Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

SPE-195085-MS

A New Decline Curve Analysis Method for Layered Reservoirs

Kittiphong Jongkittinarukorn, Chulalongkorn University; Nick Last, Well Test Knowledge International; Freddy
Escobar, Universidad Surcolombiana; Kreangkrai Maneeintr, Chulalongkorn University

Copyright 2019, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Middle East Oil and Gas Show and Conference held in Manama, Bahrain, 18-21 March 2019.

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 have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper 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 SPE copyright.

Abstract
This study presents a new method to improve production forecasts and reserves estimation for a multilayer
well in the early stages of production, using Arps’ hyperbolic decline method to model the decline rate of
each layer. The method can be applied to both oil and gas wells.
The new approach generates the profiles of instantaneous decline rate (D) and instantaneous decline
exponent (b) from the historical flow rate (q). Because of the inherent noise in the production data, a
regression technique is applied to smooth the flow rate data and the analysis is performed on the smoothed
data. History matching is performed not only on the profile of q, but also on the profiles of D and b. The
results are the unique decline parameters (qi, Di, and b) for each layer.
For a multi-layer well, the values of D and b vary with time, which means that its performance cannot
be modeled using a conventional single-layer well approach. Furthermore, the well-known non-uniqueness
problem from history matching is magnified in a multilayer well: many models can successfully match the
production profile in the short term but fail to match it in the longer term. Only the correct model can match
the profiles of q, D, and b over both the short-term and the long-term. The proposed method provides the
correct unique decline parameters (qi, Di, and b) for each layer, during the early stage of production, and
these parameters are then valid for the life of the well. The method works well for both synthetic examples
and actual field data.
The novelty of the new methodology is the ability to provide the decline parameters for each layer at
an early stage of production, which can then be used for production forecasting in the long term. The non-
uniqueness problem from history matching is solved.

Introduction
Decline Curve Analysis (DCA) is a commonly-applied Petroleum Engineering method used for forecasting
production and for estimating reserves from production data under boundary-dominated flow. For a
volumetric, single-phase reservoir produced at a constant bottom-hole pressure, the flow rate can be
empirically modeled by the following function (Arps, 1945):
2 SPE-195085-MS

(1)

where
 qi : initial flow rate, MMscf/d or stb/d
 Di : initial decline rate, d-1
 b : decline exponent, dimensionless
The instantaneous decline rate (Johnson and Bollens (1927), Arps (1945), and Spivey et al. (2001)) is
defined as

(2)

For Arps’ hyperbolic rate decline, the instantaneous decline rate (Yu (2001), Blasingame and Rushing
(2005)) is

(3)

Exponential decline (b = 0) has a constant decline rate of Di while hyperbolic and harmonic declines have
decreasing decline rates. The instantaneous decline exponent is defined as

(4)

Note that the value of b is constant for Arps' hyperbolic rate decline. High values of b can be observed
during the infinite-acting or transient production period (Fetkovich et al. 1990 and Cheng et al. 2008), and
this part of the well's history should not be used for long-term performance predictions. During boundary-
dominated flow, a single-layer well has a value of b < 0.5 (Fetkovich et al. (1996), Okuszko et al. (2007)).
Higher values of the decline exponent (b > 0.5) can be observed for a multilayer well without crossflow
(Fetkovich et al. 1996). This is due to an early rapid rate decline of the higher-permeability layer followed
by an extended period of low rate decline of the lower-permeability layer. Normally, a multilayer well has
an unusually long production life.
Arps (1945) developed the hyperbolic rate decline empirically. However, the values of D and b could
be related to fluid properties and production conditions (Fetkovich (1980), Fetkovich et al. (1996), Chen
and Teufel (2002), Ye and Ayala (2013), Ayala and Ye (2013), Stumpf and Ayala (2016)). Fetkovich et. al.
(1996) showed that the initial decline rate can be expressed in reservoir engineering terms as

(5)

where UR is the ultimate recoverable oil or gas at q(t) = 0. Note that Di is not defined for harmonic decline
(b=1). For a well producing at very low flowing pressure (pwf), the decline exponent can be expressed as

(6)
SPE-195085-MS 3

where n is the backpressure curve exponent. Chen and Teufel (2002) proposed a model to estimate an
averaged b during boundary-dominated gas flow using pseudopressure (Lee et al. 2003).

