SPE 166111

Data Driven Analytics in Powder River Basin, WY

Mohammad Maysami, Razi Gaskari, Intelligent Solutions, Inc., Shahab D. Mohaghegh, Intelligent Solutions, Inc.
& West Virginia University

Copyright 2013, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in New Orleans, Louisiana, USA, 30 September–2 October 2013.

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

In the past few years, we have observed the introduction of smart technologies that adapt themselves to the specific needs of
individual users. There are many mobile and web-based services with learning capabilities that play the role of a personal
assistant in our daily life. The foundation of this new class of services is a paradigm shift from intensive computational
modeling and simulation of complicated phenomena toward data driven analytics. The oil and gas industry with the
uncertainties convoluted into our measurements and understanding of the subsurface should not be excluded from this recent
paradigm shift. Data driven analytics have proven to be a powerful alternative to conventional numerical and analytical
solutions. In their advanced form, data driven technologies may be used as comprehensive management tools of oil and gas
assets.

In this paper, we study Hilight field in Powder River Basin, a mature field with large number of wells. Lack of sufficient
dynamic data such as flowing pressure for mature fields is common among these types of fields. Conventional data analyses
impose a challenge in the absence of time-variant field measurements in addition to production history. Acquisition of a
comprehensive data set for oil and gas assets, in general, is a costly luxury that is not financially feasible for all investment
budget ranges. Data-driven approach along with pattern recognition techniques can introduce a potential solution to this
challenging task and extract practical and valuable insights which can be vital to identification, planning and developments of
assets and plays.

In this work, data from nearly 400 wells has been analyzed. Data from completion and workover was partially available. Well
logs for only 15 wells is accessible providing less than 10% petrophysical data attributes over the entire well sets. Available
production rate history for 185 wells starts from June 1969 and extends until April 2012. The information value of this
dataset is investigated through a multi-step workflow. The workflow includes reservoir delineation and geological modeling,
volumetric reserve and recovery factor estimations, production decline curve analysis, fuzzy pattern recognition (FPR)
analysis and key performance indicators (KPI) analysis. FPR analysis provides time-laps spatial patterns, enabling us to
qualitatively study the reservoir depletion and fluid flow in Hilight field. The result of these analyses has been used to
identify the depletion distribution over time and sweet spots for infill locations. KPI analysis identifies relative influence of
different parameters on hydrocarbon production. Top-Down Model is developed and used for field development planning and
economic analysis on proposed new wells. The workflow has a minimal computational footprint compared to conventional
methods. It has been demonstrated how these data driven techniques can be employed as a guide toward an improved
reservoir management and planning.

Introduction

Data is the most valuable element in solving scientific problems independent of the nature and complexity of problem itself
or even the approach and techniques used to obtain the solution. The theoretical branch of science appeared when the
generalization and modeling replaced empirical science and made it possible to describe simple phenomena. Data has been
used to develop, propose, verify and validate physical and mathematical theories ever since. With the technology
advancement in the last few decades and improvements in computing power and resources, a paradigm shift from theoretical
to computational problem solving was inevitable. Computational paradigm similar to earlier techniques revolves around data
2 SPE 166111

and modeling and simulation of complex phenomena with a goal of minimizing the difference error between estimated output
results and the observations and measured data from those phenomena (1; 2; 3).

In today’s world, however, the gigantic amount of data from all around the world calls for a new paradigm shift toward data-
intensive and data mining to tackle various problems in different scientific and engineering fields. The need for such problem
solving techniques becomes more visible with a glance at intelligent and targeted marketing and user-specific services and
suggestions. These recent features of our life are made possible based on all the collected historic data from multiple users on
different digital and web-based services.

Data acquisition, especially in oil and gas industry, is a costly and risky investment due to many unknowns associated with
the subsurface. The uncertainties rise in both information content and positive use of data in securing more production. Such
characteristics make the data acquisition a less attractive choice when smaller financial assets and thus less risk capacities are
present. Mature fields, as a particular example, commonly pose a challenge when evaluated for future planning and
revitalization. One of the main obstacles is the lack of any type of data other than production rates. The absence of sufficient
data is a limiting factor for many conventional reservoir simulation and management methods. This disadvantage, however,
does not stop data driven modeling techniques from providing necessary reservoir management tools based on a minimal
requirement for data availability.

Data driven analytics, unlike the conventional reservoir simulation and management techniques, focuses on hard measured
data from the asset and attempts to create a logical insight on the subsurface fluid flow in the reservoir free of any predefined
functional form. Complimentary data in addition to production rates and well logs only adds to accuracy of the model. Many
studies have shown the effectiveness and accuracy of artificial intelligence and data mining (AI&DM) techniques in
modeling and managing oil and gas assets around the world (4; 5; 6; 7). The benefits of data driven modeling become more
visible when reservoir management and future planning is the final goal. Data driven models with their minimal
computational footprint in comparison to conventional modeling methods, make it feasible to analyze multiple scenarios
within a practical time frame (2).

