Eur. Phys. J.

Special Topics 223, 2423–2437 (2014)


c EDP Sciences, Springer-Verlag 2014 THE EUROPEAN
DOI: 10.1140/epjst/e2014-02219-6 PHYSICAL JOURNAL
SPECIAL TOPICS
Review

Models for the modern power grid

Pedro H.J. Nardelli1,a , Nicolas Rubido2,3 , Chengwei Wang2 , Murilo S. Baptista2 ,
Carlos Pomalaza-Raez4 , Paulo Cardieri5 , and Matti Latva-aho1
1
Department of Communications Engineering, University of Oulu, Oulu, Finland
2
Institute for Complex Systems and Mathematical Biology, SUPA, University of
Aberdeen, Aberdeen, UK
3
Instituto de Fı́sica, Facultad de Ciencias, Universidad de la República, Montevideo,
Uruguay
4
Department of Engineering, Indiana University – Purdue University Fort Wayne,
Fort Wayne, USA
5
School of Electrical and Computer Engineering, University of Campinas, Campinas,
Brazil

Received 31 December 2013 / Received in final form 8 May 2014

Published online 9 July 2014

Abstract. This article reviews different kinds of models for the electric
power grid that can be used to understand the modern power system,
the smart grid. From the physical network to abstract energy mar-
kets, we identify in the literature different aspects that co-determine
the spatio-temporal multilayer dynamics of power system. We start
our review by showing how the generation, transmission and distribu-
tion characteristics of the traditional power grids are already subject
to complex behaviour appearing as a result of the the interplay be-
tween dynamics of the nodes and topology, namely synchronisation
and cascade effects. When dealing with smart grids, the system com-
plexity increases even more: on top of the physical network of power
lines and controllable sources of electricity, the modernisation brings
information networks, renewable intermittent generation, market lib-
eralisation, prosumers, among other aspects. In this case, we forecast a
dynamical co-evolution of the smart grid and other kind of networked
systems that cannot be understood isolated. This review compiles re-
cent results that model electric power grids as complex systems, going
beyond pure technological aspects. From this perspective, we then indi-
cate possible ways to incorporate the diverse co-evolving systems into
the smart grid model using, for example, network theory and multi-
agent simulation.

1 Introduction

Electric power grid names the system deployed to deliver electricity from the gen-
erating units (power plants) to the end users (houses, industries etc.) [1]. Such a
a
e-mail: [email protected]
2424 The European Physical Journal Special Topics

system is traditionally a one-way network, where the generators “produce” the elec-
tric energy, that needs to “travel” long distances in high-voltage transmission lines,
to arrive in the distribution network that delivers the electricity to the consumers.
These three blocks – generation, transmission, and distribution – have their own engi-
neering challenges (many of them already solved in both research and implementation
levels).
In any case, the power grid elements are interdependent, which per se may allow
for complex processes such as cascade effects [3]. Big blackouts exemplify them, where
few small events occurring at approximately the same time may lead to a collapse in
the whole interconnected system.
The scenario described above illustrate the “old-fashion” electric power grids. In
contrast, modern grids – the so-called smart grids [1] – uses information and communi-
cation technologies (ICTs) to collect information (such as the behaviours of suppliers,
consumers, and prosumers) and take actions in an automated fashion to improve the
efficiency, reliability, economics, and sustainability of the production and distribution
of electricity [2,4]. The smart-grid, still an idea on the designing table, would enhance
the reliability of the energy infrastructure, decrease the impact of supply disruptions
and even create an internet of energy. A change of paradigm in the way energy is
produced, distributed, traded, and consumed, is set to take place.
Technological development of signal processing and two-way communication tech-
niques will be fundamental to securely distribute information among the grid ele-
ments. Besides, few big and many small generators are expected to coexist in smart
grids. This allows for more distributed energy sources (e.g. solar panels or wind tur-
bines) and self-sufficient micro-grids. The control of these variable sources of en-
ergy is also a big technological challenge, where real-time action is required in many
situations [1].
From another perspective those aspects of smart grids enable new products, ser-
vices, and markets. For instance, any person who has a solar panel at home may
become an energy trader. Therefore a new market appears within smart grids, where
consumers, or groups of consumers, play an active role in the electric power system:
they have evolved from simple consumers to active prosumers. The social collective
behaviour of the consumers and traders might bring drastic effects into the stability
of the grid.
From this brief description, we can also foresee possible issues beyond technical-
ities. For instance, markets are known to be unstable where few speculative agents
(persons or computer programs) may collapse the whole economic system – we are
nowadays experiencing this in the current crisis triggered by the US housing market
crash [5]. This would not be different in the smart grid. For instance, one might think
about speculative agents that buy energy when the price is low, store it and then
resell when the price is high. If this is the case, how this simple speculative behaviour
will change the power flows inside the grid? How the abstract market network affects,
and is affected by, the physical power network? The answers to those questions are
unknown and generally even not investigated.
In this review paper, we attempt to describe a wide range of existing network
models for the structures of smart grids beyond technical aspects and discuss how
they are inter-related, aiming at bridging the gap between these models, each coming
from different disciplines.
To do this, we divide our presentation into different sections. Section 2 focuses on
physical models of the grid and the complex behaviours and events that can happen
in the traditional grid, such as frequency synchronisation and cascade effects. Section
3 deals with the challenges of building a dedicated communication network for smart
grids. In Sect. 4, we present the benefit and problematic issues in using renewable
sources of energy and other emerging technologies that may change the grid dynamics
 Resilient Power Grids and Extreme Events 2425

(e.g. electric vehicles). Section 5 reviews the complexity related to the smart grid
management and how it is context-dependent (e.g. the smart grids designed for rural
zones are different from the ones for urban areas). In Sect. 6, we present theoretical
approaches that can be used to combine all these aspects of the energy system in order
to model and study the smart grid dynamics as a complex multiplex networks [6], a
network of networks formed by different layers, with different dynamical behaviours,
and different connections. Section 7 concludes this review, pointing out some possible
ways to go.

