The document describes the Energy Management Adviser (EMA) developed by Electricité De France to provide electricity consumption analysis and energy management advice to customers. The EMA was initially developed to include personalized advice but has expanded to include consumption analysis and alerts. It is modeled using ontologies developed in collaboration with Oxford University and reasons over customer data using the HermiT reasoner. Ongoing work includes developing modular tips and capturing customer questionnaires and preferences in a triple store for additional personalization and comparisons.
1. The Energy Management
Adviser at Electricité De
France
Pierre Chaussecourte, Birte Glimm, Ian Horrocks,
Boris Motik and Laurent Pierre
vendredi 25 octobre 13
2. •
•
•
•
One of the world’s leading electric utilities
72.7 billion sales 2012
39.3 million customers wordwide
159 740 employees
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
3. •
•
•
•
One of the world’s leading CS departments
Leading role in OWL development
OWL reasoners inc. HermiT, ELK & RDFox
Collaborating with EDF since 2010
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
4. Summary
Presentation of the EMA
Modelling the EMA
Tools
Ongoing Developments
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
5. Presentation of the
EMA
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
6. • Sending a “bare” invoice to our
customers is not possible anymore
• We must include an analysis of electricity
consumption and advice on energy
management
• We must also be able to send an alert
about electric consumption over a
period of time
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
11. The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
12. Vous pouvez constater que votre consommation d’électricité a été moins élevée en
moyenne cette année que l’année dernière. Pourtant la température a été en moyenne
moins élevée (- 1.6°C par rapport à l’année dernière), ce qui a pu augmenter votre
consommation de chauffage.
La raison de cette diminution est donc certainement due à une utilisation moins fréquente et
plus économe de vos équipements électriques que l’année dernière.
You can notice that your electricity consumption has been less high on average this year
than last year. However, the temperature has been less high on average (- 1.6°C compared
to last year), that could have increased your electricity consumption dedicated to heating.
This decrease is therefore certainly due to a less frequent and and more energy-efficient use
of your electric appliances than last year.
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
13. Variations
You can notice that your electricity consumption has been higher on average this year than
last year. Considering that the temperature has been less high on average (- 1.6°C
compared to last year), this could have increased your electricity consumption dedicated to
heating.
You can notice that your electricity consumption has been higher on average this year than
last year. This increase is certainly due to a more frequent and less energy-efficient use of
your electric appliances than last year.
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
14. Modelling the EMA
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
15. • We started by giving personalised advice
to our customers
• We added electricity consumption
analysis later
• Finally, we redesigned everything in
collaboration with Oxford
• Now EMA serves 300,000 customers
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
16. Energy Management Adviser
Weather Report Service
Electric Heating
ColdWeather
OverConsuming
Owner
...
Yo can notice that ....
Forcasted Electricity
Consumption
Electricity Consumption Data
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
24. Lower
∃ isTipOf.S ⊑ C1
lower
hasTemperatureWinter
True
a_house
t
C1
S
electricHeating
livesIn
a_client
hasTip
a_tip
Higher
isTipOf
higher
hasConsumptionCompareWithPrevious
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
25. Lower
∃ isTipOf.S ⊑ C1
lower
∃ isTipOf.S ⊑ C2
hasTemperatureWinter
True
a_house
t
C1
S
electricHeating
livesIn
a_client
hasTip
a_tip
Higher
isTipOf
higher
hasConsumptionCompareWithPrevious
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
C2
26. SELECT ?t
WHERE {
a_tip a ?t .
?t rdf:subClassOf ActualTip.
}
Lower
lower
∃ isTipOf.S ⊑ C1
∃ isTipOf.S ⊑ C2
hasTemperatureWinter
True
a_house
t
C1
S
electricHeating
livesIn
a_client
hasTip
a_tip
Higher
isTipOf
higher
hasConsumptionCompareWithPrevious
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
C2
27. Reasoning & Formulae
•
How to compute a formula in OWL ?
-
•
computation in advance
annotations on properties
What if a formula depends on inferred facts ?
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
28. Special Situations
•
Not all the tips are valid for every customers
e.g. : Do not tell elderly customers to reduce their home
temperature
•
Define a new ontology vs Complicate the
existent one
•
A SPARQL query used to subtract inaplicable
tips
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
30. ∃ hasAge.[integer >= 75]
a_non_tip
S’
notIsTipOf
T1
a_client
a_tip
isTipOf
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
31. ∃ hasAge.[integer >= 75]
nT1
∃ notIsTipOf.S’ ⊑ nT1
a_non_tip
S’
notIsTipOf
T1
a_client
a_tip
isTipOf
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
32. ∃ hasAge.[integer >= 75]
nT1
∃ notIsTipOf.S’ ⊑ nT1
a_non_tip
S’
notIsTipOf
T1
a_client
a_tip
SELECT ?t
WHERE {
a_tip a ?t .
?t rdf:subClassOf ActualTip.
MINUS {
a_non_tip a ?t .
?t rdf:subClassOf ActualTip.}
}
isTipOf
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
33. Avoiding Nominals
•
As soon as assertions describing a customer change,
the ontology hierarchy has to be recomputed
S ⊑ ∃ hasTip.{c1}
•
∃ electricConsumption.{higher}
We use ‘simulated nominal class’ instead
∃ isTipOf.S ⊑ C1
∃ electricConsumption.Higher
higher : Higher
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
34. Tools : Hermit
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
35. • HermiT is an open source OWL reasoner
• Implemented in Java
• We experimented with new solutions in
order to speed up processing of
customers
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
36. • Incremental changes in the ABox
(no nominals in the TBox)
• Batch of customers (4 is the best)
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
38. Modular Tips
•
Tips are represented as single class and are
associated with a customer situation
•
Combinatorial explosion of tips/customer situation
classes
•
Most of the tips share the same parts of text
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
39. «Your electric consumption has been higher but this winter has been colder than usual and this
could have increase your consumption»
higherElectricConsumption
explains
colderWinter
associatedTipFragment
«Your electric consumption has been higher»
associatedTipFragment
«this winter has been colder than usual»
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
40. «Your electric consumption has been higher but this winter has been colder than usual and this
could have increase your consumption»
higherElectricConsumption
explains
colderWinter
associatedTipFragment
«Your electric consumption has been higher»
associatedTipFragment
«this winter has been colder than usual»
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
44. Questionnaire
3
increase of heated surface
100
m2
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
45. Questionnaire
3
increase of heated surface
100
m2
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
46. Questionnaire
3
a new freezer
increase of heated surface
100
m2
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
47. Questionnaire
3
a new freezer
increase of heated surface
100
m2
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
48. Questionnaire
3
a new freezer
increase of heated surface
100
m2
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
49. Questionnaire
3
a new freezer
increase of heated surface
100
m2
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
50. Triple Store
•
•
•
•
We have started developments with RDFOx
Customers information can be historicized
Comparison between customers is available
Using rules allows us to capture formulae in a
declarative way
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13
51. Thank you
The Energy Management Adviser at Electricité De France - ISWC 2013
vendredi 25 octobre 13