Re-creating 170 million years of geologic evolution, computer
simulations are homing in on new petroleum reserves.
In Brief:
Software tools that simulate realistically how oil is generated, and how it
migrates through sediment and accumulates in oil fields, can reduce the
costs of oil exploration and production. Pioneering simulation methods
developed at IBM Research are able to model complex geological processes
such as the dynamics of entire sedimentary basins. The tools are poised
both to guide oil companies to new reserves and to help them pump more out
of old ones.
Looking for new oil is an expensive process. With a dwindling number of
easily accessible oil fields to be found on land, oil production has come
to rely heavily on offshore wells, which must be drilled in deeper and
deeper water. Each such drilling operation costs $20 million to $50
million, and in the United States an average of 10 dry exploratory wells
are drilled for every producing one. In view of numbers like these,
petroleum geologists are increasingly turning to computer simulations that
can cut the costs of exploration or help squeeze more oil out of each
underground reservoir.
Ulisses Mello, a researcher at IBM's Thomas J. Watson Research Center, is
developing tools that can assist on both fronts. Mello is on the vanguard
of the relatively new field known as integrated basin simulation, which is
becoming an important risk-assessment resource for oil exploration. Using
geological data, scientists in this field can reconstruct the evolution of
sedimentary basins through geologic time, simulating earth processes such
as sedimentation, structural evolution, heat flow, fluid flow and the
generation and migration of hydrocarbons.
Mello has developed novel solutions to some of the more challenging
problems in this field, including large-scale simulation and the
representation of evolving three-dimensional geological structures. Using
innovative techniques and up to 200 nodes of a powerful IBM RS/6000®
SP™ parallel computer, Mello has created the first simulation of the
evolution of the oil-rich Gulf of Mexico. By modeling how sediment laden
with organic material has been deposited since the Gulf's formation, Mello
attempts to pinpoint locations where the combination of heat, pressure and
time has "cooked" the sediment into petroleum. In the process, he has
devised techniques for modeling the shifts that occur in the boundaries
between moving masses -- techniques that can be applied in geological
research generally and even in other fields such as materials science.
Another set of simulations can model how oil is pumped out of a reservoir,
and where pockets of oil have been inadvertently bypassed. Overall, Mello
aims to "help the decision-making process in the oil business and reduce
the risks of exploration."
A "LIVING MODEL"
In his Gulf of Mexico simulation, Mello began with what is known about the
region today and attempted to infer the course of its evolution since the
beginning of its drift phase 170 million years ago, in the Late Jurassic
Period, when the Yucatan block was pulled away from North America. "In
geological simulations," he says, "we know what the end point is -- the
sediments that now exist -- but we need to simulate how they got there, so
as to determine where conditions were right for petroleum to mature. Once
we know that, we can identify promising drilling sites."
Mello started with data -- gathered from seismic imaging and from actual
drilling sites -- on the thickness and porosity of sediments of each
geological epoch at many points throughout the Gulf. The data allowed him
to calculate the rate of sedimentation at each location and time, and
thereby to generate a simulation that shows sediment gradually being
deposited in the correct amounts throughout the Gulf. Taking into account
the present characteristics of the various layers of sediment and resulting
rock, the simulation then models how the rock compresses underlying layers,
how heat from the earth's interior is transferred and how water trapped in
the sediment is squeezed out.
As one part of the model generates a history of temperatures and pressures,
a second part takes these results and simulates the formation of oil. Each
layer of sediment contains a known amount of organic material. It is this
organic material -- the remains of marine organisms as well as organic
matter washed out to sea by rivers -- that gives rise to petroleum. The
transformation takes place when this matter is buried by the layers of
sediment above and cooked by heat from the earth's interior over millions
of years.
To determine the distribution of oil, Mello had to model the complex
dynamics of heat flow in the region. "One of the most important processes
in the Gulf," Mello points out,"occurs when heavy mud is laid down and
becomes shale, trapping water within the layers of rock." As layers build
up, they exert tremendous pressure on the water. The pressurized water in
turn retards the upward flow of heat. This is because pores in the rock
remain open, retaining more water, which effectively insulates the
overlying rock from heat rising from below.
The resulting map of simulated oil maturity is, says Mello, "in reasonably
good agreement" with what is known of oil distribution in the Gulf --
perhaps accurate enough "to allow an oil company to assume a most
optimistic and a most pessimistic scenario in deciding whether to drill in
a particular spot." But a model is only as reliable as the data on which it
is built. And knowledge of the Gulf's present geology is incomplete -- most
detailed in areas where many wells have been drilled, much less so where
only seismic data is available. Consequently, Mello regards his Gulf
simulation as a "living model" that links data to processes: as new seismic
or drilling data becomes available, he plugs it into the simulation,
gradually improving accuracy.
Already, though, Mello's simulations are contributing to knowledge of the
Gulf of Mexico basin. They have helped researchers to understand how the
basin produced its known reserves of 112 billion barrels of oil and 524
trillion cubic feet of gas, as well as to begin quantifying and locating
large additional reserves that are believed to exist. And the work on the
effects of highly pressurized water on oil maturation has earned Mello and
a colleague -- Garry D. Karner, of Columbia University's Lamont-Doherty
Earth Observatory -- the Wallace Pratt Memorial Award of the American
Association of Petroleum Geologists.
