Publications
544 results found
Rücker M, Bartels W-B, Bultreys T, et al., 2020, Workflow for upscaling wettability from the nano- to core-scales, International Symposium of the Society of Core Analysts
Shams R, Masihi M, Boozarjomehry RB, et al., 2020, Coupled generative adversarial and auto-encoder neural networks to reconstruct three-dimensional multi-scale porous media, JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING, Vol: 186, ISSN: 0920-4105
- Author Web Link
- Cite
- Citations: 47
Akai T, Bijeljic B, Blunt M, 2020, Local Capillary Pressure Estimation Based on Curvature of the Fluid Interface-Validation with Two-Phase Direct Numerical Simulations, ISSN: 2555-0403
With the advancement of high-resolution three-dimensional X-ray imaging, it is now possible to directly calculate the curvature of the interface of two phases extracted from segmented CT images during two-phase flow experiments to derive capillary pressure. However, there is an inherent difficulty of this image-based curvature measurement: The use of voxelized image data for the calculation of curvature can cause significant errors. To address this, we first perform two-phase direct numerical simulations to obtain the oil and water phase distribution, the exact location of the interface, and local fluid pressure. We then investigate a method to compute curvature on the oil/water interface. The interface is defined in two ways. In one case the simulated interface which has a sub-resolution smoothness is used, while the other is a smoothed interface which is extracted from synthetic segmented data based on the simulated phase distribution. Computed mean curvature on these surfaces are compared with that obtained from the fluid pressure computed directly in the simulation. We discuss the accuracy of image-based curvature measurements for the calculation of capillary pressure and propose the best way to extract an accurate curvature measurement, quantifying the likely uncertainties.
Scanziani A, Singh K, Menke H, et al., 2020, Dynamics of enhanced gas trapping applied to CO2 storage in the presence of oil using synchrotron X-ray micro tomography, Applied Energy, Vol: 259, ISSN: 0306-2619
During CO2 storage in depleted oil fields, under immiscible conditions, CO2 can be trapped in the pore space by capillary forces, providing safe storage over geological times - a phenomenon named capillary trapping. Synchrotron X-ray imaging was used to obtain dynamic three-dimensional images of the flow of the three phases involved in this process - brine, oil and gas (nitrogen) - at high pressure and temperature, inside the pore space of Ketton limestone. First, using continuous imaging of the porous medium during gas injection, performed after waterflooding, we observed chains of multiple displacements between the three phases, caused by the connectivity of the pore space. Then, brine was re-injected and double capillary trapping - gas trapping by oil and oil trapping by brine - was the dominant double displacement event. We computed pore occupancy, saturations, interfacial area, mean curvature and Euler characteristic to elucidate these double capillary trapping phenomena, which lead to a high residual gas saturation. Pore occupancy and saturation results show an enhancement of gas trapping in the presence of both oil and brine, which potentially makes CO2 storage in depleted oil reservoirs attractive, combining safe storage with enhanced oil recovery through immiscible gas injection. Mean curvature measurements were used to assess the capillary pressures between fluid pairs during double displacements and these confirmed the stability of the spreading oil layers observed, which facilitated double capillary trapping. Interfacial area and Euler characteristic increased, indicating lower oil and gas connectivity, due to the capillary trapping events.
Gao Y, Lin Q, Bijeljic B, et al., 2020, Pore-scale dynamics and the multiphase Darcy law, Physical Review Fluids, Vol: 5, Pages: 1-12, ISSN: 2469-990X
Synchrotron x-ray microtomography combined with sensitive pressure differential measurements were used to study flow during steady-state injection of equal volume fractions of two immiscible fluids of similar viscosity through a 57-mm-long porous sandstone sample for a wide range of flow rates. We found three flow regimes. (1) At low capillary numbers, Ca, representing the balance of viscous to capillary forces, the pressure gradient, ∇P, across the sample was stable and proportional to the flow rate (total Darcy flux) qt (and hence capillary number), confirming the traditional conceptual picture of fixed multiphase flow pathways in porous media. (2) Beyond Ca∗≈10−6, pressure fluctuations were observed, while retaining a linear dependence between flow rate and pressure gradient for the same fractional flow. (3) Above a critical value Ca>Cai≈10−5 we observed a power-law dependence with ∇P∼qat with a≈0.6 associated with rapid fluctuations of the pressure differential of a magnitude equal to the capillary pressure. At the pore scale a transient or intermittent occupancy of portions of the pore space was captured, where locally flow paths were opened to increase the conductivity of the phases. We quantify the amount of this intermittent flow and identify the onset of rapid pore-space rearrangements as the point when the Darcy law becomes nonlinear. We suggest an empirical form of the multiphase Darcy law applicable for all flow rates, consistent with the experimental results.
