Reducing geothermal energy recovery costs and environmental concerns has led to interest in modeling new borehole heat exchanger designs. These heat exchangers are environment-friendly because they inject all the geofluid back into the reservoir after harvesting its energy, resulting in zero mass withdrawal from the system. This study develops reduced-order reservoir models for quickly estimating production temperature and thermal recovery from borehole heat exchangers using inspectional analysis and predictive modeling. An inspectional analysis suggests that there are nineteen dimensionless numbers necessary to fully scale this design. These dimensionless numbers were used in the statistical modeling. A Box-Behnken experimental design was used to create the simulation runs. A rigorous sensitivity analysis was performed on these designed series of runs to identify the most important dimensionless groups in predicting the dimensionless response for different segments of time. The selected dimensionless groups along with the dimensionless production temperature and thermal recovery factor (response), were used in the regression models. The models were reduced and assessed using training and testing runs. Applications of the developed models are also presented. The approach presented in this paper is general and provides a means to interpret and develop predictive models from simulator outputs in other research areas.