You are on page 1of 13

.

SPE 15069
A Complete Integrated Model for Design and RealTime Analysis
of Hydraulic Fracturing Operations
by AR. Crockett and N.M. Okusu, Resources

Massachusetts

Engineering Systems Inc., and M.P. Cleary,

Inst. of Technology

SPE Members
~opyright 1986, Society of Petroleum Engineers
rhis paper was prepared for presentation at the 5Sth Califomis Regional Meeting of the Society of Petroleum Engineers held m Oakland, CA, April 24,
!9ss,
rhis pa~r was selected for presentation by an SPE Program Commmes following review of mlormation contained in an abstract submitted by the
suthor(a). Contents of the psper, as presented, have not bean reviewad by the Sosiefy of Petroleum Enginaare and are aubjact to correction by tha
guthor(a). The material, aa presented, doas not necessarily reflect any poaifion of the $ociaty of Petroleum Engineers, its offiiere, or members. Papera
praaentad at SPE meetings are subject to publication review by Editorial Committees of the Socialy of Petroleum Enginaera. Permission to copy is
restricted to an abstract of not more than 200 words, Illustrations may not be copied. The abstract should contain conspicuous acknowledgment of where
nndby whom the pa~r is presented. Write Publicatiina Manager, SPE, P.O. Sox 83SSSS,FUchardaon, TX 7S0S3-3S3S. Telex, 7309S9, SPEDAL.

Abstract

Introduction

A new comprehensive model of hydraulic fracturing is presented which has been developed for the Gas Research Institute
(GRI) mobile fracture monitoring and analysis facility. The
main purpose of the model is to simulate the hydraulic fracturing process in real-time, that is on+ite during the fracturing
operation, but the model can also be used for pre-fracture design and post-fracture analysis. Sensor data obtained during
the course of the job such as wellhead pressure, flow rates,
frac-fluid viscosity, and proppant staging can be received
directly by the model se input, superseding the prefrac job
design schedule, and making possible more accurate model estimatea of current fracturing conditions and predictions of final
fracture geometry, as the job proceeds.

A substantial amount of effort has been invested, especially over the past decade, in the development of models for
design and analysis of hydraulic fracturing. The resulting models have varied in at least three major aspects: realism and
generality of the assumptions made in formulating the modela; complexity of the resulting computer codes and machine
requirements; and flexibility of the input-output characterietics, especially in relation to real job conditions and operator
.
interfacing. Although some good progress haa been made by
many groups, there have not been any models which were satisfactory in all of the three important areas, and most models
h ave, at beet, been adequate in one aspect only.

Examples of previous work may be found in the considThe overall model haa four major components describing:
erable literature which has evolved on this subject. The sim flow of fluids and slurry in tubular goods
plest models, assuming 2D geometry with a constant specified
creation and propagation of the hydraulic fracture
height, were those of Christianovich, Geertsma, de Klerk and
transport of proppant, deposition, and fracture closure
Daneshy (CGD, Refs. 1, 2), having width dependent on length,
heat and fluid exchange between fracture and reservoir.
and those of Perkine, Kern and Nordgren (PKN, Refs. 3, 4),
This fully-integrated, numerically robust model of the hy- having width dependent on height: these two models are really
draulic frauturing process takes directly into account as much valid only when length greatly exceeds height (or vita-versa if
of the essential physics as possible, given the computational
the models are turned on end) so clearly neither are approlimitations of the real-time application. Additional informa- priate for the typical field condition where the fracture is not
tion, pertaining to very complex reservoir characteristics, can very @l contained.
be indirectly supplied to the model through data-based results
Since both the height and length may grow substantially
obtained prior to the job from other more comprehensive [e,g.,
3D, croes+wctional) fracture simulations. Having thereby em- during the job in contradiction to these 2D models an
a priori specification of height is hardly acceptable, even if
bedded more elaborate fracture analysis into a simple lumped
based
on (usually tenuow) deductions from well 10SS. To remodel, accurate pre~lctions are achieved at execution speeds
faster than real-time, allowing on-the-job analysis and even solve thm problem, Cleary (Ref. 5) developed a pseudo-three.
dlmeneional hydrafrac model (P3DH) of simultaneow length
real-time history matching for unknown reservoir parameters.
and height growth, having width dependent mainly on the
Sample results are presented from model simulations of a
lesser of height or length. This model wee adopted by varifracture treatment performed in the Travis Peak formation of
ous groups in industry (e.g., Refis. 6, 7) while others developed
East Texas. Actual sensor data from the job was used as input
specialii
variationa on the same theme (e.g., R@. 8, 9). Aland model predictions are compared with field measurements.
though these P3DH-type models, if fully developed, do indeed
References and illustrations at end of paper.

9
m

.A
. fY3MPl,F!TF!
. . . . . . ----

REAL-TIME
.. ..

1. Frac-fluid/slurry
flow in tubular ~oods and perforations, including effects of cross-linking, foaming and proppant
concentration in gelled fluid flow rheology. This incorporates
both laboratory and field data on pipe flow, the latter based
on simultaneous uphole and downhole pressure measurement
wherever such data has been collected: a major purpose is
to eliminate the need for downhole data collection, by being
confidently able to calculate true downhole fracturing pressure
from measured uphole pressures.

jerve to make first-order estimates of fracture geometry for


.athe: uniform reservoir and pumping conditions, the crossIeciional approach implicit in these models 10SSSits advantage
or fractures which are not moderately to well contained and
nore generally, they do not easily allow the representation of
:omplex geometries and treatment schedules.
To provide a fracture height growth criterion valid for all
evels of confinement and a structure favorable to the eventual
~corporation of general and complex reservoir characteristics,
lore comprehensive 3D models have been developed, using
VOdistinct approaches (Refs. 10, 11). Although the former
lef. 10) haa in many ways become a practical tool for analysis
nd design (Ref. 12), providing comprehensive capabilities, the
lrrent generation of fully 3D models lacks the kind of computational efficiency and ease of operation required for many
?al-time applications, especially those which involve repeated
Kecutions of the model, at rates faster than real-time (e, g.,
m the determination of unknown reservoir characteristics by
istory-matching, or for the prediction of final fracture configrat ions corresponding to alternate treatment schedules).

