Explainable AI for Wind Energy Systems: State-of-the-art Techniques, Challenges, and Future Directions

This review paper offers a thorough assessment of Explainable Artificial Intelligence (XAI) methodologies applied to wind energy systems, which are crucial for improving transparency, trust, and operational performance in wind energy-related areas including wind power forecasting, fault detection and predictive maintenance, wind farm optimization and control, and Supervisory Control and Data Acquisition (SCADA) data analysis. It elaborates on model-agnostic and model-specific XAI methods and more recently emerging methods such as counterfactual explanation and concept-based reasoning, and the potential of these approaches to explain the more complicated AI models used in wind turbine applications. We also review the important issues of the lack of benchmarking datasets, limited temporal explainability, human factors integration, and hardware limitations for real-world real-time deployment. Furthermore, we include the current evaluation measures, actual on-site deployments, and suggest future research to develop lightweight, temporally aware, human-centered, and causally interpretable AI systems for safer, more reliable, and efficient wind energy systems. • XAI boosts transparency and efficiency in wind energy operations. • Covers forecasting, fault detection, and predictive maintenance. • Reviews model-agnostic vs model-specific methods for interpretability. • Notes challenges: data limits, real-time use, and human-factor issues. • Suggests lightweight, human-centered, causally interpretable XAI designs.