基于神经网络模型的地球同步卫星高精度轨道预报

发布时间：2018-04-21 00:43

本文选题：GEO + IGSO　；参考：《空间科学学报》2016年01期

【摘要】：利用动力学模型得到的预报轨道精度随时间推移衰减较快,针对这一问题,提出一种改进地球同步卫星轨道预报精度的新方法.将神经网络作为工具,结合轨道动力学特性建立一个训练样本集,根据当前时刻预报误差特征在样本集中搜索最佳训练样本,利用训练得到的神经网络模型补偿和改进当前时刻的预报轨道,达到提高预报精度的目的.基于实测数据的试验分析表明,不同卫星在不同初始时刻下的改进效果是不同的.预报4 d的轨道精度由43m提高至15m,预报8d的轨道精度由183m提高至80m.基于神经网络模型预报4d和8d的改进成功率分别为78.33%和88.33%.
[Abstract]:The prediction orbit accuracy obtained from the dynamic model decreases rapidly with time. To solve this problem, a new method to improve the prediction accuracy of geostationary satellite orbit is proposed. The neural network is used as a tool to establish a training sample set in combination with orbital dynamic characteristics. According to the prediction error feature of the current time, the best training sample is searched in the sample set. The training neural network model is used to compensate and improve the prediction track at the present time to improve the prediction accuracy. The experimental analysis based on the measured data shows that the improvement effect of different satellites at different initial times is different. The orbit precision of 4 days is improved from 43m to 15m, and that of 8 days is improved from 183m to 80m. The improved success rates of 4 d and 8 d prediction based on neural network model were 78.33% and 88.33%, respectively.
【作者单位】：北京市遥感信息研究所;北京卫星导航中心;
【基金】：上海市空间导航与定位技术重点实验室基金项目资助(12DZ2273300)
【分类号】：V412.41

【相似文献】