Anomaly Detection Method Based on Granger Causality Modeling

Satellites are very expensive to manufacture and require high reliability. Monitoring a large amount of telemetry data during the satellite orbit operation, the telemetry data are an important data source for analyzing the internal correlation of the satellite system and detecting anomalies. Telemetry data is in the form of time series, and there may be mutual influence and correlation between these time series. Due to the diversity of its influence and association forms, it is necessary to establish an effective model to determine the association relationship between them in order to detect anomalies on this basis and identify the cause of anomalies. In this paper, we use Granger causality model to analyze correlation between time series of telemetry data and establish a causality model. Detecting anomalies according to the causality which under normal circumstances and find out the cause of the anomalies. The case study shows that our method is effective.