Sie sind auf Seite 1von 14

COMPLETION/STIMULATION

Cracking Rock: Progress in Fracture Treatment Design

In the 1950s, hydraulic fracturing


was a hit-or-miss proposition.
Through the 60s and 70s, better
data quality and more sophisticated models of rock mechanics
improved control over the fracture
job. Today, with cost-effective,
high-power computing, two-dimensional (2D) models of fracture
propagation are giving way to a
three-dimensional (3D) approach.
Fracture treatment design has
never before been so powerful or
flexible a tool.

Barry Brady
Bobby Poe
Jack Elbel
Houston, Texas, USA
Mark Mack
Hugo Morales
Ken Nolte
Tulsa, Oklahoma, USA

For their help with this article, thanks to Larry Behrmann,


Schlumberger Perforating Center, Rosharon, Texas, USA;
Simon Bittleston, Schlumberger Cambridge Research,
Cambridge, England; CJ de Pater, Delft Technical University, The Netherlands; Cor Kenter and Jacob Shlyapobersky Koninklijke/Shell Exploratie en Produktie Laboratorium, Rijswijk, The Netherlands; Paul Martins, BP
Exploration (Alaska) Inc., Anchorage, USA; and George
K. Wong, Shell Bellaire Research, Houston, Texas, USA.
In this article, NODAL, DataFRAC and ZODIAC (Zoned
Dynamic Interpretation Analysis and Computation) are
marks of Schlumberger. VAX is a mark of Digital Equipment Corp. and Sun is a mark of Sun Microsystems, Inc.

The idea of hydraulically creating cracks in


a pay zone to enhance production was
developed in the 1920s by R.F. Farris of
Stanolind Oil and Gas Corp. He evolved the
concept following a study of pressures
encountered during squeezing of cement,
oil and water into formations. In 1947,
Stanolind (now Amoco Production Co.) per-

formed the first experimental hydraulic fracture in the Klepper #1 gas well in Grant
County, Kansas, USA. Deliverability of the
well did not improve appreciably, but the
technique showed promise, and the following year Stanolind presented a paper on the
Hydrafrac process.1 Halliburton Oil Well
Cementing Company obtained a license to
the process and, in 1949, performed the first
commercial fracturing treatments, raising
production of two wells outstandingly.2

Oilfield Review

The method took off. By 1955, treatments


reached 3000 wells per month, and by
1968, more than a half-million jobs had
been performed. Today, hydraulic fracturing
is used in 35 to 40% of wells, and in the
United States, where the procedure is most
widespread, it has increased oil reserves by
25 to 30%.3 Interest in hydraulic fracturing
shows no signs of abating.4 Application of
the technology is expanding from mainly

October 1992

1. Clark JB: A Hydraulic Process for Increasing the Productivity of Wells, Transactions of the AIME 186
(1949): 1-8.
2. Waters AB: History of Hydraulic Fracturing, presented at the SPE Hydraulic Fracturing Symposium,
Lubbock, Texas, USA, 1982.
3. Veatch RW Jr, Moschovidis ZA and Fast CR: An
Overview of Hydraulic Fracturing, in Gidley JL,
Holditch SA, Nierode DE and Veatch RW Jr (eds):
Recent Advances in Hydraulic Fracturing, Monograph
12. Richardson, Texas, USA: Society of Petroleum
Engineers (1989): 1-38.

4. Warpinski NR: Invited Paper: Rock Mechanics Issues


in Completion and Stimulation Operations, in Tillerson JR and Wawersik WR (eds): Proceedings of the
33rd US Symposium on Rock Mechanics. Santa Fe,
New Mexico, USA (June 3-5, 1992): 375-386.

4000

Fracture treatments/yr

3000

Remove
damage

Tight gas;
goal of 10
increase

North American
activity declines;
gas deregulation

Moderate/high
perm; goal
of 2
increase

2000

Middle East
imports to
North America

1000

Improved
materials,
understanding
OPEC supply restrictions

0
1950

1960

1970

1980

1990

2000

Year

nChanging motivation for hydraulic fracturing. The three parts of the graph with positive slope indicate three motivations: initially, to remove damage, then to improve tenfold the productivity of tight gas sands, and today, to double productivity of mediumto high-permeability formations.
low-permeability reservoirs to medium-to
high-permeability settings (above ).
Hydraulic fracturing is the pumping of fluids at rates and pressures sufficient to break
the rock, ideally forming a fracture with two
wings of equal length on both sides of the
borehole. If pumping were stopped after the
fracture was created, the fluids would gradually leak off into the formation. Pressure
inside the fracture would fall and the fracture would close, generating no additional
conductivity. To preserve a fracture once it
has been opened, either acid is used to etch

the faces of the fracture and prevent them


from fitting closely together, or the fracture
is packed with proppant (usually sand) to
hold it open. This article concentrates on
the latter technique.
Today, a typical fracturing treatment uses
thickened fluids pumped in stages. The first
stage is a pad of water, a polymer and
additives. Then comes the slurry, which is
pad plus proppantgenerally sandin suspension. Different concentrations of proppant and volumes of slurry are pumped as
the job progresses (below ).

Pressure exerted by the pad initiates and


propagates the fracture. The slurry helps
extend the fracture and transports proppant.
The fracture gradually fills until the proppant packs into the fracture tip (next page ).
At this point, the fracture treatment is finished and pumping stops. As pressure
within the fracture declines, the fracture
closes on the proppant pack, ensuring that it
remains in place, providing a conduit for
hydrocarbons. Productivity would be inhibited by viscous fluid in the pad and slurry
that remains in the formation. However,
when the fluids high viscosity is no longer
needed, the high temperature of the formation or special oxidizers cause the fluid
break to a lower viscosity, allowing it to
be produced back.5
Hydraulic fracturing lies at the interface of
fluid mechanics and rock mechanics. In the
45 years since the first fracture job, fluid science has advanced significantly. Treatment
fluids have been diversified to handle many
temperature, chemical and permeability
conditions (see Rewriting the Rules for
High-Permeability Stimulation, page 18).
Additives control a range of fluid properties,
such as viscosity, pH, stability and loss of
fluid to the formation, called leakoff.6 Many
proppants have been developed, from the
standard silica sand to high-strength proppants, like sintered bauxite and zirconium
oxide particles, used where fracture closure
stress would crush sand.

Job Description Information


Stage
Name

Pump
Rate

Fluid
Name

bbl/min.

