Sylvain Frey
1, 2
, François Huguet
, Cédric Mivielle
, David Menga
, Ada Diaconescu
and Isabelle
Infres & SES Departments, CNRS LTCI – Télécom ParisTech, 46 rue Barrault, 75013, Paris, France
ICAME Department, EDF R&D, 1 avenue du Général De Gaulle, 92141, Clamart, France
Keywords: Computing, Autonomic, Scenario, Vision, Decentralised Systems, Micro Smart Grids.
Abstract: Autonomic computing is a bio-inspired vision elaborated to manage the increasing complexity of
contemporary heterogeneous, large scale, dynamic computer systems. This paper presents a series of
scenarios relative to micro smart grids – district-size “smart” electricity networks. These scenarios involve
situations where autonomic management approaches could provide promising solutions. They therefore
appear as short stories of a possible autonomic micro smart grid, that illustrate the concepts of autonomic
computing as well as the potential behind this vision. At the same time, these scenarios reveal open issues as
well as novel perspectives on the future of micro smart grids.
Autonomic computing is a vision that proposes
conceptual and practical solutions to complexity
management in computer systems (Horn, 2001). An
autonomic computer system is defined by its ability
to manage some of its own properties without the
help of a human administrator. Namely, an
autonomic system is capable of self-configuring,
self-protecting, self-healing and self-optimising
(Kephart & Chess, 2003), relieving users from
difficult management tasks.
So far, realisations of autonomic computing remain
restricted to the management of simple, isolated
systems: the domain still needs major contributions
in order to create large-scale, heterogeneous,
dynamic, open autonomic systems (Dobson et al.,
2010). In particular, “bottom up” engineering
methods propose to build such complex systems out
of integration of simpler ones (Ulieru & Doursat,
2011). However, unpredictable and possibly
ambiguous interactions between autonomic
sub-systems make it difficult to comprehend the
functioning of autonomic systems-of-systems.
The goal of this paper is to elaborate realistic
scenarios for illustrating autonomic approaches,
understanding the issues behind autonomic
management and paving the way to an engineering
process for complex autonomic systems (Frey et al.,
2012). The domain of application chosen here is that
of micro smart grids. After a short presentation of
management requirements for micro smart grids, a
series of scenarios will be exposed, that show the
capabilities of autonomic systems and allow the
analysis of required features for such possible
autonomic solutions.
The constant increase of energy consumption makes
electricity grid management expensive and complex.
Differences between consumption troughs and peaks
(e.g. during winter nights) lead national providers to
over-dimension means of production in order to face
ever-increasing but still rare maximal demand.
Meanwhile, production means and energy storages
(solar panels, windmills, batteries, etc.) start being
available for private use. Private users were so far
exclusively consumers, now they may get the
capability to participate to active load management.
This situation raises the issue of integrating millions
of production means and storage capacities under
private control and therefore less controllable than
traditional means. Indeed, for reasons of privacy and
Frey S., Huguet F., Mivielle C., Menga D., Diaconescu A. and Demeure I..
DOI: 10.5220/0003952001370140
In Proceedings of the 1st International Conference on Smart Grids and Green IT Systems (SMARTGREENS-2012), pages 137-140
ISBN: 978-989-8565-09-9
2012 SCITEPRESS (Science and Technology Publications, Lda.)
private property, direct interventions on equipments
may often be impossible.
In order to cope with the challenges future electricity
grids face, the smart grid vision proposes to couple
the traditional electricity network with an
infor-mation network allowing a “smart”
management of the electrical load (Schiller &
Fassmann, 2010).
In a micro smart grid, each home participates to the
network by selling its production or adapting its
consumption according to the load. The system that
manages the district grid gives live information on
the load that determines current purchase and sale
prices, according to customer contracts. Domestic
productions are connected to consumptions,
allowing a fine management of supply and demand,
practically use by use. Thus, grid management is not
only about limiting consumptions and maximising
productions, but also scheduling these streams
throughout the day, while taking usages into
The vision of micro smart grids requires
sophisticated management capabilities, in particular
with respect to systems integration. Therefore this is
a promising use case for autonomic computing to
illustrate its capabilities while addressing concrete
issues of the domain. The rest of this paper presents
scenarios that detail how autonomic approaches
could be a solution to micro smart grid management,
with analysis of the technical requirements possible
autonomic solutions should fulfil.
Taking advantage of renovation work, Mark has
invested in an autonomic underfloor heating. Thanks
to a set of sensors and actuators, the system can
monitor temperature in the room and receive
instructions. A hot water tank allows accumulating
and returning heat; this is done according to
electricity rates, to Mark's orders, to the weather, to
the actual temperatures in the rooms and to whether
or not someone is present in the apartment. Mark
has also activated a geo-location functionality:
aware of Mark's position, the system can reduce
temperature while he is out and ensures everything
returns to normal when he comes back.
