Content uploaded by Markus Tauber
Author content
All content in this area was uploaded by Markus Tauber on Aug 17, 2014
Content may be subject to copyright.
A Self-Organising Approach for Smart Meter
Communication Systems
Markus Tauber1, Florian Skopik1, Thomas Bleier1and David Hutchison2
1AIT, Austrian Institute of Technology
{markus.tauber,florian.skopik,thomas.bleier}@ait.ac.at
2Lancaster University,
d.hutchison@lancaster.ac.uk
Abstract. Future energy grids will need to cope with a multitude of
new, dynamic situations. Having sufficient information about energy us-
age patterns is of paramount importance for the grid to react to changing
situations and to make the grid ‘smart’. We present preliminary results
from an investigation on whether autonomic adaptation of intervals with
which individual smart meters report their meter readings can be more
effective than commonly used static configurations. A small reporting
interval provides close to real-time knowledge about load changes and
thus gives the opportunity to balance the energy demand amongst con-
sumers rather than ‘burning’ surplus capacities. On the other hand, a
small interval results in a waste of processing power and bandwidth in
case of customers that have rather static energy usage behaviour. Hence,
an ideal interval cannot be predicted apriori, but needs to be adapted
dynamically. We provide an analytical investigation of the effects of auto-
nomic management of smart meter reading intervals, and we make some
recommendations on how this scheme can be implemented.
1 Introduction
Emerging alternative forms of energy are increasingly allowing consumers to
produce electricity and to feed surplus capacities back into the power grid. This
will turn them from consumers into producers. Additionally, power grids will
become more important for mobility as electric cars will be connected to the grid
for charging their batteries. These aspects will turn traditional customers into
prosumers. If, for instance, a customer, produces surplus energy via solar panels
during the day, (s)he would be in the producer role. When however charging her
or his car over night, (s)he would become a (heavy) consumer. Such emerging
scenarios will contribute to a highly dynamic overall power grid usage which
requires the traditional power grid to become smarter by adding more control
–theSmart Grid. A smart grid utility provider needs to be able to detect
over-usage or under-provision in (real-)time to manage demand by, for instance,
2c
2014 Springer, Published at IWSOS 2013, LNCS 8221, pp. 169–175
2 Markus Tauber, Florian Skopik, Thomas Bleier and David Hutchison
scheduling the charging time of consumers’ e-cars. This would avoid ‘burning
off’3surplus capacities and hence increase sustainability of energy usage.
Motivation. Fine-grained information about energy consumption patterns
is of paramount importance for grid providers to react to a changing environment
and to maintain high sustainability by, for instance, managing demands flexibly.
Today’s smart meters send consumption values to the grid provider at constant
intervals [5, 7]. A small interval is beneficial for energy sustainability and a power
grid’s efficiency, as it allows fine-grained demand management [1]. This can also
be directly beneficial for the customer if the provider’s ability to control charging
periods of her/his heavy usage appliances (e.g. an electric car battery charger),
is incentivised by reduced energy prices for the consumer. However, if consumers
without heavy usage appliances (i.e. those who do not exhibit a high degree of
variability in energy consumption), frequently report their energy usage, they
reduce the benefit of small intervals as unnecessarily gathered monitoring data
increases the overall operational load on the grid’s ICT infrastructure. Uniformly
applied intervals may not be sufficient as energy usage patterns and energy
requirements vary from household to household, and over time.
Secondary Effects: Sustainability vs. Privacy Trade-off. Furthermore,
a small interval threatens individual customer’s privacy – especially if it allows
the derivation of behaviour patterns [5] from energy consumption readings. Thus,
depending on the situation, a small interval may require privacy to be traded
off against sustainable energy usage. We are able to provide some information
to prepare further investigations even though this concern is outside the scope
of this paper.
Research Contributions. As outlined above, different situations can be
identified in which various grid aspects depend on the flexible usage patterns de-
termined by the user base. We pick the reporting interval as first representative
example to investigate how to apply more self-* properties to the smart-grid.
