Вы находитесь на странице: 1из 6

SPE 52224

Reservoir Management Using Production Data


T. Soeriawinata and M. Kelkar, University of Tulsa

Copyright 1999, Society of Petroleum Engineers, Inc.


signals passed between wells measured by rate correlations
This paper was prepared for presentation at the 1999 SPE Mid-Continent Operation had at least some component coupled to geomechanics.
Symposiom held in Oklahoma City, Oklahoma, U.S.A., 2831 March 1999.
Refunjol and Lake2 presented a practical technique to
This paper was selected for presentation by an SPE Program Committee following review of determine preferential flow trends in a reservoir utilizing
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 Spearman rank correlation analysis. In their work, they
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 correlated time series of injection and production rates ranks
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Permission to copy is restricted to an abstract of not more than 300
from pairs composed of each injection well and all adjacent
words. Illustrations may not be copied. The abstract should contain conspicuous producers. Different time lags are utilized to find an extreme
acknolwedgment 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. coefficient value. The extreme rank correlation coefficient
values of all injectors are then utilized to infer preferential
flow directions. The inference is done by grouping the pairs
based on their spatial orientation, then constructing histogram
Abstract
based on the orientation of well pairs with maximum
This paper presents a methodology to effectively use the
correlation. One alternative is by linearly interpolating along a
production data to improve the performance of the existing
line from an injector to the adjacent producers and creating
water flood. Using simple statistical procedures and method of
contour map for each injector. They applied their technique to
superposition (e.g., one producing well could be affected by
a field and compared the result with tracer data.
several injection wells), the injection wells which are affecting
Jansen3 stated that if the rate series were not prewhiten, the
the production of oil and water from individual producing
correlations were no longer between local variations in the two
wells are determined. The connection or the lack of
rates, but rather between the trends of the series. Thus, without
connection between injection and producing wells based on
obtaining stationary series through prewhitening (i.e.,
the previous step can be utilized in optimizing water flood
differencing the series), it is not possible to determine a direct
process and in modeling the reservoir. The method is tested in
causal relationship between the two series. Unfortunately,
a synthetic water flood field and in a mature water flood field
prewhitening would amplify the noise in the data. He also
in West Texas.
suggested that pressure superposition between the injection
wells and producers caused an effect where mainly large
positive rate changes could be correlated, but not negative one.
Introduction This is due to the fact that other injectors would try to
For small operators and producers, the most abundant compensate the lost pressure and mass support when one
information available is the production data. Typically, for injectors rate was reduced.
water flood operation, based on monthly well tests and overall Panda and Chopra4 utilized different approach to
production, a great deal about individual oil and water determine injector-producer interaction. An artificial neural
production rates as well as water injection rates is known. network was trained to estimate the well interaction between
Using this information as well as any supporting information, different injectors-producers pairs. The approach was then
the operator is ultimately interested in optimizing the water applied to numerical simulation of a water flood
flood. This may include changing the flood pattern, re- In this study, following the previous studies1-3, we also
completing the wells or drilling in-fill wells. The typical utilize Spearman rank cross correlation between injectors and
applications of the production data are in the form of relating producers. Furthermore, we account for the superposition
the water injection rates from injection wells to the production effect in the reservoir, caused by influence of multiple
rates from surrounding producing wells. injection wells on a producing well. Jansen3 described that
Heffer et al.1 utilized Spearman rank correlations to relate reservoir can be viewed as a system that converts the input
pairs of wells, each pair consisted of an injector and a signals, in the form of injection rates, into output signals, in
producer. They concluded that the communications through the form of production data. The fact that there are a number
the reservoir were reflected by the correlations in rates of injectors and producers operating simultaneously causes
between injectors and producers. They also suggested that the
2 SOERIAWINATA AND KELKAR 52224

