Professional Documents
Culture Documents
Summary
Over the last decade, horizontal wells have become a well-established technology for the recovery of oil and gas. A considerable
amount of analytical and experimental work has been published on
various aspects of horizontal-well production, including transientflow and stabilized-inflow models, productivity indices, and coning
and cresting behavior. Although these methods provide insight into
the behaviors of horizontal wells, only a few of them consider
pressure drop along the wellbore and essentially infinite conductivity is assumed. In 1990, Dikken1 proposed the first semianalytical model to evaluate the production performance of a horizontal
well with the consideration of the wellbore-pressure drop resulting
from turbulent flow. Since then, others2-9 have presented different
coupling models for wellbore flow and reservoir inflow through
perforations. However, even in cases where pressure drop along a
wellbore is considered, only the frictional component is included
under most circumstances; pressure drop because of acceleration
and other effects is neglected. We show that, because of the
existence of perforation inflow, accelerational pressure drop can be
important relative to the frictional component and can significantly
influence the well-flow rates under some flow conditions. Furthermore, traditional methods to determine frictional pressure drop in
pipe flow that do not account for inflow are used in most of the
coupling models, which is not justified, because inflow changes the
wall friction in the wellbore.
In this paper, a single-phase wellbore-flow model is presented
that incorporates not only frictional, accelerational, and gravitational pressure drops, but also the pressure drop caused by inflow.
The new model is readily applicable to different wellbore-perforation patterns and well completions and can be easily incorporated
into reservoir simulators or analytical reservoir-inflow models. To
obtain the precise contribution of the frictional pressure drop along
Copyright 1998 Society of Petroleum Engineers
Original SPE manuscript received for review 6 October 1996. Revised manuscript
received 19 December 1997. Paper peer approved 13 January 1998. Paper (SPE
36608) first presented at the 1996 Annual Technical Conference and Exhibition held
in Denver, Colorado, 69 October.
124
Reynolds number. Merkine et al.18 extended these results to account for the effect of suction on the level of turbulence in the fluid
(suction decreases turbulence level). Their results show good agreement with the measurements of velocity profiles by Weissberg and
Berman.19 Unfortunately, no results of the friction factor were
given. By modifying Kinney and Sparrows analysis, Doshi and
Gill20 obtained an excellent agreement with the experimental velocity profiles of Weissberg and Berman. Lombardi et al.21 investigated the fully developed turbulent flow along a pipe with air
injection through its porous wall and found that the Nusselt number
in the downstream region of the pipe could be related to local
parameters, such as the local axial-flow Reynolds number and the
injection Reynolds number, without additional dependence upon
the inlet-axial Reynolds number and axial position.
It is quite interesting to note that the characterisics of pipe flow
with wall mass transfer are different from those of channel flow or
flow past a flat plate. For example, considering the laminar flow
case, the local friction factor increases with an increase of wall
Reynolds number for pipe flow but decreases for channel flow.
Moreover, the suction-induced transition to turbulent flow (suctioninduced instability) and the existence of multiple solutions for pipe
flow have no counterpart in channel flow.22
The different behavior of the local friction-factor change with
inflow for laminar flow and for turbulent flow is another interesting
fact about pipe flow. For laminar flow, the local friction factor
increases with an increase in the injection Reynolds number,
whereas it decreases for turbulent flow. The underlying mechanisms will be discussed later in this paper.
Fluid flow in pipes with wall mass transfer did not interest many
petroleum engineers until the horizontal-well technology was introduced and widely applied in the petroleum industry starting in
the 1980s. For fluid flow in horizontal wells, the flow format
is quite similar to pipe flow with mass transfer through its porous wall. The main differences between these two types of flow
are as follows.
In horizontal wells, the mass transfer is normally through
perforations, whereas, in the case of pipe flow, it is through pores
in the wall. In other words, the effective perforation density is very
large (theoretically infinite) for the porous-pipe flow case. Nevertheless, in the case of openhole completions, the horizontal-well
and porous-pipe flow problems are conceptually identical.
The injection rates are usually quite small in the case of
porous-pipe flow; this is not necessarily the case for wellbore flow.
When there is no mass transfer through the pipe wall, the
effective pipe roughness may be very different from the actual pipe
roughness in a horizontal well because of the effect of perforations
on the axial flow (such as flow separation, cavity flow, or secondary
flow), but it changes only slightly from the actual value for the
porous-pipe flow case.
