You are on page 1of 9

SPE 108176

Improving Reserves Estimates From Decline Curve Analysis of Tight and Multilayer
Gas Wells
Y. Cheng, SPE, W.J. Lee, SPE, and D.A. McVay, SPE, Texas A&M U.

Copyright 2007, Society of Petroleum Engineers


regimes between layers lead to abnormal decline behavior
This paper was prepared for presentation at the 2007 SPE Hydrocarbon Economics and (b>1.0) in multilayer tight gas wells, which leads to errors in
Evaluation Symposium held in Dallas, Texas, U.S.A., 1–3 April 2007.
production forecasts. Our new procedure provides a method to
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
minimize these errors.
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 Introduction
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 Decline curve analysis is one of the most commonly used
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 techniques to predict future production performance and
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.
estimate reserves from routinely available production data.
Box 833836, Richardson, Texas 75083-3836 U.S.A., fax 01-972-952-9435. Although Arps’ decline equations were developed
empirically,1 the parameter b in the decline equations was
Abstract proved to be related to fluid properties and production
Decline curve analysis is one of the most commonly used conditions.2-3 Conventional decline curve analysis inherently
techniques to estimate reserves from production data. In tight assumes a single-layer reservoir, the well producing at
formations that have been stimulated, especially when there constant bottomhole pressure, and stabilized flow conditions.3
are multiple layers that communicate only at the wellbore, the In addition, use of Arps’ equations implies that there are no
uncertainty in reserves estimates from this technique is quite changes in completion and operating conditions. It is well
large because forecasting future performance is quite difficult. documented that the decline exponent should range between
This uncertainty can affect the classification of reserves, and zero and 1.0 when these assumptions are satisfied.1,3
could limit what we should call “proved developed reserves.” Tight gas reservoirs are characterized by permeabilities
In this paper, we present new procedures to mitigate the less than 0.1 md. Gas wells in tight formations usually require
complexity of decline curve analysis in multilayer tight gas hydraulic fracturing of multiple layers to be commercially
wells. Using synthetic and field examples, we demonstrate viable. Therefore, analysis of decline behavior in tight gas
how reserves can be estimated more reliably. wells presents unique technical challenges.4-5 It is often very
For tight and multilayer gas wells, it is not uncommon that difficult, if not impossible, to estimate reserves accurately in a
decline curve analysis yields an Arps decline-curve parameter timely and consistent fashion with decline curve analysis.
‘b’ larger than 1. Single-layer hydraulically fractured tight gas Long times, often years, are required to reach so called
wells also appear to have b values greater than one. Different pseudo-steady state flow (actually, boundary-dominated flow,
practices are used to handle such complexity. For forecasts, since the term “pseudo-steady state” strictly applies only to
some analysts simply use the ‘b’ value obtained from constant-rate production). The production data available for
matching production data, while others force the ‘b’ value to decline curve analysis are, therefore, typically not stabilized.
be 1. Still others use the hyperbolic decline and the matched As a result, it is not uncommon for tight gas wells to exhibit
value of ‘b,’ but, when the decline rate reaches a Arps’ decline constants, b, that exceed 1.0.6 With b-values
predetermined limit, they switch to exponential decline for the greater than 1.0, future performance and remaining reserves
remainder of the forecast. Thus, forecasted performance will be greatly overestimated. In conventional practice, some
differs significantly when different analysts analyze the data. analysts simply use the b value obtained from matching of
Consequently, the reserves estimate has large uncertainty, production data, while others force the b value to be 1.0. Still
which can, in turn, affect its classification. others use the hyperbolic decline and the matched value of b,
In this paper, we first present an in-depth investigation of but, when the decline rate reaches a predetermined limit, they
decline behavior of tight, single-layer and multilayer gas wells switch to exponential decline for the remainder of the forecast.
by analyzing depletion characteristics using simulated data However, this procedure has no physical basis. This type of
sets. We illustrate the long-duration transient effects present in decline behavior is highly unlikely in nature.
single-layer stimulated tight gas wells and the complex flow
Another complication in analysis of decline data in tight
regimes present when wells in layered reservoirs are produced
gas wells is that, in most cases, production is commingled
commingled. Our work indicates that, as observed in field
from multilayered formations that are hydraulically fractured
data, transient effects and coexistence of different flow
with multiple stages. Due to variations in formation
2 SPE 108176

