provide a connected view of the end-
to-end network of assets, based on
real operational data.
Operators and owners can imple-
ment these digital twins in several
ways: either they can purchase the
relevant tools from GE Digital and
build it themselves; or buy a twin
from GE Digital’s catalogue of twins
and input their own data. “We have
over 300 pre-built digital twins of
components in our APM systems, so
they can feed their data into the twin,
which then learns about their sys-
tem,” said Parris. The third way, he
notes, is for GE Digital to take the
customer’s data and build the twin.
One of the biggest challenges that
companies often face, however, is to
rst collect the necessary data, and
this to some degree is determining the
prevalence of the technology in the
various parts of the power sector.
Looking forward, Parris highlights
a few key areas of advancement in
digital twins and ways in which GE
Digital is working to accelerate their
use.
Although digital twins can bring
value and deliver savings through
early warning, prediction and optimi-
sation, he noted that operators are of-
ten not comfortable with basing their
strategies on twins to, for example,
predict the lifetime of a $20 million
sensor in a turbine.
“Getting people to adopt it is the
hardest thing. So about three years
ago we created something called
Humble AI, which takes into account
the zone of competency for a particu-
lar [digital] model; so you use the
model inside the zone of competency,
and when outside that zone you use a
different model or human and feed
that data back in so the AI system gets
smarter. That’s why it’s humble; it
knows what it doesn’t know and it
wants to learn.”
The technology has already been
developed for gas turbines and wind
turbines and Parris notes that it is
giving operators greater comfort in
terms of reducing risk.
Another area that Parris says GE
Digital is currently focusing on is
how to put this “all into a process
that people like”. The company is
therefore combining digital with
Lean methodology.
He explained: “Lean takes any
process you have and says: ‘tell me
what you are trying to solve.’ In
power, you might be trying to reduce
the cost of maintenance or increase
how much power you deliver at a
certain fuel level. So there’s a process
behind it. Lean will call for a value
map of the process, whereby all the
data will be pooled from the experts.
Lean is about pulling the data, and
that same data is what a data scientist
D
igitalisation has opened up all
kinds of possibilities in the
power sector. Yet there is one
area of digitalisation that can make a
profound difference – the concept of
the digital twin, a mirror of the physi-
cal world.
The digital twin is most commonly
dened as a software representation
of a physical asset, system or process
designed to detect, prevent, predict
and optimise through real-time ana-
lytics to deliver business value. The
technology has been around for
some time but with the Internet and
progress in technologies such as arti-
cial intelligence (AI) and machine
learning, digital twins are entering a
new phase, bringing new possibili-
ties to owners and operators of
power assets.
Colin Parris, Senior Vice President
and Chief Technology Ofcer, GE
Digital, has seen the technology
grow from its infancy to become an
important tool in GE Digital’s arse-
nal to better serve its customers’ ef-
forts to improve the operation and
value of their assets – whether in
power generation or transmission
and distribution.
He said: “The digital twin actually
came out of aviation and in particular
the military, perhaps a decade or
more ago. The Navy was looking at
how to understand the readiness of
an aircraft sitting on one of its carri-
ers. It’s not like a plane at an airport;
if you don’t plan for parts or service,
the aircraft doesn’t y and the mis-
sion is compromised.
“So the notion was: can I have a
digital model that can tell me the state
of readiness of an aircraft…? GE then
began thinking about how it could do
something like that, initially for its
Aviation business. And because we
also have turbines running every-
where for the electricity and energy
sectors – where typically we had to
give suppliers six or seven months
lead-time before the parts were needed
– it made sense to have digital twins.”
GE Digital then began to investigate
what else a twin might be able to do.
“Because we have engines that are in
the air, engines that are producing
electricity or engines that are pump-
ing oil out of the ground, we began to
see a pattern of what customers
wanted to do.
“First, they want an early warning
of a problem; with a jet engine you
need an early warning about failure.
In the energy sector, you want to be
warned about any anomalies – it’s
much better to x a bearing or blade
early rather than to get to a point
where there is damage that can cause
an engine to be out for six months.
The second is continuous predictions
on the remaining life of a part, to
understand what parts I need in my
inventory for when it has to be re-
placed. And the third thing is optimi-
sation: optimising a turbine for
highest energy delivery and lowest
fuel cost.”
