A Modeling Study of ≥2 MeV Electron Fluxes at Different Prediction Time Scales Based on LSTM and Transformer Networks

The geostationary orbit is located in the Earth's radiation belts where the geostationary satellites may encounter high-energy electrons that could potentially damage them. With the number of geostationary satellites steadily increasing, there is an increasing need for services that provide accurate predictions of relativistic electron fluxes and provide hazard warning to satellite operators. We develop prediction models based on LSTM and transformer networks to provide the predictions at different prediction time scales of ≥2 MeV electron fluxes with 5-minute resolution at the geostationary orbit. Compared with previous models, our models show great performances for the 1-hour and 3-hour predictions.

地球静止轨道（geostationary orbit）坐落于地球辐射带内，在此运行的地球静止卫星（geostationary satellites）可能遭遇潜在可造成设备损伤的高能电子。随着地球静止卫星数量稳步攀升，为卫星运营商提供精准相对论电子通量（relativistic electron fluxes）预测与灾害预警的服务需求日益增长。本研究基于长短期记忆网络（LSTM）与Transformer网络构建预测模型，可在地球静止轨道处以5分钟分辨率，针对≥2 MeV电子通量开展不同预测提前时长的预报。相较于既往模型，本模型在1小时与3小时提前预测任务中展现出更优异的性能。