基于视觉Transformer多模型融合的风电机组异常状态监测  

ABNORMAL CONDITION MONITORING OF WIND TURBINE BASED ON VISION TRANSFORMER MULTI-MODEL FUSION

作　　者：向玲[1] 高鑫[1] 姚青陶 苏浩[1] 胡爱军[1] 程砺锋 Xiang Ling;Gao Xin;Yao Qingtao;Su hao;Hu Aijun;Cheng Lifeng(Department of Mechanical Engineering,North China Electric Power University,Baoding 071003,China)

机构地区：[1]华北电力大学(保定)机械工程系,保定071003

出　　处：《太阳能学报》2025年第4期522-529,共8页Acta Energiae Solaris Sinica

基　　金：国家自然科学基金(52075170,52175092);河北省自然科学基金(A2022502005);河北省在读研究生创新能力培养资助项目(CXZZBS2024169)。

摘　　要：为实现风电机组的异常状态监测并用于其故障诊断和日常维护,提出一种新的监测方法,该方法基于视觉Transformer(ViT)模型与长短期记忆(LSTM)网络融合,能有效识别风电机组的运行状态。首先,利用箱线图法和Spearman相关性分析对原始SCADA数据进行预处理,去除无效数据并选择输入参数。然后,构建融合LSTM的ViT预测模型,并引入统计学中KL散度作为检测指标,对目标参数预测值与真实值进行计算分析。最后采用核密度估计确定安全阈值,根据检测指标是否越过安全阈值来识别风电机组异常状态。通过将该模型应用于华北某风场进行实例分析,并与其他深度学习模型对比。结果表明:该方法相较于其他模型能更好识别出风电机组异常状态。In order to realize the abnormal state monitoring of wind turbines and use them for fault diagnosis and routine maintenance,a new monitoring method is proposed in this paper,which is based on the fusion of Vision Transformer model and long short-term memory(LSTM)network,which can effectively identify the operating status of wind turbines.Firstly,the box plot method and Spearman correlation analysis were used to preprocess the original SCADA data,remove the invalid data and select the input parameters.Then,the Vision Transformer prediction model fused with LSTM was constructed,and the KL divergence in statistics was introduced as the detection index to calculate and analyze the predicted value and the real value of the target parameter.Finally,the kernel density estimation is used to determine the safety threshold,and the abnormal state of the wind turbine is identified according to whether the detection index exceeds the safety threshold.The model was applied to a wind farm in North China for case analysis,and compared with other deep Xi models.The results show that the proposed method can better identify the abnormal state of wind turbines than other models.

关 键 词：风电机组 状态监测 长短期记忆网络 视觉Transformer KL散度 

分 类 号：TK83[动力工程及工程热物理—流体机械及工程]