摘要
针对卫星钟差具有趋势项和随机项变化的特征问题,提出了GM-RBF组合模型的方法。该模型首先用GM(1,1)提取预处理后的卫星钟差趋势项部分并进行建模预报,得到相应的残差序列,通过RBF神经网络训练用灰色模型预报所获得的残差序列,然后将GM(1,1)模型的钟差后续预报值与RBF神经网络的残差预报值对应相加可得组合模型的预报结果。为验证组合模型的有效性和可行性,将组合模型预报结果与GM(1,1)模型、ARIMA模型、RBF神经网络模型预报结果进行对比实验。实验结果表明:组合模型预报精度要高于其他单一模型,其在不同时段的平均预报精度可提高46.4%~86.2%。
In order to address the eigenproblem of the change of the trend term and random term of the satellite clock error,this paper proposes a method of the GM-RBF combined model.First,this model uses GM(1,1)to extract the pre-processed trend term of the satellite clock error,conducts modeling and forecasting to obtain the corresponding residual sequence,and uses the grey model to predict the obtained residual sequence by RBF neural network training.Then,the prediction result of the combined model can be obtained by adding the subsequent prediction value of the clock error of the GM(1,1)model and the residual prediction value of the RBF neural network.In order to verify the validity and feasibility of the combined model,the prediction results of the combined model are compared with those of the GM(1,1)model,the ARIMA model and the RBF neural network model.Experimental results show that the forecast accuracy of the combined model is higher than that of other single models,and that its average forecast accuracy can be increased by 46.4%~86.2%in different periods.
作者
唐彦
李豫
李特
TANG Yan;LI Yu;LI Te(Xinjiang Institute of Engineering,Urumqi,Xinjiang Uyghur Autonomous Region,830000 China;Yingkou Chuangxue Jiaoyu,Yingkou,Liaoning Province,115000 China;Heilongjiang Institute of Technology,Harbin,Heilongjiang Province,150000 China)
出处
《科技资讯》
2024年第7期27-31,共5页
Science & Technology Information
基金
2021年度黑龙江省省属本科高校基本科研业务费科研项目“电离层动态变化与地震相关性分析研究”(项目编号:2021QJ02)。
关键词
BDS
卫星钟差
灰色模型
RBF
神经网络
组合模型
钟差预报
BDS satellite clock error
Grey model
RBF neural network
Combination model
Clock error forecast