In this article, we study the Powder River basin in Wyoming using data driven techniques for modeling, evaluation and
analysis. The available data from the field is qualitatively analyzed by various techniques such as fuzzy pattern recognition
and the fluid drainage over the years is probed. Next, the economic impact of new wells in future is studied using the
developed data driven model for the field.

Hilight Field in Powder River Basin

This study was performed on Hilight Field in the Powder River Basin about 20 miles south of Gillette, Wyoming. Hilight
Field is a large stratigraphic trap in the Muddy sandstone that was discovered in 1969. Waterflooding was initiated in the
early 1970s and expanded through the late 1990s. Raw data was acquired from IHS Energy’s database. Initial review of the
data revealed that there are a total of 396 wells in Hilight field and production rate data for 185 wells is available. Well logs
for only 15 wells are available providing limited petrophysical data over entire reservoir.

A geographical distribution of wells, color-coded based on their type and availability of data attributes is presented in Figure
1. Wells with production rate data are shown as filled orange circles and the wells with well log data are marked by red rings.
Green circles represent the wells producing from different formations. Non-vertical producing wells and geographically
outlier producing wells are shown by blue and purple circles, respectively. Note that most of provided well logs fall into west
side of the field. In order to increase certainty and integrity of the analysis and extracted information content, we have picked
160 wells for our analysis excluding wells completed in different formations or remote locations comparing to majority of
wells in addition to non-vertical producers. Figure 2 depicts a field development timeline where the wells are clustered based
on their date of first production. This map summarizes the location of initial production wells in the field and how new wells
added up over time until 2012. Circles with darker fill colors represent older production wells with earlier date of first
production (DOFP).

Data Driven Analytics

The techniques used in our analysis and workflows are based on integration of reservoir engineering with Artificial
Intelligence and Data Mining (AI&DM) technology. Following is a list of processes that were implemented in the study of
Hilight field as steps of data driven analytics workflow.

 Hilight field dataset review and preparation:

– Well Logs for certain wells and Reservoir Characteristics (Petrophysical Properties)
SPE 166111 3

– Production history
– Completion details
– Workovers and Stimulations

 Delineation and Volumetric Calculations (on a per-well basis)

 Fuzzy Pattern Analysis to identify hidden patterns and trend

– Well Quality Analysis and Fuzzy Trend Analysis
– Key Performance Indicators Analysis (KPI) to study relative influence of different reservoir properties as
well as operational constraints.

 Field-Wide Fuzzy Pattern Recognition to qualitatively identify field development strategies:

– Well Performance during fixed interval from date of first production
– Remaining reserves as a function of time
– Recovery factor and reservoir depletion

 Top-Down Model Development:

– Forecast the production of proposed wells
– Economic Analysis of Proposed New Wells

The first and most fundamental step in a data driven approach is data compilation, cleansing, quality control. Preparation of a
comprehensive spatio-temporal dataset which can represent the fluid flow in the field is the basis of the analysis (3). In this
step, production data along with average reservoir properties for each well based on provided well logs is compiled into the
dataset. Well specifications such as coordinates, perforation depths, and workover details are also integrated to secure more
information in the dataset.

HILIGHT Field, WY ‐ All Wells
All Wells Wells with Production Well Logs Non‐Vertical Producers Producing From Other Formations Outlier Producers
44.05

44

43.95

43.9
Latitude

43.85

43.8

43.75

43.7

43.65
‐105.5 ‐105.45 ‐105.4 ‐105.35 ‐105.3 ‐105.25 ‐105.2 ‐105.15
Longtitude

Figure 1 - Map of all wells in Hilight field dataset, divided into categories: Note the location of wells with production (orange circles)
as well as provided well logs (red rings).
4 SPE 166111

Date of First Production
All Wells DOF Before 1970 DOF 1970‐80 DOF 1980‐90 DOF 1990‐2000 DOF 2000‐12
44.05

44

43.95

43.9
Latitude

43.85

43.8

43.75

43.7

43.65
‐105.5 ‐105.45 ‐105.4 ‐105.35 ‐105.3 ‐105.25 ‐105.2 ‐105.15
Longtitude

Figure 2 - Map of all wells in Hilight field dataset, color-coded based on their date of first production: Darker fill colors represent
wells with earlier dates of first production.