2 The physics of electric power grid

The power grid networks are still subject to a balance between supply and demand,
or load balance, since batteries and other related technologies that exist nowadays can
only store small amounts of electric energy. As we have mentioned, power grids are
topologically interconnected. This fact, summed with the inherent need for load bal-
ance within the network, characterizes complex dynamics of the grid. In this section,
we review some of the main results in the field. Specifically, focusing on synchroniza-
tion issues and cascade effects.

2.1 Synchronization

In power grids, electricity is normally transmitted in alternating current (in 50 or

60 Hz depending on the country). Therefore, in addition to load balance, synchroni-
sation between all elements is a crucial aspect to the system [7]. A famous description
of how synchronisation can appear in networks of coupled oscillators was made by
Kuramoto in [8]. Such description has been used as a model for many complex sys-
tems, ranging from biological clocks to power grids. For instance in [9], the authors
have shown that, if machines can be modelled as having the same dissipative coeffi-
cient α, the same moment of inertia, and power is not lost during transmission, then
a power grid where power-conservation laws holds can be cast into a second order
Kuramoto-like model. Namely,

θ̇i = ωi ,
N
 (1)
ω̇i = −αi ωi + Pi + Pmax aij sin(θj − θi ),
j=1

where i = 1, 2, . . . , N represents a consumer or a generating unit of the system, θi

the time-dependent phase of unit i, ωi is the time-dependent frequency centred in the
power grid one of Ω Hz, Pi is related to the power consumed (if Pi < 0) or generated
(if Pi > 0), aij indicates whether two nodes are connected and Pmax is related to the
maximum power that can be transmitted in a direct link, αi represents the damping
coefficient of the machine i. Since Eq. (1) considers that the power dissipated in
transmission is zero, i.e., whatever is transmitted is received, this model should be
considered as a model not for the whole grid, but as a model for the high-power
network, responsible for the transmission of power.
Noticing that the uncoupled dynamics of Eq. (1) is linear (ωi and −αωi ) and
has constant parameters (Pi ), and that the magnitude of the term representing the
power being transferred Pmax is constant for every edge, then, the linear stability of
the synchronous solution (θ̇i = θ̇j and ω̇i = 0) is independent of the parameters that
2426 The European Physical Journal Special Topics

make the network heterogeneous (Pi ). If a further assumption is made [10], i.e., the
network has nodes that are highly connected (large degree) and Pmax is sufficiently
strong, resulting in a network whose nodes have small phase difference, then the
stability of the synchronous nodes is determined by the eigenvalues of the matrix
describing the orthogonalised variational equations:
 
0 1
(2)
Pmax γi −α

where γi are the eigenvalues of the Laplacian matrix of the network, defined as
Lij = Aij − Iki (Aij is adjacency matrix, I is identity matrix, and ki is the de-
gree of node i). α is the average damping coefficient by assuming αi = αj , ∀i, j. In
this sense, the Master Stability Function [11] is therefore given by the values of Pmax
and α such that the matrix (2) has only eigenvalues with negative real part. This
function allows one to understand how the topology of the grid naturally (without
control) favour synchronous states. Notice that Eq. (2) is obtained by expanding cos
about zero (due to the small phase difference), which means that the power-grid is
operating in a mode where small amounts of power are transferred in the cable lines.
In [10], the authors have also studied long-term global stability by numerical means,
by determining coupling strengths for which the long-term behaviour of Eq. (1) is
frequency synchronisation, even after large perturbations are introduced in the sys-
tem. Their results were applied to study the natural stability of the Brazilian power
grid. Using the same grid model, the authors in [12] studied the collective dynamics
in large-scale oscillatory networks, using the British grid to exemplify their results.
Interestingly the authors showed a somehow counter-intuitive result: the more decen-
tralized the power sources are, the more robust the grid is against topological faults.
On the other hand, they also showed that a more decentralised system is at the same
time more sensitive to perturbations.
In [13] Witthaut and Timme have shown how the inclusion of new links can affect
the power grid synchronisation. The authors argument is that new links may lead
to Braess’s paradoxes [14] – phenomenon where adding new links reduces the overall
performance of the system. The authors recently extended this work in [15] where
they characterise other consequences of such a paradox in the network such as cascade
effects (which is the topic of the next subsection).
The swing equation was used in [16,17] as the dynamic model of generators:
2Hi Di
Ω θ̈i = Pmi − Pei − Ω ω. The parameters in the swing equation have a clear physi-
cal meaning. Hi is the inertia constant of the generator, Pmi is the mechanical power
provided by the generator. Pei is the electrical power transmitted to the network from
N
the generator reading Pei = Ei j=1 Ej (Bij sin(θi − θj ) + Gij cos(θi − θj )), where Ei
is the internal voltage of ith generator, Bij and Gij are the conductance and suscep-
tance between node i and j. Di is the damping coefficient with typical values between
0 and 2. In the Kuramoto-like model, the parameters are normalised. For example,
Pi is proportional to the power produced or generated, but not equal. If the power
dissipation on the transmission line is set to zero (i.e. Gij = 0), power grids modelled
by the swing equation become algebraic equivalent to power grids modelled by the
Kuramoto-like equation ([9]). One only needs to divide the constant value 2H i
Ω on
both sides of the swing equation, noticing that θ˙i = ωi . There is however a funda-
mental difference between both models regarding the way consumers are modelled.
Consumers are described in the Kuramoto-like model ([9]) by simply setting Pi < 0.
If one wants to model the dynamics of a power grid by the swing equation, one mod-
els a consumer by a unit that uses either a constant impedance, power, or current.
A reduced network is obtained by eliminating load buses (responsible to bring power
 Resilient Power Grids and Extreme Events 2427

