Anda di halaman 1dari 5

Application of Stream-Conversion

Methods To Generate Compositional

Streams From the Results of a
Multimillion-Cell Black-Oil-Simulation Study
of the Shaybah Field
Bassam Al-Awami, SPE, K. Hemanthkumar, SPE, Fatema Al-Awami, SPE, and Mansour Mohammedali, SPE,
Saudi Aramco

Summary work, we apply an efficient method6,7 to generate the composi-

Detailed compositional simulation of a giant reservoir with many tional rates from a black-oil simulation of the giant Shaybah field.
components is not practical. However, detailed multimillion-cell The theoretical basis for this method is presented in detail in Ref.
black-oil simulation of giant reservoirs is now quite feasible. In 7. Herein, we present only the pertinent information to elucidate its
this work, we apply an efficient method to generate the composi- application in this work.
tional rates from a black-oil simulation of the giant Shaybah field.
In situations in which the reservoir recovery mechanism is not Stream-Conversion Methods
dominated by compositional effects, an equation-of-state (EOS)- The stream-conversion method, as the name suggests, is used to
based stream-conversion method can be used. This stream- convert data from one form to another. Although the application of
conversion method relies on the fact that when laboratory pressure/ this method is general, we apply this method here to petroleum
volume/temperature (PVT) data measured on available well- streams and, more specifically, to the conversion of surface volu-
stream compositions are used to generate the black-oil PVT tables, metric rates of oil (qo) and gas (qg) from the black-oil simulator
some of the compositional information is lost. The stream- into overall compositional rates (ni) for the desired number of
conversion model retains this valuable compositional information components (i=1, nc). The stream-conversion equation for this
and applies it to each producing-well completion in the black-oil case is as follows:
simulation at every timestep.
ni = Sioqo + Sigqg, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (1)
As proof of the concept, the stream-conversion method was
applied to a black-oil simulation and to a limited (eight- where Sio is the stream-conversion factor for component i in the oil
component) compositional simulation to generate a 17-component phase and Sig is the stream-conversion factor for component i in
compositional stream, and the results were compared to the re- the gas phase. This stream-conversion process is essentially the
spective full EOS compositional simulation for a relatively small recombination of the reservoir equilibrium oil (xi) and gas (yi)
sector (250,000 cells) of the giant Shaybah field. The composi- compositions into a well-stream composition. The stream-
tional stream rates are in excellent agreement with the stream- conversion factors Sio and Sig, in terms of properties generated by
converted black-oil results. As would be expected, the computa- an EOS-based PVT software when simulating a laboratory deple-
tional costs of using the EOS-based compositional simulator (with tion-type (CVD, DLE, CCE) experiment, are given by Eqs. 2 and
17 components) are in excess of 40 times the black-oil-simulation 3 (please refer to Ref. 7 for a full derivation from the overall
time for the small-sector model. In general, the stream-conversion component material-balance equation):
method can be used to generate the dynamically varying compo-
sitional streams from any black-oil simulation for use in the design Sio = 关共Coo + Rs兲xi − Rs共rsCog + 1兲yi兴 Ⲑ k共1 − rs Rs兲 . . . . . . . . . . . (2)
and operation of surface facilities and in calculating the amounts and
of a certain cut [e.g., natural gas liquids (NGL)] from the produc-
tion streams. Sig = 关Rs共rsCog + 1兲yi − rs 共Coo + Rs兲xi兴 Ⲑ k共1 − rs Rs兲, . . . . . . . . . (3)
Introduction where Coo and Cog are the conversions from surface oil volume to
Recent advances in parallel-reservoir-simulation technology1 have an “equivalent” surface gas volume, defined as
made it feasible to model the performance of giant hydrocarbon Cox = K␥ox Ⲑ Mox, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4)
reservoirs with simulation models that retain the full geologic-
model resolution.2–5 These multimillion-cell simulation/geologic where subscript o denotes surface oil, subscript x denotes either o
models, when carefully conditioned to engineering data, lend them- for reservoir oil or g for reservoir gas, ␥ is the specific gravity of
selves to rapid history matching, despite their size.2–4 More impor- surface oil from the relevant phase, M is the molecular weight of
tantly, they are used actively in optimizing field development with surface oil from the relevant phase, and K is a constant. A full EOS
more confidence and in day-to-day reservoir management.3,4 fluid characterization is performed using all the laboratory PVT
The above-mentioned multimillion-cell simulation models use data, which preferably cover the full operating range of pressures
a black-oil treatment of the hydrocarbon fluids. Where composi- and compositions. This EOS is used to generate the black-oil PVT
tional treatment of the hydrocarbon fluids is desired, a conven- tables and the stream-conversion factors Sio and Sig from depletion
tional full EOS-based compositional simulation of a giant hydro- experiments covering the whole operating-pressure range in the
carbon reservoir with many components is not yet practical. In this reservoir with a high frequency of pressure points for the desired
number of components. The maximum number of components is
usually equal to or less than the most detailed number of compo-
nents in the fluid characterization.
Copyright © 2005 Society of Petroleum Engineers
The black-oil reservoir simulation is run as usual, but with the
This paper (SPE 84361) was first presented at the 2003 SPE Annual Technical Conference output of well completion rates (layer rates) for oil, gas, and water
and Exhibition, Denver, 5–8 October, and revised for publication. Original manuscript re-
ceived for review 28 January 2004. Revised manuscript received 18 February 2005. Paper
and completion grid-cell pressure at the frequency of the reporting
peer approved 30 March 2005. interval, which is usually once a month. The Streams conversion