Little information is available with regard to the significance of
the above-mentioned differences on the prediction of horizontalwell behavior. Although much research work has been completed
for the porous-pipe flow problem, these results may not be directly
applicable to horizontal wells. Recognizing this fact, petroleum
engineers began to study the horizontal-well flow problem from
the late 1980s.
Kloster23 studied flow resistance in a perforated pipe, both with
and without flow injection through the pipe wall, by conducting
experiments on a pipe of 656 in. outside diameter and 17 ft in length.
Asheim et al.24 stated that the total pressure drop along a perforated
pipe is made up of wall friction and inflow acceleration and
computed the wall friction factor in the same way as for a regular,
unperforated pipe. Ihara et al.25 studied channel flow with continuous influx into the horizontal channel from an oil-reservoir model.
They stated that pressure gradients increase almost uniformly in the
test channel because of the confluence of influx and axial flow, and
the resulting pressure drop increases linearly with influx velocity.
Because of perforations, effective pipe roughness may be different from the original roughness even without inflow or outflow
through perforations. Su and Gudmundsson26 measured the mass
flow rates and water-column heights for water flow along a vertical
pipe. Data for flow, both with and without perforations, were
SPE Journal, June 1998
collected. They introduced a roughness function in the frictionfactor calculation to account for the effect of perforations on pipe
roughness and correlated the roughness function as seven times the
perforation-to-casing-diameter ratio, based on their measurement.
For fluid flow in a pipe with one single perforation, Yuan et al.27
developed a correlation for the apparent friction factor by matching
their experimental data. (Apparent friction factor is equivalent to
dimensionless total pressure drop along a pipe section. It includes
different pressure-drop components, such as frictional, accelerational, and gravitational. Therefore, the apparent friction factor is
not exclusively related to wall friction. All the new correlations
presented in this paper are for the Fanning friction factor, which is
defined as the dimensionless pressure drop caused solely by wall
friction.) However, there is significant difference between wellbore
flow and pipe flow with one single perforation. Moreover, pipe
flow with one single perforation is not applicable to real horizontal
wells; hence, it is questionable whether their correlation can be used
for wellbore flow. Recognizing this deficiency, they extended their
experiments to pipe flow with more perforations and developed a
new correlation for the apparent friction factor.27
Although single-phase pipe flow with wall mass transfer has
been the topic of numerous research activities, more work is
necessary, considering the following observations.
No general correlation exists for determining the wallfriction factors for fluid flow in a wellbore with inflow or outflow
through perforations.
The accelerational- and inflow-directional-pressure drops are
neglected in most wellbore-flow models or wellbore/reservoir coupling models.
The wall-friction shears are usually evaluated with frictionfactor correlations for pipe flow without wall mass transfer. Therefore, the impact of mass transfer through the pipe wall is not included.
General Wellbore Flow Model
nDx
g
rAI vr nx 2 rA# Dx sin u,
BI gc
gc
1
1
rA v2 2
rA n2 2tw S Dx
B1 gc 1 1 B2 gc 2 2
. . . . . . . . . . . . . . . . . . . (1)
SD
r d n2
S
g
nr AI
dp
2 tw 1
5 2
n n 2 r sin u. . . . . . (2)
dx
gc dx B
A aI gc A r x
gc
Eq. 2 indicates that the overall pressure gradient consists of four
different components.
125
The pressure gradient caused by kinetic-energy change (accelerational effects). This term should be zero for incompressible
fluid flow in pipes with constant inside diameter (ID) and without
inflow or outflow through the pipe wall. Obviously, it will not be
zero for wellbore flow with wall-mass transfer.
The frictional-pressure gradient, which depends on both axial
and perforation flows.
The pressure gradient caused by inflow direction, which is
called inflow-directional-pressure gradient in this paper. It may
help or hinder the axial flow depending on the inflow direction.
The gravitational pressure gradient. It is reasonable to assume
that the gravitational-pressure gradient is trivial and thus negligible
for horizontal wellbore flow.