permeability and fracture half length, different flow regimes simulation results, the time at which production response
may co-exist in different layers. Lower permeability zones “feels” the boundary effects is much earlier than the
may be in transient flow while higher permeability zones have stabilization time. After 4 months of production, the pressure
established stabilized, boundary-dominated flow. The profile at the nearest boundaries (from the hydraulic fracture tips)
of production contribution by individual layers constantly dropped 10 psia. Therefore, although the flow regime is
changes with time. Given these complications, decline curve transient before stabilized flow is reached, boundaries begin to
analysis of multilayer tight gas wells is especially difficult, affect production data prior to complete stabilization. Thus,
particularly with regard to estimating long-term production the possibility exists of predicting long-term behavior with
and reserves. unstabilized data.
The objective of this study is to develop methodology to
improve reserve estimates from decline curve analysis of tight Synthetic Case 2. In this case, a gas well produces from
and multilayer gas wells. We first generated simulated data three layers with three hydraulic fracture treatment stages.
sets representing two synthetic cases – a single-layer gas well These three layers communicate only at the wellbore. As in
and a multilayer gas well. We then analyzed the characteristics Case 1, the well is centered in a square reservoir with a
of decline equation parameters resulting from fitting the drainage area of 1867*1867 ft2. The permeabilities of the three
simulated rate data. Since the true reserves for the simulated productive formations are 0.02 md, 0.006 md and 0.002 md.
cases are known, we determined the errors in projecting future The adjacent productive layers are separated by shale zones 50
performance and estimating reserves using current analysis ft thick. Each productive formation has 50 ft of net pay and a
techniques. We show that decline curve analysis using fracture half-length of 450 ft. Other basic reservoir, fracture
stabilized data is sufficient to produce accurate reserves and fluid properties are the same as in synthetic Case 1.
estimates. However, given the fact that, in most tight wells, we The 30-year decline behavior of this gas well is shown in
have to work with transient flow data to estimate reserves Fig. 2. The production rates from each of the three layers are
from early-time production data, we propose an improved also displayed in this figure. Because of the differences in
analysis technique to effectively handle such cases. layer permeabilities, the production rates in the different layers
Improvement in reserves estimates for synthetic and field vary significantly. The higher permeability layer has a
examples are illustrated using the proposed technique. relatively high early-time production rate and a steeper
decline. Lower permeability layers decline less rapidly and
Analysis of Decline Behavior dominate the later time production. Due to permeability
We generated two synthetic cases using a commercial differences, the time required to establish boundary-dominated
reservoir simulator. These two synthetic cases illustrate the flow is very different in the different layers. For the 0.02-md
decline behavior of a fractured, single-layer well and a layer, 480 days, or 1.3 year, were required to reach stabilized
fractured multilayer well. flow while the 0.002-md layer required 4,800 days, or 13
years.
Synthetic cases
Synthetic Case 1. In this case, a gas well produces from a Dynamics of decline curves and forecast errors
hydraulically fractured single-layer reservoir. The well is Decline behavior of oil and gas wells is often described
centered in a square reservoir with a drainage area of by Arps’ decline curve equations. Arp’s equation for a
1867*1867 ft2 (80 ac). Basic reservoir, fracture and fluid hyperbolic decline is
properties are listed below:
− 1b
q = qi (1 + Di bt ) (1)
Reservoir and fracture properties
Reservoir temperature = 250 oF
Initial reservoir pressure = 5,000 psi where q is production rate at time t, qi is the initial production
rate, Di is the initial decline rate, and b is the decline exponent.
Net pay thickness = 150 ft
Gas porosity = 0.06 When b approaches 0, the decline becomes exponential; when
Gas permeability = 0.006 md b approaches 1, the decline becomes harmonic.
In tight reservoirs, determination of b has often been
Fracture half-length = 450 ft
problematic. Some analysts simply take the b value obtained
Fracture conductivity = 100 md•ft
from matching of production data (which is often greater than
Bottom-hole flowing pressure = 1,000 psi
1), while others force the b value to be 1. The value of b
significantly affects the projection of future rates.
Fluid properties
Gas gravity (air=1) = 0.65
Use of b-value obtained from best fit with transient data
Initial gas viscosity = 0.025 cp
included. We first determine the b-values and other decline
Initial gas compressibility = 1.2*10-4 psi-1
equation parameters from the best fit to simulated rate data, in
Initial gas production rate = 2,000 Mscf/d
which transient data are included, with varying fraction of the
total production history starting from time zero. The regressed
The decline performance of this gas well during a 30-year
parameters are then used to forecast the production for a total
history is shown in Fig. 1. The time to achieve stabilized,
well life of 30 years. Table 1 gives regressed values of the
boundary-dominated flow is about 1,600 days. However, from
parameters (qi, b, Di) for the single-layer case. As we can see,
SPE 108176 3

