摘要
利用南方地区多个气象站和电力部门观冰站的导线覆冰逐日冰厚资料,将广义极值分布和广义帕雷托分布引入导线覆冰的概率模型研究中,通过超门限覆冰次数的泊松分布拟合检验,结合H ill图解,提出了基于超门限峰值法门限值的确定方法;对两种分布在导线覆冰极值模型拟合的适用性研究表明,广义帕雷托分布对各站覆冰冰厚极值的拟合精度最高;重现期冰厚极值估计随样本长度的变化分析表明,广义帕雷托分布模型极值估计的稳定性比广义极值分布强,一般样本容量达到25 a左右时,广义帕雷托分布重现期冰厚极值的估计趋于稳定,可以作为短序列下估计导线覆冰极值的较好方法。
The generalized extreme value distribution（GEV） and generalized Pareto distribution（GPD） are used to study the conductor icing probability model with the daily standard icing thickness data in the south of China.By the simulation test of Poisson distribution of the number of exceedance events and Hillplot,the method of identifying threshold is proposed based on Peaks Over Threshold（POT）.Numerical simulations show that GPD shows the best accuracy in simulating extremes of each site.The change of icing thickness extremes of Return period under different sample lengths demonstrates that GPD has a better stability than GEV.When the sample length is about 25 years,the estimation of GPD tends to be stable which could be considered as a good method for short array data.
出处
《大气科学学报》
CSCD
北大核心
2010年第4期385-394,共10页
Transactions of Atmospheric Sciences
基金
西南电力设计院委托项目"输电线路覆冰数学模型研究"
关键词
广义极值分布模型
广义帕雷托分布模型
导线覆冰
重现期
generalized extreme value distribution（GEV）
generalized Pareto distribution（GPD）
conductor icing
return period
