MARCH 2016

The Better Business Publication Serving the Exploration / Drilling / Production Industry

Modeling Optimizes Asset Performance

By Chad Baillie lies in its ability to whittle down a potential KPIs that emerge from multivariate
list of more than 100 attributes from 3-D analysis are correlated with other well
MISSOURI CITY, TX.As more well seismic and well data, and to identify the log, core and completion data to determine
and completion data have become available top five or so attributes that have the which reservoir properties or conditions
in unconventional U.S. oil and gas plays, greatest impact on well performance. The they may represent. Once these indicators
geoscientists and engineers have sought attributes that contribute to cumulative are defined and understood, they can be
to understand exactly which reservoir production are known as the wells key integrated into a single 3-D predictive
properties have the greatest impact on performance indicators (KPIs). Typical model of production.
production performance in each formation. KPIs include structural or geometric at-
Over time, operators also have come to tributes, attributes related to seismic fre- Integrated Modeling Workflow
depend on seismic data both to map quency or amplitude, mechanical and Figure 1 shows a generic overview of
overall structure and to delineate reservoir elastic properties, and even completion a process dubbed well prospectivity and
properties away from well control. or engineering attributes such as proppant productivity analysis (WPPA), in which
However, it has become painfully clear mass and fluid volume. multivariate statistical analysis (orange
that strong well performance cannot be
correlated with individual reservoir prop-
erties in unconventional fields. Instead, a FIGURE 1
unique combination of various properties Generic Well Prospectivity and Productivity Analysis Workflow
and conditions controls the productivity
of any particular well or reservoir. With HTI Processing Deviation Surveys, Perf
so many post-stack and prestack seismic for Azimuthal Locations, Stage Data,
Anisotropy Production Data and
attributes available today (not to mention Well Log Seismic Prestack Elastic Completion Data
Well Information
well logs and engineering information), Logs Tops,
Import/QC Preparation and Interpretation and Inversion for
G&G Data Petrophysics Well Integration Rock Properties
operators are finding it increasingly dif- Culture Data

ficult to sort through and quantify their Ambient Seismic

Formation Top Import/OC
(TFI and Velocity Modeling
meanings, relationships, and relative con- Semblance)
Lithology/Facies Engineering Data
Estimation
tributions to reservoir behavior and pro- Generate Post-Stack
duction results. Depth Conversion
Attributes
Given that there are simply too many Geology & Geophysics Data
Evaluation and Model Updated Model Extract Seismic Create Wellbore and
static and dynamic variables and potential Engineering Data Validation with New Wells Attributes and Assign to Stage Intervals
outcomes for traditional analysis, inter- Geoscience Integration
Wellbore/Stage Intervals
pretation and modeling, seismically-con- Correct for Variability
in Completions
strained multivariate statistical analysis Correlate Seismic
Extractions to Production
with integrated geoscience and engineering and Engineering Data Create Production
data can provide far more reliable pre- Metrics

dictive models of cumulative production. Multicollinearity Analysis

A case study from the Wolfcamp Shale
play in the Permian Basin shows how Build 3-D Production Multivariate Statistical Engineering
Prediction Models Analysis Best Practices
operators can apply these models to op-
timize drilling, completion design and Create Prospectivity Drilling Optimization
execution, identify refracturing candidates, and Probability Maps Analysis

and ultimately improve recovery in un-

conventional wells and fields. Well Completion Analysis

The strength of multivariate statistics

Reproduced for Global Geophysical Services with permission from The American Oil & Gas Reporter
www.aogr.com
SpecialReport: Unconventional Resource Science

