基于RBF神经网络的改进模型在电离层TEC预报中的应用  

Application of Improved Model Based on RBF Neural Network in Ionospheric TEC Prediction

作　　者：胡文权 HU Wenquan(Ningbo Alatu Digital Technology Co.,Ltd.,Ningbo 315042,China)

机构地区：[1]宁波市阿拉图数字科技有限公司,浙江宁波315042

出　　处：《测绘与空间地理信息》2023年第8期164-167,共4页Geomatics & Spatial Information Technology

摘　　要：为了提高电离层TEC值的预报精度,建立更高精度的电离层TEC预报模型,本文在RBF神经网络模型的基础上引入奇异谱分析(Singular Spectrum Analysis,SSA)方法,构建新的电离层TEC预报模型。该组合模型首先通过SSA提取原始序列中的特征分量,避免噪声分量对预报结果的影响,其次将去噪后特征分量作为RBF神经网络模型的输入值。使用IGS中心提供的TEC数据序列进行模型验证,结果表明,无论是对平静期电离层TEC预报还是磁暴期电离层TEC预报,相比于单一的RBF神经网络模型预报结果,本文提出的SSA-RBF神经网络模型的预报结果均更优,其中平静期预报残差在2 TECU以内,磁暴期预报残差在3—4 TECU以内,验证了本文提出组合模型的优越性。In order to improve the prediction accuracy of ionospheric TEC values and establish a more accurate ionospheric TEC prediction model,this paper introduces Singular Spectrum Analysis(SSA) method on the basis of RBF neural network model to construct a new ionospheric TEC prediction model.Firstly,the combined model extracts the characteristic components in the original sequence through SSA to avoid the influence of noise components on the prediction results.Secondly,the denoised characteristic components are used as the input value of RBF neural network model.The TEC data series provided by IGS center are used for model verification.The results show that the prediction results of SSA-RBF neural network model proposed in this paper are better than those of single RBF neural network model,whether for ionospheric TEC prediction in calm period or ionospheric TEC prediction in magnetic storm period.The residual prediction during the calm period is within 2 TECU,and the residual prediction during the magnetic storm period is within 3—4 TECU,which verifies the superiority of the combined model proposed in this paper.

关 键 词：奇异谱分析 RBF神经网络模型 电离层 平静期 磁暴期 

分 类 号：P228[天文地球—大地测量学与测量工程]