John Emeka Udegbunam
John Emeka Udegbunam
John Emeka Udegbunam
by
ISBN: 978-82-7644-590-9
ISSN: 1890-1387
i
Contents
Preface iv
Acknowledgements v
Summary vi
List of Papers vii
Abbreviations ix
Nomenclature x
1 Introduction 11
1.1 Well Construction 11
1.2 Conventional Drilling 12
1.3 Innovative Drilling Concepts 13
1.3.1 Managed-Pressure Drilling 14
1.3.2 Underbalanced Drilling 18
2 The Role of Models in Well Planning 21
2.1 Introduction 21
2.2 Wellbore Stability Models 21
2.2.1 In-Situ Stresses 23
2.2.2 Mechanisms of Wellbore Failures 24
2.3 Well Flow Models 25
2.3.1 Model Formulation 26
2.3.2 Numerical Solution 30
3 Stochastic Modeling 33
3.1 Introduction 33
3.2 Risk and Uncertainty 34
3.3 Monte Carlo Simulation 35
ii
3.4 Stochastic Model 37
4 Results and Discussion 39
4.1 Wellbore Stability Analyses 39
4.1.1 Example Cases 41
4.1.2 Stochastic Sensitivity Analysis 45
4.2 Underbalanced Operations 48
4.2.1 Uncertainty BHP Prediction 48
4.2.2 Mechanics of Collapse for UBD 55
4.2.3 Sensitivity Analysis for UBO 58
5 A Transient Model for Well Flows 60
5.1 Introduction 60
5.2 The AUSMV Scheme 61
5.3 Dual-Gradient Drilling 62
5.4 Underbalanced Drilling 67
5.5 Inference 70
6 Overview of the Research Papers 72
7 Conclusion and Further Work 76
7.1 Conclusion 76
7.2 Further Work 77
References 79
iii
Preface
The Norwegian Ministry of Education and Research funded the project. There
was a onetime financial support from Statoil through the Akademia program.
The thesis is divided into two parts. The first part comprises seven chapters.
Chapters 1 and 2 present background information relevant to the research.
Chapter 3 discusses stochastic modeling. The chapter describes a fully
probabilistic approach for transforming deterministic models to stochastic
models. In this case, wellbore stability models and a steady-state hydraulic
model are given as the base models.
Chapter 4 presents simulation results and discussion.
Chapter 5 discusses a transient flow model for MPD and UBD applications.
Chapter 6 gives summaries of the research publications.
The last chapter presents the conclusion of the work. The chapter also includes
a list of concerns that should be resolved in future studies.
The second part of the thesis consists of five technical papers describing the
research findings. A seminar paper that has been prepared for submission to a
conference is also included.
iv
Acknowledgements
The Norwegian Ministry of Education and Research funded this work. I thank
the ministry for its beneficence.
I also thank Statoil for providing a onetime financial support through the
Akademia program.
I thank Prof. Kjell Kåre Fjelde for his generous contribution towards the
success of this work. Prof. Fjelde is not only my academic mentor but also a
dependable friend, with whom I often share technical and personal challenges.
His three years of close supervision of this work has transformed me into a fine
petroleum engineer with a strong drive for problem solving.
I also thank Prof. Bernt Sigve Aadnøy for bringing his wealth of experience to
bear on this work. Indeed, many inspiring discussions I had with him have
enriched the content of the thesis.
Above all, I thank Almighty Chukwu Okike, the Cause of all things visible and
invisible, for His sustenance and guidance. His love and grace are ever upon
my household.
Chukwuemeka Udegbunam
September 2014
v
Summary
vi
List of Papers
Paper I John Emeka Udegbunam, Kjell Kåre Fjelde, Øystein Arild, Eric
Ford, and Hans Petter Lohne. Uncertainty-Based Approach for
Predicting the Operating Window in UBO Well Design. Paper
SPE 164916, presented at the EAGE Annual Conference &
Exhibition incorporating SPE Europec, London, UK, 10–13 June
2013.
Paper II John Emeka Udegbunam, Hans Petter Lohne Kjell Kåre Fjelde,
Øystein Arild, and Eric Ford. Improved Underbalanced
Operations with Uncertainty Analysis. Paper SPE 167954,
presented at the 2014 IADC/SPE Drilling Conference and
Exhibition, Fort Worth, Texas, USA, 4–6 March 2014.
Paper III John Emeka Udegbunam, Bernt Sigve Aadnøy, and Kjell Kåre
Fjelde. Uncertainty Evaluation of Wellbore Stability Model
Predictions. Paper SPE 166788, presented at the SPE/IADC
Middle East Drilling Technology Conference and Exhibition,
Dubai, 7–9 October 2013.
Paper IV John Emeka Udegbunam, Kjell Kåre Fjelde, Steinar Evje. The
Academic AUSMV Scheme — A Simple but Robust Model for
Predicting Highly Dynamic Well Flow Phenomena. An extended
abstract, presented at the Celle Drilling 2012, Celle, Germany, 17–
18 September.
vii
Paper V John Emeka Udegbunam, Kjell Kåre Fjelde, Steinar Evje, Gerhard
Nygaard. A Simple Transient Flow Model for MPD and UBD
Applications. Paper SPE 168960, in the proceedings of the
SPE/IADC Managed Pressure Drilling and Underbalanced
Operations Conference and Exhibition, Madrid, Spain, 8–9 April
2014.
viii
Abbreviations
ix
Nomenclature
x
Introduction
1 Introduction
11
Introduction
Figure 1.1—Well costs for 8 fields on the NCS between 2003 and 2012 (from Petoro
2014).
The need to improve well design, considering high well costs and modeling
uncertainties, has motivated the stochastic method presented in Chapter 3.
12
Introduction
20% of the total well time. To minimize the problems, Aadnøy (2010) proposed
the median line principle. This states that the mud weight should be kept close
to the in-situ stress field in a surrounding rock. The design approach minimizes
the risks of lost circulation and differential sticking, because a minimum
disturbance is introduced in the borehole wall.
Fig. 1.2 shows typical mud weight selections. The median-line mud weight will
provide a common optimum for many key parameters that influence the drilling
process. In addition, the mud weight is always a compromise—one optimum at
the top and another optimum at the bottom.
13
Introduction
troublesome zones are isolated with casings, to prevent interaction between the
well and the formation. As the well gets deeper, the diameter of a successive
casing becomes smaller until a pay zone is drilled. However, there exist some
prospects, where it will impossible to reach the target depths, or where desired
production-casing diameter cannot be achieved, by use of the conventional
technology (Breyholtz 2011). Such prospects include deep and ultra-deep-
water wells. Other problem areas include depleted fields and naturally fractured
formations.
Depending on the problem areas, the need to drill deeper with few casings and
handle depleted reservoirs, to minimize lost circulation and well kicks, has
motived the developments of managed-pressure drilling (MPD) and
underbalanced drilling (UBD).
Lost circulation and well kicks are frequently encountered while drilling in
these formations. It is also a major drilling problem associated with naturally
fractured formations.
MPD technologies make it possible to exploit the prospects that,
conventionally, would have been undrillable or difficult to drill. The methods
enable drillers to precisely manage the annular pressure profile and ‘walk’ the
thin line between the pore and fracture pressures. MPD drastically helps to cut
nonproductive time (NPT) by reducing circulation loss, gas kicks, and stuck
pipe incidents. MPD also mitigates equivalent circulating density (ECD)
problems while drilling extended reach wells and wells with narrow pressure
margins. Casing points can be extended, to limit the total number of casing
strings and the subsequent hole size reduction (Rehm et al. 2008).
According to the IADC committee on UBO and MPD (Malloy et al. 2009),
managed-pressure drilling is an adaptive drilling process used to precisely
control the annular pressure profile throughout the wellbore. The objectives are
14
Introduction
15
Introduction
16
Introduction
17
Introduction
Pressure, sg
1,00 1,20 1,40 1,60 1,80 2,00 2,20 2,40
0
Fracture pressure
2000
TVD, m
Dual-gradient well
3000 pressure
Effective MW
during circulation
4000
Static mud weight
5000
6000
7000
Figure 1.3—Total vertical depth versus pressure, for a typical SMD system. Seawater
depth = 1500 m, gradient = 1.03 sg. Mud gradient = 2.12 sg. MW = mud weight.
