中文摘要：

为获得高精度实时GPS卫星钟差，文章提出一种基于多项式和最小二乘支持向量机（Least Squares Support Vector Machines，LS—SVM）相结合的钟差预报方法。该方法采用国际GNSS服务发布的超快速观测星历建模进行短期预报，首先根据卫星钟的物理特性用附有周期项的多项式模型进行拟合以提取趋势项和周期项，然后用LS—SVM对多项式拟合残差进行建模预报，最后将预报结果加上趋势项和周期项，得到最终的钟差预报值。试验结果表明，所提算法能够实时有效地对GPS卫星钟差进行预报，且精度优于超快速预报星历。

英文摘要：

In order to get real-time GPS clock products of high accuracy, a new prediction algorithm was proposed, which incorporated a polynomial model with a few cyclic terms and least squares support vector machines （LS-SVM） for clock offset. Firstly, a polynomial model adding a few cyclic terms was utilized to fit the clock offset series according to the physical characteristics of satellite clocks so as to extract the trend term and periodic term. Then, the polynomial fitting residuals were modeled based on the LS-SVM. Finally, the trend term and periodic term were added to the forecasted residual result to produce the final prediction value for the clock offset. The IGS ultra-rapid observed （IGU-O） products were employed as the initial values of the prediction strategy. The simulation results demonstrate that the proposed algorithm can be used to predict clock offset with high accuracy in real-time, and outperforms the IGS ultra-rapid predicted （IGU-P） solutions at least on a daily basis.