一种组合模型在卫星钟差长期预报中的应用研究

Application of a combined model in satellite clock bias prediction

近年来,卫星钟差长期预报普遍采用灰色模型,该模型的预报精度虽较传统的二次多项式模型有所提高,但仍不理想。本文在其基础上提出了一种新的组合模型:首先利用灰色模型估计的残差建立二次多项式模型,预报以后历元的残差,然后和灰色模型的预报结果相加;并分析了利用不同历元个数的残差建模所得组合模型的精度,将组合模型与灰色模型、二次多项式模型的预报精度进行了比较。结果表明:组合模型相对于灰色模型的预报精度能提高一个数量级左右,验证了本文提出的组合模型的可行性和有效性。

Recently, grey model is usually used to predict satellite clock bias. This model＇ s predict precision is better than quadratic polynomial model, but it is still insufficiency. A new combined model was proposed in this paper： first, quadratic polynomial model was established using the predicted residual of grey model to predict the later epochs＇ residual, then added them on the predicted result of grey model. The amount of epochs to establish model of bias was also discussed. Then it compared the precisions between the combined model, the grey model and the quadratic polynomial. It is shown that the former is one magnitude higher than grey model and the combined model was reliable and valid.