Ismail Hossain

Thesis Name

Thesis Abstract

The global transition to renewable energy has placed solar photovoltaic (PV) systems at the centre of sustainable power generation. Despite the rapid decline in installation costs, maintaining system efficiency remains a persistent challenge. PV modules frequently suffer from faults such as micro-cracks, delamination, soiling, and hotspots, which degrade performance, shorten module lifespan, and increase maintenance costs. Small scale operators, who form a large portion of Australia’s distributed solar market, often lack the resources for regular manual inspection, leading to unrecognised losses in power output.

Traditional diagnostic methods- including manual visual inspection, infrared thermography, and electrical parameter monitoring- provide useful insights but have limitations. They are labour-intensive, depend heavily on operator expertise, and cannot easily scale to large installations. Environmental factors such as lighting, dust, and weather variations further reduce the accuracy of these conventional approaches (Akram et al., 2021). Consequently, there is an increasing need for automated, accurate, and cost-effective monitoring techniques capable of identifying faults in real time..

Recent advances in artificial intelligence, particularly deep learning, have shown great promise in automating visual-inspection tasks. Convolutional Neural Networks (CNNs) and Transformer-based architectures have achieved state-of-the-art results in object recognition, medical imaging, and industrial inspection (Wang et al., 2023). Applying these methods to PV diagnostics can significantly enhance fault identification and enable predictive maintenance. However, current research on PV fault detection using deep learning remains fragmented and uneven in quality.