Summary Volumetric maximum curvature attribute computed from 3D ocean bottom cable (OBC) seismic data, production logging tool (PLT), inorganic chemical tracer data, and fractures observed from core and full-bore formation microimager (FMI) logs were integrated to characterize fractured carbonate reservoirs of an offshore oil field in Abu Dhabi, United Arab Emirates (UAE). The extracted maximum curvature anomalies are predominantly orientated in NNE-SSW and NE-SW, a trend perpendicular to the dominant fault direction in the oil field and similar to the dominant strike directions of fractures measured from core data and FMI logs. Because the fracture strike directions of well data mimic the strike directions of curvature anomalies at corresponding reservoir levels, we interpreted the maximum curvature anomalies to represent dilatational fractured zones or fracture corridors. Integration of dynamic data, such as PLT and chemical tracers, and maximum curvature anomalies demonstrate that the inferred fracture zones can determine water breakthroughs as well as inter- and intrareservoir communications. As a result, this study highlights possible fracture zones and their internal architecture, as well as their potential flow capabilities. These results play a key role in reservoir management and monitoring of water movement through structural pathways.

Summary Forecasting production from hydrocarbon fields is challenging because of the large number of uncertain model parameters and the multitude of observed data that are measured. The large number of model parameters leads to uncertainty in the production forecast from hydrocarbon fields. Changing operating conditions [e.g., implementation of improved oil recovery or enhanced oil recovery (EOR)] results in model parameters becoming sensitive in the forecast that were not sensitive during the production history. Hence, simulation approaches need to be able to address uncertainty in model parameters as well as conditioning numerical models to a multitude of different observed data. Sampling from distributions of various geological and dynamic parameters allows for the generation of an ensemble of numerical models that could be falsified using principal-component analysis (PCA) for different observed data. If the numerical models are not falsified, machine-learning (ML) approaches can be used to generate a large set of parameter combinations that can be conditioned to the different observed data. The data conditioning is followed by a final step ensuring that parameter interactions are covered. The methodology was applied to a sandstone oil reservoir with more than 70 years of production history containing dozens of wells. The resulting ensemble of numerical models is conditioned to all observed data. Furthermore, the resulting posterior-model parameter distributions are only modified from the prior-model parameter distributions if the observed data are informative for the model parameters. Hence, changes in operating conditions can be forecast under uncertainty, which is essential if nonsensitive parameters in the history are sensitive in the forecast.

Summary Carbon dioxide (CO 2 ) enhanced oil recovery (EOR) has long been practiced in the US as an efficient mean for enhancing oil production. Many of the US CO 2 -EOR developments have been designed horizontally. This is because of a viscous-dominated CO 2 flow regime that is prevalent in these developments driven by thin and low-permeability reservoirs. Reservoirs and fluid properties are different in the North Sea. Pays are usually thicker with better petrophysical properties. Lighter oils can also be found in North Sea reservoirs. This suggests that a dissimilar flow regime might prevail CO 2 displacements in the North Sea developments, which could favor a dissimilar CO 2 -EOR process design. This study thus compares CO 2 flow regimes between several North Sea and US reservoirs. We use scaling analysis to characterize and compare CO 2 flow regimes between these two classes of reservoirs. Scaling analysis characterizes CO 2 displacement in each reservoir system using three dimensionless numbers: gravity, effective aspect ratio, and mobility ratio. Displacement experiments conducted in stochastically generated permeability fields, under exactly matched magnitudes of the derived dimensionless numbers, reveal the prevailing CO 2 flow regime in each reservoir system. Results of scaling analysis indicate that CO 2 flooding in the North Sea reservoirs can be generally characterized with a larger gravity number, smaller effective aspect ratio, and smaller mobility ratio than the average US CO 2 flooded reservoirs. Flow regime analysis indicates that unlike the majority of the US CO 2 flooded reservoirs, CO 2 flow regimes tend to be more gravity-dominated in the North Sea class of reservoirs. CO 2 flow regimes in the North Sea systems are expected to suffer from a higher degree of instability because of thicker North Sea pays, which limit effective crossflow. Understanding the differences and characteristics of CO 2 flow regimes in the North Sea prospects can help operators design their CO 2 flooding more efficiently, which could increase the recovery factor (RF) as well as address CO 2 storage requirements, a necessary consideration for CO 2 -EOR deployment in the North Sea.

