Solar Photovoltaic (PV) systems typically convert solar irradiance into electricity, thereby helping to reduce the need for fossil fuels and the amount of greenhouse gases released. They provide a reliable and continuous renewable source of energy. However, PV systems are continuously exposed to diverse and changing environmental conditions, such as temperature, humidity, dust, and rain. Exposure to such conditions creates electrical and visible faults in the PV systems. These faults may reduce the PV system’s performance, reliability, and lifetime. In this regard, this paper aims to propose a framework/methodology for reliability modeling and assessment of large-scale grid-connected PV systems using a Fault Tree Analysis (FTA) approach. Specifically, an exhaustive literature survey is carried out to acquire the failure rates of different components/faults existing on the DC side of the PV system. Then, the Fussel-Vesely (F-V) importance measure is employed to identify critical faults and their criticality ranking. Results showed that solder bond failure, broken cell, broken interconnect (finger interruption), rack structure, grounding/lightning protection system, delamination, discoloration, and partial shading are the most critical faults which severely degrade the performance of the PV systems. The recommendations and scope for further study are provided to optimize operations and maintenance costs.

Solar photovoltaic (PV) power plants represent a major pillar of the global renewable energy transition, with rapid technological advancements driving widespread adoption. As deployment increases, reliable fault detection becomes critical to ensure long-term performance, safety, and economic viability of PV systems. Common defects such as cracking, delamination, discoloration, and shadowing can substantially degrade power output and may lead to irreversible module failure if not detected early. Existing fault detection approaches largely rely on classical image processing techniques or single deep learning architectures, which struggle to handle complex visual variations caused by changing environmental and operational conditions. Furthermore, the scarcity of diverse, high-quality datasets limits model robustness and generalization. This paper proposes a hybrid deep learning framework that integrates a customized convolutional neural network (CNN) based on ResNet-50V2 with a Vision Transformer (ViT) to address these limitations. The CNN component captures fine-grained local spatial features, while the ViT leverages global self-attention to model long-range contextual dependencies. A high-quality RGB dataset was developed to analyze four major PV defects under varying lighting and weather conditions. Experimental results demonstrate strong performance, achieving 95.85% accuracy, 95.62% precision, 95.90% recall, and a 95.76% F1-score. The proposed model outperforms state-of-the-art CNN and transformer-based architectures, including EfficientNetV2, MobileNetV3, DeiT, CCT, and Swin Transformer. Overall, the framework provides a robust, scalable solution for intelligent PV monitoring, enabling early fault detection and improving system reliability and efficiency

Over time, solar Photovoltaic (PV) systems experience a decline in performance and reliability due to various environmental factors. Fault Tree Analysis (FTA) can be used to assess the reliability of these systems and identify faults and failure modes that can significantly impact the entire PV system’s performance. However, in practice, obtaining accurate failure probability values for the components of a solar PV system is challenging since systems operate in an ever-changing environment, resulting in a scarcity of data for statistical estimation. This paper proposes a fuzzy theory-based FTA approach to obtain the failure probabilities of the faults more accurately. The Fuzzy-FTA methodology converts experts’ subjective opinions expressed in linguistic terms into a Failure Possibility Score (FPS), which is then converted into a failure probability value. The results of the proposed approach are compared to those obtained through the conventional FTA to assess its effectiveness, applicability, feasibility, and efficiency. Soiling, dust accumulation, inadequate system maintenance, bad system configuration, bird dropping, delamination, improper installation, shading, grounding, and discoloration are the most critical faults and impact on the performance and reliability of solar PV systems.

Solar photovoltaic (PV) power plants are a key pillar of the global transition to renewable energy, and rapid technological development has led to widespread deployment. As deployment increases, reliable fault detection is essential to ensure long-term performance, safety, and financial sustainability of PV systems. Common defects such as cracking, delamination, discoloration, and shadowing can significantly reduce power output and cause irreversible failure of the module unless these defects are detected in the early stages. Current fault-detection systems are mostly based on classical image-processing methods or single deep-learning models, which cannot adapt to changing environmental and operational conditions and cannot cope with complex visual variations. Moreover, the limited variety and quality of datasets hinder the robustness and generalizability of the model. To examine four key PV defects under different lighting and weather conditions, a high-quality RGB dataset was created. The proposed work introduces a new hybrid CNN-ViT architecture that incorporates depth wise separable convolutions to dramatically decrease the computational complexity while keeping discriminative spatial features. Squeeze-and-excitation blocks are also added to improve defect-aware feature learning and provide adaptive channel-wise recalibration, which can be used to effectively highlight photovoltaic defects such as cracks, delamination, and discoloration. Moreover, a multi-scale attention mechanism with different kernel sizes (1×1, 3×3, and 5×5) is proposed to encode defect patterns at different spatial resolutions. To overcome these shortcomings, this paper proposes a hybrid deep learning architecture that combines a customized convolutional neural network (CNN, ResNet-50V2) and a Vision Transformer (ViT). The CNN part captures fine-grained local spatial features, while the ViT uses global self-attention to be long-range contextual dependencies. The experimental performance is high, with an accuracy of 95.85, a precision of 95.62, a recall of 95.90, and an F1-score of 95.76. The model proposed in this research paper is compared with ultramodern hybrid and transformer-based architectures, including DL-LGM, Cat Boost-GB, DeepF-SVM, DeiT, CCT, and Swin Transformer. Overall, the framework offers a powerful, scalable, intelligent PV monitoring solution that enables prompt fault detection and enhances system reliability and efficiency.