SPE 89338

A Study of IOR by CO2 Injection in the Gullfaks Field, Offshore Norway H. Agustsson, SPE, STATOIL ASA, G.H. Grinestaff, SPE, PETROTEL INC.

Copyright 2004, Society of Petroleum Engineers Inc. This paper was prepared for presentation at the 2004 SPE/DOE Fourteenth Symposium on Improved Oil Recovery held in Tulsa, Oklahoma, U.S.A., 17–21 April 2004. This paper was selected for presentation by an SPE Program Committee following review of information contained in a proposal submitted by the author(s). Contents of the paper, as presented, have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material, as presented, does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE meetings are subject to publication review by Editorial Committees of the Society of Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to a proposal of not more than 300 words; illustrations may not be copied. The proposal must contain conspicuous acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O. Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.

Abstract This paper describes the feasibility study of a large-scale miscible CO2-WAG (MWAG) injection scheme in the Gullfaks Field, offshore Norway. We describe the reservoir engineering workflow and simulation techniques, the predicted production and injection profiles, and the main infrastructure solutions under consideration.

Compositional cross-section models and recently available streamline-tracer simulation techniques are employed to scale up from element models to a fast, full-field simulator with a high degree of flexibility.

Figure 1: Gullfaks Field Location

The starting point for the workflow is a set of black oil and streamline front tracking models, history matched on coarse and fine grids. A fast, finely gridded streamline model is used to identify the MWAG injection targets, define injection well locations and completion strategy. Fine gridded cross-sections are extracted and used in a compositional simulator to study and quantify the miscible displacement process. These are the used to derive scaling parameters used in a simple, ultra-fast

streamline-tracer model, scaling the MWAG process up to field level. The streamline-tracer model interactively optimises solvent allocation and generates production predictions on a well-by-well basis. Water flood recovery and incremental IOR are predicted simultaneously in a single simulation run.

In addition, the general economic limitations and example technical solutions for implementation of a CO2 MWAG on the Gullfaks Field are briefly described.

Shetland

North Sea

Stavanger

GullfaksBergen

Norway

GreatBritain

Denmark

Shetland

North Sea

Stavanger

GullfaksBergen

Norway

GreatBritain

Denmark

Statfjord

Gullfaks

Magnus SnorreMurchison

VeslefrikkNinian

Figure 2: Gullfaks Reservoir Structure

Introduction to the Gullfaks Field The Gullfaks field has been extensively documented in the literature. For a recent overview of reservoir description and management, refer to Agustsson, et al1. To summarise, the field is located 180 km NE of Bergen, Norway, see Figure 1,

2 SPE 89338

These units also communicate when juxtapositioned across the many faults in the field. These aspects pose a challenge to both modelling and managing the reservoir.

near the median line between the British and Norwegian sectors of the North Sea. The Gullfaks is in a mature oil province, neighbouring many large oil fields. Production is from sandstones in the Brent Group and the underlying Statfjord, Cook and Lunde formations. The Brent Group represents the major part of the producing formations in Gullfaks. It consists of Jurassic sands, with shallow marine to fluvial deposits at around 1800m subsea.

The field started production in 1986, and has as of July 2003 produced roughly 300 MSm3 of the currently estimated reserves of 342 MSm3. In place reserves are around 582 MSm3. Figure 3 shows the development in reserves and production from 1986 to date. The current production strategy is through pressure maintenance at above saturation pressure (230 bar in the Brent) by water injection, augmented by gas injection and a modest natural water influx. The plans envisage continued production to 2020, with an overall recovery factor of 58%.

The field installations are shown in Figure 4. Production is to three large gravity-base concrete platforms, each with processing and water injection capabilities, and two with gas injection facilities. Gas export is by pipeline to shore and the stabilised oil is stored in tanks and exported by tanker. Third party processing of fluids from the nearby Tordis and Vigdis fields, as well as the recent development of the subsea Gullfaks Satellites, boosts production, but still leaves vacant process capacity. Primary drilling on Gullfaks is now completed. Only seven out of a total of 136, drilling slots are currently vacant, and 120 platform wells are in operation.

IOR and the Gullfaks Field The Gullfaks Field is currently well into decline and can justly be described as a mature oil field. In spite of excellent projected recovery of around 58% at the end of field life at around 2020, significant volume of oil remains in the ground.

Figure 3: Gullfaks Reserves Development

Reservoir quality is generally high to very high, with

porosity around 30-35%, and HCPV weighted average horizontal permeability of the order of 800 mD. The reservoir pressure and temperature are around 310 bar and 74OC respectively, at datum (1850m TVD SS). The oil gravity is between 32-36o API and the Brent GOR is around 100 v/v. The reservoir is extensively layered and severely faulted; see Figure 2, with numerous intercommunicating compartments. Apart from the lateral segmentation, with the major N-S faults mostly sealing, the Brent may be roughly divided into three main hydraulic units vertically, referred to as Upper, Middle and Lower Brent, nevertheless with a degree of vertical communication.

IOR activities have been focused on in Gullfaks almost from the very beginning. The immiscible hydrocarbon gas WAG implemented from the early 1990’s is one example of a successful implementation of an additional reservoir mechanism. Other uses of associated gas injection include cyclic gas injection in producers. Together, these mechanisms have recovered a very significant amount of attic oil. During high oil prices in the early 1990’s, different chemical IOR methods, such as surfactants and polymers, were studied but determined uneconomic.

Additional success has been achieved on a well-by-well basis, such as conformance control, both mechanical and chemical, sand control, hydraulic fracturing, and smart wells. One of the largest contributions has come from successful infill drilling, targeting the remaining oil mapped through strategic data gathering, where 4D seismic has lately played a major role. Inexpensive through tubing drilling is expected to account for a significant portion of future IOR.

Figure 4: Gullfaks Field Installations

In order to lift the recovery even higher, and to contact maximum possible hydrocarbon volume in one sweeping operation, the possibilities for miscible gas injection on a field scale were investigated. As the associated gas is immiscible in Brent at reservoir conditions, and enrichment by imported intermediate components was determined to be too expensive, the focus turned to CO2 injection. A series of PVT experiments and fluid studies showed the minimum miscibility pressure of CO2 in the Brent to be below the bubble point pressure of 230 bar. The current average reservoir pressure is around 300 bar; close to initial, so miscible conditions prevail.

