SPE 84214

Evaluating Well Performance and Completion Effectiveness in Hydraulically Fractured

Low-Permeability Gas Wells
David D Cramer, SPE, BJ Services Co.
Copyright 2003, Society of Petroleum Engineers Inc.
This paper was prepared for presentation at the SPE Annual Technical Conference and
Exhibition held in Denver, Colorado, U.S.A., 5 8 October 2003.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the Society of Petroleum Engineers is
prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300
words; illustrations may not be copied. The abstract must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.

Abstract
Hydraulic fracture stimulation often dictates the economic
outcome of wells completed in low permeability gas
reservoirs. Attempting to understand well performance - the
rate and pressure behavior of a well over its productive life provides the opportunity to discover the elements driving
stimulation and completion effectiveness.
This paper
demonstrates the integrated use of practical reservoir
engineering methods to evaluate well performance, identify
flow regimes and distinguish between reservoir and
completion
induced
behavior
in
low-permeability
hydraulically fractured gas wells.
Introduction
Well performance is the rate and pressure behavior of a well
throughout its productive life. It can be evaluated through
various analytic methods; most are based on the constant
terminal rate solution of the radial diffusivity equation.
Superposition can be applied as needed to account for major
geometrical and flow rate discontinuities (e.g., sealing
boundaries, shut in periods.)1-3 When analyzing gas reservoirs,
reservoir fluid properties are highly pressure dependent, and it
is necessary to replace pressure with the real gas pseudopressure function m(p) to linearize the flow equations.4

m( p ) = 2

( p) z( p)dp

.. (1)

pB

Radial
diffusivity
following equation.

simplifies

to

the

1 m( p) c m( p)
r
=
r r
r
k
t

.. (2)

Embedded in this equation is the hydraulic diffusivity

term (c / k) that determines and is proportional to the speed
that production or injection impulses (transients) travel
radially through the formation.
Transients propagate
relatively slowly in low permeability rock and this effects the
duration of the drawdown period necessary to perform
transient analysis. The hydraulic diffusivity/ time relationship
from a point or line source can be evaluated with the radius of
investigation equation.
Ri =

t=

kt
948 i ci

948i ci r 2
k

.. (3) or
.. (4)

Hydraulic diffusivity is sufficiently low in most low

permeability gas reservoirs that well stimulation is usually
desired to speed the recovery of reservoir fluid. Well
stimulation is usually achieved by a hydraulic fracture
treatment. The purpose of the treatment is to extend a high
conductivity planar conduit, usually vertical in orientation,
from the wellbore into the reservoir rock. This is usually
achieved in sandstone and many carbonate reservoirs by the
emplacement of propping agents which inhibit fracture closure
and may possess high porosity and permeability when
distributed as a multi-layer pack. Viscous delayed-breaking
gel is often used as a carrier fluid to achieve widespread
placement of proppant in the fracture. When fracturing
treatments are successful, the increased reservoir flow area
created by the fracture can greatly increase the production
capacity of the well in addition to any benefit enabled by the
mitigation of near-wellbore formation damage. To enhance
this effect, the flow capacity of the fracture must be great
compared to the flow capacity of the native reservoir exposed
to the fracture. This characteristic is reflected in the
dimensionless fracture conductivity term FCD, which is the
ratio of the ability of the fracture to deliver fluid to the
wellbore to the ability of the reservoir to deliver fluid to
the fracture.

SPE 84214

FCD =

k f bf
kX F

.. (5)

Initially, fluid flows linearly from the reservoir into

the fracture as a well is produced. Then, the pressure transient
extends from the fracture into the formation, and flow in the
reservoir transitions from a linear to pseudo-radial flow
pattern provided that the reservoir drainage radius (XE) is
relatively long compared to the fracture half-length or radius
(XF). This phenomena was evaluated by Gringarten et al, who
evaluated the basic case of a well centered in a bounded
square-sided isotropic reservoir possessing an infinite
conductivity fracture (i.e., pressure drop down the fracture =
0.)5-6 Their analysis for a reservoir producing a slightly
compressible liquid resulted in the log-log type curve plot
shown in Figure 1. Substituting m(p) for pressure enables the
use of this plot for analyzing gas reservoirs.
As indicated in the Gringarten
dimensionless time in a fractured well is

TD =

0.000264kt
i ci X F2

type

curve,

.(6)

Solving for t, the transposed equation is.

t=

3788TDi ci X F 2
k

.. (7)

The dimensionless pressure term in gas units is

PD =

kh
(m( pi ) m( pwf ))
1422QG i

.. (8)

In the Gringarten analysis, linear flow is present at

the start of post-fracture production, and is revealed by a halfslope trend in log-log coordinates. The end of linear flow
occurs at a dimensionless time of only 0.016 (in the case of a
uniform-flux fracture, the end of linear flow occurs at a
dimensionless time of 0.16.) After a long transition period,
pseudo-radial flow occurs at a TD~3. For pseudo-radial to
develop in a fractured well, the drainage radius/ fracture halflength ratio (XE /XF) should be 5 or greater.6 Similar work
done by Wattenbarger and Ramey found that the onset of
pseudo-radial flow in a well with an infinite conductivity
fracture occurs at a TD ~ 1, with correspondingly less stringent
XE /XF requirements.7 During pseudo-radial flow, the infinite
conductivity fracture acts as an expanded wellbore with an
apparent radius Rwa having the following relationship.

The type curve method is not well suited by itself for this
purpose since, in employing a logarithmic pressure scale, the
resolution is pressure diminished and the transient response is
non-linear.8 The Gringarten plot in semi-log coordinates in
shown in Figure 2.
The slope of the correct straight semi-log line is
proportional to k as follows.
slope = m = 1637

QG T
kh

.. (11)

QGT
.. (12)
mh
The slope and 1-hour intercept of the correct semilog straight line is proportional to skin as follows.
k = 1637

s = s + D QG =
m( p ) m( p )

k
i
1hr
log
1.151
2

i ci rw

.. (9) or
.. (10)

The significance of pseudo-radial flow is that if it is

present, the straightforward semi-log linear analysis method
(MDH plot) may be used to analyze field data and derive
reservoir flow parameters and hydraulic fracture effectiveness.

