This work presents a new open-source carbonate reservoir case study, the COSTA model, that uniquely considers significant uncertainties inherent to carbonate reservoirs, providing a far more challenging and realistic benchmarking test for a range of geo-energy applications. The COSTA field is large, with many wells and large associated volumes. The dataset embeds many interacting geological and petrophysical uncertainties in an ensemble of model concepts with realistic geological and model complexity levels and varying production profiles. The resulting number of different models and long run times creates a more demanding computational challenge than current benchmarking models. The COSTA model takes inspiration from the shelf-to-basin geological setting of the Upper Kharaib Member (Early Cretaceous), one of the most prolific aggradational parasequence carbonate formations sets in the world. The dataset is based on 43 wells and the corresponding fully anonymised data from the northeastern part of the Rub Al Khali basin, a sub-basin of the wider Arabian Basin. Our model encapsulates the large-scale geological setting and reservoir heterogeneities found across the shelf-to-basin profile, into one single model, for geological modelling and reservoir simulation studies. The result of this research is a semi-synthetic but geologically realistic suite of carbonate reservoir models that capture a wide range of geological, petrophysical, and geomodelling uncertainties and that can be history-matched against an undisclosed, synthetic ‘truth case’. The models and dataset are made available as open-source to analyse several issues related to testing new numerical algorithms for geological modelling, uncertainty quantification, reservoir simulation, history matching, optimization and machine learning.

@en

Geothermal energy has a high potential to contribute to a more sustainable energy system if the associated economic risks can be overcome in the design process. The development planning of deep geothermal reservoirs (over 1000 m depth) relies on computer models to forecast and then optimize system design. Optimization is easy where all the objective's (e.g. NPV) optimization parameters and, most importantly, the geology are considered as known, but this is almost always not the case. Where the complex engineering design (e.g. well placement) meets significant geological uncertainty every development option should be tested using an expensive simulation against the range of geological possibilities. The impracticality of simulating so many models results in a limited exploration of geological uncertainties and development options. Consequently, the risk of improper system design cannot be properly assessed. This paper presents an approach to understand the trade-offs in maximizing heat extraction while minimizing energy usage in re-injection for a new geothermal reservoir development while considering the uncertainty from 18 different geological models. Our approach is computationally feasible because we apply multi-objective particle swarm optimization (MOPSO), to an ensemble of response surface models, built using Gaussian process regression (GPR), for each and every geological scenario. MOPSO explores the trade-off surface for the competing objectives using the mean reservoir responses (covering the geological uncertainty). Our results highlight the impact of geological uncertainty on the optimal well placement and show the need to consider geological uncertainties adequately in optimization. The work demonstrates the shortcomings of using only one geological model of a geothermal reservoir and/or a single objective in optimization. We additionally demonstrate the practicalities of using response surface models in this way for geothermal systems. We anticipate that our work raises awareness for the scope of optimization of geothermal reservoir design under geological uncertainty.@en

This work presents a new open access carbonate reservoir case study that uniquely considers the majoruncertainties inherent to carbonate reservoirs using one of the most prolific aggradational parasequencecarbonate formation sets in the U.A.E., the Upper Kharaib Member (Early Cretaceous), as an analogue. Theensemble considers a range of interpretational scenarios and geomodelling techniques to capture the maincomponents of its reservoir architecture, stratal geometries, facies, pore systems, diagenetic overprints andwettability variations across its shelf-To-basin profile. Fully anonymized data from 43 wells across 22 fields in the Bab Basin in the U.A.E from different geo-depositional settings and height above the free water level (FWL) was used. The data comprises of a fullsuite of open-hole logs and core data which has been anonymized, rescaled, repositioned and structurallydeformed; FWLs were normalized and the entire model was placed in a unique coordinate system. Theresultant static and dynamic models(s) capture the geological setting and reservoir heterogeneities ofselected fields but now at a manageable scale. Synthetic production data has been generated by adding wellsto an undisclosed -truth case- model to obtain field-wide and well-by-well production data (oil, gas, andwater rates, bottom-hole pressures etc.) from simulation runs. The original oil in place (OOIP) and reserves that have been computed from these models are syntheticand unique. Here we present an initial field development plan and corresponding reservoir simulations thatshowcase the heterogeneity inherent to the model and demonstrate the variability of the flow and storagecapacity of the different reservoir architectures found in and around the Bab Basin. This is an exampleapplication of how we can use synthetic production data to improve our understanding of flow behavioursin carbonates. The novelty of our work is the provision of a unique open access dataset which enables reproduciblescience in the field of reservoir characterisation and simulation, and helps training new generations ofgeoscientists and reservoir engineers in the art of characterising, simulating and predicting the reservoirperformance of carbonate reservoirs under different recovery processes.

