Professional Documents
Culture Documents
SPE 17939 Procedures For Initializing Reservoir Simulation Models
SPE 17939 Procedures For Initializing Reservoir Simulation Models
SPE 17939
This paper was prepared for presentation at the SPE Middle East Oil Technical Conference and Exhibition held in Manama, Bahrain, 11-14 March, 1989 ..
This paper was selected for presentation by an SPE Program Committee following review of information contained in an abstract 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. Permission to copy is restricted to an abstract of not more than 300 words. Illustrations may not be copied. The abstract should contain conspicuous acknowledgment
of where and by whom the paper is presented. Write Publications Manager, SPE, P.O. Box 833836, Richardson, TX 75083-3836. Telex, 730989 SPEDAL.
875
2 PROCEDURES ~R INITIALIZING RESERVOIR SIMULATION MODELS SPE 17939
AREAL GRID
used to compute the property arrays calculated
The selection of the areal grid and at grid block centers.
orientation follow established principles4.
To name a few, these are aimed at representing Typical mapping problems include the
the flow geometry, directional rock properties non-representativeness of some of the core and
or discontinuities, and defining the locations log data. For example, the log-derived
at which saturations and pressures are to be porosity (and saturation) across tight thin
calculated. Of specific significance in intervals is largely dominated by shoulder
applying these principles to large reservoirs effects. In addition, the selection of core
is that due to the large number of wells, the plugs for core analysis is usually made in
resulting areal grid has to be large in order 'clean' or homogeneous sections that overlooks
to maintain detail between neighbouring wells both the highly permeable streaks and the
and in areas of special interest, such as tight or stylolitic intervals. such problems
around fluid contacts and in areally thin are tackled by extensive screening and
transition zones, etc. Accordingly, ext.ra scrutinization of the data during vertical
effort is needed to minimize the size of the layering.
areal grid in a way that does not compromise·
on the main objectives of the study. The horizontal and vertical permeability in
the porous intervals, and the horizontal
VERTICAL GRID permeability in the tight intervals are taken
form core data, whenever available. In most
As shown in Figure-3, the high contrast in cases, the vertical permeability in the porous
rock properties, such as horizontal intervals is roughly _equal to horizontal
permeability, is not always reflected on permeability. However, and in addition to the
porosity. Accordingly, and based on problems in core plug selection that are
established geological models, fine layering mentioned above, the core data of the tight
simulation models are constructed to represent and/or stylolitic intervals is mostly not
such contrast taking into account all applicable, as values of less than 0.01 md are
available data on rock type, horizontal and not possible to measure by conventional
vertical permeability, and fluid saturations. methods. Special whole core analysis across
In addition, thin tight and/or stylolitic these intervals, on the other hand, has shown
intervals exist in the reservoirs which may kv values of as low as 1. 0 E-4 md in the
act as barriers to vertical flow. These strongly developed stylolitic zones. Based on
intervals are mostly continuous but have a such data, and using core-log transforms, a
wide range of vertical permeability between method has been established to derive the
1.0 E-4 md, to about 10 md, as measured by vertical permeability from the porosity
special whole core analysis5. As a result, profile across such intervals5.
these layers are modelled as separate layers
in order to avoid apriori assumptions on their Due to the presence of in-situ vugs and
ability to form total or partial barriers to hairline fractures, it is worthy to note that
vertical flow. In the case that these layers more often than not, the test-derived
are known to be discontinuous, and in absence horizontal permeability is a few folds higher
of a layer vanishing option in most than the core-derived permeability, and that
simulators, equivalent porous intervals are simple correlations between the two sets of
used to extend these layers throughout the data are not possible. This problem is
model. The modelling of the tight layers as compounded in presence of partial or selective
such is necessary in order to avoid lwuping well completions that are based on the
their data with the adjacent porous layers stratified nature of the .reservoirs.
that have much better rock and fluid
characteristics. In addition, the presence of Accordingly, the initial core-based
these layers, specially when exhibiting permeability maps are usually upgraded by
variable rock properties, allows for a more results of single well models that use the
realistic simulation of the flood front same full field layering, and through which
advance, as vertical cross flow between the transient pressure tests are matched.