(13)

The D term in the apparent skin factor s is the nonDarcy velocity-dependent pressure drop coefficient. For high
flow rate wells where velocity-dependent pressure drop could
be important, a multi-rate test should be performed to
determine the true skin factor.3 Skin (s) is related to Rwa and
XF as follows.
Rwa = Rw e s
X F = 2 Rw e

.. (14)

(15)

Table 1 shows the exponential relationship of semilog derived skin to the apparent or effective (i.e., infinite
conductivity proxy) XF. With very negative skin values (e.g.,
-6 or less), the computed XF is very sensitive to adjustments in
the semi-log trend line (slope) and the resulting 1-hr intercept
(i.e., a slight slope adjustment results in a large XF change.)
XF(ft)

skin
0
-1
-2
-3
-4
-5
-6
-7
-8

1
2
5
13
36
99
269
731
1987
-s

Rwa=0.5XF
XF=2 Rwa.

+ 3.23

rwa/XF = rw e /XF =
XF
rw =

0.5

-s

= 2rw e

4 in.

Table 1: Relationship of Skin to Effective XF

SPE 84214

Inspection of the semi-log type curve plot in Figure 2

shows that if the semi-log trend line is selected from data
measured prior to the development of pseudo-radial flow (e.g.,
TD<1-3), k will be over-predicted (flatter slope) and XF will be
under-predicted. Conversely, if the semi-log slope was
selected after the onset of pseudo-steady state (depletion)
flow, or at any time after no-flow boundaries effect the
pressure response, k will be under-predicted (steeper slope)
and effective XF will be over-predicted. A variety of analysis
strategies are necessary to have the best chance of identifying
the various flow regimes during the production life of a well.
Plotting the derivative of the pressure response has been used
to enhance the diagnostic capability of type curves in
identifying flow regimes9, but this method requires the use of
accurate, high-resolution data with a low noise level.
Gringartens analysis presumes infinite fracture
conductivity (a solution for uniform flux fracture conductivity
is also covered in the paper) yet in many field cases there can
be a significant pressure drop down the fracture, indicative of
finite conductivity fractures with a moderate to low FDC.
There are various type curves that evaluate the global pressure
response of finite conductivity fractures, but these logarithmic
pressure scale plots suffer from the same issues noted in the
discussion on the Gringarten plot.10-12 Fortunately, finite
conductivity fractures can also be evaluated with the semi-log
plotting method during the pseudo-radial flow period.
According to Cinco-Ley, TD at the start of pseudo-radial flow
(TDbs) for a finite conductivity fracture can be estimated by the
following equation.11
TDbs = 5exp 0.5 FCD 0.6

.. (16)

Using this equation, for FCD ranging from 1 to 300

(the latter essentially an infinite conductivity fracture), TD at
the start of pseudo-radial flow ranges from 0.45 to 0.97. For a
finite-conductivity fracture, the relationship between skin,
apparent wellbore radius and actual fracture half-length is not
as straightforward as with an infinite conductivity fracture.
Cinco evaluated this situation and the results are charted in
Figure 3.10 The y-axis scalar RW/XF is the ratio of apparent
wellbore radius to fracture half-length (Rwa/XF). As shown
previously, for an infinite conductivity fracture, this ratio is
0.5, and is marked by the superimposed horizontal line. For a
FCD of 1, the ratio is 0.2. In this case, the Rwa/ XF relationship
is as follows.
Rwa = 0.2XF
XF =5 Rwa
X F = 5 Rw e s

.. (17) or
(18) and
.. (19)

The actual fracture half-length at a FCD=1 will be 2.5

times the fracture half-length of an infinite conductivity
fracture exhibiting the same value of skin or apparent wellbore
radius. This is only the case for a fracture with an FCD=1. The
Rwa for other FCD values can be derived from Figure 3. Figure
3 has limited use as a tool to evaluate field data because the
fracture conductivity in the field is not known a priori. As

with fracture length, there are factors that can cause the
modeled fracture conductivity response to deviate
significantly from the actual response. If permeability can be
resolved from semi-log analysis, there is a possibility that the
previously mentioned finite conductivity fracture type curves,
which focus on the early TD response of idealized fractures,
could resolve apparent fracture half-length, based on the
infinite conductivity assumption, from the actual fracture halflength at some specified and finite value of fracture
conductivity. Also, specialty plots pressure change versus
quarter root or square root of time - exist that enable
calculation of XF from the bilinear (this flow period ideally
occurs prior to the linear flow period in a finite conductivity
fracture) and linear flow periods, respectively.10 An issue with
any plotting technique that relies upon early-time field
production data, is that fracturing fluid cleanup, changing nonDarcy pressure drop component, rapid pressure and rate
changes, and operational system changes often distort the data
for days and sometimes weeks, regardless of the type of
coordinate system used. Middle-time data has the best chance
of avoiding these pitfalls. And it has been shown that except
for very early time well performance, the apparent or effective
XF derived from evaluation of the pseudo-radial flow period
with the semi-log plot, describes very well the overall
performance of even ultra low conductivity fractures.13
Late time data is often useful for material balance
calculations. When all reservoir boundaries have been reached
and pressure is depleting uniformly in the reservoir, data will
trend as a unit slope in log log coordinates, as shown with the
Gringarten depletion stems in Figure 1. Comparing the XE/XF
ratio of field data going up a particular Gringarten depletion
stem offers a check on XF computed in the semi-log plot and
drainage area (XE) computed by other material balance
methods, such as the Cartesian plot.3 However, at this lowenergy stage of well performance, liquid slugging often
distorts data trend.
Working With Production Data
The precision of well performance solutions depends on the
quality and quantity of the obtained data, especially in the case
of commercial production, which is analogous to a drawdown
well test.5,6 Production data consists of gas, condensate/oil
and water production volumes, and well pressure, recorded
over time. The measured pressure is normally wellhead
pressure at the tubing or annulus conduit. Correlations are
used to calculate hydrostatic and friction components to
convert wellhead pressure to a bottomhole pressure [M(pWF).]
Production data is the most available type of well performance
data. In most low permeability cases, cost or test-duration
considerations (due to very low hydraulic diffusivity) prohibit
the application of the more controlled pressure buildup test
(liquid loading problems often preclude shut-in tests as well.)
The methodology and application of production data analysis
is the main tool used in the quest to understand well
performance in low permeability gas reservoirs and is the
focus of this paper. Other data sources and engineering
methods should be used when appropriate to enhance, evaluate
or expand the results of production data analysis;

