基于阈值模型(POT)的广义帕累托分布(GPD)能充分利用样本中较大的观测值,在极值计算上有广泛的应用。本文讨论了广义Pareto分布以及该分布在风压极值计算中的应用。采用极大似然估计方法,对风压样本进行GPD拟合,得到具有一定保证率的风压极值。通过与峰值因子法(PFM)和经典极值方法(GEVD)计算的结果相比较,证明GPD拟合风压数据的效果较好,高保证率下的极值估计合理。对于高斯分布的风压样本,三种方法计算得到的结果比较接近;对于非高斯分布的样本,三种计算方法的结果存在较大差别,以GPD拟合的结果最优。应用GPD方法计算风压极值无需风压分布满足高斯假定,因而具有更广泛的适用性。
This paper introduces the generalized Pareto distribution (GPD) and discusses its application in the a- nalysis of extreme wind pressure. Compared with the generalized extreme value distribution, to build peaks over threshold (POT) model can makes use of all data in the high wind pressure and provides an accurate calculation model with more data. The parameters can be acquired by using maximum likelihood estimation, which is a com- monly recognized method in obtaining the parameters. Based on the fitted distribution, the fractile of certain guar- antee rate is calculated. With comparison of results obtained from the peak factor method and generalized extreme value distribution method, it proves the GPD fits the data well and gets reliable extreme value. In addition, the study confirms that the peaks factor method is not suitable in the field of non-Gaussian distribution. The GPD meth- od does not assume the wind pressure obeys Gaussian distribution, and it performs well in both Ganssian distribu- tion and non-Gaussian distribution. Therefore, the GPD is more suitable in the calculation of extreme value in wind pressure.