Modeling Interdependent Socio-technical Networks via ABM
Smart Grid Case
Daniël Worm
1
, David Langley
2
and Julianna Becker
2
1
Performance of Networks and Systems, TNO, Delft, The Netherlands
2
Strategic Business Analysis, TNO, Delft, The Netherlands
Keywords: Agent Based Modeling, Interdependent Socio-Technical Networks, Smart Grid.
Abstract: The objective of this paper is to improve scientific modeling of interdependent socio-technical networks. In
these networks the interplay between technical or infrastructural elements on the one hand and social and
behavioral aspects on the other hand, is of importance. Examples include electricity networks, financial
networks, residential choice networks. We propose an Agent-Based Model approach to simulate
interdependent technical and social network behavior, the effects of potential policy measures and the
societal impact when disturbances occur, where we focus on a use case concerning the smart grid, an
intelligent system for matching supply and demand of electricity.
1 INTRODUCTION
The objective of this paper is to improve scientific
modeling of interdependent socio-technical
networks. This is important in the field of designing
critical infrastructures. Failures in these systems are
rare events which may have catastrophic
consequences. Society requires resilient
infrastructure which can cope with a wide variety of
threats. Examples of failures include natural
disasters like Hurricane Sandy, and technical failures
like cable burnout in the energy network in Germany
which has a highly distributed renewable energy
production. According to the German Federal
Network Agency, at the end of March 2013 the
electricity network threatened to collapse: “The
security of the network can no longer be guaranteed.
[...] We have had to intervene more than forty times
to prevent surges in wind and solar power from
compromising the entire electricity system. The
stress generated by these situations is becoming
increasingly difficult to handle.”
Since Holling’s (1973) seminal work on the
resilience of systems, engineering scientists have
endeavored to design critical infrastructures capable
of coping with disturbances (McDaniels et al 2008;
Boin & McConnell, 2007). However, social
components are often missing in models of critical
infrastructure. This is a problem for two reasons.
First, human behavior can influence the system, and
thus the likelihood of failure. Second, effects of
disturbances have human dimensions, whereby
strategic decisions can best take account of the
disruption that people experience (and the perceived
effect thereof) rather than use solely technical
parameters. This provides our motivation for this
research into interdependent socio-technical
modeling.
In this study we propose an Agent-Based Model
(ABM) approach to simulate interdependent
technical and social network behavior, the effects of
potential policy measures and the societal impact
when disturbances occur. The use of individual or
agent based approaches are common in the study of
complex adaptive systems (Holland 1995),
especially where the interactions between the agents
are complex, nonlinear, discontinuous, or discrete,
where the population is heterogeneous and where the
topology of the interactions is heterogeneous and
complex (Bonabeau 2002). This applies increasingly
to networks, whether physical or social. Using
ABM, its structure and behavior have potential to
resemble reality better than simple mathematical
models, especially when the underlying real
relationships are complex (Remondino 2004).
In order to obtain our objective, we focus on a
specific use case: the smart grid, a future intelligent
system that helps to match demand and supply of
electricity in a sustainable and secure manner. In
such a system, both social and technical aspects play
310
Worm D., Langley D. and Becker J..
Modeling Interdependent Socio-technical Networks via ABM - Smart Grid Case.
DOI: 10.5220/0004622503100317
In Proceedings of the 3rd International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH-2013),
pages 310-317
ISBN: 978-989-8565-69-3
Copyright
c
2013 SCITEPRESS (Science and Technology Publications, Lda.)
an important role. The model we obtain for this use
case, and that we will describe in this paper, helps to
give insight in certain effects arising from the
interplay between these aspects. Furthermore, from
it we obtain generic insights into interdependent
socio-technical network modeling, contributing to
our main objective.
2 RELATED WORK
The topic of simulation models for interdependent
socio-technical networks is receiving attention in a
wide range of scientific domains. This development
is based on the enormous amount of data related to
social, economic, technology and biological
networks, which is increasingly available for
research, as well as readily accessible computing
power for carrying out the necessary computations
(Kleinberg 2008; Jackson 2008; Reed et al 2009;
Khanin & Wit 2006). We briefly address a number
of the most relevant streams of literature.
