Research On Satellite Telemetry Data Prediction Based On ARIMA-SVR Combination Model

With the continuous development of the technological age,the lifetime of satellites in orbit is getting longer and longer.The state of the satellite is affected by the space environment and the duration of the operation.These effects may cause the satellite to malfunction during operation,resulting in irreparable accident losses.The telemetry data generated during the operation of the satellite can directly express the state of the satellite.Forecasting the trend of changes in satellite telemetry data can predict the possibility of failure in advance,leaving more time for satellite operation and control management personnel to deal with the failure.In this paper,a satellite telemetry data prediction model is established to predict the change trend of the actual satellite telemetry data,improve the prediction accuracy,and enable the staff to more accurately analyze and interpret the status of satellites in orbit.The characteristics of satellite telemetry data are more complex.In order to improve the prediction accuracy,this paper conducts prediction research from the combined model level.In view of the periodicity and nonlinearity of satellite time series telemetry data,combined with the advantages of Autoregressive Integrated Moving Average model(ARIMA)for processing periodic samples and support vector regression model(SVR)for processing nonlinear samples,this article uses the ARIMA-SVR combined model to predict the true satellite telemetry data.This paper takes the actual on-orbit operation data of a certain type of space science satellite as the research object,and uses the ARIMA model,the SVR model,Long Short-Term Memory model(LSTM)and the ARIMA-SVR combined model to make short-term and medium-term predictions.In order to ensure the reliability of the selection of the parameters of the kernel function of the SVR model,this paper uses the Particle Swarm Optimization(PSO)algorithm to optimize the parameters of the kernel function of the SVR model.In the last of the experiment,this paper uses R-Square(R~2),Mean Square Error(MSE),Root Mean Square Error(RMSE)and Mean Absolute Relative Error(MAPE)evaluation criteria to evaluate the prediction effects of four different models.The experimental results show that the prediction accuracy of the ARIMA-SVR combination model is higher than that of the ARIMA,SVR and LSTM prediction models whether it is short-term or medium-term prediction.Comparing the results of the short-term forecast and the mid-term forecast,it is found that the accuracy of the ARIMA-SVR combined model in the short-term forecast is higher than the medium-term forecast,indicating that the ARIMA-SVR combined model is more suitable for the short-term forecast of satellite telemetry data.In this paper,the actual satellite telemetry data is used as the research object,combined with the prediction results obtained from the experiment,it proves that ARIMA-SVR model can effectively improve the precision of satellite telemetry data,shows that the model has practical value in predicting the telemetry data,in the future it can provide effective decision analysis support for satellite anomaly detection.