Summary An approach to the analysis of production data from waterflooded oil fields is proposed in this paper. The method builds on the established techniques of rate-transient analysis (RTA) and extends the analysis period to include the transient- and steady-state effects caused by a water-injection well. This includes the initial rate transient during primary production, the depletion period of boundary-dominated flow (BDF), a transient period after injection starts and diffuses across the reservoir, and the steady-state production that follows. RTA will be applied to immiscible displacement using a graph that can be used to ascertain reservoir properties and evaluate performance aspects of the waterflood. The developed solutions can also be used for accurate and rapid forecasting of all production transience and boundary-dominated behavior at all stages of field life. Rigorous solutions are derived for the transient unit mobility displacement of a reservoir fluid, and for both constant-rate-injection and constant-pressure-injection after a period of reservoir depletion. A simple treatment of two-phase flow is given to extend this to the water/oil-displacement problem. The solutions are analytical and are validated using reservoir simulation and applied to field cases. Individual wells or total fields can be studied with this technique; several examples of both will be given. Practical cases are given for use of the new theory. The equations can be applied to production-data interpretation, production forecasting, injection-water allocation, and for the diagnosis of waterflood-performance problems.

Summary Reduction of fracture/well spacing and increases in hydraulic fracture stimulation treatment size are popular strategies for improving hydrocarbon recovery from multifractured horizontal wells (MFHWs). However, these strategies can also increase the chance of fracture interference, which can not only negatively impact the overall production but also introduce complexities for production data analysis. A semianalytical model is therefore developed to analyze production data from two communicating MFHWs and applied to a field case. The new semianalytical model uses the dynamic drainage area (DDA) concept and assumes three porosity regions. The three-region model is comprised of a primary hydraulic fracture (PHF), an enhanced fractured region (EFR) adjacent to the PHF, and a nonstimulated region (NSR). Assuming a well pair primarily communicates through PHFs, the equations for two communicating wells are coupled and solved simultaneously to model the fluid transfer between the wells. This method is used within a history-matching framework to estimate the communication between the wells by matching the production data. The semianalytical model is first verified against a more rigorous, fully numerical simulation model for a range of fracture/reservoir properties. These comparisons demonstrate that there is excellent agreement between the fully numerical simulation model results and the new semianalytical model. The semianalytical model is then employed to history-match production data from six MFHWs (drilled from two adjacent well pads) exhibiting different degrees of communication. For the purpose of history matching the data, only strong communication between pairs of wells (intrapair communication) is considered in the three-region model, and the results show good agreement with the field data. A flexible, yet simple, semianalytical model is developed for the first time that can accurately model the communication between multiple well pairs. This approach can be used by reservoir engineers to analyze the production data from communicating MFHWs.