SPE 89338 3

Simulation Tools for the Gullfaks Field. The Gullfaks field is a large oil field, and operative numerical simulation models have recently been constructed for the Brent Group, with which this project is solely concerned. Figure 5 shows a schematic of the various simulation models currently in use, the number of blocks they contain and what has been their primary function. They are all upscaled from one 3D geological (structural and sedimentological) model gridded to some 25 million blocks. The model construction process is described in detail by Jacobsen et al2. The models are variously run in black oil, compositional or in streamline formulation. In all the models, the blocks are nominally 80x100m in areal dimension. Those models that are employed in the MWAG evaluation are highlighted.

The Brent group represents the single, most extensive target of remaining hydrocarbons in the field. Other parts of the field may represent additional upside potential, but are outside the scope of the current study. Feasibility Study of Miscible CO2 Injection (MWAG) in the Brent Group of the Gullfaks Main Field Traditionally, a study of miscible gas injection on a full-field scale relies on some form of upscaling from detailed segment or sector models. These can range from simply multiplying up the recovery from one or more pattern models in the case of a pattern flood, to analytically scaling up wrt hydrocarbon pore volume from representative segment models, or to extend an already existing full field model using pseudo-parameters to describe the effect of the miscible process and calibrated against fine gridded reference models. Of the latter, the Todd-Langstaff (T-L) solvent option is one alternative.

It is generally not considered feasible to run large, full-field compositional simulation models directly due to the computational demands associated with fine grids and simulator options essential to adequately model the miscible displacement process.

The T-L options is also computationally demanding, and additionally has inherent difficulty in simultaneously and reliably predicting gas utilisation and oil recovery. The T-L omega model works best for first contact miscible process where well rates are similar and stable, It was therefore decided to adopt a technique recently made available and based on streamline-tracer simulation5,6. The main motivation for this decision was the reported speed of the computation, its flexibility wrt handling and optimizing injection strategy, and not least, its track record in application in other reported studies and actual projects. Conveniently, the technique and tools had just become commercially available for the first time.

Figure 6: Full-field Model FFM104

A short description follows of only the models used in this MWAG study, in the logical order in which they are used:

Figure 5: An Overview of Gullfaks Simulation Models

FFM19: 19 Layer Black Oil Model.

FFM19 is the primary history matched black oil, finite difference model for water injection. The simulation layer thickness varies from around 10 to 60m. Key history match parameters are fault transmissibility and pseudo relative oil-water permeability curves. Elapsed run times range from 15-20 hours for the 13 years history and 20 years prediction, respectively. We consider this model unsuitable for compositional simulation due to the coarse grid and the associated physical and numerical dispersion problems3,4, but it is the primary reference, to which the streamline models described below are compared to

verify a history match and water flood prediction to start up of MWAG.

Gullfaks Simulation Model FamilyMain Field, Brent Group

FFM1972k blocks

FFM52240k blocks

Thruth Models2D and 3D

Cross-sections

FFM3Streamline Modelwith Tracer Logic

FFM104460k blocks

FFM166Upper Brent

Cross-sections only

FFM189Lower Brent

Cross-sections only

3D IntegratedGeological Model

25 mill. blocks

4 SPE 89338

Figure 7: Reference “Truth” Model Locations

FFM104: 104 Layer Streamline Front Tracking Model.

FFM104 has been converted from finite difference to streamline formulation. The grid has an average layer thickness of 2 meters. An acceptable history match, wrt FFM19, on a field level, has been verified when using rock relative permeability curves and directly upscaled properties

and transmissibilities without any modification. Run with 6 months time steps, elapsed run times of 4-6 hours are achieved. This front tracking model is used to establish the following items:

a) Identify miscible gas injection target areas b) Develop well perforation strategy c) Calculate the connected hydrocarbon pore volume

(HCPV) for each MWAG injector location d) Extract 2D cross-sections and 3D sector Reference

Models to study the displacement process FFM104 has the advantage of detailed resolution,

capturing essential geological aspects, structural and sedimentological, governing the fluid flow in the reservoir. Figure 6 shows the model viewed from the southeast and above. It shows the extensive layering and contrast in reservoir quality, as well as the major fault segments and the top erosion on the eastern flank of the field.

The Compositional Reference Models. The physics of the

MWAG injection and displacement process are studied in a fine grid, compositional, finite difference simulator. Reference models, sometimes referred to as “Truth Models” provide recovery and timing information for calibrating the streamline-tracer model used to upscale to field level. These models are mechanistic 2D vertical sections of a single injection well, and primarily represent reservoir description changes and WAG injection conformance for each area of the field.

Model Description and Fluid Characterization. A total of six 2D Reference Models were extracted from the high resolution, streamline, front tracking model FFM104. The cross-section locations, shown in Figure 7, were chosen to represent the range of geological variation in each fault block and the field as a whole. The cross-section orientation was selected to obtain the best possible alignment with the streamlines, i.e. flooding pattern, in the full-field model. The grid was refined by a factor of 8 horizontally, and by 2 vertically in some layers, resulting in an average block size of 12 and 2 meters horizontally and vertically. A refinement by 8 is based upon our past experience with grid resolution effects in miscible gas displacement simulation. Figure 8 shows each model grid and the gas saturation after CO2 injection. The large difference in vertical sweep and solvent distribution between the models is evident. This shows clearly the need to use several models, from different locations across the field, to capture the effect of variation in reservoir architecture on recovery. The models also have different injector-producer well spacing to cover the range that exists in the field. The Reference Models incorporate physics generally considered essential when simulating miscible displacement such as hysteresis and gas trapping.

We feel that in this situation, and especially if resources are limited, it is preferable to employ many computationally light 2D cross-sections in preference to one or few ‘heavy’ 3D sector models as the basis for full-field upscaling. A self- contained study of 2D vs 3D behaviour can additionally be conducted, if desired or deemed necessary.

