A silver lining

1 May 2019



Large oil and gas companies are teaming up with major providers of cloud computing services to deploy Artificial Intelligence (AI) to improve the interpretation of subsoil data. Jim Banks discusses how AI could facilitate a paradigm shift in the industry’s approach to geoscience with Dr Eirik Larsen, an expert in AI and data management.


Industries of all kinds are investing heavily in Artificial Intelligence (AI). There is widespread anticipation that its ability to not only crunch data, but also learn to interpret that data effectively and at high speed to give accurate models of current conditions and future trends, will deliver big dividends. The automotive industry is already well down the road of using AI to develop self-driving cars, the retail industry is putting it to good effect in generating appropriate recommendations for customers, and the oil and gas industry is now turning AI towards the subsoil environment.

Last year, Total announced a partnership with Google Cloud with the aim of using AI to analyse subsoil data. The endeavour makes it possible to interpret subsurface images, including seismic studies, using Computer Vision technology and to automate the analysis of technical documents using natural language processing capability. As the project develops, its goal is to enable Total’s geologists, geophysicists, and reservoir and geoinformation engineers to explore and assess oil and gas fields faster and more effectively.

“The world is changing,” said Jean-Michel Lavergne, senior vice-president of strategy, business development and R&D at Total Exploration & Production when the partnership was announced. “Looking ahead, one driver of change will be artificial intelligence. We’re starting to apply it to geosciences, the field that differentiates oil companies from each other.

“Geosciences allow us to find new sources of oil and gas faster and better than our competitors,” he added. “The purpose of the agreement with Google is to develop systems that make us more efficient. Today, geoscience interpreters spend more than half their time gathering the data they need to be able to perform the value-added work. The aim of this AI project is to shorten the time our teams spend on prep work so they can focus on value-added tasks.”

In the UK, the Oil & Gas Technology Centre, that focuses on innovation to maximise recovery from the UK Continental Shelf, has put out a call for ideas that might enable machine learning (ML) to unlock the full potential of the North Sea; the move comes as operators face difficult decisions about end of field life and the decommissioning of key infrastructure. Its aim is to help develop tools that will open up a better understanding of the remaining hydrocarbon potential in the area, and it has a huge volume of data from decades of successful exploration and field development that innovators can use to build and refine sophisticated analytics algorithms.

With these initiatives, and others backed by small and large energy companies, it is clear that AI and ML have a huge role to play in the future of the industry.

A close eye on AI

As co-founder and CEO of Earth Science Analytics, that focuses on the commercial application of AI in petroleum geoscience, Dr Eirik Larsen is well placed to answer the questions surrounding AI. His company’s goal is to integrate data management, data analytics and ML with the petroleum geoscience workflow. He has 19 years’ experience from the E&P industry, having held technical and managerial roles in oil companies and consultancy firms including Statoil, Rocksource and Geokonsulentene, which has seen him involved in research, exploration, field development and production on the Norwegian Continental Shelf (NCS) and internationally.

“The industry is very excited about the opportunities it sees with this technology, though its use is not yet mainstream in exploration and production,” he remarks. “Currently, companies are planning, making roadmaps and running a lot of proof of concept projects. They are rushing to embrace the open-source AI tools and neural network tools that are available and then working out how to apply them.”

The key reason for adopting the new technology is, for now, the potential increase in efficiency that it offers. Larsen sees AI delivering a 100-fold increase in the speed at which subsoil data can be analysed, but also believes that this is just scratching the surface of what AI can do.

“As we progress, we will see it used for predicting rock and fluid properties in wells,” he says. “This is now done using equations that need to be tuned by experts, but there are not enough of them. AI algorithms can train on well logs and measures from ground cores, and they can learn. The efficiency improvement means that you can obtain fluid property curves for all of the wells in which you are interested, not just a sample.”

“This can fuel creativity in exploration on a datadriven basis,” he adds. “When a new licensing round is announced, there is a huge acreage to explore, which is difficult given that there are always resource constraints. Now, with AI, you can look at everything. You can examine more leads and prospects and select the best high-grade opportunities from a large set rather than from a small set.”

Not only does this have major implications for the exploration of new discoveries, but it also helps the industry with the growing need to make existing assets last for longer.

“There are many mature facilities that owners and authorities want to keep going, and in which they want to open up new volumes. With AI, they can identify exploration opportunities in a datadriven way. Many decisions have been made in the past because people were digging into historical data and found that wells that were believed to be dry could come into production again. The list is long. We can now automate that analysis using well data,” says Larsen.