Summary Low-frequency distributed acoustic sensing (LF-DAS) data are a powerful attribute to detect fracture hits and characterize fracture geometry during multistage hydraulic fracturing treatments in unconventional reservoirs. The DAS data in low-frequency bands linearly correlate with strain and strain rate induced by dynamic fracture propagation. Due to the complexity of multiple-fracture propagation in unconventional reservoirs, the measured signals from different wells exhibit various characteristics. Mechanisms causing the differences are not well understood, which makes the interpretation of real LF-DAS data and detection of fracture hits very challenging. Hence, it is necessary to relate the observed strain/strain-rate signatures to specific fracture patterns based on the physical model of rock deformation during fracture propagation and to quantitatively characterize signatures surrounding fracture hits. In this study, we have applied our in-house fracture propagation model to simulate simultaneous multiple-fracture propagation as well as fracture-induced strain and strain-rate responses along an offset monitor well. Then a general guideline for fracture-hit detection is proposed based on quantitative analysis of strain/strain-rate responses during multiple-fracture propagation. Finally, a set of field examples are presented to demonstrate the potential of LF-DAS data on hydraulic fracturing monitoring. During multiple-fracture propagation, a “heart-shaped” zone with positive strain rates may be identified for each fracture before the fracture hit. Immediately after the fracture encounters the monitor well, part of the fiber within the fracture path keeps being extended, while the fiber sections off the path become compressed. Three 1D features along the channel (location) axis are designed to detect fracture hits. The features are maximum strain rate, the summation of strain rates, and summation of strain-rate amplitudes. Channels with fracture hits usually exhibit significant peak values of the three features. However, the characteristic signatures can be less detectable when the gauge length is close to the cluster spacing. Connections between fracture-hit locations and cluster perforations clearly reflect the fracture propagation direction. The field examples illustrate the complexity of real LF-DAS signals and demonstrate the adequacy of the proposed guideline for fracture-hit detection with multicluster completion. The fractures propagate nearly perpendicular to the horizontal wellbore in this unconventional shale formation. In addition, four to five fractures out of eight perforation clusters can propagate 396.24 m (1,300 ft) and hit the monitor well, and the “heel-biased” fracture pattern is observed (fractures that do not hit the monitor well are usually close to the toe side). Fracturing fluid leaking off into the previous stage can also be diagnosed, which could negatively affect the completion efficiency.

Summary In this work, a probabilistic methodology for decline curve analysis (DCA) in unconventional reservoirs is presented using several Bayesian model-fitting algorithms and deterministic models. The deterministic models considered are the power law exponential (PLE) model, the stretched exponential production decline (SEPD) model, Duong’s model, and the logistic growth analysis (LGA) model. Accurate production forecasting and uncertainty quantification were the primary objectives of this study. The Bayesian inferencing techniques described in this work utilize three sampling vehicles, namely the Gibbs sampling (implemented in OpenBUGS, an open-source software), the Metropolis-Hastings (MH) algorithm, and approximate Bayesian computation (ABC) to sample parameter values from their posterior distributions. These different sampling algorithms are applied in conjunction with DCA models to estimate DCA parameter prediction intervals. Using these prediction levels, production is forecasted, and uncertainty bounds are established. To examine its reliability, the methodology was tested on over 74 oil and gas wells located in the three main subplays of the Permian Basin, namely, the Delaware play, the Central Basin Platform, and the Midland play. Results show that the examined DCA-Bayesian models are well calibrated, result in low production errors, and narrow uncertainty bounds for the production history data sets. Also, the LGA model is best in terms of prediction errors for all algorithms except MH. The Gibbs algorithm is nearly the best algorithm in terms of prediction error for all DCA models except the Arps model. Prediction errors are the highest in the Central Basin Platform. The methodology was also successfully applied to unconventional reservoirs with as low as six months of available production history. Depending on the amount of production history data available, the probabilistic model that provides the best fit can vary. It is therefore recommended that all possible combinations of the deterministic and Bayesian model-fitting algorithms be applied to the available production history. This is to obtain more confidence in the conclusions related to the production forecasts, reserves estimate, and uncertainty bounds. The novelty of this methodology relies on using multiple combinations of DCA-Bayesian models to achieve accurate reserves estimates and narrow uncertainty bounds. This paper can help assess shale plays because some of the shale plays are in the early stages of development when productivity estimations are carried out.

Summary Due to ultralow permeability, there is practically no pressure interference between wells producing from tight oil (≈0.1 md average permeability) and shale oil reservoirs (∼0.001 md average permeability). This renders type well methodology as a commonly used tool for forecasting production performance of these “unconventionals.” Several authors proposed different methods for constructing type wells for unconventional reservoirs, but none compared them. In this study, we compare three of the most common types of well generation methods. In the absence of real field data, individual production histories for wells are established by picking Arps decline curve parameters q i , b , and D i , which govern a well’s production performance, from respective distributions of these parameters. Next, we compare type wells constructed using the Monte Carlo (MC) method, final cumulative (FC) method, and time slice (TS) method. Moreover, we study the impact of a possible linear correlation (also linear C) between b and D i (comparable to field observations) on MC, FC, and TS type wells. We also study the effect of well count, and D min value, which is the 4th parameter introduced with modified Arps equation for unconventionals. In this paper, we show that type wells generated with TS and FC methods have almost the same behavior, whereas the MC method is affected the most by stochastic experimentation, well count, and D min .