SPE 89338 5

S ector M odel S um m ary fo r C O 2 S R K 8M odel W ater cut R ecovery % S TO IIP M odel Type A nalogy

N am e W A G S tart W AG S tart B ase W tfld W AG (S 2) V ertica l S w eep M echan ismS E C 2 89.81 66.51 75.42 81.83 P oor A ttick O ilS E C 3 95.04 54.27 67.09 77.05 G ood S hale under-runS E C 4 93.02 52.47 65.45 72.92 F air S hale under-runS E C 5 92.31 48.86 62.14 76.35 G ood K h Dom inatedS E C 6 93.55 63.35 70.26 78.17 P oor Th ie f Dom inatedS E C 7 93.95 61.86 71.99 77.22 F air K h Dom inated

S E C2 C O 2 S RK 8

S E C 3 CO 2 S R K 8

S E C4 C O 2 S RK 8

S E C 6 C O 2 S R K 8

S E C 5 C O 2 S RK 8

S E C 7 C O 2 S R K 8

G AS Saturatio n Year 2030

G S ATLE G E ND

Inc R ec%IO R6.429.967.47

14.217.925.23

Figure 8: Fine Gridded Compositional Reference Model

Three EoS (Equation of State) models were developed and

tuned to CCE, CVD, swelling and MMP experiments with reservoir fluid and CO2. An SRK EoS with 8 pseudo components was selected for the reservoir simulation work. Peng-Robinson EoS models with 7 and 10 pseudo components respectively also matched the fluid experimental data well, but gave higher miscible recovery from the cross-sections. These results were used as sensitivities in the subsequent uncertainty analysis.

Reference Models Simulation Setup. Cross-section models

are individual entities even though they clearly represent the

geology at their location. However, the models must also be representative of the actual dynamic field conditions, such as pressure, saturations, recovery factor, and throughput rate for when they are used to establish incremental recovery. With this in mind, the Reference Models were conditioned to represent the field state at expected MWAG start up. Well perforation strategy is also adapted to reflect of what can reasonably be anticipated achieved in the field. The models are run with constant reservoir pressure control, and inject solvent up to slug sizes of 0.1, 0.5, and 1.0 HPV (hydrocarbon pore

volume) fraction at a constant rate of 4.5% TPV (total pore volume throughput), with a WAG ratio equal to 1 with quarterly cycles. Rate dependency was also evaluated using

Dependence of Recovery on RateGullfaks Lower Brent I-1 Segment

0

0.072

0.1000.117 0.123

0.00

0.05

0.10

0.15

0.20

0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0% TPV Flood Rate

Rec

oery

% H

CPV

Rec %-HPV

Match Curve

Dependence of Recovery on Slug SizeGullfaks Lower Brent I-1 Segment

0

0.0520.069

0.0930.110

0.00

0.05

0.10

0.15

0.20

0.0 0.2 0.4 0.6 0.8 1.0 1.2Slug Size, % TPV

Rec

oery

% H

CPV

Rec %-HPV

Match Curve

Dependence of Recovery on RateGullfaks Lower Brent I-1 Segment

0

0.072

0.1000.117 0.123

0.00

0.05

0.10

0.15

0.20

0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0% TPV Flood Rate

Rec

oery

% H

CPV

Rec %-HPV

Match Curve

Dependence of Recovery on Slug SizeGullfaks Lower Brent I-1 Segment

0

0.0520.069

0.0930.110

0.00

0.05

0.10

0.15

0.20

0.0 0.2 0.4 0.6 0.8 1.0 1.2Slug Size, % TPV

Rec

oery

% H

CPV

Rec %-HPV

Match Curve

Figure 9: A set of Performance, or Solvent Efficiency, Curves

6 SPE 89338

3.0%, 4.5%, 8.0%, and 10% TPV/year, where the base case for water flood is 4.5% TPV/year. This results in 6 WAG simulation runs and 1 base case water flood run to evaluate MWAG performance, per cross-section.

From these runs, dimensionless performance curves are computed, quantifying IOR in relation to solvent injection in terms of slug size as percentage of hydrocarbon pore volume (%HCPV), and solvent injection rate as percentage of pore volume throughput (%TPV). Figure 9 shows a set of typical performance curves. The performance curves describe the cumulative and incremental effect, i.e. IOR oil production, as a function of cumulative and incremental solvent injection. They can also be thought of as describing solvent efficiency. That is how they are subsequently used in optimizing solvent allocation to injection wells during the scale up to full-field level.

Figure 10: Tracer Matching, IOR Oil

Tuning the Tracer Models to the Reference Models. The

Reference Model cross-sections are scaled up to the level of the streamline-tracer model to calibrate the streamline-tracer parameters. The tracer adsorption and pore fraction parameters in the upscaled cross-sections are adjusted to match oil and solvent rates, cumulative production and solvent breakthrough time, as predicted by the fine gridded cross-sections. Figure 10 shows a typical set of matches for a single cross-section. Injection and production of gas and oil are individually matched.

2D vs. 3D Reference Models. A 3D sector model in a

highly dipping fault block was built in order to quantify the effect of transversely lateral fluid movement in a dipping structure, absent in a 2D model where lateral fluid movement is forced along the cross-section. Areal sweep was evaluated by running the 3D model in Streamline Front Tracking mode and evaluating the connected HCPV during MWAG injection based on time-of-flight. That volume was then used to scale dimensionless recovery and compare to the 2D reference model. A comparison of 2D vs. 3D recovery is shown in Figure 11. The bottom yellow curve is for Ethane injection, not CO2. The difference between the two top curves, CO2 injection into 2D and 3D models is of the order of 0.1, which is much smaller than the differences observed between different 2D models in Figure 10. The dominant factor in 2D – 3D comparison is generally likely to be the reservoir

properties at the WAG injector. Other authors have reported similar findings, e.g. from the Forties6 field.