The MPD system is more expensive compared with the conventional drilling.
Yet it delivers significant cost savings by cutting NPT associated with kicks,
losses, and well-control events, increasing rate of penetration (ROP) and
making previously undrillable wells drillable. Operators find that using MPD
cuts their NPT from 20 to 100%. The system usually requires only minor
modifications to the conventional rig. It permits all drilling, logging and
completion operations to be executed safely and efficiently (Nas, 2010).
18
Introduction
19
Introduction
20
The Role of Models in Well Planning
2.1 Introduction
Models are developed to approximate or mimic systems and processes of
different natures and of varying complexity. Many processes are so complex
that physical experimentation is too time-consuming, too expensive, or even
impossible. Therefore, to investigate systems and processes, investigators often
turn to mathematical or computational models (Saltelli et al. 2008).
Well planning and subsequent drilling operations require the use of
models. These come in form of wellbore stability models and hydraulic models
for calculating well pressures. The hydraulic models are further subdivided into
steady-state and transient models.
The models, however, have some limitations because the modeling processes
only approximate physical phenomena. There also uncertainties related to the
model input parameters. Thus, modelers should be aware of the imprecision of
these models. In this chapter, the mathematical models used in the present
research will be discussed.
Stochastic modeling will be presented in the Chapter 3. The chapter describes
how to transform the deterministic models to stochastic models.
21
The Role of Models in Well Planning
22
The Role of Models in Well Planning
23
The Role of Models in Well Planning
Pwf = 3σ h − σ H − Po (2.1)
Shear Failure. The von Mises Yield Condition and Mohr-Coulomb Shear
Failure Criterion are the most commonly used hypotheses for shear failure
analysis. However, the discussion will be limited to Mohr-Coulomb Criterion,
which is the model adopted for wellbore collapse.
Formations at depth exist under a state of compressive in-situ stress (Bradley
1979). Wellbore collapse is often caused by shear failure of rock around the
24
The Role of Models in Well Planning
borehole. To keep the rock from failing during drilling operation, mud pressure
must be sufficiently high, to support the load imposed on the borehole wall by
the in-situ stresses. The mud pressure must not be too high as to fracture the
formation.
The Mohr-Coulomb failure model for borehole collapse neglects the
intermediate principal stress but include the effect of the directional strengths
of shales (Aadnøy and Chenevert 1987). The model predicts the minimum mud
pressure that can cause wellbore collapse.
Eq. 2.2 gives the shear failure model for collapse.
1
Pwc = (3σ H − σ h )(1 − sin α ) + Po sin α − τ o cos α (2.2)
2
The equation expresses the collapse gradient in terms of the horizontal stresses
, pore pressure, rock friction angle, and cohesive rock strength.
25
The Role of Models in Well Planning
220
Pore pressure
210 Bottomhole pressure
Collapse pressure
200
190
Pressure, bar
180
170
160
150
140
0 10 20 30 40 50 60
Gas-Injection Rate, m3/min
∂
( Aρ l α l v l ) = 0 (2.3)
∂z
∂
( Aρ g α g v g ) = 0 (2.4)
∂z
26
The Role of Models in Well Planning
∂P ∆Pf
= −( ρ l α l + ρ g α g ) g − (2.5)
∂z ∆z
Eqs. 2.3 and 2.4 represent the steady-state mass transport of liquid and gas. This
means that mass rate of each phase is constant along the well.
Eq. 2.5 represents the conservation of total momentum of the fluid mixture. It
expresses the annular well pressure as a combination of hydrostatic pressure
gradient and frictional effects.
This system of equations will be solved for each cell in a discretized well. The
solution procedure will be described in detail in Section 2.3.2.
Closure Laws. From the three conservation equations, the unknowns are the
liquid and gas volume fractions, designated as αl and αg; phase velocities, vl and
vg; phase densities, ρl and ρg; and the wellbore pressure, P. This implies that
there are three equations but seven unknowns. To find a solution to the system,
four additional constraints or closure laws are required.
Eqs. 2.6 through 2.8 represent phase densities and volume fraction relation.
( P − Patm )
ρ l = ρ l ,0 + (2.6)
a l2
P
ρg = (2.7)
a g2
(α l + α g ) = 1 (2.8)
The constants, ρl,0 = 1000 kg/m3 is the density of liquid at standard conditions,
Patm = 1bar (105 Pa.) is the atmospheric pressure, al = 1000 m/s is the sound
velocity in the liquid phase, and ag = 316 m/s is the sound velocity in the gas
phase. The sound velocity is related to the compressibility of the fluid.
27
The Role of Models in Well Planning
v g = C0 vm + Vd = C0 ( vls + v gs ) + Vd → C0 (α l v l +α g v g ) + Vd (2.9)
Both coefficients, C0 and Vd, are flow-dependent parameters. Eq. 2.9 is also
valid for a vertical liquid-gas flow in an annulus.
The gas phase moves faster than liquid because of two mechanisms:(i) higher
concentration of gas near the center of a pipe, where velocity is higher, with the
effect captured by the centerline velocity, C0vm, and (ii) the tendency of the gas
to rise in the pipe due to buoyancy, given by Vd (Livescu et al. 2009). The values
of C0 ranges from 1.0 to 1.2 for most vertical flow patterns such as bubble,
dispersed bubble, churn and slug flows. The values of the parameter, Vd, may
typically range from zero to 0.55 m/s.
For the dispersed bubble flow, C0 = 1.0, whereas Vd = zero, that is, no-slip
conditions. This occurs at high superficial liquid velocities, whereby turbulent
forces break large bubbles and disperse the gas phase in a continuous stream of
liquid. This can even occur for GVFs larger than 0.25 but not exceeding 0.52
(Lage and Time 2000).
Higher GVFs cause transition to slug flow. For the slug flow, the conditions for
the rise of the Taylor bubble are given by C0 = 1.2 and Vd = 0.35 (g (do + di)) 0.5
(Lage and Time 2000).
In the bubble flow, the C0 values in the range of 1.0 to 1.1 have been mentioned
in the literature according to Lage and Time 2000.
Frictional pressure-loss gradient is estimated with Eq. 2.10. Eq. 2.11 gives an
expression for calculating the hydrostatic pressure-loss gradient. The equations
are based on the drift-flux formulation, where flow condition is assumed
homogeneous, and the variables are averaged over a cell or well segment.
∆Pf 2 fρ m v m abs( v m )
= (2.10)
∆z (d o − d i )
28
The Role of Models in Well Planning
∆Ph
= ρm g (2.11)
∆z
ρ m abs( v m )( d o − d i )
N Re = (2.12)
µm
Eqs. 2.13 through 2.15 give the multiphase mixture variables: mixture velocity,
mixture density, and mixture viscosity.
v m = v ls + v gs = α l v l + α g v g (2.13)
ρm = αl ρl + α g ρ g (2.14)
µm = α l µl + α g µ g (2.15)
For NRe ≥ 3000, the flow is turbulent, and the friction factor is defined by:
For NRe ≤ 2000, the flow is laminar, and the friction factor is given as:
24
f = (2.17)
N Re
29
The Role of Models in Well Planning
Eq. 2.17 is used for estimating the Fanning friction factor, for laminar flow in
a concentric annulus. The proposed correlation is given in Caetano (1986).
Q = PI ( Po − Pb ) (2.18)
Pb = Ph + Pf + Pchoke (2.19)
30
The Role of Models in Well Planning
A pressure, Pguess, is guessed at the lower boundary of the first cell. The phase
densities in this cell are calculated based on the pressure value.
At the injection point (well bottom), the fluid-mass rates, ql and qg, are known.
With these values, the phase superficial velocities are calculated by use of Eqs.
2.20 and 2.21.
ql
vls ,1 = (2.20)
Aρ l ( Pguess )
qg
v gs ,1 = (2.21)
Aρ g ( Pguess )
Gas velocity, vg,1, is calculated with Eq. 2.9. The phase volume fractions and
liquid velocity are determined by solving Eq. 2.22 through 2.24.
v gs ,1
α g ,1 = (2.22)
v g ,1
31
The Role of Models in Well Planning
α l ,1 = 1 − α g ,1 (2.23)
vls ,1
vl ,1 = (2.24)
α l ,1
The numerical solver estimates the pressure in the next cell by considering the
pressure drops across the previous cell. The frictional and hydrostatic pressure-
loss gradients are calculated with Eqs. 2.10 and 2.11.
