Finding the optimal approach for injecting carbon dioxide into deep reservoirs
Peering 500 years into the future of a simulated carbon sequestration site, School of Earth Sciences researchers are using speedy parallel computing in their quest to find the optimal approach for injecting carbon dioxide into deep reservoirs.
Carbon dioxide from coal-fired power plants is a major contributor to global warming. Corralling that CO2 before it enters the atmosphere and pumping it deep underground could help slow the pace of climate change – if the gas stays put.
The most promising sites for storing carbon dioxide lie thousands of feet below the surface in saline aquifers found in porous sandstone. An ideal CO2 aquifer would be overlain by a low-porosity, low-permeability layer called a “caprock,” that effectively seals in the CO2.
But a caprock – often shale– could have undetected fractures, which could become conduits for CO2 leakage, especially if too much gas is injected into the aquifer too quickly.
Any CO2 making its way back into the atmosphere would reduce the effectiveness of carbon capture and storage as a means of combating global warming and could also potentially contaminate freshwater resources it encounters as it migrates upward.
Escaping CO2 would likely carry along trace amounts of combustion byproducts such as sulfur dioxide that were captured along with it, as well as entraining naturally occurring contaminants such as arsenic that might be present in the rock it passes through. CO2 itself would acidify freshwater aquifers and, if it reached surface lakes and rivers, could impact aquatic life and affect the potability of the water.
David Cameron, graduate student in Environmental Earth System Science, is working to avoid those outcomes.
“If we can quantify the risk of leakage, then we might be able to optimize the way we inject carbon dioxide into the storage reservoir to minimize that risk,” he said.
When carbon dioxide is pumped into a reservoir, it is dense, but not quite as dense as the brine already present. So the CO2 tends to rise through the reservoir, ponding directly under the caprock. Minimizing the amount of ponded CO2 is important for minimizing the potential for leakage.
Cameron’s advisor, Energy Resources Engineering Professor Louis Durlofsky, specializes in reservoir modeling – specifically flow simulation and optimization. Optimization for CO2 storage involves using a geological model of a given site and running simulations to determine the best locations and rates for CO2 injection wells.
To develop a model of a particular carbon storage site, geoscientists would gather data to characterize the site by performing seismic surveys and drilling exploratory wells to reveal the site’s geological structure. Instruments lowered down boreholes would measure porosity and other rock characteristics, and cores extracted from the boreholes would be analyzed in the lab to refine the picture painted from the field data.
The models Cameron and Durlofsky work with are not for a specific site, but represent conditions that might be encountered in a real reservoir.
But no matter how detailed the data, there are always many uncertainties, which modelers deal with by creating multiple different plausible versions, or “realizations,” of the geology, which makes the whole endeavor pretty computation-intensive.
“We have to simulate these realizations out for about 500 years into the future to track the movement of CO2,” Durlofsky said. “And typically Dave has to run over a thousand flow simulations for each realization in each optimization run.”
Until recently, that much “model crunching” would have taken quite a while to do, as simulations had to be run sequentially – one at a time. But the Center for Computational Earth and Environmental Science (CEES) in the School of Earth Sciences has the capacity to offer parallel computing to researchers, which has enabled Cameron to run more than 50 simulations at once.
“Parallel computation is revolutionizing the way people do optimization problems,” Cameron said.
In the past, researchers often worked with complicated algorithms designed to require the fewest number of simulations to minimize computational time. But the speed of parallel computing allows Cameron to use simpler, more reliable algorithms that take advantage of performing multiple simulations at a time.
If they were working on an actual injection site, Cameron and Durlofsky would be continually receiving data from the site as CO2 injection progressed, which they would use to refine their geological models.
Cameron said the rules that govern how their optimization algorithm proceeds are loosely based on how swarms of insects and animals interact, a topic he got interested in as an undergraduate in mathematics at Sydney University in Australia.
“Social insects obey a lot of very interesting mathematical principles,” he said. “Flocks of animals such as bees and birds communicate in certain ways that turn out to be quite successful at finding optima, such as food sources.”
Cameron’s senior honors thesis was on optimizing termite reproductive behavior. His successful prediction of how termites strike a balance between producing workers to perpetuate the colony and sending other termites off to establish new colonies was published in The Bulletin of Mathematical Biology.
“Now I’m using the principles of social insect behavior – particle swarm theory – to optimize a completely different problem,” he said.
He presented his current research at the annual meeting of the American Geophysical Union last December and won an “Outstanding Student Paper Award.”
Prior to arriving at Stanford, Cameron spent time in the corporate world as a management consultant. He wanted to combine his math skills with his interest in the environment and sustainability, so he applied to and was accepted at Stanford.
Once he completes his PhD, Cameron said he’d like to continue applying his optimization skills to environmental and sustainability problems. He’d like to continue teaching, which he enjoys, but would also like to be engaged in research or consulting work. But, that’s an optimization problem for another day. Right now, he’s staying focused on carbon sequestration.
Cameron and Durlofsky both see carbon capture and storage as one of several technologies that can cut CO2 emissions in the U.S., but say its greatest impact may be overseas.
“China is burning huge amounts of coal, for example,” said Durlofsky. “If carbon sequestration gets implemented at large scale there, that could have a big impact on reducing global CO2 emissions.”