Professional Documents
Culture Documents
net/publication/240738185
CITATIONS READS
13 1,843
3 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Virendra Singh on 20 December 2016.
D ue to limited availability
and quality of data, the
complexity of the physics in-
volved in predicting flow, and
uncertainties in the modeling
process, the reservoir model is
always plagued by uncertainties.
It is, therefore, important to
analyze the input data in terms
of quality, quantity, and
complexities of different scales,
review basic assumptions,
appropriateness of the modeling
workflow with reference to
static and dynamic reservoir
uncertainties, desired hardware
and software tools to reposition
the base case model with
reasonable certainty for better Figure 1. Global level reservoir modeling process workflow.
project evaluation, risk analysis,
and integrated decision making. 1) Geophysical uncertainties about migration, time-to-
Under these circumstances, it is necessary to achieve the depth conversion, picking, fault positioning, and well
right balance between short-term gains in production and ties
optimizing long-term development. We propose a workflow 2) Geologic uncertainties about in-place hydrocarbon
that achieves this goal in a cost-effective way. This paper de- volume, sedimentary depositional environments, rock
scribes the workflow, and its effectiveness is demonstrated types and their heterogeneities, spatial distribution, and
through an example. particle size
3) Petrophysical uncertainties about net reservoir thick-
Integrated reservoir modeling workflow ness, Vshale, porosity, permeability, water saturation,
The need for better decisions in field development has made and fluid contact locations
it necessary to utilize multidisciplinary teams and integrate 4) Dynamic uncertainties about absolute and relative per-
their functions/resources to maximize the economic value of meabilities, fault transmissibilities, horizontal barriers,
the project. thermodynamics, injectivity, productivity indexes, and
Figure 1 summarizes the process, shows the complex in- well skin.
terrelation of various functions commonly implemented in a
modeling effort and its iterative nature. Each process gener- There are no rules to systematically rank these uncer-
ates an output which is an input to the reservoir modeling or tainties, but an incorrectly selected or neglected uncertainty
to other parts of the workflow. may result in incorrect quantification of development-related
Implementation of many of these processes tends to be risks. Therefore, all possibilities should be considered. To ap-
iterative as simple models evolve to more advanced models as propriately assess these uncertainties, we suggest a workflow
additional data are collected during field development and as that focuses on six major areas:
more understanding of a particular model is developed. The
outputs from these processes are incorporated into a reservoir 1) Distribution of gross rock volume (GRV). It is not possible to
characterization model developed using geostatistical meth- generate perfectly sharp seismic images of the subsurface;
ods. The geologic framework forms the “static” portion of the due to the nature of the seismic method, blurring and lat-
reservoir model. Upscaling and application of flow analysis eral smearing of images occur. Therefore, it is important
provides the “dynamic” part. to estimate uncertainties related to the seismic data and
Major sources of reservoir uncertainty at the early stage categorize them by their impact on the structural frame-
of development can be broadly grouped into four main cat- work. The seismic workflow involved in the construction
egories: of a structural framework includes acquisition and prepro-
sion.
Migration places reflectors in their true
position and collapses diffractions. Therefore,
the impact in terms of the structural image is
crucial especially if reflectors are dipping. Mi-
gration uncertainty depends on the accuracy of
the velocity and on the migration algorithms.
The velocity used for time and depth migra-
tion is derived from a smoothed version of the
stacking velocity. The uncertainty associated
with the velocity should be determined prior to
and during smoothing.
Seismic interpretation includes many sub-
processes: well-to-seismic ties, detection of seis-
mic markers, picking horizons, gridding hori-
zons, fault detection, and building a complex
structural framework. These operations com-
bine seismic data and prior geologic knowl-
edge.
Time-to-depth conversion may involve a
Figure 2. Time structure map corresponding to reservoir top. vertical stretch from time to depth plus residual
corrections, or tying seismic depth to well depth.
The uncertainty related to this step is generally
very high and, as it shifts the whole reservoir up
or down, can represent 50% of the total uncer-
tainty related to gross rock volume estimation.
