A review of data fusion technology for fault diagnoses in wind turbine components

Abstract

Traditional fault diagnostic methods in wind turbines (WTs), which rely on single-source data, suffer from significant limitations, including incomplete component coverage, susceptibility to sensor failures, and inadequate noise suppression. As WTs become increasingly complex with vast data sources, there is a growing need for more robust diagnostic approaches that integrate multiple data sources. Despite the rising interest in data fusion technology, a systematic review of its application across WT components remains absent in the literature, which is the gap this study aims to bridge. The novelty of this study lies in its broad scope, encompassing a wider range of WT components, data fusion methodologies, data combinations, fault conditions, and critical limitations in existing methods. This review highlights the current technological advances in data fusion for WT fault diagnosis, exposing the benefits gained and challenges encountered. Our findings reveal common misconceptions in this field, which restrict the full potential of advanced diagnostic systems. To further enhance the potential benefits of this approach, future research must explore the fusion of multimodal, heterogeneous data from diverse sources, enabling WTs to achieve greater diagnostic accuracy and robustness. This review also underscores the critical need for comprehensive monitoring of all essential WT components, ensuring that technology is not limited to fault detection in a subset of the system. Finally, advancing data fusion algorithms and approaches to manage the ever-growing volume and complexity of multisource data would ultimately enhance turbine reliability, reduce maintenance costs, and increase operational efficiency.