to the consumers). This reduced network has values of Bij and Gij that incorporate
the existence of consumers.
We cite the following recent works that deal with synchronization of power grids as
a complex dynamical network: [16–20]. In particular, in [16], the authors have derived
a type of Master Stability Function for the Swing equation, an equivalent mechanical
model for a power system, more general than the model in (1), since it takes into
consideration that power is dissipated in the power lines. Therefore, it can be used to
study the stability of a power network that models not only the transmission but also
the distribution of energy. The authors have shown this type of Master Stability Func-
tion in terms of the eigenvalues of a matrix constructed from the eigenvalues of the
coupling matrix and parameters of the generators. The coupling matrix is a matrix
that contains terms related to the adjacency matrix (topology), but also terms due
to the phase difference between the nodes (dynamics). Even though the stability can
then be determined in terms of the network topology, the information obtained from
one type of network cannot be extended directly to another arbitrary network topol-
ogy. For this reason, we coin this result as a Master Stability type-Function, which
in fact can help one to understand how to enhance the stability of the synchronous
state of a particular network, by changing parameters but not the structure. Another
work of relevance is the one in [17], which has studied the global long term stability
of the swing equation, neglecting the dissipation in the transmission of power. Their
result is general, but still is a function of the phase diference, and therefore cannot be
used to infer the stability of a power grid by studying another power grid. Inferring
the stability of a network out of another network is a crucial point if one wants to
grow a power network, an issue that becomes even more needed now that the grid will
need to be adapted to become the smart grid. Finally, in the work of [18], the authors
provide a general global long-term stability condition for a mechanical system anal-
ogous to a structure preserving model for the power-grid. This model considers that
nodes have different dynamics. Each dynamical behaviour can be seen as a model for
the dynamics of different elements of a power-grid (see the discussions of the work
of [105], in Sect. 6). Some nodes are modelled by a second-order equation (similar to
Eq. (1)) and others are modelled by a first-order equation (similar to making ω̇i = 0,
in Eq. (1)). The condition states that the network will be frequency synchronous
ωi = ωj or phase coherent (|θi − θj | < π/2) if the eigenvalues of L+ ω are smaller
or equal then sin (γ), where L+ is the pseudo-inverse of the Laplacian matrix of the
network, and |θi − θj | ≤ γ < π/2.

2.2 Cascade effects

Cascade effects denotes a series of sequential events that is triggered by one or few
initial disturbances within the system, causing in some cases its collapse [22]. In power
grids, many blackouts are known to be caused as a result of cascades of small outages
events [23–26]. The best reported examples are the blackouts in North America in
2003, Europe in 2006, Brazil 2009 [25], and India 2012 [27], where a sequence of
outages shut down large parts of the electric grid.
Within complex systems, cascade effects are closely related to its topology [28–31]:
random, preferential-attachment, small-world, or scale free lead to particular cascade
characteristics. For example, scale-free networks are known to be unresilient against
selective attacks. In particular, scale-free networks are more sensitive to attacks on
short-range than on long-range links [32].
While there is still no consensus about whether there is a typical topology for real
power grids, in recent studies it was found that the power grid follows an exponential
distribution of degrees [31, 34]. There are however exceptions. In [35] the authors have
2428 The European Physical Journal Special Topics

shown evidences that the North America power grid could have a scale-free topology
[35]. Contributing to this controversy, the authors in [39] compared the data from
the US power grid using different fitting models. Based on the topological distance
analysis, they have shown that the power grid is neither scale-free nor small-word.
In fact, it is shown that there are topological differences in the grids worldwide, as
demonstrated by the works in [15,36, 37]. It is worth to mention that [31, 33], the
authors have suggested that the controversy about the type of network seems to exist
in the topology of the high tension power network, while the medium and low voltage
networks seem far from being a small-world network.
The analyses carried out in [38] are based on the “electric distance”, the equivalent
resistant of a Kirchoff network. The authors in this work claimed that they are able to
better capture the specificities of power grids (where Ohm and Kirchoff laws prevails)
than when topological distances or connectivity measures are employed. The electric
distance, whose analytical calculation was derived in [40], indeed provides a powerful
way to understand the flows and cascade effects in networks. The electrical distance
between two nodes – also known as equivalent resistance – multiplied by the amount
of flow entering a network is the load capacity of the edge, where load and flow are
linearly related. It is also shown in [41] that one strategy to avoid cascade effects in
DC models of power grid is to design a network topology such that its input flows
avoid surpassing the individual load capacities of the edges.
Using the electric distance, Eppstein et al. proposed in [42] an algorithm to identify
events that cause cascading failures in a power system. This work was further extended
in [43] for partitioning the power grid networks that may be used as a tool for avoiding
cascade effects. In a recent paper [44], the authors proposed a self-healing strategy for
Kirchoff networks, which can be potentially employed to avoid cascade fails in power
grids.
In most of the cases, cascade failures are related to connections or nodes that
are dropped in the power system. However, as described before, power grids are also
subject to Braess paradoxes when the topology of the network changes [15]. The
authors showed that the inclusion of a link may also lead to cascade events and
illustrate this in the British grid (Fig. 6 in [15]). At this point, we have illustrated
that the traditional power grids face a complex dynamics even when looking only at
simplified systems.
In most of the works presented in the previous section, it is was usual practice
to neglect the dissipation in power lines (the dissipative term α appearing in Eq. (1)
refers to the dissipation in the machines). Additional analytical difficulties appear if
one wants to take into account power loss due to dissipation. The work in [16] has
considered dissipation in the power lines. However, only local linear stability of the
frequency synchronisation is stated in this work. By neglecting dissipation, theoretical
results obtained from models of the power-grid such as the ones from [10, 13, 18] remain
only valid to study the power transmitted throughout the high-voltage transmission
lines.
Cascade models provide a way understand how the resilience of a grid is related
to both its topology and the dissipation in its power lines. In [21], the authors have
studied a model of a power-grid whose power lines have a distribution of heterogeneous
effectiveness (power-loss). The authors showed results of loads redistributions after
failures of nodes in the UK power grid. They showed that if generators fail in a
power network with the same topology of the UK power grid, the cascading effect
appearing would be less abrupt than the one in a network with a scale-free topology.
As mentioned, scale-free networks are unresilient to selective attacks, such as an attack
only on the generators. Compared with some existing models, the proposed model in
[21] requires larger node capacity to suppress cascading failure, being more realistic
to actual power grids.
 Resilient Power Grids and Extreme Events 2429