(7)

The normalized pseudopressure (pp,n(p)) is defined as

(8)

Fetkovich (1980) used the departure curve method (differencing) to analyze decline curve data from
the East Side Coalinga field, which consists of upper and lower oil sands separated by shale. During late
time, the production was dominated by layer 1. History matching during this period yielded the decline
parameters (qi1 and N1) for layer 1. The production rate for layer 1 was extrapolated backwards in time,
and the difference between the actual rates and the rates from backwards extrapolation of layer 1 was the
production from layer 2. History matching the resulting layer 2 data yielded the decline parameter (qi2 and
N2) for layer 2. The results were consistent with the geologic description. However, the method is useful
only when the well is at the late stage of production, when the production is mainly from one layer. The
method cannot be applied early in the well's life.
Cheng et al. (2008) developed a method to improve reserves estimates from decline-curve analysis of tight
and multilayer gas wells: the b-value is estimated during boundary-dominated flow. and history matching
is performed using this b value in a backward data-fitting fashion to determine the decline parameters
(qi and Di) at the end of production. These parameters are then used for production forecasts. However,
Cheng's proposed method cannot be used to estimate decline parameters for each layer: it assumes that the
performances of a multilayer well can be represented by a single-layer well model
The production performance for a multi-layer gas well is complicated since the production contribution
from each layer changes constantly with time (Fetkovich (1980), Cheng et al. (2008)). Several previous
studies (Arps 1945, Fetkovich et al. (1996), Cheng et al. (2008)) showed that DCA for multilayer gas wells
could yield an Arps decline-curve exponent (b) greater than 1.0. While b is constant for a single-layer
well, b is changing with time for a multilayer well. The non-uniqueness problem from history matching is
more prominent in a multi-layer system. The available methods could not provide a unique set of decline
parameters (qi, Di, and b) for individual layers.
The objective of this study is to propose a new method to provide valid production forecasts and reserves
estimation for multilayer wells (both oil and gas) based on early-stage production. The method's innovation
is to estimate the instantaneous values of decline rate (D) and decline exponent (b) from the production
data during boundary-dominated flow, and to perform history matching on the profiles of q, D, and b to
determine a set of decline parameters (qi, Di, and b) for each layer.
Both synthetic examples and field data are used to validate the proposed method. The non-uniqueness
problem from history matching is solved and the best decline models for individual layers are identified.
Consequently, better production forecasts and more accurate reserves estimation can be achieved for both
the short- and the long-term.

A New Methodology
In order to best illustrate the principles of the method, this study considers a 2-layer well. Layer 1 has a
higher productivity than Layer 2. The main assumptions are:
4 SPE-195085-MS

• Constant flowing wellbore pressure

• Single-phase flow

• No aquifer support

• No reservoir crossflow

• No wellbore crossflow

• No changes in completion and operating conditions.

For commingled production at a constant wellbore pressure, production from any one layer is not
dependent on the production from other layers. Therefore, the well flow rate is the sum of the flow rates
from each layer (Russell and Prats, 1962). The flow rate of the well is:
(9)
At t = 0, the initial flow rate of a well is:
(10)
Following Fetkovich (1980), the decline-curve dimensionless rate of a well is defined as:

(11)

The decline-curve dimensionless rate is a weighted average of the decline-curve dimensionless rates of
layers 1 and 2. The weights of layers 1 and 2 are the initial production contribution of layers 1 and 2,
respectively. Note that the weights are constant.
Then the instantaneous decline rate of the well is

(12)

The instantaneous decline rate of the well is a weighted average of the decline rates of layers 1 and 2.
The weights of layers 1 and 2 are the instantaneous production contributions of layers 1 and 2, respectively.
These weights are not constant.
The instantaneous decline exponent of the well is

(13)

While Eq. (12) is similar to the result from Spivey et. al (2001), Eq.(13) is different. However, the result
from Eq.(13) is perfectly matched with their Fig. 2. Fetkovich et al. (1996) showed that the average value
of b(t) is not less than the values of b1 and b2.
To understand Eqs. (11)-(13), let's consider a 2-layer gas well with the following properties for layers
1 and 2, respectively.

• Initial flow rate (qi) = 5.0 and 2.0 MMscf/d

• Initial decline rate (D) = 0.010 and 0.001 d-1

• Decline exponent (b) = 0.01 and 0.50


SPE-195085-MS 5

During the boundary-dominated flow period, the production profile and the production forecasts based
on the production data of 100 days, 1, 2, and 4 years are shown in Figure 1. The first forecasting method is
from Cheng et al. (2008), labeled as "Historical b". The two key components in their method are:

• The b-value is the averaged value during the production period.