GE Digital then moved to see how
this could be expanded across an
electricity network, looking at all the
components on the grid to optimise
the maximum amount of generation
for the lowest cost. This was then
extended to processes, such as smelt-
ing in order to consume the least
amount of electricity and least amount
and materials.
The digital twins that are increas-
ingly being adopted today are not like
the static models of the past that were
used to perhaps predict the behaviour
of a network at a given moment in
time. Today’s digital twins are what
Parris calls “living, learning models”
that take in a steady stream of data to
continuously update their models.
He said: “While there is widespread
use of things we call twins, which just
give insights from data coming in, we
are now moving into combining the
physics and AI to give us deeper and
deeper insights into what’s happen-
ing. There are twins in generation,
transmission and distribution and
there are especially new twins for
what is going with distributed energy
resources. People are wondering how
to model all of the electric vehicles
and battery sources that are coming
on line – with all the volatility it cre-
ates, you need twins and analytics to
help you.”
GE Digital is focused on how digital
twins can help its customers across
three core areas: assets, networks and
processes.
Addressing the power generation
sector, the company’s Asset Perfor-
mance Management (APM) software
solution creates digital twins based
on operational/eet data of: compo-
nents such as pumps or compressors;
critical assets, like turbines; or sys-
tems of assets such as an entire
power station. This type of digital
twin is an increasingly common tool
for operators of large equipment to
optimise their maintenance sched-
ules and to predict and avoid un-
planned downtime.
For transmission and distribution,
its Advanced Distribution Manage-
ment Solution (ADMS) and Geo-
graphic Information System (GIS)
use operational data from across the
network to create network digital
twins that can create virtual models.
These allow grid operators to better
manage and optimise networks, for
example, in the face of increasingly
extreme weather, aging infrastructure,
and the growing use of renewables
on the grid. Such twins essentially
THE ENERGY INDUSTRY TIMES - OCTOBER 2020
Executive Interview
14
Digital twins are an
exciting technology
with incredible
potential.
Junior Isles catches
up with GE Digital’s
Colin Parris for his
take on some of the
benets they bring
and a glimpse of
what’s to come.
A mirror to the future
A mirror to the future
needs... to create a model for embed-
ding in the process.
“Engineers like this combination of
digital and Lean because they all
know Lean, and now they can see
Lean inside of digital. Lean helps fo-
cus on the amount of money you will
save, or whatever it is you want to
change, while digital does the digital
transformation inside the process.
This is allowing the technology to
gain more traction in the industry.”
So what is the future of the technol-
ogy itself? Parris offered a glimpse of
a few research projects he has been
working on with investment from or-
ganisations such as the US Defense
Advanced Research Projects Agency
(DARPA) and the Intelligence Ad-
vanced Research Projects Activity
(IARPA).
Over the last three years they have
been investing money with GE Digi-
tal in an area called ‘Emerging Lan-
guages’. About ve years ago, GE
Digital began exploring the idea of
assets that could talk to each other and
solve problems.
Parris explained: “What if one wind
turbine could show another turbine
its sensor readings, ask if it has seen
these readings before and then ask:
‘what was the problem?’ And that
turbine could respond, saying for
example, I have seen these readings
before and it was a bearing problem.
And what if then, that asset could
communicate with us and tell us what
it thinks the problem is? This would
be tremendously helpful. It would
allow us to identify problems very
early on.”
GE began developing a language
between the turbines and has been
experimenting for the last year, with
“some interesting results”.
Parris said: “It can communicate
simple things like: there was a storm,
damage to a blade, this sensor reading
looks wrong and I think it’s this. It’s
at an early stage but what begins to
get me excited is the speed at which
they communicate, and the things that
they say is interesting.
“If you think about the next 4-5
years of this and get to a point where
machines are diagnosing themselves,
although humans will still be in-
volved, it will all be a lot faster.”
It’s an exciting future. On a wider
scale machines talking to each other
in such a way offers an incredible
opportunity in the ght against cli-
mate change.
Parris concluded: “It’s especially
relevant to me because of the decar-
bonisation problem. If you ever get
to a point where these assets are go-
ing to have to work together to reduce
carbon in the atmosphere, you want
them working together in the most
optimal way.”
Parris: we are now moving into
combining the physics and AI
to give us deeper and deeper
insights into what’s happening