Uphole and bottomhole pressures are related by dividing


the slurry in the wellbore into batches or sections along the
length of the wellbore (Fig. 1) and so!ving the flow equations
for each section. The flow equation for the jtb section is
1 dp
4+-=9T;
v dz2
p dz2

of Model

2fv2

p
=exp(~)~l+~

(1)

Pr

where p is the pressure along the wellbore, and f the friction


factor characterizing the viscous drag (Ref. 13). The kinetic
energy or convective term udu/dz2 haa not been neglected
and, in order to model compressible flows ,whlch occur during foam treatments, a closure relation is assumed in which
the slurry density p is written in terms of pressure, the slurry
bulk modulus B, and the slurry density p, determined for some
reference pressure (usually either atmospheric or bottomhole).
This expression for p is linearized as shown because p is sufficiently less than 13;the resulting linearized form of Eqn, 1 and

The foregoing 3D, P3DH and 2D efforts have substanally enhanced abilities to model many aspects of hydrafrac;
owever, for different reasons, none has produced an adequate
~odel for realtime analysis of fracture operations. Thus, the
urpose of the model described in this paper is not only to be
sed for comprehensive hydrafrsc design, but also for real-time
nalysis of the treatment. To achi~ve this, the models had to
atisfy the three major conditions ~iet!orih at the outset: they
re realistic and general in the I,hysics and geometry which
hey represent; they run effectively on all levels of computer
ystem, from mainframe to portable; and they operate on the
ctual data generated by the job itself, running in real-time as
ceded and providing an instantaneous view of the job status to
he operator including pre-frac design and post-frac analyis. By offering all of these essential capabilities, the rnodelling
f hydrafrac finally takes on a greater and more meaningful
ractical significance.

Overall Description

SPE 15069

OF HYDRAULIC FRACTURING

MODEL

its solution is
dp
~-C1p=CO;

Pj=

co =

(Pr9

Cl = (Prg +

Pj-1(1

cl~B)

cO~B,

(2b]

uj-1)2/Pr~)/c2,

zf(Pj-1

zf(Pj-lvj-1)2/Pr~)

(Z4

/~c2t

(2C]

and
(2d:

C2 = 1 (~j-lvj-1)2/prB

where pj - 1 and pj are the pressures at the top and bottom o


and b?B is the batch length, SinC
the J%h batch respectively,
the fluid is assumed to be compressible, the batch length is un
known and must be determined from an integral expression fo]
the mass of the batch MB, given values for the j -1 variables

Capabilities

The overall modelling capability consists of four fully integrated modules, each representing a major aspect of the hydraulic fracturing process. In order to be run during the fracture treatment at rates faster than real-.me and :.ing aCtual sensor data as input, these modules weie .. . kematical ly
formulated to accept a wide range of input values, and were
efficiently implemented on the computer to minimize execution time. None of thw was accomplished, however, at t he
expense of accuracy or completeness; on the contrary, we ha ve
taken considerz- ; care to incorporate as much of the easential physics as possible dkectly in the modules an indir ect
method of data-baaing results provides a means to account for
whatever remains. The physics which have been incorporateed
into each of these modules is described in the following four
sect ions:

en
MB =

/0

pAwdzz

(38

yielding
(Pj-lcl

co)t~

+ (~+ pj_,/B)tB

2B

999
.-.

=o.45-

2,75
n

(3b

deification of Clapp

The frictim factor is ba


equation (Ref. 13)
f;lfa

B
-=0.
prAw

4.-.
n

q(ReCfC

(n-.2)/2)

(4a

in which jc ia the Clapps friction factor and Rec is the Re}


nolds number for power law fluids, The friction factor f i
related b fc by

RPF.
a
-

R flRf)f!KRTT-

1*V
MIRO

.A. . . . .

-A

-----

-.

N.M. (3K1JS1J (!z M.P. CT,TCAR.Y


-----

-------

..-

.-.

-----

E =

acbclOg
Re

R,,

= ;;:_:m

(4b)

vhere ac and be are coefficients found to be necessary in order


o match laboratory and field pressure data for pipe flow, The
:ffect of proppant on viscosity is described by

n which p: is the apparent viscosity for gel and s is the volunetric sand concentration. The effect of foam quality on visosity is accounted for in a similar fashion; however, the precise
orm of this effect has not yet been finalised,
!. Fracture creation. extension. shutin and closure:
rhis component is a completely general integrated 3D model of
Lydraulic fracturing, incorporating explicitly to first crder all
~fthe essential physics, including the rock mechanics of deformation and fracture, the coupled fluid flow within the fracture,
md the fluid and heat exchange with the surrounding reserfoir. Any required degree of accuracy may be obtained by
lmbedding, in functions or data-bases, the results of comprehensive (e, g,, fully 3D or cross%ectional) fracture simulations
nto the integration parameters or gamma (-Y)factors of the
nodel, thus taking into account the effects of complex reser{oircharacteristics (i. c,, multiple strata, inhomogeneities, etc.)
m the spatial variation of fracture attributes (i. e., crack openng, excess pressure, fracture height, etc.). Default values for
~amma factors have also been determined, which are either
:onstants when variation in the factors does not greatly afkct fracture propagation, or functions of relevant parameters
when the parameter variation dominantly affects the fracture
configuration (i. e., confinement, frictional drag, etc.). Having
relegated the complex numerical analysis to the determination
~f functions or data-bases for the integration parameters, it
is possible to perform accurate, faster-than-real-time
analyses during the fracture treatment, as well as routine pre- and
post-frac analyses on PC-based systems.

In all practical hydrafrac operations, the pumping-rate


is specified and the pressure is a consequence of the specified
flow-rate. Hence, the most important equation determining
pressure and crack-opening is the theological law for flow of
the fluid; although this can be quite complex, for instance involving viscoelastic memory effects for cross-linked polymers
(e.g., Ref. 5), we will employ here the more conventional powerIaw model,
tii m
=
()PF

A 2n+1

~pF

~Xi

i=l,2

(7a)

in which @is the effective channel-flow viscosity


P = 2K(4

+2/n)*

r = K~n

(7b)

phrased in terms of the consistency index K and the flow


behavior index n which appear in the relationship between
shear-stress r and shear strain-rate ~. Both K and n can
be measured with a conventional viscometer. In previous efforts (Ref. 5, 15), the lateral/vertical mass flow rates per unit
height/length, W1and W2,respectively, have been written as
Wl = W/2 L2

(8a)

where the mass flow rate into the fracture wing is divided by
the fracture height, and
(8b)
where W2equals the time rate of change of the vertical crosssection of the fracture plus the mass loss rate per unit length.
7x and 72s are the vertical crocxwwction and fluid loss shape
f~tors respectively. Specifying WI as the mass flow rate input
W, however, reduces the order of the governing set of equations
by one; this causes discontinuous jumps in estimates oi fracture
height and width when the flowrate W varies with time. We
use another closure condition similar to Eqn. (8b) for both
lateral and vertical directions

One of the primary equations is overall mass conservation,


W 2WL = F V. A (2L2) LI

7i4

(5)

in which W and 2W& are injected and lost fluid masses , pf


the fracture slurry density, while Ll, 2L2 and A are the length,
total height and width of one fracture wing. ~Vis the numerical
factor, which describes the volumetric shape of the fracture.

