Sie sind auf Seite 1von 14

SPE 62618

Fresh Water Injection Stimulation in a Deep Tight Oil Reservoir


S.W. Wong*, P.M. O'Dell, Petroleum Development Oman LLC,
C.J. de Pater and J. Shaoul, Pinnacle Technologies Delft BV

Copyright 2000, Society of Petroleum Engineers Inc.


This paper was prepared for presentation at the 2000 SPE/AAPG Western Regional Meeting
held in Long Beach, California, 1923 June 2000.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, 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 an abstract of not more than 300
words; illustrations may not be copied. The abstract 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
The Athel exploration play in the South of Oman contains
a huge oil reserve, but in rather unusual reservoirs with very
low permeability. Two discovered fields, Al Noor and Al
Shomou carry a total oil in-place in excess of 2 billion barrels.
These reservoirs contain thick silicilyte slabs that are encased
in salt and shale. Due to low permeability, effective well
stimulation plays a crucial role in economic field
development.
There was evidence that silicilyte cores contained salt and
it was believed that removing the salt with a simple fresh
water soak could increase the permeability. Subsequently, a
series of fresh water injection experiments was conducted in
an Athel well, in 1995-97. In total, more than 16000 m3
(100,000 bbl) were injected. The aim was to wash the salt
from the matrix, increasing the near wellbore permeability and
hence, improving productivity. The well was indeed
stimulated by these treatments, but the improved productivity
did not last for very long.
Later, we suspected that the fresh water injection could
have resulted in fracture propagation, or other production
improvement mechanisms. A fresh look at the injection and
production test data was conducted in 1998, with the objective
of assessing the applicability of water-frac treatments. This
could potentially eliminate or significantly reduce the need for
proppant in fracture treatments.
The study showed that the production improvement was
not due to matrix salt washing. In fact, hydraulic fractures
were propagated. But the production response after the final
and largest volume fresh water injection was complex and
difficult to model. Apparently, in addition to fracture

* Presently with Shell International E&P in Rijswijk

propagation, a large cavity was also dissolved in the overlying


salt layer. The back production of salt saturated water, and
increasing storage coefficient as derived from successive well
test analysis could be related to the creation of such a cavity.
In addition, the rock might have dilated, contributing to the
short lived, but significant production increase. The study
concluded that so-called water-frac treatments would need
proppant to maintain fracture conductivity in the Athel.
This paper presents a field case of a novel stimulation
method in tight reservoir rock. The methodology of integrating
well tests analyses, production logs and fracture geometry
simulations to obtain a comprehensive reservoir stimulation
model is outlined. The production mechanisms induced by the
fresh water injection and the cause of the rapid productivity
decline is described.
Introduction
The Athel formation in South Oman is a self generating
source rock and reservoir. This unusual rock is generally
called Silicilyte, although the reservoir is laminated on many
scales and several lithofacies can be distinguished. Some 8090% of the grains consist of microcrystalline quartz. The
reservoirs consist of big slabs (some 2 by 6 km) that are
encased in salt and shale, see Figure 1. The rock contains
much organic material and the matrix is very fine-grained (2-3
m). Its matrix permeability is very low, in the order of
0.02mD.
The reservoir is highly overpressured: at a depth of 4000 m
the reservoir pressure is 800 bar. The Athel contains volatile
oil of 48o API, with a bubble point of 277 bar. The initial
solution gas-oil-ratio is 400 to 550 m/m and the initial oil
viscosity is 0.3 to 0.36 mPa-s. The reservoir is strongly
faulted. Some vertical fractures have been found in core,
suggesting that these features must be common throughout the
reservoir. These vertical fractures were completely cemented
with quartz, halite and clay and they may hinder flow
communication in the reservoir.
Two fields that have been appraised are Al Noor and Al
Shomou. The latter was first tested with Al Shomou-1 (drilled
in 1995 on the crest of the structure, see Figure 2). This well
encountered 328 m net pay. Reserve booking of Al Shomou
followed the successful fracture stimulation of Al Shomou 3.
The treatment was performed in two massive stages to cover
3
the entire interval. Each stage used 1280 m of cross-linked

S.W. WONG, P.M. O'DELL, C.J. DE PATER, J. SHAOUL

gel and 250,000 kg of proppant. The production was increased


from 60 m/day to an initial stabilized rate (after 7 days of
production) of 395 m/day. This implied an improvement of
the productivity index by a factor of seven 1.
Recently, it has become apparent (e.g. in gas reservoirs in
East-Texas) that such massive fracture stimulations may be
sub-optimal in low-permeability reservoirs where fracture
length is important. In normal gel fracture treatments, the high
pressure that results from using viscous fluids with high
proppant concentrations may limit the fracture length. This
effect may be exacerbated by reduced proppant pack clean-up
and low conductivity2. Under some conditions, pumping large
volumes of water and adding only low proppant concentration
of 1-2 ppg may provide a more economical stimulation of
these East-Texas reservoirs.
Therefore, a study was initiated to investigate the
feasibility of using minimum proppant concentration in future
Athel well stimulation. The extrapolation of these field data
and present practice to other reservoirs is not straightforward,
especially for Athel. For instance, the success of these socalled water-frac treatments may depend on the fracture
conductivity that can be sustained, and this is in turn related to
the residual fracture width. If a fracture in the Athel would
heal very soon, the benefit of stimulation cannot be
maintained if no proppant was used. Residual fracture width
can be related to fracture roughness in core tests, but healing
of fractures should ideally be determined under in-situ
conditions. Such a field test has not been carried out in the
Athel, but the water washes (or squeezes) trial in Al Shomou-1
well in the mid 90s may now provide valuable information.
The original intention in Al Shomou-1 was not to fracture
stimulate the well, but to dissolve the salt from the reservoir
rock by injecting fresh water at matrix flow conditions. The
dissolution of salt was expected to increase the porosity and
enhance the matrix permeability. The well was indeed
stimulated by the water squeezes: production increased from
40 m3/day to 270 m3/day, which was comparable to the result
of the massive fracture stimulation. The production rate,
however, declined sharply. ODell3 reported the results of
pressure buildup analysis (PBUs) that suggested the presence
of a fracture. Perhaps, the fracture was indeed propagated
during the water injections, and the healing of the fracture
might explain the deterioration of the productivity.
The injection experiment was never intended as fracture
stimulation, therefore not all pumping pressures were recorded
and there was no planned sequence of step rate tests and
pump-in shut-ins, needed to provide better insights on the
stimulation mechanism. Still, we shall attempt to analyse the
available data to determine the reason for productivity
enhancement and the subsequent rapid productivity decline. A
number of postulations are plausible. At one extreme we have
the matrix salt dissolution mechanism and at the other, the
fracturing mechanism. We shall try to interpret the data as
much as possible without making unnecessary assumptions.
In this paper, we will briefly describe the entire
experiment. The available data are pressure and production
records and production logs. Then, we present the well tests

