Professional Documents
Culture Documents
JPT - 2163 Fracture Gradient Prediction and Its Application in Oilfield Operations - Ben A Eaton PDF
JPT - 2163 Fracture Gradient Prediction and Its Application in Oilfield Operations - Ben A Eaton PDF
Introduction
The subject of many discussions and technical papers the fracture extension pressure gradient in areas of
in the last 20 years has been the prediction of the well- incipient normal fauMng.* Overburden stress gradi-
bore pressure gradients that are required to induce ent, formation pore pressure gradient and Poissons
or extend fractures in subsurface formations. The sub- ratio of rocks were the independent variables that
ject merits this attention because of the frequently were shown to control fracture pressure gradient, the
recurring problems that arise from an inabtity to dependent variable.
predict fracture pressure gradients. In 1967, Matthews and Kelly pubIished another
Encountered in several common types of opera- fracture pressure gradient equation that is different
tions in the oil indust~ are problems associated with from that of Hubbert and Willis in that a variable
the prediction of formation fracture pressure gradi- matrix stress coefficient concept was utilized.8 Later
ents. When wells are being drilled in both new and tie same year, Costley wrote about a similar idea.5
old fields, lost circulation is often a very troublesome Goldsmith and Wilson used a least-squares curve-
and expensive problem. Complete loss of circulation fitting technique and field data from the Gulf Coast
has been disastrous in some cases. Many times, such aree to correlate fracture pressure gradient with for-
disasters could have been avoided if techniques for mation pore pressure gradient and formation depth.4
calculating fracture pressure gradient had been em- They noted that the fracture pressure gradient in-
ployed in the well plans, and if casing strings had been creased with increasing depth while the pore pressure
set, and mud weight plans had been followed accord- gradient remained constant+
ingly. In areas of abnormally pressured formations, In each of these cases, the problem for which a so-
the prediction of fracture gradients during the well- lution was sought was to determine the bottom-hole
planning stage is extremely important. In fact, it is as pressure gradient required to initiate or extend a frac-
important as the prediction of formation pressure ture. Results of the previous work show that fracture
gradients, which has received a great deal of attention pressure gradient is a function primarily of overbur-
in recent years. den stress gradient, pore pressure gradient, and the
There are several published methods used to deter- ratio of horizontal to vertical stress. There is argu-
mine fracture pressure gradients. However, none of ment for a fourth variable in that in many cases break-
these methods appears to be general enough to be down fracture pressure gradient is greater than the
used with much reliabdity in all areas. In 1957, Hub- fracture extension pressure gradient. However, if the
bert and Willis published a classical paper that in-. fracturing fluid is able to penetrate the formation
eluded the development of an equation used to predict through the pores or existing cracks, there is very little
s=(oo+w->-- s - 2)
which k known as the Hubbert and WNis equation.
Under these conditions, the fracture gradjent k cal-
culated to be a constant with increasing depth for all
normally pressured formations. This is known to be
untrue for many cases in the Gulf Coast. Actually,
Eq. 2 predicts values that are usually too !OWcom-
pared with values from field data. However, this does
not mean that Eq. 1 is invalid. It is more probable
that the assumptions used to obtain Eq. 2 are in error. -LO 1.s 20 3.0 4.0 5.0
Eq. 2 was used for comparative purposes to calculate NORMAL
R,,II,/OBSERVl?D
R,,tI,
Curve 2 of Fig. 2, using the pressure gradient curve Fig. lB-Relationship betwean shaie resistivity parameter
data (Curw.1 nfll,g,. 2] as determined from the.log .. ....... RMtM/R.wMand reservoirfiuid pressuregradient.
1354 JOURNALOF PETROLEUMTECHNOI.0t3Y
havior, until the abnormal pressure section is trav- load is 1.0 psi/ft. To calculate a fracture gradient by
ersed. Then pJD may decrease as shown by the this method one must use the following procedure.
other curves in Figs. 2 and 5. 1. Obtain the formation pore pressure.
Matthews and Kellyg presented a fracture gradient 2. Determine the effective stress, u = 1.0 D p.
equation similar to Eq. 1. However, they introduced 3. Determine the depth D{ for which the matrix
the concept of a variable horizontal-to-vertical stress stress u would be the normal value:
ratio. Fig. 3 is a reproduction of their curves showing
the variable stress ratio as a function of depth for two pi = .
areas. Again one must assume that the overburden 0.;35
4. Use the value of Dt from the preceding step with
o- Fig. 3 to determine K~.
