Professional Documents
Culture Documents
Abstract
When the first frac pack was performed in the Gulf of Mexico
(GOM) over a decade ago, very little was known about the
effects this form of sand control would have on the intended
formation. Even less was known about how to optimize the
treatment to get the most benefit for the formation. Since that
time, the sand control community has learned a great deal
about the effects and benefits of frac packing various
unconsolidated formations throughout the world. However,
most of the knowledge and design criteria have been housed
within the minds of individuals and cannot be looked at as a
whole to find trends and fine tune the design methods
currently being used. Another complicating factor is the
number of frac models being used within the industry with
varying degrees of complexity. Therefore, even though
thousands of frac packs have been performed globally, fracpack redesign methods are still subjective and differ from
individual to individual and model to model.
The recent creation of a database that houses selected
formation evaluation test (FET) and frac data, along with
model-specific parameters, allows full-scale analysis of a large
number of jobs pumped in the Gulf of Mexico. With a
consistent analysis procedure in place, the database, populated
with numerous treatments by engineers working throughout
the GOM, can be analyzed objectively. The data contained in
this database include rock mechanics, net pressures, pumping
trend data, tip screenout (TSO) times, etc.
This paper explains the methodology and discusses the
results of the database analysis, using case studies to
determine the best method for analysis of the jobs. Crossplots
show the correlation between TSO prediction and actual
events and suggests recommendations for more successful
future design work. This paper is meant to give up-to-date
guidelines to help design better frac packs.
Introduction
Within the sand control community, the ability of an engineer
to redesign a frac pack from data generated during the
minifrac can sometimes be considered more art than science.
Often the engineer whose job it is to formulate the frac pack
treatment will use several different methods to arrive at a
solution deemed most correct. Many hours are dedicated to
determining the proper design for a frac-pack treatment. While
often the results cannot be argued, it is unwise and possibly a
waste of time to reinvent the wheel for every job. Rather, it
should be the goal of the sand control and frac-pack
communities to develop a design method that can determine
the most important parameters necessary to complete the job.
This method would be easily repeatable and could be used
throughout the GOM and possibly in other high-permeability,
unconsolidated regions of the world.
The purpose of this paper is to solve the problem
previously described. The goal of the project was to accurately
predict the TSO event such that the fracture geometry could be
better understood. In fracture theory, the TSO1 is the point at
which there is no longer any propagation of the fracture
length. Most fracture models treat this as the time at which the
first grain of sand is exposed to the tip of the fracture, ceasing
any further growth and allowing net pressure to accumulate
and build width throughout the length of the fracture.
The creation of a standard reliable method for predicting
the onset of the TSO event would enable engineers involved in
designing frac-pack operations to become more aligned in
their procedures, as well as provide more accurate results.
Knowing which knobs within the fracture model should be
manipulated in order to get the most accurate results would be
invaluable. This would allow for much quicker analysis of the
data accumulated from the minifrac thereby saving expensive
rig time. In addition, should multiple engineers be analyzing
the data, a single reliable method would allow a much more
concrete determination of any issues because everyone should
see very similar results. There have been many efforts to do
this in the past.2 This paper presents the methodology behind
developing a database system to track frac-pack treatment
data, what data was deemed necessary to a reliable model, and
the procedure for procuring that information from various
fracture treatments. Then, case studies are presented that prove
that the model, populated following the recommended
procedure, accurately predicts the TSO event in various
situations.
SPE 98171
Procedure
In fracture mechanics there are many variables that the
engineer can control, such as fluid viscosity, fluid density,
injection rates, volumes, and concentrations. However, there
are even more variables over which engineers have absolutely
no control, such as depth, stresses, formation modulus,
pressure, porosity, permeability, and fluid loss characteristics.3
It is the second set of variablesthose that engineers cannot
controlthat have plagued the sand control community. It is
essential when utilizing the most accurate models (3-D
models) to have a handle on the variables involved. To do this,
a database was set up to record all essential parameters,
recorded from minifracs and frac-pack treatments. However,
due to the large number of unknowns, it was necessary to
determine certain variables based on previously known
information. For this task, equations were developed for
certain parameters, namely Youngs Modulus, in each of the
varying rock layers:
E=
compressive stress
linear strain
Gradient
(psi/ft)
Young's
Modulus (psi)
Shale
0.637
2.19E+05
Shaley
0.631
2.04E+05
Sandy
0.625
1.87E+05
Sand
7651
0.619
1.71E+05
Formation
Type
SPE 98171
a.
3.
4.
Analysis of Data
The data from each frac-pack treatment was run through the
previously discussed procedure and the outputs described in
Part 4 of the procedure were entered into a Microsoft
Access database for ease of manipulation. Table 2 shows an
example of an input table used by the sand control engineers
for populating the database.