both superposition and noise, which must be accounted for in 5. Check if the cumulative cross correlation is greater than a
analyzing the data. threshold value. If it is greater, then the set of injectors is
accepted, otherwise all injectors are rejected. As in step 2,
the choice of threshold is an arbitrary one. For our cases,
Procedure we used a threshold value of 0.5.
Superposition between several injectors has a significant effect 6. Repeat the step 1 to 5 for other producers in the field.
to the response of a producer. It can be constructive or The difficulties in the above procedure are to determine the
destructive. Figure 1 shows a typical five-spot pattern where cross correlation threshold values. There are two thresholds
one producer could be influenced by four injectors. In reality, applied, the first is the minimum cross correlation for the first
the producer can also be influenced by injectors that are injector [step 2], the second is the minimum cumulative cross
outside the pattern. The effect of multiple injectors on the correlation for the set of injectors [step 5]. Besides these
producer can be constructive or destructive, as shown in thresholds, we may need to use another one, the minimum
Figure 2. The basic assumption for this work is that if two increment in cumulative cross correlation to be considered
injectors are in good communication with a common producer, significant. If an injector is added and the increase in
it is expected that the cross correlation of the summation of the cumulative cross correlation is lower than this threshold, we
water injection rates to the liquid production rate is higher need to check on the trend of cumulative cross correlation
than the cross correlation of each single injector. from after this injector (X) is added to when the last injector
Unfortunately, it is practically impossible to distinguish (Y) is accepted in the set of supporting injectors. If the
whether a slight increase in cross correlation is due to increment in cumulative cross correlation when each injector
superposition effect or due to noise in field rate data. (from X to Y) is added is always lower than the minimum
However, if there is a significant jump in cross correlation increment threshold, then all injectors from X to Y are
value when rate from an injector is added, most probably it is insignificant, thus should be rejected. However, if one injector
caused by constructive interference in injection rates rather (Z), in between X and Y, is added and the cumulative cross
than caused by the noise. correlation increment is higher than the increment threshold,
The procedure utilized in this study is summarized below. then all injectors from X to Z are accepted. Injectors between
1. Select one producer from the field. Z to Y should be analyzed by using similar process. We will
2. Determine an injector that has the highest cross illustrate the process by using the data in Table 1. Consider
correlation value to the selected producer among all the Prod-1, when Inj-3 and Inj-4 were added, the cumulative cross
directly adjacent wells. For a five-spot pattern water correlation increments were lower than the minimum
flood, we select all the four injectors surrounding the increment (0.005 for this case) for both Inj-3 and Inj-4. Thus,
producer. The highest cross correlation value has to be both injectors were rejected. However, for Prod-2, when Inj-4
greater than a minimum threshold. The choice of a was added, the cumulative cross correlation increased more
threshold value is an arbitrary one; in our study, we than the minimum threshold. Thus, both Inj-3 and Inj-4 were
selected a value of 0.25. However, when other values are accepted.
selected, the results are not significantly different. If no Below, we will show the application of the procedure
adjacent injector has cross correlation higher than the explained above, first to a synthetic reservoir, and second to
threshold, the search proceeds to step 5, i.e., search the real field data from North Robertson Unit, Texas.
injectors outside the pattern.
3. After determining the injector with the highest cross
correlation, find an injector that causes highest significant Application to synthetic field
improvement in cross correlation. This is achieved by The field is a rectangular homogeneous reservoir, which is
adding the injection rates series from the the previous compartmentalized into several zones by sealing barriers.
injector [step 2] and a newly selected injector, and Figure 4 shows the map of the field. There are 12 5-spot
calculating cross correlation between the total injection patterns with 12 producers and 20 injectors. The water flood
rates series and liquid production rates series from the performance was simulated using ECLIPSE simulator. The
producer. The newly selected injector is changed, one at a water injection rates at all injectors were varied monthly, and
time, until we find an injector which provides the most the responses of producers were recorded as monthly liquid
improvement in cross correlation. The selected injectors production rates. The above procedure was applied to the
are restricted to certain pre-defined regions from a field. Table 2 shows the results of all producers in the field.
producer. An example is shown in Figure 3. This distance Note that cross correlation values are very high for all wells.
limit is utilized to restrict the search, as it is unlikely that Clearly, we actually did not need to utilize any threshold in
an injector too far away from a producer has a significant this case. The reason of these high values is that the synthetic
impact on a producer. reservoir has closed outside boundaries. The water injected
4. Repeat step 3 until there is no increase in cross correlation could not flow outside the reservoir boundaries. Real fields
when a new injector is added. almost never have perfect closed boundaries, consequently
some of the injected water travels outside the boundaries and
52224 RESERVOIR MANAGEMENT USING PRODUCTION DATA 3