Proppant
Concentration

gal

lbm/gal

Proppant Type
+ Mesh

Estimated Surface
Pressure
psi

Pad

35

YF140

5000

INTERPROP + 20/40

5630

Slurry

35

YF140

9000

INTERPROP + 20/40

4610

Slurry

35

YF140

14,000

INTERPROP + 20/40

3760

Slurry

35

YF140

23,000

INTERPROP + 20/40

3080

Slurry

35

YF140

15,000

INTERPROP + 20/40

2460

Slurry

35

YF140

13,200

6170

nA typical pumping schedule for a


hydrofrac in a gas well in east Oklahoma, USA. Each unit of fluid that
represents a change in proppant
concentration or flow rate or both is
called a stage; a specific sequence
of stages is called a pumping
schedule. This is a pumping schedule to produce a 909-foot [277-m]
fracture. The pad fractures the rock
and helps transport the proppant,
which holds the fracture open after
pressure is released. A major component of fracture design is establishing the volume and chemistry of
pad and slurry. Generally, the pad
6

Stage Fluid
Volume

is the largest stage, accounting for


30 to 50% of fluid, and, rarely, up to
70%. Ideally, to optimize the
propped fracture length, the pad is
completely leaked off at the
moment the fracture reaches its
intended length. If the pad leaks off
too soon, the fracture will be too
short; if too late, the fracture is not
effectively propped. In this well, five
slurry stages with different proppant
concentrations and volumes are
used, but as many as 17 or 20 slurry
stages may be used in large frac
jobs. The later slurry stages have
higher proppant concentrations
than earlier stages because the
slurry fluid leaks off as it travels
along the fracture. Therefore, a

slurry concentration that starts at the


wellbore as 2 lb of proppant per gallon of fluid [240 kg/m3], may end up
as 8 lbm/gal [960 kg/m3] at the end
of pumping, and 44 lbm/gal [5270
kg/m3] when the fracture closes. In
this job, one proppant size is used
(20/40 refers to a standard sieve
mesh size that permits passage of a
particle with an average diameter of
0.63 mm [0.025 in.] ). A larger proppant is sometimes used near the wellbore to minimize turbulent flow,
which would decrease hydrocarbon
flow rate.

Oilfield Review

October 1992

25% slurry volume pumped

Height, m

30

15

50% slurry volume pumped

Height, m

30

15

75% slurry volume pumped


30

Height, m

15

0
0

50

100

Distance, m
Proppant concentration, vol %

Until recently, advances in rock mechanics lagged somewhat behind those in fluid
technology. In the 1950s, there was no need
for a rigorous theory of fracture propagation,
the backbone of fracture treatment design.
Low-volume, low-rate and low proppant
concentration fracture stimulation succeeded without careful design. But as treatments grew in size and complexity, operators needed more control. Today more than
ever, the expense of hydraulic fracturing
requires that the operator knows how the
formation will respond to treatment, and
whether the treatment designthe selection
of pump rates, fluid properties, pumping
schedule and fracture propagation model
will create the intended fracture (see To
Frac or Not to Frac? next page ).
Pivotal to designing the treatmentand to
deciding whether to do one at allis costbenefit analysis, relating cost of the fracture
job to increased well productivity. The more
fracture length for a given fracture conductivity, the more productivity, but also the
more costly the fracture job. This analysis,
called net present value, is done with simulators that find the optimum fracture length
and conductivity for a given payback schedule. Too short a fracture, or too low a conductivity, and the increase in well productivity wont cover the cost of the fracture
treatment; too long, and the extra fracture
length will add significantly to cost but negligibly to production. Some simulators
model fracturing economics in longer terms;
they tell, for example, for a well with a
given deliverability, amortized at a certain
rate, how much should be spent on
hydraulic fracturing given a future oil price.
In the past few years, improvements in
fracture design have come from developments in several areas:
Fracture geometry modeling. Mathematical models today can better predict how
in-situ rock responds to fracturing.
Relationship of perforation design and
fracture initiation (see The Shape of Perforation Strategy, page 54 ). Careful
design of perforations can minimize pressure drop at the borehole.
Fracture treatment evaluation. Mathematical advances have also made evaluation
tools more powerful. There is a growing
practice of testing the validity of the fracture geometry model against postfracture
well test data, then refining the model.
This back analysis permits prediction of
fracture parameters, particularly fracture
length and conductivity, to be compared
with independent field measurements.

0
5
10
15
20

Initial
fracture
geometry
at wellbore

25
30
35
65

5. Gulbis J, Hawkins G, King M, Pulsinelli R, Brown E


and Elphick J: Taking the Brakes off Proppant-Pack
Conductivity, Oilfield Review 3, no. 1 (January
1991): 18-26.
6. Overviews of fracturing fluids:
Constien VG: Fracturing Fluid and Proppant Characterization, in Economides MJ and Nolte KG (eds):
Reservoir Stimulation, 2nd ed. Englewood Cliffs, New
Jersey, USA: Prentice Hall (1989): 5-15-23.
Ely JW: Fracturing Fluids and Additives, in Gidley
JL, Holditch SA, Nierode DE and Veatch RW Jr (eds):
Recent Advances in Hydraulic Fracturing, Monograph
12. Richardson, Texas, USA: Society of Petroleum
Engineers (1989): 130-146.

nAn investigational proppant transport


model, showing variation of proppant
concentration at three times during fracturing. This simulation, by Simon Bittleston at Schlumberger Cambridge
Research in England, predicts the final
distribution of proppant, used for quantifying fracture conductivity. Yellow is no
proppant, green to dark blue is low to
high proppant concentrations, respectively, and red is packed proppant. Slurry
is denser than pad so it tends to slump,
called gravity current. After 50% of the
slurry volume is pumped, a shower of settling proppant appears as a light blue fog
near the tip of the propagating slurry.
Falling proppant results in a packed bed
(red) along the bottom of the fracture. This
packed bed restricts downward growth of
the fracture. As a result of this proppant
distribution modeling, the pumping
schedule can be modified to optimize
fracture design. Although still a research
tool, it may later be integrated into fracture design programs.
7

To Frac or Not to Frac?

Fracture Geometry Modeling

Determine if the well is providing the maximum benefit, indicated


by return on investment and net present value.

Evaluate permeability and skin (near well damage) from well test.

Determine benefit using NODAL


analysis for various
combinations of:
Recompletions (tubing size,
perforations, surface
equipment, artificial lift)
and
Matrix treatments
(different materials and sizes)
or
Fracture treatments
(different material and sizes).

Maximum benefit achieved for


recompletions only?

Yes

Perform recompletion.

No
Perform matrix
treatment
(see Trends in Matrix
Acidizing, page 24).

Yes

Maximum benefit achieved after


matrix treatment only?
No
Is maximum benefit achieved after
matrix treatment with recompletion?

Yes
Perform recompletion.

No
Perform fracture
treatment.

Yes
Is maximum benefit achieved
after fracturing only?
No
Is maximum benefit achieved after
fracturing with recompletion?

Yes
Perform recompletion.

No
Fracturing not needed.