Summary Prediction of well production from unconventional reservoirs is often a complex problem with an incomplete understanding of physics and a considerable amount of data. The most effective way for dealing with it is to use the gray-box approach that combines the strengths of physics-based models and machine learning (ML) used for dealing with certain components of the prediction where physical understanding is poor or difficult. However, the development of methodologies for the incorporation of physics into ML is still in its infancy, not only in the oil and gas industry, but also in other scientific and engineering communities, including the physics community. To set the stage for further advancing the use of combining physics-based models with ML for predicting well production, in this paper we present a brief review of the current developments in this area in the industry, including ML representation of numerical simulation results, determination of parameters for decline curve analysis (DCA) models with ML, physics-informed ML (PIML) that provides an efficient and gridless method for solving differential equations and for discovering governing equations from observations, and physics-constrained ML (PCML) that directly embeds a physics-based model into a neural network. The advantages and potential limitations of the methods are discussed. The future research directions in this area include, but are not limited to, further developing and refining methodologies, including algorithm development, to directly embed physics-based models into ML; exploring the usefulness of PIML for reservoir simulations; and adapting the new developments of how the physics and ML are incorporated in other communities to the well-production prediction. Finally, the methodologies we discuss in the paper can be generally applied to conventional reservoirs as well, although the focus here is on unconventional reservoirs.

Summary Seawater injection is widely used to maintain offshore-oil-reservoir pressure and improve oil recovery. However, injecting seawater into reservoirs can cause many issues, such as reservoir souring and scaling, which are strongly related to the seawater-breakthrough percentage. Accurately calculating the seawater-breakthrough percentage is important for estimating the severity of those problems and further developing effective strategies to mitigate those issues. The validation of using natural-ion boron as a tracer to calculate seawater-breakthrough percentage was investigated. Boron can interact with clays, which can influence the accuracy in seawater-breakthrough calculation. Therefore, the interaction between boron and different clays at various conditions was first studied, and the Freundlich adsorption equation was used to describe the boron-adsorption isotherms. Then, the boron-adsorption isotherms were coupled into the reservoir simulator to investigate the boron transport in porous media, and the results in turn were further analyzed to calculate the accurate seawater-breakthrough percentage. Results indicated that boron adsorption by different clays varied. pH value of solution can significantly influence the amount of boron adsorbed. As a result, the boron-concentration profile was delayed in coreflood tests. The accuracy of the new model was verified by convergence rate tests and comparison with analytical results. Furthermore, model results fit well with experimental data. On the basis of the reservoir-simulation results, the boron-concentration profile in produced water can be used to calculate the seawater-breakthrough percentage by considering the clay-content distribution. However, the seawater-breakthrough point cannot be determined by boron because the boron concentration is still at the formation level after seawater breakthrough due to boron desorption.

Summary A potential enhanced oil recovery technique is to inject alkali into a reservoir with a high-total acid number (TAN) crude to generate soap in situ and reduce interfacial tension (IFT) without the need to inject surfactant. The method may be cost-effective if the IFT can be lowered enough to cause significant mobilization of trapped oil while also avoiding formation of gels and viscous phases. This paper investigates the potential field application of injecting alkali to generate in-situ soap and favorable phase behavior for a high-TAN oil. Oil analyses show that the acids in the crude are a complex mixture of various polar acids and not mainly carboxylic acids. The results from phase behavior experiments do not undergo typical Winsor microemulsion behavior transition and subsequent ultralow IFTs below 1×10 −3 mN/m that are conventionally observed. Instead, mixing of alkali and crude/brine generate water-in-oil macroemulsions that can be highly viscous. For a specific range of alkali concentrations, however, phases are not too viscous, and IFTs are reduced by several orders of magnitude. Incremental coreflood recoveries in this alkali range are excellent, even though not all trapped oil is mobilized. The viscous phase behavior at high alkali concentrations is explained by the formation of salt-crude complexes, created by acids from the crude oil under the alkali environment. These hydrophobic molecules tend to agglomerate at the oil-water interface. Together with polar components from the crude oil, they can organize into a highly viscous network and stabilize water droplets in the oleic phase. Oil-soluble alcohol was added to counter those two phenomena at large concentrations, but typical Winsor phase behavior was still not observed. A physicochemical model is proposed to explain the salt-crude complex formation at the oil-water interface that inhibits classical Winsor behavior.