2 .
1~ CURVEL FO:~fl:NmPRESSURE
LOUISIANA OULFCOAST
Throughout the remainder of this work, it is postu-
lated that the assumptions leading to Eq. 1 are valid
and that all of the independent variables are functions
16
of depth. The problem is to determine the relationship
of overburden stress, pore pressure, and Poissons
ratio with depth. Since it is accepted here that abnor-
i mal formation pressure gradients may be determined
from logs, that aspect of the problem is solved. The
20
0,3 0.4 0.5 0.6 0.7 0.8 0.s i .0 next steps are to assume that the overburden stress
MATRIX STRESS COEFFICIENT- Ki gradient is 1.0 psi/ft, then to solve Ea. 1 for the stress
Fig.3-Matrix stress coefficient?
OCTOBER,1969
a
Pw
. P available. These logs were used to plot bulk density
DD vs depth, which is shown in Fig. 6. The values for
(4) bulk density were read at the mid-point of each 1,000-
l: V= Sad
D ft iriterval and averaged step by step downward to
20,000 ft of depth. In this manner, the overburden
and to evaluate Eq. 4 with field data. stress curve of F]g. 7 was produced. The wdue of
A great deal of data from the analysis of hydraulic overburden stress read from the curve at any depth
fracturing treatments in West Texas was published by represents the real average overburden gradient at
Critkndon. These data were used to develop the left that specific depth. Further averaging need not be
curve of Fig. 4. It can be seen that for the producing done.
formations of the West Texas area, the assumptions The same procedure was used for similar data from
S/D = 1.0 and v = 0.25 are valid. wells in the Santa Barbara Channel. Bulk densities
Data given by Costley were used to back-calculate from logs and the resulting overburden stress gradient
the middle Poisson ratio curve of Fig. 4. Note the curve are shown in Figs. 10 and 11, respectively, with
curvature of the trend of Poisson ratio vs depth for the results being similar to those given in Figs. 6 and
the Gulf Coast area. This is caused by the sediments 7. It was concluded that variable overburden stress
being younger and more compressible near the sur- gradient curves, as determined from density logs of
face, but less compressible and more plastic with good quality, are far superior to any assumed constant
depth. For this reason the curve approaches 0.5 as an number.
upper limit. This limit is the Poisson ratio of an incom- Based on Eq. 4, the same field data, and Fig. 7,
pressible material in the plastic failure environment, the Poisson ratio trend for the Gulf Coast area was
Example results of fracture pressure gradient cal- back-calculated and plotted on the right side of Fig.
culations using the middle Poisson ratio curve of Fig. 4. Note that this curve approaches the middle curve
4 and Eq. 1 are included for comparative purposes. at greater depths where the overburden stress gradi-
These are shown by Curve 6 of Fig. 2 and Curve 4 ent does approach 1.0 psi/ft. It was concluded that
of Fig. 5. Figs. 4 and 7 could be used with formation pressure
From the preceding calculations it becomes evident data and Eq. 1 to predict accurately the fracture gra-
that the variation of overburden stress with well depth dients in the Gulf Coast. The same method will apply
must be determined where formations are compressi- in other areas, provided that the overburden stress
ble, such as in the Gulf Coast area. A composite gradient and Poisson ratio curves are determined from
group of density logs from many Gulf Coast wells was good data.
\ 2. FRACTUREGRAOIENT
!im
tymmgT a WILLIS
3. FRAGTUREGRAOIENT
MATTHEWSa KELLY
4. FRAGTUREQRAOIENT
PRESENT METHOO
=
88
g
t
~ \
b 10
n
Is.
16
0 0.2 0.4 0.6 o.e 1.0 1,2 I .4
(3RAOIENT- @ft
The preceding discussion illustrates a method according to the curve of Fig. 7. However, all that
whereby all three variables that control the fracture must be done to arrive at the correct fracture gradient
pressure gradient are determined from well data. A curve is to use Fig. 7 and interpolate between the
nomograph for solving Eq. 1 to predict fracture pres- various gradients of Fig. 9. The interpolation is shown
sure gradients is shown, with an example calculation, as the correct fracture gradient curve on Fig. 9. Note
in Fig. 12. that near the surface, the curve is between the 0.8 and
0.9 curves, and at the 4,000 ft mark it crosses the 0.9
How to Apply the New Techuique curve and approaches the 1.0 curve with depth. This
To demonstrate better how to apply this method of illustrates somewhat elaborately the effect of over-
predicting fracture gradients, an example follows, burden stress gradient on fracture pressure gradient.
showing how the technique should be employed in However, in actual practice only the true fracture
drilling well plans. gradient should be used.
