The thermal conductivity of rock is an essential thermophysical parameter in geothermal engineering. Experiments have shown that when rock changes from an unsaturated to a saturated state, thermal conductivity can be increased by 50% at most. In this paper, the variation of thermal conductivity with saturation of 101 samples of six different rock types from Songliao Basin, Gonghe Basin, and Ordos Basin in China was measured in the laboratory. Correlations between saturation, porosity, internal structural characteristics of rocks and thermal conductivity were investigated. The results show that the effect of saturation on the thermal conductivity of rock increases exponentially with porosity. Based on three classical models and the relationship between porosity, saturation, and thermal conductivity, improved models and a fitted model of thermal conductivity are proposed. By comparing the improved models to the fitted model, the applicable scope of each model is obtained. The errors of each model are analyzed in detail. The average MAE value of the improved models is 0.405 and the average RMSE value is 1.207 over the applicable model scopes. The R2 of the fitted model is 0.94. Compared with experimental data from literature, MAE values of the improved models and the fitted model are 0.040 and 0.003 respectively. Unsaturated rock with a single fracture heat transfer model is established by simulation. A thermal-Hydrological coupling model is used to simulate seepage heat transfer through fracture. The effect of variable thermal conductivity simulation is the best with the experimental value, and the MAPE value was 8.3%. This study fills a gap in available models of unsaturated thermal conductivity of rock, and provide theoretical support for the accurate calculation of heat transfer efficiencies in related geothermal engineering applications.