• Other decline parameters (qi and Di) are the values at the end of production period.

Figure 1—Flow rate (q) for a 2-layer gas well

This set of decline parameters is used for the production forecast. Note that some values of b are greater
than 1.0. The forecasts can match the production profile only in the short term and fail to do so in the longer
term. If the value of b is restricted to be less than or equal to 1.0 (b ≤ 1), the production forecast is improved
significantly. Fetkovich et al. (1990) recommended the b value of 0.6 (b = 0.6), which further improves the
forecast. The errors in reserve estimations at the end of each production period are also shown in Figure
1. With complex production, the error percentage does not decrease monotonically with time. The reserves
are underestimated during the early stage of production and tend to be overestimated during the late stage
of production. Yu (2011) pointed out the slope change (flattening and then increasing again) in the log-log
plot. This is due to a large permeability contrast between layers. With a long production life, the layer with
lower permeability dominates the well production in the long term. It is apparent that we cannot predict the
performance of a multilayer gas well using a single-layer well approach.
The profiles of flow rate (q), decline-curve dimensionless rate (q/qi), decline rate (D), and decline
exponent (b) are illustrated in Figure 2. We can make the following observations:
6 SPE-195085-MS

• Production contribution from each layer varies with time. The higher-permeability layer has a
relatively high early-time production rate with a steeper decline. The lower-permeability layer
declines less rapidly and dominates the later-time production. The production contribution from
layer 2 is increasing with time. This phenomenon was clearly explained by the previous studies (Yu
(2011), Fetkovich et al. (1990), Cheng et al. (2008)). For a system with large permeability contrast,
the early rate decline is large. Then the production is dominated by layer 2 with a much flatter rate
decline (Yu (2011)). This characteristic can be identified by the log-log plot of flow rate vs time.
• As described in Eq. (11), the decline-curve dimensionless rate (q/qi) of a well is a weighted average
of the decline-curve dimensionless rate of layers 1 and 2. The weights are constant, not a function of
time. As a result, on the plot of decline-curve dimensionless rate (q/qi) vs t, the respective distances
between well and each layer are always equal for any point in time, regardless of the instantaneous
production contribution from each layer.
• The instantaneous decline rate of a well (D) is a weighted average of the decline rate of layers 1 and
2 (D1 and D2). The weights are not constant. The value of D(t) is always between the values of D1(t)
and D2(t). As the contribution of layer 1 decreases, the value of D(t) approaches the value for D2(t).
• Eq. (13) explains the instantaneous decline exponent of a well (b). The value of b is increasing and
then decreasing. The maximum value of b can be greater than 1.0. As the contribution of layer 1
diminishes, the value of b(t) approaches the value for b2(t). (Fetkovich et al. (1990))
• This example clearly shows that the values of decline rate (D) and decline exponent (b) are not
constant but vary with time.

Figure 2—Flow rate (q), decline-curve dimensionless rate (q/qi), decline rate (D), and decline exponent (b)
SPE-195085-MS 7

The methodology of this study improves production forecasts and reserves estimation from DCA for a
two-layer gas well. The available production data are:

• Flow rate (q(t))

• Cumulative production (Gp(t)).

The unknown decline parameters for layers 1 & 2 are:

• Initial flow rates (qi1, qi2)

• Initial decline rates (Di1, Di2)

• Decline exponents (b1, b2)

To find models for layers 1 & 2, history matching is performed to match q(t), D(t), b(t), and Gp(t) of
the well.
(9)
(14)
(15)

(16)

where

(17)

(18)

(19)

(20)

(21)

(22)

Additional history matching on decline rate D(t) and decline exponent b(t) provides the decline
parameters for each layer. Therefore, the new method will solve the non-uniqueness problem. The model
will yield good production forecasts and reserves estimation at the early stage of production.
8 SPE-195085-MS

Applications
Example 1: Non-uniqueness problem from history matching
The non-uniqueness problem from history matching has been well documented in the literature. It becomes
highly problematic when we try to match performance from a multi-layer well with a single-layer system.
The objective of this example is to show how the proposed method can be applied to solve the non-
uniqueness problem with a limited amount of production data. Let's consider a two-layer gas well with the
properties listed under columns "well" Table 1. This model has the following characteristics:

• Layer 1 has higher initial production rate than layer 2 (qi1 > qi2)

• Layer 1 has higher initial decline rate than layer 2 (Di1 > Di2)

• Layer 1 has lower decline exponent than layer 2 (b1 < b2)

Table 1—Decline parameters (qi, Di, and b) from history matching for Example 1

Based on these parameters, layer 1 will be depleted faster than layer 2. In this case, we assume that
we have only 400 days of production data. Four different cases of history matching are considered in this
example:

• Case I: The initial flow was incorrect and layer 2 has the higher initial production rate.

