摘要
针对大型风力发电机组偏航系统故障问题,提出一种基于麦克风的非接触式声学诊断方法。通过对偏航系统的故障机理和拾取声信号的特点分析,研究提取无量纲的高低频能量比与线性预测倒谱系数来表征偏航系统的状态。然后,引入主成分分析法自适应的优化初始特征向量,在此基础上设计一种基于支持向量机的故障分类器,并采用网格法寻优分类器的参数。最后,基于实测数据集验证了算法的有效性。
In view of the shortcomings of current fault detection methods for wind turbine yaw systems,this paper proposes a noncontact detection method based on acoustic signals.Through analysis of the fault mechanism of yaw systems and measuring acoustic signal,two different kinds of features including energy ratio of high frequency to low frequency and linear prediction cepstrum coefficients are presented for characterizing yaw system.Subsequently,original features are adaptively optimized using principal component analysis,the support vector machine classifier is designed to recognize the state of yaw system and the parameters of the classifier are optimized by grid method.Finally,effectiveness of proposed method is validated with dataset measured from wind field.
作者
李永战
谢磊
夏政
高宝成
Li Yongzhan;Xie Lei;Xia Zheng;Gao Baocheng(Tianjin Mingyang Wind Power Equipment Co.Ltd,Tianjin 300300,China;Automation School,Beijing University of Posts and Telecommunications,Beijing 100876,China)
出处
《电子测量技术》
2019年第17期169-173,共5页
Electronic Measurement Technology
关键词
偏航系统
故障诊断
高低频能量比
线性预测倒谱系数
主成分分析
支持向量机
yaw system
fault diagnosis
energy ratio of high frequency to low frequency
linear prediction cepstrum coefficients
principal component analysis
support vector machine