Table 2Re-Analysis Project Input Sheet
Job #
Input filename:
WellNameMiniMatch.inp
MiniMatch Stress Sand
MiniMatch Stress
Gradient Sand
MiniMatch Stress Shale
MiniMatch Stress
Gradient Shale
MiniMatch Composite
Layering Effect
Net Pressure Change
psi
psi/ft
psi
psi/ft
SPE 98171
Job #
Due to CLE
manipulation
MiniMatch Ct (Ctmini)
Trend of BHTP during
Minifrac Injection
MiniMatch DeltaP
MiniMatch Fluid Eff
Fluid Pumped after
MiniFrac
psi
1/2
ft/min
psi
%
Input filename:
WellNameForwardModel.inp
Forward Model Tip
Event
Actual Model Tip Event
min
min
Input filename:
WellNameJobMatch.inp
FracMatch Ct (Ctfrac)
FracMatch
PropDragCoefficient
(PDC)
Net Pressure Change
due to PDC
manipulation
1/2
ft/min
psi
4.35
5.24
5.5
7.133
7.43
8.21
9.65
10.2
8.7
10.8
11.1
11.5
11
13.5
10
14
15.75
16
16.7
17
15
17
34
17
30
22
28.6
30
26.5
Graphing the data from the first column against the data
from the second column shows how close the times are to
being an exact match. Given the fact that fracturing in reality
is an inexact science, the graph in Fig. 2 shows a very close
match for each point.
The orange line in Fig. 2 represents a true 1:1 slope and
would require a perfect match for the points to align exactly.
As is evident from the graph, the majority of data falls on or
very near the orange line, but there are a couple of points that
are not as closely aligned.
Fig. 3 shows the same graph with the unaligned points
highlighted. The reason these points are not as closely
matched as the rest of the data can be determined from some
of the other data points found in the database. One of these
points is a record of whether any fluid was displaced into the
formation following the minifrac. Table 2 shows a data point
labeled Fluid Pumped After Minifrac. This input was
included to determine if there were any benefits to displacing
fluid other than crosslinked fluid into the formation to attempt
a regain of the original leakoff characterization5 permeability.
Obviously, crosslinked gels cause, at the very least, minor
damage to the formation and therefore alter the leakoff from
the minifrac to the frac-pack treatment. The two points in
question were the only ones that did not have any fluid
displaced into the formation following the pumping of the
minifrac. All of the minifrac displacement fluid was reversed
out prior to pumping the frac-pack treatment. The data from
Fig. 3 indicates some sort of correlation between cleaning
the formation and being able to determine at what time the
TSO event is going to occur.
SPE 98171
Case Studies
This section provides a closer look at the minifrac analysis, the
forward modeling, and the post-job analysis of three wells
following the procedure detailed in the previous section. A
variety of information is presented, including the pressure
matches, the forward modeling charts, and the fracture cross-
sections. The first two cases show how effective the procedure
can be when used correctly and the third case is an example of
how not using a formation flush following the minifrac results
in less than adequate pressure responses. The properties for
the three wells used in the case studies are given in Table 4.
Depths (MD)
Perf Interval
(ft)
Closure
Pressure
(psi/ft)
BHP (psi)
Young's
Modulus (psi)
Permeability
(md)
13,106
14,378
40
0.72
8,085
367,000
275
16,115
21,434
138
0.63
6,700
202,500
500
12,036
12,994
156
0.75
8,042
236,000
100
Case
Study
Case Study 1
The first case study concentrates on a well found in the
deepwater GOM. The water depth for this particular well was
5,373 ft with a total depth of 14,378 ft MD. The frac-pack
treatment for this well was pumped using state-of-the-art
pumping equipment from a stimulation vessel. Using the
procedure described at the beginning of this paper, Youngs
Modulus was calculated to be 367,000 psi. This number was
left constant throughout the analysis of the treatment. The
minifrac test consisted of 12,600 gal of crosslinked fluid
pumped at a rate of 30 bbl/min. Fig. 4 shows the leakoff of the
test after being matched with the computer model using the
described procedure. Instantaneous shut-in pressure (ISIP) and
closure pressure (PC) match extremely well.
Following the match of the net pressures generated during
the minifrac test, the proposed job was redesigned. This new
treatment, consisting of the schedule shown in Table 5, was
pumped into the formation at a rate of 30 bbl/min until
screenout occurred.
Table 5Pumping Procedure for Case Study 1
Stage
Rate
(bbl/min)
Beginning
Proppant
Conc.
(ppg)
Ending
Proppant
Conc.
(ppg)
Clean
Volume
(gal)
30
30,471
30
0.5
0.5
4,089
30
4,024
30
4,634
30
3,730
30
4,210
30
3,367
30
30
computer model was then run, and the TSO event generated
by the computer matched almost identically to the actual job.