@en

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 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 semi-synthetic reservoir that contains a wide range of geological and interpretational uncertainties. An ensemble of reservoir models is available for the Watt Field, and history matching was carried out for the entire ensemble using both techniques. 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 a Latin Hypercube to generate training runs from which a proxy model was created. The proxy model was constructed using polynomial regression and validated using further full-physics simulations. A particle swarm optimisation algorithm was then used to optimize the NPV for the polymer flood. The same approach was used to optimise a standard water flood for comparison. Optimisations of the polymer flood and water flood were performed for the history matched model ensemble and the original ensemble. The sensitivity studies showed that polymer concentration, location of polymer injection wells and time to commence polymer injection are key to optimizing the polymer flood. The optimal strategy to deploy the polymer flood and maximize NPV varies based on the history matching technique. The average NPV is predicted to be higher in the stochastic history matching compared to the adjoint technique. The variance in NPV is also higher for the stochastic history matching technique. This is due to the ability of the stochastic algorithm to explore the parameter space more broadly, which created situations where the oil in place is shifted upwards, resulting in higher NPV. Optimizing a history matched ensemble leads to a narrow variance in absolute NPV compared to history matching the original ensemble. This is because the uncertainties associated with polymer degradation are not captured during history matching. The result of cross comparison, where an optimal polymer design strategy for one ensemble member is deployed to the other ensemble members, predicted a decline in NPV but surprisingly still shows that the overall NPV is higher than for an optimized water food. This indicates that a polymer flood could be beneficial compared to a water flood, even if geological uncertainties are not captured properly.

@en

This work presents a new open access carbonate reservoir case study that uniquely considers the major uncertainties inherent to carbonate reservoirs using one of the most prolific aggradational parasequence carbonate formation sets in the U.A.E., the Upper Kharaib Member (Early Cretaceous), as an analogue. The ensemble considers a range of interpretational scenarios and geomodelling techniques to capture the main components of its reservoir architecture, stratal geometries, facies, pore systems, diagenetic overprints and wettability variations across its shelf-to-basin profile. Fully anonymized data from 43 wells across 22 fields in the Bab Basin in the U.A.E from different geodepositional settings and height above the free water level (FWL) was used. The data comprises of a full suite of open-hole logs and core data which has been anonymized, rescaled, repositioned and structurally deformed; FWLs were normalized and the entire model was placed in a unique coordinate system. The resultant static and dynamic models(s) capture the geological setting and reservoir heterogeneities of selected fields but now at a manageable scale. Synthetic production data has been generated by adding wells to an undisclosed 'truth case' model to obtain field-wide and well-by-well production data (oil, gas, and water rates, bottom-hole pressures etc.) from simulation runs. The original oil in place (OOIP) and reserves that have been computed from these models are synthetic and unique. Here we present an initial field development plan and corresponding reservoir simulations that showcase the heterogeneity inherent to the model and demonstrate the variability of the flow and storage capacity of the different reservoir architectures found in and around the Bab Basin. This is an example application of how we can use synthetic production data to improve our understanding of flow behaviours in carbonates. The novelty of our work is the provision of a unique open access dataset which enables reproducible science in the field of reservoir characterisation and simulation, and helps training new generations of geoscientists and reservoir engineers in the art of characterising, simulating and predicting the reservoir performance of carbonate reservoirs under different recovery processes.@en

