These are the slides for my keynote lecture "AI Techniques for Smart Grids" at the 2014 IEEE Innovative Smart Grid Technologies - Asia conference where I discussed the role and potential of self-organization in the smart grid.

1. AI Techniques for Smart Grids
Networked and Embedded Systems
Wilfried Elmenreich | 2014-05-22
Keynote lecture, ISGT-ASIA 2014

2. Introduction
• Many AI techniques are already in use
– Artificial neural networks (Modeling)
– Fuzzy logic (Control)
– Evolutionary algorithms,
– Swarm algorithms (Optimization)
• Now we go for the real thing
– should we change the way the system is
controlled?
Must?

3. Building Self-Organizing Systems 3
Wilfried Elmenreich
Self-Organzing
Systems

4. What is a Self-Organizing System
„A self-organizing system (SOS) is a set of
entities that obtains global system behavior via
local interactions without centralized control.“

5. Adaptation Robustness
Scalability
from C. Bettstetter, „Lakeside Labs“
Self-Organizing Systems are Effective!
Image: Imgur.com

6. Adaptation Robustness
Scalability
from C. Bettstetter, „Lakeside Labs“
Self-Organizing Systems are Effective!
Image: Imgur.com

7. Characteristics
• System of many interconnected parts
• Degree of difficulty in predicting the system behavior
• Emergent properties
• Dynamic
• Decentralized control
• Global behavior from local interactions
• Robustness, adaptivity
• Non-linearity (small causes might have large effects)
7
Wilfried Elmenreich

8. SOS and Smart Grid
• Why a self-organizing approach?
8
Wilfried Elmenreich
Why Self-
Organzation?
Image: Creative Commons, Wikipedia
Figure: Creative Commons, Wikipedia

9. Transferring control to the network
• Counter-arguments
– Giving up control makes the system instable,
– untrustable,
– harder to maintain…
• Pro arguments
– Stability for complex system can
be only achieved by control
approach at same complexitiy level
– Self-organizing systems are more robust…
– and provide inherent scalability
• Sometimes you do not have this choice!
9
Wilfried Elmenreich
Image: Creative Commons, Wikipedia

10. Example: Wide Area Synchronous Grids
(Interconnections)
10
Wilfried Elmenreich
Figure: Creative Commons, transmission data based on European Joint
Research Center/Institute for Energy and Transport
• Operate at synchronized
frequency
• UCTE grid (Continental Europe)
is largest synchronous grid in
the world in terms of
generation capacity (667 GW)
• Unbundling process of
power generation and
Transmission System Operators
(TSO)  many players

11. Oscillations in wide area grids
On Saturday, 19 February 2011 around 8:00 in the morning, inter-area
oscillations within the Continental Europe power system occurred. The
highest impact of these 0.25 Hz oscillations was observed in the
middle-south part of the system with amplitudes of +/- 100 mHz in
southern Italy and related power oscillation on several north-south
corridor lines of up to +/- 150 MW and with resulting voltage
oscillation on the 400 kV system of +/- 5 kV respectively.
ENTSO-E, ANALYSIS OF CE INTER-AREA OSCILLATIONS OF 19 AND 24 FEBRUARY 2011, 2011
Almost the same event reappeared on 24 February 2011 during
midnight hours
11
Wilfried Elmenreich

12. System frequency oscillations
12
Wilfried Elmenreich
• Superposition of 0.18 Hz (East-West Mode) and 0.25 Hz
(North-South Mode) modes
• Frequency and damping continously oscillates
Figure: ENTSO-E, ANALYSIS OF CE INTER-AREA OSCILLATIONS OF 19 AND 24 FEBRUARY 2011, 2011

13. Investigation of the oscillation events
• Transmission system operators (TSOs) Amprion, Mavir, TenneT
DE, Swissgrid,... exchanged power recordings
• Event was not predictable, no single cause
• Oscillations started around the change of the hour
– Turkey had changed mode displacement
• Total system load was low
• Absence of industrial load
• Dispersed generation (PV, Wind) provides less stabilized
inertia than classical generators
• Italian system currently more sensitive to oscillation modes
– Power system stabilisers in Italy had been reinforced
13
Wilfried Elmenreich

14. Observations from this example
• Liberalization of power market has decreased the scope of
control
• New approach is to carefully and knowledgeable interact with
the system in order to guide it
• We can can observe the main properties of a SOS here
• Understanding this system in a new way became a necessity
14
Wilfried Elmenreich