SPE 62618

analysis. The well test data will be scrutinised independent of


any modelling assumptions. However, parameters such as
permeability can only be determined if one assumes a
reservoir model. Moreover, the prediction of production
behaviour also requires a reservoir model. Based on this
reservoir model we will look at the injection pressure and the
production behaviour. Finally, we will discuss the
reconciliation of all data with the model and outline the
remaining uncertainties.
Injection and Production Data
Table 1 lists the most important test activities during the
water squeeze and the data collected before and after each
injection. The main periods were:
0. Initial production.
1. 20 m injection of fresh water and 20 m3 injection of acid,
2. 200 m injection followed by a period of production.
3. 1000 m injection followed by a period of production.
4. 5000 m injection followed by a period of production.
5. 16800 m injection followed by a period of production.
We have PBUs after the most significant injections and
production logging tools (PLTs) were also run during
injection and production. Figure 3 shows the pressure and
production over the entire test period. Apparently some
stimulation occurred during the first washes and a significant
stimulation was achieved after the 5000 m injection. After
that stimulation, the well was produced for a short period,
which showed a fast decline in productivity. After a prolonged
shut-in of 3 months, the well was produced for nearly 3
months (June-Aug 1996). The production shows some sharp
spikes owing to production metering being at the end of a 50
km flowline. In this flowline the velocities of oil and gas were
neither constant nor uniform but the integral of any sufficient
time interval shows the well to be stable and declining towards
the original, unstimulated rate. In Dec 1996 a large injection
of 16800 m was performed. After this injection, the
production character changed completely. Initially, only (salt
saturated) water was produced (10,000 m) and after that oil
was produced at a rate of 120 m per day.
Looking at the cumulative water injection and production
(Table 2) we see that the fraction of water produced increased
with each injection cycle, except for the first injection, which
was much smaller than all the subsequent injections.
Production Logs
Initially, the production came from the bottom and top of
the reservoir. The first large injection of 200 m was into the
top layer. The production after the injection came from the
bottom and top of the reservoir.
The 1000 m injection occurred again into the top layer.
There was evidence that some flow occurred behind the
casing, since the flow profile extended to the top of the open
hole section. The production was now more spread out over
the reservoir height. The top zone was particularly stimulated.
The 5000 m injection showed that the fluid was more
evenly distributed over the reservoir. The production also
came from the middle of the reservoir after this injection. The

SPE 62618

FRESH WATER INJECTION STIMULATION IN A DEEP TIGHT OIL RESERVOIR

16800 m injection was again concentrated at the top, but also


significant injection occurred into deeper layers. We conclude
that the production profile changed a great deal over the entire
test period, although most of the water was injected into the
top layer. The deeper layers also accepted some injection
water; in particular during the later injections.
Pressure Buildup Analysis
We re-analysed the four pressure build-up data sets
available, from December 1995, February 1996, May 1996
and February 1997.
Test Analysis: The first PBU analysis shows that the well
clearly had a negative skin from the start. This may have been
caused by drilling induced fractures, since a high overbalance
was used and there was evidence for circulation losses. We
were able to match the data with a fractured well model and
decreasing wellbore storage. This gave a kh of 1.64 md-m,
along with a fracture half-length of 8 m. This is quite
reasonable, because the PLT at this stage shows only a small
part of the interval contributing and vertical permeability is
believed to be negligible. So we would expect to see only a
small part of the total kh. (Typical total kh for the entire
reservoir is in the order of 10 md-m.)
The second and third build-ups were more difficult to
analyse. This is largely because prior to the buildup, the
production rate was not constant, and there was also mixed
production of water and oil. This was further complicated by
the unknown effect of water injection performed just two
weeks before. Nevertheless, good matches were obtained in
both cases, and the match of these two pressure buildups
provided valuable information about the changing kh, fracture
lengths and reservoir boundaries.
The second build-up took place after the 1000 m3 water
injection. We again used the infinite conductivity fracture
model. The estimated kh in this case was slightly higher, at
2.95 md-m, and the fracture half length estimate was 10 m.
Again, the higher kh estimate is in agreement with the PLT
results, which showed more of the interval contributing to
flow after the 200 and 1000 m3 injections.
The third pressure build-up, in May 1996, was after the
3
5000 m water injection. This is the most important of all the
buildup data, because it is the longest and it followed the
injection that achieved the most dramatic production
enhancement. Once again, it was necessary to use a fractured
reservoir model to match the build-up data. Figure 4 shows the
half slope plot and provides the most important piece of
evidence for the existence of a fracture. This behaviour is also
evident in the fourth build-up, shown in Figure 5.
Uniqueness of solution: In matching the pressure build-up
data, there is always a trade-off between permeability and
fracture length. We constrained this problem of nonuniqueness to some extent by matching both the buildup
pressure and the initial reservoir pressure. A sensitivity
analysis was carried out and for these cases, we found that it
was impossible to match both the final buildup data as well as
the initial reservoir pressure with a different combination of kh
and fracture half-length.

