Oil and gas production generates enormous amounts of information – a seismic survey in New Mexico, for instance, can generate a petabyte-sized file alone.
“To turn this right into a picture on which to base a call requires a 100-exaflop operation,” says Bill Braun. Chevron CIO, told the audience at this yr's VB Transform: “This is incredible computing power.”
To support this data processing, the multinational oil and gas company has been using GPUs since 2008 – long before many other industries needed and even considered the sort of processing power for complex workloads.
Now Chevron is using the most recent generative AI tools to achieve much more insights and value from its massive data sets.
“AI is an ideal fit for big, established corporations with huge data sets – it’s precisely the tool we’d like,” said Braun.
Gaining insights from Permian Basin data
But it isn’t just individual corporations which might be sitting on huge (and consistently growing) data sets – Braun pointed to the Permian Basin Oil and Gas Project in western Texas and southeastern New Mexico.
Chevron is certainly one of the most important landowners within the basin, which is about 250 miles wide and 300 miles long. With an estimated 20 billion barrels remaining, it comprises about 40% of oil production and 15% of natural gas production within the USA
“They have been a giant a part of U.S. manufacturing history over the past decade,” Braun said.
He noted that the “real gem” was that the Texas Railroad Commission requires all operators to publish the whole lot they on site.
“Everything is publicly available,” Braun said. “You can access it, and your competition can access it.”
Artificial intelligence (AI) will be useful here because it may possibly analyze enormous amounts of information and quickly provide insights.
Overall, the publicly available data sets have “developed into a chance to learn from the competition, and when you don't, the competition learns from you,” says Braun. “This greatly accelerates the way in which through which everyone learns from one another.”
Enable proactive collaboration and protect people
Chevron operates in a big, distributed area, and while there may be good data in certain places, “it's not available for your entire area,” Braun noted. But new-generation AI can overlay these different data points to fill in gaps within the geology between them.
“It’s the right application to fill out the remainder of the model,” he said.
This will be helpful, for instance, with wells which might be several kilometers long. Other corporations could possibly be working within the areas around those wells, and the AI could alert to disturbances so human users can proactively intervene to forestall disruptions for each parties, Braun explained.
Chevron also uses large language models (LLMs) to create technical standards, specifications, safety bulletins and other warnings, and AI scientists are consistently refining the models, he said.
“If there are going to be six exact designs, we don't want our generative AI to get creative and give you 12 of them,” he said. “They should be really fine-tuned.”
Braun's team can be studying how best to optimize models when it comes to geology and equipment in order that AI can, for instance, make a guess about where the subsequent basin may be.
The company can be beginning to use robotic models and Braun sees this as a “huge profit” when it comes to safety.
“The idea is to let robots do the harmful work while humans stand at a secure distance and make sure that the duty is finished well,” he said. “It may very well be more cost effective and fewer dangerous to let the robot do it.”
The boundaries between previously different teams are blurring
In the energy sector, on-site teams and in-office teams are sometimes separated from one another – each physically and digitally. Chevron has worked hard to bridge that divide, Braun explained. The company has merged teams to blur the lines.
“To me, those are the very best performing teams. If the machine learning engineer is talking a couple of problem with a pump and the mechanical engineer is talking a couple of problem with the algorithm and the API, you possibly can't tell who’s who,” he said.
A number of years ago, the corporate also began sending engineers back to school to pursue advanced degrees in data science and systems engineering to refresh and update their skills. Data scientists – or “digital scholars” – are all the time integrated into work teams “to act as a catalyst for a unique way of working.”
“We have crossed that threshold when it comes to our maturity,” Braun said. “We began with small successes and moved on.”
Synthetic data and digital twins help reduce CO2 emissions
Of course, there are major concerns about environmental impacts within the energy sector, as in every other sector. Carbon sequestration – the means of capturing, removing and permanently storing CO2 – is playing an increasingly essential role here, explained Braun.
Chevron has a few of the largest carbon sequestration facilities on the planet, Braun claims, but the method remains to be under development and the industry doesn't know exactly how the reservoirs where the carbon is sequestered will behave over time. Chevron has been running digital twin simulations to make sure the carbon stays where it belongs and generating synthetic data to make those predictions.
The enormous energy consumption of information centers and AI can be a vital consideration, noted Braun. How these often distant locations “will be managed as cleanly as possible is all the time the start line of the discussion,” he said.
