KIT – The Research University in the Helmholtz Association
ENERGY INFORMATICS SEMINARINSTITUTE FOR AUTOMATION AND APPLIED INFORMATICS (IAI)
www.kit.edu
Deep Learning for Energy Time SeriesKai Schmieder
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Motivation
Buildings are major energy consumer worldwide(Pérez-Lombard et al. 2008)
Minimize the energy wastage and making power generation and distribution more efficient by intelligent control decisions(Marino et al. 2016, p. 7046)
Load forecasting crucial for mitigating uncertainties of the future
Kai Schmieder: Deep Learning for Energy Time Series2 08.01.19
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Motivation
Deep Learning proved to be useful in manifold fieldsComputer visionSpeech and audio processingNatural language processingRoboticsBioinformatics and chemistryFinance…
(Goodfellow et al. 2016, p.8f)
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series3
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Agenda
Foundation
Long Short-Term Memory for Energy Time Series
Evaluation
Discussion
Conclusion
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series4
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
FOUNDATION
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series5
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Energy Time Series
Aggregate level and building level forecasting categorized byShort-term à one hour to one weekMedium-term à one week to one yearLong-term à ranges longer than one year
(Mocanu et al. 2016, p. 91)
Two main approaches to forecast Energy Time SeriesStatistical and Machine Learning based modelsPhysical Principles based models
(Mocanu et al. 2016, p. 91; Marino et al. 2016, p. 7046)
à here: Short-term & Statistical and Machine Learning based models
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
6
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Deep Learning
Central goal of Machine LearningLearn useful representations of input dataThat get us closer to expected output
Deep LearningSpecific subfield of Machine LearningIdea of successive layers of representationsDepth: how many layers contribute to a model of the dataLayered representations (almost always) learned via neural networks
(Chollet 2018, p. 6ff)
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series7
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
Artificial Intelligence
Machine Learning
DeepLearning
Figure 1: Own representation based on Chollet, 2018, p.4
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Overview of relevant Deep Architectures
08.01.19
DeepArchitectures
Deep Forward Networks
ConvolutionalNeural Networks
RecurrentNeural Networks
Deep Generative Models
Autoencoders
Mixes ofarchitectures
Kai Schmieder: Deep Learning for Energy Time Series8
Figure 2: Own representation based on Goodfellow et al., 2016, p. 11
e.g. Multi-Layer-Perceptrons
e.g. AlexNet for image classification
e.g. Long Short-Term Memory
e.g. Boltzman Machines
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
e.g. Convolutional Neural Networks with Recurrent Neural Network
e.g. For Feature Extraction
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Recurrent Neural Networks
Family of neural networks for processing sequential dataBasic principle
Each member of the output is a function of the previous member of the outputUnrolled Recursive Neural Network
(Goodfellow et al. 2016, p. 363f)Vanishing gradient problem: challenge to learn long-term dependencies(Hochreiter 1991; Bengio et al. 1993, 1994)
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series9
Figure 3: Olah, 2015, http://colah.github.io/posts/2015-08-Understanding-LSTMs/img/RNN-unrolled.png, Retrieved on 16.12.18
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
LONG SHORT-TERM MEMORY FOR ENERGY TIME SERIES
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series10
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Short abstract of Marino et al. (2016)
Application of two architectural variations of Long Short-Term Memory
Stacked Long Short-Term Memory Sequence-to-Sequence Long Short-Term Memory
(Marino et al. 2016)
Experiments on benchmark dataset of electricity consumption for a single residential customer (“Individual household electric power consumption”)(Dheeru and Taniskidou 2017)
Kai Schmieder: Deep Learning for Energy Time Series11 08.01.19
Building Energy Load Forecasting using Deep Neural Networks
Daniel L. Marino, Kasun Amarasinghe, Milos Manic
Department of Computer Science
Virginia Commonwealth University
Richmond, Virginia [email protected], [email protected], [email protected]
Abstract—Ensuring sustainability demands more efficient
energy management with minimized energy wastage. Therefore, the power grid of the future should provide an unprecedented level of flexibility in energy management. To that end, intelligent decision making requires accurate predictions of future energy demand/load, both at aggregate and individual site level. Thus, energy load forecasting have received increased attention in the recent past. However, it has proven to be a difficult problem. This paper presents a novel energy load forecasting methodology based on Deep Neural Networks, specifically, Long Short Term Memory (LSTM) algorithms. The presented work investigates two LSTM based architectures: 1) standard LSTM and 2) LSTM-based Sequence to Sequence (S2S) architecture. Both methods were implemented on a benchmark data set of electricity consumption data from one residential customer. Both architectures were trained and tested on one hour and one-minute time-step resolution datasets. Experimental results showed that the standard LSTM failed at one-minute resolution data while performing well in one-hour resolution data. It was shown that S2S architecture performed well on both datasets. Further, it was shown that the presented methods produced comparable results with the other deep learning methods for energy forecasting in literature.
Keywords—Deep Learning; Deep Neural Networks; Long-Short-Term memory; LSTM; Energy; Building Energy; Energy Load forecasting
I. INTRODUCTION
Buildings are identified as a major energy consumer worldwide, accounting for 20%-40% of the total energy production [1]-[3]. In addition to being a major energy consumer, buildings are shown to account for a significant portion of energy wastage as well [4]. As energy wastage poses a threat to sustainability, making buildings energy efficient is extremely crucial. Therefore, in making building energy consumption more efficient, it is necessary to have accurate predictions of its future energy consumption.
At the grid level, to minimizing the energy wastage and making the power generation and distribution more efficient, the future of the power grid is moving to a new paradigm of smart grids [5], [6]. Smart grids are promising, unprecedented flexibility in energy generation and distribution [7]. In order to provide that flexibility, the power grid has to be able to dynamically adapt to the changes in demand and efficiently distribute the generated energy from the various sources such as renewables [8]. Therefore, intelligent control decisions should
be made continuously at aggregate level as well as modular level in the grid. In achieving that goal and ensuring the reliability of the grid, the ability of forecasting the future demands is important. [6], [9].
Further, demand or load forecasting is crucial for mitigating uncertainties of the future [6]. In that, individual building level demand forecasting is crucial as well as forecasting aggregate loads. In terms of demand response, building level forecasting helps carry out demand response locally since the smart grids incorporate distributed energy generation [6]. The advent of smart meters have made the acquisition of energy consumption data at building and individual site level feasible. Thus data driven and statistical forecasting models are made possible [7].
Aggregate level and building level load forecasting can be viewed in three different categories: 1) Short-term 2) Medium-term and 3) Long-term [6]. It has been determined that the load forecasting is a hard problem and in that, individual building level load forecasting is even harder than aggregate load forecasting [6], [10]. Thus, it has received increased attention from researchers. In literature, two main methods can be found for performing energy load forecasting: 1) Physics principles based models and 2) Statistical and machine learning based models. Focus of the presented work is on the second category of statistical load forecasting. In [7], the authors used Artificial Neural Network (ANN) ensembles to perform the building level load forecasting. ANNs have been explored in detail for the purpose of all three categories of load forecasting [9], [11]-[13]. In [14], the authors use a support vector machines based regression model coupled with empirical mode decomposition to for long-term load forecasting. In [15], electricity demand is forecast using a kernel based multi-task learning methodologies. In [10], authors model individual household electricity loads using sparse coding to perform medium term load forecasting. In the interest of brevity, not all methods in literature are introduced in the paper. For surveys of different techniques used for load forecasting, readers are referred to [16], [17] and [8]. Despite the extensive research carried out in the area, individual site level load forecasting remains to be a difficult problem.