adjacent porous layers will depend on the Alternatively, if enough test-derived data are
physical · properties of the dense layers available for the total reservoir, or
themselves, such as mobility, capillary sub-reservoir, the horizontal test
pressure and absolute pressure6. permeability maps are first converted into
absolute maps using ko/ka correlations from
ROCK PROPERTIES SCAL data, after which new test-based maps per
layer are generated using the vertical
All rock properties, such as structure tops, distribution of the core maps. The generated
iso-pachs, porosity, horizontal and vertical horizontal and vertical permeability maps are
permeability are mapped using contouring again checked and corrected by single well
packages. The generated surfaces are then models.
876
SPE 17939 A. Y. FAWAKHIRI, I. M. HASSAN, A. F. ABED 3
As for porosity, the severity of the log the transition zone limits in each model
shoulder effects in the tight i.ntervals is layer. In this regard, it should be noted
evaluated by comparing the averaged porosity that derivation of the transition zone from
across the whole interval with the minimum log-derived saturations is more superior than
recorded porosity. If high discrepancies are that derived from SCAL data due to the usual
noted, indicating significant shoulder high scatter and even some inconsistency in
effects, then the log-derived porosity maps the SCAL data.
are generated using minimum values only.
FLUIDS SATURATIONS
FLUIDS CONTACTS
The fluids saturations are mapped using
Due to the heterogeneous rock properties, the contouring packages following the definition
observed fluids contacts (gas-oil or of the common free fluids levels for the total
oil-wate:r) typically show vertical variations reservoir, and a consistent set of log-derived
that conform to surfaces of contacts rather apparent fluid contacts for each layer. A
than to planar .contacts. such a phenomenon is common encountered problem is the computed
caused by vartations i.n the capillary saturation in the tight or stylolitic
pressures as a result of the heterogeneous intervals, which is affected by shoulder
rock properties. Accordingly, the free fluids effects. If such intervals are strongly
contacts (free gas and water levels) are first developed, the computed water saturation is
defined for the reservoir using the discovery usually close to 100\, which is in line with
pressure profiles ·or regional data that are the expected high capillary pressure resulting
verified by PVT. results, as preRented in ·from rock tightness. In such a case, and if
Figure-4. The intersection points of the this phenomenon is consistent with the
gas-oil and oil-water lines define such free permeability and porosity maps, the layer, or
levels, at which the capillary pressure is portions thereof, could be keyed-out directly
zero by.definition. Any rise of fluids above as non-reservoir rock. In the more difficult
such levels, which is exhibited.as an apparent case where this could not be justified, the
contact on the logs, is then related to maximum water saturation, rather than the
capillary threshold pressure, as outlined computed average saturation in each layer is
below. mapped.
877
4 PROCEDURES FOR INITIALIZING RESERVOIR SIMULATION MODELS SPE 11939
The above findings are to be expected within The range of the laboratory-measured
upper and lower limits of rock properties. In k4/¢, may not cover the actual range of
all, they indicate that improved recovery is the input arrays to the simulator, or
obtained with improved rock properties following modification of permeability
exhibited in the core plugs as absolute during history matching. Hence, man-made
permeability and porosity. upper and lower limit curves are added to
the family that conform with the actual
CONSTRUCTION OF BND-PODIT CORRBLATIONS AND data trends.
FMILY CURVBS
Similarly, the drainage threshold
Based on the above, and .following quality capillary pressure is correlated with a
control and screening of the laboratoy data, a function of porosity and permeability,
set of SCAL end-points correlations and family e.g. ¢ v k, as presented in Figure-11.
curves is constructed for each rock type as As in the relative permeability
follows: end-points, the data is smoothed by
regression and upper and lower limits are
The .laboratory-measured end points, swc, imposed.