One of the first processes of the workflow on the dataset is performing delineation and geostatic modeling. Delineation
process is a specific implementation of Voronoi graph theory (5; 8) to generate so called Estimated Ultimate Drainage Area
(EUDA). Volumetric reserve calculation as a common reservoir engineering practice is the next step after delineation. The
geological model, also known as geo-cellular model, static model or earth model, includes the major reservoir characteristics
that are used as the basis of all hydro-dynamic models including the reservoir models (2; 3). A major advantageous
characteristic of data driven modeling and analytics is that it incorporates field measurements and tries to avoid
interpretations as much as possible. As such, our data driven also uses a geostatic model using only measured reservoir
properties for its analysis. Figure 3 shows an example of such isolation of the production wells as well as the map of depth
for Hilight field as example of geostatistical modeling.

Figure 3 - Delineation of Hilight field (Left) and geological models for depth [ft] (Right).
SPE 166111 5

Decline Curve Analysis (DCA) is a classic Production Data Analysis (PDA) technique based on fitting the recorded
production rates by a mathematical function (4; 9; 10). These fitted curves for each individual well represent the general
decreasing trend of production rate and can be used to predict the future production performance of the well by extrapolation.
Examples of DCA for two individual wells in Hilight field are depicted in Figure 4. DCA provides a starting point in
processing production data and also a basis for calculating Estimated Ultimate Recovery (EUR) (9; 10; 11). Decline curve
analysis, however, falls short in providing meaningful link between production rates and reservoir properties. Therefore,
sensitivity or uncertainty analysis based on measured properties of the field is a missing feature when simplistic techniques
such as DCA or regression analysis are used (12; 13).

Figure 4 – Decline Curve Analysis (DCA) for individual wells in Hilight field.

Fuzzy Pattern Recognition Analysis

Conventional statistical analysis such as cross-plots and regressions frequently fails to find hidden trends and patterns in large
and complex datasets (Figure 5 and (12)). Fuzzy pattern recognition, as a modern and superior technology, compensates for
shortcomings of traditional methods in revealing non-trivial and useful patterns. Fuzzy pattern recognition is a collection of
multi-dimensional classification methods applied to groups of measurements or observations based on artificial intelligence
and fuzzy set theory (14).

In “Well Quality Analysis (WQA)” or “Step Analysis”, a type of fuzzy pattern recognition, multiple fuzzy classes of wells
are defined based on production indicator parameters. In Figure 6, three well quality classes of “Poor, Average and Good” are
defined for cumulative production in Hilight field. Three classes of wells have overlapping regions and thus wells can have
partial membership in more than one class. The well in Figure 6 for example, has 0.71 and 0.29 partial membership in Poor
and Average classes, respectively. Impact of single or multiple attributes, such as reservoir properties, on these defined fuzzy
classes can be demonstrated in step form (Figure 7). It can be concluded from the step analysis shown in Figure 7 that
cumulative production is monotonically increasing with an increase in completion footage or decrease in shot density.
Similar behavior patterns can be easily extracted using this type of analysis for various production indicators and independent
attributes.

“Fuzzy Trend Analysis (FTA)”, a similar analysis where distinct individual wells in the dataset are considered as unique well
qualifiers, extracts hidden trends in complex datasets in the form of continuous curve (as oppose to discrete steps in WQA).
Result of Fuzzy Trend Analysis for same production indicator (cumulative production) and attributes (completion and shot
density) is shown in Figure 8, where the continuous trends support the result of step analysis. A comparison between Figure 7
and Figure 8 versus Figure 5 manifests the potential advantage of fuzzy pattern recognition analysis over statistical analysis
widely used in many data mining analyses.

Furthermore, qualitative and quantitative patterns are extracted over entire field using fuzzy pattern recognition techniques
(4; 5; 6; 7; 9). The field-wide pattern recognition analysis allows dividing the reservoir into regions with different quality
indices in order to indentify hidden and non-trivial distribution patterns of different attributes throughout the reservoir. The
idea behind this type of analysis is to be able to integrate and put all information in perspective and thus draw conclusions
and design field development strategies. Figure 9 shows an example of such analysis for best 3 months of cumulative
production in Hilight field. Performance of the different zones of the field through time can be investigated in this type of
analysis. Figure 10 summarizes time-lapse variations of the field for certain time interval past first date of production for all
6 SPE 166111

the wells in the field. Darker shades represent more productive regions in the field. Note the consistency of the regions over
time. Fuzzy Pattern Recognition analysis can be used to assist the decision making process and designing field development
strategies such as identifying sweet spots (Figure 11).

Completion Shot Density

Figure 5 – Cross Plots of cumulative production versus Completion (Left) and Shot Density (Right) lack of any visible trend.

Figure 6 – Schematic and numeric presentation of fuzzy cluster set for cumulative production.

Completion Shot Density

Figure 7 - Well Quality Analysis (WQA) for completion (Left) and shot density (Right) in cumulative production.