Polymer flooding offers the potential to recover more oil from reservoirs but requires significant investmentswhich necessitate a robust analysis of economic upsides and downsides. Key uncertainties in designinga polymer flood are often reservoir geology and polymer degradation. The objective of this study is tounderstand the impact of geological uncertainties and history matching techniques on designing the optimalstrategy 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 matchdata from a prolonged waterflood in the Watt Field, a semi-synthetic reservoir that contains a wide rangeof geological and interpretational uncertainties. An ensemble of reservoir models is available for the WattField, and history matching was carried out for the entire ensemble using both techniques. Next, sensitivitystudies were carried out to identify first-order parameters that impact the Net Present Value (NPV). Theseparameters were then deployed in an experimental design study using a Latin Hypercube to generate trainingruns from which a proxy model was created. The proxy model was constructed using polynomial regressionand validated using further full-physics simulations. A particle swarm optimisation algorithm was then usedto optimize the NPV for the polymer flood. The same approach was used to optimise a standard water floodfor comparison. Optimisations of the polymer flood and water flood were performed for the history matchedmodel ensemble and the original ensemble. The sensitivity studies showed that polymer concentration, location of polymer injection wells and timeto commence polymer injection are key to optimizing the polymer flood. The optimal strategy to deploythe polymer flood and maximize NPV varies based on the history matching technique. The average NPVis predicted to be higher in the stochastic history matching compared to the adjoint technique. The variancein NPV is also higher for the stochastic history matching technique. This is due to the ability of thestochastic algorithm to explore the parameter space more broadly, which created situations where the oilin place is shifted upwards, resulting in higher NPV. Optimizing a history matched ensemble leads to anarrow variance in absolute NPV compared to history matching the original ensemble. This is because theuncertainties associated with polymer degradation are not captured during history matching. The result ofcross comparison, where an optimal polymer design strategy for one ensemble member is deployed to the other ensemble members, predicted a decline in NPV but surprisingly still shows that the overall NPV ishigher than for an optimized water food. This indicates that a polymer flood could be beneficial comparedto a water flood, even if geological uncertainties are not captured properly.

@en

There are a number of benchmark studies for clastic reservoirs (e.g. the PUNQ, SPE10, Brugge Field, Watt Field, or Norne Field models) for challenges such as EOR, testing new numerical algorithms for reservoir simulation and uncertainty quantification. To date, the only carbonate benchmark reservoir is UNISIM-11-D which focuses on the Brazilian Pre-Salt however, neglects any geological uncertainties. This paper presents an update on ongoing work to create an open access reservoir model ensemble that considers the major ranges of interpretational uncertainties inherent to carbonate reservoirs covering interpretational scenarios and modelling choices such as geomodelling techniques. The result of this research is a semi-synthetic but geologically realistic suite of carbonate reservoir models that capture a wide range of geological, petrophysical, and geomodelling uncertainties and can be history-matched against a synthetic “truth case”. Eventually the models will be made available as open access data to provide the opportunity for reproducible science in the field of reservoir simulation and help training new generations of reservoir engineers in the art of simulating and history matching carbonate reservoirs over the next decade.

@en

Reliable production forecasting for fractured carbonate reservoirs is a challenge. Natural fractures, adverse wettability and complex matrix heterogeneity are all uncertain and can all negatively impact upon recovery. Ideally, we should consider different reservoir concepts encapsulated in a large ensemble of reservoir models to quantify the impact of these and other geological uncertainties on reservoir performance. However, the computational cost of considering many scenarios can be significant, especially for dual porosity/permeability models, rendering robust uncertainty quantification impractical for most asset teams. Flow diagnostics provide a complement to full-physics simulations for comparing models. Flow diagnostics approximate the dynamic response of the reservoir in seconds. In this paper we describe the extension of flow diagnostics to dual porosity models for naturally fractured reservoirs. Our new diagnostic tools link the advective time of flight in the fractures to the transfer from the matrix, identifying regions where transfer and flux are not in balance leading to poor matrix oil sweep and early breakthrough. Our new diagnostics tools have been applied to a real field case and are shown to compare well with full-physics simulation results.

