基于密度聚类算法的风电机组异常数据点筛选  

Screening of abnormal data points of wind turbine based on density clustering algorithm

作　　者：王克挺 WANG Keting(CHN Energy Gansu Electric Power Co.,Ltd.,Lanzhou 730010,China)

机构地区：[1]国家能源集团甘肃电力有限公司,甘肃兰州730010

出　　处：《电子设计工程》2024年第18期127-131,共5页Electronic Design Engineering

基　　金：国网甘肃省电力公司科技项目(53262825001B)。

摘　　要：风电机组异常数据点筛选受到大规模高维噪声数据干扰,导致数据点筛选结果不全面。为精准筛选风电机组异常数据点,提出了基于密度聚类算法的风电机组异常数据点筛选方法。根据风电机组异常数据特征密度聚类,将多维向量空间中的数据形式化为特征值邻域,避免高维噪声影响异常数据点筛选过程。计算邻域半径和邻域密度,以反映数据分布紧密程度,确定密度低的点为噪声点。采用云分段最优熵算法,分析风速、功率数据样本关系,并计算信息熵。将样本熵计算结果输入到云发生器中,获取熵所在云序列坐标点,实现异常数据点筛选。由实验结果可知,所提出方法能够精准筛选出1号和2号风电机组异常数据点,为风电机组的安全运行提供精准数据。The screening of abnormal data points of wind turbines is disturbed by large-scale high-dimensional noise data,resulting in incomplete screening results of data points.In order to accurately screen abnormal data points of wind turbines,a method based on density clustering algorithm is proposed to screen abnormal data points of wind turbines.According to the feature density clustering of wind turbine abnormal data,the data in multi-dimensional vector space is formalized into eigenvalues neighborhood to avoid the influence of high-dimensional noise on the screening process of abnormal data points.Calculate the neighborhood radius and neighborhood density to reflect the tightness of data distribution,and determine the low density points as noise points.The cloud segmentation optimal entropy algorithm is used to analyze the sample relationship of wind speed and power data,and calculate the information entropy.Input the sample entropy calculation result into the cloud generator,obtain the cloud sequence coordinate point where the entropy is located,and realize the abnormal data point screening.According to the experimental results,the proposed method can accurately screen the abnormal data points of No.1 and No.2 wind turbines,and provide accurate data for the safe operation of wind turbines.

关 键 词：密度聚类 风电机组 异常数据点 云分段最优熵 数据点筛选 

分 类 号：TN914[电子电信—通信与信息系统]