort in improving the prediction of each type of irradiance considered in this work. The results obtained for spectral solar irradiance (SSI) point out that photosphere images do not have the same influence on the prediction of all wavelengths tested but encourage the bet on new spectral lines prediction.