SPE 96395
Numerical Investigation of Horizontal Well Performance With Selective Completion
R. Kalita, SPE, and Y. Jalali, SPE, Schlumberger
Copyright 2005, Society of Petroleum Engineers
This paper was prepared for presentation at Offshore Europe 2005 held in Aberdeen,
Scotland, U.K., 69 September 2005.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in a proposal submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the Society of Petroleum Engineers is
prohibited. Permission to reproduce in print is restricted to a proposal of not more than 300
words; illustrations may not be copied. The proposal must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
Abstract
This study presents a numerical method of determining the
sandface completion profile of a horizontal well to achieve a
prescribed production objective. A technique is developed to
determine the completion profile that maximizes an objective
function (e.g., cumulative production) for a given well
operating program in a defined reservoir environment.
Sandface completion is modeled in terms of a connection
factor between the reservoir gridblocks adjacent to the well
and wellbore segments. Well performance is evaluated in
terms of cumulative production within specified operational
constraints (maximum rate, minimum bottomhole pressure).
The method uses a finite-difference technique to determine the
sensitivity of well performance to changes in the wellbore-
reservoir connection profile. The optimum connection profile
is determined by a gradient-based technique using the
sensitivity coefficients from the finite-difference method.
This approach can be used to either plan the sandface
completion during the well design stage, or to optimize the
completion after drilling based on measured petrophysical and
flow properties of the formation. Examples are given in the
paper covering both water- and gas-drive problems.
Introduction
Diagnosis of the cause of premature water-cut or excessive
gas-oil ratio in producing wells, and determination of
appropriate mitigation measures, is vital to field operations.
This is particularly important for horizontal wells, as these
wells are prone to early appearance of water or gas, due to the
extensive contact with the reservoir.
One possible cause of this behavior is the completion of the
well in the transition zone, where the saturation of the
displacing phase is mobile from initial production. This is a
risk in tight formations, particularly in oil-water systems,
where the transition zone can be large. Such problems can
also occur when open-hole logs are difficult to interpret or
when depth uncertainties exist in the formation. To cope with
this situation the basic options are to prorate the production of
the well or to sidetrack and recomplete the well.
When the well is completed in relatively clean and permeable
sand in the oil column, however, such behavior must be
attributed to adverse wellbore hydraulics. The pressure profile
in the wellbore exerts a non-uniform pull on the formation,
thus inducing an uneven sweep and local breakthrough of gas
or water. Due to the rate-dependent nature of this problem,
the remedy is to choke the production. A trade-off, therefore,
has to be made between reduction of gas or water production,
and corresponding loss of oil production.
When the well is completed in formations with bedding or
stratification, the natural dip or the high-angle trajectory of the
well can cause it to traverse multiple facies with contrasting
petrophysical and flow characteristics. Under these conditions,
a natural tendency exists for the preferential encroachment of
the displacing phases along the conductive strata, or
channeling of injection fluid from neighboring wells. The
remedy for such cases is to seal intervals of adverse fluid ratio,
which may be identified by production logging methods.
Working over nearby injectors may also solve this problem.
Active methods of inflow control in horizontal wells have thus
far remained limited to multi-reservoir applications. When the
well strikes multiple geological targets, as in channel streams
or compartmentalized reservoirs, completions with
independent control of each reservoir unit have been
successfully used by operators.1-2 Active control in wells
penetrating a contiguous geological formation, however, is a
rarity. The reason for this seems to be the complexity that
active control introduces in the completion string (e.g.,
external casing packers), and the uncertainty that this scheme
can prevent the spread of water or gas from one interval to
another.
A practical approach to mitigation of preferential
encroachment of water or gas in horizontal wells is selective
completion of the sandface. Intuitively, non-completion of
conductive intervals is the best way of coping with this
problem. This approach is sometimes formulated as the
standard practice in certain operating companies. Although
this approach is in principle quite sound, proper
implementation of this approach requires a methodology for
determining the completion profile that is most appropriate for
2 SPE 96395
a given reservoir environment and well operating program.
The object of this paper is to present such a methodology.
Problem Formulation
A standard reservoir simulator is used to model the drainage
