A Detailed Review of Photovoltaic Dust and Soiling Detection Techniques Using UAV, Image Processing, Machine Learning and Deep Learning Approaches

Computer Engineering

Solar photovoltaic systems are becoming increasingly important because of the growing demand for renewable and sustainable energy generation technologies. However, the efficiency and performance of photovoltaic panels are sig nificantly affected by environmental contaminants such as dust accumulation, bird droppings, mud, pollution par ticles, and sand deposition. These contaminants reduce sunlight transmission, decrease power generation efficiency, and create hotspot regions that may permanently damage photovoltaic cells. Traditional photovoltaic inspection and maintenance methods mainly depend on manual monitoring operations which are labor-intensive, expensive, time-consuming, and inefficient for large-scale photovoltaic farms. Researchers introduced intelligent photovoltaic monitoring systems using image processing, machine learning, deep learning, thermal imaging, Internet of Things technologies, unmanned aerial vehicles, and robotic cleaning systems to overcome these limitations. UAV systems equipped with RGB and thermal cameras provide high-resolution real-time monitoring capability for large pho tovoltaic installations. Deep learning algorithms such as CNN, YOLOv5, YOLOv7, YOLOv8, EfficientDet, Mask R-CNN, and SDS-YOLO provide highly accurate photovoltaic dust detection and fault diagnosis performance. Transfer learning approaches and attention mechanisms further improve photovoltaic inspection accuracy under varying environmental conditions. This review paper presents a comprehensive overview of photovoltaic dust and soiling detection techniques using image processing, machine learning, deep learning, UAV monitoring systems, thermal imaging, IoT technologies, and robotic cleaning systems. Comparative analysis, challenges, limitations, and future research directions are also discussed in detail.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.