Upstream Technology

See all articles

Uncracking the code

New 3D multi-phase flow simulation technology makes it possible to predict the flow of both oil and water through rocks using only rotary sidewall cores

Doing less with more is a mantra in low-price environments. Reducing reservoir uncertainty is simultaneously a driving need when devising production plans, so its vital to know as much as possible about how much oil is in the ground and how fluids will move through the reservoir rocks.

Over the last decade, supermajor BP has invested heavily in its Digital Rocks programme, which aims to describe reservoir rocks with a computer in a matter of weeks or months, instead of in a lab, where experiments can take up to a year and a half at BP's Sunbury facility in the UK.

BP Joanne Fredrich, senior adviser for upstream technology at BP.
Both processes provide information about relative permeability and residual oil saturation to help engineers fine-tune drilling and production plans.

"We like to be able to predict the performance of the reservoir," says Joanne Fredrich, senior adviser for upstream technology at BP, who leads the Digital Rocks initiative.

"We like technology that's going to enable us to make the best decisions we can for the business, so we like things to be predictive."

The programme's initial focus revolved around describing static reservoir properties and predicting single-phase flow.

Understanding the static rock properties – how much oil is in the ground, for instance – via the digital platform served as the foundation for simulating how two different fluids would flow through the rocks. Simply predicting how oil alone would flow through the reservoir took huge amounts of computing power.

Multi-phase flow is the simultaneous flow of distinct fluid phases with different properties, such as oil and water, while relative permeability is the measure of how oil flows relative to water when both exist in the reservoir.

BP RESEARCH: The left 3D image is of a reservoir sandstone from a deep-water Gulf of Mexico field. The image contains 1050 by 1050 by 1500 voxels (the 3D equivalent of a pixel) with a resolution of ~2 microns, i.e. the physical dimension is 2.5 by 2.5 by 3.5 millimetres. The pore space is black, and different minerals can be distinguished due to their differing attenuation. This original image domain is mirrored to yield a simulation domain of 1050 by 1050 by 3000 voxels, or alternatively 2.5 by 2.5 by seven millimetres. Mirroring the domain yields a model with more than 3 billion voxels, but this facilitates application of a physically realistic boundary condition for the simulations. The middle image shows the same sample but with the cut-out now showing the location of residual oil following an unsteady-state flow simulation after injection of about two pore volumes of the water phase. The oil clusters are coloured according to size, with blue being small (barely visible) and red the largest. The right image shows the distribution of residual oil in the entire volume following injection of two pore volumes of the water phase. The residual oil saturation is about 25%. Note that the simulation domain is actually twice that shown in the middle and right images due to mirroring. The mirrored domain is not shown here.


Fredrich has described the ability to predict multi-phase flow as the "holy grail" of reservoir engineering (Upstream Technology 1/2014).

At the beginning of the journey, Fredrich says, BP had a person working on the two-phase problem.

"We coded up our own solver, and we could do some problems, but the numerical stability was a big issue," she recalls.

Without that numerical reliability, the simulations generated insights but not predictions.


In 2013, BP began collaborating on that complex physics problem with Exa, a company focused on engineering software for simulation-based design.

"We knew our underlying flow physics technology had underlying advantages for multi-phase flow simulation," says David Freed, vice president of oil and gas at Exa.

"We felt we had the right technology basis to tackle that problem, but we knew we needed a partner with domain expertise and suitable compute facility. We found that with BP."

The collaboration used Exa's base simulation engine as a starting point but also drew on work BP had already done.

Exa SCAN TO SIMULATION: A high resolution 3D image of Berea sandstone (bottom) is used to create the pore volume reconstruction (centre) needed for the Exa DigitalROCK absolute permeability flow simulation (top).

"We thought we'd be able to replicate the physics earlier on. It's a very, very difficult problem," Fredrich says.

"When we started this, we envisioned 18 months, maximum. It's been three years."

