Automated Thermal Imaging Inspection of Solar Fields

Abstract

Utility-scale solar is growing across the globe due to the increased demand for clean, renewable energy sources and decreased costs of solar panels. Failing solar cells that are not producing electricity reduce a solar farm’s revenue. As a result, solar farm managers must conduct regular maintenance checks to ensure that all the solar panels are fully functional. A common way of checking whether a solar panel is performing at a sub-optimal level is to check its heat signature with a thermal camera. Many types of failures produce “hotspots,” areas on the solar panel surface that are hotter than the rest of the panel’s surface. Hotspots indicate that a panel is performing sub- optimally. Some utility-scale solar farms contain over 100,000 panels making manual heat signature checks with a hand-held thermal camera impractical. A more feasible solution is to attach a thermal camera to a quadcopter drone and check the heat signatures of each panel by flying the drone over all the panels. The aim of this thesis is to demonstrate that one can build one of these types of solar panel inspecting systems from scratch. I implemented this approach by building a quadcopter drone equipped with a thermal camera and used it to take pictures of solar modules at Princeton’s Solar Collector Field to see if my drone could capture any thermal images of hotspots. In addition to manually inspecting the thermal images of the panels, I also devised an algorithm using image segmentation to automate the analysis of these thermal images. I conclude that no hotspots were detected in the panels I inspected with my drone. I close with a discussion on some of the limitations of my implementation and how one might extend this thesis topic in future research.