The increased implementation of Grid connected PV Systems calls for complex fault detection and diagnosis solutions to ensure system reliability, safety and efficiency. Traditional methods such as threshold-based and rule-based methods perform poorly with complicated fault scenarios and variations in changing environmental conditions. Artificial intelligence techniques such as machine learning, deep learning, and ensemble learning methodologies represent strong candidates to diagnose faults in PV systems connected to a grid. This paper presents an overview of significant developments in Fault detection and diagnosis techniques, such as Convolutional Neural Networks (CNNs), Artificial Neural Networks (ANNs), Support Vector Machines (SVMs), and Random Forest (RF) classifiers. CNN models have shown high levels of performance in spatial and temporal fault feature extraction, with ANN-based methods offering adaptable learning for non-linear fault patterns. SVMs provide better classification accuracy with clearly defined decision boundaries, with RF-based methods taking advantage of ensemble learning to address large volumes of information with unevenly distributed fault types. Hybrid models integrating optimized systems with deep learning configurations, with additional use of statistical methodologies, have improved fault classification with increased efficiencies. Comparative studies present an overview of their merits and demerits, which point to accuracy, computational complexity, and applicability in real time. Though there has been considerable development, there is still scope for improvement in availability, scalability, and readability of the data sets. Future studies should address the integration of hybrid intelligent models with optimization-based systems to design scalable real-time fault detection models for large photovoltaic systems.
Ratsheola et al. (Mon,) studied this question.