In the field of resilience engineering a main focus is
on the effects of natural disasters on a range of
infrastructural networks. There is a general
recognition that interdependencies between
networks are both an important driver of cascading
failures and a significant modeling challenge (Reed,
Kapur and Christie, 2009). Recent work is including
the ‘human factor’ as one of the interdependent
networks, in recognition of the importance of
modeling the socio-technical system as a whole, e.g.
Johnsen and Veen’s (2013) assessment of the key
communication infrastructure used in emergency
communication in railways in Norway, although this
is not yet widespread practice.
A second relevant scientific domain is the
sociology of the housing market, where methods are
developed for analyzing housing price dynamics
(Erlingsson et al, 2013), urban sprawl and individual
choices about where to live, and the implications of
these choices for residential patterns (Devisch et al,
2009). Individual choices respond to the relative
attractiveness of residential areas, but they also
change that attractiveness (Bruch and Mare, 2012).
ABM have been used to model these choices (Macy
and Wilier, 2002; Benenson, 2004) including the
interdependencies of different market segments,
such as racial residential segregation (Zhang, 2004).
Finally, in direct relation to the case study we
address in this paper is the smart electricity grid.
Much literature on this topic which implements
ABM is focused on multi-agent systems to control
distributed smart grid technology, rather than
simulate the socio-technical networks including
household choices. Studies which do include human
behavior include simulating load profiles for
households equipped with smart appliances under
conditions of real-time variable-price tariffs
(Gottwalt et al., 2011), and micro-level models
of household capacity adaptation allowing for
occupants to vary their achieved comfort by
foregoing electricity when the price is too high
(Guo et al. 2008). Whether such behavior is realistic
in the real world has yet to be demonstrated. ABMs
of the smart grid demonstrate herding behavior
where many agents independently converge their
loads the time intervals they expect to have lower
prices, thus leading to undesirable load peaks which
can cause network failure (Ramchurn et al., 2011).
To prevent such herding behavior developing,
simulations have shown that introducing inertia can
help, for example by imposing penalties for
deviation from past behavior (Voice et al., 2011) or
more complex algorithms for spreading load across a
number of expected future low-price time intervals
(Reddy and Veloso, 2012).
The model we present in this paper builds on the
work from these scientific domains, adding
particularly to the theoretical grounding of the social
model from psychology as a way of improving the
combined socio-technical approach.
3 OBJECTIVES
Our aim is to model, in a quantitative manner,
interdependent socio-technical networks and the
effects that failure cascades can have. Of key
importance is the link between infrastructural,
(technical) networks and human behavior. In this
paper we focus on the smart grid case and in future
research we consider other cases and attempt to
uncover generic elements one should take into
account when modeling socio-technical
interdependent networks and their societal impact.
These models need not be highly accurate at this
stage, but they should be able to generate the types
of network behaviors arising from the
interdependency between the social and technical
systems based on the characteristics of the different
networks and on potential policy interventions.
Our research questions read:
How can we model cascading effects between a
technical and a social network model, whereby
changes and disturbances in a technical
network affect human behavior and vice versa?
ModelingInterdependentSocio-technicalNetworksviaABM-SmartGridCase
311
How can we model the societal impact of these
mutually interacting networks, both in hard,
financial terms and in a soft, reputational
sense?
In the long term our objectives are twofold. First, to
examine to what extent (re)routing / steering /
consolidating human behavior is possible when a
disturbance in a network occurs. Second, to examine
how policy interventions can influence failure
cascades between interdependent networks so as to
minimize negative societal impact.
4 SOCIO-TECHNICAL SMART
GRID MODEL
In this use case we model a future residential-level
electricity network, including smart grid elements. A
smart grid is an electricity network that intelligently
reacts to the behavior of different stakeholders, such
as generators, consumers and those that do both,
with the aim of efficiently supplying sustainable,
economic and secure electricity and coping with
disruptions (Clastres 2011). An important element in
achieving this is flexible pricing, triggering adaptive
consumer behavior.