Sgc, Srow and Srog are correlated against
a function of porosity and permeability Drainage and imbibition capillary
of the core plugs, e.g. ¢ v- k, as pressure family curves, presented in
~ presented for an oil-water system in Figures-12 and 13, are constructed using
Figures-5 and 6. The scatter which may a similar technique to that of the
still persist in the data is smoothed by relative permeability. The ordinate of
a regeression line, ensuring that the sum the drainage capillary pressure, however,
of the end points and the mobile phase excludes the threshold pressure section
saturation is unity. Depending on the of the original curves, as this is
range of the data, upper and loWer limits obtained separately by Figure-11.
of the generated line may also be imposed.
The constucted SCAL end-point correlations and
The increasing trend of relative family curves for each rock type represent an
permeability with improved rock absolute data base that is used for generating
properties is checked by plotting the the saturation tables for the simulator at any
relative permeability results of each stage in the study.
plug against a function of porosity and
permeability, e.g., k4/~ at all GENERATION OF SATURATION TABLES
normalized fluid saturations, as
presented in Figures-? and 8. The use of The saturation tables, comprising of the
k4!9J in such plots is made in order to relative permeability and capillary pressure
spread the data of the plugs for curves, are generated by an in-house method
subsequent interpolation. At this stage, that uses the family curves and end-point
further screening of the data is made by correlations, together with the established
excluding the· plugs of a given rock type rock and fluid arrays (cell center depths,
whose data do not conform with the porosity, permeability, tops of transition
general increasing trends. zones, apparent fluid contacts, free water
level, and the log-derived fluid
A set of relative permeability family of saturations). In the process, the data of the
curves pertaining to a given rock type is derived tables are made to match the observed
then constructed, as presented for an log-derived data, such as tops of transition
oil-water system in Figures-9 and 10. zones and apparent fluid contacts. In fact,
Each curve in these plots represents the the method can generate a saturation table for
data of one core plug plotted in terms of each grid block. Rather than that, however,
k4/~. The abscissa is the normalized +-he saturation tables are generated for groups
water saturation. The ordinate, on the of grid blocks falling within user-defined
other hand, is the product (k4/~) (krow) ranges of k4/¢.
for the oil relative permeability family,
and (k4/¢) (krwo) for the water relative Taking an oil/water system as an example, and
permeability family. In each family, the working on given a group of grid blocks with
curves of the different plugs are an average k4/¢, the z-axis of the drainage
878
SPE 11939 . A. Y. FAWAKHIRI, I. f'l. HASSAN, A. F. ABED 5
As in the saturation tables, the in-house Maps of the input rock data, such as
method used for generating the PVT tables thickness, porosity, horizontal and
relies on the concept of generating an vertical permeability are produced by the
absolute data base from which the input PVT simulator, and are checked for
tables are derived. such a data base includes consistency.
all the differential expansion PVT properties
in addition to differential-to-flash relations.
879
6 PROCEDURES FOR INITIALIZING RESERVOIR SIMULATION MODELS SPE 11939
The initial calculated reservoir pressure Bo(a) Oil formation volume factor at
and bubble point pressure profiles are atmospheric pressure and reservoir
plotted versus depth. This plot is used temperature (vol/vol)
to check the location of the gas-oil Bo(bp)= Oil formation volume factor at bubble
contact and free .water level, and the point pressure
calculated fluids gradients for k Absolute permeability (mD)
comparison with field measured data. krow Relative permeability to oil, o/w
system
Th~ simulator-generated initial fluid krwo Relative permeability to water, o/w
saturation maps are produced and compared system
with the log-derived maps for p Pressure (psia)
verification. Input mapping problems P(bp) Bubble point pressure (psia)
and/or initialization data problems could PCI Imbibition capillary pressure, o/w
be identified by such comparisons.Using system (psi)
the calculated initial pressure arrays, PCD Drainage capillary pressure, o/w
the model is dynamically re-initialized system (psi)
to the log-derived saturation maps, and PTH Drainage capillary threshold pressure,
the resulting initial fluids-in-place per o/w system (psi)
model layer and for the whole model are Porosity (fraction)
compared with those obtained by the sw Water saturation (fraction)
equilibrium initialization. swc Connate water saturation (fraction)
Sgc Critical gas saturation (fraction)
Using the equilibrium initialization, the swn Normalized water saturation (fraction)
moveable hydrocarbons volume in each Srow Residual oil saturation (fraction)
model layer and in the whole model are Srog Residual oil saturation (fraction)
produced for checking.
