Vous êtes sur la page 1sur 16

SPE-180383-MS

Development and Application of Probabilistic Fluid Property PVT Models


for Stochastic Black Oil Reservoir Simulation
Hussein Alboudwarej and Jonathan M. Sheffield, Chevron Energy Technology Company, Richmond, CA

Copyright 2016, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Western Regional Meeting held in Anchorage, Alaska, USA, 23–26 May 2016.

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 have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
Fluid properties are often one of the key uncertainties in reservoir simulation studies. A workflow and
guidelines are presented to develop an overall probabilistic forecasting framework, through which likely
fluid models are validated, uncertainty ranges are estimated, and probabilistic PVT models are generated
for use in stochastic black oil reservoir simulation modeling.
Workflows are provided for two distinct scenarios. For scenarios with a large number of samples from
the reservoir or a representative analog fluid property database, standard statistical methodologies are
applied to generate probabilistic fluid models. However, in most cases where there are only limited data
available, a unique approach is adopted to generate a representative pseudo-database for uncertainty
assessment workflows. The pseudo-database is generated based on fluid property correlations that are
screened for the conditions of the specific fluid system. The pseudo-database is used to both validate a
reliable likely case and to determine ranges to generate probabilistic models. Numerical formulations of
probabilistic models are provided.
Application of the workflow is demonstrated for three different cases of fluid property data. For a case
based on a large set of PVT measurements, standard deviation of specific fluid properties (FVF, GOR,
Viscosity, and Compressibility) is calculated and used in standard statistical methods to calculate P10/P90
cases. In another example, the validity of fluid viscosity data based on an assumed viscosity correlation
was investigated, suggesting that fluid viscosity was actually closer to a P20 model as opposed to a P50
(likely case). In another case study, where only limited PVT data were available, a number of screened/
filtered fluid property correlations are used to represent available PVT data, and then the input parameters
are perturbed to generate a pseudo-database to capture fluid property ranges as well as variation in a field.
Currently, there is a lack of a systematic way to determine uncertainty ranges for fluid properties, as
well as validating the most likely case (best case or mid case). The novelty of our framework is in the
provision of a consistent workflow to generate probabilistic fluid models that are not based on experiential
methods employed in the past.
Introduction
Probabilistic reservoir simulation studies are often used to assess impact of various geological and
dynamic engineering parameters on reservoir processing rates, different aspect of oil and gas production
2 SPE-180383-MS

from hydrocarbon reservoirs, and/or reservoir development alternatives (Walstrom et al., 1967; Steagall
and Schiozer, 2001). Most of the input parameters into simulation models are subject to uncertainty,
particularly rock and fluid properties which are often key uncertainties involved in such studies (Zabel et
al., 2010). Variability and large uncertainty in PVT measurements is a big hitter in some of the studies.
However, there is a lack of a systematic way to determine uncertainty ranges, as well as validate the most
likely case (best case or mid case) of fluid properties. In most cases, there are only a limited number of
fluid samples to represent the fluid system for the entire field, as well as the challenge that the existing
PVT data for various analog fields are not readily accessible or reliable. In addition to probabilistic
reservoir simulation studies, there is also an increasing demand for analog data for fast track and business
development projects, and/or identification of outliers in existent measurements.
In this context, it is important to distinguish the natural variation of fluid properties in a field due to
geological settings versus the uncertainty in a specific fluid property that is used for fluid system
representation. Geological phenomena such as progressive source rock maturation, presence of multiple
source rocks, biodegradation (particularly non-uniform biodegradation) across the field, and/or gas
leakage from the reservoir (particularly local leakage) will affect fluid property variation in a field.
Additionally, fluid mixing within in the reservoir as part of the maturation process, and filling history of
the reservoir, together with variation of pressure and temperature in the reservoir affect homogeneity of
fluid properties across the vertical and lateral expanse of the reservoir. A good example is variation of
viscosity in a heavy oil reservoir due to degree of biodegradation (Seccombe, 2005) and variation in
temperature in both vertical and lateral directions (Sibley, 1970).
Another uncertainty interjecting into the overall assessment of fluid properties is that of fluid sampling
and sample transfer. Integrity of the sample could be compromised if proper sampling or sample transfer
protocols are not followed (Bon et al., 2007). If this occurs, samples will not be representative of the
reservoir from which they were retrieved. For example, excessive pressure draw down causing two phase
flow of hydrocarbon gas and liquid, collection of samples contaminated with oil-base mud, or loss of H2S/
CO2 to collection container hardware during downhole sampling, or erroneous production GORs for
recombination of separator oil and gas samples will lead to non-representative samples.
Errors associated with measurements also contribute to uncertainty of PVT data (Nagarajan et al.,
2007). Accuracy of measurements is highly dependent on; i) Experimental procedure, ii) Suitability of
experimental apparatus, iii) Proper set up of experimental apparatus, and iv) Measurement being done by
experienced personnel with track record of delivering high quality data among other factors. A PVT
specialist is able to check for the consistency of the PVT experimental data and assess its quality using
available tools such as EoS modeling, fluid property correlations, multiple measurements and or even an
analog database.
In an ideal situation, there exist a large number of high quality data points for uncertainty assessment
and statistical analysis. These data are generated based on a large number of samples that properly
represent the fluid system in the reservoir. In addition to the fluid property data, complementary data such
as geological, geophysical and geochemical data will also help assess fluid variation or potentially
existence of multiple PVT regions in a reservoir/field. Such a database can be used to quickly evaluate
quality of experimental data of similar fluids, or in case of no sample, a representative analog sample may
be used to perform some preliminary analysis. Despite the high cost of sample collection and PVT
analyses, fluid property data are not consistently collected and properly maintained. In some cases,
measured fluid properties are not properly quality controlled (QCed) before use.
The focus of this paper is to present a workflow and guidelines to consistently generate reliable ranges
and thus probabilistic fluid property black oil models for use in reservoir simulators.
SPE-180383-MS 3

