Ensuring Photovoltaic Module Integrity through Electroluminescence Imaging and Machine Learning Solutions

This paper presents a tool for detecting and preventing defects in photovoltaic (PV) modules using Electroluminescence (EL) imaging and machine learning. The tool employs adaptive defect detection models and convolutional neural networks (CNN) to improve detection accuracy and efficiency. Our dataset includes diverse images from different inspections across different regions, enhancing the robustness of the training process. The tool's phases include synchronized defect detectors, cell segmentation, and automated criticality assessment, allowing for precise defect identification and prioritization of maintenance actions. Testing demonstrates significant improvements in defect detection and criticality evaluation, reducing the need for manual inspections and optimizing PV system maintenance.