Then the pressure in cell, i, is given by Eq. 2.25
∆Ph ,i −1 ∆Pf ,i −1
Pi = Pi −1 − ∆z ( + ) (2.25)
∆z ∆z
Again, based on the value of Pi, the phase densities, superficial velocities,
velocities, and volume fractions in this cell are obtained by solving the mass
conservation laws (Eqs. 2.3 and 2.4) and constraints (Eqs. 2.6 through 2.9)
The same algorithm is followed in calculating the pressure and other flow
variables in the next cell until the outlet pressure, PN, in the last segment is
estimated. An exact solution must satisfy the function expressed in Eq. 2.26. If
not, the solver repeats the iteration once again until a solution is found. The
principle is to ensure that the numerical solution satisfies the physical reality.
However, a tolerance limit, tol = 1000 Pa, is specified in the solver such that if
G (Pguess) < tol, then the guessed pressure, Pguess, is the real bottomhole pressure.
The two-phase flow model is also described in the Papers I and II. The model
forms the basis for the uncertainty prediction of UBD operating window
presented in Chapter 4.
32
Stochastic Modeling
3 Stochastic Modeling
3.1 Introduction
Physical and geologic models are the bases for constructing mathematical
models. While the physical model is a tangible object that represents a natural
phenomenon or process, the geologic model is an abstract formulation of a
geologic concept that may be tested by collecting geologic data (Koch and Link
1980). Modeling geological processes is subject to uncertainty because of
scarcity and inaccurate nature of input data (de Rocquigny et al. 2008).
Deterministic models are mainly used in the industry, but they do not consider
uncertainty propagation. The problem with these models is that they usually
involve single-value assessments of averages or expected values, which
effectively obscure risk (Nersesian 2013). An over-simplification of input
parameters results in loss of variability information and inability to analyze the
associated uncertainties and risks quantitatively (Liang 2002).
Uncertainty-based methods for drilling and well design (Morita 1995;
Ottesen et al. 1999; Liang 2002; de Fontoura et al. 2002; Sheng et al. 2006;
Aadnøy 2011) are not relatively new concepts. The industry frequently uses
scenario analysis to predict the most likely, the best and the worst cases through
project cycles. Traditionally, sensitivity analysis is used to ascertain the
contributions of input-parameter uncertainties to output uncertainties. The
method is often used in hydrocarbon volumetric estimation and well
forecasting. One limitation of this approach is that the determining factors, for
example, price may fluctuate from time to time. Thus, it will not adequately
capture the possible range of the variable distributions.
As quoted in Nersesian (2013), risk lies in the tail of a probability
distribution. Only by making the transformation from deterministic and
scenario models to stochastic models can one evaluate risk hidden in the tail.
A novel stochastic modeling approach is proposed here. This shows how to
propagate uncertainties from assessable input parameters to output realizations.
Monte Carlo technique provides a means for the uncertainty propagation. With
the approach, the deterministic models presented in Sections 2.2 and 2.3 are
transformed to stochastic models. Although the relationships among the
variables in the deterministic models are easily predictable, introduction of one
or more random elements alters the relationship. Then, instead of having single-
33
Stochastic Modeling
point estimates, the models (now stochastic) predict outputs that follow
probability distributions. This idea forms the basis of the stochastic modeling.
34
Stochastic Modeling
represent the modeler’s knowledge (or lack of it) regarding the system and its
parameterization.
35
Stochastic Modeling
36
Stochastic Modeling
37
Stochastic Modeling
Y = G ( x, d ) (3.1)
Fig. 3.2 gives a graphic representation of the Monte Carlo frame, with the
preexisting model linking the output variables, Y, to a number of uncertain and
fixed input parameters, x and d.
38
Results and Discussion
The two example cases given in Chapter 2—UBD well planning and wellbore
stability analyses—will be used to demonstrate the applications of the
stochastic modeling.
The probability distributions used in the stochastic simulations are selected
such that they are consistent with the range of input data they are modeling.
The triangular distribution is represented by T (a, c, b), the uniform distribution
is represented by U (a, b), and the normal distribution by N (µ, σ). Input
uncertainties, which are expressed in percentages, quantitatively indicate how
much the mean differ from the extremes values. However, they are only
reasonable guesses based on field data and values quoted in the literature. There
is also a room for further improvements.
39
Results and Discussion
2013a). The novel approach describes how uncertainty can be propagated from
accessible input data to output variables.
The formation considered in the work is assumed homogeneous and isotropic.
The model (Eq. 2.1), non-penetrating Kirsch solution for wellbore fracturing,
and Mohr-Coulomb Shear-Failure Criterion for collapse (Eq. 2.2) are defined
as the preexisting models for the stochastic WSA.
Fig. 4.1 presents a general procedure for stochastic wellbore stability analyses.
The functions, y = f (v, w, x) and z = f (v, w), are the base models. The first step
involves the random sampling of input variables. Then the random numbers are
applied in the base models, to generate single-point outputs. This process is
repeated for n times and the output histograms are constructed. If the output
data are not realistic, the inputs are redefined or the models are calibrated
against offset data.
40
Results and Discussion
Deterministic Prediction
Let σH = 1.8 sg; σh = 1.5 sg; Po = 1.05 sg; α = 30°; τo = 0.5 sg.
Applying the input data in Eqs. 2.1 and 2.2, the wellbore fracture gradient is
1.65 sg, and the collapse gradient is 1.07 sg.
In this case, the single-point estimates of the geopressures will give too
optimistic a drilling window. This may lead to drilling problems. However, it
is impossible to assess the associated risk and uncertainty based on these fixed
input data. The only way to achieve this is by running stochastic simulations.
Stochastic Prediction
Case A. Table 4.1 presents the input parameters (now random variables) with
assumed uncertainties. All the inputs are assigned a triangular distribution, as
shown in Table 4.2.
41
Results and Discussion
The input distributions are applied in the deterministic stability models. After
600,000 Monte Carlo trials, the fracture and collapse pressure distributions are
generated. The simulation time is approximately 46 s.
Fig. 4.2 gives the geopressure distributions.
0.06
Fracture pressure
Collapse pressure
0.05
0.04
Probability
0.03
0.02
0.01
0
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
Pressure Gradient , sg
Figure 4.2—Fracture and collapse pressure distributions. Number of Monte Carlo trials
= 600,000.
42
Results and Discussion
Case B. Table 4.3 gives the stress bounds for the three types of in-situ stress
state described in Anderson (2012). Here, a normal fault stress state is assumed,
based on the magnitudes of the stress data.
Table 4.4 gives the redefined input-parameter distributions.
The stochastic simulations are repeated with the redefined input distributions.
Fig. 4.3 presents the resulting fracture and collapse pressure distributions after
running 600,000 Monte Carlo trials.
43
Results and Discussion
0.06
Fracture pressure
Collapse pressure
0.05
0.04
Probability
0.03
0.02
0.01
0
1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1
Pressure Gradient, sg
Figure 4.3—Fracture and collapse pressure distributions. Number of Monte Carlo trials
= 600,000.
There is no crossing of the stability curves this time. Therefore, the stochastic
models have predicted solutions that are more realistic.
Tables 4.5 and 4.6 compare the cumulative uncertainties in the stochastic
geopressure prognoses for the Cases A and B respectively.
Pressure
Data Statistics Fracture Collapse
P10 (sg) 1.43 0.91
P50 (sg) 1.66 1.07
P90 (sg) 1.88 1.23
Mean (sg) 1.66 1.07
Standard deviation (sg) 0.174 0.123
Cumulative uncertainty (%) ±38 -44/+45
Cumulative uncertainty = (max/min value – expected value)/expected value
44
Results and Discussion
Pressure
Data Statistics Fracture Collapse
P10 (sg) 1.81 1.42
P50 (sg) 1.86 1.48
P90 (sg) 1.91 1.55
Mean (sg) 1.86 1.48
Standard deviation (sg) 0.0397 0.0503
Cumulative Uncertainty (%) ±8 ±12
Cumulative uncertainty = (max/min value – expected value)/expected value
45
Results and Discussion
Fig 4.4 shows the plots of the geopressure distributions for the cases.
