Abstract: The thermal conductivity of geotechnical media is a key indicator in the research and practice of surface environment and geothermal energy theory.The study aims to establish a model for the random distribution of three-dimensional porous media in geotechnical soils to improve the prediction accuracy of thermal conductivity and reduce the difficulty and workload of testing.Combined with the properties of geotechnical porous media,the percentage of each phase in the geotechnical soil is determined using a low-field NMR analyzer.After a three-dimensional stochastic model for geotechnical porous media was established using programming software and loaded into the simulation software,the NMR data and material parameters were entered to calculate the thermal conductivity.Then,the portable thermal property analyzer was used to measure the thermal conductivity of soils with different water contents and analyze the variation relationship between water content and thermal conductivity.Finally,the validity of the established random distribution model for porous media is verified by comparing it with the experimental data.The results show that the pore water in the geotechnical soil is an important indicator of thermal conductivity and that the thermal conductivity increases with water content.The calculated value of the stochastic model is consistent with the measured value of the instrument.The average PBIAS value between the calculated and measured values is 0.0139,and the maximum PBIAS value is 0.021.The purpose of this paper is to provide a reliable way to predict the thermal conductivity of geotechnical soils.