the b-values decrease as more production data are used in the analysis and transient flow or unstabilized data are
regression, but they are all larger than 1. We also notice that excluded from history matching.
Di is positively correlated with b. Fig. 3 shows a set of For the single-layer case, boundary-dominated flow is
production forecasts using the b-values and corresponding qi reached at about 1,600 days, or 4.4 years. We discarded the
and Di from best curve fits of one year, ten years and twenty data for the first five years to exclude transient data and began
years of production data. Note that the quality of the matches history matching starting at 1,825 days, or 5 years. Table 5
is good for 1-year production data, but deteriorates as longer provides the regressed values of the three decline parameters
production data (such as 10 years and 20 years) are matched. based on history matching to the end of the 6th through the
Fig. 3 clearly indicates that future production is greatly 10th year. The value of b generally decreases as we include
overestimated in all cases. The b-values greater than 1 leads to more late-time data in history matching, but all are less than
these overestimated future production values. 1.0. Di also decreases as b decreases. Future production was
For the multilayer commingled production case, the b- forecasted using the regressed parameters (as shown in Fig. 7),
values obtained from history matching are also all greater than and the relative errors in estimated remaining reserves are
1, as shown in Table 2. Again, Di is positively correlated with presented in Table 5. The range of relative errors is from 1%
b. Fig. 4 is a set of production forecasts with these b-values to 4%.
and other regressed parameters. Future projections are For the multilayer production case, we also used only data
overestimated for all scenarios investigated. after the fifth year for decline curve analysis. Recall that the
As a conclusive observation, for both single-layer and two layers with higher permeabilities have reached boundary-
multiple layer cases, decline curve fitting with transient flow dominated flow while the bottom layer with the smallest
data included leads to b-value greater than 1, and forecasting permeability is still in transient flow at the end of the fifth
using b-values≥1.0 overestimates remaining reserves. year. Nevertheless, the total well production appears to be
stabilized after five years of production. The regressed
Use of b-value forced to be 1 with transient data included. parameters obtained from history matching of one year’s
With b constrained between 0 and 1 during history matching, production data (from the end of the fifth year to the end of the
we obtained other sets of production forecasts. Table 3 gives sixth year) through five years of data (from the end of the fifth
regressed values of the decline parameters (qi, b, Di) for year to the end of the tenth year) are shown in Table 6. The b-
synthetic Case 1. Di continuously decreases as production time value changes slightly, but remains near 0.6. The error in the
increases whereas b is constant at 1 for all history match forecast is quite small, varying from 0.2% to 1.2%. In contrast
periods. Fig. 5 shows the forecasted production rates. When to the single-layer case, the forecast errors decrease steadily as
one year of rate data is used in history matching, future more data are available for history matching. The forecasted
production is greatly underestimated. When 5 years of rate future production is shown in Fig. 8.
data are used for history matching, projected performance Results from this section demonstrate that decline curve
approaches the actual performance at late times, but most of analysis using only stabilized data is sufficient to produce
the projection is still less than actual. History matching and accurate reserves estimates.
forecasted production change only slightly when history
matching duration increases from 5 years to 10 years. The Improved Analysis Technique
projected rate curve crosses the actual rate curve and leads to For tight gas wells, transient flow may last many years. We
an overestimated prediction. often have to forecast future production and estimate
For synthetic Case 2, multilayer production, Table 4 remaining reserves when we have only transient data.
provides the regressed values of the decline parameters (qi, b, Methodology to improve the reliability of decline curve
Di) for various senarios. Future production is also analysis based on only transient flow data would be of
underestimated for short history match periods, but is much significant value to the industry.
closer to the true value than the single-layer case (Fig. 6). The
projections almost overlap for history matching durations of 5 Methodology
years and 10 years. In the previous section, we saw that b-values obtained from
These observations indicate that b constrained to be less regression are all greater than 1 when transient data are
than or equal to 1 can result in either underestimated or included in the history match. Use of b-values greater than 1
overestimated production forecasts. With values of qi and Di leads to a significantly overestimated production forecast.
obtained from history matching of very early-time data (such Constraining b-values to be less than or equal to 1 also fails to
as one year), future production would typically be provide accurate projections. To project future performance
underestimated. When more late-time data are included, accurately, the b-value that represents stabilized flow is
forecasts approach and surpass the true values. Forecasts tend required. Unfortunately, such a b-value cannot be obtained
to become more stable as more late-time data are included in from analysis of transient flow data.
the analysis. Note that as b is constrained to be less than 1.0, We propose an improved technique to analyze transient-
the quality of the matches becomes worse (as we compare the flow-dominated production data. There are two key
matches from the best curve fit). components in our proposed method. First, we estimate a
priori a b-value that can reasonably represent stabilized
Use of b-value obtained from best fit of only stabilized (boundary-dominated) flow. Second, we fix this b-value in
data. We also examined b-value variation when only history matching of production data to determine other decline
stabilized data during boundary-dominated flow are used in parameters (qi and Di) in a backward data fitting fashion. The
4 SPE 108176

