Artificial intelligence in wind turbine fault diagnosis: A systematic knowledge mapping and trend analysis

Abstract

Over the past decade, fault diagnosis technology in the wind energy sector has advanced rapidly, yet existing reviews exhibit methodological and data source fragmentation. This study employs bibliometrics and content analysis to systematically trace the conceptual evolution and technological trajectory of intelligent fault diagnosis for wind turbines. Based on 1736 fault diagnosis papers published in mainstream journals and conferences between 2016 and 2025, it quantitatively reveals trends in research themes, methodologies, and focal points. This study critically examines the strengths, limitations, and application boundaries of existing diagnostic frameworks, highlighting practical challenges such as data imbalance, insufficient open benchmarking, and obstacles to digital twin deployment. It also evaluates emerging trends in integrating foundational models with digital twins, noting their potential for enhancing component-level precision diagnostics. The proposed roadmap centers on deep learning and multimodal models, leveraging robust shared data, unified industry standards, and comprehensive security–privacy protection mechanisms. It emphasizes causal inference and lightweight edge deployment technologies. By tracing historical developments and evaluating existing diagnostic approaches, this study aims to accelerate data-driven practices in wind power operations and maintenance while establishing reliable intelligent fault diagnosis systems—critical for ensuring stable power generation from wind turbines.