Computer Communications 36 (2013) 1665–1697

Contents lists available at ScienceDirect

Computer Communications
journal homepage: www.elsevier.com/locate/comcom

Review

The role of communication systems in smart grids: Architectures,

technical solutions and research challenges
Emilio Ancillotti, Raffaele Bruno ⇑, Marco Conti
Italian National Research Council (CNR), Institute for Informatics and Telematics (IIT), Via G. Moruzzi 1, 56124 Pisa, Italy

a r t i c l e i n f o a b s t r a c t

Article history: The purpose of this survey is to present a critical overview of smart grid concepts, with a special focus on
Received 3 February 2013 the role that communication, networking and middleware technologies will have in the transformation of
Received in revised form 3 September 2013 existing electric power systems into smart grids. First of all we elaborate on the key technological, eco-
Accepted 6 September 2013
nomical and societal drivers for the development of smart grids. By adopting a data-centric perspective
Available online 19 September 2013
we present a conceptual model of communication systems for smart grids, and we identify functional
components, technologies, network topologies and communication services that are needed to support
Keywords:
smart grid communications. Then, we introduce the fundamental research challenges in this field includ-
Smart grid
Communication architecture and protocols
ing communication reliability and timeliness, QoS support, data management services, and autonomic
Control and management behaviors. Finally, we discuss the main solutions proposed in the literature for each of them, and we iden-
Middleware tify possible future research directions.
Security Ó 2013 Elsevier B.V. All rights reserved.

1. Introduction  To enable the massive deployment and efficient use of dis-

The term smart grid is commonly used to refer to a modernized
electrical system, in which new and more sustainable models of
energy production, distribution and usage will be made possible
by incorporating in the power system: (a) pervasive communica-
tion and monitoring capabilities, and (b) more distributed and
autonomous control and management functionalities [1,2]. As a
matter of fact existing electric grids are a large-scale, unidirec-
tional and centralized systems in which the electricity is delivered
from remote power plants through a tree-based distribution
system to local customers with pre-established load profiles [3].
However, a number of technological innovations, as well as
environmental and economical concerns, have emerged in the last
decade that have made traditional electric power systems outdated
and not well suited to meet the reliability, efficiency and sustain-
ability requirements posed by those changes [4].
Although there might be different views on what will be the
definitive model of a smart grid, the following key capabilities
are widely recognized as essential for the successful implementa-
tion of smart grids [5]:
⇑ Corresponding author.
E-mail addresses: e.ancillotti@iit.cnr.it (E. Ancillotti), r.bruno@iit.cnr.it (R.
Bruno), m.conti@iit.cnr.it (M. Conti).
tributed energy resources, including renewable energy
sources and energy storage systems;
 To enhance the efficiency, resilience and sustainability of
an electric grid by incorporating real-time distributed intel-
ligence enabling automated protection, optimization and
control functions;
 To allow the interaction of consumers with energy manage-
ment systems to enable demand-response and load shap-
ing functionalities;
 To enable real-time, scalable situational awareness of grid
status and operations through the deployment of advanced
metering and monitoring systems;
 To support the electrification of transportation systems by
facilitating the deployment of plug-in electric vehicles
and their use as mobile energy resources.
From a practical point of view the above vision requires the per-
vasive deployment of ‘‘intelligent’’ devices [6] (e.g., sensors, actua-
tors, smart appliances, smart meters, embedded computers, etc.)
that are capable of collecting real-time and fine-grained informa-
tion about electricity usage patterns, as well as about the status
of distributed energy resources and other components of the elec-
tric grid. This huge amount of heterogeneous information collected
by the metering and monitoring infrastructures that will be incor-
porated in a smart grid must be shared in a reliable and secure

0140-3664/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.comcom.2013.09.004
1666 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697

