To understand the implications of cement corrosion for CO2 storage, simulations must be carried out both to predict the composition of fluids that will interact with wells and to calculate formation-scale leakage rates that could result from cement deterioration at individual wells. To accomplish these goals, Jean Prevost’s group has been expanding the capabilities of the in-house numerical simulator Dynaflow, while Mike Celia and colleagues have developed fast analytical models and new methods for including well properties in complex numerical models.
If cements in existing wells were exposed to large volumes of CO2-rich fluids, degradation like that seen in laboratory experiments would be a serious drawback for underground carbon storage. However, carbonated brines might become saturated and less aggressive by the time they travel through a formation and come into contact with cements. To assess fluid compositions at wells, the Storage Group decided at the outset of the project to develop in-house codes to simulate CO2 flow physics, geochemistry,and geomechanics.
Since the beginning of the grant, a team led by Jean Prevost has expanded the capabilities of the finite element code Dynaflow to include multiphase, multicomponent flow. Prior to this year, the simulator development team had added modules for two-phase flow, implemented equations of state appropriate for CO2 -brine systems, and developed the capability to calculate pH.
This year, refinements in the treatment of pressure and relative permeability in the model have improved its accuracy and stability. Work was also completed on an efficient flash calculation to simulate 3-phase equilibrium conditions encountered in supercritical CO2 sequestration. However, since leakage along well stems is a vertical flow where pressure and temperature can vary over a wide range, 4 phases must be taken into account: aqueous brine, CO2-rich liquid and gas, and solid salt. During the past year, the group has made significant progress in simulating subcritical CO2 thermodynamics essential to modeling such 4-phase systems and addressing shallow transport in leakage calculations.
The team has also implemented a geochemistry module to account for interaction of brine with the formation and cement in injection scenarios. Having incorporated equilibrium geochemistry, the next step is to incorporate cement chemistry kinetics to allow simulation of corrosion of cement during leakage along a well. Collaborators at the Los Alamos and Lawrence Livermore National Laboratories will determine reaction parameters by simulating the cement experiments performed in our lab. The results will be incorporated into the Dynaflow geochemistry module and the model will be used to study the rate of attack of brine flowing through an annulus around a cement plug in a well. The next step in development will be to incorporate the thermal effect of reactions and phase changes, which will play an important role during leakage, into the model.
Improved upscaling in numerical models
Since well dimensions are so much smaller than the gridscale in most reservoir simulators, incorporating the impacts of multiple leaky wells in this type of simulation is challenging. Fine-grid simulations can be used to differentiate flow within a well from that in the surrounding rock, but are computationally expensive and impractical for modeling large numbers of wells. The Celia group has been working to include numerous abandoned wells in numerical simulations of CO2 injection in a way that reduces computation time without compromising accuracy in leakage rates.
Sarah Gasda has investigated two methods which take advantage of modeling the CO2 system with a relatively coarse grid while still capturing the effect of sub-scale leaky wells. The first method treats the well as a structured heterogeneity occurring at the sub-scale and determines the correct effective, or pseudo, relative permeability function needed at the coarse scale. Use of standard pseudo functions for upscaling can overestimate flow by hundreds of percent, but Gasda and colleagues showed last year that their new method significantly reduces the inflated flow rates produced by standard methods, bringing leakage estimates into agreement with fine-grid predictions. This year, a second method using domain decomposition to refine the coarse grid only around leaky wells was shown to similarly improve model predictions.
The group has now developed a standard finite difference simulator for solution of two-phase (CO2 and brine) flow equations that accommodates both the upscaled relative permeability functions and domain decomposition. The code is being used to simulate multi-well leakage, and to test some of the assumptions used in deriving the group’s analytical models (described below). The team has also implemented a vertically-averaged version of the governing equations, with an assumption of a sharp interface separating the two fluids.
Another aspect of Gasda’s work this year has been to determine the importance of a sloping caprock on the migration and shape of a buoyant plume of injected CO2 during the injection phase. The results of her study will provide a methodology to determine whether slope is a significant factor in simplified models of CO2 injection systems.