A perfect storm

The trend among energy companies to partner with providers of cloud computing services is growing stronger. Total and Google began sounding out the use of AI last year. Equinor, formerly Statoil, is working with Microsoft Cloud in Norway to push forward its digital transformation initiative and drive cloud-enabled innovation. Shell is deploying C3 IoT with Microsoft Azure as its AI platform, and intends to rapidly scale and replicate AI and ML applications across its upstream and downstream businesses.

“Companies are rushing to embrace the opensource AI tools and neural network tools that are available and then working out how to apply them.”

At the start of 2019, Pandion Energy in Oslo kickstarted the process of digitalisation in its subsurface capabilities to drive innovation and enhance efficiency in the exploration of the NCS. It has partnered with Computas and Google, and has mobilised a dedicated team of ML, data science and geoscience experts in an initiative driven by its financial backer, Kerogen Capital.

“Pandion has always had a holistic approach to our exploration activities,” remarked Jan Christian Ellefsen, CEO of Pandion. “Having Kerogen to select us as their first pilot case for developing an advanced digital subsurface platform also reflects on the potential in digitalisation on the NCS.”

For Larsen, these initiatives show how a combination of technologies – AI, ML and cloud services – have come together to enable a step change in the industry’s capability.

“Cloud platforms are a big enabler for this,” he says. “You are working with really large data sets, so you could choose to buy hardware that could be out of date in three or four years, or you can rent that hardware in the cloud, which is what most people choose to do because of the cloud’s data storage and computing power. That is why they are partnering with internet giants like Amazon and Google.”

“Many decisions have been made in the past because people were digging into historical data and found that wells that were believed to be dry could come into production again.”

“We have seen it work well in other domains like retail,” he adds. “It is why Amazon grew so big. It built that capability for itself as a retailer and that allows it to now be the biggest provider of cloud services. The oil and gas industry has been digital for 20 years or more, having used computers to interpret well logs, but now it can do much more through automation.”

The potential for collaboration

Larsen’s company is set to continue developing its AI solution beyond its current capabilities. It chose to focus on predicting rock and fluid properties in wells, then to go through the same process with seismic data and marry the two together.

“Computers consume measured data, and generate rock and fluid curves back to the platform,” he explains. “Then you create a loop because you have built a repository of predicted data that can be checked for accuracy against the measured data that comes back from the well. In the next level of analytics, queries can be made on that data to develop better models. That is the field we are moving into now.”

“What the industry wants out of it is better decisions,” he adds. “It wants to make more discoveries, drill fewer dry wells, increase production from existing wells and increase the return on investment. Those are the end goals. Human scientists could not consume all of the data, partly because of time constraints and partly because the traditional tools cannot handle the full volume of data. Now, we can use the biggest supercomputers in the world.”

Next steps to take

From here, the next step is to allow AI to learn, as it has in other industries. In image analysis in the automotive industry, for example, AI is learning to move from photographs to real-world models through image segmentation. Autonomous cars need to learn how to recognise other vehicles, pedestrians, cyclists and all of the other features they will encounter on the road. They do this by working on a large image library that has been labelled with the necessary parameters.

“In geoscience, you need a lot of labour to mark up different images with properties on which the algorithms can train,” Larsen explains. “There are some national data repositories in the UK, Norway, the Netherlands and elsewhere that we can use.

At the moment, however, companies are doing this on their own. Companies are guarding their projects closely because knowledge and technology give a competitive advantage, so they are reluctant to open up. There is a lot of talk about sharing, however, so perhaps the industry will open up in the future.”

Whether or not the industry is able to find a way to collaborate to deliver the potential that AI promises remains to be seen. Regardless, AI and ML are set to have a momentous impact. The key issue will be how fast the industry can progress without the sharing of innovation, as well as insight.

“As far as I am concerned, this is a gamechanger,” Larsen says. “It is beyond anything the industry has seen in 20 years. It is bigger than 3D interpretation software because of the scale and the 100-fold increase in speed. When we build analytics on top of what we already have, we will open up a new domain.”

Across the industry, oil companies are using AI to better hone the reading of subsoil data.
Dr Eirik Larsen


Privacy Policy
We have updated our privacy policy. In the latest update it explains what cookies are and how we use them on our site. To learn more about cookies and their benefits, please view our privacy policy. Please be aware that parts of this site will not function correctly if you disable cookies. By continuing to use this site, you consent to our use of cookies in accordance with our privacy policy unless you have disabled them.