3 Information networks and smart grids

The information and communication technologies (ICTs) are changing the dynamics
of our society [45]. For example, the internet, which was built to connect researchers
back in 1980, is now part of everyday life [46]. Together with its benefits, the inter-
net is also creating new issues related to security, privacy, criminal activities amongst
others [47]. New generations tend to be globally networked, which produces new chal-
lenges, new risks and even changes in the scales of events [48]. And yet, new concepts
that machines, appliances, humans etc. must be always connected have been devel-
oping building the internet of things [49] or even the internet of everything [50].
In this way, humans and its social relations start being even more relevant to
understand the system behaviour [51]. The complexity involved in ICT-based sys-
tems is expect to be much higher than in “traditional” ones; for instance, tendencies
tend to propagate faster as people are able to exchange more information and then
time-scales of events tend to decreases, which in turn may facilitate cascade effects
[48]. These topics have been recently covered in a whole special issue of the European
Physical Journal Special Topics [52].
These kind of ICT-based solutions are also in the core of the smart grid – some
people even claim that smart grid will become the internet of energy [53, 54]. The
authors in [55] analysed wireless mesh networks for time-critical communications in
smart grids, targeting always the minimal delay. In [56] Galli et al. provided an exten-
sive analysis of power line communication in the context of smart grids, also indicating
the importance of the grid topology since “(...)the power grid is not only the infor-
mation source but also the information delivery system”.
As a reference in the topic, we cite [57] as a complete survey on the technolog-
ical challenges to design a communication network suitable for smart grids. They
described systematically (refer to Fig. 2 therein) the advantages and drawbacks of
possible technologies, from different wireless systems to power line communication.
Interestingly, they indicated the communication requirements needed for different
smart grid applications (e.g. demand response, outage management, distribution au-
tomation, advanced metering etc.) as well as the best communication standards for
each context.
From another perspective, the authors in [58] stated the challenges in the deploy-
ment of communication systems for smart grids: complexity, efficiency, reliability and
security. In the following sections, we will cover some of these challenges and try to
understand how the electric grid and the communication networks may co-evolve in
different contexts.

4 Challenges of renewable energy and other technologies

The discussions about CO2 emissions and their effects in climate change have been
growing [59]. One of the key points to decrease such emissions regards the use of
renewable sources of energy instead of oil- or coal-based sources. Looking to the
consumption side, electric/hybrid vehicles are seen as another important change to
decrease emissions. In this section, we will review the challenges in implementing
these technologies1 .

1
A counterpoint of those solutions are found in the analysis provided by Zahner and
compiled in [60], where the author claims that such celebrated solutions are only “green
illusions”.
2430 The European Physical Journal Special Topics

4.1 Renewable energy

In [59] Jacobson identified the following options as the most important renewable
sources of electricity: “solar-photovoltaics (PV), concentrated solar power, wind,
geothermal, hydroelectric, wave, tidal, nuclear, and coal with carbon capture and
storage technology”. Herein we focus on the characteristics of PV and wind sources
due to their wider use nowadays.
PV technology is used to convert solar radiation into electricity using semicon-
ductors that present photovoltaic effect [61]. While PVs have been employed for some
time in space industry as the source of energy to spacecraft, its commercial use start
growing with the energy crises in the 70’s [62].
Wind power plants is the technology that uses the wind movements to produce
energy. As a source of electricity, it dates 1888 [63], where “(...)the Brush wind turbine
in Cleveland, Ohio had produced 12 kW of direct current power for battery charging
at variable speed”.
In both cases, the idea is to capture the natural potential of energy and convert to
electricity so that the dependency of “dirty” sources like oil and coal decreases. In this
way, the objective is to switch the large-scale oil/coal power plants to “clean” PVs
and wind-mills technology. However, as a natural phenomenon, wind and sun shine
are unpredictable in small-scales and consequently their integration in the power grid
network may cause instability in the power supply (remind the need for load balance
within the power grid): sometimes peaks of production will not be followed by the
demand, or vice-versa [64].
Synchronization issues described in the previous section would also worsen when
such renewable sources integrate in the grid. This can be clearly seem if we study the
N
global mean field of Eq. (1). Defining the global mean field as Y = θi  = N1 j=1 θj ,
 N
and noticing that N1 i=1 j=1 aij sin (θj − θi ) = 0, then the mean field equation for
the Kuramoto-like power-grid network is given by Ÿ = −αẎ + Pi . If Pi  = 0,
meaning that demand matches production of energy, the mean field equation de-
scribes a steady-state solution, and the system is stable. If however Pi  = 0, for
example Pi  >> 0 due to a peak of generation, then the mean field variable be-
comes unstable and might oscillate. Each node in the grid is somehow coupled to the
global mean field. That causes oscillations in ωi making nodes to deviate from the
natural frequency of the grid Ω. To see this, we define the local mean field  of node
i as θi  = k1i j aij θj , and the coordinate transformation ri eiθi  = k1i j aij eiθj .
Neglecting imaginary terms [65], Eq. (1) can be written in terms of the local mean
field as
θ̈i = −αθ̇i + Pi + Pmax ki ri sin (θi  − θi ) (3)