Summary The heavy-oil- and bitumen-recovery process by injection of a pure condensing solvent in a solvent vapor chamber provides an alternative to steam-based recovery techniques such as steam-assisted gravity drainage (SAGD). Because of the lower operating temperature between 40 and 80°C, the process uses a much lower energy budget than a steam process and thus results in significantly reduced greenhouse-gas emissions. This paper describes the route to a successful production function with the physical processes at play and using analytical tools. Physical relationships are derived for the solvent/bitumen (S/B) ratio, the bitumen drainage from the roof of the solvent vapor chamber, and for bitumen extraction from both sides of the solvent chamber by the draining condensed solvent. The fast diffusion of bitumen into this narrow liquid solvent zone is likely subtly enhanced by transverse dispersion. The speed of bitumen extraction from the roof of the solvent vapor chamber is constrained by the gas/oil capillary pressure. Extraction from the side of the chamber is approximately three times faster by the action of the thin gravity-draining liquid solvent film. Several equations are provided to enable creation of a heat balance for this condensing solvent process. Laboratory and field observations are matched, including the rates, the heat balance, and the S/B ratio. The model can explain constrained production performance by identifying the rate-limiting steps (e.g., when insufficient solvent condenses). The model predicts high solvent holdup during the rise of the solvent chamber. A method to estimate this solvent liquid saturation is provided. The S/B ratio depends on injector-wellbore heat losses, the (high) liquid saturation in the rising solvent chamber, and the process properties (operating temperature), reservoir properties (heat capacity, porosity, and oil saturation), and solvent properties (density and latent heat). In the existing body of literature, no satisfactory analytical model was available; this new approach helps to constrain production performance and to estimate solvent and heat requirements. The methods in this paper can be used in the future for subsurface project design and performance predictions.

Summary An immiscible hydrocarbon‐gas foam enhanced-oil-recovery (EOR) pilot has been designed and implemented in a hydraulically fractured tight reservoir in the Woodbine Field in Texas. Although gas injection is considered the main EOR technology for tight fractured unconventional reservoirs, gaseous foam has not previously been considered as an effective conformance solution in fractured formations. In this paper, we present experimental evaluation of the surfactant, field‐pilot design, and implementation and performance analysis of the pilot toward developing an unconventional hydrocarbon-gas foam EOR conformance solution. Several surfactants were screened through a bulk‐foam test for the harsh reservoir conditions (120°C, 3.23% salinity, and ∼27% clay). The selected surfactant was further evaluated for static adsorption on reservoir rocks at room temperature to ensure an economic field pilot. The surfactant was also evaluated for oil‐brine emulsion tendency to mitigate any field‐implementation issues. A single horizontal injector and two surrounding horizontal producer pads were developed for an improved oil recovery/EOR pilot implementation in the Woodbine Field. Water and produced hydrocarbon gases were injected alternately, as well as in coinjection mode; however, no consistent incremental oil was observed. Injected gas and water broke through on an order of hours and days, respectively. The injector showed more connectivity with one of the producers, suggesting a strong areal conformance problem. A steady baseline operation was established by coinjecting gas and water at a constant gas fraction and total constant rate that resulted in a steady production baseline. The baseline injection was continued with surfactant injection in brine for in‐situ foam generation. During 5 weeks of surfactant injection, foam generation and mobility reduction were confirmed with the increase in the measured bottomhole pressure. Mobility control resulted in out‐of‐zone (OOZ) injection elimination for both gas and water and gas diversion to bypassed areas. With conformance corrected at the injector and deeper in the reservoir, oil‐production rates more than doubled, gas utilization was improved, and a low gas/oil ratio (GOR) and improved volumetric sweep were confirmed. The increased oil production continued for at least 6 weeks after completing surfactant injection. More than 2,000 bbl of incremental oil was recovered in 11 weeks of pilot operation. This current work confirms the technical efficacy and potential of the gaseous foam conformance solution for incremental oil production with gas‐EOR in unconventional plays.