Drainage Area Boundaries: The apparent reservoir


pressure seen in the last build-up test in February 1997 was
much lower than the initial reservoir pressure. This is often
seen in cases where the drainage area is a long narrow
rectangle. Following significant production, it takes a long
time for the area near the well to recharge, compared to a
radial flow situation. In order to match this decline in apparent
reservoir pressure, it was necessary to use a bounded reservoir
model. Using the distance to the boundaries obtained from the
match of the third buildup, we were able to match the buildup
with the correct initial pressure. The reservoir boundary
effects became apparent when looking at the derivative plot
for this buildup. The final change in slope could be matched
with a rectangular reservoir model with a total width of 40 m.
The length of the drainage area could not be determined at this
point, so we made it a large number. In order to match the
buildup data, we used a larger kh, of 5 md-m and a fracture
half-length of 24 m.
Possible mechanisms of Reservoir Stimulation
Originally, the intention of the water injection trial was to
dissolve salt in the rock matrix and thereby increase the
permeability. Indeed, salt saturated water was produced and
the reservoir was significantly stimulated. However, the
existence of salt in the matrix has not been proven. The
concept started with the circumstantial evidence that a core
slab dipped in fresh water had a glaze of salt after a few hours.
However, all Athel wells are drilled with salt saturated mud.
What are the alternative explanations for the apparent
stimulation? The injections could have induced a fracture in
the reservoir. The fracture could explain the high injectivity
and the productivity improvement. Since the fracture would be
expected to grow towards the salt (and the cement quality was
poor) the fresh water could have come in contact with the salt
overburden. Dissolution of the salt could have created a
cavity. This mechanism could explain the large injectivity (at
relatively low pressure) and the production of salt saturated
water. Upon opening of the well, most of the water in the
cavity would be produced, and the cavity might even have
collapsed. Reservoir oil can enter a cavity at the reservoir top
to replace voidage created by producing water from the cavity.
In addition, the rock surrounding the cavity might have
dilated, contributing to the short lived production increase.
What volume could we expect for this cavity? If the 16800
m injection of fresh water would dissolve 250 gr. salt per
litre, a cavity of 2000 m could have been created. The
previous injection might also have created a cavity. Since the
casing cement bond was poor, some of the injected water also
could be stored in the annulus around the casing. Assuming a
stored volume of 2000 m, this would account only for a small
part of the total injection. Therefore most of the injected water
would need to be stored in the reservoir. This would require a
large fracture, because the injectivity of the reservoir itself is
much too low.
If there is a residual fracture due to the previous injections,
this would explain why a large amount of water can be
injected into the reservoir below fracturing pressure.

S.W. WONG, P.M. O'DELL, C.J. DE PATER, J. SHAOUL

The well tests show that there is linear flow after the 5000
m injection. This behaviour might still be explained by the
matrix salt dissolution mechanism, since the PBU just
indicates that there is a planar permeable feature in the
reservoir. If the dissolution process would be very anisotropic,
it could have created a planar zone that is very permeable.
This mechanism is however, very unlikely because the
stimulated interval would need to have a permeability much
higher than values obtained from cleaned core material or
from correlation.
Alternatively, the water might have dissolved salt from the
cemented fractures. If the fractures would be aligned in a
plane, the result would look like fracture stimulation in a PBU.
The only difference will be that the dissolution process might
have occurred at a lower pressure.
All the production mechanisms postulating salt dissolution
from the reservoir matrix lead to a stimulated well. However,
it is difficult to reconcile well performance with matrix
stimulation. Well performance on October 1st, 1996 (after all
but the last and largest fresh water stimulation) is the same as
well performance on December 1st, 1995 (prior to any
stimulation).
The most likely explanation of the PBU behaviour,
however, is the propagation of a large fracture in the 5000 m
injection. Unfortunately, we lack the pumping pressure record
for this injection. It is very well possible that the pumping
pressure during this injection was sufficiently high to create a
fracture, since the injection rate during this injection was
much higher than the injection rate of the previous injections.
Therefore, for the analysis of the production data, we
postulate the following scenario (sketched in Figure 6).
Initially, a fracture was induced in a layer with a low stress.
This would explain the injection profile during the 200 m
injection, where most of the water was injected at a depth of
3700 m. This resulted in a slight stimulation of the upper zone.
The 5000 m injection induced a large fracture that gradually
healed when the well was produced.
A completely different production behaviour was seen in
the 16800 m injection. That might be explained by the
formation of a cavity in the salt overburden. Either the water
flowed behind pipe, or the water reached the salt through the
fracture. When the well was produced, the cavity collapsed,
resulting in a quick flowback of the water (which was now salt
saturated). Later, the cavity stabilised and the reservoir started
to produce oil from the fracture.
There are many uncertainties in this picture of the events,
but it honours the data that we have. We will now use this last
scenario to match the injection behaviour during the initial and
last (16800 m injection) injection. We will also estimate the
size of the fracture created in the 5000 m injection.
Then we will look at the production behaviour, which is
the purpose of the entire study. We will compare the actual
production with the production of a well stimulated with
propped fractures, to determine the extent of any fracture
damage that could explain the fast decrease of the oil rate.

SPE 62618

Modelling Fracture Growth


In modelling fracture growth during water injection, the key
question is whether we are in a regime of fast fracture growth
and slow leak-off or in the regime of dominant matrix flow
(fast leakoff) and slow fracture growth. The two regimes can
be determined from the dimensionless injection rate, which is
the injection rate scaled on the reservoir transmissivity:
iD =

i
...................................(1)
2 khp

The dimensionless fracture length is obtained by dividing


the fracture length by the approximate size of the pressure
penetration (given by the square root of the product of
hydraulic diffusivity and time):
LD =

Lf
t

.....................................(2)