can not be recovered. This will reduce the cross correlations of communication in section 329. Since the primary recovery
real fields. Figure 5 shows the mapping of all the from this section was good, this area is a prime candidate for
connectivities between injectors and producers in the synthetic infill wells. In 1996, six producers and two injectors were
field. It is shown that all connectivities occured within each drilled in the section. The cumulative production from each of
compartment, except for connectivities between Producer 204 the new producer is shown in the box. All the producers were
and Injector 107 and between Producer 208 and Injector 113. economically succesful. Figure 9 shows a plot of monthly
However, if we analyzed carefully, the increments in cross production versus time for section 329. As evident, the new
correlations are insignificant when the injectors were added. In wells have added new reserves to the section without robbing
both cases, by adding an injector, the correlation coefficient any production from old wells.
increased by less than 0.002. These two connections represent In contrast, the evaluation of section 5 shows strong
the smallest increase among all pairs. In Figure 5, we connectivities. Figure 10 shows that there are east-west
differentiated a weak connectivity as a dashed-line connecting connectivities trends in section 5, especially in middle part of
the injector with the producer. Here we defined a weak the section. This is consistent with the field observation about
injector as an injector that increased the cross correation lower the response of producers when injection rates at injectors
than 0.005. If we use this threshold to eliminate any injectors, were altered. Although section 5 was also good primary
the revised map is shown in Figure 6. If a minimum producer, it may not be a good candidate for infill wells.
increment threshold were applied, the weak injectors would be By applying the above procedure, we can obtain the idea
rejected, thus the connectivities dissappeared. For a real field about the trends in reservoir connectivities. This information,
data, it is necessary to apply this threshold since noise in data along with other available production data, such as flowing
can be deceiving. pressures, water cut, etc., and data from other sources (for
Since communications between injectors and producers are example geological data and logs), can be utilized to optimize
localized within each compartment, it is possible to infer the performance of water flood, such as determining the
possible barriers in the reservoir from the lack of connectivity, possible locations to drill infill wells, water allocations, etc. It
see Figures 5 and 6. We can see that the inferred barriers are can also be utilized for reservoir modeling purpose.
good approximations to the real barriers.

Conclusions
Application to real field data Based on this study, following conclusions can be drawn:
Observing that the procedure worked for synthetic case, we A new methodology for quantifying connectivity between
tested the procedure into real field data. As previously injectors and producers in a mature water flood is
mentioned, for field cases, the injected water can flow outside proposed.
the lease boundary, resulting in lower cross correlations The methodology identifies strong connectivities and
between injectors and producers. Furthermore, field data potential barriers in the field.
contains noise due to inaccuracy in measurements. Due to The method is validated by applying it to a synthetic as
these factors, the cross correlations are generally lower than well as to field data.
the synthetic case. For the synthetic case, the cumulative cross
correlation values are very high (greater than 0.9).
Considering these factors, we assumed that the meaningful References
cumulative cross correlation for the field data utilized here 1. Heffer, K.J., Fox, R.J., and McGill, C.A.: Novel Techniques
was 0.5, i.e., we accepted the set of injectors if their ShowLinks between Reservoir Flow Directionality, Earth
cumulative cross correlation was greater than this threshold. Stress, Fault Structure and Geomechanical Changes in Mature
For threshold of accepting the first adjacent well, we use 0.25. Waterfloods, paper SPE 30711 presented at the SPE ATCE,
We also did not accept weak connectivities, i.e., if the Dallas, TX, October 1995.
increments in cross correlations were lower than 0.005. 2. Refunjol, B.T., and Lake, L.W.: Reservoir Characterization
The field data were taken from North Robertson Unit Based on Tracer Response and Rank Analysis of Production and
Injection Rates, presented at the 4th International Reservoir
section 329 and Section 5 in Gaines County, Texas, see Characterization Technical Conference, Houston, TX, March 2-
Figure 7. Production in this field began in the early 1950s 4, 1997.
with 40 acres spacing. The water flood was initiated in 1987 3. Jansen, F.E.: Reservoir Description from Production Data,
on a converted 20 acres spacing. All producers were on Ph.D. dissertation, University of Tulsa, 1996.
artificial lift, where the wellbore pressure was considered 4. Panda, M.N., and Chopra, A.K.: An Integrated Approach to
fairly constant. This made the unit suitable for the above Estimate Well Interactions, paper SPE 39563 presented at the
procedure since changes in production rates at producers were SPE India OGCE, New Delhi, India, February 1998.
likely caused by changes in injection rates at injectors only,
not by changes in wellbore pressure.
Figure 8 shows the map of section 329. The connectivities
between injectors and producers illustrate lack of direct
4 SOERIAWINATA AND KELKAR 52224