Therefore, the work presented in this paper investigates a deep learning based methodology for performing individual building level load forecasting. Deep learning allows models composed of multiple layers to learn representations in data. The use of multiple layers allow the learning process to be carried out with multiple layers of abstraction. A comprehensive overview and a review of deep learning methodologies can be
978-1-5090-3474-1/16/$31.00 ©2016 IEEE 7046
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
Figure 4: Marino et al., 2016, p. 7046
Energy Informatics Seminar
Institute for Automation and Applied Informatics (IAI)
Overview of presented Long Short-Term Memory architectures
Stacked
Long Short-Term Memory
Sequence-to-Sequence
Long Short-Term Memory
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series12
Foundation DiscussionLong Short-Term Memory for
Energy Time Series Evaluation Conclusion
Figure 5:
Own representation
based on
Marino et al., 2016,
p. 7047Figure 6:
Own representation
Energy Informatics Seminar
Institute for Automation and Applied Informatics (IAI)
Long Short-Term Memory
Type of Recurrent Neural Network introduced by Sepp Hochreiter and
Jürgen Schmidhuber (1997)
Great success in various applications (Goodfellow et al. 2016, p. 363f)
Solves vanishing gradient problem
Key concepts
Cell State
Gates
Input
Forget
Output
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series13
Foundation DiscussionLong Short-Term Memory for
Energy Time Series Evaluation Conclusion
Figure 7: Olah, 2015,
http://colah.github.io/posts/2015-08-Understanding-
LSTMs/img/LSTM3-chain.png, Retrieved on 03.12.2018
Energy Informatics Seminar
Institute for Automation and Applied Informatics (IAI)
Long Short-Term Memory Cells
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series14
0 Cell State
Legend
Foundation DiscussionLong Short-Term Memory for
Energy Time Series Evaluation Conclusion
Figure 8: Olah, 2015,
http://colah.github.io/posts/2015-08-
Understanding-LSTMs/img/LSTM3-C-
line.png, Retrieved on 03.12.2018
1 Forget Gate
Figure 9: Adapted from Olah, 2015,
http://colah.github.io/posts/2015-08-
Understanding-LSTMs/img/LSTM3-
focus-f.png, Retrieved on 03.12.2018
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Long Short-Term Memory Cells
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series15
2 Input Gate
Legend
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
Figure 10: Adapted from Olah, 2015, http://colah.github.io/posts/2015-08-Understanding-LSTMs/img/LSTM3-focus-i.png, Retrieved on 03.12.2018
3 Update Cell State
Figure 11: Adapted from Olah, 2015, http://colah.github.io/posts/2015-08-Understanding-LSTMs/img/LSTM3-focus-C.png, Retrieved on 03.12.2018
Energy Informatics Seminar
Institute for Automation and Applied Informatics (IAI)
Long Short-Term Memory Cells
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series16
Legend
4 Output Gate
Foundation DiscussionLong Short-Term Memory for
Energy Time Series Evaluation Conclusion
Figure 12: Adapted from Olah, 2015,
http://colah.github.io/posts/2015-08-
Understanding-LSTMs/img/LSTM3-
focus-o.png, Retrieved on 03.12.2018
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Stacked Long Short-Term Memory architecture
Stacked architectureStack multiple Long Short-Term Memory Cells into a multi-layer architectureDense layer
Input vector: i[t] = [ y[t-1] day[t] day-of-week[t] hour[t] ]Use also more than one time step as “context”
Output vector:
(Marino et al. 2016)
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series17
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
f[t]
Figure 5: Own representation based on Marino et al., 2016, p. 7047
y[t]^
Energy Informatics Seminar
Institute for Automation and Applied Informatics (IAI)
Overview of presented Long Short-Term Memory architectures
Stacked
Long Short-Term Memory
Sequence-to-Sequence
Long Short-Term Memory
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series18
Foundation DiscussionLong Short-Term Memory for
Energy Time Series Evaluation Conclusion
Figure 5:
Own representation
based on
Marino et al., 2016,
p. 7047
Slide from before
Figure 6:
Own representation
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
The idea of Sequence-to-Sequence Learning
Introduced by Sutskever et al. (2014) to map sequences of different lengths
Comprised of two sub-modelsEncoder
Convert input sequences of variable length andEncode them in a fixed length vector, which is then used as input state for the decoder
DecoderGenerates an output sequence of fixed length
(Marino et al. 2016, p. 7048)
Applications in speech recognition, machine translation and question answering (Goodfellow et al. 2016, p. 385)
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series19
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Sequence-to-Sequence Long Short-Term Memory architecture
EncoderInput vector: i[t] = [ y[t-1] day[t] day-of-week[t] hour[t] ]
“Output”: Cell State
DecoderInput vector: f[t]
Output vector:
(Marino et al. 2016)
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series20
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
Figure 6: Own representation
f[t]
y[t]^
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
EVALUATION
Kai Schmieder: Deep Learning for Energy Time Series21 09.01.1908.01.19
Energy Informatics Seminar
Institute for Automation and Applied Informatics (IAI)
Horizon
Data
Experiments in Marino et al. (2016)
“Individual household electric power consumption” (Dheeru and Taniskidou 2017)
Aggregation Level
1 minute (original)
1 hour
Training set: 3 years; Test set: 1 year
Kai Schmieder: Deep Learning for Energy Time Series22 09.01.19
One Step
60 Steps
Evaluation and results
Model
Foundation DiscussionLong Short-Term Memory for
Energy Time SeriesEvaluation Conclusion
Stacked LSTM
Sequence-to-Sequence LSTM
Stacked LSTM performs well with inputs further in the past for hourly data
Sequence-to-Sequence LSTM performs well in both datasets
Comparable results to Mocanu et al. (2016)
08.01.19
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Implementation and hardware resources
Jupyter Notebooks at https://github.com/nicoleludwig/EnergyInformatics
Implementation with PythonPandas NumPy scikit-learnKeras with TensorFlow as backendChartify
Hardware resources for trainingGPU: NVIDIA Tesla V100*
Kai Schmieder: Deep Learning for Energy Time Series23
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
09.01.19
* Made possible by Andreas Bartschat (IAI), thanks again! J
08.01.19
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Data for evaluation
Energy consumption data of KIT Campus NordBuilding 12415 minutes resolutionTraining set of three years(Mid May 2012 until Mid May 2015)Test set of one year(Mid May 2015 until Mid May 2016)
Kai Schmieder: Deep Learning for Energy Time Series24
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
09.01.19
36
36
L559
L559
Einfahrt Nord
100m
1600
906
L559
L559
Karlsruher Allee
Grabener Straße
Eggensteiner Straße
Liedolsheimer Straße
eßartS remiehnekniL
eßartS rettetshcoH
ztalP-ztlohmle
H-nov-nnamre
H
eellA renefahsdlopoeL
eßartS relatshcirdeirF
.rtS rehcolneknalB
Spöcker Straße
Weingartener Straße
eßartS remiehnekniL
Unt
ergr
omba
cher
Str
aße
Blankenlocher Straße
Blan
kenl
oche
r St
raße
meteorologischerMessmast
Büchenauer StraßeBüchenauer Straße
Neureuter Straße
Eggensteiner Straße
Karlsruher Allee
Weingartener Straße
Spöcker Straße
Grabener Straße
eellA renefahsdlopoeL
CASINO
Werkfeuerwehr
KIT-Shuttle
Anmeldung
trha
fnie
tpua
H
Lieferzufahrt
H
Administration
Bibliothek
FTU
KITA
H
Einfahrt Nord
KIT-Shuttle
Anmeldung ITU
Anmeldung WAK
FIZ
Anmeldung WAK
Eltern-Kind-Büro
FIZ
FIZ
FIZ
FIZ
H2
Strahlenpassstelle KIT
211
212
210
213
232
231
226224222
227225
223
223/1
221
121122
123
101103
5131
131
126
124
142
144
141241
240
230
238239
244
242
249
248
245
243
220 2505252
256
257
23723
5
251
261
260
260/1
252
234
253
254
266
258
259
255
278
272
274273276
400
440
412413
415
420
414
410
418 419409
406
411
408
405
404
5404
403
407430
5402
401A
402
401
301
300
309
307
310
302
308303
305
304
317
312
306
316
152
151
311
330
332
333 315
324
329
327
326
322
321
421
422
429
428
423
424
9627
434 426
425
443
447436
433
432 444
438 446
448
435439
459
454
458
453
456
464
461
462
460
4529652
9653450
451
445442441
441/1
341
343
345
3499642
348
352
351
353
517 500518
511
510
513
512515
514
516
527
573
574575
549
544
541542
522523
521
5523
605 606
601 602
603
607
640
611
615
604
631633
629 620
5621
608
618
650614
690
663
660
670
692
691693 689
688
696
695
682
684/1684
684/2
686
708
704
706707
713 719
714
712
718
9670
733
724
723
726
736
732
721
725
727
730
729
229
323
681
441
1600
720
710 717
449
9651
264
737
457
9615
247
701
630
908
746
747
356
9602
9675
9683
612
687
9900
145
416
127
906
705
Richtung B36, Leopoldshafen und Campus Süd
Richtung Stutensee/Blankenloch
S1/ S11 von/nach
Karlsruhe
KIT-Campus NordLieferanschrift: Hermann-von-Helmholtz-Platz 176344 Eggenstein-Leopoldshafen
Richtung Stutensee/FriedrichstalRichtung B36, Linkenheim-Hochstetten
Web
erstr
aße
Beet
hofe
nstr
aße
Mozartstraße
Haydnplatz
Moltkestraße
Maximilianstraße
Nördliche Hildapromenade
100m
Nöthnitzer Str.
Hallwachsstraße
Helm
ho
ltzstraße
Zeuner Str.
