摘要
为了进一步提升导航高精度实时定位的精度,提出一种预测全球电离层的多通道ConvLSTM模型:根据全球电离层地图(GIM)时空变化非线性的特征以及电离层总电子含量(TEC)与太阳活动和地磁活动的相关性,提出以行星际三小时磁情指数(Kp)、太阳黑子数(SSN)以及TEC作为多通道输入的基于编码器-解码器的卷积-长短期记忆神经网络(ConvLSTM)模型;然后将2018-2020年的电离层TEC及相关数据作为数据集,提前1 d预测GIM。结果表明,基于多通道输入的模型在预测任务上具有显著优势,且不同输入的ConvLSTM模型皆优于欧洲定轨中心1 d预报GIM(C1PG)模型;在地磁平静期和磁暴期,多通道输入的模型表现良好。
In order to further improve the precision of high-precision real-time positioning in navigation,the paper proposed a multi-channel ConvLSTM model for predicting the global ionosphere:according to the nonlinear characteristics of the temporal and spatial variations of the global ionospheric map(GIM)and the correlation of the ionospheric total electron content(TEC)with solar and geomagnetic activities,the convolutional long short-term memory(ConvLSTM)neural network based on encoderdecoder was given with the multichannel inputs containing interplanetary three-hour index(Kp),sunspot number(SSN)and TEC;then,the ionospheric TEC and related data for 2018 to 2020 were used as a dataset to forecast GIM 1 day in advance.Results showed that the model based on multi-channel inputs would have a significant advantage in the forecasting task,and the ConvLSTM model with different inputs could outperform the 1-day prediction of GIM(C1PG)of the Center Orbit Determination Europe;moreover,the model with multi-channel inputs would perform well during geomagnetically quiet and stormy periods.
作者
陈鑫鑫
李淑慧
陈栋
胡翔宇
CHEN Xinxin;LI Shuhui;CHEN Dong;HU Xiangyu(School of Land Science and Technology,China University of Geosciences(Beijing),Beijing 100083,China)
出处
《导航定位学报》
CSCD
北大核心
2024年第5期125-131,共7页
Journal of Navigation and Positioning
基金
中国地质大学(北京)大学生创新创业训练计划项目(202311415064)。
关键词
电离层总电子含量
多通道
卷积网络
长短期记忆神经网络
预测
ionospheric total electron content
multichannel
convolutional network
long short-term memory neural network
prediction
