组合模型预报导航卫星钟差

摘要

导航卫星原子钟在运行过程中,不可避免地受到环境和其他因素的影响而产生噪声,导致卫星钟差不规则变化.研究表明钟差是5种噪声共同作用的结果,并且有多种模型可以预报钟差,但它们各具特点和适用性,因此可以考虑组合这些模型预测的结果,以提高预报结果的可靠性.讨论了经典权、最优权、神经网络3种组合方法,并运用IGS精密卫星钟差数据,以二次多项式和灰色模型预报为基础,计算了组合模型的预报结果.研究表明:3种方法都能有效地组合单一模型预报结果,预报精度也大体相当,其中,经典权法最优、最优权法次之、神经网络法稍逊.%Affected by environmental and other factors, the navigation satellite atomic clocks must produce error and bias. The research shows that clock errors and bias result from five type noise and there are several models with then the error and bias can be predicted. Each model has its disadvantages and advantages. In order to improve the reliability of the prediction of clock error and bias, it is useful to combine the forecasting result with each single model. This paper discusses the method for combining polynomial and grey model by use of the ICS precision clock series atomic clocks as an example. Finally, a useful idea from the result is adopted that the model combination is used to prediction the navigation satellite clock error and bias.