Dül
fers
traß
e
Wei
ßbac
hstr
aße
Mommsenstraße
100m
Kre
uzec
kbah
nstr
aße
100m
3101
StandortOstendorfhausLieferanschrift:Weberstraße 576133 Karlsruhe
Standort DresdenProjektträger Karlsruhe – Außenstelle Dresden
Lieferanschrift:Hallwachsstraße 301069 Dresden
KIT-Campus AlpinInstitut für Meteorologie und Klimaforschung (IMK-IFU)
Lieferanschrift:Kreuzeckbahnstraße 1982467 Garmisch-Partenkirchen
Richtung Garmisch-Partenkirchen und B23
RichtungKreuzeck- / Alpspitzbach
906
36
L559
L559
Karlsruher Allee
Grabener Straße
Eggensteiner Straße
Liedolsheimer Straße
eßartS remiehnekniL
eßartS rettetshcoH
ztalP-ztlohmle
H-nov-nnamre
H
eellA renefahsdlopoeL
eßartS relatshcirdeirF
.rtS rehcolneknalB
Spöcker Straße
Weingartener Straße
eßartS remiehnekniL
Unt
ergr
omba
cher
Str
aße
Blankenlocher Straße
Blan
kenl
oche
r St
raße
meteorologischerMessmast
Büchenauer StraßeBüchenauer Straße
Neureuter Straße
Eggensteiner Straße
Karlsruher Allee
Weingartener Straße
Spöcker Straße
Grabener Straße
eellA renefahsdlopoeL
CASINO
Werkfeuerwehr
KIT-Shuttle
Anmeldung
trha
fnie
tpua
H
Lieferzufahrt
H
Administration
Bibliothek
FTU
KITA
H
Einfahrt Nord
KIT-Shuttle
Anmeldung ITU
Anmeldung WAK
FIZ
Anmeldung WAK
Eltern-Kind-Büro
FIZ
FIZ
FIZ
FIZ
H2
Strahlenpassstelle KIT
211
212
210
213
232
231
226224222
227225
223
223/1
221
121122
123
101103
5131
131
126
124
142
144
141241
240
230
238239
244
242
249
248
245
243
220 2505252
256
257
23723
5
251
261
260
260/1
252
234
253
254
266
258
259
255
278
272
274273276
400
440
412413
415
420
414
410
418 419409
406
411
408
405
404
5404
403
407430
5402
401A
402
401
301
300
309
307
310
302
308303
305
304
317
312
306
316
152
151
311
330
332
333 315
324
329
327
326
322
321
421
422
429
428
423
424
9627
434 426
425
443
447436
433
432 444
438 446
448
435439
459
454
458
453
456
464
461
462
460
4529652
9653450
451
445442441
441/1
341
343
345
3499642
348
352
351
353
517 500518
511
510
513
512515
514
516
527
573
574575
549
544
541542
522523
521
5523
605 606
601 602
603
607
640
611
615
604
631633
629 620
5621
608
618
650614
690
663
660
670
692
691693 689
688
696
695
682
684/1684
684/2
686
708
705704
706707
713
719
714
712
718
9670
733
724
723
726
736
732
721
725
727
730
729
229
323
681
441
1600
720
710 717
449
9651
264
737
457
9615
247
701
630
908
746
747
356
9602
9675
9683
612
687
9950
145
416
127
906
Richtung B36, Leopoldshafen und Campus Süd
Richtung Stutensee/Blankenloch
S1/ S11 von/nach
Karlsruhe
KIT-Campus NordLieferanschrift: Hermann-von-Helmholtz-Platz 176344 Eggenstein-Leopoldshafen
Richtung Stutensee/FriedrichstalRichtung B36, Linkenheim-Hochstetten
Web
erstr
aße
Beet
hofe
nstr
aße
Mozartstraße
Haydnplatz
Moltkestraße
Maximilianstraße
Nördliche Hildapromenade
Nöthnitzer Str.
Hallwachsstraße
eßar
tszt
loh
mle
H
Zeuner Str.
Dül
fers
traß
e
Mommsenstraße
100m
Kre
uzec
kbah
nstr
aße
100m
3101
StandortOstendorfhausLieferanschrift:Weberstraße 576133 Karlsruhe
Standort DresdenProjektträger Karlsruhe – Außenstelle Dresden
Lieferanschrift:Hallwachsstraße 301069 Dresden
KIT-Campus AlpinInstitut für Meteorologie und Klimaforschung (IMK-IFU)
Lieferanschrift:Kreuzeckbahnstraße 1982467 Garmisch-Partenkirchen
Richtung Garmisch-Partenkirchen und B23
RichtungKreuzeck- / Alpspitzbach
HerausgeberKarlsruher Institut für TechnologieKaiserstraße 12, 76131 Karlsruhe
Kontakt
Facility Management (FM)Bauprojekte und Immobilien – Immobilienmanagement (BPI-IMM)[email protected]
Stand:Mai 2016
Karl-Wilhelm-Platz4, 5
H
H
H
H
H
Engesserstraße
Richard-Willstätter-Allee
Engl
er-B
unte
-Rin
g
Hagsfelder AlleeWaldparkplatz
Gotthard-Franz-Straße
Schönfeldstraße
Karl-Wilhelm-Stra
ße
Haid-und-Neu-Straße
EdelsheimstraßeAm Fasanengarten
Parkstraße
eßar
tS-z
tinß
eirP
-zne
cni
V
Ade
naue
rrin
g
Leonhard-Sohncke-Weg
Ernst-Gaber-Str.
Engelbert-Arnold-Straße
Wilhelm-Nußelt-Weg
Rudo
lf-P
lank
-Str
.
Frit
z-H
aber
-Weg
Wol
fgan
g-G
aede
-Str
aße
Iren
e.-R
osen
berg
.-Str
.
R.-Baumeister-Platz
Otto-Ammann-PlatzEhrenhof
Paulcke-Platz
W.-
Jord
an
-Weg
Stra
ße a
m F
orum
Haupteinfahrt
Nordeinfahrtmit Codekarte
Einfahrt
Einfahrt
Einf
ahrt
10.50
Studentenwerk
StudentenzentrumKIT-Shuttle
Forstschule
Sportanlagen
10.30
10.7
0
FORUM
MENSA
AUDIMAX
SPORT
BIBLIOTHEK
30.21
11.20
11.1
1
11.10
11.30
10.9
1
10.92 10.93 10.94
10.9
5
10.96
11.2
3
11.21
11.2
2
10.81
10.82
10.83 10.84 10.85
10.86 10.87
10.8
8 10.89
10.6
110
.64
10.6310.62
02.17
02.12
02.1
3
02.1
402.1
5
02.16
02.18
30.91
30.9
3
30.90
30.80
30.4
1
30.4
2
30.43
30.44
30.45
30.4
6
30.31
30.3
2
30.33
30.3
430.35
30.36
30.81 30.8230.51
30.5
0
01.12
01.1
3
30.60 30.6130.95
30.9
6
30.70
30.79
30.29
40.11
40.12
40.13
40.1
4
40.15
40.16
40.17 40.18
40.1
9
40.0
2
40.0
4
40.21 40.22 40.2
340
.24
40.2
5
40.2
6
40.27
40.28
40.29
40.3
140
.32
40.33
40.40
40.43
50.20
50.2150.22
50.2350.2450.25
50.1050.12
50.31
50.32
07.21 07.30
50.33
50.34
50.35
50.3650.38
50.4
0
50.41
7
51
13
11b
11a
4
15
42
2-6
68
17
7
31
8
6
4
2
7
1 1a
2
3
4
651
7
9 3
22
2
1
4
6 8 102
7
2018
5
5a
2
8
19
4
2
5
5
32
7
9
109
8
7
20
20b
71
3
59b
9a11
a 11c
11a
11b
1412
24
8
15
23
22
21
3
20a StandortsucheService-Telefon +49 721 608-100008:00 – 18:00 Uhr
Campuspläne als App www.pkm.kit.edu/navigator
Anmeldung / Information .......................................................................................................................... 221 ......B 09Casino ....................................................................................................................................................... 145 ......B 09Dienstleistungsrotunde .............................................................................................................................. 210 ......B 09
Fortbildungszentrum für Technik und Umwelt (FTU) ................................................................................. 101 ......B 09KiTa „Nanos“ ........................................................................................................................................... 213 ..... C 09KIT-Bibliothek Campus Nord ....................................................................................................................... 303 ......B 08 Ostendorfhaus .......................................................................................................................................... 191 ..... D 01
Administration:
Administration ........................................................................................................................................... 141 ......B 09Medizinische Abteilung ............................................................................................................................ 124 ......B 09Sicherheitsmanagement ........................................................................................................................... 439 ..... D 08Strahlenpassstelle des KIT Campus Nord .................................................................................................... 436 ..... C 08Werkfeuerwehr .......................................................................................................................................... 315 ..... A 08
Projektträger Karlsruhe (PTKA):
PTKA – Baden-Württemberg Programme (BWP) .......................................................................................... 438 ..C/D 08PTKA – Produktion und Fertigungstechnologien (PFT) ................................................................................. 436 ..... C 08PTKA – Standort Dresden ....................................................................................................................................... D 01PTKA – Wassertechnologie und Entsorgung (WTE) ................................................................................... 9675 ..C/D 08
Institute:
