May 5, 2025 • 10 mins
Get ready to explore the future of energy innovation! In this episode of Energy Transition Talks, CGI experts Diane Gutiw, Lukas Krappmann, and Peter Warren dive into how digital triplets—an evolution of digital twins—are transforming the energy industry. Learn how companies are using AI-powered, interactive models to unlock insights faster, optimize systems, and deliver smarter energy solutions.
From hydrogen electrolyzers in Germany to grid optimization and predictive maintenance, discover how agentic AI and specialized data models are accelerating the shift to a more resilient and sustainable energy landscape. Perfect for tech enthusiasts, energy professionals, and innovation leaders—don’t miss this look into the next wave of digital transformation in energy.
Visit our Energy Transition Talks page
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Peter Warren (00:00):
Hello everyone and welcome back to our ongoing
series of conversations aboutenergy transition and how things
are changing in industry.
We actually just came back fromthe Hanover Messe, which is the
Hanover Fair for manufacturing,and there's a lot of overlap
between manufacturing and theenergy markets.
We're going to touch on that,but the big dive today is
talking about a concept calleddigital twins and digital
(00:22):
triplets.
So with me I have two greatexperts, diane and Lukas, and
let's start with Diane.
Do you want to introduceyourself?
Diane Gutiw (00:30):
Thanks, peter, and thanks for inviting me to the
podcast.
My name is Diane Gutu.
I lead our AI Global ResearchCenter and a lot of our focus
has been on digital twins andextending them to digital
triplets.
So great to be joining theconversation twins and extending
them to digital triplets.
Lukas Krappman (00:48):
So great to be joining the conversation, yeah,
thanks, thanks, pete.
My name is Lukas Kruppmann,from Germany here and I'm
responsible for one of theclients active in as well, the
manufacturing and energy andutilities industry, and
therefore we already worked on acouple of concepts according to
digital twins, and also someideas and triplets.
And thanks, pete, happy to alsobe here today.
Peter Warren (01:05):
That's a great thing.
Thanks for joining me.
So we're covered from the farcoast of Canada over to Germany
and middle part of Canada.
So thanks very much, diane.
Since you're the residentexpert in all things digital
twins and you've come up withthe concept of digital triplets,
do you want to give us sort ofa baseline conversation on what
those are?
Diane Gutiw (01:22):
Sure, absolutely so .
The concept of digital tripletis actually quite simple, but
it's a fantastic way to extendan existing investment in data
and a data ecosystem.
So if we look at the differentlayers, you know you have your
physical asset or a group ofassets infrastructure that
you're monitoring as yourphysical layer.
(01:44):
The digital twin is thatdigital representation of those
things, that person, thosepeople, those pieces of
equipment.
So when you're looking at adigital twin, it's collecting
the data from operationalsystems, from historic systems,
information about each of thoseassets, as well as anything
(02:04):
diagnostic coming from edgecomputing, iot devices, so that
you have a really good holisticview of that ecosystem or that
particular item that you want tomonitor.
Digital twins are not new.
They're a great way of lookingat operations and interactions
between different assets, aswell as starting to do with
(02:25):
traditional AI, some analysis ondifferent scenarios.
So why would digital tripletsbe important?
Well, it's extending that.
It's leveraging newertechnology, so generative AI,
large language models and, insome cases, small language
models, which I think we'll gointo a bit to be able to explore
that digital twin layer andprovide an opportunity for an
(02:49):
operator to have a conversationusing natural language with that
data.
So the way that the digitaltriplet works.
It's a form of a Gentic AIwhich, of course we heard in
Hanover and other places is areally hot topic which is a
group of collaborative agentsthat are monitoring the layer of
data on the digital twin.
(03:09):
They're continuously listeningand they are able to work
autonomously on defined tasks.
So, for example, one may bereally good at understanding
diagnostic information, anotherone on parsing IoT information
and looking for anomalies.
You might have one that's ableto generate different types of
information and then provide itback to the person that's asking
(03:31):
the questions.
The real value of the digitaltriplet to me are two things.
