论文部分内容阅读
提出了一种改进的自适应重要抽样方法,以广义极值分布为例,引入L-矩法,建立样本统计特性与分布参数的联系,估算极限事件的发生概率.以浙江省云港流域的24h设计暴雨为例,计算金竹岭和仙人潭两个站点降雨量分别大于213mm和200mm的概率.计算结果表明改进的自适应重要抽样方法能很好地模拟水文极限事件,叠代次数随着抽样个数的增加逐渐减小.与常规的MC法比较,重要抽样的效率有显著提高.另外,此改进的自适应重要抽样方法还能推广到其他的分布函数.
An improved adaptive importance sampling method is proposed. Taking the generalized extreme value distribution as an example, the L-moment method is introduced to establish the relationship between the statistical characteristics of samples and the distribution parameters and estimate the occurrence probability of extreme events. 24 h design torrential rain as an example to calculate the probability of rainfall greater than 213mm and 200mm respectively at the two sites of Jinzhuling and Xianren Lake.The calculated results show that the modified adaptive important sampling method can well simulate the hydrological extreme events, Compared with the conventional MC method, the efficiency of important sampling has been significantly improved.In addition, this improved adaptive important sampling method can also be extended to other distribution functions.