Download as pdf or txt
Download as pdf or txt
You are on page 1of 26

URTeC: 2460295

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


Geological and Geomechanical Modeling of the Haynesville Shale:
A Full Loop for Unconventional Fractured Reservoirs
W. Sebastian Bayer*, Marcus Wunderle, Ewerton Araujo, Rene Alcalde, Calvin Yao,
Fred Suhy, Thomas Jo, Fleur Bases, Abu M. Sani, Yiwei Ma, Abhishek Bansal, Eric
Peterson, Rohan Goudge, Ankur Awasthi, and Mukul Bhatia BHP Billiton Petroleum.
Copyright 2016, Unconventional Resources Technology Conference (URTeC) DOI 10.15530-urtec-2016-2460295

This paper was prepared for presentation at the Unconventional Resources Technology Conference held in San Antonio, Texas, USA, 1-3 August 2016.

The URTeC Technical Program Committee accepted this presentation on the basis of information contained in an abstract submitted by the author(s). The contents of this paper
have not been reviewed by URTeC and URTeC does not warrant the accuracy, reliability, or timeliness of any information herein. All information is the responsibility of, and, is
subject to corrections by the author(s). Any person or entity that relies on any information obtained from this paper does so at their own risk. The information herein does not
necessarily reflect any position of URTeC. Any reproduction, distribution, or storage of any part of this paper without the written consent of URTeC is prohibited.

Summary

The Haynesville Shale remains a prolific gas resource amongst the Unconventional Plays in the US. The continued
viability and the commercial success of the play are highly dependent on the optimization of field development
plans through drilling, completions, and production improvements.

This paper presents an integrated solution that includes geologic, geophysical, and geomechanical properties. The
workflow includes a Discrete Fracture Network (DFN), modeled hydraulic fractures, and well diagnostics data, to
improve the understanding of the subsurface. The goal is to provide valuable input to optimize the development plan
and completions strategy in the Haynesville Shale.

The development of the integration platform (3D geo-cellular model) involved detailed seismic interpretation based
on a sequence stratigraphic framework, definition of stratigraphic-mechanical units, and incorporation of a robust
petrophysical analysis set in a structurally controlled grid. The structural framework of the model was enhanced
using over 100 carefully interpreted geo-steered horizontal wells to improve accuracy and grid calibration to the
well paths. The natural fracture analysis included core description and fracture counts complemented by borehole
image data, which coupled with a geomechanical stratigraphic characterization study, assisted in understanding the
field wide fracture intensity distribution and orientation.

The hydraulic fracture conductivity and net pressure profiles, along hydraulic fracture planes, were developed using
a planar geometry fracture simulator. The results served as input to the geomechanical model and as the basis for
hydraulic fracture stage design setup in the dynamic model.

A 3D geomechanical model was constructed using the geologic model, based on the pore pressure and mechanical
properties from calibrated 1D-geomechanical models. Computational geomechanical simulations allowed us to
identify reactivated natural fractures, which produced synthetic-microseismic events, and the Critically Stressed
Fracture Volumes (CSFV). These inputs were used in the subsequent identification of Stimulated Rock Volumes
(SRV). Interpretations are supported by tracer data and other field observations that assisted in establishing inter-
well connectivity.

The products from these processes will be incorporated into a reservoir simulation model. History matching of
production data will be conducted for validation and refinement. History matched models will be used to identify
and evaluate the impact of key drivers of optimization studies to various field development scenarios in order to
enhance well completion and well spacing strategies in the development plan.
URTeC: 2460295 2

Introduction

The Haynesville Shale remains an extremely prolific shale gas-producing reservoir. The question is no longer how
we produce gas from the Haynesville, but how we produce gas more efficiently. We believe the primary tactic to
answer that question is simple—through multi-disciplinary integration. We have designed and implemented a multi-

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


disciplinary project to optimize completions and potentially increase productivity. Our detailed reservoir
characterization takes into account hundreds of wells, thousands of quality data points, core studies, seismic
interpretation, geologic, petrophysical, and geomechanical models. The Haynesville play is known as a “fractured-
play” due to low permeabilities, high occurrence of fractures in core and high flow rates. Understanding the matrix
and fracture components of unconventional fractured reservoir systems is critical. In this case study, we attempt to
demonstrate our methodology in the simplest form. We do this by looking at each component independently, with a
geocellular reservoir model for the matrix (storage) and a discrete fracture network (DFN) for the natural fractures
that we integrate in our unconventional reservoir characterization loop (Static-to-Geomechanics-to-Flow).

Geologic Setting / Study Area

The Haynesville Shale play covers a very large area across Northwest Louisiana and East Texas (Figure 1). Various
descriptions of the play have been published by many authors, most notably Hammes et al. (2011) and the Bureau of
Economic Geology (2015). We define the play beyond what most have done to completely characterize the reservoir
including areas where the Haynesville may not be the mudrock facies we commonly think of as the target interval.
The play itself lies between the East Texas and Northwest Louisiana Salt Basins on the Sabine Uplift, straddling the
Texas/Louisiana border and extending across multiple Texas counties and Louisiana parishes (Figure 1).

Figure 1: The Haynesville Shale Basin lies between the East Texas and Northwest Louisiana Salt Basins on the Sabine Uplift. Texas/Louisiana
border. For regional understanding, we cover the entire play area (Brown Polygon). Higher resolution “sub-regional model” (Blue Polygon). All
models use same well tops, petrophysical analysis, seismic data, and workflows ensuring model integration.

To visualize and attempt to understand what is driving the productivity of the play, we started by defining a type log
and building a very large, yet simple geologic model (Santacruz et al., 2015). The basic framework was built based
on a seismic horizon we referred to as Base Haynesville. The Base Haynesville is characterized as a strong seismic
reflector referred to as the Gilmer Lime and informally as Smackover by many in the literature and the industry
(Ewing 2001, Hammes et al. 2011, Cicero et al. 2014). That strong seismic reflector was picked in over 120 2D
lines and 11 3D volumes, tied together and mapped in the time domain, depth converted using a velocity model, and
subsequently calibrated to well penetrations using both vertical and horizontal wells (Figure 2c).
URTeC: 2460295 3

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


Figure 2: (a) Louisiana Type Log modified after Cicero & Steinhoff (2010, 2014) and adapted to Louisiana. (b) Haynesville Type Log. The base
of the section is defined by a sequence boundary followed by an upward increasing gamma ray into a maximum flooding zone. The end of
Haynesville Time is capped by a basin-wide maximum flooding surface marking the end of carbonate deposition, as seen in the lithology track.
Internal picks, current Unit Stratigraphy include sequence stratigraphy and Mechanical Stratigraphic Relationships. (c) Base Haynesville
Structure Map. The type log’s location is marked by a red star. Outline of the Haynesville modified from internal work and from the literature
(Hammes et al., 2011; EIA, 2011; Texas Bureau of Economic Geology, 2015).

Figure 3: Gross Depositional Environment (GDE) Maps. (a) Lower Haynesville. (b) Upper Haynesville. Modified from Cicero and Steinhoff,
2014. Area of interest shown as Blue Polygon overlyed on each GDE. Larger version of type log in figure 2
URTeC: 2460295 4

Static Modeling

The goal of our modeling effort was to prepare the best possible description of the subsurface (Cook et al. 2014).
We aim to be predictive to the best of our knowledge based on the understanding and contributions from each one of
the disciplines on our team. To achieve this we constantly build on learnings, maintaining an evergreen approach

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


(Kleban et al. 2016), and evolving the modeling methodology (Santacruz et al., 2015) for Unconventional Fractured
Mudrock reservoirs in BHP Billiton.