Uncertainty Assessment Workflow


Figure 1 shows the overall fluid property uncertainty assessment workflow. The main elements of the
workflow include:
● Assessment of the amount and quality of available fluid properties data in the short term. Ideally
a statistically significant database specific to a fluid system is available. This data QC and integrity
evaluation is a critical part of workflow.

Figure 1—Uncertainty assessment workflow

● In case of no or limited quality data, access to an analog database is the underlying assumption
● Development of a database of quality PVT data in the long term is required for reliable uncertainty
assessment.
It is clear that in case of a large of number of QCed PVT data, standard statistical analysis techniques
are used to generate uncertainty ranges. However, the main challenge is in cases where there is no or very
limited fluid property data.
Correlations have been used to represent a wide range of fluid properties. These correlations are based
on diverse fluid property databases developed either geographically or based on the characteristics of the
fluids (light, medium, heavy, gas condensate, etc.). In general, fluid property correlations only need
limited characterization properties to estimate PVT properties. Fluid property correlations are only valid
over a specific range of variables (e.g. pressure and temperature).
Since a fluid property correlation is developed using an existing PVT database, the reverse process may
be used to generate a ⬙pseudo⬙-database of fluid property data that can be used for statistical analyses
(Figure 2). A factorial combination of relevant and valid correlations may be used to generate such a
database for statistical analysis. With this approach, limited information from an existing analog may be
used as an input to fluid property correlations. Figure 3 shows a graphical representation of full factorial
generation of a pseudo-database. An example is given to demonstrate a viscosity pseudo-database, where
4 SPE-180383-MS

a full factorial combination of dead oil viscosity (Vod), gas oil ratio (GOR or Rs), and saturated oil
viscosities (Vos) are sequentially used to generate under-saturated viscosity (Vou) data points. Again, the
basic input data such as pressure, temperature, API gravity of oil, gas gravity, etc., are given based on a
limited number of available samples or may be provided by assuming an analog database. The latter
approach in general may be used to determine possible uncertainty ranges for exploration applications.
The other application of such a ⬙pseudo⬙ database is to QC any available experimental data. This becomes
a very valuable tool particularly in case of viscosity data where available models need tuning data before
they can be used for prediction even for models which need extended fluid characterization input values
such as composition.

Figure 2—Idea of Fluid Property Pseudo-Database generation

Figure 3—Schematic representation of pseudo-database generation


SPE-180383-MS 5

The general workflow to generate a pseudo-database for uncertainty analysis has 4 main steps:
1. Input basic fluid data and generate preliminary pseudo-database
a. Provide basic fluid properties such as API gravity of oil, gas gravity, and pressure and
temperature conditions. This information may come from one analyzed sample or could be
extracted from an assumed analog database in cases where no actual sample is available (e.g.
exploration).
b. Generate a preliminary pseudo-database with a full factorial combination using valid fluid
property correlations. Validity range of fluid property correlations should be determined based
on the input parameters.
2. Screen for representative models
a. Select representative correlations to include in uncertainty analysis. In cases where limited
experimental data is available, valid correlations representing experimental data within typical
measurement error should be selected. Table 1 shows typical range of experimental error for
various fluid properties. For example, GOR correlations that are within (10% of experimental
data should be used for next step. In cases of no experimental data, all valid correlations may
be used, and thus a larger uncertainty range is expected.