1. Fig. 8 shows a plot of resistivities for a well in 3. The casing points can be selected by making use
the East Cameron area. Data from Fig. 8 are used of the formation pressure gradient curve (minimum
with Fig. ) B to produce the formation pressure curve mud weight curve) and the true fracture gradient
of Fig. 9. Mud weights for this area should be deter- curve. In the example of Fig. 9, surface pipe was set
mined in this manner during the well planning stage. at 3,100 ft. If we assume that minimum mud weights
The mud weight scale at the bottom of Fig. 9 and the were used during drilling, a protective string of pipe
formation pressure gradient curve dictate the mini- must be set at no deeper than 10,400 ft. A vertical
mum mud weight program. line from the pressure gradient curve at 10,400 ft
2. The next step is to determine and plot fracture extends upward to a point slightly above that repre-
gradient vs depth. (This is illustrated in Fig, 9, which senting the surface pipe shoe. In this case, break-
shows five fracture gradient curves. ) It is assumed down and lost circulation would not occur. However,
that the overburden gradient averages 0.8 psi/ft all if drilling were to continue to 11,000 ft before a pro-
the way down. The graphic solution of Eq. 1 shown tective string is set, breakdown would occur. A verti-
by Fig. 12 is used with the variable Poissons ratio cal line from the pressure gradient curve at the 11,000
curve (right-hand curve of Fig. 4) to find the fracture ft mark extends up and intersects the fracture gradient
gradient vs depth. The process is repeated for assumed curve at the 5,000 ft mark. Breakdown would occur
overburden stress gradients of 0.9, 1.0, and 1.1 psi/ft, anywhere between 5,000 and 3,100 ft of depth. This
These four fracture gradient curves are not correct, is simply to say that casing points should be selected
because the overburden gradient varies with depth by using both the formation pressure and the fracture
o o
v
2 2
\
* *
\ !
6 6
\
\
= 8
ii
ij:
j! 10
&
w
o 1, z
*
1
\ \
w
o
12
\ 12
\\\ \
1
14 14
\
\
\ \,
16 \
16
\
\
\
18 te
Fig. 6-Composite bulk density curvefrom density log Fig.7-Composite overburdenstressgradient for all
data for the Gulf Coast. normally compacted Gulf Coast formations.
OCTOBER,1%9 1357
. o
tl
12
14
1
o
2
2
3
- FRAGTURE
4 -
PRES6URE
4
6
+
&
08
f *
ii
76
E \
& 10 h
w E
o \ w
FORMATION
o
PRESSURE - \, 7
Ii?
\
e
\
I4 $
\
t,
s
I6
30UlVALENTl#UD:E10~ LS#OA L
9 10 II II* I7 19 1B 10
t ,
r 0.76 o.e o.e5 0.9 0.s5 IQ h s
0.3 0.4 0.6 0s 0.7 Os o~ Lo
OVERBURDEN LOAD - PS1/FT
ORADIENT- ml/ft
Fig.9-Fracture gradient with va.iableoverburdenand fig. 11--overburdan stress gradiant,
Poissonsratio inciuded,Well C, East Cameron. Santa Barbara Channel.
sl~ y
D
0.9 ~
0.8- .
0.7- -
\< .
VD
0.6 L 1s.0 1.0
0.0
!8.0 I
\ :1 ;
0.90
0.4
OCTOBER,1%9 1359
Fracturing occurred when a gradient of 0.57 psi/ft smoothly to 1.0 psi/ft at about 20,000 ft of depth. A
was reached, as shown by the sudden change in slope similar trend has been established for the Santa Bar-
at about 700 B/D injection. New in-fill wells in the bara Channel area.
same area are hydraulically fractured with gradients 5. The present method is a modification of the Hub-
of O68 to 0.70 psi/ft. The relatively low fracture bert and Willis approach, using a variable Poisson
gradients of old producers are due to low formation ratio and a variable overburden stress gradient with
pressures. Many old producing wells are fractured depth. The method will predict the fracture gradient
unintentionally when converted to injection wells. for the Gulf Coast area with as much accuracy as the
The pressure in the producing formation is so low log-derived pressure gradient,
that it cannot withstand even the gradient of fresh 6. Using this approach and field data, the same
water. method can be developed for other areas.
