Player is loading...
Pr. Dominique Genoud & Jérôme Treboux - Integrating machine learning and AI in practical energy applications : a review of these last 10 years activities at the Institute of Information Systems
Pr. Dominique Genoud & Jérôme Treboux, HES-SO Valais-Wallis
Monday, Nov. 15, 2021 · 10:46 a.m. · 13m 09s
Since 10 years the Institute of Information Systems is integrating machine learning and Artificial intelligence sytems in real world applications like Energy systems. We will present our success and how we did overcome the difficulties in collecting proper data, developing new algorithms, retraining them and maintening them and integrating them in real information systems through research and applied research projects.
Embed
Transcriptions
Note: this content has been automatically generated.
00:00:00
lot of lot of forecasting them and well that's probably
00:00:05
cost on the obvious choice for coffee bar also well i
00:00:10
and so we're not the first that's coming up it's so now
00:00:14
we call it is real and i'm not too fought it off
00:00:18
uh it's age actors on stage to sing actually oh oh big trout
00:00:30
oh us might but it yeah okay okay okay
00:00:48
i would hate to put up with a hello can you hear me
00:00:59
cool yeah um so hello everybody yeah um so when the manager move from
00:01:07
a h.'s is source why as you can see on the no void now
00:01:11
so uh am i yes since a long time and the other thing is that we have a lab in uh
00:01:18
it ha so that is doing a a lot of different machine and stuff
00:01:23
and the uh so my i we not talk about that uh in details but my trajectory
00:01:28
comes from the jabber long time ago uh and we were experimenting a neural networks at that time
00:01:36
so we're still using them for me different things and we use the not the a lot of different records too
00:01:43
and uh so uh uh during during the uh
00:01:47
the us why work there uh we also add you know started up uh many different startup so
00:01:53
the idea is to be uh between uh research and application always all the time
00:01:59
and that's what we do yeah we just go a bit more global so everybody
00:02:02
talked about these days and the detailed research uh is your home is there for uh
00:02:08
talking about that but very quickly the and and um what i wanted to uh to propose you
00:02:15
today is is just the story of what we did here in why we did it and the um
00:02:23
everything started uh ten years with a eleven years are going to
00:02:25
doesn't turning bruce's when uh yeah actually pretty organised a conference and
00:02:31
about energy i smart breeds electricity and the i thought this
00:02:37
is starting point was very interesting because the war and you know
00:02:40
trying to anticipate and forecast what would be the the word in ten twenty years after that
00:02:46
and uh so the idea was to how to work you know
00:02:50
put some concept on on what word electricity agreed at that time
00:02:55
of course the energy the that goes with that and many of the
00:02:59
of the concept that for uh it worked at that time are now
00:03:02
you you start to apply them and and companies and and utilities are
00:03:06
starting to use them and understand them it's more difficult than we thought
00:03:11
but still uh it's very interesting because we had no bitter
00:03:14
power plant we have microwaves we'll all this kind of concept
00:03:18
and the at that time right by day and not many people were talking about that that didn't know
00:03:24
about that so uh the thing was how can we you know we our uh a big producer of energy
00:03:30
uh we do have a different utilities companies so how was it possible to to mix the two
00:03:36
and so uh of for me the the the the thing
00:03:39
was interesting when you know i'm i'm a computer scientist so
00:03:43
uh that's the place what data are that are interesting me and the the the main starting point
00:03:49
was there to to track with the think about where i would be to put is the more data
00:03:54
and what you produce the data is what we call the crow because the place where you do a can some things
00:03:59
you do produce things if you use a you know a a for one or something like that and try to store fit
00:04:07
and uh also that's the place where you uh though the most the time so
00:04:10
you can do something and then apply our or men's feet of algorithm that too yeah
00:04:16
and so the the idea was really fun and we do the digits start doing