Since fluid contacts remain at constant depth,
a small variation in depth conversion implies a
relatively huge difference in the initial in-place
volume of hydrocarbons. The usual way to de-
termine depth uncertainties is cross validation.
In the areas of less well control, stacking veloc-
ity can be used for time-to-depth conversion.
This will have two uncertainty terms, one from
calibration to well velocities and another that is
intrinsic to stacking velocities.
3D structural model building consists of a
stack of horizons and a set of faults. In addi-
tion to this static geometric representation, to-
pological relationships exist between faults and
horizons. These relationships are part of the de-
Figure 3. Average velocity map derived from seismic stacking velocity cube. scription of the structural framework.
It is important to analyze the generated 3D
cessing; stacking; migration; interpretation; time-to-depth structural framework using basic structural geologic con-
conversion; and building a 3D structural model. cepts to understand the history of deformation, under-
Data acquisition is the most uncertain phase of the stand and interpret displacements, stress and rates, pres-
seismic workflow, and inadequate parameters may lead to sures and temperatures, impact of structural uncertainty
erroneous deliverables that permeate the entire modeling on hydrocarbon migration, and accumulation and fluid
process. Substandard preprocessing may blur or obliterate flow during production. From this 3D structural model,
the image. Incorrect statics will impact the shape and posi- GRV can be computed using top and base of the reservoir
tion of reflectors. Inadequate reflection strength will lead or hydrocarbon-water contacts.
to inaccurate rock property analysis. 2) Distribution, definition, and continuity of lithofacies. Litho-
Stacking is a most important phase of seismic process- facies may be used to define flow units within the struc-
ing because at this point the image is constructed. If an tural framework to build the reservoir architecture. Defin-
incorrect velocity is used, the energy of the reflector will be ing the reservoir architecture is a highly complex process
reduced when positioning the data. The stacking velocity that requires integration of all data (well logs, core data,
is also an input for migration and time-to-depth conver- seismic attributes, conceptual depositional model, sedi-
mentary facies types and their depositional environment, portant to include the dependency relationships between
sedimentary body orientation, and sequence stratigraphic input parameters (e.g., area versus net pay, porosity ver-
evolution). Facies distribution interpreted from well logs sus water saturation) if any exist. The estimation of HIIP
and core data has very high vertical resolution but very distribution combines reservoir geometrical, geologic and
poor lateral resolution. petrophysical uncertainties and uses reservoir geometry
Identifying reservoir stratigraphy from 3D seismic data (structural/stratigraphic framework), facies distribution,
allows quantitative facies mapping away from the well. rock types and porosity, net-to-gross, permeability and
Many seismic attributes, derived from pre- and poststack saturation). In the case of scarce data, it is necessary to
data, are used for facies distribution in the 3D geologic define the range of uncertainties to obtain the probability
model. distribution of HIIP.