of oil toward a horizontal well, driven by bottom-water or gas-
cap or both.
An optimisation problem is formulated in terms of an
objective function, constraints, and variables. The objective
function in this study is the cumulative production of oil from
the well, over a specified time horizon (several years).
The constraints include specification of reservoir properties,
well geometry, and well operating program (e.g., production at
constant liquid rate subject to minimum BHP).
The variable is the connectivity between the reservoir
gridblocks and the wellbore. The wellbore is discretized into
segments to model wellbore pressure drops. The reservoir-
wellbore connectivity is modeled with a normalized
connection factor, with zero indicating non-completed cased-
hole and one indicating open-hole; see Appendix for definition
of connection factor. As a reservoir simulator is used in this
work for forward modeling purposes, the resulting completion
profile is basically a profile of connection factors between the
well segments and adjacent gridblocks. Converting this profile
to a completion profile (e.g., for pre-perforated liner or
perforated casing) is a separate problem, which is not
addressed in this study. However, the resulting connection
factor profile should reveal the completion pattern that may be
appropriate.
The optimisation problem therefore consists of finding the
connection factor profile that maximizes cumulative oil
production from the well within specified constraints.
The constraints can also be varied to see what impact they
have on the final result. In this study, the impact of reservoir
property constraints (e.g., permeability anisotropy) and
constraints on well operating program (e.g., rate target) on the
resulting completion profile are investigated.
The mathematical formulation of this problem and solution
methodology is described in the Appendix. The methodology
is an adaptation of the work presented by Fengjun Zhang on
optimisation of valve placement in segmented horizontal
wells.3
Testing the Optimization Algorithm
In this section two cases are presented where the required
selective completion pattern is intuitively evident. These
cases are run in order to verify that the optimisation algorithm
is robust.
Case 1 -- A 2400-ft horizontal well is placed in the center of a
4200- x 4200- x 240-ft box-shaped reservoir with oil layers
underlain by water. There are 21 x 21 x 12 gridblocks. Basic
model parameters are summarized in Table 1. The well is
completed in the fifth layer from gridblock (5,11,5) to
(17,11,5) and is put under constant liquid flow rate of 10,000
STB/day. The well is modeled with the multisegment option
of a commercial simulator and uses the drift flux correlation
for friction drop calculations in the wellbore.4 Figure 1 and
Figure 2 show successive iterations of the algorithm moving
from the initial guess of constant partial connectivity along the
well length (0.1 and 0.5 connection factor), to complete
connectivity as the optimum solution (1.0 connection factor).
For both the initial guesses, the optimum connection factor
profile is the same.
Case 2 This is a modification of the previous example with
an embedded streak with high vertical permeability in the x-
gridblocks 14 and 15 so as to simulate a good connectivity
between the well and the bottom water layer. The bottom layer
porosity is multiplied by a factor of 10 to simulate a strong
water zone. The rest of the data is same as in the previous
example. The high-transmissibility streak will contribute at
high water-cut and thus it may not be efficient to complete this
segment along the well path. The optimisation algorithm is run
on this model and the results are shown in Figure 3. The
optimum results suggest full completion in blocks 1-13, no
completion in block 14, and partial completion in blocks 15-
17. The cumulative oil production figures are shown in Figure
4, which shows an increase in oil production from 9.533
MMSTB to 10.481 MMSTB (for non-selective and selective
completions). Figure 5 shows the water saturation profile
along the well plane at the end of the simulation run for the
non-selective (left) and the selective cases (right). It is clear
that the optimizer tries to reduce the water channeling along
the vertical streak, so as to reduce water production and
increase the oil production (the well control is total liquid
production).
Sensitivity to Well Operating Program
Two operating scenarios are considered here. One is short-
term with high offtake rate and the other is long-term with a
moderate offtake rate. This is examined for both water-drive
and gas-drive problems.
Case 3 This is a strong bottom water-drive reservoir that is
susceptible to coning. The modeling parameters are shown in
Table 2. Figure 6 shows the completion profile for the case
when the well is produced at a liquid rate of 13000 STB/day
and optimised for a period of 3 years. The optimizer suggests
non-completion of the well towards the heel. The change in
cumulative oil production from 13.81 MMSTB (unoptmized)
to 14.01 MMSTB (optimized) is shown in Figure 7. However
when the well is produced at a liquid rate of 7000 STB/day
and optimised for a period of 5 years, the optimizer suggests a
full completion along the total well length (Figure 8). Thus