Summary Injecting water with chemicals to generate emulsions in the reservoir is a promising method in the enhancement of heavy‐oil recovery because the formation of oil‐in‐water (O/W) emulsions significantly reduces oil viscosity. Nanoparticles (NPs) (Pickering emulsions) can be used for this purpose as a cost‐effective alternative to expensive surfactants; however, such Pickering emulsions need to be stable for successful applications. The objective of this study is to screen the effective emulsifier for O/W emulsions from a broad range of solid NPs and identify suitable Pickering emulsifying agents (e.g., adjusting pH or salt concentration) that can render emulsions stable at relevant conditions, and to investigate how a range of physical parameters, such as particle concentration, water/oil ratio (WOR), and temperature affect emulsion stability. Five NPs—including cellulose nanocrystals (CNCs), silica, alumina, magnetite, and zirconia—were tested on their capabilities of stabilizing O/W emulsions through glass vial screening tests under various pH and salinity conditions. The screening results showed that the CNC could become an effective emulsifier by either adjusting pH or salinity. In addition, zeta potential measurements were conducted to explain the observations. The stabilization mechanisms of CNCs were studied through an epifluorescent transmitted microscope showing that the formation of a dense particle layer around the oil droplets, as well as a network in the continuous phase, were the two main mechanisms accounting for the high stability of the emulsions stabilized by CNCs. The effects of particle concentrations on the emulsion stability were studied quantitatively by analyzing the droplet‐size distributions calculated by the open-source ImageJ software, with the results showing a sharp decrease in droplet size, followed by a smooth change as the particle concentration increased. For the WOR effect, phase inversion from O/W to water‐in‐oil (W/O) emulsions was observed when the oil content was more than 0.6. The thermal stability of emulsions was studied both macroscopically by glass bottle tests and microscopically through a microscope, both of which show that the CNC‐stabilized emulsions remained thermally stable up to 100°C. The rheological properties of both aqueous dispersions of CNCs and the corresponding O/W emulsions were also measured under various salinity conditions. The results showed that the salinity had a great impact on the viscosity of the CNC suspension and the typical shear‐thinning behavior of Pickering emulsions. This study provides an option to enhance emulsion stability without surfactants, which will reduce the costs and facilitate field applications of emulsion flooding in heavy‐oil recovery.

Summary The determination of optimal well locations and well controls in the horizontal-well steamflooding of heavy-oil reservoirs is a meaningful but also challenging task for the complex well types and complicated mechanisms. In this paper, a framework that combines optimization algorithms and the reservoir simulator together is proposed to solve this problem. Two typical algorithms, particle-swarm optimization (PSO) and mesh adaptive direct search (MADS), are both used to study their performance on well-placement optimization, well-control optimization, and the joint optimization of these two aspects. For the joint-optimization problem, both the simultaneous approach and sequential approach are considered. A net-present-value (NPV) formula for evaluation of the steamflooding project is proposed, and optimization runs are conducted for an offshore heavy-oil reservoir by maximizing the NPV of the horizontal steamflooding pilot. The results show that both the PSO and MADS were effective for well-placement/control optimization of the horizontal steamflooding wells. The NPVs were greatly improved throughout the optimization process. The control frequency has great influence on the optimal NPV. Intermittent steamflooding might be a better choice than continuous steamflooding. The steam-injection rate and oil-production rates need to be controlled and decreased at the latter stage for mitigating steam channeling and an ineffective steam cycle between injection and production wells. For the joint-optimization problem, the simultaneous procedure finds the best solution for a case with smaller variable numbers and the sequential procedure performs better for a case with larger variable numbers. The PSO algorithm performs better than the MADS algorithm for more complex problems with larger variable numbers in both the simultaneous and sequential procedures. The sequential procedure is worth considering in practice for problems with large scale and a limited computational budget.

