Non-intrusive fault detection in shipboard power systems using wavelet graph neural networks

• DOI: 10.1016/j.meaene.2024.100009
• Publication Date: 2024-06-20T07:47:56Z
• Authors: Soroush Senemmar; Roshni Anna Jacob; Jie Zhang

Naval shipboard power systems (SPS) are rapidly embracing electrification, resulting in loads that generate pulsation currents and encounter substantial transients. However, conventional time-based features alone inadequate for effectively monitoring safeguarding these against faults. This highlights the critical requirement advanced machine learning based methods to discern differentiate between various transient stages within load profile. In this paper, we propose a Wavelet Graph Neural Network (WGNN) model non-intrusive fault detection SPS. The system leverages dynamic of SPS train test performance with varying scenarios. underlying structure interdependence among component states network captured using WGNN model, accuracies over 99% intrusive 97% detection. developed has also shown be robust presence pulse noise, achieving an accuracy 95%. At end, real-time simulation proposed method is validated on Hardware-in-the-loop system, guaranteeing high fidelity low latency approach. These findings validate effectiveness real-world applications