Identification of key performance indicators is one of the most important tasks in understanding any system’s behavior. Most
of the traditional methods evaluate the parameters, one parameter at a time, and mostly use simple regression analysis. The
collective interaction between parameters must be taken into account, for a comprehensive analysis and modeling of any
complex, non-linear and dynamic process. Therefore, one should identify the key performance drivers in a process by
analyzing the parameters in a combinatorial fashion. Key Performance Indicator (KPI) analysis based on fuzzy cluster
analysis has been tested and validated on known problems. This technique outperforms all other techniques, especially when
applied to complex, dynamic and non-linear problems.
SPE 166111 7

While KPI analyses extract valuable information regarding behavior and relationship in data, it allows for more detailed
probing of dataset. Relative influence of different values of certain data attributes is an example of such possibilities. Figure
12 shows an example of such detailed analysis studying effect of different workover injection fluid types on production rate
in Hilight field. KPI analysis can be paired with cluster analysis to go further and look a frequency distribution of best 9
months Cumulative Production for specific value of workover injection fluid (Figure 12).

Completion Shot Density

Figure 8 - Fuzzy Trend Analysis for Completion (Left) and Shot Density (Right) in Cumulative Production.

Figure 9 - Fuzzy Pattern Recognition (FPR) analysis based on cumulative production of best 3 months: Top map of field with fuzzy
patterns of production indicator (Cumulative production of best 3 months) along latitude and longitude is shown and field is divided
to partitions with different quality indices ranging from excellent to poor (Top-Left). 3D view of the field partitioning based on quality
indices is presented (Top-Right). Table is summarizing the average value of production indicator (PI) for each partition (Bottom).
8 SPE 166111

First 3 Months First Year of First 5 Years of First 10 Years

of Production Production Production of Production

Figure 10 - Analyzing time-lapse field performance. Field is divided to partitions with different quality indices ranging from excellent
(darker shades) to poor (lighter shades).

Best 12 Months of Cumulative Production Recovery Factor Remaining Reserve as of

Cumulative Production 2015

Figure 11 - Decision making process (identifying sweet spots) using multiple fuzzy pattern recognition analysis: Warmer and colder
color shades represent better and worse quality indices, respectively.

Figure 12 – KPI Analysis: Relative influence of different workover injection fluid on cumulative production of best 9 months (Left);
Frequency distribution of best 9 months of cumulative production for wells with acid as workover injection fluid (Right).
SPE 166111 9

Top-Down Model (TDM)

Top-Down Model (TDM) is a class of AI-based models based on actual field data (2; 5; 6; 7; 15; 16). The physical laws
governing the fluid flow in the reservoir is deduced from the spatio-temporal dataset rather than being imposed directly.
Classic reservoir engineering practices are used in preparation of the comprehensive dataset to be used in development of
top-down models. TDM has an advantage to be operational with minimal interpreted data and has been tested and verified in
numerous cases (2; 5; 6; 7; 17; 15). Inputs to TDM are commonly divided into two types of statics that are single values and
constant over time and dynamics that are varying over time. Figure 13 lists the inputs for developed TDM for the Barrel of
Oil Equivalent (BOE) in Hilight field. Field properties such as well coordinates, top depth, pay thickness and completion fall
into static class. Note that the only dynamic property available beside the production history is dates and types of workovers
on individual wells. Figure 13 shows the history matched result of developed TDM for all the included wells in Hilight field.
Gray dots and shades represent annual production history and actual cumulative production in Hilight field, respectively. The
green line represents the matching result obtained from TDM model developed. Estimated cumulative values from TDM are
depicted in green shades.

Top-Down models with the minimal computational footprint and strong dependency on measured data rather than interpreted
data, are a preferred candidate for data driven analytics and modeling. Furthermore, Top-Down Models are known to be able
to provide a solution in the absence of sufficient data which is common in mature fields. Figure 14, similarly, provides
history matched results for individual wells in Hilight field. Sensitivity of the output (production) in TDM can also be probed
versus single or multiple input variables (reservoir properties) using Monte-Carlo simulation (Figure 15). This is a
considerable advantage over classic techniques such as DCA or simplistic or complex regression analysis (12; 13) which is
possible because of the negligible computation footprint and the systematic nature (input-output relationships) of TDM.

Top-Down Model Inputs

Static Data Dynamic Data

Location - X Q-Oil (t-1)
Location - Y Q-Gas (t-1)
Depth Workover (t-1) Date
Pay Thickness Workover (t-1) Type
Completion
Porosity

Figure 13 - Top-Down Model developed for Hilight field: TDM inputs are listed (Left). TDM estimates for entire field from 1975 to 2017
are presented (Right). Gray dots represent actual annual oil productions and the green line shows the model estimates and
prediction. Gray and green shaded area, represent the actual and model cumulative production, respectively. The red bars at bottom
show the number of active wells in each year.