Mark may monitor his system live state: electrical
consumption, desired temperatures compared to
actual temperatures, money saved thanks to the
water tank stock. The initial investment was
substantial but Mark sees a difference on his bill and
can be confident he will get return on investment. In
addition to ecological considerations on energy
consumption reduction, the regulator adapts to the
occupants' desires via simple instructions. Reading
his system history, Mark has realised that some
occasional adjustments he did by hand – for
instance, in his bathroom – have been learnt by the
autonomic system that anticipates his morning
shower since then.
Mark's inquiring mind has led him to play with his
regulator's most advanced functionalities. He has
been offered a choice of management profiles,
favouring comfort, energy savings or bill reduction.
After some experimentations, Mark eventually
decided to follow an highly economical policy that
reduced again his electricity bill. And when guests
are expected, the system can automatically restore
more suitable comfort conditions.
Analysis: This scenario raises the following issues:
conflicts between objectives and an
unpredictable context;
necessity to anticipate the context, in
particular, to develop learning features.
The equipment this scenario describes follows two
contradictory objectives: “maintain temperature”
versus “limit electrical consumption”, in a dynamic,
unpredictable context. This context must be
monitored by the system via temperature sensors,
measurements of the electricity consumption, grid
load sensors, hot water stock probes, possibly
presence sensors and geo-location system. Since
measurements are imprecise by nature, the system
must take possible errors (unidentified moves in a
room, GPS failure, etc.) into account.
In addition to adapting to the context (temperature,
domestic load) the autonomic system must be able to
forecast it in order to achieve efficient management
of its stock. In the scenario, the system must
anticipate Mark's return home after work; it reloads
its stock during the day, according to a consumption
forecast for the night. This forecast may imply mid
to long-term learning capabilities, in order to
evaluate a standard night profile, possible variations
of Mark's consumption, etc.
Unexpected events (unforeseen load peak, unusual
order from the user, heat wave, cold wave, etc.) can
challenge previous forecasts and threaten the
system's functioning. However, it must be able to
anticipate such variations and allow room for
manoeuvre in order to face the unexpected with
appropriate reactions. Sophisticated objectives can
help the system solve the “margin of error versus
optimisation” dilemma.
Mark keeps testing new equipments in his flat and
installed a sophisticated skylight featuring a
controllable power-driven actuator and an
autonomic controller. The latter fetches data
produced by temperature and presence sensors in
the room, reads objectives received by the heating
system and follows local weather forecasts online.
Aside from manual remote control, the controller
proposes several programs for airing the room while
maintaining its temperature.
After making sure that the skylight's functioning
does not create security problems, Mark gives it
order to open for 10 minutes a day, provided nobody
is in the room and external conditions (temperature,
rain, wind) do not threaten the room's temperature
objective. In case unfavourable conditions persist,
the skylight does not open and reports the event.
From studying its systems' journals, Mark has
noticed that far from disrupting energy saving
objectives, the skylights participates to temperature
regulation: when outside temperature condition are
favourable, the skylight opening eases the heating
system's task.
Analysis: The scenario raises the following issues:
integration of different equipments, on a
small scale;
collaboration between equipments in
order to reach a common objective
(here, temperature).
In order to simplify the scenario, we assume that the
skylight's electrical consumption is negligible. In a
simple case, integration between the heating system
and the skylight may be minimal: the two
equipments have no direct relation with each other,
they simply interact through the physical world by
influencing temperature in the room. This kind of
integration, however limited, may lead to a valid
system (Frey et al., 2010).
In a more advanced case, the equipments share
predictions on their behaviour and take each other
into account in their management logic. For instance,
on a cold day, the skylight warns the heating system
that a temperature drop may be necessary during the
day, due to air renewal. The heating system answers
with possible moments of the day it estimates it
could face a temperature drop better (because its
stock would be full, the grid's load should be low,
Mark would be away). The two equipments
eventually agree on the best solution and both reach
their objectives with minimal disruption of each
other ones'.
This “social ability” of autonomic equipments
influences the way they are conceived from the very
beginning, since they must be able to interact and
cope with other systems that are potentially
unpredictable and sources of errors. Thus, a properly
designed autonomic systems should be able to run
alone as well as to integrate a wider system, without
having to undergo fundamental changes in its
internal functioning. Therefore this “situatedness” of
autonomic systems must be a key concern at the
design phase.
Mark's neighbour, Sal, just started her washing
machine. Her flat not being equipped with means of
production nor stocking, it is part of the consuming
homes in the district.
Mark's flat is now equipped with an autonomic
battery, in addition to existing equipments, therefore
it is capable of stocking and reselling electricity.