310 August 2005 SPE Reservoir Evaluation & Engineering

Fig. 2—A 3D view of the sector model (Sabkhah 10).
Fig. 1—A 3D view of the Shaybah field-simulation model.
surface oil and gas rates are again converted to compositional rates
software (which uses generalized forms of Eqs. 1 through 3) is for process engineering, and the outlet streams from the surface
then used to generate the compositional stream rates for each well facilities are again converted to familiar quantities, such as crude
completion as a function of time. grade and the oil-tanker manifest items. If a single stream-
This process is almost equivalent to performing an EOS-based conversion method is used at all these steps, then all of the con-
compositional simulation at each production-well completion, ex- versions become consistent and accurate. Thus, as a routine, it is
cept that the actual simulation model is a black-oil simulation and, recommended to use the stream-conversion method for all black-
hence, is very fast, and the simulation model can retain the full oil simulation results to convert to compositional streams for sur-
geologic resolution. This approach is suitable for reservoirs in face process engineering and other applications.
which the recovery mechanism is not dominated by compositional Shaybah Field
effects. In situations in which the latter is the case,9,10 this ap-
proach can be applied to a compositional simulation with a limited The Shaybah field was discovered in 1968 and is located in the
number of components, and the stream-conversion method can be southeastern part of Saudi Arabia. It is a large carbonate field with
applied to convert the results into an extended number of compo- several billion barrels of oil in place and several trillion standard
nents for use in facilities design or to maximize the recovery of a cubic feet of free gas. The oil in the Shu’aiba reservoir is overlain
certain cut of the compositional stream (e.g., NGL). by a huge gas cap, which is the primary drive mechanism. Because
Consistency and accuracy are common threads in this stream- of the existence of this large gas cap and the rock variations in
conversion process. The hydrocarbon fluids from the reservoir go Shu’aiba reservoir, gas-coning problems are expected. Oil produc-
through a series of “conversions.” First, the reservoir simulator tion started in mid-1998 at 500,000 B/D. Approximately 85% of
(e.g., a black-oil simulator) uses the converted laboratory data to the produced gas is being recycled into the gas cap. A simulation
black-oil tables and generates the surface production rates. The model (with a black-oil fluid characterization) at full geologic-
model resolution was built, resulting in 3.5 million grid cells. A 3D
view of the Shaybah simulation model is shown in Fig. 1. The
historical field performance was matched, and a comprehensive
evaluation of alternative gas-production forecast scenarios and
their impact on sweep efficiency and ultimate economic oil recov-
ery was carried out using our in-house parallel reservoir simula-
tor.1 For evaluating the long-term effects of gas cycling and sur-
face process engineering/design and other optimizations, a com-
positional simulation of the field is desired. But because a full-field
EOS-based compositional simulation is not yet practical, the
stream-conversion method to convert the black-oil-simulation re-
sults into compositional results is being evaluated.
Shaybah Sector Model
Before applying the stream-conversion method for the full field,
this approach was evaluated with a relatively small sector model
(Sabkhah 10) in the Shu’aiba reservoir. The Sabkhah 10 area is in
the northwest part of the Shaybah field. A 3D view of the sector
simulation model displaying initial oil (green), water (blue),
and gas (red) saturations is shown in Fig. 2. This model has 81
gridblocks in the x-direction and 106 gridblocks in the y-direction,
with 30 layers resulting in 257,580 grid cells. This sector model
was chosen so that a full compositional simulation could be per-
formed and then compared with the black-oil converted composi-
tional results.
An EOS fluid characterization was performed with laboratory
data, resulting in a 17-component characterization. The 17-
component and eight-component names used in the limited com-
positional simulations are given in Table 1.