0.00%

3.92%

6.66%

8.73%

10.39%

0.00%

3.37%

4.93%

6.48%

8.57%

0.00%

3.47%

7.32%

9.01%9.01%

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00

3D E300

2D E300

CO2 E300

Rec

over

y fr

actio

n hp

v

HPV Slug Size

Dimensionless Recovery vs Solvent InjectedBase Rate=4.5%TPV

Figure 11: Comparison of Dimensionless Recovery from 2D and 3D Reference Models

Streamline Front Tracking: 104 Layers

Black Oil Model: 19 Layers

IOR Scale Tracer Model: 3 Layers

DX (Meters)

Streamline Front Tracking: 104 Layers

Black Oil Model: 19 Layers

IOR Scale Tracer Model: 3 Layers

DX (Meters)

Figure 12: Comparison of Simulation Grids

FFM3: 3 Layer Streamline-Tracer Model. FFM3 is

scaled up from FFM104, reducing the number of layers from 104 to 3. The 3 layers correspond to the vertical division of the field into hydraulic flow units as described earlier. The model includes a tracer option, the purpose of which is to scale up the miscible displacement process to a full-field level. The principles are briefly described below, with full details supplied by Giordano5.

Vertical transmissibility is set to zero and the layers only communicate across fault juxtapositions. The motivation for having three layers in the model is to have the possibility of individual assessment of recovery from these three reservoir units, whereas conventionally, application of the technique makes use of only a single layer for the whole reservoir.

SPE 89338 7

Figure 13: Comparison of History Match, different Models

Input to FFM3 comes from the FFM104 and the Reference

Models: From FFM104: • • •

• •

•

• •

MWAG Flooding Units MWAG well locations, type and definition Connected HCPV by MWAG-injector

From the compositional Reference Models:

Tracer Adsorption and Timing parameters Solvent Efficiency Curves for different slug sizes and injection rates

In the FFM3 MWAG simulations, the IOR performance is

determined from the cumulative solvent injection into each injector, the associated set of performance curves and the tracer control parameters calibrated to each Reference Model. Each MWAG injector is assigned a performance curve based upon location or reservoir description. A built-in algorithm optimises the solvent allocation to the injection wells each time step by ranking the incremental yield for each MWAG

injector, using the performance curves, slug size and target WAG ratio.

Oil and water flows (the simulator is 2-phase, so associated gas production is calculated from oil production using a fixed solution GOR) are computed simultaneously for water flood in the conventional front tracking manner, and for the miscible displacement by propagating tracers along oil-water streamlines. The model runs the combined history and 30-year prediction on a single CPU in less than half an hour, elapsed time. The water flood history in FFM3 is matched to the FFM19 results by adjusting the pseudo relative permeability curves only. However, this does not affect the MWAG IOR prediction because IOR oil and solvent streams are not subject to model saturations, but are represented by tracers, independently tuned to match the results of the compositional Reference Models previously described. The Streamline-Tracer model thus does not model the miscible process physics. Rather using three tracers, it mimics the behaviour as modelled by the compositional Reference Models. The tracers are defined as follows:

Effective solvent, which mobilises IOR oil The IOR oil, and Ineffective solvent, which does not mobilise IOR oil

86 88 90 92 94 96 98 000

10000

20000

30000

40000

50000

60000

70000

Oil

Rat

e (S

M3/

DA

Y)

0

5.0x10 7

1.0x10 8

1.5x10 8

2.0x10 8

2.5x10 8

3.0x10 8

3.5x10 8

4.0x10 8

Cum

mulative O

il (SM3)

Field Oil Production Rate & Total

FFM19 HISTORY FFM104 FFM3

86 88 90 92 94 96 98 000

10000

20000

30000

40000

50000

60000

70000

Wat

er R

ate

(SM

3/D

AY)

0

2.0x10 7

4.0x10 7

6.0x10 7

8.0x10 7

1.0x10 8

1.2x10 8

1.4x10 8

1.6x10 8

1.8x10 8

2.0x10 8

2.2x10 8

Cum

mulative W

ater (SM3)

Field Water Production Rate & Total

FFM19 HISTORY FFM104 FFM3

86 88 90 92 94 96 98 000

10000

20000

30000

40000

50000

60000

70000

Oil

Rat

e (S

M3/

DA

Y)

0

5.0x10 7

1.0x10 8

1.5x10 8

2.0x10 8

2.5x10 8

3.0x10 8

3.5x10 8

4.0x10 8

Cum

mulative O

il (SM3)

Field Oil Production Rate & Total

FFM19 HISTORY FFM104 FFM3

86 88 90 92 94 96 98 000

10000

20000

30000

40000

50000

60000

70000

Wat

er R

ate

(SM

3/D

AY)

0

2.0x10 7

4.0x10 7

6.0x10 7

8.0x10 7

1.0x10 8

1.2x10 8

1.4x10 8

1.6x10 8

1.8x10 8

2.0x10 8

2.2x10 8

Cum

mulative W

ater (SM3)

Field Water Production Rate & Total

FFM19 HISTORY FFM104 FFM3

Note that the sum of the effective and ineffective tracer injected equals the total solvent injection. With time, the ratio of ineffective to effective tracer increases, representing the reducing solvent efficiency with cumulative injection.

Model and Type No. CPUs

Elapsed Time (hrs)

FFM104 Black Oil, finite diff. 8 40

FFM19 Black Oil, finite diff. 4 15

FFM104 Streamline 1 7

FFM3 Streamline-Tracer 1 0,25

Table 1: Comparison of Model Run Times

The tracer model accounts directly for vertical sweep,

perforation conformance, geology, and other physical aspects incorporated into the Reference Models. However, areal sweep in FFM3 is entirely determined by the conventional water-oil front tracking method. The assumption that areal sweep for MWAG will follow water-oil streamlines has not been reported to be a problem for predictions with streamline-tracer models in Alaska, where pattern floods and low dip formations are prevalent. In the case of Gullfaks where formation layers dip typically between 15-25 degrees, additional 3D sector model simulations were run to verify that essential areal

8 SPE 89338

displacement behaviour of the solvent was satisfactorily taken into account using only 2D Reference Models. As mentioned earlier, this result was also observed in the case of the Forties work6, although it should not be taken as a general observation, but checked in each individual case.

Summary of the Simulation Models. The simulation models that have been described represent a wide range of grid resolution and geological detail. The upscaling processes, both as regards static reservoir parameters and the dynamic miscible fluid displacement process, are intended to capture the essential aspects of how these phenomena affect oil recovery and solvent utilisation in such a way as to be able to be brought forward to a fast, accurate and practical tool to scale detailed Reference Model results to field level.

