基于WGAN-GP的高风速区概率风谱建模  

Modelling of probabilistic wind spectra in high wind speed domain based on the WGAN-GP

作　　者：刘芸[1] 王浩[1] 林禹轩 LIU Yun;WANG Hao;LIN Yuxuan(Key Laboratory of Concrete and Prestressed Concrete Structures of Ministry of Education,Southeast University,Nanjing 211189,China)

机构地区：[1]东南大学混凝土及预应力混凝土结构教育部重点实验室,南京211189

出　　处：《东南大学学报(自然科学版)》2025年第2期364-369,共6页Journal of Southeast University:Natural Science Edition

基　　金：国家自然科学基金资助项目(52338011);中央高校基本科研业务费专项资金资助项目(2242022k30030,2242022k30031);中国留学基金委国家奖学金资助项目(202306090285).

摘　　要：针对实测高风速样本匮乏、难以准确捕获强/台风概率信息的问题,提出了基于带有梯度惩罚的Wasserstein距离生成对抗网络(WGAN⁃GP)的高风速区概率风谱建模方法。以江阴长江大桥2019—2020年的高风速实测数据为基础,将平均风速与风谱参数同时作为WGAN⁃GP的输入变量,生成符合实测数据分布规律的高质量新样本,扩充了高风速样本数据集,并建立了高风速区概率风谱模型。结果表明,高风速区间的谱参数服从对数正态分布,其中标准差σ_(u)及无量纲参数A_(u)和B_(u)的均值分布参数与平均风速线性正相关,Au和Bu的标准差分布参数与平均风速线性正相关,σ_(u)的标准差分布参数与平均风速线性负相关。To solve the problem of the lack of measured high wind speed samples and the difficulty of accurate capture of statistical probability information for strong winds or typhoons,a method for constructing probabi⁃listic wind spectra model in high wind speed domain is proposed based on the Wasserstein generative adver⁃sarial network with gradient penalty(WGAN⁃GP).Based on the measured high wind speeds between 2019 to 2020 in Jiangyin Bridge,the measured mean wind velocity and wind spectral parameters are taken as the input variables in the WGAN⁃GP.The new high⁃quality generated data conformed to the distribution law of mea⁃sured input variables are obtained,and the samples in the high wind speed domain are enhanced.The probabi⁃listic wind spectra model in the high wind speed domain is constructed.The results show that the spectral pa⁃rameters in the high wind speed domain obey lognormal distribution.The mean value distribution parameters of the standard deviationσ_(u),non⁃dimensional parameters A_(u) and B_(u) have the linear positive correlation with the mean wind speed.The standard deviation distribution parameters of Au and Bu show linear positive correlation with the mean wind speed,and the standard deviation distribution parameters ofσ_(u)have the linear negative correlation with the mean wind speed.

关 键 词：概率风谱 生成对抗网络(GAN) 强/台风 实测 大跨度桥梁 

分 类 号：TU312.1[建筑科学—结构工程]