SPE 84214

demonstrations of this are shown in the subsequent

case studies.
With production analysis, the incremental time step is
critical and its necessary magnitude is a function of
permeability and reservoir size. Relatively high permeability
intervals may require a sampling rate in terms of minutes or
hours; with the typical low permeability well, a daily sampling
rate is usually sufficient. Any shut-in period should be
accounted for, although with cyclic production methods such
as plunger-lift, this may not be practical. The reciprocal
productivity index semi-log method is an effective method of
normalizing the inevitable rate and pressure variations that
characterize production from the typical low permeability gas
well.3,14-16 With this method, a modified semi-log graph of
[m(pi)-m(pwf)]QG versus log (t) will generate a straight line
or slope (m) in the middle-time or pseudo-radial flow region.
Permeability and skin can be derived from the slope and onehour intercept of that line.
T
m'h
s ' = s + D QG =

k = 1637

.. (20)

m( pi ) m( pwf )

1
k
1.151
log

QG m

1hr m '
i ci rw

..(21)

+ 3.23

The objectives of analyzing well performance include

assessing reservoir flow capacity and size, deducing the
presence of geologic discontinuities, forecasting future well
performance, and identifying operational and completion
related effects. As noted previously, most low-permeability
gas wells require hydraulic fracture stimulation to extend the
influence of the wellbore over a large surface area, which can
significantly improve well productivity (i.e., rate (QG)
pressure drawdown from reservoir extremities to wellbore
[m(pi)-m(pwf)] and accelerate hydrocarbon recovery.
Simulators coupling solutions of hydraulic fracture geometry
and proppant transport are used to optimize the design of
fracture treatments and offer a means to estimate the
differential benefits of various treatment alternatives.
However, it is often the case that the forecasted well
performance based on the idealized fracture and wellbore
condition falls short of the achieved well performance. The
appropriate application of well performance analysis can offer
significant qualitative and sometimes quantitative insight into
reservoir and fracture properties and completion efficiency. In
the best cases, the separate effects can be isolated and this
information can be uploaded into more involved methods of
analysis such as numerical simulation for enhanced insight.
The Basics of the RPI Production Data
Analysis Method
Crafton introduced a mixed solution method of production
data analysis that uses the abovementioned variable
rate/pressure normalization technique and incorporates
additional analytic and numerical solutions to verify and

enhance the semi-log solution.14-15 The user interface is a

graphical analysis system known as RPI (which stands for
reciprocal productivity index.) The rate-normalized semi-log
(MDH) plot allows the separate resolution of the reservoir
flow parameter, the permeability-height product (kh) and the
fracture flow parameter, which is the apparent or effective
(infinite conductivity proxy) fracture half-length (XF).
Considering that well production is a special form of a
variable-rate drawdown test, if the production rate change is
smooth (i.e., no more than 20% change per time step)17 and
the production data is of reasonable quantity and quality, the
RPI method can be used to evaluate well productivity using
surface-measured data.
To augment the modified MDH analysis, companion
diagnostic plots are used in RPI and are shown in an example
from an actual analysis in Figure 4. The Type Curve (TC) plot
(northwest plot), displays different curves on one graph: halfslope, Agarwal finite fracture conductivity, and Gringarten
infinite fracture conductivity/bounded reservoir type curves.
It allows the identification of flow regimes (half-slope for
linear formation flow), reservoir boundaries (unit slope),
clean-up effects, layered reservoir effects, and provides a
check on the MDH results. In reservoirs displaying radialflow behavior, data corresponding to the MDH straight line
should fall on the infinite conductivity line of the Type Curve
plots, since the MDH solution is an effective fracture halflength solution and presumes infinite conductivity flow
behavior in the fracture. The Pseudo-Steady State (PSS) plot
(southwest plot) gives another check on the transition from
infinite-acting (curvilinear trend) to bounded or depletion flow
(straight-line trend), and it decompresses events with its
Cartesian time scale. The slope of the Cartesian plot after the
development of depletion can be used to compute reservoir
drainage volume/ area.3 The Equivalent Rate and Split Rate
Production plots (south-central and southeast plots) compare
the actual production data with simulated data (fractured well
production simulator) based on the semi-log derived kh and
XF parameters. In a thorough mixed solution analysis, all
plots are checked for consistency.
In the event of shut-in periods, RPI reinitializes the
production, taking into account shut-in time on the pressuretime relationship and starting a new transient when production
resumes. In this way, RPI enables pattern recognition of the
effect of reservoir and mechanical events/changes in the life of
the well.
As noted above, RPI enables the ability to recognize
the transition from infinite-acting to bounded (depletion) flow.
When this event is identified in producing time, the program
then allows the resolution of drained area, which combined
with input petrophysical and fluid propeties, internally
calculates initial hydrocarbons-in-place and estimated
ultimate recovery.
Finite-conductivity fractures can be approximated by
introducing a positive skin term and increasing a specified
fracture half-length, shifting up in a northwest-trending
manner on the Type Curve to a finite-conductivity Agarwal

