Professional Documents
Culture Documents
SPE-Application of Probabilistic Techniques For Assesing San
SPE-Application of Probabilistic Techniques For Assesing San
SPE-Application of Probabilistic Techniques For Assesing San
SPE 47334
accomplished with the use of model derivations by Detournay parameter will be exceeded. The typical procedure is to
and Fairhurst17 and Tokar18, as well as validation and generate a series of output distributions for a range of
calibration by comparison to results obtained using finite controllable operating conditions such as the equivalent
element and finite difference geomechanical models. As such, circulating density (ECD) for drilling problems, or the
the model has enhanced flexibility for accurately modeling drawdown pressure for hydrocarbon production problems. It
both horizontal and vertical well problems in a variety of may be possible from these outputs to determine a safe limit
tectonic settings. The first version of this Windows-based for the critical operating parameter. An example for a borehole
program will be available from Advanced Geotechnology in stability analysis is shown in Fig. 3. In this case, it is
late 1998. concluded that the risk of borehole instability is too high for
An example of the type of output generated by the current ECD’s less than 1200 kg/m3. This conclusion is based on the
version of this model is shown in Fig. 1. Both the size and selection of a 25% probability that NYZA exceeds 1.0.
shape of the zone of rock which has undergone brittle-plastic For sand production risk assessments in open-hole
yield (referred to hereafter as the yielded zone) may be used as completions, this a type of output might indicate that the risk
indicators of the magnitude of borehole instability risk for of sand production is high for all economically feasible
drilling problems, as well as sand production risk for open- drawdown pressures. The operator might then conclude that
hole production problems. For the examples in this paper, the sand control measures are required. This type of analysis is
Normalized Yielded Zone Area (NYZA) is used as the risk also applicable to perforated completions, although the 3D
parameter. NYZA is defined as the cross-sectional area of geometry of this type of problem makes the analysis more
yielded rock divided by the original cross-sectional area of the complex.
drilled borehole.
Some Practical Aspects of Probabilistic Modeling
Monte Carlo Simulation Tools. The model described above Selection of Input PDF’s. In the case of input parameters for
has been implemented in a PC-based spreadsheet application. which data are available, the use of probabilistic input
The solution is accomplished using a Visual Basic macro, represents a means of accounting for the spatial variations
which obtains the required input parameters from selected which are expected to occur for a parameter. For example,
spreadsheet cells and returns the values of the output Fig. 4 shows the distribution of the peak cohesion strength
parameters to the spreadsheet. Calculation of the size of the parameter from a near-vertical well penetrating a weak
yielded zone is generally achieved in a fraction of a second. sandstone reservoir. These values were calculated from log-
Implementation of Monte Carlo simulation is achieved derived mechanical properties that had been calibrated to
using the EXCEL add-in called @RISK19. This component laboratory-measured static mechanical properties from cores
allows the user to specify model input parameters using a taken from this well. A number of commercially available
number of built-in probability density functions (PDF’s). software programs exist which can be used for identifying and
Likewise, the user specifies which spreadsheet cells contain ranking the probability distribution functions that best match
output parameters that should be captured and stored for each the actual data. In this case the best fit PDF was the Extreme
model simulation. Some comments regarding the choice of Value distribution, with a location parameter a of 2.01 and a
PDF’s for various input parameters as well as the most shape parameter b of 0.75.
important simulation settings are discussed later in this paper. When there are little or no data available for some input
@RISK provides a number of useful built-in tools for parameters, the use of probabilistic input allows the user to
analyzing the simulation output. These tools include: (1) tables vary the value of each parameter over a plausible range. The
of detailed and summary statistics for all of the input and choice of the form of the input function may be somewhat
output parameters, (2) parameter sensitivity statistics and arbitrary in this case. The use of a uniform PDF is most
charts, and (3) histograms showing the distributions of the cautious where a reasonable estimate of the most likely value
output parameters. is not available. The triangular PDF is useful in cases where
the user has greater confidence regarding a most likely value.
