高轨卫星轨道预报中神经网络模型优化设计

高轨卫星是我国卫星导航系统的重要组成部分。提升该类卫星的轨道预报精度有利于用户定位精度的提高。提出了一种改进高轨卫星轨道预报精度的新方法。该方法避开了精化动力学模型的困难,尝试从轨道预报误差的规律中寻找突破。利用神经网络作为建立预报模型的工具,将某历史时刻的轨道预报误差作为训练样本,利用训练好的神经网络模型补偿当前时刻的预报轨道以提高轨道预报精度。对影响神经网络模型补偿效果的各因素进行了详细分析,制定了适应于高轨卫星短期、中期和长期预报的神经网络最优模型。利用实测数据进行了试验分析,结果表明:预报8,15及30 d应选择的训练步长分别为10,20及25 min;轨道预报8~30 d时,训练噪声均选取0.01。神经网络模型有效地改进了高轨卫星的轨道预报精度,预报4~30 d,轨道精度提高幅度为34.67%~82.37%不等。

Optimal Design of ANN Model in Orbit Prediction for High-Altitude Orbit Satellite

High-altitude obit satellite is an important component in satellite navigation system of our country, high-precision predicted orbit is good for the positioning of users. A new method of satellite orbit prediction based on Ar-tificial Neural Network ( ANN) model was proposed in order to improve the precision of orbit prediction. Avoiding the difficulty of amending the dynamical model, it is attemped to make use of ANN as the tool of constituting orbit prediction model, search the rule of orbit prediction, then utilize ANN model to compensate the orbit prediction er-ror. Several complications influencing on the compensation effection were analyzed particularly. The short-term, middle-term and long-term forecast project were established. The experiment results showed that, the training step should be 10, 20, 25 minutes when the predictive arc respectively is 8, 15, 30 day; the noise should be 0. 01 when the predictive arc is among 8~30 days. The orbit prediction error based on ANN model are less than those based on dynamical model, the rates of improvement of predicting 4~30 days are 34.67%~82.37%.