Transfer Learning-based Enhanced Detection of Hotspot and Bypass Diode Faults in Photovoltaic Arrays

Due to the increasing importance of solar systems and their growing role in the world, ensuring their proper and flawless operation is of utmost importance. Solar photovoltaic (PV) systems are susceptible to various faults, making early fault detection a critical task. Persistent faults in PV systems can lead to irreversible damages. In this regard, this paper focuses on PV fault detection and classification using the transfer learning method. Initially, a dataset of thermographic images from PV arrays was collected. Subsequently, fine-tuning was performed, and finally, fault detection and classification were carried out using a pre-trained model. The model achieved 100% accuracy in detecting and classifying bypass diode and hotspot faults, as well as distinguishing healthy modules.