00:04:21
two thousand ten with uh you know just i figure you thing and we we
00:04:25
uh we started to a to a different kind of project so to interest people at
00:04:30
different level than we did start with uh uh it she says with that time with uh
00:04:35
people the torah i so i am in the business could part uh is because informatics is there
00:04:41
but to also a lot of ingenious work interested in and uh
00:04:46
yeah it's one of them and we'll talk about that a later is now the
00:04:49
head of uh the institute of energy and and run and so everything progress there
00:04:55
so there was you know technical things research things there was
00:04:58
structured things that that just no rule again that a on the
00:05:02
uh that that's very interesting and also the last the last step was how to interest the the the people that were the most
00:05:09
uh important there which were the people who are producing data how we're using these data after that
00:05:15
and so we started the uh many years ago uh with you know real projects
00:05:21
and the this project so we didn't uh do some research to improve what
00:05:26
we yeah there's more those and understanding of of of how energy systems works
00:05:32
like a like a logo modelling uh i've seen to date yet like murdering a house
00:05:37
is about what to me during the behaviour of users understand withering electricity a consumption and production
00:05:44
and the uh what we did try to this kind of things and so
00:05:48
yeah it was just really to to put those in touch with the
00:05:52
research and to be able to trends in in the applications after that
00:05:56
and so we started with a collecting data on our own and one which was fun because we'd
00:06:03
we had to convince people locally that's uh we wanted to to collect data to be able to
00:06:08
read something that to you know as you know we for our models we need a we need data
00:06:13
and so uh they're reluctant to do at the beginning so they were really uh yeah you know uh
00:06:19
you know it's used less we never be able to do a production something like that
00:06:24
and but some of them say okay yeah okay you rock out slightly say
00:06:29
yeah we do we strive things clues just checked it's that it that it works
00:06:33
and so what we did and after the first two or second um iteration there
00:06:40
then we started with some tracks everybody starting to to to work on this
00:06:45
and the and now we are ten years eleven years after that and we still
00:06:49
have a and we have a good traction on so that i was more
00:06:52
lies so use the now more i see more less because sometimes more sometimes less
00:06:57
uh but so they see the it they they start to see the really the the the thing they can do a uh with
00:07:04
us and we do we have now a lot of implementation that
00:07:07
are uh over us with someone i would say and the um
00:07:12
it's crucial because as you know and there's you you do uh uh we need a town we need reliable
00:07:17
data to do uh to be able to uh to train our models and do things that are that are useful
00:07:23
and uh so that that that was all this work uh did oh that's and as you can see
00:07:28
we did you know project in different levels of from very practical a project that will just you know
00:07:34
taking care of data uh on taking your very small things
00:07:37
a very applied to a a european projects or a research project
00:07:43
and so of course uh many people now start start to work on that and
00:07:47
uh if i come from the the uh so just to talk about
00:07:50
the institute of informatics so we have now three different labs that are working
00:07:54
more lies on a on a on energy some more some less
00:07:59
uh you have your uh the the heads of this of this lab so yeah there's
00:08:03
that then yeah there's our presenter from before uh when a a shoe man and the um
00:08:12
so that was that was that that is the mechanic on our side but what is
00:08:15
also very interesting is that we we are connected now to with the the engineering part
00:08:21
and i have to put it just because i forgot the the uh the second part of our uh
00:08:26
don't thing which is the economical portland i mean talked about that just a right before so there's also a
00:08:32
uh people from coming from the economical parts but uh we get us different through
00:08:37
a that is a guy that isn't winning here so we are meeting a knowledge there
00:08:43