To describe quantitatively the relative importance of different
pressure-gradient components, the following three dimensionless
numbers are introduced: Raf 5 the ratio of the accelerational and
frictional pressure gradients; Rgf 5 the ratio of the gravitational and
frictional pressure gradients; and Rda 5 the ratio of the directional
and accelerational pressure gradients. With these definitions, the
total pressure gradient takes the simple form,
dp
4tw
5 2
@1 1 Raf ~1 2 Rda ! 1 Rgf #. . . . . . . . . . . . . . . . . . (3)
dx
d
The closer the wellbore location is to the horizontal-well, toe
(x 5 0), the smaller the local wellbore-production rate qw; thus, the
larger the Raf and the more important the accelerational pressure
gradient. In contrast, near the heel of the horizontal well (x 5 L),
the local production rate qw becomes large and close to the total
well-production rate, whereas the specific influx (inflow rate/
wellbore length) qe does not change significantly, so Raf is small,
and the accelerational-pressure gradient is small and may be negligible.
For the uniform inflow case, it can be shown that28:
In the laminar-flow regime, Raf is dependent only on fluid
properties, inflow rate, and pipe ID, but it is independent of location
x and pipe roughness . The larger the qe, the larger the Raf, but the
shorter the laminar-flow length Lwf.
In the turbulent-flow regime, the friction factor, f, depends on
the local Reynolds number, the wall Reynolds number, and the
relative pipe roughness; consequently, the Raf depends on location
x, pipe geometry (pipe ID and pipe roughness), fluid properties, and
inflow rates.
Because turbulent flow occurs along almost the whole wellbore
section for most practical situations, it is anticipated that the
momentum-correction factor B does not change much for different
velocity profiles29; therefore, B can be taken as constant. Besides,
the derivative, dv/dx, can be easily obtained from the mass-balance
equation. Hence, Eq. 2 can be rewritten as
tw 5
g
d dp 2nr AI
nr AI 2
2 2
nn 1
n sin 2g 2 r sin u .
4 dx B gc A I 2aI gc A I
gc
. . . . . . . . . . . . . . . . . . . . . . . . . . (4)
On the basis of Eq. 4, the wall friction shear stress and, hence,
the wall friction factor, can be calculated. The wall friction factor
is expected to depend on the local axial Reynolds number, the
injection Reynolds number, the effective relative pipe roughness,
and, possibly, the inflow/axial flow rate ratios.
Effects of Wall Mass Transfer on Friction
As discussed earlier, mass transfer through the pipe wall affects the
wall-friction shear. The influence of either inflow or outflow
depends on the flow regime present in the wellbore. The inflow
(production well) increases the wall friction for laminar flow while
decreasing it for turbulent flow. In contrast, outflow (injection well)
decreases the wall friction for laminar flow while increasing it for
turbulent flow. In other words, the wall friction is different from
that of pipe flow with no inflow or outflow. Therefore, frictionfactor correlations for pipe flow without inflow or outflow cannot
be used for wellbore flow with both axial flow in the pipe and
inflow or outflow through perforations.
Laminar Flow. Kinney15 numerically solved the problem of fully
developed laminar flow in a porous pipe where both axial flow and
inflow or outflow through the pipe wall were present. The results
show that the ratio between the local friction factor, f, and the
no-wall-flow friction factor, f0 (which is defined, in this paper, as
the friction factor computed from the correlations for pipe flow with
no mass transfer through the pipe wall by use of local Reynolds
number and effective pipe roughness), is dependent only on the wall
Reynolds number. Unfortunately, this relationship was given in a
graphical form and no equation was supplied. Hence, it cannot be
used easily in a wellbore-flow model.
Figs. 2 and 3 show the calculated ratios between the local friction
factor and the no-wall-flow friction factor corresponding to different wall Reynolds numbers NRe,w, based on the reduced ordinary
differential equation for the dimensionless stream function and the
numerical procedure provided in Kinney.15 Fig. 2 also shows the
prediction of the local wall friction factor by the Yuan and Finkelstein equation.14 Because the Yuan and Finkelstein equation was
obtained by means of a perturbation method in which the injection
wall Reynolds number, NRe,w, was assumed to be sufficiently small,
it is anticipated that this equation can only be used for small NRe,w
(say NRe,w , 2.0).
These data have also been used to develop the following frictionfactor correlations by means of a fast and effective nonlinear
regression procedure called the Polytope method.30, 31
For inflow (production well) (NRe,w . 0),
f5
16
0.6142
~1 1 0.04304 NRe,w
!.