Angewandte Informatik (IAI) ...................................................................................................................... 449 ..... C 07Angewandte Materialien – Angewandte Werkstoffphysik (IAM-AWP) ........................................................ 681 C 05/06Angewandte Materialien – Keramische Werkstoffe und Technologien (IAM-KWT) ...................................... 575 ..... A 06Angewandte Materialien – Werkstoff- und Biomechanik (IAM-WBM) ......................................................... 696 ..... C 06Angewandte Materialien – Werkstoffkunde (IAM-WK) ............................................................................... 695 ..... C 06Angewandte Materialien – Energiespeichersysteme (IAM-ESS) ................................................................. 441/1 .. B/C 07Biologische Grenzfl ächen 1 – Biomolekulare Micro- und Nanostrukturen (IBG 1) ........................................ 601 ..... C 07Biologische Grenzfl ächen 2 – Molekulare Biophysik (IBG 2) ....................................................................... 352 ..... A 07Biologische Grenzfl ächen 3 – Makromolekulare Materialien (IBG 3) ............................................................ 601 ..... C 07Biologische Grenzfl ächen 4 – Magnetische Resonanz (IBG 4) ..................................................................... 351 ..... A 07Biologische Grenzfl ächen 5 – Biotechnologie und Mikrobielle Genetik (IBG 5) ........................................... 601 ..... C 07Beschleunigerphysik und Technologie (IBPT) ............................................................................................... 329 . B 07/08Experimentelle Kernphysik (IEKP) ................................................................................................................ 401 .. B/C 08Funktionelle Grenzfl ächen (IFG) .................................................................................................................. 330 ..... A 08Festkörperphysik (IFP) ................................................................................................................................. 425 ..... C 08Hochleistungsimpuls- und Mikrowellentechnik (IHM) ................................................................................. 421 .. B/C 08Katalyseforschung und -technologie (IKFT) ................................................................................................. 727 ......B 05Kernphysik (IKP) ......................................................................................................................................... 401 .. B/C 08Meteorologie und Klimaforschung – Atmosphärische Spurengase und Fernerkundung (IMK-ASF) ............... 435 ..... D 08Meteorologie und Klimaforschung – Atmosphärische Umweltforschung (IMK-IFU) ................... Campus Alpin ..... D 02Meteorologie und Klimaforschung – Atmosphärische Aerosol Forschung (IMK-AAF) ................................... 326 ......B 08Meteorologie und Klimaforschung – Forschungsbereich Troposphäre (IMK-TRO) ........................................ 435 ..... D 08Mikrostrukturtechnik (IMT) ......................................................................................................................... 301 ......B 08Mikroverfahrenstechnik (IMVT) .................................................................................................................. 605 .. B/C 07Nukleare Entsorgung (INE) ......................................................................................................................... 714 ..... A 05Neutronenphysik und Reaktortechnik (INR) ................................................................................................ 521 ......B 06Nanotechnologie (INT) ............................................................................................................................... 640 .C 06/07Prozessdatenverarbeitung und Elektronik (IPE) ............................................................................................ 242 ..... C 09Photon Science and Synchrotron Radiation (IPS) ......................................................................................... 329 . B 07/08Technikfolgenabschätzung und Systemanalyse (ITAS) ............................................................................... 3300 ......K‘heTechnische Chemie (ITC) ............................................................................................................................ 403 ..... C 08Technische Physik (ITEP) .............................................................................................................................. 418 ..... C 08Tritiumlabor Karlsruhe (ITEP-TLK) ................................................................................................................ 451 ..... C 07Toxikologie und Genetik (ITG) .................................................................................................................... 305 ......B 08Laboratorium für Applikationen der Synchrotronstrahlung (LAS)................................................................. 329 . B 07/08Steinbuch Centre for Computing (SCC) ...................................................................................................... 449 ..... C 07
Helmholtz-Programme:
Atmosphäre und Klima (ATMO) ................................................................................................................. 435 ..... D 08BioInterfaces in Technology and Medicine (BIFTM) ...................................................................................... 440 ..... D 08Energieeffizienz, Materialien und Ressourcen (EMR) ................................................................................... 435 ..... D 08Erneuerbare Energien (EE) ......................................................................................................................... 435 ..... D 08Kernfusion (FUSION) .................................................................................................................................. 451 ..... C 07Materie und Technologie (MT) .................................................................................................................... 242 ..... C 09Materie und Universum (MU) .................................................................................................................... .. B/C 08Nukleare Entsorgung und Sicherheit sowie Strahlenforschung (NUSAFE) .................................................... 433 ..... C 08Science and Technology of Nanosystems (STN) ........................................................................................... 440 ..... D 08Speicher und vernetzte Infrastrukturen (SCI) ............................................................................................... 418 ..... C 08Supercomputing & Big Data (SBD) .............................................................................................................. 449 ..... C 07Technologie, Innovation und Gesellschaft (TIG) ......................................................................................... 3300 ..... K‘heVon Materie zu Materialien und Leben (MML) ........................................................................................... .....B 7/8
Warnhinweise
Interne Notrufnummer 3333
Auf dem Campus gilt die Straßenverkehrsordnung.
Höchstgeschwindigkeit 50 km/h.
Vorfahrtsregel „rechts vor links“ beachten.
Sicherheitsbereiche dürfen nur unter Beachtung örtlicher Regularien betreten werden.
Zutritt ist tageszeitlich begrenzt.
Karlsruher SchlossKarlsruher SchlossKarlsruher Schloss
Karl-Wilhelm-Platz4, 5
Kronenplatz
Marktplatz2, 3, 5, S1, S11, S4, S41, S5
Marktplatz1, 2, 4, 5, S2, S4, S41, S5
1, 2, 4, 5, S2, S4, S41, S5
H
H
Durlacher Tor /KIT Campus-Süd1, 2, 4, 5, S2, S4, S41, S5
01.85
01.90
01.9201.93
01.94
Waldparkplatz
Ernst-Gaber-Str.