Table 1—Typical measurement error for PVT properties.


Uncertainty in PVT
Fluid Property Measurement (%)

GOR 3-10
BO 1-4
␮O 10-30
CGR 5-15
BG 1-4
␮G 10-30

3. Perturb basic fluid input and regenerate pseudo-database


a. Perturbation of basic input properties comes from a knowledge of typical errors for measure-
ment of such properties (e.g. 1-3% in API), or from an assumption of variation in the reservoir.
For example, an assumed variation in API in a reservoir based on an analog field.
b. The pseudo-database will be regenerated using only selected correlations.
4. Standard statistical analysis and particular standard deviation and the number of data points used
in the analysis are used to determine uncertainty ranges for key pressure points (reservoir pressure,
saturation pressure).
a. Based on the regenerated pseudo-database, either only uncertainty ranges are determined or the
mean together with associated uncertainty ranges are calculated.
A graphical representation of the entire process of pseudo-database generation and use to generate
probabilistic models is shown in Figure 4, while Figure 5 shows an example of a viscosity database where
both experimental data and data points using correlations are statistically analyzed. TIBCO™ Spotfire®
was used for data analysis. The same pseudo-database may also be used for QC of experimental data
particularly in the case of viscosity where no predictive model is available.
6 SPE-180383-MS

Figure 4 —Detailed workflow for PVT uncertainty assessment and determination of uncertainty ranges

Figure 5—An example of viscosity pseudo-database and statistical analysis


SPE-180383-MS 7

Probabilistic Fluid Models


Black Oil fluid model is the most common approach to define fluid properties for use in reservoir
simulators. In this approach fluid properties for separate oil and gas phases are provided in a tabular form.
The tabular data are presented in two formats:
1. Saturated oil and gas property tables where properties of a saturated system are represented as a
function of pressure. In such tables, under-saturated properties are calculated mathematically using
the slope of change with pressure. For example, oil formation volume factor and viscosity at
under-saturated condition is calculated using an exponential function for formation volume factor
or a semi-log extension for viscosity.
2. Saturated and under-saturated properties are tabulated for the fluid system with changing com-
position (GOR), as a function of pressure.
In either of the black oil formulation representations (1 or 2 above), fluid property data for the mid case
will be used to develop P10 and P90 black oil fluid models using calculated uncertainty. Again, the
uncertainty ranges are calculated based on a statistically significant existing database, or using a
pseudo-database generated using fluid property correlations. Assuming a normal cumulative distribution
function, 10th and 90th percentile values are calculated using (Taylor, 1997):
(Equation 1)

where SD stands for standard deviation of the particular fluid property. Coefficient 1.282 corresponds
to an infinite number of data point assuming a normal distribution. If a relatively smaller number of data
points are available, student’s t-distribution probability density function may be used. Figure 6 shows the
Impact of sample size on cumulative probability distribution function for student’s t-distribution at 90th
percentile compared to normal distribution. In this case, the sample size is represented by a parameter
called degree of freedom which is defined as the number of sample points minus one. The graph simply
shows that if the degree of freedom (or # of data points) is more than 10-15, a normal distribution of
probability density function may be used for statistical analysis. Otherwise, the coefficient in equation 1
will need to be adjusted based on a student’s t-distribution. For example, for a case where the number of
data points is 7 (corresponding to a degree of freedom of 6), the coefficient in equation 1 will be 1.44.
Table 2 shows the standard deviation multiplier value in equation 1 as a function of number of data points.
Notice that for a very large number of data points, student’s t-distribution function will approach that of
a normal distribution function.
8 SPE-180383-MS

Figure 6 —Impact of sample size on probability distribution function for student’s t-distribution

Table 2—Student’s t cumulative distribution probability values


for 90th percentile.
Degree of Freedom (n-1) 90 % Probability

1 3.078
2 1.886
3 1.638
4 1.533
5 1.476
6 1.440
8 1.397
10 1.372
15 1.341
20 1.325
30 1.310
50 1.299
100 1.290
Infinity 1.282