For the considered case we identify situations where small reporting intervals
are beneficial with respect to sustainability whereas large intervals can be dis-
advantageous. The contrary view applies with respect to efficiency in processing
the reported data. Thus an ideal interval cannot be predicted apriorias it de-
pends on the variability of an individual’s energy usage over time. Autonomic
management [2] is a (policy-driven) approach to adapt the metering report in-
terval for individual smart meters in response to different usage patterns and
requirements, in order to improve the grid’s overall efficiency, in contrast to
contemporary statically configured systems. Eventually, this will turn the smart
grid into a self-organising system. As a direct effect, complexity is moved from
a central point of computation (at the utility provider) to the individual smart
meters, which increases efficiency in data processing and makes the system as
a whole more scalable, more resilient and has a positive impact on the above
mentioned secondary effects.
3Where the term ‘burning off’ energy in this case refers to the losses due to reduced
degree of efficiency when employing battery buffers or pumped storage hydro power
stations. See: Energy storage - Packing some power. The Economist. 2011-03-03
A Self-Organising Approach for Smart Meter Communication Systems 3
Structure of this paper. We provide some information on background and
related work in Section 2, followed by an analysis of the problem and our ap-
proach towards autonomic management in Section 3. We report on a preliminary
evaluation of our approach in Section 4, and provide some concluding remarks
and outlook on future work in Section 5.
2 Background and Related Work
Smart Grids in general represent a popular research topic; however, neither our
primary focus, which is to improve efficiency by autonomic interval adaptation
nor our secondary topic of interest, which is the (autonomic) management of the
trade-off between sustainability and privacy via reporting interval adaptation, is
well investigated. For instance, current state of the art regarding meter reading
reporting involves a static configuration and only mentions different statically
configured intervals [5], but no autonomic adaptation of those is being discussed.
[5] mentions that Canada supports meter readings at 5 to 60 minute intervals
and that the next generation of smart meters will reduce these time intervals
to one minute or less. [7]. Other existing work [9] has been conducted to apply
autonomic management to multiple domains. This also includes the adaptation
of house keeping intervals in order to improve routing overlays efficiency. Despite
being in a different field it shares some similarities to the approach we propose.
With respect to secondary effects of our approach, threats to and vulnera-
bilities of smart metering systems are widely discussed topics [4, 3, 10]. While
communication security [3] is widely studied, the aspects of privacy and poten-
tial threats [5] to customers through smart meter data exploitation are not fully
covered up till now [8]. An important first step towards a privacy-enabled smart
grid has been made by NIST [6], when defining problems related to privacy
protection and legal constraints.
3 Problem Analysis and Approach
3.1 Scenario
Based on the demand management use case introduced in Section 1, we model
the ideal frequency with which meter readings are being reported in terms of the
dependency of the fluctuation of power usage over time, (i.e. in energy usage).
If the power consumption and its fluctuation over time are high it is beneficial
for the provider if meter readings are sent at small intervals. This is, however,
only applicable if power usage of the individual customer exhibits some degree
of fluctuation and intensity. Autonomic management is an approach to control
systems in the presence of changing situations and requirements.
3.2 Approach
Autonomic management approaches in general are based on the autonomic man-
agement cycle. This comprises a monitoring, analysis, planning and execution
4 Markus Tauber, Florian Skopik, Thomas Bleier and David Hutchison
phase. During a monitoring phase relevant events are captured, of which metrics
are derived in an analysis phase. Based on these metrics a policy determines how
the system is modified in a planning and an execution phase.
At a high level: Our autonomic management mechanism is intended to op-
erate on each smart meter in a grid individually, requiring local data only in
order to achieve an overall benefit. It is designed to detect cases when reporting
effort is being wasted, and to increase the current reporting interval accordingly.