If we inject water at a rate that is close to the rate that the


reservoir could take in matrix flow we are in the fast leakoff
regime. However, if we inject at a rate that is much higher
than the matrix flow, we will quickly propagate a fracture,
since the injected water cannot leak into the rock and the water
has to be stored inside the fracture. The latter regime is the
hydraulic fracturing regime. In well stimulation with hydraulic
fracturing, one attempts to overwhelm the leak-off and to
create a large fracture. In waterflooding, one tries to keep the
fracture relatively small with respect to the flood front. It is
important to determine the regime, because conventional
hydraulic fracturing simulators are not suitable for modelling
water flood fractures. In a waterflood fracture, the poro-elastic
stresses (and possibly thermal stresses) may dominate the
fracture growth.
Figure 7 shows a plot of dimensionless fracture length
versus dimensionless injection rate. For Athel conditions, we
see that an injection rate close to 50 m/day would correspond
with the fast leak-off limit, whereas the injections at a rate of
1000 m/day would induce a fast growing fracture. In that case
we can use a normal hydraulic fracture propagation model,
since fluid leak-off will be limited and the poro-elastic (and
thermo-elastic) effect will also be small.
It is possible that some of the injections were in an
intermediate regime, since the production periods were much
longer than the injections. If the depletion extended some 50
m after 3-4 months production, the water injection would not
change that much in 20 days. In that case we could treat the
pressure drop due to depletion as a lower background reservoir
pressure, and consequently the poro-elastic stress reduction
might be significant. For modelling the poro-elastic effects,
we have applied an analytical model that goes back to the
work of Perkins and Gonzales5. Koning6 considerably
improved the poro-elastic stress calculation. We use the model
to estimate the fracture size and the associated stress changes.
This model considers fracture growth in a reservoir, assuming
that the fracture stays in the reservoir and that all stress

SPE 62618

FRESH WATER INJECTION STIMULATION IN A DEEP TIGHT OIL RESERVOIR

changes occur inside the reservoir. Both poro-elastic and


thermal stresses are calculated for elliptical cooled and
flooded zones.
The fracture pressure is calculated from the pressure drops
(for a given injection rate) inside the cooled zone, the warm
flooded zone and the oil zone, respectively. The fracture
propagation criterion is based on a balance between the
fracture pressure and the rock stress and toughness. For
approximately constant stress and pressure along the fracture,
this results in:
p f = p flooded + poil + pres ,i = min,i + p +

K Ic
Df

.....(3)

For a sizeable fracture, the toughness-related term can be


neglected and the fracture propagation is based on the
calculated pressure drop and the stress on the fracture face. Of
course, both the pressure drop and the stress depend on the
fracture length. At a given time, the fracture length is found
from iteration on the fracture length to satisfy Eq. 3.
The following parameters must be estimated in order to
determine the fracture length:
pres : Reservoir pressure at start (and during injection). We
can assume that this is known from independent
measurements.
smin,i : Minimum stress before injection. This is a key
parameter to estimate fracture growth. We assume that the
stress is known from the Al Shomou-3 data-fracs.
Ap : Poro-elastic coefficient. Calculated from Biots
constant and the Poisson ratio. This becomes very important,
if it approaches unity. For the Athel reservoir, the Biot
constant is 0.5 and the Poisson Ratio is estimated as 0.2. This
yields a poro-elastic coefficient of 0.38.
kh : Transmissivity of the formation. Permeability must
often be determined from well tests, since core data are not
representative for large scale permeabilities. Core data of Al
Noor indicate a range for k of 0.001-3 mD.
re : Drainage radius. As yet, this is very uncertain in view
of the discontinuities due to cemented fractures.
Fracture Simulation Results
In the initial injection with 200 m, the pressure rose and fell
suddenly, indicating a breakdown. Such a breakdown would
only occur if there was no initial fracture (or only a very small
fracture). If the breakdown occurred in a thin layer with a
lower stress, we might explain the pressure with a fracture
propagating in the thin layer. However, there was evidence for
a sizeable initial fracture, so it is more likely that the
breakdown was due to fracturing into the cement annulus. The
pressure level with an initial fracture of some 5 m is indeed
expected to be relatively low.
Lets assume that a fracture was indeed created in the 5000
m3 injection. Applying the injection rate and a leak-off
coefficient from permeability and reservoir pressure, we have
modelled the fracture growth using both a conventional
hydraulic fracturing model as well as a waterflood fracture

model assuming constant fracture height. The significant


result is that both of these simulations predict a fracture that is
at least 100 m long.
Finally, we have simulated the pressure behaviour in the
last (16800 m) injection. We first looked at a simulation with
a static fracture. The injectivity is already high with a fracture
of 25 m, but after some time, the predicted pressure would be
much higher than the measured pressure (Figure 8). Also, the
character of the pressure is not well represented. The
measured pressure rises sharply and then stays almost
constant, whereas the simulated pressure rises slowly during
the whole period (a logarithmic curve that is expected for
constant transmissivity). The observed pressure is indicative
of fracture propagation, but the pressure was much lower than
the virgin closure pressure. It is possible to simulate the
pressure with a strong reduction in reservoir pressure, which
reduces the stress.
Since the production period was much longer than the
injection period, we could treat the depleted zone as a lowered
reservoir pressure. It was apparent from the simulations that
there is a significant depletion around the well. A block of
some 40 by 50 m has a pressure of some 330 bar. Of course,
during the injection the pressure would redistribute again and
we approximated that with a rise in the background reservoir
pressure. Figure 9 shows the results of the simulation with the
depletion and recharging of the reservoir. We see that the
fracture now propagates to 25 m. The flat pressure is caused
by the fracture extension. The low bottom hole pressure can
then be explained by the lowered horizontal stress due to the
depletion. In view of all the uncertainties in the reservoir
parameters and the actual flow geometry (e.g. flow behind the
casing), this simulation result is somewhat speculative, but it
offers a plausible mechanism of the fracture propagation in the
large injection. This can explain the enhanced reservoir
transmissivity after this injection, as shown by the last well
test.
Reservoir Simulation of Production Data
Reservoir simulation of the production periods following the
water washes should give us valuable information about any
changes in the properties of an induced fracture over time.
Reservoir simulation in this case was tricky for several
reasons:
Multi-phase production with only surface pressure
data and rapidly changing water cut during cleanup.
The production logs show a complicated, timedependent production profile.
Complicated history of alternating production and
injection
Possible changes in the flow geometry, due to the
formation of a cavity in the salt, would further
enhance the uncertainties in the actual reservoir
injection and production rates.
The ideal simulation would have been to model all
injection and production data with a 3-D multi-phase reservoir
simulation, including the effects of changing fracture
parameters over time. This was not practical within the scope