Analysis Procedure. The basic analysis procedure is
summarized in Fig. 2. The possible sources of input data are Minimum Number of Model Iterations. A crucial step in
also included in this figure, although it is not uncommon to probabilistic modeling is to ensure that enough simulations are
have access to only a few of these data sources. The in-situ run so that the true form of the output cumulative probability
stresses, rock properties and reservoir fluid properties are function is predicted, while at the same time preserving the
examples of parameters that can have a critical influence on economy of the simulation process by running as few
the model results but are often poorly constrained or are iterations as possible. The Latin Hypercube sampling method,
known to vary throughout the drilled formations due to although the subject of some criticisms, is the preferred
geological heterogeneity. sampling method because it reduces the necessary number of
For borehole stability and sand production risk analyses, a model iterations by a factor of approximately 1/3 20.
useful form of the output is a histogram of the NYZA. The Simulations should be run until the output parameter
cumulative probability curve on such a plot allows the user to distribution remains essentially unchanged with additional
assess the probability that a critical value of this risk
SPE 47334 APPLICATION OF PROBABILISTIC TECHNIQUES FOR ASSESSING SAND PRODUCTION AND BOREHOLE INSTABILITY RISKS 3
model iterations. @RISK can monitor the convergence of the hole enlargement measured with 4-arm caliper logs from an
output parameter distributions. For the poro-elastoplastic the offset well, and leak-off test results from another field in
model described in this paper, it has been found that iteration the same sedimentary basin. It was noted by the operator that a
counts in the 1000 to 2000 range satisfy the criteria described slightly overpressured formation pressure gradient of
above. These simulations generally require a few minutes to 11.2 kPa/m existed in the reservoir.
run on a Pentium-based PC. The operator wished to assess two issues: (1) would it be
possible to drill the horizontal leg of the well at underbalanced
Interdependency Among Input Parameters. For many bottomhole pressures, and (2) would an open-hole completion
physical problems, there may exist some degree of in this reservoir be feasible.
interdependence between two or more input parameters. In
such cases, it is undesirable to allow each of the input Analysis Results. Monte Carlo simulations were
parameters to be randomly selected from independently performed over a range of equivalent circulating densities
prescribed PDF’s. For example, it is known that the ranging from underbalanced to overbalanced conditions. The
unconfined compressive strength (UCS) of a rock is generally results are shown in Fig. 5 through Fig. 8. It is clear from
related to its Young’s Modulus, although the reliability of the these results that there is a trend of decreasing NYZA with
correlation will vary for different lithologies. In such a case, it increasing ECD. As shown in Fig. 9, the mean NYZA is very
is desirable to specify these input parameters in such a way large (5.6) at the lowest ECD, and ECD’s in the 1250 to
that they have some measure of interdependence, while 1300 kg/m3 are range are required to obtain mean NYZA’s of
preserving a degree of randomness in the association between the order of 1.0. When applying this type of analysis in
them. The simplest way to accomplish this is to specify an previous open-hole stability investigations, it has been
@RISK PDF for one of the parameters (Young’s Modulus, for observed that NYZA values of the order of 1.0 are often
example), and then specify the second parameter (e.g., UCS) associated with the onset of borehole instability-related
as a function of the first. A PDF may be included in this difficulties such as stuck pipe. Hence, these results indicate
function, or a random function generic to the spreadsheet in that the risk of severe borehole instability is unacceptably
use, in order to introduce some variability into the parameter large for underbalanced drilling operations, and that ECD’s of
interdependence. the order of at least 1250 kg/m3 are required to reduce the risk
@RISK also provides built-in tools for generating of borehole instability problems to an acceptable level.
