Anomalies frequently occur during the operation of spacecraft in orbit, and studying anomaly detection methods is crucial to ensure the normal operation of spacecraft. Due to the complexity of spacecraft structures, telemetry data possess characteristics such as high dimensionality, complexity, and large scale. Existing methods frequently ignore or fail to explicitly extract the correlation between variables, and due to the lack of prior knowledge, it is difficult to obtain the initial relationship of variables. To address these issues, this paper proposes a new method, namely spatio-temporal graph learning reconstruction (STGLR), for spacecraft anomaly detection. STGLR employs a dynamic graph learning module to infer the initial relationships among telemetry variables. It then constructs a spatio-temporal feature extraction module to capture complex spatio-temporal dependencies among variables, leveraging a graph sample and aggregation network to learn embedded features and incorporating an attention mechanism to adaptively select salient features. Finally, a reconstruction module is used to learn the latent representations of features, capturing the normal patterns in telemetry data and achieving anomaly detection. To validate the effectiveness of the proposed method, experiments were conducted on two public spacecraft datasets, and the results demonstrate that the performance of the STGLR method surpasses existing anomaly detection methods, with an average F1 score exceeding 0.97.

The rotational mechanism, which plays a critical role in energy supply, payload antenna pointing, and attitude stabilization in satellites is essential for the overall functionality and performance stability of the satellite. This paper takes the space turntable of a specific satellite model as an example, utilizing high-frequency and high-dimensional telemetry data. An improved informer model is used to predict and diagnose features related to the turntable’s operational health, including temperature, rotational speed, and current. In this paper, we present a forecasting method for turntable temperature data using a hybrid model that combines singular spectrum analysis with an enhanced informer model (SSA-Informer), comparing the results with threshold limits to determine if faults occur in the satellite’s rotational mechanism. First, during telemetry data processing, singular spectrum analysis (SSA) is proposed to retain the long-term and oscillatory trends in the original data while filtering out noise from interference. Next, the improved informer model predicts the turntable temperature based on the mapping relationship between the turntable subsystem’s motor current and temperature, with multiple experiments conducted to obtain optimal parameters. Finally, temperature thresholds generated from the prediction results are used to forecast faults in the rotational mechanism over different time periods. The proposed method is compared with current popular time-series prediction models. The experimental results show that the model achieves high prediction accuracy, with reductions of at least 10% in both the MAE and MSE than CNN-LSTM, DA-RNN, TCN-SE and informer, demonstrating the outstanding advantages of the SSA and improved informer-based method in predicting temperature faults in satellite rotational mechanisms.