Krevor S, Blunt MJ, Trusler JPM, et al., 2020, Chapter 8: An introduction to subsurface CO<inf>2</inf> storage, RSC Energy and Environment Series, Pages: 238-295, ISBN: 9781788014700
The costs of carbon capture and storage are driven by the capture of CO2 from exhaust streams or the atmosphere. However, its role in climate change mitigation is underpinned by the potential of the vast capacity for storage in subsurface geologic formations. This storage potential is confined to sedimentary rocks, which have substantial porosity and high permeability in comparison to crystalline igneous and metamorphic rocks. These in turn occur in the sedimentary basins of the Earth's continents and near shore. However, the specific capacity for storage is not correlated simply to the existence of a basin. Consideration must also be made of reservoir permeability, caprock integrity, injectivity, fluid dynamics, and geomechanical properties of pressurisation and faulting. These are the topics addressed in this chapter. These processes and properties will combine in complex ways in a wide range of settings to govern the practicality of storing large volumes of CO2. There is clear potential for storage at the scale required to mitigate the worst impacts of global climate change, estimated to be in the order of 10 Gt CO2 per year by 2050. However, until at least dozens of commercial projects have been built in a range of geologic environments, the upper reaches of what can be achieved, and how quickly, will remain uncertain.
Ladipo L, Blunt MJ, King PR, 2020, A salinity cut-off method to control numerical dispersion in low-salinity waterflooding simulation, Journal of Petroleum Science and Engineering, Vol: 184, Pages: 1-19, ISSN: 0920-4105
Low-salinity or controlled salinity waterflooding (LSWF) is a promising enhanced oil recovery (EOR) technique. In simulations of this process, numerical dispersion smears saturation fronts, causing errors in the results. The objective of this work is to control these effects in LSWF simulation. We examine the impact of numerical dispersion on simulated LSWF performance. The low-salinity (LS) front is smeared even at unfeasibly fine grids. The velocities of the water fronts are altered. A numerical mixed zone forms around the interface between the injected and resident brines. This mimics a typical physical mixing effect.In reservoir simulation, threshold salinities are defined where the low salinity effect (LSE) is first encountered. It has been suggested that numerical dispersion effects can be corrected by imposing effective thresholds. We demonstrate that existing methods to evaluate these effective salinities do not accurately predict the salt front movement especially when dispersion is significant.We propose a simple simulation-based approach to evaluate the effective salinities based on the conservation of volumes of the resident and injected brines in the reference and upscaled solutions. After comparing analytical and corrected coarse-grid solutions in one-dimension, the effectiveness of the approach is demonstrated in multi-dimensional systems.A method is proposed to control the numerical mixed zone to replicate a physical longitudinal mixing effect. This method is demonstrated in one-dimension and does not require a fine-grid numerical solution as a benchmark. We investigate the effects of effective thresholds on the modeled transverse mixing or dispersion. A method to model transverse dispersion in simulations with an effective longitudinal component is suggested. This method is extended to the explicit modeling of physical dispersion in systems with transverse flows.We can now simply evaluate the effective salinities for a simulation grid; and control i
Ladipo L, Blunt M, King P, 2020, Optimizing low salinity waterflooding with controlled numerical influence of physical mixing considering uncertainty
Controlled/Low Salinity Waterflooding (LSWF) is an augmented waterflood with well-reported improved displacement efficiency compared with conventional waterfloods. Physical mixing or dispersion of the injected low-salinity (LS) brine with the formation high-salinity (HS) brine substantially reduces the low-salinity effect. Numerical dispersion often misrepresents this mixing in conventional LSWF-simulations, causing errors in the results. Uncertainty in the reservoir description further makes the evaluated performance questionable. Existing studies have suggested optimal amounts for the injected LS-brine to sustain its displacement stability during interwell flows with physical mixing, but with poor or no consideration of uncertainty. This work focuses on optimizing the injected LS-brine amount considering reported flow uncertainties while ensuring adequate correction of the erroneous influence of numerical dispersion on physical mixing. We investigate the impacts of flow