Figure 12 compares the different full field models FFM104, FFM19 and FFM3, illustrating the difference in resolution and complexity. It shows a schematic of plan view and cross-sections from the 3 full-field simulation models used for the Gullfaks gas displacement IOR evaluation. The plan view on the left is coloured by block grid increment which ranges from 50-150 meters, where each cross-section of the right of Figure 12 is coloured by PermZ showing shale in dark blue. As may be seen, the model with 3 layers does not have vertical heterogeneity, whereas 19 layers result in thick blocks of sand and shale that are highly upscaled. Using 104 layers has minimal upscaling of vertical variation and shows discrete sand and shale packages.

Figure 14: Overview of Simulation Workflow

Figure 13 shows the water flood history and prediction for these models, showing that they all give a result at field level that is acceptable. Table 1 lists the elapsed times for running these models. Although the absolute numbers will depend on the type of hardware, a comparison of the relative speed is instructive. Simulation Workflow. Figure 14 shows a somewhat simplified diagram of the simulation workflow. FFM104 is the 104-layer streamline front tracking model used to develop the injection strategy through selection of injector locations,

identify areas of bypassed oil and to divide the field into MWAG flooding regions. It is also used to determine the connected HCPV for each MWAG injector, a key input for the optimisation algorithm for solvent allocation in FFM3.

Six 2D cross-sections, the Reference Models, are extracted from FFM104, representative of areas previously identifies as MWAG flood regions. They are run in a compositional simulator to generate the Performance Curves for each region. The curves are entered into FFM3, the 3-layer full-field streamline-tracer simulator, where each MWAG injector is assigned a particular performance curve according to its location in the reservoir. The performance curves enable the estimation of incremental IOR and solvent efficiency, each simulation time step, for the MWAG injectors. The solvent efficiency is used to optimize solvent allocation to the injectors throughout the simulation.

The Reference Model cross-sections are also upscaled to layering analogous to that in FFM3, and used to calibrate the tracers to match IOR oil and solvent production in the corresponding finely gridded, compositional Reference Models. The calibrated tracer parameters are then exported to FFM3 to control tracer propagation on a field scale. The solvent is propagated along all the streamlines emanating from each MWAG injector, resulting in mobilisation of IOR oil, solvent entrapment and breakthrough. As it is a 2-phase simulator, the associated gas production is calculated from the oil production using a fixed solution GOR. The water flood process is calculated in the conventional streamline – front

tracking manner, with the miscible process mimicked by the tracers.

In summary, this technique has the unique ability to simultaneously generate profiles for the water and miscible flood recovery. The production and injection constraints are the individual platform and well capacities and the total solvent supply. The last consists of imported solvent and reinjection fraction for the back produced solvent. Solvent allocation during each simulation time step is determined by the number of available injection wells, the WAG-ratio and predicted IOR oil production according to the efficiency given by the individual wells’ performance curves. Development of the MWAG Injection Strategy. Although it is essential for a successful miscible injection project to employ a thermodynamically efficient injectant, it is probably even more important to

have an injection strategy that places the injectant optimally in the reservoir, over time. This is in fact true of any recovery mechanism; good well locations are a must for efficient drainage. With costly injectant, this is paramount to obtain maximum reserves for every ton or cubic metre of injectant.

Streamline-Tracer Model

1D Cross-sections (1-layer)

Export MWAG areas and connected

pore volume / injector

CalibrateTracer Controls

to match oil and solventRecovery and Timing

FFM3Streamline-

Tracer Model(3-layers)

Upscale 104 to 3 layers

Simultaneous Full-Field

WI and MWAG Recovery Estimates

FFM104Streamline

Model(104 layers)

Extract X-sectionsupscale to 1 layer

Reference ModelsCompositional

2D Cross-sections (100-200 layers)

Export Tracer Controls

Extract X-sectionsand refine

ComputePerformanceCurves

Streamline-Tracer Model

1D Cross-sections (1-layer)

Export MWAG areas and connected

pore volume / injector

CalibrateTracer Controls

to match oil and solventRecovery and Timing

FFM3Streamline-

Tracer Model(3-layers)

Upscale 104 to 3 layers

Simultaneous Full-Field

WI and MWAG Recovery Estimates

FFM104Streamline

Model(104 layers)

Extract X-sectionsupscale to 1 layer

Reference ModelsCompositional

2D Cross-sections (100-200 layers)

Export Tracer Controls

Extract X-sectionsand refine

ComputePerformanceCurves

Export Performance

Curves

Export Performance

Curves

Gullfaks has traditionally been developed with water injection as the primary reservoir energy source. The water has been largely injected, into key ‘super injectors’ in the aquifer, below the original OWC (oil-water contact). This way, reservoir pressure has been maintained successfully through voidage replacement. Some over-injection has also taken place, resulting in partial pressure support through the regional aquifer to nearby fields. The current well network in the field is therefore primarily designed for long distance volume

SPE 89338 9

replacement with water through the aquifer, but with some injection wells placed closer to the producers, in what with time has become the flushed zone. This strategy is unsuitable for an MWAG for a variety of reasons, some of the most important ones being:

1. The long distance between peripheral injectors and oil zone producers would result in late solvent breakthrough. A shorter distance is desirable in order to have better vertical sweep and rapid reuse of the expensive solvent.

2. Significant volumes of solvent would be trapped and lost out in the aquifer and would not mobilise any oil

3. Control over the slug size applied to each flood pattern becomes impossible with the long throughput time and uncertain flow paths in the absence of breakthrough in wells

Figure 15: Classical Aquifer Water Injection Pattern

Rapid circulation of solvent through the IOR target accelerates the recovery, and each flooding area can receive concentrated flood treatment before moving on to the next, rapidly expanding the treatment as more solvent is available from back production. Figures 15 and 16 illustrate the difference in the reservoir flood pattern, where on one hand the energy is expended and dissipated over long distances by peripheral aquifer injection, and on the other hand through closely spaced injector-producer pairs in the hydrocarbon zone, resulting in a more even coverage aerially and contacting more of the oil in place.