Summary Identification of vuggy intervals and understanding their connectivity are critical for predicting carbonate reservoir performance. Although core samples and conventional well logs have been traditionally used to classify vuggy facies, this process is labor intensive and often suffers from data inadequacies. Recently, convolutional neural network (CNN) algorithms have approached human-level performance at multiimage classification and identification tasks. In this study, CNNs were trained to identify vuggy facies from a well in the Arbuckle Group in Kansas, USA. Borehole-resistivity images were preprocessed into half-foot intervals; this complete data set was culled by removing poor-quality images to generate a cleaned data set for comparison. Core descriptions along with conventional gamma ray, neutron/density porosity, photoelectric factor (PEF), and nuclear magnetic resonance (NMR) T2 data were used to label these data sets for supervised learning. Hyperparameters defining the CNN network size (numbers of convolutional layers/filters and the numbers of fully connected layers/neurons) and minimize overfitting (dropout rates, patience, and minimum delta) were optimized. The median losses and accuracies from five Monte Carlo realizations of each hyperparameter combination were the metrics defining CNN performance. After hyperparameter optimization, median accuracy for vuggy/nonvuggy facies classification was 0.847 for the cleaned data set (0.813 for the complete data set). This study demonstrated the effectiveness of using microresistivity image logs in a CNN to classify facies as either vuggy or nonvuggy, while highlighting the importance of data quality control. This effort lays the foundation for developing CNNs to segment images to estimate vuggy porosity.

Summary Polymer flooding offers the potential to recover more oil from reservoirs but requires significant investments, which necessitate a robust analysis of economic upsides and downsides. Key uncertainties in designing a polymer flood are often reservoir geology and polymer degradation. The objective of this study is to understand the impact of geological uncertainties and history matching techniques on designing the optimal strategy for, and quantifying the economic risks of, polymer flooding in a heterogeneous clastic reservoir. We applied two different history matching techniques (adjoint-based and a stochastic algorithm) to match data from a prolonged waterflood in the Watt Field, a semisynthetic reservoir that contains a wide range of geological and interpretational uncertainties. Next, sensitivity studies were carried out to identify first-order parameters that impact the net present value (NPV). These parameters were then deployed in an experimental design study using Latin hypercube sampling (LHS) to generate training runs from which a proxy model was created using polynomial regression. A particle swarm optimization (PSO) algorithm was employed to optimize the NPV for the polymer flood. The same approach was used to optimize a standard waterflood for comparison. Optimizations of the polymer flood and waterflood were performed for the history-matched model ensemble and the original ensemble. The optimal strategy to deploy the polymer flood and maximize NPV varies based on the history matching technique. The average NPV and the variance are predicted to be higher in the stochastic history matching compared to the adjoint technique. This difference is due to the ability of the stochastic algorithm to explore the parameter space more broadly, which created situations in which the oil in place is shifted upward, resulting in a higher NPV. Optimizing a history-matched ensemble leads to a narrow range in absolute NPV compared to optimizing the original ensemble. This difference is because the uncertainties associated with polymer degradation are not captured during history matching. The result of cross comparison, in which an optimal polymer design strategy for one ensemble member is deployed to the other ensemble members, predicted a decline in NPV but surprisingly still showed that the overall NPV is higher than for an optimized waterflood, even for suboptimal polymer injection strategies. This observation indicates that a polymer flood could be beneficial compared to a waterflood, even if geological uncertainties are not captured properly.

