SPE 138229

Performance Model Analysis for Candidate Recognition

J.S. Tan, SPE, and Y. Del Castillo, SPE, Schlumberger DCS, and R.D. Reese, SPE, Schlumberger, and C. Pinzon, SPE

Copyright 2010, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Latin American & Caribbean Petroleum Engineering Conference held in Lima, Peru, 1–3 December 2010.

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
The main objective of the project was to increase operation efficiency through monitoring production and injection performance in
18 Permian Basin fields (waterflood and natural depletion). A performance model (PM) technique has been developed to
efficiently analyze massive amounts of production and injection data from thousands of wells. Using the PM technique, under-
performing wells and patterns were rapidly identified and ranked for workover opportunities. Additionally, non-responsive
injection areas (composed of several patterns) were also identified to enhance injection efficiency. The PM technique was
implemented in 18 fields in the Permian Basin. The largest field having more than 1300 wells was evaluated.

The PM technique described here is based on a modified heterogeneity index (MHI) concept. This improvement was necessary
since calculations using traditional heterogeneity index (HI) skewed the results and incorrectly quantified the performance
comparison. The MHI has successfully corrected and overcome traditional HI weakness.

The PM is an improved candidate recognition technique that uses binary codes and personality concepts to effectively monitor
wells and injection patterns. The personality characterization process creates and uses several interpretation scenarios to identify
problematic wells, patterns or both.

PM methodology and personality concepts are discussed in detail and field implementation results are presented.

Introduction
Traditional Heterogeneity Index.
R. D. Reese (1996) initially showed the method of heterogeneity index in a completion ranking system:

t max ⎡
Fluid well ⎤
HI Fluid = ∑ ⎢ ⎥ , ..................................... (1)
t =0 ⎣⎢ Fluid avgwell ⎦⎥

where HIFluid is the heterogeneity index for any type of fluid production ratio. Fluid may be oil, water, gas, oil equivalent, total
liquid, gas/oil ratio, or water cut and may consist of either “rate” or “cumulative” numbers (Reese 1996).

According to Eq. 1, a well with no heterogeneity (an average well for the field or formation) has an HI value equal to exactly
1. When the HI is examined over n (periods of time), a heterogeneity signature is obtained (Reese 1996).

To illustrate the use of heterogeneity indexing, consider a well with an oil production rate, qo, at an instance of time t, assume
during that instance of time t, the average oil production rate of all the wells in that field is qo avg. If the well is performing above
average, the HI value at that instant of t shall be greater than 1. If the HI value is normalized to zero, a positive HI value indicates
2 SPE 138229

an above average condition. On the other hand, a negative HI value indicates a below average condition. Finally a zero HI value
indicates an average condition. Therefore, introducing the normalization concept, the Eq. 1 becomes:

⎡ Fluid well
t max ⎤
HI Fluid = ∑ ⎢ − 1⎥ , ............................... (2)
⎢ Fluid avgwell ⎥⎦
t =0 ⎣

Rearranging Eq. 2 gives,

t max ⎡
Fluid well − Fluid avgwell ⎤
HI Fluid = ∑ ⎢ ⎥ , ................ (3)
t =0 ⎢⎣ Fluid avgwell ⎥⎦

However, a close examination of Eq. 1 indicates that the HI values are bound between 0 and infinity. Similarly, the
normalization of Eq. 3 indicates that the HI values are bound between -1 and infinity. This is because Fluidwell values are bound
between 0 and infinity.

If the above observation is ignored, the heterogeneity signature calculated from the summation over n periods of time will tend
to amplify HI values when the well has above-average conditions and curtail the HI values when the well has below-average
conditions. Consequently, this will skew the analysis result and prevent quantitative comparison. Therefore, a modification to Eq.
3 is necessary to correctively quantify the heterogeneity index comparison (Del Castillo et al. 2009).

Modified Heterogeneity Index.

An introduction of Fluidmax and Fluidmin values into the MHI equation has successfully enabled quantitative comparison.

t max
⎡ Fluid well − Fluid avgwell ⎤
MHI Fluid = ∑ ⎢ ⎥ , ....... (4)
t = 0 ⎣ Fluid max well − Fluid min well ⎦

Eq. 4 shows that MHI values are always bound between -1 and 1 at every instance of t.
The values of Fluidwell have the characteristic of existing between 0 and infinity. At any instant of t, the Fluidwell value could be
equal to or greater than Fluidmin well (Fluidwell ≥ Fluidmin well). As Fluidwell approaches a value of 0, Fluidmin well value also approaches
0, reducing Eq. 4 to:

− Fluid avgwell
MHI Fluid = , ..................................... (5)
Fluid max well t

Additionally, since Fluidmax well is always greater than Fluidavg well, the MHI value is always greater than -1.

