基于ResNet-Transformer模型的风电机组偏航异响故障诊断  

Yaw Noise Fault Detection for Wind Turbines Based on ResNet-Transformer Model

作　　者：陈亚楠 胡凯凯 李籽圆 王立鹏 CHEN Yanan;HU Kaikai;LI Ziyuan;WANG Lipeng(CRRC Zhuzhou Institute Co.,Ltd.,Zhuzhou,Hunan 412001,China)

机构地区：[1]中车株洲电力机车研究所有限公司,湖南株洲412001

出　　处：《控制与信息技术》2025年第1期21-26,共6页CONTROL AND INFORMATION TECHNOLOGY

基　　金：湖南省十大技术攻关项目(2023GK1030)。

摘　　要：风电机组作为新型电力系统的核心设备,运行在恶劣的工作环境下,并受到多种不确定因素的影响,易发生故障。其中,作为风电机组的关键部件,偏航系统的故障显得尤为严重。为提高风电机组偏航系统异响故障诊断的准确性,文章提出了一种基于声音信号的ResNet-Transformer模型。首先,利用ResNet的残差结构进行声音信号的局部特征提取,以增强对微小信号变化的感知能力;其次,结合Transformer的多头注意力机制捕捉全局特征信息,从不同时间步长的原始信号中提取全局信息,使模型在捕捉长时间跨度特征时具备更高的灵活性和鲁棒性;最后,通过整合局部信息和全局信息,使模型同时关注微观细节和宏观依赖,从而精确识别和分类复杂声音信号中的关键特征。实验结果表明,该模型基于声音信号的偏航故障诊断准确率可达96.88%,为后续快速、有针对性的风电机组维护和安全诊断提供了新的技术途径。Wind turbines are among the core components of the new power system,operating in harsh environments and subjected to various uncertainties that make them prone to faults.The yaw system,as a critical component of wind turbines,is particularly susceptible to faults.To enhance the accuracy of diagnosing yaw noise faults in wind turbines,this paper develops a deep residual network(ResNet)-Transformer model based on acoustic signals.Firstly,the model employs the residual structure of ResNet to extract local features from the acoustic signals,enhancing sensitivity to subtle signal variations.Secondly,it utilizes the multi-head attention mechanism of the Transformer to capture global features,facilitating the extraction of global information from original signals across different time steps and thereby improving the model's flexibility and robustness in capturing long-term features.Finally,by integrating both local and global information,the model effectively balances micro-level details with macrolevel dependencies,enabling accurate identification and classification of critical features in complex acoustic signals.Experimental results demonstrated that the proposed model achieved an accuracy of up to 96.88%in diagnosing yaw faults based on acoustic signals,providing a novel technical approach for future rapid and targeted diagnostics to ensure maintenance and operational safety in wind turbines.

关 键 词：风电机组 偏航异响 故障诊断 深度残差网络 TRANSFORMER 

分 类 号：TM614[电气工程—电力系统及自动化]