基于GM-ARIMA模型的遥测数据中长期预测

Long-Term Prediction Method of Telemetry Data Based on GM-ARIMA Model

为了实现对在轨卫星遥测数据的高精度预测,提出了一种基于时间序列分解的遥测数据预测方法,即应用HP滤波(Hodrick-Prescott Filter)将遥测数据的时间序列分解成趋势项和波动项,并根据各项特点分别使用灰色GM(1,1)模型和季节型ARIMA模型进行预测,然后叠加趋势项和波动项各自的预测值,得到最终预测结果。此方法可以有效降低趋势性、波动性相互影响产生的误差,提高预测精度。通过对某卫星阳压在轨数据的实证分析,验证了上述预测方法的有效性,在半年的预测期内,达到了很高的预测精度。所提方法在卫星健康评估、故障诊断和预警等方面具有重要应用价值。

In order to achieve high-precision prediction of satellite telemetry data in orbit,a telemetry data pre-diction method based on time series decomposition is proposed in the paper.The time series of telemetry data was de-composed into trend term and fluctuation term by using HP filter.According to various characteristics,the grey GM(1,1)model and seasonal ARIMA model were used for prediction;Then,the respective predicted values of trend term and fluctuation term were superimposed to obtain the final prediction result.This method can effectively reduce the error caused by the interaction of trend and fluctuation,and improve the prediction accuracy.Through the empiri-cal analysis of the Anode Voltagetelemetry data of a satellite,the effectiveness of the prediction method was verified,and the prediction accuracy was very high in the prediction period of half a year.This method has important applica-tion value in satellite health assessment,fault diagnosis and early warning.