One is the ability to accessinformation which goes beyond
the traditional discrete datacoming from operational systems,
because we can look atnarrative data and images and
videos and and have aconversation with that data the
same way traditionally we couldwith discrete data, without
(03:52):
having to spend a lot of energyon modeling it.
But the most important thing isthe accessibility, the fact
that you can have a conversationwith your data.
You can ask what would happenacross an energy grid if this
part went down or if I need todo some maintenance on a piece
of equipment.
How then would I readjust bothmy human resource load as well
(04:15):
as the energy load to compensatefor anything that needs to
change.
You can ask information onalerts and you're having that
conversation either by text orby phone in natural language.
So it's like having your bestgroup of advisors and specialist
assistants being able to pullthat information for you real
time.
So a bit of a long-windedexplanation, but it's something
(04:36):
that's really definitely takingoff in the energy space.
Peter Warren (04:39):
I appreciate that definition and I guess it's
really a case of theseinvestments in digital models
people already have.
It's really layering those intoa way that's more consumable
from a business standpoint,maybe, than from a technical
standpoint.
Would that be a fair summary?
Diane Gutiw (04:52):
Absolutely, and that's why this is a great place
for organizations to start withtheir AI journey.
And accelerating that AIjourney is because you are
extending your investment inwhat you already have in place.
That may show you opportunitiesto be able to do more, but by
adding a layer of generative AInot just on your documents, the
way we're seeing in RAG models,but across your whole data
(05:14):
ecosystem you're able to reallyget very quick insights into
information that in the past,were really complex and
expensive to be able to do.
Peter Warren (05:22):
Well, that's really interesting and, Lucas,
you've got a practicalapplication of this, originally
from a manufacturer of energysystems, but do you want to give
us a summary of your storythere?
Lukas Krappman (05:32):
Yeah, so we basically start with the classic
, first, I would say digitalrepresentation, for example, of
an hydrogen electrolyzer.
So most of our clients inGermany are actually building
the hydrogen electrolyzers.
With regards to the PEMelectrolyzers, and what
eventually we did is wedigitized first the physical
model Diane was alreadyreferring to and then adding
(05:55):
first the telemedicator to it,but also the business processes.
So what is, for example,happening before a stack in
terms of pressure, temperature,but also current or voltage, and
what is happening afterwards,able to looking at that from
(06:15):
like a 3d representation andalso talking with the data.
For example, you're seeing thatthere's a stack being colored,
for example, in red or orange,indicating that something is
wrong, and then you can click onit, investigate it and then
talk with the data.
This is what diane referred toearlier, like in the right hand
side, said hey, can you pleasehave a look over the last day,
what happens in this particularstack?
Do you need to replace it?
Was there some I don't knowkind of pressure leakage?
(06:36):
Was there a spike into voltage?
Really, about the digital twinsfor the hydrogen electrolyzer
operations, this is how westarted and we're now actually
trying to elevate the conceptand letting the data which is
going in, so telemetry data,speak and connect it to all the
historic data.
Peter Warren (06:54):
So this was really came from their desire to
digitize and be sort of moremodern in their platform, so
that they're trying to be moreof a data-driven system to bring
up a higher level of value.
Would that be a good example?
They're trying to operatethings and optimize their full
production.
What is the outcome they'relooking to get?
Lukas Krappman (07:13):
Yeah, I mean that's the main objective.
With the current market and allthe new companies putting their
products in the market.
It's not only about, let's putit, the physical product they're
selling, but also about thedigital experience.
In the past, you were referringto that like about the user
experience or how differentstakeholders and people can
interact with the product.
(07:33):
So what we are trying to do, orwhat we aim to do with the
children for the operations, is,as well as provide the internal
manufacturers the possibilityto improve the product, but also
provide, let's say, a digitalservice to their customers in
terms of, okay, accessing thedata, integrating it into, for
example, other components likethe energy grid, but also, for
(07:57):
example, the question whatshould I do really with an
electron right?
So should I put the remainingenergy of the electrons back
into it?
