The EAGE Forum Brains, Machines & Rocks: Assessing the Digital Revolution heard from Gabriel Guerra (Shell), Maitri Erwin (Microsoft Azure for Energy) and Steve Freeman (Schlumberger) with EAGE First Break Editor Andrew McBarnet in the chair. To summarize the debate, all were agreed that the cloud and artificial intelligence were ‘big’, and that the future will be ‘open’, with reference to Shell’s OSDU push and Schlumberger’s contribution thereto. McBarnet suggested that geoscientists were skeptical, considering machine learning as ‘just another tool’. All immediately agreed… ‘ML is good for specific simple tasks, you still need a person in the loop’ … ‘explorationists think in broader and different ways’ … ‘in highly faulted deep thrust will still need a human’. The quality of the data underpinning AI is critical, this is a ‘huge issue’ … ‘a fundamental problem’ … ‘the rational for OSDU’ … ‘ML has a huge role, but we need to get beyond the mediaeval stage of data management’. ‘This is a huge task where companies are struggling – especially in the current cost cutting environment’.
The discussion turned to the required skill set for the new workforce. For Erwin, Universities still offer ‘siloed, over-specialized’ approaches to education. Industry does not necessarily need ‘fully-fledged’ data scientists, but some coding is needed*, along with a ‘passion for data science’. For Guerra, there is still a lot of resistance to the new data-driven technology even though this is changing. ‘Using a pen and pencil to interpret seismic is not going to last very long’. For Schlumberger, the real resource is people. You can’t just buy data science expertise. It’s much better to provide petrotechs with new skills. If you need a head of IT then the service companies have failed you’.
The zoom format meant that some of interesting questions and comments were asked but not answered. Inter alia …
‘We are pressurized by the IT industry to buy, buy, buy. The IT people do not see how much science needs to be migrated into the new systems’.
‘What are the benefits of AI for seismic processing? Project turnover time? Noise reduction? Better imaging?’
‘How does having a big data mess in the cloud solve anything?’
‘How will digitization help with the energy transition?’
‘How should a medium size company gets into the digital transformation space? What are the fundamental questions should one ask before thinking or deciding on moving with a digital transformation?’
The EAGE Forum Session titled ‘Energy Transition: How Fast Really?’ purported to explore the reality of the energy transition and what it means for the geoscience community. Host Andrew McBarnet introduced panelists Bob Fryklund (IHS Markit), Philip Ringrose (Equinor) and Iain Stewart (Plymouth University).
Ringrose observed that many oil and gas companies are reporting carbon reduction targets ... but are they real? Businesses have to reposture (sic). There is now a perception that there is money to be made in low carbon energy systems.
Fryklund added that while in the past, the important thing was to please shareholders financially. Now the license to operate is critical and has changed the dialog.
Stewart thought the repositioning was unconvincing and could not see an ‘oil and gas’ company surviving as such.
Ringrose acknowledged that 10 years ago, ‘oils could be accused of greenwashing’. Things are now changing with the move into renewables.
Fryklund sees industry fragmentation with a splinter group of large Europeans that are transitioning to become ‘energy’ companies. Other smaller companies think that they won’t be able to compete with the utilities. ‘What we are good at is producing oil and gas’.
McBarnett asked if carbon capture would get the job done.
Ringrose hopes that we will do CCS in reservoirs and rock formations, building a ‘new, exciting hybrid industry’.
Stewart agreed, the Paris agreement requires negative emissions/CCS. Biotech, AI and quantum computing will all help.
Fryklund was skeptical that folks looking to do CCS would hire an ex-oil geoscientist!
Given the interest in AI and ML in seismics, we dipped in to the papers presented in the special session on machine learning. Topics included Reconstructing missing seismic data through deep learning with recurrent inference machines – Ivan Vasconcelos (Utrecht University). Source de-ghosting of coarsely-sampled data using a machine-learning approach – Jan-Willem Vrolijk (Delft University Of Technology). Automatic parameter selection using k-fold test – Nihed El Allouche (Schlumberger) and others in a similarly specialist vein. We have yet to read a truly killer application of AI in oil and gas. Perhaps the operators are keeping these to themselves.
Read the EAGE 2020 abstracts on EarthDoc.
* Just a little anecdote here. To our personal knowledge, ‘coding’ has been a part of the geophysical curriculum in the UK for at least 50 years.
This article originally appeared in Oil IT Journal 2021 Issue # 2.
For more information or to comment on this topic email here.