Professional Documents
Culture Documents
SPE 1832-'3
Summary. This paper presents a critical study of a 3D stochastic simulation of a fluvial reservoir and of the transfer of the geological
model to a reservoir simulation grid. The stochastic model is conditioned by sand-body thickness and position in wellbores. Geological
input parameters-sand-body orientation and width/thickness ratios-are often difficult to determine, and are invariably subject to in-
terpretation. Net/gross ratio (NGR) and sand-body thickness are more easily estimated. Sand-body connectedness varies, depending
on the modeling procedure; however, a sedimentary process-related model gives intermediate values for connectedness between the
values for a regular packing model and the stochastic model. The geological model is transferred to a reservoir simulation grid by
use of transmissibility multipliers and an NGR value for each block. The transfer of data smooths out much of the detailed geological
information, and the calculated recovery factors are insensitive to the continuity measured in the geological model. Hence, we propose
improvements to the interface between geological and reservoir simulation models. Input for our conditional, statistical, sedimentary,
process-related model includes regional geological parameters (rates of subsidence and aggradlltion and grossly contoured paleotopog-
raphy) and local values (depth, thickness, and direction and width of sand bodies ascertained from well-test and core analyses). In-
tegration of sedimentological understanding of the reservoir's depositional environment is emphasized.
-
Sand Content 0.80 ~ Group 1
Sand Channel' Dimensions
Sand Bod y Orientation Group 2
Reservoir Dimensions
">-
:0
~
0.60
.c
0 0.40
.t
MODEL OUTPUT
0.20
(Group 1)
1600
. 1.0
........300:1 -;:'0.5
1400
- - - - 287:1 ....
........;,........
~.","
1200 --159:1
........:,..."
E --66.6:1 ........;;,..;,.;
::;; 1000
'0 ..
~....:;;,."
~~ ......:;:;;." .,..,.,..
()
600
..
,;;'"
,;.;;",
.........
.................
..........
400
~~ 0.0
200 ~;
~" ..........
......... ,
o~~~~~~~--~~--~---.---,--.---,--,
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00
Channel Thickness (m)
AX
Fig. 3-Conditlonal probability for sand-channel width given
Fig. 1-Schematic of the geological simulation model. its thickness.
4. Sand bodies are placed randomly within the reservoir with their where Ve=the contact volume between wells, and Vsd=the total
orientation determined from a probability distribution obtained from sand volume in the simulation reservoir. The volumes are corrected
provenance studies and, to a lesser degree, from dipmeter meas- for overlapping of sand channels. Calculation of Fe demands con-
urements. siderable computer resources because of the reservoir's complex
5. The NGR for a given reservoir zone is constant and user- geometry. Continuity is closely related to the concepts of connect-
specified. edness and average fractional contact, which will be covered later.
6. Sand bodies can be conditioned to specified vertical wells within Estimation of continuity is possible and depends on the user's spec-
the reservoir boundaries. ification of the well pattern in the reservoir, the flow length be-
With this model (Le., well data, user-defined probability functions tween the injection and production wells, and the maximum number
for input parameters, and a random-number generator), many of sand channels the user considers reasonable for flow between
reservoir representations can be simulated. The procedure involves the injectors and producers. The continuity ratio is the ratio of the
the following steps. displaceable reserves to the sum ofthe depletable and displaceable
Step 1. Assign a position for the sand-channel centroid obtained reserves. An example of a calculated continuity ratio is illustrated
from the random-number generator. later.
Step 2. Choose the channel thickness from the user-given proba- The sand bodies generated consist of several connected sand
bility distribution. channels and have geometries that are too complex to be transferred
Step 3. Choose the channel width for the given thickness from directly into a numerical reservoir simulator. Thus, a rectangular
the user-given conditional-probability distribution. grid is laid over the chosen geological simulation, and parameters
Step 4. Choose the channel orientation from the user-given prob- (NGR and TMM's for neighboring gridblocks in the flow direction)
ability distribution. are calculated.
Step 5. Calculate the NGR for the reservoir. NGR is calculated as the average sand content of each individual
If the target NGR has not been obtained, a new iteration from Step gridblock. The NGR value is used with porosity to calculate PV
1 is performed. in the reservoir simulator.
As mentioned previously, the particular reservoir to which this
TMM's between neighboring gridblocks are calculated by use
modeling tool was applied required water injection to maintain
of control planes between the gridblock centers. A harmonic mean
reservoir pressure and to displace the oil. Because we needed to
of the sand content in these control planes defines each TMM. These
discriminate between the many possible reservoir realizations, we
numbers are combined with the permeabilities in the reservoir simu-
defined a continuity concept. The continuity ratio, Fe> is
lator to calculate transmissibility between gridblocks in the flow
Fe = Vel Vsd , ...................................... (1) direction.
