摘要
为有效进行风电机组叶片运行时的裂纹状态评估,提出一种基于连续小波变换(Continue Wavelet Transform,CWT)和残差神经网络(Residual Networks,ResNet)结合的叶片裂纹状态评估方法。首先对叶片加速度振动信号做CWT后生成二维彩色时频图像,然后将图像分别作为训练集和测试集,使用34层ResNet进行训练和诊断,最后选取天津某风电场提供的1.5 MW风力发电机作为研究对象,根据其样本数据将叶片故障程度按照裂纹长度和宽度分为健康、轻微、中等、严重、危险5种状态,评估平均准确率高达98.23%,方法的有效性和可行性得到验证。
In order to effectively evaluate the crack state of wind turbine blades during operation,a blade crack state evaluation method based on CWT and ResNet is proposed.Firstly,the two-dimensional colorful time-frequency image is generated by CWT of the blade acceleration vibration signal.Then,with the image as the training set and the test set respectively,the 34-layer ResNet is used for training and diagnosis.Finally,the 1.5 MW wind turbine provided by a wind farm in Tianjin is selected as the research object.The degree of blade failure based on sample data is classified into five states according to crack length and width,such as healthy,minor,moderate,severe,and dangerous.The average accuracy of the evaluation is as high as 98.23%,which verifies the effectiveness and feasibility of the proposed method.
作者
李练兵
肖亚泽
张萍
张国峰
吴伟强
陈程
LI Lianbing;XIAO Yaze;ZHANG Ping;ZHANG Guofeng;WU Weiqiang;CHEN Cheng(State Key Laboratory of Reliability and Intelligence of Electrical Equipment,Hebei University of Technology,Tianjin 300130,China;School of Electrical Engineering,Hebei University of Technology,Tianjin 300130,China;Hebei Construction Investment OffshoreWind Power Co.,Ltd.,Tangshan 063000,Hebei,China)
出处
《噪声与振动控制》
CSCD
北大核心
2024年第2期143-148,293,共7页
Noise and Vibration Control
关键词
故障诊断
风电机组
状态评估
小波变换
残差神经网络
数据预处理
fault diagnosi
wind turbine
state estimation
wavelet transform
residual neural network
data preprocessing
