Player is loading...
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
Monday, Nov. 15, 2021 · 10:16 a.m. · 10m 40s
In this presentation, we combine forecasting with machine learning techniques with spatial predictions on meteorological data.This methodology is being tested on a photovoltaic (PV) station in a set of 164 power plants in the Czech Republic.Our results show that this approach can be valid when we do not have access to real weather conditions.
Embed
Transcriptions
Note: this content has been automatically generated.
00:00:00
it's changed perspective mentions amount to supply
00:00:06
and we go from these things
00:00:12
on on the on side of the stage is if you the second starts up
00:00:22
short introduction is uh
00:00:44
okay so
00:00:47
um so hi everyone uh i'm i mean
00:00:50
i'm the research uh associate assistant in the
00:00:54
both institute off uh energy management and energy and and environment
00:00:59
up five stresses so i work there since around the year
00:01:04
ah i'm holding a bachelor's degree in uh energy
00:01:08
and i had followed some courses in a data science
00:01:14
so today i will be presenting the work out francis
00:01:18
casino which is that correct of mine in the lab
00:01:22
and he up early guys for not being here today so i want to
00:01:27
uh mentioned that and not what we're neither quarter of the work so i will
00:01:33
i took the freedom to split the presentation into two parts first part more about his paper
00:01:40
and then the last part about to some project related to
00:01:44
a ion energy yeah in our lab that time uh working on
00:01:50
so as i say that i recall the first
00:01:55
part these uh this paper from francesca listen you know
00:01:59
um it's a paper that uh composes a doctoral
00:02:05
cases that he will be defending a in december
00:02:09
and dan for the second part uh as i
00:02:13
say it uh a few project related to a
00:02:16
high energy that we're conducting in our lab so
00:02:20
francesco whole the double master in statistic and econometrics
00:02:24
uh and he is working for around five years four years in the in the lab
00:02:30
and he defend uh he will be defending his
00:02:34
this in december uh this year and next month
00:02:38
so the paper entitled mission earning forecasts for for the what type power station
00:02:43
with a special interpretation cosmo seven model with a use case in czech republic
00:02:50
so uh a company and industrial uh ask them
00:02:56
uh if they could improve for cost of their hundred around hundred and fifty and two hundred
00:03:02
a t. v. station localise in czech republic in many location
00:03:07
and they had as constraint uh no real measurement data uh of whether
00:03:14
in the site location and and very few or
00:03:17
no a characteristic of this installation beside historical data so
00:03:25
first uh they will uh use the cosmos seven model which is an
00:03:30
atmospheric uh production model from at those trees based on based on physics equation
00:03:36
uh we can have a glance on the table that is returned by
00:03:41
by this model and he is computed or provide our the prediction
00:03:47
and some like cloud cover age uh
00:03:50
radiation a solar radiation temperature like very classic
00:03:55
uh features that say and the t. v. production
00:04:00
on side of this hundred and fifty around side
00:04:04
uh our our daily basis uh for around one and a half year
00:04:10
so at the time in twenty seventeen uh
00:04:13
according to the state of the art uh some
00:04:19
some paper would have visit base modelling for the
00:04:22
for cost uh has some also with neural networks
00:04:27
and uh uh arts features different
00:04:31
many technical note go into details but
00:04:34
some of them also use already metallurgical cited satellite data and the one ah
00:04:43
so to sum up um the methodology
00:04:47
uh it's very uh very common that's a pipeline
00:04:53
and uh um with a first that of reprocessing
00:04:57
the data are cleaning uh handling missing values uh
00:05:02
and so on dan the main idea in this paper was to interpol late
00:05:09
uh the made their o. g. call the top pain
00:05:12
uh from the cosmo mother into the p. v. location site
00:05:17
and the uh we've all their uh common feature and invented technique
00:05:24
and then use an a four distinct model
00:05:29
autoregressive model with seasonal and exertion those values
00:05:34
some other neural network uh architecture uh at this time
00:05:42
so quickly about uh the methodology
00:05:47
so inverse distant waiting was used
00:05:50
to interpol late though metallurgical data from cosmo seven into the p. v. location
00:05:56
uh uh he hard around four thousand point about czech republic
00:06:02
and uh it he dues use this point to interpol wait at the location
00:06:07
then he evaluate autocorrelation function um and to
00:06:11
evaluate the like in importance of seasonal autoregressive
00:06:16
i mean average model and then the feature the
00:06:20
covariance or correlation of the of metal data uh
00:06:28
then um for their resort uh here
00:06:32
on top last weekend see an example of
00:06:36
a single out a single output a singles bad but for
00:06:41
a test set around two hundred and fifty time step and uh
00:06:47
he evaluate or compare the different metrics and it's
00:06:52
pacific only the root mean square error or divided by the max value
00:06:56
and he opting quite similar resort with the different models
00:07:02
and the heat then compare this result to the
00:07:07
other papers obtain uh but by averaging each models
00:07:13
and he allies that or he and notice that you
00:07:18
obtain the quite a smaller or a little improvement in the
00:07:23
in the uh i chrissie off his uh of heat of these come combination of model
00:07:29
and uh he tried to explain uh the resort one by
00:07:35
evaluating the different distribution not the residual of each a model
00:07:43
so he here for
00:07:46
so i invite everyone for any further detailed question
00:07:50
about any of the topics to a right to him
00:07:58
and some project in our lab related to a an
00:08:02
energy both are in a swiss project with different partners
00:08:07
uh and the second one is
00:08:09
specifically we've also many other academy partners
00:08:15
um one of them is the localise energy
00:08:19
focused first mark we which is kind of similar
00:08:22
uh with the presented paper but different location and with different goals
00:08:29
but the idea here will be to use and graph neural network
00:08:34
we've uh he protocol material features and to again tried
00:08:40
to pretty this time multi output and new to step uh
00:08:46
for different um girls or values
00:08:52
and the large a project uh use act clean energy
00:08:58
act change which is with uh also this many partners
00:09:02
the goal here is to um they're blob uh two platform
00:09:07
and uh the idea is to make possible
00:09:10
peer to peer energy trading in local communities
00:09:14
and there will be one platform for the
00:09:17
us to to manage these local energy market
00:09:21
and uh uh a participant application for
00:09:25
a each participant in the market too
00:09:29
to be in this market and is project is specifically
00:09:34
uh taking into a consideration many layer of constraint
00:09:40
physical economy caught and technology caught our
00:09:43
eyes we are also targeted in a two
00:09:48
to certify it changes we've we've i'm blocking
00:09:53
and the that's a bit more in detail
00:09:57
uh the da so part or platform will have as objective
00:10:03
to uh input panoramic creed she's pricing advise or
00:10:08
to steer market in some degree and
00:10:12
the participant application which will be drive by
00:10:17
how to make to try than the
00:10:19
um automated agent strategy and a two
00:10:26
be selected by the party sit and hence therefore
00:10:30
have many different objective uh in these uh trading
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
