Speaker
Description
Forecasting solar wind conditions at the Lagrange L1 point is determinant for mitigating space weather risks caused by high-velocity streams. As stated in several recent papers, the variation of the solar wind velocity is a key proxy for space weather events like solar storms and geomagnetic perturbations. In the light of Augura Space's mission to make AI driven space weather forecasting operational, we are developing DeepHelio, a project investigating deep learning approaches to predict solar wind speed at L1.
Our pipeline combines AIA solar images and HMI magnetograms from the Solar Dynamics Observatory (SDO), OMNI in-situ measurements of the solar wind and geomagnetic indices to train a multimodal deep learning regression model. In this poster, we present our studies on novel model architectures - combining traditional neural networks, CNNs and transformers - and the results we obtained.
Augura Space’s DeepHelio project seeks to fill the gap between state-of-the-art computer vision models and heliophysics.