Wi =

PF

~is A hi +

7i6 wL/Lkv

i=l,2;

k=2,1

(8c)

which yields the same solution to withh a constant as Eqns. (8a


b) for steady pumping, while allowing smooth transitions in
height and width through surges.

The width A determined by the pressure in the fracture


and the results of numerical modelling (e. g,, as described extensively in Ref. 10) may be phrased as an integrated crackopening relation

The factor 7i4 is intended to account for the differenc~


in the flow from that between parallel plates; in addition
Eqn, (7a) must be embedded in a spatially-dhtributed
mode
to numerically determine the pressure gra&ent (which is a WC.
tor -F
in the full 3D simulations, (e.g., Re!%. 10, 12) thai
support the present model). However, the results of these sim
ulationa can be recorded as coefflciente qi2 and simply used u
the form
~pf
=
(9:
-7i2u/Li,
i = 1, 2
aZi

in which pF is the fracture pressure, ac is the closure stress


and E is the crack-opening modulus, e.g., E/4(1 - U2) for an
isotropic homogeneous rock structure. Of course, the actual
opening depends in a complex way on the geometry (not just
the shortest length scale 4) and on the rock structural variation:
these characteristics are embedded in the coefficient VI.

where the pressure gradient coefficient 7i2 reflects all of th~


complexities associated with stratification, tluid rheology, Wlc
tional drag and earth streaa gradenta.
---I

-----

..- .-.

Equations 5-9 can be combined to create two governing


rst order differential equations in the length and height Li,
s follows:
(lea)
d (L?) + BiL~ = Ai
E
here
lV = (3n + 6)/n,
(lOb)
L
~i2~i~E
WZWL
Li
+2Li

A _ ~ .
i
(Ioc)
[[71 -Y&D
2pF7u ~
~

nd
Bi = N_

~w2~w~)

i = 1, 2; k = 2, 1. (lOd)

Once the length and height are determined from Eqns. 10,
le other parametem, such as the width and excess pressure,
}Iloweasily from bad substitution into Eqns, 5-9. In actualy, there are three differential equations governing length and
oth upper and lower height extensions as shown in Fig. 1; our
lodel possesses this general feature, but, for simplicity, only
~e symmetric fracture formulation is presented.
C1early, the fluid 10SS2WL plays a dominant role in the
dutiona just derived and fluid 10SSwill be discussed in the
ext section. However, it is interesting to derive some analyt:al results which can be obtained in the special case where a
Vewtonian fluid is assumed (n = 1) and fluid 10SSis neglected.
rhe solution for a circular fracture (no confinement) is found
,0 be
=

A=

(ha)

[:-:[+13],

[[i%w2[a]

These solutions repraent the two extremes bounding the


aspect ratios of fractmes created during moat treatments; intermediate levels of confinement occur smoothly as 7M -+ O.
Note that the excess preaeure u of the uncontained and complete!y contained fractures diffem substantially
u decreases
by t-lis and is ~wiant
to flowrate for the former; whereas
it incre~~ by tlis and depends on flowrate for the latter.
Interpretation of this pressure renpome has become a popular method of determining the extent of fracture confinement
(e.g., Ref. 14). Cleary has previously derived a number of special analytical solutions (Ref. 15) using the closure conditions
expressed in Eqne. (8a, b) and has compared hia solutions with
those of earlier models (e.g., Refk. 1-4) whkh, incidently, are
special cases of the present model. Clearys solutions correspond very closely to Eqna. (11, 12) even though he altem
Eqn. 8a to account for radial flow patterns in a circular fracture.
What has been described so far ieh spatially integrated
statement of the governing equatione to be solved, but equally
important is the issue oi values for the gamma factors. Previous approaches (e. g,, Refs, 6, 7) which developed forma for the
722, were baaed on croesvertical pressure gradient coefficient,
sectional models of the khid described in Refi. 6, 16; them
forms have now been developed into the following structure
w~l~ include alterations not only to 722, but also to 712, ~1,
and Vis:
(13a)
71 = S, &S9qf,
~i2

*d

E 4 T1T13 P lt_~
71 [9 qlzqlq ~ 1

(llC

uhereaa the solution for a constant height fracture (complete


confinement) becomes

1=

[[:-lwi]i~

[:-;(&)2+

(12a)

qis = Li7~8*

%)9

12

1* 29

(13b)

(13C)

The shape factom occuring in Eqns. 13 are determined


from detailed (3D and croes+ectional) numerical simulations
(e.g., Refa. 12, 17) and comparison with laboratory and field
data (e.g., Ret%. 12, 18). The information derived from these
studka can be either dwectly incorporated through a numerical data-base, or more Preferably, condemsad into ftmctimal
repreaentations of the S-fwtore whkh depend on the essential
parameters of the physics involved. Cleary origiidly devel.
oped this latter approach, whkh was implemented by Settari
and Ckary (Ref. 6), and further studkd in the context oi
P3DH and croes-eectiond models by Settari (Ref. 16) and
Narendran (Ref. 17). The determination of comprehenaiw
and accurate functional forma for the shape factors ia at th
heart of 0U2 model development, and many detdb have not

(12b)

md

u=%~*:($2#

S.i Sdi Spi Spi

The shape factors (S) appearing in Eqns. 13 denote the


following influences on fracture evolution: 1) confining stress
variation S,; 2) modulus stratification Sd ; 3) spatial variation
of frac fluid viscosity SP ; 4) frictional drag induced by proppant SP ; S) stratified fluid 10SSSL ; d
6) certain
geometric
effects (e.g., crack-opening dependence on geometry) S~. 7:,
7?w ad V$ denote b~e VZLIUSS
of the g~a
f~tors when
none of these effects is significant (in which case all of the Sfactore equal unity). If any additional physical influences need
to be incorporated, representative shape factom can simply be
added to the strings of factom in Eqne. 13.