SPE 84214

line, while maintaining a trend match on the MDH plot with

the actual well data. However, in terms of comparing a group
of similar wells and forecasting future well performance, the
effective fracture half-length concept is adequate and
more reliable.13
RPI is a single-layer, single-phase simulator that
combines the oil, gas and water rates into a combined mass
flow rate (this feature is adjustable). In-house comparisons of
RPI results with well test analysis and numerical 3D reservoir
simulator results have been in good agreement, even in multilayer, variable-layer-property cases, especially in regard to
deriving total-well kh.
As in all well test analysis methods, there is no
practical way to uncouple permeability from pay height.
Incorrect selection of pay height will affect both the
permeability and effective fracture half-length calculations.
The effect of an incorrect pay-height calculation can be
quickly evaluated with the following equations.
XF = Initial XF (initial height corrected height)0.5 ..(20)
k = Initial k (initial height corrected height) (21)
DA = Initial DA (initial height corrected height) (22)
Other Considerations
Long-term formation linear flow is an indication of
permeability anisotropy such as swarms of natural fracture
that have a preferred direction, or the existence of a long
narrow reservoir such as a fluvial channel or turbidite deposit.
The linear flow model presumes flow into a plane as opposed
to flow towards a line as in radial flow. The computed XF is
the half-width of the reservoir unit rather than the hydraulicfracture XF that is evaluated in the radial-flow model.though
the calculation of skin. The MDH semi-log plot is not
applicable. The appropriate specialty plot is the t plot, yet
kh and XF are inseparable as both parameters appear in the
equation for the slope of the line, so k and XF cannot be
uniquely solved for. Often the best that can be done in a linear
flow case is a material balance solution for gas-in-place and
reservoir drainage volume. In a parametric-comparison study
(such as referred to in Case Study 3), wells exhibiting linear
flow normally must be excluded from the analysis (see Tables
2 and 3.)
The required frequency and resolution of rate and
pressure data is a function of permeability and requirements
increase as permeability increases. In higher permeability
reservoirs, production data should be collected at a time
frequency of minutes or hours rather than days. This type of
information is rarely available in public databases and needs to
be diligently measured, collected and stored in a computerready form by the well operator.
Simplifying assumptions in the RPI production
analysis scheme include single layer, single-phase fluid (with
the option of converting produced condensate and/or water to
a gas equivalent volume), cylindrical or square reservoir

shape, isotropic petrophysical properties, and stable

permeability over time. Natural fractures are often the
dominant flow conduit in low permeability gas reservoirs. A
well-connected natural fracture network will exhibit radial
flow. As noted above, strongly anisotropic fracture systems
will exhibit linear flow, shown by a long-term half-slope trend
in the RPI Type Curve plot. In the later case, semi-log or any
specialty plot analysis to isolate formation (kh) and fracture/
wellbore properties will not be possible. Additionally, if
reservoir drawdown increases significantly during the
productive life of the well, the resulting increased compressive
stress acting against open natural fractures may tend to reduce
their effective width, sometimes reducing permeability
significantly.18 This phenomenon seems to be most prevalent
in high pressure formations, during the time period when
wellhead pressure is greater than line pressure but is
decreasing over time. In these cases, it is imperative to defer
evaluation until wellhead pressure and drawdown stabilizes.
Departures from these assumptions can be acceptable
within certain limits. For instance, field data will rarely transit
perfectly up a Type Curve unit-slope depletion stem. This is
because of factors such as a well being imperfectly centered in
the local drainage system, irregularly shaped no-flow
boundaries, composite reservoir behavior (i.e., changing
reservoir flow quality with distance from the well), and multilayer completions with differing layer size and petrophysical
properties19 (therefore, differential depletion among layers).
Often, there is not a priori knowledge as to the location of the
well in the drainage system and the specific size of individual
layers. Yet most of these issues can be accounted for as all
zones trend toward a pure depletion or semi-steady state flow
response. It is important that the initial reservoir pressure be
fairly uniform among layers.20 For instance, if the pressure
distribution in an individual layer(s) is affected by production
from previously completed offset wells, this could negate the
validity of the analysis. Another consideration is that the
reservoir unit drainage radius (XE) must be lengthy in relation
to XF, at least three times as long (Gringarten and others.) In
essence, all or most layers should exhibit a period of infiniteacting, psuedo-radial flow in the same general real time frame;
permeability should be well within an order of magnitude
among all layers.
Case Study #1
The subject well produces from the Frontier formation in the
LaBarge Platform trend of the Greater Green River Basin of
Wyoming. It was drilled as a field extension edge well; two
wells in adjacent 160-acre locations had each produced in
excess of 1 bcf of gas from the same pay horizon, so some
degree of pressure depletion from the initial field pore
pressure gradient of 0.53 psi/ft was expected. The main pay
zone at 10,450 ft is an upper shoreface marine sand deposit. A
bioturbated lower shoreface marine interval below this pay
zone was selected to be perforated yet the poor grain sorting
and high apparent irreducible water saturation make this
interval a dubious pay pick; it is probably marginally
productive at best. A thin fluvial sand above the marine sands
showed some pay characteristics.

