基于灰色模型与小波神经网络的卫星钟差预报方法  

A METHOD OF SATELLITE CLOCK BIAS PREDICTION BASED ON GRAY MODEL AND WAVELET NEURAL NETWORK

作　　者：王宇谱[1] 吕志平[1] 陈正生[1] 黄令勇[1] 

机构地区：[1]信息工程大学地理空间信息学院,郑州450052

出　　处：《大地测量与地球动力学》2014年第3期155-159,共5页Journal of Geodesy and Geodynamics

基　　金：国家自然科学基金项目(41274015);国家863计划项目(2013AA122501)

摘　　要：提出一种基于灰色模型和小波神经网络相结合的卫星钟差预报方法。首先通过灰色模型对钟差进行拟合并确定钟差预报的灰色模型,然后根据拟合残差对小波神经网络进行建模,最后将两种模型对应的预报结果结合得到钟差预报值。使用IGS精密钟差进行实验,证明该方法的预报效果优于二次多项式模型和灰色模型,特别是对于ⅡR型铷钟和ⅡF型铷钟,其预报精度可以提高2倍以上。A new prediction method based on the combination of gray system model and wavelet neural network is proposed. Firstly, fitting given satellite clock bias values with gray model, and determining parameters of gray model with the values. Then, establishing model of wavelet neural network with fitting residuals. Finally, combining the prediction results obtained from the two models mentioned above to get prediction results. The results of prediction tests using the precise satellite clock bias data from IGS show that the method can get better satellite clock bias prediction results, compared to the quadratic polynomial model and the grey model. Especially for the rubidium clock of Ⅱ R type and Ⅱ F, the prediction precision can be provided with two times or more.

关 键 词：卫星钟差预报 灰色模型 小波神经网络 惯性制导系统 组合预报模型 

分 类 号：P228.41[天文地球—大地测量学与测量工程]