Injection Well Locations for the MWAG. In order to

achieve timely and efficient sweep of the IOR target, many watered-out producers were reassigned to MWAG injection. In all, 33 wells are assigned to MWAG injection from 2008. Most of the wells will require a degree of workover such as plugging, zone isolation, re- and additional perforating and tubing replacements. Of the 33 wells, 14 are currently water injectors (including one WAG injector) and the rest are

producers, including 5 future well targets that will be turned to MWAG injection, either immediately or in due course.

The operations required to prepare each well for injection are routine on the Gullfaks field, and the time, equipment and costs associated well known from experience. It is assumed that all the well work can be completed in 2007 and 2008.

Figure 16: MWAG Injection Pattern

The MWAG flooding areas are shown in Figure 17. There are 11 areas represented by the 6 different Performance Curves applied to the 33 MWAG injectors. Reservoir coverage with this well configuration is estimated at 72%, using FFM104.

Results CO2 MWAG Reservoir Management Impact. The current development strategy for the Gullfaks Field is to inject water and modest quantities of associated gas to 2006, and only water thereafter. Considerable effort has been put into optimizing this scenario, but mostly by optimizing injection, production and well utilization (completion strategy, workovers, etc) within the constraints of the existing injection well pattern and from drilling infill producers. The full field simulation tools are useful aids in exploring alternative management strategies for the field.

The MWAG scenario represents a considerable change in reservoir management. Injection would now take place mostly above the original OWC, utilizing many watered-out producers. This will result in a much more rapid circulation of injected fluids due to closer well spacing. The simulator predicts accelerated oil production, as well as more rapid sweep to lower oil saturation in the inter-well space. The effect of the water injection alone is significantly enhanced with the new strategy, with additional benefits being brought by the miscible displacement by CO2.

10 SPE 89338

I1

I2

I3I5

H5

H7

G6

G4

G2F3

F4

AQ

I1

I2

I3I5

H5

H7

G6

G4

G2F3

F4

AQ

I1

I2

I3I5

H5

H7

G6

G4

G2F3

F4

AQ

Figure 17: MWAG Flooding Regions

Segregated Processing and Direct Reinjection. Several

options exist for implementation of a CO2 MWAG in the Gullfaks field. We will describe one that is relatively simple in terms of infrastructure upgrades and modification is termed “Segregated Processing and Direct Injection”.

Figure 19: Solvent Saturation in 2008 at startup

The Gullfaks field is a processing centre for 3rd party production from nearby fields. Many of these rely on gas sales as a main source of revenue. If contamination of the sales gas by CO2 is to be avoided, it can be achieved by dedicating one of the two processing trains to the production stream from the MWAG areas on the Gullfaks-A and -C platforms, whilst the other trains is reserved for non-CO2 contaminated production. The capacities of the processing trains conveniently enable such an arrangement.

Figure 20: Solvent Saturation in 2029

Calculations show that the inclusion of the hydrocarbons in the injectant has only minor negative consequences. The maximum dilution of the injectant is 40% on GFA in the first year, and then rapidly reduced to less than 20% and further, as increasing amounts of CO2 are backproduced. In order to approximate this strategy in the 2-phase streamline-tracer simulator, the imported solvent mass is adjusted up to reflect the contribution of the hydrocarbon gas in the injectant. In the current study, the hydrocarbon gas production is simply averaged, assuming that the return rate of hydrocarbon injection gas is the same as for solvent, over the entire period and – in the simulator - adding a corresponding reservoir volume of solvent to the import. By avoiding having to extract the solvent from the production stream for reinjection, the solvent utilization is also increased. Some CO2 must nevertheless be expected lost, and the reinjection fraction in the simulator is set at 95%.

Figure 18: Oil Saturation in 2008, Top Reservoir

Additionally, to simplify the recovery of back produced injectant, the entire gas production; solvent and hydrocarbon gas, from the MWAG trains is mixed with the CO2 import from shore and injected.

SPE 89338 11

At that level, a useful reservoir slug size is also achieved during the field’s operating lifetime, and the resulting injection and production profiles can be accommodated within the platform capacities, with manageable upgrades where necessary. Greater injection rates and an extended injection period can be accommodated, but a cost-benefit analysis would reveal the optimum. The project parameters presented are considered a reasonable starting point. Table 2 shows the simulation constraints and results.

Results / Up to Year: 2020 2030

Cumulative Injectant GSm3 47 60 (includes hydrocarbon gas) Mt 87.4 111.6 Imported Injectant GSm3 25 25 Mt 46.5 46.5 Backproduced Injectant GSm3 18 32 Mt 33,5 57.7 Incremental Oil, MWAG MSm3 48 58 due to CO2 alone, MSm3 24 31 Net Efficiency ton CO2 / Sm3 oil 2.0 1.5 Gross Efficiency ton CO2 / Sm3 oil 3.7 3.6

Table 2: CO2 MWAG Project Parameters, Directly Simulated Figures

Figure 18 shows the oil saturation at top reservoir in 2008 and Figures 19 and 20 show the solvent saturation in 2008 and 2029, showing the high degree of reservoir coverage

Figure 21 shows comparison of the oil production profiles for three separate cases:

1. Conventional water and gas injection (current

strategy) as predicted by FFM19. The profile is run to 2020 only, as the economic limit for the field life according to current strategy is shorter.

CO2 MWAG Injection and Production Profiles. The CO2 supply level of approximately 5 Mt/year over a 10-year period is based on an evaluation of quantities that can be realized within the project time frame.

2. Water flood only using the 33 well injection pattern from 2008, and

3. CO2 MWAG. The two latter profiles are run to 2030 in the expectation that economic life for the field in these scenarios can be extended well beyond 2020.

12/86 12/90 12/94 12/98 12/02 12/06 12/10 12/14 12/18 12/22 12/26 12/300

10K

20K

30K

40K

50K

60K

Oil

Prod

uctio

n R

ate

(SM

3/D

AY)

0

100M

200M

300M

400M

Cum

ulative Oil Production, Sm

3

In the simulations, all the new injection wells come on line

simultaneously, even though it is acknowledged that it will take two years to prepare the extra wells for injection. However, at the current stage of the investigation, this is not considered to be of major importance, but will be corrected if and when plans proceed towards implementation.

