Computational Models for Evaluating Long Term CO2 Storage in Saline Aquifers

(funded by NSF and KAUST through the Academic Excellence Alliance Program)

Geologic sequestration is a proven means of permanent CO₂ greenhouse gas storage, but it is difficult to design and manage such efforts. Predictive computational simulation may be the only means to account for the lack of complete characterization of the subsurface environment, the multiple scales of the various interacting processes, the large areal extent of saline aquifers, and the need for long time predictions. This project investigates high fidelity multiscale and multiphysics algorithms necessary for simulation of multiphase flow and transport coupled with geochemical reactions and related mineralogy, and geomechanical deformation in porous media to predict changes in rock properties during sequestration. The work will result in a prototypical computational framework with advanced numerical algorithms and underlying technology for research in CO₂ applications, which has been validated and verified against field-scale experimental tests.

BIGDATA: Collaborative Research: IA: F: Fractured Subsurface Characterization Using High Performance Computing & Guided by Big Data (Funded by NSF)

Natural fractures act as major heterogeneity in the subsurface that control flow and transport of subsurface fluids and chemical species. Their importance cannot be underestimated, because their transmissivity may result in undesired migration during geologic sequestration of CO₂, they strongly control heat recovery from geothermal reservoirs, and they may lead to induced seismicity due to fluid injection into the subsurface. Advanced computational methods are critical to design subsurface processes in fractured media for successful environmental and energy applications. Continue reading →

Simulation of the Cranfield CO2 Injection Site with a Drucker-Prager Plasticity Model

Coupled fluid flow and geomechanics simulations have strongly supported CO₂ injection planning and operations. Linear elasticity has been the popular material model in CO₂ simulation for addressing rock solid material behaviors. On the other hand, nonlinear constitutive models can take into account more realistic rock formation behaviors to model complex, chemically active, and fast injecting operations. For example, failure or damage may occur for rock formation near wellbores due to high fluid injection pressures or flow rates. The damaged formation near wellbores results in the changes in rock porosity or permeability, which impacts fluid flow behaviors. Such failure or damage of rock formations can be well described by the Drucker-Prager plasticity theory.

The Drucker-Prager plasticity solid mechanics module has been implemented into IPARS (Integrated Parallel Accurate Reservoir Simulators developed at the Center for Subsurface Modeling, The University of Texas at Austin). The coupled poro-plasticity system is solved using an iterative coupling scheme: the nonlinear flow and mechanics systems are solved sequentially using the fixed-stress splitting, and iterates until convergence is obtained in the fluid fraction. To the best of our knowledge, the application of this algorithm is new for poro-plasticity. To achieve fast convergence rates for solving the nonlinear solid mechanics problems, a material integrator is consistently formulated and implemented in the IPARS geomechanics module. An enhanced parallel module for general hexahedral finite elements is also developed for IPARS for solving large-scale problems in parallel. A driver for the direct solver SuperLU has been implemented in cases when the linear systems are difficult to converge. A Cranfield CO₂ injection model is set up according to the reservoir geological field data and rock plasticity parameters based on Sandia national lab experimental results. This Cranfield model is solved using IPARS and the prediction on CO₂ flow and formation deformation is presented.

Figure 1 shows a comparison between poro-elastic and poro-plastic results in an injection well case with homogeneous Cranfield data and rectangular geometry. Figure 2 shows poro-elastic results in an injection well case with heterogeneous Cranfield properties and geometry. Future work includes the incorporation of plastic yielding into the initial condition so that the heterogeneous Cranfield properties and geometry can be used in the poro-plastic case. Together with stress dependent permeability and fluid fraction, this will give us an accurate predictive geomechanics model for matching CO2 injection results at the Cranfield site.

Figure 1. Comparison of elasticity (left) and plasticity (right) models with homogeneous parameters and rectangular geometry. Fluid pressure, vertical displacement, and plastic strain are shown (below).

Figure 2. Elasticity model with heterogeneous Cranfield properties and geometry. 3D (left) and 2D (right) plots of the vertical displacement component at final simulation time (below).