where θi  = Y + ξi (t). Therefore, oscillations in Y might destabilise the machine i,

making its frequency to largely deviate from the natural frequency of the grid.
These engineering aspects are in fact the target of many research groups world-
wide. For example, the authors in [66] propose a battery-based approach to smooth
the supply variations in PV and wind generators. In [67] Riffonneau et al. studied a
dynamical programming approach to deal with power flow when PV generators are
present in the power grid.
All these solutions have in common the advance of storage technologies to filter
the unbalances inherent of PVs and windmills [68]. In other words, the problem is
that these generators have a dynamic that depends on factors out of engineering
control such as presence of clouds in the region where the PV is installed or wind
speed in the region of the wind-mill. Therefore, the engineering problem is to build a
network that can cope with these aspects while the power grid maintain its desirable
 Resilient Power Grids and Extreme Events 2431

characteristics (e.g. load balance, synchronisation etc.). As we saw in Sect. 2, these

aspects are closely related to the network topology and its power flows. Later in
Sect. 6 we will come back to this topic when providing the guidelines to assess the
system dynamics.

4.2 Electric vehicles

Another important element in the modern grids is the electric vehicle (EV) and its
unique load characteristics [69]: EVs can be seen as a mobile battery that can be
plugged anywhere in the power grid to be charged or even provide electricity in de-
mand peaks. In other words, the topology of the network may be highly mobile if EVs
start being widely used. This indicates that the challenges regarding load balance and
power flows will become even more complex.
One possible solution is to use scheduling algorithms to make use of the benefits
of EVs while avoiding charging them in demand peaks [70]. Hilshey et al. proposed
in [71] an algorithm to cope with the impact of charging EVs based on the so-called
smart charging methods. Marshall et al. gave in [72] another interesting perspective
of the problem by showing the effects of EVs on the heat transfer in the electric
cables. In [73], the authors assessed the impact of charging vehicles in residential
regions, comparing different ways to cope with it. Spatio-temporal dynamics of EVs
were analysed in [74] based on deterministic fluid equations, queueing systems and
stochastic models.
Some other studies evaluated the impact of EVs based on actual data. For exam-
ple, the authors prepared in [75] a report about how EVs affect the regional US power
grid. In some chapters of the book [76], MacKay dealt with many technical aspects
of EVs based on vehicle specification or available power grid data.
As pointed in [76], EVs are not only limited to cars but also collective transporta-
tion networks like trams, trains etc. In this case, one should understand much more
than technical aspects related to EVs and further evaluate their impact looking at
traffic patterns of people and goods in the transportation network, which is closely
related to when and where the vehicles need to be plugged in the power grid. In other
words, the dynamics of the transportation network will affect, and will be affected
by, the power grid.

5 Contextual aspects of smart grids

In recent years there has been a considerable amount of work carried out in sev-
eral countries to understand the best ways to modernize their power grids [77–79].
These studies have mainly focused on the technical aspects of the proposed smart
grid, showing several real case studies that have been used to evaluate its expected
advantages. Most of them, however, only evaluate the deployment of smart meters
across a neighbourhood or a small city, using then the information collected from
those meters to manage the energy use, in cooperation with the consumers.
The inclusion of consumers – who are people part of social contexts – is indeed a
key aspect in the success of smart grids. In fact, many results about network sciences
have been first presented by social scientists [80]. For example, the famous six-degrees
of separation that characterizes small-world topology were introduced empirically by
Milgram, a social psychologist, in [81]. After this seminal work, many other models
of social networks have been proposed, as indicated in the Chapter 3 of [82].
Processes of diffusion in networks are also very important to understand how a
technology such as the smart grid applications will propagate or not throughout the
2432 The European Physical Journal Special Topics

social system. Diffusion have been extensively studied in epidemiology to understand

how diseases spread over a network based on social relations (refer to Chapter 17 of
[82]). Dealing with new technologies, Rogers studied the diffusion of innovations in
the society [83].
Market (de)regulation is an important point in the deployment of smart grids.
Many specialists claim that real-time pricing or other market-based incentive strate-
gies can indeed change behaviours of consumption, decreasing the demand for electric-
ity in peak periods [84–87]. In other words, the smart grid system would be designed
upon market strategies to flat the demand load curves. Consumers who only demand
electricity in the traditional electricity market would be also expected to be able to
trade (buy and sell) energy. In this case, the new prosumers can use the technologies
described in Sect. 4 to produce energy and sell the surpluses to the grid.
While these solutions are optimal under the optics of the mainstream (neoclassical)
economic theory, they may fail when implemented in actual systems [88]. Amongst the
main critiques, we can cite rationality and equilibrium as assumptions that produce
interesting mathematical results [89], but many times fail to model the actual socio-
economical phenomenon. Some authors go even further: the utility theory of value
– which is the basis of neoclassical economic theory – cannot be either proved or
disproved since their basic assumptions cannot be tested [90]. Without going into the
merits of these works, we can conjecture some possible issues in deregulated electricity
markets such as speculation (e.g. buy electricity when it is cheaper, use batteries to
store it and sell when it is more expensive) or rational, autonomous demand control
(e.g. turn off air-conditioning when it is too hot or the heating when it is too cold).
These issues might contribute to stabilise the market, but might also destabilise it.
The energy market can present the same problematic issues that stock markets have.
For example, phenomena such as the “Minsky moment”. The longer the period of sta-
bility, the larger the potential risk for greater instabilities when market participants
must change their behavior. This effect in the stock market could cause a major col-
lapse in the value of energy.
Yet from the demand side, the inner city dynamics as well as its long-term evo-
lution are important to understand the grid implementation. When researching the
“science of cities”, social science researchers, together with urban planners and physi-
cists, have been developing analytical and simulation tools to study the evolution of
urban areas as in [91–93]. Their focus has been on large cities trying to find universal
features for places across the world. The main conclusion from those studies is that
factors such as land use, transportation, and communications play very important
roles in the social dynamics of cities, which in turn affects the energy consumption
and the power grid dynamics [94, 95].
The different results presented in this section suggest that the success of the smart
grid goes beyond the already complex technological challenges. In this case, there is
an increasing awareness that the proper study, modelling, and planning of smart grids
should not be limited to the technical aspects. In this way, the research should include
not only the power grid network itself, but also the different networks that affect, and
are affected by, the power system. These studies should also include the social, eco-
nomical and other contextual networks where the smart grid is operating as pointed
out in [94–99]. All in all the generation, transmission and distribution of electricity
do not exist on its own isolated domain.