An early incident that confirmed to Fredrich that her team was on the right path was when simulations accurately predicted two-phase displacement in lab experiments in a 2D micro-model with idealised geometry.

"I'm still stunned today at how accurately we captured the shape of the displacement front. We were able to replicate that perfectly," she says.

Being able to predict the reservoir performance requires inputting reliable data about relative permeability.

"To solve the two-phase flow problem, you have to account for the fact that the liquids are attracted or repelled by the solid surface.

"Also, there is an energy potential associated with the two-liquid interface," Fredrich says.

"It needs a whole new layer of physics to simulate as compared to just a single fluid. It's been extremely challenging in the industry and academia. A lot of groups have chased after this."

More than a few scientists think the problem is "uncrackable", she says.



Exa SIMULATION: Water (light blue) and oil (orange) are shown before (top), during (middle), and after (bottom) the Exa DigitalROCK simulated waterflood experiment. Injected water at the top displaces the oil and leads to its production at the bottom of the domain. Globular oil blobs at the end of the water flooding highlight the water wet wall properties

However, in early May the two companies essentially announced they had cracked the uncrackable when they publicised a multi-year commercial agreement centred on multi-phase flow simulation technology.

BP and Exa moved aggressively two years ago to validate the simulation predictions against lab experiments using cores.

Exa's DigitalROCK software supplements the operator's Digital Rocks programme, which includes experts in 3D imaging, fluid mechanics, numerical modelling, computational physics, high performance computing, rock physics and reservoir engineering.

The Digital Rocks technology uses images of rock core samples created using ultra-high resolution CT scans to generate a 3D digital model of the rock.

The CT scanner functions to resolutions equivalent to 1/50th of the thickness of a human hair and images the sample from about 3000 different angles.



"If you don't have relative permeability data, you have to guess."
David Freed, Exa

The resulting images are then put through proprietary algorithms at BP's Center for High Performance Computing (CHPC) in Houston to characterise rock properties.

BP's aim is for the new capability to help engineering teams make more informed decisions on wells, production facilities and resource progression.

Fredrich says the predictions generated through the two-phase flow programme will also help optimise enhanced oil recovery (EOR) programmes.

Reservoir performance under waterflood and EOR is gaining importance as some of BP's premier oilfields enter new stages of life, including Mad Dog, Thunder Horse and Atlantis in the deep-water Gulf of Mexico, and several North Sea fields.

"The reservoir engineers, it's like they're beating down the door. They want the technology now. It really is the holy grail. It's the cream at the top. It's a big win for us," Fredrich says.

"This is a key enabler for advanced waterflood recovery. It will change our ability to manage our reservoirs as they enter the mid to late field life."


Freed says the DigitalROCK software represents a real breakthrough in flow simulation technology as it is the first time that multi-phase physics has successfully been applied to relative permeability.

"If you don't have relative permeability data, you have to guess," he says.

Stephen Remondi, Exa's president and chief executive, adds: "If you don't have confidence in your simulation model data, you can't rely on it, and if you can't rely on it, you can't make operational decisions."

Exa's predictive technology is based on the Lattice-Boltzmann method, which simulates flow by tracking movement and collisions of fluid particles in space and time.

DigitalROCK needs a precise digital representation of the piece of rock input into the system, and that forms the geometric input to the flow simulator. It can simulate hydrocarbons, or oil phase, and brine, or the water phase.

"This is vastly more complicated than single phase flow," Freed says.

Exa SINGLE FLOW: Exa's DigitalROCK software performs detailed flow analysis for the exact pore space geometry and allows the identification of critical flow paths of higher flow rate (red) within the rock.

Remondi says the complication stems not just from the complexity of the interfaces between the two fluids but also because it is then combined with the complex pore space geometries of the reservoir rock.

"The simulations predict how much oil will remain trapped during waterfloods with different amounts of water as well as relative permeability.

"If I push a certain amount or with a certain force, how well do the fluids move under that driving force?" Freed asks.

