Optimizing Photovoltaic System Diagnostics: Integrating Machine Learning and DBFLA for Advanced Fault Detection and Classification

The rapid growth in photovoltaic (PV) power plant installations has rendered traditional inspection methods inefficient, necessitating advanced approaches for fault detection and classification. This study introduces a novel hybrid metaheuristic method, the Dung Beetle Optimization Algorithm combined with Fick’s Law of Diffusion Algorithm (DBFLA), to address these challenges. The DBFLA enhances the performance of machine learning models, including artificial neural networks (ANNs), support vector machines (SVMs), and ensemble methods, by fine-tuning their parameters to improve fault detection rates. It effectively identifies critical faults such as module mismatches, open circuits, and short circuits. The research demonstrates that DBFLA significantly improves the performance of conventional machine learning techniques by forming a stacking classifier, achieving an individual meta-learner accuracy of approximately 98.75% on real PV datasets. This approach not only accommodates new operating modes and an expanded range of fault conditions but also enhances the reliability of fault detection schemes. The primary contribution of DBFLA lies in its ability to balance exploration and exploitation efficiently, resulting in superior classification accuracy compared to existing optimization techniques. By combining real and simulated datasets, the proposed hybrid method showcases its potential to substantially improve the precision and speed of PV fault detection models. Future work will focus on integrating these advanced models into real-time PV monitoring systems, aiming to reduce detection times and further enhance the reliability and operational efficiency of PV systems.