6 Modelling smart grids

We have described many challenges related to the modernisation of the electric power
grids. This section indicates possible ways to include these diverse factors into an
 Resilient Power Grids and Extreme Events 2433

unified framework inspired by cutting-edge research in network theory, multi-agent

simulations and other tools developed to assess complex systems. Following this line of
thought, researchers from different fields participated in the workshop “Power Grids
as Complex Networks” (Santa Fe Institute on May, 2012). The conclusion of such a
transdisciplinary event was [97]: “Integrating and transcending disciplines will enable
new ideas to shape the power system(...)”.
For example, Kremers presented in his doctoral thesis [100] a way to model smart
grids as agent-based simulation, including diverse aspects such as people behaviour
using home appliances and time series equations to model the wind-power generators
in a given place. A research group from the US Department of Energy at Pacific
Northwest National Laboratory have been developing the GridLAB-D agent-based
simulator for distribution power system [101]. The authors in [102] recently studied
the control of renewable sources of energy using agent-based simulation.
In [94], the authors proposed a cellular automaton to model the expansion of a
city so as to predict the spatial demand of electricity based on the theory presented
in [91]. This work was further extended in [95], modelling the city expansion using
multi-agent models instead of the simple automaton. In both cases, however, the
modernisation aspects of the power grid (as the ones cited in Sects. 3 and 4) have
been neglected, and the authors only explore the growing in consumption based on
the city growth.
In [104], the authors study how different network topologies and growth mod-
els contribute to a network appropriated to a decentralised electricity market. They
showed which is an optimal growth evolution strategy that balances best the ratio
between increased connectivity and costs.
Other possibility to model smart grids comes from the last advances in control the-
ory for networked systems, regarding the observability of a system, i.e., the partial or
full determination of the state variables of a system from limited measurements. For
instance, the authors in [103] have analysed the observability of power grids based on
the so-called “network observability transition”, indicating when observable islands
appear in the system as the number of measurement nodes, randomly selected, in-
creases. These islands are group of nodes that can have their states known (complex
voltages) by a reduced number of observations using phasor measurement units.
In [106] Brummitt et al. studied how interdependence between networks can sup-
press cascade effects in power grids. They showed based on the Bak-Tang-Wiesenfeld
model [107] that there exist a number of interconnections that minimize the risk of
cascade effect whereas the systems still benefit of being interconnected.
Along these lines, a recent work [108] has proposed a tensorial mathematical for-
malism to tackle multiplex networks where nodes that are connected in many different
ways. They have extended the quantities characterizing traditional one layer networks
such as degree centrality, modularity, clustering coefficient, eigenvector centrality, to
multilayer networks, where each layer represent a different type of connections be-
tween nodes. Other results on network theory applied in smart grids were presented
back in Sect. 2, when discussing synchronization and cascade effects.
Finally, the work in [105] offers a bridge into how engineering and physics can
come together to understand the power-grid with a communication structure. They
showed how a power-grid composed of generators (modelled by the swing equation),
nodes representing the AC energy obtained from DC generation (coming from wind
and solar), and load buses (that distributes the power) can be controlled by applying
linear feed-back controllers in a reduced number of nodes. By control, it is meant that
one can either achieve frequency synchronisation (as previously defined, θ̇i = θ̇j ) or
phase coherence (|θi − θj | < γ). The authors showed how these states can be reached
as a function of the control strength and eigenvalues of a Laplacian matrix constructed
from the power-grid structure and the structure of the network formed by the nodes
2434 The European Physical Journal Special Topics

Cities

Contextual
Cascade Power Grid
Smart Grids Aspects Markets
Effects (Section 2)
(Section 5)

Diffusion
Syncroni-
of
zation
Innovation

Information New
Network Technologies
(Section 3) (Section 4)

Electric Renewable
Vehicles Sources

Fig. 1. Schematic figure to represent some aspects of the smart grid that can be modelled as
(complex) networks. Each topic is related to the respective section it appears in this review.

that are being observed and controlled. They also provided conditions for the control
when the observations are sampled, a situation that can often be present when for
example communication is interrupted.

7 Final remarks

In this review article, we presented many challenges that arise with the moderniza-
tion of power grids. From purely technical issues of the topology of the generation,
transmission and distribution networks (Sect. 2) to social issues of electricity market
deregulation, we revisited different phenomena related to smart grids that affect each
other in their co-evolving dynamics. We also identified in the recent literature of net-
works and multi-agent systems the tools needed to tackle these challenges. Figure 1
compiles the topics covered herein.
To conclude this paper, it is worth saying that the understanding of the smart
grids is still limited – although growing – due to the lack of transdisciplinary knowl-
edge in the field. We expect that this review also serves as a roadmap of possible ways
to carry out research in smart grids. In Sects. 1–5, we present the approaches being
used to model sub-units or elements that will be part of the smart grid. In Sect. 6, we
present theoretical approaches and ideas of how these sub-units and elements could be
theoretical merged into a unifying framework in order to understand the behaviour of
the power system as a complex multiplex networks [6], a network of networks formed
by different layers, with different dynamical behaviours, and different connections.
A complex network that is smart: resilient, energy efficient, stable, flexible and cheap
to evolve, making the energy to arrive at the end-user with a reduced amount of
greenhouse production.