The geologic models constructed became an integration platform (3D geo-cellular model). The multidisciplinary
strengths of the team were combined resulting in the best possible interpretation of the subsurface. The platform
includes detailed seismic interpretation, definition of mechanical character, a robust petrophysical analysis, and a
model framework calibrated to over 100 horizontal wells with structural curtains (Figures 4 and 6). These structural
curtains significantly enhance the framework of the model due to the careful quality control we impart on every re-
steered horizontal well. These are the same wells used in studies by our reservoir engineers working on various
projects, including simulation (Figure 4).

Figure 4: Example of a cluster of Structural Curtains in one of our study areas. Spheres are additional points of constraint for each one of the
geologic units (standardized mechanical unit framework and well tops). Structural Curtain: Official unit stratigraphy along horizontal wells as
defined by geo-steering based on representative offset pilot.

Integrated Modeling

The construction of the 3D geo-cellular models, for the matrix component of the system at the regional and sub-
regional scales, incorporates detailed seismic interpretation, the stratigraphic framework, and the endorsed /official
petrophysical data (Figure 5). For the fracture component of the system, we included mechanical stratigraphy and
detailed natural fracture studies, which included core and borehole image logs. For each mechanical unit, the team
focused on identification of fracture families, related fracture sets, their fracture intensity distribution, and
orientation. This is further expanded in the DFN and Natural Fracture Characterization sections of this paper.

Figure 5: The integrated reservoir modeling workflow from Basin to Field to Well/Pad.
URTeC: 2460295 5

The integrated geologic and DFN descriptions were used to feed geomechanical models. These models were
validated through qualitative calibration with field tests using tracers (Figure 22). The plan for these calibrated
models is to optimize development planning and completion strategies as the work continues in the Haynesville.

Play Scale Modeling

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


The regional model has gone through several iterations, from a simple full Haynesville section model (Santacruz et
al., 2015) to the most recent that includes all mechanical units and matches the resolution of the sub-regional scale
model. Figure 7 shows, in map view, the extents of our play scale model built on a seismic horizon at the Base
Haynesville, a simple model for field-wide evaluation (Beeson et al, 2014). The modeling resolution is enough to
capture higher structural detail for areas covered with our 3D seismic volumes. Re-steered horizontal wells are
incorporated both at the regional and local modeling scales.

Figure 6: Unit stratigraphy correlation along horizontal wells–for well location, see Figure 7c. (a) Correlation to the top of Haynesville interval.
(b) Correlation markers extended down the wellbore, as experience of the team increases. Using patterns in the gamma ray curve recognizable to
experienced operations geologists working the project. (c) Most recent and crucial step in our workflow: vertical and horizontal wells correlated
to official unit stratigraphy in addition to re-steered horizontal wells with the entire lateral tied to unit stratigraphy (see also structural
curtains in Figure 4). (d) Updating structural framework with new interpretation data. Displayed above is a horizontal well as compared to unit
stratigraphy. Colors of the picked tops match the colors of the surfaces. The brown surface is a modeled property, displayed at the base of the
stratigraphic section.

An essential milestone achieved in the project was correlating all available wells to the unit stratigraphy (Figure 2b)
and using the units for the model framework. In most cases, when new data became available it was possible to
correlate horizontal wells to at least, the top of Haynesville interval (Figure 6a). First pass work required picking the
top of the reservoir section using the gamma ray log and comparing the new well to the existing modeled unit
stratigraphy. As the project continued and the experience of the team grew, correlation markers extended down the
wellbore, using patterns in the gamma ray curve recognizable to experienced operations geologists working the
project (Figure 6b). These additional correlation points added confidence to the interpretation of well placement.

The most recent and crucial step in our work includes re-steered horizontal wells with the entire lateral tied to the
official unit stratigraphy from geosteering (Figures 6d, also structural curtains in Figure 4). With this type of
interpretation, there is uncertainty, however the confidence in well placement goes from low to high confidence.
URTeC: 2460295 6

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


Figure 7: Integrated Modeling. (a) Insert showing regional extension of models, outlines of Texas, Louisiana (yellow), and footprint of counties
for reference (black). (b) Regional (~107 miles EW extent) and Sub-regional (~50 miles EW extent) scale models. Smaller areas of interest
(colored boxes) inside blue area, highlight sector models extracted from the sub-regional model. (c) Regional play-scale model. Example of a
property and its map view field-wide distribution. This model honors the field boundary. Areas showing more detail correspond to available 3D
seismic data in the models. An example of this is the Northeast portion of the model (area covered by a 450 square mile 3D survey). The black
rectangle is highlighting a Haynesville horizontal well (seen in Figure 6) that has been re-steered and incorporated into the geologic model, at all
scales.

Model framework and Reservoir Properties

The most important goal in our project is to meet the needs of end-users for everyday work in the Production Unit.
Our fit-for-purpose modeling approach achieves this. A daunting but crucial structural data gathering campaign that
included more than 765 wells correlated to the official Unit Stratigraphy and over 100 horizontal wells re-steered to
obtain structural curtains, resulted in more than 64,000 structural and stratigraphic high quality calibration control
points, available for any project (Figure 4).

Our models take into account high-quality well log information for the entire regional area of interest (Figure 8),
only including wells with full petrophysical analysis that have been peer reviewed by our Petrophysics team. Once
required quality is achieved, the well log becomes part of our official dataset and only this data is used as input for
our models (Figure 5). Our official log database consists of 181 wells that include logs to prepare the properties
selected by the integrated reservoir characterization team (Figure 10a). These properties are Total Porosity, Water
Saturation, Gas-Filled Porosity, Permeability (GRI: matrix), Total Organic Content, Gamma Ray, Volume of Clay,
Quartz Content, Carbonate Content, Static Young’s Modulus, Dynamic Young’s Modulus, and Poisson’s Ratio.

Petrophysical properties were estimated (Figure 10a) with a multi-mineral linear solver. In this method, all available
well-logs (density, neutron, and resistivity) are solved simultaneously to estimate mineralogy (clay, quartz, calcite,
kerogen, and pyrite) and fluid constituents of the rock (porosity, water saturation, and hydrocarbon saturation).
Multi-mineral analysis is under-determined due to fewer equations (density, neutron, and resistivity) than number of
unknowns (mineralogy, water and gas saturation). In-order to reduce non-uniqueness of the solution, TOC, pyrite,
and calcite concentrations were predetermined by deriving correlations with core and well-log data. Estimated
petrophysical properties were calibrated to core and ECS data (Figures 9 and 10a). TOC was estimated using a
modified Passey’s (1981) and Schmoker’s (1979) method, then further calibrated to TOC values from core. A
reasonable agreement was observed between core TOC and calculated TOC (Figure 9a). Pyrite concentration was
estimated by generating a correlation between core pyrite and core TOC. High uncertainty was observed in the
estimation of pyrite concentration. Water saturation was calculated using Waxman-Smit’s equation (Waxman and
Smits, 1968).
URTeC: 2460295 7

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


Figure 8: All available Petrophysical data for calibration. Database includes 181 wells with full Petrophysical analysis. The goal of gathering
and analyzing a large dataset of wells was two-fold—1) To be able to describe the entire play. 2) To ensure data trends between the local model
and the regional-scale model were fully integrated. The Brown Area covers the Regional model (entire play). The Blue Polygon covers the sub-
regional model.

