Artificial Intelligence for Fault Detection and Diagnosis in Wind Turbines: A Comprehensive Survey

Abstract

Wind turbines are pivotal to renewable energy, yet their complexity and deployment amplify vulnerability to faults in critical components like gearboxes, bearings, and blades, driving high maintenance costs. This survey reviews the advancements in fault detection and diagnosis in wind turbines, with emphasis on traditional machine learning, deep learning, and hybrid approaches. The review includes data sources and preprocessing techniques for fault detection, and diagnosis, such as: denoising, fusion, and handling imbalanced data, as well as a description of the common fault types, ensemble models, and the use of explainable AI (LIME and Shapley Value) for performance evaluation using metrics (Accuracy, F1-score, RMSE, and Early Warning) to assess the overall effectiveness and the challenges faced when deploying AI-based fault detection and diagnosis systems in practice. Limitations including data scarcity, poor generalizability, and black-box opacity are discussed, alongside future directions like digital twins, few-shot learning, and edge computing. This work provides an industry-focused review to guide scalable, interpretable fault detection and diagnosis systems for cost-effective wind energy.