46
Results and Discussion
0.06
Base Case
0.05 Case 3
Case 2
0.04 Case 1
Probability
0.03
0.02
0.01
0
1.7 1.75 1.8 1.85 1.9 1.95 2 2.05 2.1
Fracture Pressure Gradient, sg
0.06
Base Case
0.05 Case 4
Case 5
0.04 Case 3
Probability
Case 2
0.03 Case 1
0.02
0.01
0
1.3 1.35 1.4 1.45 1.5 1.55 1.6 1.65 1.7 1.75
Collapse Pressure Gradient, sg
Figure 4.4—Monte Carlo sensitivity Analyses. The plots show the pressure distributions
for different cases. Top: fracture pressure; Bottom: collapse pressure.
From Fig. 4.4 (Top), the Case 2 is the fracture pressure distribution when the
only uncertain input is the minimum horizontal stress. This distribution has the
highest data spread compared with the Cases 1 and 3. Therefore, the minimum
horizontal stress is the most significant input parameter responsible for the
variability in the critical fracture pressure prediction (Base Case). The second
most influential parameter is the pore pressure (Case 3).
For the collapse pressure, the cohesive rock strength (Case 4) proves to be the
most influential parameter in the critical collapse pressure prognosis. This is
47
Results and Discussion
48
Results and Discussion
• Fluid-injection rates: liquid rate = 1400 L/min (0.0233 m3/s); gas rate
ranges from zero to 60 m3/min (1 m3/s)
• Liquid viscosity, µ l , = 1 cp (0.001 Pa.s); gas viscosity, µ g = 0.01 cp
(0.00001 Pa.s)
• The liquid and gas volume rates are measured at standard surface
conditions.
• The reservoir fluid is a gas with the same properties as the injected gas.
The average productivity index, PI, is 2.83E-08 m3/s/Pa.
Uncertain Input Parameters. Some of the model input parameters that are
subject to randomness include gas-slip parameters, C0 and Vd, parameter related
to friction factor calculation, fc, pore pressure, Po, and collapse pressure, Pcoll.
There are also uncertainties associated with choke operability and reservoir
fluid influx. They are implemented with parameters Pcho and PI. Other inputs
are considered fixed parameters.
The parameter, fc, is included in Eq. 2.10 as:
∗
∆Pf ∆Pf
= fc (4.1)
∆z ∆z
In general, the distributions should be chosen such that they reflect the physics
of the problem. The focus here is to demonstrate a stochastic methodology for
the UBD well planning. Investigating the goodness of fit of the input
distributions will be a discussion for future work.
Triangular distributions are assigned to the gas-slip parameters and fc. As
indicated in Adams et al. (1993) and Nilsen et al. (2001), all pressure
parameters are modeled with normal distributions.
Table 4.9 presents the random inputs with their assumed uncertainties and
probability distributions.
49
Results and Discussion
All simulations are run in MATLAB. First, the deterministic BHP prediction
will be presented as a base case. In this scenario, the fixed values of the input
parameters are used in the simulations.
220
Pore pressure
210 Bottomhole pressure
Collapse pressure
200
190
Pressure, bar
180
170
160
150
140
0 10 20 30 40 50 60
Gas-Injection Rate, m3/min
Figure 4.5—Annular bottomhole pressure versus gas-injection rates. Liquid rate = 1400 L/min.
Fig 4.5 shows that the predicted BHP falls within the UBD operational window
for the injection-gas rate greater than 10 m3/min. During unloading, the BHP
decreases due to the decrease in the mixture density of the multiphase fluids.
The well friction builds up as GVF increases. The system, according to Saponja
(1998), may eventually reach an optimum point where the reduced hydrostatic
50
Results and Discussion
Stochastic Results. The uncertain and fixed inputs are applied in the Monte
Carlo frame described in Fig. 3.2, with the deterministic model as the basis. For
each gas-injection rate, the stochastic model predicts BHP that follows a
probability distribution. The criterion for convergence of the results is by visual
inspection. After repeating 100,000 Monte Carlo trials for several times, the
resulting histograms appear nearly identical. This number is good enough for
convergence in the present case. For a more complex hydraulic model, the
Cramér-von Mises Criterion described in Anderson (1962) can be used as a
measure for testing the result convergence.
The P10, P50, and P10 of the BHP distribution are plotted against the gas-
injection rate. The results are the three BHP bands—the lower limit, the most
likely, and the upper limit—presented in Fig. 4.6. It takes approximately 11
hours to run the entire simulations.
51
Results and Discussion
220
P10-PP
P50-PP
210 P90-PP
P10-BHP
P50-BHP
200
P90-BHP
P10-CP
190 P50-CP
Pressure, bar
P90-CP
180
170
160
150
140
0 10 20 30 40 50 60
Gas-Injection Rate, m3/min
Figure 4.6—Annular BHP distributions at different gas-injection rates. Liquid rate = 1400
L/min; number of Monte Carlo trials = 100,000; PP = pore pressure; CP = collapse
pressure.
Fig. 4.6 shows that uncertainty in the prediction of the BHP increases with the
annular gas volume. This effect is not evident from the deterministic results. In
addition, the UBD operating window is further limited by the extreme values
of the pore and collapse pressures, that is, P10-PP and P90-CP bands. The
probabilistic results indicate that some BHP values fall outside this window.
Again, this is not the case with the deterministic prediction.
To maintain an underbalanced mode, considering the wellbore stability, the
gas-injection rates in the range of 30–40 m3/s is recommended for this case.
This shows that a probabilistic approach can give a different
recommendation—regarding the UBD operational window—from a pure
deterministic method. While selecting the optimum gas-injection rate, it is also
important that cutting transport and downhole motor requirements be taken into
consideration.
52
Results and Discussion
UBD operating limits. During the simulation, the numerical counter records the
number of times there is a pressure overbalance or collapse condition. In the
end, the ratio of this number to the number of Monte Carlo trials gives the
probability of overbalance or collapse.
This can be demonstrated with the three gas-injection gas rates—15, 30, and 40
m3/min. The number of Monte Carlo trials is 100,000.
Figs. 4.7 through 4.9 present the results of the simulations.
0.09
Pore pressure
0.08 Bottomhole pressure
Collapse pressure
0.07
0.06
Probability
0.05
0.04
0.03
0.02
0.01
0
120 140 160 180 200 220 240
Pressure Distribution, bar
53
Results and Discussion
0.09
Pore pressure
0.08 Bottomhole pressure
Collapse pressure
0.07
0.06
Probability
0.05
0.04
0.03
0.02
0.01
0
120 140 160 180 200 220 240
Pressure Distribution, bar
0.09
Pore pressure
0.08 Bottomhole pressure
Collapse pressure
0.07
0.06
Probability
0.05
0.04
0.03
0.02
0.01
0
120 140 160 180 200 220 240
Pressure Distribution, bar
54
Results and Discussion
Fig. 4.7 shows that the chance of having pressure overbalance is high at lower
gas rates, as indicated by the probability. This likelihood decreases as the gas-
injection rate is doubled (Fig 4.8).
By contrast, the probability of collapse increases as more gas is pumped into
the annulus. There is a higher chance of having wellbore collapse when the gas-
injection rate is 40 m3/s (Fig. 4.9), compared with the zero probability when the
gas-injection rate is 15 m3/s.
Table 4.10 presents a summary of the simulation results.
The probabilistic approach not only provides a means for the uncertainty
propagation, it also ensures that the annular well pressure remains
underbalanced in the target section.
Overbalance (OB)
For a conventional overbalanced drilling, Po < Pw. Eq. 2.2 defines the critical
well pressure that will cause wellbore collapse as:
1
Pwc = (3σ H − σ h )(1 − sin α ) + Po sin α − τ o cos α
2
55
Results and Discussion
Underbalance (UB)
Permeable Rocks. In underbalanced drilling, Po > Pw. For permeable rocks
such as sandstones and carbonates, the formation fluid will flow into the
wellbore when Pw = Po.