On the other hand, the Fluidwell value could be equal to or less than Fluidmax well (Fluidwell ≤ Fluidmax well). As Fluidwell approaches
to infinity, it can be replaced in Eq. 4 by Fluidmax well leading to the following expression:

(Fluid max well − Fluid avgwell ) ≤ (Fluid max well − Fluid min well ) , (6)
SPE 138229 3

Eq. 6 clearly shows that the denominator of Eq. 4 is always greater than the numerator; this is because Fluidavg well is always
greater than Fluidmin well. Therefore, the MHI value is always less than 1.

Table 1 shows a comparative calculation of HI with traditional and modified methods. Note that MHI method clearly limits the
values between -1 and 1 while the traditional HI method has limiting values between -1 (lower limit) and infinity (upper limit).
Also notice that the total sum of HI values is equal to zero in both methods.
Table 1 —Heterogeneity Index comparison.

Problem Analysis and Solution

Since the MHI value in Eq. 4 gives a quantitative comparison, it is used as a production indicator for oil, water, gas, water
injection and gas injection rates in the PM analysis. For a given period of study time, a positive MHI value at the end of the time
period means the well is outperforming the average well; and a negative MHI implies an underperforming well. Similarly, the
MHI value can be calculated for a pattern (using previously assigned geometric factors); as before, a positive MHI value indicates
the pattern is outperforming the average pattern, and a negative MHI value implies an underperforming pattern.

Quadrant Analysis.
Cross-hair scatter plots from a production analysis software are used to graphically present the results of the MHI calculations;
these plots are capable of showing the traces of heterogeneity signatures over a period of study time.

The use of cross-hair scatter plots enables analysis of 3 variables simultaneously. Fig. 1 shows typical MHI results for oil,
water production and water injection rates at pattern level. The left plot shows the water MHI vs. water injection MHI and the right
plot shows water MHI vs. oil MHI. Using both graphs significant pattern performance can be identified / analyzed in each
quadrant.

Patterns in Quadrant 1 inject more water than the average pattern and also produce more water than the average pattern (left
plot). Additionally, when analyzing these same patterns in terms of oil (right plot), two different groups can be identified: patterns
that produce above the average pattern oil production (i.e. S-092) and patterns that produce below the average pattern oil
production (i.e. S-052). For the last group, water injection seems to be channeling directly from injectors to producers.

Patterns in Quadrant 2 inject more water and also produce less water than the average pattern (left plot). Also, as before, when
analyzing these same patterns in terms of oil (right plot), two different groups can be identified. Water injection has a positive
impact on only some of these patterns (ones above the average pattern oil production). However, some of these patterns still
perform below average in either oil or water production in spite of having above-average water injection. The injected water in
these last patterns is possibly being lost to the formation as these patterns do not have high water production.

Patterns in Quadrant 3 inject less water and also produce less water than the average pattern (left plot). Also, as before, when
analyzing these same patterns in terms of oil (right plot), two different groups can be identified. The well performing oil patterns
are probably getting support from nearby injectors (outside the immediate pattern). On the other hand, patterns that are performing
below average in terms of oil, water or water injection present workover opportunities for producers and injectors as well as target
areas for increases in water injection.
4 SPE 138229

Patterns in Quadrant 4 inject less water and also produce more water than the average pattern (left plot). The produced water is
probably coming from nearby injectors outside the immediate patterns.

Although the quadrant analysis technique shows significant pattern performance, it is not an efficient process for analyzing
large amounts of production and injection data. Another limitation is that the technique can analyze only three variables (in this
case qo, qw, and iw). Fig. 2 shows very crowded MHI results for quadrant analysis. To find a way to efficiently analyze these results
is quite challenging especially in fields that have gas production and gas injection in addition to the previous three variables (qo,
qw, qg, iw, and ig).

The introduction of personalities as part of the characterization process and also the development of the PM (built inside the
production analysis software) have significantly simplified the workflows and allow the engineer to focus more on interpretation
and implementation of projects based on the results of the analysis.

