基于时空特征深度融合的双馈式风力发电机故障早期预测  

Early fault prediction for doubly-fed wind turbine generator based on the temporal spatial feature fusion

作　　者：胡爱军[1] 邢磊 张鑫[1] 孙俊豪 向玲[1] HU Aijun;XING Lei;ZHANG Xin;SUN Junhao;XIANG Ling(Department of Mechanical Engineering,Hebei Key Laboratory of Electric Machinery Health Maintenance&Failure Prevention,North China Electric Power University,Baoding 071003,Hebei,China)

机构地区：[1]华北电力大学机械工程系,河北省电力机械装备健康维护与失效预防重点实验室,河北保定071003

出　　处：《中国工程机械学报》2023年第4期353-357,共5页Chinese Journal of Construction Machinery

基　　金：国家自然科学基金资助项目(52175092)。

摘　　要：风力发电机组监控与数据采集(SCADA)系统记录了机组运行参数,包含丰富的运行状态变化信息,为风力发电机组的健康状态识别提供了数据来源。提出一种基于自注意力机制级联的时空特征深度融合网络模型(SC-TSFN),模型采用空洞卷积层增大感受野,提取不同变量之间的空间特征,通过双向长短期记忆网络学习空间特征前后时间依赖性,自注意力机制突出关键变量空间特征权重以及动态调节局部与全局时间特征信息。对2.0 MW风电机组10 min间隔SCADA数据分析结果表明,该方法实现了风力发电机故障的早期预测并消除了误报。The supervisory control and data acquisition(SCADA)system records the operating parameters of the wind turbine,which contains rich information about the change of the operating state,and provides a data source for the identification of the health state of the wind turbine.To solve this problem,this paper proposes a temporal spatial feature fusion network based on self-attention mechanism cascade.The spatial features between different variables are extracted by increasing the receptive field with atrous convolution layer,and the temporal dependence of spatial features are learned through the bidirectional long short-term memory network,the self-attention mechanism highlights the spatial feature weight of key variables and dynamically adjusts the local and global temporal feature information.The analysis results of the SCADA data of a 2.0 MW wind turbine at 10 min interval show that this method can achieve the early prediction of the wind turbine faults and eliminate false alarms.

关 键 词：风力发电机 风力发电机组监控与数据采集(SCADA) 深度学习 故障误报 时空特征 

分 类 号：TK83[动力工程及工程热物理—流体机械及工程]