Detection of Faults in Solar Panels Using Deep Learning
Seung Heon Han, Tariq Rahim, Soo Young Shin
Abstract
Renewable energies, carbon neutrality, and sustainable practices have become an important aim for many countries. Solar power generation has drawn much consideration where maintenance of solar panels is an essential task due to the natural and other mechanical circumstances. In this paper, we have proposed a deep learning (DL) approach for the detection of faults in solar panels. The proposed system uses an unmanned aerial vehicle (UAV) equipped with a thermal camera and GPS for acquiring thermal images and localization of the fault in solar panels. An improved version of You only look once (YOLOv3-tiny) is employed as a DL model for the detection of the fault and then transmitted that information using Long-Term Evolution (LTE) to a remote server for visualization. The performance of the proposed model is compared with the current default YOLOv3-tiny, where high performance was achieved by the proposed DL model.