Upstream oil and gas offers some of the most promising opportunities to apply artificial intelligence, including speeding up the identification, targeting and optimization of hydrocarbons while also saving untold millions of dollars and hours of labor.
Functions such as assessing the geology of petroleum basins, drilling with continually improving precision and efficiency, gauging a much wider range of potential outcomes, and forecasting the production of specific oil and gas basins naturally lend themselves to AI applications because of the sheer volume of data and imaging required, experts said.
India-based consultants at Mordor Intelligence estimate spending on AI in the oil and gas sector to be $2.4 billion in 2023 and nearly double by 2028 at $4.2 billion, with North America accounting for the largest share of the market.
Oil price reductions have, in part, driven increased demand for AI in the upstream oil and gas sector, Mordor said on its website.
"Margin constraints ... forced oil and gas operators to change their priorities from raising overall output to successfully optimizing it," the consultancy said. "The factors propelling the growth of the global artificial intelligence market in the oil and gas industry include eliminating the costly risk of drilling, utilizing big data to improve operational performance, and transforming the traditional production system into new predictive technologies."
To be useful, data that is generated must be properly categorized and analyzed. In some cases, sizable volumes of data and complex functions have historically taken months to process, but AI experts said they see that time frame being cut to minutes or seconds.
Tailoring models
What makes AI so different from advanced computing is the ability to tailor models to extremely specific needs, regions and schedules, according to Alex Fender, a partner and AI expert at consultancy Kuiper XYZ.
For example, in reservoir modeling, a one-size-fits-all mechanical solution to running simulations is no longer acceptable, Fender said. Exploration and production operators need to be able to process and analyze massive amounts of specific data synthesis geared not just to a single large reservoir "but [also] the sub-layers" and intervals within geological horizons, according to Fender.
"It would be tailor-made to this particular square mile or [interval] or whatever area of rock, giving you thousands and thousands of combinations to figure out the most accurate way" of how to proceed, Fender said.
The machine-learning aspect of AI may not yield apparent early results in some cases — seismic imaging, for example — but subsequent results are likely to improve as more variable data are continuously added to the models.
AI can also be deployed in critical functions such as predictive maintenance of oil and gas installations both onshore and offshore, where AI "can track toxicity levels and leaks and warn users of issues that must be fixed," Mordor said. And AI can identify temperature fluctuations and automatically modify cooling and heating systems where needed throughout the year, the consultancy said.
Digital twins and high speeds
Several companies are deploying so-called "digital twins" — virtual copies of real-life product, system or process — to attain the overall health of the oil wells, said Mordor. Digital twins calibrate well requirements and analyze conditions to improve overall functional quality.
For instance, Mordor noted that Chevron Corp. is rolling out digital twin technology for its oil fields and refineries. The major oil company "expects to save millions worth of quality and other maintenance costs," Mordor said.
In a paper published in "Energy and AI" in March 2021, Dmitry Koroteev, founder and CEO of Digital Petroleum LLC, detailed some of the ways AI can be used to address existing industry challenges:
– An AI tool to accelerate conventional reservoir simulations via deep neural networks — which seek complex nonlinear relationships between data sets — makes it possible to screen through a very large number of field development scenarios before selecting the most optimal one, potentially accelerating that process by a factor of 200 to 2,000.
– AI-assisted geological assessment of reservoir rocks, also using deep neural networks, has the potential to remove human errors that cause wrong mapping. This results in more accurate definitions of correct hydrocarbon targets, cutting the time needed for manual mapping from several weeks to several seconds.
– A tool that detects types of drilled rock and potential failure using real-time drilling telemetry via a combination of machine-learning algorithms can maximize the contact between the wellbore and the pay zone. It also offers time savings of up to 20% and money savings of up to 15% for well construction.
– Another tool developed for rock typing — based on images of rock samples extracted from wells — cuts the time needed to achieve responses, which can be sped up by a factor of over 1 million.
Generative capabilities a challenge
Jeff Fleece, chief information officer for oilfield services provider Baker Hughes Co., said figuring out how to best apply the "new, generative capabilities" of AI remains a challenge.
"While some form of AI has been around for decades, generative AI is where the machine learning capabilities are now taking in massive sets of data in the billions if not trillions of data points, so the scalability of the answers and solutions that can be computed from those are much more powerful than we've had from traditional models," Fleece said. "But where they can be applied has been the biggest challenge."
"Companies are exploring to determine what the direct applicability would be — and also the risks and downsides" that could result from these powerful tools, Fleece said.
Cybersecurity, for example, is an area where AI lends itself not only to solutions but also to risks, Fleece said.
"If code can be auto-generated, then you can also have malicious code generated, and the risks of that scaling much faster, much more quickly than a traditional cyberattack are one of the areas that cyber-professionals are concerned with," Fleece said.
"You also have concerns with companies that have a high degree of intellectual property and also allow broad use of generative AI tools by their employees, where the employees don't necessarily understand the security ramifications and end up leaking [data] out of the company," Fleece said.
S&P Global Commodity Insights reporter Starr Spencer produces content for distribution on Platts Connect. S&P Global Commodity Insights is a division of S&P Global Inc.
S&P Global Commodity Insights produces content for distribution on S&P Capital IQ Pro.