FIGURE 2 Based on the input data available, the

Location of 3-D Survey Area and Permian Basin Geologic Setting process also can determine engineering
best practices, and can facilitate additional
steps to validate the models.
The integrated WPPA workflow can

be repeated and the models can be carried

forward throughout the life of a particular
reservoir or field to identify ideal landing
points for laterals and completion zones
to optimize production.
Permian Basin Application
To demonstrate the potential value of
multivariate analysis and WPPA tech-
nologies, the integrated workflow was
applied to predict cumulative production
in the Wolfcamp formation in West Texas
using a unique dataset consisting of a
multiclient 3-D seismic survey and public
well and completion data. The objective
was to determine if the success of pro-
duction models created with multivariate
analysis and WPPA for more detailed
projects elsewhere within the basin and
in other unconventional plays could be
box in the lower right) plays such a model is used to convert all seismic in- reproduced using these data inputs.
critical role in integrating geological and terpretations and attributes from time to The 3-D multiclient survey was ac-
geophysical data (green boxes) with en- depth. quired in 2012, targeting the Permian
gineering data (yellow boxes). The first step on the engineering side section of the Wolfcamp. Figure 2 shows
The first step is to load and quality of the workflow is to load and QC all rel- the location of the University Lands 3-D
control all the available well information evant deviation surveys, perforation loca- dataset and the major subdivisions of the
needed for production modeling. Primary tions, stages, completion and production Permian Basin. The dataset covers 356
well data typically consists of well logs data. Stage and completion information square miles in Reagan, Upton and Crock-
for geologic interpretation and petrophys- is used to define productive zones for in- ett counties, including the eastern edge
ical analysis, wells with velocity control tegrating with the geoscience data. of the Central Basin Platform and part of
for seismic interpretation and velocity The initial step in integrating the geo- the geologically complex Ozona Arch
modeling, and horizontal wells with pro- science and engineering data is to extract area.
duction and completion data. Following 3-D seismic attributes along each wellbore The entire thickness of the Permian
the QC (quality control) of the well in- at or near the stage locations using a sta- section is roughly 4,000 feet, and is
formation, a rigorous interpretation of tistical sampling method. Next, the ex- shown in color in Figure 2 overlaid on
the formation tops is performed, making tracted attributes are correlated with a the seismic section. The survey was de-
sure the markers are consistently picked range of production metrics, completion signed, based on a primary depth of
at each well. data and well data to determine the degree 6,000-8,000 feet to allow proper imaging
After defining reservoir zones and sur- to which variables are correlated. This of azimuthal anisotropy. This ensured
rounding formations, the seismic is tied also allows the interpreter to understand that stable amplitudes within the formation
to the wells using available sonic infor- the individual relationships each variable were adequate for interpretation and in-
tegration. The surveys dense acquisition
mation. If sonic data are not available, has with production and what type of
geometry also resulted in an uplift in the
either sonic or density can be estimated transformation may be needed in the mul-
nominal fold, allowing an increase in the
from the petrophysics. Obtaining accurate tivariate analysis.
signal-to-noise ratio and providing more
well-ties is essential to further analysis, The goal for this stage is to leverage robust gathers for future inversion work.
since any errors that occur at this stage the integration, allowing the statistics to The seismic attribute analysis began
will propagate, causing significant prob- highlight the meaning of various rela- with computing a suite of 45 post-stack
lems later in the workflow. tionships between production and seismic seismic attributes, which included a set
Following well-ties and error analysis, attributes. The result of multivariate analy- of incoherence volumes, several fault
formation tops, seismic time interpreta- sis is to identify the KPIs and determine probability volumes, 15 volumetric cur-
tions, and well velocity control are inte- the optimal transformation required for vature attributes, several complex trace
grated to build the velocity model. In each attribute. Finally, the KPIs are com- attributes, and a set of spectral decompo-
parallel, a large number of both post- bined to create a 3-D seismically-calibrated sition frequency volumes. All attributes
stack and prestack seismic attributes are predictive model and maps of production were analyzed over different intervals
computed. Finally, the integrated velocity (or indeed any other response metric). within the Wolfcamp section for variations
SpecialReport: Unconventional Resource Science

FIGURE 3 transformations of each one.