August 2005 SPE Reservoir Evaluation & Engineering 311

Fig. 3—Comparison of nonhydrocarbon rates. Fig. 4—Comparison of the lighter-hydrocarbon rates.

To evaluate the extension of a limited compositional simulation (i.e., N2, CO2, and H2S) vs. time. In all the comparison figures, the
using the stream-conversion method, the 17-component fluid char- symbols represent values from the EOS 17-component composi-
acterization was pseudoized to eight components. tional simulation (EOS17), and the solid lines represent values
The black-oil PVT tables were generated from the 17- from the stream-converted black-oil simulation (BO-CMP17).
component fluid characterization. This same EOS fluid character- We notice that the stream-converted black-oil results are in ex-
ization was also used in the Streams program to generate the cellent agreement with the results from the compositional simulation
stream-conversion-factor tables. for these relatively small amounts of nonhydrocarbon components.
In Fig. 4, we present the comparisons for the lighter compo-
Sector-Model Compositional Simulations nents: methane (C1), ethane (C2), and propane (C3) rates.
The sector-model simulations were run for 6.5 years, which in- All the components are being tracked accurately by the stream-
cluded a 1.5-year historical field-performance period and a 5-year converted black-oil simulation. In Fig. 5, we present the compari-
forecast period. Two compositional simulation runs were made. In sons for the intermediate component rates: n-butane (n-C4), isobu-
the first simulation run, the hydrocarbon fluid was represented tane (i-C4), n-pentane (n-C5), and isopentane (i-C5).
with 17 components. In the second simulation, the hydrocarbon The agreement between the EOS and the stream-converted re-
fluid was represented with eight components. sults is excellent. In Fig. 6, we present the comparisons for the
hexanes, heptanes, octanes, and nonanes.
Sector-Model Black-Oil Simulation The agreement continues to be very good for these components.
In Fig. 7, we present the results for the heavy components: n-
As noted previously, the black-oil PVT tables were generated from decane (C10), undecane (C11), and the lumped C12+ component.
the 17-component fluid characterization. The black-oil simulation For all the components, the stream-converted black-oil results
of the Sabkhah-10 model was run for the same simulation period are in excellent agreement with the results from the compositional
as the compositional simulations. However, the run time for the simulation. One of the key reasons is that the black-oil rates at the
black-oil model is approximately 40 times faster than the 17- well-completion level are used.7 Because the rates of all the com-
component compositional model. The Streams program was used ponents are in excellent agreement, any surface-processed compo-
to convert the surface oil and gas volumetric rates of each well nent rates should also be in very good agreement because the rates
completion into a 17-component compositional rate. will be processed through the same separator stages for both com-
positional- and black-oil-simulation results.
Comparison of Compositional Stream Rates
We now compare the EOS compositional-simulation rates for a Comparison for Limited-Component
typical well (Well A) in the Sabkhah-10 area with the composi- EOS Simulation
tional rates obtained by converting the black-oil-simulation results.
In cases in which compositional effects dominate the recovery
Fig. 3 shows a comparison of the nonhydrocarbon component rates
mechanism, a limited-component compositional simulation can be

Fig. 5—Comparison of the intermediate-hydrocarbon rates. Fig. 6—Comparison of C6 to C9 hydrocarbon rates.

312 August 2005 SPE Reservoir Evaluation & Engineering

Fig. 7—Comparison of the heavy-hydrocarbon rates.
Fig. 8—Limited-component to 17-component—C1 to C3.

