基于多通道CNN-GRU的低纬度区域电离层预测研究  

Research on Low-latitude Small-area Ionospheric Prediction Based on Multi-channel CNN-GRU

作　　者：张仁中 杨嘉祎 李家乐 陈冠宇 刘佳悦 李豪瑞 申云萧 李旺[1,2] ZHANG Renzhong;YANG Jiayi;LI Jiale;CHEN Guanyu;LIU Jiayue;LI Haorui;SHEN Yunxiao;LI Wang(Faculty of Land Resources Engineering,Kunming University of Science and Technology,Kunming 650093,China;Yunnan Province Key Laboratory of Intelligent Monitoring of Natural Resources and Spatiotemporal Big Data Governance(Under Preparation),Kunming 650093,China)

机构地区：[1]昆明理工大学国土资源工程学院,云南昆明650093 [2]云南省自然资源智能监测与时空大数据治理重点实验室(筹),云南昆明650093

出　　处：《测绘科学技术学报》2025年第1期27-36,共10页Journal of Geomatics Science and Technology

基　　金：国家自然科学基金项目(42204030);云南省“兴滇英才支持计划”项目;云南省基础研究计划项目(202201BE070001-035,202301AU070062);昆明理工大学学生课外学术科技创新基金项目(2024ZK093)。

摘　　要：赤道电离北冠带覆盖中国南部区域,受“喷泉效应”影响该区域电离层动力学特征非常复杂,严重影响卫星导航定位的精度。本文选取云南和四川为研究区域(以下简称川滇区域),基于中国陆态网的48个GNSS观测站数据,采用具有多通道特征的卷积神经网络-门控循环单元算法针对低纬度区域电离层预测展开研究。空间特征上,CNN-GRU能够有效预测多种尺度的电离层空间结构,相关系数优于0.9。但是在南部边界地区,存在明显预测误差。时间特征上,该模型在0~12 h的时间尺度内预测误差小于12~24 h的时间尺度,整体优于1.7 TECu,且在双至日内的预测精度优于双分日。验证结果表明,该TEC模型能够显著提升低纬区域的电离层预测精度,有效服务于高精度导航定位和空间环境监测。The northern crest of the equatorial ionization anomaly covers southern China,where the“fountain effect”creates highly complex ionospheric dynamics that significantly affect the accuracy of satellite navigation and positioning.In this paper,Yunnan and Sichuan are selected as study areas.Based on data from 48 GNSS observation stations of China's Crustal Movement Observation Network,a convolutional neural network-gated recurrent unit(CNN-GRU)algorithm with multi-channel characteristics is applied to study ionospheric prediction in low-latitude regions.In terms of spatial features,the CNN-GRU can effectively predict ionospheric spatial structures at various scales,with a correlation coefficient better than 0.9.However,there is a noticeable prediction error near the southern boundary region.Temporally,the model's prediction error in the 0~12 h time scale is smaller than that in the 12~24 h time scale,with an overall error of less than 1.7 TECu.Moreover,the prediction accuracy during solstices is better than that during equinoxes.The validation results show that this TEC model can significantly improve the ionospheric prediction accuracy in low-latitude regions,providing effective support for high-precision navigation,positioning,and space environment monitoring.

关 键 词：电离层模型 深度学习 总电子含量 川滇区域 多通道 低纬区域 

分 类 号：P228[天文地球—大地测量学与测量工程]