manner to a multitude of distributed energy management systems. is responsible for maintaining communication connectivity among
These systems are responsible for analyzing the received data, pre- systems, devices, and applications, which is similar to the role of
dicting and detecting possible issues (e.g., failures, energy short- the communication infrastructure described in this survey. In gen-
ages, disturbances), and making decisions to control and eral, some of the cited surveys can be regarded as complementary
optimize the operations of the power system. Note that most of to our work. However, our survey is distinct from those papers in
these management and control functions will be executed in re- the sense that it adopts a data-centric perspective and it tries to ex-
sponse to local events but they may have an impact on the resil- plain how communication networks will allow smart grid applica-
iency and efficiency of large portions of the electric grid. tions to collect, share and use information data in a timely, reliable
To support information collection, distribution and analysis, as and secure manner.
well as automated control and optimization of the power system, The rest of this paper, is organized as follows. In Section 2 we
we argue that the smart grid communication system will rely on first outline key drivers for smart grids and we provide an outlook
two major subsystems: a communication infrastructure and a mid- on the transition from existing power systems to future smart
dleware platform. The communication infrastructure consists of a grids. We also describe the characteristics of the most innovative
set of communication technologies, networks and protocols that: services and applications that smart grids can enable in Section 3.
(i) support communication connectivity amongst devices or grid In Section 4 we introduce a conceptual framework of the main
sub-systems, and (ii) enable the distribution of information and components of a smart grid communication system by following
commands within the power system. As better explained in the fol- a bottom-up approach. Specifically, we will start by outlining the
lowing, basic requirements for such communication infrastructure most popular communication technologies for smart grids. Then,
are scalability, reliability, timeliness and security. The middleware we continue with describing in details the various proposals avail-
platform consists of a software layer, which is situated between able in literature to implement communication infrastructures
the applications and the underlying communication infrastructure, (Section 5), middleware platforms (Section 6) and and security
providing the services needed to build efficient distributed func- mechanisms (Section 7) for smart grids, along with the most
tions and systems. Specifically, a middleware runs on the devices important open issues. Finally, in Section 8 we conclude our survey
that are part of the smart grid communication infrastructure to by summarizing lessons learned. In addition, refer to Table 1 for
support: (i) data management services (e.g., data sharing, storage, the abbreviations used in this survey.
and processing), (ii) standard communication and programming
interfaces for distributed applications, and (iii) computational
2. The evolution path towards the smart grid
intelligence and autonomic management capabilities. Further-
more, it is of paramount importance to ensure secure and reliable
In this section we overview the infrastructure that underlies
operations of the smart grid communication system to protect the
existing power systems to clarify its major shortcomings and the
entire smart grid. However, security solutions can not be confined
key drivers for smart grids. Furthermore, we discuss the most
to a single component of the smart grid communication system but
important new paradigms for electricity generation, delivery and
security is a cross-layer issue because it is equally important to: (i)
consumption that are envisioned for smart grids.
secure devices, information, and services, (ii) preserve data integ-
rity, confidentiality and authenticity, and (iii) ensure very high
availability of electricity provision. Then, the purpose of this survey 2.1. Design principles and structure of traditional electricity grids
is to provide a general reference architecture of the smart grid
communication system and its major components. We utilize our The essential purpose of an electric grid is to deliver electricity
reference model to identify basic system requirements and key to customers, which are the termination points of the power distri-
technologies, with special focus on communication technologies bution system. In the early days of electricity distribution, electric
and protocols, data management services, autonomic control func- grids were isolated systems in which electric power was produced
tions and security mechanisms. Furthermore, this survey provides by small generators using direct current (DC). However, DC-based
a comprehensive and up-to-date survey of state-of-the-art solu- electricity could only be transmitted over short distances due to
tions for each of the above research challenges, and we discuss
Table 1
advantages and shortcomings of each proposed technique. Finally, Abbreviations.
we also identify main open issues and future research directions.
AMI Advanced metering infrastructure
It is important to point out that other surveys exist targeting
AMR Automatic meter reading
various aspects of smart grids. There are a number of positional pa- BAN body area network
pers that focus on explaining what makes an electric grid a smart DER Distribute Energy Resources
grid. Those papers have the purpose of setting a common back- DG Distributed generation
ground and vision about smart grids, and to specify roadmaps DR Demand response
FAN Field area networks
and guidelines for the development of smart grids [1,2,7–10]. EMS Energy management system
Other surveys have investigated the communication networks for EV Electric vehicle
smart grids in terms of requirements, enabling technologies and HAN Home area network
possible network architectures [11–18]. In [19] the research chal- HMI Human machine interface
IED Intelligent electronic device
lenges for smart grid communications are analyzed from an indus-
P2P Peer-to-peer
trial perspective in terms of interoperability, scalability and PLC Programmable logic controller/ Power line communication
security. There are also surveys that focus exclusively on cyber- PMU Phasor measurement unit
security issues for smart grids [20–24], smart grid standardization RTU Remote telemetry unit
[25], and other aspects of smart grids, such as future utility control SCADA Supervisory control and data acquisition
SDO Standards developing organization
centers [26] or future technologies for the transmission grids [27]. SMG Smart micro-grid
The work in [28] provides a detailed overview of routing solutions V2G Vehicle-to-grid
for smart grid communications. The survey in [29] focuses on three VPP Virtual power plant
major systems of smart grids, namely the smart infrastructure sys- WAN Wide area network
WASA Wide-area situational awareness
tem, the smart management system, and the smart protection sys-
WLAN Wireless local area network
tem. According to the view in [29] the smart infrastructure system
 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
power line losses. With the advent of alternating current (AC) at
the end of the 19th century it was possible to distribute electricity
over long distances using high voltages more efficiently than with
DC. Then, rapid industrialization, urbanization and economic
development required the deployment of large-scale infrastruc-
tures for electricity delivery, which were operated through na-
tional monopolies. The structure of those national electric grids
has remained (almost) unchanged till nowadays, and it consists
of three main components: (1) generation, (2) transmission and
(3) distribution [2].
In a traditional power grid, the generation sub-system relies on
a small number of large power plants using conventional (coal, oil,
natural gas, and nuclear) resources to produce electricity. Then,
high-voltage transmission lines, which form the transmission net-
work, are used to transfer electricity across long distances from
power plants to electric substations. A substation includes trans-
formers to change voltage levels from high transmission voltages
to lower distribution voltages. Furthermore, substations perform
several other important functions, such as grid protection and
power control. Substations, medium- and low-voltage power lines,
and electric meters form the distribution network. It is important
to point out that a power system must be engineered such that
electricity production always matches electricity demands. This
implies that any change in power demands should be accommo-
dated through an equal change in power supply. To deal with
unexpected peak demands special power generators are deployed
in existing power grids, which have very short start-up delays.
Note that battery technologies are still expensive and inefficient,
thus energy storage systems cannot be employed on a large scale
to mitigate the impact of sudden load changes and fluctuations.
Even if electric utilities keep generation capacity in reserves that
can be accessed quickly, electricity imbalances are still possible
for a number of reasons, and existing power systems may suffer
from grid instabilities and power outages. Furthermore, as power
grids are complex highly interconnected systems, a failure at one
location can easily trigger a cascade effect that could result into re-
gional blackouts, such as in the case of the Northeast US and Can-
ada blackouts of 1965 and 2003 [30].
To improve the reliability of electricity provision and to support
distribution automation, in the 1960s electric utilities started inte-
grating Supervisory Control And Data Acquisition (SCADA) systems
in their grid infrastructures. SCADA is not a specific technology but
Fig. 1. An example of a SCADA-based control system.
1667
generally refers to computer-based centralized systems imple-
menting control applications for industrial processes. As shown
in Fig. 1, a SCADA system usually consists of the following
components:
 Remote telemetry units (RTUs): microprocessor-controlled
devices responsible for: (i) interacting with sensors, (ii)
converting sensor readings to standard data formats, and
(iii) delivering sensed data to monitoring stations;
 Programmable logic controller (PLCs): minicomputers that
are used as field devices for process control and machine
automation (e.g., to open or close circuit breakers at
substations);
 A supervisory computer-based system that is composed of
several remote units and a master station, and it is used to
collect the data from RTUs, perform data analysis and send
commands to PLCs;
 Databases for storing historical data, measuring trends and
deriving forecasts;
 Human–Machine Interfaces (HMIs) to present a simplified
representation of the system and its status to a human
operator, who can make supervisory decisions;
 A communication infrastructure connecting the supervi-
sory system to the other SCADA components.
Early SCADA systems were installed in substations and trans-
mission networks, and they were using dedicated point-to-point
communication links (e.g. telephone lines or optical fibers) to con-
nect the remote stations with the utility control center. Most ven-
dors of SCADA products had developed simple proprietary
communication protocols to connect the master stations with RTUs
and PLCs [31]. However, as SCADA solutions were widely deployed
by electrical utilities to control and protect the power grid, some of
those protocols, such as Modbus [32], emerged as standard de facto
recognized by all major SCADA vendors, while power engineering
standardization bodies started to release public standard protocols,
such as IEC 61850 [33]. Furthermore, the SCADA centralized archi-
tecture has also evolved over time, allowing more distributed pro-
cessing and control, and the interconnection of different SCADA
systems through wide-area networks [3,31]. However, as ex-
plained in following sections, SCADA systems do not appear suit-
able for implementing the fully decentralized and autonomous
control functions demanded by future smart-grid applications.
2.2. Key challenges for existing electric grids
From the beginning of 1990s we are witnessing a rapid transfor-
mation of existing electric grids, which is driven not only by tech-
nological innovations but also by economical, regulatory and
societal factors. Those changes impose new operational scenarios
and technical challenges to power systems, which we outline in
the following.
First of all, the soaring demand for new power supplies, and the
increasing public awareness of the need of more sustainable
sources of energy, are promoting the use of renewable energy re-
sources (e.g., wind and solar technologies) for power generation
[34]. According to various reports [35,36], half of the estimated
194 gigawatts (GW) of new electric capacity added globally during
2010 was derived by renewable resources. Although most of the
electricity produced by renewable energy sources still comes from
large-scale facilities (e.g. wind farms, solar parks, and biomass
power plants), situation is rapidly changing with the proliferation
of small-scale distributed generators using renewable resources.
Many argue that there might be many advantages in the distrib-
uted generation (DG) model [37]. For instance, DG ensures that
electricity generation is located near the place where energy is
1668 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
used, reducing power losses and congestion on transmission lines
[38–40]. However, as explained in Section 2.1, the existing distri-
bution networks are not designed to handle power flowing from
end users to substations, and this may negatively impact on distri-
bution stability [41]. Most importantly, renewable energy re-
sources are intermittent and highly variable, and the uncertainty
in energy supply can cause reliability problems or power quality
degradations (e.g., undesired voltage fluctuations) [42,43]. This
necessarily requires more sophisticated coordination and control
techniques than those supported in current electric grids [44]. Re-
cently, utility companies have started introducing automated me-
ter reading (AMR) systems in their distribution networks to
remotely collect data from the meters (e.g., consumption records,
alarms) at customers’ premises [45,46]. However, AMR systems
are typically designed using simple one-way communication infra-
structures that do not allow pervasive control of an electric grid
[15].
Another factor that is contributing to the uptake of DG technol-
ogies is the deregulation of energy markets [47]. More specifically,
power systems are no longer national monopolies, but there are
independent power producers selling electricity to utility compa-
nies, and independent operators maintaining and controlling re-
gional transmission and distribution networks. Then, electricity
prices are determined in an electronic auction market according
to demand and supply principles [48]. At least two types of elec-
tricity markets exist, which are regulated by different rules: (i)
the wholesale electricity market for trading large amounts of en-
ergy between generators, system operators and retailers, and (ii)
the retail electricity market, in which electricity retailers sell en-
ergy to end users [49]. Following the dynamics of those markets,
energy costs fluctuate, and prices raise as demands increase. The
first consequence of this competitive market structure is that cur-
rent power systems have increasingly ‘‘meshed’’ topologies,
emerging from the interconnection of several smaller grids. Fur-
thermore, energy trading across regional power grids is causing
uncertainties in energy delivery that current electric grids are not
well suited to handle. Finally, in liberalized market environments,
small electricity consumers can exploit DG technologies to become
potential producers. However, this requires improved system flex-
ibility to preserve the reliability and stability of the power system
[42].
The desire of reducing the environmental impact of our lifestyle
is also fostering the market penetration of electric vehicles (EVs).
However, the mass adoption of EVs is not without challenges for
electric grids [7,50]. On the one hand, EVs represent new mobile
loads for the power system, which can significantly alter typical
patterns of energy usage in households, while causing a dramatic
soaring of electricity demands. For instance, the simultaneous
charging of several EVs located in the same area can easily deter-
mine unexpected peak loads and rapid fluctuations of power de-
mands in different parts of the electric grid [51]. Therefore, new
management capabilities are needed to regulate the recharging
process of EVs with the objective of flattening aggregated power de-
mands and mitigating load imbalances in power systems [52,53].
2.3. The next-generation electric grid
As introduced in Section 1 next-generation electric grids are
commonly denoted as smart grids. Although there are slightly dif-
ferent views about the ultimate model of a smart grid, a consensus
is forming about the new technologies and paradigms that are
essential for the successful deployment of a smart grid, namely:
ð1Þ advanced metering infrastructure, ð2Þ distributed energy re-
sources, ð3Þ smart micro-grids, and ð4Þ vehicle-to-grid technologies
[1,2,7,9,10,54]. Fig. 2 provides a schematic view of the smart grid
reference architecture used for the following discussion.
2.3.1. Advanced metering infrastructure (AMI)
As previously observed SCADA and AMR systems were the first
attempts to introduce simple digital communication capabilities in
power systems. However, smart grid applications will need perva-
sive and real-time control of each grid component and not only of
smart meters and substations. For this reason, a smart grid should
incorporate a pervasive and scalable two-way communication
infrastructure to enable more distributed command-and-control
functionalities. One of the most important components of this
communication infrastructure is the advanced metering infrastruc-
ture (AMI) [55], which will be used to interconnect the smart me-
ters (i.e., electricity meters that incorporate networking and data
management functionalities) installed at end customers’ premises
with other control systems and data aggregators (see Figs. 2 and 5
for illustrative examples). Then, AMI systems can contribute in sev-
eral ways to the realization of the smart grid vision. First, electric
utilities can use AMI as data acquisition networks to monitor: (i)
power quality, (ii) how much electricity is produced/stored by
DER (distributed energy resources) units, and (iii) power consump-
tion of household appliances. This large amount of metering data
can be exploited to proactively identify failure conditions and
anomalies and to take appropriate countermeasures, or to imple-
ment sophisticated techniques to regulate electricity usage pat-
terns (e.g., dynamic pricing or scheduling of residential loads). In
a more general view, AMI networks can be foreseen to interconnect
not only smart meters but also a variety of intelligent electronic de-
vices (IEDs),1 which will be massively dispersed within smart grids.
Communication architectures and technologies suitable for AMI will
be analyzed in details in Section 5. It is also important to point out
that AMI will allow smart grids to collect a huge volume of hetero-
geneous data from a large number of sources. How to efficiently
aggregate, store and analyze this data is the subject of intensive re-
search (see Section 6 for an overview on such body of work).
2.3.2. Distributed energy resources (DER)
As discussed in Section 2.2, in existing power systems it is
becoming increasingly common a more distributed generation of
electricity. This trend is rapidly gaining momentum as DG technol-
ogies improve, and utilities envision that a salient feature of smart
grids could be the massive deployment of decentralized power stor-
age and generation systems, also called distributed energy re-
sources or DERs. We have already mentioned that smart grids
will facilitate the integration of small-scale renewable energy
sources (e.g., solar panels in residential applications), which will
help to reduce demand for fossil-fuel power plants and to increase
supply redundancy [34]. Distributed energy storage is widely rec-
ognized as a key enabler of smart grids for its role in complement-
ing renewable generation by smoothing out power fluctuations
[56,57]. For instance, surplus energy can be stored during condi-
tions of low demand and supplied back during periods of heavy
load. Moreover, energy storage systems generally have a quicker
response than conventional power generators, and they can be
used to increase system reliability [58,59]. However, how to coor-
dinate diverse DER technologies, which have different capabilities
and characteristics (e.g., energy capacity and time response func-
tions), is not well understood yet.
2.3.3. Smart micro-grid (SMG)
As shown in Fig. 2, a smart micro-grid (or micro-grid for brev-
ity) is a single, autonomous, self-sustainable power system formed
by an interconnection of distributed energy resources, which
serves various electricity customers (e.g., residential buildings,
1
IED is a term used to indicate microprocessor-based controllers of power system
equipment that are provided with advanced monitoring and communication capa-
bilities. IEDs are expected to replace RTUs and PLCs of SCADA systems [3].
 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
Fig. 2. Reference model of a smart grid.
commercial premises and small industries) located near one an-
other [60,61]. A micro-grid should be able to either operate inte-
grated with the utility grid or disconnected (i.e., islanded) from
the distribution system. This also implies that a micro-grid should
have its own management system to support the control functions
needed to autonomously regulate electricity flows [62], as well as
to participate into the energy market for electricity trading [63]. It
is important to notice that the micro-grid concept is not totally
new and there are already many practical examples of micro-grid
applications, such as industrial micro-grids, which provide high-
quality and reliable power to large industrial loads, or utility mi-
cro-grids, which serve loads in either densely populated urban
areas or rural regions [64]. However, we can expect a proliferation
of SMGs in future power systems, and in particular of customer-
driven micro-grids, which will be established though contractual
agreements among residential customers [57]. On the one hand,
customer-driven (or community) micro-grids appear as the most
suitable technology to optimize the deployment of DERs. On the
other hand, micro-grids can also represent the best approach to
tackle the complexity of pervasive deployment of intelligent and
distributed control functions in electric grids. Indeed, each micro-
grid can be seen as a smart grid at a much smaller scale. Thus,
many experts believe that a smart grid can progressively emerge
through the peer-to-peer interconnection of an increasing number
of micro-grids [2,65,66].
2.3.4. Vehicle-to-grid (V2G)
As pointed out in Section 2.2, the increasing use of EVs pose
new challenges to electric grids because electricity demands will
rise significantly. Thus, a considerable body of work is concentrat-
ing on defining control strategies to distribute spatially and tempo-
rally EV charging in order to avoid peak loads and to optimally
utilize grid capacity [52,67,53]. On the other hand, EVs can also
1669
offer many advantages to smart grids. Specifically, many studies
have shown that during the day cars are parked most of the time,
thus a large number of EVs can be assumed to be connected to a
power socket at any given point in time [68]. Consequently, when
EVs are plugged into the electric grid their batteries can be used as
backup energy storage systems. For instance, EVs can supply back
part of their stored electric power to stabilize the electricity pro-
duced by intermittent renewable energy sources. This new
power-generation paradigm is generally known as vehicle-to-grid
(V2G) technology [54,69]. To some extent V2G technologies can
be seen as a special case of DERs with the additional complexity
that power sources are mobile [9].
3. Smart grid applications
The smart grid vision entails innovative services and applica-
tions in addition to technological transformations. In the following
we summarize the salient features of four major smart grid appli-
cations, which are also illustrated in Fig. 2. This is useful to identify
the key requirements for smart grid communication systems.
3.1. Wide-area situational awareness (WASA)
One of the most important applications that smart grids should
support is real-time wide-area situational awareness (WASA), which
is defined as the ability to build a high-resolution description of the
current state of the power grid over a wide area. Then, this infor-
mation can be analyzed, for instance to predict the evolution of
the power grid state under different operational conditions and en-
ergy control strategies [26,70–72].
Intuitively, WASA applications rely on a pervasive monitoring
infrastructure to collect real-time data from widely dispersed
sensors. The WASA monitoring system is expected to incorporate
1670 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
diverse types of sensors, including sources of non-electrical data,
such as weather information that is used to predict changes in
wind and solar power generation. Specifically, synchronized mea-
surement technologies are emerging as an essential enabler for
WASA applications, especially for power control and protection
[70]. The most advanced type of sensors based on this technology
are the Phasor Measurement Units (PMUs), which provide syn-
chronized, real-time and high-resolution (up to 60 samples per
second) measures of voltage and frequency parameters for the
transmission lines they are connected to [73]. Note that many sys-
tem operators have already started deploying PMUs in their trans-
mission networks [74,75]. It is also important to point out that
WASA applications impose stringent latency requirements, and
communication delays are the dominant component in the total
system delays. Thus, special attention should be dedicated to the
proper design of a communication infrastructure that is capable
of ensuring delay guarantees.
Another key component of a WASA application is the set of tools
that are used to analyze the huge amount of raw and heteroge-
neous data received from multiple sources that exist in the power
grid. Specifically, simulation tools suitable for large-scale power
systems are needed to identify trends, diagnose undesired behav-
iors and to detect in advance potential vulnerabilities [1]. How to
combine simulation-based analysis of the power system with
model-based analysis, with the objective of improving accuracy
and convergence times of state estimation, is an important re-
search topic [26]. Finally, visualization tools should be developed
to present the output of data analysis in an effective and flexible
manner, e.g. by combining geospatial information with electric
information [76].
3.2. Energy management systems (EMS)
As observed in Section 2.1, in current electric grids, manage-
ment functionalities are supported by utility control centers, which
are typically implemented through distributed SCADA systems [3].
However, it is widely recognized that in smart grids it will be nec-
essary to adopt a more decentralized control model in which mul-
tiple autonomous and independent energy management systems
(EMSs) cooperate to achieve desired control objectives by exploit-
ing the information collected by WASA applications. Furthermore,
EMSs should be capable of operating on different parts of the grid
and at different scales [9]. For instance, there will be home EMSs
(HEMSs) to monitor and control the electricity usage in house-
holds, and building EMSs (BEMSs) coordinating multiple HEMSs
to maximize the energy efficiency of entire buildings [77]. At a lar-
ger scale, there will be EMSs providing advanced management and
control services for substations, micro-grids [63] or other subsys-
tems of the power grid [44,78]. In general, several technologies be-
yond classical SCADA solutions can be considered to implement
those EMSs, including service-oriented architectures (SOA), grid/
cloud computing, multi-agent systems, etc. In particular, in Sec-
tion 6.3 we will focus our attention on the large body of research
work dedicated to the development of multi-agent control systems
for smart grids.
3.3. Demand response (DR)
The term demand response (DR) is used to indicate a variety of
mechanisms that smart grids will utilize to dynamically shape
the electricity consumption of end users in response to energy sup-
ply conditions with the objective of improving the efficiency and
reliability of electricity provision [49,79]. As an example, signifi-
cant savings in power generation costs can be achieved by flatten-
ing out peak demands. Furthermore, approximately 20% of the
total power generation capacity is deployed as backup reserve for
coping with peak demands but it is in use only 5% of the time [66].
DR applications could help to obtain a more efficient utilization of
the existing energy resources. The approach most commonly pro-
posed for implementing DR applications is dynamic pricing. Basi-
cally, by adopting variable electricity tariffs the utility companies
can incentivize customers to lower peak demands and smooth de-
mand profiles by: (i) shifting some power-demanding household
activities (e.g., dishwashers) to off-peak periods, (ii) temporarily
changing the settings of energy-demanding appliances, such as
heaters and air conditioners, or (iii) activating residential power
generators to supplement the electricity provided by the distribu-
tion grid [49,80]. In the long term the expected potential benefits
of DR would be to reduce overall electricity prices, and to signifi-
cantly improve grid reliability and stability through a better
spreading of electricity demands [79,81]. It is also useful to point
out that different control approaches can be used to achieve DR
objectives [77]. On the one end of the spectrum of proposed strat-
egies there are pure customer-side techniques in which individual
houses use real-time prices to make independent decisions about
their energy usage profiles [82]. On the other hand of the spectrum
there are utility-driven techniques, commonly known as direct load
control (DLC) mechanisms, which allow electric utilities to sche-
dule power consumption of residential appliances [83,84]. In gen-
eral, the former approach requires more intelligent HEMSs, while
the latter approach is easier to implement but it can suffer from
scalability issues.
3.4. Virtual power plant (VPP)
It is expected that the proliferation of DG technologies in smart
grids could lead to a new paradigm for power generation, called
virtual power plant (VPP). More specifically, a VPP consists of a large
cluster of distributed power generators co-located with each other
in a single site, which are jointly managed and controlled. A VPP
produces a total capacity similar to the one of a conventional
power plant [85]. On the other hand, a VPP appears to the electric-
ity markets and the grid system operators as a variable-size power
plant ensuring more flexibility than conventional power plants
[86–88]. For instance, a VPP is expected to be able to react to
changes in customers’ load conditions much faster than traditional
power plants [61]. In addition, the VPP approach could allow an
easier integration of intermittent renewable resources into grid
operations, because it implements the control functions needed
to deal with power fluctuations. Finally, many studies argue that
the VPP model provides an effective aggregation technique to con-
trol fleets of EVs as a single entity, thus facilitating the integration
of V2G services into smart grids [69,89,90]. However, a VPP is also
a complex system and the development of suitable energy man-
agement systems that can be used to implement the VPP model
is still an open issue.
3.5. Smart grid standardization efforts
There is a general consensus that standardization is one of the
key issues in the design of smart grids [25,5,91,92]. Indeed, the
adoption of inter-operability standards for the overall system is a
critical prerequisite for making the smart grid system a reality.
However, a smart grid is a complex system that requires different
layers of interoperability. For instance, standards for smart meters,
smart devices, charging interfaces with electric vehicles are essen-
tial to facilitate market penetration of new smart grid products and
services, as well as seamless interoperability between them. Simi-
larly, all smart grid applications illustrated in previous sections re-
quire exchanges of information, for which interoperability
standards are needed. Finally, a smart grid consists of many differ-
ent domains and actors (i.e., generation, transmission, distribution,
 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
markets, operations, service provider, and customer), and standard
interfaces are needed among them.
There are many internationally recognized professional associ-
ations, e.g, Institute of Electrical and Electronics Engineers (IEEE),
Internet Engineering Task Force (IETF) or International Electrotech-
nical Commission (IEC), international standardization bodies, e.g.,
International Telecommunication Union (ITU), regional standardi-
zation organizations, e.g., National Institute for Standards and
Technology (NIST), and industrial alliances, e.g., ZigBee, IPSO and
HomePlug, working towards the development of standards for
smart electricity systems. However, each entity typically targets
a different market sector or application scenario. For example, IEEE
focuses mainly on standards that cover media-related interopera-
bility problems (i.e. message delivery over different communica-
tion technologies), while the activity of the IETF mainly covers
interoperability issues related to transport and application areas.
It is also evident that electrical industries can leverage many Inter-
net-related standards to build a highly interoperable information
and communication infrastructure for smart grids [91,92]. In Sec-
tion 5 we will explain the details of some of the most important
communication standards that can be applied to the smart grid do-
main, while in the following we focus on the most representative
standards for smart grid applications.
Within the IEC there are many Technical Committees (TCs) that
are working on standards for smart grid applications, such as TC
57, which prepares standards for equipment and systems used in
the control, protection and automation of power systems, includ-
ing EMS and SCADA [93], or TC 13, which prepares standards for
smart metering systems, for electrical energy measurement, and
customer information and payment [94]. Up to now, more than a
hundred IEC standards can be identified relevant to the smart grid
[95]. In these core standards, IEC/TR 62357 describes a SOA-based
reference architecture for power system automation and provides
a framework to illustrate the interdependencies between all the
existing object models, services and protocols in TC 57 [96]. At
the bottom level of this architecture there is the IEC 61850 family
of standards, which specify communication networks and systems
for power utility automation, with a special focus on substation
automation [33]. Note that the abstract data models defined in
IEC 61850 have been mapped to a number of protocols that can
run over generic TCP/IP networks and/or switched Ethernet. Other
series of core standards define a Common Information Model (CIM)
to standardize the exchange of information between different clas-
ses of applications, such as Energy Management related applica-
tions (IEC 61970 family [97]) and electrical distribution systems
(IEC 61968 family [98]). Data exchange for meter reading, tariff
and load control is specifically addressed in the IEC 62056 series
of standards [99]. Finally, security aspects for the TC57 series of
protocols are handled in the standard IEC 62351.
IEEE is also very active in the smart grid sector and it has re-
cently launched a new project called P2030 to establish a frame-
work for achieving smart grid interoperability [100]. Up to now,
nearly a hundred standards have been released or are under devel-
opment, which cover different issues relevant to the smart grid,
ranging from monitoring and control applications to communica-
tions over power lines [101]. Of particular importance is the IEEE
1815 standard, which has ratified the Distributed Network Proto-
col (DNP3) standard for communications in electric power systems
[102]. DNP3 plays a crucial role in modern SCADA systems, because
it is primarily used for reliable and secure communications be-
tween control centers and RTUs or IEDs. In addition, interoperabil-
ity standards are under development to support a transparent
mapping between IEEE 1815 and IEC 61850. With regard to mon-
itoring applications, it is also important to mention the IEEE C37
family of standards, which defines communication protocols for
real-time PMU data exchange [103]. The series of standards IEEE
1671
1547 provide guidelines criteria and requirements for intercon-
necting distributed energy resources to electric power systems,
including standards for monitoring, information exchange and
control [104]. Finally, IEEE is actively working towards standards
for power line communications considering both broadband and
narrowband technologies through the IEEE 1901 [105] and IEEE
1901.2 working groups, respectively.
4. Reference model of communication systems for smart grids
In the vision presented so far, and widely accepted by utility
companies, regulators and academia, the smart grid architecture
requires the integration of the energy infrastructure, which is
responsible for electricity generation, delivery and consumption,
with a communication system, which provides support for auto-
mated and distributed monitoring, management and optimization
functions [5]. The communication system entails different technol-
ogies, components and services, and in this survey we analyze two
of its major parts:
Communication infrastructure. The communication infrastruc-
ture (or network) is responsible for providing the connectivity ser-
vice among individual electric devices or entire grid sub-systems. In
the context of smart grids, the key priorities of this communication
network are: (a) to ensure reliable and real-time data collection
from an enormous number of widely dispersed data sources, and
(b) to support the various communication services (e.g., multicast
and group communications) that are needed by power control
applications to distribute commands and configuration instructions
in the power system. As explained in following sections, this com-
munication infrastructure is envisioned as a collection of intercon-
nected networks that will be structured into a hierarchy of at least
three main tiers or domains: (1) local area networks for the access
grid segment and the end customers, (2) field area networks for the
distribution segment, and (3) wide area networks for the utility
backbone. A variety of technologies, network topologies and com-
munication protocols are considered for each of these categories
[14,16].
Middleware platform. The middleware is a software layer run-
ning above the communication network, which provides commu-
nication and data management services for distributed
applications, as well as standard interfaces between applications
and smart grid devices. Different types of middleware solutions ex-
ist that differentiate from each other for the set of abstractions and
programming interfaces they provide to applications, such as dis-
tributed objects, event notifications, distributed content manage-
ment, synchronous/asynchronous communication functions, etc.
[106]. Furthermore, middleware is increasingly used to create
peer-to-peer (P2P) overlays, i.e., distributed systems in which de-
vices (peers) self-organize into a network and cooperate with each
other by contributing part of their (storage, computing, band-
width) resources to offer useful services, such as data search, dis-
tributed storage, or computational intelligence [107,108]. Given
the ability of P2P technologies to scale with increasing numbers
of devices and services, several studies have proposed to use
P2P-based middleware technologies to deal with the complexity
of managing and controlling smart grids [109].
In the remaining of this survey we take a bottom-up approach
to describe the main approaches proposed in the literature to build
communication and middleware solutions suitable for smart grids.
Fig. 3 provides a scheme to guide the following discussion. We
start in Section 5.1 from the communication infrastructure by
analyzing the requirements that this component must meet in or-
der to support smart grid applications. For the sake of clarity we
categorize these requirements into two classes: (i) quantitative
requirements, which define target performance metrics for data
1672 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697

communications, and (ii) qualitative requirements, which define

the functional characteristics that this communication infrastruc-
ture should exhibit [110,14,111]. Then, we continue in Section 5.2
by overviewing the wired and wireless communication technolo-
gies that are applicable in the smart grid context, and we identify
major advantages and limitations of each solution. Furthermore,
the full specification of a communication architecture requires to
define: (a) which are the communicating entities and how they
are organized in network topologies, and (b) which are the
communication protocols used to exchange messages in standard-
ized formats and to support various communication services.
Section 5.3 deals with the first topic, while Section 5.4 deals this
the second one.
In the second part of this survey we concentrate on middleware
platforms for smart grids. In particular, in Section 6.1 we focus on
middleware techniques specifically designed for supporting data
management services that are able to meet the scalability, reliabil-
ity and real-time requirements of applications that must process,
store and share data from a large number of heterogeneous data Fig. 4. Protocol stack specified in ZigBee SEP 2.0.
sources. For instance, as explained in the following sections, most
applications for grid protection require that monitored data is
delivered immediately and automatically to appropriate control smart grids must operate in a timely manner and without inter-
entities. To ensure data availability and timeliness in such large- rupting the grid services.
scale and heterogeneous systems it is necessary to adopt more ad-
vanced data management models than classical client–server or
polling techniques. Furthermore, in Section 6.3 we also explore 5. Communication infrastructures for smart grids
the key role of middleware technologies in supporting intelligent
systems, and in particular multi-agent systems, which are used There are several important issues in communication network
to provide flexible, decentralized and autonomous management design, such as:
functions in smart grids.
To conclude this survey in Section 7 we analyze a cross-layer is-  Which communication technologies should be used to
sue affecting both communication infrastructures and middleware establish links between devices?
platforms, that is security and privacy. Some challenges will not be  Which network topologies are applicable in the context of
dissimilar from those of other large-scale telecommunications net- electric grid infrastructures, and how communication tech-
works, but new vulnerabilities are due to the fact that smart grids nologies and grid geography affect the topology of the
are complex systems that integrate physical infrastructures with network?
information and communication technologies. It is also important  Which networking and transport protocols are the most
to point out that electricity must always be available in power sys- appropriate for meeting the requirements of smart grid
tems. Thus, continuous availability must be considered the most communications?
important security objective in smart grids. On the contrary, in
most telecommunications networks confidentiality and integrity There is a general agreement that it is not possible to give a
are often considered more critical than availability. As described unique answer to the above questions because smart grids will
in following sections, this implies that security mechanisms for operate in different environments and use cases. In the rest of this
section we focus on discussing the main advantages and disadvan-
tages of the different options that have been proposed in the liter-
ature, along with open issues and future research directions.