performed. The results from this simulation can be extended to

detailed compositional rates. We illustrate this process by running (250,000 grid cells) sector model. Before this work, as stated in a
an eight-component fluid-characterized compositional simulation previous section, a comprehensive black-oil-simulation study of
of the Sabkhah-10 area. The Streams software is then used to the full Shaybah field was completed using a 3.5-million-cell
convert the component rates at each well completion to a detailed simulation model. One of the optimized cases was rerun, invoking
17-component output. The results are presented for a typical well the option to output well-completion (layer) rates and completion
(Well B) in the Sabkhah-10 area in Figs. 8 and 9 for the light grid-cell pressures (obviously, there is no need to rerun the black-
hydrocarbons and the heavy hydrocarbons, respectively. For the oil simulation if these quantities were output during the original
intermediate and nonhydrocarbons, the comparisions are similar. run). The stream-conversion method was then used to generate the
In these figures, the symbols represent the component rates from compositional streams for the entire simulation period (1998 to
the 17-component EOS (EOS17) compositional simulation. The 2042). In Fig. 10, we present the normalized component produc-
solid, dashed, and dotted lines represent the component rates from tion rates for one of the gas/oil-separation-plant (GOSP) feed
the eight-component EOS (EOS 8–CMP 17) compositional simu- streams in the Shaybah field.
lation converted by the Streams method into 17 components. For the sake of brevity, the individual-component rates are
In Fig. 8, we present the results for the light-hydrocarbon com- lumped into nonhydrocarbons, light components, intermediate
ponents (methane, ethane, and propane) that make up approxi- components, C6-to-C11 components, and C12+ components. The
mately 50 mol% of the initial overall composition. only additional time required to calculate the individual-
In Fig. 9, we present the comparison for the heavy components component rates is the time taken by the Streams software to
(C10 through C12+); these components make up approximately 22 convert the black-oil well completion stream rates. This process
mol% of the initial overall composition. takes only a few minutes for the several hundred wells in the
The good agreement in the component rates indicates that this Shaybah field.
approach can be used for situations in which the compositional
effects dominate the recovery mechanism. The EOS compositional Conclusions
simulation can be carried out with four or five pseudocomponents, In this work, we apply an efficient method to generate composi-
and the results can be converted (at each well completion), using tional results from a multimillion-cell black-oil simulation of the
the stream-conversion method, to the desired number of compo- giant Shaybah field.
nents for subsequent processing. The following conclusions can be drawn from this study:
1. By applying the stream-conversion method, we have been able
Shaybah Full-Field to generate compositional results for the giant Shaybah field by
Black-Oil-Simulation Conversion running a black-oil model of the field at geologic resolution (3.5
In a previous section, we presented a satisfactory comparison of million cells). Currently, it is not practical to run a 3.5-million-
the EOS-based compositional-simulation results with the stream- cell, 17-component compositional model.
converted black-oil results (Figs. 3 through 7) for a relatively small 2. The application of the stream-conversion method to the black-
oil-simulation results of the 250,000-cell sector model to gen-

Fig. 9—Limited-component to 17-component—C10 to C12+. Fig. 10—Normalized Shaybah GOSP component rates.

August 2005 SPE Reservoir Evaluation & Engineering 313

erate the component rates produces accurate results and thus 5. Al-Shaalan, T.M., Fung, L.S., and Dogru, A.H.: “A Scalable Massively
validates the approach for a real field. Parallel Dual-Porosity Dual-Permeability Simulator for Fractured Res-
3. For situations in which compositional effects may be important, ervoirs With Super-K Permeability,” paper SPE 84371 presented at the
the stream-conversion method applies equally well, as illus- 2003 SPE Annual Technical Conference and Exhibition, Denver, 5–8
trated in the eight-component EOS simulation of the sector October.
model and subsequent conversion to 17 components. 6. Whitson, C.H. and Brule, M.R.: Phase Behavior, Monograph Series,
4. One of the key reasons for the success of this method is its SPE, Richardson, Texas (2000) 20.
application at the well-completion stream level. This process is 7. Hoda, M.F.: “The Engineering of Petroleum Streams,” Doktor Ingenior
almost equivalent to performing an EOS-based compositional thesis, Norwegian U. of Science and Technology (NTNU), Trondheim,
simulation at each production-well completion, but without ac- Norway (June 2002).
tually performing a compositional simulation. 8. PetroStream Management, Users Manual, Pera A/S, Trondheim, Nor-
5. As expected, this approach to generating the component rates is way (September 2002).
very efficient. In the case of the sector model, the black-oil- 9. Coats, K.H.: “Simulation of Gas Condensate Reservoir Performance,”
simulation run is approximately 40 times faster than the 17- JPT (October 1985) 1870.
component EOS compositional simulation. In fact, very little 10. Fevang, O., Singh, K., and Whitson, C.H.: “Guidelines for Choosing
additional run time (in the order of minutes) was required to Compositional and Black-Oil Models for Volatile Oil and Gas-
generate the component rates for all 17 components from the Condensate Reservoirs,” paper SPE 63087 presented at the 2000 SPE
3.5-million grid-cell black-oil-simulation results for a 40-year Annual Technical Conference and Exhibition, Dallas, 1–4 October.
forecast period.