uncertainties on the optimal LS slug-size. The sensitivity of the optimal slug-size to heterogeneity is examined under uncertainty. We evaluate how the interaction between physical mixing and geological heterogeneity influences slug integrity and performance. We propose an improved 'effective salinities' concept to evaluate appropriate effective salinities to characterize the desired representative physical mixing supressing the large numerical dispersion effects usually encountered in coarse-grid LSWF-simulations. This ensures reliable representation of physical dispersion in such grids. We consider different models with characterized levels of heterogeneity and essential variables that control the impact of mixing on LSWF performance based mainly on reported data. New indicators are defined to evaluate the displacement stability and performance of injected LS-brine thereby relating its technical and economic performance. Slug performance is evaluated at different injection times to examine the se
Panagiotis K, Blunt MJ, 2020, A Hubbert analysis on natural gas production of the top producers. How the carbon budget is affected under unconstrained extraction
The Hubbert curve was first introduced seventy years ago, to estimate oil reserves and production in the US. In this paper, Hubbert’s logistic function is used to estimate the peak production of natural gas of the top producers worldwide. The aim is to manage and fit the historic data with the minimum error and eventually project the CO2 emissions that result if the estimated reserves are extracted. Finally, we try to answer how the carbon budget is affected if production continues unconstrained. To that extend, historic data of the major producers were fitted and both production and expected emissions, were estimated. For several countries, the logistic function presented an adequate fit, while for others, it did not. The countries that didn’t fall under the bell-shaped (Hubbert) curve, have made political decisions to constrain their production. Continuing with the other countries (so called reference countries) we estimate that their cumulative emissions from natural gas production, will account for 59% of worldwide emissions by 2050, with China and the US dominating. Most importantly, in the case of no action for mitigating the emissions, total CO2 emitted, from natural gas production only, will consume 85% of the available carbon budget by 2050 to limit expected temperature increases to1.5 C and 31% of the budget in the case of a 2 C temperature increase.
Mosser L, Dubrule O, Blunt MJ, 2020, Stochastic Seismic Waveform Inversion Using Generative Adversarial Networks as a Geological Prior, MATHEMATICAL GEOSCIENCES, Vol: 52, Pages: 53-79, ISSN: 1874-8961
- Author Web Link
- Cite
- Citations: 88
Krevor S, Blunt MJ, Trusler AJPM, et al., 2020, An Introduction to Subsurface CO<sub>2</sub> Storage, CARBON CAPTURE AND STORAGE, Editors: Bui, Dowell, Publisher: ROYAL SOC CHEMISTRY, Pages: 238-295, ISBN: 978-1-78801-145-7
- Author Web Link
- Cite
- Citations: 3
Akai T, Alhammadi AM, Blunt MJ, et al., 2019, Mechanisms of microscopic displacement during enhanced oil recovery in mixed-wet rocks revealed using direct numerical simulation, Transport in Porous Media, Vol: 130, Pages: 731-749, ISSN: 0169-3913
We demonstrate how to use numerical simulation models directly on micro-CT images to understand the impact of several enhanced oil recovery (EOR) methods on microscopic displacement efficiency. To describe the physics with high-fidelity, we calibrate the model to match a water-flooding experiment conducted on the same rock sample (Akai et al. in Transp Porous Media 127(2):393–414, 2019. https://doi.org/10.1007/s11242-018-1198-8). First we show comparisons of water-flooding processes between the experiment and simulation, focusing on the characteristics of remaining oil after water-flooding in a mixed-wet state. In both the experiment and simulation, oil is mainly present as thin oil layers confined to pore walls. Then, taking this calibrated simulation model as a base case, we examine the application of three EOR processes: low salinity water-flooding, surfactant flooding and polymer flooding. In low salinity water-flooding, the increase in oil recovery was caused by displacement of oil from the centers of pores without leaving oil layers behind. Surfactant flooding gave the best improvement in the recovery factor of 16% by reducing the amount of oil trapped by capillary forces. Polymer flooding indicated improvement in microscopic sweep efficiency at a higher capillary number, while it did not show an improvement at a low capillary number. Overall, this work quantifies the impact of different EOR processes on local displacement efficiency and establishes a workflow based on combining experiment and modeling to design optimal recovery processes.