One more example of the effect of formation pres- 7. The method is very simple, and Fig. 12 maybe
sure on fracture gradient is that three pressure-rate used as a working chart for any area.
tests were run on a California injector on Dec. 13,
1961; June 11, 1962; and Dec. 3, 1962. Fracture NomencIatnre
pressure gradients were determined at sudden changes D = depth, ft
in slope of pressure-rate plots. These gradients were Di = equivalent depth of lowermost normally
0.748, 0.864, and 0.993 psi/ft on their respective pressured formation, ft
dates, These data illustrate how the fract-urepressure Ki = matrix stress coefficient
gradient is affected by changes in formation pressure P,, wellbore pressure, psi
=
around a wellbore. P = formation pressure, psi
S = overburden stress, psi
How to DeveIop the TecIndque u = net effective overburden stress, u = S p,
for Other Areas psi
The data necessary to develop this method for other v = Poissons ratio
tectonically relaxed areas of the earth are as follows:
1. Overburden stress gradient vs deuth, Such data Acknowledgments
can be derived from bulk densities taken from logs, I should like to express appreciation to the manage-
seismic data or shale density measurements. A plot ment of Continental Oil Co. for permission to publish
of bulk density vs depth can then be converted to a this oaper. Special thanks are given to C. H. Hefner
plot of average overburden stress gradient vs depth, for his preparation of the nomograph, Fig. 12.
2, Actual fracture pressure gradients for several
depths. These can be lost-circulation or squeeze data References
or actual fracturing data. 1. Hubbert, M. King and Willis, D. (3,: Meehanics of Hy-
draulic Fracturing, Tram., AIME (1957) 210, 1S3-166.
3, Formation pressures that apply to the data in
2, Howard, G. C. and Scott, P. P.: An Analysis and the
Item 2. (Irl Items 2 and 3, the depths must corre- ~lntr;; of Lost Circulation, Trans., AIME (1951) 192,
spond.) -.
With these data and Eq. 4, the Poissons ratio curve 3. Matthews, W. R. and Kelly, John: How to Predict Forma-
tion Pressure and Fracture Gradient, Oil and Gas J.
for the area can be back-calculated and plotted vs (Feb. 20. 1967).
deuth. The result will be a curve similar to those of 4. Goldsmith, R. and Wilson, G.: staff engineers, Rodu~ii~n
Fig. 4 and a curve similar to that of Fig, 7. Whh these Engineering Serviees, Continental Oil Co., Hm.sston, Tex.,
curves and Fig. 12, fracture gradients can be pre- private communication.
5. Costley, R. D.: Hazards and Costs Cut by Planned Drill-
dicted quite easily and quickly. These values can be ing Programs, World Oil (Ott., 1967 ) 92.
plotted as a function of depth and the resulting curves 6. Crittendon, B. C.: The Mechanics of Design and Inter-
can be used in all t.k operations previously described, pretation of Hydraulic Fracture Treatments, J. Pet. Tech
(Oct., 1959) 21-29.
Conclusions 7. Wuerker, R. G,: Annotated Tables of Strength and Prop-
ert ies of Rocks, Drilling, SPE Petroleum Transactions
In summary, several conclusions have been drawn. Reprint Series No. 6 (1963) 23-45.
1, In drilling well plans, well stimulation plans, and 8. Hottman, C. E. and Johnson, R. K.: Estimation of For-
secondary recovery plans, fracture pressure gradient mation Pressures from Log-Derived ShaIe Properties,
1. Pet. Tech. (June, 1965) 717-722. iJWT
should be considered.
2. Poissons ratio for rocks increases with depth in
the Gulf Coast area. Original manuscript receivad In Socfety of Petroleum Err@neara
3, The Poissons ratio trend, which can be back- office July 2, 1968. Revised manuscript received June 26, 1969.
Paper (SPE 2163) was presented at SPE 43rd Annual Fall Meetln~
calculated using field data, will not be exactly the same held in Houston, Tex., Sept. 29.Ott. 2, 196% end at SPE FifUI Annual
for data from different areas. Eastern ReftIonal Meeting held Irr Charleston, W. Vs., Nov. 7.6,
1968. @ Copyright 1969 Amarlcen Institute of Mlnlng, MatallurgL
4. In the Gulf Coast area, the average overburden cal, and Petroleum Enslneere, Inc.
stress gradient does not equal 1.0 psi/ft, but instead This paper will be printed In Traneactiona volume 246, which
is about 0.85 psi/ft near the surface and increases will cover 1969.