Common Failures in Photovoltaic Systems and Their Diagnostic Methods With Applications of Artificial Intelligence Techniques: A Review

Abstract

Photovoltaic (PV) systems are susceptible to faults and degradation mechanisms arising from electrical, mechanical, and environmental stressors that can reduce performance, reliability, and safety. This review provides a comprehensive assessment of common PV system faults and degradation mechanisms by examining their causes, observable signatures, performance and safety consequences, and diagnostic pathways. Imaging and electrical-based techniques used for fault detection and diagnosis are reviewed and compared in terms of their working principles, diagnostic applications, advantages, limitations, and field operating requirements, while artificial intelligence (AI)-based developments associated with these diagnostic approaches are also discussed. The review further covers AI-based approaches, including machine learning and deep learning models, dataset generation, acquisition and preparation, comparative analysis of AI model families and their suitability for different PV diagnostic tasks, and evaluation metrics for performance assessment. The reviewed studies demonstrate the broad applicability of AI across electrical, environmental, imaging, and time-series data for PV fault detection and diagnosis, while the comparative analysis shows that AI model effectiveness is closely linked to the structure of the diagnostic data and the intended fault-detection task. The review further identifies key challenges and future research directions in PV fault detection and diagnosis, providing a comprehensive reference for researchers and practitioners seeking to improve the reliability, efficiency, and intelligence of PV system fault monitoring and maintenance.