Figure 22 shows the solvent injection and back production profiles. The MWAG contribution to the field’s economic life can be roughly estimated, albeit in a simplified manner, by examining Figure 21. Assuming that the field life with the ‘no-action’ strategy is around 2018, one can see that with the MWAG, that level of production, or more, is maintained to 2029. If the future OPEX of the field per unit production volume is maintained or even reduced through efficiency gains, extension of field life could be realized, other things being equal.

Figure 21: Oil Production, Base Case Water flood and MWAG

Sensitivity Parameters and Uncertainty

12/00 12/04 12/08 12/12 12/16 12/20 12/24 12/280

2M

4M

6M

8M

10M

12M

14M

16M

(SM

3/D

AY)

0

10G

20G

30G

40G

50G

60G

70G Field Cum

ulative Solvent Product (SM3)

Several sensitivities have been run, and others incorporated from previous and separate studies of water flooding. The main sensitivities are listed in Table 3.

“Attic Oil” refers to a subtraction from simulated IOR recovery due to attic oil already produced from dry gas injection. The other parameters are associated with the ability to maintain and run the offshore infrastructure, including wells. The CO2 flooding parameters are those specifically related to the prediction of results from the miscible displacement.

As can be seen from the P50 recoverable reserves in Table 3, a significant portion of the MWAG benefit is predicted to be attainable through water injection alone, using the well pattern that has been developed with miscible CO2 flooding in mind. Whereas this may represent a significant opportunity for Gullfaks, it detracts to a large extent from the potential for a miscible CO2 flood.

Figure 22: Solvent Injection and Production, MWAG

12 SPE 89338

It remains, however, to ascertain that such intensive water flooding of the Gullfaks field is manageable. This work is currently underway. Some of the major obstacles are:

•

•

• •

Increased power requirements for water injection systems Increased water production, liquid processing capacity, cleaning and disposing of greater volumes of produced water Increased potential for H2S production Increased Sand Production. Increased liquid production through fewer wells (many producers are reemployed as water injectors) restrained by MSFR (Maximum Sand Free Rate) is a challenge

Parameters Comments

Well Economic Limit Well reliability, wear and tear

Processing Capacity Platform gas and liquid production and injection capacity

Ope

ratio

nal

WCT Well production limit due to water cut

Reservoir Performance

Reflects the maximum variation in reservoir performance from E300 Truth Models

CO2 Supply Uncertainty in CO2 supply C

O2 F

lood

Pa

ram

eter

s

EoS Fluid Description PR07 results in higher oil recovery than SRK8

Att

ic O

il

Attic Oil that is already produced before CO2 MWAG

The streamline-tracer simulator is two phase, and cannot directly account for attic oil that has been removed by dry gas injection

P50

Rec

over

able

R

eser

ves,

MSm

3

34 47

22 28

CO2 MWAG to 2020 CO2 MWAG to 2030 CO2 part only to 2020 CO2 part only to 2030

Table 3: Main Sensitivity Parameters and Results from Risk analysis

Example Solution for Implementation of CO2 MWAG in the Gullfaks Field CO2 Supply Scenarios. Commercial injection using external CO2 sources is not currently taking place anywhere in the North Sea. There are no natural sources of CO2 in the North Sea or vicinity that are significant in the current context. Nor

does any infrastructure exist on land for CO2 removal or transport either on land or at sea. An entire infrastructure from capture, to transport to injection has therefore to be built nearly from scratch to realize a commercial IOR project such as that being considered here.

Melkøya, 0,7

Esbjerg, 2,0

Kårstø, 0,2 +

Sleipner

Brunsbüttel , 0,65

Antwerp /Rotterdam, 2 - 3

Gullfaks

Brae

Mongstad, 2,2

Grenland, 0,5 - 2

Ålborg, 3 + 2

Drax~9,0

Egernsund3,0

Kalundborg, 2,0 +

Melkøya, 0,7

Esbjerg, 2,0

Kårstø, 0,2 +

Sleipner

Brunsbüttel , 0,65

Antwerp /Rotterdam, 2 - 3

Gullfaks

Brae

Mongstad, 2,2

Grenland, 0,5 - 2

Ålborg, 3 + 2

Drax~9,0

Egernsund3,0

Kalundborg, 2,0 +

Figure 23: Sources of CO2 around the North Sea Region

Figure 23 shows a map of the North Sea and the

surrounding region. Current sources of industrial CO2 are marked on the figure, showing the quantity in MT/year. The sources are geographically scattered over a large, densely populated area. This makes the gathering of CO2 to a single point of export very difficult, if not impossible. Parallel transport, e.g. pipelines from several sources to Gullfaks, is likewise not an economically viable alternative.

In the technical solution alternatives where the backproduced CO2 is transported onshore with the hydrocarbon gas, the captured CO2 from the associated gas is returned from shore to Gullfaks. However, this adds cost. The option of mixing and reinjecting the contaminated hydrocarbon gas with the imported CO2 is considered to be the preferred alternative. Simulations have shown that the effect on recovery is minimal.

A number of alternatives have been evaluated, with two described briefly below:

CO2 from Danish Coal Fired Power Plants. This scenario,

illustrated in Figure 24, assumes supply from Danish coal fired power plants with a single point of export. Two transport options have been considered for this scenario:

•

•

Option 1: Ship transport to a CO2 hub located onshore in western Norway, and pipeline transport from there to the field. Option 2: Transport by pipeline directly to the field.

SPE 89338 13

The first option has greater flexibility with regard to adding other sources, but involves more complex logistics. It is also

CO2 from Multiple Continental and Norwegian Sources.

This scenario illustrated in Figure 25 assumes that CO2 supplied from two continental chemical plants and captured for disposal at the Norwegian Snøhvit condensate field in the Arctic, is delivered to a hub at the Mongstad oil terminal on the Norwegian west coast. Additional CO2 is captured from existing emissions at Mongstad to make up the volume to around 5 Mt/year. Transport is by ship from the external sources to Mongstad, and by pipeline from Mongstad to the field.

