Convolutional Neural Network-Based Solar Panel Dust Detection and Automatic Cleaning with Integrated Temperature Control
摘要
The efficiency of solar panels can diminish due to various factors, including soiling, hotspot formation, elevated temperatures, and manufacturing discrepancies. Among these, soiling and high temperatures emerge as a significant culprit. This paper endeavors to explore the development of a portable, user-friendly, and cost-effective solar panel cleaning device and temperature control device to address the issue of high soiling and maintenance cost for solar power plants especially in deserted regions like the Middle East. Additionally, it delves into the methodology for detecting dust accumulation on solar panels, proposing an image processing object detection machine learning software trained on more than 3400 images, capable of understanding the cleanliness status of solar panels and autonomously triggering cleaning processes with the assistance of a microprocessor. This system also includes the automatic sprinkling of water for temperature reduction once the thermal IR sensor detects a temperature higher than the specified range. It comprises three integral components: software for dust detection, signal transmission for panel cleaning, temperature control with the use of microprocessor, and hardware featuring a specially crafted prototype composed of 3D laser printed acrylic components which allows for easy movement of device across the solar panel rails and sprinkle water efficiently. This paper also looks into the economic feasibility and the future prospects of the cleaning device. Comprehensive coverage of both software and hardware aspects is provided within this paper.