Beruflich Dokumente
Kultur Dokumente
properties such as porosity and permeability, constructing and demonstrated to correlate well with long-term production,8
calibrating a reservoir simulation model of the area, and then shown in Fig. 3. Virgin best year, VBY, is the BY of a well at
using the reservoir model to predict future production and virgin conditions. Depletion effects are removed by computing
reserves at potential infill well locations. While it may be the BY of a local area at a time before depletion using a 2D
accurate, this approach can be prohibitively time-consuming regression of BY vs. well start date. VBY is used as a proxy
and expensive. For some large, low-permeability gas basins for kh in the pseudosteady state flow equation. In this method,
with large data sets (sometimes over 1,000 wells) and complex well spacing is used as a proxy for drainage area and is
geology, the cost and time requirements of a conventional calculated from the area of voronoi polygon around each well
reservoir evaluation study are not acceptable. based on well locations.
As an alternative approach to conducting detailed studies,
various authors have used empirical or statistical analyses to 10,000
model variable well performance.4-12 In particular, McCain et
al.7 used a statistical, moving-window method to determine Maximum Month
1,000
Production (MMscf)
10-Year Cumulative
100
10
0.1
0.01 0.1 1 10 100
Fig. 1 - Diagram of the fast method showing how the window Monthly Rate During Best Year (MMscf/M)
moves across area. The small blue circles are the well
locations and big circles are the moving domains. Fig. 3 - Best Year vs. 10-year cumulative production.8
In this work, however, a more rigorous, model-based
Regression coefficients for each window are determined
analysis is employed in each moving domain. The model is
by regressing the parameters (well spacing, Gp/A, VBY) for
based on a combination of the material balance equation and
the wells within each window. The windows are limited in
the pseudosteady state flow equation, simplified by assuming
size, e.g., 3000 acres, and generally contain 5 to 20 wells. If
that many properties are constant within an individual moving
the number of wells in a window is less than a minimum
domain. The result is a linear regression equation that is
value, e.g., 3-5, a regional or global regression is used instead
applied within each window.
of a local regression.
In this rigorous, model-based method, BY is used as a
Once the regression equation coefficients are determined
proxy for production rate in the pseudosteady state flow
for each window, performance can be estimated for infill wells
model. BY, bestyear, is simply the best 12 consecutive months
by substituting the appropriate values for candidate infill well
of production divided by 12, shown in Fig. 2 and it has been
conditions. The result of this analysis is a prediction of BY for
SPE 91755 3
a new infill well offsetting each existing well. The primary 100
advantages of the moving domain technique are its speed and 95
its reliance upon only well location and production data. It is 90
85
routinely used to conduct infill screening studies of projects
80
consisting of 1000’s of wells and can be used to evaluate an 75
entire basin in a few man-days. 70
65
60
Generation of Test Permeability Fields and 55
Production Data Sets 50
-50% -30%
performance, it was first necessary to determine the true infill 2000
well performance from simulation. To do this, we placed a 1800
new well in the first grid block and made a 1-year projection, 1600
from the end of history, to determine the production to be 1400
+30%
gained by a new well in this location. We then repeated this 1200
for each grid block in the system, thus generating a map, or
1000
distribution, of the additional production attributable to one +50%
800
new well at all the possible grid locations in the reservoir.
600
The infill-well performance estimates from simulation are
400
on a cell basis, while the fast method results are on a well
basis. Therefore, the simulation results must be converted to a 200
Relative error, %
-10
determine the effect of permeability level on the estimation
results, we also ran another homogeneous reservoir case. We
increased the average permeability to 1.0 md and compared -15
the results with the 0.2md reservoir case.
For the 0.2 md Case 1, the average infill BY from the fast
-20
method is 513 MSCM/M and the average infill BY from
simulation is 543 MSCM/M in Case 1. Thus, on average, the
fast method underestimates the infill BY by about 6%. This -25
small error is noteworthy, considering that the reservoir has 0.1 1 10
experienced depletion resulting in a 12% decrease in well Search area/correlation length, dimensionless
productivity.
But for the 1.0 md homogenous case, the error increases. Fig. 6 - Relative error of infill BY vs. search area/correlation
The estimates become more scattered in this case and several length for Case 3.
wells differ by more than 50% from the simulated values. The
fast method underestimates the average infill BY by 25% for Effect of Well Spacing
this case. The larger error is attributed to greater depletion, The fast method will calculate the initial well spacing and
40% for the 1.0 md case, which results when only current well spacing in the data processing. The current well
permeability is increased and other reservoir parameters spacing is used as a proxy for the drainage area. We want to
remain the same. determine the effect of well spacing on the estimation results
of the moving domain technique. But the well spacings of the
Effect of Search Area 100 wells in the four cases that we ran are not the same. It is
The default value for the local search area in the fast method is very difficult to draw any conclusions with regard to well
3,000 acres. To find the effect of the search area, we varied the spacing from these 100-well cases.