Predicted versus Actual Production for 21 Input Wells (Left) Key Performance Indicators
And Predicted Input and Blind Wells versus Actual Production (Right)
The most important KPIs driving the
statistical predictions of three-month oil
per foot, along with the reservoir properties
or conditions they represented, were:
Root mean square curvature and
dip curvature, by themselves, were not
highly correlated with cumulative pro-
duction. However, combined with other
variables, they made a significant contri-
bution to well performance. This indicated
that structural features associated with
faulting and fracturing, seals and migration
pathways were impacting production.
Relative amplitude change and 24-
28 hertz spectral decomposition frequency
volumes, were used interchangeably, since
relative amplitude change had a high
linear relationship with midfrequency
volumes. The lower values of relative
amplitude change indicated an increase
in both water and oil production, which
was verified with hydrocarbon and water
in stratigraphy, thickness, faulting and tribute with various production metrics
fracturing, fluid and lateral facies. saturation logs in several wells across
to determine the best response or predictor
From an analytics standpoint, even variable for multivariate analysis. We also the field. Significantly higher fracture
though the interpreter may be interested evaluated how changing these production porosity in these zones of higher fluid
only in the reservoir surrounding the pro- metrics over time affected the correlations saturation also was noticed.
ducing wells, it is also important to un- with each attribute. Based on strong linear 10-14 Hz spectral decomposition
derstand how adjacent formations may and nonlinear correlations, three-month attributes correlated with both total organic
impact reservoir properties and conditions, oil production scaled per foot of completion carbon and the brittle/ductile quality that
and how the seismic data may respond to length was selected as the variable to pre- was calculated at several wells. Correla-
these changes. Therefore, all seismic at- dict. Using a nonlinear multivariate re- tions of nine-month gas production versus
tributes also were extracted and analyzed gression method, we identified three pri- low-frequency volumes also supported
over other horizons and intervals to see mary performance indicators out of a this relationship. As the 10-14 Hz com-
how they varied laterally across the Uni- wide range of attributes, and tested various ponent increased, so did both gas pro-
versity Lands dataset.
The next step was integrating the seis- FIGURE 4
mic with the well data used for the mod-
eling, which consisted of 21 horizontal Arbitrary Cross-Section through Six Vertical Wells
wells in the Wolfcamp with production With Single-Zone Completions
data and limited completion information.
In addition, 25 other producing wells
were left out of the analysis, reserving
them for blind testing to validate or revise
predications later. A single horizontal
section was created for the 21 input wells,
and defined both the completion interval
and the interval for the well-to-seismic
extractions.
Formation tops and the corresponding
time horizons were used with well ve-
locities from two wells to convert all
seismic attributes and horizons to depth.
The depth-converted seismic attributes
then were sampled along the target com-
pletion interval and averaged for the next
stage of analysis.
The geoscience/engineering integration
included individually correlating each at-
SpecialReport: Unconventional Resource Science

