Non-Isothermal Fluid Flow Reservoir Simulation Using Dumux Software

Geothermal energy has received global attention for being a clean, sustainable and cheap energy source. Numerical modeling is currently an integral part of all the stages of geothermal processes such as exploration and development. In this 3D simulation study, a low-temperature geothermal doublet has been considered. In such system, hot water is produced from one well (producer) and cooled water whose heat content has been depleted by a surface heat exchanger is re-injected underground through another well (injector). This work tests the practicality of employing the Finite-Volume based Dumux simulator for modeling the development of the cooled-water thermal front over the years. Dumux is a relatively new multipurpose open-source simulator that is based on the C++ programming language. The Dumux simulation outcome has been validated by a comparison against that of the Finite-Difference based commercial reservoir simulator ECLIPSE. The computational performances of both simulators have also been analyzed. Two major scenarios were adopted: a convection-dominated scenario in which only the geothermal aquifer layer was modeled and a convection plus conduction scenario in which the caprock and bedrock layers were also modeled. The inclusion of the caprock and bedrock layers into the model geometry accounts for the added thermal conduction effects across those layers. The outlining of two non-isothermal scenarios was meant to evaluate whether the added computational cost of modeling the caprock and bedrock layers was justified. The consistency of the Dumux solution for a varying maximum allowed time step size was investigated. Furthermore, the sensitivity of the simulation variables to different injection and production rate schemes was examined. The studied simulation variables are pressure and temperature as primary variables in addition to water viscosity as a secondary variable. Both temporal and spatial variations of variables were inspected according to the modeled scenario.