5.1. System requirements

To identify which communication technologies are suitable for

smart grids first of all it is necessary to specify the basic require-
ments that smart grid communication infrastructures should sat-
isfy. Therefore, utility companies, research organizations and
governments have elaborated several reports on the communica-
tion needs of smart grid services [110,111,14,17]. In the following
we list the most important requirements focusing on two catego-
ries: (i) quantitative requirements, which specify in a measurable
manner the target communication performance demanded by the
applications, and (ii) qualitatively requirements, which identify
the capabilities that must be supported by the communication
system.

5.1.1. Quantitative requirements

The key communication requirements for smart grid applica-
Fig. 3. Reference model of the communication system for smart grids. tions can be summarized as follows:
 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
 Latency: In general most control and protection functions in
power systems have tight delay constraints and require
prompt transmission of information. For instance, in the
case of distribution automation the IEDs that are deployed
in substations should send their measurements to data
aggregators within 4 ms, while communications between
data aggregators and utility control centers require a net-
work latency 68–12 ms [112]. Other applications are less
time-critical and can tolerate higher network delays (e.g.,
most smart meters today send their readings periodically
every 15 min). It is also important to point out that net-
work latencies depend on several factors. Therefore, com-
munications with low and stable latencies require a
communication network specifically designed to optimize
delay performance [14].
 Reliability: Critical functionalities in smart grid require very
high levels of reliability (up to 99.9999% reliability which
corresponds to a total outage period shorter than one sec-
ond per year) [110], which cannot be obtained if the under-
lying communication infrastructure does not support very
reliable communications [113]. Note that there is a number
of possible causes for network failures, including link/node
failures, routing inconsistencies, overloading, etc., which
make necessary the use of different techniques to cope
with them. However, redundancy (e.g., multiple copies of
the same messages, multiple paths used for the same infor-
mation flow, multiple servers to execute a task, etc.) will be
of paramount importance to ensure reliability in large-
scale networks. In addition, data may have different levels
of criticality, and there are messages that can occasionally
tolerate losses [114]. Thus, the communication network
should provide applications with the ability to select
between different priority levels for data transmissions.
 Data rate: The bandwidth requirements of smart meters
and other sensors that are used in electric grids are typi-
cally modest (each meter reading requires about 300 kbps).
However, bandwidth demands are growing and the band-
width used by smart meters, PMUs and other IEDs will
probably be in the range between 10kbps and 100kbps
[110]. Then, the data rate of the communication channel
can become a serious concern, especially in the utility core
backbone, due to the high number of IEDs that are expected
to be connected to smart grids. It is also interesting to note
that many studies argue that it will be necessary to over-
provision the communication infrastructure in terms of
channel bandwidth to ensure low transmission delays
and reduce packet losses on transmission buffers [16].
5.1.2. Qualitative requirements
The key capabilities for smart grid communications can be sum-
marized as follows:
 Scalability: A smart grid can involve millions of users and
even more devices. Thus, scalability is probably one of the
most intuitive requirements for the smart grid communica-
tion system. However, there are different types of
scalability, like load scalability (i.e., the ability of the com-
munication system to easily handle an increasing amount
of data traffic or service requests) or geographic scalability
(i.e., a network that is deployable in a wide range of sizes
and configurations). Similarly, different measurements of
scalability can be considered, such as the size of routing
tables as the number of nodes increases, or the amount of
communication resources used by each node [115,16]. Dis-
tributed communication architectures have emerged in the
past to support Internet services in a scalable manner, such
1673
as peer-to-peer networks (P2P) [107], which could also be
applied in the smart grid context. However, the scalability
issue in smart grids is further exacerbated by the fact that
most of the grid devices will be limited in terms of storage,
computing and communication capabilities.
 Interoperability: As pointed out above, and better described
in the following sections, many different devices, commu-
nication technologies and networking protocols will be
used in smart grids. Therefore, it is essential to ensure the
interoperability between those different communication
networks. Standards and open network architectures (i.e.,
using non proprietary protocols like IP-based networks)
will play a key role for interoperability at all the layers of
the network architecture [5]. On the other hand, network-
ing elements could be introduced to provide translation
services between different standards (e.g., network gate-
ways that route packets between different networks that
use separate protocols). It is important to note that net-
work interoperability is one face t of the interoperability
problem, but application interoperability is also essential.
For instance, application interoperability requires stan-
dards to ensure that applications assign the same meaning
to exchanged messages.
 Flexibility: Flexibility of the smart grid communication sys-
tem is a multi-faceted concept. On the one hand, flexibility
entails the ability to support heterogeneous smart grid ser-
vices, which have different reliability and timeliness
requirements. On the other hand, flexibility also implies
the ability to provide different communication models.
For instance, multipoint-to-point (MP2P) communications
are important in monitoring applications, which require
to periodically and simultaneously collect status informa-
tion from a large number of sensors. Point-to-multipoint
(P2MP) communications, or more in general group-based
communications, are equally important because are used
to distribute commands and configuration instructions to
electric devices [116,14]. In summary, there is a need for
networking technologies and protocols with a high degree
of flexibility and (self-) adaptability because the same com-
munication infrastructure must satisfy the requirements of
different applications running on top of it.
 Security: A smart grid is a critical infrastructure that needs
to be robust against failures and attacks. Thus, stringent
security requirements are imposed on its communication
infrastructure in terms of resilience to cyber attacks and
protection of customers’ privacy [117]. For instance, the
communication system must ensure that devices are well
protected by physical attacks, unauthorized entities cannot
have access to the metering information, or that sensitive
data cannot be modified while in transit in the network.
Similarly, the communication system must be robust
against network attacks that aim at disrupting the commu-
nication services to damage electricity provision. Authenti-
cation, encryption, trust management, and intrusion
detection are examples of important security mechanisms
that must be supported in smart grids to prevent, detect
and mitigate such network attacks [20].
5.2. Communication technologies and standards
Communication technologies can be classified into two broad
categories: wired technologies and wireless technologies. Gener-
ally speaking wired technologies are considered superior to wire-
less technologies in terms of reliability, security and bandwidth
because cables are easier to protect from interference and
eavesdroppers. Furthermore, the equipment is generally cheaper
1674 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
compared to wireless solutions, as well as the cost of maintenance.
On the other hand, wireless networks ensures low installation
costs and flexible deployments with minimal cabling, which are
essential characteristics to rapidly provide network connectivity
over wide areas or in areas where there is not a pre-existing com-
munication infrastructure. Moreover, new approaches have been
proposed to improve the energy efficiency of mobile-connected de-
vices [118]. In addition, recent advances in broadband wireless
technologies are providing data rates and network capacities com-
parable to those of popular wired networks. For these reasons,
electric utilities are increasingly relying on wireless technologies
to build their communication infrastructures [12].
The rest of this section is dedicated to analyze the most impor-
tant wired and wireless technologies and standards that can be
utilized in a smart grid. Furthermore, a summary of the main
features of wired and wireless communication technologies are re-
ported in Table 2 and Table 3, which compare those two classes of
communication technologies in terms of: standardization activi-
ties, maximum data rates, transmission ranges, frequency bands,
and applicability scope in the smart grid communication system.
The tables also summarize the advantages and disadvantages of
each technology.
5.2.1. Wired technologies
Traditionally, wired communication technologies were
preferred by utility operators because were considered the most
reliable option for a communication network. The most important
wired technologies that are used in smart grids are:
Power line communications (PLC). PLC technologies utilize exist-
ing power cables for information exchange [55]. This allows utility
companies to use a single infrastructure for both power and data
transmission. For this reason, PLC systems have been proposed as
a cost-effective and straightforward solution to grid communica-
tion needs. As an example, the majority of AMR deployments
around the world are using PLC technologies for transmitting
metering data [46]. However, the use of power lines to provide
reliable data transmissions has to face a number of technical chal-
lenges due to the signal propagation characteristics of typical
power cables, such as high signal attenuation, disruptive interfer-
ence from other power signals, including nearby electric appli-
ances or external electromagnetic sources [119,120]. It is also
important to note that there are two major families of PLC technol-
ogies that operate in different bands and have different capabili-
ties. More precisely, there are narrowband PLC (NB-PLC)
technologies that operate in transmission frequencies of up to
500 kHz. Within this frequency range, the resulting data rates are
modest, from 1 bps to 10 Kbps up to 500 Kbps. NB-PLC can be
used on both high and low voltage lines, and trial deployments
have demonstrated that it is possible to cover very large distances
(150 km or more) [75]. The other category of PLC solutions includes
broadband PLC (BB-PLC) technologies, which target significantly
higher bandwidth performance than NB-PLC (up to 200 Mbps) by
operating over much higher frequency bands (2–30 MHz). On the
downside, the higher frequency bands that are used in BB-PLC
reduce the maximum coverage and reliability of data communica-
tions. Thus, BB-PLC is mainly considered for in-home applications.
Nevertheless, device manufacturers have recently announced new
BB-PLC modems that can support data rates of about 10 Mbps up
to distances of 8 km over transmission lines [55].
Many standards have been developed or are under specification
for the PLC systems described above. The most important indus-
trial association that provides widely-adopted technology specifi-
cations for in-home PLC systems is the HomePlug Powerline
Alliance[121]. Over the past years this standardization group has
released several standards, which have progressively enabled
increasing channel rates, starting from 4 Mbps (HomePlug 1.0) to
85 Mbps (HomePlug Turbo), and, more recently, 200 Mbps (Home-
Plug AV and AV2). It is useful to note that HomePlug AV is a quite
advanced technology that not only provides high-quality, multi-
stream networking over power lines, but it also offers several
co-existence operational modes. For instance, it is backward com-
patible with HomePlug 1.0, it enables inter-networking with
devices using the IEEE 1901 standard (also known as Broadband
over Powerline (BPL) technology) [119,105], and it is the first
standard to allow hybrid home networks, combining wired and
wireless devices [122]. Finally, the HomePlug Alliance has also
designed the HomePlug Green PHY (GP) Specification, which is a
lower data rate, lower power version of HomePlug AV, fully
interoperable with HomePlug AV and IEEE 1901 products,
which is expected to significantly reduce power consumptions
and costs.
Optical communications. In the last decades optical communica-
tion technologies have been widely used by electric utilities to
build the communication backbone interconnecting substations
with control centers [3]. The major advantages of this communica-
tion technology are: (a) its ability to transmit data packets over
distances in the order of several kilometers providing a total band-
width of tens of Gbps (by aggregating multiple individual fibers),
and (b) its robustness against electromagnetic and radio interfer-
ence, making it suitable for high-voltage environments. For in-
stance, several restoration and protection schemes have been
devised for optical grids, which can overcome simple network fail-
ures by providing backup paths [123]. In addition, a special type of
optical cables, called Optical Power Ground Wire (OPGW), com-
bines the functions of grounding and optical communications,
allowing long-distance transmissions at high data rates. Thus,
OPGWs have been used in the construction of transmission and
distribution lines [124]. It is reasonable to believe that fiber-optic
communications will play a key role also in smart grids. Recent
studies are also expanding the scope of optical communications
by proposing the use of optical fibers to provide smart grid services
directly to end customers [125,126], although the cost of fiber
installation is recognized as an obstacle for the adoption of this
technology. It is important to point out that the use of optical com-
munications in access networks, also known as fiber-to-the-home
(FTTH), is made possible by the advent of passive optical network
(PON) technologies. Specifically, PONs do not require electrically
powered switching equipment but they use only optical splitters
to separate and collect optical signals as they move through the
network. Furthermore, PONs enable a single optical fiber to serve
multiple premises in a point-to-multipoint fashion, which permits
to support network topologies suitable for access networks (e.g.,
tree-based topologies) [127]. Among the many variants of PON
technologies, Ethernet PON (EPON) has been attracting much inter-
est from grid operators because it enables the use of the standard
Ethernet communication protocol over an optical network. This of-
fers significant benefits over other PON solutions because it facili-
tates the interoperability with existing IP-based networks.
Digital Subscriber Lines (DSL). DSL generally refers to a suite of
communication technologies that enable digital data transmissions
over telephone lines. The main advantage of DSL technologies is
that electric utilities can interconnect residential users to control
centers avoiding the additional cost of deploying their own com-
munication infrastructure. On the downside, a communication
fee must be paid to the telecommunications operators maintain
the network infrastructure. Note that there are a number of DSL
variants, ranging from basic Asymmetric DSL (ADSL), which sup-
ports up to 8 Mbps in the downstream and up to 640 kbps in the
upstream, to ADSL2+ with a maximum theoretical download and
upload speed of 24 Mbps and 1 Mbps respectively. Very-high-bit-
rate DSL (VDSL or VHDSL) provides faster data transmission over
copper wires (up to 52 Mbps downstream and 16 Mbps/s
Table 2
Summary of wired communication technologies for smart grids.

Family Standards Data rate Coverage Scope Advantages Disadvantages

PLC  NB-PLC: ISO/IEC 14908–3 (Lon-  NB-PLC: 1–10 Kbps for low data-  NB-PLC: 150 km or  NB-PLC: large-scale  Large-scale communication infra-  Multiple non-interoperable
Works), ISO/IEC 14543–3-5 rate PHYs, 10–500 Kbps for high more AMI, FAN, WAN structure is already established technologies
(KNX), CEA-600.31 (CEBus), IEC data-rate PHYs  BB-PLC: 1.5 km  BB-PLC: HAN/  Physical separation from other tele-  High signal attenuation and
61334–3-1, IEC 61334–5 (FSK and  BB-PLC: 1–10 Mbps (up to small-scale AMI communications networks channel distortion

E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697

Spread-FSK)
 BB-PLC: TIA-1113 (HomePlug 1.0),
IEEE 1901, ITU-T G.hn (G.9960/
G.9961)
 non-SDO NB-PLC: Insteon, X10,
G3-PLC, PRIME
 non-SDO BB-PLC: HomePlug AV/
Extended, HomePlug Green PHY,
HD-PLC
Optical Fibers  AON: IEEE 802.3ah
 PON: ITU-T G.983 (BPON), ITU-T
G.984 (GPON) IEEE 1901, IEEE
802.3ah (EPON)
DSL  ITU G.991.1 (HDSL)
 ITU G.992.1 (ADSL), ITU G.992.3
(ADSL2), ITU G.992.5 (ADSL2+)
 ITU G.993.1 (VDSL), ITU G.993.1
(VDSL2)
200 Mbps on very short distance)
 IEEE 802.3ah (AON): 100 Mbps up/
down
 BPON: 155–622 Mbps up/down
 GPON: 155–2448 Mbps up, 1.244–
2.448 Gbps down
 EPON: 1 Gpbs up/down
 ADSL: 8 Mbps down and 1.3 Mbps
up
 ADSL2: 12 Mbps down and up to
3.5 Mbps up
 ADSL2+: 24 Mbps down and up to
3.3 Mbps up
 VDSL: 52–85 Mbps down and 16–
85 Mbps up
 VDSL2: up to 200 Mbps down/up
 Low operational costs  Disruptive interference from
electric appliances and other
electromagnetic sources
 Difficult to support high bit
rates
 Routing is complex
 Standards evolve relatively
slowly
 IEEE 802.3ah  WAN  Long-distance communications  High network deployment costs
(AON): up to 10 Km  AMI (with (much longer than DSL) (lower with PONs than AONs)
 BPON, GPON: up to FTTH systems)  Ultra-high bandwidth (suitable for  High cost of terminal
20–60 Km supporting multimedia services to equipment
 EPON: up to 10– residential customers)  Difficult to upgrade
20 Km  Robustness against electromag-  Not suitable for metering
netic and radio interference applications
 ADSL: up to 4 km  AMI  Large-scale communication infra-  Telco operators can charge util-
 ADSL2: up to 7 km  FAN structure is already established ities high prices to use their
 ADSL2+: up to 7 km  Most commonly deployed broad- networks
 VDSL: up to 1.2 km band technology for residential  Not suitable for network back-
 VDSL2: 300 m customers haul (long distances result into
(maximum rate) – data rate degradation)
1 Km (50 Mbps)

1675
 1676
Table 3
Summary of wireless communication technologies for smart grids.

Family Standards Data rate Coverage Scope Advantages Disadvantages

WPAN  IEEE 802.15.4  IEEE 802.15.4: 256 Kbps  IEEE 802.15.4: Between 10 and 75 m  V2G  Very low power consumption  Low bandwidth
 Non-SDO: ZigBee, WirelessHART,  HAN  Cheap equipment  Do not scale to large networks
ISA 100.11a (all based on IEEE  AMI  Suitable for devices with low
802.15.4) memory and computing power
 New standards provide full
interoperability with IPv6-
based networks