but what was very interesting there is that in fact uh we were able to create a the these
00:08:49
are internal stuff for the one that that not at it chooses so but we were able to create a
00:08:54
a new winter domain institutes that that was good i think all the knowledge that we and
00:08:59
and so we have to have that is the that we took this afternoon on it is the the
00:09:03
head of these uh uh into them institute and i think this this is a really really interesting things because
00:09:10
to be able to make our work uh efficient and useful we
00:09:15
need to label this components so we don't need only a a
00:09:19
a research uh parameters uh the high less highest of it uh of of modest
00:09:24
you can have but you also have to implement them and to to have to
00:09:27
collect information uh you have to to create the information system that goes with that
00:09:33
and so it's it's a multi uh i'm pretty low wage a cistern and the um
00:09:40
so what we yeah done the you know i've seen this this morning i think you would see these
00:09:44
uh it so we had to collect a lot of different kind of data and data are so crucial
00:09:49
and the uh but what kind of data what level of data uh what quantity uh is is
00:09:55
very is very important and then to did collect things like since or output which are really uh uh
00:10:02
you know the the basics of it but also uh of course production storage consumption but also things that will
00:10:08
put important to be able to to set up until now more there's a properly what we're also you know
00:10:15
we've seen a administrative data georgia geographical data we have
00:10:19
seen the just just before that the da really useful
00:10:22
but also uh uh the weather will so it it's in this uh today but also a user
00:10:29
behaviour so we have the uh your pen project there and the um yeah it's okay and the um
00:10:36
all this makes um so we we were able to experiment a lot of
00:10:40
different kind of a machine learning models from you know from gaussian mixture to uh
00:10:46
the decision tree around them for us to uh and of course a neural networks
00:10:51
and then uh we'll just collects she home talk about uh one of the little
00:10:56
thing we are doing now which are just no continuing uh uh does does general worked
00:11:02
the two of them into today and just one thing maybe uh that is interesting is
00:11:07
that we we we are used to use different kind of models uh and the um
00:11:15
when you have too many put a lot of data are then then we need this kind of things and approach that i will uh jerome like spring
00:11:22
okay so it hello everyone um which you really quickly if somebody like a real working a
00:11:29
lot and uh it's it's it's yell a lot
00:11:33
she does it in a lot of ah soldiers
00:11:39
you know it's it's uh are very efficient your loss or a set
00:11:46
that's a unit that train it's a
00:11:51
i have this huge meals on what rock stations um
00:12:01
she and small that sold right yeah but i didn't
00:12:05
as well as that it it's all had a huge and
00:12:11
which are different at of a national news
00:12:16
the idea is to use its select friends uh features
00:12:23
right it's it's s. e. t. s.
00:12:27
i think it's it's it's a fast and uh
00:12:34
you can use it as an example to uh_huh
00:12:39
a false rejection of put just a hundred bucks just fine
00:12:48
uh the final uh oh it's just it's it's it's it's it's it just ends folded energy
00:12:57
and it's really a interesting full means but
Conference Program
Workshop introduction by Jérôme Kämpf & Henning Müller + Keynote speech Pr. Dr. Guglielmina Mutani - Place-based energy models for sustainable cities
Pr. Dr. Guglielmina Mutani, Politecnico di Torino
Nov. 15, 2021 · 9 a.m.
333 views
Q&A (Keynote speech: Pr. Dr. Guglielmina Mutani)
Pr. Dr. Guglielmina Mutani, Politecnico di Torino
Nov. 15, 2021 · 9:49 a.m.
Roberto Boghetti - Simulating District Heating Networks’ hydraulics with residual physics neural networks
Roberto Boghetti, Idiap Research Institute
Nov. 15, 2021 · 10 a.m.
Q&A (Roberto Boghetti)
Roberto Boghetti, Idiap Research Institute
Nov. 15, 2021 · 10:13 a.m.
Amine Weibel - Machine learning forecast for photovoltaic power stations (PV) with spatial interpolation on the COSMO-7 model. A real ‘use-case’ on the Czech Republic
Amine Weibel, HES-SO
Nov. 15, 2021 · 10:16 a.m.
Q&A (Amine Weibel, HES-SO Valais-Wallis)
Amine Weibel, HES-SO
Nov. 15, 2021 · 10:27 a.m.
