Automatic Fault detection of solar system based on data analysis using IRT images and deep learning CNN - LSTM strategy.

These arrays are essential in the ongoing shift toward sustainable energy sources, reducing our dependency on fossil fuels. However, despite their importance, solar panels are vulnerable to various types of faults that can significantly degrade their performance and energy conversion efficiency. One of the primary challenges in photovoltaic power generation is ensuring that these systems maintain optimal performance and continue to operate as intended over time. Recent advancements in renewable energy research have transformed the issue of reliance on fossil fuels into an opportunity for growth in sustainable energy solutions. Within this context, the timely and accurate detection of faults within solar installations becomes a critical component in ensuring the reliability, safety, and economic viability of solar power systems. Fault detection not only minimizes costly downtime and maintenance but also contributes to improved system performance. This study investigates an Intelligent Fault Detection and Diagnosis (IFDD) framework that uses radiometric infrared thermography (IRT) for predictive maintenance of SPV arrays. This approach enables efficient, remote diagnostics and continuous monitoring of system health. A deep learning model, based on a Convolutional Neural Network (CNN), was developed to automatically identify and classify faults in photovoltaic modules through thermal imagery. The research leveraged a robust dataset of 20,000 infrared (IR) images from operational solar farms, covering 12 different fault types, including partial shading, electrical shorting, and surface contamination. The proposed CNN model achieved a remarkable average accuracy of 99.06% in fault detection and classification, demonstrating its potential to significantly enhance diagnostic precision and support predictive maintenance strategies, ultimately improving the reliability of solar power systems