Alhosani A, Scanziani A, Lin Q, et al., 2019, In situ pore-scale analysis of oil recovery during three-phase near-miscible CO2 injection in a water-wet carbonate rock, Advances in Water Resources, Vol: 134, ISSN: 0309-1708
We study in situ three-phase near-miscible CO2 injection in a water-wet carbonate rock at elevated temperature and pressure using X-ray microtomography. We examine the recovery mechanisms, presence or absence of oil layers, pore occupancy and interfacial areas during a secondary gas injection process. In contrast to an equivalent immiscible system, we did not observe layers of oil sandwiched between gas in the centre of the pore space and water in the corners. At near-miscible conditions, the measured contact angle between oil and gas was approximately 73°, indicating only weak oil wettability in the presence of gas. Oil flows in the centres of large pores, rather than in layers for immiscible injection, when displaced by gas. This allows for a rapid production of oil since it is no longer confined to movement in thin layers. A significant recovery factor of 80% was obtained and the residual oil saturation existed as disconnected blobs in the corners of the pore space. At equilibrium, gas occupied the biggest pores, while oil and water occupied pores of varying sizes (small, medium and large). Again, this was different from an immiscible system, where water occupied only the smallest pores. We suggest that a double displacement mechanism, where gas displaces water that displaces oil is responsible for shuffling water into larger pores than that seen after initial oil injection. This is only possible since, in the absence of oil layers, gas can contact water directly. The gas-oil and oil-water interfacial areas are lower than in the immiscible case, since there are no oil layers and even water layers in the macro-pore space become disconnected; in contrast, there is a larger direct contact of oil to the solid. These results could serve as benchmarks for developing near-miscible pore-scale modelling tools.
Raeini AQ, Yang J, Bondino I, et al., 2019, Validating the generalized pore network model using micro-CT images of two-phase flow, Transport in Porous Media, Vol: 130, Pages: 405-424, ISSN: 0169-3913
A reliable prediction of two-phase flow through porous media requires the development and validation of models for flow across multiple length scales. The generalized network model is a step towards efficient and accurate upscaling of flow from the pore to the core scale. This paper presents a validation of the generalized network model using micro-CT images of two-phase flow experiments on a pore-by-pore basis. Three experimental secondary imbibition datasets are studied for both sandstone and carbonate rock samples. We first present a quantification of uncertainties in the experimental measurements. Then, we show that the model can reproduce the experimental fluid occupancies and saturations with a good accuracy, which in some cases is comparable with the similarity between repeat experiments. However, high-resolution images need to be acquired to characterize the pore geometry for modelling, while the results are sensitive to the initial condition at the end of primary drainage. The results provide a methodology for improving our physical models using large experimental datasets which, at the pore scale, can be generated using micro-CT imaging of multiphase flow.
Spurin C, Bultreys T, Bijeljic B, et al., 2019, Mechanisms controlling fluid breakup and reconnection during two-phase flow in porous media, Physical Review E, Vol: 100, ISSN: 2470-0045
The use of Darcy's law to describe steady-state multiphase flow in porous media has been justified by the assumption that the fluids flow in continuously connected pathways. However, a range of complex interface dynamics have been observed during macroscopically steady-state flow, including intermittent pathway flow where flow pathways periodically disconnect and reconnect. The physical mechanisms controlling this behavior have remained unclear, leading to uncertainty concerning the occurrence of the different flow regimes. We observe that the fraction of intermittent flow pathways is dependent on the capillary number and viscosity ratio. We propose a phase diagram within this parameter space to quantify the degree of intermittent flow.
Spurin C, Bultreys T, Bijeljic B, et al., 2019, Intermittent fluid connectivity during two-phase flow in a heterogeneous carbonate rock, Physical Review E, Vol: 100, ISSN: 2470-0045
Subsurface fluid flow is ubiquitous in nature, and understanding the interaction of multiple fluids as they flow within a porous medium is central to many geological, environmental, and industrial processes. It is assumed that the flow pathways of each phase are invariant when modeling subsurface flow using Darcy's law extended to multiphase flow, a condition that is assumed to be valid during steady-state flow. However, it has been observed that intermittent flow pathways exist at steady state even at the low capillary numbers typically encountered in the subsurface. Little is known about the pore structure controls or the impact of intermittency on continuum scale flow properties. Here we investigate the impact of intermittent pathways on the connectivity of the fluids for a carbonate rock. Using laboratory-based micro computed tomography imaging we observe that intermittent pathway flow occurs in intermediate-sized pores due to the competition between both flowing fluids. This competition moves to smaller pores when the flow rate of the nonwetting phase increases. Intermittency occurs in poorly connected pores or in regions where the nonwetting phase itself is poorly connected. Intermittent pathways lead to the interrupted transport of the fluids; this means they are important in determining continuum scale flow properties, such as relative permeability. The impact of intermittency on flow properties is significant because it occurs at key locations, whereby the nonwetting phase is otherwise disconnected.