E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697

WiFi  IEEE 802.11e (QoS enhancements)  IEEE 802.11e/s: up to  IEEE 802.11e/s/n: up to 300 m  V2G  Low-cost network deployments  High interference since it operates
 IEEE 802.11n (ultra-high network 54 Mbps (outdoors)  HAN (unlicensed spectrum) in a very crowded unlicensed
throughput)  IEEE 802.11n: up to  IEEE 802.11p: up to 1 Km  AMI  Cheap equipment spectrum
 IEEE 802.11s (mesh networking) 600 Mbps  High flexibility, suitable for dif-  power consumption might be too
 IEEE 802.11p (WAVE - wireless ferent use cases high for many smart grid devices
access in vehicular environments)  Simple QoS support (basically traf-
fic prioritization)
WiMAX  IEEE 802.16 (fixed and mobile  802.16: 128 Mbps down  IEEE 802.16: 0–10 km  AMI  Suitable for thousands of simul-  Network management is complex
broadband wireless access) and 28 Mbps up  IEEE 802.16 m: 0–5 (optimum), 5–30  FAN taneous users  High cost of terminal equipment
 IEEE 802.16j (multihop relay)  802.16 m: 100 Mbps for (acceptable), 30–100 (reduced perfor-  WAN  Longer distances that WiFi  Use of licensed spectrum
 IEEE 802.16 m (advanced air mobile users, 1 Gbps for mance) km  A connection-oriented control
interface) fixed users of the channel bandwidth
 More sophisticated QoS mecha-
nisms than 802.11e.
3G/4G  3G: UMTS (HSPA, HSPA+)  HSPA: 14.4 Mbps down and  HSPA+: 0–5 km  V2G  Able to support tens of millions  Cellular operators can charge util-
 4G: LTE, LTE-Advanced 5.75 Mbps up  LTE-Advanced: 0–5 (optimum), 5–30  HAN of devices ities high prices to use their
 HSPA+: 84 Mbps down and (acceptable),30–100 (reduced perfor-  AMI  Low power consumption of ter- networks
22 Mbps up mance) km minal equipment  Use of licensed spectrum increases
 LTE: 326 Mbps down and  Cellular operators are launching cost
86 Mbps up smart gridspecific service  Difficult tot ensure delay
 LTE-Advanced: 1 Gbps solutions
down and 500 Mbps up  High flexibility, suitable for dif-
ferent use cases
 Use of licensed spectrum
reduces interference
 Open industry standards
Satellite  LEO: Iridium, Globalstar,  Iridium: 2.4 to 28 Kbps  Depend on number of satellites and  AMI  Long distance  High cost of terminal equipment
 MEO: New ICO  Inmarsat-B: 9.6 up to their beams  WAN  Highly reliable  High latency
 GEO: Inmarsat, BGAN, Swift, 128 Kbps
MPDS  BGAN: 384 up to 450 Kbps
 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
upstream), and on coaxial cable (up to 85 Mbps down- and up-
stream), but it can only operate over short distances (about
1.2 km). Second generation systems (VDSL2) promise to obtain
data rates exceeding 100 Mbps in both upstream and downstream
directions at a range of 300 m.
5.2.2. Wireless technologies
Nowadays, there are different technologies and standards for
wireless communications, which can be easily classified based on
their transmission ranges. In the following we overview the most
important wireless technologies that are applicable to smart grids
by going from the technology with the smallest coverage area to
the technology with the largest one:
802.15.4-based networks. The IEEE 802.15.4 technology [128] is
the reference standard that specifies the physical and MAC layers
for low-rate, low-power and low-cost wireless personal area net-
works (LR-WPANs). Indeed, the basic IEEE 802.15.4 physical layer
offers data rates of 250 Kbps over distances of about 10 m,
although alternate physical layer standards have been proposed
that allow higher communication throughput [129]. The supported
network topologies are star (single-hop), cluster-tree, and mesh
(multi-hop). In each type of topology there is a special node, called
PAN coordinator, which is responsible for managing the entire net-
work. Networks with tree or mesh topologies have also special
nodes, called routers, that relay messages by establishing multi-
hop connections between end devices and PAN coordinator. The
standard also defines two different channel access methods that
provide support for different power management mechanisms
and channel sharing algorithms. Furthermore, within the 802.11
working group, a new task group, called 802.15.4 g, has been estab-
lished to design PHY layer enhancements to legacy 802.15.4 suit-
able for smart utility networks (SUNs) [130]. However, it is also
known that IEEE 802.15.4 power management mechanism can lead
to very low packet delivery ratios if the MAC parameters setting is
not appropriate [131]. A distributed algorithm is proposed in [132]
to autonomously configure the 802.15.4 MAC layer to minimize
the power consumption while meeting the reliability requirements
of applications.
It is important to point out that the IEEE 802.15.4 standard is
the basis for many other industrial standards for monitoring and
control applications. Among these industrial efforts the most
important ones are the ISA 100.11a standard [133], the Wireless-
HART standard [134], and the ZigBee standards [135]. More specif-
ically, the first two standards are primarily designed for industrial
automation and control systems. They both use 802.15.4-based
radios but they replace the 802.15.4 MAC protocol with a colli-
sion-free TDMA-based scheme. Furthermore, they both include
additional adaptation layers to support distributed command-
and-control applications. Note that WirelessHART is not a
completely new technology but it is a backward-compatible
enhancement of the HART Communication Protocol, which is an
open standard commonly used in the automation industry for
the last 20 years. In addition, in 2010 WirelessHART has been ap-
proved by the International Electrotechnical Commission (IEC) as
an international standard (IEC 62591 [134]). Note that one of the
main advantages of IEEE 802.15.4 over other short-range radio
technologies is the very low power consumption [136]. However,
ZigBee is certainly the most widely adopted technology for LR-
WPANs in both industrial and commercial environments because
it is considered simpler and less expensive than other solutions.
Specifically, ZigBee is a specification for a suite of protocols that ex-
tend the IEEE 802.15.4 standard with additional network manage-
ment capabilities, security functions and application support
sublayers. One of the most important features of the ZigBee
standard is the definition of application profiles that allow multiple
vendors to create interoperable products. These profiles provide a
1677
description of: (a) the devices supported for a specific application,
and (b) the data formats, message types and communication mod-
els to be used by those devices. Relevant to smart grid is the ZigBee
Smart Energy Profile (SEP), which provides an interface for manag-
ing appliances that monitor, control, and automate the delivery
and use of energy [137]. At the time of the writing two SEP versions
exist, SEP 1.x and SEP 2.0. SEP 2.0 was developed in cooperation
with other standardization groups, such as IPSO (IP for Smart Ob-
jects) [138] and HomePlug [121] industrial alliances. It offers
new capabilities, such as control of plug-in hybrid electric vehicles
and apartment buildings. In addition, it integrates some IETF stan-
dards, such as 6LowPAN and ROLL [139], which will ensure full
interoperability between ZigBee networks and IPv6-based net-
works (see Fig. 4). More details on those standards are reported
in Section 5.4.1.
IEEE 802.11-based networks (WiFi). The family of IEEE 802.11
standards, also known as WiFi, is certainly the suite of wireless
communication technologies mostly used for home and local area
networking (i.e., WLANs). The main reasons of this success are: (i)
WiFi operates in unlicensed 2.4 GHz and 5 GHz frequency bands,
(ii) WiFi uses simple and flexible access schemes based on CSMA/
CA principles, and (iii) low-cost radio interfaces exist. The original
version of the IEEE 802.11 standard was released in 1997 and clar-
ified in 1999, but since then several amendments have been ap-
proved adding new features and extended capabilities. Today the
vast majority of WiFi radio interfaces are dual band and they have
the capability to transmit on the 5 GHz band using 802.11a physi-
cal mode, and also in the 2.4 GHz band using 802.11b/g/n physical
modes [140,141]. The highest data rates are supported by 802.11n,
which integrates the OFDM-based transmission schemes used in
802.11a/g with multiple-input multiple-output (MIMO) antennas
to boost the maximum data rate of 54 Mbps supported in
802.11a/g up to 150 Mbps. Given the large number of data rates
that are available, several algorithms have been proposed for dy-
namic rate adaptation [142]. Note that the transmission ranges de-
pend on a number of factors including transmission powers,
antenna types, indoor or outdoor environments, and modulation
schemes. Experimental studies indicate that reasonable outdoor
ranges can be up to 300 m for 802.11n-based radio interfaces. Fi-
nally, the flexibility of 802.11 standard allow their use in various
parts of the smart grid communication system, as better explained
in Section 5.3. On the downside, CSMA/CA access schemes are less
energy-efficient than TDMA-based schemes (e.g., 802.15.4 MAC).
Thus, various techniques have been devised to minimize the en-
ergy consumption of the WLAN interface, including power saving
modes, packet compression and aggregation, or duty cycling dur-
ing contention periods [143].
Besides 802.11n, there are three other standards in the IEEE
802.11 family that are expected to be important for smart grid
communications. The first one is the 802.11e standard [144] be-
cause it offers QoS features (e.g, traffic prioritization, scheduling
and admission control) that are suitable for delay-sensitive appli-
cations. The second one is the 802.11s standard [145], which de-
fines mechanisms to support multi-hop transmissions and to
build wireless mesh networks on top of the 802.11 physical layer.
Finally, the third one is the 802.11p standard [146], which defines
enhancements to basic 802.11 standard to support wireless access
in vehicular environments. Thus, the 802.11p standard will be one
of the key enabling technologies for V2G systems.
IEEE 802.16-based networks (WiMAX). The IEEE 802.16 standard,
commercialized under the name of WiMAX, was firstly released in
2001 to support long-distance (up to 7–10 km) broadband (up to
100 Mbps) wireless communications, especially in rural and sub-
urban areas [147]. Conceptually, IEEE 802.16 is conceived as a
complementary technology to IEEE 802.11 because it is designed
to support: (i) thousands of simultaneous users over larger areas,
1678 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
(ii) a connection-oriented control of the channel bandwidth, and
(iii) more sophisticated QoS mechanisms than the traffic categories
defined in 802.11e. On the downside, 802.16-based networks re-
quire a more complex network management and they operate on
licensed frequency bands, which make 802.16 technology more
suitable for network operators. As for 802.11, different versions
of WiMAX technologies exist. The most recent version is the
2009 release [148], which includes many advanced features such
as: (i) OFDMA, MIMO and various types of adaptive modulation
and coding schemes, (ii) support for multicast and broadcast ser-
vices, and (iii) seamless handover for nomadic users. In addition,
different types of multi-hop relaying techniques are specified in
the 802.16j standard [149] to enable larger coverage areas and
more flexible deployments. Finally, an important evolution of the
802.16 standard family currently under development is the
802.16 m amendment [150], whose goal is to provide at least
100 Mbps data throughput at high mobility (350 km/h) and 1 Gbps
at low mobility. Furthermore, 802.16 m will support handover
with other radio access technologies, including 802.11 and cellular.
Note that data rates and coverage ranges of 802.16 technologies
make them suitable for connecting large facilities (e.g., power
plants) to utility control centers, as well as to deploy AMI networks
in scarcely populated areas.
3G/4G cellular networks. One of the main advantages of public
cellular networks over other wireless communication technologies
is the larger coverage area. For these reasons, in the past, utilities
have extensively used cellular technologies, such as GSM, GPRS
and EDGE, for data communications in SCADA and AMR systems
[151,152,45]. A shortcoming of cellular data services is that they
are relatively expensive. Furthermore, cellular networks generally
provide variable throughput and latency performance, depending
on the number of other users served by the same base station.
On the other hand, cellular networks are witnessing a rapid evolu-
tion and new generations of cellular technologies supporting high-
er data rates and more sophisticated data communication services
are being developed. Today, the most widely commercialized mo-
bile cellular systems are based on the third generation (3G for
short) of cellular technologies. 3G standards are developed and
maintained by the 3GPP industrial organization. UMTS systems,
first offered in 2001, are still the most popular 3G standards.
Among the various radio interfaces used in UMTS systems, the
highest speeds are provided by the Evolved High-Speed Packet Ac-
cess standard (HSPA+) [15], which can support data rates up to
168 Mbps in the downlink and 22 Mbps in the uplink. The succes-
sors of 3G standards will be 4G systems, which are designed to en-
able mobile ultra-broadband Internet access. A candidate 4G
system under development by 3GPP is the Long Term Evolution
(LTE) - Advanced standard, a major enhancement of currently de-
ployed LTE systems [153]. The main new capabilities introduced
in LTE-Advanced with respect to previous 3G technologies include:
(i) bandwidth and spectrum flexibility, (ii) an easier handoff be-
tween different networks, (iii) better support for heterogeneous
network architectures ranging from macro-cells to femtocells,
and (iv) more advanced mobile networking capabilities. Note that
another candidate for 4G system is the 802.16 m technology,
which was described in the previous section, although it is not tar-
geting ubiquitous connectivity as LTE does.
Satellite. Satellite systems support communications with vari-
able bandwidth and latency performance using satellites stationed
on orbits at different altitudes, including Low Earth Orbits (LEO),
Medium Earth Orbit (MEO), and Geostationary Earth Orbits (GEO)
[154]. Satellite systems using high altitude orbits have the advan-
tage of not requiring tracking antennas, which are expensive. The
downside is that they are affected by higher transmission delays
(up to one second if link-layer acknowledgments are used). On
the other hand, LEO satellite systems are less expensive to deploy.
Traditionally, electric utilities have considered satellite communi-
cations for SCADA systems and other services provided in rural
or geographically remote locations, which are either beyond the
coverage of terrestrial communication networks, or it is difficult
(and costly) to reach with dedicated fibers. Recent advances in sa-
tellite systems, with the development of smaller and low-cost sta-
tions, can open up new opportunities for the use of satellite
communications in smart grids. For instance, satellite systems
can be used to provide backup communication services at critical
substations or backhaul transport services for AMI networks.
5.3. Communication network architecture
In Section 4 we have anticipated that the smart grid communi-
cation infrastructure is commonly envisioned as a hierarchical net-
work with a three-tier architecture, consisting of: (1) an access tier,
(2) a distribution tier, and (3) a core tier. The purpose of this sec-
tion is to further elaborate on this view and to present a detailed
reference model for such communication infrastructure. This refer-
ence architecture describes the communication capabilities that
should be provided to electric devices at different tiers. Further-
more, it explains how devices should be organized into network
topologies. In addition, we discuss how small-scale networks could
be interconnected to build a large-scale communication system
providing end-to-end connectivity over a regional or national area.
Finally, this reference network architecture allows us to easily map
the communication technologies described in Section 5.2 onto the
different components of the communication infrastructure (this
mapping is also reported in Table 2 and Table 3). To guide the fol-
lowing discussion in Fig. 5 we illustrate our reference architectural
model by giving an example of a possible smart grid communica-
tion network. This example has been obtained by considering the
smart grid illustrated in Fig. 2 and substituting the energy flows
with information flows.
Access tier. The communication networks deployed in the access
tier of the smart grid communication infrastructure are responsible
for: (i) enabling real-time information flows between end custom-
ers and energy management systems, and (ii) allowing a more ac-
tive role of end customers in the electricity market and the power
grid management. Therefore, home area networks (HANs) are very
important for the access tier because they can provide low-cost
solutions to support monitoring and control of electric devices that
are deployed at customers’ premises. Both low-range wired and
wireless technologies can be used to build such networks.
However, WiFi is generally considered too expensive for most
HAN devices and it consumes too much power, while ZigBee is
the de facto standard used by many manufactures to provide
low-power wireless communication capabilities to a variety of tiny
devices (e.g., sensors) [155]. Nevertheless, HAN gateways should
be equipped with multiple radio interfaces to facilitate the
integration of different classes of devices. It is also important to
point out that similar concepts can be applied to networks with
larger scales such as Building Area Networks (BANs) and Industrial
Area Networks (IANs), which will be used in the access tier to
monitor and control the electricity consumption in buildings and
industrial facilities.
The access tier must also provide connectivity services suitable
for electric vehicles, which will have the twofold roles of mobile
consumers and mobile storage of electricity. In this case wireless
communication technologies are the most natural choice for sup-
porting V2G scenarios. However, different use cases can be envis-
aged that require different networking solutions. For instance,
electric vehicles need to interact with HANs when parked at homes
or offices, and WiFi-based LAN solutions appear the best option. On
the other hand, electric vehicles can stay connected to the smart
grid communication system when on the move by using public
 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
Fig. 5. An example of an end-to-end communication infrastructure for smart grids.
3G/4G cellular networks or a WiFi-enabled roadside communica-
tion infrastructure [156]. More generally, vertical handover tech-
niques can be used by moving vehicles to switch from one
wireless network to another while maintaining seamless connec-
tivity. A survey of the main vertical and horizontal handover ap-
proaches in the literature can be found in [157,158].
Distribution tier. The communication networks deployed in the
distribution tier of the smart grid communication infrastructure
are responsible for: (i) enabling the state estimation and real-time
control of the distribution grid, (ii) supporting the interconnection
of local area networks (i.e., HANs, BANs and IANs) with the smart
grid communication backbone, and (iii) providing the communica-
tion support for implementing the data management services that
are needed to efficiently handle the large amount of data collected
in the distribution grid. As observed in Section 2.3, AMI networks
will be a key component of the distribution tier because they inter-
connect smart meters with data aggregators and control systems
deployed in the distribution grid. It is important to point out that
there is not a single communication technology that can meet
the requirements of all AMI deployment scenarios, and utility
companies are considering both wired and wireless communica-
tion technologies for building their AMI systems. Options include
conventional PLC technologies, point-to-point communications
using cellular or medium-range wireless (e.g., WiMax or WiFi)
technologies; multi-hop wireless technologies, such as mesh net-
working solutions, which can provide a flexible and easy to deploy
extension of the existing wired networks [159]. Given the
heterogeneity of communication technologies, specific technolo-
gies are needed to support internetworking and to provide
seamless service provisioning. The IP Multimedia Subsystem
(IMS) is a promising solution to address this issue, since it offers
the needed interworking environment and service flexibility for
the integration of wireless access technologies [160]. The 802.21
standard is also supporting various mechanisms that can be used
to enable the seamless interoperation between heterogeneous
technologies [161].
1679
In addition to AMI networks the distribution tier will include
specialized networks to provide reliable communications to a large
number of heterogeneous sensors and actuators that will be de-
ployed in smart grids to monitor and control power system equip-
ment (e.g., circuit breakers, feeders and substation transformers,
DER units, etc.). These networks are commonly named Field Area
Networks (FANs), or Neighborhood Area Networks (NANs) [14,16].
In some cases the communication technologies used in FANs will
not be dissimilar from the ones that are considered for AMI net-
works. However, FANs can also be considered an evolution of exist-
ing SCADA-based networks that are used for power grid protection,
and they will have more stringent real-time requirements than
AMI networks. Furthermore, elements in a FAN can be physically
distant from each other. Thus, 4G technologies (e.g., LTE) will have
a key role in most FAN deployments, while they are less important
in AMI networks [15].
Core tier. The core tier of the smart grid communication system
consists of a Wide Area Network (WAN). Such high-capacity com-
munication backbone is used to deliver the large amounts of data
collected by the highly dispersed AMI systems and FANs to remote
control centers over long distances. Various options have been con-
sidered for the deployment of a WAN meeting the requirements of
smart grid communications, such as all-IP core networks or MPLS-
based networks. However, the fundamental choice that electric
utilities are facing is between the deployment of private WANs
and the use of public data networks. Several factors are influencing
the decision of grid operators and the need of high reliability, secu-
rity and low latency are the most important ones, along with eco-
nomical affordability. As a matter of fact, a growing number of
utilities are choosing to deploy a private hybrid fiber/wireless net-
work as the backbone for their electric grid [162].
So far we have presented a general reference model of the net-
work topology for the smart grid communication system. To con-
clude this section we further elaborate on some specific areas
that are worth exploring: (i) the applicability of Internet concepts
and models to smart grid communications, (ii) AMI deployments,
1680 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
and (iii) the applicability of multi-hop wireless technologies for
smart grid communications.
5.3.1. Similarities between the Internet and the smart grid
communication infrastructure
In this section we discuss the fundamental similarities that may
exist between the architectural model of the Internet and the ref-
erence model of the smart grid communication network that we
have described above [163,164]. To recognize those similarities is
important because they motivate the adoption of Internet design
principles when designing scalable, reliable and secure networking
solutions for the smart grid [14,165,16]. For instance, both the
Internet and the power grid have emerged from the incremental
interconnection of an enormous number of (computing, sensing
and electric) devices. Both the Internet and the power grid are
highly heterogeneous and wide-area complex systems, which must
support various degrees of autonomous control at different time
scales. Finally, both the Internet and the power grid are witnessing