This work was partly supported by the Science without Boarders Special Visiting
Researcher fellowship CAPES/Brazil 076/2012, SUSTAIN Finnish Academy and
CNPq/Brazil 490235/2012-3 jointly funded project, CNPq/Brazil 312146/2012-4. MSB
acknowledges EPSRC grant EP/I032606/1.
 Resilient Power Grids and Extreme Events 2435

References

1. X. Fang, et al., IEEE Comm. Surv. Tutorials 14, 944 (2012)

2. U.S. Department of Energy, Smart Grid/Department of Energy, Retrieved
2012-06-18
3. P. Hines, et al., SIAM Math. Awareness Month (2011)
4. H.-J. Appelrath, et al., Business Inf. Syst. Eng. 4, 1 (2012)
5. A. Jones-Rooy, S. Page, Business Critical Rev. 24, 313 (2012)
6. S. Gómez, et al., Phys. Rev. Lett. 110, 028701 (2013)
7. F. Blaabjerg, et al., IEEE Trans. Ind. Electr. 53, 1398 (2006)
8. Y. Kuramoto, in International Symposium on Mathematical Problems in Theoretical
Physics (Springer-Verlag, USA, 1975), p. 420
9. G. Filatrella, et al., Eur. Phys. J. B 61, 485 (2008)
10. R. Carareto, et al., Comm. Nonlinear Sci. Numer. Simul. 18, 1035 (2013)
11. L. Pecora, T. Carroll, Phys. Rev. Lett. 80, 2109 (1998)
12. M. Rohden, et al., Phys. Rev. Lett. 109, 064101 (2012)
13. D. Witthaut, Marc Timme, New J. Phys. 14, 083036 (2012)
14. D. Braess, et al., Transp. Sci. 39, 446 (2005)
15. D. Witthaut, M. Timme, Eur. Phys. J. B 86, 377 (2013)
16. A. Motter, et al., Nat. Phys. 9, 191 (2013)
17. F. Dorfler, F. Bullo, SIAM J. Cont. Opt. 50, 1616 (2012)
18. F. Dorfler, et al., Proc. Nat. Acad. Sci. 110, 2005 (2013)
19. J. Simpson-Porco, et al., Automatica 49, 2603 (2013)
20. S. Lozano, et al., Eur. Phys. J. B 85, 231 (2012)
21. H.-P. Ren, et al., Cascade Failure Analysis of Power Grid Using New Load Distribution
Law (submitted)
22. L. Duenas-Osorio, S. Vemuru, Struct. Safety 31, 157 (2009)
23. I. Dobson, et al., Chaos: An Interdisciplinary J. Nonlinear Sci. 17, 026103 (2007)
24. P. Hines, et al., IEEE Potentials 28, 24 (2009)
25. M. Vaiman, et al., IEEE Trans. Power Syst. 27, 631 (2012)
26. S. Pahwa, et al., Scientific Reports 4 (2014)
27. G. Zhang, et al., Physica A 392, 3273 (2013)
28. S. Strogatz, Nature 410, 268 (2001)
29. D. Watts, Proc. Nat. Acad. Sci. 99, 5766 (2002)
30. R. Albert, H. Jeong, A.-L. Barabási, Nature 406, 378 (2000)
31. G.A. Pagani, M. Aiello,Physica A 392, 2688 (2013)
32. Y.-C. Lai, A.E. Motter, T. Nishikawa, Lect. Notes Phys. 650, 299 (2004)
33. G.A. Pagani, M. Aiello, IEEE Trans. Smart Grid 2, 538 (2011)
34. R. Albert, et al., Phys. Rev. E 69, 025103 (2004)
35. D. Chassin, C. Posse. Physica A: Stat. Mech. Appl. 355, 667 (2005)
36. P. Crucitti, et al., Physica A: Stat. Mech. Appl. 338, 92 (2004)
37. M. Rosas-Casals, et al., Int. J. Bif. Chaos 17, 2465 (2007)
38. P. Hines, et al., Proceedings of the 43th Hawau International Conference on Systems.
Sciences (2010)
39. E. Cotilla-Sanchez, et al., IEEE Syst. J. 6, 616 (2012)
40. N. Rubido, et al., EPL (Europhysics Letters) 101, 68001 (2013)
41. N. Rubido, et al., Phys. Rev. E 89, 012801 (2014)
42. M. Eppstein, et al., IEEE Trans. Power Syst. 27, 1698 (2012)
43. E. Cotilla-Sanchez, et al., IEEE Trans. Power Syst. 28, 4979 (2013)
44. W. Quattrociocchi, et al., PLoS ONE 9, e87986 (2014)
45. D. Helbing, Eur. Phys. J. Special Topics 214, 5 (2012)
46. B. Leiner, et al., ACM SIGCOMM Computer Comm. Rev. 39, 22 (2009)
47. J. van den Hoven, et al., Eur. Phys. J. Special Topics 214, 153 (2012)
48. D. Helbing, Nature 497, 51 (2013)
49. L. Atzori, et al., Computer Networks 54, 2787 (2010)
2436 The European Physical Journal Special Topics