Economic Analysis

One of the main advantages of data driven analytics is the independence of interpreted data as well as flexibility with lack of
dynamic production data (time-lapse bottom-hole flowing pressure and etc.). As discussed earlier, these advantages along
with the negligible computational requirement to traditional reservoir management techniques, allows for a reliable and
practical reservoir management and development tool.

We have integrated the insights gained through the field-wide pattern recognition analysis (Figure 9, Figure 10, and Figure
11) and the forecasting capability of the developed Top-Down Model (Figure 13 and Figure 14) to showcase development
strategies in Hilight field. As shown in Figure 16, eight new wells in various locations of the field were planned to start
producing in 2013. The location of these wells was inspired by the field-wide analysis performed on cumulative production,
recovery factor, and remaining reserve. In order to be able to compare the performance of the new wells, they were
distributed in regions of the field with relatively more and less productivity expectation. Excellent quality index for
remaining reserve in north east parts of the field, suggest higher production values for new wells in that region (Well 1, 2,
10 SPE 166111

and 3) in comparison to the rest of new wells. Similarly from average quality indices in Figure 11, it can be argued that
central wells (Well 4 and 5) stand next after new wells in north east in term of production values. Finally, southern new wells
are expected to have lowest production performance among the new wells given the higher recovery factors. Table 1 lists the
predicted cumulative production of new wells from 2013 through 2017 and confirms the expected performances explained
above. Note that cumulative production for new wells in north east regions of the field is about twice this value for the rest of
new wells.

Figure 14 – TDM estimates for individual wells from 1975 to 2017: Gray dots, green line, gray area and green area represent actual
annual oil productions, the model estimates and prediction, the actual and model cumulative production, respectively.

Figure 15 – Uncertainty Analysis using Monte-Carlo Simulation and TDM: Estimated value of TDM is presented with green line
versus actual annual production values in gray dots. Low and High bounding values are shown in highlighted shade and P50 is
shown as dashed line. Effect of uncertainty of porosity on production of Well-262360000 (Top); Effect of uncertainty of pay
thickness and completion on production of Well-25327000 (Bottom).
SPE 166111 11

Net Present Value
New 1 New 2 New 3 New 4 New 5 New 6 New 7 New 8
$3,000 

Thousands

New 1
New 2
New 3
$2,500 

New 1
New 2
New 3
$2,000 

New 1
New 2
New 3
Net Present Value $1,500 

New 5
New 4
New 5

New 6
New 4

New 8
New 1

New 7
New 2
New 3
$1,000 

New 6
New 5

New 8
New 4

New 7
New 6

New 8
New 7
New 5
New 4
$500 

New 1

New 6
New 2
New 3

New 8
New 7
$‐

New 5
New 4

New 6

New 8
New 7
$(500)
New 1
New 2
New 3
New 4
New 5
New 6
New 7
New 8

$(1,000)
2012 2013 2014 2015 2016 2017
Years

Figure 16 –Analysis of new wells: Location of new wells are marked with green circles and numbers (Left). Annual Net Present Value
(NPV) is compared for new wells (Right). Details of Predicted production for New Wells in Hilight field are listed in Table 1.

In order to evaluate the proposed development we have performed economic analysis and calculated net present value (NPV)
using cash flow model (18). The average cost of drilling and completion of new wells in Powder River basin can range
between $400,000 and $600,000 (19; 20; 21). For the practical purposes, we will assume an average value of $500,000 as the
capital expenditure for a new well in Hilight field. For the analysis purposes, we consider the price of crude oil from
Wyoming to fluctuate around $70 per bbl (19; 22). Cost of production for each barrel of oil is estimated to be $5 and
Discount rate and tax in cash flow model is assumed to be 8% and 20%, respectively.

The economic analysis is presented in Table 1 in details and the annual Net Present Values for new wells are depicted in
Figure 16. Note the better economic value of new wells in north east regions (Wells 1, 2, 3) that pay back the capital
expenditure (drilling and completion). New well 1, for example, turns enough revenue in first 9 months to pay back the
capital expenditure (Break even time). This time period for wells with lower performance can be extended up to 21 months.

Conclusion

Developing a cohesive and reasonably accurate predictive model for an asset, even when professionals have access to all the
internal and private datasets available in a company, is a complex and time consuming task. In view of that, it becomes
evident that evaluating prospects using readily available (public) data is not a trivial matter. Just like any other disciplines,
addressing complex problems usually requires development of sophisticated tools. Expecting to get good results in evaluating
complex prospects in the oil and gas industry with simple tools (such as Decline Curve Analysis) will prove to be, to put it
mildly, wishful thinking.