Fig. 5 shows the job net pressure vs. the net pressure
generated by the model. The scaling has been adjusted to show
how close the two different TSO events mimicked one
another.
Case Study 2
The second case study demonstrates how the procedure works
on a shallow water shelf well in the GOM. The water depth for
this particular well was 350 ft with a total depth of 21,434 ft
MD. The sand control/stimulation treatment was also pumped
off a stimulation vessel. The Youngs Modulus from the
previously described spreadsheet was calculated at 202,500
psi. This formation was less consolidated than the previous
case, causing the prediction of the TSO event to be more
difficult. The minifrac test for Case Study 2 consisted of 7,560
gal of crosslinked fluid pumped at a rate of 18 bbl/min. Fig. 6
shows the leak-off of the test after being matched with the
computer model using the described procedure. The two main
match points (ISIP and PC) again match extremely well.
Following the match of the net pressures generated during
the minifrac test, the proposed job was redesigned. The
redesigned treatment consisted of the schedule shown in
Table 6. This amount of fluid and proppant was pumped into
the formation at a rate of 18 bbl/min until screenout occurred.
Table 6Pumping Procedure for Case Study 2
Stage
Rate
(bbl/min)
Beginning
Proppant
Conc.
(ppg)
Ending
Proppant
Conc.
(ppg)
Clean
Volume
(gal)
18
10,769
3,325
18
0.5
1,375
1,555
18
1,196
18
10
2,509
30
1,539
30
11,447
18
10
10
3,862
18
12,850
SPE 98171
Stage
Rate
(bbl/min)
Beginning
Proppant
Conc.
(ppg)
Ending
Proppant
Conc.
(ppg)
Clean
Volume
(gal)
20
3,973
20
0.5
2,096
20
3,016
20
5,864
20
4,994
20
4,003
20
5,405
20
10
10
8,496
20
11,368
Conclusions
Several conclusions were made from the study of the case
histories uploaded into the database and documented in this
paper.
1. The availability of a database, which encompasses a large
cross-section of the GOM, is invaluable when attempting
to generalize about frac-pack treatments.
2. The equations used to estimate rock properties prove
sufficient when there is a need to standardize across a
large cross section.
3. The data indicates that use of a displacement linear fluid
pumped into the formation following the minifrac allows
the formation to return to a more natural state, thereby
making the information attained from the model more
relevant to the frac-pack treatment.
4. Stimulation models do not necessarily rely on the
information entered as much as the users ability to use
the data the simulator outputs. In other words, the same
data entered can give different answers that can vary
according to the experiences of the user.
5. The procedures detailed in the paper take away much of
the subjectivity of designing a frac pack treatment and
point toward a single answer, thereby removing much
confusion from the process.
References
1. Ellis, R.C.: An Overview of Frac Packs: A Technical
Revolution (Evolution) Process, Journal of Petroleum
Technology (Jan. 1998) 66-68.
2. Dusterhoft, R. et al: Improved Minifrac Analysis
Technique in High Permeability Formations, paper SPE
30103 presented at the SPE European Damage
Symposium, The Hague, The Netherlands, May 15-16,
1995.
3. Barree, R.D. et al: Use of Frac-Pack Pressure Data to
Determine Breakdown Conditions and Reservoir
Properties, paper SPE 36423 presented at the 1996 SPE
Annual Technical Conference and Exhibition, Denver,
CO, October.
4. Smith, M.B. et al: Multiple Frac Height Measurements:
A Case History, paper SPE 19092 presented at the Gas
Technology Symposium, Dallas, TX, June 1989.
5. Vitthal, S. et al: How Minifrac Alters Leak-off and
Ways to Counteract It, paper SPE 58767 presented at the
International Symposium on Formation Damage Control,
Lafayette, LA, February 2000.
SPE 98171
40
35
30
Actual (min)
25
20
15
10
0
0
10
15
20
25
30
35
SPE 98171
40
35
Outlying Points
30
Actual (min)
25
20
15
10
0
0
10
15
20
25
30
Fig. 4Minifrac Analysis Match using the described procedure. Model net pressure and the observed net
match nicely at the ISIP and the closure point.
35
SPE 98171
Fig. 5Case Study 1 post-job match. Notice how the model net pressure (white) and the job net pressure (yellow)
match at the point of TSO.
Fig. 6Case Study 2 Minifrac Analysis Match where the model net pressure and the job net pressure match.
10
Fig. 7Case Study 2 post-job match. The TSO event of both the model net pressure and the actual net pressure
match very closely.
Fig. 8Case Study 3 Minifrac Analysis Match where the model net pressure and job net pressure match.
SPE 98171
SPE 98171
Fig. 9Case Study 3 post-job match. The TSO event of the model net pressure occurs much earlier than the actual
net pressure.
11