parameters qi and Di are extrapolated backward to time zero, Synthetic Case 1. We assume that only the production data
and the resulting set of decline parameters including the from the first three years of well life are available for the
predetermined b-value is then used for future projections. analysis, and that we need to forecast the production
For a gas well producing from a single layer, the performance for the remaining 27 years.
representative b-value for stabilized flow has been shown to Table 8 provides multiple sets of regressed parameters (qi
be related to the reservoir drive mechanism,3 and also to the and Di) by backward analysis. Note that b is fixed at 0.433,
fluid properties and reservoir conditions.2-3 b is not a constant which is estimated using Eq. 2. We then plot log qi and log Di
but decreases as the reservoir depletes at constant BHP. vs. backward time interval, respectively, as shown in Figs. 9
Although instantaneous values of b can be larger than 1 under and 10. Extrapolating these two plots to time zero, we obtain
some conditions, the average b during the entire depletion the values of qi and Di at the end of the third year, 494
phase will be less than 1.2 As b is a variable, but depends on MSCF/D and 0.000284 1/D, respectively. Taking the end of
the time segment used for history matching, Chen2 proposed a the third year as the initial time for prediction, we used qi =
model to estimate an average b as a best approximation during 494 MSCF/D, b = 0.433 and Di = 0.000284 1/D to predict
boundary-dominated depletion. From commonly available production performance for the remaining 27 years, as shown
reservoir properties, Chen proposed that we calculate an in Fig. 11. The production performance predicted using
average b, called bE, using pseudopressure7: conventional practices of decline curve analysis – (1) using the
b-value from the best fit regardless of the value, or (2)
pi c gi p p ( pi ) − p p ( p wf ) constraining b to less than or equal to 1, as discussed
bE = 1 − ………………...…(2) previously – are also presented in Fig. 11 for comparison. The
zi pi p wf relative errors in estimated remaining reserves are -2.8% using
( )−( ) our method, compared to 56.2% and -35.2% using
zi z wf
conventional practices.
This synthetic case indicates that our method significantly
For a gas well producing from multiple layers, finding the improves future projection with decline analysis for single-
representative value of b is more difficult. Table 7 gives the layer, hydraulically fractured tight gas wells.
range of b-value obtained from matching the remaining well
life at different production stages using data from synthetic Synthetic Case 2. In this case, we assume that only the first
Case 2. This table provides insight to help us identify the year’s production data are available for decline analysis, and
“best” representative value of b for the duration of the production performance for the remaining 29 years needs to be
projection because we obtained b directly by matching the forecasted. We fixed b at 0.6 and history matched the data to
data in the projection period. As we can see, b demonstrated a determine qi and Di. With the backward scheme, we obtained
quite stable value of about 0.6 during most stages of depletion. multiple sets of qi and Di, as summarized in Table 9. Plotting
Therefore, we suggest using b = 0.6 for multilayer production log qi and log Di vs. backward time interval, we obtained the
cases. We will further validate the choice of this b-value later, values of qi and Di at the end of the first year by extrapolation.
and we note that this value is supported by the work of Using b = 0.6 and the extrapolated values qi = 861 MSCF/D
Fetkovich et al.8 and Di = 0.000815 1/D, we projected future production
When b is determined, we need to find appropriate performance (Fig. 12). The predicted performance from two
values of qi and Di for forecasting. We used a backward conventional practices used in decline curve analysis is also
analysis scheme.9 In this approach, we start at the latest point presented in Fig. 12. The relative errors of estimated
of available production data and go back a selected time remaining reserves are -14.7% from our method, compared
interval to obtain a segment of data for history matching. We with 109.9% and -28.3% from other methods. It is obvious
obtain one set of parameters from fitting the selected data that our method provides much better estimation than
segment. If we start at the end and go back further to another, conventional practices of decline analysis.
earlier time point, we can obtain another segment of data for We examined the effectiveness and superiority of our
history matching and thus obtain another set of parameters. In method when more production data are available. Fig. 13
this way, we obtain multiple sets of regressed parameters with presents the forecasts for 28 and 27 years. The 28-year
available data. Each set of parameters (qi and Di) correspond projection corresponds to two years of production data
to a given backward time interval. If we then plot log Di vs. available for analysis, while the 27-year projection
backward time interval and extrapolate the plot to time zero, corresponds to three years of production data available for
we obtain the Di corresponding to the latest time for which we analysis. The 29-year projection is also included in Fig. 13.
have production data. The same procedure is also applied to qi The error in the production forecast is reduced significantly
to obtain the value of qi corresponding to the latest time. when more production data are available. Using decline
Forecasting is then initiated at the latest point of production analysis of two years of production data, we obtained the
data with extrapolated values of qi and Di, and the preselected estimated remaining reserves with a relative error of -4.5%,
b-value. compared to -14.7% when only one year of data was available.
Using decline analysis of three years of production data, we
Validation with synthetic cases obtained the estimated remaining reserves with a relative error
We used the two synthetic cases given in the previous of -2.0%.
section to validate the proposed method. Synthetic Case 2 indicates that our method greatly
improved production forecasts and reserves estimates for
SPE 108176 5

