Anomaly Detection of Satellite Telemetry in Orbit Based on Sequence and Point Feature Combination

Aiming at the detection problems of low accuracy, high false alarm rate in the detection of satellite telemetry timing data, a satellite telemetry anomaly detection model is presented based on combination of sequence and point features. The model firstly makes steady separation in telemetry data to obtain data trends and steady residual; as the data trends contain most of telemetry sequence information, the non-stationary telemetry sequence feature is reconstructed on each separation layer by performing inverse data trends transform on these coefficients. And stable residual is mainly composed of telemetry random point, the telemetry point feature is reconstructed on the final separation layer by performing inverse steady residual transform on these coefficients. Finally, the integrated anomaly detection model, that is double autoregressive combined model, is constructed by fuse the two features. The results of anomaly detection experiment in the telemetry data of a certain type of satellite power system show that the model can adapt to the complex changes of satellite telemetry data in orbit, reduce decision interference from noise and outliers effectively, reduce the false alarm rate, improve detection precision, achieve effective detection in the anomaly threshold of satellite telemetry data on orbit.