15. 15
Wilfried Elmenreich
Another Example
Image: Creative Commons, Wikipedia

16. Smart Meter Rollout
• Energy Services Directive
(2006/32/EC) and the electricity
directive (2009/72/EC) require
the implementation of
"intelligent metering systems".
• Such systems ought to be in
place for 80% of electricity
consumers by end 2020
16
Wilfried Elmenreich
Source: The Smart Grid in Europe, 2012-2016: Technologies, Market
Forecasts and Utility Profiles (GTM Research), August 2011

17. The Smart Grid, as the Providers Envision it
• Smart meters
– Read meters remotely (save money for data acquisition)
– Get metering data at a high resolution
• Controllability of the loads
– Send „off“ signals to customer appliances at peak load situations
– Cut off a customer that does not pay the bill
• Having a system supporting different types of energy sources
and storage
in overall: get more comprehensive information and control
over the system
17
Wilfried Elmenreich

18. The Smart Grid, as the Customers want it
• Magically save energy / reduce bill
• Connect own generators (plug-in PV system)
• Get more reliable energy service
• Get green energy
• Don‘t give up privacy or control
in overall: only positive things should arise,
nothing must get worse
18
Wilfried Elmenreich

19. How Self-Organization can help
• Handling complexity: Provides scalable approaches for a high
number of interacting components
 providers will like that
• „Bossless structure“: Allow bottom-up processes, keep
responsibility and decisions at customer („I can decide“)
 customers will like that
Building Self-Organizing Systems 19
Wilfried Elmenreich

20. What holds us?
• Reluctance to give up (central) control
• Hard to understand – hard to trust
– Many proponents miss a Non-linear thinking (© Alessandro
Vespignani), a.k.a. complex system goggles
• How can be design self-organizing systems?
This is our quest:
• Provide models, proofs, case studies, etc. showing that self-
organizing approaches work
– Sufficiently large, realistic case studies
Building Self-Organizing Systems 20
Wilfried Elmenreich

21. Building a self-
organizing system
Image: Creative Commons, Wikipe

22. Rules of an SOS may be simple…
• ..but finding the right rules is difficult!
• Complex systems are hard
to predict
• Counter-intuitive
dependencies
22
Image: USGOV-NOAA (Public Domain)
Wilfried Elmenreich – Building Self-Organizing Systems

23. Evolutionary Design Approach
Building Self-Organizing Systems 23
Wilfried Elmenreich
• Evolution applied during design phase
• We don‘t refer to evolution/development of a system
at run time

24. Search Algorithm
Building Self-Organizing Systems 24
Wilfried Elmenreich
• Figuratively and literally a zoo on metaheuristic
optimization algorithms
• Ability to find global optimum
• Number of tweaking parameters?

25. FREVO: A Software for Designing SOS
• FREVO (Framework for Evolutionary Design)
• Operates on a simulation of the problem
• Interface for sensor/actuator connections to the agents
• Feedback from a simulation run -> fitness value
• Open-source, system-independent http://frevo.sourceforge.net

26. System architecture
Building Self-Organizing Systems 26
Wilfried Elmenreich
6 major components:
task description, simulation setup, interaction
interface, evolvable decision unit, objective function,
search algorithm

27. Application
examples
Image: Creative Commons, Wikipedia

28. Application example: Trader (1)
• Evolving an energy trader algorithm at consumer/prosumer
level
• Simulation
• Java module added to FREVO
• Market rules
• Simulated Market
• Agent
• No initial knowledge
about market rules
• Trader rules are learned implicitly
• This way also counter-intuitive strategies are considered

29. Application example: Trader (2)
• Tradeoff between performance, complexity and
comprehensibility
 There is no free lunch!
Performance of
evolved market agents

30. WiP: Evolving system of device-level traders
• Model HEMS devices as agents with independent controllers
• Constraints are given by a budget per device and the importance of a device
for the user

31. Summary
• AI techniques can be used as a tool but as well
contribute to a change in system design
• Self-organizing systems are promising for handling
complex systems
• Design challenge
– Evolutionary approach in combination with modelling
techniques
• Validation challenge
– Verification techniques, simulation
– Need for more case studies
31
Wilfried Elmenreich

32. Thank you very much for your attention! 
Building Self-Organizing Systems 32
Wilfried Elmenreich
Thank you very much
for your attention!
Image: Creative Commons, Wikipedia