a transition from a structure with a clear distinction between the
core network and the access network (with almost all the system
intelligence residing in the core) to a more federated system where
the intelligence of the network (i.e., its ability to distribute, store,
or modify information and energy, respectively) can be migrated
to the periphery [166–168]. To cope with scalability, heterogeneity
and decentralization requirements, the Internet architecture has
relied on the interconnection of small-scale subnetworks, which
were then organized into a hierarchy of networks covering larger
geographic areas. As we have described earlier, such structure is
also conceived as the most suitable for smart grids in which indi-
vidual electric subsystems have the responsibility of controlling
separate geographical regions (e.g. micro-grids). Then, such electric
subsystems can be easily mapped into separated communication
subsystems interconnected with each other so as to form a hierar-
chy of communication networks [169]. Furthermore, the IP stack
has proven its interoperability and extensibility as the basis of
the global Internet over the last 30 years. Thus, Internet technolo-
gies may be seen as a promising solution for the interoperability
problem in the smart grid communication infrastructure [91]. On
the downside, there are well-known Internet problems (like secu-
rity) that may hinder a tighter integration between the smart grid
communication infrastructure and the Internet.
5.3.2. Examples of communication networks for AMI systems
The network architecture model we have outlined above pro-
vides a unified conceptual framework to support end-to-end com-
munication services in smart grids. However, in the literature there
are also several examples of specialized communication networks
that have been designed for specific components of the smart grid
communication infrastructure. In particular, the design of commu-
nication networks for AMI systems has attracted much research
interest. A full-fledged network architecture for AMI systems is de-
scribed in [170]. In that architecture the smallest communication
subsystem is the HAN, which corresponds to an individual apart-
ment. Then, a number of HANs are aggregated into a BAN, while
several BANs form a NAN. Distinguishing features of this network
architecture are the use of: (i) low-power wireless communication
technologies in the HANs, (ii) different types of gateways to control
the communications of increasingly larger areas and to facilitate
the interconnections of different subsystems, and (iii) cellular
(3G) technologies to enable communications between multiple
BAN gateways and a NAN gateway. Note that in [170] it is also
envisaged that the 3G network interconnecting the gateways is a
dedicated network separated from public cellular systems to en-
sure improved safety and reliability.
AMI networks similar to the one proposed in [170] have been
also described in [171,83,45,172]. In particular, in [171] data
management services are deployed within the NAN gateways,
which are called concentrators, to support aggregation and storing
of the information flows coming from multiple BANs. Furthermore,
a base station is deployed to serve an area containing multiple
NANs, which receives the data sent by NAN concentrators over
dedicated WiMAX channels, and forwards this information to the
regional control center over a wired network (e.g., using the Inter-
net core backbone). In general, the trend is to prefer licensed wire-
less technologies, such as WiMAX [11,83] and GSM/GPRS [45,172],
for the communications in the neighborhood area networks, be-
cause grid operators can establish their own private network
infrastructure, which makes easier to guarantee the required com-
munication reliability and QoS. On the other hand, unlicensed
wireless technologies, such as ZigBee [173,174,172,175,176] and
WiFi [177], are the main standards that are being considered for
communications within houses or buildings, although they have
been also proposed as less expensive options for supporting com-
munications in NANs [178]. As observed in Section 5.2, some of
those standards have been also recently extended to provide spe-
cific support for the interoperability of devices that will be used
in smart grids, such as the ZigBee SEP [179,137].
5.3.3. Multi-hop wireless networking in smart grids
In this section we discuss the potential role that the multi-hop
communication paradigm can have in a smart grid. Specifically, we
focus our attention on two of the most mature and consolidated
examples of networking technologies using multi-hop wireless
communications: wireless mesh networks and wireless sensor net-
works. Interested readers can find an up-to-date overview of re-
search and development in the broader area of networks based
on the multi-hop ad hoc networking paradigm in [180].
Wireless mesh networks. A wireless mesh network (WMN) is a
dynamically self-organized network in which nodes automatically
establish and maintain network connectivity through multi-hop
wireless paths. There is a general agreement that wireless mesh
networking is a key enabling technology for most next-generation
wireless networks [159]. WMNs have the potential to bring many
advantages also to the smart grid communication system and
many studies have advocated the use of WMNs in smart grids
[11,16]. For instance, a multi-hop wireless communication system
inherently provides multiple redundant paths between any pair of
communicating nodes. This eliminates single point of failures and
mitigates the occurrence of bottleneck links, increasing the com-
munication reliability with respect to conventional infrastruc-
ture-based wireless networks [181,182]. Furthermore, multiple
paths can be used to achieve a more balanced distribution of traffic
over the network [183] or to minimize the energy used by the
mesh backbone to route traffic [184]. Note that the use of self-
organizing networking technologies allows to provide self-healing
capabilities to the communication network, which is necessary to
meet the reliability requirements of smart grid communications.
At the same time, wireless multi-hop communications are a low-
cost solution to increase the coverage of existing wired networks
and to deal with the difficulties of deploying new cables in existing
facilities. Finally, as observed in Section 5.2 most popular wireless
technologies, including WiFi, ZigBee and WiMAX, are now includ-
ing amendments to their standards to provide support for
multi-hop communications (e.g, 802.11s and 802.16j). Thus, mesh
networking capabilities are expected to be supported by most
commodity products. Recently, millimeter-wave links that operate
in the 71–86 GHz frequency band have also been proposed as a
cost-effective high-speed alternative for fixed wireless mesh net-
works, with data rates up to 1 Gbps [185].
Wireless sensor networks. A wireless sensor network (WSN)
consists of autonomous sensor devices monitoring physical or
environmental phenomena, which employ multi-hop wireless
 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
communication technologies to cooperatively transfer their read-
ings to a common sink or gateway. Most advanced sensors can also
perform simple actions (e.g., close a circuit). In this case the term
wireless sensor and actuator network (WSAN) is generally pre-
ferred to denote a wireless network that is able to perform distrib-
uted sensing and acting tasks. As observed in Section 3.1, WSNs are
expected to play a key role in enabling the pervasive monitoring of
electric grids. It is important to point out that WSN is not a novel
concept. There are more than two decades of intensive research
in this field and several surveys exist in the literature that outline
existing approaches for routing and data collection in WSNs [186–
188]. In particular, energy awareness is a very important design
consideration for protocols and algorithms in sensor networks,
and the design of a WSN should consider the amount of energy
each protocol can spend to perform its tasks [189]. In addition to
energy efficiency, there are many other technical challenges to ad-
dress before successfully applying WSN technologies in the smart
grid domain. Specifically, the harsh environmental conditions of
typical power systems, such as medium-voltage substations or
power control rooms, pose significant issues to the reliability and
latency of wireless communications, which also impact on the de-
sign of communication protocols for WSNs. Most importantly, in
many classical applications for wireless sensor networks the goal
is to detect the occurrence of an event (e.g., when the ambient tem-
perature is above a given threshold), and it is sufficient that a small
subset of the sensors that have observed the same event are able to
report it to the application. This characteristic enables the use of
mechanisms such as duty-cycling and data fusion to reduce spa-
tially-correlated contention between nodes in the same neighbor-
hood, as well as to improve energy efficiency and network lifetime
[190]. On the contrary, in many smart grid applications the sensed
data of each individual sensor might be needed. For instance, this is
the case of smart metering applications, which must collect infor-
mation from each smart meter. This means that existing data col-
lection techniques for wireless sensor networks should be
modified to meet this additional constraint. Furthermore, in wire-
less sensor and actor networks there is a need for suitable coordi-
nation mechanisms between sensing and control functions to
ensure that correct actions can be taken within the strict time con-
straints of most power grid processes.
We conclude this section by pointing out that most WSNs for
smart grids will be deployed in conditions that can make very dif-
ficult to replace batteries. Furthermore, maintenance costs due to
the replacement of a large number of batteries may be excessive.
To cope with these difficulties, many studies advocate the use of
energy harvesting techniques to allow sensor nodes to take power
from the surrounding environment. Energy harvesting covers a
wide range of different technologies. Solar energy is the most obvi-
ous power source and photovoltaic cells can be used to obtain
power from it. However, this solution is not applicable indoors.
In smart grid environments electromagnetic induction seems the
most feasible technology, but there are other approaches that can
work as well. For instance, using the difference in temperature
Fig. 6. Classification of research trends in the area of routing for WMNs and WSNs in the smart grid context.
1681
with thermal energy harvesting can be effective in sub-stations,
where there is a significant temperature gradient. For sensors de-
ployed in buildings, vibrational energy can also be used. The down-
side of energy harvesting technologies is that they are still unable
to provide a sustained energy supply to support continuous oper-
ation. Therefore, new power management schemes, transmission
techniques, data aggregation and link scheduling algorithms, and
routing protocols are needed to exploit the sporadic availability
of energy [191–193].
5.4. Communication protocols
To allow communications in a distributed network formed by
devices from different vendors and using heterogeneous communi-
cation technologies, it is necessary to specify a suite of protocols
to: (i) associate node identities to network devices (naming and
addressing functions), (ii) establish network paths between nodes
(routing function); (iii) define message formats and rules to ex-
change those messages (transport function); and (iv) support ad-
vanced networking services such as broadcast, multicast, QoS
and security. A comprehensive overview of all the network proto-
cols and services that have been proposed in the literature for
smart grid communications might require a separate survey per
se. Thus, we focus on three important areas that we believe still en-
tail major research challenges: (1) routing protocols, (2) QoS sup-
port, and (3) transport protocols.
5.4.1. Routing protocols
As observed above, routing protocols are responsible for estab-
lishing communication paths among nodes in a network. As dis-
cussed in Section 5.3, in a smart grid different classes of devices,
communication technologies and network topologies will be used
depending on applications and use cases. This implies that multi-
ple routing protocols should be employed to enable smart grid
communications in the different networks forming the smart grid
communication infrastructure. It is important to point out that in
the past electric utilities preferred to avoid using routing protocols
to communicate with field devices due to potential ðaÞ real-time
violations and ðbÞ security attacks. However, as motivated in Sec-
tion 5.3 electrical grids are progressively migrating towards IP-
based network architectures. Therefore, it is expected that routing
protocols will be increasingly more important, especially in the ac-
cess tier of the smart grid communication system.
It is out of the scope of this survey to outline all the different
routing protocols that have been proposed so far for the various
network categories described in Section 5.3. Interested readers
can find a comprehensive survey of routing protocols for HAN
and NANs in [28]. On the contrary, in the following we focus our
attention on routing and data forwarding schemes for networks
using wireless multi-hop communications (i.e., WMNs and WSNs)
because we believe this is the area in which most research
challenges are yet to be addressed, especially as far as QoS support
is concerned. For the sake of clarity, in Fig. 6 we provide a
1682 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
classification of the research trends in the area of routing for smart
grids, which we have analyzed in the following.
In the literature there exists a large amount of work on routing
for ad hoc networks, as testified by the considerable number of dif-
ferent protocols that have been proposed, and the standardization
of few of them by the IETF MANET working group. Furthermore, a
number of surveys can be found on this area, which provide com-
prehensive classifications and comparisons of existing routing ap-
proaches [187,194–196]. Relevant to this survey are the studies
that are dedicated to the development of a more solid understand-
ing on the applicability of those routing approaches in the smart
grid domain, and AMI systems in particular [197,177,198]. One of
the main findings of those studies is that geographic routing strat-
egies, which use geographical coordinates of networks nodes to
optimize the path selection (e.g., selecting as intermediate relay
the node that is geographically the closest one to the final destina-
tion), seem the best schemes for smart grid communications
because they minimize routing overheads and they work particu-
larly well in static networks. For instance, the authors in [197]
consider an exemplifying PLC network used in low- and
medium-voltage distribution grids, and show how different loca-
tion-based routing strategies cope with variable link qualities
and node failures.
Another research area related to routing in smart grid networks
include the adaptation of routing schemes designed for general-
purpose WMNs to the context of smart grid communications. In
[199], the reliability of the AODV routing protocol is investigated
in a distribution grid topology that spans many kilometers. A
new multi-path routing protocol for multi-channel WMNs, called
DMMR, is proposed in [200], which tends to choose paths that do
not share network nodes to reduce the probability that multiple
paths are broken at the same time when a single node fails. In this
way, DMMR ensures stable end-to-end connections. Other works
have explored the applicability of the Hybrid Wireless Mesh Proto-
col, or HWMP, specified in the IEEE 802.11s standard [145], for
routing in WMNs that are used for the monitoring and manage-
ment of electric grids. HWMP operates directly at the link layer
and it uses MAC addresses for routing instead of IP addresses. Fur-
thermore, it is an hybrid protocol that adopts a tree-based hierar-
chical routing scheme for data transfers from network nodes to
mesh gateways, and an AODV-like routing scheme for data trans-
fers between network nodes. A known issue of HWMP is route
instability in case of link failures and the work in [201] proposes
new routing metrics to reduce route fluctuations. A multi-path
extension to the tree-based routing strategy used in HWMP is also
described in [202]. Alternative tree-based routing schemes are pro-
posed in [203,204] by employing data forwarding mechanisms in-
spired by distributed depth-first search algorithms. Furthermore,
to reduce network control overheads those schemes use data pack-
ets to detect loops and to propagate the information on failed links.
In this way, routing tables can be updated quickly and with little
overhead.
Another considerable amount of work has been conducted on
the design of routing protocols for WSNs that are able to cope with
the limitation of computing power and memory size of embedded
devices. Indeed, many of the devices installed in the last mile of an
AMI network, such as home appliances, IEDs, and smart meters
will be based on tiny micro controllers. Most of the above-de-
scribed routing protocols for WMNs are not applicable to those de-
vices because they perform complex computations and store a
large amount of routing information. For these reasons, popular
wireless standards for sensor and control systems, such as ZigBee
and WirelessHART, have implemented much simpler routing
schemes. Specifically, ZigBee specification include three routing
protocols: (1), a tree-based routing scheme for data collection at
the network sink, (2) a variant of the AODV algorithm that is used
to establish on-demand multi-hop network paths, and (3) a source
routing for communications between the sink node and the end
devices. It is also important to point out that in the last few years
there has been a growing interest from the research community in
designing optimized networking protocols for WSNs, which could
potentially match the requirements of the smart grid (interested
readers are referred to [186,187,205] for detailed surveys). IETF
had recently released a new standard specifying an IPv6-based
routing protocol, called Routing Protocol for Low Power and Lossy
Networks (RPL) [139], which is rapidly emerging as the most ma-
ture and commercially viable routing solution for large-scale AMI
systems [206]. Specifically, RPL organizes the network topology
into a tree-based structure (called Directed Acyclic Graph, DAG)
and it adopts a gradient-based data forwarding scheme to mini-
mize memory usage and protocol complexity. Given the impor-
tance of RPL applications, a few papers have addressed the
performance evaluation of RPL in different use cases [207–210].
In particular, results shown in [209] for a medium-scale outdoor
network deployment indicate that RPL ensures smaller routing ta-
bles and lower control overheads than conventional shortest-path
routing algorithms. On the other hand, the study in [210] points
out that RPL nodes may suffer from severe unreliability problems,
mainly because RPL often selects sub-optimal paths with low qual-
ity links. Other studies have focused on proposing enhancements
to the basic RPL standard. The work in [211] proposes to construct
the DAG assigning to each link in the network a cost depending on
the ETX metric, a popular routing metric initially proposed for
wireless multi-hop routing protocols [212]. In [178] extensions
to RPL are proposed to enable smart meters to automatically dis-
cover connectivity and recover from loss of connectivity. The effec-
tiveness of the local repair mechanisms used in RPL is also
investigated in [213]. However, RPL applicability in large-scale
(thousands of nodes) networks with stringent reliability require-
ments is still an open issue. In addition, the stability of DAG struc-
ture with accurate PLC channel models and real-world traffic
patterns needs to be examined.
5.4.2. QoS support
In Section 5.1 we have observed that reliability and timeliness
are key requirements for smart grid communications. However,
different smart grid applications may have different constraints
for latencies and communication reliability. For instance, in basic
metering applications a delay of few seconds when collecting me-
tered data is tolerable, while applications monitoring transmission
lines should operate on a time scale of few milliseconds. Similarly,
most grid protection applications, which involve the remote con-
trol of critical grid components such as breakers and switches, re-
quire very high levels of communication reliability to avoid power
grid instability. Consequently, the smart grid communication infra-
structure should adopt suitable mechanisms to enforce different
QoS guarantees to network flows depending on the application
constraints. In telecommunications networks QoS differentiation
is typically achieved through resource reservation and traffic prior-
itization. Specifically, various approaches can be employed to pri-
oritize important and time-critical network flows over less
critical data traffic. For instance, many MAC layers (e.g., 802.11e
and 802.16) support the specification of different traffic categories
and they use scheduling algorithms to provide bandwidth differen-
tiation [214,215]. However, MAC-based solutions are generally
limited to provide QoS guarantees on single communication links.
For this reason, there is an increasing awareness that a full-fledged
QoS-based architecture is needed to satisfy the different
requirements of smart grid applications. Several QoS-based
frameworks have been proposed for the Internet, such as Integrated
Services (IntServ) [216], Differentiated Services (DiffServ) [217], and
Multi-Protocol Label Switching (MPLS) [218], which could be applied
 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
also in smart grid communication networks. It is out of the scope of
this work to present a description of those solutions. However, we
want to remark that it is still an open issue to decide which of
those QoS-based architectures is the most appropriate for the
smart grid domain. The study in [219] proposes to use MPLS tech-
nologies to handle fine-grained bandwidth management in smart
grids communication systems, while the study in [220] discusses
how to integrate MPLS with DiffServ. A multi-service routing archi-
tecture is described in [221], which uses the DiffServ model in the
data plane, the RTCP protocol for performance monitoring and
multi-service path calculation in the control plane [221]. A detailed
and fair comparison of different QoS approaches, and their testing
in real-world deployment are still missing. Note that recently most
of the research efforts in this area have been focused on the devel-
opment of optimization frameworks to compute network paths
that satisfy multiple QoS constraints simultaneously (also known
as constrained-based routing or QoS routing), because this is an
essential feature for any communication infrastructure that aims
at guaranteeing QoS. A comprehensive survey of existing algo-
rithms for constrained-based routing can be found in [222]. Given
the heterogeneity of the smart grid, traditional methodologies can-
not be directly applied due to the requirements of high computing
and storage capabilities. Thus, new schemes are needed that can be
implemented by both powerful and resource-limited devices. Pre-
liminary results on QoS routing solutions specific to smart grids
have been developed in [223,224]. However, it is still an open
problem to understand the impact of power system dynamics on
the stability of QoS routing, or how to define QoS requirements
in the context of smart grid.
5.4.3. Transport protocol
Transport protocols are responsible for delivering data among
application processes running on separate hosts in a network
[225]. Besides simple inter-process communications there are
many other services that can be optionally provided by a transport
protocol, such as data integrity, congestion avoidance and flow
control. The simplest transport protocol in the Internet protocol
suite is UDP, which is connection-less (i.e., it does not set up a ded-
icated end-to-end connection) and it does not provide any guaran-
tee on packet delivery. However, UDP can be used tighter with the
