关键词: 风力发电机, 状态检修, 多层极限学习机, 希尔伯特-黄变换, 频域分析

Abstract: During the operation of wind power generation machines,multiple source monitoring signals exhibit complex characteristics due to strong noise. The fault characteristics in the frequency domain overlap and are confusing, making it prone to misjudgment and missed detection. Therefore,a state intelligent detection method based on multi-sensor information fusion and multi-layer extreme learning machine is proposed. By integrating multi-dimensional monitoring data,the HilbertHuang transform is used for adaptive frequency domain decomposition to extract frequency domain energy features;a multilayer extreme learning machine model is constructed to achieve hierarchical identification and precise classification.
Experimental results show that in the early degradation state recognition, only one misjudgment occurred. The F₁ score in various states is stable at above 0. 9,providing a reliable path for intelligent operation and maintenance of the machines.

Key words: wind power generators, condition-based maintenance, multilayer extreme learning machine, hilbert-huang transform, frequency-domain analysis