AI and machine learning (ML) are used as a tool in the oilfield to optimize hydrocarbon output and predict failures before they happen. But the industry still has some kinks to work out.

Engineers still have to make adjustments through trial and error to find the sweet spot between technology and human intervention, industry experts said at the Society of Professional Engineers’ (SPE) Artificial Lift Conference and Exhibition on Aug 20.

Advanced data analytics provide the necessary information to design artificial lift systems. But the complex data also shows the industry is far from being completely hands-off, Courtney Richardson, an engineer at Occidental Petroleum, said at the conference.

“The reason that I say that is if you've looked at any high-resolution surveys for wells lately, we are trying to pump through some very complex wellbore geometries and sucker rod pumping that creates a unique challenge,” Richardson said.

Producing multiple wells from a single pad complicates the problem.

“If you're looking at multi-well pads, those four well trajectories are very different well by well, so we take those case by case,” she said.

While software and related algorithms are helpful, there is still no substitute for human input, Richardson said.

“I think that the predictive design software that we use is great and it gives us a fundamental base for designing our wells,” she said. However, there is a large volume of input within the predictive design software that the industry does not fully understand, she added.

At the design stage, a hands-on approach is necessary to understand the logic and build around possible scenarios. But that doesn’t mean Occidental isn’t taking advantage of AI and ML where it can—such as in surveillance and optimization, Richardson said.

Although high-level data analytics, ML and modeling technology are economically feasible for many wells, there is a cost-based limit on its application, especially for those managed by a small group of engineers.

“We've got 8,000 sucker rod pumped wells and the vast majority of those are very low marginal producers,” said Chevron engineer Amine Zejli. From an economic standpoint, it doesn’t make sense for Chevron to spend money to optimize every well, Zejli said. Reliable automation is imperative.

Richardson noted that high frequency data sampling and cloud computing is transforming the way that “we classify our failures, and it's revolutionizing the way that we design and mitigate those failures,” Richardson said.

The combination of technologies revealed previously unrecognized well dynamics, enabling the industry to be more aggressive with failure mitigation efforts.

Ultimately, the goal is to build capabilities to address specific field needs and engineering problems as they arise.

Occidental is also trying to optimize power usage by applying synchronous electric motors—where the rotor rotates at the same speed as the magnetic field in the stator—in their beam pumping units.

“So we've installed a couple of these permanent magnet motors on pumping units, and we're really early in those trials, but we know that it's been wildly successful in the ESP [electric submersible pumps] world,” Richardson said.

Artificial lift has come a long way from the days of pump-jack driven sucker-rod pumps. The path for future generations will be paved by digital technologies and advanced surveillance methods to adapt to an ever-changing landscape.

“Artificial lift needs digital twin technology to grow,” said Lawrence Camilleri, CEO of Camilleri & Associates. Digital reproductions need to be physics-based so that the fundamentals are captured. The value of this approach lies in comparing potential to actual production as the roadmap for optimizing hydrocarbon output, he said.

The industry also needs to embrace a holistic approach.

“We won't just be artificial lift engineers, we won't be production engineers, we won't be reservoir engineers, we will act as one engineer that sees no limits to the application of math and physics to solve the industry’s most challenging problems,” Camilleri said.

Comments