Blunt MJ, Lin Q, Akai T, et al., 2019, A thermodynamically consistent characterization of wettability in porous media using high-resolution imaging, Journal of Colloid and Interface Science, Vol: 552, Pages: 59-65, ISSN: 0021-9797
Conservation of energy is used to derive a thermodynamically-consistent contact angle, θt, when fluid phase 1 displaces phase 2 in a porous medium. Assuming no change in Helmholtz free energy between two local equilibrium states we find that Δa1scosθt=κϕΔS1+Δa12, where a is the interfacial area per unit volume, ϕ is the porosity, S is the saturation and κ the curvature of the fluid-fluid interface. The subscript s denotes the solid, and we consider changes, Δ, in saturation and area. With the advent of high-resolution time-resolved three-dimensional X-ray imaging, all the terms in this expression can be measured directly. We analyse imaging datasets for displacement of oil by water in a water-wet and a mixed-wet sandstone. For the water-wet sample, the curvature is positive and oil bulges into the brine with almost spherical interfaces. In the mixed-wet case, larger interfacial areas are found, as the oil resides in layers. The mean curvature is close to zero, but the interface tends to bulge into brine in one direction, while brine bulges into oil in the other. We compare θt with the values measured geometrically in situ on the pore-scale images, θg. The thermodynamic angle θt provides a robust and consistent characterization of wettability. For the water-wet case the calculated value of θt gives an accurate prediction of multiphase flow properties using pore-scale modelling.
Abd AS, Elhafyan E, Siddiqui AR, et al., 2019, A review of the phenomenon of counter-current spontaneous imbibition: Analysis and data interpretation, JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING, Vol: 180, Pages: 456-470, ISSN: 0920-4105
- Author Web Link
- Cite
- Citations: 49
Akai T, Lin Q, Alhosani A, et al., 2019, Quantification of uncertainty and best practice in computing interfacial curvature from complex pore space images., Materials (Basel), Vol: 12, Pages: 1-21, ISSN: 1996-1944
Recent advances in high-resolution three-dimensional X-ray CT imaging have made it possible to visualize fluid configurations during multiphase displacement at the pore-scale. However, there is an inherited difficulty in image-based curvature measurements: the use of voxelized image data may introduce significant error, which has not-to date-been quantified. To find the best method to compute curvature from micro-CT images and quantify the likely error, we performed drainage and imbibition direct numerical simulations for an oil/water system on a bead pack and a Bentheimer sandstone. From the simulations, local fluid configurations and fluid pressures were obtained. We then investigated methods to compute curvature on the oil/water interface. The interface was defined in two ways; in one case the simulated interface with a sub-resolution smoothness was used, while the other was a smoothed interface extracted from synthetic segmented data based on the simulated phase distribution. The curvature computed on these surfaces was compared with that obtained from the simulated capillary pressure, which does not depend on the explicit consideration of the shape of the interface. As distinguished from previous studies which compared an average or peak curvature with the value derived from the measured macroscopic capillary pressure, our approach can also be used to study the pore-by-pore variation. This paper suggests the best method to compute curvature on images with a quantification of likely errors: local capillary pressures for each pore can be estimated to within 30% if the average radius of curvature is more than 6 times the image resolution, while the average capillary pressure can also be estimated to within 11% if the average radius of curvature is more than 10 times the image resolution.
Gao Y, Qaseminejad Raeini A, Blunt MJ, et al., 2019, Pore occupancy, relative permeability and flow intermittency measurements using X-ray micro-tomography in a complex carbonate, Advances in Water Resources, Vol: 129, Pages: 56-69, ISSN: 0309-1708
We imaged the steady-state flow of brine and decane (oil) at different fractional flows during dual injection in a micro-porous limestone, Estaillades, using X-ray micro-tomography. We applied differential imaging to: (a) distinguish micro-porous regions from macro-pores, and (b) determine fluid pore occupancy in both regions, and relative permeability at a capillary number, Ca = 7.3 × 10 −6 . The sample porosity was approximately 28%, with 7% in macro-pores and 21% in pores that could not be directly resolved (micro-porosity). Fluid occupancy in micro-porosity was classified into three sub-phases: micro-pore space with oil, micro-pore space with brine, and micro-pores partially filled with oil and brine. Our method indicated an initially higher oil recovery from micro-porosity, consistent with waterflooding in a water-wet rock. The fractional flow and relative permeabilities of the two fluids were obtained from measurements of the pressure differential across the sample and the saturation calculated from the images. The brine saturation and relative permeabilities are impacted by the presence of water-wet micro-porosity which provides additional connectivity to the phases. Furthermore, we find that in addition to brine and decane, a fraction of the macroscopic pore space contains an intermittent phase, which is occupied either by brine or decane during the hour-long scan time. Pore and throat occupancy of oil, brine and intermittent phase were obtained from images at different fractional flows using the generalized pore network extracted from the image of macro-pores. The intermittent phase, where the occupancy fluctuated between oil-filled and brine-filled, was predominantly located in the small and intermediate size pores and throats. Overall, we establish a new experimental methodology to: (i) quantify initial and recovered oil in micro-pores, (ii) characterise intermittent flow, and (iii) measure steady-state relative permeability in carbonates, whi
