Robust anomaly identification algorithm for noisy signals: spacecraft solar panels model

Prony method is an efficient feature extraction technique when applied to signals at noiseless environment. However, space environment is very rich with different sources of noise. Noise has a massive effect on Prony method that is noticeably degrades its performance. The objective of this research is to introduce an efficient and robust technique for fault detection and identification at noisy environment to one of the most vital subsystems in spacecraft, which is electrical power subsystem (EPS). Occurrence of anomaly in spacecraft EPS is mainly related to some parameters that are voltage, current and delivered power. This will directly affect the overall spacecraft operation and may result in partial or total mission loss. It is crucial for spacecraft onboard computer to be provided with a diagnosis task that keeps track of EPS parameters and efficiently detect and identify anomalies at different noise levels. The proposed approach is based on noisy signal energy contents. The signal energy contents are large enough compared with noise energy contents. The algorithm uses the short-time energy for noisy signals to robustly identify photovoltaic (PV) anomalies according to their causes like short circuit, open circuit and shading. The proposed approach is proved to be able to discriminate among different types of PV anomalies, and the results ensure the robustness of the proposed algorithm. This is carried out with different noise levels, with minimal task execution time and with small memory footprint. The proposed algorithm is considered to be a generic approach for noisy signal identification.