• Case II: The initial decline rate for layer 1 is 12% overestimated.

• Case III: The initial decline rate for layer 1 is 12% underestimated.

• Case IV: The decline exponent was incorrect, and layer 1 has a higher decline exponent.

The parameters (qi, Di, and b) and the errors in reserves estimation from history matching are reported
in Table 1. It is apparent that the value of Di has a significant effect on the estimated reserves. The results of
history matching are shown in Figures 3-8. Figure 3 plots profiles of q for the different cases. It is apparent
that these four models can all match the 400-days of production data pretty well. This is the well-known
non-uniqueness problem from history matching. But the four models fail to match well performance in the
long term. Plots of q vs Gp for the different cases are illustrated in Figure 4. These four models predict
significantly different ultimate recovery from the actual performance. History matching on q and Gp leads
to the non-uniqueness problem. The profiles of decline-curve dimensionless rate (q/qi) for different cases
are illustrated in Figure 5. The models for all cases can match (q/qi) of a well but fail to match the decline-
curve dimensionless rate of an individual layer. Note that in the actual DCA we cannot see the actual profile
of the decline-curve dimensionless rate of individual layers.
SPE-195085-MS 9

Figure 3—Flow rate (q) for Example 1

Figure 4—Flow rate (q) vs Cumulative gas production (Gp) for Example 1
10 SPE-195085-MS

Figure 5—Decline-curve dimensionless rate (q/qi) for Example 1

Figure 6—Decline rate (D) for Example 1


SPE-195085-MS 11

Figure 7—Decline exponent (b) for Example 1

Figure 8—Production contribution from layer 2 (q2/q) for Example 1


12 SPE-195085-MS

The profiles of D, b, and (q2/q) are illustrated in Figures 6-8, respectively. Note that the profile of (q2/
q) can be obtained when production-logging data is available. Different models yield different profiles of
these parameters. Therefore, we could use this set of parameters to solve the non-uniqueness problem and
to identify the best model for production forecasting and for reserves estimation.
The results of this example illustate that history matching on q, Gp, D, and b simultaneously yields a
unique model for a two-layer gas well. It can match not only the short-term production data, but also the
long-term production data. The profiles of D and b have a greater ability to identify the valid model than
the profile of q does. However, the effect of noise inherent in production data is more pronounced in the
profiles of D and b.

Example 2: An oil well from Coalinga oil field


The data in this example is from the East Side Coalinga field, California, United States. (Ref. 18, 27, 29,)
There are isolated (no crossflow) upper and lower oil sands. The departure curve method (differencing)
is applied to analyze the production data (Fetkovich 1980). Note that the departure curve method requires
the last portion of the production data to differentiate the performances of upper and lower layers. This is
the period during when only one layer significantly contributes to the production. Therefore, it cannot be
applied to analyze production data during the early of well life when both layers significantly contribute to
the production. The result is the following oil rate.

(23)

where
 q(t) : the total production rate from both layers, bbl/y
 t : time, years
Fetkovich (1980) mentioned that a single-layer model with b-value of 0.2 can match nearly all of the
data points but can't be explained by any of the drive mechanisms in this field. We tried history matching
with a single-layer with the following result.

(24)

The proposed method is also applied to analyze the production from this field. It is apparent that the
production data is very noisy. Therefore, we perform regression on q and t, yielding a profile for the oil
rate. This profile is then used to establish the profiles for D and b. History matching on q, D, and b yields
the following model:

(25)

The profiles of q, D, b and the production contribution from layer 2 (q2/q) are illustrated in Figure 9. Note
that the profile of b is slightly distorted by the polynomial function used in the regression. It is clearly seen
that the model in this study can fit the profiles of q, D, and b better than the ones from the departure curve
method and from a single-layer model. For the model from the departure curve method, layer 2 has lower
initial contribution (qi2 < qi1) but was depleted faster (Di2 > Di1). Layer 1 with higher productivity dominated
the production during the late production period. However, the model in this study indicates that layer 2
with lower productivity should be depleted more slowly (Di2 < Di1) and should dominate the production
during the late period.
SPE-195085-MS 13

Figure 9—Flow rate (q), decline rate (D), decline exponent (b), and production contribution from layer 2 (q2/q) for Example 2

Discussion
For a multi-layer well
Fetkovich (1980) recommended that two or more layers can be perfectly represented by a single only when
these layers share the same values of a decline-curve dimensionless rate (qDd).

