"Understanding contaminant movement and enhanced oil recovery techniques can save billions of dollars in clean-up as well as production," said Wheeler, director of the Center for Subsurface Modeling (CSM) at the Texas Institute for Computational and Applied Mathematics (TICAM) in the College of Engineering at the University of Texas at Austin. "Our participation in NPACI involves collaboration on many projects, ranging from pollution remediation in bays and estuaries to education and outreach."

The CSM, founded by Wheeler in 1995, comprises more than a dozen faculty and research scientists with expertise in applied mathematics, engineering, computer science, and physical, chemical, and geological sciences. "Our approach to simulation is necessarily interdisciplinary," said Wheeler, who was elected this year to the National Academy of Engineering. Many members of the team came with Wheeler from Rice University, where she had taught for more than 20 years. At Texas, she holds the Ernest and Virginia Cockrell Chair in Engineering.

Oil and gas reservoirs are complex geological formations. To simulate them, it is necessary to take solid, liquid, and gas phases into account, as well as any chemical reactions occurring among and between the phases. "This is the main challenge for porous media modeling," Wheeler said, "to handle highly nonlinear and hysteretic coupled flows with multiscale, multiphase, multicomponent, and multiphysics features."

At CSM, they are developing accurate and efficient parallel algorithms to address this challenge. One of their main efforts is the design and implementation of an Integrated Parallel Accurate Reservoir Simulator (IPARS).

"IPARS is not a simulator in the traditional sense, not a large code that simulates a particular reservoir recovery process, but rather a problem-solving environment--an 'umbrella' with hooks for the various pieces of code that are common to reservoir simulation," said Steven Bryant, a chemical engineer who is the associate director of CSM. The hooks allow users to integrate codes to handle multiple physical models, generalized well management, multiple fault blocks, and dynamic, locally adaptive mesh refinement. All of these components of IPARS are reusable and reconfigurable within the high-level programming interface. "The primary goal is to support realistic, high-resolution reservoir studies with a million or more grid elements on scalable, parallel computer systems," Bryant said.

The IPARS framework supports representation of the 3-D transient flow of multiple phases containing multiple components plus immobile phases (rock) and adsorbed components. The basic model may include one or more "fault blocks," modeled on a grid, corresponding to the gross geometric characterization of many subsurface reservoirs. Each block may have an independent user-defined coordinate system and gravity vector. The primary grid imposed on each block may be geometrically irregular, but is smoothly deformable to a "logically rectangular" reference block.

"The representation of multiple fault blocks is considerably facilitated by the parallel implementation, which enables different processors or groups of processors to operate on different blocks," Bryant said. "Special 'mortar' grids are defined to couple adjoining fault blocks and permit the representation of flows from block to block."

The IPARS framework now supports a two-phase, a three-phase (also known as "black oil"), and a compositional model. "We recently implemented a new formulation of our two-phase model," Bryant said. "As a demonstration of the ease of use of the framework, one person was able to produce a working, tested, debugged model within a month, beginning with no knowledge of the framework at all." This is in itself a major achievement, according to Wheeler, because typical reservoir simulators in the oil industry may require 20,000 lines of code, much of which is not related to the physical problem to be simulated but rather to data distribution and management, load-balancing, and similar concerns. These are the reusable and reconfigurable portions of IPARS.

"We've run a million-gridblock black-oil model," Bryant said, "and our collaborators on this project have run a million-gridblock compositional model, simulating a complicated gas-injection process for 1,000 days, which would have taken months on a workstation, in just 30 minutes on 64 nodes of an IBM SP." Developers within the CSM group include Joe Eaton, Ivan Yotov, Malgorzata Peszynska, Srinivas Chippada, Carter Edwards, John Wheeler, and collaborators Peng Wang and Wei Xu of the Center for Petroleum and Geosystems Engineering at Texas. One NPACI collaborator is former group member Manish Parashar, now at Rutgers University.

Top | Contents | Next

Both oil well management and pollution remediation schemes have used the mechanism of pumping something down into the Earth, such as water or acid, to facilitate the recovery of something else. Only a code that does all the biogeochemistry can predict the full consequences of such measures, let alone permit scientists to understand what takes place naturally.

Of particular use in pollution remediation studies is the CSM code ParSSim, which stands for Parallel Subsurface Simulator, the development of which was begun under a Grand Challenge grant from the Department of Energy.

"A professor of mine at Texas, Bob Schechter, was fond of saying that the Earth's crust is really a giant chromatographic column--a combiner and separator of minerals, hydrocarbons, and solvents," Bryant said. "ParSSim is a reactive transport simulator that we think of in much the same way. It handles general biogeochemistry, including adsorption, desorption, ion exchange, mineral precipitation and dissolution, absorption into or out of fluid phases, oxidation and reduction, speciation, radionuclide decay, and biodegradation. In particular, the biological reactions are fully coupled with the abiotic reactions."

The present ParSSim implementation exhibits excellent parallel scaling and dynamic load balancing. It can represent a single flowing phase and an arbitrary number of immobile solid and fluid phases, as well as an arbitrary number of chemical species and reactions. In addition to Bryant, the current developers include Todd Arbogast, Clint Dawson, and Joe Eaton.

Now in a full-fledged NPACI collaboration, the flow portion of ParSSim is being implemented within the KeLP environment (see page 12) under development by Scott Baden and colleagues in the UC San Diego Computational Science and Engineering Department. "We're planning to use the experience we've gained porting the flow code to help port the transport and chemistry codes of ParSSim next," Bryant said. Baden visited Austin in December 1997, installing KeLP on the Texas parallel platforms. Max Orgiyan from Baden's group visited in February 1998 and the collaborators designed a "mortar" grid to be implemented under KeLP.

"The characterization and remediation of contaminated sites of all sorts is difficult and expensive," Wheeler said. "Our efforts at CSM and within NPACI can benefit the understanding, design, and testing of economically feasible decontamination strategies. These are problems we are all eager to work on." --MM

Top | Contents | Next

Top | Contents | Next