The main goal of the petrophysical analysis is to improve our understanding of storage, hydrocarbon pore volume,
influence of the matrix-fabric, and the stiffness of identified mechanical units (Santacruz et al., 2015). All of these
are part of the matrix component of the Haynesville system and are covered by the preparation of reservoir
properties. Building on this premise, we go beyond the matrix in our current effort and attempt to describe the
natural fracture component of the system. We interpret the Haynesville interval for our area of interest as a type 2
fractured reservoir (Figure 10b), where the matrix provides the bulk of the storage and the fractures provide the
essential reservoir permeability (Nelson, 2001). We address this portion through the construction of a DFN, taking
into account mechanically similar units (Figure 15).

As the project increased the level of detail, we made the decision to investigate multiple resolutions and agreed to
redefine our model for property distribution. The model included finer resolution cells, in terms of area (330ft X
330ft), as compared to our previous regional model (Santacruz et al., 2015), but with a lower vertical resolution
(5ft). As a characterization team, it was clear, that going through upscaling with our models was going to result in
the loss of geologic fidelity and the difficulty of controlling end-results of upscaled distributed properties. This was
a key compromise between geoscientists and engineers, that allowed the grid to be constructed at the resolution of
the dynamic model, while maintaining resolution to reasonably honor relevant trends observed in the logs (Figure
11b).

Taking into account that no subsequent upscaling was going to be performed and that future simulation efforts
would honor the agreed geologic model resolution, the grid was designed to treat each one of the mechanical
stratigraphic units of the Haynesville independently at the agreed vertical resolution of 5ft (Figure 11b). This was
done to prevent data contamination between independent units and to be able to distribute the properties with data
analysis specific to each unit.
URTeC: 2460295 8

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


Figure 9: Comparison between core and calculated petrophysical properties. Blue arrows along axis indicate direction of increasing values.

Integration and continuous communication prevented misunderstandings and lost work, common amongst isolated
teams. By doing this, we ensured that we would only have to build one seamless geologic model that was tied to the
regional geologic trends observed (Figure 3). This proved very valuable, allowing us to extract sector models with
Petrophysical properties, at the proper resolution anywhere within the area of interest and on-demand to any
professional working the asset. The large amount of seismic data incorporated, the logs in our official petrophysical
data set, and the number of calibration points we have added as we work the project gives us high confidence in the
subsurface structure and property distributions in our areas of interest.

Figure 10: Matrix and Fracture components of the Haynesville System. (Top, a) Example of Petrophysical logs selected by integrated team for
reservoir characterization of matrix component. Type well (Top left). Numbers identify log tracts of interest. 1.) Gamma Ray, 4.) Total Organic
Carbon, 5.) Total Porosity, 6.) Water Saturation, 7.) Gas-Filled Porosity, 10.) Permeability (GRI: matrix), 12.) Volume of Clay (dark gray),
Volume of Quartz (yellow), and Volume of Carbonate (light blue). Sample well (Top right: 1.) Gamma Ray, 6.) Static and Dynamic Young’s
Modulus, 7.) Poisson’s Ratio. Blue arrows under the name of each property indicate direction of increasing values. (Bottom, b) Fractured
reservoir types. Haynesville section for our area of interest, interpreted as a type 2 fractured reservoir. Matrix provides storage and fractures
provide essential reservoir permeability (Nelson, 2001). Classification of fractured reservoirs, based on fracture permeability and fracture
porosity.
URTeC: 2460295 9

As part of our current standard unconventional workflow, we modeled the reservoir and rock properties
stochastically (Gaussian random function simulation) with an isotropic area variogram, allowing the data from pilot
wells to control the distribution radially from the location of a well (x,y) to where another pilot well with full
petrophysical analysis was encountered (Santacruz et al., 2015). These occurrences, pilots with full petrophysical

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


sets, became more common in our area of interest with the addition of selectively located vertical wells to our
official database. As new wells are added with the support of our Petrophysics group, the knowledge of the
subsurface in our models continues improving with every iteration (Kleban et al. 2016).

Detailed data analysis was conducted using well log distributions in our official dataset. This calibration step was
conducted for each one of the reservoir properties for each unit (Figure 12a). A new methodology in our workflow is
the ranking of reservoir properties into primary, secondary, and third-ranking-order properties. Even though we
conducted and assigned rankings to all properties in the model, in this paper we only describe four examples.

Figure 11: Sub-regional (high Resolution) model well log upscaling. (a) 3D Property distribution example of water saturation. Match of log
Petrophysical data shown along vertical well. Logs are upscaled into the grid for property distribution. No further upscaling is performed
between geo-cellular properties. (b) Match of high resolution model to logs (logs are shown as bold lines along with color filled geocellular
properties). Gamma Ray is shown in track #1. Upscaled Properties: Total Porosity, Water Saturation, and Matrix Permeability. All logs have been
previously calibrated to core data. Blue arrows show direction of increasing property values. Vertical size cell size ~5ft.

1.) Bulk Density was designated as a primary property, tied to well logs, honoring the shape and limits of
petrophysical distributions, with no imposed spatial trends to the modeled property (Figure 12).

2.) Total Porosity was assigned a secondary property ranking. Just like primary properties, it was tied to the well
logs, the shape, and limits of the distributions in the logs. A secondary spatial trend using the previously generated
Bulk Density property was imposed (Figure 13a). Dominant porosity type in the Haynesville is organic porosity.
Organic-rich Haynesville mudrock that approaches lower density (Hammes et al. 2011) shows a clear control in the
regional depositional trends of porosity that is also observed when distributing the property using only logs
(Santacruz et al., 2015). Based on these observations we recognize the expected strong correlation between Bulk
density and Total Porosity as a useful control to the distribution. Furthermore, bulk density is less affected by user
influence in the petrophysical analysis.

3.) For the case of Water Saturation and 4.) Permeability (Third-ranking-order properties), the difference is that, the
imposed spatial trend was performed using Porosity. A positive correlation was used between Matrix Permeability
and Porosity. An inverse correlation was used for Water Saturation; taking into account that organic porosity is the
dominant porosity type (Figure 13b).
URTeC: 2460295 10

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


Figure 12: Data analysis; Bulk Density, primary property example. (a) Calibration of shape and limits of petrophysical properties to distributions
in official log database. (b) Distribution result of completed property. Distribution of full resolution logs (pink), distribution of upscaled logs
(green), and distribution of modeled property (blue). Data analysis always done separately for each unit.

Once petrophysical properties are distributed, our workflow provides the ability to extract sector models anywhere
in the areas of interest (regional 7c or sub-regional 7b). Additional to reservoir characterization studies or reviews in
core acreage, this ability becomes important to the company in evaluations of areas with operating or working
interest that may not be prime acreage, or in blocks away from the play activity core. If data was to be ignored due
to its location outside the operating area of interest, the decision to be made would be extremely time consuming and
very difficult.

Figure 13: Property distribution methodology. (a) Secondary property ranking. Total Porosity distributed using Bulk Density as an imposed
positively correlated spatial trend. (b) Third-ranking-order properties. Water Saturation distributed using Total Porosity as an imposed inversely
correlated spatial trend. (In histograms) Distribution of full resolution logs (pink bars), distribution of upscaled logs (green), and distribution of
modeled property (blue).

This simple but innovative way of modeling is more effective, practical, and quality controlled than the alternative
of constructing numerous, unrelated, independent sector models. A critical downside of constructing small-
disconnected models is that they rely on a very limited number of pilot wells restricted to their reduced area of
interest, lacking the connection to important regional trends. Previous work and experience with the workflow in
other shale plays (Cook et al. 2014) allowed us to visualize and understand the need for a seamless tie between
smaller models extracted from the sub-regional model (colored boxes in Figure 7b and example in Figure 14).