Conversely, it decreases the interval in situations when a higher reporting rate
is appropriate. A high variability within reported (aggregated) values suggests
a decrease of the interval which other wise could be increased, to reduce unnec-
essary reporting activities. The magnitude of change for each of these interval
adaptations (increase/decrease) is determined by our autonomic manager’s pol-
icy. During each planning phase, the policy considers metric values derived from
events received during the current autonomic cycle. These events are based solely
on locally gathered data, thus no additional network traffic is generated by the
autonomic manager.
In detail this means that: In the monitoring phase energy consumption is
measured in (close to) real-time, and such an individual measurement is referred
to as Raw Energy Measurement (REM). A number of such REM values will be
measured, maintained and aggregated. We refer to the aggregate values as Ag-
gregated Raw Measurements (ARM) ; the latter values are sent in smart meter
reports. The metric we consider as an appropriate measure for variability is the
standard deviation of ARM values (σARM ). Based on σARM the policy deter-
mines the proportion Pby which the current interval should be decreased. In our
preliminary investigation here we define a threshold tafter which we consider
the variability (i.e. σARM ) as high enough for decreasing the reporting interval.
The new interval is then calculated as:
new interval =current interval ×(1 −P)(1)
The proportion of change Plies between zero and one, and is calculated as:
P=1−1
metric−ideal
k+1 (2)
where metric denotes σARM and ideal is zero in our case. kis a positive constant
that controls the rate of change of Pwith respect to the difference between
the metric value and its ideal value. The higher the value of k, the lower the
resulting proportion of change, and hence the slower the resulting response by
the manager. kcan be used to consider consumers’ reporting preferences in the
smart meter configuration. Further, we define that, if σARM is smaller than the
variability threshold we increment the current value by 10 (arbitrarily chosen).
We constrain ourselves here to values between 1 seconds (the lowest possible
reporting interval) and 1 hour (the maximum interval [7] – see Section 2).
A Self-Organising Approach for Smart Meter Communication Systems 5
4 Evaluation and Results
We have evaluated our autonomic management approach based on usage pat-
terns derived from [6]. These show the energy consumption of a number of ap-
pliances over a normal day. We reproduce this as shown in Figure 1, which
also represents measurements at the smallest possible interval (i.e. 1 sec.). The
exhibited usage pattern represents an average user into which we factor in an
e-car battery4to represent a regular use case with phases of heavy usage (after-
noon/evening) and light usage (nights). Figure 1 shows the energy usage over
the elapsed time during a day when a static energy reading interval (1 sec) is
defined and also shows how the reporting interval is autonomically adapted,
based on our approach (as outlined in Section 3). We choose a number of values
for the policy parameters t(threshold) and k(adaptation rate control), the pa-
rameter values are given in the individual plots. We also configure our policy
0
0.5
1
1.5
2
2.5
0 2 4 6 8 10 12 14 16 18 20 22 24
energy (Wh)
time (h)
Refigerator
Toaster
Kettle
Washing
machine
Hob heater
Oven
Kettle
E-vehicle
charging
energy usage reports at each second
0
50
100
150
200
250
300
350
0 2 4 6 8 10 12 14 16 18 20 22 24
reporting interval (s)
time (h)
autn. pol.: k=10, t=100
autn. pol.: k=10, t=100
0
50
100
150
200
250
300
350
0 2 4 6 8 10 12 14 16 18 20 22 24
reporting interval (s)
time (h)
autn. pol.: k=100, t=100
autn. pol.: k=100, t=500
0
50
100
150
200
250
300
350
0 2 4 6 8 10 12 14 16 18 20 22 24
reporting interval (s)
time (h)
autn. pol.: k=1000, t=100
autn. pol.: k=1000, t=500
Fig. 1. Raw energy usage over time, at the left top and autonomically adapted intervals.
to only consider a sample size of the latest 30 ARM values to compute σARM .