Paulcke-Platz
W.-J
ord
an-W
eg
Engesserstraße
ZähringerstraßeKreuzstraße
Schlossplatz
Zirkel
Kronenstraße
Waldhornstraße
Fritz-Erler-Straße
Waldhornstraße
Berliner Platz
Kaiserstraße
Kapell
enstr
aße
Durlacher Allee
R.-Baumeister-Platz
Otto-Ammann-PlatzRu
dolf
-Pla
nk-S
traß
e
Wilhelm-Nusselt-WegEhrenhof
Engelbert-Arnold-Straße
Iren
e-Ro
senb
erg-
Stra
ße
Engesserstraße
Englerstraße
Lehm
anns
traß
e Wol
fgan
g-G
aede
-Str
aße
Leonhard-Sohncke-Weg
Neu
er Z
irkel
geW-reba
H-ztirF
Schlossbezirk
Simon-M
oser-Weg
Richard-Willstätter-Allee
muroF ma eßartS
Engl
er-B
unte
-Rin
g
gnirreuanedA
Hagsfelder Allee
Am Fasanengarten
Gotthard-Franz-StraßeKornblumenstraße
Fraunhoferstraße
eßar
tS-z
tinß
eirP
-zne
cni
V
Edelsheimstraße
Parkstraße
Schönfeldstraße
Haid-und-Neu-Straße
Karl-Wilhelm-Stra
ße
Lärchenallee
Wes
tein
fahrt
mit
Codekar
te
Haupteinfahrt
Nordeinfahrtmit Codekarte
Einfahrt
Einfahrt
trhafniE
10.50
Studierendenwerk
StudentenzentrumKIT-Shuttle
Administration
Studierendenservice
Steinbuch Centre forComputing (SCC)
BauamtGastdozentenhaus
Forstliches Bildungs-zentrum Karlsruhe
Sportanlagen
10.40
10.30
10.7
0
FORUM
MENSA
Audimax
SPORT
10.31
10.32 10.34
10.33
KIT-BIBLIOTHEK
01.5101.52
30.12 30.11
30.10
20.1120.12
20.13
20.14
01.53
20.20
20.21
20.50 20.51 20.52
02.1
1
02.1
0
20.54
20.5302.02 02.03 02.04
20.30
02.95
30.21
30.22
30.2
3
30.24
30.2
5
20.4011.20
11.1
1
11.10
11.30
11.4
010
.1201
.80
10.11
10.9
1
10.92 10.93 10.94
10.9
5
10.96
10.21
01.95
10.23
11.2
3
11.21
11.2
2
10.81
10.82
10.83 10.84 10.85
10.86 10.87
10.8
8 10.89
10.6
110
.64
10.6310.62
02.17
02.12
02.1
3
02.1
402.1
5
02.16
02.18
30.91
30.9
3
30.90
30.41
30.42
30.43
30.44
30.45
30.31
30.33
30.3
430.35
30.36
30.81 30.8230.51
30.5
0
01.12
01.1
3
30.60 30.6130.95
30.9
6
30.70
30.79
30.29
40.12
40.13
40.1
4
40.15
40.16
40.17 40.18
40.1
9
40.0
2
40.0
4
40.21 40.22 40.2
340
.24
40.2
5
40.2
6
40.27
40.28
40.29
40.3
140
.32
40.33
40.40
40.44
40.41
40.43
50.20
50.2150.22
50.2650.25
50.1050.12
50.3150.32
07.21
50.33
50.34
50.35
50.3650.38
50.4
0
50.41
01.70
05.01 05.20
01.96
07.0707.08
07.30
07.31
07.01
Adolf-Würth-Gebäude
30.28
PRÄSIDIUM
11.30
10.9
1
10.9610.96
11.30
10.9
1
10.9610.96
KinderUniversum
Studienberatung
30.46
30.3
2
11
12
13
14
19
1
2a
2
2
3
5
5a 5b
24
6
13
2
7
7
9
7
51
13
11b
11a
4
15
42
2-6
68
17
31
8
64
2
7
1 1a
2
3
4
651
7
9
8a4a
8 422a 2b
3
22
1042
1
4
6 8 10
12
2
13
14
11
7
2018
5
5a
2
8
1
6
4
2
5
5
3
13
2
7
9
109
8
7
65
11
71
3
59b
9a11
a 11c
11a
11b
14122
4
8
15
23
2221
3
89-93
1-3
27
3032
34
53
3
6
14
13
31
1a
13
30.49
30.48
KIT-Campus Süd
Lieferanschrift:Kaiserstraße 1276131 Karlsruhe
Eggenstein-Leopoldshafen
Karlsruhe A5Durlacher Allee
Moltkestr.
Basel/Stuttgart
Ausfahrt Karlsruhe Durlach
Frankfurt
Karlsruhe
Graben-Neudorf
36
L558
L559
Adenauerring
KIT-Campus Nord
Ostendorfhaus
KIT-Campus West
KIT-Campus Ost
KIT-Campus Süd
KIT-StandorteKarlsruhe
LAGEPLAN CAMPUS NORD
KIT-StandorteDeutschland
Karlsruhe
Dresden
Garmisch
Ulm
Black Forest Observatory
Helmholtz-Institut Ulm (HIU)Lieferanschrift:Helmholtzstraße 1189081 Ulm
Richtung Ulm-West und Stuttgart (A8)
B10/A8
B10
Richtung Neu-Ulm/Wiblingen/Biberach
Helmholtzstraße
Berliner Ring
Albert-Einstein-Allee
Blaubeuren
B10
Richtung Ulm-Donautal
Ulm Eselsberg
1
2
3
4
5
6
7
8
9
10
A B C D
D
A B C D
1
2
1
2
3
4
5
6
7
8
9
10
3
KIT – Die Forschungsuniversität in der Helmholtz-Gemeinschaft
100 mN
Neubaumaßnahmen
Fremdinstitutionen
Zaunanlage/Arealbegrenzung Fremdinstitution
4
401
329
Eltern-Kind-Büro .............................................................................................................................. 226 C 09 .....
HerausgeberKarlsruher Institut für TechnologieKaiserstraße 12, 76131 Karlsruhe
Kontakt
Facility Management (FM)Bauprojekte und Immobilien – Immobilienmanagement (BPI-IMM)[email protected]
Stand:Juli 2016
Karlsruher Schloss
Durlacher Tor4, 5, S2
Karl-Wilhelm-Platz4, 5
Durlacher Tor1, 2, S4, S5
Kronenplatz
Marktplatz2, 5, S1, S11, S4, S41
Marktplatz1, 2, 3, 4, 5, S1, S11, S2, S4, S41, S5
1, 3, 4, 5, S2, S4, S5
3
H
H
H
H
H
H
Kaiserstraße
Durlacher Allee
Engesserstraße
Schlossbezirk
Engesserstraße
Richard-Willstätter-Allee
Eng
ler-
Bu
nte
-Rin
g
Hagsfelder AlleeWaldparkplatz
Gotthard-Franz-Straße
Schönfeldstraße
Karl-Wilhelm-Stra
ße
Haid-und-Neu-Straße
EdelsheimstraßeAm Fasanengarten
Parkstraße
Vin
cenz-Prieß
nitz-Straß
e
Ad
enau
erri
ng
Leonhard-Sohncke-Weg
Ernst-Gaber-Str.
Engelbert-Arnold-Straße
Wilhelm-Nußelt-Weg
Ru
do
lf-P
lan
k-St
r.
Frit
z-H
aber
-Weg
Wo
lfg
ang
-Gae
de-
Stra
ße
Iren
e.-R
ose
nb
erg
.-St
r.
Waldhornstraße
Kronenstraße
Fritz-Erler-Straße
Waldhornstraße
Schlossplatz
Zirkel
Englerstraße
Simon-M
oser-Weg
Lehm
anns
traß
e
Neu
er Z
irkel
R.-Baumeister-Platz
Otto-Ammann-Platz
Berliner Platz
Zähringerstraße
Kreuzstraße
Ehrenhof
Paulcke-Platz
Kapell
enstr
aße
W.-
Jord
an
-Weg
Moltkestraße
Stra
ße
am F
oru
m
Wes
tein
fahrt
mit
Codekar
te
Haupteinfahrt
Nordeinfahrtmit Codekarte
Einfahrt
Einfahrt
Ein
fah
rt
10.50
Studentenwerk
StudentenzentrumKIT-Shuttle
Administration
Studentensekretariat Prüfungsamt
Rechenzentrum
BauamtGastdozentenhaus
Forstschule
Sportanlagen
10.40
10.30
10.70
FORUM
MENSA
AUDIMAX
SPORT
10.31
10.32 10.34
10.33
BIBLIOTHEK
01.5101.52
30.12 30.11
30.10
20.1120.12
20.13
20.14
01.53
20.20
20.21
20.50 20.51 20.52
20.53
02.11
02.10
02.02 02.03 02.04
20.30
02.95
30.21
30.22
30.23
30.24
30.25
20.40
11.20
11.11
11.10
11.30
11.40
10.12
01.80
10.11
10.91
10.92 10.93 10.94
10.95
10.96
10.2110.23
11.23
11.21
11.22
10.81
10.82
10.83 10.84 10.85
10.86 10.87
10.88 10.89
10.61
10.64
10.63
10.62
02.17
02.12
02.13
02.1402.15
02.16
02.18
30.91
30.93
30.90
30.80
30.41
30.42
30.43
30.44
30.45
30.46
30.31
30.32
30.33
30.3430.35
30.36
30.81 30.82
30.51
30.50
01.12
01.13
30.60 30.6130.95
30.96
30.70
30.79
30.29
40.11
40.12
40.13
40.14
40.15
40.16
40.17 40.18
40.19
40.02
40.04
40.21 40.22 40.23
40.24
40.25
40.26
40.27
40.28
40.29
40.31
40.32
40.33
40.40
40.43
50.20
50.21
50.22
50.23
50.2450.25
50.1050.12
50.31
50.32
07.21 07.30
50.33
50.34
50.35
50.3650.38
50.40
50.41
01.70
05.01 05.20
01.96
11
12
13
14
19
1
2a
2
2
3
5
5a 5b
24
6
13
2
7
7
9
3
11a
7
51
13
11b
11a
4
15
42
2-6
68
17
7
31
8
6
4
2
7
1 1a
2
3
4
651
7
9
9a9b
8 422a 2b
3
22
1042
1
4
6 8 10
12
2
13
14
11
7
2018
5
5a
2
8
19
4
2
5
5
32
7
9
109
8
7
20
20b
65
11
71
3
59b
9a
11a 11c
11a
11b
14
12
24
8
15
23
22
21
3
89-9
3
27
20a
40.1140.11
40.1240.1240.1240.1240.1240.12
40.1340.13
40.14
40.19
40.02
40.02
40.02
40.21
40.2740.27
771
3
5
24
StandortsucheService-Telefon +49 721 608-100008:00 – 18:00 Uhr
Campuspläne als App www.pkm.kit.edu/navigator
Anmeldung / Information ............................................................................................................................ 221 ......B 09Casino ....................................................................................................................................................... 145 ......B 09Dienstleistungsrotunde .............................................................................................................................. 210 ......B 09Eltern-Kind-Büro ........................................................................................................................................ 226 .... C 09Fortbildungszentrum für Technik und Umwelt (FTU) ................................................................................... 101 ......B 09KiTa „Nanos“ ........................................................................................................................................... 213 ..... C 09KIT-Bibliothek Campus Nord ....................................................................................................................... 303 ......B 08 Ostendorfhaus .......................................................................................................................................... 191 ..... D 01
Administration:
Administration ........................................................................................................................................... 141 ......B 09Medizinische Abteilung ............................................................................................................................ 124 ......B 09Sicherheitsmanagement ........................................................................................................................... 439 ..... D 08Strahlenpassstelle des KIT Campus Nord .................................................................................................... 436 ..... C 08Werkfeuerwehr .......................................................................................................................................... 315 ..... A 08
