Optical Experiments Prediction of the Quantum Science Experiment Satellite Based on Gradient Boosting Decision Tree
-
摘要: 量子科学实验卫星在轨运行期间完成4种光学实验,地面监测人员通过遥测参数阈值判断卫星是否进行光学实验、实验类型及实验结果.这种方法需要预先设定大量阈值,并且这些阈值需要根据在轨卫星重新设定,可扩展性较差.针对以上问题,提出一种基于机器学习的光学实验判别方法,将量子科学实验卫星的光学实验监测任务抽象为机器学习中的多元分类问题,构建分类模型,利用量子科学实验卫星的真实历史遥测数据对模型进行训练,并通过真实实验计划对训练得到的模型进行验证.实验结果表明,本文提出的方法在没有专家先验知识的前提下,判别准确率达到99%,可用于量子科学实验卫星光学实验的实时监测任务.提出的基于机器学习的判别方法具有较强的可扩展性,可应用于卫星在轨运行的其他监测任务.Abstract: The quantum science experimental satellite mainly carry out four kinds of optical experiments during the orbital operation. The ground monitoring personnel mainly judged whether the satellite carried out optical experiments, experimental types and experimental results through the telemetry parameter threshold. This method requires a large number of thresholds to be set in advance, which requires a lot of manpower, and these thresholds need to be reset according to the on-orbit satellite, and the scalability is poor. Aiming at the above problems, this paper proposes an optical experiment discriminating method based on machine learning. Firstly, the optical experiment monitoring task of quantum science experimental satellite is abstracted into a multi-classification problem in machine learning. A classification model is constructed, and then the quantum science experimental satellite is used. The real historical telemetry data is used to train the model, and finally the trained model is verified by the real experimental plan. The experimental results show that the proposed method can achieve 99% accurate accuracy without the expert prior knowledge, and can be used for real-time monitoring tasks of quantum science experimental satellite optical experiments. The machine learning-based discriminant method proposed in this paper has strong scalability and can be widely extended to other monitoring tasks of satellite orbit operation.
-
Key words:
- Optical experiment /
- Telemetry data /
- Machine learning /
- Thresholds value
-
[1] WANG Fan. Research on Space Payload Fault Diagnosis System Based on Fault Tree[D]. Beijing:Graduate School of Chinese Academy of Sciences, 2007(王帆. 基于故障树的空间有效载荷故障诊断系统研究[D]. 北京:中国科学院研究生院, 2007) [2] JI Changwei, HUANG Wenhu, RONG Jili. Fault diagnosis technology for manned spacecraft in foreign countries[J]. Chin. Space Sci. Technol. 1995(1):23-29, 52(纪常伟, 黄文虎, 荣吉利. 国外载人航天器故障诊断技术[J]. 中国空间科学技术, 1995(1):23-29, 52) [3] GEIB P L, COX D D, TOMASI A M. Payload state of health monitoring design for next generation satellite constellations[C]//IEEE Xplore. Conference:Aerospace Conference Proceedings, 2002, 1:5, DOI:10.1109/AERO. 2002.1036841 [4] GEORGES H M, WANG D, XIAO Z. GNSS/Low-Cost MEMS-INS integration using variational Bayesian adaptive cubature Kalman smoother and ensemble regularized ELM[C]//Mathematical Problems in Engineering, 2015:1-13 [5] IVERSON D L. Inductive system health monitoring with statistical metrics[C]//International Conference on Artificial Intelligence, 2004:605-611 [6] JEROME H F. Greedy function approximation:a gradient boosting machine[J]. Ann. Statis., 2001, 29(5):1189-1232
点击查看大图
计量
- 文章访问数: 672
- HTML全文浏览量: 59
- PDF下载量: 52
- 被引次数: 0