(26)

In other words, these layers will be depleted at the same rate when they have the same values of Di and b.
The decline-curve dimensionless rate is more sensitive to the value of Di than to the value of b. For practical
purposes, we may concentrate on Di and ignore the effect of b. Fetkovich et al. (1996) recommended
reducing a multi-layer system to a 2-layer system by combining layers with similar value of (qi/G). Many
studies recommend not having more than 2 layers for history matching since it will complicate the matching
process and magnify the non-uniqueness problem.

Noise on production data


The proposed method uses the parameters D and b which involve the derivative of production data. The
effect of inherent noise in production data is magnified by taking derivative (Ilk et al. (2011) and Varma et
14 SPE-195085-MS

al. (2018)). Therefore, data smoothing and/or modern differentiation algorithms are required to improve the
derivative calculations (Kupchenko et al. (2008), Varma et al. (2018)). The rate-cumulative profile is always
smoother than the rate-time profile. Ilk et al. (2008) recommended the following alternative to compute the
values of D and b for noisy production data.

(27)

(28)

Only data in the hyperbolic window can be used for history matching
The decline exponent (b) in Arps’ hyperbolic decline is assumed to be a constant throughout the well
life. Stumpf and Ayala (2016) showed that the decline exponent (b) is approximately constant only in
the hyperbolic window. This window occurs during the early boundary-dominated flow (BDF) period.
Furthermore, the value of b approaches zero during the late BDF period. Therefore, only the data in the
hyperbolic window can be used for history matching. Prediction outside this window yields invalid matching
results because the requirement that b is approximately constant is not honored.

Production logging (PL)


Production logging is a cased-hole logging technique that can be used to allocate production on a layer-by-
layer basis. As well as defining zonal flow contributions at a given point in time, PL can define the inflow
performance relationship and average drainage area pressure separately for each zone based on the zones’
individual flow contributions at different total-well flow rates. PL surveys can be repeated multiple times
and can be used to define the variation of layer flow contributions with time. If the magnitude of inherent
noise in the production data is too large and the profiles of D and b cannot be established, then PL can be
applied to solve the non-uniqueness problem.

Conclusions
1. Both decline rate (D) and decline exponent (b) of a multi-layer system are not constant, but vary
with time. The production profile of this system cannot be modeled by a 1-layer system (with a
constant value of b), especially when there is a significant contrast in permeability. Consequently, the
conventional DCA method yields unreliable production forecasts and reserves estimation.
2. The non-uniqueness problem of history matching for a multi-layer system is illustrated. Many
different 1-layer models can match production from a multi-layer system, especially when a well is at
an early stage of production. The model with correct decline parameters (qi, Di, and b) in each layer
could not be identified from the methods available in the literature.
3. This study proposes a new method to improve production forecasts and reserves estimation and to
solve the non-uniqueness problem for a well in multi-layer system. With this new methodology,
history matching is performed not only on the flow rate (q), but also on decline rate (D) and decline
exponent (b). This method yields the set of decline parameters (qi, Di, and b) of each individual layer
at the early stage of production.
4. The proposed method is successfully validated using several synthetic examples and an actual field
data set.
5. Improvements in production forecasting and reserves estimation, especially during the early stage of
production, can be achieved by applying the proposed method.
SPE-195085-MS 15

Nomenclature
b : decline exponent, dimensionless
bE : equivalent or averaged decline exponent, dimensionless
cg : isothermal gas compressibility, psi-1
D : decline rate, day-1
G : ultimate gas recovery, MMscf
Gp : cumulative gas production, MMscf
n : backpressure curve exponent, dimensionless
N : ultimate oil recovery, stb
p : pressure, psia
pp,n : normalized pseudopressure, psia
q : flow rate, MMscf/d or bbl/year
qDd : decline-curve dimensionless rate, dimensionless
t : time, days
UR : ultimate recoverable oil or gas, stb or MMscf
z : gas compressibility factor, dimensionless
μg : gas viscosity, cp