By following our methodology, the technical group will always be able to provide an answer, even if that answer is
at varying resolution, or with a different degree of certainty. At any rate, the team will not be blind due to a lack of
model coverage. To do this effectively, the data cannot stop at an arbitrary line and needs to be encapsulated
seamlessly in the regional play scale model (Figure 14).
URTeC: 2460295 11

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


Figure 14: On-demand modeling. Seamless tie between smaller “shoe-box” models extracted from the sub-regional model. Shoebox models are
extractions that include a single horizontal well and immediate surrounding area. Smaller model extractions allow for the ability to obtain3D
representative properties anywhere in the sub-regional area of interest, with no edge effects and the benefits of the full petrophysical set (181
wells). This type of extractions is possible from all of our models at any scale desired as required.

Geomechanical Units

The identification of different materials based on their stiffness is critical to understanding the way they respond to
stress. The key is to separate materials based on mechanical stiffness (hard vs. soft, Ferrill and Morris, 2008; Ferrill
et al., 2011; Smart et al., 2014; Slatt and Abousleiman, 2011; and many others). A geomechanical study was
conducted for the Haynesville section in our area of interest. The study was based on core (rebound hammer) and
logs (Young’s modulus) to identify the variability of stiffness of the rocks to be modeled. Our modeling effort is
based on the assumption that under stress, stiffer materials undergo failure, resulting in higher fracture intensity, and
softer materials accommodate more strain as “ductile” deformation, in turn showing lower fracture intensity.

Figure 15: Mechanical and DFN Units. Values in log tracts (sample well) increase o the right (a) Mechanical Stratigraphy Type Log. Track 1:
Wireline GR (color filled), Depth Shifted Core GR for 2 Coring Runs. Run 1 (blue) Run 2 (red). Track 2: Depth Shifted Hardness measured on
core using Bambino. Track 3: Young’s Modulus. Track 4: Two iterations of smoothing of the raw hardness measurements (see track 2). Data is
smoothed to easily compare to Wireline logs. Track 5: Upscaled mechanical stiffness. Track 6: Geologic modeling Units. (b) Example of 12ft of
core – 1 box of Haynesville Core. Orange circles represent sample locations using a mechanical rebound hammer (Bambimo). (c) Example of 6
DFN units selected for modeling (color bar to the left). Stratigraphic unit tops (colored spheres) and Young’s Modulus log in 3D model (hot
colors high YM, cool colors low YM). Bold colored lines show average YM used to combine units with similar Mechanical Stiffness.
URTeC: 2460295 12

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


Figure 16: Representation of the workflow followed to implement Mechanical properties in construction of DFN (a) Mechanical Stratigraphy
Type Log. Comparison of young’s Modulus (tract 2) and rebound Hammer (tract 3) in type well. Young’s modulus shows very good relationship
with mechanical stiffness. (b) DFN units selected for model. (c) Example of YM property distribution in model. Sample Pilot shows cells at a
vertical well location tied to the logs. Modeled YM property along horizontal well; Cross section shows the distributed property in the model,
based on our official log database.

We used a sampling interval of 6 inches along the core, to closely match the wireline log resolution. Data covered
for this part of the project includes around 7000 feet of core. The end result of this work, was a Hardness Value
record along the core (measured in Leebs), that was depth shifted using the core gamma ray. We used the data as a
proxy for Mechanical Stiffness, to be compared to Young’s Modulus. See Zahm and Enderlin (2010), McClave
(2014), Lee et al. (2014), Lee, (2015), Ferrill and Morris (2008), Ferrill et al, (2011), (Figure 15a).

For the purpose of DFN construction, we combined the geologic units into 6 zones, based on our understanding of
mechanical stiffness. We qualitatively implemented the average young’s modulus as the parameter to combine units
that show a similar response (Figure 15c). The result was the grouping of mechanically related units into compatible
zones (Figures 15a and 15c). Using our official log database we distributed Young’s modulus (YM) and other
geocellular mechanical properties in our model. This provided us with readily available mechanical properties as a
proxy for mechanical stiffness (Figure 16a) and the ability to extract them along horizontal wells (Figure 16c) in the
area of interest (Figure 7b, blue polygon).

Core Study, Fracture Characterization

The natural fracture analysis encompassed a comprehensive campaign, covering more than ~3000ft of core. For the
purposes of feeding the DFN model, we identified five critical tasks. Natural fracture validation, verification of
mineral fill, recording the width of mineralized fractures, fracture spacing, and fracture intensity.

The most important part was to ensure that the features identified were indeed natural fractures that originated in the
subsurface and not artificially induced, or fossils masquerading as fractures. In our study, we only counted clearly
mineralized natural fractures (Figure 17c). Fossils in the Haynesville section can resemble mineralized fracture fills
to the untrained eye. This is due to the crystalline appearance of their carbonate shells when observed in side view.
Further inspection of the features, in samples partitioned along the shell, clearly revealed the fossils and served as
analogues for fossil identification in other sections of the core (Figure 17a). A very experienced geologist in our
group was fully devoted to the study for its entire duration, ensuring the quality of the natural fractures identified.
URTeC: 2460295 13

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


Figure 17: Five key tasks in fracture analysis for DFN modeling. Natural fracture validation, verification of mineral fill, recording the width of
mineralized fractures, fracture spacing, and fracture intensity. (a) Identification of fossils masquerade as fractures. (b) Natural Fracture families.
Core piece showing 2 vertical fractures (Family 3) and 1 horizontal fracture (Family 1). (c) Photo of a large mineralized vertical fracture (Family
3) and a smaller, also mineralized, oblique fracture (Family 2).

We documented the width of the fracture fill (Figure 17c) and the spacing of the natural fractures for each set
(Figure 17b) in detail. In this paper we only cover the fracture intensity (fracture count by foot of core), which is the
key input for the construction of our fracture intensity logs that feed the DFN model.

In addition to the 26 wells with core that we included in the study we contacted an experienced consultant that
reprocessed and interpreted 11 wells with borehole image logs. This encompassed the database that was used to
identify the 4 natural fracture families we used in the model and to assign the dip azimuth and magnitude to the
fracture sets. The borehole image data (oriented data) supplemented the natural fractures in each of the families
identified in core (non-oriented). The four families are: 1) horizontal fractures, 2) oblique fractures, 3) vertical
fractures, and 4) horizontal micro fracture swarms (Figure 18).

Figure 18: Fracture and Fill Types. Four fracture families identified in core study that we used to model the fracture component of our system
(type 2, fractured reservoir). (a) Family 1. (b) Family 2, (c) Family 3, (d) Family 4.
URTeC: 2460295 14

Natural Fracture Modeling, DFN

A discrete fracture network is the representation of the natural fractures in a fractured reservoir system. The DFN
model aims to represent the characteristics of naturally fractured reservoirs from the detailed scale to the scale of the
system. The characteristics of natural fractures are based on detailed observations in core, well-logs, and borehole

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


image logs, amongst others. Large scale components and behaviors of the system are interpreted from well to well
tracer studies (Sani et al., 2015), pressure data, seismic attributes, and microseismic clouds, amongst other
technologies for which we find new uses every day, in the development of unconventional resources.

The DFN is the most critical variable in the study of type 2, fractured reservoirs (Figure 10b). In these systems, the
matrix provides the hydrocarbon storage and fractures provide the essential reservoir permeability (Nelson, 2001).
Natural fractures provide permeability, that is orders of magnitude higher than the matrix, for hydrocarbon
molecules to move from a nano-darcy region to another, until we connect to the system and accelerate flow, with the
hydraulic fracture treatments we conduct in each of the stages of a horizontal well. For this reason, it is very
important to understand to the best of our ability and available technology, both the matrix and the natural fracture
components of the integrated system. The characteristics of the hydraulic fractures were also addressed by our
proposed methodology and we cover them in the planar fracturing simulator (Hydraulic Fracture Modeling) section
of this paper.