Let Pw = Po. Replacing Po with Pw, that is Pwc, in the collapse model (Eq. 2.2),
the resulting equation is given by Eq. 4.2. The equation gives the critical well
pressure that will initiate wellbore collapse during underbalanced operations.
During UBD, the reservoir flow will stabilize the wellbore at Pw = P(UBD)wc.
If the well pressure falls under the critical value, the collapse of the wellbore
wall initiates and eventually causes its failure (Aadnøy and Reza 2010).
Impermeable Rocks. Impervious rocks like shales may not allow inward flow
of the formation fluid into the wellbore, even if the well pressure is less than
the pore pressure. Therefore, Eq. 2.2 defines the critical well pressure that will
cause wellbore collapse in this situation.
Figs. 4.10 and 4.11 respectively give the boundary conditions for wellbore
collapse, for permeable and impermeable rocks. The figures also indicate the
reservoir conditions for conventional drilling and UBD.
56
Results and Discussion
Figure 4.10—Boundary conditions for wellbore collapse, for permeable rocks such as
sandstones and carbonates.
Figure 4.11—Boundary conditions for wellbore collapse, for impermeable rocks like
shales.
57
Results and Discussion
Impermeable
OB Non-penetrating Eq. 2.2
UB Non-penetrating Eq. 2.2
Input Parameter
Case C0 Vd fc
Base case C0 = T(1.01,1.2,1.22) Vd = T(0.35,0.5,0.55) fc = T(0.9,1.0,1.1)
BHP-C0 C0 = T(1.01,1.2,1.22) 0.50 1.0
BHP-Vd 1.20 Vd = T(0.35,0.5,0.55) 1.0
BHP-fc 1.20 0.50 fc = T(0.9,1.0,1.1)
Det case 1.20 0.50 1.0
PI = 244.5 m3/day/bar; Po = 200 bar; Pcoll = 160 bar; Pcho (surface backpressure) = 6
bar
58
Results and Discussion
220
210
200
190 Pore
Pressure, bar
Det case
Base case
180
BHP-C0
BHP-Vd
170 BHP-fc
Collapse
160
150
140
0 10 20 30 40 50 60
Gas-Injection Rate, m3/min
Figure 4.12—Sensitivity analysis. Each plot of the annular pressure versus gas rates
depicts the contribution of the respective input to the uncertainty in the BHP prediction.
From the plots above, it can be seen that the BHP-fc profile overlaps with the
deterministic case (Det case). This clearly shows that fc is the least significant
input parameter in this simulation scenario. The BHP-Vd profile nearly
approximates the Det case. Hence, the gas drift velocity relative to the liquid is
the second least significant input.
On the contrary, BHP-C0 profile approaches the Base case. Therefore, C0 is the
most influential input parameter largely responsible for the cumulative
uncertainty in the BHP prediction.
The UBD multiphase circulating system represents a highly dynamic flow
situation. Because the values of C0 and Vd depend on the prevailing two-phase
flow pattern, it is pertinent that comprehensive mechanistic models (Lage and
Time 2000) be included in future studies. This also requires the use of more
realistic closure laws and inclusion of wellbore temperature effects. A similar
sensitivity analysis can then be carried out, to determine the most dominating
parameters with respect to the uncertainty in the BHP.
59
A Transient Model for Well Flows
5.1 Introduction
Hybrid flux-splitting schemes for solving the compressible Euler and Navier-
Stokes equations are a new development in upwind schemes. The schemes
combine the efficiency of flux-vector splitting (FVS) schemes and the accuracy
of flux-difference splitting schemes (Evje and Fjelde 2003). An example of
such schemes is the Advection Upstream Splitting Method (AUSM) described
in Liou and Steffen (1993). Liou (1996) extended the AUSM scheme to
AUSM+.
The attractive attributes of AUSM in comparison with other upwind schemes
are accuracy, efficiency and robustness. The AUSM+ scheme has demonstrated
that it can produce exact resolution of contact and shock discontinuities and
preserve positivity property of some primitive variables. It can also be easily
extended to other hyperbolic systems due to its simplicity.
Based on the AUSM, Evje and Fjelde (2003) proposed a hybrid scheme
termed ‘AUSMV’ that can be used to solve a hyperbolic system of conservation
laws. The two-phase model was a sequel to the scheme proposed in Evje and
Fjelde (2002). The model has a generic form:
∂ tW + ∂ x F ( x , W ) = G ( x , W ) (5.1)
60
A Transient Model for Well Flows
∂ ∂
(α l ρ l ) + (α l ρ l v l ) = Γl (5.2)
∂t ∂x
∂ ∂
(α g ρ g ) + (α g ρ g v g ) = Γg (5.3)
∂t ∂x
∂ ∂
(α l ρ l v l + α g ρ g v g ) + (α l ρ l v l2 + α g ρ gν g2 + p ) = − q (5.4)
∂t ∂x
The unknowns are the liquid and gas densities (ρ l , ρ g ), the phase volume
fractions (α l , α g ), the phase velocities (v l , v g ), and well pressure, p.
Assuming that there is no mass exchange between the liquid and gas, the phase
mass transfer coefficients, Г l and Г g , become zero.
61
A Transient Model for Well Flows
reservoir fluid influx, unloading, and connection. This will be illustrated with
DGD and UBD example cases.
Wellbore and Fluid Data. The vertical well considered in this example is 2000
m deep. The drill string consists of a drill pipe whose outer diameter is 5'' (0.127
m). There is no BHA. The inner annular diameter is 8 ½'' (0.216 m).
The marine riser is 1000 m long, with an inner diameter of 20'' (0.508 m). Thus,
an area discontinuity is expected because different flow areas will exist between
the drill pipe and riser and between the drill pipe and the annulus.
The liquid sonic velocity, a l = 1500 m/s, and viscosity, µ l = 5×10–2 Pa.s. The
reservoir fluid is a gas with viscosity, µ g = 5×10–6 Pa.s and the sonic velocity,
a g = 316 m/s. The gas-slip parameters, C 0 = 1.1 and V d = 0.5 m/s. The fluid
densities vary with the well pressure. All fluid volumes are measured at
standard surface conditions.
Effect of Gas Kick. The kick volumes considered in the numerical simulations
are 0.5 and 1 m3. At the onset, the liquid-injection rate is 600 L/min. From 300
to 790 s, the mud level in the riser is gradually lowered at the suction rate of
4500 L/min. The reservoir gas flows into the well between 1000 and 1070 s. At
1100 s, the kick is circulated out of the well with a suction rate of 600 L/min.
This ensures that the volume of the drilling fluid in the annulus is maintained.
Fig 5.1 represents a simplified well for the DGD system.
62
A Transient Model for Well Flows
Figure 5.1—A simplified well for the DGD system. The kick migrates upward and is
separated in the riser
Fig. 5.2 presents annular BHP development with time, for the two kick
volumes.
195
3
Kick volume = 0.5 m
3
Kick volume = 1 m
190
185
BHP, bar
180
175
170
100 500 1000 1500 2000 2500 3000 3500 4000
Time, s
63
A Transient Model for Well Flows
The figure shows that there is a sudden increase in the annular BHP at 1000 s,
when the gas enters the well. This is probably due to increase in the well
friction. It is also observed that the pressure decreases, and then increases,
before it stabilizes after 2500 s. The effect of the gas kick on the BHP in the
well is noticeable due to area-to-volume effect. The kick will have a lower area-
to-volume ratio in the well than in the riser, but the mud-level increase in the
riser will correspond to the kick volume. Thus, the height of the kick is larger
in the well than when it enters the riser with a larger flow area.
Though the well pressure profiles presented in Fig. 5.2 exhibit similar trend,
the pressure drop for 1 m3 kick is larger. This is because the area-to-volume
effect becomes more pronounced as the kick volume becomes larger. The well,
however, is under static conditions when taking the kick in each case.
In addition, the BHPs at 1000 s and at the end of the simulation are the same
for the two cases. This is because the mass of drilling fluid is conserved.