7. Hubbert MK and Willis DG: Mechanics of Hydraulic


Fracturing, Transactions of the AIME 210 (1957):
153-166.
8. Barree RD: A New Look at Fracture Tip Screenout
Behavior, paper SPE 18955, presented at the SPE
Joint Rocky Mountain Regional/Low Permeability
Reservoirs Symposium and Exhibition, Denver, Colorado, USA, March 6-8, 1989; Journal of Petroleum
Technology 43 (February 1991): 138-143.
Clifton RJ and Abou-Sayed AS: A Variational
Approach to the Prediction of the Three-Dimensional
Geometry of Hydraulic Fractures, paper SPE/DOE
9879, presented at the SPE/DOE Low-Permeability
Gas Reservoirs Symposium, Denver, Colorado, USA,
May 27-29, 1981.

Clifton RJ: Three-Dimensional Fracture-Propagation


Models, in Gidley JL, Holditch SA, Nierode DE and
Veatch RW Jr (eds): Recent Advances in Hydraulic
Fracturing, Monograph 12. Richardson, Texas, USA:
Society of Petroleum Engineers (1989): 95-108.
Hongren G and Leung KH: Three-Dimensional
Numerical Simulation of Hydraulic Fracture Closure
with Application to Minifrac Analysis, paper SPE
20657, presented at the 65th SPE Annual Technical
Conference and Exhibition, New Orleans, Louisiana,
USA, September 23-26, 1990.
9. The PKN model is from the work of Perkins and Kern,
revised by Nordgren to account for flow rate gradients
in the fracture.
Nordgren RP: Propagation of a Vertical Hydraulic
Fracture, Society of Petroleum Engineers Journal 12
(August 1972): 306-314; Transactions of the AIME 253.
Perkins TK and Kern LR: Widths of Hydraulic Fractures, Journal of Petroleum Technology 13 (September 1961): 937-949; Transactions of the AIME 222.

The need to understand hydraulic fracturing


stimulated advances in basic rock mechanics. A key finding was of Hubbert and
Willis, in 1957, showing that fractures in the
earth are usually vertical, not horizontal.7
They reasoned that because a fracture is a
plane of parting in rock, the rock will open
in the direction of least resistance. At the
depth of most pay zones, overburden exerts
the greatest stress, so the direction of least
stress is therefore horizontal (next page,
top). Fractures open perpendicular to this
direction and are therefore vertical. In shallow wells, or where thrusting is active, horizontal stress may exceed vertical stress and
horizontal fractures may form.
By the 1960s, fractures created below
1000 or 2000 ft [300 to 600 m] were
accepted as vertical. Operators then posed
some difficult questions: How high does the
fracture grow? How can we prevent it from
extending into the gas or water zone? How
does fracture height relate to fracture width
and length? And how do we optimize fracture dimensions?
A major task of rock mechanics became
the prediction of fracture height, length and
width for a given injection rate, duration of
injection and fluid leakoff. Needed for this
prediction is a model of how a fracture
propagates in rock.
Today, a number of models occupy a continuum from 2D to pseudo-three-dimensional (P3D) and fully 3D. The basic difference between 2D and P3D/3D models is
that in 2D models, fracture height is fixed or
set equal to length (that is, a semicircular
shape), whereas in P3D and 3D models,
fracture height, length and width can all
vary somewhat independently. Two-dimensional models have been around for about
30 years; three-dimensional for about ten
years. Increased computing power has
recently made pseudo-3D models practical
for routine design. Fully 3D models have

10. Khristianovic SA and Zheltov YP: Formation of Vertical Fractures by Means of Highly Viscous Liquid,
Proceedings, Fourth World Petroleum Congress,
Rome, Italy, section 2 (1955): 579-586.
Geertsma J and de Klerk FA: Rapid Method of Predicting Width and Extent of Hydraulically Induced
Fractures, Journal of Petroleum Technology 19
(December 1969): 1571-1581; Transactions of the
AIME 246.
11. Ahmed U: Fracture-Height Predictions and PostTreatment Measurements, in Economides MJ and
Nolte KG (eds): Reservoir Stimulation, 2nd ed.
Englewood Cliffs, New Jersey, USA: Prentice Hall
(1989): 10-110-13.
12. Van Eekelen HAM: Hydraulic Fracture Geometry:
Fracture Containment in Layered Formations, paper
SPE 9261, presented at the 55th SPE Annual Technical Conference and Exhibition, Dallas, Texas, USA,
September 21-24, 1980.

Oilfield Review

limited use because of lengthy computation


time, but they are the way of the future.
State-of-the-art fully 3D models simulate
nonplanar fractures, but most commercial
versions are planar.8
Most 2D models are based on three common models: the Perkins-Kern-Nordgren9
(PKN) model, the Khristianovic-Geertsma-de
Klerk10 (KGD) model and the radial model
(below). The PKN and KGD models assume
fracture height is constant along the length
of the fracture; height is usually picked by
lithologic boundaries. Fracture length and
width are then calculated from height
(which may be estimated using acoustic log
data combined with modeling of fracture
mechanics and elastic properties11 ), Youngs
modulus, fluid viscosity, injection rate and
time and leakoff. In the radial model, fracture length and height are equal and are
jointly allowed to vary. Width is also
allowed to vary.
The 3D approach is more realistic
because fracture height is not determined by
lithology but by vertical variation in the
magnitude of least principal stresses, which
often but not always follow lithologic units.
(The greater the vertical contrast in least
principal stresses, the better fracture height
is contained.12 )

Vertical
stress

Sv
St

Max
horiz.
stress

Min.
horiz.
stress

Sr

nStresses in the earth act in three principal directions, one vertical, and two horizontal, a maximum and a minimum. At
the borehole wall, these are vertical, S v ,
radial, S r , and tangential, S t . Vertical
stress induced by overburden usually
exceeds the two horizontal components.
This means a fracture will have the least
resistance to opening along a plane normal to the smallest principal stress.
Because this stress is horizontal, the fracture will orient vertically. In areas of
active thrusting, and in some shallow
wells, a horizontal stress may exceed
overburden and the fracture will form
horizontally. Regional tectonic forces
determine the azimuthal orientation of the
least principal stresses and thus of the
fracture wings.

The emergence of 3D models has not


eclipsed 2D models. Two-dimensional models work where:
The fracture grows in a formation of homogeneous stress and mechanical properties
so that fracture height is small compared
to formation layer thickness. The radial
model is appropriate in this setting.
Stress contrasts are high between the pay
layer and neighboring formations and
these contrasts follow lithologic boundaries. The PKN or KGD models, which
assume constant height, are appropriate in
this setting.
When these conditions are absent, use of
2D models requires estimation of fracture
height based on the users experience and
knowledge. The consequences of underestimating fracture height, for example, range
from disastrous to troublesome but manageable. The fracture may extend into a gas or
water leg, which can ruin a well. Underpredicting fracture height overpredicts fracture
length because, for a given pump rate,
unanticipated doubling of fracture height
decreases length by about 50%, depending
on leakoff. If the fracture is shorter than predicted, it may not be as productive as forecast. The pump schedule may be inappropriate, further cutting fracture conductivity.

