风电机组运行工况辨识与变桨系统故障诊断

发布时间：2018-01-07 06:35

本文关键词：风电机组运行工况辨识与变桨系统故障诊断　出处：《沈阳工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着风力发电技术的高速发展,风电机组的单机容量越来越大。然而,随着风电场的大规模建设,风力发电机组的运行维护费用高和故障率高等问题也凸显了出来,如何提高风电机组运行可靠性及利用率成为风力发电急需解决的问题。变桨系统是风电机组的重要组成部分,但由于其运行环境恶劣、组件繁多、启停频繁,导致故障频发,本文在分析风电机组及其变桨系统工作原理的基础上,利用风电机组历史运行数据信息及变桨系统故障信息,研究基于SCADA数据的风电机组运行工况辨识和变桨系统运行状态故障诊断方法,其研究内容主要包括:1)简述直驱风电机组及电动变桨系统以及SCADA系统的构成和工作原理,针对风电机组运行参数进行分析,并对SCADA监控系统所采集的历史数据进行数据预处理。采用基于信息熵的特征参数相关性分析对变桨系统运行参数进行分析。2)将风电机组运行参数能量控制模式和限功率标志作为分类特征参数,提出了基于自组织神经网络的混合属性聚类方法进行风电机组运行工况划分,该算法以自组织特征映射神经网络为框架,采用基于样本概率的异构值差度量风电机组运行参数混合属性数据的相异性。利用分类特征项在Voronoi集合中出现频率作为分类属性数据参考向量更新规则的基础,通过混合更新规则实现数值属性和分类属性数据规则的更新。在此基础上,在自组织神经网络的竞争层后增加一层输出层,使其变为有监督的分类学习网络,提出有监督的混合属性数据自组织映射分类模型,实现风电机组运行的工况辨识。3)针对风电机组电动变桨系统故障诊断问题,对风电变桨系统进行特征属性筛选,建立在不同工况下的风电变桨系统异常识别模型,该模型以主元分析为基础,将电动变桨系统运行数据投影到主元模型的主元子空间和残差子空间上,通过判断其对应的T2和SPE是否超出对应的控制限来进行异常识别;同时根据贡献图法找出异常的特征属性;最后通过故障子空间理论进行基于SPE的故障重构,实现风电变桨系统故障诊断。
[Abstract]:With the rapid development of wind power generation technology, the single unit capacity of wind turbine is increasing. However, with the large-scale construction of wind farm. The problems of high operating and maintenance cost and high failure rate of wind turbine are also highlighted. How to improve the operational reliability and utilization ratio of wind turbine has become an urgent problem for wind power generation. Variable propeller system is an important part of wind turbine, but because of its poor operating environment, various components, frequent start and stop. On the basis of analyzing the working principle of wind turbine and its variable propeller system, this paper makes use of the historical operation data of wind turbine and the fault information of variable propeller system. The method of wind turbine operating condition identification and fault diagnosis of variable propeller system based on SCADA data is studied. The main research contents include: (1) introduce the composition and working principle of direct-drive wind turbine, electric variable propeller system and SCADA system, and analyze the operating parameters of wind turbine unit. The historical data collected by the SCADA monitoring system are preprocessed and the operational parameters of the propeller system are analyzed by the correlation analysis of characteristic parameters based on information entropy. The energy control mode and the limited power mark of the wind turbine operating parameters are taken as the classification characteristic parameters. A hybrid attribute clustering method based on self-organizing neural network is proposed to partition the operating conditions of wind turbine. The algorithm is based on self-organizing feature mapping neural network. The heterogeneity value difference based on sample probability is used to measure the heterogeneity of the mixed attribute data of wind turbine operating parameters. The frequency of classification feature in Voronoi set is used to update the reference vector of classification attribute data. The basis of the rules. The data rules of numerical attributes and classification attributes are updated by mixed updating rules. On this basis, a layer of output layer is added after the competition layer of self-organized neural networks to make it a supervised classification learning network. A supervised self-organizing mapping classification model of hybrid attribute data is proposed to realize the operating condition identification of wind turbine. 3) to solve the problem of fault diagnosis of electric variable propeller system of wind turbine. The characteristic attributes of wind turbine variable propeller system are screened, and the abnormal identification model of wind power variable propeller system under different working conditions is established. The model is based on principal component analysis. The operation data of the electric propeller system are projected onto the principal subspace and residual subspace of the principal component model, and the abnormal identification is carried out by judging whether the corresponding T2 and SPE exceed the corresponding control limit. At the same time, according to the contribution diagram method to find out the characteristic attributes of the anomaly; Finally, fault reconstruction based on SPE is carried out by fault subspace theory, and fault diagnosis of wind power variable propeller system is realized.
【学位授予单位】：沈阳工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM315

【参考文献】