interdependent input function distributions based on the As in the case of borehole instability risk, the tendency of
Spearman rank order correlation coefficient. This technique is an barefoot-completed horizontal well to produce sand or at
implemented by specifying the PDF’s for each interdependent worst, collapse entirely, is strongly related to the extent of
input parameter, as well as the correlation coefficient yielding around the borehole. The results from the borehole
describing the dependence between them. Upon initiating the instability risk analysis were also useful for evaluating sand
simulation, @RISK randomly generates the interdependent production risk. Fig. 5 and Fig. 6 show the NYZA histograms
input parameters required to perform all of the model for bottomhole pressures corresponding to drawdowns of 2.2
iterations. Prior to performing the iterations, however, @RISK and 0.6 MPa, respectively. These figures indicate that a
then pairs off the individual values from each of these significant amount of yielding will occur even at these
distributions in such a way as to achieve the desired relatively low drawdown pressures. It is difficult to
correlation coefficient. quantitatively assess the exact extent of yielding that will
Murtha20 describes an example where the variance in a result in unacceptable amounts of sand production because
simulation output is underestimated if input parameter factors such as fluid viscosities, saturations and flow rates as
interdependence is ignored. Similar results have been well as sand grain characteristics will affect the mobilization
observed in our borehole stability and sand production risk of yielded sand. However, it was decided based on these
simulations. results that the risk of experiencing problematic amounts of
sand production was unacceptably high, and sand control
Borehole Instability and Sand Production Risk for a measures were required.
Horizontal Well Based on recommendations arising from this investigation,
This section summarizes the application of the techniques the operator drilled the horizontal well using ECD’s in the
described above to a borehole instability and sand production 1200 to 1250 kg/m3 range. No serious borehole instability
risk assessment study for a horizontal well in a weak problems were encountered. Slotted liners were installed to
sandstone reservoir. The input parameters used in this analysis control sand production, although the performance of this well
are listed in Table 1. Note that several of the inputs are during production has not yet been established.
specified as PDF’s rather than discrete values. Rock
mechanical properties were estimated from log-derived Conclusions
dynamic elastic properties from a nearby near-vertical offset 1. Monte Carlo simulation techniques have been applied to
well. The horizontal in-situ stress magnitudes and orientations poro-elastoplastic borehole stability analyses for over- or
were estimated based on the tectonic setting of the field, the underbalanced drilling conditions and to assess sand
trend of normal faults mapped in this region, back-analysis of production risk with drawdown.
4 P.J. MCLELLAN, C.D. HAWKES SPE 47334
2. This technique allows one to account for uncertainty and 15. McLellan, P.J. and Wang, Y.: “Predicting the effects of pore
variability in several rock mechanical input parameters used in pressure penetration on the extent of borehole instability:
the stability analysis Application of a versatile poro-elastoplastic model,” In Proc.
SPE/ISRM Eurock’94, Balkema, Rotterdam (1994) 205.
3. The end result of a Monte Carlo simulation for this class
16. Hawkes, C.D. and McLellan, P.J.: “A new model for predicting
of problem is a Cumulative Distribution Function (CDF) time-dependent failure of shales: Theory and application,” paper
which provides an estimate of the probability of a certain CIM 97-131 presented at the 1997 Annual Technical Meeting of
outcome, e.g., the yielded zone size exceeding a threshold the CIM Petroleum Society, Calgary, June 8-11.
value known to result in drilling or production problems. 17. Detournay, E. and Fairhurst, C.: “Two-dimensional elastoplastic
4. Further work is needed to better understand the acceptable analysis of a long, cylindrical cavity under non-hydrostatic
level of risk associated with borehole stability or sand loading,” Int. J. Rock Mech. Min. Sci. & Geomech. Abstr. (24(4)
production for various operating conditions. 1987) 197.
18. Tokar, G.: “Generalization of Galin’s problem to frictional
materials and discontinuous stress fields,” Int. J. Solids
References
Structures (26(2) 1990) 129.
1. Bradley, W.B.: “Failure of inclined boreholes,” Transactions of
19. Palisade Corp.: Guide to Using @RISK, Windows Version (Jul.
the ASME (Dec. 1979) 232.
1997) 319.
2. Mody, F.K. and Hale, A.H.: “Borehole-stability model to couple
20. Murtha, J.A.: Decisions Involving Uncertainty: An @RISK
the mechanics and chemistry of drilling-fluid/shale
Tutorial for the Petroleum Industry, James A. Murtha, Houston
interactions,” JPT (Nov. 1993) 1093.