And these are the questions Ican ask the digital twin.
Peter Warren (08:08):
That's really interesting.
And, diane, you actuallypublished an article recently
hitting on that very same point,which really gets to how do
people overcome the barriers andmove ahead and looking ahead to
this sort of area.
But in that article you hit alot of interesting points about
how to optimize not just part ofthe energy system but the whole
energy system.
Diane Gutiw (08:28):
Yeah, I think you know, when you look at how to
optimize any system, anyinfrastructure, including the
energy system, focusing onwhat's the problem you want to
solve right, identifying whereit is that there is an
opportunity to improve and thenextending that.
A big example that we see wouldbe all of that information
(08:49):
that's collected from IoTdevices and edge computing.
You know, right now we are notreally getting the value out of
that data because this isinformation that's collected in
a second or a couple of secondsor minutes or a couple of
minutes, and there's somefantastic insights, but so much
volume of data it's nearimpossible to be able to pull
(09:09):
that.
So being able to leverage thesenewer tools that are able to
see patterns without having tospend the time on that level of
monitoring really is what'staking it forward.
The other thing I think that'sshifting is having small
language models, which Imentioned in the beginning,
which are not using the entireinternet.
You couldn't ask it how do Imake the best apple pie?
(09:32):
The way you can with the largelanguage models, but it's
absolutely trained on thecontext of what it needs to do,
so it's improving the computepower and the efficiency of the
models and it's being able togive much more precise answers
to questions on very specific towhat the problem is you're
trying to solve.
For example, you can have asmall language model that's
(09:55):
fine-tuned and trainedspecifically on that different
types of IoT data to get thereal value out of that.
So I think, starting withwhat's a problem where there's
an opportunity, and then be ableto look at refining each part
of that value chain is reallycritical to getting that return
on investment.
Peter Warren (10:14):
Well, that's outstanding.
Thank you for that.
So we're going to end thisfirst part of a two-part series,
so hopefully you guys can catchup with us in the next part,
where we're going to talk aboutmore of the innovation that's
moving forward here, the impactsof the ongoing geopolitical
issues that we are facing rightnow and really the vision for
how things are moving forward.
So we'll catch up to you all inpart two.
(10:34):
Thank you Lucas, Thank youDiane, We'll talk to you again.
Lukas Krappman (10:37):
Thank you, have a nice day, bye-bye.
Thanks, peter.
Hello everyone and welcome back to our ongoing
series of conversations aboutenergy transition and how things
are changing in industry.
We actually just came back fromthe Hanover Messe, which is the
Hanover Fair for manufacturing,and there's a lot of overlap
between manufacturing and theenergy markets.
We're going to touch on that,but the big dive today is
talking about a concept calleddigital twins and digital
(00:22):
triplets.
So with me I have two greatexperts, diane and Lukas, and
let's start with Diane.
Do you want to introduceyourself?
Diane Gutiw (00:30):
Thanks, peter, and thanks for inviting me to the
podcast.
My name is Diane Gutu.
I lead our AI Global ResearchCenter and a lot of our focus
has been on digital twins andextending them to digital
triplets.
So great to be joining theconversation twins and extending
them to digital triplets.
Lukas Krappman (00:48):
So great to be joining the conversation, yeah,
thanks, thanks, pete.
My name is Lukas Kruppmann,from Germany here and I'm
responsible for one of theclients active in as well, the
manufacturing and energy andutilities industry, and
therefore we already worked on acouple of concepts according to
digital twins, and also someideas and triplets.
And thanks, pete, happy to alsobe here today.
Peter Warren (01:05):
That's a great thing.
Thanks for joining me.
So we're covered from the farcoast of Canada over to Germany
and middle part of Canada.
So thanks very much, diane.
Since you're the residentexpert in all things digital
twins and you've come up withthe concept of digital triplets,
do you want to give us sort ofa baseline conversation on what
those are?
Diane Gutiw (01:22):
Sure, absolutely so .