search areas from 1,000 to 12,000 acres in the analysis of Case To investigate the effects of non-uniform well spacing,
3 and the result is presented in Fig. 6. we ran four additional cases with the same permeability
It seems that the smaller domain size results are better for distribution and a uniform arrangement of wells. The results
this case. Generally speaking, there is a larger variation in show that the well spacing does not have a significant effect
reservoir properties in bigger domains than in smaller on the estimation accuracy of the fast method.
domains. These results indicate that, if the domain size is
greater than about 3,000 acres, the relative error of the infill Effect of the Date for the Infill Well on the Analysis
BY begins to increase significantly. The objective of this part of the study was to determine if the
From this study, we know that the size of the search area time difference between the date for the infill wells and the
will have a significant effect on the estimation accuracy of the latest wells' date of first production in our study area
fast method. The optimum search-area size will be the first influences the estimation results.
parameter we need to determine when we apply this Fig. 7 shows the time-production performance for a
technology to a new area. typical domain in our study. The solid line is the best-fit line
We also note the non-monotonic behavior of the relative that the fast method used to calculate infill well performance.
error of infill BY with respect to search area size in Fig 6, but But, after the second round of infill drilling, the depletion rate
we are not sure what cause that behavior. However, we do of the closed-boundary reservoir increased from that of the
know that the search area of 2,000 acres is very close to the period after the first round of infill drilling. The reservoir
correlation length of the permeability distribution for Case 3. depletion rate follows the dashed line after the second round
Thus, it appears that the search area should be less than or of infill drilling; it will not follow the best-fit line any more.
equal to the correlation length of the permeability field and the The ongoing infill drilling campaigns will deplete the
optimal search area is the correlation length of the fields. reservoir more rapidly than before because of the large
Geologically, the search area size should be small enough to number of infill wells. The performance of the latest wells will
avoid large changes in regional reservoir properties, e.g., give us more valuable information on the current reservoir
permeability. However, statistically, it needs to be large than the old wells.
enough to have at least several wells per window. The difference between the two lines in Fig. 7 is not very
large at first but increases with time. This suggests that the
most accurate estimation time for the infill wells may be soon
after the latest round of infill drilling.
SPE 91755 5
120 Case 4 Field8-9 show limited sand continuity among wells and large
various in sand qualities over short distances. Therefore, well
100 interference was not expected in the majority of the field. The
large number of existing wells and the compartmentalized
80
nature of the sands precluded detailed reservoir analysis to
60 determine the infill drilling potential in the Ozona field.
Voneiff, et al.8 first developed the fast statistical method
40 and used it to determine the infill drilling potentials in the
Ozona field. The results of their study identified 1,246 infill
20 candidates representing 636 BCF of additional reserves in the
field. Using this method, not only were they able to quantify
0
the number of infill wells and infill reserves, but they were
0.05 0.1 0.15 0.2 0.25 0.3 0.35
also able to identify the location of the infill wells in a short
1/SQRT(number of wells) time frame. Fig. 9 shows the location of the infill potential
Fig. 8 - Variability of difference between the fast method and across the Ozona field.
simulation infill potential estimates decreases as number of wells
increases. The lines are least-squares fit to the data points from
different cases.15
6 SPE 91755
• This technology is developed mainly for single-phase 3. Wu, C.H., et al.: “An Evaluation of Waterflood Infill Drilling in
flow of gas in tight-gas reservoirs. We do not know the West Texas Clearfork and San Andres Carbonate Reserves,”
accuracy of the results from this technology under SPE paper 19783 presented at the 1989 SPE Annual Technical
Conference and Exhibition, San Antonio, 8-11 October.
multiphase flow conditions, such as from gas reservoirs 4. French, R.L., Brimhall, R.M., and Wu, C.H.: “A Statistical and
with significant water production or from oil reservoirs. Economic Analysis of Incremental Waterflood Infill Drilling
Recoveries in West Texas Carbonate Reservoirs,” paper SPE
Conclusions 22624 presented at the 1991 SPE Annual Technical Conference
In this paper we systematically evaluated the accuracy of a and Exhibition, Dallas, 6-9 October.
fast method which can rapidly assess the infill drilling 5. Wu, C.H., Lu, G.F., Gillespie, W., and Yen, J.: “Statistical and
Fuzzy Infill Drilling Models for Carbonate Reservoirs,” paper
potentials in large tight gas basins. After the parameter SPE 37728 presented at the 1997 SPE Middle East Oil Show &
sensitivity study, the strength and weakness, and field Conference, Bahrain, 15-18 March.