(1995) 171.
3. Santarelli, F.J., Brown, E.T. and Maury, V.: “Analysis of
borehole stresses using pressure-dependent, linear elasticity,”
Int. J. Rock Mech. Min. Sci. & Geomech. Abstr. (23(6) 1986)
445.
4. Nawrocki, P. and Dusseault, M.B.: “Modelling of damaged
zones around openings using radius-dependent Young’s
modulus,” J. Can. Petrol. Tech. (Mar. 1996) 31.
5. Cheng, H. and Dusseault, M.B.: “Deformation and diffusion
behaviour in a solid experiencing damage: a continuous damage
model and its numerical implementation,” Int. J. Rock Mech.
Min. Sci. & Geomech. Abstr. (30 1993) 1323.
6. Shao, J.F. and Khazraei, R.: “Borehole stability analysis in
brittle rocks with continuous damage model,” In Proc.
SPE/ISRM Eurock’94, Balkema, Rotterdam (1994) 215.
7. Papamichos, E., Vardoulakis, I. and Sulem, J.: “Generalized
continuum models for borehole stability analysis,” In Proc.
SPE/ISRM Eurock’94, Balkema, Rotterdam (1994) 37.
8. Thallak, S., Rothenburg, L. and Dusseault, M.B.: “Analysis of
borehole stability using discrete element models,” In Proc. 7th
Int. Congress on Rock Mechanics, Aachen, Germany (1991)
813.
9. Rawlings, C.G., Barton, N.R., Bandis, S.C., Addis, M.A. and
Gutierrez, M.S.: “Laboratory and numerical discontinuum
modeling of borehole stability,” JPT (Nov. 1993) 1086.
10. Zheng, Z., Kemeny, J. and Cook, N.G.W.: “Analysis of
borehole breakouts,” JGR (B94 1989) 7171.
11. Bradford, I.D.R. and Cook, J.M.: “A semi-analytic elastoplastic
model for borehole stability with applications to sanding,” In
Proc. SPE/ISRM Eurock’94, Balkema, Rotterdam (1994) 347.
12. Ladanyi, B.: “Use of the long-term strength concept in the
determination of ground pressure on tunnel linings,” In Proc. 3rd
Int. Congress on Rock Mechanics, National Academy of
Sciences, Washington (1974) 1150.
13. Wang, Y. and Dusseault, M.B.: “Borehole yield and hydraulic
fracture initiation in poorly consolidated rock strata - Part II.
Permeable media,” Int. J. Rock Mech. Min. Sci. & Geomech.
Abstr. (28 1991) 247.
14. Murtha, J.: “Incorporating historical data into Monte Carlo
simulation,” paper SPE 26245 presented at the 1993 Petroleum
Computer Conference, New Orleans, Jul. 11-14.
SPE 47334 APPLICATION OF PROBABILISTIC TECHNIQUES FOR ASSESSING SAND PRODUCTION AND BOREHOLE INSTABILITY RISKS 5
TABLE 1: BASE CASE INPUT PARAMETERS USED FOR BOREHOLE YIELD MODELING,
HORIZONTAL WELL IN A WEAK SANDSTONE RESERVOIR
Parameter Value*
peak friction angle {30, 35, 40}°
residual friction angle peak friction angle - {3, 5, 7}°
peak cohesion {0.75, 0.9, 1.05} MPa
residual cohesion {0.25, 0.75, 1} × peak cohesion MPa
Biot’s poroelastic coefficient 1
poroplastic coefficient 1
Young’s modulus 500 MPa
Poisson's ratio 0.3
Elastic zone permeability 5000 mD
Yielded zone permeability {0.2, 1, 4} × kelastic mD
filter cake efficiency {0.5, 0.8, 1}
yielded zone fluid viscosity 3 mPa·s
elastic zone fluid viscosity 3 mPa·s
drainage radius 6m
vertical stress gradient {21, 22, 23} kPa/m
maximum horizontal stress gradient {1.0, 1.05, 1.1} × min. horizontal stress grad.