3) Distribution of petrophysical properties. Petrophysical inter- 5) Distribution of recoverable reserves and statistical production
pretation identifies a consistent set of rock properties from profiles. In order to quantify the impact of dynamic uncer-
the log and core data with the objective of developing tainties on recoverable reserves and production, it is im-
flexible petrophysical models. Typical deliverables are net portant to identify the main uncertainty parameters based
reservoir thickness, shale volume, porosity, permeability, on flow simulation. The workflow for estimating dynamic
water saturation and fluid contact locations. These data uncertainties is:
are usually provided without quantitative determination
of their uncertainties. • Identifying the main uncertainty parameters such as
A wide range of probability distribution functions (pdfs) main faults sealing capacity, secondary faults sealing capac-
are used and uncertainty on their mean value is introduced ity, possible permeability barriers within reservoir, absolute
to fully cover the range of petrophysical uncertainties. In permeability, fluid type, composition and its properties (oil
the case of limited well data, it is important to capture or gas or both, gravity, GOR, GCR, viscosity, mobility ra-
the realistic range of uncertainty for each petrophysical pa- tio, initial condensate yield), water/oil and gas/oil relative
rameter. To avoid underestimation of uncertainty ranges, permeability, skin of producers and injectors, PVT, specific
well-log petrophysical properties can be calibrated. transmissibility reduction for low NTG regions, existence
When sufficient well data are available, Monte Carlo of a possible barrier layer, aquifer influx and recovery fac-
simulation can calculate petrophysical uncertainty. Monte tor
Carlo modeling is flexible and quick, allowing different • Identifying ranges of the selected uncertainties
interpretation models to be built and quick testing of the • Quantification of probability density function (PDF) as-
results. However, Monte Carlo simulation requires many sociated with each parameter
iterations for a stable and meaningful set of statistics. This • Establishing a methodology to identify the parameters
can be time consuming and needs to be evaluated for the with significant impact on well productivity and recover-
specific case. able reserves
This integrated petrophysical interpretation assesses the • Forecast surface process dependent production rates and
possible range of petrophysical parameters to select reason- production plateau duration for different envisaged mod-
able cut-off parameters for defining productive zones in els
the reservoir, to obtain probabilistic hydrocarbon-initially-
in-place (HIIP) estimates, to understand and solve reser- 6) Evaluating the impact of uncertainties on project economics
voir flow simulation problems, and to determine reservoir and risks. The production forecast obtained from reservoir
performance. simulation forms the basis for assessing project feasibility
4) Quantification of the distribution of hydrocarbon-initially- and business value, estimating the economic performance
in-place (HIIP). The hydrocarbon volume in a reservoir of various production strategies, and prioritizing them
can be expressed as: based on return on investment. The schedule of wells to be
drilled, the pressure maintenance strategy, the investment
Vhc = A*h* φ* (1-Sw) * B
schedule for the new pipelines and equipment facilities are
or a few of the important components of the forecasting sce-
nario. Additionally, economic parameters, such as costs for
Vhc = GRV* N/G* φ* (1-Sw) * B
drilling, laying new pipelines and investing in new equip-
where A = area of the reservoir filled with hydrocarbon, h ment, along with pricing forecasts for oil and gas for rev-
= reservoir thickness, GRV = gross rock volume, N/G = enue estimation, are needed for the evaluation of alternate
net to gross thickness, φ = porosity, Sw = water saturation, development strategies. The monetary value of a project
B = hydrocarbon surface transformation factor (1/FVF), can be normally determined by five factors: (1) the amount
V hc = volume of hydrocarbon, and FVF = formation vol- of hydrocarbon recoverable under a specific development
ume factor. scheme; (2) the rate at which these hydrocarbons can be
After all parameters are determined, the volumetric produced: (3) the cost of producing hydrocarbons; (4) the
equation can compute the the distribution of HIIP. To price that the hydrocarbons will fetch at the market; and
avoid overestimates or underestimates of HIIP, it is im- (5) the fiscal regime under which the hydrocarbons will
to quantify the reservoir architecture. Based on the available different possible distributions as inputs to capture their un-
information, the reservoir is assumed to be deposited in a ma- certainty range. The inferred porosity distribution from well
rine environment as beach sandbar and oriented NE-SW. logs will have significant uncertainty due to the estimation
This reservoir consists of coarse-, medium-, and fine-grain method from gamma ray, density, neutron density, and/or
sands. Shale and carbonates are considered nonreservoir fa- sonic logs. To estimate the uncertainty ranges in petrophysi-
cies. All facies were used to populate the 3D static reservoir cal parameters distribution, the porosity and water saturation
models for the four cases (Figure 6). distribution models were simulated using well-log data and
Figure 7 shows the facies modeling which uses truncated were validated from the core properties (Figure 8). The poros-
Gaussian simulation with NE-SW orientation to represent ity and saturation were distributed using sequential Gaussian
the sandbar geometry and respect the vertical proportion simulation for different facies. The permeability distribution
curves derived from wells. N/G thickness was 0 for nonreser- used an empirical permeability-porosity relation. 3D views
voir and 1 for reservoir facies. The N/G ratio was computed of porosity and permeability distribution (Figures 9 and 10)
from the different facies. The gross reservoir thickness in the showed similar trends to that of facies distribution.
