A Generative Adversarial Network Based Approach for Accurate Detection of Bright Spots in Photovoltaic Panels

Ensuring the safe and efficient operation of photovoltaic (PV) solar panels requires early detection of defects during the manufacturing process itself. Among these defects, the presence of 'bright spots' on the surface of PV solar panels poses a major concern as they can lead to excessive energy production which further leads to short circuits, increased heating, and potentially even fires. In this paper, we propose a robust approach to reliably detect bright spots on the surface of PV cells that are based on image classification techniques enhanced by novel feature extraction and generative adversarial networks (GANs). The proposed approach demonstrated robust performance on Electro-Luminescence (EL) images from a real, operating PV solar panel manufacturing unit with a recall of 0.93 and an F1-score of 0.93, offering a reliable solution to detect bright spots in PV cells.