minimum horizontal stress gradient {13.5, 14, 15} kPa/m
azimuth of minimum horizontal stress NE-SW
reservoir pressure gradient 11.2 kPa/m
equivalent circulating density 1000 to 1300 kg/m3
true vertical depth 1530 mKB
well radius 0.108 m
wellbore orientation parallel to maximum horizontal stress
KB height 5.4 m above ground level
*Numbers in parentheses represent the left, center and right points of a triangular probability density function
NYZA = A1
A2
A1
minimum intact rock
horizontal A2
stress predicted yielded
zone size
Well Design
Rock Mechanical Data
In-Situ Stress • Depth
• Regional data • Azimuth • Dynamic elastic properties
• Fracturing data • Inclination • Lab strength tests
• Breakouts • Hole size • Log analysis
• Drawdown • Back analysis
Petrophysical Data
Reservoir Data
• porosity Borehole Stability
• permeability & • Reservoir pressures
• sonic velocity • Fluid viscosity
Sand Production
• density • PVT
• saturations Risk Model
Probabilistic
Analysis Using
Monte Carlo
Techniques
Prediction of the
Probability of Various
Outcomes
e.g. Severe Hole Collapse
to Minor Sanding
PROBABILITY
Fig. 2.—Procedure for probabilistic modeling of borehole instability or sand production risk.
SPE 47334 APPLICATION OF PROBABILISTIC TECHNIQUES FOR ASSESSING SAND PRODUCTION AND BOREHOLE INSTABILITY RISKS 7
100
90
80 1100 kg/m^3
1200 kg/m^3
1300 kg/m^3
Cumulative Probability (%)
70
1400 kg/m^3
60
ECD required to reduce probability of instability-
related problems to less than 25%
50
40
10
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Normalized Yielded Zone Area
Fig. 3.—Example showing the cumulative probability of NYZA for a range of mud equivalent circulating densities.
0.8
0.7
0.6
Best Fit PDF:
Extreme Value Function (a = 2.01, b = 0.75)
0.5
Frequency
0.4
0.3
0.2
0.1
0
1.2
1.6
2.0
2.3
2.7
3.0
3.4
3.7
4.1
4.4
4.8
5.2
5.5
5.9
6.2
6.6
6.9
7.3
7.7
8.0
Cohesion (MPa)
Fig. 4.—Histogram of log-derived peak cohesion values for a weak sandstone reservoir.
8 P.J. MCLELLAN, C.D. HAWKES SPE 47334
90% 90%
300
250
80% 80%
Cumulative Frequency
Cumulative Frequency
Frequency
Frequency
50% 150 50%
150
40% 40%
100
30% 30% 100
20% 20%
50
50
10% 10%
0% 0 0% 0
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
Normalized Yielded Zone Area Normalized Yielded Zone Area
Fig.5—Predicted extent of yielding in a weak sandstone reservoir Fig.6—Predicted extent of yielding in a weak sandstone reservoir
3 3
at 1000 kg/m ECD (drawdown pressure = 2.2 MPa). at 1100 kg/m ECD (drawdown pressure = 0.6 MPa).
90% 90%
250 500
80% 80%
60% 60%
Frequency
Frequency
40% 40%
100 200
30% 30%
20% 20%
50 100
10% 10%
0% 0 0% 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Normalized Yielded Zone Area Normalized Yielded Zone Area
Fig.7—Predicted extent of yielding in a weak sandstone reservoir Fig.8—Predicted extent of yielding in a weak sandstone reservoir
3 3
at 1200 kg/m ECD. at 1300 kg/m ECD.
SPE 47334 APPLICATION OF PROBABILISTIC TECHNIQUES FOR ASSESSING SAND PRODUCTION AND BOREHOLE INSTABILITY RISKS 9
5
NYZA
1
underbalanced overbalanced
0
1000 1050 1100 1150 1200 1250 1300
3
ECD (kg/m )