The number of values determines how much outliers may be compensated for
and the semi-arbitrarily chosen number (based on test runs of our harness) was
considered sufficient for this initial demonstration. We leave an investigation on
ideal sample sizes for future work as this would be beyond the scope of this work.
The presented results show that our simplistic autonomic manager detects
phases of little variability and increases the reporting interval accordingly until
4http://www.pluginrecharge.com/p/calculator-how-long-to-charge-ev.html
6 Markus Tauber, Florian Skopik, Thomas Bleier and David Hutchison
a peak – this suggests potential for improvement with respect to how aggres-
sively the manager deals with variability and high intervals. We also see that the
configuration with high kvalues (1000) reacted most desirable by keeping the
interval low when fluctuations occurred. Only little difference can be observed
between the effects due to t. However, an holistic analysis (e.g. gradient decent)
for all above mentioned policy parameters is required as next step in future work.
5Conclusion
As outlined in this paper, the smart-grid will have to be increasingly flexible to
cope with varying usage patters and hence the identification of aspects which
can be managed in an autonomic manner is an important step to improve the
smart-grid further. In a first step to add some self-* properties to the smart grid
we have proposed an adaptation of an individual entity of the grid to achieve
some overall benefit. We have shown that autonomic adaptation of the reporting
interval in individual smart meters will result in significantly fewer reports in
phases with very little variability of energy consumption behaviour between the
reports that are send to a central control unit. As this represents only an initial
investigation, we have limited ourselves to show a rather simplistic policy. A
multitude of adaptations of our policy is possible, e.g. reducing the historical
data analysed for deriving metrics, or evaluating of different parameters (e.g. k
see Equation 2) to consider consumer preferences, and also considering energy
generation at the consumer/prosumer side in our policy design. We also plan to
implement and experimentally evaluate our approach using available tools, as e.g.
[9]. Future work also includes an analysis of other self-* aspects, e.g., the trade-
off between privacy vs. sustainability due to interval adaptation (see Section 1).
This seems promising as we already see that adapted reporting intervals may
make it harder to derive usage patterns and hence to compromise privacy.
Acknowledgments
This work was funded by the Austrian security research programme KIRAS, by
the Austrian Ministry for Transport, Innovation and Technology and by the UK
EPSRC Current project (reference EP/I00016X).
References
1. Ibars, C., Navarro, M., Giupponi, L.: Distributed demand management in smart
grid with a congestion game. In: SmartGridComm. pp. 495 –500 (oct 2010)
2. Kephart, J.O., Chess, D.M.: The Vision of Autonomic Computing. IEEE Computer
36(1), 41–50 (2003)
3. Khurana, H., Hadley, M., Lu, N., Frincke, D.A.: Smart-grid security issues. IEEE
Security & Privacy 8(1), 81–85 (2010)
4. Metke, A.R., Ekl, R.L.: Security technology for smart grid networks. IEEE Trans-
actions on Smart Grid 1(1), 99–107 (2010)
A Self-Organising Approach for Smart Meter Communication Systems 7
5. Molina-Markham, A., Shenoy, P., Fu, K., Cecchet, E., Irwin, D.: Private memoirs
of a smart meter. In: ACM BuildSys. pp. 61–66 (2010)
6. NIST: Nistir 7628: Guidelines for smart grid cyber security: Vol. 2, privacy and
the smart grid. Tech. rep. (2010)
7. Sensus: Flexnet ami system, http://sensus.com/
8. Skopik, F.: Security is not enough! on privacy challenges in smart grids. Int’l J. of
Smart Grid and Clean Energy 1(1), 7–14 (2012)
9. Tauber, M., et. al.: Self-adaptation applied to peer-set maintenance in chord via a
generic autonomic management framework. In: SASOW. pp. 9–16 (2010)
10. Wei, D., et. al.: An integrated security system of protecting smart grid against
cyber attacks. In: Innov. Smart Grid Tech. pp. 1–7 (2010)
