A multi-stage review framework for AI-driven predictive maintenance and fault diagnosis in photovoltaic systems
Ali Hamza, Zunaib Ali, Sandra Dudley, Komal Saleem, Muhammad Uneeb, Nicholas Christofides
Abstract
The photovoltaic (PV) sector encounters challenges such as high initial costs, reliance on weather, susceptibility to faults, irregularities in the grid, and degradation of components. Predictive maintenance (PdM) aims to proactively identify issues, thereby enhancing reliability and efficiency but may lack specific fault details without additional diagnostic efforts. This research presents an advanced PdM and fault diagnosis framework that integrates fault pattern analysis, severity assessments, and critical fault predictions. It aims to improve the functionality of PV systems, minimize downtime, and enhance reliability by identifying and analyzing specific fault patterns. Consequently, our article provides a critical review of current Artificial Intelligence (AI) methodologies for PdM and fault diagnosis in PV systems. Moreover, this study highlights the significance of data standardization and offers recommendations on how PdM, when combined with fault diagnosis, can utilize various data sources to anticipate faults in advance, assess their severity, and optimize system performance and maintenance activities. To the best of the authors’ knowledge, no such review study exists.