IPARS: A New-Generation Framework for Reservoir Modeling

PARTICIPANTS
Steven Bryant, University of Texas at Austin

o wring the last drop of oil from an oil field and to avoid groundwater pollution, oil researchers have gone to great lengths to model the complex physics and structures of underground reservoirs. How do they go from their best research information to an efficient and nonpolluting recovery strategy? At least one comprehensive solution--a framework for linking several physical models into one simulation--is emerging at the Texas Institute for Computational and Applied Mathematics (TICAM) as part of an NPACI Engineering project on Oil Recovery and Pollution Remediation in collaboration with computer scientists at UC San Diego and the University of Maryland.

THE RESERVOIR PROBLEM

A MODULAR FRAMEWORK

GEOMECHANICS AND FRACTURE ZONES

THE RESERVOIR PROBLEM

"Effective production management of oil reservoirs is placing new demands on the mathematics and science of reservoir simulation," said Mary Wheeler of the Center for Subsurface Modeling (CSM) at TICAM, located at the University of Texas at Austin. The CSM, founded by Wheeler in 1995, unites a dozen scientists with expertise in applied mathematics, engineering, computer science, and physical, chemical, and geological sciences. "We need to deploy an accurate modeling strategy."

The traditional mode of simulator development begins with the physical problem of interest, defines the governing equations, discretizes them, and builds a solver and a front end specific to the physical problem. Tackling a problem with different physics meant repeating the entire process. Meanwhile, oil fields are presenting problems of ever greater complexity. Length scales of practical and economic interest range from tens of meters to kilometers, while the governing phenomena may vary over millimeter scales. Resolving these processes presents a huge computational challenge. Moreover, different physical processes occur simultaneously in different parts of the model.

To address the problems with traditional models and incorporate advances from the new generation of models, the CSM is developing an Integrated Parallel Accurate Reservoir Simulator (IPARS). One of the principal design features of the IPARS framework is its modular structure, enabling new or more detailed physical information to be studied with only incremental coding development. The IPARS development team recently reported on its work at the Fifth SIAM Conference on Mathematical and Computational Issues in the Geosciences in San Antonio, Texas, March 24-27.

REFERENCE
Researchers interested in exploring the use of IPARS in their research should contact Steve Bryant at sbryant@ticam.utexas.edu.

Figure 1. IPARS modeled a random permeability field in a three-block reservoir example (right) in which different numerical schemes are used in the blocks (left).

A MODULAR FRAMEWORK

The next-generation IPARS framework is able to include multiscale analysis, multiblock discretizations, and advanced solvers, all implemented on parallel platforms. "IPARS reverses the traditional mode of simulator development," said Steve Bryant, CSM associate director. "Any physical model can be run under the framework, and the developer need only provide routines that compute a time step for the physical model of interest."

"IPARS uses two levels of domain decomposition because of the complexity of underground reservoirs," Bryant said. "We must take solid, liquid, and gas phases into account, as well as reactions among and between phases." First, the physical domain is decomposed into subdomains or blocks according to the geometry and geology of the reservoir, the placement of wells, the physical, chemical, and biological
processes taking place, and the temporal and spatial scales salient to the solution. In the second stage, the computations are further decomposed on a parallel machine to achieve computational efficiency.

In IPARS, then, an oil reservoir is a union of blocks. The interfaces between blocks are mortar spaces--thin domains tied together by physically meaningful interface boundary conditions that permit input and output from one block to neighboring blocks or domains (Figure 1).

The efficient comparison of multiple realizations is essential. "IPARS lets us do this for remediation models as well as for oil and gas production models," Bryant said. "But the ideal computational environment surrounding IPARS will have database support for carrying out spatial queries, with user-specified processing of very large data sets, plus tools to support the development of multiblock applications that allow individual blocks to be structured or unstructured. Thus in NPACI, we are linking IPARS development with two other elements of this ideal environment."

The IPARS team is implementing new codes using software projects from NPACI's Programming Tools and Environments thrust area. The KeLP programming environment, under development by Scott Baden of the Computer Science and Engineering Department at UC San Diego, is a run-time library that supplies optimal implementation on arbitrary architectures. The KeLP library is especially important for highly heterogeneous execution environments, including the teraflops IBM SP that will be installed at SDSC.

Also being linked with IPARS is the Active Data Repository (ADR) under development by Joel Saltz and Alan Sussman of the University of Maryland. ADR is an infrastructure for building parallel database systems that enable the integration of storage, retrieval, and processing of multidimensional data sets on parallel machines. It has already been used successfully by CSM scientists in the simulation of oil spills in bays and estuaries (see the January-March 1999 enVision). "For pollution remediation calculations with IPARS, ADR will help by storing, retrieving, and processing data generated by surface and groundwater models, as well as sensor data sets," Bryant said.

Figure 2. Oil Concentration

Oil concentrations from the IPARS model shown in Figure 1 after 1,500 days of injection. Red is higher concentration, blue is low.

GEOMECHANICS AND FRACTURE ZONES

The IPARS development process was illustrated recently with two types of geomechanics applications. In the first, an IPARS black oil model was coupled with JAS3D, a geomechanics code from Sandia National Laboratories, to simulate flows in weak, highly compactible rock formations. Pressures computed in the black oil flow simulation are passed to the geomechanics code, which computes the resulting changes in rock porosity. The updated porosity is passed to the flow simulation, and the cycle is repeated. The models were integrated by simply adding a call to JAS3D within the IPARS driver routine of the black oil model.

In the second application, a two-phase flow model within IPARS was coupled with UTWID, a model that simulates physical phenomena near injection wells. "We wanted to incorporate the dynamic computation of injectivity with the full-field performance of interacting wells, but without adding the overhead of detailed near-well simulation to the full-field model," Bryant said. "The coupled code reveals very rich behavior, even for simple injection schemes." All that was added to IPARS was a short subroutine that updates one parameter for each injection well, computed in a call to the UTWID code. No other alteration was required.

Bryant also described successful tests of IPARS, recently reported in a paper written by the group for the Journal of the Society of Petroleum Engineers. To model two-phase flow through a fracture zone, the zone was approximated by a union of three layers, two horizontal and one vertical, in a model containing five blocks and six mortars. Since permeability in the fracture zone was two orders of magnitude higher than in the rest of the reservoir, the fracture layers were discretized on a much finer grid than the rest of the domain. "This relatively simple example suggests that the specification of real networks of fractures and faults can be done to any suitable degree of complexity," Bryant said.

Extending the same methods to a two-phase, 3-D, horseshoe-shaped reservoir, the Wheeler group was able to represent flow conditions created by two injection wells and three production wells situated around the horseshoe (Figure 2). Finally, the group considered a U-shaped reservoir consisting of three blocks, coupling different numerical models across the block interfaces. This produced an extremely complex domain, and the variation in permeability made the problem much more difficult at the mortar interfaces.

"Our models need to capture the complex underground structures so we can test alternative injection or drilling strategies," Wheeler said. "We think IPARS will be a boon for researchers in this field who want to try out new ideas, a one-stop shop for engineers wanting to examine the behavior of different physical models of a particular problem, or of combinations of models." --MM