multilayer tight gas wells when only transient flow data are 3. The proposed method mitigates the complexity of decline
available for the analysis. It is superior to conventional curve analysis in tight gas wells. Our method was
practices used in decline analysis. This new method validated using synthetic examples for single-layer and
consistently provides more accurate forecasts when more multilayer gas wells. Accurate reserves estimates and
production data are available. performance forecasts were obtained.
4. Practical applicability of our method was demonstrated
Field example with an actual tight gas well producing from multiple
We next apply our new method to a field example and zones fractured with multistage treatments. The forecast is
demonstrate its applicability to typical well data. quite consistent with the actual behavior. Similar to the
This field example is a gas well producing from a tight gas synthetic examples, an accurate reserves estimate and
reservoir with permeabilities ranging from 0.002 to 0.02 md. performance forecast were obtained.
The initial reservoir pressure was about 4,500 psi and the net 5. The new method provides a useful tool to estimate
pay is estimated to be 150 ft. Three staged hydraulic fracture reserves and forecast future performance from decline
treatments were implemented on multiple productive zones. analysis of the early-time production data in tight gas
The fracture half-lengths are estimated to range between 300 ft wells. Althrough there are some limitations, it should still
and 600 ft. This well was plugged after 20 years of production. be far better than conventional practices.
Fig. 14 shows the monthly gas rate history over the well life.
We applied our decline analysis method to the first three Nomenclature
years of data and then predicted performance for the b = Arps’ decline exponent
remaining 17 years. Fig. 15 is the plot of log qi vs backward bE = average decline exponent
time interval. The value of qi extrapolated to time zero is cgi = initial gas compressibility, 1/psia
7,315 MSCF/mo. Fig. 16 is the plot of log Di vs backward Di = initial decline rate, 1/D or 1/mo
time interval. The value of Di extrapolated to time zero is p = initial pressure, psia
0.0087 1/mo. Note that outliers were removed for pp = pseudopressure, psia2/cp
extrapolation because we need to characterize the overall pwf = bottom-hole flowing pressure, psia
decline trend instead of localized variation. We took the end of q = gas production rate, MSCF/D
the third year as the initial point of projection, and used the set qi = initial gas production rate, MSCF/D
of parameters qi = 7,315 MSCF/mo, b = 0.6 and Di = 0.0087 t = time, days or months
1/mo for forecasting. The projected performance is shown in zi = initial gas compressibility factor, 1/psi
