基于ConvLSTM的北京区域电离层延迟建模  

Modeling of ionospheric delay in Beijing region based on ConvLSTM

作　　者：谭宗佩 白征东[1] 张强 郭锦萍 段博文 TAN Zongpei;BAI Zhengdong;ZHANG Qiang;GUO Jinping;DUAN Bowen(School of Civil Engineering,Tsinghua University,Beijing 100084,China)

机构地区：[1]清华大学土木工程系,北京100084

出　　处：《测绘工程》2024年第1期25-31,46,共8页Engineering of Surveying and Mapping

摘　　要：电离层延迟是影响卫星定位精度的重要因素,针对GNSS单频定位中电离层延迟改正精度较低的问题,文中利用Bernese5.2软件处理北京13个CORS站从2016-09-15—2016-10-14的GNSS观测数据,得到北京市及周边区域(31°~47°N,108°~124°E)的VTEC值,并基于ConvLSTM神经网络建立北京市及周边区域的电离层延迟模型VclNet,并将该模型与Klobuchar模型和GIM(c1pg)、GIM(c2pg)、三角级数模型、多项式模型的VTEC预报值进行了精度对比分析。6种预报中,VclNet的效果最好,其对北京市中心点VTEC预报值精度为1.99 TECU,区域VTEC预报值精度为2.09 TECU,Klobuchar模型的预报效果最差,中心点精度和区域精度分别为5.92 TECU和5.99 TECU。Ionospheric delay is the main factor that affects the accuracy of satellite positioning.Aiming at the low accuracy of ionospheric delay correction in GNSS single frequency positioning,this paper uses Bernese5.2 to process data of 13 stations in Beijing from September 15,2016 to October 14,2016.According to the GNSS observation data,the VTEC values of Beijing and surrounding areas(31°N~47°N,108°E~124°E)are obtained.Based on the ConvLSTM neural network,the ionospheric delay model VclNet of Beijing and surrounding areas was established,and was compared with the Klobuchar model,GIM(c1pg)and GIM(c2pg),trigonometric series model and polynomial model.Among the six models,VclNet has the best prediction.Its accuracy for the Beijing central point VTEC prediction is 1.99 TECU,and the regional VTEC prediction accuracy is 2.09 TECU.The Klobuchar model has the worst prediction.The central point accuracy and regional accuracy are respectively 5.92 TECU and 5.99 TECU.

关 键 词：全球导航卫星系统 天顶方向的总电子含量 卷积长短期记忆网络 区域模型 精度分析 

分 类 号：P228[天文地球—大地测量学与测量工程]