In this paper, we demonstrated sophisticated technologies such as Fuzzy Pattern Recognition and Top-Down Modeling that
are categorized as advanced data driven analytics in order to evaluate a given prospect in Powder River Basin. Comparing
advanced data-driven analytics with traditional techniques such as DCA, regression and etc. clearly demonstrates the value
added by these technologies.

Large quantity of existing wells coupled with limited data (parameters) availability in the mature fields imposes serious
challenges with traditional simulation and modeling techniques. However, advanced data driven analytics techniques such as
Fuzzy Pattern Recognition and Top-Down Modeling do not suffer from similar limitation. They have completely flexible
data architecture and can use “any type” of data that is available in order to accomplish their task. These tools with practical
development and deployment time allows for integrated and versatile investigation of the assets with reasonable financial
resources while extracting valuable information.
12 SPE 166111

Table 1 – Economic Analysis for Proposed Field Developments from 2013 to 2017: Drilling Cost and Price of oil Barrel is assumed
to be fixed. Note the decrease in production from good group (New Wells 1, 2, 3) to the rest of new wells.

Ecomonic Information

Drilling and Drilling Oil Price Production Prod. Cost Discount

Tax
Completion End Date (per bbl) Cost ( bbl) Inflation Rate
$ 500,000 2012 $ 70 $ 5 0% 8% 20%

Economic Analysis (Cash Flow Model)

Cumulative Production Present Break

Production Revenue OpEx CapEx NPV (Σ PV)
Production Cost Value (PV) Even
Well Name Year [bbl/Year] [bbl/Year] [bbl] $ $ $ $ $ Months
2012 - $ - $ - $ - $ 500,000 $(500,000) $ (500,000) 9
2013 13,675 13,675 $ 5.00 $ 957,271 $ 68,377 $ - $ 658,440 $ 158,440
Well 1

2014 14,548 28,224 $ 5.00 $ 1,018,374 $ 72,741 $ - $ 648,582 $ 807,023

New

2015 14,367 42,590 $ 5.00 $ 1,005,662 $ 71,833 $ - $ 593,043 $ 1,400,066

2016 14,120 56,711 $ 5.00 $ 988,428 $ 70,602 $ - $ 539,704 $ 1,939,769
2017 13,892 70,603 $ 5.00 $ 972,461 $ 69,462 $ - $ 491,653 $ 2,431,422
2012 - $ - $ - $ - $ 500,000 $(500,000) $ (500,000) 9
2013 13,306 13,306 $ 5.00 $ 931,420 $ 66,530 $ - $ 640,659 $ 140,659
Well 2

2014 14,120 27,426 $ 5.00 $ 988,407 $ 70,601 $ - $ 629,497 $ 770,156

New

2015 13,956 41,382 $ 5.00 $ 976,927 $ 69,781 $ - $ 576,098 $ 1,346,254

2016 13,736 55,118 $ 5.00 $ 961,520 $ 68,680 $ - $ 525,011 $ 1,871,265
2017 13,533 68,652 $ 5.00 $ 947,338 $ 67,667 $ - $ 478,951 $ 2,350,217
2012 - $ - $ - $ - $ 500,000 $(500,000) $ (500,000) 9
2013 13,237 13,237 $ 5.00 $ 926,604 $ 66,186 $ - $ 637,347 $ 137,347
Well 3

2014 13,932 27,169 $ 5.00 $ 975,226 $ 69,659 $ - $ 621,102 $ 758,449

New

2015 13,857 41,026 $ 5.00 $ 969,997 $ 69,286 $ - $ 572,011 $ 1,330,460

2016 13,748 54,774 $ 5.00 $ 962,381 $ 68,742 $ - $ 525,481 $ 1,855,941
2017 13,651 68,425 $ 5.00 $ 955,542 $ 68,253 $ - $ 483,099 $ 2,339,040
2012 - $ - $ - $ - $ 500,000 $(500,000) $ (500,000) 17
2013 8,007 8,007 $ 5.00 $ 560,497 $ 40,036 $ - $ 385,527 $ (114,473)
Well 4

2014 8,118 16,125 $ 5.00 $ 568,281 $ 40,592 $ - $ 361,927 $ 247,454

New

2015 7,878 24,004 $ 5.00 $ 551,467 $ 39,391 $ - $ 325,202 $ 572,656

2016 7,643 31,647 $ 5.00 $ 535,024 $ 38,216 $ - $ 292,135 $ 864,791
2017 7,430 39,076 $ 5.00 $ 520,065 $ 37,148 $ - $ 262,932 $ 1,127,724
2012 - $ - $ - $ - $ 500,000 $(500,000) $ (500,000) 16
2013 8,379 8,379 $ 5.00 $ 586,530 $ 41,895 $ - $ 403,433 $ (96,567)
Well 5