2D Fracture Models

Pressure required
to extend fracture

PKN

Elliptical cross section


Width height
Width < KGD;
length > KGD
More appropriate when
fracture length > height

nThe family of
basic 2D fracture
modelsPKN,
GDK and radial.

Time
Fracture
height fixed
Pressure required
to extend fracture

KGD

Rectangular cross section


Width length
More appropriate when
fracture length < height

Time

Radial

Appropriate when fracture


length = height

Pressure required
to extend fracture

Fracture
height not
fixed

Time

October 1992

P3D Fracture

For example, proppant concentrations may


be excessive, causing proppant to plug the
fracture before flowing its full length, and
leaving some fracture length unpropped.13
The evolutionary step after 2D modeling
is P3D modeling.14 When conditions are
ideal for a 2D modelhigh, known stress
contraststhe P3D model height prediction
may be more accurate than the estimated
height of the 2D model (below ). The advantage of the P3D approach is that it does not
require estimating fracture height, but it

2D versus P3D/3D Fracture Models


for Different Bed Boundary Stress Contrasts

High contrast

Low contrast

High contrast

Low contrast

does require input of the magnitude of minimum horizontal stress in the zone to be
fractured and in the zones immediately
above and below. (It calculates height using
this stress and the fluid pressure within the
fracture.) The stress values may be estimated
from a mechanical properties log, an indirect measurement.
On a small scale, the best direct stress
measurement is from several microfracs,15
in which small fractures are created at several wellbore locations (below ). Fracturing
fluid is usually water without proppant. On
the reservoir scale, determination of stress
and fluid loss is accomplished by a calibration treatment, in which a fracture is created
without proppant that is up to one-third the
length of the planned fracture. The engineer
analyzes the curve of pressure decline versus time after the rock has been fractured
(next page, top). Finding the fracture closure

2D
4200

P3D
/3D

High contrast

Low contrast

High contrast

Low contrast

Well depth, ft

4600

Log
derived

Microfrac test

5000

5400

5800

nA P3D fracture propagating from the borehole (top) and comparison of 2D, P3D/fully

3D models for high and low contrast in minimum horizontal stress between beds. A low
stress contrast is on the order of a 100 psi [690 kilopascals (kPa)]; a high stress contrast
is greater than 1000 psi [6895 kPa]. Here, if one assumes that fracture height of the 2D
model is selected based on lithology, not on stress contrast, then the 2D fracture model
stays within the beds. In the low-contrast case, the 2D model will probably overestimate fracture length and underestimate height, compared to the P3D/fully 3D models.
In the low-contrast case, there would be a slight length and height difference between
the P3D and fully 3D models. In the high-contrast case, the P3D and fully 3D models
would predict about the same geometry.
13. Nierode DE: Fracture Treatment Design, in Gidley
JL, Holditch SA, Nierode DE and Veatch RW Jr (eds):
Recent Advances in Hydraulic Fracturing, Monograph 12. Richardson, Texas, USA: Society of
Petroleum Engineers (1989): 223-244.
14. Ben-Naceur K: Modeling of Hydraulic Fractures,
in Economides MJ and Nolte KG (eds): Reservoir
Stimulation, 2nd ed. Englewood Cliffs, New Jersey,
USA: Prentice Hall (1989): 3-13-31.

10

15. Daneshy AA, Slusher GL, Chisholm PT and Magee


DA: In-Situ Stress Measurements During Drilling,
Journal of Petroleum Engineering 38 (August 1986):
891-898.
Sarda JP, Detienne JL and Lassus-Dessus J, Recommendations for Microfracturing Implementations
and the Interpretation of Micro- and Pre-Fracturing, Revue de lInstitut Franais du Ptrole 47, no.
2 (March-April 1992): 179-204.
16. Nolte KG: Fracture Pressure Analysis: Deviations
from Ideal Assumptions, paper SPE 20704, presented at the 65th SPE Annual Technical Conference and Exhibition, New Orleans, Louisiana, USA,
September 23-26, 1990.

2200

2600

3000

3400

Minimum horizontal stress, psi

nStress profile measured by


microfrac and derived from wireline log data. Most correlations
between log-derived and measured stresses are linear and
show more deviation than this
example.
17. Martins JP, Bartel PA, Kelly RT, Ibe OE and Collins PJ:
Small Highly Conductive Hydraulic Fractures Near
Reservoir Fluid Contacts: Application to Prudhoe
Bay, paper SPE 24856, presented at the 67th SPE
Annual Technical Conference and Exhibition, Washington DC, USA, October 4-7, 1992.

Oilfield Review

October 1992

Bottomhole pressure, psi

9000

nEffect of closure
stress on a pressure/time curve. In
this idealized
example, interpretation of the slope
to find horizontal
stress is straightforward. Changes in
curve slope are not
always so clear.

Pressure decline

Fracture
treatment

Fracture
closing

8000

Fracture closes
on proppant
7000

6000

Reservoir
pressure

Closure pressure =
minimum horizontal rock stress

5000
38

40

42

44

46

48

50

56

58

Time, hr
Pressure required to extend fracture, psi

pressure, which equals the minimum horizontal stress, requires interpretation of the
slopes, which is open to ambiguity.16 The
drawback of the microfrac method is its
high cost and insensitivity to stress variation
from well to well and across a field. The
leakoff estimation is also complicated when
fractures grow into impermeable layers,
where leakoff will not be proportional to
fracture area.
P3D models assume a simplified representation of fluid flow in the fracture. This
assumption is made mainly to shorten computation time, but it may result in inaccurate
estimation of fracture height. This is because
pressure distribution in the fracture, which
controls growth of fracture height, is generated by the fluid flow.
Although this problem seems simple
enough to solve, it requires the leap to fully
3D modeling of fracture geometry. Fully 3D
simulators are difficult to usethey require
accurate stress contrast dataand so are not
widely employed, but the theory permits the
closest approximation of what fractures
really do. The two main differences
between fully 3D and P3D are in how they
handle fluid flow and pressure calculation
along the fracture. Fully 3D geometry models use a fully 2D model of fluid flow,
whereas P3D models approximate the 2D
fluid flow. In a fully 3D geometry model,
pressure everywhere is used to calculate
fracture width at any point. Width is generally calculated using the pressure integral
along the total fracture length and height. In
the P3D model, the pressure-width relation
is simplified to improve efficiency, usually
by considering only particular shapes, such
as ellipses, or by neglecting variation of
pressure along the fracture length.
At BP, fully 3D models are not used routinely because of lack of appropriate input

nPressure versus
time for lateral
coupling compared with traditional fracture
models.