S.W. WONG, P.M. O'DELL, C.J. DE PATER, J. SHAOUL

of this study. Instead, we made several simplifying


assumptions and used a 2-D reservoir simulator developed
specifically for fractured well conditions.
The key assumptions made in order to make reservoir
simulation practical were as follows:
Use calculated bottomhole pressure from a vertical lift
program.
Ignore water production (cleanup) as a transient
phenomenon (except for calculations of bottomhole
pressure).
Model production separately after each water
injection. Use the same initial reservoir pressure (720
bar) in all cases, but keep the correct mass balance by
modelling the total cumulative oil production up to
that time.
Calculation of Bottomhole Pressure: In order to calculate
the bottomhole flowing pressure for history matching, we used
a vertical lift program. We used this program to calculate the
wellbore pressure drop at different points in the data,
whenever the physical properties were changing significantly
(i.e. every 10% increment in water cut). Between these
calculated points, we used interpolation. Since the relevant
properties were changing smoothly and in only one direction
(i.e. water cut is on a decreasing trend), this interpolation
should capture the correct physical behaviour. Of course, as
with any multi-phase flow problem, there are some
simplifying assumptions made in such a simulation.
Reservoir Simulation: In order to be consistent with the
welltest analysis of the various build-up tests, we used the
same kh value and the same k from the corresponding build-up
analysis. As explained in the welltest section, the apparent kh
increased after each water injection. We wanted to match the
initial production in each case, and we iterated on fracture
half-length to achieve this. The fracture half-lengths obtained
by this method are seen to increase following each water
injection, as was also seen in the buildup analysis. The
fracture lengths from reservoir simulation are a factor of 2 or 3
smaller than the estimates derived from the welltest analysis.
This is because the welltest model also included a positive
skin, whereas the reservoir simulator we used cannot model a
positive skin in the presence of a fracture.
The reservoir simulator (embedded in the hydraulic
fracture simulator 4) employed was developed especially for
the simulation of hydraulically fractured reservoirs. It is a 2dimensional finite difference model. The fracture is modelled
as 6 gridblocks. In order to model the overall change in
reservoir pressure seen between the first and last pressure
build-ups, we used a reservoir extent of 40 meters by 3048
meters; the long side was chosen to make it effectively
infinite. This gave a drainage area of 122,000 m. We used a
net pay height of 328 meters. This gives a total drained
volume of approximately 40 x 106 m.
To model the production data, we constrained the model
with calculated bottomhole flowing pressure, and matched on
rate. The rate used for matching purposes was only the oil rate,
adjusted to bottomhole conditions. The GOR varied between

SPE 62618

400 and 500 sm /sm. This gave a formation volume factor of


approximately 2.4.
We did not attempt to match the initial transient behaviour
at the start of production following each water injection. This
was not practical because this period was dominated by
cleanup of the injected water, which we were ignoring in our
simulations. Since this water was not originally in the
reservoir to begin with, producing it back is truly a transient
phenomenon.
We modelled the induced fracture using an equivalent
propped fracture with a proppant concentration of 0.25 to 0.5
kg/m. This gave an effectively infinite fracture conductivity,
with a dimensionless conductivity always greater than 10. For
such a low permeability reservoir, even a tiny fracture width
gives an effectively infinite permeability. In reality, the
fracture properties are probably changing over time. We
modelled this with a changing fracture length. We were able to
obtain quite good production history matches. All but the last
of the history match periods were about one week long. The
last period was about 60 days long. In the shorter production
periods, there was not a significant decline in the production
rate, compared to what we would expect for a very short
infinite conductivity propped fracture. This gives some
indication that over this short time scale (1 to 2 weeks); the
length and conductivity of the water-induced fracture are not
declining more rapidly than a propped fracture with the same
initial properties.
In the last production match (Figure 14) there is a very
steep decline in the first week, which is not matched. If we
match the initial production rate, the decline in the simulated
rate is much less than the decline in the actual rate. This may
be due to a decrease in the effective fracture length over time.
We matched the initial production with a half-length of 15
meters. To match the middle and final production, we had to
use a half-length of only 7 meters. This rapid decline was not
apparent in any of the earlier production data. The final
production data is quite noisy, due to pipeline storage effects
since during this time the well test unit was not being used.
We smoothed this data because we are interested in the
average rate. A summary of these results is shown in Table 4.
Plots of the production history matches are shown in Figures
10 to 14.
Reservoir Simulation of Al-Shomou 3: For comparison
purposes, we examined the results of the post-frac pressure
build-up data from Al Shomou 3. A much longer fracture halflength was observed in this well. The depletion effects
observed in Al Shomou 1 are not apparent in this well,
although the total production here was much less at the time of
this build-up. Perhaps with a propped fracture, several
reservoir blocks may be connected and drained, while with the
water wash, the fracture either cannot grow through cemented
faults, or it cannot stay open through the block boundaries.
Water production: Since we could not model the water
production in our reservoir simulations, we undertook a simple
volumetric analysis. Two interesting trends are visible in the
volume of water produced back after each injection. In the
initial production, the water cut is zero. After each water

SPE 62618

FRESH WATER INJECTION STIMULATION IN A DEEP TIGHT OIL RESERVOIR

injection, the fraction of water produced back changes. This


data is shown in Table 2. With the exception of the first small
injection, the percentage of water recovered increases for each
larger injection. With the exception of the last injection, the
average water production rate increases for each injection. If
the water is injected into a fracture, the rate at which the water
is produced back is also a function of the fracture
conductivity. This is another indication that an increase in
stimulation has occurred in each case with the larger injected
volume, except for the last and largest injection. In the welltest
analysis rate history, the water injections are modelled, as well
as the water production following the injection.
Discussion
We have shown much evidence that the reservoir was
stimulated by fracture propagation, originally suggested by
ODell3. The well test showed the half slope response typical
of a fracture. Initially, a fracture was present probably caused
by the high mud weight during drilling. Given the reservoir
transmissivity and the fracture size, we could predict the
pressure during injection of the 20 and 200 m injections. The
pressure was not expected to rise above fracture pressure.
However, the 1000 m and 5000 m injections probably
induced fracture extension. This agrees with the stimulation of
the reservoir, which was most apparent after the 5000 m
injection, which had the highest flow rate. This resulted in an
effective fracture length of some 30 m. Given this fracture size
we could simulate the pressure response during the large
(16800 m) injection. The pressure was low and hardly any
further fracture extension was expected. Alternatively, we can
explain this low injection pressure by the reduction of the
stress by depletion. At a much lower stress, the pressure may
have been sufficient to induce some fracture propagation.
Although the evidence is in favour of fracture stimulation,
contrary to salt dissolution from the matrix, one important
piece of evidence is lacking: the pressure during the 5000 m
injection was not recorded and we would need to confirm that
in this injection the fracture pressure was really exceeded.
It is possible that the stimulation of the reservoir was a
mixture of fracturing and salt dissolution: if the injections
leached salt from the cemented natural fractures, leading to a
highly anisotropic network of fractures with an enhanced
conductivity. In a well test, this would look similar to the
response of a hydraulic fracture. The main difference might be
that the dissolution could occur at a much lower pressure than
the horizontal stress. However, for the prediction of the
decline in productivity, a fracture would be equivalent to a
leached fracture system that is aligned in a plane.
Given the fracture size and transmissivity, we predicted the
productivity decline. To match the initial oil rate, we had to
use a fracture length that was much smaller than the length
that was obtained from the fracture growth simulations. We
found that the expected decrease in productivity was slower
than the actual fall in productivity. Both the small initial
fracture length and the decline in productivity can easily be
explained by healing of the fracture. However, if the
stimulation would be caused by salt dissolution from the

