基于极限学习机的风电机组主轴承故障诊断方法研究

发布时间：2018-11-03 15:33

【摘要】：风能是一种无污染、可再生的新能源,因此近年来风力发电得到了越来越多的研究。然而,大多数风电机组工作环境较恶劣,主轴承是风电机组中至关重要的传动部件,起支撑和导向作用。如果主轴承发生故障,机组将停止运行,给风电场带来极大的经济损失。因此,快速有效地诊断风电机组主轴承的故障是提高风电机组利用率,进而提高风电场经济效益的有效措施。论文主要研究内容如下:给出了小波包能量特征提取方法以提取风电机组主轴承振动信号的特征向量。通过比较不同小波基函数与阈值函数组合的降噪效果,选出最佳降噪组合,并通过分析说明软阈值处理方式的降噪效果更优。采用小波包能量特征提取方法提取风电机组主轴承振动信号的能量向量,并分析不同故障类型特征向量的相似之处,为后文的故障识别奠定了基础。给出了基于极限学习机的风电机组主轴承故障诊断方法。通过比较不同激活函数对极限学习机诊断效果的影响,选择出最优激活函数。分析了极限学习机参数对极限学习机诊断效果的影响,并给出了诊断的具体实现过程。实验证明,与最小二乘支持向量机算法相比,基于极限学习机算法的风电机组主轴承故障诊断方法具有更好的诊断效果。给出了基于核极限学习机的风电机组主轴承故障诊断方法。采用遗传算法优化核极限学习机参数,以进一步提高诊断准确率。给出了诊断的具体实施过程。通过比较基于核极限学习机与极限学习机的故障诊断混淆矩阵,发现核极限学习机的诊断效果更好。给出了基于混合核极限学习机的风电机组主轴承故障诊断方法。利用全局核函数和局部核函数的线性组合构造混合核函数以使核函数兼有全局特性和局部特性。采用遗传算法结合交叉验证的参数优化方法,对混合核极限学习机参数进行优化。给出了诊断的具体实现过程。实验证明,基于遗传算法和交叉验证优化的混合核极限学习机的风电机组主轴承故障诊断方法具有更好的诊断效果。
[Abstract]:Wind energy is a kind of non-pollution, renewable new energy, so wind power generation has been more and more studied in recent years. However, the working environment of most wind turbine units is poor, and the main bearing is the most important driving part in wind turbine, which plays a supporting and guiding role. If the main bearing fails, the unit will stop running, which will bring great economic loss to the wind farm. Therefore, rapid and effective diagnosis of the main bearing fault of wind turbine is an effective measure to improve the utilization ratio of wind turbine and increase the economic benefit of wind farm. The main contents of this paper are as follows: a wavelet packet energy feature extraction method is presented to extract the eigenvector of the main bearing vibration signal of wind turbine. By comparing the noise reduction effect of different wavelet basis function and threshold function, the optimal denoising combination is selected, and the result of soft threshold processing is better. The wavelet packet energy feature extraction method is used to extract the energy vector of the main bearing vibration signal of wind turbine unit, and the similarity of different fault type characteristic vectors is analyzed, which lays a foundation for the later fault identification. The fault diagnosis method of main bearing of wind turbine based on ultimate learning machine is presented. By comparing the influence of different activation functions on the diagnostic effect of LLM, the optimal activation function is selected. The influence of the parameters of the ultimate learning machine on the diagnostic effect of the ultimate learning machine is analyzed, and the realization process of the diagnosis is given. Compared with the least square support vector machine (LS-SVM) algorithm, the method of main bearing fault diagnosis of wind turbine based on LLM-algorithm is proved to be more effective. The main bearing fault diagnosis method of wind turbine based on nuclear limit learning machine is presented. Genetic algorithm is used to optimize the parameters of the kernel limit learning machine to further improve the diagnostic accuracy. The implementation process of diagnosis is given. By comparing the fault diagnosis confusion matrix based on nuclear extreme learning machine and ultimate learning machine, it is found that the diagnosis effect of nuclear extreme learning machine is better. A fault diagnosis method for main bearing of wind turbine based on hybrid core ultimate learning machine is presented. The mixed kernel function is constructed by using the linear combination of global kernel function and local kernel function to make the kernel function have both global and local properties. The parameters of hybrid kernel ultimate learning machine are optimized by genetic algorithm and cross-validation. The realization process of diagnosis is given. The experiments show that the fault diagnosis method of main bearing of wind turbine based on genetic algorithm and crossover verification optimization is more effective.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM315

【参考文献】