applications of the fast method was discussed. The specific 6. Soto, B.R., Wu, C.H, and Buleba, A.M.: “Infill Drilling
conclusions of this study are as follows. Recovery Models for Carbonate Reservoirs – A Multiple
Statistical, Non-Parametric Regression, and Neural Network
• For the cases examined in this work, the predicted Approach,” paper SPE 57458 presented at the 1999 SPE Eastern
Regional Conference and Exhibition, Charleston, West Virginia,
average infill performance was either very close to or less
21-22 October.
than the simulated infill performance. This suggests that 7. McCain, W.D. Jr., Voneiff, G.W., Hunt, E.R., and Semmelbeck,
the fast method tends to underestimate the infill drilling M.E.: “A Tight Gas Field Study: Carthage (Cotton Valley)
potential, providing a conservative estimate. Field,” paper SPE 26141 presented at the 1993 SPE Gas
• The accuracy of predicted infill well performance, for Technology Symposium, Calgary, 28-30 June.
either individual wells or the average of a group of wells, 8. Voneiff, G.W. and Cipolla, C.: “A New Approach to Large-
Scale Infill Evaluations Applied to the Ozona (Canyon) Gas
decreases as heterogeneity increases, and increases as the
Field,” paper SPE 35203 presented at the 1996 SPE Permian Oil
number of wells in the group increases. and Gas Recovery Conference, Midland, Texas, 27-29 March.
• The search area used in this fast method should be either 9. Cipolla, C. L., and Wood, M.C.: “A Statistical Approach to
close to or less than the reservoir permeability correlation Infill-Drilling Studies: Case History of the Ozona Canyon
length. Sands,” SPE Reservoir Engineering, August 1996, p196-202.
• The estimation error of the fast method is related to the 10. Hudson, J.W., Jochen, J.E., and Jochen, V.A.: “Practical
Technique To Identify Infill Potential in Low-Permeability Gas
time difference between the date for the new infill wells
Reservoirs Applied to the Milk River Formation in Canada,”
and the date of the last round of infill wells. This suggests paper SPE 59779 presented at the 2000 SPE/CERI Gas
that the most accurate estimation time for the infill wells Symposium, Calgary, 3-5 April.
may be soon after the latest round of infill drilling. 11. Hudson, J.W., Jochen, J.E., and Spivey, J.P.: “Practical Methods
• The primary advantages of the fast method are its speed to High-Grade Infill Opportunities Applied to the Mesaverde,
and its reliance upon only well location and production Morrow, and Cotton Valley Formations,” paper SPE 68598
presented at the 2001 SPE Hydrocarbon Economics and
data. It can be used to conduct infill-screening studies of
Evaluation Symposium, Dallas, 2-3 April 2001.
projects consisting of thousands of wells and can be used 12. Kyte, D.G. and Meehan, D.N.: “Horizontal Spacing, Depletion,
to evaluate an entire basin in a matter of man-days. and Infill Potential in the Austin Chalk,” paper SPE 36721
presented at the 1996 SPE Annual Technical Conference and
Acknowledgement Exhibition, Denver, CO, October 6-9.
13. Guan, L., McVay, D. A., Jensen, J. L., and Voneiff, G. W.,
The authors of this paper wish to thank Dr. Duane A. McVay and Dr. “Evaluation of a Statistical Infill Candidate Selection
Jerry L. Jensen for their guidance in this work and George Voneiff Technique”, SPE paper 75718 presented at the 2002 SPE Gas
for valuable input in this paper and MGV Energy Inc for providing Technology Symposium, Galgary, Alberta, April 30-May 2.
the financial support for this research. 14. Guan, L., McVay, D. A., and Jensen, J. L.:“Parameter
Sensitivity Study of a Statistical Technique for Fast Infill
NOMENCLATURE Evaluation of Tight Gas Reservoirs”, CIPC paper 2004-163
presented at 2004 Canadian International Petroleum Conference,
A = Well spacing, acre Calgary, June 8-10.
Avg = Average 15. Guan, L., McVay, D. A., and Jensen, J. L. and Voneiff, G. W.,
BY = BestYear, MSCM/M “Evaluation of a Statistical Method for Assessing Production
Gp = Cumulative production Potential in Mature, Low-Permeability Gas Reservoirs”, to
km = Kilometer appear in Journal of Energy Resources Technology.
16. Weber, K.J., and VanGenus, L.C.: “Framework for Constructing
MMScf/M = Million standard cubic feet per month
Clastic Reservoir Simulation Models,” Journal of Petroleum
MSCM/M = Thousand standard cubic meter per month Technology, 42 (1990).
STDEV = Standard deviation 17. Hecker, M.T. and Downie, R. C.: “Process Changes Improve
VBY = Vigin best year, MSCM/M Fracture Treatment Designs in the Hugoton Gas Field,” paper
SPE 35259 presented at the 1996 Gas Technology Conference,
REFERENCES Calgary, 28 April - 1 May.
1. The IOGCC Annual Report 2002-2003.
2. Thakur, G.C. and Satter, A.: Integrated Waterflood Asset
Management, PennWell, 1998.