Design and analysis of photovoltaic array temperature attributes using thermal imaging sensor device for ai-based defect detection system

Kamaruzaman, Luqman Al Hakim (2024) Design and analysis of photovoltaic array temperature attributes using thermal imaging sensor device for ai-based defect detection system. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Text (Full Text) Design and analysis of photovoltaic array temperature attributes using thermal imaging sensor device for ai-based defect detection system.pdf - Submitted Version Download (8MB)

Abstract

The need for sustainable and renewable energy alternatives is driven by the depletion of fossil fuels, which supply over 80% of the world's energy. Tidal, wind, geothermal, and solar energy are examples of clean energy, often known as renewable energy, because they can all be recycled naturally. As a renewable energy source, solar energy (including concentrating solar power (CSP) and solar photovoltaic (PV) power) contributes 3.6% of global electricity output. Nonetheless, it has cemented its position among other renewable energy technologies, accounting for more than 31% of total installed renewable energy capacity in 2022. Nonetheless, various faults, such as hotspots have an impact on the effectiveness and performance of solar panels. Underperformance of solar projects is becoming an increasing concern for solar energy system owners. The purpose of this study is to design a system that can easily detect defects on photovoltaic arrays of varying sizes and environmental conditions by developing an AI-based defect detection system using thermal imaging sensors to collect real-time temperature data from photovoltaic arrays and implement efficient algorithms for accurate defect identification, making it accessible and practical for the solar energy industry. This study is also to analyze the effectiveness and consistency of the AI-based defect detection system which uses YOLO v8. At first, The datasets are acquired from Google Images. Then, the datasets are acquired by flying a drone autonomously and capture thermal images of solar panels at a solar farm. The datasets are annotated using Roboflow. The AI model is trained and tested the AI model at 25, 50, 75 and 100 epochs. The effects of the number of epochs and the size of the datasets on the performance of the AI model was also analyzed. These findings are important in selecting the optimum object detection model.

Actions (login required)

View Item