"If it moves easily, I'm able to get the fluids out. If it's a low relative permeability situation, it's hard to get them out."

The basic DigitalROCK workflow calls for first scanning a rotary sidewall core and processing the slices of the image into a 3D structure.

The initial conditions, such as the oil saturation, are determined, and the rock image and basic parameters are uploaded into the software.

The engineer determines which conditions to test, such as pressure or inlet flow rate, and the software simulates the multiphase flow rate.

The operator may specify parameters such as the wettability of rock, flow rate of water, and the viscosity ratio, or how heavy the oil is.

"The engineers only need to specify those few qualities and everything else is automated. They don't have to have computational fluid dynamics expertise.

"That's all handled automatically by our technology. The interface is simple," Freed says.

The goal is to ascertain the range of permeability, which makes it possible to "bound the uncertainties", Freed adds.

"Relative permeability is amongst the highest uncertainties, and you get a lot of value from bounding those uncertainties."

From there, Remondi says, the engineers can work to see how changing certain variables yields different results for enhanced recovery.

He expects that initially customers will provide rock samples to Exa for processing and then compare those results with their own lab results.

"BP clearly has a big head start and has more detailed understanding of what it took to get here," he says.

Freed believes future development of the DigitalROCK software could focus on expanding flow simulation capabilities to other things going on in the wellbore.

Core problem

Obtaining a whole core in deep water can run $15 million to $20 million, whereas the digital rocks technology makes it possible to predict reservoir performance using rotary sidewall core.

"We can work with very small pieces of rock," Fredrich says, down to a piece the size of a thumbnail.

Doing lab work using old core can also be problematic. The core must be treated, which typically involves cleaning with solvents, and then restored to its original state of wettability before scientists begin measurements and experiments.

But there is no confidence in the ability to restore old core. Some are "reluctant to do experiments on old rocks because time can 'age' surfaces," Fredrich says.

"People are reluctant to work with old core because they are worried they will not be able to get it back to the reservoir state."

The high-fidelity data available through the digital effort enables the full range of rock types in the reservoir to be characterised.

Exa MULTIPHASE FLOW: Exa DigitalROCK multiphase flow simulation of water (blue) and oil (brown) in a Berea sandstone pore space with water wet walls. Steady state and unsteady state relative permeability simulations can be used to obtain the relative permeability curve for a given rock sample, which is valuable information.

"We can understand the lower-quality rocks," she says.

"Then we can build a subsurface model that much more accurately predicts the reservoir performance in later life when were not just sweeping the nicest sands in the reservoir."

Fredrich says a major advantage in the digital approach boils down to throughput.

The lab in Sunbury can run about a dozen experiments at a time, each taking 12 to 18 months to complete, while in early May BP was running about 50 digital experiments that each take about a month to complete.

"A limited number of [lab] tests can be going on at any time," she says. "Digitally, the only constraint is the HPC," a "beautiful facility" with six petaflops of computing - capacity.

During validation of the two-phase flow software, the Digital Rocks effort was using about 20% of the HPC computing power, with about 30,000 processing cores running.

Generally, Fredrich says, the programme uses around a tenth of the HPC's computing power.

The code behind the predictive software scales well, she says, so if BP needed an answer on something in two weeks rather than a month, the company could double the number of processors devoted to the task.

In May, BP was carrying out final validation of the dual-phase flow prediction software and is now preparing to roll out the technology across its portfolio.

"Every three months, people in assets were checking in with me, wanting to use this technology," she says.

As it stands, DigitalROCK only predicts oil and water flow. "Adding gas is a later stage of development," Fredrich says, noting BP has many reservoirs that have three-phase flow.

"That's a massive additional set of physics. But we have line of sight to getting there."

Read Next:

Going with the flow

On a quiet cul-de-sac on the leafy outskirts of Cambridge, UK, scientists are studying the physical and chemical processes of flow. Their work has consequences that go well beyond the oil and gas industry’s EOR needs

6 Jun 2017
Latest news
Most read