摘要
卫星遥测大数据是卫星地面站判断其运行状态的唯一依据,遥测参数的有效判读对监测在轨卫星健康状态具有重要意义。而遥测数据维度高、数据量大、专业性强的特点为高精度、低误检率的多维遥测数据判读带来严峻挑战。因此,提出了一种基于数据驱动的卫星遥测大数据智能判读方法。该方法以极限学习机(ELM)预测模型为基础,对目标参数进行高精度的单步预测,同时,基于目标参数在时间维度上的变化趋势对预测结果进行修正。最后,基于集成学习的方法针对目标参数的不同类别分别给出判读策略。利用卫星电源子系统仿真数据和真实卫星遥测数据对所提出的参数判读方法的有效性进行验证,实验表明,该方法对不同类型的监测数据具有较强的自适应能力和鲁棒性。
Satellite telemetric big data is the only basis for the satellite ground station to determine the operating state, the effective interpretation of telemetric parameters is of great significance for monitoring the satellite healthy state on orbit. However, telemetric data has the characteristics of high dimension, large data volume and strong professionalism, which bring severe challenges for the interpretation of the multidimensional telemetric data with high accuracy and low false detection rate. Therefore, this paper proposes a data-driven based intelligent interpretation method for satellite telemetric big data. Based on the extreme learning machine(ELM) prediction model, the proposed method performs high accuracy one-step-ahead prediction of the target parameters. At the same time, the prediction results are corrected based on the changing trend of the target parameters in time dimension. Finally, the interpretation strategies are given for different categories of target parameters based on ensemble learning method. Satellite power subsystem simulation data and actual satellite telemetric data were used to verify the effectiveness of the proposed parameter interpretation method. The experiment result shows that the method has strong adaptive ability and robustness for different types of monitoring data.
作者
史欣田
庞景月
张新
彭宇
刘大同
Shi Xintian;Pang Jingyue;Zhang Xin;Peng Yu;Liu Datong(Automatic Test and Control Institute,Harbin Institute of Technology,Harbin 150080,China;Shanghai Institute of Satellite Engineering,Shanghai 201109,China)
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2018年第12期81-91,共11页
Chinese Journal of Scientific Instrument
基金
国家自然科学基金(61571160
61771157)项目资助
关键词
卫星
遥测
大数据
判读
极限学习机
集成学习
satellite
telemetry
big data
interpretation
extreme learning machine
ensemble learning
