| 一种LSTM模型预测BC值的空间碎片无控再入预报方法 |
| |
| 引用本文： | 蒯家伟,赵柯昕,孙立刚,廖名传.一种LSTM模型预测BC值的空间碎片无控再入预报方法[J].宇航学报,2022,43(12):1731-1738. |
| |
| 作者姓名： | 蒯家伟 赵柯昕 孙立刚 廖名传 |
| |
| 作者单位： | 1. 南京航空航天大学航天学院，南京 211106； 2. 北京电子工程总体研究所，北京 100854
3. 北京遥测技术研究所，北京 100094； 4. 中国科学院微小卫星创新研究院，上海 201304 |
| |
| 基金项目： | 国家自然科学基金面上项目（12073045）；中国科协第七届青年人才托举工程（2021QNRC001）；南京航空航天大学基本科研业务费项目（NP2022449）；上海航天科技创新基金（SAST2021 075）；空间微波技术重点实验室开放项目（HTKJ2022KL504017） |
| |
| 摘 要： | 提出一种利用长短周期记忆(LSTM)神经网络模型动态预测无控再入过程中弹道系数(BC)值实现空间碎片高精度再入时刻预报。通过利用空间碎片两行根数(TLE)、简化通用摄动模型(SGP4)与公开的物体陨落时间作为实测数据样本，利用迭代修正BC值方法构建预测模型的训练集，由此构造用于预测BC值的LSTM模型预测BC，再采用高精度轨道外推动力学模型配合预测BC值预报再入时刻，结果表明基于LSTM模型预测BC的空间碎片再入时刻预报方法是可行的，在95%的置信度内，90天以上的再入时刻预报精度小于10%，30天预报精度小于8%。
|
| 关 键 词： | 空间碎片 弹道系数(BC) 长短期记忆(LSTM)神经网络 两行根数(TLE) 再入预报 |
| 收稿时间： | 2022-09-23 |
A Method of Space Debris Re entry Time Prediction Using LSTM Neural Network Based Ballistic Coefficient Pre estimation |
| |
| KUAI Jiawei,ZHAO Kexin,SUN Ligang,LIAO Mingchuan.A Method of Space Debris Re entry Time Prediction Using LSTM Neural Network Based Ballistic Coefficient Pre estimation[J].Journal of Astronautics,2022,43(12):1731-1738. |
| |
| Authors: | KUAI Jiawei ZHAO Kexin SUN Ligang LIAO Mingchuan |
| |
| Affiliation: | 1.College of Astronautics, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China;
2. Beijing Institute of Electronic System Engineering, Beijing 100854, China;
3. Beijing Research Institute of Telemetry, Beijing 100094, China;
4. Innovation Academy for Microsatellites of Chinese Academy of Sciences, Shanghai 201304, China |
| |
| Abstract: | A high precision re entry time prediction method for space debris is proposed by using the long short term memory (LSTM) neural network model to dynamically predict the ballistic coefficient (BC) during uncontrolled re entry. The two line element (TLE) number of space debris, the simplified general perturbations 4 (SGP4) and the published object’s falling time are used as the actual measured data samples, and the training set of the prediction model is constructed based on the iteratively corrected BC values, then the LSTM model used to predict the BC values is constructed to predict the BC. The high precision orbit extrapolation dynamic model is used to predict the re entry time with the BC values. The results show that the space debris re entry time prediction method based on the BC values predicted by LSTM model is feasible. Within 95% confidence, the prediction deviation for more than 90 days is less than 10%, and the prediction accuracy for 30 days is less than 8%. |
| |
| Keywords: | Space debris Ballistic coefficient (BC) Long short term memory (LSTM) neural network Two line element (TLE) Re entry time prediction |
|
| 点击此处可从《宇航学报》浏览原始摘要信息 |
|
点击此处可从《宇航学报》下载全文 |