C ⳱ surface oil volume conversion to equivalent gas SI Metric Conversion Factors
k ⳱ constant gas volume, 23.69024531 (m3/kmol) bbl × 1.589 873 E–01 ⳱ m3
K ⳱ constant, 23667.52637 in metric units
M ⳱ molecular weight
n ⳱ component molar rate Bassam Al-Awami is a petroleum engineering specialist in the
Reservoir Description & Simulation Dept. at Saudi Aramco in
q ⳱ surface volumetric rate
Dhahran, Saudi Arabia. e-mail:
rs ⳱ gas/liquid content He worked for 3 years in rotational assignments in petroleum
Rs ⳱ solution gas/oil ratio engineering departments within the company; he also worked
S ⳱ stream-conversion factor in reservoir management for 8 years and reservoir simulation
x ⳱ liquid mole fraction for the last 9 years, conducting reservoir-development and
performance-optimization studies. Currently, he is involved in
y ⳱ gas mole fraction fluid characterization and compositional modeling. Al-Awami
␥ ⳱ specific gravity holds a BS degree in petroleum engineering from King Fahad
U. of Petroleum & Minerals in Dhahran. He served on the 2004
Subscripts SPE ATW “Water Control/Management” Program Committee.
K. Hemanthkumar is a petroleum engineering consultant in the
g ⳱ gas phase
Exploration & Petroleum Engineering Technology Dept. at
i ⳱ component number Saudi Aramco in Dhahran. e-mail: kesavalu.hemanthkumar@
nc ⳱ number of components After 9 years in the fluid properties equations-of-
o ⳱ oil phase state development area, he has been involved in reservoir
x ⳱ either oil or gas simulator development, support, and application for the last
20 years. Before joining Saudi Aramco, he worked for Roxar,
Reservoir Simulation Research Corp., and EG&G in reservoir
simulation. Hemanthkumar holds a B.Tech. degree from the U.
The authors would like to thank Saudi Aramco management for of Madras and MS and PhD degrees from the U. of Oklahoma,
permission to publish this paper. We would like to thank Curtis all in chemical engineering. He served on the 1993 and 2005
Whitson and Faizul Hoda at Pera for introducing us to the stream- SPE Reservoir Simulation Symposium program committees.
conversion method and for the numerous discussions on this sub- Fatema Al-Awami is a petroleum engineering specialist in the
ject. We especially would like to thank Knut Uleberg at Pera for Reservoir Description & Simulation Dept. at Saudi Aramco in
helping us set up the output data-conversion process from our Dhahran. e-mail: After initial ro-
tational assignments in various petroleum engineering depart-
in-house reservoir simulator and for consultations thereafter.
ments within the company, she has focused on conducting
reservoir simulation studies for the last 17 years for field man-
agement, development, and future performance forecasts.
1. Dogru, A.H. et al.: “A Parallel Reservoir Simulator for Large-Scale Currently, she is the lead simulation engineer for the Shaybah
Reservoir Simulation,” SPEREE (February 2002) 11. field study. Al-Awami holds a BS degree in petroleum engineer-
2. Dogru, A.H. et al.: “Simulation of Super-K Behavior in Ghawar by a ing from the U. of Southern California. Mansour Mohammedali
Multi-Million Cell Parallel Simulator,” paper SPE 68066 presented at is Unit Head in the Reservoir Description & Simulation Dept. at
the 2001 SPE Middle East Oil Show, Bahrain, 17–20 March. Saudi Aramco in Dhahran. e-mail: mansour.mohammedali@
3. Pavlas, E.J. Jr.: “Fine-Scale Simulation of Complex Water Encroach- During his 22 years in the petroleum industry, he
has been involved in many reservoir engineering and simula-
ment in a Large Carbonate Reservoir in Saudi Arabia,” SPEREE (Oc-
tion studies addressing reservoir development and perfor-
tober 2002) 346. mance optimization. His main area of interest is reservoir simu-
4. Al-Garni, S. et al.: “Multimillion-Cell Simulation Study of the Khurais lation. Mohammedali holds a BS degree in petroleum engi-
Complex,” paper SPE 84081 presented at the 2003 SPE Annual Tech- neering from King Fahad U. of Petroleum & Minerals in
nical Conference and Exhibition, Denver, 5–8 October. Dhahran, Saudi Arabia.

314 August 2005 SPE Reservoir Evaluation & Engineering