水力发电2025，Vol.51Issue(6)：61-68,8.
基于LightGBM-IGWO的深厚覆盖层地基渗透系数智能反演分析
Intelligent Inversion of Permeability Coefficient for Deep Overburden Foundation Based on LightGBM-IGWO
摘要
Abstract
The geological permeability coefficient is an important factor for accurately analyzing seepage in hydraulic engineering and has a significant impact on the design and safety assessment of projects.However,traditional inversion methods often have limitations in computational efficiency and accuracy when dealing with complex geological conditions and large-scale data.To address this issue,this research proposes an intelligent permeability coefficient inversion method based on the Light Gradient Boosting Machine(LightGBM)and an improved Grey Wolf Optimization(IGWO)algorithm.The method first combines a finite element forward model with orthogonal experimental design to generate an inversion sample set,and then uses LightGBM to construct a seepage calculation surrogate model.On this basis,the IGWO algorithm,enhanced with the Levy flight strategy,is introduced to improve the efficiency and accuracy of searching for the optimal permeability coefficient.The method is validated through a hydropower station project,located in a deep overburden layer.The results show that the LightGBM model performs better in borehole water level predictions,with the predicted values closely matching the measured data.The maximum absolute error of borehole water level is 8.29 m,and the relative error is only 0.27%,meeting the precision requirements for engineering applications.Furthermore,the simulated natural seepage field distribution aligns with the typical seepage pattern of mountainous areas,further demonstrating the reliability and practical application value of the model in geological permeability coefficient inversion.关键词
渗透系数/智能反演/有限元模型/莱维飞行策略/LightGBM/IGWOKey words
permeability coefficient/intelligent inversion/finite element model/Levy flight strategy/LightGBM/IGWO分类
建筑与水利引用本文复制引用
唐杰,陈新根,王安城,殷乔刚..基于LightGBM-IGWO的深厚覆盖层地基渗透系数智能反演分析[J].水力发电,2025,51(6):61-68,8.基金项目
国家自然科学基金资助项目(51909215) （51909215）
中国博士后科学基金项目(2021T140554,2020M683527) （2021T140554,2020M683527）
