DETECTION AND CLASSIFICATION OF DEFECTS IN SOLAR PANEL CELLS USING DEEP LEARNING MODEL

Abstract

This study presents an advanced approach for the detection and classification of defects in solar cells through Electroluminescence (EL) imaging using a Lightweight Convolutional Neural Network (LwNet) model. The main objective is to increase the precision and effectiveness of solar cell defect detection, which is essential for quality assurance in the production of solar panels. Both mono-crystalline and poly-crystalline silicon solar cells were included in the updated EL-Photovoltaic (ELPV) dataset, which was then subjected to augmentation approaches to resolve class imbalance. With a compact architecture tailored for defect identification, the suggested LwNet model produced remarkable outcomes, including 97.53% accuracy, 96.94% precision, 97.26% recall, and 97.08% F1-score. This study shows that the LwNet model is a very good way to detect defects in solar cells. It provides a lightweight, accurate, and efficient substitute for real-time manufacturing and quality evaluation applications.