Water Production
Continuity Breakthrough Time Recovery
Ratios (years) (years) (%)
NGR=30%
Best 1.75 5.9 37
Middle 1.50 4.8 29
6 Worst 1.00 6.3 35
NGR=45%
4
Best 2.50 6.3 49
Middle 2.50 6.7 51
Worst 1.50 6.7 48
NGR=60%
Best 3.00 7.2 56
30 40 50 60 70 80 90
Reservoir Continuity (%)
Middle 3.00 7.2 56
Worst 2.50 6.4 53
Inconsistent Input Parameters simulations were run for different NGR's (Fig. 4). Three reservoirs
The input parameters for the geologic simulation include proba- for each NGR were then simulated in a black-oil reservoir simu-
bility distributions for channel thickness, channel width as a function lator. Three continuity ratios were run for each NGR: the best, the
of channel thickness, and channel orientation. It is our experience middle, and the worst (Table 1). The ultimate recovery (at 90%
that obtaining probability distributions for the three geological pa- water cut) has little variance over the given range of continuity-
rameters is difficult. Data for channel-thickness distribution (Fig. ratio values. Even more serious, the simulated reservoirs with the
2) were derived from two independent geological studies of the same highest continuity may have the poorest recovery. With increased
core and log data from the Statfjord reservoir. NGR, the geological realizations give increasingly similar break-
Probability distributions for channel width are derived from the through, production time, and recovery (Table 1). A single run with
geological literature, the sedimentological interpretation of the sand- an 80% NGR put extreme demands on computer time. Therefore,
body geometry, and well-test analysis. The relevance of published we concluded that the continuity ratio was not a good production-
data to a specific subsurface situation is always ambiguous; i.e., potential measure for stochastically generated reservoirs. We rec-
it is difficult to know how similar analog data are to the reservoir ognized, however, that a qualitative method for differentiating the
under scrutiny. Fig. 3 shows one interpretation of all known data reservoir representations must be used when such stochastic simu-
needed to give the conditional probability of a given sand-channel lators are used.
width when the thickness is given for a specific formation. For a
thickness of 10 ft [3 m], for example, the probability of a width Sand.Body Connectedness
> 2,821 ft [> 860 m] and < 2,953 ft [ < 900 m] is 50%. An alter- Sand-body connectedness, loosely defined as a 2D attribute that
native interpretation (Group 2 in Fig. 2) of the same data was that describes the degree of sand-body overlap in a cross section of the
every sand-channel width was 60 times its thickness. This 60: 1 reservoir, is a concept that, though intuitively obvious, is difficult
value actually had a probability of 0 % in the first study. to define unequivocally. We adopted Allen's8 definition because
Most geological data focus on relationships between sand-channel
of its direct relevance to reservoir management purposes. Allen de-
thickness and width. Well-test analyses, however, give sand-body
fines the average fractional contact as the average fraction of the
dimensions and do not reveal whether they are single channels or
sand-body perimeter in contact with contiguous sand bodies. For
composite units. Sand bodies of significance in a production test
the regularly packed model described in this paper, 8 the connect-
will invariably comprise several overlapping channels.
edness, K, can be approximated as follows.
Dipmeter data are an information source that may be useful for
determining sand-body orientation. Sand bodies, however, are com- K=O for F nl /:s;0.50 ............................... (2)
posite units with an architecture that probably comprises a number
and K=2x(Fnl/-0.50) for Fnl/~0.50, ................. (3)
of channels, which in turn comprise a range of smaller-scale sand
units, such as dunes. Orientations obtained from dipmeter data may where Fnl/=the sand content of the simulated reservoir.
represent only the orientation of structures within a limited volume The stochastic model discussed here generates reservoir reali-
of the whole sand body rather than the general orientation of the zations that show that when sand-body connectedness is measured
body. Dipmeter data may thus become the subject of geological with Allen's method, it is an approximately linear function of the
interpretation. In this instance, one group interpreted the dipmeter sand content in the reservoir (Fig. 5).
data to give orientations of 30 east from north with a 60 dis- Bridge and Leeder 5 introduced a 2D sedimentary process-related
persion, while another group interpreted the same data as 60 east model that permits sand-body connectedness to be modeled with
from north with a 40 dispersion. a greater degree of geological integrity. For computational reasons,
We concluded that input parameters in the geological simulations they define interconnectedness as the ratio of the total number of
are ambiguous; therefore, simulations generated should be treated erosive contacts between channel belts to the total number of channel
circumspectly. belts encountered in vertical linear scannings of the cross section
generated. Though average fractional contact 8 and the intercon-
Simulation Examples nectedness rati05 are specific to the methods in which the reali-
In the sand-body simulations, the continuity ratio is assumed to be zations are generated, they are closely related to the sand-body
a measure of reservoir qUality. When evaluating reservoir quality connectedness defined previously. Measures of sand-body connect-
with a stochastic simulator, it is important to understand the devi- edness' average fractional contact, and the interconnectedness ratio
ations from the average reservoir generated. To check the relia- are independent of well spacing and are measures of the reservoir
bility of the continuity ratio as a discriminator, several stochastic architecture itself.
. . . . . .:>/
<f)
0.20
0.00 ~-.,--.,.-.-,..--,...--f--,----,--------,,--------,---,
.00 .10 .20 .30 .40 .50 .60 .70 .BO .90 1.00
Reservoir NGR
Layer 3 Layer 4
Fig. 5-Sand-body connectedness as calculated by different
models; the definition is different for the process-related
model.
NGR NGR
.978 .998
.652j.<:..~~'l 5 .665 5
.332
Layer 3 Layer 4
NGR
.971
.666 .677
.361 .354
0~15----~~____- ,____~
'" 10
5
Grid Blocks in X-direction Grid Blocks in X-direction
Fig. 7-Sand bodies from Fig. 6 after being transferred to the 15x5x4 simulation grid.
Flow Direction