The development of enhanced geothermal systems using CO2 (CO2-EGS) is a promising idea for expanding geothermal energy production (especially in areas with scarce water resources) when large supplies of captured anthropogenic CO2 may be available in the future. Implementing this concept relies on replacing the natural geothermal brine in the reservoir with injected CO2 to achieve enhanced energy recovery, and raises the questions of the fate of dissolved salts in the brine as CO2 dries out the system, and how any precipitated salt could affect fluid flow. In this case, a new TOUGH2 equation of state module (ECO2H) was used to simulate CO2 injection in an EGS with a brine system comprised of H2O and NaCl. This so called CO2-EGS reservoir is at a depth of 3.5-4.5 km with normal pressure (hydrostatic) and temperature (160-200 degrees C) gradients. A classic "five-well" geometry is assumed in our 706 m x 706 m x 1 km block, of which only one eighth of the area needs to be modeled due to symmetry. The fractured EGS reservoir was modeled using the multiple interacting continua (MINC) conceptual model with fracture spacing of 10 m. Dry CO2 was injected at the bottom of the initially brine-saturated reservoir and hot fluids were produced from the top of the reservoir. Simulations show that the brine contained in the fractures is produced initially, and only a few weeks later, the CO2 plume breaks through at the production well. The two-phase nature of flow at this time causes a reduction in flow rate. Fluid production increases again as the reservoir dries out and the injected CO2 fills the fractures (and more slowly the matrix). As the produced fluid becomes single-phase CO2, energy production is enhanced. For salt mass fractions of the order of 0.01 (salinity of 10,000 ppm), total heat produced during the lifetime of the well (about 6 years) is 270% more than that achievable with H2O as the working fluid. This result is probably at the lower end of what had been previously suggested by Randolph and Saar (2011). Simulation results show that as the brine is driven out of the matrix by capillary pressure, H2O evaporates into the CO2 plume and salt precipitates in the fractures clogging up the flow system. At the highest salt mass fraction modeled here (0.15), enhanced energy production is inhibited by halite precipitation in the fractures. Our simulations suggest that for low-salinity systems, significant clogging occurs close to the production well after less than 10 years, while at high salinities clogging occurs close to the injection well in less than one year. Even though clogging of the reservoir is an apparently inevitable consequence of the drying of the saline geothermal reservoir, the fact that clogging occurs in specific reservoir regions could imply that remediation strategies could be developed to mitigate clogging. Published by Elsevier Ltd.