期刊文献+

基于GA_XGBoost的山区公路风吹雪灾害降维评估模型

Dimensionality reduction assessment model for blowing snow disasters on mountain highways based on GA_XGBoost
原文传递
导出
摘要 新疆G218线那拉提至巴仑台段(那巴高速公路)位于天山中部地区,区域风吹雪灾害发生频率高且危害程度大,风吹雪危险度评估对于公路雪灾防治十分关键。然而,目前山区公路风吹雪危险度预测指标选取繁杂,直接应用于工程设计可操作性不高。本研究依托那巴高速气象站数据,结合再分析雪深数据和其他辅助数据,获取了影响风吹雪危险度的13个指标;基于这些指标构建了高等级公路风吹雪危险度预测模型GA_XGBoost,该模型通过Lasso特征筛选得到积雪深度、风路夹角、海拔、植被覆盖率、曲线半径等5个高贡献度致灾指标,使用基于遗传算法的超参数优化方法搜索XGBoost模型的最优解,构建了风吹雪危险度预测模型。最后,以那巴高速调研实测数据对模拟结果进行验证。验证结果表明:(1)目前模型预测的风吹雪危险度准确,提出的风吹雪评估模型优于传统模型;(2)那巴高速风吹雪危险度分布较为复杂,模型计算和现场调查均表明,高危险度核心区位于艾肯达坂和查汗努尔达坂,区域盛行以西风为主的风吹雪。 The Narati-Baluntai section of the G218 Highway(Naba Highway)in Xinjiang is located in the cen⁃tral region of Tianshan Mountain.The region experiences frequent and severe blowing snow disasters,making risk level assessment critical for the prevention and control of highway snowstorms.However,current blowing snow risk level prediction methods for mountain highways suffer from overly complex indicator selection,result⁃ing in low operational applicability for engineering design.This study utilized meteorological station data from Naba Highway,combined with reanalysis of snow depth data and other auxiliary data,to identify 13 indicators influencing blowing snow risk level.Based on the above indicators,this study established the GA_XGBoost pre⁃diction model for blowing snow risk level on highways.This model employed Lasso feature selection to identify five high-contribution disaster-inducing indicators,including snow depth and wind-road angle.Then,using ge⁃netic algorithm-based hyperparameter optimization,this study searched for the optimal solution of the XGBoost model to establish the wind-blown snow risk level prediction model.Finally,the simulation results were validat⁃ed using field data from the Naba Highway.The validation results showed that:(1)the blowing snow risk level predicted by the proposed model was accurate,and the model outperformed traditional models.(2)The distribu⁃tion of blowing snow risk levels on the Naba Highway was relatively complex.Both model calculations and field surveys indicated that the core high-risk zones were located at Aiken Daban and Chahan Nuur Daban,where westerly winds dominated blowing snow.
作者 马磊 胡智轩 韦振勋 李杰 李肽脂 吴燕 赵静 MA Lei;HU Zhixuan;WEI Zhenxun;LI Jie;LI Taizhi;WU Yan;ZHAO Jing(Xinjiang Academy of Transportation Sciences Co.,Ltd.,Urumqi 830000,China;Xinjiang Naba Expressway Development Co.,Ltd.,Korla 841000,Xinjiang,China;Chang’an University,Xi’an 710064,China)
出处 《冰川冻土》 2025年第4期1112-1124,共13页 Journal of Glaciology and Geocryology
基金 新疆维吾尔自治区自然科学基金项目(2023D01B25) 科技部重大专项(2022xjkk060203) 新疆维吾尔自治区重点研发计划项目(2023B03004-1)资助作。
关键词 风吹雪 XGBoost 遗传算法 特征筛选 灾害评估 blowing snow XGBoost genetic algorithm feature selection disaster assessment
  • 相关文献

参考文献19

二级参考文献361

共引文献105

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部