Power June 2023 - 19

CONNECTED PLANT
on advanced analytics. Duke Energy,
for example, in November 2022 began
a " multiyear strategic collaboration " to
build new smart grid software and services
on AWS and expand its Intelligent
Grid Services-a suite of custom-built
applications that help the utility anticipate
future energy demand and identify
where and how to update the power
grid. The utility also agreed to migrate
its information technology and grid analytics
workloads to AWS.
" Our digital evolution is foundational
to our clean energy transition, " said Bonnie
Titone, senior vice president and
chief information officer at Duke Energy.
" For example, to accurately simulate future
energy needs and plan investments
for the grid, we need to run hundreds
of millions of power flow calculations-
a process that would take weeks using
traditional IT [information technology]
hardware. By using cloud technologies
that AWS is developing for Duke Energy,
we aim to run those same simulations in
15 minutes or less. "
Robotic Advancements
As a big boost to power plant operations
and maintenance, advancements in robotics
are meanwhile poised to reshape
difficult operations and maintenance
tasks. Autonomous robotic inspection is
already being used in a fleet of diverse
robots that are suited to handle difficult
or dangerous environments. Consumers
Energy and National Grid are separately
using Boston Dynamics' Spot, the
company's iconic dog-like agile robot, to
inspect assets at the company's critical
electrical substations, capturing thermal
scans and gauge reading autonomously.
Ontario Power Generation and Duke
Energy are meanwhile using Spot for
inspection missions at nuclear power
plants. Spot has also been deployed at
hydropower plants and mining areas, assessing
areas that are hazardous due to
risk of seismic activity, rock stability, or
the presence of poisonous gases.
Pittsburgh-based Gecko Robotics
has also seen a ramp up in interest in
its robots. Armed with multi-angle ultrasonic
scans and visual cameras, its
climbing TOKA series robots (Figure 2)
can inspect tanks, boilers, scrubbers,
piping, wall-thickness, and pitting. The
business has grown especially pivotal
given the power sector's aging assets.
" We're in the business of taking companies
from reactive maintenance with
very linear or poor reliability to proactivity,
but even beyond that, " Jake LooJune
2023 | POWER
2. Gecko's TOKA 4 robot scales a boiler
collecting ultrasonic data to fuel an artificial
intelligence-powered software platform.
Courtesy: Gecko Robotics
sararian, CEO and co-founder at Gecko
Robotics, told POWER.
Gecko also offers AI-enabled data
analysis collected from the multifaceted
inspections enabled by robots, which
has proven invaluable to some of its
power plant customers. " We've collected
and own data on tens of thousands
of assets. We're able to give insights
using some of the advanced software
into what we think will fulfill, predicting
obviously from the data sets what some
of the problems are, but also be able to
ensure that there is prudent spend of
capex [capital expenditure] to fix the
problems, not just today but into the future,
so that you can extend how often
you are shutting down. "
Connected Devices and the Drawback
As defined by the National Institute of
Standards and Technology (NIST), internet
of
things
(IoT)
comprises the
interconnection of consumer-level digital
devices (such as smart appliances
and cellular devices); whereas, indutrial
IoT (IIoT) covers the interconnection of
digital devices, such as sensors, instruments,
and machines in industrial, manufacturing,
and business processes, for
example, safety, performance tuning, access
control, facilities control, security,
and maintenance.
IIoT is also notable for its integration
of information and communication technology
to enhance operational technology
(OT), which are " programmable
systems or devices that interact with the
physical environment. " OT systems, for
example, include
instrumentation and
control (I&C) systems, supervisory control
and data acquisition (SCADA) systems,
and distributed control systems.
While traditional OT typically interconnects
via wired networks, IIoT devices
www.powermag.com
Power Industry Use
Cases for ChatGPT-Like
Models
OpenAI's release in November 2022 of
ChatGPT, a commercially available chatbot
that is based on an artificial intelligence (AI)
large language model (LLM), ushered in
widespread recognition that generative AI
could deliver a step change in the humanto-data
interface. And while generative
AI is already disrupting many industries,
it could hold substantial potential for the
power sector, including to help automate,
develop, communicate, and analyze data,
several experts have suggested.
As Geir Engdahl, chief technology officer
at Cognite, explained to POWER,
" What's going to be very valuable is the
combination of deep domain knowledge
and data, and a system that is able to
carry out the task. So, it has all the piping
in place, data operations, if you will,
where data is flowing from your actual
power plant operations into the system
continuously, and we have the ability to
carry out tasks that have value within
the system, such as creating inspection
orders, or do root cause analysis with all
of the data available to you, " he said. " I
think the combination of this 'data ops'
and ChatGPT and similar large language
models, that is going to really change
the way that knowledge workers in
these industries work. "
The biggest gains from custom-designed
generative AI programs could be
applied to all phases of asset performance
management, in tasks such as identifying
equipment for which automated monitoring
has triggered alarms, Engdahl suggested.
" It's able to kind of synthesize all
this industrial knowledge and make really
clever suggestions, " he noted. " Of course,
the human will be in the loop and make the
final call as to what to actually do. "
In addition, generative AI models
could provide references to documents,
or assist in onboarding data into a knowledge
graph. " What that means is as you
are ingesting, for instance, unstructured
data, which could be documents or drawings,
you could derive more context from
the data. " The model could, for example,
identify the specific type of document,
and it could
allow workers,
through
detailed " human language, " to quickly
query and get specific details about a
document. Engdahl, however, cautioned
that the model has shortcomings and
flaws including " hallucinations, " lying by
omission, data security, and intellectual
property protection, he said.
19
http://www.powermag.com
Power June 2023

Table of Contents for the Digital Edition of Power June 2023
