Reservoir Simulations Use AI to Change with the Times

GALVESTON, Texas — The demands E&Ps are heaping on the shoulders of artificial intelligence (AI), cloud computing and machine learning are daunting. Good thing AI doesn’t get nervous.

Among both the practical and fanciful goals of companies such as Exxon Mobil, Equinor, Chevron Corp. and others: using simulations to unlock frac geometry; accelerating greenfield and brownfield development; ever more automation; and even eschewing data to let AI play with engineers’ ideas. 

Just how reliant are companies on their technology? Exxon Mobil Chief Reservoir Engineer James Hacker calls reservoir simulation “the single most important tool we use in making every single upstream decision.”

Oil and gas companies have long relied on reservoir modeling and simulations to plan E&P strategies. Now E&Ps are turning to the same tools to test carbon capture and storage (CCS) projects and geothermal exploration, experts said during a plenary session focused on the future of reservoir simulation at the SPE Reservoir Simulation Conference in Galveston on March 28.

From supermajors to geothermal firms, companies are also turning to commercial software and open source solutions as they continue to experiment in the digital realm.

Exxon is exploring ways to build models faster, link simulations to other models and create simulations that can optimize a project’s full life cycle. In doing so, the supermajor recently pivoted away from using its own proprietary full-field simulation software and started using a commercially available program, Hacker said.

“That was a difficult decision,” he said. “We’d been building simulations in-house for more than 50 years.”

But the supermajor decided a common software package made it easier to share models with partners and regulators. Plus, the talent pool for people able to build and maintain detailed simulators is shallow. And commercial tools now available are “very good,” Hacker said.

In fact, he said, the programs make it possible to “take a sparse bit of data” and help the company understand unconventional oil and gas reservoirs to develop a concept around it.

Baris Guyaguler, Chevron’s reservoir simulation development and environment chapter manager, said the operator wants to accelerate turnaround time for both greenfield and brownfield simulations, “automating the entire chain” to decrease project turnaround time and decision-making, he said. 

“We’re trying to do more with less, trying to recover more with a low recovery footprint,” he said, adding that simulations informing enhanced oil recovery for unconventionals is important for the industry. “The percentage of reserves we leave behind in unconventionals is very disturbing.”

Unlocking frac geometry 

Hacker said, he has reached the point where he is “starting to think of unconventionals as becoming conventional, because I think we understand it well enough that I think we can optimize it.”

And the industry is getting “closer to understanding frac geometry” using simulations, said Fervo Energy co-founder and CTO Jack Norbeck. 

Using simulations to understand frac geometry is important not just for unconventional oil and gas but also for geothermals. 

“We need to encourage frac hits to occur,” he said. “What’s happening with far field frac connectivity between wells” controls the flow rates possible for geothermal wells and is a “big question that I think is one that is ripe for exploring in the R&D [research and development] community. I think we’re getting closer to understanding frac geometry.”

Norbeck said in the world of simulations for geothermal activity, Fervo is interested in models that help with well spacing and optimizing completions design. Fervo just drilled a pair of high-temp geothermal wells and carried out stimulation treatments on them. 

“The Permian, from my perspective, has a lot of data,” he said, but that’s not the case for geothermal. “When you really have almost zero field data, then simulation is really all you have to plan.” 

Reservoir simulations are also critical for CCS, said Shashi Menon, SLB vice president of digital.

“It’s a different problem to solve than trying to produce oil and gas, versus sticking something back in” the reservoir, he said. 

Companies use simulation technology to understand how CCS will affect the integrity of the formation and “mak[e] sure the carbon is going to stay there forever,” he said. “I don’t think anybody has cracked that problem yet.”

Ola Miljeteig, Equinor’s manager for CCS technology, said simulations for CCS purposes are “a little ahead of geothermal” but “still nowhere near where they are on oil and gas.”

Mutually beneficial future 

Equinor is working with open source simulations in both oil and gas and CCS in hopes that the community can help solve these problems together, Miljeteig said.

“Open source has a future,” he said. “It can be mutually beneficial for both of us.”

Likewise, machine learning could aid in carrying out computationally intensive simulations “at orders of magnitude less cost,” Guyaguler said.

Hacker said he is “cautiously optimistic” about the possibilities machine learning brings to the table.

“I’ve seen machine learning mostly applied where we have actual data,” he said. “I’m more interested in how we can apply those tools, including AI, to better replicate what’s in our heads.”

AI is also evolving, and the industry is increasingly embracing that technology. But what does it mean for workers?

“AI won’t replace lawyers,” Guyaguler said. “But lawyers using AI will replace lawyers not using AI, and something similar may happen in the reservoir engineering space.”