This use case is relevant in relation to our
objectives described in Section 3, because of the
strong interdependency between the technical and
the human element in this socio-technical system.
The behavior of consumers plays a key role in the
performance of the future electricity network, since
this behavior directly determines electricity demand
and decentralized supply, which then affects the
pressures placed on the physical electricity network.
We choose to model this human element at the
individual level, rather than the aggregated level, so
that we can include heterogeneous effects per
household, such as the price each pays, the comfort
(i.e. the fulfilled demand) and power failures each
experiences, as well as the peer influence working
via the social network. Therefore, the electricity
network at the residential level (low voltage) is
relevant to our purposes, although the results can be
extrapolated to the neighborhood and regional level.
4.1 Description of Model Framework
To model the interaction between a residential-level
electricity network and human behavior, we made
three separate models. These models interact with
each other as shown by Figure 1. The behavioral
model is split up into two models:
1. Short Term Choice Model: This model covers
the short term choices consumers make based on
their electricity needs and fluctuating electricity
prices. We assume a power-management application
adapts demand real-time and that the consumers can
choose one of three profiles: maximum comfort
(electricity is used irrespective of the price), medium
comfort (a price cap is selected but only for a limited
time) and minimum comfort (a price cap is selected
and usage is halted above that price). Besides this,
consumers can choose two other one-off measures:
to install a solar panel and to insulate their home.
Time steps in this model are intervals of 15 minutes
and the model calculates how much electricity each
household demands (or supplies) per time step.
2. Long Term behavioral Model: This model
determines the attitudes and behavioral intentions of
the consumers, which in turn influence their
behavioral choices in the short term model. We
model five attitudes which are influenced by both
that household’s own experiences and by the
attitudes of others in their social network.
The five attitudes are: Attitude about price paid
for electricity, attitude about personal comfort (i.e.
the willingness to forego electricity), attitude about
personal energy efficiency, attitude about renewable
energy production, and attitude about confidence in
the electricity supply. These attitudes are continuous
variables with value between 0 and 1. Time steps in
this model are days, weeks, or months (set by an
adjustable parameter) and the model calculates the
five attitudes per household and what this means for
their behavioral intention.
This model is based on the psychological Theory
of Reasoned Action (Ajzen and Fishbein, 1980;
Bagozzi, 1992; Fishbein and Ajzen, 1975) which
states that behavioral intention is driven by attitude
and social influence. Social influence is the person's
perception that most people who are important to
him think he should or should not perform the
behavior in question. Later theories of social
influence go beyond this normative pressure to
include other forms of influence, such as imitation
(Langley et al, 2012). As for the link between
behavioral intention and actual behavior, in a meta-
analysis of 87 studies, Sheppard, Hartwick and
Warshaw (1988) report an average correlation
between intention and behavior to be 0.53, which
means that on average consumers’ answers to
questions about their intentions account for only 28
percent of the variance in their actual behavior. For
low-involvement products, such as electricity, this
link may be even weaker (Quester and Lim, 2003).
Therefore, we introduce a probability for linking
SIMULTECH2013-3rdInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
312
behavioral intention in the long term model to
choice behavior in the short term model.
We do not include the extended Theory of
Planned Behavior, which includes the role of
perceived behavioral control, as the behaviors in our
model are well within the behavioral control of the
agents (Ajzen, 1991).
Finally we have a technical network model:
3. Electrical Network Model: This model
computes power flows in a residential-level (low
voltage) network, based on demand and supply. It
also determines if and where disruptions occur in the
electrical network, for example if the supply in a
given part of the electricity network exceeds the
demand whereby a physical cable burns through.
Time steps in this model are intervals of 15 minutes.
Figure 1: Graphical overview of the relationships between
the three models.