Real-time Transport Protocol (RTP), which provides specific mech-
anisms to improve the delivery of audio and video over IP networks
[226]. For instance, RTP packets carry information (e.g., time
stamp) that allows to implement jitter compensation at the receiv-
ers. Furthermore, RTP supports data transfer to multiple destina-
tions through IP multicast. On the other hand, the most common
alternative to UDP is the Transport Control Protocol (TCP). TCP is
more sophisticated than UDP because: (i) it provides connection-
oriented communications, (ii) it uses message numbering and
retransmissions to recover packet losses and suppressed duplicate
data, (iii) it supports reordering of out-of-order data, and (iv) it
implements flow and congestion control techniques. Since the
majority of smart grid applications require reliable communica-
tions, TCP seems the natural choice also for the smart grid commu-
nication system. However, TCP does not provide guarantees on
network delays experienced by transmitted packets. In addition,
the timeouts used to detect some packet losses can cause notice-
ably delay spikes. Thus, TCP cannot adequately meet the require-
ments of smart grid communications in terms of timely data
delivery, especially for data traffic that it is inherently periodic
(e.g., metering readings, measurements from PMUs, etc.). Further-
more, data sources in most smart grid applications generate small-
sized packets at a low rate. In this case, TCP congestion control can
be ineffective and it can cause useless retransmissions of packets
and throughput degradations. For these reasons, a few studies have
been recently conducted to design either suitable modifications to
1683
TCP or totally new transport protocols in order to achieve lower la-
tency while preserving reliability in data-collection applications.
One example is the scalable and secure transport protocol that is
proposed in [227], called SSTP, which is suitable for monitoring
applications in which a large number of clients infrequently com-
municate with servers. More specifically, one of the main goals of
SSTP is scalability, and this is achieved by assuming that the server
does not continuously maintain state information for each session
(i.e., sensor device) but it encrypts the session state and it trans-
mits the encrypted information to the associated client, which
temporarily stores it. Then, the client returns to the server the en-
crypted state with his next message. Furthermore, SSTP assumes
that in-order delivery is not required as data is always time-
stamped at the sender side. Finally, a SSTP client can immediately
send a message without any delay unless its sending window,
whose size is determined by the number of clients associated to
the same server, is full. Results shown in [227] indicate that SSTP
can provide much lower end-to-end delays than TCP. An alterna-
tive approach, called Split and Aggregated TCP (SA-TCP), is pro-
posed in [228], which employs split and aggregation techniques.
More specifically, in SA-TCP each client creates a TCP connection
with a single intermediate node, called transport aggregator, which
is deployed before a bottleneck link. Then, this node aggregates all
received data and it creates another TCP connection with the final
data collector. In this way, a bottleneck can be shared between
many meters more fairly and the number of packet retransmis-
sions is reduced. On the downside, TCP splitting approaches are
more vulnerable to security attacks, and packets may suffer longer
delays. An alternative congestion control technique is proposed in
[229] by adapting the monitoring rate of smart meters to the avail-
able bandwidth. The basic idea is to formulate an optimization
problem to determine which is the amount of traffic that can be re-
duced at different locations in the network without affecting the
grid operations. It is also useful to point out that there is a large
body of work focusing on reliable data collection protocols for
WSNs (see [230,188] for survey papers on this area). However,
those transport protocols can not be applied to smart grid monitor-
ing applications without substantial modifications for a number of
reasons: (i) in smart grid each packet conveys unique information
associated to a specific meter at a given time instant, thus data
aggregation techniques cannot significantly reduce the total vol-
ume of data traffic to be delivered, and (ii) redundant deployment
is not a feasible solution for achieving reliability because smart
meters identify specific end users.
6. Middleware platforms for smart grids
Many smart grid applications will require a precise estimation
of the state of large portions of the power system in order to con-
trol and optimize the electricity delivery and usage. However, not
all smart grid applications require the same sets of data, nor the
same reporting frequency. For instance, off-line trend analysis does
not impose tight delay constraints, while control and protection
functions need real-time transmission of data (and commands).
Therefore, scalable data management systems specialized for
smart grids are needed to share this large amount of information,
and to timely deliver relevant data to applications that really need
it [231].
In Section 2.1 we have pointed out that SCADA systems already
include solutions for storing and analyzing data collected from a
large number of RTUs. However, those solutions rely primarily on
centralized databases, which do not provide the level of scalability
and flexibility needed to handle the huge increase in data storage
and processing that will occur with smart grids. On the other hand,
there is an increasing consensus among electric utilities that
1684 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
Fig. 7. Classification of research trends in the area of middleware for smart grids.
middleware technologies are needed to build large-scale data
sharing systems because they provide efficient data management
services suitable for distributed environments. In addition, middle-
ware can help to successfully handle the inherent complexity of
building robust and autonomous control applications in heteroge-
neous distributed systems [232]. For instance, existing middleware
platforms provide various optimized tools and mechanisms for
cooperatively controlling complex systems consisting of a large
number of highly interconnected and interdependent components
operating in dynamic environments. In particular, multi-agent
middleware platforms are drawing much attention from research-
ers in the smart grid area because they appear the most suitable
technology to provide self-adaptation and self-healing capabilities
to smart grids [233].
In the following sections we will outline the most important
middleware platforms proposed so far for smart grid control appli-
cations. Note that most middleware platforms are designed follow-
ing a given model, or paradigm, which describes the approach used
to manage communications and distribution of resources. Thus, to
guide the following discussion in Fig. 7 we also provide a classifi-
cation of the related work based on three main categories: (i) mid-
dleware services for data management; (ii) object-oriented
middleware, and (iii) multi-agent systems.
6.1. Middleware services for data management
One of the earliest examples of a full-fledged middleware plat-
form supporting data management for smart grid applications is
known as GridStat [234–236]. More specifically, GridStat provides
a publish/subscribe communication model in which a grid object
can be either a producer or a consumer of data. The consumers
can declare to the middleware system their interests in one or
more types of data (e.g., a data aggregator may be interested in
receiving at a higher frequency measurements from a set of smart
meters), while the producers simply announce the availability of
their data. Then, the middleware platform is responsible for dis-
tributing the relevant data to the subscribed consumers [237]. It
is important to observe that the communication primitives pro-
vided by the middleware simplify the interactions between pub-
lishers and subscribers because subscribers do not need to know
the identity nor the location of publishers, and vice versa. This
decoupling is one of the main features that facilitate large-scale
deployments. In the case of GridStat, the publishers are either
sources of measurements and control settings (status variables),
which are periodically updated by the middleware, or sources of
alarm conditions (status alerts) that require immediate attention.
On the other hand, subscribers are application programs that use
these variables and alerts. Then, the GridStat middleware provides
the services needed to control the way information is disseminated
and accessed. For instance, a directory service is implemented to
allow subscribers to find particular status variables of interest.
Moreover, the publishers have the ability to inform the middle-
ware infrastructure about the semantic of the status variables they
monitor, such as type, availability frequency, and priority. Simi-
larly, the applications can specify their QoS requirements for data
access, such as the maximum delay for receiving status updates,
the minimum update frequency, or the maximum number of up-
dates that can be lost per unit time. In order to meet these QoS
requirements directly at the middleware layer, the GridStat archi-
tecture requires the deployment of two classes of special devices
within the smart grid:
 status routers, which establish redundant paths between
the middleware peers to support QoS-aware multicast of
periodic status updates;
 brokers, which are organized into a hierarchy of manage-
ment entities that negotiate with the subscribers their ini-
tial QoS requirements in order to make them less stringent
if necessary. In addition, brokers also handle the resource
allocation in the communication network to ensure that
the selected paths meet the QoS requirements of each sub-
scription request.
A Java-based implementation of key mechanisms in the Grid-
Stat framework is described and experimentally evaluated in terms
of delivery latency and throughput in [238]. Although GridStat of-
fers a flexible and robust communication framework, it also suffers
from a series of limitations. First of all, it only supports the publish/
subscribe communication model, which is appropriate for deliver-
ing status updates or alerts, while alternative communication ser-
vices (e.g., remote procedure calls for invoking commands) are
needed in other use cases. In addition, GridStat relies on a rigid
hierarchy of brokers that allocate resources at status routers and
configure the network paths to meet the QoS application require-
ments. This scheme can generate significant signaling overheads.
Therefore, other studies have considered alternative approaches
based on self-organizing P2P technologies. For instance, a data-
centric information infrastructure that uses a publish/subscribe
communication model to deliver time-sensitive data between
EMSs and distribution substations is designed in [109]. However,
in this case the publish/subscribe system is not implemented using
a centralized directory service as in [234], but through a distrib-
uted content overlay created over a networked pool of storage
disks. In [109] it is assumed that these storage units are installed
at substations but also sensors can be allowed to share part of their
memory resources. Then, a distributed hash table (DHT) is employed
for efficient and scalable data retrieval. More specifically, a hash
function is used to generate a unique key per each data item, which
is then stored and/or replicated among the peers in the network
based on their identifiers. In this way, the overhead of both data
storage and data search is evenly distributed among peers. This ap-
proach not only addresses the scalability issues caused by central-
ized storage repositories, but it also ensures improved reliability by
avoiding single points of failures. In general, DHT-based overlay
 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
networks can ensure a very efficient retrieval of single data items
due to the use of key-based routing approaches [107]. Therefore,
other papers have proposed to use DHTs for supporting large-scale
data sharing and decentralized data repositories [231,239]. An
alternative approach is proposed in [240], in which a middleware,
called SeDAX, is developed to enable secure, large-scale data shar-
ing to support both transaction and query-based communications.
Specifically, SeDAX leverages on the properties of Delaunay Trian-
gulation (DT) graphs to design efficient message forwarding
schemes based on geometric hash functions. However, the creation
and maintenance of a structured overlay network that maps each
data key to a peer in the overlay is generally incurring high signal-
ing overheads. Thus, other studies argue that unstructured P2P
networks, which do not maintain a rigid overlay network, are more
suitable for smart grid applications in which data must be distrib-
uted to a large number of interested parties at the same time [241].
6.2. Object-oriented middleware
The middleware solutions described in the previous section
were primarily targeting the scalability of data dissemination and
data sharing functions. However, as discussed in Section 5.1, inter-
operability between different devices, networks and communica-
tion technologies is another key requirement that smart grid
communication systems must satisfy. The middleware paradigm
that is commonly considered the most suitable for supporting
interoperability between heterogeneous distributed systems is
the object-oriented paradigm [106]. Specifically, in object-oriented
middleware platforms, like CORBA [242] or Ice [243], resources,
processes and components are abstract objects that implement
standard interfaces, which hide all internal implementation details
of the object. Furthermore, the object abstraction allows to invoke
services or make calls to procedures on remote systems by also
hiding the differences of underlying networking technologies. This
distributed object-oriented paradigm is used in CoSGrid (Control-
ling the Smart Grid) [244], a middleware that provides support
to specify special remote objects called embedding metering de-
vices (EMDs). An EMD implements sensing and controlling capabil-
ities, share information via remote method invocations, and it can
be used to manage arbitrary smart grid components, from individ-
ual appliances to entire substations. EMD prototypes for different
types of devices that work with CORBA and Ice are described in
[232]. In addition, CoSGrid defines a set of interfaces for basic
smart grid services, such as metering, notification, node and data
aggregation, which can be used to ease the development of new
distributed smart grid applications. A similar approach is followed
by OHNet (Objected-based Middleware for Home Network) [245],
an object-oriented middleware for supporting the interoperability
between home devices and smart grid devices. In this system home
devices can be associated to state, function, control or streaming
objects on the basis of their features, while discovery, connection
and management objects are used to activate related functions
on home devices. Then, OHNet ensures interoperability among de-
vices, which may adopt different communication protocols and
technologies, through the definition of the Virtual Network Adap-
ter (VNA). Specifically, VNA is an abstraction layer that instantiates
the invocations of abstract methods into protocol-dependent
implementations. An alternative approach to build a Cyber-Physi-
cal Home Control System, i.e., a system that allows users to control
appliances in the physical environment by intuitive operation
through a virtual home on the network, is proposed in [246]. The
key idea is to use an OSGi-based service architecture to support
service-oriented remote control methods for home appliances.
Note that the design of user-friendly and service-oriented HEMSs
is a very active research area. For instance, the authors of [247] de-
scribe a system to control and manage home appliances in the
1685
framework of Universal Plug-and-Play (UPnP). Finally, the OSGi
platform is also used in [248] to construct a general middleware
for IoT applications involving RFID tagged objects, sensor networks
and pervasive computing technologies.
Finally, it is worth noting that in the power engineering com-
munity there is an increasing consensus on the importance of the
object-oriented paradigm, especially for the specification of inter-
operability standards. One example is the Common Information
Model (CIM) [97], which is a series of open standards released by
the International Electrotechnical Commission (IEC) to allow dif-
ferent applications to exchange information about the status of
electric grid components. For instance, CIM-based models have
been proposed to exchange power system data between: (i) energy
management systems (IEC 61970 standards [97]), (ii) electrical dis-
tribution systems (IEC 61968 standards [98]), and (iii) intelligent
electric devices within substations (EC 61850 standards [249]).
Then, CIM-based standards provide a common vocabulary to
formally describe all the major components of a power system.
Support is also provided to define multiple object classes and
attributes, as well as the relationships between them [250,251].
Furthermore, the CIM architecture supports composition of
different classes and attributes, which facilitates model flexibility
and extensibility. However, it is important to note that CIM only
facilitates the exchange of power system data in object-oriented
communication systems, but how to extract useful information
from an abstract representation of the power data is still an open
issue.
6.3. Multi-agent intelligent systems
As observed at the beginning of this section, smart grids are
complex systems consisting of a large number of heterogeneous
and interdependent components operating in dynamic environ-
ments. Therefore, decentralized management is considered the
only feasible approach to control the operations of electric grids
over multiple time-scales [252,235,2]. However, with decentral-
ized control it is difficult to achieve global optimization objectives,
such as efficiency and stability of the entire grid. On the other
hand, recently the emerging field of autonomic distributed comput-
ing has produced innovative middleware technologies to build
intelligent distributed systems that are capable of managing,
repairing, optimizing and protecting themselves without human
intervention [253]. Among the various approaches that have been
proposed so far to implement autonomic management capabilities
in distributed environments, multi-agent systems (MAS in short) are
the most popular in the smart grid research community
[254,4,255,233]. This is also demonstrated by the large number
of multi-agent systems that have been designed for a variety of
smart grid applications, including power system restoration
[256,257], fault diagnosis [258–261], management of distributed
energy resources [262–264], demand-side management [265],
management of energy storage systems [266,78], optimization of
EV operations[267–269,53], substation automation [270], distribu-
tion control [271], network monitoring [272,261] and visualization
[273], electricity market simulation [274,275], profiling of power
generation and energy usage patterns [276], management of mi-
cro-grids and VPPs [277,262,62,278,279].
A MAS is a software system that is composed of multiple auton-
omous components – the intelligent agents – that interact with
each other and react to environmental changes in order to accom-
plish a given task (e.g., to control a physical resource, or to solve a
complex problem in a distributed manner) [280]. In the simplest
case agents are reactive objects that can only respond to signals
from the environment. In more advanced systems, agents can also
be proactive in the sense that they are programmed to execute dif-
ferent actions in order to achieve a global goal. In addition, agents
1686 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
can cooperate and coordinate with each other in order to find the
best sequence of actions that could achieve that goal. To this end
agents can employ various artificial intelligence techniques,
including machine learning, fuzzy logic, neural networks or genetic
algorithms for local decision making. Although an agent can show
a high degree of flexibility and autonomy because it can dynami-
cally change its behavior to achieve a given goal, it is also subject
to an important limitation. Specifically, given the scale of the sys-
tem to control the agent can only observe and measure small por-
tions of the grid (i.e., the environment is only partially observable).
This implies that an agent can not use global knowledge to make
its decisions. Finally, it is useful to observe that the environment
that is observable by a MAS can be either physical (e.g., transmis-
sion lines, generators of renewable energy, electrical appliances,
etc.), and in this case it is observed through sensors; or virtual
(e.g., databases, computing facilities, other agents), and in this case
it is observable through programming interfaces [255].
From the above discussion we can observe that the main advan-
tage of agent-based technologies in smart grids is to provide a
decentralized management solution based on autonomous local
decisions that can ensure a high level of flexibility and robustness.
Although there are already many smart grid applications where
MAS technologies have been investigated, there are also unsolved
technical issues, which must be addressed in order to use this tech-
nology in real-world deployments. In the following we outline
three of the most important technical challenges.
6.3.1. Functional architectures for MAS
A number of different functional architectures have been pro-
posed in literature to build multi-agent systems [254], but there
is not yet a consensus on which one of those approaches is the
most suitable for smart grid applications. It is important to observe
that flexible, extensible, and open architectures, where agents can be
easily added or removed, should be preferred to closed architec-
tures, where agent interactions are fixed at design time. The
architectural model most commonly adopted in power systems is
the multi-layered architecture. For instance, in [264] a three-layer
architecture is proposed for managing distributed energy re-
sources. In this model, the bottom layer consists of agents manag-
ing physical resources (e.g., energy generators and power storage
systems). The middle layer includes agents that provide high-level
management services (e.g., fault diagnosis, protection and restora-
tion, optimization of power parameters, etc.) to the agents con-
nected to the physical resources. Finally, the top layer contains
the agents handling the user interfaces. To improve the scalability
of MAS solutions many studies have proposed to group agents,
especially those operating on physical resources, into clusters
(e.g., a micro-grid cluster or a VPP cluster). Indeed, clustering and
coalition formation are common techniques in multi-agent sys-
tems for reducing architecture complexity. For instance in [269]
coalitions are used to integrate in a more efficient way electric
vehicles into the electricity grid. In this case an aggregator agent,
called coalition server, is used to hide the details about the individ-
ual vehicles, and to present a group of vehicles as a single resource
to grid operators.
Alternative two-layer architecture models are considered in
[257] for power system restoration, and in [272,281] for the man-
agement of the distribution grid. More precisely, in [257,272] two
types of agents are considered: equipment agents and facilitator
agents. The former are controlling physical resources, such as
transmission lines, transformers and phase controllers, while the
latter are used to promote the cooperation of the equipment agents
that are associated to them. For instance, in [272] a facilitator agent
is responsible for controlling an entire electric substation. A more
elaborate architecture is described in [270], which defines multiple
types of facilitators that perform different tasks such as: device
control, data acquisition and transfer, data analysis and data que-
rying. The use of facilitator agents is also proposed in [259]. How-