the completion profile is dependent on the operating program.
Case 4 This case shows the performance of the optimization
algorithm in a gas-cap scenario The reservoir model is
basically the same as the one used in the previous case (Table
2), in which a gas-cap is introduced and the bottom water is
removed. The GOC is at 6040-ft, the well is completed in the
7th layer from gridblock (5,11,7) to (17,11,7). Two operating
programs are considered for the well. First the well is
SPE 96395
produced at a liquid rate of 15000 STB/day and optimised for
a period of 4 yrs. Then the well is produced at 7000 STB/day
and still optimised for a period of 4 years. Figure 9 and
Figure 10 show the optimized completion factors along the
well for the two programs. Greater non-completion of the heel
is suggested for the higher rate.
Sensitivity to Reservoir Parameters
Prior to the drilling operation, the information on subsurface
characteristics is typically derived from offset wells or
reservoir models with varying degrees of history and
refinement. Therefore, well planning needs to consider the
range of parameters and reservoir scenarios that are plausible.
In this section the impact of several key reservoir properties
on sandface completion profile are exanmined.
Case 5: This case includes both bottom-water and gas-cap.
The production control is constant liquid rate of 15,000
STB/day optimised over 2 years. In one scenario Kv/Kh is
0.1, in the other 0.3. Figure 11 and Figure 12 show the
impact of this variation on the optimum completion profile.
Case 6: This case also includes both bottom-water and gas-
cap. But the permeability field is quite heterogeneous. The
reservoir has 40 x 30 x 12 grids and the well is completed in
the 5th layer from gridblock (13,15,5) to (25,15,5). The well is
produced at 15000 STB/day and optimised over 4 years. The
rest of the parameters remain same as in Table 2. The
permeability field is shown in Figure 13. As shown the well is
drilled through a highly heterogeneous section and it intersects
a highly permeable zone in the middle. The optimized
connection factors for this case are shown in Figure 14.
Figure 15 shows the increase in cumulative oil production
because of non-completion of the high permeability zone that
is prone to water production.
Conclusions
1. A robust gradient-based optimization algorithm is
developed to determine the optimum sandface
completion profile of a horizontal well.
2. The completion profile is determined in terms of the
connection factor between wellbore segments and
reservoir gridblocks adjacent to the well.
3. The optimum connection factor profile depends on
the operating program for the well and basic
reservoir properties and drive mechanism.
4. In general, the degree of required completion
selectivity increases with the strength of the drive
mechanism (bottom-water or gas-cap), the target
production rate, and the degree of heterogeneity of
the formation.
Acknowledgements
We thank the management of Schlumberger Well Completions
and Productivity for permission to publish the results. We also
thank Fengjun Zhang for his insight into the problem.9
3
Nomenclature
Bob Oil formation volume factor at bubble point pressure,
RB/STB
GOC Gas-Oil Contact
ID Internal Diameter
k Permeability, md
N Number of connection factors
OWC Oil-Water Contact
P Number of fluid phases of the wellbore flow problem
Pb Bubble point pressure, psi
Pi Initial reservoir pressure at specified datum, psi
Q Cumulative oil production
Rsb Solution gas-oil ratio at bubble point pressure,
scf/STB
Sg Gas saturation
Sw Water saturation
Swc Connate water saturation
Porosity
ob Oil viscosity at bubble point pressure, cp
o Oil density, lbm/ft3
References
1. Algeroy J, Pollock R: Equipment and Operation of
Advanced Completions in the M-15 Wytch Farm
Multilateral Well, paper SPE 62951 presented at the
2000 SPE Annual Technical Conference and
Exhibition, Dallas, 1-4 October.
2. Rundgren G, Algeroy J, Hestenes L.E, Jokela T, Raw
I: Installation of Advanced Completions in the
Oseberg 30/9-B-38 B Well, paper SPE 71677
presented at the 2001 SPE Annual Technical
Conference and Exhibition, New Orleans, 30
September 3 October.
3. Kharghoria A, Zhang F, Li R, Jalali Y: Application
of Distributed Electrical Measurements and Inflow
Control in Horizontal Wells under Bottom-Water
Drive, paper SPE 78275 presented at the 2002 SPE
European Petroleum Conference, Aberdeen, 29-31
October.
4. Holmes, J.A., Barkve, T., Lund, : Application of a
Multisegment Well Model to Simulate Flow in
Advanced Wells, paper SPE 50645 presented at the
European Petroleum Conference held in the Hague,
the Netherlands, 20-22 October 1998.
5. Eclipse Technical Description 2003A, pp 955.
6. Li R.: Conditioning Geostatistical Models to Three-
Dimensional, Three-Phase Flow Production Data by
Automatic History Matching, Ph.D. dissertation, U.
of Tulsa, 2001.
7. Schewchuk J.R.: An Introduction to the Conjugate
Gradient Method without Agonizing Pain, Lecture
Notes, Carnegie Mellon University, 1994.
8. Kalita R.: Conditioning a Three Dimensional
Reservoir Model to Gas Production Data, MS
Thesis, University of Tulsa, 2000.
9. Zhang F: Flow Control Optimization, Internal
Report, Schlumberger Reservoir Completions,
August 2001.
4 SPE 96395

