Investigating simplified modeling choices for numerical simulation of CO2 storage with thermal effects

Temperature is a key variable for the modeling of geological CO2 storage, and any simulation model must take it explicitly or implicitly into account. In the most basic case, the assumption of reservoir temperature is simply reflected in the particular choice of fluid properties. More advanced models represent temperature as an imposed field that vary with spatial coordinates but remains unchanged in time. Yet more advanced models include thermal fields that evolve in time, obtained from explicit modeling of heat transport based on energy conservation. When modeling geological CO2 storage at large spatial and temporal scales, it is beneficial to employ simplified flow models with significantly lower computational requirements than fully resolved 3D simulations. At the large spatial scales and low flow rates associated with CO2 migration studies, temperature can be reasonably considered as an imposed external field dictated by the geothermal gradient, with CO2 and brine assumed to be in thermal equilibrium with the surrounding rock. Closer to an injection well, the picture is different. CO2 may be injected at a temperature significantly different from that of the aquifer, which leads to an expanding thermal front around the injection well. This local change in temperature not only affects fluid properties, but also geomechanical stresses and the rate of geochemical reactions. For these reasons, an adequate model of temperature evolution should be included in the simulation model. In this article, we examine whether certain simplified approaches used to simulate CO2 storage at the large scale may be adapted and applied to model the regions affected by the thermal front. We do this by comparing the results from upscaled (vertically integrated) flow models extended with heat transport and different choices of overburden representations against highly resolved 3D models. These comparisons are carried out for a number of test cases spanning a wide range of scenarios constructed to minimize or maximize specific characteristics of the coupled flow-thermal system, namely the Peclet number, the gravity number and the amount of thermal bleed. Our results suggest that the thermal front can be reasonably modeled in many practical cases using a vertically integrated flow model with constant vertical temperature. The results also suggest that a simplified overburden representation can often be adequate, particularly for scenarios with low thermal bleed, although there will be some amount of overestimation on the advancement of the thermal front and distortion of the thermal front shape. We also argue that the impapct of a simplified overburden representation is very similar to the use of linear heat transfer coefficients. On the other hand, while models that completely neglect thermal bleed may perform acceptably in some low-bleed settings, they can lead to very large errors in other cases.