These models are connected as follows: Each 15
minute interval, the short term choice model is
executed, computing the demand (or supply) of each
consumer, based on the price at that moment (which
in turn is influenced by total demand and supply),
their comfort profile, their devices needing
electricity, their insulation level and the production
of their solar panel, if applicable. The output is
passed to the electrical network model, that
determines how the demand is met and if any
disruptions in the network occur. This is
communicated back to the short term choice model,
because disruptions affect the remaining demand of
each consumer.
This process continues until one time step of the
long term model has been reached. Then the long
term behavioral model is executed, taking into
account output from both the short term behavioral
model as well as the electrical network model over
the past time.
4.2 Model Specifics
Agents with their Social Network
The agents in our model are 208 households in a
fictional residential area, divided into 13 streets.
Each household has a number of electrical devices
that require different amounts of energy and have
different time windows within a day in which the
demand of the device should be fulfilled.
The agents are linked with each other via a social
network (‘friends’), which is randomly drawn via
the following principles: The number of friends of
each household is Poisson distributed with mean λ,
and distributed in such a way that two households in
the same street (‘neighbors’) are n times more likely
to be friends than two households in different streets.
(We chose λ =8 and n=4) This social network will
influence the agents’ attitudes. We randomly divide
the agents into three different types, which fixes
initial attitudes of the agents: Comfort (willing to
pay for high comfort), Budget (wants to pay as little
as possible), Eco (aims towards sustainable energy).
Technical Network
The electricity model consists of a network with 14
nodes, taken from an actual low-voltage network.
One of the nodes is the main generator that connects
the low-voltage network to the medium-voltage
network and thus supplies all demanded electricity
which the households do not produce themselves via
solar panels. The model computes the power flows
over each link needed to fulfill the demands, based
on DC power flow methods (Wood and Wollenberg,
1996). Each link is endowed with a maximum
capacity which, if crossed, will cause the link to
break, leading to rerouted power flows which may
cause new failures in turn and possibly derive
households of electricity.
The behavioral and technical models connect via
the electricity demands of the agents. The demand
and supply within a street (16 households per street,
13 streets in total) are aggregated and communicated
as input to one of the nodes in the electrical network.
In turn, failures in the electrical network influence
the behavioral model, by changing attitudes due to
unfulfilled electricity demands.
4.3 Implementation and Verification
The two behavioral models were implemented in
Repast Simphony 2.0 Beta, a java-based toolkit for
agent-based modeling and simulation (North, Tatara,
Collier and Ozik 2007). For the technical electricity
ModelingInterdependentSocio-technicalNetworksviaABM-SmartGridCase
313
network model, an existing load balancing low-
voltage model is used, which has been implemented
in a MATLAB package called MATPOWER
(Zimmerman e.a., 2011). The Java package
MatlabControl enables the connection between the
different models.
In order to verify if the implemented models
correspond with our conceptual design and work in
the desired way, we follow the verification process
proposed by Rand and Rust (2011), which includes
documentation, programmatic testing, and test cases.
Due to space constraints we do not go into detail in
this paper. One issue which we experienced in the
verification process is interesting to note: that some
of the proposed verification steps are difficult to
carry out in the case of modeling interdependent
networks. For example, one of the test case
approaches these authors recommend is the use of
corner cases, whereby extreme values are used as
inputs and the behavior of the model is examined for
unexpected output (Gilbert, 2008). However, due to
the interdependencies incorporated into our model,
interpreting the output of corner cases is non-trivial.
4.4 Scenario Analysis
In order to address our research questions we ran a
number of scenario’s whereby different conditions
were assessed. We highlight a number of the most
interesting results here.
Crossover Effects
We ran the model in a ’default’ setting (Figure 2)
and in a setting where network cables are more
likely to fail (Figure 3), in order to investigate
crossover effects from the technical model into the
behavioral model.
The figures show the dynamics in the behavioral
model regarding agents choices for the different
power-management comfort profiles. We see a clear
distinction between these two cases: in the more
fragile network setting, the minimum comfort profile
is less popular than in the default setting, and there is
an increase in the maximum comfort profiles. A
reason for this is that disruptions lead to less fulfilled
electricity demands than usual, causing more people
to wish higher comfort. This in turn may cause even
more disruptions in the network, due to increased
demand, which shows a crossover effect from the
behavioral model back into the social model.