The gross interval was perforated and stimulated in

two stages. The lower marine sands were treated with
150,000 gallons of fluid and 300,000 pounds of proppant.
That treatment was cleaned up for 1-1/2 days, then lower
perforations were isolated with a bridge plug and the upper
fluvial sand was perforated and treated with a relatively small
job, consisting of 13,500 gallons of fluid and 31,500 pounds of
proppant. Both treatments utilized N2-assisted (20% N2)
crosslinked carboxymethyl guar (CMG) gel and 20/40 mesh
proppant. The rate/ pressure history matched results of the
treatments using a lumped parameter 3D fracture simulator are
shown in Figure 5.21 As seen in previous analyses in the area,
extensive fracture height growth was exhibited in the
simulations. The first treatment is shown to communicate to
the upper fluvial sand and the second treatment is shown to
have communicated back into the lower marine sand. This
indicates that the main pay zone suffered a dump of additional
load fluid after depleting its pressure during prior load
recovery. The outlet for flowback fluid following the second
treatment was far uphole through the fluvial perforations,
further impairing liquid cleanup in this zone.
Well productivity following the treatments was
disappointing and worsened after several weeks of production.
After six months of production, the well was shut in to
conduct a pressure buildup test (PBU); bottomhole pressure
was estimated from measured surface pressure data. Both the
pressure buildup during shut in and the production data were
analyzed (Figures 6-9). The results from the two methods
agree well and indicate that the reservoir accessed by the
wellbore and fracture shows some depletion from the
offsetting wells (p*=4600 psi, 0.44 psi/ft gradient), reservoir
flow capacity is very low (kh=0.40 md-ft), and the effective
fracture half-length is short at the time the well was shut in for
the pressure buildup (XF ~20 ft). The results from the pressure
buildup helped guide the selection of the semi-log straight line
used in the production analysis. The production analysis is
very instructive in showing several major changes in well/
fracture condition. Initially, after a short cleanup period, the
well exhibited an effective XF ~ 90 ft, despite the use of a
premium fracturing fluid system. This is much less than the
300 to 350 ft calculated from the fracture simulations, and less
than the 200 ft effective XF commonly measured in production
analysis of other Frontier wells in the area. Production fell off
sharply after day 70, dropping below 100 mcf/d. A shift of the
MDH trend line up to that portion of the production data
indicated that skin increased to 0, and therefore, the effective
XF dropped to 0. The improvement in well productivity
starting at day 150 is coincident with the installation of a
plunger lift, which assisted in removing water from the
wellbore. A shift of the MDH trend line down to that event
showed that skin decreased and thus fracture half-length
increased to about 20 ft.
These events suggest that liquid loading can have a
dramatic impact on the effective XF of low energy, low kh
wells. Gas is extremely buoyant and much more mobile than
water. If it is flowing at a low velocity in the fracture, the gas
will tend to rapidly migrate through any standing liquid and
flow along the upper portion of the fracture. Constraining the

SPE 84214

vertical flow area in the fracture in this manner is physically

the same as reducing the fracture length. It will be manifested
as a reduced effective XF in the well performance analysis.
The plunger lift installation was helpful, but is probably not
dewatering very much of the water logged fracture. An
alternate production analysis is shown in Figures 10-11. The
results from this analysis do not agree well with the PBU
analysis. But without the benefit of the PBU analysis, it
would have been a credible solution.
In retrospect, well performance problems were set up
by the approach taken in the stimulation plan. The marine
sand interval was impaired in unloading the fracturing fluid
load water, and only a portion of the fracture was ever open to
gas flow. Because of this, the gas flow rate quickly dropped
below critical flow velocity needed to continuously remove
liquid from the wellbore. Once a standing column of liquid
accumulated in the wellbore, it likely migrated back into the
fracture, further reducing the flow area open to flow and thus
reducing the effective XF, ultimately to zero. A more effective
approach for this relatively simple multi-zone completion is
the limited entry perforating and stimulation method, in which
all zones are stimulated and cleaned up all at once.22-23
Case Study #2
The subject area is the Baldy Butte field in the Washakie
Basin of Wyoming, which is productive in the upper
Mesaverde (Almond) formation. A performance study of 29
wells was conducted in 1998 and the results were chronicalled
in paper SPE 49046.24 For each well, wireline log analysis was
used to assign a net pay value and hydrocarbon porosityheight product (h) and the RPI method of production analysis
was used to attempt derivation of the permeability-height
product (kh), permeability (k) , effective fracture half-length
(XF) and gas-in-place (drainage area). One of the most
notable findings in the paper was that the economic
dominance of Company A was attributable to achieving
significantly greater effective XF per well than Companies B
and C, even though Company A used less fracturing fluid and
proppant in their treatments (see Figure 12). In turn, this was
attributed to paying fastidious attention to liquid removal
throughout the completion and production life of the wells.
Another significant finding was the small apparent drainage
area per well (80 acres) as compared to well spacing (160
acres), suggesting the viability of increased well density for
the field. A factor reducing confidence in the analysis was the
short duration of production of several of the wells in study
(only 100 days of production in two cases.) To reassess the
results of the study, 30 months of additional daily well
performance data was obtained for the five original wells of
Company A and 8 months of data on Baldy Butte 7, a downspaced well drilled by Company A to access the viability of
in-fill drilling on smaller well spacing. The RPI method of
production analysis was used to derive reservoir and
completion parameters. Summary of the results are contained
in Tables 2 and 3.

SPE 84214
Well
Baldy Butte
Baldy Butte
Baldy Butte
Baldy Butte
Baldy Butte
Averages:

2
3
4
5
6

7
pay (ft) kh (md-ft) kh (md-ft) k (md) k (md-ft) % change flow
original
revised
original revised
geometry
56
4.42
1.01 0.079
0.018
-77% radial
14
1.38
1.03 0.086
0.074
-25% radial
43
0.90
0.82 0.021
0.019
-9% radial
42
linear
34
1.29
0.99 0.038
0.029
-23% radial
37.8
2.00
0.96 0.056
0.035
-52%

Table 2: Revised Study, Reservoir Flow Parameter,

Company A Wells, Case Study 3
Well

xf (ft)
xf (ft)
% change drainage area gas-in-place
original revised
(acres)
(bcf)
Baldy Butte 2
172
216
20%
37
1.959
Baldy Butte 3
224
263
15%
185
2.775
Baldy Butte 4
163
217
25%
40
1.529
Baldy Butte 5
50
2.132
Baldy Butte 6
108
155
30%
38
1.15
Averages:
167
213
22%
70
1.909

