基于卷积神经网络的风电机组气动不平衡故障诊断方法研究  

STUDY ON DETECTION METHOD OF ROTOR AERODYNAMIC IMBALANCE BASED ON CONVOLUTIONAL NEURAL NETWORK

作　　者：杨旺春 梁雪 孙传宗 Yang Wangchun;Liang Xue;Sun Chuanzong(Guodian Power Guangxi Wind Power Development Co.,Ltd.,Beihai 536000,China;Liaoning Zhongke Jingchuang Intelligent Energy Co.,Ltd.,Shenyang 110027,China;School of Mechanical Engineering,Shenyang University of Technology,Shenyang 110870,China)

机构地区：[1]国电电力广西风电开发有限公司,北海536000 [2]辽宁中科经创智慧能源有限公司,沈阳110027 [3]沈阳工业大学机械工程学院,沈阳110870

出　　处：《太阳能学报》2025年第3期531-537,共7页Acta Energiae Solaris Sinica

基　　金：辽宁省科技厅博士启动基金(2019-BS-182);航空发动机双转子系统非线性动力学建模及振动机理研究。

摘　　要：为解决风电机组中风轮气动不平衡的诊断问题,降低风电机组的运维成本,提出一种基于一维卷积神经网络的风轮不平衡识别方法。融合变分模态分解和相关峭度计算实现风轮气动不平衡的感知。并提出基于一维卷积神经网络的气动不平衡识别方法,以机舱的振动加速度作为输入,识别气动不平衡的具体程度。在不同湍流强度和噪声环境下进行交叉验证,识别结果的准确率在95%以上,证明该方法可应用于风轮不平衡的诊断中,提升风电机组运行的安全性。For the problem of identification for rotor imbalance in wind turbine,and to reduce the operation and maintenance cost of wind turbine,an identification method of rotor imbalance based on one-dimensional convolutional neural network is proposed.Firstly,the combination of variational mode decomposition(VMD)and correlation kurtosis calculation is used to realize the perception of the rotor aerodynamic imbalance.Secondly,a recognition method of aerodynamic imbalance based on one-dimensional convolutional neural network is proposed,and the vibration acceleration of the nacelle is taken as the input to identify the specific magnitude of the rotor aerodynamic imbalance.Finally cross-validation was performed in different turbulence intensity and noise environments,and the identification accuracy of the cross validation was more than 95%,which proved that the method could be applied to the diagnosis of rotor imbalance and improve the safety of wind turbine.

关 键 词：风电机组 机器学习 故障诊断 风轮不平衡 卷积神经网络 

分 类 号：TK81[动力工程及工程热物理—流体机械及工程]