(llb)

md
0=-=

(12c)

---

-.

----

~een finalised; however, to consolidate ideas, several illustraive examples based on cross+ectional analyzes are described
~elow:
1) If the confining stress is greater in the adjacent strata
,han in the reservoir by an amount Auc, vertical height growth
s impeded. Such a step stress contrast affects the pressure
yadient 7i2 and the crack opening 71 coefficients when Lz >
H,; the corresponding S-factors take on the following forms:
S~l = ~ [~- AoC[l -e~p(~.i(l

S82= : u[[

AuC 1-

- ~))]],

(14a)

sin-(~,z~)
sin-l (0,2)

(14b)

1]

and

s, = S*2+

2 Au=
;~~,$b!n

l-t

(1-(LLE)2)*
~)2

Spi =

p{ew]
p(eR)
+

viscosity variation

kwlexp(-~ia

f)

in which both the characteristic distance from the wellbore Lej


(e.g., as in Ref. 20) over which the temperature varies from
wellbore e~ to reservoir 9R temperature, and the exponential
coefficient ~Pi, determine the scale of the thermal dependence.
It is also true of viscosity variation that it can only impede,
not stop fracture growth; however, the breakdown of the fracfluid viscosity, from heat-up and other causes, aids confinement
when wting in conduction with stress contrasts because a leSS
viscous fluid produces a smaller excess pressure which in turn
decreazes the equilibrium height,
4) The geometric shape factor Sg, represented by

+
.

[ 1- (1-(/3,

3) The effect of temperature-dependent


along the frmture b expr~sed ss

(14C)

$1

S. and S.i have been constructed so that the excess pressure


u in the governing equations (Eqns. 5-9) is replaced with the
proper values for confined fracture growth: note that in the
Iirzt limiting case where the fracture height just reaches the
confining barrier (L2 = l..,) the three S-factors are all unity
(homogeneous reservoir conditions); for the second case where
the fracture has grown far out of zone (L2 >> H,) all of the
factors approximately equal (0 - Au,)/u, In the latter case,
the characteristics of the adjacent strata predominate and the
fracture again grows ae if were in a homogeneous medium,
except the conlining stress UChaa been replaced with UC+AuC.
The coefficients ~,~ and /3, are determined from more detailed analyses; however, several constraints on them can be
identified. First, @,l and Pa2 must be chosen so that S82 < Sal,
in order that the height growth maybe condned. Second, when
Au. >0 there exists an equilibrium height, beyond which
no further growth occurs at the particular value of G for thl s
case, and especially for highly elongated well contained fractures, @,l = B, = 1.

% = 1 + [~ -1]
L

exp(-~c(%

L1/Lz, Lz < LI
= { L2/Ll, L1 < L2

1)),
(lg)

changes the crack opening coefficient between limiting values


appropriate for a well contained VS. a circular fracture, depending on the aspect ratio of the fracture.
9. Re servoir Simulation:
This component represents,
as accurately as needed for practical purposes, the fluid flow
and heat transfer withb the rock in the strata surrounding
the fracture. It thus allows calculation of fluid Ioas and f&ture efficiency; heat-up of the the-fluid and, consequently,
ita changing thermorheolog~ poro-induced backstrezsea; and
eventually, production of oil and gas through the propped fracture.

The main featurea of the fiuid-leas model are shown in


Fig. 2 in which the fracture surroundings are characterized
by four major zones indicated previously (e.g., Refs. 19, 20):
a filter-cake zone 6C, a leaked-off fluid zone 6L, a thermally
cooled zone 6h if the thermal front exceeds the leak-off front,
and a distributed reservoir zone. The penetration depths out
2) Modulus stratification also affects both qi2 and -yl:
into the reservoir of the thermal and the leak-off fluid fronts
are small relative to the scale of the fhc-pressure diffusion so
(1d ) that the incompressible flow through these zones a~acent to
d = :
+ [1 -:1
xd-@d *)
the fracture can be described by straightforward expressions oi
and
DArtys law:
Sdi = 1 + [S~ -1] exp(-~di ~).
(14e)
~L_2~pF
pc=2kL
p&pL=2kRp&-pJ
(15a)

~~
PR 6h 6L
PC
&
Since there is no equilibrium height, the modulus contras t relating the fracture pressure pp to the interface preesure PI
can only impede, not stop fkacture growth; consequently, in th e The viscosity of the fluid occupying each of the zones is mitabll
limit of a very large fracture the adjacent strata characterizti Cs averaged to take into account its thermal sensitivity
again predominate arid the fracture grows as if it seen only
1
(lsb;
~a (0) dzs
the modulus of the adjacent strata in thw knit both S~l an d
*=C6
/
sdz must have returned to a value of unity and $j ta E/EA
in
where a may stand for the filtercake, the leak~ff, or the ther
order to have replaced E with ~A everywhere in the govemin g
mal zone (C, L, and h subscript respectively).
equations. The exponential coefiiciente fid and ~di control th e
rate of thu replacement and are determined from the rnore
detailed &nalyaes.

Zza

A CX3MPLETE
REAL-TIME
----.-- -
.

Since the scale of thermal penetration into the reservoir


relative to the fracture length is also small, heat transfer in the
reservoir is assumed to be ID and normal to the fracture surface. Further assuming that the reservoir possesses thermally
homogeneous properties, a Greens function solution can be
written for the governing thermal transport equation, analogous to that for fluid loss (Eqn. 15c)

Beyond the thermal and leak-off zones, DArty flow and


DmpressibiliQ must both be taken into account; the Greens
Influence) function solution to the resulting diffusion equaIon which governs pore pressure is used to relate the interface
ressure PI to the loss rate q~:
t
PI

PR

~ qL (r)

7~ (~, T) d?,

Yf(h ~) = *

[4n :_

~)

(15C)

eff-t3(Z3,

6L=A ~d

t
~ qL

(~)

7h(~3,

The

T)] ~h (z3,

dX d?,

Z; t, T)

z; ~, ~)

(a)

= s& (tN] ~= (~n).