The concept of digital tripletis actually quite simple, but
it's a fantastic way to extendan existing investment in data
and a data ecosystem.
So if we look at the differentlayers, you know you have your
physical asset or a group ofassets infrastructure that
you're monitoring as yourphysical layer.
(01:44):
The digital twin is thatdigital representation of those
things, that person, thosepeople, those pieces of
equipment.
So when you're looking at adigital twin, it's collecting
the data from operationalsystems, from historic systems,
information about each of thoseassets, as well as anything
(02:04):
diagnostic coming from edgecomputing, iot devices, so that
you have a really good holisticview of that ecosystem or that
particular item that you want tomonitor.
Digital twins are not new.
They're a great way of lookingat operations and interactions
between different assets, aswell as starting to do with
(02:25):
traditional AI, some analysis ondifferent scenarios.
So why would digital tripletsbe important?
Well, it's extending that.
It's leveraging newertechnology, so generative AI,
large language models and, insome cases, small language
models, which I think we'll gointo a bit to be able to explore
that digital twin layer andprovide an opportunity for an
(02:49):
operator to have a conversationusing natural language with that
data.
So the way that the digitaltriplet works.
It's a form of a Gentic AIwhich, of course we heard in
Hanover and other places is areally hot topic which is a
group of collaborative agentsthat are monitoring the layer of
data on the digital twin.
(03:09):
They're continuously listeningand they are able to work
autonomously on defined tasks.
So, for example, one may bereally good at understanding
diagnostic information, anotherone on parsing IoT information
and looking for anomalies.
You might have one that's ableto generate different types of
information and then provide itback to the person that's asking
(03:31):
the questions.
The real value of the digitaltriplet to me are two things.
One is the ability to accessinformation which goes beyond
the traditional discrete datacoming from operational systems,
because we can look atnarrative data and images and
videos and and have aconversation with that data the
same way traditionally we couldwith discrete data, without
(03:52):
having to spend a lot of energyon modeling it.
But the most important thing isthe accessibility, the fact
that you can have a conversationwith your data.
You can ask what would happenacross an energy grid if this
part went down or if I need todo some maintenance on a piece
of equipment.
How then would I readjust bothmy human resource load as well
(04:15):
as the energy load to compensatefor anything that needs to
change.
You can ask information onalerts and you're having that
conversation either by text orby phone in natural language.
So it's like having your bestgroup of advisors and specialist
assistants being able to pullthat information for you real
time.
So a bit of a long-windedexplanation, but it's something
(04:36):
that's really definitely takingoff in the energy space.
Peter Warren (04:39):
I appreciate that definition and I guess it's
really a case of theseinvestments in digital models
people already have.
It's really layering those intoa way that's more consumable
from a business standpoint,maybe, than from a technical
standpoint.
Would that be a fair summary?
Diane Gutiw (04:52):
Absolutely, and that's why this is a great place
for organizations to start withtheir AI journey.
And accelerating that AIjourney is because you are
extending your investment inwhat you already have in place.
That may show you opportunitiesto be able to do more, but by
adding a layer of generative AInot just on your documents, the
way we're seeing in RAG models,but across your whole data
(05:14):
ecosystem you're able to reallyget very quick insights into
information that in the past,were really complex and
expensive to be able to do.
Peter Warren (05:22):
Well, that's really interesting and, Lucas,
you've got a practicalapplication of this, originally
from a manufacturer of energysystems, but do you want to give
us a summary of your storythere?
Lukas Krappman (05:32):
Yeah, so we basically start with the classic
, first, I would say digitalrepresentation, for example, of
an hydrogen electrolyzer.
So most of our clients inGermany are actually building
the hydrogen electrolyzers.
With regards to the PEMelectrolyzers, and what
eventually we did is wedigitized first the physical
model Diane was alreadyreferring to and then adding
(05:55):
first the telemedicator to it,but also the business processes.
So what is, for example,happening before a stack in
terms of pressure, temperature,but also current or voltage, and
what is happening afterwards,able to looking at that from
(06:15):
like a 3d representation andalso talking with the data.