REFERENCES
A five-year zero rate simulation run is
performed to check the satability of the 1. R. A. Nelson: "Significance of Fracture
model with respect to the initial fluids Sets Associated with Stylolite Zones,"
contents, saturations and pressures The American Association of Petroleum
against time. Geologists Bulletin, Volume 65 No.ll
(November 1981), 2411-25.
Once satisfactory results are obtained, the
model initialization is concluded and the 2. Won c. Park and Erick H. Schot:
history matching phase of the study is started. "Stylolites: Their Nature and Origin,"
Journal of Sedimentary Petrology Volume
CONCLUSIONS 38 No.1, 115-91.
Both detail and consistency in the input data 3. Abu Dhabi company for onshore Oil
of large heterogeneous simulation · models are Operations: "A Critical Review of
necessary in order to ensure proper Stylolites: Their Formation and
representation of the complex flow regimes in Significance to Abu Dhabi Reservoirs,"
such reservoirs. Our experience in using Stylolites and Associated Phenomena
simplifying assumptions (such as lumping of Relevance to Hydrocarbon Reservoirs, Abu
layers, pseudo-functions, etc) has shown that Dhabi National Reservoir Research
although numerically satisfactory simulation 'Foundation (ADREF), ADREF, Abu Dhabi
results could sometimes be obtained, the (1984), 219-90.
resulting models fail to achieve the level of
detail that is required to optimize the 4. Staggs, H.M. and Herbeck, E. F.:
performance of such reservoirs. "Reservoir simulation Models An
Engineering overview," JPT (December
The resulting ·detailed simulation models· are 1911), 1428-36.
naturally large and costly. The presented
initialization concepts and procedures for 5. Mitri, K. T., Abed, A. F., "A
such models have been designed to maintain Semi-Empirical Lumped Parameter Model for
consistency of the input data with physical Estimation of Vertical Permeability and
principles, without compromising on the Threshold Pressure of Dense Stylolitic
required detail. The high cost of running such Limestone Intervals," paper SPE 11941
models is a natural consequence of the large presented at the 1988 Middle East Oil
model sizes and the required detail. However, Show, Bahrain, March 11-14.
these costs are continuously decreasing with
the current improvements in the speed and 6. Martin, J. H., Cooper, J. A.: "An
options of reservoir simulators, and the Integrated Approach to the Modelling of
significant advances in computer speeds. Permeability Barrier Distribution in a
Sedimentologically Complex Reservoir,"
NOMENCLATURE paper SPE 13051 presented at the 1984
59th Annual Technical and Exhibition of
SPE, Houston, September 16-19.
Bo Oil formation volume factor (vol/vol)
880
seE !1939
~
,.,
r:f ~~
•
•
•
8450
• 0 • •
,., • • • • •
Fig. 1-Structure contour map on top, Reservoir A. Fig. 2-Petrophyslcal characteristics and layering model for the top third of Reservoir B.
DEPTH
(FT)
7800
8000
!="·
~ 8400
7850 l:
1-
c.
w
c
8800
7900
WATER (0.49 PSI/FT)..,
9200+----------.--------~~----~~
4000 4200 4400 4600
PRESSURE (PSIA)
Fig. 3-Fiood front advance In a porous Interval underlain by a stylolitic Interval In Fig. 4-lnltlal pressure profile, Reservoir D.
Reservoir C.
881
',~o
,). f_ t·,
·~
'
·~ j i ~t •. ;;·~'
!·~
...
1 1
10,--.,------------------~ 10.-------~----------------------~
•• ..
