南京信息工程大学学报2025，Vol.17Issue(3)：374-383,10.DOI:10.13878/j.cnki.jnuist.20240509001
具有注意力机制的CNN-GRU模型在风电机组异常状态预警中的应用
Application of CNN-GRU model with attention mechanism in anomaly warning of wind turbines
摘要
Abstract
To tackle the frequent failures of wind turbines resulting from their long-term operation in harsh environ-ments,an anomaly warning method using Convolutional Neural Network(CNN)and Gated Recurrent Unit(GRU)with attention mechanism is proposed.First,the Clustering by Fast Search and Find of Density Peaks(CFSFDP)al-gorithm and the Local Outlier Factor(LOF)algorithm are jointly utilized to preprocess the abnormal data in the wind turbine Supervisory Control And Data Acquisition(SCADA)system.Second,the input and output parameters are determined based on mechanism analysis and feature importance evaluation by Extreme Gradient Boosting(XG-Boost)algorithm.Then a CNN-GRU model with attention mechanism is employed to establish the performance pre-diction model for wind turbines under normal operating conditions.Based on this prediction model,the state evalua-tion indicators for wind turbines are constructed using the time-lapse sliding window,and the warning threshold is determined with the statistical interval estimation to realize the early warning of abnormal conditions.Finally,fault warning experiments are carried out with real historical fault data from a wind power unit,which show that the pro-posed method can accurately identify and warn the abnormal states in advance,facilitating timely fault handling by operation and maintenance personnel and ensuring the safe and stable operation of the wind turbine generators.关键词
风电机组/卷积神经网络/门控循环单元/注意力机制/故障预警Key words
wind turbine/convolutional neural network(CNN)/gated recurrent unit(GRU)/attention mechanism/fault warning分类
动力与电气工程引用本文复制引用
马良玉,胡景琛,段晓冲,黄日灏..具有注意力机制的CNN-GRU模型在风电机组异常状态预警中的应用[J].南京信息工程大学学报,2025,17(3):374-383,10.基金项目
河北省中央引导地方科技发展资金项目(226Z2103G) （226Z2103G）