2014 8,472 16,851 $ 5.00 $ 593,040 $ 42,360 $ - $ 377,695 $ 281,129

New

2015 8,143 24,994 $ 5.00 $ 570,038 $ 40,717 $ - $ 336,154 $ 617,282

2016 7,821 32,815 $ 5.00 $ 547,470 $ 39,105 $ - $ 298,931 $ 916,213
2017 7,527 40,342 $ 5.00 $ 526,862 $ 37,633 $ - $ 266,369 $ 1,182,582
2012 - $ - $ - $ - $ 500,000 $(500,000) $ (500,000) 19
2013 6,975 6,975 $ 5.00 $ 488,215 $ 34,873 $ - $ 335,809 $ (164,191)
Well 6

2014 7,051 14,025 $ 5.00 $ 493,542 $ 35,253 $ - $ 314,327 $ 150,136

New

2015 6,830 20,855 $ 5.00 $ 478,079 $ 34,149 $ - $ 281,925 $ 432,061

2016 6,615 27,470 $ 5.00 $ 463,078 $ 33,077 $ - $ 252,851 $ 684,912
2017 6,421 33,891 $ 5.00 $ 449,456 $ 32,104 $ - $ 227,234 $ 912,147
2012 - $ - $ - $ - $ 500,000 $(500,000) $ (500,000) 21
2013 6,159 6,159 $ 5.00 $ 431,102 $ 30,793 $ - $ 296,525 $ (203,475)
Well 7

2014 6,284 12,443 $ 5.00 $ 439,873 $ 31,420 $ - $ 280,146 $ 76,672

New

2015 6,175 18,618 $ 5.00 $ 432,257 $ 30,876 $ - $ 254,904 $ 331,575

2016 6,071 24,689 $ 5.00 $ 424,970 $ 30,355 $ - $ 232,043 $ 563,618
2017 5,981 30,670 $ 5.00 $ 418,663 $ 29,905 $ - $ 211,666 $ 775,284
2012 - $ - $ - $ - $ 500,000 $(500,000) $ (500,000) 21
2013 6,429 6,429 $ 5.00 $ 450,044 $ 32,146 $ - $ 309,554 $ (190,446)
Well 8

2014 6,544 12,973 $ 5.00 $ 458,073 $ 32,720 $ - $ 291,738 $ 101,292

New

2015 6,395 19,369 $ 5.00 $ 447,678 $ 31,977 $ - $ 263,997 $ 365,289

2016 6,251 25,620 $ 5.00 $ 437,591 $ 31,257 $ - $ 238,934 $ 604,224
2017 6,123 31,743 $ 5.00 $ 428,596 $ 30,614 $ - $ 216,688 $ 820,911
SPE 166111 13

References

1. Hey, T., Tansley, S. and Tolle, K. The Fourth Paradigm: Data-Intensive Scientific Discovery. s.l. : Microsoft Research,
2009.

2. Reservoir simulation and modeling based on artificial intelligence and data mining (AI&DM). Mohaghegh, S. D. [ed.] M.
J. Economides. 6, s.l. : Elsevier, 2011, Journal of Natural Gas Science and Engineering, Vol. 3. ISSN 1875-5100.

3. Reservoir Simulation and Modeling Based on Pattern Recognition. Mohaghegh, S. D. Woodlands, Texas : SPE, 19-21
April 2011. SPE Digital Energy Conference and Exhibition. SPE 143179.

4. An Integrated Technique for Production Data Analysis (PDA) with Application to Mature Fields. Gaskari, R.,
Mohaghegh, S. D. and and Jalali, J. 4, Novermber 2007, SPE Production & Operations Journal, Vol. 22, pp. 403-416.

5. Top-Down Intelligent Reservoir Modeling (TDIRM). Gomez, Y., Khazaeni, Y. and Mohaghegh, S. D. New Orleans,
Louisiana : SPE, 4-7 October 2009. SPE Annual Technical Conference and Exhibition. SPE 124204.

6. New Insight into Integrated Reservoir Management using Top-Down, Intelligent Reservoir Modeling Technique;
Application to a Giant and Complex Oil Field in the Middle East. Kalantari-Dehaghi, A., Mohaghegh, S. D. and
Khazaeni, Y. Anaheim, California : SPE, 27-29 May 2010. SPE Western Regional Meeting. SPE 132621.

7. Top-Down Intelligent Reservoir Modeling (TDIRM); A New Approach In Reservoir Modeling By Integrating Classic
Reservoir Engineering With Artificial Intelligence & Data Mining Techniques. Mohaghegh, S. D. Denver, Colorado :
AAPG, June 7-10 2009. AAPG Annual Convention and Exhibition.

