一种基于灰色神经网络的卫星钟差预报策略

A clock offset prediction strategy for satellites based on grey neural network

针对钟差预报中灰色神经网络模型种类较多、性质和适用范围尚未具体分析的问题,根据其预报特点,该文提出了一种基于灰色神经网络的自适应钟差预报策略。基于MGEX精密钟差数据进行预报实验,采用不同建模钟差数据量进行相同时间段钟差预报,对3种不同的灰色神经网络模型钟差预报效果进行对比,总结了几种预报模型的性质与适用范围。该文提出的自适应预报策略较好地平衡了几种灰色神经网络模型的特点,提升了钟差预报效果。基于该文策略的实验结果表明:所提策略能够有效利用不同灰色神经网络模型特点,提高钟差预报精度。在1d预报中,该策略较本文提及的其他可靠方法精度提升1%~3%;6h预报中,该策略较较灰色模型等精度提高0.02~0.09ns。

In order to solve the problem that there are many kinds of grey neural network models,and their properties and application scope have not been analyzed concretely in the clock offset prediction,an adaptive clock offset prediction strategy based on grey neural network was proposed according to the prediction characteristics of different grey neural network models.Based on the precise clock offset data of MGEX(the multi-GNSS experiment and pilot project),prediction was carried out in the same time period by using different size of clock offset data.The clock offset prediction results of three different grey neural network models were compared,and the properties and application scopes of these models were summarized.The adaptive prediction strategy proposed in this paper balanced the characteristics of several grey neural network models and improved the prediction effect of clock offset.The experimental results based on the proposed strategy showed that the strategy could effectively utilize the characteristics of different grey neural network models and improved the accuracy of clock offset prediction.Compared with other reliable methods mentioned in this paper,the precision of this strategy was improved by 1%~3%in 1day forecast,and 0.02~0.09ns in 6hours forecast,which was better than that of grey model.