0
10 • 0
10
•
•
•
• •
10 +------r'~----.-----.------,----__,
1 -1
10+------.-----.-L----.-----.-----~
0.00 0.20 0.40 0.60 0.80 1.00 0.00 0.20 0.40 0.60 0.80 1.00
Swc Srow
Fig. 5-Correletlon ol connate water saturation with q,../k. Fig. &-Correlation ol residual oil saturation with q,..[k.
10 ~~~~~--~----~--~-------------------,
8
10 ,-------~~--------------~------~---.~----,
8
-4 -4
10 ~~----.-------.-------.-------.-----~ 10 +4~._--~------~--._----~--~--~------~
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
KRWO KROW
Fig. 7-Correlatlon ol relative permeability to water with k 4 {q,, Fig. a-Correlation ol relative permeability to oil with k 4 /¢.
882
:(1 '-~
:i .1: l·.'
8 8 4
10 8 10 K 1¢= 1 X
4
K tc1>= 1 X 10
6
6 1.9 X 10
1.9 X 10
4 4
10 10
-e.
-v' -e.
~ -v'~
0*
3: 3:*
c: 10° 0 10°
~ 2.2 X 100 c:
~
-2
6.2 X 10
10- 4 10-4
-4
1.0 X 10
10 +--------.--------.---------.-------~------~
8
10 +--------.--------.--------.--------.-------~
8
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
SWN SWN
Fig. 9-Water relative permeability family curves as function of k 4 /t/>. Fig. 10-011 relative permeability family curves as function of k 4 /t/>.
1~~------------------------------~
• •
•
10 +--------.--------.---------.--------.--------~
8
Flg._11-Correlatlon of drainage oil/water capillary threshold, pressure with q,Jk. Fig. 12-Dralnage oil/water capillary pressure family curves as function of k 4 /t/J.
883
SPE 179J9
10~
8.0 X 103
4
10
'P 80 Ill
!:. ~
::z:: ~
1- 8800 60
a. ~
UJ a.
Q
40
9000 20
8 FWL._.
10 +-------~--------~------~--------~------~ 9100 0
0 0.2 0.4 0.6 0.8 1.0 0.0 0.20 0.40 0.60 0.80 1.00
SWN Sw
Fig. 13-lmblbltlon oil/water capillary pressure family curves as function of k 4 /<f>, Fig. 14-Correctlon of the acale·derlved drainage capillary pressure profile to match the log·
derived profile.
8200
INITIAL GOC
__\,
1.00
~
co
INITIAL
'P 8400 it-RESERVOIR 0
!:. PRESSURE co 0.99
::z:: .......
I- 0
ll. co
UJ
Q
LL. 0.98
>
LL.
-1
8600 0
• Q
0.97
UJ
!:::!
-1
<(
::i!! 0.96
a::
0
z
8800+-------------~------------.-----------~
2000 3000 4000 5000 0.95+-----~------~-----r------r-----~
0.0 2.0 4.0 6.0 8.0 10.0
PRESSURE (PSIA) NORMALIZED PRESSURE ( P I P(BP) )
Fig. 15-Varlallon of bubblepolnt pressure with depth, Reaerv(llr E. Fig. 16-0IIformatlon volume factor, above bubblepolnt pressure, as function of bubblepolnt pressure.
884
EE 17939
1.20 1.20
~
0
10
I
~
10
~ 1.00 ena: 1.00
10
0
e (/)
' 0.80 5
~ 0
0
10
~
a:
I
0 0.60 .J 0.60
e 0
(/)
Ll.
>
Ll.
<
(!)
.J 0.40 c 0.40
0 w
N
c ::::i
w <
N :E
::::i 0.20 a: 0.20
< 0
:E
a: z
0
z 0.00 0.00~--~~---T-----r----~----r---~
0.00 0.20 0.40 0.60 0.80 1.00 1.20 0.00 0.20 0.40 0.60 0.80 1.00 1.20
NORMALIZED PRESSURE ( P/P(BP) ) NORMALIZED PRESSURE ( P/P(BP) )
Fig. 17-011 formation volume factor, below bubblepolnt pressure, as function of bubblepolnt presaure. Fig. 18-Gas/oll ratio as function of bubblepolnt pressure.
885