Lin Q, Bijeljic B, Berg S, et al., 2019, Minimal surfaces in porous media: Pore-scale imaging of multiphase flow in an altered-wettability Bentheimer sandstone, Physical Review E, Vol: 99, Pages: 063105-1-063105-13, ISSN: 1539-3755
High-resolution x-ray imaging was used in combination with differential pressure measurements to measurerelative permeability and capillary pressure simultaneously during a steady-state waterflood experiment on asample of Bentheimer sandstone 51.6 mm long and 6.1 mm in diameter. After prolonged contact with crude oil toalter the surface wettability, a refined oil and formation brine were injected through the sample at a fixed total flowrate but in a sequence of increasing brine fractional flows. When the pressure across the system stabilized, x-raytomographic images were taken. The images were used to compute saturation, interfacial area, curvature, andcontact angle. From this information relative permeability and capillary pressure were determined as functionsof saturation. We compare our results with a previously published experiment under water-wet conditions. Theoil relative permeability was lower than in the water-wet case, although a smaller residual oil saturation, ofapproximately 0.11, was obtained, since the oil remained connected in layers in the altered wettability rock.The capillary pressure was slightly negative and 10 times smaller in magnitude than for the water-wet rock,and approximately constant over a wide range of intermediate saturation. The oil-brine interfacial area wasalso largely constant in this saturation range. The measured static contact angles had an average of 80◦ with astandard deviation of 17◦. We observed that the oil-brine interfaces were not flat, as may be expected for a verylow mean curvature, but had two approximately equal, but opposite, curvatures in orthogonal directions. Theseinterfaces were approximately minimal surfaces, which implies well-connected phases. Saddle-shaped menisciswept through the pore space at a constant capillary pressure and with an almost fixed area, removing most ofthe oil.
Mosser L, Dubrule O, Blunt M, 2019, Deep stochastic inversion
© 81st EAGE Conference and Exhibition 2019. All rights reserved. Numerous modelling tasks require the solution of ill-posed inverse problems where we seek to find a distribution of earth models that match observed data such as reflected acoustic waveforms or produced hydrocarbon volumes. We present a deep learning framework to create stochastic samples of posterior property distributions for ill-posed inverse problems using a gradient-based approach. The spatial distribution of petrophysical properties is created by a deep generative model and controlled by a set of latent variables. A generative adversarial network (GAN) is used to represent a prior distribution of geological models based on a training set of object-based models. Then we minimize the mismatch between observed ground-truth data and numerical forward-models of the generator output by first computing gradients of the objective function with respect to grid-block properties and then using neural network backpropagation to obtain gradients with respect to the latent variables. Synthetic test cases of acoustic waveform inversion and reservoir history matching are presented. In seismic inversion, we use a Metropolis adjusted Langevin algorithm (MALA) to obtain posterior samples. For both synthetic cases, we show that deep generative models such as GANs can be combined in an end-to-end framework to obtain stochastic solutions to geophysical inverse problems.
Mosser L, Dubrule O, Blunt M, 2019, Deep stochastic inversion
Numerous modelling tasks require the solution of ill-posed inverse problems where we seek to find a distribution of earth models that match observed data such as reflected acoustic waveforms or produced hydrocarbon volumes. We present a deep learning framework to create stochastic samples of posterior property distributions for ill-posed inverse problems using a gradient-based approach. The spatial distribution of petrophysical properties is created by a deep generative model and controlled by a set of latent variables. A generative adversarial network (GAN) is used to represent a prior distribution of geological models based on a training set of object-based models. Then we minimize the mismatch between observed ground-truth data and numerical forward-models of the generator output by first computing gradients of the objective function with respect to grid-block properties and then using neural network backpropagation to obtain gradients with respect to the latent variables. Synthetic test cases of acoustic waveform inversion and reservoir history matching are presented. In seismic inversion, we use a Metropolis adjusted Langevin algorithm (MALA) to obtain posterior samples. For both synthetic cases, we show that deep generative models such as GANs can be combined in an end-to-end framework to obtain stochastic solutions to geophysical inverse problems.