ever, in this case the facilitator agent is responsible for maintaining
a list of services (or resources) that other agents in the system can
offer (or control). With the support of a nameserver agent, which
maintains the names and locations of each agent (e.g., IP ad-
dresses), a facilitator allows other agents to dynamically enter/
leave the system and register/deregister their locations and
capabilities.
6.3.2. Interoperability
When developing a multi-agent system it is essential to use
standardized agent models to allow agents to easily cooperate,
irrespective of their different capabilities and functions, or the plat-
forms used to develop them. For these reasons in the MAS commu-
nity there is a significant body of work focusing on the
formalization of:
(a) agent specification languages, which are standards for defin-
ing messages types and agent interaction models. Today
FIPA-ACL is used by MAS developers as the de facto standard
for message exchange and interaction protocols [282];
(b) ontologies, which define a common ‘‘vocabulary’’ of terms
and concepts that agents are able to exchange and interpret.
Currently, the trend in the community of MAS developers is to
implement application-specific ontologies. However, interopera-
bility between multi-agent systems using different ontologies is
difficult to obtain, even if they run on the same platform or they
are based on similar concepts [283]. A solution to this problem is
to employ a two-layer model for ontology specification. Specifi-
cally, the ontology in the top level, called ‘‘upper’’ ontology in
[283], is responsible for defining the basic concepts that are in
common to most smart grid applications (e.g., substation, switch,
voltage, etc.), and it is used as a template for defining more specific
ontologies for different applications. It is also useful to note that
existing object-oriented standards for information exchange in
power engineering applications, such as CIM [97] and IEC 61850
[249], can be easily used as reference models when developing
an upper ontology for smart grid applications.
6.3.3. Implementation platforms
In recent years several tools and programming frameworks
have been provided to develop multi-agent systems in the context
of smart grid applications, and interested readers can refer to [284]
for a complete survey. However, the open-source Java Agent Devel-
opment Framework (JADE) [285] is probably the most common
platform for developing MAS solutions for smart grid applications.
Most of existing MAS implementations are prototypes used for lab-
based experiments [286,277], which aim at demonstrating the fea-
sibility of the MAS approach. Recently some multi-agent systems
have been tested into operational electric grids. For instance, a
multi-agent system, called Protection Engineering Diagnostic
Agents (PEDA) [259], has been used by a transmission system
operator in the U.K. to interpret SCADA-related data and to provide
online diagnostic information and alarms [287]. In [269] a multi-
agent system is implemented using JADE to integrate a group of
electric vehicles into the electric grid and to support both demand
response and V2G services. However, to demonstrate that MAS
technologies meet the robustness and reliability requirements of
typical power grid applications, large scale testing is still needed.
A valuable option to avoid the costs and risks associated to test
new technologies in real power grids is to employ realistic
simulation tools that simultaneously model electric power scenar-
ios and the behaviors of computer communication protocols. An
example is provided by EPOCHS [288], which is a combined
 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
simulation environment that federates commercial electric simula-
tors with ns-2, a popular open-source communication-network
simulator [289].
Before concluding this section it is necessary to observe that
alternative technologies have also been considered for building
large-scale and interoperable power system applications, including
service-oriented architectures (SOA) [290], grid and cloud comput-
ing [291–293] and web services [294]. However, agent-based tech-
nologies appear the most suitable solution for distributed control
applications that must support autonomous behaviors, as also
demonstrated by the wide range of applications in which multi-
agent systems are being developed. Interested readers are also re-
ferred to [290–292,294] for a more detailed discussion about the
technical challenges that should be addressed to apply in the smart
grid domain the other approaches mentioned above to build large-
scale interoperable dynamic systems.
7. Security mechanisms for smart grids
Electric utilities consider security, protection and reliability of
the electric infrastructures one of their key priorities because fail-
ures or malfunctions in power systems not only would have an
economical impact, but could also cause serious damages to peo-
ple. Historically security in power systems was obtained by ensur-
ing physical protection of power generators and distribution grids,
as well as physical isolation of utility control centers. Furthermore,
closed and dedicated communication networks running proprie-
tary protocols were typically used in SCADA-based power control
systems [3]. On the other hand, it is reasonable to expect that in
smart grids the deployment of distributed and autonomous control
functions, as well as the adoption of open network architectures,
will bring new security vulnerabilities to smart grids and they will
allow a variety of unforeseen security attacks. Different categories
of security attacks specific to the smart grid domain have been
discussed and identified in previous survey papers
[117,21,295,296,24,297]. Based on those studies, the most impor-
tant vulnerabilities of the smart grid communication system can
be broadly classified as follows:
 Device vulnerabilities: IEDs will be widely deployed in smart
grids to monitor and remotely control electricity production
and distribution processes. However, malicious users or attack-
ers can compromise these devices, e.g., to manipulate sensed
data or to disrupt normal grid operations. Furthermore, many
IEDs will support wireless communications to facilitate deploy-
ment and simplify the access to information. An intuitive down-
side of wireless communication technologies is that they rely on
an inherently unprotected physical medium. This makes easier
to capture private information (eavesdropping), disrupt commu-
nications with noise signals (jamming attack), or generate fake
messages (injecting attack). Solutions of these attacks include
encryption of messages for data integrity and confidentiality,
authentication of network access, and various randomized
transmission methods, such as spread spectrum techniques [23].
 Network vulnerabilities: The adoption of open network architec-
tures, off-the-shelf network devices and publicly available com-
munication standards is needed to meet the flexibility,
scalability and interoperability requirements discussed in Sec-
tion 5.1. However, this also causes various security problems
observed in other telecommunications networks adopting open
architectures (e.g., Internet), such as malicious modification of
routing information, DNS hacking, and various types of denial
of service (DoS) attacks. More specifically, in a typical DoS attack
an attacker can control a set of nodes to overwhelm other nodes
with data traffic, resulting in excessive network delays or even
communication failures. For instance, compromised smart
1687
meters can be forced to flood an EMS with meaningless mes-
sages. In this way an EMS would consume all its computational
and communication resources and it would not be able to timely
react to legitimate requests. To prevent these network attacks,
access control and intrusion detection are essential security
mechanisms. Similarly, authentication and authorization
schemes are also necessary to support secure remote configura-
tion and control of geographically dispersed devices [298].
 Data vulnerabilities: Data manipulation is an important security
issue because an attacker can modify data or control commands
to compromise electric grid reliability. However, smart grids
will also be increasingly vulnerable to data attacks designed
to compromise the privacy of customers. For instance, the
man-in-the-middle attack is a popular method used by mali-
cious users to gain access to information without physically
compromising the target. If an attacker is able to snoop the
metering data transmitted from the consumer’s home it could
infer consumer’s habits and activities. For instance, it would
be possible to learn whether a home is occupied or not by
detecting power consumption signatures of specific activities
(e.g., watching television or using the microwave) [299]. As bet-
ter explained in following sections, encryption, message
authentication and intrusion detection schemes are needed to
preserve data integrity and confidentiality in smart grids [21].
Finally, it is important to point out that security techniques
developed for other telecommunications networks can fail to meet
the security requirements of smart grid communication systems
because [300]:
(i) Many security attacks envisioned for smart grid communica-
tion systems do not have a counterpart in other computer
networks because they can target either the physical power
system or the communication infrastructure or both [301].
For instance, in the power system there are several control
loops used to control physical aspects of generation, trans-
mission, and distribution processes, and an attack directed
at the communication networks underlying these control
loops can have a system-wide impact on power system
stability.
(ii) Typical security objectives and priorities for smart grids are
different from those of most telecommunications networks.
Indeed, in most networks data confidentiality and integrity
are more important than communication availability. On
the contrary, in a power system electricity must always be
available. Thus, availability is the most important security
objective along with the protection of consumers’ privacy.
(iii) Most devices that are used in smart grids are expected to be
heavily constrained in terms of computation capabilities and
data storage. Thus, security techniques developed for high-
capability Internet-enabled devices (e.g., servers, laptops,
or smartphones) are not suitable for smart grid devices.
In the rest of this section we will present some fundamental
security techniques that should be integrated in smart grids for
improving the robustness of smart grids against various security
attacks and privacy risks. Furthermore, we will also discuss a few
representative solutions in each category to clarify interesting
open problems. For the sake of clarity, in Fig. 8 we provide a clas-
sification of major research trends in the area of security for smart
grids, which we have also outlined in following sections.
7.1. PKI management
A public key infrastructure (PKI) is a fundamental security tool
in most telecommunications networks and distributed systems in
1688 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
Fig. 8. Classification of research trends in the area of security for smart grids.
general. Recently the design of PKIs suitable for smart grids has
been extensively studied [302,303]. Basically, a PKI is a system of
software and hardware components providing digital signatures,
called certificates, which are used to identify a certain entity
(e.g., a device, a program, an organization). These certificates are
securely stored by a trusted certificate authority (CA), which also
decides when and how to renew or revoke them. For instance,
smart meters can register their serial numbers with a CA. Typically,
a PKI employs public key cryptography techniques, which are
based on the simultaneous creation of a pair of public and private
keys, to securely store and exchange certificates. The public key
is publicly available (as part of the digital certificates) in central
repositories and it is used to encrypt messages, while the private
key is known only to the recipient of the message and it is used
for decryption. This is substantially different from symmetric key
crypto-systems, where there is an initial exchange of a shared se-
cret key, which would require to pre-establish a secure communi-
cation channel. Finally, PKI also supports message authentication
because the sender can use its private key to encrypt a digital cer-
tificate, which is then utilized to sign transmitted messages.
Although PKI is the most popular key management scheme in
the Internet there are some issues to apply existing PKI solutions
to the smart grid domain [302,304]. First of all, there is a scalability
problem because a PKI for smart grids should maintain certificates
for millions of devices. Furthermore, differently from classical PKI
systems, we cannot assume that identities are the only relevant
properties that should be certified in smart grid devices. For in-
stance, context information such as the installation location of
the devices could also be included in the certificate. However, this
information is known only after the installation and it may change
dynamically. Hence, digital certificates associated to smart grid de-
vices should also be dynamic. Finally, smart grids must support
real-time and reliable operations. Therefore, PKI technologies
should meet strict delay constraints and provide fault tolerance
to ensure high availability. However, a PKI involves centralized
authorities and complex certification policies and it might be quite
challenging to satisfy latency requirements.
To address the above issues a number of research directions are
currently being investigated. For instance, it is generally agreed
that more automated configuration tools are needed for PKI sys-
tems in smart grids. These tools will be used to allow each organi-
zation to set its own security policies (how private keys are
protected, which is the validity time of a certificate, how certifi-
cates are revoked, etc.) and to easily adapt these policies to the dif-
ferent requirements of smart grid devices and applications [302]. A
large body of work is also dedicated to the design of PKI architec-
tures that are able to meet the reliability, scalability and delay
requirements of smart grids. For instance, a novel key management
scheme is proposed in [305], which combines public-key and
symmetric key approaches to achieve simplicity and to improve
scalability. Basically, trust anchors are deployed in the PKI system,
which employ robust public key methods to establish symmetric
keys between data aggregators and data collectors. Then, trust del-
egation mechanisms are used to allow simple sensors nodes to ac-
cess the grid and to communicate with data agents. However, the
key distribution scheme proposed in [305] can suffer from the
man-in-the-middle attack during the initial authentication phase.
To solve this issue, in [306] it is proposed to use trusted third par-
ties to distribute shared keys among the components of the smart
grid. The focus of the work in [307] is the scalability and interop-
erability of the key management system. In particular, the key
management scheme proposed in [307] uses existing standard pro-
tocols for network access authentication (i.e., EAP) to avoid that
multiple authentication and key establishment processes are per-
formed across different protocols and link-layer technologies. A
simple PKI system is designed in [308], which assumes that a un-
ique machine number is available at each device, which can be
used by a central server holding a master key to generate unique
private keys. The main advantage of this scheme is that it does
not require to separately configure each device. On the other hand,
unique machine identifiers are not always available. Finally, a pub-
lic key crypto-system is proposed in [309], which exploits homo-
morphic encryption techniques to avoid that a pair of public and
private keys is generated for each communication link between
customers and utility control centers. In general, an aspect that
needs much further investigation is the design of enhancements
to existing PKI systems to support new use cases, in particular
those due to the increasing adoption of mobile electric vehicles.
7.2. Authentication and access control
An essential security mechanism is the authentication method,
which is used to verify device identities and data validity. As ob-
served above PKI systems inherently provide device authentication
through digital signatures. The authenticated device identity can
be exploited to establish shared secret keys that are used for
encrypting and authenticating data packets. It is important to note
that symmetric key cryptography is generally preferred to public
key cryptography for data authentication because it uses the same
key for both decryption and encryption, and this results in faster
cryptographic algorithms that requires less processing power.
However, authentication schemes suitable for smart grids should
take into account the limited resources (i.e., low memory and com-
putational capacity) of most smart grid devices, as well as the
stringent delay requirements of smart grid applications. A light-
weight message authentication scheme is proposed in [310], which
aims at minimizing the number of messages exchanged during
authentication. This is achieved by combining the Diffie-Hellman
 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
protocol for key establishment and an hash-based message
authentication code (MAC) technique. An authentication protocol
is proposed in [311] for multicast communications, which reduces
storage and bandwidth overhead with respect to other schemes
that use public-key signature over multiple messages. An alterna-
tive authentication approach is proposed in [312], which ensures
both authentication and data integrity by jointly using digital sig-
natures and timestamps. Finally, three authentication mechanisms
are specified in [313] for devices typically used in HANs, such as
smart meters, smart household appliances, and electric vehicles
moving. Note that EVs pose the most complex authentication prob-
lems because they can move between different home area
networks.
A data protection mechanism that it is closely related to device
authentication is access control, which aims at ensuring that access
to objects (i.e., any entity containing information or providing ser-
vices) is permitted only to authorized users. The most common ac-
cess control schemes in telecommunications networks are based
on access control lists (ACLs), which specify the identities of users
allowed to access shared objects and what privileges they have.
However, ACL-based approaches may not scale with the huge
number of customers and resources involved in smart grid ser-
vices. Therefore, most of the research in this area is focusing on
the formulation of role-based access control methods, in which per-
missions to perform certain operations are assigned to specific
roles rather than user identities. A network access control model
for power system with micro-grids is described in [314], in which
each micro-grid is a separate network domain with an indepen-
dent control center that maintains XML-based security policies
and roles. In this scheme a role is defined as a collection of privi-
leges that can be executed by authorized users in local and remote
domains. Note that each user can be assigned with multiple roles.
Furthermore, to improve scalability roles are structured into a hier-
archy in which the ‘‘parent’’ role directly inherits all the privileges
of its ‘‘children’’ roles. In [315] it is proposed to store data in an en-
crypted form that depends on the data attributes. Then, only users
having the same attributes as the data can decrypt those messages.
Finally, an alternative scheme is proposed in [316] to use smart
meters as firewalls between home area networks and the utility
core backbone.
7.3. Privacy protection and data integrity
The most intuitive method to ensure data privacy is to use a
cryptographic protocol, which preserves data confidentiality by
encrypting data messages between data originators (e.g., smart
meters) and utility providers. The device authentication schemes
described in the previous section can be used to create the shared
secret keys that are needed for message encryption. However, the
massive deployment of smart meters and smart appliances also
raises new privacy concerns. Specifically, it becomes an important
issue to avoid that malicious users can learn information about
customers’ behaviors and habits. For these reasons several privacy
protection methods specific for smart metering have been recently
proposed [22]. These solutions can be categorized into two sepa-
rated classes: anonymization techniques and randomization
techniques,
Anonymization techniques achieve privacy protection for
metering data by anonymizing energy usage measurements in or-
der to make more difficult to associate metering information to a
particular smart meter or customer. In [317] this is achieved by
assuming that a trusted third party provides pseudonymous iden-
tities, which are then used to anonymize high-frequency metering
data. Note that this metering data is the most critical to infer usage
patterns of specific electrical appliances, while low-frequency
metering readings used for billing purposes are rarely enough to
1689
offer adequate privacy. The use of a trusted third party is also pro-
posed in [318], although this approach introduces additional com-
plexities and requirements to the security framework. More
efficient schemes can be designed by exploiting secure data aggre-
gation techniques. For instance a privacy-preserving aggregation
scheme, called EPPA, is designed in [319], which permits to per-
form data aggregation at intermediate proxies without decrypting
the data received by individual smart meters. A new encryption
mechanism is also proposed in [299], which allows a group of
smart meters to encrypt their data in such a way that electric util-
ities can reconstruct aggregated energy usage information without
decrypting them. Note that the main open issue in this class of pri-
vacy protection schemes is how to preserve the ability of electric
utilities to know the instantaneous electricity consumption, which
is needed to support demand response and direct load control.
Randomization techniques are used in particular to protect the
confidentiality of usage patterns of household electric appliances.
More specifically, this class of schemes achieve privacy protection
by manipulating the metering data with some stochastic perturba-
tion function so that it is difficult to infer patterns of energy con-
sumption. For instance, a simple privacy scheme is developed in
[318], which adds random Gaussian noises to each smart meter.
This approach is simple but not very robust because it applies ran-
domization on true readings and statistical methods could be used
to extract load profiles. More sophisticated schemes are developed
in [320], which mix actual smart meter readings with faked read-
ings sampled from arbitrary distributions. However, further stud-
ies are needed to understand which is the level of uncertainty
that is added by such randomization methods, and whether useful
information about energy consumption can be maintained in the
modified data. Finally, an alternative approach is proposed in
[321,322] by assuming that smart homes will contain a variety of
energy storage and generation systems, which can be used to
obfuscate appliance load signatures by randomly mixing utility
electricity with electricity from local generators.
7.4. Intrusion detection
Although several security mechanisms can be used to deal with
security attacks, more sophisticated attackers can exploit unknown
system vulnerabilities and succeed in compromising network com-
ponents. In this case an intrusion detection system (IDS) is essen-
tial to identify the attack and trigger appropriate countermeasures.
There are different approaches that can be applied to design intru-
sion detection techniques, but at least three major methodologies
can be identified [323]:
 Anomaly detection: anomaly detection recognizes intrusive
actions by looking for deviations from normal behaviors
of protocols, programs or processes. This approach requires
the specification of accurate statistical models of normal
profiles, which can be quite difficult and costly to obtain.
