Skip to main content

S-Cube algorithm helps Tullow discover oil in Guyana

Published by , Editor
Oilfield Technology,


Discovered volumes from high impact drilling fell overall by 50% in the 2014 – 2018 period compared to the previous five years. Only a little over half of the decline can be accounted for by a lower well count which fell by 28%; falling success rates and average discovery sizes accounted for the rest. Better subsurface predictive analytics is the key to turning this around according to many explorers like Tullow Oil. For the most data intensive critical aspect of this process - constructing the underlying rock velocity model from seismic field data - Tullow have turned to S-Cube and Amazon Web Services (AWS) for next generation waveform inversion on the cloud known as XWI.

Seismic data from active seismic surveys is by far and away the most important way to probe the interior of the earth. The exploration and production (E&P) industry generates exabytes of it per year for evaluating the subsurface potential as it strives to better define buried rock formations, prospective structures and hydrocarbon trapping configurations. Waveform inversion of seismic data is the data intensive computation for optimising an earth model requiring vast compute resources to utilise the full recorded response of the earth. When performed successfully, the underlying unobservable subsurface properties it estimates are predictions of the subsurface rock velocity that would be encountered along a wellbore drilled in a given location.

AWS and S-Cube exhibited XWI In June at the International Supercomputing exhibition ISC19 in Frankfurt, Germany. Consisting of an industry-leading nonlinear optimisation toolbox which takes standard methodology to the next level, it enables entirely data-driven earth model building with more automation and more accuracy when starting further from the true answer. The basis for Tullow Oil to select XWI for its waveform inversion were the four guiding principles of accuracy, resolution, differentiated capability and predictive power. For Tullow, this capability can be applied to significantly diminish risk and uncertainty in subsurface imaging and hence in business risks.

Gareth O’Brien, Principal Geophysicist at Tullow Oil, said: “The sheer flexibility of the XWI cloud solution, as well as the advanced nature of the optimisation algorithm that drives, is exactly what’s required to gain an edge in the competitive area of E&P. Where other traditional approaches are seen to diverge and suffer from cycle skipping, XWI succeeds. The velocity prediction it offers can be applied across our portfolio from block screening to drill or drop to optimal well placement in exploration and production. We are in total control with the ability to create, operate and tear down secure optimised HPC clusters in minutes. By running multiple scenarios simultaneously, we were able to increase quality score from a baseline of 57% to 75% by narrowing uncertainty in the sea floor position and density ratio. This directly impacted the resolution we obtained deeper down at the target level. This is an example of gaining extra value through intelligent use of predictive analytics.”

Nikhil Shah, COO of S-Cube, said: “The XWI algorithm marks the start of the big data parameter learning and HPC convergence using cloud compute infrastructure for the energy industry. We continue to advance our search platform to solve for unknowns which affect the quality and accuracy of the final result maintaining predictive power of earth rock property trends below 3 km depth.

“With our cloud-native architecture, we are able to run multiple concurrent initialisations simultaneously and travel through search space in a way which redefines what is possible for resolving deep structures containing significant prospective zones with large volumes of untapped resources.”

Read the article online at: https://www.oilfieldtechnology.com/digital-oilfield/29082019/s-cube-algorithm-helps-tullow-discover-oil-in-guyana/

You might also like

 
 

Embed article link: (copy the HTML code below):