Projektträger Karlsruhe (PTKA):
PTKA – Baden-Württemberg Programme (BWP .......................................................................................... 438 ..C/D 08PTKA – Produktion und Fertigungstechnologien (PFT) ................................................................................. 436 ..... C 08PTKA – Standort Dresden ....................................................................................................................................... D 01PTKA – Wassertechnologie und Entsorgung (WTE) ................................................................................... 9675 ..C/D 08
Institute:
Angewandte Informatik (IAI) ...................................................................................................................... 449 ..... C 07Angewandte Materialien – Angewandte Werkstoffphysik (IAM-AWP) ........................................................ 681 C 05/06Angewandte Materialien – Keramische Werkstoffe und Technologien (IAM-KWT) ...................................... 575 ..... A 06Angewandte Materialien – Werkstoff- und Biomechanik (IAM-WBM) ......................................................... 696 ..... C 06Angewandte Materialien – Werkstoffkunde (IAM-WK) ............................................................................... 695 ..... C 06Angewandte Materialien – Energiespeichersysteme (IAM-ESS) ................................................................. 441/1 .. B/C 07Biologische Grenzfl ächen 1 – Biomolekulare Micro- und Nanostrukturen (IBG 1) ........................................ 601 ..... C 07Biologische Grenzfl ächen 2 – Molekulare Biophysik (IBG 2 ) ....................................................................... 352 ..... A 07Biologische Grenzfl ächen 3 – Makromolekulare Materialien (IBG 3) ............................................................ 601 ..... C 07Biologische Grenzfl ächen 4 – Magnetische Resonanz (IBG 4) ..................................................................... 351 ..... A 07Biologische Grenzfl ächen 5 – Biotechnologie und Mikrobielle Genetik (IBG 5 ) ........................................... 601 ..... C 07Beschleunigerphysik und Technologie (IBPT) ............................................................................................... 329 . B 07/08Experimentelle Kernphysik (IEKP) ................................................................................................................ 401 .. B/C 08Funktionelle Grenzfl ächen (IFG) .................................................................................................................. 330 ..... A 08Festkörperphysik (IFP) ................................................................................................................................. 425 ..... C 08Hochleistungsimpuls- und Mikrowellentechnik (IHM) ................................................................................. 421 .. B/C 08Katalyseforschung und -technologie (IKFT) ................................................................................................. 727 ......B 05Kernphysik (IKP) ......................................................................................................................................... 401 .. B/C 08Meteorologie und Klimaforschung - Atmosphärische Spurengase und Fernerkundung (IMK-ASF) ............... 435 ..... D 08Meteorologie und Klimaforschung – Atmosphärische Umweltforschung (IMK-IFU) ................... Campus Alpin ..... D 02Meteorologie und Klimaforschung – Atmosphärische Aerosol Forschung (IMK-AAF) ................................... 326 ......B 08Meteorologie und Klimaforschung – Forschungsbereich Troposphäre (IMK-TRO) ........................................ 435 ..... D 08Mikrostrukturtechnik (IMT) ......................................................................................................................... 301 ......B 08Mikroverfahrenstechnik (IMVT) .................................................................................................................. 605 .. B/C 07Nukleare Entsorgung (INE) ......................................................................................................................... 714 ..... A 05Neutronenphysik und Reaktortechnik (INR) ................................................................................................ 521 ......B 06Nanotechnologie (INT) ............................................................................................................................... 640 .C 06/07Prozessdatenverarbeitung und Elektronik (IPE) ............................................................................................ 242 ..... C 09Photon Science and Synchrotron Radiation (IPS) ......................................................................................... 329 . B 07/08Technikfolgenabschätzung und Systemanalyse (ITAS) ............................................................................... 3300 ......K‘heTechnische Chemie (ITC) ............................................................................................................................ 403 ..... C 08Technische Physik (ITEP) .............................................................................................................................. 418 ..... C 08Tritiumlabor Karlsruhe (ITEP-TLK) ................................................................................................................ 451 ..... C 07Toxikologie und Genetik (ITG) .................................................................................................................... 305 ......B 08Laboratorium für Applikationen der Synchrotronstrahlung (LAS)................................................................. 329 . B 07/08Steinbuch Centre for Computing (SCC) ...................................................................................................... 449 ..... C 07
Helmholtz-Programme:
Atmosphäre und Klima (ATMO) ................................................................................................................. 435 ..... D 08BioInterfaces in Technology and Medicine (BIFTM) ...................................................................................... 440 ..... D 08Energieeffizienz, Materialien und Ressourcen (EMR) ................................................................................... 435 ..... D 08Erneuerbare Energien (EE) ......................................................................................................................... 435 ..... D 08Kernfusion (FUSION) .................................................................................................................................. 451 ..... C 07Materie und Technologie (MT) .................................................................................................................... 242 ..... C 09Materie und Universum (MU) .................................................................................................................... 401 .. B/C 08Nukleare Entsorgung und Sicherheit sowie Strahlenforschung (NUSAFE) .................................................... 433 ..... C 08Science and Technology of Nanosystems (STN) ........................................................................................... 440 ..... D 08Speicher und vernetzte Infrastrukturen (SCI) ............................................................................................... 418 ..... C 08Supercomputing & Big Data (SBD) .............................................................................................................. 449 ..... C 07Technologie, Innovation und Gesellschaft (TIG) ......................................................................................... 3300 ..... K‘heVon Materie zu Materialien und Leben (MML) ........................................................................................... 329 .....B 7/8
Warnhinweise
Interne Notrufnummer 3333
Auf dem Campus gilt die Straßenverkehrsordnung.
Höchstgeschwindigkeit 50 km/h.
Vorfahrtsregel „rechts vor links“ beachten.
Sicherheitsbereiche dürfen nur unter Beachtung örtlicher Regularien betreten werden.
Zutritt ist tageszeitlich begrenzt.
Karlsruher SchlossKarlsruher SchlossKarlsruher Schloss
Karl-Wilhelm-Platz4, 5
Kronenplatz
Marktplatz2, 3, 5, S1, S11, S4, S41, S5
Marktplatz1, 2, 4, 5, S2, S4, S41, S5
1, 2, 4, 5, S2, S4, S41, S5
H
H
Durlacher Tor /KIT Campus-Süd1, 2, 4, 5, S2, S4, S41, S5
01.85
01.90
01.9201.93
01.94
Waldparkplatz
Ernst-Gaber-Str.