Personalities Analysis.
a) Pattern Level.
A simplified pattern personality analysis shows the relationships of three variables (qo, qw, and iw) which are initially developed
and summarized into eight types of pattern personalities as shown in Table 2. From the table, a variable performing above average
is assigned “HI” and coded as 1, and a variable performing below average is assigned “LO” and coded as 0. Since each variable
can be assigned 0 and 1, taking base 2 of three variables, 23 will equal 8. This means that there are eight characteristics from the
combinations of the three variables, which are defined as pattern personalities.

Table 2 —Eight pattern personalities from three variables.

The first type of pattern personality is called “lazy” because water is injected at rates below the average pattern, and oil and
water production are also produced below average pattern rates. The consequence of low injection is low production; therefore,
these patterns are categorized as “lazy” patterns. This identification indicates an opportunity to further increase injection in these
patterns. The cause of low injection must be investigated to determine if the injectors are impaired by water supply or facilities
issues or if the producers in these patterns have developed positive skin.

The “waster” pattern personality strongly indicates that the water injected into the patterns at rates above those of the average
pattern does not increase the oil production of the patterns (qo is performing below the average pattern). The fact that these patterns
also have below average water production suggests that the injected water is being wasted in the formation. A typical diagnostic is
to check for perforation conformance and geological features surrounding the producers and injectors in the patterns.

Patterns with “thief” personalities may indicate that water is being stolen from other area. These patterns are characterized by
high water production and low water injection compared to the average pattern.

Patterns that are injecting and producing water above the average pattern, but producing oil below the average are characterized
as having “short cutter” personalities; the injected water is taking shortcuts from injectors to producers and is not effectively
contributing to oil production. In contrast to the “thief” pattern, there could be a direct communication between the injectors and
producers within the “short cutter” patterns. Production logging of the producers and injectors may provide a better understanding
of these phenomena.
SPE 138229 5

Patterns with “perfect” personalities achieve high oil production performance while water injection and water production rates
remain below those of the average pattern. The “perfect” patterns require the least attention of all pattern types.

“Hard working” patterns have high oil production compared to the average pattern and above average water injection rates.
These patterns have lower water production than the average pattern. Through the analysis of these patterns, an empirical optimal
water injection rate can be determined.

On the other hand, the “celebrity” patterns have low water injection but high oil production. They also have high operating
overhead because of over production of water. In summary, these patterns have high oil productivity at the expense of high water
production with minimum hard work in the water injection. The above average water production in “celebrity” patterns may come
from strong injectors outside the patterns. Controlling the injection rates from these nearby injectors or performing water control
techniques on the “celebrity” producers may reduce the water problem.

Finally, “hyperactive” patterns have a great deal of energy and deliver high oil and water at above average pattern rates; the
water injection rate is also above that of the average pattern. This may suggest that the injectors in these patterns do not need
“hyper” water injection activity and that the producers may be candidates for water control.

The eight-pattern personality types just described are simplified versions of PM pattern level personaly analysis that is based
on three variables (qo, qw, and iw). When taking into consideration two additional gas variables (gas production, qg and gas
injection, ig) the analysis will generate 32 pattern personality types (25). Table 3 summarizes these 32 pattern personalities from
binary code combinations of five variables (qo, qw, qg, iw, and ig).

The pattern personality analysis in PM is applicable to fields that exhibit consistent pattern shape. However, if fields have
either irregular or no apparent pattern shape, a well level personality analysis technique can be used.

b) Well Level.
In this study, the PM analysis was run separately on producers and injectors. Each run generates well personalities.
Superimposing the producer and injector personalities on the base map delivers various scenarios for engineering interpretation.

Producer Personalities.
There are three variables related to producers – oil, water, gas (qo, qw, qg), therefore, there are eight producer personalities (23).
Table 4 summarizes the producer personalities, which can be subdivided into poor performance and superior performance.

Table 4 —Eight producer personalities from three variables.

The under-performing producers produce oil at rates below the average producer. They are called “lag” producers and can be
divided into four subgroups:

The “lazy” producers have below average production of oil, water, and gas. This group may hold hidden potential for workover
opportunities.

The “lag high gas” producers are producing gas at rates above that of the average producer. Normally, these producers are gas
wells or they may have a perforation near the gas cap. Expansion of gas cap or depletion of oil zone, or both, may have changed
the gas/oil contact level. Gas coning may have caused this problem.
6 SPE 138229

The “lag high water” producers are producing water at rates above that of the average producer. These producers may have
water coning or channeling problems or high water rates caused by a change in the oil/water contact due to waterflooding.

