POWER February 2023 - 28

SMART GRID
the establishment of microgrids. The Bipartisan
Infrastructure
Law authorizes
the DOE, through the program, to borrow
up to $2.5 billion to assist in the construction
of high-capacity transmission
lines with an innovative approach that
can spur valuable new lines that otherwise
would not get built or increase the
capacity of already planned lines.
New Technology Available
Meanwhile, as those programs get
rolling, other smart grid technology is
available today. For example, Chicago, Illinois-headquartered
S&C Electric Co., a
leader in grid modernization, will launch
its new EdgeRestore System at the DISTRIBUTECH
International conference in
San Diego, California, Feb. 7-9, 2023. The
company provided POWER with an exclusive
early look at the EdgeRestore System
(Figure 2), which is designed to dramatically
improve power reliability for customers
by mitigating outages and eliminating
emergency truck rolls on underground
residential distribution systems.
When an underground fault or single
loss of source occurs on an underground
residential loop, the EdgeRestore system
automatically locates and isolates
the fault and reroutes power from an
alternate source-all within 60 seconds.
This intelligent system reduces the impact
of sustained outages that commonly
last for hours and eliminates the
need for immediate crew intervention,
keeping power on until the underlying
issue can be resolved. The automatic
isolation of faults also greatly simplifies
the process for locating faults on underground
residential circuits and reduces
medium-voltage exposure
for
firmware
crews.
Furthermore, the devices do not need
any programming,
updates,
computers, internet connectivity, batteries,
radios, or antennas.
" EdgeRestore truly modernizes the
last mile of the underground power distribution
system, reducing sustained
power outages to momentary interruptions
and making the grid safer, smarter,
and more resilient throughout weather
challenges and growing electricity demands, "
Anders Sjoelin, S&C president
and CEO, said in a statement issued to
POWER. " S&C's long history of grid innovation
and customer focus help us
deliver solutions that address the challenges
utilities face today and anticipate
the challenges they will face tomorrow. "
Florida Power and Light Co. (FPL) is expected
to be an early adopter of the technology.
" For nearly two decades, Florida
28
2. S&C Electric's new EdgeRestore system
is easy to deploy with smart communication
and fits inside the transformer enclosures
on new or existing underground residential
loop circuits with no changes to protection
schemes. Courtesy: S&C Electric
Power and Light has been on a journey to
build a stronger, smarter and more stormresilient
energy grid to better serve our
customers, " Manny Miranda, executive
vice president of power delivery for FPL,
said in the statement issued by S&C.
" We're pleased to work with S&C, with
whom we've had a long relationship, to
help develop modern solutions that advance
the grid and have a positive impact
for our customers by keeping the power
on, now and in the decades to come. "
Artificial Intelligence for the Smart Grid
" Beyond Limits is an industrial AI [artificial
intelligence] company specializing in developing
hybrid AI solutions by combining
data and expert knowledge, " Azarang Golmohammadi,
machine learning manager
with Beyond Limits, told POWER. " The
company has developed solutions for the
energy industry including power production
optimization, " he added.
Mehdi Hosseini, a data scientist with
Beyond Limits, explained that AI can
help the performance of the smart grid
in two major ways. " One of them is accurate
prediction of demand and generation,
and the other is the real-time grid
operation, " he said.
" Power utility companies rely heavily
on long-term and short-term predictions, "
Hosseini explained. " For example, every
day they require an accurate prediction
of the electrical demand of the next day.
The electrical demand is heavily associated
with weather data, business activities,
and everything that affect those
business activities. So, they go through
historical data to find patterns and point
associations between those two to have
an accurate prediction.
" One of the important applications of
AI that is already being used by many
power utilities-and is also a hot research
topic-is to encode these predictions using
different AI tools, whether they are in
www.powermag.com
the form of neural networks or tree-based
models. These models go through historical
data and are trained on any feature
that might affect the electrical demand of
the next day, " Hosseini said.
Having accurate predictions helps
companies plan power plant unit operation
much better and minimize spinning
reserves. " That reduces the operation
cost and also reduces the carbon emissions
caused by running unnecessary
power generation units. So, that's an important
factor, " said Hosseini.
Having accurate weather predictions
is also very important. Currently, most
companies use numerical methods like
numerical optimization to improve power
plant operation. But as the number of
renewable generation units increases,
the numerical calculations become more
complex with many more decision variables,
often greatly affected by weather.
If utilities want to solve those problems
in real time, numerical optimization falls
short, especially when there is high uncertainty
in the power sources.
" The way that AI helps, as suggested
by many recent studies, is to replace numerical
optimization with methods like
reinforcement learning. Reinforcement
learning models can be trained on numerous
operation scenarios, and once trained,
they can make scalable decisions in very
short time-they can reduce the runtime
by at least three orders of magnitude. They
can make decisions in a fraction of a second,
as opposed to numerical methods
that require minutes or even hours for a
larger-scale problem, " Hosseini said.
One of the concerns surrounding AI,
however, involves feasibility and safety
of the solution. " With mathematical optimizations,
you have hard-coded constraints
in the system, but when you're
using AI, they only rely on training. So,
it doesn't guarantee that you never have
an unsafe solution, especially when it
comes to energy systems or operational
power systems, " said Hosseini.
But work is being done to address
the dilemma. " There are many methods
that try to combine AI in different forms
with numerical methods, " Hosseini said.
There have also been strategies developed
to use AI as a decision support tool
rather than as a decision-maker. " As a
recommender system to human operators, "
Hosseini explained. There may also
be ways to encode constraints into AI
models, so a solution to overcome this
challenge may be very near. ■
-Aaron Larson is POWER's
executive editor.
POWER | February 2023
http://www.powermag.com
POWER February 2023

Table of Contents for the Digital Edition of POWER February 2023