300

250

Lateral
coupling
200

PKN
150

KGD
100

50
0

20

40

60

80

Time, min

data. They are used to understand fracture


propagation in a particular field.17 Where
fracture containment is poor, 3D models
have been used to assist microfrac interpretations and to generate simple models for
routine fracture design. These simple models are refined by posttreatment evaluation.
The pressure integral advantage of the
fully 3D model has been introduced to PKN
and P3D models using a method called lateral coupling. This is a way to introduce 3D
elasticity to models that dont include it.
Mathematically, lateral coupling puts back a
gross approximation of the pressure integral
along the fracture length. This poor-mans
integral couples pressures at points along
the fracture, instead of considering them in
isolation. Compared with conventional PKN

and P3D modeling, it doubles or triples


computation time, but improves estimation
of fracture height and fracture pressure during treatment (above ).
A third evolutionary stage, multilayer fracture (MLF) modeling, takes one step back in
order to take two steps forward. The MLF
simulator is a revision of PKN modeling that
permits describing the geometry of more
than one fracture forming in more than one
layer and then planning the appropriate

11

pumping schedule.18 (below ). Multilayer


modeling was needed as more reservoirs
were exploited in which conventional modeling has limitations. This is often the case
when stress barriers prevent the coalescing
of fractures in multiple zones or where layers of varying thicknesses and stress magnitudes are to be fractured.
The MLF approach indicates whether a
single treatment or separate treatments are
needed to achieve optimum geometry of
fractures in multiple zones. If separate treatments are needed for the desired penetration in each layer, the MLF simulator may
be used to determine how many are
required. It can also help in planning limited entry perforatingvarying the number
of perforations in each layer, depending on
layer thickness and stress state, to achieve
the desired fracture geometry. (Fewer perforations in the layer taking the most fluid
restricts flow and diverts it into other layers.)
Inputs to the MLF model are the same as
for P3D: stress profile, Youngs Modulus and
leakoff for each formation. The model differs from existing descriptions of multilayer
fracturing in that it quantifies transient fluid
partitioning during pumping as a function of
fracturing fluid and formation properties.
Existing models calculate partitioning only
Gamma
ray

at a single time or for a limited number of


formation characteristics.19
The MLF model also allows the prediction
of crossflow between fractures after pumping stops and before all the fractures close.
Matching the predicted and measured crossflow permits a more accurate prediction of
the parameters that determine fluid volume
that enters each zone, and the resulting fracture length and height.
With the arrival of the MLF model, the
engineer can choose from five general types
of fracture propagation models. Selection of
the right model is critical. Even slight differences between modeled and actual fracture
dimensions can translate to dramatic differences in required proppant concentration
and weight, and pad volume (next page ).
Usually, PKN, KGD and radial models are
chosen with a chain of empirical deductions. The engineer estimates the shape of
the induced fractureif length exceeds
height, its PKN; if length is less than height,
its KGD. This value is based the sand thickness to be fractured, proximity to gas, water
or other fractures and estimation of the
stress contrast between the reservoir section
and abutting formations, usually shales. The
stress contrast estimate is often valid when
the well has clean sands and clean shales.

The Perf and the Frac: Whats the Link?

Field wisdom holds that the ideal perforation lies in the plane normal to the minimum far-field stress direction. This perforation links most directly with the induced
fracture, minimizing pressure drop near the
borehole. Other perforations probably connect with the fracture indirectly, if at all. But
because fracture azimuth is generally not
known and because alignable perforating
guns are not readily available, conventional
guns shooting at closely spaced angles
around 360 are generally used. These are
called phased guns. The closer the angle
(phasing) between perforations, the better
chance of having more perforations in or
near the ideal plane. Not until recently,
however, were large-scale experiments performed to evaluate the relationship between
perforations and hydraulic fractures.
Behrmann and Elbel of Schlumberger and
Dowell Schlumberger, respectively, used
full-scale perforators on steel casing
cemented into sandstone blocks placed in a

2D

P3D

MLF

Perfs

Perfs

Layered beds

The estimate becomes tenuous in silty shale,


which may have the same stress magnitude
as sand but may poorly contain fracture
height. Again, the best measurement of
stress is obtained from a microfrac.

Shale

Sand

nComparison of 2D, P3D and multilayer fracture (MLF) models in a multilayer setting. In the 2D model, fracture

height is selected to be limited by the top of the upper sand and bottom of the lower sand. The fracture is considered to grow simultaneously from both sands and to be of uniform length. Youngs Modulus is averaged for the
two sands and the shale between them. In the P3D model, the fracture grows from one sand to the other, but not
simultaneously as in the 2D model. In both the 2D and P3D models, fracture lengths are equal for both the thick
and thin sands. In the MLF model, which uses a modified PKN model, fracture lengths and heights are unequal.
Length depends on fracture height, stress magnitude and Youngs Modulus. As with other 2D models, height is
selected for each layer, here by lithologic boundaries. The next generation MLF model will adapt P3D modeling.

12

Oilfield Review

2000

0.75

KGD
0.50

PKN
0.25

Fracture penetration, ft

Treatment cost, $ 106

1.0

PKN
1500

KGD
1000

500
0

750

1500

2250

3000

2.5
2.0

KGD

1.5
1.0

PKN

0.5
0
0

750

1500

80,000

160,000

240,000

Fluid volume, gal

2250

3000

Fracture conductivity, md-ft

Proppant weight, lb 106

Fracture half-length, ft
2900

KGD

2400
1900
1400

PKN
900
400
0

750

1500

2250

3000

Fracture half-length, ft

Fracture half-length, ft

Comparison of Fracture-Design Calculations for Different Fracturing Models


KGD
Pad volume, bbl
Proppant-laden fluid volume, bbl
Average sand concentration, lbm/gal

Perkins-Kern

Nordgren

750

1,350

1,650

1,250

650

350

2.5

3.5

157,500

68,350

51,000

Viscosity after pad, cp

36

36

36

Created fracture length, ft

698

804

845

Total amount of sand, lbm

Effective fracture length, ft

486

240

185

Created fracture width, in.

0.22

0.17

0.16

Effective fracture width, in.