(lb)

The average fracture temperature is calculated simply from


an expression of energy balance for the fracture
$

(P C) FVF

+ AFhL

(eF

eR)]

- (@) UILAF(eF

(Pc)wd(%

eR)

(17C)

- eR)

where each term respectively takes into account: 1) thermal


storage in the fracture; 2) convection flow from the wellbore,
3) conduction from the reservoir rock; and 4) convection into
the reservoir.
The loss of fluid from the fracture into the surrounding
reservoir causes a swelling or strain in the rock which in
turn increases the confining stress on the fracture. A ?ormal
expression for this back-stressn or poro-induced stress UB is

If the pressure difference pp pff is assumed constant and


the thermal effects are ignored (lf = O), the resulting special
solution to Eqns. (15d, e) for the fluid loss rate qL has square
root time behavior,
(16a )
=
KL/&
qL

(18a)

OB =Ug+u:+ug

in which the three regions of Figure 2 contribute to the overall


stress (Ref. 20). Each of these stresses takes a different form,
represented by the following expressions:

where KL is the overall fluid loss coefficient. This !OSScoefficient is related to filtercake, kink-off, and reservoir zone properties, in addition to the preaure dMerence PF - pff, by the
following quadratic expression in KL:

(18b)

Ug = 2J#CCi.?dL/%t,
= 2fi~(1 - #C,)~(pF

0;
u:

+pR@iZ
~R

O; t, r)-

~L (k)

md using this expression in the loss rate integral of Eqn, 15d.


Previous comparisons with fully distributed calculations of loss
(Ref. 20) have shown this technique to be sufficiently accurate
Formost practical applications.

$[&(-@c)+

eff e (Z,

(15e)

(lsf)

-qL(~J

~ hL (~) ~~ (2s,

except there is a second integral representing the effect of thermal convection which is absent in the case of fiuid-diffusion.
The effect of fracture area creation is taken into account just
as in fluid loss:

Yote that since the depth of the frac-pressure diffusion into the
:eservoir is typically small relative to the fracture length during
;he treatment time, it is sufficient to aesume ID loss normal to
;he fracture, provided that the creation of new fracture area is
Lakeninto account; this is accomplished by dividing the volume
!bssrate which occured at time level tm by the current fracture
mea Ap(tN):
qL (tn)

A
~k

[-rexp[~~~:l

W)

d~.

I%[r&[

vhere If=lfor
6~-6rj>Oand
H=0 for6h-6LS0.
eak-off penetration can be found from the expression

t)

Combining Eqns. (15a, c) produces the overall governing


quation relating fluid loss to fracture pressure:
t
fiR ~h
- 6L)H
PF
PR r)dr = F(
~ 9~(~)7j(t,
/
[

+ [@ce)+&cc],L

SPE 15069

MODEL OF HYDRAULIC FRACTURING

bn

&(PI -

PR)[l

ep(-t-2)

-pi)/?rt(l
-t

- V),
$erfc

(~)]

(MC)
(18d)

@c]K~

KL(pF-pR)=O.

(16b )

The easy solution of this quadratic is provided and discusse d


further in Ref. 18.

---

where ~ = 2 @/t
is the dimeaionless diffwion time; 4 is the
lesser of the fracture height or length; q = (1 - ~/&)(l
2u)/2(1 - v) is the poroelaetic induced stress factor; and b =
0.56 is a numerical factor (Ref. 20). Thermally-induced strasws
do not greatly affect hydraulic fracture growth, except in steam
and water fioodhg, because the time scale of the thermal
convective-&ffuaion in reservoir rock is much larger thau the
fracture growth time.

4. Prormant transmort and rdacement: A rigorous


malysis of proppant transport would require the knowledge
}f detailed flow patterns in the fracture and the spatial distribution of leak-off, as in Ref. 12; however, such a scheme
s at present too computationally demanding for many real,ime applications. Therefore, in keeping with the approach of
his model, the essential characteristics of proppant transport
:e.g., mass conservation, concentration distribution, differential settling, bank formation, screen-out, and closure) are in:luded explicitly, and gamma coefficients provide a means to
Iata-bsae the more accurate modelling results.
In order to calculate the penetration into the fracture of
:ach proppant concentration band (Fig. 3) corresponding to
;he treatment staging, the slurry volumes in the fracture VP~fter leak-off must be calculated:

vP.

t
=

Q-

/[

(19)

Qbn]~r

tpm

where tpm is the time when the rnth proppant stage entered
the fracture, and Q~~ is the volume leak~ff rate through the
hcture area adjacent to VP-. Once VP- is obtained, the extension into the fracture of each concentration band Lpm is
Foundfrom mess conservation yielding

v
(723 -

%)

(%)3

-723

(%)2

(Zoa)

where VF is the fracture volume. The height extension of each


band lfPm is then related to LP~ by a ratio of fracture height
to length
LZ
(20b)
Hp. = 7P LP. ~
h which :iw default value of 7P is 1. The proppant volume ss
a whole ISthen settled downward in accordance with relations
for the unhindered and hindered settling velocities (e.g., as in
Ref. 21):
V8 =

(Pp-Pf)~

(h+l)d

108 n

l(J1.s2(l-f.)

72 K

(21a)

and
Vef:
h

(21b)

where V*and v~b are respectively the unhindered and hindered


settling velocities, and fv is the fractional volume of suspended
proppant.

Sample

Implementation

of the Models

To illustrate the capabilities of the overall integrated motlel, we analyse a fracture treatment performed in the Travis
Peak formation of Eeet Texas on which the models were implemented in real-time, using the Gas Research Institute (GRI)
fracture treatment monitoring and analysis facility developed
by Resources Engineering Systems (RES) and described in
Ref. 22. The well was one of many monitored and analysed
under the auspices of the GRI Tight Gas Sand Program, in conjunction with a number of ~,ther GRI contractors (e.g., Ref. 23).