flow
geometry
radial
radial
radial
linear
radial

Table 3: Revised Study, Fracture Flow Parameter

and Material Balance, Company A Wells, Case
Study 3
The average drainage area per well equaled 70 acres
although late time liquid slugging makes material balance
calculations difficult (see Figure 13 for example.) The
average kh product decreased 50% as compared to the original
study, from 2 md-ft to1 md-ft. Average effective XF increased
22%, from 167 ft to 213 ft. The analytic results changed
because with more data and producing time, the apparent
MDH trend line became steeper, the y (or 1 hr) intercept
lowered, therefore kh decreased and effective XF increased. In
the initial analysis, in some of the wells, the flow regime was
still in transition from linear to pseudo-radial flow, and a false,
too-flat slope was selected (see Figure 2).
The Baldy Butte 7 was completed in a manner similar
to the original wells listed in the initial study, except the gelant
loading of the crosslinked guar-borate fracturing gel was
reduced by 28%, from 35 lb/1000 gal to 25 lb/1000 gal.
Production analysis of this well indicates infinite-acting flow
still prevails, the kh product is 30% below the Company A
field average, yet depletion was not encountered and the well
exhibits a XF very close to the field average (Figure 14).
Case Study #3
The subject well is completed in a 24 ft thick upper shoreface
bar sand deposit in the upper Mesaverde (Almond)
formation in the Washakie Basin, Wyoming (Figure 15). Well
depth is 11,200 ft. Porosity in this zone is about 14%, well
above the arbritrary 10% pay cut-off for the area, and initial
pore pressure was estimated to be 6000 psi. It was treated
with 105,000 gallons of a 40 lb/1000 gal guar-borate gel
system and 316,000 pounds of 20/40 mesh intermediate
strength ceramic proppant. Post-treatment daily production
and wellhead pressure data were obtained for the first 540
days of production and monthly production data for the last
183 days. The production analysis is contained in Figure 4
(shown previously). This is an excellent example of a
relatively large reservoir of moderate flow capacity (kh=2.4
md-ft) that exhibited excellent and sustained post-treatment
clean-up, and very good fracture stimulation results (effective
XF = 318 ft). The various analytic plots are consistent in

interpretation and show a very smooth response over time.

Although a drainage area was assigned in the analysis (290
acres) to better match the late time monthly data, there is no
clear evidence of bounded flow in the Type Curve plot and it
is possible that infinite acting flow is still exhibited at the end
of data. If this is the case, using the radius of investigation
equation shown above, the minimum drainage radius is 2227
ft and the minimum drainage area is 358 acre. Subsequent
development of the field proved that the areal extent of this
pay sand was larger than 358 acres.
Note the increased slope of the semi-log plot as the
flow regime transits from linear to pseudo-radial flow. If the
semi-log slope exhibited during the linear and early transistion
time period was used to derive parameters, kh would have
been over-estimated and effective XF under-estimated. The
field data lines up within the first week of production with the
infinite conductivity Gringarten type curve (base curve),
indicating that dimensionless fracture conductivity is high;
data associated with finite conductivity fractures10-12 will lay
above the base curve during the early stages of production.
To check the results of this interpretation, a finitedifference fractured-well production simulator was used.25
Wellhead pressure measurement and the permeability derived
from the RPI production analysis were used as a fixed model
inputs. The proppant distribution calculated by the lumped
parameter 3D hydraulic fracturing simulator was initially input
into the production simulator and iteratively reduced in length
until the best early and middle time rate match was achieved.
The drainage area was adjusted to get the best match of the
later (monthly) data. The results are shown graphically in
Figure 16. The finite-simulator history-matched XF of 348 ft
is reasonably close to the semi-log derived effective (infinite
conductivity proxy) XF of 318 ft (Figure 4.) The small
difference between the two values verifies the previous
qualitative observation that fracture conductivity is high and
further establishes the validity of the RPI method, at least in
straightforward cases. Single zone completions in extensive,
moderate to high permeability, geo-pressured sands show the
best potential for achieving a long effective XF to due the
greater ease of keeping the wellbore and fracture unloaded
of liquid.
Case Study #4
The subject well is a 13,000 ft gas well completed in a 75 net
ft sandstone interval in Middle East. Initial reservoir pore
pressure was 8500 psi and initial condensate and water yields
are approximately 20 bbl/ MMscf gas each. The well was
stimulated with 38,000 gallons of a 40 lb/1000 gal guar-borate
crosslinked gel and 143,000 pounds of 16/30 mesh high
strength ceramic proppant.
Following cleanup of the
treatment, a modified isochronal test was performed over a 7
day period at flow rates varying from 11,700 mcf/d to 37,000
mcf/d gas equivalent. The results of the production analysis
are depicted in Figures 17-18. In the MDH plot, note the
presence of different trend lines that correspond to different
flow periods. The skin values computed from the separate
flow tests are plotted versus the stabilized flow rate for each
test in Figure 19. A strong rate-dependent apparent-skin effect

is observed in this relatively high flow capacity well (kh= 462

md-ft, k= 6.2 md.) If the multi-rate test had not been
performed to evaluate this non-Darcy pressure drop
component, it may have been falsely surmised that the fracture
stimulation did not produce any benefit. In fact, the zero rate
skin is 5, which equates to an effective XF of 60 ft. The
emplaced fracture serves a very important function by taking
the non-Darcy flow pressure drop away from the reservoir,
thereby eliminating or delaying the drop-out of condensate in
the near-wellbore area of the reservoir.
Conclusions
1. Systematically using multiple engineering tools refines
the interpretation of well performance.
2. Periodically updating the analysis increases the level of
confidence in the interpretation.
3. Liquid removal, or lack of it, has a profound impact on
hydraulic fracture stimulation effectiveness in low
permeability gas wells.
4. Material balance analysis shows opportunities for
in-fill drilling.
Nomenclature
bf
= fracture width, ft
c
= effective compressibility, psi-1
D
= non-Darcy flow constant, D/Mscf
DA
= drainage area, acres
h
= pay thickness, ft
k
= reservoir permeability, md
kf
= fracture permeability, md
m
= semi-log slope, m(p)/cycle
m
= semi-log slope, [m(pi) m(pwf)]/QG as abscissa
m(p)
= real gas pseudo-pressure, eq.1
p
= pressure, psi
p*
= semi-log trend extrapolated to infinite shut-in
time,
psi
QG
= gas rate, Mscf/d
R,r
= radius, ft
= true wellbore radius, ft
RW
Rwa
= apparent wellbore radius, ft
s
= skin factor, dimensionless
s
= s + DQG = apparent skin factor
t
= time, hr (unless otherwise specified)
T
= temperature, R
XE
= drainage radius, ft
XF
= fracture half-length, ft
z
= gas deviation factor