Figs. 5.3a through 5.3c present snapshots of the gas kick at different time
intervals, as it migrates up the well.
t = 1000 s t = 1500 s
0 0
-200 -200
-400 -400
-600 -600
-800 -800
Depth, m
Depth, m
-1000 -1000
-1200 -1200
-1400 -1400
-1600 -1600
-1800 -1800
-2000 -2000
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
GVF GVF
Figure 5.3a—Depth versus gas volume fraction at t = 1000 s and t = 1500 s. Kick volume
= 0.5 m3, number of cells = 25.
64
A Transient Model for Well Flows
t = 2000 s t = 2500 s
0 0
-200 -200
-400 -400
-600 -600
-800 -800
Depth, m
Depth, m
-1000 -1000
-1200 -1200
-1400 -1400
-1600 -1600
-1800 -1800
-2000 -2000
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
GVF GVF
Figure 5.3b—Depth versus gas volume fraction at t = 2000 s and t = 2500 s. Kick volume
= 0.5 m3, number of cells = 25.
t = 3000 s t = 4000 s
0 0
-200 -200
-400 -400
-600 -600
-800 -800
Depth, m
Depth, m
-1000 -1000
-1200 -1200
-1400 -1400
-1600 -1600
-1800 -1800
-2000 -2000
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
GVF GVF
Figure 5.3c—Depth versus gas volume fraction at t = 3000 s and t = 4000 s. Kick volume
= 0.5 m3, number of cells = 25.
In Fig. 5.3b (Left), the kick volume appears to shrink, even though the mass is
the same. This is caused by the area change, as the kick enters the riser with a
much larger cross-sectional area. At 4000 s, a stable air/drilling fluid interface
65
A Transient Model for Well Flows
is restored as shown in Fig.5.3c (Right). This time, the gas kick has been
circulated out of the system.
200
N = 25 cells
N = 50 cells
190
BHP, bar
185
180
175
170
100 500 1 000 1 500 2 000 2 500 3 000 3 500 4 000
Time, s
Fig. 5.4—Annular bottomhole pressure versus time. The results show the effect of
discretization error on the BHP.
The difference in the results shows that accuracy in the prediction of BHP
improves when a higher number of grid cells is used in the simulation.
Increasing the cells from 25 to 100, results in a pressure difference of 2 bar at
4000 s. In addition, as the grid is refined, improvement in accuracy decreases
for each refinement. This reflects the convergence of the results. Therefore, a
large increase in the BHP may not be expected, if the number of cells is set to
400. A cell size smaller than 20 m, or 200 cells, is sufficient for this case.
66
A Transient Model for Well Flows
However, increasing the number of cells to 200 will cause the simulation to go
very slow. The convergence will also take much longer time than necessary.
Alternatively, the AUSMV scheme can be extended to a higher order scheme
with slope limiter techniques (Fjelde et al. 2003; Lorentzen and Fjelde 2005).
In this case, a less computational effort will be required, and convergence
towards a solution will be faster with few cells.
67
A Transient Model for Well Flows
200
180
160
BHP, bar
140
120
100
80
200 500 1000 1500 2000 2500 3000
Time, s
Figure 5.5—The behavior of the annular bottomhole pressure at different simulation time
intervals. Choke pressure = 3 bar.
After unloading, the BHP decreases but stabilizes at 122 bar after 1000 s.
Before this point, hydrostatic effects dominate the flow conditions. The highly
productive fractured zone is drilled at 1200 s. The reservoir gas flows into the
well and the annular gas volume increases correspondingly. Then the BHP
suddenly increases due to the increase in the annular friction. The flow will
approach an optimal point (Saponja 1998), where friction will begin to
dominate the pressure conditions.
The BHP gradually decreases as the reservoir gas migrates upward, and the
system eventually reaches new steady-state conditions. Thus, the region beyond
the point (1200 s, 122 bar) is more friction dominated. This flow condition is
more beneficial in controlling the gas influx.
Fig. 5.6 shows that the frictional and hydrostatic components of the BHP
approach the optimal point as the system stabilizes.
68
A Transient Model for Well Flows
200
Friction
180 Hydrostatic Pressure
160
140
120
Pressure, bar
100
80
60
40
20
0
100 500 1000 1500 2000 2500 3000
Time, s
The BHP (Fig. 5.5) is the sum of the frictional and hydrostatic components
presented in Fig. 5.6. The visual representation of the pressure components in
the transient simulations is invaluable because it shows when the pressure
condition is hydrostatic or friction dominated.
69
A Transient Model for Well Flows
Open connection
Variable connection
0.6
Closed connection
0.5
0.4
Gas Influx, kg/s
0.3
0.2
0.1
-0.1
0 500 1000 1500 2000 2500 3000
Time, s
From Fig. 5.7, it can be observed that the gas influx volume is smallest for the
closed connection and largest for the open connection. The variable choke
pressure connection has the least fluctuations in the mass influx and is the most
advantageous in this example case. This connection procedure also gives the
least fluctuation in the BHP as presented in Paper V (Fig. 14).
Again, the AUSMV scheme has shown that it is capable of handling the well
dynamics associated with pipe connection.
5.5 Inference
Based on the simulation results, the AUSMV scheme has demonstrated that it
can handle highly dynamic flow situations associated with DGD and UBD. For
the DGD system, the scheme gives insights into the annular BHP behavior and
the dynamics of well control.
For the UBD system, the AUSMV scheme is used to study unloading scenario.
The scheme is able to predict when steady-state conditions can be achieved. It
also shows that it can handle different connection procedures.
70
A Transient Model for Well Flows
The transient model can also be used to investigate the effect of reservoir inflow
on the annular BHP. This is important for situations where there are
possibilities to encounter highly productive fractured zones.
In addition, the scheme can be used for training and educational purposes.
However, it is necessary to consider field data in future works, to prove the
practical relevance of the theoretical model. Then, the model predictions can be
compared with simulation and experimental results from previous studies.
71
Overview of the Research Papers
The first paper proposes a novel stochastic approach for predicting UBD
operational window. Because underbalanced operations involve a highly
dynamic flow situation, one inconsistency or the other affects the annular BHP
prediction. This can be attributed to modeling uncertainties, which result from
the modeling process and input parameters.
This work only considers uncertainties in the flow-dependent parameters
such as gas-slip parameters and a factor related to frictional pressure loss
gradient. A simple steady-state two-phase model is used to study the UBD
system. Probability distributions are assigned to the uncertain inputs, and the
uncertainty propagation is by means of Monte Carlo simulation.
Compared with the deterministic method, the simulation results show that
the stochastic approach predicts a more realistic operational window. The
stochastic model also demonstrates that it has the capability to predict when the
BHP is outside the UBD operational limits with a certain probability.
A sensitivity analysis is used to rank the uncertain inputs according to their
relative importance. This is based on their individual contributions to the
cumulative uncertainty in the BHP. The gas-slip parameters are the most
dominant uncertain inputs in this case.
In a way, the stochastic approach can help to reduce uncertainty in the BHP
prediction. As a result, well planners can make better decision as they select
critical operating parameters during underbalanced drilling.
72
Overview of the Research Papers
parameters—on the BHP prediction. The paper also discusses the uncertainty
associated with two-phase flow patterns. However, slug flow is considered as
the most dominant flow pattern in this example.
The paper investigates typical fracture and collapse models with respect to
uncertainties in the input parameters. The inputs considered uncertain are in-
situ stresses, pore pressure, cohesive rock strength and angle of internal friction.
Inaccuracy of the input data may be attributed to noise in measurements and
measurement errors. There is a human error as well.
In the stochastic analyses, non-penetrating Kirsch solution for wellbore
fracturing and Mohr-Coulomb Shear Failure Criterion for collapse are defined
as the base models. The underlying principle is to assign probability
distributions to the uncertain input parameters. Then the input uncertainties are
propagated by means of Monte Carlo simulation. Traditionally, the
deterministic stability models predict single-point estimates of critical
fracturing and collapse pressures. In the uncertainty approach, exact solutions
become probability distributions.
The sensitivity analyses indicate that the minimum horizontal stress is the most
significant random input that influences the critical fracture pressure prediction.
For the collapse pressure prediction, the rock cohesive strength is the most
dominant input parameter.
