基于机器学习的NWP ZTD长短期预测模型  

NWP ZTD long-term and short-term prediction model based on machine learning

作　　者：白子仪 徐莹 冯健 于浩 张方照 BAI Ziyi;XU Ying;FENG Jian;YU Hao;ZHANG Fangzhao(College of Geodesy and Geomatics,Shandong University of Science and Technology,Qingdao,Shandong 266590,China;Coalfield Geological Center of Geological Bureau of Xinjiang Uygur Autonomous Region,Urumqi 830009,China;36th Research Institute of China Electronics Technology Group Corporation,Jiaxing,Zhejiang 314000,China)

机构地区：[1]山东科技大学测绘与空间信息学院,山东青岛266590 [2]新疆维吾尔自治区地质局煤田地质中心,乌鲁木齐830009 [3]中国电子科技集团公司第三十六研究所,浙江嘉兴314000

出　　处：《导航定位学报》2024年第4期34-44,共11页Journal of Navigation and Positioning

基　　金：国家自然科学基金项目(42174035)。

摘　　要：对流层延迟是影响全球卫星导航系统(GNSS)定位精度的主要误差源之一,利用数值天气预报(NWP)模型估计天顶对流层延迟(ZTD)是常用的方法之一,但NWP模型预报资料估计的ZTD精度有限;NWP模型再分析资料估计的ZTD不能用于GNSS实时定位,且目前大多数文献未能对ZTD长短期预测分别进行研究。因此,利用欧洲中期天气预报中心(ECMWF)的第五代全球气候再分析资料数据集(ERA5)和国际GNSS服务组织(IGS)的高精度ZTD数据,研究基于反向传播(BP)神经网络、支持向量机和长短期记忆网络3种机器学习算法构建以年为时间窗口的ZTD长期预测模型和以24h为时间窗口的ZTD短期预测模型的可行性。实验结果表明:构建的ZTD长期预测模型和短期预测模型可以有效提高预测ZTD的精度。Tropospheric delay is one of the main error sources that affects the positioning accuracy of global navigation satellite system(GNSS).It is one of the commonly used methods to estimate the zenith tropospheric delay(ZTD)by using the numerical weather prediction(NWP)model.However,the accuracy of ZTD estimated by NWP model forecast data is limited.The ZTD estimated by the NWP model reanalysis data cannot be used for GNSS real-time positioning,and most of the current literatures fail to study the long-term and short-term prediction of ZTD separately.In this regard,using the fifth-generation global climate reanalysis data set(ERA5)of the European center for medium-range weather forecasts(ECMWF)and the high-precision ZTD data of the international GNSS service(IGS),the feasibility of constructing ZTD long-term prediction model with year as time window and ZTD short-term prediction model with 24h time window based on three machine learning algorithms of back propagation(BP)neural network,support vector machine and long-term and short-term memory network was studied.The experimental results showed that the constructed ZTD long-term prediction model and short-term prediction model could effectively improve the accuracy of ZTD prediction.

关 键 词：全球卫星导航系统(GNSS) 天顶对流层延迟(ZTD) 数值天气预报(NWP) 机器学习算法 预测模型 

分 类 号：P228[天文地球—大地测量学与测量工程]