Summary Hydrocarbon (re‐)development projects need to be evaluated under uncertainty. Forecasting oil and gas production needs to capture the ranges of the multitude of uncertain parameters and their impact on the forecast to maximize the value of the project for the company. Several authors showed, however, that the oil and gas industry has challenges in adequately assessing the distributions of hydrocarbon production forecasts. The methods for forecasting hydrocarbon production developed with digitalization from using analytical solutions to numerical models with an increasing number of gridblocks (“digital twins”) toward ensembles of models covering the uncertainty of the various parameters. Analytical solutions and single numerical models allow calculation of incremental production for a single case. However, neither the uncertainty of the forecasts nor the question in which the distribution of various outcomes the single model is located can be determined. Ensemble‐based forecasts are able to address these questions, but they need to be able to cover a large number of uncertain parameters and the amount of data that is generated accordingly. Theory‐guided data science (TGDS) approaches have recently been used to overcome these challenges. Such approaches make use of the scientific knowledge captured in numerical models to generate a sufficiently large data set to apply data science approaches. These approaches can be combined with economics to determine the desirability of a project for a company (expected utility). Quantitative decision analysis, including a value of information (VoI) calculation, can be done addressing the uncertainty range but also the risk hurdles as required by the decision‐maker (DM). The next step is the development of learning agent systems (agent: autonomous, goal‐directed entity that observes and acts upon an environment) that are able to cope with the large amount of data generated by sensors and to use them for conditioning models to data and use the data in decision analysis. Companies need to address the challenges of data democratization to integrate and use the available data, organizational agility, and the development of data science skills but making sure that the technical skills, which are required for the TGDS approach, are kept.

Summary Offshore reservoirs are subjected to pressure loading from the ocean tide. The resulting pressure fluctuation, notably its amplitude and phase, provides valuable information regarding the formation compressibility and heterogeneity. The purpose of the present study is twofold: First, to propose a method for calculating tidal efficiency from harmonic analysis of regional tide stations and detrended bottomhole pressure (BHP), and second, to compare the compressibility from tidal analysis with that obtained from rock‐mechanics measurements and material balance. This case study is on a fractured oil field for which matrix laboratory measurements alone cannot capture the large‐scale formation compressibility that is driven by the fracture distribution. This paper will show how, in the absence of seabed‐pressure measurements, a synthetic diurnal tide can be simulated by interpolating the harmonic constituents of neighboring tide stations. The validity of this method was confirmed on two offshore fields. A new procedure that combines a Savitzky and Golay ( 1964 ) (SG) filter and cubic splines gave satisfactory results to filter out the tidal signal residual from the reservoir‐transient response for both buildup and interference tests. In addition, this paper found that wells in fractured areas of the field have higher rock compressibility and exhibit a higher tidal efficiency. The same effect is observed in flank wells with higher water saturation. Conversely, the tidal efficiency is dramatically reduced in wells experiencing gas breakthrough.

Summary Closed‐loop reservoir management (CLRM) consists of continuous application of history matching and optimization of model‐predictive control to maximize production or reservoir net present value (NPV) in any given period. Traditional field‐scale implementation of CLRM by using a large number of reservoir models, in particular when uncertainty is accounted for, is computationally impractical. This presented machine-learning-assisted workflow uses the echo state network (ESN) coupled with an empirical water fractional flow relationship as a proxy to replace time‐consuming simulations and improve the computational efficiency of the CLRM. The ESN, under the paradigm of reservoir computing, provides a specific architecture and supervised learning principle for recurrent neural networks (RNNs). ESNs, with randomly generated and invariant input weights and recurrent weights, greatly minimize the computational load and solve potential problems during typical backpropagation through time in traditional RNNs while it still obtains the benefits of RNNs to memorize temporal dependencies. Also, the linear readout layer makes the training much faster using analytical ridge regression. Field‐level well control and production‐response data are fed into the workflow to obtain a trained ESN and fitted fractional‐flow relationship, which will represent/reproduce the dynamics of the reservoir under various well‐control scenarios. Further production optimization is directly applied to the matched models to maximize reservoir NPV. The optimized well‐control scenario is applied, and further observation is obtained to update the models. History matching and production optimization are performed again in a closed‐loop fashion. The previously mentioned advantages make ESN a very powerful tool for CLRM, with both history matching and production optimization quickly accomplished, and make near‐real‐time CLRM possible. In this paper, two case studies will be presented to prove the effectiveness of the proposed workflow.