0.20

0.16

0.16

Effective fracture height, ft

98

94

85

Average fracture conductivity, darcy-ft

7.1

6.5

6.5

Adapted from Veatch RW Jr, et al, reference 3.

nComparison of fracture properties for PKN and KGD fractures (top four graphs) and for
three fracture models (bottom).

triaxial stress cell. 20 They made several


observations about the relationship between
perforation orientation and stress direction.
They found that fractures initiate from the
wellbore wall in the optimum hydraulic
fracture direction, from perforations nearest
this direction, or both. Fractures tend not to
form at other perforations.
The best perforation-to-fracture communication is achieved when perforations are
within 10 of the far-field minimum horizontal stress. This means that perforations
not optimally oriented may result in a large
pressure drop, or proppant bridging, when

October 1992

pad and slurry flow around the annulus to


the fracture. As expected, the maximum
number of perforations in communication
with the fracture is achieved with a perforating gun having the smallest possible angle
between shots.
Another finding of Berhmann and Elbel
concerns pump rate and viscosity of the
prepad, a low-viscosity fluid sometimes
pumped ahead of the pad. It has been long
recognized that a prepad can increase pore
pressure, and thereby decrease fracture initiation pressure. The lower the initiation pressure, the lower the pressure required.
Behrmann and Elbel, after cutting apart the

sandstone blocks, found that slow pumping


of low-viscosity prepad has another effect: it
maximizes the number of fractures initiated
at perforations suboptimally aligned. More
work is needed to determine whether
increasing suboptimally aligned fractures
reduces pressure drop at the well, which
would improve deliverability.
Pearson and colleagues at ARCO Alaska
Inc. aligned perforations normal to the minimum far-field stress in deviated wells. They
used perforating guns with a downhole orientation motor in conjunction with realtime navigation tools. This enabled placement of larger, more productive fractures.21
Pearson and colleagues suspect that posttreatment skin damage may be associated
with pressure drops from poor communication between the main fracture and fractures from perforations that are not aligned
normal to the minimum far-field stress.
Analysis of the ARCO results by CJ de Pater
and colleagues at Delft Technical University in The Netherlands suggests that Pearsons results may be inconclusive.22 Pearson and colleagues changed a number of
parameters (such as multiple zone to single
zone perforation and gun size) that may
have equally explained their ability to place
larger treatments.
18. Elbel JL, Piggott AR and Mack MG: Numerical
Modeling of Multilayer Fracture Treatments, paper
SPE 23982, presented at the SPE Permian Basin Oil
and Gas Recovery Conference, Midland, Texas,
USA, March 18-20, 1992; Journal of Petroleum
Technology 43 (May 1991): 608-615.
19. Ahmed U, Newberry BM and Cannon DE: Hydraulic
Fracture Treatment Design of Wells with Multiple
Zones, paper SPE/DOE 13857, presented at the
SPE/DOE 1985 Low Permeability Gas Reservoirs Symposium, Denver, Colorado, USA, May 19-22, 1985.
Ben-Naceur K and Roegiers J-C: Design of Fracturing Treatments in Multilayered Formations, SPE
Production Engineering 5 (February 1990): 21-26.
20. Berhmann LA and Elbel JL: Effect of Perforations on
Fracture Initiation, paper SPE 20661, presented at
the 65th SPE Annual Technical Conference and
Exhibition, New Orleans, Louisiana, USA, September 23-26, 1990.
21. Pearson CM, Bond AJ, Eck ME and Schmidt JH:
Results of Stress-Oriented and Aligned Perforating
in Fracturing Deviated Wells, paper SPE 22836,
presented at the 66th SPE Annual Technical Conference and Exhibition, Dallas, Texas, USA, October 69, 1991.
For details of the aligned and oriented perforating
technique:
Yew CH, Schmidt JH and Yi L: On Fracture Design
of Deviated Wells, paper SPE 19722, presented at
the 64th SPE Annual Technical Conference and Exhibition, San Antonio, Texas, USA, October 8-11, 1989.
22. de Pater CJ, personal communication, 1992.

13

Today, the center of controversy in fracturing is a fundamental concept called fracture


toughness, a measure of energy dissipated
by fracture growth. Established thinking
holds that fracture toughness is a material
property that is independent of fracture size.
The focus is on energy dissipated at the fracture tip, considered to be a very small zone.

Another school of thought, led by investigators at Shell, mainly Jacob Shlyapobersky,


maintains that fracture toughness is not a
material property, and that it increases with
fracture size.26 This point of view holds that
fracture toughness is the release of energy
not at the fracture tip but within a large
zone of irreversible deformation around the
fracture tip. The volume of this zone is
thought to increase with fracture size.
These two views lead to different explanations for the creation of fracture width,
which is directly related to net pressure
(fracture propagation pressure minus closure
pressure). The size-dependent school says
fracture width is larger and only weakly
affected by fracture fluid viscositythat is,
that net pressure is not sensitive to viscosity.
This is because net pressure, in order to
overcome the large, size-dependent toughness, creates a fracture width large enough
to make viscous flow effects negligible.
According to established thinking, because
toughness is not size-dependent and has a
conventional magnitude, pressure gradients
from viscous flow dominate the toughness
effect and fracturing, and create smaller
fractures than those modeled by the sizedependent toughness school.
The two schools, therefore, have different
calculations of fracture length and required
pad volume. The size-dependent school
maintains that the established view will
underestimate width and therefore overestimate fracture length for a given fracture volume. This is because net pressure, according to the established view, is determined
mainly by viscosity and not, as the size
school holds, by viscosity and increasing
fracture toughness. The established view
maintains that apparent error in estimation
of fracture length and width does not result
from size-dependent toughness but from use
of an inappropriate fracture geometry or
reservoir model.27
Another area of investigation concerns the
assumption that rock behaves as a purely

26. Shlyapobersky J, Walhaug WW, Sheffield RE and


Huckabee PT: Field Determination of Fracturing
Parameters for Overpressure Calibrated Design of
Hydraulic Fracturing, paper SPE 18195, presented
at the 63rd SPE Annual Technical Conference and
Exhibition, Houston, Texas, USA, October 2-5, 1988.
Shlyapobersky J, Wong GK and Walhaung WW:
Overpressure Calibrated Design of Hydraulic Fracturing, paper SPE 18194, presented at the 63rd SPE
Annual Technical Conference and Exhibition, Houston, Texas, USA, October 2-5, 1988.
Lewis PE: Analysis of Treatment Data Yields CostEffective Fracturing, The American Oil and Gas
Reporter 35, no. 1 (January 1992): 32-34, 36-38.
Shlyapobersky J: Energy Analysis of Hydraulic Fracturing, Proceedings of the 26th US Symposium on
Rock Mechanics, Rapid City, South Dakota, USA
(June 26-28, 1985): 539-546.

