基于神经网络的卫星钟差预报研究

Research of satellites clock error prediction based on neural networks

针对卫星钟差的数学预测补偿问题,研究基于径向基函数神经网络的卫星钟差预报方法。以国际全球定位系统服务机构数据中心网站公布的卫星钟差数据作为输入样本,利用径向基函数神经网络进行学习训练,进行卫星钟差预报,将结果与线性插值法和多项式插值法以及数据中心发布的精密卫星钟差数据做比较,分析卫星钟差的预报精度。仿真结果表明,RBF神经网络逼近非线性函数的能力强、收敛速度快,卫星钟差预报精度高、残差波动小、拟合平滑度好。

Aim at the satellite clock differs of mathematics predict an in expiation of problem and studied to differ a forecast method to the satellite clock of the radicle function nerve network according to the path, position system service with the international world organization data website in the center announcement of satellite clock bad the data is to input sample and make use of the path carries on study training to the radicle function nerve network and carries on a satellite clock and differs forecast, and put result and line to be worth method and polynomial and put and be worth a method and do than, with compare the data center release of precise satellite clock bad the data did than, analyze the satellite clock differs of forecast accuracy, imitate and really prove as a result, RBF nerve network＇s approaching line ability of the function doesn＇t is strong, refrain from rash action speed quickly, the satellite clock differs and forecasts accuracy Gao, cripple bad undulate small, draw up and match a smooth degree good.