Detecting that the district grid load increases, the
system performs an internal analysis, in order to
decide what to do next. It gathers the following
analysis: the battery is fully loaded, the flat's room
has been aired already, no particular consumption
is planned for the time being and a priori, the hot
water tank should be able to face any thermal
hazard by itself.
Thus, Mark's system decides to dedicate part of the
battery's stock to external demand. The buyback
rates are favourable to Mark; on her side, Sal
benefits from cheaper local electricity.
Later on, Mark's unexpected return increases his
home's consumption. Starting an intense housework
session usually happening on week-ends, Mark has
opened the windows wide. The heating system,
unprepared, is compelled to spend its stock and even
get power from the battery in order to restore a
comfortable temperature.
Mark's system not being able to provide the district
grid without threatening its own interests (on both
thermal and economical aspects) it ceases being a
producer for its neighbours. Sal's home has lost a
potential provider.
Analysis: The district grid is a typical example of
distributed organisations bringing together
concurrent entities with their own objectives. Each
home tries to minimise its electricity bill and
maximise its profits, although some less egoistic
criteria may be involved as well. The district grid
has no direct control on homes and their equipments;
however it can encourage collaboration via
advantageous fares. Buying and selling is decided at
the home level, according to specific internal
constraints, as shown in the scenarios.
The scenario introduces several issues:
scaling up;
global vs. local management;
heterogeneity of the domains inter-acting.
The first of these difficulties excludes completely
centralised solutions that would intend managing
hundreds of homes and all their appliances with a
single controller – which is, at best, an extremely
difficult task. Furthermore, due to privacy respect, it
is not conceivable that users entrust an external
entity with managing their everyday appliances.
Global vs. local management is related to balancing
interests of the different parties involved (homes,
district and national grids). Variations of production
means, average offer, average demand and energy
costs are inevitable. Therefore the district grid has to
feature a management system that equilibrates
relations between the parties on the long run, via
regulation means – energy fares and contracts,
production means, storage. The policies driving this
regulation depend on the rules (laws, institutions)
that control the district grid.
Finally, heterogeneity of the domains interacting
appears through the succession of perturbations at
the end of the scenario: Mark unexpectedly returning
home, disrupting the flat's temperature, hence local
consumption increasing, hence district production
decreasing, hence possibly Sal's system seeing
energy prices rising. This chain of consequences
illustrates unpredictable behaviours happening in
heterogeneous distributed systems, and again pleads
for decentralising decisions, in association with
separation of concerns (usages management,
temperature management, home energy
manage-ment, district grid management).
This paper proposed a series of autonomic micro
smart grid scenarios, illustrating the potential of
autonomic systems and analysing requirements for
possible autonomic solutions to micro smart grid
management issues. Technical contributions, namely
integration design patterns for autonomic
management systems, have been published based on
this work (Frey et al., 2012). Yet the authors hope
that designing and exposing scenarios contributes to
the understanding of micro smart grids and
autonomic systems, unveiling conceptual and
technical challenges that will be encountered by
future engineers and researchers.
Becker, B., Allerding, F., Reiner, U., Kahl, M., Richter,
U., Pathmaperuma, D., Schmeck, H., Leibfried, T.,
2010. Decentralized Energy-Management to Control
Smart-Home Architectures. Lecture Notes in
Computer Science, 2010, Volume 5974/2010.
Dobson, S.; Sterritt, R.; Nixon, P.; Hinchey, M.; 2010.
Fulfilling the Vision of Autonomic Computing.
Computer, vol.43, no.1, pp.35-41, Jan. 2010.
Frey, S., Diaconescu, A., Lalanda, P., 2010. A
Decentralised Architecture for Multi-Objective
Autonomic Management. In Proceedings of the 2010
Fourth IEEE International Conference on
Self-Adaptive and Self-Organizing Systems (SASO
Frey, S., Diaconescu, A., Demeure, I., 2012. Architectural
Integration Patterns for Autonomic Management
Systems. In Proceedings of the 9th IEEE International
Conference and Workshops on the Engineering of
Autonomic and Autonomous Systems (EASe '12).
Horn, P., 2001. Autonomic Computing: IBM's perspective
on the State of Information Technology. IBM.
Kephart, J.O., Chess, D.M., 2003. The Vision of
Autonomic Computing, Computer 36,1 (Jan. 2003).
Schiller, C.A., Fassmann, S., 2010. The Smart Micro Grid:
IT Challenges for Energy Distribution Grid Operators.
White paper, IBM.
Ulieru, M., Doursat, R., 2011. Emergent engineering: a
radical paradigm shift. International Journal of
Autonomous and Adaptive Communications Systems
(IJAACS), 4(1): 39-60. 22 pages.