Oil and Gas Companies Turn to AI to Cut Costs
Artificial intelligence helps predict equipment glitches, manage workers and increase output
By Neanda Salvaterra
Amid a growing push to cut operating costs, big oil is looking to artificial intelligence for help with automating functions, predicting equipment problems and increasing the output of oil and gas.
AI tools can quickly find solutions for the costly problems that can disrupt the business of searching for and extracting hydrocarbons. For example, a faulty well pump at an unmanned platform in the North Sea repeatedly disrupted production earlier this year for Aker BP, a Norwegian oil company in which BP owns a stake. The company finally fixed the problem by installing an AI program that monitors data from sensors attached to the pump and flags glitches before they cause a shutdown, says Lars Atle Andersen, vice president of operations for technology and digitalization. Engineers now can fly in to fix such problems ahead of time and avert a shutdown, Mr. Andersen says.
While Aker BP got the help it needed from a small Austin, Texas-based AI software firm, SparkCognition, some bigger oil-and-gas companies are working with giants in the tech industry. Exxon Mobil Corp. in February started a partnership with Microsoft Corp. to deploy AI programs to optimize its operations in the Permian, or West Texas Basin. The oil giant also recently installed an AI program to gain insights from data coming from millions of sensors that monitor its global refineries. Total SA, meanwhile, is linking up with Google Inc. to better interpret seismic data, and is set to increase its investment in AI to squeeze more hydrocarbons from existing assets.
Royal Dutch Shell PLC, for its part, has tested an AI program that monitors sensors on equipment at its Rotterdam refinery, the largest in Europe, to help figure out where to better direct maintenance staff and dollars. And through a subsidiary in California, Shell has an AI program that helps drivers of electric vehicles shift their charging times to when electricity is less costly.
Advances in machine learning and the falling cost of storing data are key factors in big oil’s motivation to harness the potential of AI. Since 2017, there has been an industrywide drive to move data such as geological information into digital formats that in turn has created vast troves of information that companies can mine for insights using powerful data-crunching programs.
“When you mention data at this scale to data scientists, you can see them start salivating,” says Sarah Karthigan, data science manager at Exxon Mobil, which says it has a database consisting of about five trillion data points. “The intent here is that we can run our plants more efficiently, more safely and potentially with fewer emissions.”
The company deployed an AI program in January to comb through all of the data generated by its 42 refineries and chemical processing plants around the globe. Feeding into what the company calls its “data lake,” all of its refineries now have sensors monitoring things like how much oil is flowing through the system. Exxon uses a machine-learning algorithm to mine its data for problems and solutions such as how to best blend hydrocarbons to get different types of petroleum products. The insights are available to teams of human experts throughout the company.
The drive by oil-and-gas companies to cut costs has grown in importance since oil prices plummeted in mid-2014, disrupting the viability of upstream exploration projects that were planned when crude was still fetching $100 a barrel. While the price of Brent, the global benchmark, is up roughly 6% in the past year, companies are being pressured by investors to keep up capital discipline and find additional cost savings.
AI can help find those cost reductions by tackling a range of problems. Its deployment in upstream operations could yield collective savings in capital and operating expenditures of $100 billion to $1 trillion by 2025, according to a 2018 report by PricewaterhouseCoopers. Most companies declined to discuss their exact spending on AI.
“Combining data and analytics can create new business models,” says Martin Kelly, head of corporate analysis at the consulting firm Wood Mackenzie, who adds, “AI is a component of a broader digital transformation that the oil-and-gas industry is undergoing.”
Exxon, the world’s largest listed oil-and-gas company, is plowing about $1 billion a year into research where machine learning is included and says it intends to increase its spending on AI in the future.
Another factor driving the introduction of AI is its promise in helping capture the knowledge of retiring workers as the industry’s workforce ages. U.K.-based BP has a project called Hands, in which experts in different fields such as water or sand management are training an algorithm that will be able to dispense advice in the future. Researchers at Exxon are also considering creating an algorithm to retain the expertise of its workers, Ms. Karthigan says.
“We have experts in many areas, and they will retire someday,” says Ms. Karthigan. “A lot of their knowledge has been gained through experience, which isn’t captured anywhere.”
France’s Total this year is investing about €200 million ($219 million), or 30% of its research and development budget, into digital technology, of which AI is an increasing part, says Philippe Cordier, a scientific computing program director at the company. In its partnership with Google, Total is testing an AI program in the Gulf of Guinea, off western Africa, that will help interpret data from three-dimensional images of the subsurface and any potential hydrocarbon reservoirs found there.
Finding oil and gas, especially offshore, is a costly process that can sometimes take years. Geologists spend a lot of time looking at seismic graphics to learn about the geological composition of areas they are exploring. The AI program Total is using will help organize such data and identify imagery like fault lines, Mr. Cordier says. The company will continue to rely on humans to identify patterns that might lead to oil and gas discoveries, he says. But by using the technology, Total hopes to get faster, more reliable production forecasts and boost the productivity of its engineers by freeing them from repetitive tasks.