Mosser L, Dubrule O, Blunt M, 2019, Deep stochastic inversion
© 81st EAGE Conference and Exhibition 2019. All rights reserved. Numerous modelling tasks require the solution of ill-posed inverse problems where we seek to find a distribution of earth models that match observed data such as reflected acoustic waveforms or produced hydrocarbon volumes. We present a deep learning framework to create stochastic samples of posterior property distributions for ill-posed inverse problems using a gradient-based approach. The spatial distribution of petrophysical properties is created by a deep generative model and controlled by a set of latent variables. A generative adversarial network (GAN) is used to represent a prior distribution of geological models based on a training set of object-based models. Then we minimize the mismatch between observed ground-truth data and numerical forward-models of the generator output by first computing gradients of the objective function with respect to grid-block properties and then using neural network backpropagation to obtain gradients with respect to the latent variables. Synthetic test cases of acoustic waveform inversion and reservoir history matching are presented. In seismic inversion, we use a Metropolis adjusted Langevin algorithm (MALA) to obtain posterior samples. For both synthetic cases, we show that deep generative models such as GANs can be combined in an end-to-end framework to obtain stochastic solutions to geophysical inverse problems.
Oliveira TDS, Blunt MJ, Bijeljic B, 2019, Modelling of multispecies reactive transport on pore-space images, Advances in Water Resources, Vol: 127, Pages: 192-208, ISSN: 0309-1708
We present a new model, named poreReact, to simulate multispecies reactive transport on pore space images. We solve the Navier–Stokes equations and the advection-diffusion equation for concentration on an unstructured grid using the finite volume method implemented in OpenFOAM. We couple it with the chemical model Reaktoro, which we use to calculate the chemical equilibrium of homogeneous reactions in each grid cell, considered as a completely mixed batch reactor. We validate the model against analytical solutions and experimental data, and investigate, for a range of Péclet numbers, the interplay between transport and reaction for multispecies reactive transport in a 3D bead pack where two streams of reactants at different pH are injected in parallel. We analyse the distribution of species and the rates of formation and consumption in the pore space and find that, despite the relative homogeneity of the bead pack and symmetry in injection conditions, the concentration fields of the products can be asymmetric because of the interplay between transport and chemical equilibrium. For different Péclet numbers, we calculate relative yields (the ratio between the observed change in concentration and the change that would be obtained if the reactants were completely mixed). We observe that lower Péclet numbers give rise to higher relative yields because of increased transverse mixing by diffusion. However, higher absolute yields are obtained at higher injection velocities because of the larger amount of matter available for reaction in a given time. Reaction is more favoured in the faster-flowing regions of the pore space. However, this effect is more marked for species for which advection is the dominant mechanism of transport to reactive sites, as opposed to diffusion-mediated reactions where the full velocity distribution is sampled before reaction occurs.
Al-Khulaifi Y, Lin Q, Blunt MJ, et al., 2019, Pore-scale dissolution by CO₂ saturated brine in a multimineral carbonate at reservoir conditions: impact of physical and chemical heterogeneity, Water Resources Research, Vol: 55, Pages: 3171-3193, ISSN: 0043-1397
We study the impact of physical and chemical heterogeneity on reaction rates in multimineral porous media. We selected two pairs of carbonate samples of different physical heterogeneity in accordance with their initial computed velocity distributions and then injected CO 2 saturated brine at reservoir conditions at two flow rates. We periodically imaged the samples using X-ray microtomography. The mineralogical composition was similar (a ratio of dolomite to calcite of 8:1), but the intrinsic reaction rates and mineral spatial distribution were profoundly different. Visualizations of velocity fields and reacted mineral distributions revealed that a dominant flow channel formed in all cases. The more physically homogeneous samples had a narrower velocity distribution and more preexisting fast channels, which promoted dominant channel formation in their proximity. In contrast, the heterogeneous samples exhibit a broader distribution of velocities and fewer fast channels, which accentuated nonuniform calcite distribution and favored calcite dissolution away from the initially fast pathways. We quantify the impact of physical and chemical heterogeneity by computing the proximity of reacted minerals to the fast flow pathways. The average reaction rates were an order of magnitude lower than the intrinsic ones due to mass transfer limitations. The effective reaction rate of calcite decreased by an order of magnitude, in both fast channels and slow regions. After channel formation calcite was shielded by dolomite whose effective rate in slow regions could even increase. Overall, the preferential channeling effect, as opposed to uniform dissolution, was promoted by a higher degree of physical and/or chemical heterogeneity.