A better confidence level can be established because the risks associated
with the geopressure prognoses are fully assessed. This may lead to an
improved mud window prediction and casing design. Therefore, the approach
can improve well planning and help to reduce drilling problems such as lost
circulation and stuck pipe.
73
Overview of the Research Papers
This paper presents a simple and robust transient model that can handle highly
dynamic flow systems. The model, AUSMV scheme, is developed for
academic and training purposes. Students can use the model to investigate
different flow scenarios and pressure control challenges encountered in general
well control, managed pressure drilling and underbalanced drilling.
The capability of this model to handle highly changing flow scenarios is
demonstrated with the examples taken from managed pressure drilling and
underbalanced operations.
In the DGD example case, the AUSMV scheme demonstrates that it can
simulate a dynamic well control scenario. It is also used in gaining insight into
the effects of flow area change and numerical discretization on the annular BHP
development.
For the UBD case, the intent is to show the difference between hydrostatic-
dominated and friction-dominated flow conditions. The simulation results show
how the pressure components can be visualized. In addition, the scheme proves
that it can be used to study the dynamic effect related to different connection
procedures.
74
Overview of the Research Papers
Paper V was revised and submitted to SPE Drilling & Completion, with a
revised title:
On the AUSMV Scheme: A Simple Transient Flow Model for MPD and UBD
Applications.
75
Conclusion and Further Work
7.1 Conclusion
This work presents a novel stochastic approach for drilling and well planning.
Some vital operational concerns have motivated the proposition of the
methodology.
Mathematical models are but perfect tools for investigating subsurface
phenomena, processes and so forth. This also applies to hydraulic flow models
and wellbore stability models. Generally, modeling uncertainties may result
from uncertainties in input parameters. The input uncertainties may be due to
noise in measurements and measurement errors. There is a human error as well.
In some situations, the input data are scarce and inadequate. In addition, there
is uncertainty resulting from modeling processes, because models only
approximate physical phenomena.
76
Conclusion and Further Work
Judging from the foregoing discussion, the stochastic modeling may be a future
and cost-efficient drilling solution for exploiting depleted reservoirs, fractured
carbonate formations and deepwater targets. In essence, the industry will
always use deterministic and scenario methods. Yet their predictions should be
supplemented with the results from stochastic simulations.
77
Conclusion and Further Work
Generally, the input distributions should be chosen such that they reflect the
physics of the problem. No formal test is performed in this work, to determine
the goodness of fit of the input distributions and convergence of the model
results. The focus here is to present a simple stochastic approach for drilling
and well planning.
Investigating the goodness of fit of the input distributions should be discussed
in a future study. This can be performed with Cramér-von Mises Criterion
(Anderson 1962) or any other formal tests. Therefore, more literature and
experimental data should be considered, to ascertain what could be the most
appropriate distributions for the uncertain input parameters.
Finally, one can integrate the steady-state flow model with the geopressure
models and investigate the integrated system with the probabilistic approach.
78
References
References
Aadnøy, B.S. 1988. Inversion Technique to Determine the In-Situ Stress Field
from Fracturing Data. Paper SPE 18023, presented at the 63rd SPE Annual
Technical Conference and Exhibition, Houston, Texas, USA, 2–5 October.
http://dx.doi.org/10.2118/18023-MS.
Aadnøy, B.S. 2010. Modern Well Design, second edition. Leiden, The
Netherlands: CRC Press/Balkema.
Aadnøy, B.S., Cooper, I., Miska, S.Z., Mitchell, R.F, and Payne, M.L. 2009.
Advanced Drilling and Well Technology. Richardson, Texas, USA: Society of
Petroleum Engineers.
Aadnøy, B.S. and Hansen, A.K. 2005. Bounds on In-Situ Stress Magnitudes
Improve Wellbore Stability Analyses. SPE Journal 10 (2): 115–120.
http://dx.doi.org/10.2118/87223-PA.
Adams, A.J, Parfitt, S.H.L., Reeves, T.B., Thorogood, J.L. 1993. Casing
System Risk Analysis Using Structural Reliability. Paper SPE 25693, presented
79
References
Addis, M.A., Hanssen, T.H, Yassir, N., Willoughby, D.R, and Enever, J. 1998.
A Comparison of Leak-off Test and Extended Leak-off Test Data for Stress
Estimation. Paper SPE 47235, presented at the SPE/ISRM Eurock 98,
Trondheim, Norway, 8–10 July. http://dx.doi.org/10.2118/47235-MS.
Arild, Ø., Fjelde, K.K., Merlo, A., and Løberg, T. 2008. BlowFlow — A New
Tool within Blowout Risk Management. Paper SPE 114568, presented at
IADC/SPE Asia Pacific Drilling Technology Conference and Exhibition,
Jakarta, Indonesia, 25–27 August. http://dx.doi.org/10.2118/114568-MS.
Arild, Ø., Ford, E.P., Løberg, T., and Baringbing, J.W.T. 2009. KickRisk — A
Well Specific Approach to the Quantification of Well Control Risks. Paper SPE
124024, presented at the SPE Asia Pacific Oil and Gas Conference and
Exhibition, Jakarta, Indonesia, 4–6 August. http://dx.doi.org/10.2118/124024-
MS.
Bennion, D.B., Thomas, F.B., Bietz, R.F., and Bennion, W.D. 1998.
Underbalanced Drilling, Praises and Perils. SPE Drill & Compl 13(4): 214–
222. http://dx.doi.org/10.2118/52889-PA.
Bradley, W.B. 1979. Failure of Inclined Boreholes. J. Energy Resources
Technology 101 (4): 232–239. http://dx.doi.org/10.1115/1.3446925.
80
References
Bratvold, R.B. and Begg, S.H., 2010. Making Good Decisions. Richardson,
Texas, USA: Society of Petroleum Engineers.
Brown, J.D, Urvant, V.V., Thorogood, J.L., and Rolland, N.L. 2007.
Deployment of a Riserless Mud-Recovery System Offshore Sakhalin Island.
Paper SPE 105212, presented at SPE/IADC Drilling Conference, Amsterdam,
The Netherlands, 20–22 February. http://dx.doi.org/10.2118/105212-MS.
Caetano, E.F. 1986. Upward Two-Phase Flow through an Annulus. PhD thesis,
University of Tulsa, USA.
Cohen, J.H. and Deskins, G. 2006. Use of lightweight Solid additives to Reduce
the Weight of Drilling Fluid in the Riser. Paper SPE 99174, presented at the
IADC/SPE Drilling Conference, Miami, Florida, USA, 21–23
February. http://dx.doi.org/10.2118/99174-MS.
81
References
Eggemeyer, J.C., Akins, M.E., Brainard, R.R., et al. 2001. SubSea MudLift
Drilling: Design and Implementation of a Dual-Gradient Drilling System. Paper
SPE 71359, presented at the 2001 SPE Annual Technical Conference and
Exhibition, New Orleans, Louisiana, USA, 30 September–3 October.
http://dx.doi.org/10.2118/71359-MS.
Evje, S. and Fjelde, K.K. 2002. Hybrid Flux-Splitting Schemes for a Two-
Phase Flow Model. J. Computational Physics. 175 (2) 674–701.
http://dx.doi.org/10.1006/jcph.2001.6962.
Evje, S. and Fjelde, K.K. 2003. On a Rough AUSM Scheme for a One-
Dimensional Two-Phase Model. Computers & Fluids J. 32 (10): 1497–1530.
http://dx.doi.org/10.1016/S0045-7930(02)00113-5.
Falk, K., Fossli, B., Lagerberg, C., Handal, A., and Sangesland, S. 2011. Well
Control When Drilling with a Partly Evacuated Marine Drilling Riser. Paper
SPE 143095, presented at the IADC/SPE Managed Pressure Drilling and
Underbalanced Operations Conference and Exhibition, Denver, Colorado,
USA, 5–6 April. http://dx.doi.org/10.2118/143095-MS.
Forrest, N. and Bailey, T. 2001. Subsea Equipment for Deep Water Drilling
Using Dual Gradient Mud System. Paper SPE 67707, presented at the
SPE/IADC Drilling Conference, Amsterdam, The Netherlands, 27 February–1
March. http://dx.doi.org/10.2118/67707-MS.