@en

This paper describes the application and testing of innovative dual porosity flow diagnostics to quantitatively rank large ensembles of fractured reservoir models. Flow diagnostics can approximate the dynamic response of multi-million cell models in seconds on standard hardware. The need for new faster screening methods stems from the challenge of making robust forecasts for naturally fractured carbonate reservoirs. First order uncertainties including the distribution and properties of natural fractures, matrix heterogeneity and wettability can all negatively impact on recovery. A robust multi-realisation approach to production forecasting is often rendered impractical due to the time cost for simulating many models. We have extended existing flow diagnostics techniques to dual porosity systems by accounting for the matrix-fracture exchange. New metrics combine the transfer rate with the advective time of flight in the fractures identifying risk factors for early water breakthrough and providing quantitative measures of dynamic heterogeneity. We have compared ranking a large ensemble of synthetic fractured reservoir models using dual porosity flow diagnostics and using full-physics simulation. The synthetic ensemble explores a number of different geological concepts around the fracture distributions, wettability and matrix heterogeneity which can. Not only does the flow diagnostic ranking agree well with the cumulative oil ranking the run time for the flow diagnostics is <0.25% of the total simulation time. This significant reduction in the time to compare models allows more time to spend running full physics simulation on the important and geologically diverse cases that offer the most insight.

@en

In this study we have applied new dual-porosity flow diagnostics to a recently discovered offshore fractured carbonate reservoir undergoing charactersiation studies. Carbonate reservoirs are typically highly hetrogeneous, naturally fractured and often mixed to oil wet, all of these factors are uncertain and can negatively impact upon recovery. With few wells drilled at the time of this study significant uncertainty hinders robust decision making. A robust multi-realisation approach is rendered impractical as the run time for a single realisation of the dual-porosity model is in excess of several days. Instead of using brute force we have utilised grid based flow diagnostics as a fast screening tool to select reservoir models for further full-physics simulations. The CPU time of flow diagnostics is almost negligable. Flow diagnostics are numerical tests perfomed on the static model that provide the time-of-flight, tracer partitions, drained/swept volumes and well pairs. In addition, dual-porosity metrics link the advective flow in the fractures to transfer from the matrix, indicating regions where flow and transfer are unbalanced and hence at risk of early breakthrough. Over 30 flow diagnostic tests were performed in under 10 minutes, the equivalent screening would take weeks using simulation. Results have shown that the fracture intensity and wettability are the most signifcant uncertainties that impact upon transfer and recovery, this effect would be missed by an equivalent single-porosity model. Well placement in this reservoir is very sensitive; the results show the proposed placement is effective for the assumed yet uncertain facies distribution. This broad sensitity screening has guided the ongoing modelling strategy, in partcular pinpointing the need for detailed characterisation of the fracture and facies distributions. Flow diagnostics are hence an excellent way to complement production forecasting workflows by providing a tool for quickly ranking and selecting scenarios for further detailed full-physics simulation, allowing us to focus more computational resources on reservoir models that are of particular interest.@en