Fig. 14, along with the actual production performance. Fig. 14 zwf = gas compressibility factor at BHP, 1/psi
shows that the predicted decline behavior is very consistent
with the observed behavior. The relative errors of estimated References
remaining reserves and ultimate reserves are -2.97% and - 1. Arps, J.J.: “Analysis of Decline Curves,” Tran., AIME (1945)
1.97%, respectively. Our method led to a forecast that is a 160, 228.
good representation of actual production history. 2. Chen, Her-Yuan.: “Estimating Gas Decline-Exponent before
Fig. 17 presents the projections obtained using Decline-Curve Analysis,” SPE 75693 presented at the 2002 SPE
Gas Technology Symposium, Calgary, Alberta, Canada, 30
conventional practices in decline analysis, with the same April-2 May.
production data as we used in our method, i.e, the first three 3. Fetkovich, M.J., Fetkovich, E.J. and Fetkovich, M.D.: “Useful
years of production data. One forecasts far too much future Concepts for Decline-Curve Forecasting, Reserve Estimation
production, and the other forecasts far too little future and Analysis,” SPERE (Feb. 1996) 13.
production. The relative errors in estimated remaining reserves 4. Cox, S. A.: “Reserve Analysis for Tight Gas,” paper SPE 78695
are 36.50% and -47.18%, respectively, for these other presented at the 2002 SPE Eastern Regional Meeting,
methods. Lexington, KT, 23-25 October.
This field example indicates that our method can 5. Neal, D.B. and Mian, M. A..: “Early-Time Tight Gas Production
significantly improve the reliability of decline curve analysis Forecasting Technique Improves Reserves and Reservoir
Description,” SPEFE (Mar. 1989) 25.
in multilayer tight gas wells, since the error of reserves 6. Maley, S.: “The Use of Conventional Decline Curve Analysis in
estimation and production forecast is greatly reduced. Tight Gas Well Applications,” paper SPE/DOE 13898 presented
at the 1985 SPE/DOE Low Permeability Gas Reservoirs
Conclusions Symposium, Denver, CO, 19-22 May.
1. Conventional practices of decline curve analysis in tight 7. Lee, W.J., Rollins, J.B., and Spivey, J.P.: Pressure Transient
gas wells are problematic in forecasting long-term Testing, Textbook Series, SPE Richardson, TX (2003) Vol. 9.
performance, and typically provide quite inaccurate 8. Fetkovich, M.J. et al: “Depletion Performance of Layered
reserves estimates. Reservoirs without Crossflow,” SPEFE (Sept. 1990) 310.
2. We have developed a new method to accurately project 9. Cheng, Y., Wang, Y., McVay, D.A. and Lee, W.J.: “Practical
Application of a Probabilistic Approach to Estimate Reservoirs
production performance and reliably estimate remaining Using Production Decline Data,” SPE 95974 presented at the
reserves from decline analysis of the early-time 2005 Annual Technical Conference and Exhibition, Dallas, 9-12
production data. October.
6 SPE 108176

