基于支持向量机和多源信息的直驱风力发电机组故障诊断

Direct-Drive Wind Turbine Fault Diagnosis Based on Support Vector Machine and Multi-Source Information

提出了一种综合考虑风速、转速以及主轴水平方向和垂直方向振动的时域特征参数、频域特征参数等多源信息的基于支持向量机(support vector machine,SVM)的直驱风力发电机组故障诊断方法。对直驱风电机组正常状态、风轮质量不平衡、风轮气动不平衡、偏航和断叶片等5种状态进行实验分析,研究不同状态下的机组特征。根据实验分析结论,将风电机组主轴水平方向、垂直方向振动的时域参数、频域参数以及风速、转速选为描述机组运行状态的特征参数,对机组进行故障识别。将风电机组5种状态下的特征参数作为学习样本,在SVM中训练,建立不同特征的参数向量和故障类型的映射关系,从而达到故障诊断的目的。根据风电机组不同故障的实验数据,对考虑多源信息的故障模型进行应用检验。结果表明,该方法简单有效,具有很好的故障识别能力和良好的鲁棒性,适合直驱风电机组故障诊断,同时可以满足在线故障诊断的要求。

A support vector machine （SVM）-based diagno：,is method of direct-drive wind turbine generation set is proposed. In this diagnosis method the multi-source information such as wind speed, rotational speed of wind turbine, time-domain and frequency-domain feature parameters of vibration signal in horizontal and vertical direction are synthetically considered. Experimental analysis of direct-drive wind power generation set under five conditions, i.e., normal condition, wind wheel mass imbalance, wind wheel aerodynamic imbalance, yaw and blade break, is carried out to research the features of the wind turbine under different states： Based on the results of experimental analysis, the time-domain and frequency domain parameters of vibration of main shaft of wind turbine generation set in horizontal and vertical direction as well as wind speed and rotational speed are selected as feature parameters to describe operation state of the wind turbine for the fault identification. Taking the feature parameters of wind turbine under the five conditions as learning samples, the SVM is trained and the mapping relations between different feature parameters and fault types are built for the fault diagnosis. According to experimental data of different faults of wind turbine generation set, the application testing of fault model considering multi-source information is performed, and the testing results show that the proposed method is simple and effective; the proposed method possesses satisfied fault identifying ability and good robustness, so it is suitable for fault diagnosis of direct-drive wind turbine, meanwhile it can meet the requirement of online fault diagnosis.