SHIFTING BOUNDARIES
While developing the Gulf simulation, Mello has also been hard at work on a
problem that has long plagued geological modeling: how to represent the
formation of new faults and other moving boundaries over time. Simulating a
process such as flight is relatively easy, because it involves variables
that change smoothly and continuously within fixed boundaries (an airplane
wing). But in geology, faulting causes sedimentary blocks to slip past each
other, sedimentary layers deform and buckle; and salt or magma can intrude
on the layers from below. For a model to reflect this changing topology
adds formidable levels of complexity.
Applying geometric modeling tools developed by various researchers over the
past 15 years, Mello and Watson researchers Mike Henderson and Paulo
Cavalcanti have produced a novel three-dimensional geological model that is
topologically flexible. The basic unit of the model is a geological block,
defined by topological entities representing faults and surfaces between
different sedimentary layers. This block, which incorporates both
geometrical and geological characteristics, is depicted by a mesh of
elements that deform in response to the modeled forces that act on the
block.
To solve the problem of how to create new topological blocks when a fault
develops, the model incorporates all the existing geological boundaries
within a given region. At the start of the simulation, before present-day
faults have formed, the boundaries are inactive. They remain in this latent
state until geological forces are strong enough to initiate the fault. At
that point, the model activates the fault boundary and the blocks on either
side of the fault are free to move as separate entities.
To simulate these dynamic conditions, Mello had to develop a means of
handling the often severe distortion in the numerical meshes as the rocks
move. Mello defined discrete fault zones, narrow areas where the mesh
distortion occurs. Meanwhile, the meshes in the blocks moving on each side
of the fault are almost undisturbed. This mirrors the real world, where
geological blocks can move past each other for hundreds of miles without
major distortion.
Mello envisions uses for this flexible modeling technique far outside
petroleum geology -- in simulating the dynamics of continental drift, for
example. Modeling the motion of geological plates requires just the sort of
shifting boundaries and changing alignments that Mello's simulation tools
provide. Farther afield, many problems that involve large deformations --
such as studies of metal failure in bridge collapses or car collisions --
might benefit from such tools.
So far, however, Mello has restricted his use of the new technique to the
pursuit of oil. He has, for example, successfully modeled the formation of
salt domes -- a process that often entails just the kind of deformation the
model can handle well -- and their effect on oil generation. Salt domes are
formed when salt rock moves upward because of its buoyancy with respect to
the surrounding rocks. Because they efficiently transport heat from the
rock below to the rock above, salt domes exert a major effect on
temperatures. Mello's work has shown, for example, that, contrary to prior
assumptions, oil could still be forming beneath salt domes in areas
previously considered overcooked, and may prove accessible to future
drilling.
GETTING THE OIL OUT
Finding the oil is only half the battle. As petroleum is pumped out of a
field, the natural gas (formed with the oil) that drives it toward the
wells exerts less and less pressure. Inevitably, pockets of oil are left
behind. If the pockets are large enough, new wells can be drilled to tap
them, but it's crucial to learn where the oil is and how much is left. The
standard way of determining this is to make seismic maps of a region and
compare them with similar maps made when the fields were new. Since
petroleum in the rock affects the speed of seismic signals traveling
through them, changes in those speeds can in theory produce a map of
changes in oil distribution, and thus point to missed pockets.
But, says Mello, there's a big problem with comparing seismic images. "The
older images were made, in some cases, decades ago with different
technology, often with a different amount of spacing between receiving
stations. These differences can produce artifacts when the images are
compared -- places where there appear to have been changes in the oil
distribution but really there are just changes in how you got the data."
To provide a check on this data, Mello, together with scientists led by
Roger Anderson at Lamont-Doherty, is working to integrate a set of
simulations that models how oil should flow, given the initial conditions
and the wells drilled, and where it should be at present. First, a
stochastic, or probability-based, simulation takes the old seismic imaging
data and calculates
the likely distribution of oil before well production. A second, fluid-flow
simulation models how the oil would move as it was pumped. And a third
simulation models how the oil motion affects the seismic image.
A component of this technology -- "time-lapse seismic" -- has already been
pressed into service. Texaco has drilled a well in the Gulf of Mexico that
pumps out 1,500 barrels of bypassed oil per day. The well has so far
yielded 1.4 million barrels, worth some $24 million. Besides helping to
zero in on bypassed oil, the technology lets the oil companies continuously
monitor production.
Currently, Mello is working with oil and service companies, including
Petrobras -- the national oil company of his native Brazil -- and Western
Geophysical to develop still more sophisticated models. Mello believes that
this is just the beginning. New advances taking place at IBM Research in
simulation, visualization and data mining, he predicts, will achieve more
and more realistic pictures of the subsurface and make hunting for oil a
more exact and less costly proposition.
FYI:
http://www.research.ibm.com/imaging/#geo
Eric Lerner is a freelance science writer based in Lawrenceville, New
Jersey.
More Information:
Simulating Nature
Computer simulations attempt to emulate nature by breaking up continuous
space and time into discrete units. Space is defined as a mesh and time as
individual time steps, like the frames in a motion picture. In certain
simulations, including some created by Ulisses Mello, the mesh elements are
not fixed in space; instead, they migrate as the material they represent
shifts about. The simulation keeps track of a set of conditions at each
mesh point - pressure, heat, chemical composition, rate of deposition of
sediment and so on.
The physical and chemical laws governing the geological processes are
approximated by a set of mathematical and computational rules that allow
the computer to calculate the conditions at each mesh point at one time
step, given the conditions at all mesh points from previously calculated
time steps. The initial set of conditions must be provided by whoever is
performing the simulation.
Generally, conditions at one mesh element depend mainly on conditions at
nearby elements. As a result, a simulation's space domain can be split up
into many subdomains to run on hundreds of processors in parallel.
