American Oil and Gas Reporter - January 2021 - 85

Tech Trends 2021
Approach Predicts Trouble Frac Stages
By Eric Romberg,
Aaron Fisher,
Chris Niedz,
Brian Wehner,
Joel Mazza
and Alex Zhou
DENVER-Tracker Resource Development III holds a block of 26,000 contiguous acres in the southeastern Midland
Basin, where it is deploying a 3-D reservoir
model to develop the Wolfcamp formation
using laterals often exceeding 10,000 feet
in length. From an early field planning
stage, Tracker has attempted to identify
and reduce risks to field development
economics, including a large variance in
production results, unproductive sections
of carbonate within the target zone, and
casing failures and completion problems
related to high treating pressures.
To mitigate completion risks, Tracker
is applying machine learning methods
using geomechanical collected at or near
the drill bit, wellbore trajectory and stagelevel completion data to develop models
capable of predicting which stages are at
risk of completion difficulties. Tracker
collected drill bit geomechanical data on
26 laterals, several of which had experienced various problems during completion.
Based on data from these wells, engineers
developed two objective proxies for trouble
stages: a high fracture gradient and low
sand placement.
To predict end-of-treatment fracture
gradients and sand-to-water ratios of each
stage, the data first was subdivided by
target horizon to ensure the model results
were not skewed toward a specific horizon
and the models then were cross-validated,
hyperparameter-tuned, and tested against
a subset of the data that had been withheld
from the training algorithm.
Multiple metrics were applied to measure
the effectiveness of the models. For the
fracture gradient regression, a model that
minimized the root mean squared error
(RMSE) was chosen. This model was compared with a prediction that each stage
would have a fracture gradient equal to the
average fracture gradient. The correlation
between predicted and measured frac gradients also was calculated. These metrics
were analyzed for both the entire dataset
and the test data subset. By all these meas-

ures, the model performed significantly
better than the control prediction.
For the sand-to-water ratio metric, the
model that maximized the receiver operating characteristic area under the curve
(ROC-AUC or AUC) score was chosen.
The model's success was measured by
the precision of its predicted outlier (trouble) stages on the test data subset. The
model performed with a precision of at
least 73%. Both models identified stages
where casing failures had occurred.
By predicting the probability of a trouble stage, mitigation plans can be implemented pre-emptively to reduce completion costs. By further analyzing the models,
engineers can identify the major drivers
for the trouble stages and work to reduce
them as early as the well planning stage.
Horizontal Well Planning
Tracker Resource's acreage block is
situated along the Reagan and Irion county
line near Barnhart, Tx. In this area, the
Wolfcamp B and C intervals often contain
significant amounts of slope sediments
and carbonate debris flows that can cause
drilling and completion issues. The company uses a depth-converted 3-D reservoir
model (consisting of seismic data, cores,
log suites from pilot wells, and log and
well data from offset horizontal and vertical producers) to map a sequence of
stacked landing targets and plan horizontal
wellbores.
Wells are planned in 3-D depth using
a combination of multiple stochastic inversions and neural network-derived geologic models. Acoustic impedance also
is very helpful in identifying carbonate
debris flows (high impedance) as well as
intervals with high porosity and total organic carbon. Wellbores are planned to
avoid geohazards (i.e., interpreted faults,
prominent lineaments and dissolution
collapse features), optimize exposure to
the high-porosity, high-TOC intervals,
avoid carbonate debris flows and drill
with minimal to no dip changes.
Co-blending seismic features of similarity and curvature has yielded a new
tool for identifying faults and fractures
not easily seen in seismic cross-sections.
Tracker evaluates the pre-completion well
bores for any lineaments or trends seen
in the 3-D volume. Completion engineers

are given the depth intervals where any
disturbance was identified and relay that
information to field engineers during
completion. If these identified stages are
problematic during treatment, decisions
easily can be made to change the completion design or simply move on to the
next stage.
Geomechanical data collected while
drilling further reduces risks during plugand-perf completions. Treatments are
planned to reduce variations in minimum
horizontal stress (Shmin) within each stage.
The geomechanical data also identifies
carbonate debris flows within the well
path, allowing completion engineers to
bypass rock considered unproductive.
Even with robust 3-D modeling and
well planning, however, Tracker's wells
have encountered difficult stages with
high breakdown pressures, high pumping
pressures and the inability to place proppant. On a few occasions, drilling out all
plugs was not possible because of casing
obstructions, possibly related to fault activation during stimulation. The operator
began analyzing curvature and similarity
volumes for potential fault/fracture identification near the difficult completion
stages and compromised casing intervals.
Completion Challenges
Tracker drilled and completed 28
Wolfcamp horizontal wells in Irion
County between February 2017 and November 2019. The development proceeded in three phases with eight wells
in phases one and two, and 12 wells in
phase three. While drilling out composite
frac plugs, casing obstruction was encountered on six of the 28 wells (two
wells in each of the three phases). These
obstructions limited the working internal
diameter of the production casing and
either prevented or inhibited access beyond the obstruction.
On two of the wells in phase one,
conventional drill-out assemblies were
not able to pass the obstructions. Several
unsuccessful attempts were made to dress
off the obstructions, but multiple frac
plugs had to be left in wellbores. These
wells underperformed expectations by
15%-60%.
After these issues were observed on
the early wells, several mitigation strategies
JANUARY 2021 85

American Oil and Gas Reporter - January 2021

Table of Contents for the Digital Edition of American Oil and Gas Reporter - January 2021