cemented natural fractures, we can also expect healing of


these fractures.
The results of the experiment indicate that water washes
are not very effective as stimulation technique for the Athel.
The propagated fracture heals quickly after the injection stops.
Apparently, the residual fracture width is insufficient and we
need to enhance the conductivity in the fracture. This can be
achieved with a standard propped fracture treatment.
However, it is also possible that addition of a small amount of
proppant would have given a better stimulation. As yet, it is
uncertain whether proppant is essential for the success of the
waterfrac treatments in Athel. The Athel reservoir conditions
appear to be less favourable for creating conductive fractures
with little proppant. Residual fracture width and fracture
roughness are related to the stress difference between the
stresses acting on the fracture plane. Since the horizontal
stresses are expected to be almost equal, we would expect
rather smooth fractures. Moreover, the success of the
waterfracs in East-Texas was largely caused by the reduced
cost of the treatments. In deep wells in Oman, the treatment
cost will be comparatively smaller compared to the total well
cost. This makes it less attractive to reduce the treatment cost.
Moreover, the cross-linked gel fractures in East-Texas showed
rising pressure that limited the fracture length. In Al Shomou3 the pressure was flat, which helps to create a long fracture.
Considering all these factors and the result of the water
squeezes, it is less likely that water fracs will be successful in
the Athel.
However, it may be feasible to test a waterfrac in a future
well. If it would be decided to try the waterfrac stimulation,
the treatment schedule would need to be drastically changed
with respect to the water squeezes. Proppant will need to be
added. Furthermore, an appropriate datafrac should be
performed. Also, the completion would need to be designed
for fracture stimulation. Given all these changes there is
potential benefit from further testing of the waterfrac
approach.
Further work is needed to develop a consistent method for
interpreting the well tests. In particular, the significance of the
faults that are apparent in the well tests should be investigated.
Analysis of the 32 day build-up as well as the final short
build-up in February 1997 show that the drainage area is most
likely an elongated rectangular shape, with a total width of
about 40 m. This boundary is most likely caused by cemented
natural fractures. Cores show evidence of cemented fractures
that could inhibit lateral flow. It will be important to determine
the lateral continuity of the reservoir for further development
planning. If lateral discontinuities would also limit the
drainage area around conventionally stimulated wells, it may
be beneficial to re-frac so that the secondfracture makes a
large angle with the original fracture plane. This may be
feasible in the Athel, since the original stress contrast between
the horizontal stresses is probably small. Therefore, a propped
fracture treatment could change the stress difference to give a
strong re-orientation of a subsequent re-fracture treatment.

S.W. WONG, P.M. O'DELL, C.J. DE PATER, J. SHAOUL

Conclusions
1. The initial fracture/reservoir transmissivity was able to
take the fluid from the 20 m injection without going
above fracturing pressure. For the 200 m and 1000 m
injections, the injection pressure would have been far
above fracture pressure, therefore we believe that these
two injections caused a small amount of fracture growth,
as well as stimulation of additional transmissivity.
2. We conclude that a large fracture of some 100 m must
have been propagated in the 5000 m injection, although
we have no record of the injection pressure.
3. Fracture simulations showed that the fracture should have
propagated to a length of some 100 m. However, the
effective fracture length from production data and
pressure build-ups was only 10-30 m. This indicates that
the fracture healed over most of its length after injection
stopped.
4. The initial 2 weeks of production after the 5000 m
injection did not show any accelerated production decline.
The production decline following the 32 day shut-in was
faster than expected for a fracture with a constant
conductivity. This shows that fractures should be propped
for optimum performance.
5. A fundamentally different production response was seen
after the largest water injection of 16,881 m (Dec. 96).
This response could not be explained with a simple
reservoir simulation. It is possible that a large cavity was
dissolved in the overlying salt layer during this injection.
Based on comparing water production rates after each
injection, it appears that there was not a comparable
increase in stimulation following the largest injection,
compared to after the 5000 m injection.
Nomenclature
C = wellbore storage coefficient.
Df = minimum dimension of fracture.
Lf = fracture half-length
k = permeability
h = reservoir net height
i = injection rate
pf = fracture pressure
pres = reservoir pressure.
Pi = initial pressure
Rinv = radius of investigation
Skin = skin factor
re = drainage radius.
t = time.
Xf = fracture half-length
Dpflooded = pressure drop in flooded zone
Dpoil = pressure drop in oil zone
Dsp = poro-elastic stress change
h = hydraulic diffusivity
m = fluid viscosity
smin,i = initial minimum horizontal stress

SPE 62618

References
1. Wong, S.W., Ford, S., and Turner, R., Massive Hydraulic
Fracture Stimulation in Deep, High-Pressure Athel Formation,
SPE 506124, SPE European Petroleum Conference, The Hague,
1998.
2. Mayerhofer, M.J., Richardson, M.F., Walker Jr., R.N., Meehan,
D.N., Oehler, M.W., and Browning Jr., R.R., Proppants? We
Dont Need No Proppants, SPE 38611, SPE Annual Technical
Conference, San Antonio, 1997.
3. ODell, P.M., The Athel: A challenging formation in South
Oman, SPE Reservoir Evaluation & Engineering, pp354-358,
1998
4. Crockett, A.R., N.M. Okusu and M.P. Cleary, "A complete
Integrated model for design and real-time analysis of hydraulic
fracture operations", SPE Paper 15069, Proc. 56th California
Regional Meeting, Oakland, CA, (US), 1986.
5. Perkins, T.K. and J.A. Gonzalez, Changes in earth stresses around
a wellbore caused by radially symmetrical pressure and
temperature gradients, SPEJ, pp129-140, 1984.
6. Koning, E.J.L., Waterflooding under Fracturing Conditions, PhD
Thesis, Delft University, 1988.