TABLE 1-ILLUSTRATION FOR UTILIZING THE MINIMUM


INCREMENT THRESHOLD

Producer # Supporting Cumulative cross correlation


Injectors value when the injector at the
2nd column is added
Prod-1 Inj-1 0.3500
Inj-2 0.6000
Inj-3 0.6020
Inj-4 0.6025
Prod-2 Inj-1 0.3100
Inj-2 0.5500
Inj-3 0.5510
Inj-4 0.6000

TABLE 2-THE LIST OF PRODUCERS AND THEIR


SUPPORTING INJECTORS FOR SYNTHETIC CASE

Producer # Supporting Cumulative cross correlation


Injectors value when the injector at the
2nd column is added
201 101 0.8663
102 0.9892
202 106 0.7706 Figure 1-A typical five-spot pattern where one producer is
107 0.8195 surrounded by four injectors.
104 0.8831
108 0.9060
103 0.9427
203 108 0.7785
104 0.8531
106 0.8773
103 0.9070
107 0.9599
112 0.9802
204 109 0.8166
105 0.9745
113 0.9833
107 0.9847
205 106 0.7732
107 0.8633
108 0.9122 Constructive Interference
112 0.9339
206 112 0.8298
108 0.8948
106 0.9291
207 114 0.9374 =>
110 0.9788
208 114 0.9297
110 0.9689
115 0.9731
118 0.9757 Destructive Interference
113 0.9763
209 111 0.9010
116 0.9248 =>
120 0.9629
210 114 0.8992
117 0.9351
110 0.9530
118 0.9589
211 114 0.8767 Figure 2-Constructive and Destructive Wave Interference analogy
115 0.8825 for Superposition effects in reservoir.
119 0.9735
117 0.9842
118 0.9858
110 0.9862
212 120 0.8536
116 0.9047
111 0.9873
52224 RESERVOIR MANAGEMENT USING PRODUCTION DATA 5

101 102 103 104


201
203
202
105 106 107 108
Producer x 205 206
204
112
109 110
111
207 208 209
113 114 116
115
210 211 212

117 118 119 120

Figure 3-Pre-defined search neighboorhood for Producer x. The Figure 4-Areal map of the synthetic reservoir, triangles represent
black triangles represent the injectors that are inside this Injectors, circles represent producers.
producer pre-defined region, while white triangles represent
injectors outside the neighboorhood.

101 102 103 104 101 102 103 104

201 202 203 201 202 203

105 106 107 108 105 106 107 108

204 205 206 204 205 206

109 110 111 112 109 110 111 112

207 208 209 207 208 209

113 114 115 116 113 114 115 116

210 211 212 210 211 212

117 118 119 120 117 118 119 120

Figure 5-Connectivities between Injectors and Producers in Figure 6-Connectivities between Injectors and Producers in
synthetic field, solid lines represent significant correlations, synthetic field if the weak correlations were eliminated.
dashed lines stand for weak correlations. The bold lines are the
reservoir barriers inferred from the lack of connectivity.
6 SOERIAWINATA AND KELKAR 52224

Section 329

North Robertson Unit


3501 3502 3503 3504 Section 329

Injector
3517 3518 3519 3520
Producer

Infill producer
3505 3506 3507 3508 3802
Infill injector

3533 3523 3532 3524


3521 3538 3522

37713 50832 35710

Section 5 3509 3510 3529 3511 3536 3512 3691

3525 3537 3526 3535 3527 3534 3528 3604

30208 56853 23462

3513 3514 3515 3516 3602 3603 3701

3207
3531
1404 2217 2218

Figure 7-The sketch map of North Robertson Unit, Gaines County, Figure 8-Section 329 of North Robertson Unit. There is lack of
Texas. connectivity between Injectors and Producers. The numbers
inside the boxes are the cumulative oil production for the infill
wells in Stock Tank Barrels.

25000

301 302 401 492


North Robertson Unit
402
Section 5
20000 2604 307 308 406
Injector
monthly oil rate (STB/month)

Producer
303 304 403 404
15000

803 103 207 208

10000 293
101 201 202

203

804 104 209 210


5000

204 205 206


102

295
0
1006 1007 1008 1009
94/01 95/01 96/01 97/01 98/01
time (years)

Figure 9-Aggregate oil rate of NRU Section 329. The solid line Figure 10-Section 5 of North Robertson Unit. The bold lines are
represents total oil rate from all producers including the infill the possible barriers inferred from lack of connectivity.
wells, the dashed line represents total oil rate without the infill
wells.

Вам также может понравиться