Table 1. Regressed parameters from history Table 5. Regressed parameters from history
matching different duration periods for synthetic matching of only boundary-dominated (stabilized)
case 1 (b obtained from best curve fit) flow data for synthetic Case 1
HM* period, year HM period, year
q i , MSCF/D b D i , 1/D q i , MSCF/D b D i , 1/D Error*
Start End Start End
5 6 428.5 0.362 0.0002620 1.0%
0 1 2025 2.93 0.0164
5 7 428.5 0.432 0.0002621 3.7%
0 2 2030 3.01 0.0171
5 8 428.5 0.432 0.0002620 4.1%
0 3 2031 3.02 0.0173 5 9 428.5 0.419 0.0002617 3.1%
0 5 2023 2.96 0.0164 5 10 428.4 0.397 0.0002608 3.6%
0 10 1945 2.59 0.0114 * Relative error in estimated remaining reserves
0 20 1675 1.90 0.0045
* History matching
Table 6. Regressed parameters from history
matching only boundary-dominated (stabilized) flow
Table 2. Regressed parameters from history data for synthetic Case 2
matching different duration periods for synthetic HM period, year
q i , MSCF/D b D i , 1/D Error*
case 2 (b obtained from best curve fit) Start End
HM period, year 5 6 401.4 0.615 0.000413 1.2%
q i , MSCF/D b D i , 1/D
Start End 5 7 401.4 0.606 0.000413 0.9%
0 1 2027.6 3.16 0.0100 5 8 401.3 0.598 0.000412 0.6%
0 2 2006.3 2.84 0.0086 5 9 401.3 0.592 0.000412 0.4%
0 3 1971.5 2.52 0.0071 5 10 401.3 0.588 0.000412 0.2%
0 5 1891.9 2.03 0.0050 * Relative error in estimated remaining reserves
0 10 1744.2 1.43 0.0029
0 20 1622.5 1.06 0.0019
Table 7. Regressed parameters from history
matching late-time data for synthetic Case 2
HM period, year
Table 3. Regressed parameters from history q i , MSCF/D b D i , 1/D
Start End
matching different duration periods for synthetic
1 30 885 0.638 0.00071
case 1 (b constrained to be < 1)
2 30 687 0.605 0.000578
HM period, year
q i , MSCF/D b D i , 1/D 5 30 401 0.589 0.000412
Start End
10 30 215 0.594 0.000287
0 1 1745 1.0 0.0046
13 30 161 0.599 0.000242
0 2 1581 1.0 0.0030 16 30 126 0.592 0.000208
0 3 1474 1.0 0.0022 19 30 102 0.571 0.000182
0 5 1348 1.0 0.0016 20 30 95 0.554 0.000175
0 10 1229 1.0 0.0011
0 20 1199 1.0 0.0010
Table 8. Regressed qi and Di for synthetic Case 1
HM period, days Time interval qi b Di
Table 4. Regressed parameters from history
Start End days MSCF/D 1/D
matching different duration periods for synthetic 182 1095 913 864 0.433 0.00067
case 2 (b constrained to be < 1) 365 1095 730 733 0.433 0.00051
HM period, year q i , MSCF/D b D i , 1/D 547 1095 548 658 0.433 0.00043
Start End 730 1095 365 605 0.433 0.00038
0 1 1810 1.0 0.0033 912 1095 183 564 0.433 0.00035
0 2 1699 1.0 0.0024
0 3 1643 1.0 0.0020
0 5 1594 1.0 0.0018 Table 9. Regressed qi and Di for synthetic Case 2
0 10 1573 1.0 0.0017 HM period, days Time interval qi b Di
0 20 1589 1.0 0.0018 Start End days MSCF/D 1/D
91 365 274 1285 0.6 0.00152
182 365 183 1104 0.6 0.00119
273 365 92 989 0.6 0.00101
SPE 108176 7

