[P-EM10] Space Weather and Space Climate

[PEM10-10] Predicting Extreme Solar Energetic Particle Events with a Transformer Network

*Si Chen¹ (1.Aerospace Information Research Institute, Chinese Academy of Sciences)

Keywords:SEP events, transformer network, human space missions, solar activity

Accurate prediction of Solar Energetic Particle (SEP) events is essential for mitigating their adverse effects on spacecraft operations, aviation, and human spaceflight. This study employs a Transformer network model to predict extreme SEP events by treating them as changepoints, focusing specifically on forecasting the onset timing of these events. The model leverages historical data from 1997 to 2023, incorporating Coronal Mass Ejections (CMEs) and solar flare occurrence times as connected inputs to enhance the stability and robustness of predictions. The inclusion of these solar activity indicators significantly improved prediction accuracy. Results from the validation set demonstrate the model's effectiveness in capturing SEP event onset and intensity, particularly during periods of heightened solar activity. These findings underscore the potential of Transformer-based algorithms to provide reliable operational forecasting of SEP events, contributing to improved preparedness in SEP forecasting and future human space missions.