IEEE Electrification - December 2022 - 6

TECHNOLOGY LEADERS
Bringing Artificial Intelligence
to the Grid Edge
By Yingchen (Y.C.) Zhang and Marc Spieler
S THE DEMAND FOR ENERA
gy
increases along with the
requirements for decarbonization,
there is more complexity
in which the world produces, moves,
and consumes energy. The way we
supply and consume energy is
becoming increasingly complex and
requires us to leverage technology
and learn from the success of other
industries. From the fuel that powers
cars and planes, to the gas used for
stove top cooking, to the electricity
that keeps the lights on in homes
and businesses, energy powers our
daily lives. Oil, gas, and electricity are
mature commodity markets, but as
the mix changes at a rate never seen
before, the industry must adapt at a
similar pace. This includes leveraging
artificial intelligence (AI), realtime
analytics, and machine learning
to create autonomous energy systems
that can increase reliability and
resiliency in a more complicated
world. However, AI alone is not
enough to transform the processes
used to produce, transport, and
deliver these resources.
Enter edge AI, deployed where
decisions need to be made, where
the data are being produced, where
communications networks may not
provide for the low latency required
to make real-time decisions-such
as on oil rigs, within power plants,
Digital Object Identifier 10.1109/MELE.2022.3210778
Date of current version: 24 November 2022
6
along transmission and distribution
grids, riding along utility trucks,
and embedded in industrial facilities.
As the grid edge
becomes exponentially
more complex,
there is a massive
opportunity for electric
utility companies
to use edge AI to
improve operational
efficiency, enhance
functional safety,
integrate renewable
energy, increase grid
resiliency, and provide
reliable and affordable power.
What Is Edge AI, and
How Does It Work?
Edge AI is the deployment of embedded
AI applications in connected
devices in the physical world. The
" edge " component is a result of AI
computation done near the user at
the edge of the network, close to
where the data are located, rather
than centrally in a cloud computing
facility or private data center.
Take the example of autonomous
vehicles (AVs). Self-driving cars need
AI to process large volumes of onboard
sensor data from cameras, lidar,
and radar to inform real-time
driving decisions in split seconds, not
minutes or hours. This requires
continuous perception, mapping,
planning, and driver monitoring
capabilities. Imagine the conseIEEE
Electrification Magazine / DECEMBER 2022
The way we supply and
consume energy is
becoming increasingly
complex and requires
us to leverage
technology and learn
from the success of
other industries.
quences if an AV needed to wait for a
response from a data center before
applying brakes or swerving to avoid
an obstacle in the
road. Future electric
grid infrastructure
will leverage edge AI
in a similar way to
dynamically manage
distributed energy resources
(DERs), meet
charging demands,
balance the grid to
avoid waste, and
integrate more renewables,
local production,
and storage from increased
rooftop solar and electric vehicle
(EV) adoption.
The same approach for edge AI
applies to other trillion-dollar industries,
from energy to health care, to
manufacturing and beyond. Three
recent innovations have accelerated
the development and deployment of
these AI models to the edge. The first
is the maturation of neural networks
to enable generalized machine learning.
The second is GPU-accelerated
compute infrastructure that uses
distributed parallel computing to run
AI at the edge. The third is the widespread
adoption of Internet-ofThings
(IoT) devices-including
sensors, smart cameras, and autonomous
mobile robots-combined
with big data to enable the deployment
of AI models at the edge over
5G networks.
2325-5897/22©2022IEEE
https://orcid.org/0000-0002-5559-0971
IEEE Electrification - December 2022

Table of Contents for the Digital Edition of IEEE Electrification - December 2022