101 views
Dr. Giuseppe Peronato - HeatLearn: a convolutional neural network to assess the building energy consumption at regional scale from open land-use maps
Dr. Giuseppe Peronato, Idiap Research Institute
Nov. 15, 2021 · 10:30 a.m.
Q&A (Dr. Giuseppe Peronato)
Dr. Giuseppe Peronato, Idiap Research Institute
Nov. 15, 2021 · 10:43 a.m.
Pr. Dominique Genoud & Jérôme Treboux - Integrating machine learning and AI in practical energy applications : a review of these last 10 years activities at the Institute of Information Systems
Pr. Dominique Genoud & Jérôme Treboux, HES-SO Valais-Wallis
Nov. 15, 2021 · 10:46 a.m.
Q&A (Pr. Dominique Genoud & Jérôme Treboux)
Pr. Dominique Genoud & Jérôme Treboux, HES-SO Valais-Wallis
Nov. 15, 2021 · 10:59 a.m.
Dr. Marco Tognoli - Fuel-saving potential simulation and estimation when optimally predicting the supply set-point conditions of a District Heating Network with detailed system data-driven modelling
Dr. Marco Tognoli, Politecnico di Milano
Nov. 15, 2021 · 11:31 a.m.
Q&A (Dr. Marco Tognoli)
Dr. Marco Tognoli, Politecnico di Milano
Nov. 15, 2021 · 11:44 a.m.
Pr. David Wannier & Jean-Marie Allder - Research advance in Energy prediction for electrical vehicles and green mobility hotels, the results of the eVIP project
Pr. David Wannier & Jean-Marie Allder, HES-SO Valais-Wallis
Nov. 15, 2021 · 11:46 a.m.
Q&A (Pr. David Wannier & Jean-Marie Allder)
Pr. David Wannier & Jean-Marie Allder, HES-SO Valais-Wallis
Nov. 15, 2021 · 11:58 a.m.
Dr. Kavan Javanroodi - Extending the concept of energy hub to facilitate sector and spatial coupling
Dr. Kavan Javanroodi, Solar Energy and Building Physics Laboratory (LESO-PB) at EPFL
Nov. 15, 2021 · 11:59 a.m.
143 views
Q&A (Dr. Kavan Javanroodi)
Dr. Kavan Javanroodi, Solar Energy and Building Physics Laboratory (LESO-PB) at EPFL
Nov. 15, 2021 · 12:16 p.m.
Pr. Pierre Roduit - Prediction of the electricity flexibility of households using Machine Learning
Pr. Pierre Roduit, HES-SO Valais-Wallis
Nov. 15, 2021 · 12:18 p.m.
Q&A (Pr. Pierre Roduit)
Pr. Pierre Roduit, HES-SO Valais-Wallis
Nov. 15, 2021 · 12:32 p.m.
Loïc Puthod - An open-data acquisition toolchain for AI applications
Loïc Puthod, Centre de recherche Crem
Nov. 15, 2021 · 1:59 p.m.
Q&A (Loïc Puthod)
Loïc Puthod, Centre de recherche Crem
Nov. 15, 2021 · 2:10 p.m.
Cédric Mugabo Serugendo - EnerMaps: The open data tool empowering your energy transition
Cédric Mugabo Serugendo, Centre de recherche Crem
Nov. 15, 2021 · 2:16 p.m.
&Q (Cédric Mugabo Serugendo)
Cédric Mugabo Serugendo, Centre de recherche Crem
Nov. 15, 2021 · 2:27 p.m.
Recommended talks
Intelligent feed-back on user comfort in low energy buildings
Rui Oliveira, University of Aveiro, PT
May 7, 2019 · 10:09 a.m.
169 views
2ème prix: eSMART Technologies
Fabrizio Lo Conte, Laurent Fabre, eSMART Technologies
Nov. 18, 2011 · 8:45 a.m.
157 views