Innovative high-speed method for detecting hotspots in high-density solar panels by machine vision

Abstract The occurrence of hotspots in photovoltaic panels is one of the most common problems of solar power plants, which reduces the output power of photovoltaic arrays and can also cause irreparable damage to the solar cells. There are several ways to identify hotspots, including using custom datasets using thermographic camera images, which will be later used to teach YOLO and Faster R-CNN computer vision algorithms. In practice, it is observed that the YOLO algorithm is many times faster than the Faster R-CNN in high-density solar panels. Therefore, the applied method is the safest choice for automatic hotspot detection in large-scale photovoltaic power plants to improve overall efficiency. In this paper, by comparing the performance of methods such as Faster R-CNN with YOLO, we concluded that the YOLO algorithm has far better advantages in terms of quality of detection, and speed. Therefore, this factor makes the use of YOLO significantly helps to speed up the troubleshooting of solar modules caused by hotspots, and this factor improves the efficiency of solar power plants in the long run. Meanwhile, in the studies for this paper, the results extracted by Python have been optimized as an algorithm to be used for hotspot detection.