Paulcke-Platz
W.-J
orda
n-W
eg
Moltkestraße
Engesserstraße
ZähringerstraßeKreuzstraße
Schlossplatz
Zirkel
Kronenstraße
Waldhornstraße
Fritz-Erler-Straße
Waldhornstraße
Berliner Platz
Kaiserstraße
Kapell
enstr
aße
Durlacher Allee
R.-Baumeister-Platz
Otto-Ammann-PlatzRu
dolf
-Pla
nk-S
traß
e
Wilhelm-Nusselt-WegEhrenhof
Engelbert-Arnold-Straße
Iren
e-Ro
senb
erg-
Stra
ße
Engesserstraße
Englerstraße
Lehm
anns
traß
e Wol
fgan
g-G
aede
-Str
aße
Leonhard-Sohncke-Weg
Neu
er Z
irkel
Frit
z-H
aber
-Weg
Schlossbezirk
Simon-M
oser-Weg
Richard-Willstätter-Allee
Stra
ße a
m F
orum
Engl
er-B
unte
-Rin
g
Ade
naue
rrin
g
Hagsfelder Allee
Am Fasanengarten
Gotthard-Franz-StraßeKornblumenstraße
Fraunhoferstraße
Vincenz-Prießnitz-Straße
Edelsheimstraße
Parkstraße
Schönfeldstraße
Haid-und-Neu-Straße
Karl-Wilhelm-Stra
ße
Lärchenallee
Wes
tein
fahrt
mit
Codekar
te
Haupteinfahrt
Nordeinfahrtmit Codekarte
Einfahrt
Einfahrt
Einf
ahrt
10.50
Studierendenwerk
StudentenzentrumKIT-Shuttle
Administration
Studierendenservice
Steinbuch Centre forComputing (SCC)
BauamtGastdozentenhaus
Forstliches Bildungs-zentrum Karlsruhe
Sportanlagen
10.40
10.30
10.7
0
FORUM
MENSA
Audimax
SPORT
10.31
10.32 10.34
10.33
KIT-BIBLIOTHEK
01.5101.52
30.12 30.11
30.10
20.1120.12
20.13
20.14
01.53
20.20
20.21
20.50 20.51 20.52
02.1
1
02.1
0
20.54
20.5302.02 02.03 02.04
20.30
02.95
30.21
30.22
30.2
3
30.24
30.2
5
20.4011.20
11.1
1
11.10
11.30
11.4
010
.1201
.80
10.11
10.9
1
10.92 10.93 10.94
10.9
5
10.96
10.21
01.95
10.23
11.2
3
11.21
11.2
2
10.81
10.82
10.83 10.84 10.85
10.86 10.87
10.8
8 10.89
10.6
110
.64
10.6310.62
02.17
02.12
02.1
3
02.1
402.1
5
02.16
02.18
30.91
30.9
3
30.90
30.41
30.42
30.43
30.44
30.45
30.31
30.33
30.3
430.35
30.36
30.81 30.8230.51
30.5
0
01.1
3
30.60 30.6130.95
30.9
6
30.70
30.79
30.29
40.12
40.13
40.1
4
40.15
40.16
40.17 40.18
40.1
9
40.0
240
.51
40.0
4
40.21
40.50
40.22 40.2
340
.24
40.2
5
40.2
6
40.27
40.28
40.29
40.3
140
.32
40.33
40.40
40.44
40.41
40.43
50.20
50.2150.22
50.2650.25
50.1050.12 07.21
50.33
50.34
50.35
50.3650.38
50.4
0
50.41
01.70
05.01 05.20
01.96
07.0707.08
07.30
07.31
07.01
Adolf-Würth-Gebäude
30.28
PRÄSIDIUM
11.30
10.9
1
10.9610.96
11.30
10.9
1
10.9610.96
KinderUniversum
Studienberatung
30.46
30.32
01.1
2
11
12
13
14
19
1
2a
2
2
3
5
5a 5b
24
6
13
2
7
7
9
7
1a
51
13
11b
7
5
11a
4
15
42
2-6
68
17
31
8
64
2
7
1 1a
2
3
4
651
7
9
8a4a
8 422a 2b
3
22
1042
1
12
4
6 8 10
12
2
13
14
11
7
2018
5
14
5a
2
8
1
6
4
2
5
5
3
13
2
7
9
109
8
7
65
11
71
3
59b
9a11
a 11c
11a
11b
14122
4
8
15
23
2221
3
1-3
27
3032
34
53
3
6
14
13
31
1a
13
30.4
9
30.4
8
50.3
1 50.32
KIT-Campus Süd
Lieferanschrift:Kaiserstraße 1276131 Karlsruhe
Eggenstein-Leopoldshafen
Karlsruhe A5Durlacher Allee
Moltkestr.
Basel/Stuttgart
Ausfahrt Karlsruhe Durlach
Frankfurt
Karlsruhe
Graben-Neudorf
36
L558
L559
Adenauerring
KIT-Campus Nord
Ostendorfhaus
KIT-Campus West
KIT-Campus Ost
KIT-Campus Süd
KIT-StandorteKarlsruhe
LAGEPLAN CAMPUS NORD
KIT-StandorteDeutschland
Karlsruhe
Dresden
Garmisch
Ulm
Black Forest Observatory
1
2
3
44
5
6
7
8
9
10
A B C D
D
A B C D
1
2
1
2
3
4
5
6
7
8
9
10
3
KIT – Die Forschungsuniversität in der Helmholtz-Gemeinschaft
100 mN
Neubaumaßnahmen
Fremdinstitutionen
Zaunanlage/Arealbegrenzung Fremdinstitution
Figure 13: https://www.kit.edu/downloads/Campus-Nord.pdf,Retrieved on 05.01.2019
08.01.19
Energy Informatics Seminar
Institute for Automation and Applied Informatics (IAI)
Additional time series approaches
Stationarity
“A stationary time series is one whose properties do not depend on the
time at which the series is observed.” (Hyndman and Athanasopoulos
2018)
Implementation with differencing to remove effects
Normalization (Brownlee 2018, p. 249, Chollet 2018, p. 210f)
Min-Max-Scaling
Transform features by scaling each feature to a given range
Implementation with range [0, 1]
Standardization
Standardize features by removing the mean and scaling to unit variance (z-
transformation)
Implementation with mean = 0 and standard deviation = 1
Kai Schmieder: Deep Learning for Energy Time Series25
Foundation DiscussionLong Short-Term Memory for
Energy Time SeriesEvaluation Conclusion
09.01.1908.01.19
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Overview of evaluation scenario characteristics
Kai Schmieder: Deep Learning for Energy Time Series26
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
Characteristic Category
Level of Aggregation 15 minutes 1 hour
Stationarity Stationary Non-Stationary
Normalization None Min-Max-Scaling Standardization
Models Stacked Long Short-Term Memory
Sequence-to-Sequence Long Short-Term Memory
Horizon One step One Day One Week
à 72 scenarios per building
09.01.1908.01.19
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Evaluation metric
Root mean squared error (RMSE)
(Mocanu et al. 2016, p. 95)
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series27
RMSE = 1' ∗ )*
+,-.
/+0-.
12(40,, − 740,,)9
)* à number of steps in horizon' à total number of steps predicted into the future40 à real values for time-step t740 à value predicted by the model for time-step t
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Baseline models
Naïve ForecastValue of the last week on the same day of the week at the same timePrediction for timestep t: (t - 672 / t - 168)
Ordinary RegressionRegression of the respective load time series with the following regressors
Load before one day (t - 96 / t - 24)Load before two days (t - 192 / t - 48)Load before one week (t - 672 / t - 168)Dummy variables for weekend, month, hour and minute (only for 15 minutes)
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series28
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
Legend: (15 minutes / 1 hour)
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Results for Building 124
Kai Schmieder: Deep Learning for Energy Time Series30 08.01.19
Level of Aggre-gation
Sta-tionary
Normal-ization
RMSE per time horizon
Stacked-LSTM S2S-LSTM Naïve ForecastOrdinary
Regression
15 Minutes
No
None (0,015; 0,853; 0,893) (1,474; 1,496; 1,491)
0,909 (0,397; 0,789; 0,825)
Min-Max (0,01; 0,792; 0,865) (1,482; 1,473; 1,491)
Standard (0,011; 0,816; 0,868) (1,475; 1,489; 1,494)
Yes
None (0,007; 1,193; 1,726) (0,427; 1,755; 2,022)
Min-Max (0,016; 1,039; 3,196) (0,427; 1,761; 2,093)
Standard (0,008; 1,177; 1,693) (0,427; 1,755; 2,023)
1 Hour
No
None (0,021; 0,742; 0,871) (1,453; 1,449; 1,452)
0,829 (0,311; 0,724; 0,76)
Min-Max (0,026; 0,712; 0,81) (1,433; 1,447; 1,453)
Standard (0,03; 0,73; 0,864) (1,443; 1,456; 1,452)
Yes
None (0,008; 0,839; 1,291) (0,558; 1,693; 1,964)
Min-Max (0,049; 0,885; 1,899) (0,558; 1,695; 2,015)
Standard (0,009; 0,856; 1,208) (0,558; 1,7; 1,964)
Legend Root Mean Squared Error (One Step; One Day; One Week) | LSTM – Long Short-Term Memory | S2S – Sequence-to-Sequence
Foundation DiscussionLong Short-Term Memory for
Energy Time SeriesEvaluation Conclusion
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Remarks on results
Technical and practical insightsHandling time shift and some missing days in 2012Data pre-processing pipeline
Traditional time series analysis: 2D [samples, features]Deep Learning: 3D tensor [samples, time-steps, features]
Distinction between one-step and multi-step forecastsParameters
Time for training per Long Short-Term Memory modelapprox. 6,5 hours for all scenarios
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series31
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
DISCUSSION
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series32
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Comparable results with Mocanu et al. (2016)
Missing parameter configurations
Critical reflection on Marino et al. (2016)
Training batch and epochsNorm clippingDropoutActivation function at Dense layer…
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series33
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
In Mocanu et al. (2016) seven experiments with different resolutions and time horizons on the same dataset (Dheeru and Taniskidou 2017)Implementation and evaluation of Conditional Restricted Boltzmann Machine and Factored Conditional Restricted Boltzmann MachineComparable result in one of seven scenarios à Results of other scenarios?
