American Oil and Gas Reporter - December 2021 - 73

SpecialReport: Stimulation & Completions
require less iron on site. We instrument
these ground iron networks so we can
detect leaks and make adjustments without
putting people in the red zone, which we
are 100% against. "
Increasing horsepower density may
ease simulfracs, which Sagar says generally
require 1.5 times the horsepower of
a standard spread but make it possible to
fracture two wells at the same time.
Getting that much horsepower on location
can be challenging in regions with hilly
terrain and small pads.
Despite those challenges, Universal
is performing simulfracs in both the
Northeast and Texas. " Simulfracs generally
improve the lateral feet treated each day
by 50%, reduce the amount of horsepower
required per wellbore by 22% and shorten
the time on pad by 30%, so they offer
tremendous value, " Sagar says.
For simulfracs to make sense, the
wells' geology needs to be homogenous
enough that they will respond positively
to the same chemistries and pressures,
pad size and wellhead layout need to accommodate
the equipment, and logistical
networks must be in place to deliver
water and proppant, Sagar outlines.
" We think 25%-30% of future activity
will be a good candidate for simulfracs, "
he says. " In many areas, operators who
want to move to simulfracs need to redesign
their drilling programs and pad layouts.
Also, simulfracs require enough planning
and equipment that they generally only
add value with long development programs
that enable continuous operations. "
Getting sand, chemicals and other
products to site for simulfracs takes some
finesse, Sagar adds. " The oil and gas industry
is not immune to the supply chain
issues that are affecting other industries, "
he observes. " We think these issues will
only last six months or so, but in the
meantime, we are doing what we can to
navigate them intelligently. "
That means maintaining long-term relationships
with trucking companies, proppant
suppliers and parts manufacturers,
Sagar says. " The age-old tool of planning
as much as possible and communicating
with suppliers is critical, but we also are
optimizing our product mix and supply
chain timing using cutting-edge machine
learning platforms, " he reports.
Machine Learning
Machine learning can help engineers
work more efficiently and increase fracs'
effectiveness, says Jeremy Miller, the
ML-driven data labeling denoted with
ML-driven event alerts denoted with
Because machine learning algorithms can detect and label subtle events quickly, they
can enable or accelerate completion analysis, Well Data Labs observes. In this image
from a sealed wellbore pressure monitoring project, an algorithm has labeled pressure
waves arriving at either of two monitor wells from the two active treatment wells. Well
Data Labs says SWPM helps completion engineers evaluate several metrics, including
fluid distribution, far-field fracture counts and fracture height and half-length.
vice president of product for Well Data
Labs. " We have had customers get 10x
returns on investment by using ML to
automate simple but common analyses, "
he reports.
As an example, Miller cites the time
it takes for each stage to reach the desired
pump rate. " To calculate time-to-target
rate manually, an engineer would go
through data to pick out the start of each
stage and when it hit the target rate, then
subtract the end time from the start time.
With ML, we can do that automatically
as soon as data is loaded.
" The ML algorithms also compare
each stage to historical data to highlight
any that took longer than planned, as
well as possible causes for the delay, "
Miller continues. " The goal is speed routine
data analysis so engineers can spend
time on more valuable work. "
According to Miller, ML empowers
increasingly busy engineers to optimize
performance by assessing drilling and
subsurface data alongside frac data. " Many
operators would like to analyze drilling
and geosteering data to see whether they
are in zone and detect changes in rock
characteristics that will influence how
the frac behaves, " he says. " With ML,
the foundation of our sealed wellbore
pressure monitoring service, they can
perform these analyses rapidly enough
to apply them to more wells. "
As ML algorithms become more efficient
and computing power grows, operators
will increasingly analyze data in
real time and tweak their completion parameters
during or between stages, Miller
says. " When we first started offering realtime
services several years ago, operators
did not know what they could change because
they never had the capability to analyze
data so quickly, " he recalls. " Today,
they have a better sense of what adjustments
are beneficial and possible to perform. "
Miller says many service companies
are leveraging ML to improve their products.
" For example, perf gun providers
have asked us how they can use machine
learning to understand failures and optimize
design. These lessons will inform product
development and potentially enable them
to advise customers on how to increase
the chances of a smooth frac treatment.
" Today, one of the most common conversations
we have with customers is
how ML can help them evaluate and improve
their ESG metrics, " Miller remarks.
" We are working with operators and service
companies to automate ESG reports
and identify situations where they can
reduce waste and emissions without sacrificing
performance. "
Deploying Machine Learning
Machine learning has a vast opportunity
set in the oil field, Miller argues. " ML
can be a poor fit when data is gathered
manually and sporadically, but when the
data is available and the primary challenge
is aggregating and examining it, it's worth
exploring whether ML can partially or
completely automate the process. This is
true whether that process is simple and
repeatable or involves deep analyses and
DECEMBER 2021 73
Treating Pressure
Slurry Rate
Treating Pressure
Slurry Rate
Surface Pressure
MONITOR
Surface Pressure
MONITOR
American Oil and Gas Reporter - December 2021

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