SPE 62618

FRESH WATER INJECTION STIMULATION IN A DEEP TIGHT OIL RESERVOIR

Table 1 - Activities during the water squeeze tests. The labels A-S refer to the labels in Figure 3.
Date

Activity

22-Nov-95 0
05-Dec-96
15-Dec-95
20-Dec-95

A
B
C

36 m perforations
8 day pressure build-up
PLT / downhole sampling

26-Dec-95 1
03-Jan-96

D
E

20m3 fresh water wash


20m3 acid squeeze

07-Jan-96
18-Jan-96

F 200 m3 water wash / PLT


G PLT

Volume Injection PLT/PBU Remarks


rate
3
3
m
m /day
PBU1

negative skin
PLT1 Producing 42 m3/day
PLT1a

20
20

420

200

267
333

24-Jan-96 3
26-Jan-96
01-Feb-96
15-Feb-96

Start 1000 m3 wash


1000 m3 water wash / PLT
PLT / TDT
16 day pressure build-up

1000

H
I
J

17-Mar-96 4
19-Mar-96
12-Apr-96
22-Jun-96
20-Sep-96
10-Oct-96
22-Oct-96

Start 5000 m3 wash


5000 m3 water wash / PLT
32 day pressure build-up
Produce to Birba Station
Hook-up Well Test Unit
PLT
PVT Sampling

5000

K
L
M
N
O
P

29-Nov-96 5
01-Dec-96
19-Dec-96
20-Feb-96
20-Mar-96
07-Jan-97

Start 16000 m3 wash


Q 16,881 m3 water wash

PLT4 Injecting 1000 m3


PLT5 Producing 90 m3/day
PBU2
negative skin
1429
PBU3

PLT6 Injecting 5000 m3


fracture response from half slope

PLT7 Producing 40 m3/day


16881

844

fracture response from half slope


PLT8 Injecting 16000 m3

R
S

Well Test Unit Inspection


shut-in for long term build-up

Injected Water Volume

PBU4
PLT9 Producing

Table 2: Summary of water production following each water injection.


Cleanup Period
Vol of Prod Water
Avg Water Prod. Rate

(m )
20
200
1000
5000
16881

PLT2 Injecting 200 m3


PLT3 Producing 46 m3/day

(days)
4
7
17
21
80

(m )
17
68
341
2515
10570

Water Recovered

(m /day)
4.25
9.7
20
120
132

%
85
34
34
50
63

10

S.W. WONG, P.M. O'DELL, C.J. DE PATER, J. SHAOUL

SPE 62618

Table 3 - Results of well tests during the water injection experiment.


Test
Date
Model

Pi

bar

pre-frac

pre-frac

05-Dec-96
One fault at 4 m

05-Dec-96
Bounded reservoir;
infinite conductivity
fracture
720

723

following 1000 m3
injection
15-Feb-96
Bounded reservoir;
infinite conductivity
fracture
752*

following 5000 m3 Following 16800 m3


injection.
injection.
12-Apr-96
20-Mar-96
Bounded reservoir; Bounded reservoir;
infinite conductivity infinite conductivity
fracture
fracture
834*
728

Tmatch

(day)-1

202

16.1

18.2

18.5

36.1

Pmatch

(bar)-1

.0115

0.00359

0.00319

0.00207

0.00628

m3/bar

0.0331

0.055

0.0497

.00504

0.0208

Ci/Cf

2.7

1.81

Alpha

46.3

16.1

Skin

-.83

.67

0.0798

0.0916

.32

Kh

md.m

4.99

1.64

2.95

4.98

8.53

md

.0152

0.005

0.009

0.0152

0.026

R.inv

15.2

8.7

16.2

30.4

14.7

Xf

10

24

32

Fault at

4.15

20

20

20

20

Fault at

20

20

20

20

* This value was not matched in these two cases, as explained in the text.
Table 4: Summary of reservoir simulation results.

Test Description
Prod. After 200 m3 water injection.
Prod. After 1000 m3 water injection
Prod. After 16 day pressure buildup
Prod. After 5000 m3 water injection
Prod. After 32 day buildup (matching only initial prod.)
Prod. After 32 day buildup (matching only later production)

Date

Jan 9-18, 1996


Jan 29 Feb 9, 1996
Mar 12-16, 1996
Mar 21- April 12, 1996
July 10 Sept. 16, 1996
July 10 Sept. 16, 1996

Fracture Half
Length
(m)
1.5
4.0
4.0
18.0
15.0
7.0

SPE 62618

FRESH WATER INJECTION STIMULATION IN A DEEP TIGHT OIL RESERVOIR

11

Top Natih-a
Base tertiary
Top Natih-e
Top Haima
Top Salt

ASH-3ST
ASH-2
ASH-1 ST2 ASH-3

Top Silicilyte
Top Buah

1000 m
Athel reservoir extension, mapped on
3D seismic
Mean case reservoir Area

Figure 1: Schematic cross-section through the Al Shomou slab.

Figure 2: Al Shomou well locations.