The foundation for the construction of the DFN in our study is twofold; the mechanical properties generated in the
sub-regional static model and the field wide natural fracture property distributions. All properties that feed the DFN
were generated independently for each mechanical unit. This guaranteed that only characteristics from compatible
materials (rocks) were included in the statistics of each zone. We did this to achieve as close as possible,
distributions with no data contamination.

Fracture Families and Sets

This portion of the project was completed as a full time teamwork exercise, between very experienced professionals;
the reservoir geologist building the fractured reservoir model and the geologist that recorded all fracture properties
in the core study. The key elements of success were patience, discipline, and having a clear methodology to follow.
For this DFN we consider that 4 families, with 2 fracture sets each, modeled independently by mechanical unit, are
sufficient to describe the fracture component of the system (Figure 19) in what we believe is a quiet structural area.
The DFN was constructed away from any seismic resolution faults, in an area with no major folds. This is
corroborated by horizontal drilling of multiple wells (>4000 ft., laterals) landed in the Haynesville, with no large
structural folds or faults reported along the target lines (Figures 4 and 6). The reasons for the area selection are very
simple; the area is very prolific (pressure gradient of 0.8 to 0.9 psi/ft) and it is representative of acreage of interest to
the business, where we are most active (Figures 4 and 19a).

The four fracture families identified in the core study were used in the DFN model (Figure 18).

Family-1 groups horizontal fractures. These fractures originated through catagenesis (Al Duhailan, et al., 2015), as
the source rock matured promoting the ability for overpressure to generate horizontal cracks that became
mineralized by supersaturated fluids. This occurred repeatedly as evidenced by the multiple events recorded in the
mineral fill of well-developed horizontal fractures (family 1, Figure 18a). Fractures in this family have dip
magnitudes that go from 0o to 16o and a dip azimuth from 0o to 360o with no preferential trend (Figure 20a). The
random distribution of dip azimuth is characteristic of all fractures in this family. We believe this is a consequence
of the mechanism responsible for the fractures (catagenesis). Horizontal fractures in family-1 display a clear sharp
contact between the fracture fill material and the matrix host rock (Figure 18a). Organic-rich-softer rocks
(mechanical units) display higher horizontal fracture intensity than stiffer and less organic rich rocks (see
comparison between 19b and 19e).

Family-2 groups oblique fractures. We interpret family-2 fractures observed in our study area, as the result of
mechanical units reaching their elastic limits and breaking under stress. Fractures in this family have dip magnitudes
from more than 16o to less than 60o (Figure 18b and 20b).
URTeC: 2460295 15

Family-3 groups the vertical fractures. Just as we observed in the oblique fractures, we currently interpret them to be
the response of the consolidated rock to accommodate stress. Fractures in this family have dip magnitudes that go
from 60o to 90o (Figures 17c, 18c and 20c).

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


Figure 19: Haynesville Natural Fracture Official Database in Model. (a) This DFN area. (b) Fracture intensity (FI), horizontal fractures.
(c) Planes interpreted from borhole image logs, discs. (d) Official tops, notice yellow sphere above last disc. (e) FI, vertical and oblique fractures.

A relationship is clearly observed for Families 2 and 3, between the mechanical stiffness of the material (rock) and
their fracture intensity, both in core and borehole image logs. Stiffer rocks (mechanical units) display higher oblique
and vertical fracture intensities than softer rocks (see core sample in Figure 17c. compare 19b and 19e).

The dip azimuth distributions for both families (family 2 and 3) showed clear trends that allowed the grouping of
fractures into two sets; a primary set (systematic) and a secondary set (non-systematic).

The primary set includes fractures with dip azimuths between 90 o and less than 180o as well as fractures with dip
azimuths between 270o and less than 360o (Figure 20b and 20c). In our area of interest, primary sets display the
closest plane orientations to the observed maximum horizontal stress direction that is N86 oE (Figure 20c). The
secondary set is generally orthogonal to the primary set of both families (family 2 and 3). Dip azimuths for fractures
in the secondary set go from 0 o to less than 90o and between 180o and less than 270o (Figures 20b and 22b).

Family-4 groups horizontal micro fracture swarms. Based on core observations (Figure 18d) we currently interpret
this family, as an early stage of fracture development, compared to the fully developed horizontal fractures in
family-1. The only difference observed, for the purpose of modeling, between family-4 and family-1 fractures, was
the swarm like nature of the aligned micro cracks within the host rock, characteristic of family 4 (Figure 18d). Dip
magnitude and azimuth ranges in the model are the same as used for family-1 (Figure 20a).

Fracture counts were recorded in the core study, documenting number of fractures observed per foot of core. This
was done separately for each family, to generate individual core-fracture intensity logs (Figure 19b). The core study
data (26 wells) was complemented by 11 wells with fracture interpretations from image logs (Figure 19c and 19e).
This dataset included logs with fracture dips and dip-azimuths (Figure 19c) and borehole-image-fracture intensities
(Figure 19e). All borehole image intensity logs, received from experienced interpreter, had been corrected for well
path orientation. This was confirmed with the interpreter, to avoid using the apparent fracture spacing that would
have been present if the deviation survey had not been used in conjunction with interpreted fracture discs to perform
the correction (Figure 19c).

To obtain fracture aperture data for the model, in this project we used placeholder initial values. These feed the first
pass distribution of apertures for each fracture set. These values are a starting point that in our workflow are tuned
later, as we move to simulation phase. In our workflow we calibrate apertures based on dynamic well tests that are
URTeC: 2460295 16

far less uncertain than traditional and time consuming static measurements on un-stressed non-preserved core
(Figures 17c and 18a).

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


Figure 20: Orientation and dip magnitude for sets of planes in DFN model. Rose diagram showing trends of planes (Strike). Points are the poles
of planes (dip azimuth direction and dip magnitude). (a) Horizontal planes plot in the center. (c) Vertical planes plot toward outer edge of plot.

DFN plane dimensions (Average ~650ft) are based on proppant tracer studies, that will be covered in the Tracer
Analysis section of this paper. Sizes of discrete plane elements used in this project are based on recommendations in
the SPE Liquids Rich Shale Workshop, from experiences in various type 2, fractured reservoirs in USA (Knitter et
al. 2015). The key principle here was to build the DFN at the scale of system, to model the effect of the fracture
corridors in the model and their interaction with hydraulic fractures along the well (Figures 21top and 30a).

We completed the DFN for this study using widely available commercial modeling software, in addition to a
discrete fracture modeling plugin. We use the same modeling package as our production unit teams. In this paper,
we do not show the seismic in our acreage of interest, but area trends as interpreted from seismic attribute analyses
indicate fracture swarms extending approximately 2,500 ft. (on average) can be interpreted. Specifically, in
enhanced similarity volumes as described by Dewett and Henza (2016). Similar features are documented in the
literature for two unconventional plays (Dewett et al. 2016).

In this paper, we focus on the DFN planes modeled for the immediate surrounding area along the wells selected for
our future simulation phase (Figures 21top and 30a).

Figure 21: Construction of Discrete Fracture Network (DFN) Model. The Haynesville shale is a naturally fractured reservoir, hence hydraulic
fractures interact with natural fractures. Core photos show the 3 major families in the model.
URTeC: 2460295 17

Tracer Analysis

The process to incorporate tracer data in our study focuses on the analysis of proppant tracer migration and fluid
tracer communication to identify potential pathways (Sani et al. 2015). The results confirm that in this area, fracture
network elements need to be long enough to match the diagnostics studies. Four DFN models were constructed,

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


achieving best fit to observations on the fourth iteration. A brief description of the process followed to use well
diagnostics data for the selected DFN model is presented in this paper (V4, Figure 21top and 30a). Plane elements in
our initial high resolution DFN models (versions V1, V2, and V3) were too small to match the connectivity
observed.

