Enhancing Solar Plant Efficiency: A Review of Vision-Based Monitoring and Fault Detection Techniques

Abstract

Over the last decades, environmental awareness has provoked scientific interest in green energy, produced, among others, from solar sources. However, for the efficient operation and longevity of green solar plants, regular inspection and maintenance are required. This work aims to review vision-based monitoring techniques for the fault detection of photovoltaic (PV) plants, i.e., solar panels. Practical implications of such systems include timely fault identification based on data-driven insights and problem resolution, resulting in enhanced energy outputs, extended lifetime spans for PV panels, cost savings, as well as safe and scalable inspections. Details regarding the main components of PV systems, operation principles and key non-destructive fault detection technologies are included. Advancements in unmanned aerial vehicles (UAVs), as well as in artificial intelligence (AI), machine learning (ML) and deep learning (DL) methods, offering enhanced monitoring opportunities, are in focus. A comparative analysis and an overall evaluation of state-of-the-art vision-based methods for detecting specific types of defects on PVs is conducted. The current performance and failures of vision-based algorithms for solar panel fault detection are identified, raising their capabilities, limitations and research gaps, towards effectively guiding future research. The results indicate that shading anomalies significantly impact the performance of PV units, while the top five fault detection methodologies, according to preset evaluation criteria, involve deep learning methods, such as CNNs and YOLO variations.