For Shell, AI and data are as critical as oil

At Shell, there are many reasons to use AI and data to transform their business.

From rising energy needs and unconnected environments to increasing pressure to fight climate change, the oil and gas industry is at a crossroads. Energy companies like Shell can either stick to the status quo or embrace the idea of ​​a low-carbon energy future.

The transition to a more distributed, diverse and decentralized energy system means optimizing end-to-end processes and maintaining them on a scale. This means that solutions that can be deployed at a rapid pace worldwide are crucial. And that means Shell had to become an AI-powered technology company.

Accelerate digital transformation

For example, last November, Shell co-founded the Open AI Energy Initiative (OAI) with Baker Hughes, Microsoft, and AI company C3 AI to help accelerate the energy industry’s digital transformation.

According to Dan Jeavons, Vice President of Computer Science and Digital Innovation at Shell, the OAI offers industry leaders the opportunity to work together openly, fairly and transparently. This enables them to create interoperable standards between AI applications and accelerate the adoption of digital technologies and achieve net zero emissions in the future.

“We are committed to being net zero by 2050 or earlier and to achieving a 50% reduction in magnitude one and two emissions by 2030,” he said.

While digital technology may not be the silver bullet, it is one of the core levers that Shell uses to accelerate the energy transition. Jeavons adds, “While we are going to have to transform a lot of hardware to change the energy sector, we can also utilize the data we have today and use it to transform the system.”

AI plays a critical role in Shell’s business strategy

Shell has already implemented several AI initiatives over the years, including the deployment of reinforcement learning in its exploration and drilling program; the roll-out of AI at public electric car charging stations; and installing computer vision-enabled cameras at service stations.

Recently, the company also launched the Shell.ai Residency program, which enables computer scientists and AI engineers to gain experience with a variety of AI projects in all Shell businesses.

Shell is currently deploying north of 100 AI applications into production each year. They have also developed a central community of more than 350 AI professionals who design AI solutions using large pools of data available in the many businesses within Shell.

AI helps Shell with predictive maintenance

“Reliability and security are absolutely fundamental,” Jeavons said. “Having the ability to identify when things are going wrong and intervene proactively was a priority for us.”

AI allowed Shell to use predictive monitoring to complement monitoring techniques they already had in place.

To put this in perspective, Jeavons claims that it has more than 10,000 pieces of equipment currently monitored by AI – from valves and compressors to dry gas seals, instrumentation and pumps, while AI also provides predictions about possible failure events. To monitor all that equipment, 3 million sensors collect 20 billion rows of data each week, while nearly 11,000 machine learning models allow the system to make more than 15 million predictions each day.

Historically, Shell has relied on physics-based models to make these predictions. Before the advent of a predictive maintenance program run by C3 AI, the company will usually replace parts after a certain period of time. This approach meant that parts were often replaced while still in good condition. An alternative strategy was to wait until something failed. With equipment failure, assets needed to be temporarily closed for repairs, affecting production.

AI-based predictive maintenance has enabled the company to reduce equipment and maintenance costs by using resources more efficiently, reducing production interruptions and avoiding unplanned downtime.

Tom Siebel, CEO of C3 AI, explained that there are numerous infrastructure and orchestration issues surrounding AI.

“It’s not that difficult to build machine learning models,” he said. “What is difficult is to put two million machine learning models into production in one application.”

However, with a proactive technical monitoring approach, Shell’s data scientists were able to analyze thousands of data points simultaneously, enabling engineers and others to draw insights from that data.

“Our team uses that data to understand what normal behavior across our asset base looks like in specific cases, including equipment such as compressors, valves and pumps,” says Jeavons. “Then we create predictions of what we think will be normal in the coming periods. From that prediction we can identify when normal conditions no longer occur and then link it to historical events. ”

AI for optimization is next for Shell

Now, Shell has commercialized its AI predictive maintenance applications built with C3 AI software. From now on, Jeavons says the company is now laser-focused on optimization.

“This means we can identify ways to produce more efficiently, generate more output for the same cost and more importantly, we can also look at the CO2 footprint of these processes and start optimizing accordingly,” Jeavons said.

In the near future, he added, Shell is also exploring how AI can be used to monitor carbon capture, storage facilities and methane levels.

“These businesses involve making our existing business more efficient and effective, but also play a key role in our energy transition strategy,” he said.