基于TimeGAN-Stacking的风电机组变桨系统故障诊断方法  

FAULT DIAGNOSIS METHOD FOR WIND TURBINE PITCH SYSTEM BASED ON TimeGAN-Stacking

作　　者：潘美琪 贺兴[2] Pan Meiqi;He Xing(College of Smart Energy,Shanghai Jiao Tong University,Shanghai 200240,China;Research Center for Big Data and Artificial Intelligence Engineering and Technologies,Shanghai Jiao Tong University,Shanghai 200240,China)

机构地区：[1]上海交通大学国家电投智慧能源创新学院,上海200240 [2]上海交通大学电力传输与功率变换控制教育部重点实验室,上海200240

出　　处：《太阳能学报》2025年第1期192-200,共9页Acta Energiae Solaris Sinica

基　　金：国家自然科学基金(52277111);上海市科学技术委员会(21DZ1208300)。

摘　　要：风电机组变桨系统的少量不均衡故障样本难以训练基于数据驱动的故障诊断模型,导致监测系统常常漏报或误报故障。针对上述问题,提出一种基于TimeGAN-Stacking的风电机组变桨系统故障诊断方法。在数据层面,由于原始样本类别不平衡,基于时序生成对抗网络(TimeGAN)跟踪风电机组运行数据逐步概率分布的动态变化特征,同时优化生成样本的全局分布与局部分布,有效平衡且扩容风电机组多种故障综合样本集;在模型层面,建立Stacking集成模型,融合多个故障诊断器的优势,进一步提高故障诊断能力。最后,基于实际风场数据对所提方法进行测试,结果表明,所提出的TimeGAN-Stacking故障识别方法可有效诊断4种变桨故障。The small number of unbalanced fault samples in the variable pitch system of wind turbines makes it difficult to train data-driven fault diagnosis models,leading to frequent missed or false alarms in monitoring systems..In response to the above issues,this article proposes a fault diagnosis method for wind turbine pitch system based on TimeGAN-Stacking.At the data level,due to the imbalance of the original sample categories,the dynamic change characteristics of the gradual probability distribution of the fan operation data are tracked based on the time-series Generative adversarial network(TimeGAN),and the global and local distribution of the generated samples are optimized to effectively balance and expand the comprehensive sample set of multiple faults of the fan;At the model level,establish a Stacking integrated model to integrate the advantages of multiple fault diagnosis devices and further improve fault diagnosis capabilities.Finally,the proposed method was tested based on actual wind field data,and the results showed that the proposed TimeGAN-Stacking fault identification method can effectively diagnose four types of pitch faults.

关 键 词：风电机组 数据挖掘 故障分析 深度学习 时序生成对抗网络 样本增强 

分 类 号：TM614[电气工程—电力系统及自动化]