Real-Time Anomaly Detection in Satellite Communication Systems Using Machine Learning Techniques

Satellite are critical systems that have to survive from space environment effects and failure risks in systems. Satellite delivers to the ground operator an abundance data related to system status telemetry, the telemetry parameters are monitored to indicate satellite performance. Effective detection of anomaly data in massive satellite telemetry is vital process to effective utilization of satellite. Efficient anomaly data detection from massive satellite telemetry is a pivotal process for effective utilization of the satellite. In this paper, a technique in Machine Learning (ML) for detection of anomalies for satellite in real time. The procedure for detection of possible future statistical problems based on the technique of autoregressive integrated moving average analysis involves the application of Prophet, long short-term memory algorithm (LSTM), and auto-encoder algorithms. An automated framework finds itself relying on time-series analysis-based anomalies detection for satellite telemetry data as per the actual observed values with respect to their predicted ones so as to trace out anomalous events.The paper composed of four section. During second section, problem description of satellite attitude control system and telemetry example for our study. In Third second, the description of time series models and the proposed methodology relies on time-series analysis and includes four main stages (preprocessing of raw telemetry data using linear interpolation for missing values, down-casting to optimize storage, transformation and scaling for normalization, and implementing predictive models such as ARIMA, Prophet, and LSTM). The section fourth, the proposed system contains, transformation, Scaling and normalization applies. Experimental results using Prophet model predictions show that it is very efficient, consistent in anomaly detection, and takes very little computational time. Prophet results compared with other algorithms employed display the efficiency and superiority of the proposed model.