The big buzzwords of the day are “artificial intelligence” and energy experts say AI will be increasingly useful in the oilfield only if guided by human intelligence and experience.
Odessa oilman Kirk Edwards, Waco economist Ray Perryman and 2023 Permian Basin International Oil Show President Larry Richards say an over-reliance on AI could lead to disaster.
Edwards said the technology “has been sneaking into the oil and gas industry for the past few decades or so in so many ways.
“The most obvious to me is the new horizontal drilling technology whereby we are seeing signals from the earth some 15,000 feet or more deep where only a computer program can disseminate the data to perfectly process it and tell us what the earth is trying to say,” he said.
“This makes the difference in staying in zone while drilling or getting outside the zone and making a poor well. Schlumberger has always used computer technology early in their well logging process and it has made a huge difference in what that technology is doing for us today.”
Edwards said AI “will continue to grow in our industry as there is too much money at stake not to make things more efficient and profitable ahead.”
Perryman said the oilfield is in the early stage of applying the technology.
“We’re only beginning to scratch the surface of how AI can be effectively incorporated into businesses of all types, including oil and gas,” he said. “Clearly there are some things AI does very well such as assisting with analyzing the vast amounts of data now utilized in oil and gas exploration.
“It can also potentially make certain processes more efficient, monitoring and analyzing in real time.”
Nonetheless, Perryman said, it will be some time before AI can replace substantial human functions in the energy sector.
“Obviously parts of the business are physical and require human labor and much of the other parts of the process are now run by skilled technicians,” he said. “I have been on rigs in every decade since the 1980s and it is a completely different world in terms of sophistication and automation.
“At this point AI is not particularly adept at creativity and it has no judgment, which is often required in real time on a rig. You have likely read the recent stories of the attorney who filed a case using AI to draft parts of it only to realize that the AI had completely fabricated case law.”
Perryman said the technology can help draft some parts of reports, articles or other documents, but it can’t do so with a human voice “or even an eye toward the specifics of the truth.
“I have even had a humorous experience where I was introduced using an AI bio,” he said. “We’ve made amazing strides, but at this point AI still has notable limitations for an industry with the precision requirements of oil and gas.
“For the oil business, it will take some time to identify and test the best ways AI can be used effectively, efficiently, reliably and safely,” Perryman said. “It’s obvious that there will be ways in which it can improve many aspects of the industry that will be identified and implemented over time.
“However, it is still an emerging technology and companies will rely on human judgment, skills and instinct for the foreseeable future. It’s probable that producers will begin to use AI tools for certain tasks, but I expect that any major shift remains years away.”
Richards said the technology “has enormous potential and opportunities but equally daunting risks.
“As the oil and gas industry incorporates the use of AI to crunch massive sets of data and improve algorithms, we’ll need to constantly manage that risk-reward equation,” he said. “We have used technology and automation, coupled with data collection and algorithms, for years to improve field productivity.
“For example you have automated valves tied into an oil company’s computer system on the pipelines that are taking produced water from a well site to a nearby disposal well,” Richards said. “Sensors with transmitters on the line can sense when pressure on the line drops dramatically, suggesting a leak, or spikes, suggesting a blockage.
“In either case you can write an algorithm in the program that allows a signal to be sent to the appropriate valve to automatically shut the valve and stop water flow until it can be researched by a field team.”
He said these types of automation “are incredibly helpful to reduce spills and increase efficiency.
“With machine learning you constantly feed all the data of years of disruptions and shutdowns to the computer system and the program is actually rewriting algorithms and programs on its own based on pre-set directions from lessons learned and past experiences,” Richards said. “One of the more promising uses of AI machine learning is to identify the root causes of equipment failures in a system and develop predictive maintenance models and programs.”
An algorithm is a finite sequence of rigorous instructions that are typically used to solve a class of specific problems or to perform a computation.
Richards cited his experience in leading teams challenged with putting together large-scale preventive maintenance programs for rigs, big trucks and compressor packages. “The Holy Grail of these type projects is to achieve a truly predictive maintenance program that constantly learns from each day’s new data and allows better decisions to be made by the team,” he said.
“Huge savings are found by reducing unplanned failures and downtime in our industry and the key to success is always to utilize the experience and expertise of your field personnel.”
“They usually know more about the failure points and timelines of equipment than the manufacturers or engineers and their expertise is often under-appreciated,” Richards said. “These individuals will be the key to developing AI programs that actually work.
“Their experience will be key not only to develop the information driving these programs but also to correctly evaluate the information over the first several years of implementation.”
The industry has learned that with automation increased sophistication often means more potential for malfunctions “especially in the brutal field conditions of the Permian,” Richards said.
“The challenge will be maintaining a healthy level of common sense at every step of the process,” he said. “We also need to understand that the risks are very real. Computer systems can be hacked and giving control authority to any third party, much less a machine, can have fatal consequences.”
Richards said it makes perfect sense to use AI machine learning to crunch massive geological and seismic data sets to find oil deposits and produce them.
“Utilizing AI to monitor and control some level of field automation probably makes sense with appropriate controls and over-rides in place,” he said. “Tying the automation of pipelines and blending programs at the intersection and choke point of over half of our current U.S. oil production in Cushing, Okla., would not.
“As with most things we need to go in with our eyes wide open, understand the risks as well as the potential benefits and use a lot of common sense at every step in the process.”
