基于NARX回归神经网络的岸基GNSS-IR有效波高反演模型分析  

Study on inversion model of significant wave height from shore-based GNSS-IR by using NARX recurrent neural network

作　　者：张一 周立 ZHANG Yi;ZHOU Li(School of Marine Technology and Geomatics,Jiangsu Ocean University,Lianyungang 222005,China)

机构地区：[1]江苏海洋大学海洋技术与测绘学院,江苏连云港222005

出　　处：《测绘通报》2022年第2期90-94,共5页Bulletin of Surveying and Mapping

基　　金：国家重点研发计划“海洋环境安全保障”重点专项(2018YFC1405702)。

摘　　要：有效波高是海洋动力环境的主要参数,针对现有岸基双天线GNSS-R有效波高反演方法必须使用专用型接收机,以及相关测站布设较少、数据获取困难等限制,本文提出了一种利用普通单天线测量型GNSS接收机的GNSS-IR数据反演有效波高的方法。首先,利用NARX回归神经网络构建潮高反演误差与有效波高大小之间的相关模型,并给出了从数据选取、数据集制作到反演模型构建的相关流程与方法。然后,使用斯坦福大学哈勃肯斯海洋实验站中P231站两年的数据,开展了基于NARX回归神经网络的岸基单天线GNSS-IR有效波高反演模型试验。试验结果表明,该模型适用于0.1~2.5 m(即二~四级海况之间)范围内有效波高的反演应用,反演有效波高的MSE最小为0.01 m。Significant wave height is the main parameter of marine dynamic environment. This paper proposes a method to build the shore-based GNSS-IR significant wave height inversion model using NARX regression neural network. The method aims to explore how to use the data of a common single antenna measurement GNSS receiver to invert the significant wave height, and this paper gives the relevant processes and methods from data selecting, data processing to model construction. At the same time, this paper uses the two-year data from P231 station in the Stanford University Hopkins marine station for shore-based experiments. The results show that this method is suitable for the inversion application of significant wave height in the range of 0.1~2.5 m, and the min mean squared error of the inversion results of significant wave height is 0.01 m.

关 键 词：GNSS-IR 潮高 NARX 海面粗糙度 有效波高 

分 类 号：P228[天文地球—大地测量学与测量工程]