基于历史数据的测控系统指标预测方法探索

Research on Index Forecasting Methods for Space TT&C System Based on Historical Data

目前对航天测控系统进行故障预测与健康管理成为提高系统安全和可靠性的新发展方向。但已有的方式,如监控系统、自动化测试等,只能对系统进行实时监测或者事后分析。如何让测控系统在实时监测自身状态的基础上,能提前预估系统未来指标状态,从而更好地预防系统可能发生的故障,成为一条新的测控系统发展思路。在经典测控系统的故障识别和预测技术基础之上,针对测控系统关键指标预测的问题,探索基于历史数据的指标预测方法,具体分析基于SVR的单步和多步指标预测方法,并使用实测数据进行试验和分析。与Ridge、Lasso等传统线性回归方法相比,使用SVR进行指标预测更为准确。

At present Prognostic and Health Management(PHM)becomes the new development direction to improve the safety and reliability of space TT&C system.But the existing means,such as monitoring system and automated testing,can only monitor the system timely.How to predict the future state of the TT&C system based on the real-time monitoring,so as to prevent the occurrence of system failure,becomes a new idea for TT&C system development.On the basis of the classical fault identification and prediction technology of TT&C system,this paper studies the index forecasting methods based on the historical data for the key index prediction of TT&C system.The single step and multi-step prediction methods based on SVR are analyzed in detail and they are verified by the measurement data of TT&C system.Compared with the current Ridge,Lasso and other traditional linear regression methods,the prediction methods based on SVR are more accurate for the index forecasting of the TT&C system.