NRe
. . . . . . . . . . . . . . . . . . . . . . . (5)
~2NRe, w !1.3056
16
1 2 0.0625
.
NRe
~NRe, w 1 4.626!20.2724
. . . . . . . . . . (6)
S D G
f 5 f0 1 2 29.03
NRe, w
NRe
0.8003
. . . . . . . . . . . . . . . . . . . . . . . (7)
f 5 f0 1 2 17.5
NRe,w
.
N0.75
Re
. . . . . . . . . . . . . . . . . . . . . . . . . . . . (8)
shorter the length of the laminar-flow region. For the fully developed turbulent-flow region, all the Raf curves merge into one, which
varies only with wellbore location. Between the laminar and the
fully developed turbulent-flow region lies the partially developed
turbulent-flow region, where Raf varies with wellbore location, fluid
properties, production rate, and pipe geometry.
128
Well 1
Well 2
Well 3
Wellbore length, ft
Perforation length, ft
Pipe ID, in
Relative pipe roughness
Perforation ID, in
Perforation density,
shots/ft
Inflow direction, degrees
Fluid density, lbm/ft3
Fluid viscosity, cp
Wellbore rate at
perforation start point,
B/D
Inflow rate, B/D
Specific production
index, B-D/ft-psi
1,000
100
6.18
0.0002
0.18
8
1,870
1,870
4.5
0.0001
0.18
10
1,000
1,000
6.0
0.0002
0.18
10
90
62.4
0.878
7,000
45
52.4
2.5
0
45
62.4
1.0
0
7,000
N/A
7,500
N/A
N/A
2.0
pressure drop. The pressure drawdown near the well toe is less than
that near the well heel; therefore, the specific inflow rate near the
well toe should be smaller than that near the well heel. On the one
hand, wellbore pressure, or pressure drop, is needed to determine
the specific inflow-rate distribution along the wellbore; on the other
hand, the specific inflow-rate distribution is an indispensable parameter to calculate wellbore pressure, or pressure drop. Hence, a
model that incorporates both the wellbore-fluid flow and the fluid
flow from reservoir into wellbore is required for determining
specific inflow-rate distribution, wellbore-pressure drop, and overall well-production rate.
Because the reservoir-inflow model is not the main concern of
this paper, a simple model used by Dikken1 and Novy9 is applied
to describe the reservoir-fluid flow.
qe ~x! 5 Js @pe 2 pw ~x!#,
. . . . . . . . . . . . . . . . . . . . . . . . . . . (10)
shows the production-rate change with wellbore length. Wellproduction rate decreases significantly because of wellbore-pressure drops and is likely to reach an asymptotic value at a certain
wellbore length. Beyond this, no additional production will be
contributed by additional well length.
Conclusions
28. Ouyang, L.-B., Arbabi, S. and Aziz, K.: General Single Phase Wellbore
Flow Model, topical report, Contract No. DE-FG2293BC14862 U.S.
DOE, Washington, DC (February 1997).
29. Ouyang, L.-B. and Aziz, K.: Steady-State Gas Flow in Pipes, J. Petroleum Science and Engineering (1996) 14, No. 1, 137.
30. Gill, P.E., Murray, W., and Wright, W.H.: Practical Optimization,
Academic Press, London (1981).
31. Ouyang, L.-B.: Stratified Flow Model and Interfacial Friction Factor
Correlations, Masters thesis, Stanford U., Stanford, California (1995).
32. Wallis, G. B. Discussion of Ref. 17, J. Heat Transfer (1970) 92, No. 1,
124.
33. Ouyang, L.-B., Arbabi, S., and Aziz, K.: Preliminary Analysis of the
1995 Stanford Horizontal Wellbore Experiments, Productivity and
Injectivity of Horizontal Wells, annual technical report, Contract No.
DE-FG2293BC14862, U.S. DOE, Washington, DC (March 1996)
124.
cp 3 1.0*
ft 3 3.048*
lbm 3 4.535 924
psi 3 6.894 757
*Conversion factor is exact.
E203
E201
E201
E100
5
5
5
5
Pazsec
m
kg
kPa
SPEJ
Ouyang
Arbabi
Aziz
133