Norwegian energy company Equinor ASA is on a mission to drive the global energy transition with the help of innovative technologies. To do this, the organization has launched an initiative to digitalize its gas trading activities and guide decision-making processes around them using data and insights. Culminated with the creation of a machine learning platform, the project is driving long-needed cultural and digital transformation across Equinor – bringing them one step closer to their sustainable pledges.
“There are many things you can do when you’re driving change. You can raise awareness, run demos and pilots, test certain concepts. But for it to really succeed, you must always put the people you do it for at the heart of it.”
Robin Steen, Delivery Lead at Equinor ASA, is discussing digital transformation at his company and the cultural shift this has triggered.
“At Equinor, we want to play our part in driving the energy transition all over the world,” says Steen. “And we believe that technology and automation are essential to achieving it.”
It’s a mission that Equinor is bidding to achieve with help from Microsoft and machine learning. One that recently led them to build EurekaML, a machine learning platform designed to operationalize AI models and support gas traders in their decision-making processes – ultimately providing Equinor with the tools and analytics necessary to buy and sell its gas products at the best value.
“The energy sector is bracing itself for a huge period of change and evolution which heavily depends on technology to succeed,” he says. “With Microsoft on our side, we’re working to make sure that we’re prepared for that.
“We’re making the most of the opportunities that technology provides and using them to get closer to our business and sustainability goals.”
Evolving times call for a change
A Norwegian energy company, Equinor is committed to providing natural gas, petroleum and renewable energy across 36 countries.
To do this, the company’s Marketing & Supply division has spent the past few years investing in data and building the foundation to become a truly data-driven organization. “Part of what we do can be compared to a bank or a financial institution’s trading activities,” says Steen. “The difference is that in addition to those, we also do asset-backed trading – meaning that we produce the energy and gas that we sell and need to optimize profit through the production, transportation and processing of our commodities.”
But this is an extremely delicate and fast-paced market to carry out such operations. “Where crude oil and oil products can take a month from when they’re sold to when they’re delivered via petroleum tankers, gas takes minutes,” he explains. “For us, that means one thing: to stay ahead of the game we need to know more than anyone else. We need data and insights to guide us through all our trading operations.”
He says that the nature of these insights has evolved radically in recent years. “For a long time, the only kind of insights our traders could use beyond press insights and prices data were gut feelings, rumors, hints from colleagues, social media,” he says.
“Today, that’s no longer sustainable. Our traders know that to beat the competition and get the best value, they need to gather more and better insights. They need to have more robust models to understand what was happening yesterday but also to predict the future and make qualified bets.
“So three years ago, we started building a platform that helps them do that.”
The right opportunity to innovate
Created under the Commercial Digitalization Natural Gas (CDNG) digital program, Equinor’s platform was first conceived to help the company automate trade volumes execution and trading strategies by 2025.
“The CDNG program was born from the need to digitalize trading and improve our response to evolving market environments with advanced analytics,” says Steen, adding that the project initially focused on a specific portion of the company’s trading activities.
“This is what we call our work horse trading, and it’s essentially a very repetitive side of our operations that can only generate significant earnings if carried out in a non-stop, continuous way,” says Olve Austlid, Software Architect and Product Owner at Equinor. “In other words, if a bot gets instructed to do it over and over again.”
That’s exactly what Equinor did. “We had long been investing in our collaboration with Microsoft – for example with its Norwegian data centers – and had built a tenant platform in Azure called Omnia,” he continues. “So, we decided to make Omnia even more useful for trading, and add advanced analytics and machine learning capabilities to it.”
It was the birth of EurekaML.
Introducing EurekaML
EurekaML consists entirely of software components running on Microsoft Azure and is based on Azure Machine Learning services. Its main goal is to help users to perform their work in the easiest and most effective way possible.
More specifically, the platform allows the creation of analytics and machine learning products that can then generate data-driven insights aimed for Equinor traders – with the time taken to deploy an ML solution now down from 1-2 weeks to one day.
This is made possible by a solid data foundation laid with Microsoft Intelligent Data Platform, as well as a range of Azure services that includes Azure Synapse Analytics, Azure Functions, and Azure Cosmos DB.
The data is located on Azure Data Lake and integrated using Azure Data Factory. In addition, the platform benefits from an MLOPs solution that combines ML model development with deployment and monitoring.
Finally, Equinor is using Azure Web Apps and Microsoft Power BI, which allow even the least experienced developers and analysts to set up their own dynamic dashboards.
“Previously, there was no easy way for our traders and analysts to set up analysis environments apart from their laptop and spreadsheets,” explains Vidar Slåtten, Platform Team Lead and former Product Owner at Equinor ASA. “That meant that the analytics were kept locally to each person – and the system was not set up for sharing information.
“Azure Machine Learning has changed that entirely. We don't have to install anything locally and now have bigger clusters for running more compute-intensive tasks for training models.”
Robin Steen adds that while the focus of the platform remains on trading, Equinor is now also exploring potential uses cases post-trade.
“The more data we have, the more curious we become about accuracy and precision, being able to better predict and understand market conditions and trends,” says he. “On top of that, we can also better detect fraud, abnormalities and patterns in our systems.”
Powering cultural transformation with machine Learning
Machine learning has opened up a sea of new possibilities for Equinor, says Olve Austlid. “Our new data products – which generate data in the most trusted and secure way possible – mean that we can now have a more continuous flow of data when it comes to reports, insights and dashboards,” he says.
“On top of that, the platform is so intuitive and easy to configure that users can now take far greater responsibility of their work and no longer have to rely on external help.”
That, says Robin Steen, is testament to the massive cultural transformation that the company’s Marketing & Supply area is undergoing alongside the program.
“Change management is all about finding ways to positively impact the people you’re innovating for, which for us, in this case, are our traders,” he says. “You can't force technology on them. They need to believe that technology is an assistant to them.
“This has been a constant in our work from the very beginning, both in terms of the democratization of data that we’ve enabled and the collaboration between IT and traders that we’ve helped to foster.”
At the heart of it, he adds, is a long-established vision to “treat data as an asset” that needs to be safeguarded and protected, but also made available to all.
“You can’t just start with machine learning,” he concludes. “You first need to understand how you treat your data and make it available. Then you can figure out how to combine it with domain knowledge and machine learning, preferably via a platform like ours.
“That’s exactly what Microsoft is helping us to do: laying the foundation for a data-driven, intelligent future that we can build on day by day.”
“The energy sector is bracing itself for a huge period of change and evolution which heavily depends on technology to succeed. With Microsoft on our side, we’re working to make sure that we’re prepared for that.”
Robin Steen, Delivery Lead, Equinor ASA
Follow Microsoft