Lin Q, Bijeljic B, Krevor SC, et al., 2019, A New Waterflood Initialization Protocol With Wettability Alteration for Pore-Scale Multiphase Flow Experiments, Petrophysics – The SPWLA Journal of Formation Evaluation and Reservoir Description, Vol: 60, Pages: 264-272, ISSN: 1529-9074
Rücker M, Bartels WB, Singh K, et al., 2019, The Effect of Mixed Wettability on Pore-Scale Flow Regimes Based on a Flooding Experiment in Ketton Limestone, Geophysical Research Letters, Vol: 46, Pages: 3225-3234, ISSN: 0094-8276
© 2019. The Authors. Darcy-scale multiphase flow in geological formations is significantly influenced by the wettability of the fluid-solid system. So far it has not been understood how wettability impacts the pore-scale flow regimes within rocks, which were in most cases regarded as an alteration from the base case of strongly water-wet conditions by adjustment of contact angles. In this study, we directly image the pore-scale flow regime in a carbonate altered to a mixed-wet condition by aging with crude oil to represent the natural configuration in an oil reservoir with fast synchrotron-based X-ray computed tomography. We find that the pore-scale flow regime is dominated by ganglion dynamics in which the pore space is intermittently filled with oil and brine. The frequency and size of these fluctuations are greater than in water-wet rock such that their impact on the overall flow and relative permeability cannot be neglected in modeling approaches.
Singh K, Muljadi B, Raeini AQ, et al., 2019, The architectural design of smart ventilation and drainage systems in termite nests, Science Advances, Vol: 5, ISSN: 2375-2548
Termite nests have been widely studied as effective examples for ventilation and thermoregulation;however, the mechanisms by which the nest properties are controlled by the micro-structure of the outer walls remain unclear. Here, we combine multi-scale X-ray imaging with three-dimensional flow field simulations to investigate the impact of the architectural design of nest walls on CO2and heat transport as well as on water drainage. We show that termites construct an outer wall that contains both small and percolating large pores at the micro-scale. The network of larger micro-scale pores in the outer wall provides a permeability that is 1-2 orders of magnitude greater than that of the smaller pores, andaCO2diffusivitythat is a factor of up to eight times larger. The largerpores and resultant high porosity also reduce the solid mass required for nest construction by ~11-14%. This is energetically favorable and reduces the overall weight of the nest, thus lowering the risk of collapse. In addition, the pore network offers enhanced thermal insulation to the inner parts of the nest and allows quick drainage ofrainwater thereby restoring the ventilation and providing structural stability to the wet nest.
Akai T, Alhammadi AM, Blunt MJ, et al., 2019, Modeling oil recovery in mixed-wet rocks: Pore-scale comparison between experiment and simulation, Transport in Porous Media, Vol: 127, Pages: 393-414, ISSN: 0169-3913
To examine the need to incorporate in situ wettability measurements in direct numerical simulations, we compare waterflooding experiments in a mixed-wet carbonate from a producing reservoir and results of direct multiphase numerical simulations using the color-gradient lattice Boltzmann method. We study the experiments of Alhammadi et al. (Sci Rep 7(1):10753, 2017. https://doi.org/10.1038/s41598-017-10992-w) where the pore-scale distribution of remaining oil was imaged using micro-CT scanning. In the experiment, in situ contact angles were measured using an automated algorithm (AlRatrout et al. in Adv Water Resour 109:158–169, 2017. https://doi.org/10.1016/j.advwatres.2017.07.018), which indicated a mixed-wet state with spatially non-uniform angles. In our simulations, the pore structure was obtained from segmented images of the sample used in the experiment. Furthermore, in situ measured angles were also incorporated into our simulations using our previously developed wetting boundary condition (Akai et al. in Adv Water Resour 116(March):56–66, 2018. https://doi.org/10.1016/j.advwatres.2018.03.014). We designed six simulations with different contact angle assignments based on experimentally measured values. Both a constant contact angle based on the average value of the measured values and non-uniform contact angles informed by the measured values gave a good agreement for fluid pore occupancy between the simulation and the experiment. However, the constant contact angle assignment predicted 54% higher water effective permeability after waterflooding than that estimated for the experimental result, whereas the non-uniform contact angle assignment gave less than 1% relative error. This means that to correctly predict fluid conductivity in mixed-wet rocks, a spatially heterogeneous wettability state needs to be taken into account. The novelty of this work is to provide a direct pore-scale comparison between experiments and simulations employing experiment
This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.