50. M. Becker, et al., Proceedings of the 2013 ACM conference on Pervasive and Ubiquitous
Computing Adjunct Publication (ACM, USA 2013), p. 1175
51. F. Giannotti, et al., Eur. Phys. J. Special Topics 214, 49 (2012)
52. S. Bishop, Eur. Phys. J. Special Topics 214, 1 (2012)
53. H.-J. Appelrath, et al., Business Inf. Syst. Eng. 4, 1 (2012)
54. M. Ajmone-Marsan, et al. Eur. Phys. J. Special Topics 214, 547 (2012)
55. Y. Xu, et al., IEEE Trans. Wireless Comm. 12, 3360 (2013)
56. S. Galli, et al., Proc. IEEE 99, 998 (2011)
57. V. Gungor, et al., IEEE Trans. Ind. Inf. 9, 28 (2013)
58. Y. Yan, et al., IEEE Comm. Surveys Tutorials 15, 5 (2013)
59. M. Jacobson, Energy Env. Sci. 2, 148 (2009)
60. O. Zehner, Green illusions: The dirty secrets of clean energy and the future of environ-
mentalism. (University of Nebraska Press, USA, 2012)
61. B. Parida, et al., Renewable Sust. Energy Rev. 15, 1625 (2011)
62. R. Nelson, Semiconductor Sci. Technol. 18, S141 (2003)
63. P. Carlin, et al., Wind Energy 6, 129 (2003)
64. J. Carrasco, et al., IEEE Trans. Ind. Electr. 53, 1002 (2006)
65. J.G. Restrepo, E. Ott, B.R. Hunt, Phys. Rev. E 71, 036151 (2005)
66. L. Xiangjun, et al., IEEE Trans. Sust. Energy 4, 464 (2013)
67. Y. Riffonneau, et al., IEEE Trans. Sust. Energy 2, 309 (2011)
68. H. Bevrani, et al., IET Renewable Power Generation 4, 438 (2010)
69. D. Wu, et al., IEEE Trans. Power Syst. 26, 738 (2011)
70. D. Wu, et al., IEEE Trans. Smart Grid 3, 368 (2012)
71. A. Hilshey, et al., IEEE Trans. Smart Grid 4, 368 (2013) 905 (2013)
72. J. Marshall, et al., Elect. Power Syst. Res. 97, 76 (2013)
73. K. Clement-Nyns, et al., IEEE Trans. Power Syst. 25, 371 (2010)
74. S. Bae, A. Kwasinski., IEEE Trans. Smart Grid 3, 394 (2012)
75. M. Kintner-Meyer, et al., Impacts assessment of plug-in hybrid vehicles on electric
utilities and regional US power grids (Technical analysis Pacific Northwest National
Laboratory, USA, 2007)
76. D. MacKay, Sustainable Energy-without the hot air (UIT Cambridge, UK, 2008)
77. Smart Grid System Report (U.S. Department of Energy, USA, 2009)
78. V. Giordano, et al., Smart Grid projects in Europe: lessons learned and current
developments (European Commission, Joint Research Centre, Institute for Energy,
Luxembourg, 2011)
79. Redes Eletricas Inteligentes (Centro de Gestão e Estudos Estratégicos, Brazil, 2012)
80. S. Wasserman, Social network analysis: Methods and applications (Cambridge
University Press, UK, 1994)
81. S. Milgram, Psychol. Today 2, 60 (1967)
82. M. Newman, Networks: an introduction (Oxford University Press, UK, 2009)
83. E. Rogers, Diffusion of innovations (Simon and Schuster, USA, 2010)
84. A-H. Mohsenian-Rad, A. Leon-Garcia, 1, 120 (2010)
85. A. Conejo, et al., IEEE Trans. Smart Grid 1, 236 (2010)
86. W. Saad, et al., IEEE Signal Proc. Mag. 29, 86 (2012)
87. L. Qian, et al., IEEE J. Sel. Areas Comm. 31, 1268 (2013)
88. D. Helbing, A. Kirman., Real-World Economics Rev. 64, 23 (2013)
89. E. Bompard, et al., IEEE Trans. Power Syst. 26, 1231 (2011)
90. P. Cockshott, et al., Classical Econophysics (Routledge, UK, 2009)
91. M. Batty, Cities and complexity: understanding cities with cellular automata, agent-
based models, and fractals (MIT Press, USA, 2007)
92. M. Batty, The New Science of Cities (MIT Press, USA, 2013)
93. L. Bettencourt, Science 340, 1438 (2013)
94. E. Carreno, et al., IEEE Trans. Power Syst. 26, 532 (2011)
95. J. Melo, et al., IEEE Trans. Power Syst. 27, 1870 (2012)
96. P. Ribeiro, et al., IEEE Technol. Soc. Maga. 31, 34 (2012)
97. C. Brummitt, et al., Proc. Nat. Acad. Sci. 110, 12159 (2013)
 Resilient Power Grids and Extreme Events 2437

98. P. Nardelli, et al. [ArXiv 1308.3229] (2013)

99. B. De Vries, Sustainability Science (Cambridge University Press, UK, 2012)
100. E. Kremers, Modelling and Simulation of Electrical Energy Systems through a Complex
Systems Approach using Agent-Based Models (KIT Scientific Publishing, Germany,
2013)
101. D. Chassin, et al., GridLAB-D: An open-source power systems modelling and simulation
environment (Proceedings of IEEE/PES Transmission and Distribution Conference and
Exposition, US, 2008)
102. D. Divenyi, A. Dan, IEEE Trans. Sust. Energy 4, 886 (2013)
103. Y. Yang, et al., Phys. Rev. Lett. 109, 258701 (2012)
104. G.A. Pagani, M. Aiello, Physica A 396, 248 (2014)
105. J. Giraldo, E. Mojica-Nava, N. Quijano, Synchronization of Dynamical Networks with
a Communication Infrastructure: A Smart Grid Application, 52nd IEEE Conference on
Decision and Control, December 10 (2013)
106. C. Brummitt, et al., Proc. Nat. Acad. Sci. 109, E680 (2012)
107. P. Bak, et al., Phys. Rev. A 38, 364 (1988)
108. A. Sole-Ribalta, et al., Phys. Rev. E 88, 032807 (2013)