Shlyapobersky J and Chudnovsky A: Fracture


Mechanics in Hydraulic Fracturing, in Tillerson JR
and Wawersik WR (eds): Proceedings of the 33rd
US Symposium on Rock Mechanics. Santa Fe, New
Mexico, USA (June 3-5, 1992): 827-836.
27. Elbel J and Ayoub J: Evaluation of Apparent Fracture
Lengths Indicated From Transient Tests, paper
CIM/AOSTRA 91-44, presented at the CIM/AOSTRA
Technical Conference, Banff, Alberta, Canada, April
21-24, 1991; Canadian Journal of Petroleum Technology (in press).
Nolte KG and Economides MJ: Fracture Length
Determination and Implications for Treatment
Design, paper SPE 18979, presented at the SPE
Rocky Mountain Regional/Low Permeability Reservoir Symposium and Exhibition, Denver, Colorado,
USA, March 6-8, 1989; Journal of Petroleum Engineering 43 (September 1991): 1147-1155.

Enhanced Fracture Treatment


Evaluation

Fracture design may be fine-tuned by careful postjob evaluation. This tells whether the
job went as planned, and tests the validity
of the plan and the variables on which it
was based (see Design of an Ideal Fracture
Treatment, next page). Postfracture evaluation requires a drawdown and buildup test,
which indicates fracture skin and whether
the actual fracture length and conductivity
match those planned. This testing is not a
common procedure because operators are
usually hesitant to stop production for the
10 to 14 days required for the buildup. But
in some fields, the practice is becoming
more common in a few, select wells. For
example, in BPs Ravenspurn South field in
the UK sector of the North Sea, an extensive
program of data collection and analysis was
performed on the first six development
wells. This included extensive pre-and postfrac well testing, logging and recording of
bottomhole pressures during job execution.
The program helped optimization of job
design for the remainder of the field, leading
to significant reduction in the number of
wells required.23
A typical problem is that posttreatment
transient pressure analysis shows the fracture is shorter than indicated by the volume
and leakoff of pumped fluid. There could be
several reasons for the disparity. A common
reason, however, is that most postfracture
evaluation models assume ideal reservoir
conditionshomogeneous and isotropic
formations, uniform fracture width and conductivity and absence of skin damage.24
To get away from assuming ideal reservoir
conditions, Schlumberger has made several
improvements to the ZODIAC Zoned
Dynamic Interpretation, Analysis and Computation program. This program improves
evaluation by accounting for variation in
fracture conductivity and width along the
fracture length, for reservoir permeability
anisotropy and for fracture face skin dam23. Martins JP, Leung KH, Jackson MR, Stewart DR and
Carr AH: Tip Screen Out Fracturing Applied to the
Ravenspurn South Gas Field Development, paper
SPE 19766, presented at the 64th SPE Annual Technical Conference and Exhibition, San Antonio,
Texas, USA, October 8-11, 1989.
24. Walsh DM and Leung KH: Post Fracturing Gas Well
Test Analysis Using Buildup Type Curves paper SPE
19253, Offshore Europe 1989, Aberdeen, Scotland,
September 5-8, 1989.
25. Poe BD, Shah PC and Elbel JC: Pressure Transient
Behavior of a Finite Conductivity Fractured Well
With Spatially Varying Fracture Properties, paper
SPE 24707, presented at the 67th SPE Annual Technical Conference and Exhibition, Washington DC,
USA, October 4-7, 1992.

14

Conventional
postfracture well test

ZODIAC / P3D

nPostfracture interpretation of fracture


geometry by conventional pressure transient analysis and with the ZODIAC program. The main difference is that conventional analysis does not account for
spatial variation in fracture conductivity
and width, assumes fracture height
equals bed thickness, and ignores fracture face skin damage. The blue area is
ignored in the conventional analysis.
age.25 It also does not link fracture height
with bed thickness (above ), but uses a P3D
approach to permit variation in propped
fracture height and width in the analysis.
Compared to conventional postfracture
pressure transient analysis, the program
takes 10 to 15% more computer time on a
VAX or Sun workstation. In the future, it will
include capabilities to model the effects of
reservoir boundaries and high-velocity flow
on fracture length and conductivity estimates. The effects of reservoir boundaries
are often observed in transient tests of long
duration. These effects can be used to estimate the area and shape of the drainage
area of the well.
The Fracture Frontier: Rock Mechanics

Oilfield Review

Design of an Ideal Fracture Treatment

Fracture skin or lower fracture


conductivity?

Select fluids and additives that minimize


formation and proppant damage and
environmental impact.

Different reservoir model


permeability? Is reservoir
anisotropic? Layered?
Stress sensitive?

Obtain permeability and reservoir pressure


from well test; porosity from logs.

If not done earlier, perform microfrac to


determine correct model, fluid loss
coefficient and treatment efficiency (volume
of fluid pumped versus volume of fracture,
determined mainly by leakoff).

Different fracture geometry


model or length?

Stress revision.

If appropriate fracture geometry model not


known, do microfrac (1/3 to 1/2 length of
actual job, no proppant) to select fracture
geometry model (2D, P3D, MLF).

Frac model revision.

Test for different fracture


model or less length.

Fluid revision.

Improved or expanded stress


and modulus data.

Obtain stress magnitude and Youngs


Modulus1 versus depth from logs, cores.
Also collect other well and formation
information: lithology, natural fracture
locations, porosity. Check offset well data.

Iteration for revisions.

Use net present value (NPV) calculation to


select proppant, optimize pump schedule
and fracture length, and predict production.

Finalize pump schedule with PLACEMENT


program. The program gives pressure
required during job, frac length at end of
job and distribution of proppant.

Execute job.

Do well test and use ZODIAC


program to evaluate fracture
treatment and reservoir
characterization.

No
Is well producing as expected?
Yes

Analyze bottomhole pressure


during execution with various
fracture models.

No

Was bottomhole pressure


during execution as expected?

Yes

Fracture treatment
design is optimal.

1. Youngs Modulus is the ratio of stress (force per unit area) to strain (displacement per unit length).

October 1992

15

elastic continuum, meaning that deformation short of fracturing is fully reversible.


