Automated Solar Panel Fault Detection using Transfer Learning Techniques
Year: 2026
Type: conference-paper
Abstract: The rapid deployment of Photovoltaic (PV) systems has increased the need for accurate defect detection to ensure reliable operation, high efficiency and long-term durability of PV systems. The manual process is often lengthy, time-consuming and error-prone, making it ineffective for large-scale PV system monitoring. This study proposes Deep Learning (DL)-based approach with transfer learning to perform automatic fault detection in solar panels. Defects such as cracks, hotspots, shadows, and dirt can degrade the energy yield and reduce efficiency in PV modules. There is an increasing need to accurately detect defects due to the rapid growth of solar energy systems, as defects adversely affect system efficiency and performance. Traditional method for detecting defects, which are tedious, time-consuming and prone to human error. In this paper, DL with transfer learning approach is proposed to detect faults automatically in solar panels. Various faults, including cracks, hotspots, and shadows, lead to a reduction in energy production and can impact PV modules, thus reducing system efficiency and power generation. The performance of models is assessed based on accuracy, precision, recall, F1-score and area under Receiver Operating Characteristic Curve (ROC-AUC). The experimental findings show that this MobileNetV2 with accuracy of 79% and ROC-AUC score of 0.90 implies excellent generalization ability.
Institution: SRKR Engineering College
Language: English
Subfield: Industrial and Manufacturing Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7203474734