Proppant tracers were added in Well #1 and detected in offset Well #2 in 4 different stages. Well #2 was completed
a week after the completion of Well #1 (Figure 22). The distance covered by the proppant tracer migration from
Well #1 to Well #2 was about 1,000 ft. This is evidence of well-to-well migration through potential fracture network
pathways at this scale. On average, about 4.3% of the fluid tracers injected in Well #1 were collected in Well #2
continuously for about 4 months. Most of the fluid tracers recovered from Well#2 came from the toe section of Well
#1. About 76.4% of the fluid tracers injected in Well #1 were collected as flowback, with the remaining balance
collected in offset wells (not shown in this paper). These wells are located to the left (1,700 ft.) and to the right
(4,900 ft.) of Well #1. Fluid tracers recovered in these wells originated from different stages in Well #1. This is
evidence of well-to-well fluid communication.

Based on the tracer analysis, version V4 of the DFN was identified as the most likely plausible solution. One of the
most critical inputs in the analysis of type 2 fractured reservoirs is the DFN. Therefore, significant effort was taken
to develop a robust workflow facilitating the fine-tuning and calibration of models with the observed tracers’ data.
Our interpretation is that the effect of the stimulation treatment of Well#1 (Figure 22) and its interaction with large
fracture networks (fracture corridors), is very much related to the potential pathways taken by the tracers.

Figure 22: Well diagnostics. Proppant tracer case in area of Interest. (a) Proppant tracers detected in offset Well #2 in 4 stages—evidence of
proppant migration through possible pathways. Fluid tracers collected in offset Wells—evidence of significant well-to-well fluid communication.
Not all stages show proppant tracer migration. Well #2 was not completed at the time of stimulation of Well #1. Linear migration path shown
in figure, for representation purposes only. (b) Maximum horizontal stress direction in area of interest (N86oE).

Planar Fracturing Simulator (Hydraulic Fracture Modeling)

The hydraulic fractures are what connect the reservoir system to the well. We used a planar hydraulic fracturing
simulator to understand the fracture conductivity and net pressure profiles along hydraulic fracture planes. We used
this to feed our geomechanical model and prepared a hydraulic fracture connectivity property to be used as input in
our future flow simulation and history matching phase.
URTeC: 2460295 18

To translate the mechanical stiffness observations we obtained from the core and logs from our geomechanical
study, we used YM generating the stiffness breaks that control the toughness parameter used in the hydraulic
fracture simulation (Figure 23).

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


Figure 23: Implementation of layering (based on YM). Support for Toughness control added in Hydraulic fracture simulation. (Top)YM and
toughness control at most representative pilot well. (Bottom) YM and toughness control along lateral well used in the hydraulic fracturing
simulator.

For this portion of the workflow, the fracture geometry was generated by a commercial hydraulic fracturing
simulator, using the calibrated geomechanical properties, and the actual pumping and material data from the field
reports. The actual net pressure (P_net) for each stage was estimated based on instant shut-in pressure (ISIP) and
minimum horizontal stress (σ_hmin), for all clusters in one stage (Figure 24a).

Figure 24: Hydraulic Fracture Modeling (a) Net Pressure Output, all clusters in one stage. (b) Conductivity Output, all clusters in one stage.
Result for selected well for future flow simulation.

A single stage of hydraulic fracturing simulation was performed matching the net pressure to stage level for the six
clusters. This was in agreement with the actual estimated net pressure, also at the stage level. The fracture half
lengths (total un-propped frac) of the six clusters were around ~1,300 ft (this is not effective propped frac), and the
fracture heights were on average ~350 ft. (Figure 24b). Due to the stress shadowing effect, the inside clusters
showed relatively shorter half lengths and higher net pressures.

Based on the understanding we gained at the stage level and from production logging tool (PLT) data in a nearby
well, we developed a hydraulic fracture connectivity property. This property, that we call the “fish bone” allows us
URTeC: 2460295 19

to expand or contract laterally and vertically (3D) the extent of propped hydraulic fracturing in the simulator. This
can be done for a volume as extensive or reduced as needed, as a group of stages (Figures 25a and 32), or for each
stage independently (Figure 33). This property will be used as an additional tuning control for the reservoir engineer
in our future flow simulation and history matching phase.

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


Figure 25: (a) Completion clusters from Planar Hydraulic Simulator installed inside Flow simulation model “fish bone”. (b) Hydraulic fracture
geometry and conductivity simulated for a single stage. Non-propped half lengths were around ~1,300 ft, and the fracture heights ~350 ft.

Geomechanical modeling

The first step in the construction of a 3D geomechanical model is to ensure that the stresses and mechanical
properties are properly calibrated to wells inside of the volume of interest. Therefore, it is mandatory to start from a
rigorous 1D geomechanical model that is built upon the highest level of resolution at the wellbore scale (Figure 26).
This enables the appropriate level of upscaling to the 3D domain, reducing the resolution in the overburden sections,
while maintaining the needed resolution at the reservoir formation level, as well as in the immediate surroundings
above and below.

Figure 26: 1D Geomechanical Models. (To the right) Map insert showing DFN area of interest, 3 wells in close proximity with 1D
Geomechanical models. Less than ~4miles linear distance between wells with 1D models (orange crosses). On Log: Blue curve is pore pressure,
orange curve is minimum in-situ stress, purple is maximum in-situ stress, and green is over burden. (a) Haynesville to Bossier section. Relatively
low stress and YM contrast. Hydraulic-fracture height more controlled by pressurization of frac-fluid. (b) Smackover to Haynesville section.
Relatively high stress contrast, High YM contrast. Both aspects should hinder fracture growth downwards below the Haynesville.

In this study, we used three wells to constrain the stress field and mechanical properties in the area of interest
(Figure 26). The process involved the use of drilling, completion and formation evaluation, including drilling and
completion reports, geological reports, petrophysical data and rock mechanics testing.

Figure 26 shows a typical 1D-geomechanical model from one of the wells analyzed. One of the main highlights of
the Haynesville play is the high pore pressure levels, that can reach gradients above 0.8 psi/ft. The current dominant
URTeC: 2460295 20

stress state in the region is extensional, where the overburden is the maximum stress. This, in conjunction with the
high overpressure, yields a low level of stress anisotropy. Another highlight of the play is the similarity that exists in
terms of mechanical properties between the Bossier and Haynesville sections that impacts the hydraulic fracture
design and execution (Figure 26).

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


The 1D model was used to populate the mechanical properties within the 3D model, and as a calibration control of
the stress initialization process in the Finite Element Method (FEM) calculations. It was also used to carry out a first
pass evaluation of the critical stress condition on natural fracture families in the geologic model (Figure 27).

Figure 27: Identification of Critically Stressed Fracture Sets. First-pass evaluation of all sets in fracture Families. Family-1 (Horizontal
fractures), family-2 (oblique fractures), family-3 (vertical fractures), and family-4 (horizontal micro fracture swarms). Critically stressed
fractures shown as white dots in all plots. Families 1 and 4: Horizontal fractures, not-critically stressed, in our current modeling scenario.
Familities 2 and 3 : Oblique and Vertical Fractures are critically stressed in our models.