FIGURE 5 predicted production was roughly the

Arbitrary Line through Three Vertical Wells same. However, wells 1 and 3 have the
Showing Lateral Variations in Predicted Production highest probability for refracturing success
because the completions in both wells
missed the sweet spot revealed by multi-
variate analysis and predictive modeling.
Targeting these zones, each well should
be able to generate three-month production
values of about 10,000 barrels. It should
be remembered that neither of these wells
was used in the initial modeling.
This Permian Basin case study clearly
demonstrates the potential value of using
WPPA workflows to improve well plan-
ning, optimize completions, and enhance
recovery, even without a full suite of
data.
The productivity of any well is a func-
tion of the interplay among many static
and dynamic characteristics that require
duction and TOC. potential. This was accomplished by com- an integrated, multivariable solution. The
Next, these top performance drivers paring the ratio of production from the WPPA workflow represents a sophisticated
were modeled to production using the entire Wolfcamp zone with production multidisciplinary integration of the geo-
optimal transformations for each, which extracted at the actual perforation locations. physical, geological and engineering data.
then were summed to form the production According to the analysis, three of the By combining numerous diverse sets of
prediction for various models. Using no eight vertical wells with single-zone com- data, geoscientists and engineers can
more than three seismic attributes for pletions had potential zones of high pro- identify the specific attributes essential
each model, three unique 3-D models of duction not accessed by the original com- to hydrocarbon production in a particular
three-month oil per foot were generated. pletions. reservoir. Not only can they generate
The same extraction parameters were Figure 4 displays an arbitrary cross- quantitative estimates of the values that
used to extract the predictions along each section through six of the eight vertical impact production, but they also can iden-
wellbore. The average correlation coeffi- wells. The log displayed in color is the tify the most prospective areas for drilling.
cient between actual versus predicted extracted production model along the Properly applied, this workflow can reduce
production for the three models was 92 vertical wells. The red disks show the
percent for the 21 input wells (left panel actual completion interval for these wells.
in Figure 3). Wells 1, 3 and 4 (highlighted in yellow)
To test how effectively the model have the highest probability for refrac
could predict cumulative production per- success.
formance, the 25 producing wells that Looking at the models in 3-D showed
had been excluded from the modeling
how much lateral variability existed be-
workflow were introduced. Adding actual
tween these wells in the target interval.
versus predicted production from 17 blind
Figure 5 shows an arbitrary line through
horizontal wells and eight blind vertical
vertical wells 1, 3 and 6 from Figure 4. CHAD
wells with single-zone completions to
the 21 input wells achieved a correlation Note the lateral variations of the pro- BAILLIE

coefficient of 83 percent (right panel in duction prediction model. The seismic Chad Baillie is a lead reservoir
Figure 3). Looking only at the eight in the background is co-rendered with geophysicist at Global Geophysical
vertical wells, the model showed a 90 the fault probability, with the production Services. Prior to joining the company
percent correlation between predicted ribbon displayed in color over the reservoir in 2012, he worked as a geophysicist
and actual production. These high corre- zone. at Prism Seismic/Sigma3. Baillie has
lations demonstrate the predictive power The best vertical producer in the survey extensive experience in interdisciplinary
areawell 6 shown at the left in Figures reservoir characterization and inter-
of multivariate statistical analysis and the
4 and 5had 10,700 barrels of actual pro- pretation, and his areas of expertise
impact it can have on unconventional
duction over three months versus a pre- include multivariate statistics by in-
field development.
dicted production of 10,900 barrels (98 tegrating 3-D seismic attributes, rock
Refrac Candidate Wells mechanics, microseismic, engineering
percent accurate). Well 3 had 2,090 barrels
and geological data to optimize com-
The decision then was made to examine of actual versus 2,300 barrels of predicted
pletion and well performance. He
the possibility of refracturing some or all production. holds a B.S. in geophysics from the
of the vertical wells in specific zones Well 1 was one of the poorer per- Colorado School of Mines.
where the model indicated high production formers. It produced 101 barrels, and its
SpecialReport: Unconventional Resource Science

drilling risk significantly and improve Similar results have been obtained in recoveries of individual wells.
drilling programs. successful projects in both conventional
The Wolfcamp case study emphasizes and unconventional reservoirs across Editors Note: The author acknowl-
the value that multivariate statistical analy- North America. Results achieved in the edges Global Geophysical colleagues Ro-
sis of 3-D seismic, well and engineering field have proven that WPPA models can hit Singh, a petroleum engineer, and
data can bring to unconventional reservoirs. high-grade prospective drilling areas, en- Leszek Bednarski, a petrophysicist/geol-
The 3-D seismic attributes applied in the hance well planning and geosteering, op- ogist, for their contributions to the Wolf-
Permian Basin can predict well perform- timize completion placement and frac camp modeling project.
ance down to the stage level. design, and boost the estimated ultimate