脉动风速随机傅里叶谱模型的参数聚类分析

Cluster Analysis of Parameter of Stochastic Fourier Spectrum for Fluctuating Wind Speeds

摘要:

脉动风速随机Fourier谱模型为脉动风速模拟提供了一条基于物理的建模途径.为了实现对这一模型基本参数概率分布的合理建模，阐述了模型基本参数的物理含义，利用实测风速记录识别了基本参数的概率分布.分析发现，分界波数的统计分布出现典型的双峰形态；由原始识别结果的统计分布导出的功率谱与Kaimal谱在中、低频段有较大差别.深入研究表明，分界波数有显著的聚类特征.基于此，采用kmeans函数聚类方法分析了分界波数的数据结构，指出上述异常现象与参数的聚类结果密切相关.根据聚类结果，合理地筛选实测风速记录，进行了分界波数的建模.

Abstract:

The stochastic Fourier spectrum provides a physical based perspective for fluctuating wind simulation. To properly model the probability density function of the elemental variables, the basic physical meaning of the elemental variables is elaborated and the statistical distributions of them are identified from the measurements. It is revealed that the statistical distribution of the cutoff wave number displays an obvious bimodal pattern; the power spectrum density derived from the fitted distribution of the original identifications deviates from the Kaimal spectrum evidently in the low and middle frequency band. Besides, the identifications of the cutoff wave number are eminently characterized by a clustering phenomenon, based on which the kmeans cluster analysis is utilized to reveal the underlying structure of the data set. It is pointed that the cluster characteristic of the fluctuating wind speed sample is closely related to the above abnormal phenomenon. Finally, the distribution modeling for the cutoff wave number is conducted for the reasonably selected measurements based on the cluster results.