10000
10000
True production
1 year HM
10 year HM
20 year HM
Gas rate, MSCF/D

Gas rate, MSCF/D


1000 1000

100
100

10
0 2,000 4,000 6,000 8,000 10,000 12,000 10
0 2,000 4,000 6,000 8,000 10,000 12,000
Time, days
Time, days

Fig. 1—Production performance for hydraulically fractured well in Fig. 4—Predicted performance resulting from conventional
single-layer reservoir (Case 1). decline analysis techniques for synthetic Case 2 (b obtained from
best curve fit).

10000
10000
True production
Layer 1: k=0.02 md 1 year HM
Layer 2: k=0.006 md 5 year HM
Layer 3: k=0.002 md 10 year HM
1000 Total production
Gas rate, MSCF/D

1000
Gas rate, MSCF/D

100

100

10

10
1
0 2,000 4,000 6,000 8,000 10,000 12,000
0 2,000 4,000 6,000 8,000 10,000 12,000
Time, days
Time, days

Fig. 2—Production performance for hydraulically fractured well in Fig. 5—Predicted performance resulting from conventional
multilayer reservoir (Case 2). decline analysis technique for synthetic Case 1 (b constrained to
be < 1.0).

10000 10000
True production Total production
1 year HM 1 year HM
10 year HM 5 year HM
20 year HM 1000 10 year HM
Gas rate, MSCF/D
Gas rate, MSCF/D

1000

100

100
10

10 1
0 2,000 4,000 6,000 8,000 10,000 12,000 0 2,000 4,000 6,000 8,000 10,000 12,000

Time, days Time, days

Fig. 3—Predicted performance resulting from conventional Fig. 6—Predicted performance resulting from conventional
decline analysis techniques for synthetic Case 1 (b obtained from decline analysis technique for synthetic Case 2 (b constrained to
best curve fit). be < 1.0).
8 SPE 108176

10000 0.01
Series1
HM over 5 years to 8 years
HM over 5 years to 7 years y = 0.000284e0.00087x
HM over 5 years to 6 years
HM over 5 years to 9 years R2 = 0.9497
Gas rate, MSCF/D

1000 HM over 5 years to 10 years

Di , 1/day
0.001

100

0.0001
0 200 400 600 800 1,000
10
0 2,000 4,000 6,000 8,000 10,000 12,000 Backward time interval, days

Time, days
Fig. 10—Extrapolating Di to the initial time of projection for
Fig. 7—Predicted performance resulting from stabilized flow for
synthetic Case 1.
synthetic Case 1.

10000
True production 10000
HM over 5 years to 10 years True production
HM over 5 years to 9 years b=3.0 for best fit, error=56.2%
HM over 5 years to 8 years b=1.0 for b≤1 constraint, error=-35.2%
HM over 5 years to 7 years b=0.433, new method, error=2.8%
Gas rate, MSCF/D

1000 HM over 5 years to 6 years


Gas rate, MSCF/D

1000

100
100

HM Forecast

10
10
0 2,000 4,000 6,000 8,000 10,000 12,000
0 2,000 4,000 6,000 8,000 10,000 12,000
Time, days
Time, days

Fig. 11—Forecasts using different decline analysis techniques for


Fig. 8—Predicted performance resulting from stabilized flow for
synthetic Case 1.
synthetic Case 2.

10000 10000
True production
y = 494.3e0.0006x b=3.16, best fit, error=109.9%
b=1, constraint b≤1, error=-28.3%
R2 = 0.9702 b=0.6, new method, error=-14.7%
Gas rate, MSCF/D

1000
qi , MSCF/D

1000

100 HM

Forecast

100
0 200 400 600 800 1,000 10
Backward time interval, days 0 2,000 4,000 6,000 8,000 10,000 12,000
Time, days
Fig. 9—Extrapolating qi to the initial time of projection for Fig. 12—Forecasts using different decline analysis techniques for
synthetic Case 1. synthetic Case 2.
SPE 108176 9

10000
0.1
True production
1 year HM, error=-14.7%
2 year HM, error=-4.7%
3 year HM, error=-2.0% Outlier
Gas rate, MSCF/D

1000

Di , 1/mo
0.01

100
y = 0.0087e0.0375x
R2 = 0.9483

0.001
10 0 5 10 15 20 25 30 35 40
0 2,000 4,000 6,000 8,000 10,000 12,000 Backward time interval, months
Time, days

Fig. 13—Forecasts using our decline analysis technique for Fig. 16—Extrapolating Di to the initial time of projection for field
synthetic Case 2. example.

100,000 100,000
Actual data
Actual data b=0.6, new method, error=-2.97%
b=1, constraint b≤1, error=-47.18%
Predicted performance Gas rate, MSCF/mo b=2.65, best fit, error=36.50%
Gas rate, MSCF/mo

10,000 10,000

1,000 1,000

100 100
0 50 100 150 200 250 300 0 50 100 150 200 250 300

Time, months Time, months

Fig. 14—Production history and forecast for field example. Fig. 17—Forecasts using different decline analysis techniques for
field example.

100,000
y = 7315.1e0.025x
R2 = 0.9723
qi , MSCF/mo

10,000

1,000
0 5 10 15 20 25 30 35 40
Backward time interval, months

Fig. 15—Extrapolating qi to the initial time of projection for field


example.

You might also like