By testing, each of the families individually using the 1D model to verify the likelihood of shear re-activation, it was
possible to filter out the sub-horizontal Families 1 and 4 as the ones that would not re-activate under stress changes
due to hydraulic fracturing operations, as they are under the highest level of normal stress (i.e., the vertical stress).
Conversely, families 2 and 3 have the highest potential to shear due to their optimum orientation to the maximum
shear stress. For this reason, only families 2 and 3 were simulated in the FEM analysis for the calculation of
synthetic microseismic and estimation of Critically Stressed Fracture Volume (CSFV).

Figure 28: Stress shadowing during stage fracturing. Net pressure increased (toward heel) indicating stress shadowing.

After the stress initialization achieved the same magnitudes and directions of the 1D models at the pilot well
locations, the model was set to carry out the analysis of the impact of hydraulic fracturing on the stress evolution
after each stage and associated microseismic events. One term that has been used quite often in the development of
unconventional reservoirs is the so-called stress shadows, which is nothing less than the stress changes in the area
near the hydraulic fracture.

As rocks split, the stress state is required to find a new equilibrium point. As this process occurs, stresses are
transferred to the regions near the hydraulic fracture. This effect has been observed in the field, as the wells are
completed, by the increase of net pressures in hydraulic fractures from toe to heel (Figure 28). This increase in net
URTeC: 2460295 21

pressure occurs due to the stress evolution after each fracturing stage. The stress shadow effects are a function of the
mechanical properties of the rock and the virgin in situ stress state. This effect is compounded with shorter stage
lengths, the shorter the spacing between stages, the greater the effects that are observed.

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


Figure 29: Stress shadowing. 3D geomechanical model. Model is able to capture the stress shadows of each stage. Stress shadowing is the
mechanism that produces the synthetic microseismic events (white spheres).

A good match was obtained by the FEM simulations of stress shadowing effects on the net pressure, or equivalent
minimum horizontal stress, (Figure 28). A clear increase in the magnitude of the minimum horizontal stress was
captured by the simulation. The FEM simulation imposes a hydraulic fracture plane with dimensions as estimated by
the hydraulic fracture simulator with each stage simulated one at a time, aiming for the stress state to re-equilibrate
prior to the simulation of the subsequent stage. The good match observed between the increase in the net pressure in
the field and the numerical FEM simulations indicates that the geomechanical model is able to capture the stress
shadows of each stage and the related effects on the DFN. This is the mechanism that yields synthetic microseismic
events as shown in Figure 29 by the white spheres. The cloud of events shows a tendency to cluster around the
hydraulic fractures; this could be a function of the relative ductility of the Haynesville shale (Figure 29).

Figure 30: The Critically Stressed Fracture Volumes (CSFV). (a) Critically stressed planes in DFN. (b) Identified critically stressed fracture
volume (CSFV). The three colored regions (red, yellow and orange) represent low, mid and high cases of criticall stressed volumes. Green
ovals: Stimulation in subsequent wells could start in less stimulated areas. Nearby well could require fewer stages. With the knowledge of
possible CSFV, planning for zipper fracking could result in more beneficial treatment.

The synthetic microseismic events are the result of shearing along planes in the DFN. The shear re-activation, breaks
down any cementation associated to fracture planes and should enhance permeability by increasing the connectivity,
promoting fluid flow between fractures. Figure 30a shows the intensity of discrete planes within the CSFV. Notice
that the central part of the well appears to be the section that achieved the highest level of shear interconnectivity of
URTeC: 2460295 22

natural fractures. That being the case, the next well would probably have the potential to better stimulate other
regions with less interconnection in the first well, by strategically placing the hydraulic fracture stages (Figure 30b).

Flow Simulation Ready Properties

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


The results from running our workflow will be incorporated into a reservoir simulation model. History matching of
production data will be conducted for validation and refinement. History matched models will be used to evaluate
the impact of key drivers of optimization studies to various field development scenarios in order to enhance well
completion and well spacing strategies in the development plan. In order to achieve this and be ready for the
simulation engineer in our next phase, we prepared various simulation ready properties.

As presented in our geomechanical modeling section, a CSFV can be generated based on the stresses and the
completion treatments conducted on study wells. The CSFV allows the identification of zones of expected optimal
and poor drainage (Figure 31) that can positively influence the way we continue to stimulate subsequent wells, as
the pad is developed (Figures 30b and 31).

Figure 31: Application of critically stressed fracture volumes (CSFV), Preliminary Results.

Identification of Stimulated Rock Volumes (SRV)

The goal of static and dynamic models is to properly characterize the stimulated rock volume (SRV). This represents
the volume of rock that will flow as we produce the wells. In these terms, SRV is a dynamic property. With the
CSFV and the hydraulic fracture connectivity property in place (fish bone), we can expand laterally and vertically
(3D) the extent of propped hydraulic fracturing in the simulator. This can be done generating multiple versions of
the SRV by expanding all clusters together (Figure 32), or independently for each cluster (Figure 33).

Figure 32: Preparation of stimulated rock volume (SRV). Example of expansion of all clusters together. Top: instalation of fishbone. Bottom:
Four samples of SRV models. Unlimited number of models can be created by engineer based on material balance.

In the case where clusters are expanded or contracted independently, portions of the well within the CSFV can be
assigned a larger volume due to the presence of hydraulic fractures and critically stressed Natural fractures. On the
URTeC: 2460295 23

other hand, well regions outside of the CSFV can be assigned smaller volumes, to only account for the propped
portion of the hydraulic fractures, given that natural fractures will not be critically stressed (Figures 31 and 33).

With all hydraulic fracture conductivity and geometry properties ready, the DFN was translated from the plane
element domain to the geo-cellular domain that is required for the simulator. We translated the DFN into the geo-

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


cellular property domain using the ODA method (Oda, 1985). With this we obtain the 3 component natural fracture
permeability (Figure 34a, I,J,K directions), natural fracture porosity (Figure 34b), and the shape factor (Figure 34c,
Sigma) the simulator uses to communicate between the matrix and natural fracture domains (Kazemi et al. 1976).

Figure 33: Preparation of stimulated rock volume (SRV). Example of expansion, independently for each cluster. Red volume is CSFV.

All simulator ready properties, that include the hydraulic fracture geometry and conductivity, the critically stressed
fracture volume (CSFV), the scenarios for identification of stimulated rock volume (SRV), and natural fracture
reservoir properties translated from our DFN model, were loaded and reviewed in the simulator (Figure 34d). We
conducted this exercise to ensure no changes or unwanted modifications occurred to the properties during
initialization, in preparation to our next phase that includes the production history-matching portion of the project.
The dynamic flow simulation step, in our full loop: Static-to-Geomechanics-to-Flow.

Figure 34: Haynesville DFN preliminary parameters. Reservoir Properties for the Natural Fracture Component of the System and Sample
Property Loaded in the Simulator. (a) Natural fracture permeability, I-direction. (b) Natural Fracture porosity. (c) Shape factor (sigma), natural
fracture to matrix coupling property. Summary of available outputs for simulation. (d) Property loaded in simulator.
URTeC: 2460295 24

Implications

Having a properly calibrated model based on critical petrophysical properties that cover both the natural fracture and
matrix components of the system is vital in unconventional reservoirs. These properties can provide support 1) in the
selection of completion design and treatment, 2) for smart well placement (staggering), and 3) by having a better

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


understanding of the reservoir by the entire team.

Smaller sector models extracted from a robust high-resolution regional model are useful for simulation and for
implementation by development teams in planning and drilling activities. The main advantage of these models is
that they can be extracted anywhere in the area of interest, with all the benefits of the full petrophysical analysis, and
the ability to prepare them as soon as the team and the business requires.