To generate P10 and P90 black oil tables, values in the formation volume factor, GOR (gas oil ratio)
and the viscosity columns of the P50 (mid) case are modified using Equation 1. Since multiple entry
pressures (saturation points) are given in the black oil tables, generally standard deviation corresponding
to the original fluid (P50 or mid case fluid) saturation pressure is used. Depending on the slope of fluid
property variation with pressure in the black oil table, and the magnitude of standard deviation of the
particular property, P10 and/or P90 values may cross the P50 profile at some pressure points. To avoid
such a situation, P10 or P90 properties are calculated by shifting the P50 values by the ratio of the same
properties between P10/90 and P50 values at the P10/90 saturation pressure. This approach ensures fluid
models are consistent over the entire range of pressure. Equation 2 is a demonstration (example) of such
shift in the calculated P10/P90 values:
SPE-180383-MS 9

(Equation 2)

Similar P90/P10 equations for other fluid properties are given in the appendix.
To provide a consistent black oil table, statistical values for viscosity values are reversed as a fluid
system with a higher GOR and FVF will have a lower viscosity or vice versa. This means that the
statistical P90 viscosity value corresponds to a P10 fluid model (pessimistic) and the statistical P10
viscosity values corresponds to the P90 model. Viscosity is the only property that behaves as described.
For other fluid properties in the black oil table, statistical and fluid model values are similar. The slope
of change of formation volume factor and viscosity (commonly referred to as compressibility and
viscosity compressibility respectively) with pressure may be left as P50 case, since the behavior of the
fluid systems is not expected to change significantly even with perturbation due to uncertainty.

Case Studies on Uncertainty Assessment of Fluid Properties and


Application of Workflow
PVT data uncertainty assessment workflow and guidelines has been successfully applied to many
exploration and field development cases in the last few years and has become part of standard practice
when developing probabilistic fluid models. The following case studies demonstrate the application of the
workflow to determine and use uncertainty ranges for reservoir simulation and fluid property assessment.

A Reservoir Fluid System with a Large of Number of Samples


In rare cases where a statistically significant number of samples are available from a reservoir, QCed
database could be used for standard statistical analyses to determine standard deviation for a specific fluid
property. Such cases exist when the reservoir is generally large and delineation is required to better
understand fluid variation in the reservoir, or access to representative reservoir fluid samples (e.g.
wellhead single phase fluids) are readily available. The underlying assumption is that a representative mid
case or P50 case is selected and standard procedures (EoS) are used to generate P50 black oil tables.
Figure 7 shows an example of such database for a reservoir fluid system with more than 30 QCed PVT
data points. A number of samples were collected during exploration period, while majority of the samples
were collected during development and production for the reservoir. In this Figure, the percentage
variation of P10 and P90 values relative to P50 value i.e. (1.282 ⫻ SD / P50) ⫻ 100 for select fluid
properties are shown. For this particular reservoir fluid system, variation in fluid properties is in the same
order of magnitude as typical PVT measurement errors indicating a relatively uniform fluid system across
the reservoir. Also due to a large number of sample points, the difference between a normal distribution
for probability density function (infinite number of sample points) and that for Student’s t- distribution is
negligible.
10 SPE-180383-MS

Figure 7—Percentage variation of P10/P90 values relative to P50 values for a large PVT database (All samples from the same fluid
system).

Uncertainty Assessment of Limited Fluid Property Data


In case of exploration of a new prospect, very limited or no actual sample may be available. For early
assessment of reservoir performance, even in case of an existing fluid sample, a deterministic approach
may not capture possible outcomes for a high level evaluation. In this particular case study, an actual
sample was collected from an exploration well. However, other than a well-site bubble point pressure
measurement and a stock tank oil density, no other PVT measurement (particularly viscosity) of the
reservoir fluid was available. The development partner assumed a specific viscosity correlation to estimate
viscosity of fluid at reservoir conditions for further studies. In comparison to this deterministic approach,
a probabilistic approach could have been used, where possible uncertainty ranges for fluid properties are
generated using a pseudo-database.
To generate a pseudo-database for uncertainty assessment, valid fluid property correlations (in this case
GOR and viscosity correlations) were selected. Validity of correlations was assessed based on parameters
such as applicable range of pressure and temperature, API gravity of the fluid, and gas gravity of the
flashed gas among other parameters. Table 3 shows the selected fluid property correlations where a full
factorial combination of inputs and models were used to generate a viscosity database. Figure 8 shows
probabilistic values for reservoir condition viscosity of the fluid system of interest. In this figure, the
deterministic viscosity value is shown with a star relative to possible range of fluid viscosities. In this
example, the deterministic approach viscosity was only representing a P20 case leading to potential
optimistic reservoir performance (lower viscosity).
SPE-180383-MS 11