Appendix algorithm to the flow control optimization problem is the

Connection factor
In the reservoir simulator (Eclipse in this case) the flow path
between the well and the reservoir gridblocks are treated as a
connection.5 The connectivity between the well and a
simulator gridblock is determined by the transmissibility
factor. In this exercise we determine an optimum multiplying
factor for the gridblock connection transmissibility values. A
factor of one means the gridblock is fully open to flow and
zero means the gridblock is not open to flow. Please refer to
Eclispe Technical Manual5 for details on connection
transmissibility factor calculation. During the optimization
algorithms, the multiplying factor is evaluated over the
iterations and incorporated in the simulation run.
Optimization Algorithm
We use a gradient-based (Conjugate Gradient 6,7,8) algorithm
to form a minimization problem. Conjugate gradient method is
one optimization method, which can be implemented by
knowing the first derivative, i.e. the gradient, of the objective
function. That is the main reason that the conjugate gradient
method was selected as the optimization algorithm for this
work. And also conjugate gradient method has quadratic
termination property that is much better than steepest descent.
Conjugate gradient was originally developed for solving linear
equation systems, i.e.
Ax = b .
Solving this linear equation system is equivalent to
minimizing the quadratic form given by:
1
f ( x) = x T Ax b T x
2 .
The algorithm used to minimize objective function f(x) is
given below.
Select x0
r 0 = Ax 0 b (residual)
d 0 = r 0 (search direction)
For
k = 0,1,
(r k ) T r k
k = (Step size)
(d k ) T Ad k
r k +1
= r + Ad
k k k number of iterations exceeds the user specified
x k +1 = x k + k d k
calculation of Adk that is used to compute step size . Here A
is equivalent to the Hessian, i.e. the second derivative of the
objective function. It is hard to evaluate A. Even though we
can calculate the product of A and d without calculating A
explicitly, like several authors have done, it is still difficult to
get Adk because of the fact that we dont know the explicit
expression of the objective function. In our implementation,
we try a big step size first. If it is too big, it will be reduced by
a factor of two. We repeat this process until a proper step size
is reached. Because we do not have the explicit expression of
the objective function, we could not calculate the gradient of
the objective function analytically. Instead of using analytical
method, we use finite difference method to obtain the gradient
of the objective function. In our problem, the objective is to
maximize the cumulative oil production Q. Suppose the
variables are the connection factors, which are denoted by xi
for i = 1, , N where N is the total number of connection
factors. The derivative of Q with respect to xi can be obtained
as follows:
Q Q( xi + xi ) Q( xi )
=
xi xi
where xi is a small change of variable xi. Because the
variable domain is a limited domain, we need to handle the
boundary properly. In our procedure we give a very small
value for the derivatives when the variables are located at the
boundaries.
Two stop criteria are used.
1. When the following condition is satisfied, we will
stop the iterations:
O k +1 O k
<
O k + 10 14
where k denotes the iteration index and is a small
number given by the user. In our examples, we use
=0.001. When this stop criterion is satisfied, we treat
the algorithm as converged.
2. The maximum number of iterations. When the
maximum number of iterations, the algorithm will
stop.