Elasticity model with heterogeneous Cranfield properties and geometry

Data Driven Simulation of the Subsurface: Optimization and Uncertainty

Data Driven Simulation of the Subsurface: Optimization and Uncertainty EstimationParticipants

CSM: Mary F. Wheeler (PI), Hector Klie, Clint Dawson, Wolfgang Bangerth, Raul Tempone, Burak Aksoylu, Xiuli Gai.

PE & G: Carlos Torres-Verdin

RUTIG: Paul Stoffa (co-PI), Mrinal Sen (co-PI).

The Ohio State University: Joel Saltz (co-PI), Tahsin Kurc, Umit Catalyurek.

Rutgers University: Manish Parashar (co-PI).

Graphics Click thumbnail for larger image		Movies (.swf)
Optimization	K	optim_time 4.6 Mbmovie_surf 3.4 Mbmovie_contours 1 Mb
Optimizing oil production on the Grid	“Closing the loop” with optimization
DDSSF	Blocks

Project Summary

Intellectual Merit

Remote sensing is employed in science and engineering problems to infer material properties when these properties can not be directly sampled. To better understand and manage our environment for safety and economic reasons, much progress has been made in imaging the subsurface and estimating physical properties based on remote sensing data. Repeated observations over targets for environmental remediation and reservoir production have become a recognized diagnostic tool for assisting management decision. In addition, improved optimization techniques capable of responding to large, multi-resolution, disparate, dynamic datasets in a fault tolerant and adaptive fashion are a fundamental requirement for effectively estimating and minimizing the uncertainty in any data-driven application. The integrated and effective treatment of these issues motivates the present project. The assembled research team proposes to advance the mathematical, engineering and computational foundations necessary to enhance our understanding and extend the predictive capabilities of the physical processes that govern the subsurface phenomenal at multiple temporal and spatial scales Target applications include management of aquifers for water resources, optimizing oil and gas production, and monitoring environmental risks, e.g. at waste containment sites or arising from natural hazards.

The intellectual merits of the proposal include:

(1) development of the next generation of accurate, multi-scale, coupled chemical, fluid, geomechanical, and geophysical simulations for modeling instrumented subsurface environments;
(2) large scale optimization techniques (based on a hybridization of global and local approaches) to drive reliable decision-making and a dynamic symbiotic feedback between computation and data;
(3) deployment of an autonomic Grid middleware for providing the adequate processing of substrate and data management services for (1) and (2).

The realization of the above contributions will result in the Data Driven Subsurface Simulation Framework (DDSSF).

The framework will be built upon the experience of the team in developing prototype simulators and data management tools under an on-going ITR project. The team will have access to two large dynamic datasets: one from the Gilt Edge Mine Superfund site, managed by EPA and the Idaho national Engineering and Environmental Laboratory (INEEL); and the second from an instrumented oilfield of the coast of Norway. These observational data will provide a near continuous flow of remote sensor data over time that will serve as the basis for developing and deploying the mathematical and computational tools proposed in DDSSF.

Broader Impacts

Simulation of the coupled chemical, geomechanical and geophysical response of subsurface systems is an imperative for the scientific and engineering community to understand the environmental an economic effects of human activities. Immediate impact to DOE and EPA will occur through our collaboration with INEEL. The scientific, technological and educational impact of the proposed research on dynamic data analysis will extend well beyond subsurface modeling. Our tools will have immediate applications to global warming (e.g. CO2 sequestration), national security (e.g. mines and tunnels) and the biomedical sciences (e.g. blood flow, tissue engineering, analysis of radiology and microscopy imaging data).

The proposed research activity will involve the training of undergraduates, graduate students and post-doctoral fellows at The University of Texas at Austin, Ohio State University and Rutgers University in a truly cross-disciplinary subject that has far-reaching implications. The research activity also includes outreach to industry, government laboratories, K-12 students and teachers, and undergraduates in the broader community through workshops, seminars, internships, and summer programs and curricula development. Specific outreach opportunities include the industrial affiliates program at UT Austin, collaboration between Ohio State and the Capital University Computational Science Center, the Ohio Supercomputer Center Young Women’s Summer Institute and the Rutgers Governor’s School.

Back to Current Projects