基于工况辨识的风电机组故障预警方法研究

发布时间：2018-05-16 04:13

本文选题：风电机组 + 工况辨识　；参考：《华北电力大学》2017年硕士论文

【摘要】：近年来,风力发电行业的快速增长导致风力发电机组故障频发,运行维护费用持续增长。加强对风电机组运行状态的有效监测、实现风电机组早期故障的预警和诊断对于保障机组安全稳定运行、减少维修支出有着重要意义。与此同时,由于大型风电机组运行工况的多变性与复杂性使得机组运行状态难以准确评估,基于工况辨识的风电机组在线监测已经逐渐成为风电发展中的重要研究方向。本文首先对风电机组运行工况进行分析和辨识,然后以风电机组齿轮箱和滚动轴承为研究对象,从SCADA参数和振动参数两种数据来源着手,分别基于多元状态估计模型和变分模态分解方法,研究相对应故障预警模型和故障诊断方法,具体分为以下几个方面:首先,分析风电机组运行特性,利用风电场SCADA系统数据,在进行数据分析和预处理后,选择适当参数,利用模糊聚类算法,进行风电机组运行子工况的划分。其次,研究基于多元状态估计的风电机组故障预警。对于齿轮箱典型故障进行研究和分析;详细介绍MSET建模原理,研究MSET建模变量选取问题,通过实际风场数据对该模型的性能进行验证,并通过实验对比分析得出结论,工况辨识能够降低异常点误判断率从而降低误报警率。最后,研究基于变分模态分解的风电机组故障诊断。针对滚动轴承早故障信号微弱难以提取出有效信息的问题,提出基于VMD、AR模型以及奇异值分解的特征提取方法;使用实际风场数据的仿真研究中,通过与EMD分解方法进行对比,分析了VMD在抑制模态混叠方面的优越性;针对在对振动信号进行分类时特征向量维数过高的问题,提出基于FCM和KPCA的分类方法,使用KPCA进行数据降维处理,降维处理能够提高分类算法的有效性;最后使用凯斯实验室数据进行实验分析,工况的不同明显影响着故障诊断的精度,对振动信号数据进行工况划分,在相应的工况下进行故障诊断可以提高故障诊断的精确性。
[Abstract]:In recent years, the rapid growth of wind power industry has led to frequent failures of wind turbines and continuous increase in operating and maintenance costs. It is of great significance to strengthen the effective monitoring of the operating state of wind turbines and to realize the early warning and diagnosis of wind turbine faults for ensuring the safe and stable operation of the units and reducing the maintenance expenses. At the same time, due to the variability and complexity of the operating conditions of large-scale wind turbines, it is difficult to accurately evaluate the operating state of wind turbines. On-line monitoring of wind turbines based on condition identification has gradually become an important research direction in the development of wind power. In this paper, the operating conditions of the wind turbine are analyzed and identified, and then the gearbox and the rolling bearing of the wind turbine are taken as the research objects, starting from the two data sources of SCADA parameters and vibration parameters. Based on the multivariate state estimation model and variational mode decomposition method, the corresponding fault warning model and fault diagnosis method are studied, which are divided into the following aspects: firstly, the operating characteristics of wind turbine are analyzed, and the data of wind farm SCADA system are used. After data analysis and preprocessing, appropriate parameters are selected and fuzzy clustering algorithm is used to partition the operating sub-conditions of wind turbine. Secondly, the wind turbine fault warning based on multivariate state estimation is studied. The typical fault of the gearbox is studied and analyzed, the principle of MSET modeling is introduced in detail, the selection of MSET modeling variables is studied, the performance of the model is verified by the actual wind field data, and the conclusion is drawn through the comparison and analysis of experiments. Condition identification can reduce the rate of misjudgment of abnormal points and thus the rate of false alarm. Finally, the fault diagnosis of wind turbine based on variational mode decomposition is studied. Aiming at the problem that it is difficult to extract effective information from the weak fault signal of rolling bearing, a feature extraction method based on VMD-AR model and singular value decomposition is proposed, which is compared with EMD decomposition method in the simulation study of actual wind field data. This paper analyzes the superiority of VMD in suppressing modal aliasing, aiming at the problem that the dimension of eigenvector is too high when classifying vibration signals, proposes a classification method based on FCM and KPCA, and uses KPCA to reduce the dimension of data. Dimensionality reduction can improve the validity of classification algorithm. Finally, case laboratory data are used for experimental analysis. Different working conditions obviously affect the accuracy of fault diagnosis, and the vibration signal data are divided into working conditions. Fault diagnosis can improve the accuracy of fault diagnosis.
【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】