The flow, rheology and proppant data obtained, and used


or real-time input to the models, are shown in Figure 4(a).
rhe resulting model calculations of uphole pressure are shown
n Figures 4(b, c), where comparisons are made with the actual
neasured pressures.
Unknown reservoir end friction parametem were obtained
)y history-matching, a technique that involves comparing the
)bserved pressures and model predictions in order to minimize
;he error between them. When history-matchhg is performed
m the early response to pad injection, the parameter values
>btained may then be used during the rest of the job to enmre the beat model predictions possible. Use of our wellbore
transport model may be seen to give an excellent fit to uphole
pressure in Figl re 4(b). Although the form and coefficients of
the pipe friction component of the model have been chosen to
best fit an average of all industry lab and field data available
to us (e.g., as in Ref. 24), fine tuning of these coefficients by
history-matching has improved the accuracy within 5 % over
the entire duration of the fracture treatment.
Even more intereating is the compsmson of excess pres- crc) in Figure 4(c), remembering that the extent
of fracture containment may be estimated from the pressure
rise or fall-off (e.g., as in Eqne. (llc, 12c)) es long se other
influences on fracture pressure can be identified and separated
out. After the initiation phase, which is still under study with
the distributed 3D model (Ref. 12), we see that the reservoir
parametem associated with no containment produce an excellent fit during the early part of the job, including the period of
the great rise in pressure which is due to the first appearance
of 50 lb. cross-linked gel in the fracture. After thw early phase
of no containment, the fracture pressure continues rtilng the
best fit with model predictions during thw second phase seems
to indicate that the fracture has now encountered subatantiaf
confining barriers; however, thb pressure rise could also have
been caused by a greater flow resistance, due to proppant staging, than we had incorporated into the proppant shape factom
SPi of our model. Late in the job, when the heaviest proppant
loading was reached, the pressure response turned sharply upwards. A screenaut was clearly developing, as no amount of
containment in the models could have produced by itself such
an increase in pressure,
LiUreS (PF

Although we have been able to identify mechanisms such


ss sliding resistance, bank buildup, and screen-out as potentially dominant contributors to fracture (excess) pressure
accounting for much of the pressure rise in Fig. 4(c) descriptions of these mechanisms have not yet been finalised in
our models. Until enough reliable field data haa been matched,
the confinement inferred from the pressure response will remain
uncertain for many treatments. A Nolte-Smith type of analysis
(e.g., Refs. 14, 25) would also require detailed incorporation of
such modelling and mechanistic interpretatiom, if proper explanations are to be made: although such analyses are now
being conducted as a major focus of treatment monitoring vehicles in the field (e.g., R&. 24, 26), it has become clear that
much more detailed modelling will be required in order to dwtinguish between the effect on observed pressure of fracture
containment and the other equally important influences.

A COMPLETE

REAL-TIME

MODEL OF HYDRAULIC FRACTURING

Given that this is the case, the range of possible fracture


,mmetries corresponding to different levels of confinement, as
hewn in Figure 4(d), is often as close as one can come to
scertaining fracture geometry. Designing or analyzing fracure treatments based on this quality of information consists of
bounding the fracture within the limits of acceptable best and
rorst case estimates. If the fracture pressure can be historynatched with certainty, having eliminated ambiguous or un,nown influences on the pressure, it is then possible to narrow
he interval of probable fracture geometries to a single best
stimate of final fracture dimensions. The dominant reservoir
parameters determined by such historymatching on fracture
jressure are the confining stress contrmt, the fluid leak-off cofficients and the reservoir modulus. The extent of proppant
ransport is indicated in Figure 4(e), corresponding to the limts of no containment and high containment. With these esimates of the final propped length and width of the fracture,
alculations of oil or gas production can be made (e.g., Refs. 20,
;3); eventually, our program may be employed to make these
)redict ions at any stage during the job, allowing decisions to
hange schedules or stop pumping based on real-time predicions of well productivity.

Conclusions

Nomenclature
AF
A

SPE 1506$

area of fracture wing


area of wellbore crossaection

bulk modulus of wellbore slurry

cc

volumetic fraction of filtercake polymer in fracfluid

c~

pore-fluid diffusivity of the reservoir

CW, CF, CL

specific heats of the wellbore slurry, fracture


slurry, leak-off fluid
hydraulic diameter of the wellbore

Youngs modulus for the reservoir


crackapening modulus for the reservoir,
E/4(1 U2)
error fimction
complementary error function

E
E
erf
erfc

crack-opening moduli for upper, lower and adjacent strata (when symmetric)
friction factor for wellbore flow

f.

Clapps friction factor for power law fluids

9
&

gravitational constant
half height to modulus barrier

hL
Ha

conduction heat exchange rate

&J,

EL,

EA

A comprehensive integrated model of hydraulic fracturin g;


and wellbore transport has been developed and incorporated
into a system for real-time monitoring and analysis of fracture K
treatment operations. The model is baaed on a full 3D simula - K, Ka
tion of hydrafrac creation, but it has been formulated so as to
allow rapid execution, using data-based coefficients previousl y
determined by other more elaborate numerical analyses, an d &C, &L, k~
by comparison with field observations. This allows both eazy
identification of major features and isolation of possible error s
Kk
during design and evaluation, but especially it permits mult1ple executions of the models, faster than real-time, thus mak - L
ing possible the on-ite determination of reservoir and frictio n L1
parameters, by history-matching on observed pressures as th e L2, L2U, L2L
job proceeds. This real-time feature of using flow and rheolog y
field data as input, to calculate actual pressures and determin eLH
Pm !
Pm
parameters from the job response, itself sets the model apar t.
Of course, the model can also be used for pre-frac design an d
post-frac analysis, running on personal computers, minicom - Le,, Le,
puters and mainframes,
1
Since all of the essential physics of the hydrafrac proce Ss
appear explicitly in the model, routine improvements in its accuracy are simply made by updating the gamma coefficients;
thus future advances in modelling and experience gained from .t!B
field applications can be continually incorporated without any MB
restructuring of the model. Current and future efforts are fO-n
cused on improving and verifying default forms of the gamm a
coefficients by comparison with fully 3D fracture simulation s, P
accurately incorporating all of the physics (e, g., wall frictio n, PFv pR
screenaut) affecting the fracture pressure response, and opti- PC, PLY PI
mizing procedures for determination of reservoir and fricti on
parameters by real-time history-matching.
Q
.z--

half height to stress barrier


spatial coordinate index
consistency index, viscosity for linear fluid
bulk moduli of the overall porous matrix and
solid constituent, respectively, under drained
conditions
effective permeability of the filtercake, leak-off
and reservoir zonea
conductivity of reservoir rock-fluid matrix
overall fluid loss coefficient
length of one fracture wing
half-height (when symmetric), upper, lower
heights of the fracture
wing length and half height of rra:h proppant
stage penetration in the fracture
lateral, vertical thermal penetration scales im
the fracture
the lesser of the fracture wing length or halfheight; denotes generic length dominating cracl
opening
length of the j:h slurry section in the wellbor~
mass of the jth slurry section in the wellbore
flow behavior index
pressure
fracture preesure, reservoir pore pressure
pore pressures at the fihercake, Ieak*ff, ther
mal interfaces
slurry volume flowrate into one wing of th~
fracture

>W

A .R
. .

