Advanced deep learning-based methodology for multi-class diagnosis of wind turbine blade faults at the wind farm level

Chunchen Wei, Shimin Cai, Han Yang, Yanru Zhang,
Advanced deep learning-based methodology for multi-class diagnosis of wind turbine blade faults at the wind farm level,
Engineering Applications of Artificial Intelligence,
Volume 160, Part A,
2025,
111801,
ISSN 0952-1976,
https://doi.org/10.1016/j.engappai.2025.111801.
(https://www.sciencedirect.com/science/article/pii/S0952197625018032)
Abstract: Accurately diagnosing wind turbine blade faults is crucial for the efficient operation of wind farms. Recent advancements in deep learning have significantly improved fault detection and diagnosis techniques. However, a major challenge in current research is the limited generalizability of models trained on data from a single turbine, which often proves ineffective when applied to the broader context of an entire wind farm. To addressing this limitation, this work introduces a novel data-driven multi-fault diagnosis system tailored for application at the wind farm level. Our approach proposes a dropout-enhanced bidirectional long short-term (MBiLSTM-D) model that enhances generalizability across various turbines within a wind farm. Moreover, a focal loss function is employed to address data imbalance. To address the incompleteness and redundancy of the raw data, we also propose a series of innovative data preprocessing methods to handle missing values and useless information in the data. Extensive validation using supervisory control and data acquisition (SCADA) data from a real-world wind farm demonstrates that our system effectively overcomes the generalization shortcomings of previous models.
Keywords: Wind turbine blade; Multi-class fault diagnosis; Wind farm-level; Deep learning