基于SVD-CWT和CNN的水轮发电机转子故障识别  

Rotor Fault Identification of Hydro-generator Based on SVD-CWT and CNN

作　　者：张彬桥[1,2] 刘雷 杨洋 侯成伟 ZHANG Bin-qiao;LIU Lei;YANG Yang;HOU Cheng-wei(Hubei Provincial Key Laboratory for Operation and Control of Cascaded Hydropower Station,China Three Gorges University,Yichang 443002,Hubei Province,China;College of Electrical Engineering and New Energy,China Three Gorges University,Yichang 443002,Hubei Province,China;Datang Guanyinyan Hydropower Development Co.,Ltd,Kunming 650000,Yunnan Province,China)

机构地区：[1]梯级水电站运行与控制湖北省重点实验室(三峡大学),湖北宜昌443002 [2]三峡大学电气与新能源学院,湖北宜昌443002 [3]大唐观音岩水电开发有限公司,云南昆明650000

出　　处：《中国农村水利水电》2024年第2期205-209,共5页China Rural Water and Hydropower

基　　金：国家自然科学基金面上项目(52077120)。

摘　　要：水轮发电机转子振动故障识别是水电站运维的重难点问题,为此提出一种基于转子振动信号的故障识别方法。首先针对发电机转子的非平稳和非线性振动信号,采用奇异值分解(SVD)并结合能量差分谱理论进行降噪预处理;对预处理数据使用连续小波变换(CWT)转换为时频图并形成图像数据集;然后将该图像数据集作为卷积神经网络(CNN)输入,通过CNN多层池化及卷积形成分布式故障特征表达,最终实现发电机转子故障模式识别和分类。经实验验证,该方法准确率达到99.5%以上,能有效识别出发电机转子的故障类型。The rotor vibration fault identification of hydro-generator is a difficult problem in hydropower station operation and maintenance.Therefore,a fault identification method based on rotor vibration signal is proposed.Firstly,for the non-stationary and non-linear vibration signals of generator rotor,singular value decomposition(SVD)is used for denoising preprocessing combined with energy difference spectrum theory;the preprocessed data is transformed into time-frequency graph by continuous wavelet transform(CWT)and formed into image data set.Then the image data set is used as convolutional neural network(CNN)input,and distributed fault feature expression is formed by CNN multi-layer pooling and convolution.Finally hydro-generator rotor fault mode recognition and classification are realized.The experimental results show that the accuracy of this method is above 99.5%,which can effectively identify the fault types of hydro-generator rotor.

关 键 词：水轮发电机转子 故障识别 SVD CWT 卷积神经网络 

分 类 号：TK730[交通运输工程—轮机工程] TP277[动力工程及工程热物理—流体机械及工程]