There is evidence that high-permeability/
high-porosity formations may be elastoplastic, meaning they have some component of
irreversible deformation (below ). Further
work on this is becoming possible with the
increase in computer power needed to solve
equations for nonelastic behavior, which are
far more complex than those for elastic
behavior. Significant nonelastic behavior
would affect the prediction of fracture
geometry and the analysis of fracture pressure data.
The Fracture Frontier: High-Angle Wells

Field experience in highly deviated and horizontal wells shows that it is possible to perform hydraulic fracturing in these settings,
but the effect on well performance is still
uncertain. Little has been published on the
effect of fracturing on deviated well performance. 28 Shell investigators found that
reduced productivity is expected from a
fractured deviated well compared to a fractured vertical well.29 This is because the axis
of the wellbore may not lie in the preferred
fracture plane and may intersect the fracture
over only a small reservoir interval. This

empirical curve showing the maximum


borehole deviation that will allow development of a single fracture.
Hallam and Last made these observations
based on studies in which they cemented or
cast a liner in a block of rock, then loaded
it. Work by CJ de Pater and colleagues
shows that if the block is first loaded, then
the liner is cemented, fracture geometry will
be different.32
Work by Hugo Morales at Dowell
Schlumberger, using a 3D fracture simulator
that permits curved fractures, shows that
fracture initiation pressure can be calculated
for deviated wells, given well inclination,
azimuth and direction of principal stresses.
But once the fracture starts, there is not yet a
calculation for propagation pressure. This is
because fracture propagation models do not
address how multiple fractures affect nearborehole stresses. A general recommendation, however, is that flow rate should be
high enough to reduce bridging of proppant
associated with pressure drops of multiple,
small fractures (next page ).
An evolving capability is triaxial borehole
seismic imaginglistening from three directions to sound emitted by the fracture as it
closes, then triangulating its location to find

results in limited communication to the


borehole during fracturing and a pressure
drop that inhibits productivity. In the Prudhoe Bay field, BP has found that fracturing
can impair the performance of highly deviated wells.30
Nevertheless, the increasing number of
deviated and horizontal wells has inspired
work on fracture modeling in these settings.
Today, fracture treatment design in these
wells is largely by rule of thumb. But several
observations have been made by Hallam
and Last of BP that can enhance treatment
design in deviated wells:31
When perforation tunnels are not normal
to the minimum stress, fractures reorient
in the preferred direction. If tunnels are
short compared to their spacing, the fractures will curve before linking up, resulting in further pressure drop. Perforation
length should therefore be at least onethird to one-half tunnel separation, that is,
4 to 6 in. [10 to 15 centimeters (cm)].
Perforation densities should be 6 shots/ft
at 60 phasing and 360/ shots/ft for
phasing.
A single large fracture is more productive
than several smaller ones that may not
link up. Hallam and Last constructed an
Conceptual Deformation Models

Continuous
solid

Fracture

Elastic/brittle or
elastoplastic

Planes of
continuous weakness

Discrete
blocks

Elastic and discontinuous plastic

Random
fractures

Plastic

CONTINUUM

nSeveral modes of rock response to stress. In rock mechanical terms, they are elastic continuous deformation,
brittle failure, discontinuous deformation of block-jointed rock, and pseudocontinuous deformation and plastic yield of heavily fractured rock. Current theories of fracturing and treatment design are limited because
they use elastic continuous deformation and brittle failure almost exclusively.

16

Oilfield Review

fracture length.33 This would provide valuable feedback in development of fracture


propagation models. Still, the weakest link
in the models is probably stress magnitude
determination. A confident measurement of
stress, by an economical and practical
method, would provide the required data
for evolving a fracture propagation model.
Probably as important as technical
improvements is a change in the engineering mindset. If only I had a fully 3D model,
all my problems would go away is perhaps
just half true. Often, the most sophisticated
fracture propagation models and fracture
treatment designs are undermined by something as simple and elusive as bad permeability data. In 3D modeling, major limitations remain in input datait is still difficult
to obtain valid stress profiles, fluid-loss profiles and fracture conductivities.
Today, fully 3D models help generate simpler models for routine application. Careful
postfracture evaluation allows the engineer
to tune fracture design, yielding the most
from the simplest approaches. Tomorrow,
increased computer power may place the
curving fracture of varying height and width
within reach of engineers in the field. JMK

28. One notable paper on the subject to date: Ovens J:


The Performance of Hydraulically Fractured Stimulated Wells in Tight Gas Sands: A Southern North
Sea Example, paper SPE 20972, presented at
Europec 90, The Hague, The Netherlands, October
22-24, 1990.
An overview of fracturing horizontal wells:
Soliman MY, Hunt JL and El Rabaa AM: Fracturing
Aspects of Horizontal Wells, paper SPE 18542, presented at the SPE Eastern Regional Meeting,
Charleston, West Virginia, USA, November 1-4,
1988; Journal of Petroleum Technology 42 (August
1990): 966-973.
Brown E, Thomas R and Milne A: The Challenge of
Completing and Stimulating Horizontal Wells, Oilfield Review 2, no. 3 (October 1990): 52-62.
29. Veeken CAM, Davies DR and Walters JV: Limited
Communication Between Hydraulic Fracture and
(Deviated) Wellbore, paper SPE 18982, presented
at the SPE Joint Rocky Mountain Regional/Low Permeability Reservoirs Symposium and Exhibition,
Denver, Colorado, USA, March 6-8, 1989.
30. Martins JP, Dyke GC, Abel JC, Ibe OE, Stewart G,
Bartel PA and Hanna RR: Analysis of a Hydraulic
Fracturing Program Performed on the Prudhoe Bay
Oil Field, paper SPE 24858, presented at the 67th
SPE Annual Technical Conference and Exhibition,
Washington, DC, USA, October 4-7, 1992.
31. Hallam SD and Last NC: Geometry of Hydraulic
Fractures From Modestly Deviated Wellbores,
paper SPE 20656, presented at the 65th SPE Annual
Technical Conference and Exhibition, New Orleans,
Louisiana, USA, September 23-26, 1990.
32. de Pater CJ, personal communication, 1992.
33. Vinegar HJ, Willis PB, DeMartini DC, Shlyapobersky
J, Deeg WFJ, Adair RG, Woerpel JC, Fix JE and Sorrells GG: Active and Passive Seismic Imaging of
Hydraulic Fractures in Diatomite, paper SPE
22756, presented at the 66th SPE Annual Technical
Conference and Exhibition, Dallas, Texas, USA,
October 6-9, 1991.

October 1992

Min.
horizontal
stress

Max.
horizontal
stress

Max.
horizontal
stress

Min.
horizontal
stress

Minimum
horizontal stress

Time 1

Time 2

Time 3

nOrientation of hydraulic fractures in horizontal wells as a function of stress directions

(top) and, in a deviated well, evolution of small, multiple fractures that may contribute
to pressure drop at the wellbore (bottom). In the horizontal well example, only one large

fracture forms if the wellbore axis is normal to the minimum horizontal stress. If the
wellbore axis parallels the minimum horizontal stress, fractures form at each perforation. The end fractures are highest because they are affected on only one side by the
compressive stress exerted by the opening of the neighboring fracture. Height of these
end fractures tends not to exceed 2 to 3 borehole diameters. The time-lapse view (bottom) shows fractures developing tails that reach up and down the wellbore. By time 3,
they coalesce into one fracture. In so doing, rhomboids of rock are isolated between the
perforations. Small fractures develop here that may contribute to pressure drop at the
wellbore and early bridging of proppant.

17

Das könnte Ihnen auch gefallen