Designing an efficient system to extract heat from an enhanced geothermal system (EGS) requires proper understanding of the behavior of the reservoir over a long period. Five key parameters namely well spacing, fracture spacing, well inclination angle, injection temperature and injection rate are considered in this study for a doublet well system. To study and evaluate the performance of an EGS, second order surrogate models for 'produced water temperature', at certain time intervals are developed as a function of these five factors. The in-situ properties of a candidate reservoir for designing the simulations are taken from the FORGE site, Utah. Simulations are designed using 'Box-Behnken' design of experiments techniques to minimize the number of simulations. The models are trained and tested with the simulated results. Fitness of the models is calculated by estimating the errors using the coefficient of determination (R-2) and the normalized root mean square error (NRMSE). These surrogate models are used to study the sensitivity of the aforementioned factors on the temperature of the produced water and the heat recovery over a time period of 30 years. Finally, the hierarchy of factors, as they impact the total heat recovery are represented as a tornado plot. (C) 2018 Elsevier Ltd. All rights reserved.