沙漠与绿洲气象2026，Vol.20Issue(3)：49-56,8.DOI:10.12057/j.issn.2097-6801.2412.11108
基于二分类Logistic和多元回归的新疆电线积冰及厚度预报研究
Prediction Study of Wire Icing and Its Thickness in Xinjiang Based on Binary Logistic and Multiple Regression
摘要
Abstract
Utilizing wire icing and related meteorological data from Xinjiang between 1985 and 2015,this study established a wire ice accretion model based on multiple regression methods.The main conclusions are as follows:(1)The number of wire icing days in Xinjiang is generally higher in the northern region than in the southern region,and greater in mountainous areas compared to basins.The area along the Tianshan Mountains(including eastern,western,and central sections)recorded the highest frequency.The maximum standard ice thickness is greatest in northern Xinjiang,followed by western southern Xinjiang.In contrast,the southern edge of the Tarim Basin had the lowest values both in icing days and maximum standard ice thickness across Xinjiang.(2)A binary logistic regression model was used to predict icing occurrence,achieving an identification accuracy of 89.49%.A separate multiple regression model was applied to estimate ice thickness,yielding a mean absolute error of 0.98 mm for typical icing events.Both results indicate robust model performance.(3)Return period calculations revealed that ice thickness is generally larger in northern Xinjiang,the Tacheng area,the central and western Tianshan Mountains,and the western part of southern Xinjiang,while it is relatively smaller in the Tarim Basin and its surrounding areas,as well as most of the Eastern Xinjiang.关键词
电线积冰/二分类Logistic回归/多元回归/耿贝尔分布Key words
wire icing/binary logistic regression/multiple regression/Gumbel distribution分类
天文与地球科学引用本文复制引用
彭艳梅,潘新民,张新军,肯巴提∙波拉提,杨霰,肖静..基于二分类Logistic和多元回归的新疆电线积冰及厚度预报研究[J].沙漠与绿洲气象,2026,20(3):49-56,8.基金项目
中国铁路乌鲁木齐局集团有限公司重点课题(2023-kj-48) （2023-kj-48）
新疆维吾尔自治区气象局重点项目(ZD202309) （ZD202309）