N - Hook-up Well Test Unit


O - PLT
P - PVT Sampling
Q - 16,881 m3 water wash
R - Well Test Unit Inspection
S - shut-in for long term build-up

H - 1000 m3 water wash / PLT


I - PLT / TDT
J - 16 day pressure build-up
K - 5000 m3 water wash / PLT
L - 32 day pressure build-up
M - Produce to Birba Station

A - 36 m perforations
B - 8 day pressure build-up
C - PLT / downhole sampling
D - 20m3 fresh water wash
E - 20m3 acid squeeze
F - 200 m3 water wash / PLT
G - PLT
300

Gross Rate, m3/d

Oil Rate, m3/d

THP, MPa

Choke, mm

200

250

180

200

160

150

140

100

120

50

100

80

14-Nov-95

12-Feb-96

12-May-96

10-Aug-96

08-Nov-96

06-Feb-97

07-May-97

-50

60
A B C D EF G H I

N O

-100

40

-150

20

-200

Figure 3: Rate and Pressure data from production history during water injections on Al Shomou 1.

12

S.W. WONG, P.M. O'DELL, C.J. DE PATER, J. SHAOUL

600

370

p
Fitted Line

550

SPE 62618

p
Fitted Line

350

p 500
(bar)

p
(bar)

450

330

310
400

290
350

270

300
250
-5.5

-5.0

-4.5

-4.0

-3.5

-3.0

-2.5

250
-9.7 -9.5 -9.3 -9.1 -8.9 -8.7 -8.5 -8.3 -8.1

-2.0

Figure 4: Pressure versus square root of time from 3rd buildup,


with fracture half length of 29 m.

Figure 5: Pressure versus square root of time from 4th buildup,


with fracture half length of 36 m

Behaviour during and after 20-5000 m3 Injection


Salt

Borehole

Silicilyte

Fracture

Initial Fracture
at low Stress layer

Growing Fracture
during Injection

Closing Fracture
during Production

Behaviour During and after 16000 m3 Injection


Cavity

Initial Fracture

Cavity Growth
by Salt Dissolution

Figure 6: Scenario's for Fracture growth and cavity dissolution.

Production by
Cavity Collapse

SPE 62618

FRESH WATER INJECTION STIMULATION IN A DEEP TIGHT OIL RESERVOIR

10

900

pseudo radial
1D Carter

10

Waterflood Fracturing

BHPobs
BHPsim

800

L/(t)
(-)

13

700

(bar) 600

Propped Fracturing

500
400

10

300
5

40

10

15

20

25

-1

10

30
Lf
Xf,test
i

30

1500

1000

(m)

20

i=50 m3/d

i=1000 m /d

10

-1

10

500

-2

(m /day)

10

10
5

10

i/2khp (-)

0
0

800

15

20

25

30

Figure 9: Modelled pressure during 16000 m injection. A fracture


of 75 m k length was needed to explain the injectivity.

Al Shomou 1
Production After 200m3 Injection

4000
psim
pobs
isim
iobs 3000

900

10

Time (days)

Figure 7: Dimensionless fracture length versus dimensionless


injection rate. For low injection rate we have a small fracture inside
a large flooded zone, whereas for large injection rate the flooded
zone is much smaller than the fracture length.
1000

Calc'd Btm Press (bar)


Bottomhole Press (bar)

400.0
400.0

HC Rate (m/day)
Production Oil Rate (m/day)

100.0
100.0

i
pbh 700

(m /day)

(bar) 600

2000

500

80.0
80.0

240.0
240.0

60.0
60.0

160.0
160.0

40.0
40.0

80.0
80.0

20.0
20.0

1000

400
300
200
0

320.0
320.0

0
10

20

30

Time (days)
Figure 8: Saphir simulation of 16800 m
fracture of 35 m.

injection with a static

0.0
0.0

0.00

2.00

4.00

6.00

8.00

Time (days)

Athel

Jan1996
3

Figure 10: Production match after 200 m water injection.

10.00

0.0
0.0

14

S.W. WONG, P.M. O'DELL, C.J. DE PATER, J. SHAOUL

Al Shomou 1
Production After 1000m3 Injection
Calc'd Btm Press (bar)
Production Oil Rate (m/day)

SPE 62618

Al Shomou 1
Production match after 5000 m3 injection.
HC Rate (m/day)
Bottomhole Press (bar)

Calc'd Btm Press (bar)


Bottomhole Press (bar)

HC Rate (m/day)
Production Oil Rate (m/day)

100.0
400.0

500.0
500.0

370.0
80.0

80.0
370.0

450.0
450.0

320.0
320.0

340.0
60.0

60.0
340.0

400.0
400.0

240.0
240.0

310.0
40.0

40.0
310.0

350.0
350.0

160.0
160.0

280.0
20.0

20.0
280.0

300.0
300.0

80.0
80.0

400.0
100.0

250.0
0.0

0.00

2.00

4.00

6.00

8.00

10.00

0.0
250.0

250.0
250.0

0.00

2.80

Athel

Jan-Feb 1996

11.20

14.00

0.0
0.0

April 1996
3

Figure 13: Production match after 5000 m water injection.

Figure 11: Production match after 1000 m water injection.

Al Shomou 1
Production match after 32 day pressure build-up

Al Shomou 1
Production after 16 day pressure build-up
Calc'd Btm Press (bar)
Bottomhole Press (bar)

8.40

Time (days)

Time (days)

Athel

5.60

400.0
400.0

HC Rate (m/day)
Production Oil Rate (m/day)

Calc'd Btm Press (bar)


Bottomhole Press (bar)

HC Rate (m/day)
Production Oil Rate (m/day)

150.0
150.0

500.0
500.0

370.0
370.0

120.0
120.0

450.0
450.0

200.0
200.0

340.0
340.0

90.0
90.0

400.0
400.0

150.0
150.0

310.0
310.0

60.0
60.0

350.0
350.0

100.0
100.0

280.0
280.0

30.0
30.0

300.0
300.0

50.0
50.0

400.0
400.0

250.0
250.0

0.00

2.00

4.00

6.00

8.00

Time (days)

Athel

March 1996

Figure 12: Production match after 16 day shut-in.

10.00

0.0
0.0

250.0
250.0

0.00

12.00

24.00

36.00

48.00

250.0
250.0

60.00

0.0
0.0

Time (days)

Athel

July 1996

Figure 14: Production match following 32 day pressure build-up.


Initial peak production is matched. Decline is much faster than
predicted by the model.

Das könnte Ihnen auch gefallen