I KnRo

LJ

. .

J.uvvcz

fiRfW!Kl?TT.
----

-----

N.M. OKUSU & M.P. CLEARY

_..

_..

volume loss rate through fracture area adjacent


to mth proppant stage
volume loss rate per unit area of one fracture
face

!Lm
L

effective channel flow viscosity

P
P.
PC,

pL>

#R

apparent viscosity of the slurry


viscosities of the filtercake, 1eak4f, and reservoir fluids

radius of a circular fracture


Reynolds number for a power law fluid

Poissons ratio

tc=
d, Sdi

shape factors for modulus contr~t

p, pF

porosity
wellbore, fracture slurry densities

shape factor for fracture geometry

Pp,

inl

inverse sine function

Pr

*.
La

shape factors for fluid loss

Pw,

~,
Jpa

shape factors for proppant

shape factors for stress contrast

@w, @F, @R wellbore, fracture, and reservoir temperatures


excess pressure
u

,)#j
,

shape factors for thermal heatup

pi

time
time of entry of the mth proppant stage into
the fracture

Pm

proppant, frac-fluid densities


slurry density at a reference pressure

Pf

wellbore slurry, leak-off fluid densities

PL

temperature

uB

induced stress due to thermal and poro-elastic


effects

u:, Ug, Ug

poro-induced stress for the filtercake, leak-off


and reservoir zones
confining (or closure) stress in reservoir
stress contrasts for upper, lower strata

{F

fracture volume

/Pm
/

fracture volume up to the m:h proppant stage


slurry velocity down the wellbore

u.
AOCU,AOCL

l,, vah

unhindered and hindered settling velocities

Au.

stress contrast for adjacent strata (when symmetric)

shear stress

*,

ifiL , 2WL
01, tiz
cl, Z2, z~

mass flow rate and total pumped mass into one


fracture wing
mass loss rate and total mass lost out of one
fracture wiruz
lateral/vertical
mass flow rates per unit
height/length in the fracture
spatial coordlnatea in the lateral, vertical, and
normal directions with respect to fracture

dc

S-factor coefficient for fracture geometry

~d, ~di

S-factor coefficients for modulus contrast


S-factor coefficients for stress contrast

flev@.i
0Hi

Acknowledgements
The work presented in this paper has been performed as part
of the GRI Tight Gas Sands Project, a comprehensive effort to
imp we many aspects of hydraulic fracturing operations. We
are grateful to H, D, Vo and R. M. Willis for their assistance in
implementing the model on the computer and to Marie Dokoupil for helping to prepare this manuscript.

S-factor coefficients for thermal penetration

Figures

engineering shear strain-rate

Yf

Greens function for pore-fluid diffusion

Yh

Greens function for thermal conduction

7P

gamma factor for proppant

2. Zones of leak-off and thermal penetration in the reservoil


shown as ID normal to the fracture surface.

7.

gamma factor for fracture volume


gamma factor for crack-opening

3. Bands of proppant in the fracture corresponding to th


fracture treatment staging.

71
712, 722
?13,

723

714,

724

%S, 726

A
6C, lfL, dh

1. Hydraulic fracture and wellbore with reservoir surround


ings, showing principal variables used in modelling.

gamma factors for lateral, vertical stress gra- 4(a). Flow, rheology, and proppant data; time in minutes.
dient
gamma factors for lateral, vertical cross~ectio n 4(b). Uphole pressure data showing comparison of model w
actual pressurea; time in minutes.
area
gamma factors for lateral, vertical channel flow 4(c).
gamma factom for fluid 10ss
width of fracture at wellbore
4(d).
filtercake, leak-off and thermal penetration
deptha

poro-elastic induced stress factor


q = (1 - *)(1 -2u)/2(1 -v)

thermal diffusivity of the reservoir

~L

thermal diffusivity of the leak-off fluid

Downhole (excess) pressure data showing comparison f


model vo.. actual pressur=, time in minutes.
Model estimates for final hydraulic fracture length, heigh
and width showing dependence on confining stress co]
trast, the principal containment parameter.

4(e). Model eotimates of proppant extent, variation with tin


and confinement; time in minutes.

Zzv

REAL-TIME
HYDRAULIC
.-- OF
-
.A. COMP1.F!TR
-----. -.
-. -.-..- MODEL

References
. Geertsma, J. and F. de Klerk, A Rapid Method of Predicting Width and Extent of Hydraulically-Induced Fracturesn,
J. Rt. Tech., Dec. 1969, pp. 1571-1581.

FRACTURING

SPE 15069

14, Nolte, K. G. and M. B. Smith, Interpretation of Fracturing


Pressures, J, of J%t. Tech., Sept. 1981, pp. 1767-1775.

15. Cleary, M. P., Theoretical and Laboratory Simulation of


Underground Fracturing Operations, Firat Annual Report
of the MIT UFRAC Project, MIT Resource Extraction Laboratory Report, Aug. 1981,
. Perkins, T. K. and L. R. Kern, Widths of Hydraulic Fractureen, J. Rt. Tech., Sept. 1961, pp. 937-949.
16. Settari, A., Quantitative Analysis of Factors Influencing
Vertical and Lateral Fracture Growth, SPE/DOE Paper
. Nordgren, R. P., Propagation of a Vertical Hydraulic FracNo. 13862, presented at the SPE/DOE Low Permeability
ture, J. Sot. M. Eng,, Aug. 1982, pp. 306-314.
Gaa Reservoir Symposium, Denver, May 19&i,
. Cleary, M. P., Comprehensive Design Formulae for Hydraulic Fracturing, SPE Paper No. 9259, presented at the 17 Narendr~t M V*Analysis of the Growth and Interaction
of Multiple Plane Hydraulic Fractures, Ph.D. thesis, MIT,
55th SPE Annual Fall Technical Conference, Dallas, Sept.
Dept. of Mech. Eng., February 1986.
1980 (see also SPE Paper No. 9260, Analysis of Mecha. Danezhy, A. A., On the Design of Vertical Hydraulic Fractures, J. Rt. Tech., Jan. 1973, pp. 83-97.

