Research On The Prediction Method Of Flywheel Temperature Of Low-orbit Satellite

With the rapid development of China’s aerospace industry,the traditional real-time monitoring system based on threshold in satellite operation management can not meet the needs of large-scale satellite monitoring gradually,especially can’t realize the prediction of potential anomalies and faults of satellite system in advance,which directly affects the efficiency of fault handling of satellite system.In this paper,the prediction method of flywheel temperature of satellite in orbit is proposed,which can win precious crisis-response time for the ground monitoring personnel and better manage the satellite operation in orbit.Flywheel temperature,as an important parameter reflecting the operation status of control system of satellite attitude,is one of the safety indexes of long-term monitoring of ground monitoring system.The failure of flywheel not only causes the satellite mission to be unable to carry out normally,but also may interrupt the ongoing satellite business and even cause security risks.At present,the main monitoring method of flywheel temperature is based on the real-time monitoring of the threshold value.This method starts to deal with the fault after it occurs.It can’t find the potential abnormality in advance and prepare the disposal plan in time to implement the fast recovery.In order to solve the above problems,based on the telemetry data of low orbit satellites and the data of outer space environment of satellites,this paper studies the prediction method of flywheel temperature of satellites.By using the machine learning algorithm and time series algorithm,this paper establishes the flywheel temperature prediction model by combining Gradient Boosting Decision Tree(GBDT)and Autoregressive Integrated Moving Average(ARIMA),and studies the flywheel temperature prediction of different duration in the future.Based on this prediction model,an auxiliary monitoring and early warning system of satellite flywheel can be established,which can help to discover the potential fault of satellite flywheel in advance and help the uninterrupted operation of satellite in orbit.The main work of this paper are as follows:(1)A multi-resolution combination algorithm of satellite telemetry data is designed.Based on the diversity of time resolution of satellite telemetry data and space environment data,a targeted multi-resolution combination algorithm of telemetry data and process are proposed.This method can not only ensure the integrity of satellite telemetry data,but also complete the combination of multiple-resolution of time of telemetry data,which lays the foundation for the comprehensive analysis of satellite telemetry data in the future.(2)The modeling idea and prediction method of flywheel temperature prediction of satellite are proposed.This paper formalizes the problem of flywheel temperature prediction of satellite,and analyzes the modeling ideas of flywheel temperature prediction model by time series prediction and regression analysis prediction.The prediction method and process for flywheel temperature of satellite are proposed.This prediction process has certain universality for all types of satellite flywheel systems.(3)Studying and selecting the prediction algorithm which are suitable for flywheel temperature prediction.The main regression prediction algorithm and time series prediction algorithm are summarized.The advantages and disadvantages of each algorithm and its application range are compared and analyzed,and finally based on the demand of satellite flywheel temperature prediction,GBDT algorithm and ARIMA algorithm are selected as the best algorithm for flywheel temperature prediction.The selection process and results of the algorithm in this paper can provide a reference for future research.(4)Through algorithm training and model fusion,the prediction model of satellite flywheel temperature is constructed.Based on the decision tree algorithm and time series algorithm,trained with the satellite telemetry data and space environment data,the GBDT prediction model and ARIMA prediction model are established respectively.The model is adapt to different prediction scene through model combination to improve the prediction accuracy,which successfully achieves the prediction of flywheel temperature in different time intervals,and the real-time monitoring of flywheel temperature is upgraded to prediction and early warning.The experimental results show that the RMSE in the study of flywheel temperature prediction of satellite is 0.5495,which can meet the needs of prediction and early warning of ground monitoring system.The research results can reduce the monitoring pressure of the ground personnel,find out the potential anomalies of the satellite flywheel system in advance,and provide reference for the prediction of other telemetry parameters of the satellite.The main innovations of this paper are as follows:(1)The combining-prediction model based on GBDT and ARIMA algorithm is proposed and applied to the prediction of satellite flywheel temperature,which improves the prediction accuracy and adaptability.And the flywheel temperature monitoring is upgraded to prediction and early warning,which reduces the working pressure of ground monitoring personnel,strives for precious response-time for system recovery,and avoids potential safety risks in advance.(2)The combining method of multi-resolution telemetry data of satellite and space environment is proposed.While ensuring the integrity of data information,the data of various time and frequency are normalized,which lays the foundation for model training and prediction.(3)The main influencing factors of flywheel temperature are obtained from the data correlation analysis,which provides ideas for the subsequent research and fault location of flywheel temperature.