For example, you're seeing thatthere's a stack being colored,
for example, in red or orange,indicating that something is
wrong, and then you can click onit, investigate it and then
talk with the data.
This is what diane referred toearlier, like in the right hand
side, said hey, can you pleasehave a look over the last day,
what happens in this particularstack?
Do you need to replace it?
Was there some I don't knowkind of pressure leakage?
(06:36):
Was there a spike into voltage?
Really, about the digital twinsfor the hydrogen electrolyzer
operations, this is how westarted and we're now actually
trying to elevate the conceptand letting the data which is
going in, so telemetry data,speak and connect it to all the
historic data.
Peter Warren (06:54):
So this was really came from their desire to
digitize and be sort of moremodern in their platform, so
that they're trying to be moreof a data-driven system to bring
up a higher level of value.
Would that be a good example?
They're trying to operatethings and optimize their full
production.
What is the outcome they'relooking to get?
Lukas Krappman (07:13):
Yeah, I mean that's the main objective.
With the current market and allthe new companies putting their
products in the market.
It's not only about, let's putit, the physical product they're
selling, but also about thedigital experience.
In the past, you were referringto that like about the user
experience or how differentstakeholders and people can
interact with the product.
(07:33):
So what we are trying to do, orwhat we aim to do with the
children for the operations, is,as well as provide the internal
manufacturers the possibilityto improve the product, but also
provide, let's say, a digitalservice to their customers in
terms of, okay, accessing thedata, integrating it into, for
example, other components likethe energy grid, but also, for
(07:57):
example, the question whatshould I do really with an
electron right?
So should I put the remainingenergy of the electrons back
into it?
And these are the questions Ican ask the digital twin.
Peter Warren (08:08):
That's really interesting.
And, diane, you actuallypublished an article recently
hitting on that very same point,which really gets to how do
people overcome the barriers andmove ahead and looking ahead to
this sort of area.
But in that article you hit alot of interesting points about
how to optimize not just part ofthe energy system but the whole
energy system.
Diane Gutiw (08:28):
Yeah, I think you know, when you look at how to
optimize any system, anyinfrastructure, including the
energy system, focusing onwhat's the problem you want to
solve right, identifying whereit is that there is an
opportunity to improve and thenextending that.
A big example that we see wouldbe all of that information
(08:49):
that's collected from IoTdevices and edge computing.
You know, right now we are notreally getting the value out of
that data because this isinformation that's collected in
a second or a couple of secondsor minutes or a couple of
minutes, and there's somefantastic insights, but so much
volume of data it's nearimpossible to be able to pull
(09:09):
that.
So being able to leverage thesenewer tools that are able to
see patterns without having tospend the time on that level of
monitoring really is what'staking it forward.
The other thing I think that'sshifting is having small
language models, which Imentioned in the beginning,
which are not using the entireinternet.
You couldn't ask it how do Imake the best apple pie?
(09:32):
The way you can with the largelanguage models, but it's
absolutely trained on thecontext of what it needs to do,
so it's improving the computepower and the efficiency of the
models and it's being able togive much more precise answers
to questions on very specific towhat the problem is you're
trying to solve.
For example, you can have asmall language model that's
(09:55):
fine-tuned and trainedspecifically on that different
types of IoT data to get thereal value out of that.
So I think, starting withwhat's a problem where there's
an opportunity, and then be ableto look at refining each part
of that value chain is reallycritical to getting that return
on investment.
Peter Warren (10:14):
Well, that's outstanding.
Thank you for that.
So we're going to end thisfirst part of a two-part series,
so hopefully you guys can catchup with us in the next part,
where we're going to talk aboutmore of the innovation that's
moving forward here, the impactsof the ongoing geopolitical
issues that we are facing rightnow and really the vision for
how things are moving forward.
So we'll catch up to you all inpart two.
(10:34):
Thank you Lucas, Thank youDiane, We'll talk to you again.
Lukas Krappman (10:37):
Thank you, have a nice day, bye-bye.
Thanks, peter.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com