Table 3—Valid correlations representing gas oil ratio and viscosity for fluid sample of case study #2.
Rs Correlations Vod Correlations Vos Correlations Vou Correlations

Glaso (1980) Beggs & Robinson (1975) Chew/Connally/Pipeflow Bergman & Sutton (2006)
Kartoutmodjo (1990) Dindoruk & Christman (2004) Dindoruk & Christman (2004) De Ghetto (1994)
Mendez (1976) Glaso (1980) Petrosky (1990) Hossain et al (2005)
Kartoutmodjo (1990) Sutton (2007) Kouzel - API Modified (1997)

Figure 8 —Probabilistic demonstration of possible uncertainty range for a reservoir fluid viscosity compared to a deterministic value
(star).

The main impact in this particular case was that it allowed the asset team to refine the development
plan based on a broader range of possible outcomes.

Probabilistic Fluid Models to Capture Fluid Property Variation in a Field with Limited Data
As part of an appraisal program, it was desired to have probabilistic fluid models that also capture fluid
property variation in the field. In this case, one of the main field variables of interest was temperature
which had a significant impact on fluid viscosity. Fluid property data (PVT, Viscosity) from a limited
number of samples and data from a nearby analog field were QCed and used to screen for representative
fluid property correlations. Figure 9 shows the benchmarking of a number of viscosity correlations against
experimental viscosity data. Experimental viscosity data represented a variation due to difference in the
API of the stock tank oil, as well as the temperature at which viscosities were measured. In this particular
case, there was a variation of about 3 (API, and a temperature variation of about 25 (F leading to viscosity
variation of about one order of magnitude. Based on a full factorial inclusion of input data (API, and
temperature in this case), only a few valid correlations were available for further uncertainty assessment
(element 2 of the overall workflow).
12 SPE-180383-MS

Figure 9 —Benchmarking of viscosity correlations based on existing experimental data.

Since appraisal team was also interested to cover a larger range of temperature (50 (F) and API gravity
variation (5 (API) compared to existing data ranges, perturbation in the input parameters (element 3 of the
overall workflow) was used to determine the uncertainty ranges. Also to provide a consistent mid case
(P50 model) for viscosity as a function of temperature, average viscosity values were used, before
P10/P90 values are calculated using equations given in the Appendix. Figure 10 shows consistent
probabilistic viscosity models for a range of temperature, compared with limited experimental data for
samples from various part of the field. Note that the mid case is the average of the data at each pressure
point. Also note that the viscosity models are only representing saturated oil viscosity as a function of
pressure, as it is typically used in a black oil formulation. This trend is obviously different than the actual
behavior of viscosity beyond bubble point pressure, where viscosity increases with pressure (constant
GOR beyond saturation pressure).
SPE-180383-MS 13

Figure 10 —Probabilistic fluid model with inclusion of fluid property variation in the field

This approach ensured properly capturing fluid model uncertainty ranges, since the subsurface team
had to rely on limited experimental data and analog data from nearby field for fluid data. Impact of PVT
uncertainty range was determined to be significant on reservoir performance, impacting the appraisal plan
by justifying collection of a number of new fluid samples in various parts of the field.
Conclusions
In stochastic reservoir simulation studies, a consistent approach to provide uncertainty ranges for fluid
property models is lacking. The workflow described in this paper allows for such models. Specifically:
● An overall workflow highlighting different aspects of uncertainty analysis for fluid properties was
presented.
● Step by step process was described to determine uncertainty ranges (P10/P90) as well as most
likely case (P50)
14 SPE-180383-MS