8. Erwig, M. The graph Voronoi diagram with applications. s.l. : John Wiley & Soncs, Inc., 2000.

9. Field-Wide Reservoir Characterization Based on a New Technique of Production Data Analysis. Mata, D., Gaskari, R.
and Mohaghegh, S.D. Lexington, Kentucky : SPE, 17-19 October 2007. SPE Eastern Regional Conference & Exhibition.
SPE 111205.

10. New Method for Production Data Analysis to Identify New Opportunities in Mature Fields: Methodology and
Application. Mohaghegh, S. D., Gaskari, R. and Jalali, J. Morgantown, West Virginia : SPE, 14-16 September 2005.
SPE Eastern Regional Meeting. SPE 98010.

11. Identifying Infill Locations and Underperformer Wells in Mature Fields using Monthly Production Rate Data, Carthage
Field, Cotton Valley Formation, Texas. Jalali, J., Mohaghegh, S. D. and Gaskari, R. Canton, Ohio : SPE, 11-13
October 2006. SPE Eastern Regional Meeting. SPE 104550.

12. Practical Data Mining: Lessons Learned from the Barnett Shale of North Texas. LaFollette, R. F. and Holcomb, W. D.
Woodlands, Texas : SPE, 24-26 January 2011. SPE Hydraulic Fracturing Technology. SPE 140524.

13. Practical Data Mining: Analysis of Barnett Shale Production Results with Emphasis on Well Completion and Fracture
Stimulation. LaFollette, R. F., Holcomb, W. D. and Aragon, J. Woodlands, Texas : SPE, 6-8 February 2012. SPE
Hydraulic Fracturing Technlogy. SPE 152531.

14. Which Parameters Control Production in Shale Assets? A Pattern Recognition Study. Esmaili, S., Mohaghegh, S. D.
and Kalantari-Dahaghi, A. Pittsburgh, Pennsylvania : SPE, 20-22 August 2013. SPE Eastern Regional Meeting. SPE
165712.

15. Intelligent Time Successive Production Modeling. Khazaeni, Y. and Mohaghegh, S. D. Anaheim, California : SPE, 27-
29 May 2010. SPE Western Regional Meeting. SPE 132643.

16. State of the Art of Artificial Intelligence and Predictive Analytics in the E&P Industry: A Technology Survey. Bravo, C.,
et al. Bakersfield, California, USA : s.n., 2012. SPE Western Regional Meeting. SPE 150314.

17. Field Development Strategies for Bakken Shale Formation. Zargari, S. and Mohaghegh, S. D. Morgantown, West
Virginia : SPE, 12-14 October 2010. SPE Eastern Regional Meeting. SPE 139032.
14 SPE 166111

18. Thompson, R. S. and Wright, J. D. Oil Property Evaluation. s.l. : PennWell Corporation, 1985. ISBN-13:
9780318361413.

19. Bender, H. W., et al. Wyoming Oil and Gas Economic contribution Study. s.l. : Wyoming Bussiness Alliance, 2008.

20. EIA. Petroleum and Other Liquids: Costs of Crude Oil and Natural Gas Wells Drilled. U.S. Energy Information
Administration. [Online] 2012. http://www.eia.gov/dnav/pet/pet_crd_wellcost_s1_a.htm.

21. Hinton, D., et al. Petroleum: An Energy Profile. U.S. Energy Information Administration. [Online] 7 1999.
http://www.eia.gov/FTPROOT/petroleum/054599.pdf. DOE/EIA-0545(99).

22. EIA. Petroleum and Other Liquids: Domestic Crude Oil First Purchase Prices by Area. U.S. Energy Information
Administration. [Online] March 2013. http://www.eia.gov/dnav/pet/pet_pri_dfp1_k_m.htm.

23. Virtual Intelligence Applications in Petroleum Engineering: Part 1 ; Artificial Neural Networks. Mohaghegh, S. D. 9,
s.l. : SPE, September 2000, Journal of Petroleum Technology, Distinguished Author Series, Vol. 52, pp. 64-73. SPE
58046.

24. Virtual Intelligence Applications in Petroleum Engineering: Part 2 ; Evolutionary Computing. Mohaghegh, S. D. 10,
s.l. : SPE, October 2000, Journal of Petroleum Technology, Distinguished Author Series, Vol. 52, pp. 82-87. SPE 61925.

25. Virtual Intelligence Applications in Petroleum Engineering: Part 3 ; Fuzzy Logic. Mohaghegh, S. D. 11, s.l. : SPE,
November 2000, Journal of Petroleum Technology, Distinguished Author Series, Vol. 52, pp. 82-87. SPE 62415.