In this paper we describe new and innovative flow diagnostics tools for dual porosity models for naturally fractured reservoirs. Our new diagnostic tools allow us to compare and rank large numbers of geological models based on their approximate dynamic response in almost negligible time. Fast ranking methods allow us to select a representative ensemble of models that quantify geological uncertainty for robust production forecasting via full physics reservoir simulation. Reliable production forecasting for fractured carbonate reservoirs is a challenge. Natural fractures, adverse wettability and complex matrix heterogeneity are all highly uncertain and can all negatively impact upon recovery. Ideally we should consider a large and diverse ensemble of reservoir models to quantify the impact of geological uncertainty on reservoir performance. However, the computational cost can be significant, especially for dual porosity/permeability models. A brute force approach using powerful workstations, clusters or cloud computing can be taken to reduce the time investment. But this is not always possible, rendering robust uncertainty quantification impractical for many asset teams. Often only a small subset of scenarios is considered which may collapse into a single base case, from which development decisions are made. Base cases often fail to predict future production, need frequent modifications, lack geological realism and provide incomplete risk assessments, often causing asset teams to miss economic opportunities. Flow diagnostics can provide dynamic reservoir information in a fraction of the time for full physics simulation. We propose a workflow where we utilise flow diagnostics as a ranking tool to complement forecasting using reservoir simulation throughout. Our approach addresses the model run time, allowing us to use standard hardware. Flow diagnostics solve simplified physics to approximate the dynamic response of the reservoir, from this we can calculate and visualize key dynamic properties (e.g., time-of-flight, drained and swept reservoir volumes, time-to-breakthrough, decline rates, sweet spots, well-allocation factors). Flow diagnostics provide robust indicators of dynamic heterogeneity that allow us to select a diverse ensemble of models that captures the range of uncertainty. In this work, novel diagnostics utilising physically based transfer models have been developed to account for the fracture-matrix exchange, which otherwise could only be obtained from lengthy simulation. A new Damköhler number based metric DaDP links the advective time-of-flight in the fractures to the transfer from the matrix. DaDP identifies fast and slow draining regions of the matrix, stagnant regions within the fracture network and wells at risk of water breakthrough. This information can subsequently be used to optimise well placement and rates to maximise production and delay water breakthrough.

@en

Carbonate reservoirs contain much of the world's remaining hydrocarbon resources. These reservoirs are challenging to characterise and forecast production due to uncertainty in the fracture distribution and wettability. Ideally a large number of models needs to be simulated to quantify uncertainty but, due to time constraints, typically only a small subset of scenarios is considered. A dual-porosity module with a flow diagnostic toolbox has been implemented into MRST by the Carbonate Reservoir Group at Heriot-Watt University. The new diagnostics provide useful information that approximates the reservoir dynamics and matches full-physics simulation results, but can be obtained in a fraction of the time needed for simulation. Using flow diagnostics hence enables the fast screening and ranking of vast numbers of static models for fractured reservoirs. This enables us to consider the full range of geological uncertainty whilst minimising the number of models required for traditionally heavy simulation tasks such as, production forecasting, optimisation and uncertainty analysis that rely on full-physics simulations. Flow diagnostics hence complement and extend knowledge obtained from conventional fractured reservoir characterisation, calibration, and simulation workflows.@en

Fractured carbonate reservoirs are known for its complexity, this has big impact on production especially when EOR methods are used. Foam is known for its potential to improve recovery by diverting the flow to the poorly swept regions. In order to capture the uncertainties associated with foam EOR in fractured carbonates, multiple simulations of different geological scenarios must be tested.