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Related work
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series34
DeepArchitecture
Paper Reference
Deep Forward Networks
Con-volutional
Neural Networks
Recurrent Neural
Networks
Deep Generative Networks
Auto-encoders
Amarasinghe et al. 2017 CNN
Gensler et al. 2016 LSTM* AE*
He 2017 CNN* LSTM*
Heghedus et al. 2018 GRU
Jarábek et al. 2017 S2S-LSTM
Li et al. 2017 ELM* SAE*
Marino et al. 2016 Stacked- & S2S-LSTM
Mocanu et al. 2016 RNN CRBM & FCRBM
Ryu et al. 2016 RBM
Voß et al. 2018 WaveNet
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
Legend: * - Combined Approach | CNN – Convolutional Neural Networks | (F)(C)RBM – (Factored) (Conditional) Restricted Boltzmann Machine | ELM – Extreme Learning Machine | GRU – Gated Recurrent Unit | LSTM – Long Short-Term Memory | RNN – Recurrent Neural Network | S2S – Sequence-to-Sequence | (S)AE – (Stacked) Autoencoder
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
CONCLUSION
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series35
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Summary
AchievementsLong Short-Term Memory as one Deep Learning method for Energy Time Series ForecastingOverview of current Deep Architectures for Energy Time SeriesImplementation and evaluation of
Stacked Long Short-Term Memory and Sequence-to-Sequence Long Short-Term Memory
based on Marino et. al. (2016)
Summarised resultsStacked Long Short-Term Memory outperformed baseline models for hourly data in one-step and one-day predictionsGood performance of Sequence-to-Sequence Long Short-Term Memory could not be confirmed (without hyperparameter optimization)
Kai Schmieder: Deep Learning for Energy Time Series36 08.01.19
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Time Series
Deep Learning
Outlook
Hyperparameter optimizationGrid SearchRandom Search Bayesian Optimization, etc.
Variations of Long Short-Term Memory
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series37
Foundation DiscussionLong Short-Term Memory for Energy Time Series Evaluation Conclusion
Verify results with more buildings“Real” univariate Time Series ForecastMultivariate Time Series Forecast
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Questions
08.01.19
?Kai Schmieder: Deep Learning for Energy Time Series38
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Bibliography (1/5)Amarasinghe, K., Marino, D. L., & Manic, M. (2017). Deep neuralnetworks for energy load forecasting. In 26th {IEEE} International
Symposium on Industrial Electronics, {ISIE} 2017, Edinburgh, United
Kingdom, June 19-21, 2017 (pp. 1483–1488). IEEE.Bengio, Y., Frasconi, P., & Simard, P. (1993). The problem of learning long-term dependencies in recurrent networks. In Neural Networks,
1993., IEEE International Conference on (pp. 1183–1188).Bengio, Y., Simard, P., & Frasconi, P. (1994). Learning long-term dependencies with gradient descent is difficult. IEEE Transactions on
Neural Networks, 5(2), 157–166.Brownlee, J. (2018). Deep Learning for Time Series Forecasting. Machine Learning Mastery. Retrieved from https://machinelearningmastery.com/deep-learning-for-time-series-forecasting/Li, C., Ding, Z., Zhao, D., Yi, J., & Zhang, G. (2017). Building energy consumption prediction: An extreme deep learning approach. Energies, 10(10), 1–20.
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series39
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Bibliography (2/5)
Chollet, F. 2018. Deep Learning with Python. Manning. Shelter Island.Dheeru, D., & Karra Taniskidou, E. (2017). “Individual household electric power consumption”. UCI Machine Learning Repository. Retrieved from http://archive.ics.uci.edu/mlGoodfellow, I., Bengio, Y. & Courville, A. (2016). Deep Learning. Cambridge, MA, USA: MIT Press.He, W. (2017). Load Forecasting via Deep Neural Networks. In V. Ahuja, Y. Shi, D. Khazanchi, N. Abidi, Y. Tian, D. Berg, & J. M. Tien (Eds.), Proceedings of the 5th International Conference on Information Technology and Quantitative Management, {ITQM} 2017, December 8-10, 2017, New Delhi, India (Vol. 122, pp. 308–314). Elsevier.Heghedus, C., Chakravorty, A., & Rong, C. (2018). Energy Load Forecasting Using Deep Learning. In 2018 IEEE International Conference on Energy Internet (ICEI) (pp. 146–151).
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series40
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Bibliography (3/5)
Hochreiter, S. (1991). Untersuchungen zu dynamischen neuronalen Netzen. Diploma, Technische Universität München, 91(1).Hochreiter, S. & Schmidhuber, J. (1997). Long Short-Term Memory. Neural Computation, 9(8), 1735–1780.Hyndman, R. J. & Athanasopoulos, G. Forecasting: Principles and Practice. https://otexts.org/fpp2/stationarity.html#fn14, Retrieved on 17.12.18Jarábek, T., Laurinec, P., & Lucká, M. (2017). Energy load forecast using S2S deep neural networks with k-Shape clustering. In 2017 IEEE 14th International Scientific Conference on Informatics (pp. 140–145).Li, C., Ding, Z., Zhao, D., Yi, J., & Zhang, G. (2017). Building energy consumption prediction: An extreme deep learning approach. Energies, 10(10), 1–20.
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series41
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Bibliography (4/5)
Marino, D. L., Amarasinghe, K., & Manic, M. (2016). Building energy load forecasting using Deep Neural Networks. IECON 2016 - 42nd Annual Conference of the IEEE Industrial Electronics Society, 7046–7051.Mocanu, E., Nguyen, P. H., Gibescu, M., & Kling, W. L. (2016). Deep learning for estimating building energy consumption. Sustainable Energy, Grids and Networks, 6, 91–99.Pérez-Lombard, L., Ortiz, J., & Pout, C. (2008). A review on buildings energy consumption information. Energy and Buildings, 40(3), 394–398.Ryu, S., Noh, J., & Kim, H. (2016). Deep neural network based demandside short term load forecasting. 2016 IEEE International Conference on Smart Grid Communications, SmartGridComm 2016, 308–313.Sutskever, I., Vinyals, O., & Le, Q. V. (2014). Sequence to sequence learning with neural networks. Advances in Neural Information Processing Systems (NIPS), 3104–3112.
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series42
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Bibliography (5/5)
Voß, M., Bender-Saebelkampf, C., & Albayrak, S. (2018). Residential Short-Term Load Forecasting Using Convolutional Neural Networks. In IEEE International Conference on Communications, Control, andComputing Technologies for Smart Grids (SmartGridComm).
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series43
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
APPENDIX AND BACKUP
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series44
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
How does deep learning basically work?
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series45
Source: Own representation based on Chollet (2018, p.11)
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Activation functions
Sigmoid (σ)! t = $
$% &'(Value range [0; 1]
Tangens hyperbolicus (tanh)tanh , = &( - &'(
&(% &'(Value range [-1; 1]
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series46
Source: https://upload.wikimedia.org/wikipedia/commons/thumb/8/87/Hyperbolic_Tangent.svg/1920px-Hyperbolic_Tangent.svg.png (Retrieved on 06.01.2019)
Source: https://upload.wikimedia.org/wikipedia/commons/thumb/5/53/Sigmoid-function-2.svg/1920px-Sigmoid-function-2.svg.png (Retrieved on 06.01.2019)
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Details: Stacked Long Short-Term Memory
Objective function
Unrolling implemented with M=50
Optimizer Adam
Training with Backpropagation Through Time
Optimization with Norm clipping
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series47
! = #$%&
'() $ − +)[$])/
Figure 5: Own representation based on Marino et al., 2016, p. 7047
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Details: S2S Long Short-Term Memory
Similar setup as Stacked Long Short-Term Memory
Encoder objective function
Decoder objective function
Optimization with Norm clippingRegularization with Dropout
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series48
Figure 6: Own representation
!" = $%&'
((* % − ,*[%])0
!1 = $%&(2'
3(* % − ,*[%])0
Energy Informatics SeminarInstitute for Automation and Applied Informatics (IAI)
Backpropagation Through Time for S2S LSTM
08.01.19 Kai Schmieder: Deep Learning for Energy Time Series49