(r k +1 ) T r k +1 The iterations stop when either criterion is satisfied.

= k

(r k ) T r k
SI Metrix Conversion Factor
d k +1 = r k +1 + k d k bbl 1.589 874 E 01 = m3
End k F (F 32)/1.8 = C
ft 3.048* E 01 = m
By replacing the residual in the above algorithm by the in 2.54* E 00 = cm
lbm 4.535924 E 01 = kg
gradient of the objective function, this algorithm can be
psi 6.894 757 E + 00 = kPa
applied to nonlinear problems, like the flow control
optimization problem.3,9 One difficulty with applying this *Conversion factor is exact
SPE 96395 5

Table 1 Model Parameters for the Test Cases 1.2

Model Dimensions 4200- x 4200- x 240-ft
Grid Cells 21 x 21 x 12
1
Well Length 2400 ft
Liner OD 6 inch Iteration 0

Completion factor
0.8
Liquid Production Rate 10000 STB/day Iteration 1
Iteration 2
Permeability (horizontal) 100 md Iteration 3
0.6
Permeability (vertical) 10 md Iteration 4
Porosity 25% Iteration 5
Iteration 6
Connate water saturation 20% 0.4
Iteration 7
Initial Formation Pressure (at OWC) 2700 psi
Bubble Point Pressure (Pb) 2500 psi 0.2
Minimum Bottom-hole Pressure 1500 psi
Formation Volume Factor at Pb 1.32 RB/STB 0
Solution Gas-Oil Ratio at Pb 520 scf/STB 5 7 9 11 13 15 17

Oil Viscosity at Pb 0.7 cp Grid blocks along well path

Forecast duration 3 yrs

Fig. 2 Optimized completion for Case 1 (Initial factor of 0.1)

Table 2 Model Parameters for Cases Assessing
Sensitivity to Well Operating Program
Model Dimensions 4200- x 4200- x 240-ft
Grid Cells 21 x 21 x 12
Well Length 2400 ft
Liner OD 6 inch 1 .2
Permeability (horizontal) 500 md
Permeability (vertical) 250 md 1 It e r a t i o n 0
Porosity 25% Completion factor It e r a t i o n 1
0 .8
Connate water saturation 20% It e r a t i o n 2
0 .6
Initial Formation Pressure @ OWC 3000 psi It e r a t i o n 3
Depth of OWC 6220 ft 0 .4 It e r a t i o n 4
Bubble Point Pressure (Pb) 2500 psi It e r a t i o n 5
0 .2
Minimum Bottom-hole Pressure 500 psi
Formation Volume Factor at Pb 1.32 RB/STB 0
Solution Gas-Oil Ratio at Pb 520 scf/STB 5 6 7 8 9 10 11 12 13 14 15 16 17
Oil Viscosity at Pb 0.7 cp G r id b lo c k s a l o n g w e l l p a t h
Liquid Production Rate 1 130000 STB/day
Forecast duration 1 3 yrs
Liquid Production Rate 2 7000 STB/day
Forecast duration 2 6 yrs Fig. 3 Optimized completion for Case 2

