# AI-driven fault detection and classification in photovoltaic systems using deep learning techniques

## TL;DR

This paper introduces PVDefectNet, a deep learning model for detecting and classifying faults in solar panels with high accuracy and interpretability.

## Contribution

PVDefectNet is a novel deep learning framework for fault detection in photovoltaic systems with enhanced performance and interpretability.

## Key findings

- PVDefectNet achieves 98% average accuracy in fault classification.
- Grad-CAM visualizations confirm the model's focus on relevant defect areas.
- The model outperforms existing methods in precision, recall, and F1-score.

## Abstract

The growing needs of the world regarding electricity and the exhaustion of fossil fuel resources have aggravated the need to use renewable energy, especially the photovoltaic (PV) systems. Nevertheless, internal defects and external environmental conditions are often known to affect the operational efficiency and reliability of PV modules. This paper presents PVDefectNet, a deep learning-based fault detector and classifier of the PV systems. The proposed solution applies a resnet architecture with data augmentation techniques to enhance its resistance to operating in different operating environments. PVDefectNet is a process based on five stages that include data preparation and preprocessing, model architecture design, training, evaluation and visualization, and performance analysis. The experimental findings indicate that the proposed framework has a high classification performance with an average accuracy of 98, precision of 97.1, recall of 96.5 and F1-score of 96.8 that is better than some of the current methods. Moreover, the visualizations provided by Grad-CAM prove that the model is concentrated on physically significant defect areas, which increases interpretability and reliability. These results suggest that PVDefectNet is a good and clear solution in intelligent monitoring and maintenance of PV systems.