基于AR-Hankel矩阵的风力发电机早期故障诊断方法研究

Early fault diagnosis method of wind turbine based on AR-Hankel matrix

针对风力发电机故障早期阶段,故障特征不明显,继电保护装置检测不到相关电气量异常的问题,文章提出了基于AR模型、Hankel矩阵和奇异值分解的风力发电机早期故障诊断方法。该方法先对主轴径向振动信号进行总体平均经验模式分解,再按照互相关准则选择若干个固有模态函数建立AR模型,然后对自回归系数构建Hankel矩阵并作奇异值分解,将奇异值作为故障特征输入支持向量机判断发电机的运行状态。试验结果表明,该方法能对直驱风力发电机正常运行、定子线圈匝间短路、发电机主轴偏心、发电机轴承磨损4种状态进行准确诊断。

Fault characteristics of the early stage of the wind turbine fault is not obvious, the relay protection device cannot detect the abnormal electrical quantity. A fault diagnosis method based on AR model, Hankel matrix and singular value decomposition is presented in this paper. Firstly, take ensemble empirical mode decomposition for spindle vibration signal, and select some intrinsic mode functions according to the cross correlation criterion to build auto regressive model; then, auto regressive coefficients form the Hankel matrix; Finally, singular value decomposition is applied on Hankel matrix, and singular values are input into support vector machine to judge the state of the generator. Experiments show that the method can make an accurate diagnosis for normal operation state, stator winding inter turn short circuit fault, the generator spindle eccentricity fault and generator bearing wear fault.