Figure 2: Comfort profiles in the default scenario.
Figure 3: Comfort profiles in the scenario with a more
fragile electricity network.
Policy interventions
We can use the smart grid model to investigate the
effects of policy interventions. For instance, by
increasing solar panel subsidies, assuming this
influences people to buy more solar panels, we see
that more people will opt to go for the medium
comfort profile compared to the default scenario (see
Figure 4).
Figure 4: Comfort profiles when solar panel subsidies are
high.
SIMULTECH2013-3rdInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
314
This happens because their energy needs are
more easily fulfilled via their solar panels (so their
comfort is sufficient enough), and energy prices for
them will drop also due to solar panels (so the price
is cheap enough).
However, there is also an unwanted effect on the
electrical network: more disruptions occur in the
network compared to the default scenario (where
hardly any disruptions occur), throughout the
timeline. This can be explained by the fact that the
solar panels will create synchronized supply (on
sunny periods), which may disrupt the technical
network. This effect increases if the number of solar
panels people buy will be increased.
Societal Effects
Another crucial aspect in the smart grid case is the
impact that disruptions have on society, in ‘hard’
financial terms as well as a ‘soft’ reputational sense,
like trust. There are several ways to look at financial
impact. In (Baarsma, Berkhout and Kop, 2004)
several formulas are derived for financial impact for
individual households and companies based on both
frequency and duration of disruptions. Both of these
may differ per agent in our model, so applying the
formulas give insight in e.g. the variability of
financial impact in a residential area, which turns out
to be quite high in a scenario with many disruptions.
Trust (in the electricity system) is more difficult
to measure in real life. Surveys can help to give an
estimation for trust. In our model, we use the
variable attitude about confidence in the electricity
supply as a measure for trust. We use this to assess
the relationship between fraction of disturbances and
trust levels, in particular the impact trust has on the
operation of the network. The nature of the model
will reflect a level of distrust in the network when
there are more failures due to the behavioral aspects
built into the modeling. Therefore the impact looks
at the relationship between the two in terms of what
happens to the fraction of disruptions as distrust
increases and how do agents adapt to this.
For this analysis a conditional probability was
used focusing on the probability of failure given that
there is high distrust in the network, P(F|D>0.5),
compared to the overall probability of failure, P(F).
Looking at a run with many failures we found that
the probability of failure is higher when there are
high levels of distrust in the network, as would be
expected. However, this also suggests something on
the behavioral impact of these failures: When there
is lower trust in the network agents are more likely
to demand energy whenever they have access to it,
as if there were a sense of urgency to use the energy
before it goes out again as opposed to behaving in an
energy efficient way to safeguard their energy levels
(for instance by adapting a maximum comfort
profile). The relationship between trust and behavior
in this model implies a more irrational actor when
trust is lost, increasing the probability of a network
failure which would only perpetuate the cycle as
represented in the figure below.
Figure 5: Fraction of disruptions per day in extreme case
with many disruptions.
However, we also encounter other scenarios where a
loss of trust in the network occurs at a certain point
in time, but where the system was able to overcome
that to provide stable energy supply. These types of
scenarios are interesting to model in terms of
exploring alternative scenarios to restore or redirect
trust.
Overall, the societal impact of the smart grid can
be modeled to show how disruptions affect agents
under various scenarios, and, in turn, to see how this
influences the behavior of agents. This provides a
foundation for further exploration into these
interactions.
5 CONCLUSIONS AND FUTURE
RESEARCH
Some observations can be made in the smart grid
setting:
A consumer’s individual actions, e.g. to
compensate for a fragile network, may cause a
worsening effect on system level, in the end
causing more damage for the individual.
Policies with good intentions (e.g. subsidizing
increased solar production) may lead to
unwanted effects (disruptions in the network).
Possibly, other pricing strategies might enable
policy makers to better obtain the effect they
want (e.g., a stable network).