Subscripts
1 : layer 1
2 : layer 2
i : initial condition
wf : well flowing

References
Arps, J.J.: "Analysis of Decline Curves," Trans., AIME (1945) 160, 228–247.
Ayala, L.F. and Ye, P.: "Density-Based Decline Performance Analysis of Natural Gas Reservoirs Using a Universal Type-
Curve," J. Energy Res. Tech. 135, 2013
Blasingame, T.A. and Rushing, J.A.: "A Production-Based Method for Direct Estimation of Gas-in-Place and Reserves,"
paper SPE 98042 presented at the 2005 SPE Eastern Regional Meeting, Morgantown, W.V. 14-16 September 2005
Chen, H-Y. and Teufel, L.W.: "Estimating Gas Decline-Exponent Before Decline-Curve Analysis," Paper SPE 75693
presented at the SPE Gas Technology Symposium, Calgary, 30 April-2 May 2002
Cheng, Y., Lee, W.J., and McVay, D.A.: "Improving Reserves Estimates From Decline-Curve Analysis of Tight and
Multilayer Gas Wells," SPE Reservoir Evaluation & Engineering, October 2008, 912–920
Fetkovich, M.J.: "Decline Curve Analysis Using Type Curves," J. Petrol. Tech. (June 1980) 1065–1077; Trans. AIME,
269.
Fetkovich, M.J., Bradley, M.D., Works, A.M., and Thrasher, T.S.: "Depletion Performance of Layered Reservoirs Without
Crossflow," SPEFE 5 (3): 310–318; Trans., AIME, 289, 1990.
Fetkovich, M.J., Fetkovich, E.J., Fetkovich, M.D.: "Useful Concepts for Decline-Curve Forecasting, Reserve Estimation,
and Analysis," SPE Reservoir Engn. (Feb. 1996) 13–22.
Ilk, D., Rushing, J.A., Perego, A.D., and Blasingame, T.A.: "Exponential vs. Hyperbolic Decline in Tight Gas Sands-
Understanding the Origin and Implications for Reserve Estimates Using Arps’ Decline Curve," paper SPE 116731
presented at the Annual Technical Conference and Exhibition, Denver, USA, 21-24 September 2008
Ilk, D., Jenkins, C.D., and Blasingame, T.A.: "Production Analysis in Unconventional Reservoirs-Diagnostics, Challenges,
and Methodologies," paper SPE 144376, 2011
Johnson, R.H. and Bollens, A.L. 1927. The Loss Ratio Method of Extrapolating Oil Well Decline Curves. Trans. AIME
77: 771.
Kupchenko, C.L., Gault, B.W., and Mattar, L.: "Tight Gas Production Performance Using Decline Curve," paper SPE
114991 presented at the CIPC/SPE Gas Technology Symposium, Calgary, Alberta, Canada 16-19 Jun 2008
Lee, W.J., Rollins, J.B., and Spivey, J.P.: Pressure Transient Testing Textbook Series, SPE, Richardson, Texas, 2003.
16 SPE-195085-MS

Okuszko, K.E., Gault, B.W., and Mattar, L.: "Production Decline Performance of CBM Wells" paper CIPC 2007-078
presented at Canadian International Petroleum Conference, Calgary, Alberta, Canada, 12-14 June 2007
Russell, D.G. and Prats, M.: "Performance of Layered Reservoirs With Crossflow—Single-Compressible Fluid Case."
Soc. Pet. Eng. J. (March 1962) 53–67
Spivey, J.P., Frantz, J.H., Williamson, J.R., and Sawyer, W.K.: "Applications of the Transient Hyperbolic Exponent," paper
SPE 71038 presented at the SPE Rocky Mountain Petroleum Technology Conference, Keystone, Colorado, USA,
21-23 May 2001
Stumpf, T.N. and Ayala, L.F.: "Rigorous and Explicit Determination of Reserves and Hyperbolic Exponents in Gas-Well
Decline Analysis," SPE Journal, October 2016, 1843–1857.
Varma, S., Tabatabaie, S.H., Ewert, J., Mattar, L., and Markit I.: "Variation of Hyperbolic-b-parameter for Unconventional
Reservoirs, and 3-Segment Hyperbolic Decline," paper URTeC 2892966, presented at the Unconventional Resources
Technology Conference, Houston, Texas, USA, 23-25 July 2018
Ye, P. and Ayala, L.F.: "Straigit-Line Analysis of Flow Rate vs Cumulative Production Data for the Explicit Determination
of Gas Reserves," J. Can Pet Technol 54 (4): 296–305, 2013
Yu, S.: "An Improved Methodology to Obtain the Arps' Decline Curve Exponent (b) for Tight/Stacked Gas Reservoirs,"
paper SPE 143907 presented at the SPE North American Unconventional Gas Conference and Exhibition, Woodlands,
Texas, USA, 14-16 June 2001

You might also like