| Abstrakt: |
The geosynchronous orbit (GEO) is a region filled with energetic electrons and it hosts hundreds of satellites. Electron fluxes at GEO can change sharply within hours, making high‐time‐resolution prediction crucial. In this study, we develop and compare two neural networks for persistent high‐time‐resolution prediction: long short‐term memory with temporal pattern attention (TPA‐LSTM) and Transformer. Unlike most previous models, which only output electron fluxes, our models output the same parameters as the inputs, including magnetic local time, solar wind speed, solar wind dynamic pressure, AE, Kp, Dst, the north‐south component of the interplanetary magnetic field, and electron flux data from GOES‐15. The models are trained on approximately six years of data (2012–2016) and validated using about one year of data (2017–2018). We compare the TPA‐LSTM and Transformer models using > ${ >} $0.8 MeV electron fluxes and find that while the Transformer model performs slightly better, the difference is not statistically significant. Considering the Transformer's higher computational cost, we use the TPA‐LSTM model to develop prediction models for electron fluxes of 275, 475, > ${ >} $0.8 MeV, and > ${ >} $2 MeV with a 5‐min resolution at GEO, up to 3 days. The prediction efficiencies (PE) for 275, 475, > ${ >} $0.8 and > ${ >} $2 MeV electron fluxes based on about one year of test data (2018–2019) are 0.799, 0.831, 0.849, 0.881 (1‐day prediction); and 0.551, 0.618, 0.663 and 0.710 (3‐day prediction), respectively. Our high‐time‐resolution persistent models should be useful for both protecting satellites at GEO and serving as boundary conditions for physics‐based radiation belt models. Plain Language Summary: This study focuses on developing high‐time‐resolution predictions of electron fluxes at geosynchronous orbit (GEO), a critical region for satellites due to its stable position relative to Earth's surface and the presence of high‐energy electrons that can cause charging issues. We compared two machine learning models, TPA‐LSTM and Transformer, to forecast these fluxes using various inputs, including both geomagnetic indices, solar wind indices and historical flux. We find that both models can make 1‐day predictions with high efficiency, and the Transformer model demonstrated slightly better accuracy. However, the TPA‐LSTM model was chosen for its lower computational cost to build predictive models for multiple energy channels, as the difference in performance is not statistically important in our case. Our high‐time resolution models should be useful for satellite protection at GEO and improving space weather forecasting models. Key Points: TPA‐LSTM and Transformer models effectively predict high‐resolution electron fluxes at GEO, with the Transformer performing slightly betterWe developed cost‐effective TPA‐LSTM models to forecast electron fluxes across multiple energy channels at GEO for up to 3 daysThe model performs better in predicting GEO fluxes during quiet periods compared to storm times, due to the scarcity of storm‐related data [ABSTRACT FROM AUTHOR] |