= total porosity, fraction

= viscosity, cp
Subscripts
E
= drainage
F,f
= fracture
i
= initial conditions
w
= wellbore
wa
= apparent wellbore
wf
= bottomhole wellbore flowing

SPE 84214

Acknowledgements
Thanks to the management of BJ Services Co. for supporting
this project.
References
1. Slider, H.C.: Worldwide Practical Petroleum
Reservoir
Engineering
Methods,
PennWell
Publishing Co., Tulsa, OK (1983) 11-141.
2. Dake, L.P. : Fundamentals of Reservoir Engineering,
Elsevier, Amsterdam (1978) 131-139, 243-301.
3. Lee, J.: Well Testing, SPE Textbook Series Vol. 1,
Society of Petroleum Engineers of AIME, New York/
Dallas (1982).
4. Al-Hussainy, R. et al: The Flow of Real Gases
Through Porous Media, JPT (May 1966) 624-636.
5. Gringarten, A.C. et al.: Pressure Analysis for
Fractured Wells, paper SPE 20607 presented at the
1972 SPE Annual Fall Meeting, San Antonio,
Oct 8-11.
6. Gringarten, A.C. et al.: Applied Pressure Analysis
for Fractured Wells, JPT (July 1975) 887.
7. Wattenbarger, R., Ramey, H.: Well Test
Interpretation of Vertically Fractured Gas Wells,
JPT (May 1969) 625-632.
8. Dake, L.P. : The Practice of Reservoir Engineering,
Elsevier, Amsterdam (2001) 290-307.
9. Bourdet, D. et al: Use of Pressure Derivative in
Well-Test Interpretation, SPEFE (June 1989)
293-302.
10. Cinco-Ley, H., Samaniego-V., F.: Transient
Pressure Analysis for Fractured Wells, JPT (Sept.
1981) 1749-1766.
11. Cinco-Ley, H.: Evaluation of Hydraulic Fracturing
by Transient Pressure Analysis Methods, paper SPE
10043 presented at the 1982 SPE International
Petroleum Exhibition and Technical Symposium,
Bejing, China, March 18-26.
12. Agarwal, R. et al: Evaluation and Performance
Prediction of Low-Permeability Gas Wells
Stimulated by Massive Hydraulic Fracturing, JPT
(March 1979) 362-372.
13. Rietman, ND: An Integrated Method for Optimizing
Hydraulic Fracture Design for Tight Gas Wells,
paper SPE 39930 presented at the 1998 SPE Rocky
Mountain Regional/ Low-Permeability Symposium
and Exhibition, Denver, April 5-8.
14. Crafton, JW: The Reciprocal Productivity Index
Method, A Graphical Well Performance Simulator,
Southwest Petroleum Short Course-96, 314-324.
15. Crafton, JW: Oil and Gas Well Evaluation Using the
Reciprocal Productivity Index Method, paper SPE
37409 presented at the 1997 SPE Production
Operations, Oklahoma City, March 9-11.
16. Winestock, A., Colpitts, G.: Advances in Estimating
Gas Well Deliverability, J.Cdn.Pet.Tech. (July-Sept.
1965) 111-119.
17. Huffman, C.: A Mixed Solution Approach to
Production Analysis, SPE/ Montana Tech
Technology Forum, April 24-25, 2003.

SPE 84214

18. National Research Council: Rock Fractures And

Fluid Flow: Contemporary Understanding and
Applications, National Academy Press (1996)
405-453.
19. Bidaux,P. et al: Analysis of Pressure and Rate
Transient Data From Wells in Multilayered
Reservoirs: Theory and Application, paper SPE
24679 presented at the 1992 SPE Annual Technical
Conference and Exhibition, Washington, Oct 4-7.
20. Cobb, W. et al: Well Test Analysis for Wells
Producing Commingled Zones, JPT (Jan 1972)
27-37.
21. Cramer, DD: The Rewards and Pitfalls of Using
Treating Pressure Analysis for Evaluating Fracture
Design, paper SPE 36772 presented at the SPE
Annual Technical Conference & Exhibition, Denver,
6-9 October 1996.
22. Cramer, DD: Limited Entry MHF-1: Limited entry
extended to massive hydraulic fracturing, Oil & Gas
Journal, Dec.14,1987, 40-45.
23. Cramer, DD: Limited Entry MHF- Conclusion:
Study indicates guidelines improve results, Oil &
Gas Journal, Dec.21,1987, 41-46.
24. Cramer, D., and Stallings, D.: Well Performance
and Completion Effectiveness in the Baldy Butte
Field, Washakie Basin, Wyoming, paper SPE 49046
presented at the 1998 SPE Annual Technical
Conference and Exhibition, New Orleans, Sept. 27
30.
25. Cramer, D.: The Evolution of Hydraulic Fracturing
in the Almond Formation, paper SPE 30480
presented at the 1995 SPE Annual Technical
Conference and Exhibition, Dallas, Oct. 22-25.