The “troublesome” producers are producing gas and water at rates above those of the average producer. These producers
present challenging workover opportunities and are possibly candidates for production termination.

Alternatively, superior producers are producing oil at rates above that of the average producer. Called “lead” producers, they
can be divided into four subgroups.

The “perfect” producers have oil production rates above that of the average producer and require less attention than other
groups.

The “lead high gas” producers are producing oil and gas at rates above those of the average producer. Nearby gas injectors with
strong injection activity may be recycling the injected gas.

The “lead high water” producers are producing oil and water at rates above those of the average producer. Nearby water
injectors with strong injection activity may have direct communication channels.

The “hyperactive” producers are producing oil, water, and gas at rates above those of the average producer. Further
investigation of their characteristics may provide valuable understanding for field operations.

Injector Personalities.
The injection operation in the field may have three injection scenarios – water injection, gas injection, and water alternating gas
injection (WAG).

In the fields with WAG injection, there are two variables related to injectors: water, and gas (iw and ig, respectively). Modeling
will create four injector personalities (22). Weak injectors inject water and gas at rates below the average, and strong injectors
inject water and gas at rates above average. Meanwhile, if water injection is below average and gas injection is above average,
these WAG injectors are called “lag winj lead ginj”. On the other hand, those characterized as “lead winj and lag ginj” inject water
at above the average rate and inject gas at rates below the average injector. Table 5 summarizes the injector personalities.

Table 5 —Four injector personalities from two variables.

In fields where only water is injected (waterflooding) and there is no gas injection activity, only one variable (iw) is considered.
This gives two injector personalities (21): “00” and “10” – “lag winj” and “lead winj”, respectively. The binary code combinations
“01” and “11” in Table 5 are not applicable.

In gas injection fields with no water injection activity, only one variable (ig) is considered, which gives two injector
personalities (21): “00” and “01” – “lag ginj” and “lead ginj”, respectively. The binary code combinations “10” and “11” in Table 5
are not applicable.

c) Field Level.
The MHI calculation in Eq. 4 can be slightly modified for “field level” PM analysis. This technique allows comparison of
several fields with similar petrophysical and depositional characteristics within a regional area. This approach helps decision
makers determine whether to commit more resources and capital on operating fields in competitive regions and also aids in
ranking assets.

The synthetic well concept is used in the field level PM. Since field sizes are different, the average production and injection of
a field are calculated by dividing the total production or injection by the total number of producers or injectors at the instant t. Eq.
7 shows the “field level” MHI calculation:
SPE 138229 7

⎡ ⎤
t max
⎢ Fluid FieldPerWell − Fluid AvgAllFieldPerWell ⎥
MHI FieldFluid =∑ , (7)
⎢
t = 0 Fluid max AllFieldPerWell − Fluid min AllFieldPerWell
⎥
⎢⎣ ⎥⎦

Project Design Considerations

The use of production analysis software has played a large role in achieving the objectives in the Permian Basin project.

There are 18 fields in the Permian Basin with large amounts of monthly production and injection data. Effectively analyzing these
data for field surveillance is quite challenging. An important design consideration was the existing operator interface to the data
server, which mainly limited performance of the production software.

To address the performance issue, which would be further affected by the intensive MHI calculations required at every instant of
time on each variable, 18 standalone databases were created with structure to match that of the existing master source database.
The main requirement was to populate the standalone databases with data relevant to a specific field. A script was designed to
update data to the standalone databases from the data server. With the construction of standalone databases, the production
analysis software runs faster and with more stability since it no longer needs project filters to extract the data relevant to the
specific field.

To make these standalone databases “ready” for the PM, various background tasks were designed, performed and documented:
Report templates with instructions to run the PM for pattern and well level techniques, calculated variables, user functions, SQL
normal queries, SQL union queries, and table queries were setup within the software workspaces and the standalone databases.

Fig. 3 shows a sample of the software workspace created for running either pattern level PM or well level PM. Figs. 4-6 show the
personality characterization process and binary codes used for either pattern level or well level analysis (producers or injectors).
More details on the procedure are specified in the pending US.patent application (Del Castillo et al. 2009).

Macro level PM Applications

Using the PM at pattern level, we were able to easily extract the results and screen captures of the different pattern personalities.
Fig.7 shows field performance within a specific time period using the simplified three variables (qo, qw, and iw). The figure shows
many “lazy” patterns or non-responsive injection areas concentrated in the south east of the field. These areas represent
opportunities for production optimization either through increase in injection or through workover operations (i.e., stimulation on
producers).

