A hybrid statistical-machine learning framework for evaluating geomagnetic storm effects on MisrSat2 satellite power...

Area of Science:

• Space weather impacts on satellite power systems
• Application of statistical and machine learning methods in aerospace engineering

Background:

• Geomagnetic storms pose risks to satellite operations.
• MisrSat-2 satellite's power subsystem requires continuous monitoring.
• Previous studies often lack integrated analysis of statistical and ML approaches.

Purpose of the Study:

• To evaluate the impact of the May 2024 geomagnetic storm on the MisrSat-2 satellite's power subsystem.
• To develop and validate a hybrid statistical-machine learning framework for anomaly detection.
• To ensure mission assurance through robust diagnostics.

Main Methods:

• Integrated a multi-tiered statistical approach (CUSUM, z-score, event-based analysis) with machine learning (Mixture of Experts).
• Employed Welch's t-tests, bootstrapping, and Benjamini-Hochberg false discovery rate (BH-FDR) for statistical validation.
• Utilized EQUFLUX for radiation degradation modeling and compared ML model performance (R², MAE).

Main Results:

• Identified modest current deviations in solar arrays (13 events on panel-1, 17 on panel-2) during the storm, within design tolerances (<4%).
• The battery subsystem effectively buffered fluctuations, maintaining bus integrity.
• The Mixture of Experts model achieved high predictive accuracy (R²=0.921, MAE=0.063 A), validating statistical findings.

Conclusions:

• The hybrid framework successfully detected and validated minor power subsystem anomalies caused by the geomagnetic storm.
• The MisrSat-2 power subsystem demonstrated resilience, with no measurable radiation-driven degradation.
• The proposed framework offers a scalable solution for anomaly diagnostics and mission assurance in space environments.