@en

When simulating foam floods, uncertainties exist in both the foam and reservoir parameters however the combination of these uncertainties are rarely incorporated in forecasting. Foam flooding is an effective enhanced oil recovery method that controls mobility, reduces gas relative permeability, delays gas breakthrough and helps improve sweep efficiency. Thus it is often used in highly heterogeneous reservoirs where significant subsurface uncertainties exist. Foam uncertainties exist as (a) foam stability is controlled by a number of factors such as critical water and surfactant concentrations, brine salinity, and oil saturation which are unknown in the subsurface spatially and (b) foam flood simulation requires the accurate description of multiple parameters used in the foam flood models which are unknown. This study quantifies and compares the impact of uncertainties associated with foam model parameters with the heterogeneity of a fractured carbonate reservoir, an analogue to the highly prolific Arab D formation. Foam model parameters are not known a-priori but can be tuned to experimental data, which ideally represent a range of foam regimes and reservoir conditions. Geological heterogeneities in fractured carbonate reservoirs are complex and include, matrix wettability, fracture density/orientation and initial saturation distribution. To quantify uncertainties geological uncertainties in fractured carbonate reservoirs, an automated framework was used to history match the production response of a fractured carbonate field by varying geological parameters. This accounts for the geological uncertainties during a waterflood, which are then combined with foam uncertainties from experimental analysis in the optimisation step, by optimising the mean response of the model to foam flooding across a range of geological and foam scenarios. Our workflow used a combination of Particle Swarm Optimisation for history matching and manual optimisation, the final results of which show a wide range of possible impacts of foam flooding from different but equally well matched reservoirs. The novelty of our work is that it demonstrates how parameters that control foam stability and hence effectiveness in mobility control are related to both foam properties and geological uncertainty. Carrying these uncertainties into foam model properties from core to field scale will translate into considerably more robust estimates of uncertainty when predicting field-scale recovery compared to simulations that only consider uncertainty in the reservoir model.@en

Naturally Fractured Reservoirs (NFR) hold a significant fraction of remaining petroleum reserves. Recovery factors from NFR are usually less than in conventional reservoirs due to associated high uncertainty throughout the characterisation and modelling phases. This particularly includes the modelling and upscaling of the fracture domain using Discrete Fracture Networks (DFN). Computer assisted history matching and prediction is becoming increasingly popular as they help finding multiple history-matched models and probabilistic forecasts. Therefore, the associated uncertainty can now be quantified in a limited time frame. However, the results of a history match are known to depend on initial reservoir properties, including fracture permeability and matrix shape factors. Geological uncertainty in these two factors is exacerbated by the DFN upscaling errors. We show how DFN modelling can be used to increase geological prior knowledge and hence produce more geologically consistent models. To highlight DFN upscaling errors, we use a realistic dataset from an onshore fractured reservoir to show how the DFN upscaling error could propagate through to the history matching phase. We compare history matching of three models with different DFN upscaling processes. Results from state-of-the-art assisted history matching and prediction were found to depend on the static properties and particularly the computation of effective fracture permeability during DFN upscaling. This upscaling error alone leads to very different reservoir models, despite the best history matched models being of comparable quality. Hence, this leads to more uncertainty in reservoir production forecast. The identification of DFN upscaling errors is therefore crucial for better uncertainty quantification in reservoir simulation of NFR.@en

Benchmark problems have been generated to test a number of issues related to predicting reservoir behaviour (e.g. Floris et al., 2001, Christie and Blunt, 2001, Peters et al., 2010). However, such cases are usually focused on a particular aspect of the reservoir model (e.g. upscaling, property distribution, history matching, uncertainty prediction, etc.) and the other decisions in constructing the model are fixed by log values that are related to the distribution of cell properties away from the wells, fixed grids and structural features and fixed fluid properties. This is because all these features require an element of interpretation, from indirect measurements of the reservoir, noisy and incomplete data and judgments based on domain knowledge.Therefore, there is a need for a case study that would consider interpretational uncertainty integrated throughout the reservoir modelling workflow. In this benchmark study we require the modeller to make interpretational choices as well as to select the techniques applied to the case study, namely the geomodelling approach, history matching algorithm and/or uncertainty quantification technique. The interpretational choices will be around the following areas: (1)Top structure interpretation from seismic and well picks.(2)Fault location, dimensions and the connectivity of the network uncertainty.(3)Facies modelling approach.(4)Facies interpretations from well logs cutoffs.(5)Petrophysical property prediction from the available well data.(6)Grid resolution-choice between number of iterations and model resolution to capture the reservoir features adequately.A semi-synthetic study is based on real field data provided: production data, seismic sections to interpret the faults and top structures, wireline logs to identify facies correlations and saturation profile and porosity and permeability data and a host of other data. To make this problem useable in a manageable time period multiple hierarchically related gridded models were produced for a range of different interpretational choices.

@en