计及运行状态的直驱风力发电机早期故障诊断研究

发布时间：2018-06-12 20:50

本文选题：变工况 + 永磁风力发电机　；参考：《新疆大学》2017年硕士论文

【摘要】：新疆地区风能资源丰富,是我国重要的风电基地。该地区风电机组要承受高温、冰冻、风沙等极端环境,在其设计寿命周期内故障频发。风电机组机舱常安装于50~80m的高空,设备维护困难,且故障诊断结果受风速、载荷、并网运行状态影响较大,诊断精度不高。因此,研究计及并网运行状态的风电机组监测与故障诊断技术意义重大。本论文针对永磁直驱型风力发电机早期故障不易监测和诊断的难题,综合设备的电流和振动信息,在正常、风速突变、电网电压不平衡三种工况下对永磁直驱风力发电机正常运行(视为特殊故障)、主轴偏心、主轴轴承磨损三种故障进行诊断。具体内容如下:(1)对新疆地区风电机组故障状况进行了初步调研,对比了我国新疆地区和瑞典的风电机组故障特点。分析了主轴偏心和轴承磨损的故障类型及故障特性,定性分析上述故障对振动信号和电流信号的影响。对故障诊断方法进行梳理,对比分析诊断模型的优缺点,选定SVM作为初步诊断模型。(2)开发了风电机组多源信息采集系统,搭建了变工况风电机组故障检测试验台。实验采集了永磁直驱型风力发电机在正常、风速突变、电网三相电压不平衡三种工况下正常运行(视为特殊故障)、主轴偏心、主轴轴承磨损三种故障的定子电流和径向振动信号。研究了振动和电流信号的故障特征提取方法,建立了故障特征样本库,并引入马氏距离对故障特征的可分性进行评价。(3)利用logistic回归函数和逐对耦合方法改进了传统的SVM诊断模型,使得SVM能同时输出故障类型和故障概率。分别基于振动特征和电流特征建立诊断模型,在正常、风速突变、三相不平衡三种工况下对发电机故障进行初步诊断,并对初步诊断结果进行评价。(4)改进了传统D-S证据理论,利用可靠性矩阵,建立了证据可靠性系数和融合权重之间的关系。综合诊断模型的泛化能力和诊断效果,对初步诊断结果赋予不同的权重并进行融合诊断。相比基于单一征兆的初步诊断方法,加权融合后,错误诊断结果得到修正,诊断精度得到明显提高。
[Abstract]:Xinjiang is rich in wind energy resources and is an important wind power base in China. Wind turbines in this area have to withstand extreme environments such as high temperature, freezing, wind and sand, and frequent failures occur during their design life cycle. The engine room of wind turbine is usually installed at an altitude of 50m and 80m, the maintenance of the equipment is difficult, and the result of fault diagnosis is greatly affected by wind speed, load, grid-connected operation state, and the diagnostic accuracy is not high. Therefore, it is of great significance to study the monitoring and fault diagnosis technology of wind turbine taking into account the grid-connected operation state. Aiming at the difficult problem of early fault monitoring and diagnosis of permanent magnet direct-drive wind turbine, this paper synthesizes the current and vibration information of the equipment in normal, abrupt wind speed. The fault diagnosis of permanent magnet direct drive wind turbine (PMSG) is made under three working conditions, which are considered as special fault, spindle eccentricity and spindle bearing wear. The main contents are as follows: (1) A preliminary investigation is carried out on the fault status of wind turbines in Xinjiang, and the fault characteristics of wind turbines in Xinjiang and Sweden are compared. The fault types and characteristics of spindle eccentricity and bearing wear are analyzed, and the effects of these faults on vibration signal and current signal are analyzed qualitatively. The fault diagnosis method is combed, the advantages and disadvantages of the diagnosis model are compared and analyzed, the SVM is selected as the primary diagnosis model. The multi-source information acquisition system of wind turbine is developed, and the fault detection test-bed of wind power unit under different working conditions is built. The permanent magnet direct-drive wind turbine (PMSG) is tested under three working conditions: normal operation, sudden wind speed, unbalanced three-phase voltage (considered as a special fault and eccentric spindle). Stator current and radial vibration signal of three kinds of fault of spindle bearing wear. The fault feature extraction method of vibration and current signals is studied, and the fault feature sample database is established. And the Markov distance is introduced to evaluate the separability of fault features.) the traditional logistic diagnosis model is improved by using logistic regression function and pairwise coupling method, which enables SVM to output fault types and fault probability at the same time. Based on the characteristics of vibration and current, the diagnosis model is established, and the fault diagnosis of generator is carried out under the normal condition, the sudden change of wind speed and the three-phase unbalance, and the evaluation of the preliminary diagnosis result is made. (4) the traditional D-S evidence theory is improved. The relationship between evidence reliability coefficient and fusion weight is established by using reliability matrix. By synthesizing the generalization ability and diagnostic effect of the diagnostic model, different weights are given to the preliminary diagnosis results and fusion diagnosis is carried out. Compared with the initial diagnosis method based on single symptom, the error diagnosis result is corrected and the diagnostic accuracy is improved obviously after weighted fusion.
【学位授予单位】：新疆大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM315

【参考文献】