wells varied from 25 to 120 m. Logs from well 4 indicate an Using 3D structural and tectonic framework, the grid was
oil-water-contact (OWC) at 2550 m TVDSS which was used constructed with 100 × 100 m cell size and a total of 600
for hydrocarbon in-place estimates. 000 cells. The rock and fluid properties were assigned to each
Based on the internal reservoir architecture, petrophysical grid block in terms of facies, saturation, permeability, poros-
modeling then assigns rock properties (e.g., porosity, oil satu- ity, and net-to-gross ratio. Two methods were used for HIIP
ration, permeability) throughout the modeling grids. These estimation:
grids have been populated constraining with different facies. Deterministic — In this method, only mean gross rock
Petrophysical property distribution largely depends on the in- volume was taken from 3D models and average values of
put distribution of various properties, so it is crucial to assign N/G ratio, porosity, saturation and formation volume factor
Figure 11.
HIIP distribu-
tion for the Case
1 and Case 2
reservoir models.
Parameters Case Case Case Case were used to compute HIIP for the four models and is given
1 2 3 4 in Table 1.
GRV (millions m ) 3
5070 2723 2872 2645
Probabilistic — In order to evaluate the impact of reser-
voir uncertainties, the original 3D static reservoir models for
Net To Gross 0,93 0,93 0,93 0,93
case 1 and case 2 were considered for simulation to estimate
OWC (m) TVDSS 2550 2550 2550 2550 HIIP through a probabilistic approach. To obtain a stable
Mean Porosity (%) 15 15 15 15 outcome from simulation, 50 iterations were made for each
Mean SW (%) 34 34 34 34 input parameter used in HIIP computation. The probabilistic
FVF 1.23 1.23 1.23 1.23 approach has provided three HIIP values namely low (P90),
OIIP (MBO) 2486 1250 1334 1228 mean (P50) and High (P10) for each case (Figure 11).
The 3D static model was upscaled into 200 × 200 m cell
RF (%) 20 20 20 20
size and 75 000 cells before dynamic simulation. The reservoir
Reserves (MBO) 497 250 267 246
simulation was performed on this upscaled 3D static model.
Table 1. HIIP estimates for different cases using deterministic ap- It was ensured that upscaling of 3D static model does not af-
proach. fect the hydrocarbon-initially-in-place volume and does not
compromise the internal heterogeneities of
Case 1 Model Case 2 Model the reservoir.
No Normal Strong No Normal Strong
To identify, assess, and quantify the im-
Field/Activity pact of dynamic uncertainties on reservoir
aquifer aquifer aquifer aquifer aquifer aquifer
performance, the workflow discussed in
OIIP 2583 2583 2583 1336 1336 1336 the previous section was adopted. A stan-
Ultimate recovery 10.5% 25.1% 31.2% 11.7% 27.3% 33.3% dard PVT analysis in the laboratory for
EUR (MBO) till the black oil case was performed. Based on
258 310 325 133 169 173
MBO till 2018 these PVT data, the following parameters
EUR (MBO) till
271 648 806 159 365 445
were used to initialize the reservoir simu-
2048 lation process: initial reservoir pressure =
Number of producers 9 9 9 6 6 6 2704 psi; bubble point pressure = initial
Max. production rate
15 000 15 000 15 000 15 000 15 000 15 000
reservoir pressure; temperature = 200°F;
per well (BOPD) viscosity of oil = 2.0 cP; oil gravity = 32
Plateau rate 90 000 90 000 90 000 60 000 60 000 60 000 API; gas gravity = 0.68; salinity = 30 000
Plateau duration ppm; FVF = 1.23 rb/stb; gas-to-oil ratio =
5 5 5 4 4 4
(years) 450 scf/stb; maximum production rate per
Production field life well = 15 000 b/d; minimum flowing bore-
40 40 40 40 40 40 hole pressure = 1000 psi; oil water contact
(years)
(OWC) = 2550 m.
