Gibson: Common-Grid Models for Subsurface Fluid Flow
Richard L. Gibson, Jr.
For many years, as petroleum engineers, geologists, and geophysicists analyzed oil and gas reservoirs deep below the Earth’s surface, their characterizations have often been restricted by technology and computational ability. As a result, simplifications to data collection and processing methods were necessary. In recent years, however, advancements in data acquisition, data processing, and in computational modeling are opening new doors to reveal sophisticated tools that may lead us to new frontiers of our understanding of subsurface fluid flow, geological features, and how they interact.
Seismic wave imaging of deep rock formations and reservoirs is now greatly improved by using tens of thousands of sensors, contributing to large datasets with an unprecedented breadth of information. However, despite the proliferation of such data, the level of detail in processed images, or pictures, of the subsurface is limited by physics: seismic wavelengths are normally approximately 100 meters, and processing of seismic wave reflection data relies on simplifying assumptions and numerical modeling to try to achieve higher resolution of detailed geological features.
In the past, analysis from geophysicists was limited because it relied on computer models using uniform, rectangular grids representing earth structures. These small grid cells are approximately 10 to 20 meter cubes in size and are much larger than important geological structures controlling fluid flow. These approaches are used both because of limits in computer power and because the seismic models are based on data received from sensors that are sending waves from the surface of the Earth, down to the reservoir and back up.
The analysis of fluid flow in those same reservoirs, conversely, only needs to consider the reservoir structure and is able to use discretatizations that are often smaller than one meter. The simulations of fluid flow use irregular grids that often have cells that are very thin and perpendicular to geologic layering. This provides a fine scale and highly detailed, yet non-uniform, grid that is well suited to simulating movement of oil or other fluids in the reservoir.
Attempting to combine the seismic wave data with the flow data into a single reservoir “map” will present large challenges. Scientists know that the geometry and variations in rock and fluid properties on the fluid flow, or fine scale grid, affect seismic wave propagation as well, and these effects should be included on the coarser scale grid. Unfortunately, there is currently no reliable method to combine these two grids with high accuracy.
Dr. Richard Gibson, the Francesco Paolo di Gangi/Heep Endowed Professor and a professor in the Department of Geology and Geophysics, is leading a project to combine these two grids and create a simulator that is built on a common computational grid. This simulator will be a key component of an inversion scheme, a tool that will combine fluid and seismic wave data to characterize a reservoir, providing analysis of porosity and permeability in the subsurface and identifying the distributions of fluid types within the formation.
Unlike analyses of the past, Gibson’s approach does not rely on arbitrary analytic solutions that make assumptions or oversimplify subsurface geophysical characteristics and fluid distributions. He will develop integrated multiscale computer modeling techniques to bring together the data for seismic wave propagations and fluid flow analysis. Specifically, Gibson will develop and utilize innovative model reduction schemes, a way to make approximations in the data without making unfounded assumptions.
Beginning with a local model reduction, he will focus on a small area in each coarse grid, called a snapshot space. In this snapshot space, Gibson will identify key characteristics through a technique called local spectral decomposition. Using these key characteristics, he will construct multiscale basis functions, or descriptions of the dominant features of the snapshot area. Multiple basis functions are then coupled through a formulation, and the number of basis functions in each grid is selected adaptively based on estimates within the model reduction. The combination of the multiscale basis functions yields a common computational grid that will support simulation
This approach could address many problems that are found in the development of energy resources and in the mitigation of associated environmental risks. A reliable means of remotely detecting and quantifying changes to fluids or gases, as well as the pressures within rock formations, could provide critical insights into a reservoir’s capacity, aid in decision-making about advanced recovery methods, or even advise future well placements. Seismic imaging is the most important technology for measuring the structure and properties of hydrocarbon reservoir formations between well locations, and time-lapse seismic methods that acquire data repeatedly over time periods of months or years will allow the mapping of fluid movements.
This flow map could present opportunities for measuring the impact of recovery or storage methods well beyond the casing and beyond the reservoir. Extending beyond hydrocarbons, seismic waves can be very sensitive to CO2, and potential exists for time-lapse seismic methods to be employed in the monitoring of CO2 sequestration projects.
« Back to Faculty Interviews