Download

Download Full Text (828 KB)

Description

This project aims to create an automatic system for identifying faults in solar panels using machine learning and image analysis. Solar panels are used a lot for generating renewable energy, but their efficiency can decrease due to various issues like dust, bird droppings, snow, electrical problems, and physical damage. Manually checking these panels is not only slow and costly but can also be dangerous, especially in big solar farms. An intelligent fault detection system can help make maintenance more efficient and increase energy output. For this project, a publicly available dataset of solar panel images from Kaggle is used. The dataset includes images grouped into six categories: Bird-drop, Clean, Dusty, Electrical-damage, Physical-Damage, and Snow-Covered. These images show different states of solar panels and help the model learn to tell the difference between normal and faulty panels. The dataset is organized and divided into training, validation, and testing groups to make sure the model learns and is tested properly .The YOLOv8 classification model is used for this project. YOLO (You Only Look Once) is a deep learning model known for its speed and accuracy in image analysis. A pre-trained YOLO model is adapted using the solar panel dataset through transfer learning. During training, the model learns the visual features associated with each type of fault .After the model is trained, it is assessed using metrics like accuracy, precision, recall, and confusion matrix. The goal is to develop an automated system that can detect solar panel faults from images with high accuracy. This approach can help with predictive maintenance, cut down inspection costs, and boost the overall efficiency and reliability of solar energy systems.

Publication Date

4-30-2026

Keywords

solar panel, YOLO

Solar Panel Fault Detection Using Computer Vision and Deep Learning

Share

COinS