10

SPE 84214

Figure 1: Gringarten Plot, Infinite Conductivity Vertical Fracture

Figure 2: Gringarten analysis in Semi-Log Coordinates

Same as FCD
Figure 3: Rwa vs FCD for a Finite Conductivity Fracture

slope gives kh = 2.4 md-ft

Radial Flow Signature

intercept gives xf = 318 ft

xe/xf=5.6>5.0

tLfD = 0.0002637kt cTLf 2

Approximate
Approximate start
start of
of semi-log
semi-log line
line
== 222
222 days
days == Ln
Ln 5.4
5.4

Cartesian Plot

Average Reservoir Pressure

Rate Match

Tubing Pressure

Figure 4: RPI Mixed Solution Analysis

SPE 84214

11

Lodgepole Unit 20-29

Frontier Stage 1
Observed Net (psi)
Btm Rate (bpm)
Surf Press [Csg] (psi)

4000
50.00
7500

Lodgepole Unit 20-29

Frontier Stage 2

1st Stage
Net Pressure (psi)
Proppant Conc (ppg)

Observed Net (psi)

Btm Rate (bpm)
Surf Press [Csg] (psi)

Net Pressure (psi)

Proppant Conc (ppg)

4000
50.00

4000
50.00
9000

3200
40.00
6000

3200
40.00

3200
40.00
7200

3200
40.00

2400
30.00
4500

2400
30.00

2400
30.00
5400

2400
30.00

1600
20.00
3000

1600
20.00

1600
20.00
3600

1600
20.00

800
10.00
1500

800
10.00

800
10.00
1800

800
10.00

0
0.00

0
0.00
0

0
0.00
0

0.0

24.0

48.0

72.0

Time (mins)

96.0

120.0

0.00

Lodgepole Unit 20-29 Stage 1

Stress Profile

10.00

20.00

30.00

Time (mins)

4000
50.00

40.00

0
0.00

50.00

Lodgepole Unit 20-29 Stage 2

Stress Profile
10150

10250

Concentration of Proppant in Fracture (lb/ft)

Concentration of Proppant in Fracture (lb/ft)

10300

10200

10350

10250

10400

10300
Depth (ft)

Depth (ft)

2nd Stage

10450

10350

10500

10400

10550

10450

10600

10500

10550

10650

Permeability

10700
Low

Proppant Concentration (lb/ft)

High

0.00

0.44

0.88

1.3

1.8

2.2

2.6

3.1

Permeability

10600
3.5

4.0

Low

4.4

Proppant Concentration (lb/ft)

High

0.00

0.40

0.80

1.2

1.6

2.0

2.4

2.8

3.2

3.6

4.0

10650

10750
6000

7000

8000

9000

10000

Closure Stress (psi)

100

200

300

400

500

Length (ft)

7000

8250

9500

10750

12000

Closure Stress (psi)

100

200

300

400

500

Length (ft)

Figure 5: Fracture Simulation Results, Case Study 1

kh=0.40 md-ft
k= 0.019 md
xf= 90 ft (s=-4.6)

Trendline moved up to recent

production history. Effective
fracture half-length reduced to 24 ft.
Compare to pbu xf=19 ft.
Skin= -3.28.
Compare to
pbu skin = -3.37.

Production decline is not related to depletion

(stem too sharp, then starts to return to
infinite-acting curve.) Liquid-loading or other
flow restriction is the likely cause of the
production short-fall.

Good later match until this point

Figure 6: Analysis Based on Early Performance, Case 1

Figure 7: Analysis Based on Late Performance, Case Study 1

12

SPE 84214

For skin=-3.37, the effective

fracture half-length is:
Uniform flux=26 ft
Infinite conductivity=19 ft
P*=4600 psi used for the
duration of the analysis

kh= 0.335 md-ft

k= 0.016 md
skin(s)= -3.371

kh= 0.336 md-ft

k= 0.016 md
skin(s)= -3.35
p*= 4600 psi

Figure 8: Pressure Buildup Analysis, Case Study 1

Figure 9: Semi-log Analysis of Pressure Buildup, Case Study 1

Alternate analysis: flatter slope

Production decline is not related to depletion
(stem too sharp, then starts to return to
infinite-acting curve.) Liquid-loading or other
flow restriction is the likely cause of the
production short-fall.

kh=0.64 md-ft
k= 0.031 md
xf= 48 ft (s=-3.97)

Good match until this point

Figure 10: Alternate Production Analysis, Early Performance, Case Study 1

Alternate analysis: flatter slope

Trendline moved up to recent

production history. Effective
fracture half-length reduced to 5 ft.
Compare to pbu xf=19 ft.
Skin= -1.73.
Compare to
pbu skin = -3.37.

Figure 11: Alternate Production Analysis, Late Performance, Case 1

SPE 84214

13
250

Company A

200

XF (ft)

150

Company C

100
50

Company B
Company A

0
0%

Company B

20%

40%

Company C
60%

80%

100%

Cumulative Distribution (%)

Figure 12: A Comparison of Companies, Case 2

Radial flow model

Liquid Slugging

kh =0.69 md-ft
X F=234 ft

kh=1.18 md-ft
XF=263 ft

Figure 13: Revised Analysis, Baldy Butte 3, Case Study 2

Figure 14: Baldy Butte 7, In-Fill Well, Case Study 2

10%

Figure 15: Well Log, Case Study 3

14

SPE 84214

Calc'd Toph Press (psi)

Cum HC Prod (MMscf)
HC Rate (Mscf/d)

5000
1500
5000

Surf Press [Tbg] (psi)

Cum Gas (MMscf)
Production Gas Rate (Mscf/d)

k = 0.117 md (from RPI match)

xf = 348 ft
Drainage Area = 207 acres

4000
1200
4000

Simulated
Cumulative
Production

5000
1500
5000

4000
1200
4000

Actual Cumulative Production

3000
900
3000

3000
900
3000

Simulated Rate
2000
600
2000

2000
600
2000

Calculated BH
Flowing Pressure
Actual Rate

1000
300
1000

0
0
0

1000
300
1000

Simulated BH Flowing Pressure

0.0

93.0

186.0

Time (days)

279.0

372.0

465.0

0
0
0

Almond

Figure 16: History Match with Finite Difference Production Simulator

Figure 17: Modified Isochronal Test, Middle East well, Case 4

SPE 84214

15

Figure 18: MDH Plots, Case 4

9
R2 = 0.9902

RPI-Derived Skin

7
5
3
1
-1
-3

Skin value at 0-intercept = -4.5 (inertial/ non-Darcy effects negated)

Equivalent effective fracture half-length = 60 ft

-5
-7
0

5000

10000

15000

20000

25000

30000

35000

Equivalent Gas Rate (mcsf/d)

Figure 19: Variation of Apparent Skin with Flow Rate, Case 4

40000