By
James
Herd
Hart Energy

Among the giddy proponents of artificial intelligence (AI)—and the doomsayers—the oil and gas industry increasingly embraces the technology used to process vast caches of data, increase safety and even increase well performance.

As with other innovations, E&Ps have once again joined some of the early adopters to see what it can do. One key benefit: it’s a time saver in an industry drowning in data.

“One of the [philosophies] behind using AI and ML [machine learning] is, any task that requires a lot of human effort, any task that requires a lot of computational effort…can be reduced a lot,” Siddharth Misra said, associate professor of petroleum engineering at Texas A&M University, in a panel at Hart Energy’s SUPER DUG conference in Fort Worth, Texas.

That includes some necessary but mundane tasks, such as sorting through terabytes of data.

Data—raw, structured and unstructured—is being captured all the time in the industry, said Ali Raza, chief digital officer at ChampionX. Data analytics can refine the information to increase productivity and monitor a company’s assets, including compressors and engines.

The data is so voluminous that it’s “too much for [a] human to work [their] way through,” said Thomas Johnston, COO at ShearFRAC. So the company employs a real-time fracture guidance technology known formally as FracBRAIN, and the AI component behind it is nicknamed Shear-i. Johnston later told Hart Energy that the FracBRAIN technology “measures pressure patterns and interprets how the rock is fracturing,” and the implementation of Shear-i offers change suggestions to the “rate, proppant concentration and viscosity to more efficiently and effectively fracture the rock.”

The technology is expected to have practical applications in the field. During the panel, Johnston said that the utilization of the FracBRAIN technology in conjunction with the Shear-i AI could, in time, increase production by an estimated 5%.

Misra added that AI constructs can help to manage the “large volume of data” that is “coming from multiple data position sources.”

“That’s where a bot can take all the data, it can help [with] data reprocessing, data visualization, data entry,” Misra said. “Bots are really good at information retrieval.”

Learning machines

Performing even regular maintenance on the equipment that keeps the oil patch pumping can be dangerous—more so if the workers are relying on incomplete or incorrect information. Predictive analytics and predictive maintenance are two concepts that go hand-in-hand. If workers can—with the help of AI constructs and past data—predict when failures might occur, some dangerous tasks can be mitigated.

“Every time something happens, the [AI] model keeps learning,” said Raza, who noted that ChampionX promotes continuous and positive learning for its AI until it is able to recall past events with an accuracy of 97% to 98%.

Johnston illustrated machine learning with a more practical example, recalling a visit to the Houston Botanical Gardens. He noticed that sprinklers started watering the gardens after a rain shower. Intelligent usage of AI could reduce such a “pointless” waste of water, he said.

According to Johnston, the current automated process does utilize AI in some capacity—a timer is set-up that simply waters the plants at regular intervals—but it could be approached differently.

“You can have a look at—did it rain one inch in the last hour? Okay, therefore don’t water,” said Johnston. “And then you get even more intelligent and say ‘hey, in the next hour, what’s the prediction that it’s going to rain?’ and you can keep getting more and more and more [specific].”

Rise of the machines

Sebastian Gass, CTO of Quantum Energy Partners, offered a cautionary note: be careful what you share with an AI.

As ChatGPT gains traction and acceptance, though with decidedly mixed results, Gass emphasized the contrast between private AI engines and public ones.

“Make sure that you do not feed data into an AI engine that you don’t want to feed into the AI engine,” he said during the panel.

It was reported by Bloomberg on May 1 that employees at Samsung were barred from using generative AI tools such as ChatGPT. The ban was instituted after staff had “uploaded sensitive code to the platform.” The company conducted a survey among its employees, with 65% indicating they were concerned about AI as a security risk.

“I think all of us need to be very mindful” about the negative aspects of AI, said Gass.

And for all of the upsides of deploying AI in the oil patch, there’s also a nagging worry.

“I think every technology has unintended consequences,” Gass said. “If you listen to…smart people out there, the statistics [showed] 50% of AI experts believe there’s a 10% chance that AI will wipe out humanity.”