ModelingInterdependentSocio-technicalNetworksviaABM-SmartGridCase
315
Sensitivity analysis can be applied to gain more
insight in the effects of (combinations of) parameters
on the outcome.
The ABM approach seems suitable to investigate
interdependency between social and technical
networks. It allows to observe unforeseen (possibly
unwanted) effects arising from this interdependency
and certain policy interventions. It also allows to
investigate impact on society. The java-based toolkit
Repast Simphony is flexible for this purpose, and
allows for connections with other programming
languages, which is useful for embedding a specific
technical model into a social/behavioral ABM.
Our case highlights the need for a multi-
disciplinary approach to using ABM for socio-
technical networks. Each research domain has its
own ontology which typically does not readily
combine with that from other domains (van Dam,
Nikolic, and Lukszo, 2012). For example, the
concepts and entities generally included in system
models of technical electricity grids are
incompatible with social psychological models of
human behavior. And yet we attempt to combine
both ontologies in a single ABM.
An essential next step to take is validation of a
socio-technical network. Because we need to make
many simplifications (compared to reality) in both
the social and the technical network model, the
question is whether the combined model actually fits
reality reasonably well. If unforeseen events arise
from the socio-technical model, one would like to
know if these events are plausible in reality or come
from an oversimplification or wrong specification of
the model. It should be stressed that our aim is not to
create perfect accurate predictive models at this
phase; instead we would like to use our models to
generate the types of network behaviors arising from
the interdependency between the social and
technical systems based on the characteristics of the
different networks and on potential policy
interventions. In the smart grid case, the setting is
futuristic, therefore we had to use fictional data and
could not directly validate the complete system,
though we need to take further steps in this
direction.
Another relevant research direction is balancing
the level of required detail or complexity in both the
social and the technical network models, in order to
make them fit together best.
For both themes our future research focuses on
obtaining guidelines that are as generic as possible,
i.e. that should be applicable also to other socio-
technical networks. We aim to obtain these goals
through the study of different use cases, like
residential choice models and financial networks.
REFERENCES
Ajzen, I., 1991. The theory of planned behaviour,
Organizational Behavior and Human Decision
Processes, 50(2): 179–211.
Ajzen, I. and Fishbein, M., 1980. Understanding attitudes
and predicting social behavior. Englewood Cliffs, NJ:
Prentice-Hall.
Baarsma, B., Berkhout, P., Hop, P., 2004. Op prijs gesteld,
maar ook op kwaliteit. Technisch rapport, SEO,
Amsterdam.
Bagozzi, R. P., 1992. The Self-Regulation of Attitudes,
Intentions, and Behavior, Social Psychology
Quarterly, 55(2): 178–204.
Benenson, I., 2004. Agent-based modeling: From
individual residential choice to urban residential
dynamics, in Spatially Integrated Social Science:
Examples in Best Practice, Oxford Univ. Press,
Oxford, U.K.
Boin, A., McConnell, A., 2007. Preparing for Critical
Infrastructure Breakdowns: The Limits of Crisis
Management and the Need for Resilience. Journal of
Contingencies and Crisis Management, 15(1): 50–59.
Bruch, E., Mare D., 2012. Residential Mobility, and
Neighborhood Change. Sociological Methodology, 42:
103-154.
Clastres, C., 2011. Smart Grids: Another step towards
competition, energy security and climate change
objectives. Energy Policy 39(9), 5399–5408.
Dam, K. van, Nikolic, I., Lukszo, Z., 2012. Agent-based
modelling of socio-technical systems. Springer:
Dordrecht, Netherlands.
Devisch O., Timmermans H., Arentze T., Borgers A.,
2009. An agent-based model of residential choice
dynamics in nonstationary housing markets.
Environment and Planning A, 41(8): 1997 – 2013.
Erlingsson, E., Raberto, M., Stefánsson, H., Sturluson, J.,
2013. Integrating the housing market into an agent-
based economic model. Managing Market Complexity,
Lecture Notes in Economics and Mathematical
Systems 662: 65-76.
Fishbein, M., Ajzen, I., 1975. Belief, attitude, intention,
and behavior: An introduction to theory and research.
