Smarter Grids Through Clearer Vision: AI for Predictive Fault Detection

We propose an innovative adaptation of AI-based medical imaging methods for fault detection in power systems. Our framework, FusionScan AI, repurposes an enhancement pipeline originally built to improve Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) scans, applying it instead to the visual inspection of power infrastructure, including transformers, switchgear, and distribution assets. The approach employs three main phases: deep learning based denoising, adaptive contrast optimization, and edge refinement through convolutional filtering. Together, these operations significantly improve the visibility of thermal and optical sensor data that are often degraded by noise or poor contrast. When tested on infrared thermographic images of high voltage equipment and substations, the system revealed hidden indicators of faults such as insulation breakdown, overheating at joints, and structural irregularities. Drawing inspiration from medical imaging, where similar methods have successfully exposed tumors and microlesions, our study shows that these techniques can be effectively transferred to power system diagnostics. The results highlight the promise of medical imaging-inspired AI pipelines for monitoring distributed energy networks, supporting predictive maintenance, enhancing safety, and improving efficiency in the operation of future smart grids.