Conclusions

This paper provides an integrated full loop seismic-to-simulation reservoir characterization workflow for
unconventional fractured reservoirs. Our value proposition covers the static parameters that include 1) matrix:
storage and 2) natural fractures: flow. It integrates them with the dynamic parameters that include 3) hydraulic
fractures and stimulation treatment, 4) the geomechanical controls, and 5) calibration to well diagnostics which are
all key inputs for flow simulation and history matching. All of this, within a Static-to-Geomechanics-to-Flow, in-
house solution.

The critically stressed fracture volume (CSFV) proposed in this paper, allows for better identification of
contributing versus non-contributing stages along a horizontal well. This is based on a robust integrated model,
developed by the combined knowledge and experience of the professionals in the asset and the company’s group of
experts in shale reservoir characterization.

Acknowledgements

The authors would like to thank the support and technical staff in the asset team for their contributions to this work,
and committed management in the asset and the functions for promoting integration. We would also like to thank
advisors, mentors, and other coworkers for sharing their knowledge and experience, and BHP Billiton Petroleum for
their support and permission to publish this study.

References

Abu M. Sani, SPE, Seth B. Podhoretz, SPE, and Brent D. Chambers. The Use of Completion Diagnostics in
Haynesville Shale Horizontal Wells to Monitor Fracture Propagation, Well Communication, and Production Impact.
2015. SPE-175917-MS presented in SPE/CSUR Unconventional Resources Conference held in Calgary, Alberta,
Canada, 20–22 October.

Al Duhailan, M. A., Sonnenberg, S. A., Longman, M. Analyzing “Beef” Fractures: Genesis and Relationship with
Organic-Rich Shale Facies. 2015. URTeC: 2151959 presented at the Unconventional Resources Technology
Conference held in San Antonio, Texas, USA, 20-22 July 2015.

Cicero, A., Steinhoff, I., 2014, Sequence Stratigraphy and Depositional Environments of the Haynesville and
Bossier Shales, East Texas and North Louisiana, Geology of the Haynesville Gas Shale in East Texas and West
Louisiana, U.S.A, AAPG Memoir 105, pp 25-46.

Cicero, A. D. et al., 2010, Sequence stratigraphy of the Upper Jurassic mixed carbonate/siliciclastic Haynesville and
Bossier Shale depositional systems in East Texas and North Louisiana: Gulf Coast Association of Geological
Societies Transactions, v. 60, p. 133-148.

Cook, D., Downing, K., Bayer, S., Watkins, H., Sun Chee Fore, V., STansberry, M., Saksena, S., and Peck, D.
Unconventional Asset Development Workflow in the Eagle Ford Shale. 2014. SPE-168973-MS presented in SPE
Unconventional Resources Conference –The Woodlands, Texas, USA, 1-3 April.
URTeC: 2460295 25

Dewett, D. T., and Henza, A. A., 2016, Spectral Similarity Fault Enhancement, Interpretation, Vol. 4, No. 1
(February 2016); p. SB149-SB159.

Ewing, T., 2001, Review of Late Jurassic Depositional Systems and Potential Hydrocarbon Plays, Northern Gulf of

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


Mexico Basin, Gulf Coast Association of Geological Societies Transactions, Volume LI, pp 85-96.

Ferrill et al., 2011, Fault zone deformation and displacement partitioning in mechanically layered carbonates: The
Hidden Valley fault, central Texas, AAPG Bulletin v 95 no 8, pp. 1383-1397.

Ferrill and Morris, 2008, Fault zone deformation controlled by carbonate mechanical stratigraphy, Balcones fault
system, Texas, AAPG Bulletin v 92 no 3, pp 359-380.

Hammes, U., Frebourg, G., 2012, Haynesville and Bossier mudrocks: A facies and sequence stratigraphic
investigation, East Texas and Louisiana, USA, Marine and Petroleum Geology 31, pp 8-26.

Hammes, U., Hamlin, H. S., and Ewing, T. E., 2011, Geologic analysis of the Upper Jurassic Haynesville Shale in
east Texas and west Louisiana, AAPG Bulletin, V. 95, NO.10 (October 20011), PP. 1643-1666.

Kazemi, H., Merrill, Jr, L. S., Porterfield, K. L., Zeman, P. R. Numerical Simulation of Water-Oil Flow in Naturally
Fractured Reservoirs. 1976., SPEJ-5719. SOCIETY OF PETROLEUM ENGINEERS JOURNAL. p. 317-326.

Kleban D., and Fulton M., 2016., 4 Tips to Brew Like Maine Beer Company
http://www.homebrewersassociation.org/how-to-brew/4-tips-brew-like-maine-beer-company/

Knitter, C., and Doe, T., Natural Fractures and Their Importance to Shale Oil Development. 2013. Presented in SPE
Liquids Rich Shale Workshop, May 15-17, 2013.

Lee, 2015, Calibration of Rebound Hardness Numbers to Unconfined Compressive Strength in Shale Formations,
Journal of Petroleum Technology, Young Technology Focus, January 2015.

Lee et al., 2014, New Applications of Rebound Hardness Numbers to generate Logging of Unconfined Compressive
Strength in Laminated Shale Formations, American Rockk Mechanics Association, ARMA 14-6972.

McClave, 2014, Correlation of Rebound-Hammer Rock Strength With Core and Sonic-Log-Derived Mechanical
Rock Properties in Cretaceous Niobrara and Frontier Formation Cores, Piceance Basin, Colorado, URTec:
1921941.

Nelson, R.A., 2001, Geologic Analysis of Naturally Fractured Reservoirs, Gulf Professional Publishing/
Butterworth-Heinemann, Houston, 332p.

Oda, M., 1985, Permeability tensor for discontinuous rock masses: Geotechnique, v. 35, p. 483.

Passey et al., 1990, A Practical Model for Organic Richness from Porosity and Resistivity Logs, AAPG Bulletin v
74 No 12, pp 1777-1794.

Santacruz, C., Esquivel, R., Smith, J., Walker, R., Bayer, S., Soto T., Wunderle M., Bansal, A., Goudge, R., BHP
Billiton Petroleum. Integration of RTA Based Reservoir Surveillance and Analytical Flow Simulation to Forecast
Production in the Haynesville Shale. 2015. URTeC: 2153918 presented at the Unconventional Resources
Technology Conference held in San Antonio, Texas, USA, 20-22 July 2015.

Schmoker, J. 1979, Determination of Organic Content of Appalachian Devonian Shales from Formation Density
Logs, AAPG Bulletin v 63 No 9, pp 1504-1537.

Slatt and Abousleiman, 2011, Multi-scale, Brittle-Ductile Couplets in Unconventional Gas Shales: Merging
Sequence Stratigraphy and Geomechanics, Search and Discovery Article #80181.
URTeC: 2460295 26

Smart et al., 2014, Geomechanical modeling of hydraulic fracturing: Why mechanical stratigraphy, stress state, and
pre-existing structure matter, AAPG Bulletin v98 no 11, pp 2237-2261.

Unconventional Reservoir Services, GRI (Crushed Shale) Analysis.

Downloaded from http://onepetro.org/URTECONF/proceedings-pdf/16URTC/All-16URTC/URTEC-2460295-MS/1333747/urtec-2460295-ms.pdf/1 by Universiti Teknologi PETRONAS user on 30 October 2021


http://www.corelab.com/ps/gri-analysis

Waxman, M.H., and Smits, L.J.M.,1968, Electrical conductivities in oil-bearing shaly sands: Society of Petroleum
Engineers Journal, v. 8, p. 107-122.

Zahm and Enderlin, 2010, Characterization of Rock Strength in Cretaceous Strata along the Stuart City Trend,
Texas, Guld Coast Association of Geological Societies Transactions v 60, pp 693-702.

You might also like