10.6
1.2
10.4
FOPT(MM STB)

1
10.2
Completion factor

0.8 10
Iteration 0
Iteration 1
0.6
Iteration2
9.8
Iteration 3
0.4 9.6

0.2 9.4
0 1 2 3 4 5 6
0
5 6 7 8 9 10 11 12 13 14 15 16 17
Num ber of iterations
Grid blocks along well path

Fig. 4 Cumulative oil production for Case 2

Fig. 1 Optimized completion for Case 1 (Initial factor of 0.5)
6 SPE 96395

Figure 5 Final Water Saturation Profile for Case 2,

Un-optimized Completion (left), Optimized Completion (right)

Iteratio n 0
1.2 Iteratio n 1 1.2
Com pletion factor

1 Iteratio n 2
Completion factor

1
0.8 Iteratio n 3
0.8
0.6 Iteratio n 4
0.6 Iteration 0
0.4 Iteratio n 5
0.4 Iteration 1
Iteratio n 6
0.2
Iteratio n 7 0.2
0
Iteratio n 8 0
5 6 7 8 9 10 11 12 13 14 15 16 17
5 6 7 8 9 10 11 12 13 14 15 16 17
Gridblocks along w ell path
Gridblocks along w ell path

Fig. 6 Optimized completion for Case 3, Rate=13000 STB/day Fig. 8 Optimized completion for Case 3, Rate =7000 STB/day

14.05
1.2
Completion factor

1
Cumulativeoil (MMSTB)

14
Iteration 0
0.8 Iteration 1
13.95 Iteration 2
0.6
Iteration 3
13.9 0.4 Iteration 4
0.2 Iteration 5
Iteration 6
13.85
0
5 7 9 11 13 15 17
13.8
0 2 4 6 8 10 Gridblocks along w ell length
Num be r of ite rations

Fig. 7 Cumulative oil production for Case 3, Rate=13000 Fig. 9 Optimized completion for Case 4, Rate = 15000
STB/day STB/day
SPE 96395 7

1.2 Iteration 0
1.2 Iteration 0 Iteration 1
1 Iteration 1 1 Iteration 2
C om ple ion fa c tor

Com pletion factor

Iteration 2 Iteration 3
0.8 Iteration 3
0.8 Iteration 4

Iteration 4 Iteration 5
0.6 0.6 Iteration 6
Iteration 5
Iteration 7
0.4 Iteration 6 0.4
Iteration 8
Iteration 7
0.2 0.2 Iteration 9
Iteration 8 Iteration 10
Iteration 9 0 Iteration 11
0
5 6 7 8 9 10 11 12 13 14 15 16 17
5 6 7 8 9 10 11 12 13 14 15 16 17
Gridblocks along the w ellpath
Gridblocks along well path

Fig. 10 Optimized completion for Case 4, Rate = 7000 Fig. 12 Optimized completion for Case 5 (Kv/Kh=0.3)
STB/day

1.2
Iteration 0
1 Iteration 1
Completion factor

0.8 Iteration 2
Iteration 3
0.6
Iteration 4
0.4
Iteration 5
0.2 Iteration 6
0
5 7 9 11 13 15 17
Gridblocks along the w ellpath

Fig. 11 Optimized completion for Case 5 (Kv/Kh=0.1)

Fig 13: Permeability field for Case 6, heterogeneous reservoir (permeability range of 15 to 2000 md)
8 SPE 96395

1.2 20

Cumulative oil (MMSTB)

Iteration 0 19.8
1
19.6
Completion factor

Iteration 1
0.8 19.4
Iteration 2
19.2
0.6 Iteration 3
19
0.4 Iteration 4
18.8
Iteration 5
0.2 18.6
18.4
0 0 1 2 3 4 5 6
5 6 7 8 9 10 11 12 13 14 15 16 17
Num ber of iterations
Gridblocks along w ell path

Fig. 14 Optimized completion for Case 6 Fig. 15 Optimized production for Case 6