The oil and gas industry’s technological innovation over the last 150 years is truly astonishing, which is why its lackadaisical adoption of artificial intelligence is so surprising.
Geologists have figured out how to vibrate the earth and use seismic imaging to describe rocks thousands of feet below the surface. Mechanical engineers have designed tools that can steer a drill bit through a narrow band of oil for more than a mile.
Likewise, petroleum engineers have collected billions of data points from hundreds of devices to design the most productive wells. But when it comes to using that data to train an artificial intelligence to generate insights, the industry is still dabbling.
“There’s a lot of interest, there’s a lot of buzz, and there’s a lot of proof of concepts,” said Geir Engdahl, co-founder of Cognite, a Norwegian firm that works with oil industry data. “But when it comes to actually putting things into production, operationalizing it and actually realizing the value, we see very little of that happening.”
Engdahl was on a panel I moderated at Time Machine 2019, an artificial intelligence conference sponsored by SparkCognition, an Austin firm that develops machine learning tools. This is where an explanation of terms should come in handy.
Artificial intelligence is so poorly defined, universally overused and clearly over-hyped that it’s losing meaning. In the popular imagination, AI is a computer with general intelligence that can answer all of your questions in natural language.
In reality, though, AI is an umbrella term for many technologies, a subset of which is machine learning, the most useful AI application yet. This is what most people are talking about when they say AI.
In machine learning, programmers give a computer millions of data points related to a specific problem, then apply a series of algorithms to recognize patterns. The machine purportedly learns from the patterns and provides useful predictions or insights based on probability.
Engdahl’s company worked with SparkCognition and Aker BP, a Norwegian offshore oil driller, to analyze data from pumps and other critical equipment on offshore wells. After the machine was trained, it could identify which parts needed maintenance before they broke down, saving Aker money.
Data and AI firms cannot train the machines on their own, though, and oil and gas firms must commit their top experts to maintain safe operations, Paal Eirik Syvertsen, leader of the smart maintenance team at Aker BP, added.
“We as an operator must dare to use the data, but we need to define the safety boundaries,” he told the conference.
When lives — and billions of dollars — are at stake, companies are reluctant to grant access to just anyone. There are dozens of companies promising huge savings from digitalization, but without a track record, executives are unwilling to spend the money.
“One of these wells by Aker BP generates 60,000 barrels a day, so that’s $1.3 billion in revenue per year, so you don’t take chances on that very easily,” said Yash Kaman, a partner at Kerogen Capital, a Hong Kong-based private equity firm that invests in oil and gas. “Yes, there has been slow pace of change, but there are reasons for that as well because these are very high-value assets.”
High commodity prices often drove past technological breakthroughs in oil and gas. Hundred-dollar-a-barrel oil inspired experimentation with hydraulic fracking and horizontal drilling, two costly technologies at the time.
What’s different in 2019 is that low prices are the driving force. Producers must compete on price, and they know that anything that can lower drilling costs or prevent an unscheduled outage will give them an advantage.
The industry also needs machines that can analyze oil field data in real-time and reduce costs per barrel, not necessarily produce more barrels.
Hunt Oil Co. is experimenting with an algorithm modified from high-frequency stock trading to monitor its drilling rigs in the Permian Basin, said Ken Topolinsky, senior vice president of engineering for the Dallas-based company.
“We’ve got models for torque and drag … but also bit-wear, because clearly if the bit wears out when you’re just about to the end of the hole, and you’ve got to pull, that’s a very expensive operation,” he explained. “If we can get smarter about adjusting the drilling parameters to have the bit last longer, then that’s great.”
The world is awash with oil, and some argue the world has found all the oil it should ever burn. The companies with the smartest machines will survive the longest and generate the highest profits if they invest now.
Tomlinson writes commentary about business, economics and policy.