Micro level PM Applications

One of the advantages of micro level PM application is the ability to easily spot abnormal performance and plan immediate
actions. This section shows a typical scenario.

Fig. 8 shows a “water waster” pattern personality of Pattern ABC. The pattern personality code is 00010, and these five digits of
binary numbers represent qo, qw, qg, iw, and ig respectively. The code signifies above average water injection but below average
production of oil, water and gas. This may indicate that water injection in this pattern is not contributing to oil production and
water is being lost into the formation. A further look at the producers in the pattern indicates two of the abandoned producers are
perforated at similar depths as the injector. The injected water is probably channeling to the abandoned producers and is being lost
in the formation, thus, not effectively flooding the pattern. Therefore, the recommendation is to reduce the water injection rate in
this pattern.

Summary and Conclusions

1. Performance Model is a new approach in surveillance which uses binary codes and personality concepts to effectively monitor
producers, injectors and injection patterns
2. Performance Model technique is applicable to mature waterfloods and to conventional candidate recognition projects.
3. Performance Model presents an improved Heterogeneity index technique and also a new approach for production analysis
based on personality characterization.
4. In terms of operational efficiency and production optimization, this technique enables effective monitoring of producers and
injectors, identification of non-responsive injection areas for water injection optimization and ranking of workover candidates.
8 SPE 138229

5. In terms of flexibility, the software interface built specifically for well and pattern level analysis was successfully
implemented in 18 different west Texas waterflood projects.

Nomenclature
HIFluid = Heterogeneity index, dimensionless
MHIFluid = Modified heterogeneity index, dimensionless
MHIField Fluid = Modified heterogeneity index for field level, dimensionless
Fluidwell = Oil production rate per well, STB/D
Fluidavg well = Average oil production rate per field, STB/D
Fluidmax well = Maximum oil production rate per field, STB/D
Fluidmin well = Minimum oil production rate per field, STB/D
FluidField Per Well = Average oil production rate per field, STB/D
FluidAvg All Field Per Well = Average oil production rate per all fields, STB/D
Fluidmax All Field Per Well = Maximum oil production rate per all fields, STB/D
Fluidmin All Field Per Well = Minimum oil production rate per all fields, STB/D
qo = Oil production rate per well, STB/D
qo avg = Average oil production rate per field, STB/D
qw = Water production rate per well, STB/D
qg = Gas production rate per well, MSCF/D
iw = Water injection rate per well, STB/D
ig = Gas injection rate per well, MSCF/D

Acknowledgements
The authors thank the following people for the consecration of the project and suggestions that have improved this manuscript:
Antonio Acunzo, Juan Carlos Palacio, Randy Vaal, Robin Heim, Chip Corbett, Kim Hemsley, Philip Evbomoen and Randy Utech
of Schlumberger and Tom Bundy, Dan Warner, Steve Van Howe, Paul Schulz, Josh Viets, Chris Appel, Martin Perez, Doug
Pecore, Nadia Chalton, Aaron Fuhr, Cesario Torres, Bethel Strawser, Tim Harrington, William Vargas, Hien Bui, Judy Wang, Jeff
Hurley, Sandi Richardson and Pat Guthridge of ConocoPhillips.

References
Del Castillo, Y., Tan, J.T, and Reese, R.D. 2009. Statistical Determination of Historical Oilfield Data. U.S. Patent Application No.
12,361,623 (pending).

Reese, R.D. 1996. Completion Ranking Using Production Heterogeneity Index. Paper SPE 36604 presented at the SPE Annual
Technical Conference and Exhibition, Denver, Colorado, 6–9 October.
SPE 138229 9

Figure 1 — Typical MHI results for qo, qw, and iw rates.

Figure 2 — Crowded MHI results for quadrant analysis.

10 SPE 138229

Table 3 —Thirty-two pattern personalities from five variables.

SPE 138229 11

Figure 3—Sample of software workspace for pattern and well level PM (proprietary software).

Figure 4 —Pattern level PM personality codes (refer as Figure PM.6 in Table 6).
12 SPE 138229

Figure 5—Producer well level PM personality codes (refer as Figure PM.Prd6 in Table 7).

Figure 6—Injector well level PM personality codes (refer as Figure PM.Inj6 in Table8).

Figure 7—Macro level performance model application from pattern personality analysis
SPE 138229 13

Figure 8—Micro level performance model application from pattern personality analysis

Figure 23—Production software interface used for running either pattern level or well level PM.