● A consistent approach was provided to generate and use a ⬙pseudo⬙-database for fluid property
uncertainty assessment. In case of no sample or limited data, this pseudo-database could be used
to provide reliable black oil models.
● Mathematical relations required to generate probabilistic models are simple in nature and include
provision for consistent property change. These relations could easily be programmed in other
tools or applications for use.
● A few cases were presented to demonstrate the uncertainty assessment of fluid property data as
well as the application of overall workflow.
These workflow and guidelines allows reservoir engineers to generate consistent black oil models for
use in reservoir simulation studies. The main implications of these consistent models are:
1. Probabilistic fluid models are generated using an approach where guess work is taken out of the
process and will not be based on experiential methods employed in the past.
2. Applying the same consistent approach for generating probabilistic models for different fields
and/or reservoirs will allow the practitioner to be confident when comparing reservoir performance
from a portfolio of options.
3. As the workflow is standardized, general practitioners are able to apply these workflows and
guidelines, thus less dependent on subject matter experts (SME) to generate probabilistic models.
As such, both cost and time saving will be recognized.

Acknowledgments
Authors would like to acknowledge Chevron Energy Technology Company for permission to publish this
paper. Authors would also like to extend their acknowledgement to members of Chevron’s Virtual Phase
Behavior Team, particularly Dr. Jeff Creek for valuable discussions and feedbacks on workflow devel-
opment and deployment.

Nomenclature
API : Degree API Gravity
BO : Oil Formation Volume Factor
BG : Gas Formation Volume Factor
CGR : Condensate Gas Ratio
Comp : Compressibility
EoS : Equation of State
FVF : Formation Volume Factor
GOR : Gas Oil Ratio
MW : Molecular Weight
n : number of data points
Pb : Bubble Point Pressure
Pres : Reservoir Pressure
Psat : Saturation Pressure
Rs : Solution Gas, same as GOR
SD : Standard Deviation
ST : Separator Test
STO : Stock Tank Oil
Tres : Reservoir Temperature
UA : Uncertainty Assessment
Vod : Dead Oil Viscosity
SPE-180383-MS 15

Vos : Saturated Oil Viscosity


Vou : Under-saturated Viscosity
␭ : A parameter defined in Equation A-1
␮ : Viscosity
G (subscript) : Gas
O (subscript) : Oil

References
Bon, J.; Sarma, H.; Rodrigues, T.; and Bon, J., 2007, Reservoir-Fluid Sampling Revisited-A Practical Perspective, SPE
Reservoir Evaluation and Engineering, December 2007, 589 –596, SPE-101037-PA.
Meisingset, K.K., 1999, Uncertainties in Reservoir Fluid Description for Reservoir Modeling, SPE Reservoir Evaluation
& Engineering, October 1999, 2 (5), 431–435, SPE-57886-PA.
Nagarajan, N.R.; Honarpour, M.M.; and Sampath, K., 2007, Reservoir-Fluid Sampling and Characterization-Key to
Efficient Reservoir Management, JPT, August 2007, 80 –91.
Seccombe, J.; Bonnie, R.J.M.; Smith, M.; and Akkurt, R., 2005, Ranking Oil Viscosity in Heavy Oil Reservoirs, Presented
at the SPE International Thermal Operations and Heavy Oil Symposium, Calgary, 1-3 November 2005, SPE-97935-
MS.
Sibley, W.P., 1970, 1995, A method for Handling Spatially Varying Fluid Properties in a Simulation Model for a Fissured
Reservoir, SPE Journal, March 1970, 25–32, SPE-2374-PA.
Steagall, D.E.; Schioze, D.J., 2001, Uncertainty Analysis In Reservoir Production Forecasts During Appraisal And Pilot
Production Phases, Presented at the SPE Reservoir Simulation Symposium, Houston, 11–14 February, SPE-66399-
MS.
Taylor, J. R., 1997, An Introduction to Error Analysis, Second Edition. Sausalito, California: University Science Books.
Zabel, F.; Law, D.H.-S.; Taylor, S.; and Zuo, J., 2010, Impact of Uncertainty of Heavy Oil Fluid Property Measurement,
JCPT, March 2010, 49 (3), 28 –35.
Walstrom, J.E.; Mueller, T.D.; and McFarlane, R.C., 1967, Evaluating Uncertainty in Engineering Calculations, JPT,
December 1967, 1595–1603. SPE-1928-PA.
16 SPE-180383-MS

Appendix A
(Equation A-1)

Thermal expansion in the Bo formula below is to account for dead oil expansion from standard
temperature to reservoir (or test) temperature.
(Equation A-2)

(Equation A-3)

(Equation A-4)

(Equation A-5)

(Equation A-6)

Vous aimerez peut-être aussi