The prediction accuracy of soil thermal conductivity is crucial to the design of geothermal-related structures such as ground source heat pump and borehole heat. Based on the complex, random and self-similar microstructure characteristics of soil, by using fractal theory to consider the effects of soil porosity, particle arrangement and other parameters on soil curvature, a capillary physical model and a corresponding mathematical model of effective thermal conductivity of three-phase soil were established. The predicted results of the model were compared with existing experimental data. The results show that the model with the addition of soil particle arrangement and other parameters has higher accuracy in predicting the soil thermal conductivity with porosity between 0.21 ~ 1, and the three-phase soil thermal conductivity with different porosity is related to the soil particle arrangement blocking parameters. The larger the porosity and the smaller the blocking parameters, the closer the model prediction is to the experimental results. The bending degree has a direct influence on the series-parallel ratio of the thermoelectric ratio model. The influence of soil particle arrangement on the bending degree decreases with the increase of porosity, and the bending degree is inversely proportional to the change trend of the three-phase soil thermal conductivity. The relevant results are helpful to the research of the three-phase soil heat transfer efficiency of buried pipe and outside pipe.