American Oil and Gas Reporter - January 2020 - 89

Production Technology
Rauenzahn says. "These modern automation systems add more sophisticated algorithms that review incoming data and
make decisions to improve operations,"
he comments. "The best algorithms can
predict or diagnose and solve production
problems, as well as offer expert-level
optimization recommendations."
Machine Learning
Machine learning has tremendous potential to automate production optimization
workflows that require time-consuming
analyses of large datasets, says Jesse
Filipi, product manager for Ambyint. He
cites dynamometer card analysis as a
case in point.
"Humans take 10-20 minutes to review
cards for a single well, and they have responsibility for 200-300 wells, so it is
impossible for them to look at every
card," Filipi describes. "With advanced
deep learning, we can analyze all cards
for all wells in real time, find features
that indicate a problem and highlight
them for further investigation."
Using machine learning-based algorithms to classify wells and make routine
optimizations allows humans to prioritize
interventions and focus on next-level problems, Filipi reports. He adds that the algorithms codify field-level knowledge, ensuring it is both applied across wells and
preserved as throughout personnel changes.
Developing effective algorithms requires collaboration between data scientists
and subject matter experts such as petroleum engineers, geoscientists and physicists. "The algorithms also must be tuned
to learn from new data and continuously
refined," he says. "Scale that across all
wells working every day, and the return
on investment from machine learning can
be significant.
"One of the biggest benefits is stopping
overpumping, which can lead to early
workovers and high electricity bills," he
comments. "The vast majority of rod lift
wells are overpumped. By using machine
learning to optimize them, we can reduce
operating costs 20%. At the same time,
we potentially can increase production
from underpumped wells by 10%."
To increase machine learning's benefits,
Filipi says Ambyint is adding new data
and retraining its models constantly to
increase their accuracy and cover new
production scenarios. The company also
runs several models in parallel to answer

different questions.
"Our goal is to use artificial intelligence
to help our customers scale their repetitive
workflows to enable them to have a
greater impact on production," he says.
"This will lead to faster feedback loops
and more quantitative decisions that yield
higher production, longer well lives and
more efficient operations."

Multiphase Metering
After decades of use off shore, multiphase metering is expanding into unconventional plays, observes Sharon McCurdy,
director for Roxar Multiphase North
America, part of Emerson Automation
Solutions. "Like offshore operators, unconventional producers are interested in

INTRODUCING

"The Industry's Superior Choice in Wellhead Systems"

MODEL "J" STUFFING BOX

*

Superior protection
against frack mitigation

*

1500, 3000, and 5000 PSI
pressure ranges available

*

Low Profile integral stuffing box
and pressure balanced BOP,
eliminating potential leak paths

* Safety/environmental
protection with flapper valve
and leak detection systems
*

Enduro bonded coating for
corrosion protection and
replaceable bottom base

Contact us for more information on all TITAN PREMIER™ Products!

www.wellheadsystems.com
785.421.2255

JANUARY 2020 89
http://www.wellheadsystems.com
American Oil and Gas Reporter - January 2020

Table of Contents for the Digital Edition of American Oil and Gas Reporter - January 2020