Reading, MA: Addison-Wesley.
Gottwalt, S., Ketter, W., Block, C., Collins, J., Weinhardt,
C., 2011. Demand side management - A simulation of
household behavior under variable prices. Energy
Policy, 39: 8163-8174.
Guo, Y., Li, R., Poulton, G., Zeman, A., 2008. A
Simulator for Self-Adaptive Energy Demand
Management. In IEEE Conf. on Self-Adaptive and
Self-Organizing Systems.
Holling, C., 1973. Resilience and stability of ecological
systems. Annual Review of Ecological Systems 4:1–23.
Jackson, M., 2008. Social and Economic Networks
(Economics, Physics, Sociology), Princeton Univ.
Press.
Johnsen, S., Veen, M., 2013. Risk assessment and
resilience of critical communication infrastructure in
SIMULTECH2013-3rdInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
316
railways. Cognition, Technology & Work, 15(1): 95-
107.
Khanin, R., Wit, E., 2006. How scale-free are biological
networks. J. Comput. Biol., 13: 810-818.
Kleinberg, J., 2008. The convergence of social and
technological networks. Commun. ACM, 51: 66-72.
Langley, D.J., Bijmolt, T.H.A., Ortt, J.R., and Pals, N.
(2012) Determinants of Social Contagion During New
Product Adoption, Journal of Product Innovation
Management, 29(4): 623–638.
Macy, M., Wilier, R., 2002. From Factors to Actors:
Computational Sociology and Agent-Based Modeling.
Annual Review of Sociology, 28:143–66.
McDaniels, T., Chang, S., Cole, D., Mikawoz, J.,
Longstaff, H., 2008. Fostering resilience to extreme
events within infrastructure systems: Characterizing
decision contexts for mitigation and adaptation.
Global Environmental Change,18(2): 310–318.
North, M. J., Tatara, E., Collier, N.T., and Ozik, J. (2007)
Visual agent-based model development with repast
symphony, Agent 2007 Conference: Complex
Interaction and Social Emergence, Argonne National
Laboratory, Argonne, IL USA, 173–192.
Ramchurn, S., Vytelingum, P., Rogers, A., Jennings, N.,
2011. Agent-Based Control for Decentralised Demand
Side Management in the Smart Grid. In Int. Conf. on
Autonomous Agents and Multiagent Systems.
Rand, W. and Rust, R.T. (2011) Agent-based modeling in
marketing: Guidelines for rigor, International Journal
of Research in Marketing, 28: 181–193.
Reed, D., Kapur, K., Christie, R., 2009. Methodology for
assessing the resilience of networked infrastructure.
IEEE Systems Journal, 3(2): 174-180.
Reddy, P., Veloso, M., 2012. Factored Models for
Multiscale Decision-Making in Smart Grid Customers.
In Proceedings of the Twenty-Sixth AAAI Conference
on Artificial Intelligence, 363-369.
Sheppard, B., Hartwick, J. and Warshaw, P., 1988. The
Theory of Reasoned Action - a Meta-Analysis of Past
Research with Recommendations for Modifications
and Future-Research. Journal of Consumer Research,
15(3): 325–343.
Voice, T., Vytelingum, P., Ramchurn, S., Rogers, A.,
Jennings, N., 2011. Decentralised Control of Micro-
Storage in the Smart Grid. In AAAI Conf. on Artificial
Intelligence (AAAI-11).
Wood, A., Wollenberg, B., 1996. Power Generation,
Operation, and Control, 2nd ed. New York: Wiley.
Zhang, J., 2004. A Dynamic Model Of Residential
Segregation. Journal of Mathematical Sociology, 28:
147-170.
Zimmerman, R., Murillo-Sánchez, C., Thomas, R., 2011.
MATPOWER: Steady-State Operations, Planning and
Analysis Tools for Power Systems Research and
Education. Power Systems, IEEE Transactions on,
26(1): 12-19.
ModelingInterdependentSocio-technicalNetworksviaABM-SmartGridCase
317
