Top of the rock1 May 2019
Predicting what lies beneath the surface of an oil field is a complex and data-intensive task. World Expro speaks to Bill Shea, CEO of Sharp Reflections, who explains advances in data processing are making the guesswork more accurate – and how it’s all taking place within the cloud.
Seismic inversion is the process of transforming seismic reflection data into a quantitative rock/ fluid model of the subsurface. Using the raw seismic data (collected by sending a sound pulse through the ground), geoscientists create a reconstruction of what is happening beneath the surface of the earth.
While this reconstruction is inexact – digging a well would obviously prove more accurate – it does enable them to gather information about the entirety of an oil field, estimating where the boundaries lie between different kinds of rocks.
“In the biggest and most mature fields, you might have a couple hundred wells, and that’s still a tiny sampling of the area,” explains Bill Shea, CEO of Sharp Reflections. “Because geology is not so homogeneous, the inversion is used to understand the rock properties away from the well.”
He adds that, despite its mathematical sophistication, inversion is an inexact science. “You would stick a logging tool down the hole at one place in the seismic survey and directly measure the density or velocity of the rock in the well,” he says. “The measure of how good your inversion result is, is how good is your match between the inversion model and the real measurement at some specific locations.”
At Sharp Reflections, a Norway-based data analysis company, inversion is just one piece of the puzzle. The company works with clients to extract information from their seismic data, both in its raw (pre-stack) and processed (stacked) form. Its software, Pre-Stack Pro, includes an inversion module to convert data into geological modelling.
There is an inherent limitation with seismic data, in that it doesn’t measure the rock properties an engineer actually needs. If you’re looking to drill an oil well, you’ll be most concerned with permeability and porosity, since that’s what determines how the reservoir will flow. Seismic inversion, however, can only tell you about the ‘elastic’ properties of the rock.
“It’s now routine to be able to invert an entire seismic volume to estimates of velocity and density, but that’s not what people care about,” says Shea. “They have some correlation to the engineering properties of porosity and permeability, but those things can’t be modelled directly. So mathematicians keep coming up with cleverer ways to do it.”
Sharp Reflections’ new inversion module, PCube+, is one such example. It is able to predict the rock type (and hence the engineering properties) from the seismic information, along with a probability estimate that a given guess is correct.
“Normally simulating the reservoir requires two steps – first you estimate the elastic properties and then you try to figure out if you can make any sense of the rock distribution from the result,” says Shea. “What PCube+ does, is it tries to go directly from the seismic data to a colour-coded breakdown of the rock units, simulating this in one step instead of two.”
Ordinarily this would be an extremely computer-intensive process. However, PCube+ features an improved algorithm that makes the idea easier to implement. Shea provides a good analogy. If you give a basket of Lego blocks to a group of children, some of them will interweave all the colours quite regularly, while others will put all the reds together and all the greens together. There are billions of possible combinations they could make.
“You could imagine doing that with geology units as well, stacking up 1km of rocks in a huge range of different ways,” he says. “But if you know something about geology, you know not all combinations are possible. If you rule out what we call ‘illegal transitions’, which violate geologic principles, then rather than testing several billion combinations you can get it down into the hundreds and it becomes a practical technique.”
The power of five
The technique, which was developed by Statoil (now Equinor) and the Norwegian Computing Center (NR), has now found a long-term home at NR. “Statoil said, let’s open it up, and get the guys who write the code to start a ‘Geophysical Inversion to Geology’ (GIG) Consortium,” explains Shea. Five different oil companies are now engaged in improving the technique, while Sharp Reflections has a licensing agreement to access those improvements and make them commercially available. “So now it’s on healthy footing and it’s creating a very dynamic innovation space around something that was once off the shelf at Statoil.” For those who have been in the field a while, it’s nothing short of amazing to see such data-intensive processes become everyday practice. And it’s no less astonishing to see how it’s happening – not via on-site servers, but in the cloud.
Number of oil companies working on improving the PCube+ technique.
“There’s a lot of buzz about oil companies moving to cloud computing and part of it’s a cost issue, but it’s also being able to use large banks of computers for computer-intensive tasks,” says Shea. “The challenge for us wasn’t just to be able to get our software running in the cloud, it was to get a high performance version running in the cloud. So we’re creating connected banks of servers that operate as one virtual machine.”
Effectively, this means that if you want to use Sharp Reflections’ software, you don’t have to worry about owning hardware – you can simply access the tools as and when you need them. And while cloud computing has been around for a while, this kind of ultra-highperformance architecture is new.
“With this capability we’re able to do seismic inversions on huge data sets that could span tens of thousands of square kilometres, and do them more or less effortlessly,” says Shea. “We can do that computing anywhere and give you access to it anywhere.”
Lead by example
To take just one example, during the first week of January 2019, Sharp Reflections received freshly processed seismic data in Houston, which it stored in an Amazon cloud. The processing was completed on surveys in Sydney, Australia. Then it began detailed interpretation with OMV (the client) and Western Geco (the supplier) in New Zealand, all before the end of the month. “This would have been unthinkable without this cloud capability,” says Shea. “If you have software that’s written to perform very well in cloud data centres with lots of servers, it brings the data quality assurance to a faster, more collaborative place.”
He says that to some extent, Sharp Reflections has played the role of evangelist, convincing the early adopters to give high-performance cloud computing a go. Now we are approaching the tipping point where it will become mainstream practice.
“There’s still some resistance to ceding control of your data to these public cloud vendors, but it’s amazing how rapidly attitudes are evolving,” he says. “The model for how you pay is changing fast.”