基于分裂-合并策略改进多特征聚类算法的风电机组故障分析  

Fault analysis of wind turbine based on improved multi feature clustering algorithm based on split merge strategy

作　　者：梁耘 王维庆[1] 王海云[1] 

机构地区：[1]新疆大学电气工程学院教育部可再生能源发电与并网控制工程技术研究中心,新疆乌鲁木齐830047

出　　处：《可再生能源》2017年第10期1537-1543,共7页Renewable Energy Resources

基　　金：新疆维吾尔自治区重点实验室(2016D03021);国家自然科学基金(51267017)

摘　　要：大数据分析是大数据技术的主要应用,对大数据集进行多维度分析的多特征聚类技术越来越受到关注。针对风电机组重复故障,提出了一种分裂-合并策略改进多特征聚类算法技术的大数据处理方法。选用LabVIEW分析工具,利用改进多特征聚类算法对国内某风电场33台1.5 MW机组2015年全年的故障月报数据中的偏航系统故障进行了聚类分析,对原始数据与剔除重复故障后的数据进行统计,分别计算了故障发生频次与MTBF两项可靠性指标,通过对计算结果进行对比分析可知,由于重复故障的存在,使得利用原始数据统计得到的可靠性指标与真实值存在较大偏差,经过聚类分析的故障数据可以大大提高分析结果的准确性,更能反映机组真实的可靠性水平。Big data analysis is the main application of big data technology,and multi-feature clustering technology for big data sets is attracting more and more attention. Aiming at the repeated faults of wind turbine,a new big data processing method based on multi-feature clustering algorithm is proposed. The LabVIEW is used in this paper,using the improved multi-feature clustering algorithm,analyze the fault data of the yaw system of 33 sets of 1.5 MW unit in 2015. The original data and repetition of the fault are statistical,then fault occurrence frequency and MTBF are calculated respectively. By the comparison and analysis of the results,it is found that the reliability index obtained by the original data has large deviation from the real value due to the repetition of the fault. It can be seen that the fault data through cluster analysis can greatly improve the accuracy of the analysis results,and can better reflect the true reliability of the unit level.

关 键 词：LABVIEW 风力发电机组 故障 大数据 聚类算法 

分 类 号：TK81[动力工程及工程热物理—流体机械及工程]