Figure 25: Multi-source CO2 Supply and Transport

The aspect of material upgrades warrants a further

mention, as this would in many cases be a ‘show stopper’ for implementation of a project of this type in a field with ageing installations. Fortuitously, the Gullfaks installations were extensively upgraded during the 1990’s to 13% Cr specifications in anticipation of increasing H2S production, which subsequently has been mitigated to large extent. All the producers are completed with production tubing of the same material. This makes Gullfaks, amongst other fields of similar age in the North Sea, uniquely competent to handle corrosive well and process streams. Nevertheless, some material upgrades would be necessary. Although not trivial, they are considered manageable.

Figure 24: Single Source CO2 Supply and Transport Economic Constraints. Supplying CO2 to the North Sea fields is a costly affair. Some of the major reasons for this are:

Modifications and Upgrades of Field Installations. To implement CO2 MWAG injection on Gullfaks, considerable, but manageable, modifications to the field installations are required:

•

•

•

•

Long distance to sources capable of supplying CO2 in industrial quantities No transport solutions for CO2 in the requisite quantities exist, and must be built from scratch •

• •

• •

New wells, well conversions (already mentioned) De-bottlenecking of production systems Power plants that represent a significant potential

source of CO2 do not have capture implemented CO2 injection systems and manifolds on all three platforms The sources are scattered over a large area and across

national borders, in heavily populated areas with dense economic infrastructure

Additional intra-field pipelines, and Material upgrades Many of the large North Sea fields that would be desirable

to include in starting up a new phase of industrial oil exploitation that CO2 MWAG in reality is, are nearing the end of their economic life:

These modifications and upgrades are responsible for a significant portion of the total project CAPEX.

14 SPE 89338

•

•

Recovery from North Sea fields by water injection is already high, in many cases 50-70%. The remaining oil is scattered in the large reservoirs, representing a difficult MWAG target Field installations are ageing, and many couldn’t handle CO2 back production without major refurbishment

The technical potential for IOR is considered to be present

in the case of the Gullfaks field, but the present economic conditions prevent the project from going ahead on commercial grounds.

The procurement of CO2 represents the greatest single cost element. The other main cost elements are the CAPEX and OPEX. Even if some reductions could be achieved in these, they are likely to be modest. Therefore, a radical reduction in the cost of CO2 is required for commercial viability.

Considerable uncertainty is associated with the potential income from CO2 quota trading and its value to possible future sequestration in the Gullfaks field once commercial oil production is over. CO2 tax credits and delayed field abandonment costs are only modest income elements.

To make the project economically attractive, efforts could be made to form alliances to share the cost of the necessary infrastructure, consortium of several fields for implementation, also changes to national fiscal systems to adapt to late life production, increasing need for IOR.

Discussion Further Work. Because of advances in computing power in recent years there is more industry focus on field scale compositional simulation. The owners of the Gullfaks field have embarked upon a field-wide composition model using 52 layers and approximately 240,000 active cells. From working with the Reference Models, we estimate that considerably more detailed grid would be required for reliable results, although e.g. segment models with a flux option and/or local grid refinement may be of help. Nevertheless, incorporating the requisite physics, such as hysteresis and gas trapping would additionally increase computing requirements.

The streamline-tracer simulator is still only 2-phase, and so cannot model reinjection of associated gas. The addition of free gas as an additional tracer, development of a 3-phase streamline-tracer simulator, or a compositional streamline simulator might enhance the applicability of this methodology.

The attic oil produced on Gullfaks by dry gas injection cannot be taken into account directly by the streamline simulator. To correct for this, the simulated result is reduced by an amount representing historically produced attic oil. A more realistic option would be to take historical dry gas into account in the E300 truth models. Should the projected work with the compositional full field model mentioned above be successful, that would also enable this to be taken into account in an even more rational way.

Conclusions The potential for CO2 MWAG injection in The Gullfaks Field has been estimated using a commercially available streamline-tracer simulation tool for upscaling to field level.

Optimisation of the MWAG injection strategy has been

greatly aided using a streamline front tracking simulator with high grid resolution to first segment the field into flooding units, then to employ an inbuilt module for optimising solvent allocation on a well-by-well basis during runs with the upscaled streamline-tracer simulator. Both tools are extremely fast compared to other alternatives known to us.

The simulation tools and workflows developed for the

Gullfaks study are having a positive impact on the management of a reservoir that has been diligently worked for over 20 years; these tools are expected to be used well into the future for development and operational activities, irrespective of CO2 flooding goes ahead or not.

Even if the conclusion is that the economic conditions for a

CO2 MWAG are unfavourable, a significant opportunity to enhance the water flood strategy has possibly been identified through a radical redesign of the injection pattern. Acknowledgements The authors would like to thank the operator, Statoil, and partners in the Gullfaks license, Norsk Hydro and Petoro, for their support and permission to publish this paper.

References 1. Agustsson H., Strønen L.K, & Solheim O.A., "The Gullfaks

Field: Creating Value through Multidisciplinary Reservoir Management", paper OTC 10378 presented at the 1999 Offshore Technical Conference held in Houston, Texas, 3–6 May 1999.

2. Jacobsen, T., Agustsson, H., Alvestad, P. Kaas I., Opdal, S.T. & Strønen, L.K., "Identification and Modeling of Remaining Reserves in the Gullfaks Field", paper SPE 65412 poster presentation at the EUROPEC 2000 Conference held in Paris, France, October 24-25 2000.

3. Stalkup, F.I. et al, “Sensitivity to Gridding of Miscible Flood Predictions Made with Upstream Differenced Simulators”, SPE/DOE 20178, Apr. 1990.

4. Haajizadeh, M., Fayers, F., Cockin A., “Effects of Phase Behavior, Dispersion and Gridding on Sweep Patterns for Nearly Miscible Gas Displacement”, SPE 62995 ATCE Dallas, Tx October 2000

5. Giordano, R.M., Redman, R.S., and Bratvedt, F.: “A New Approach to Forecasting Miscible WAG Performance at the Field Scale,” SPEREE (June 1998) 192.

6. Turan, H., Skinner, R., Brand, P., Macdonald, C., Grinestaff, G., ”Forties CO2 IOR Evaluation Integrating Finite Difference and Streamline Simulation Techniques”, SPE 78298, presentation at the SPE 13th European Petroleum Conference held in Aberdeen, Scotland, U.K., 29–31 October 2002.