Table 2. Summary of reservoir development for Case 1 and Case 2 models.
Conclusions
This study shows that systematic quanti-
fication of technical risks is crucial for the
decision-making process to compare or
Figure 14. Initial reservoir pressure and oil and gas saturation under strong aquifer support understand the risk level in each invest-
for the Case 2 simulation model.
study” by Pramanik
et al. (Geophysics,
2004). “Quantifying
petrophysical uncer-
tainties” by Adam (SPE
paper 93125, 2005),
“Experience with the
quantification of sub-
surface uncertainties”
by Charles et al. (SPE
paper 68703, 2001).
“Integrated uncertainty
assessment for proj-
ect evaluation and risk
analysis” by Corre et
al. (SPE paper 65205,
2000). “Application
of integrated reservoir
studies and probabilis-
tic techniques to esti-
mate oil volumes and
recovery, Tengiz field,
Republic of Kazakh-
stan” by Dehghani et
al. (SPE paper 102197,
2006). “Reserves un-
Figure 15. Recovery factor, recoverable reserves, water certainty: some his-
cut, production and reservoir pressure profiles for the torical trends and
Case 2 simulation model. wider implications”
by Demirmen (First
Break, 1998). “Next
generation parallel
ment decision for reservoir surveillance and smart oil field computing for large scale reservoir simulation” by Fjerstad et
development. al. (SPE paper 97358, 2005). “Probabilistic estimation proce-
An integrated approach, as adopted in this work, for iden- dures” (SPE guidelines for evaluation of reserves and resourc-
tifying, assessing, and integrating reservoir uncertainties into es, 2001). “Integrated asset managment: Work process and
a complete reservoir modeling workflow, enables determina- data flow models” by Janale et al. (SPE paper 39712, 1998).
tion of reliable volumetric estimates, recoverable reserves and “Can we sample the complete geological uncertainty space in
production forecast profiles. Numerical reservoir simulation reservoir-modeling uncertainty estimates?” by Massonnat (SPE
forms an integral part in the probabilistic forecast process Journal, 2000). “Seismic imaging of deep hydrocarbon reser-
to improve the quality of analysis involving a wide range of voirs” by O’Brien and Lerche (PAGEOPH, 1988). “Mapping
static and dynamic uncertainties. This integrated modeling reservoir saturation with seismic resolution to improve reservoir
workflow translates input parameter uncertainties into eco- model” by Reymond et al. (SPE paper 57259, 1999). “Discus-
nomic parameter uncertainties. sion on integrating monitoring data into the reservoir manage-
It is important to utilize the multidisciplinary team’s ment process” by Rossi et al. (SPE paper 65150, 2000).
knowledge, skills, and expertise as well as advanced hardware
and software to evaluate, capture, and integrate reservoir un- Acknowledgments: The authors thank RepsolYPF for providing
certainties for a better production forecast, project economic the facilities to carry out this work and permission to present and
evaluation, risk analysis, and more informed decisions. publish this material. We also thank Robert Wilson for his guid-
ance and encouragement during the execution of this work. The
Suggested reading. “Understanding the seismic resolution technical support provided by Tamrat Worku and Ivan Yemez is
and its limit for better reservoir characterization” by Singh and also gratefully acknowledged. Sincere thanks to Rebecca Latimer
Srivastava (Geohorizons, 2004). “Structural uncertainties: Deter- for her valuable comments which were very useful in improving the
mination, management, and applications” by Thore et al. (Geo- quality of the paper and to Robert Wilson and William Harmony
physics, 2002). The Seismic Velocity Model as An Interpretation of RepsolYPF for improving the manuscript’s readability.
Asset by Schultz, (SEG, 1999). “Estimation of effective poros-
ity using geostatistics and multi-attribute transforms: A case Corresponding author: vsingh@repsol.com