nisrna and Procedures for Producing Favorable Shapes of 18. Crockett, A. R., Willis, R. M. and M. P. Cleary, lrnprovHydraulic Fractures ).
ing Hydraulic Fracture Predictions by Real-Time HietoryMatching on Observed Pressures, SPE Paper No. 15264,
1. Settari, A. and M. P, Cleary, Three-Dimensional Simulapresented at the SPE Unconventional G= Recovery Sym.
tion of Hydraulic Fracturing, J. of Rt. Tech., July 1984,
posium in Louisville, KY, May 1986.
pp. 1177-1190 (see also SPE Paper No. 10505, Development and Testing of a Pseudo-Three-Dimensional Model of 19. Settari, A., A General Model of Fluid Loss in Hydraulic
Hydraulic Fracture Geometry, presented at the SPE SymFracturingn, SPE/DOE Paper No. 11625, presented at
posium on Numerical Simulation, New Orleans, February
the SPE/DOE Joint Symposium on Low Permeabiliw Gas
1982).
Reservoirs, Denver, March 1983.
~. Meyer, B. R., Frac Model in 3D - Parts 1-4, Oil and Gas 20. Crockett, A. R., Vo, H., and M, P. Cleary, Studies of Fluid
Journal,,
June 17 (p, 87), July 1 (p. 62), July 22 (p. 83) and
Flow, Heat Transfer and Induced Stresses In and Around
JUIY 29 (p. 132), 1985.
Underground Fractures, Report No. REL-84-1, MIT Resource
Extraction Laboratory, March 1984.
J. Palmer, I. D., and C. T. Luis, A Model of the Hydraulic
Fracturing Processes for Elongated Vertical Fractures and
Comparison of Results with Other Models, SPE/DOE Paper No, 13864, presented at the SPE/DOE Low Permeability Gas Reservoir Symposium, Denver, May 1985.
). Advani, S. H., Finite Element Model Simulations Associated with Hydraulic Fracturing, SPE/DOE Paper No.
8941, presented at the SPE/DOE Unconventional Gae Recovery Symposium, Pittsburgh, May 1980,
1. Cleary, M. P., Kavvadaa, M., and K. Y. Lam, Development of a Fully Three-Dimensional Simulator for Analysis and Design of Hydraulic Fracturing, SPE Paper No.
11631, presented at the Symposium of Low Permeability
Gas Reservoirs, Denver, March 1983. (See also Report
No. REL-82-12, MIT Resource Extraction Laboratory $
Dec. 1982.)
1. Abou-Sayed, A. S., Sinha, K. P., and R. J. Cifton, Evalu ation of Hydraulic Fractures Using a 3-D Simulator: Part 1
Technical Approach, SPE/DOE/GRI Paper No, 12877,
and Abou-Sayed, A. S., Clifton, R. J., Dougherty, R. L., an d
R. H. Morales, Evaluation of Hydraulic Fractures Using a
3-D Simulator: Part 2 Caae Studies, SPE/DOE/GRI
Paper No. 12878, presented at the SPE/DOE/GRI Unconventional Gas Recovery Symposium, PM.sburgh, 1984.

21. Danezhy, A. A., Numerical Solution of Sand Transport in


Hydraulic Fracturing, J. ht. Tech., Jan, 1978, pp. 132140.
22. Cleary, M, P., Buharali, A. M., Crockett, A. R., and I. A
Salehi, Computerized Field System for Real-Time Moni
toring and Analysis of Hydraulic Fracturing Operations
SPE Paper No. 14087, presented at the 2nd International
SPE Meeting, Beijing, March 1986.
23. Holditch, S. A., Robinson, B. M., and W. S. Whitehead
Pre-Fracture and Pozt-Fracture Formation Evaluations
Necessary to Characterize the 3D Shape of Hydraulic ~
Fracture, SPE Paper No. 14086, presented at the 2n~
International SPE Meeting, Beijing, March 1986.

24. Hannah, R. R., Barrington, L. J., and L. C. Lance, Th


Real-Time Calculations of Accurate Bottomhole Fractur
ing Pressure fro: LSurface Measurements, Using Measure~
Pressured as a Base, SPE Paper No. 12062, presented a
58th SPE Annual Fall Technical Conference, San Franciscc
SerA. 1983.
25. Nolte, K. G., Determination of Proppant and Fluid Sched
ules from Fracturing Pressure Decline, SPE Paper Nc
13278, preeented at the 59th Annual Fall Teciudcal Con
ference, Houston, Sept. 1984.
2. Lam, K. Y., Barr, D. T. and M. P, Cleary, A Complete
26. Cooper, M. P., Nelson, S. G. and M. D. Schoppe-, Corn
Three-Dimensional Simulator for Analysis and Design of
parison of Methods for Determining In-Situ Leak-Off Rat
Hydraulic Fracturing, SPE Paper No. 15266, presented
Baaed on Analysis with an On-Site Computer, SPE Pape
at the SPE Unconventional Gas Recovery Sympozium in
No. 13223, presented at the 59th SPE Annual Fall Technici
Louisville, KY, May 1986.
Conference, Houston, Sept. 1984.
3. Govier, G. W. and K. Aziz, !l%enow of Cornplez Mixtures
in Hpss, N.Y. Van Nodrand RAnhold Co., 1972.

i5u69

WE

_..

.-

L,,,

. .

3
--Q

,.

-.~~

,-hi.

\::

x.?

.
L

+%L
..O

PL
+

j-&:

RESERVOIR

1
F ,:_~yR::KE

i. ...
~.1
~

SC ~

LLAKOFF
pPENETRATIoN
TAERMAL
PEN ETRATION---

4.4
Ah

a
PR

1. L,
.

,,, ~

x,

SpE
siu
I@e,

90.00

SLURRYFLOU (SPII)
m.ue

70.00

SAND CONC

2.08

(PPG )

d2.JA+&~
@
f--l

30

Fis.

TIME

@-Flow, dwo~,

(MINUIES)

md pmppant datsi tlnw in mlnutos.

5800

UPHOLE

PRESSURE

(PSI)

4500

4s00

3500

3080

2500

LENSOR

Zeea

1sea

1000

sea

15069

~ 0[ X [Ill)

UlOIfi !3W ,(13)

H19N31 .[13)

1H913H 3WlA3ti#d

tLL

1-1

nL

0
lx
L

(Ld)

(ISd)

3WISS3Ud SS32X3

U9N31

lwddoud

You might also like