Hydrocarbon Processing - June 2022 - 80

Biofuels, Alternative/Renewable Fuels
to gasoline blending stocks to reduce the
operators' carbon intensity score. Production
complexity stems from the variety
of feedstocks.
To maintain production volumes and
continuous operation, refiners must anticipate
changing feedstock types. Operators
may be working with crop residues
one day and animal fats the next, so process
parameters can change frequently.
This high degree of variability is the primary
differentiator from conventional
refining and represents the main production
challenge.
Data lost or ignored. The following is a
deeper dive into today's tools, with three
main assumptions:
* Producing renewable diesel is
subject to more process variability
than conventional refining.
* All the necessary tools are at our
disposal to overcome the challenges.
* The same tools can have the same
positive effects in many other
application areas.
All the elements connected with a renewable
diesel production unit produce
data in many forms, and it is up to each
facility to determine how much data gets
collected and retained. Critical process
variables will be kept in a historian, but
the associated storage costs can drive a
company to draw a line between the data
it considers useful and the data it does
not consider to be useful. The challenge
is that many companies underestimate
the value of their data because they cannot
imagine an impactful way to apply all
the data that they have collected.
For the most part, data gets locked in
historians that are isolated on operational
technology (OT) networks, where the
data is stored and processed for release
to management in the form of reports.
An engineer may occasionally dig into a
historian to solve a production problem,
but much of the data simply remains in
storage, never to be accessed again.
Data lakes bridge communication
gaps. Since individuals have access to
numerous tools, one may look at many
companies and conclude that they do
not have enough data, and that they are
not putting their data in the right place.
To start, historians are not the answer.
Historians store data in the format in
which it is produced, so data that is not
directly comparable stays that way, making
analysis more difficult. The solution
involves either replacing historians or interconnecting
them with a data lake.
The data lake concept is not unique
to manufacturing, but using it in OT environments
is relatively recent. The same
approach is used for the same purpose for
corporate information technology (IT)
applications; so, to many coming from the
IT side, the data lake concept will sound
like a logical solution. However, some IT
providers have encountered difficulties
trying to extend these platforms into OT
environments for a variety of reasons. An
OT-oriented approach is much more effective
and still allows easy extension into
IT applications. The ability to connect
from OT to IT has generally proven to be
easier than the reverse.
How is an OT data lake different than
a historian? A data lake aggregates operations
data (FIG. 2) from manufacturing,
process control and IT systems, without
disruption, and then contextualizes it and
transforms it into actionable information.
These results can be presented to every
decision-maker on any device, any time
and at any location, thus creating manufacturing
intelligence.
When data is stored in this type of a
central repository, users and critical systems
can find anything they need in one
place for more efficient analytics, trending
and reporting. This eliminates OT
data silos, while providing IT personnel
with a single location to manage, protect
and easily integrate OT data with IT tools
and cloud applications.
How does a data lake do these
things? A renewable diesel plant or unit
has various controllers, valves, process instruments
and other equipment types that
all individually create their own specific
type of data, using their specific format
and semantics. Historians can be added
to communicate with various devices or
classes of devices, but each only communicates
with its assigned sources, using its
own format and semantics. This leaves
data from different sources difficult to
compare or integrate, even within a single
historian, and multiple historians do not
normally communicate to each other.
This results in isolated data stored in multiple
silos that are cut off from each other.
A data lake takes a different approach.
It creates a common data transport layer
that supports all OT communications, regardless
of the device type. This layer provides
a common topology that can bridge
gaps, making it far more practical to use
digital tools for data analyses.
A modeling studio takes historical data
and " cleans " it to remove outliers and create
a more correlated data set that is suitable
for modeling purposes. Next, it uses
machine-learning algorithms to build
predictive models. Once these are assembled,
it is possible to verify a model's ability
to predict anomalies by using test data.
Once validated, the model is deployed
FIG. 2. An analytics platforma
provides information in a collaborative environment that is
able to unify data, people and systems. This can drive improvements in operational
performance and workplace efficiency through automated workflows, advanced analytics
and enhanced decision support.
80 JUNE 2022 | HydrocarbonProcessing.com
in a production environment, using a project
studio. This is the area to analyze realtime
data and to create rules for complex
events. When in operation, if a predictive
model finds a new abnormal condition or
process anomaly, the model is retrained to
consider that condition the next time.
http://www.HydrocarbonProcessing.com
Hydrocarbon Processing - June 2022

Table of Contents for the Digital Edition of Hydrocarbon Processing - June 2022
