Anomaly Detection Method for Satellite Telemetry Parameters Based on Time-Series Imputation Generative Adversarial Networks

In order to ensure the safe, reliable and smooth operation of satellites, timely data mining and situation analysis of telemetry parameters are essential for anomaly handling response. Aiming at the problems that existing detection methods cannot effectively model timing telemetry parameters, and the detection results perform poorly in the case of anomaly concentration, this paper proposes a satellite telemetry parameter anomaly detection method based on sequential interpolation generative adversarial network. This method extracts the timing distribution characteristics of telemetry parameters through one-dimensional convolutional neural networks, uses generative adversarial networks to learn the distribution of telemetry parameters, and innovatively adopts interpolation-based detection methods to discriminate anomaly data. Through testing on real satellite telemetry parameter datasets and public time series anomaly datasets, and compared with statistical methods, distance and density clustering methods, and prediction and reconstruction methods, the proposed method exhibits significant performance improvement, verifying the effectiveness of the satellite telemetry parameter anomaly detection method based on time series interpolation generative adversarial network. This research result not only improves the accuracy of anomaly detection, but also enhances the adaptability and robustness of this anomaly detection method to different types of data, providing strong decision support for satellite mission ground operation controllers in satellite situation analysis and anomaly disposal.