82
References
Gerald, C.F and Wheatley, P.O. 2004. Applied Numerical Analysis, seventh
edition. Boston, Massachusetts: Addison-Wesley Publishing Company.
Gough, G., Talbot, S., and Moos H. 2011. Offshore Underbalanced Drilling
with a Closed-Loop Circulating System. Paper SPE 143096, presented at the
IADC/SPE Managed Pressure Drilling and Underbalanced Operations
Conference and Exhibition, Denver, Colorado, USA, 5–6 April.
http://dx.doi.org/10.2118/143096-MS.
Handal, A. 2011. Gas Influx Handling for Dual Gradient Drilling. PhD thesis,
Norwegian University of Science and Technology, Trondheim, Norway.
Harten, A., Lax, P.D, van Leer, B. 1983. On Upstream Differencing and
Godunov Schemes for Hyperbolic Conservation Laws. SIAM Review 25 (1):
35–61. http://dx.doi.org/10.1137/1025002.
Kosh, G.S. and Link, R.F. 1980. Statistical Analysis of Geological data, volume
1. New York, USA: Dover Publications, Inc.
Lage, A.C.V.M. 2000. Two-Phase Flow Models and Experiments for Low-
Head and Underbalanced Drilling. PhD thesis, University of Stavanger,
Stavanger, Norway
Lage, A.C.V.M., Fjelde, K.K., and Time, R.W. 2003. Underbalanced Drilling
Dynamics: Two-Phase Flow Modeling and Experiments. SPE Journal 8 (1):
61–70 http://dx.doi.org/10.2118/62743-MS.
83
References
Liou, M.-S. 1996. A Sequel to AUSM: AUSM+. J. Comput. Phys. 129 (2):
364–382. http://dx.doi.org/10.1006/jcph.1996.0256.
Liou, M.-S. and Steffen, C.J. 1993. A New Flux Splitting Scheme. J. Comput
Phys. 107 (1): 23–34. http://dx.doi.org/10.1006/jcph.1993.1122.
Livescu, S., Durlofsky, L.J., Aziz, K., and Ginestra, J.C. 2009. A Fully-Coupled
Thermal Multiphase Wellbore Flow Model for Use in Reservoir Simulation. J.
Pet. Sci. Eng. 71 (2–3): 138–146.
http://dx.doi.org/10.1016/j.petrol.2009.11.022.
Lopes, C.A. and Bourgoyne, A.T. 1997. The Dual Density Riser Solution.
Paper SPE 37628, presented at the 1997 SPE/IADC Drilling Conference,
Amsterdam, The Netherlands, 4–6 March. http://dx.doi.org/10.2118/37628-
MS.
Løberg, T., Arild, Ø., Merlo, A., and D’Alesio, P. 2008. The How’s and Whys
of Probabilistic Well Cost Estimation. Paper SPE 114696, presented at the
IADC/SPE Asia Pacific Drilling Technology Conference and Exhibition,
Jakarta, Indonesia, 25–27 August. http://dx.doi.org/10.2118/114696-MS.
84
References
Murtha, J.A. 1997. Monte Carlo Simulation: Its Status and Future. J Pet. Tech.
49 (4) 361–373. http://dx.doi.org/10.2118/37932-JPT.
Nilsen, T., Sandøy, M., Rommetveit, R., and Guarneri, A. 2001. Risk-Based
Well Control Planning: The Integration of Random and Known Quantities in a
Computerized Risk management Tool. Paper SPE 68447, presented at the
SPE/ICoTA Coiled Tubing Roundtable, Houston, Texas, 7–8 March.
http://dx.doi.org/10.2118/68447-MS.
Osayande, N., Mammadov, E., Sephton, S., and Boucher, V. 2014. The
Successful Application of Automated Managed Pressure Drilling (MPD) to
Protect Cap Rock Integrity by Narrow margin Drilling in SAGD Wells. Paper
85
References
Rajabi, M.M., Toftevåg, K., Stave, R.S, and Ziegler, R. 2012. First Application
of EC-Drill in Ultra-Deepwater—Proven Subsea Managed Pressure Drilling
Method. Paper SPE 151100, presented at the SPE Deepwater Drilling and
Completions Conference, Galveston, Texas, USA, 20–21 June.
http://dx.doi.org/10.2118/151100-MS.
Rehm, B., Schubert, J., Haghshenas, A., Paknejad, A.S., and Hughes, J. eds.
Managed Pressure Drilling. Houston, Texas: Gulf Publishing Company.
Roe, P.L. 1980. The Use of the Riemann Problem in Finite Difference Schemes.
Lecture Notes in Physics vol. 141: 354–
359. http://link.springer.com/chapter/10.1007/3-540-10694-4_54.
86
References
Romate J.E. 1998. An Approximate Solver for a Two-Phase Flow Model with
Numerically Given Slip Relation. Computers & Fluids 27(4): 455–
477. http://dx.doi.org/10.1016/S0045-7930(97)00067-4.
Rommetveit, R., Fjelde, K.K., Aas, B, et al. 2003. HPHT Well Control; An
Integrated Approach. Paper OTC 15322, presented at the 1997 Offshore
Technology Conference, Houston, Texas, USA, 5–8 May.
http://dx.doi.org/10.4043/15322-MS.
Rommetveit, R., Fjelde, K.K., Frøyen, J., et al. 2004. Use of Dynamic Modeling
in Preparations for the Gullfaks C-5A Well. Paper SPE 91243, presented at
2004 SPE/IADC Underbalanced Technology Conference and Exhibition,
Houston, Texas, USA, 11-12 October. http://dx.doi.org/10.2118/91243-MS.
Saltelli, A., Chan, K., and Scott, E.M. eds. 2008. Sensitivity Analysis.
Chichester, UK: John Wiley & Sons, Ltd.
Schumacher, J.P., Dowell, J.D, Ribbeck, L.R, and Eggemeyer, J.C. 2001.
SubSea MudLift Drilling: Planning and Preparation for the First SubSea Field
87
References
Schumacher, J.P., Dowell, J.D., Ribbeck, L.R., and Eggemeyer, J.C. 2002.
Planning and Preparing for the First Subsea field Test of a Full-Scale Dual-
Gradient Drilling System. SPE Drill & Compl 17 (4): 194–
199. http://dx.doi.org/10.2118/80615-PA.
Stave, R., Farestveit, R., Høyland, S., Rochmann, P.O., and Rolland, N.L. 2005.
Demonstration and Quantification of a Riserless Dual Gradient System. Paper
OTC 17665, presented at the 2005 offshore Technology Conference, Houston
Texas, USA, 2–5 May. http://dx.doi.org/10.4043/17665-MS.
Udegbunam, J.E., Aadnøy, B.S., and Fjelde, K.K. 2013a. Paper SPE 166788,
presented at the SPE/IADC Middle East Drilling Technology Conference and
Exhibition, Dubai, 7–9 October. http://dx.doi.org/10.2118/166788-MS.
Udegbunam, J.E, Arild, Ø., Fjelde, K.K., Ford, E., and Lohne, H.P. 2013b.
Uncertainty-Based Approach for Predicting the Operating Window in UBO
Design. SPE Drill & Compl 28 (4): 326–337.
http://dx.doi.org/10.2118/164916-PA.
Udegbunam, J.E., Fjelde, K.K., Steinar, E., and Nygaard, G. 2014. A simple
Transient Flow Model for MPD and UBD Applications. Paper SPE 168960,
presented at the SPE/IADC Managed Pressure Drilling and Underbalanced
Operations Conference and Exhibition, Madrid, Spain, 8–9 April.
http://dx.doi.org/10.2118/168960-MS.
88
References
Walpole, R.E., Myers, R.H., Myers, S.L, and Ye, K. 2012. Probability &
Statistics for Engineers and Scientist, ninth edition. Boston, USA: Pearson
Education, Inc.
Williamson, H.S., Sawaryn, S.J., and Morrison, J.W. 2006. Monte Carlo
Techniques Applied to Well Forecasting: Some Pitfalls. SPE Drill & Compl 21
(3): 216–277. http://dx.doi.org/10.2118/89984-PA.
89