摘要
针对风力发电机早期故障时定子电流特征量难以提取的问题,提出了单子带重构改进小波变换结合BP神经网络的风力发电机故障诊断新方法。通过对风力发电机的定子电流进行单子带重构改进小波变换,消除了传统小波变换中的频率混叠现象;从小波变换后的子带信号中选取特征域、提取特征量作为BP神经网络的输入;在此基础上,结合BP神经网络的输入输出非线性映射能力,完成对故障的诊断和定位。经过仿真实验证实,该方法准确地实现了对风力发电机故障的诊断。
It is hard to extract effective feature quantities from the stator currents when some early faults occur in the wind power generator.A novel algorithm that combining the wavelet transform improved by single-band reconstruction and BP network is proposed to solve this problem.By using the wavelet transform improved by single-band reconstruction to the stator currents of wind power generator,the frequency aliasing in the traditional wavelet transform is eliminated.Then,the characteristic field and features from each sub-band signals that are generated from the wavelet transform are selected as the input of BP neural network.On this basis,by using BP neural network which has the input-output nonlinear mapping ability,the diagnosis and location of the fault are completed.The simulated results show that the method can accurately diagnose the faults of wind turbine generator.
出处
《电力系统及其自动化学报》
CSCD
北大核心
2012年第2期53-58,共6页
Proceedings of the CSU-EPSA
基金
山西省自然科学基金项目(2009011021-2)
关键词
单子带重构改进小波变换
神经网络
风力发电机
故障诊断
定子电流
反向传播网络
wavelet transform improved by single-band reconstruction
neural network（NN）
wind power generator
fault diagnosis
stator current
bacb-propagation network（BP）