The main advantage of anomaly detection is its ability to
detect unknown attacks. However, anomaly detection tech-
niques also produce a high number of false alarms if the
model is not sufficiently accurate. Furthermore, this
approach is suitable for relatively stable systems, otherwise
new models of normal operations should be developed
after any configuration change.
 Signature detection: signature detection is complementary
to anomaly detection because it uses deterministic patterns
of known attacks to characterize malicious behaviors. The
main advantage of signature-based methods is that they
provide high accuracy and a low number of false alarms.
However, there are also some drawbacks such as the
inability of these schemes to detect novel attacks, whose
1690 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
signatures are unknown. Furthermore, pattern matching
techniques and classification algorithms typically used in
signature detection methods can be quite costly in terms
of computing and data storage.
 Specification-based approach: specification-based methods
detect attacks by defining ordered sequence of events,
called policies, which corresponds to correct protocol
behaviors. A violation to this set of constraints and rules
is considered a security violation. A specification-based
IDS is able to detect unknown attacks as for anomaly detec-
tion. However, an obvious downside of this class of tech-
niques is that the development of specifications is time
consuming and protocol specific. Another issue is that
specifications are difficult to verify on devices that have
limited memory and computational power.
In the following we outline most representative solutions for
intrusion detection in the smart grid domain based on the above
categories.
A key design problem for both anomaly and signature detection
methods is to collect empirical data from smart grids and to extract
accurate models for normal and malicious behaviors, respectively.
Note that many intrusion detection schemes use neural networks
to learn normal profiles and detect deviations from these profiles
[323,324]. However, those techniques do not seem suitable for
smart grids because they are quite demanding in terms of compu-
tational and memory overheads. Furthermore, they do not guaran-
tee deterministic response times. A variety of techniques have
been proposed to address those issues. For instance, in [325] super-
vised learning models known as support vector machines (SVM)
are used to extract features (e.g., connection duration and protocol
type) and recognize patterns representing normal communications
because of their high accuracy and scalability. In [326] it is shown
that Decision Tree techniques can outperform other supervised
machine learning techniques to implement real-time intrusion
detection. A novel unsupervised network intrusion detection sys-
tem based on clustering techniques is decribed in [327]. Alterna-
tively, if service level agreements (SLAs) are established between
the service provider and its customers then SLA monitoring can
be used to detect service violations and to verify intrusion at-
tempts [328]. In addition, [325] defines a hierarchical IDS solution,
in which an IDS module is installed at each protocol layer. Then,
individual IDS modules are trained using data that are relevant
to their level, but they also cooperate to detect possible attacks.
In [329] malicious events are modeled as random variables follow-
ing a Gaussian process. For instance, a random variable may be
used to represent the number of malfunctioning smart meters in
a building, or the number of failed authentication attempts in a gi-
ven time interval. Then, conventional stochastic tools are used to
predict the occurrence of an attack. Distributed and low-complex-
ity pattern recognition algorithms are also developed in [330,331].
Finally, many anomaly detection algorithms utilize a sensitivity
threshold to detect violations of normal behaviors, and the proper
selection of this threshold is essential to reduce the occurrence of
false alarms. In [332] fuzzy-logic rules are developed for tuning
this sensitivity threshold based on an estimate of the system
knowledge obtained by the IDS module. However, an intelligent
adversary could compromise the IDS and adapt its actions to re-
duce the probability to be detected. How to dynamically adjust
the sensitivity threshold in response to more intelligent attacks
is still an open issue.
An important challenge for both anomaly-based and signature-
based IDSs is to identify which are the relevant events to monitor
in the power system, and many studies exist that try to classify
different types of attacks. Since a specification-based IDS do not
require empirical data to detect intrusions, this class of intrusion
detection schemes has been considered as a valid alternative to
more sophisticated IDS solutions, at least in the early stage of
smart grid development. For instance, in [333] a specification-
based intrusion detection method is designed for the C12.22
standard protocol, which specifies an application-level messaging
protocol to exchange power data over a network. In [334] a speci-
fication-based IDS is developed of the IEEE 802.15.4 standard ap-
plied in home area networks. However, there is still a high
overhead associated to these schemes, and it might be impractical
to run them on smart grid devices.
7.5. Security analysis
Utility companies need formal security analysis or simulation
tools to quantify the robustness provided by adopted security
mechanisms, as well as to predict the impact of attackers on smart
grid operations and reliability. Obviously there are many tools to
analyze vulnerabilities in traditional telecommunications net-
works [335] (e.g., to identify misconfiguration problems), and to
rate and score known vulnerabilities [336]. However, new specific
solutions are needed for smart grids due to the heterogeneity of
devices, protocols and configurations. An automated tool for AMI
configuration verification, called SmartAnalyzer, is developed in
[337] using formal analysis based on Satisfiability Modulo Theories
(SMT). Basically, this tool encodes AMI configuration templates,
which describe device configurations, network topologies, moni-
toring schedules, into SMT logic. Then, an SMT solver is used to
verify this configurations against a set of high-level constraints
(e.g., reachability between devices, schedule and resource con-
straints) and user-driven constraints (e.g., data protection, data
integrity, transmission reliability). The output of this verification
process is a report that indicates reasons of constraint violations
and possible remedies. An alternative approach to quantify the de-
gree of damages that cyber attacks can cause to a power systems is
designed in [338]. The key idea is that both physical components
(i.e., electric grid elements) and cyber components (i.e., network
devices and functionalities) can be modeled as objects to which
state information is associated. Then, directed graphs are used to
represent state dependencies amongst the various objects. Finally,
a system of differential equations is utilized to specify the rules
governing state evolution, and to model cause-effect relationships.
However, identifying cause-effect relationships in large-scale
power systems is a challenging task. Another class of methods
for risk analysis in smart grids rely on computational intelligence
approaches. A useful technique in this class is fuzzy logic because
it supports probabilistic risk assessment by measuring uncertain-
ties in the decision making processes and allowing for incomplete
or ambiguous data (fuzzy data). An overview of the main
algorithms in computation intelligence used in smart grid risk
assessment is reported in [339]. A key open issue in the area of risk
assessment is the design of toolkits that automatically generate
models of the power system from specifications, operating
procedures and network topologies to describe the vulnerabilities
of devices, services and network connections. Note that power grid
models are essential to identify system vulnerabilities, predict new
attacks and evaluate potential service damages [340].
Related to risk assessment (as well as intrusion detection) is the
design of mechanisms to evaluate the trustworthiness of smart
grid objects (e.g., devices, service providers, entire subsystems).
Specifically, a trust system allows an entity to assess the reliability
of another entity before deciding to interact with it (e.g. to use a
service provided by that entity) [341]. In this way, a trust model
can be used to determine risks and to give a weight to those risks.
Typically, trust models are associated to reputation systems, which
enable peers to rate each other after the completion of a certain
task. Then, a reputation system uses the aggregated ratings about
 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
a given peer to derive its reputation score (or trustworthiness). As
an example, power protection applications generally rely on the
cooperation among grid components to cope with grid failures or
grid instability. Thus, it will become very important to decide
which is the best entity to collaborate with, and the reputation
plays a key role in the decision process. For these reasons, various
studies exist that have explored how to integrate reputation-based
trust management systems in smart grids. For instance, in [342] a
trust management methodology is developed to verify if individual
smart meters generate selfish and malicious reports on energy
usage. Furthermore, one of the advantages of a trust system is to
provide an additional layer of protection in power systems imple-
menting agent-based control functions as discussed in [343,344].
Specifically, those studies consider a backup protection scheme
and propose trust metrics to improve the reliability of network
communications. Similarly, in [345] an early warning system is de-
signed for predicting and anticipating cascading failures in power
systems, which uses a reputation mechanism for controlling net-
work behaviors. Furthermore, nodes in a network can be assigned
different roles and perform different functions depending on their
trust levels [346]. Note that the design of suitable techniques for
quantifying trust levels in dynamic environments while preserving
data privacy is an open issue.
8. Summary and outlook on key research areas
In this survey we have advocated the view of a smart grid as the
outcome of an evolutionary transformation of the existing electric-
ity network towards an optimized and sustainable energy system.
We have analyzed the several factors that are contributing to this
evolution, ranging from technological developments (e.g., the
introduction of electric vehicles, the deployment of renewable en-
ergy resources, the recent advancements in energy storage sys-
tems, time-synchronized measurement technologies of electricity
quality) to economical and societal drivers (e.g., liberalization of
energy markets with the advent of prosumers, new environmental
concerns about climate changes and pollution, soaring growth of
energy demands, need for higher operating efficiency, active par-
ticipation of consumers in demand response, better resiliency
against both physical and cyber attacks). The incorporation in the
smart grid of a pervasive monitoring and communication infra-
structure is essential to: (a) measure the system state, (b) collect
the sensor information and propagate control signals, and (c) allow
distributed smart applications to control power flows in the net-
work. Thus, the goal of this survey was to present a complete mod-
el of this monitoring and communication infrastructure, covering
all layers of its architecture, ranging from communication technol-
ogies at the physical layer, to communication topologies and pro-
tocols at the network layer, and concluding with middleware
services needed to support the new applications in a distributed
fashion.
We believe that this survey could provide the reader with a
useful overview of the state-of-the-art solutions proposed at all
layers of the smart grid communication infrastructure, as well as
a guide of the open research issues. Hereafter, we aim at summa-
rizing some of the novel research directions that are still worth
exploring.
 Due to the massive number of data sets that will be col-
lected in smart grids, classical database-management tools
can be unable to process them within acceptable delays. In
addition to data analysis, other critical issues are data stor-
age, search and visualization. Nowadays there are many
application domains that are already witnessing an explo-
sion of collected data, such as healthcare, retail markets,
online social sites, but smart grids have the potential to
1691
become one of the most demanding one [347]. However,
since temporal and spatial variations of most electric pro-
cesses are small, it is reasonable to expect that most mea-
surements in this massive amount of data will be
redundant. This creates many opportunities for the design
of data mining techniques able to provide data compres-
sion. Furthermore, the design of novel knowledge-discov-
ery methodologies that are specific for electric data is a
very compelling research topic. Finally, cloud computing
is considered an important data and computing model for
data intensive smart grid applications.
 Energy routers indicate devices that allow units of energy
(locally generated, stored, and forwarded) to be dispatched
when and where it is needed [164]. The deployment of
energy routers within the smart grid can enable innovative
paradigms for energy distribution and control, in which
energy is logically packetized, buffered and forwarded over
the physical energy network. While energy routers guaran-
tee more flexible end efficient power distribution and this
facilitates an increased use of renewable energy sources,
they also pose new challenges. As an example, an energy
router should be able to dynamically redirect incoming
energy flows towards outgoing energy flows. Thus, new
power electronics are required in energy routers to imple-
ment automatic energy control. Similarly, energy routers
should implement distributed intelligence to control the
energy routing process. Furthermore, innovative ways of
dispatching energy in a smart grid can be devised taking
advantage of EVs. For instance, we can use the batteries
of EVs as a mean of physically moving electrical energy.
In this way EVs can support a delay-tolerant transfer of
energy between homes. However, to quantify which are
the potential gains of this approach a detailed cost-benefit
analysis is necessary.
 The design of optimization strategies for the smart grid is a
very active research area, and there is a rich set of method-
ologies and algorithms from different domains that have
been considered for this purpose. For instance, there are
several algorithms in the literature to determine the opti-
mal load schedules of groups of domestic appliances How-
ever, how to design management and control schemes able
to provide real-time and wide-area control with limited
computational costs is still an open problem. We believe
that flow-based congestion control algorithms, which have
been commonly applied in large-scale information net-
works, such as the Internet, may be a promising and attrac-
tive approach to mitigate power congestion by reducing
peak loads. For instance, there are electric devices that
can elastically adapt the amount of instantaneous power
they need, such as many common household appliances.
Then, those devices could intelligently increase/decrease
their power demands depending on congestion feedback
signals from the utilities.
 Due to the intrinsic differences between conventional
power grids and smart grids, existing electric power simu-
lation/analysis tools will not be able to accurately model
and predict the behavior of new power grids. For instance,
more precise models of renewable energy resources at
increasingly lower time and spatial scales are needed to
increase the reliability of protection and control systems.
Similarly, the behaviors of power systems will be increas-
ingly dependent on external factors, such as the reliability
of the smart grid communication network. Thus, power
simulators should integrate tools to model each component
of a smart grid, as well as the possible interactions between
components.
1692 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
 Security mechanisms are an essential part of a smart grid.
Although a considerable amount of research has been con-
ducted in this field many open issues still exist because the
increased interconnection and integration, e.g., between
electric grid, monitoring and communication network, data
management systems and applications, also introduce new
cyber-vulnerabilities into the smart grid. The most compel-
ling research challenges are the design of suitable mecha-
nisms to protect the confidentiality and integrity of
metering data, as well as new mechanisms to control the
access to smart-grid components and resources given that
physical isolation of the power grid might not be feasible
anymore.
References
[1] T. Garrity, Getting smart, IEEE Power and Energy Magazine 6 (2) (2008) 38–
45.
[2] H. Farhangi, The path of the smart grid, IEEE Power and Energy Magazine 8 (1)
(2010) 18–28.
[3] F. Wu, K. Moslehi, A. Bose, Power system control centers: past, present, and
future, Proceedings of the IEEE 93 (11) (2005) 1890–1908.
[4] S. Massoud Amin, B. Wollenberg, Toward a smart grid: power delivery for the
21st Century, IEEE Power and Energy Magazine 3 (5) (2005) 34–41.
[5] National Institute of Standards and Technology (NIST), NIST framework and
roadmap for smart grid interoperability standards, Release 1.0, January 2010.
[6] L. Atzori, A. Iera, G. Morabito, The internet of things: a survey, Computer
Networks 54 (15) (2010) 2787–2805.
[7] A. Ipakchi, F. Albuyeh, Grid of the future, IEEE Power and Energy Magazine 7
(2) (2009) 52–62.
[8] R. Kulyk, Smart grid taking our cue from nature, Tech. rep., REGEN Energy Inc.,
2009.
[9] E. Santacana, G. Rackliffe, L. Tang, X. Feng, Getting smart, IEEE Power and
Energy Magazine 8 (2) (2010) 41–48.
[10] R. Moreno, G. Strbac, F. Porrua, S. Mocarquer, B. Bezerra, Making room for the
boom, IEEE Power and Energy Magazine 8 (5) (2010) 36–46.
[11] V.C. Gungor, F.C. Lambert, A survey on communication networks for electric
system automation, Computer Networks 50 (7) (2006) 877–897.
[12] P. Parikh, M. Kanabar, T. Sighu, Opportunities and challenges of wireless
communication technologies for smart grid applications, in: Proc. of IEEE PES
General Meeting’10, 2010, pp. 1–7.
[13] V. Gungor, B. Lu, G. Hancke, Opportunities and challenges of wireless sensor
networks in smart grid, IEEE Transactions on Industrial Electronics 57 (10)
(2010) 3557–3564.
[14] W. Wang, Y. Xu, M. Khanna, A survey on the communication architectures in
smart grid, Computer Networks 55 (2011) 3604–3629.
[15] T. Khalifa, K. Naik, A. Nayak, A survey of communication protocols for
automatic meter reading applications, IEEE Communications Surveys &
Tutorials 13 (2) (2011) 168–182.
[16] J. Gao, Y. Xiao, J. Liu, W. Liang, C. Chen, A survey of communication/
networking in smart grids, Future Generation Computer Systems 28 (2012)
391–404.
[17] F. Bouhafs, M. Mackay, M. Merabti, Links to the future: communication
requirements and challenges in the smart grid, IEEE Power and Energy
Magazine 10 (1) (2012) 24–32.
[18] Y. Yan, Y. Qian, H. Sharif, D. Tipper, A survey on smart grid communication
infrastructures: motivations, requirements and challenges, IEEE
communications Surveys & Tutorials 15 (1) (2013) 5–20.
[19] Z. Fan, P. Kulkarni, S. Gormus, C. Efthymiou, G. Kalogridis, M. Sooriyabandara,
Z. Zhu, S. Lambotharan, W. Chin, Smart grid communications: overview of
research challenges, solutions, and standardization activities, IEEE
Communications Surveys & Tutorials 15 (1) (2013) 21–38.
[20] G. Ericsson, Cyber security and power system communication – essential
parts of a smart grid infrastructure, IEEE Transactions on Power Delivery 25
(3) (2010) 1501–1507.
[21] O. Kosut, L. Jia, R. Thomas, L. Tong, Malicious data attacks on the smart grid,
IEEE Transactions on Smart Grid 2 (4) (2011) 645–658.
[22] F. Siddiqui, S. Zeadally, C. Alcaraz, S. Galvao, Smart grid privacy: issues and
solutions, in: Proc. of IEEE ICCCN’12, 2012, pp. 1–5.
[23] E.-K. Lee, M. Gerla, S. Oh, Physical layer security in wireless smart grid, IEEE
Communications Magazine 50 (8) (2012) 46–52.
[24] Y. Yan, Y. Qian, H. Sharif, D. Tipper, A survey on cyber security for smart grid
communications, IEEE Communications Surveys & Tutorials 14 (4) (2012)
998–1010.
[25] S. Rohjans, M. Uslar, R. Bleiker, J. Gonzalez, M. Specht, T. Suding, T. Weidelt,
Survey of smart grid standardization studies and recommendations, in: Proc
of IEEE SmartGridComm’10, 2010, pp. 1–6.
[26] P. Zhang, F. Li, N. Bhatt, Next-generation monitoring, analysis, and control for
the future smart control center, IEEE Transactions on Smart Grid 1 (2) (2010)
186–192.
[27] F. Li, W. Qiao, H. Sun, H. Wan, J. Wang, Y. Xia, Z. Xu, P. Zhang, Smart
transmission grid: vision and framework, IEEE Transactions on Smart Grid 1
(2) (2010) 168–177.
[28] N. Saputro, K. Akkaya, S. Uludag, A survey of routing protocols for smart grid
communications, Computer Networks 56 (2012) 2742–2771.
[29] X. Fang, X. Misra, G. Xue, D. Yang, Smart grid – the new and improved power
grid: a survey, IEEE Communications Surveys & Tutorials 14 (4) (2012) 944–
980.
[30] G. Vassell, Northeast Blackout of 1965, IEEE Power Engineering Review 11 (1)
(1991) 4.
[31] D. Nordell, Communication systems for distribution automation, in: Proc. of
IEEE/PES Transmission and Distribution Conference’08, 2008, pp. 1–14.
[32] Modbus organization, Modbus application protocol specification – V1.1b,
December 2006. <http://www.modbus.org/specs.php>.
[33] IEC 61850-1, Communication networks and systems in substations – Part 1:
Introduction and overview, 2003.
[34] L. Barroso, H. Rudrick, F. Sensfuss, P. Linares, The green effect, IEEE Power and
Energy Magazine 8 (5) (2010) 22–35.
[35] International Energy Agency, Monthly electricity statistics, 2011.
[36] REN21, Renewables 2011 global status report. 2011. <http://www.ren21.net/
>.
[37] T. Ackermann, G. Andersson, L. Soder, Distributed generation: a definition,
Electric Power Systems Research 57 (3) (2001) 195–204.
[38] A. Borbely, J. Kreider, Distributed Generation: The Power Paradigm for the
New Millennium, Mechanical Engineering Series, CRC Press, 2001.
[39] W. El-Khattam, M.M.A. Salama, Distributed generation technologies,
definitions and benefits, Electric Power Systems Research 71 (2) (2004)
119–128.
[40] H. Zareipour, K. Bhattacharya, C. Canizares, Distributed generation: current
status and challenges, in: Proc. of IEEE NAPS ’04, 2004.
[41] J. Driesen, R. Belmans, Distributed generation: challenges and possible
solutions, in: Proc. of IEEE Power Engineering Society General Meeting’06,
2006.
[42] M. Eltawil, Z. Zhao, Grid-connected photovoltaic power systems: technical
and potential problems – a review, Renewable and Sustainable Energy
Reviews 14 (1) (2010) 112–129.
[43] K. Moslehi, R. Kumar, A reliability perspective of the smart grid, IEEE
Transactions on Smart Grid 1 (1) (2010) 57–64.
[44] International Energy Agency, Variability of wind power and other renewables
– management options and strategies, 2005.
[45] A. Mahmood, M. Aamir, M. Anis, Design and implementation of AMR smart
grid system, in: Proc. of Electric Power Conference’08, 2008.
[46] C. Zito, From smart metering to smart grids: Enel experience, September 12
2011. <http://www.pointview.com/data/files/9/8467/2087.pdf>.
[47] International Energy Agency, Distributed generation in liberalised electricity
markets, OECD/IEA, 2002.
[48] G. Rothwell, T. Gomez, Electricity Economics: Regulation and Deregulation,
Wiley, IEEE Press, 2003.
[49] K. Spees, L. Lave, Demand response and electricity market efficiency, The
Electricity Journal 20 (3) (2007) 69–85.
[50] M. Gharbaoui, L. Valcarenghi, B. Martini, P. Castoldi, R. Bruno, M. Conti,
Effective management of a public charging infrastructure through a smart
management system for electric vehicles, in: Proc. of IEEE ENERGYCON, 2012,
pp. 1095–1100.
[51] S. Shao, M. Pipattanasomporn, S. Rahman, Challenges of PHEV penetration to
the residential distribution network, in: Proc. of IEEE PES ’09, 2009.
[52] E. Sortomme, M. Hindi, S. MacPherson, S. Venkata, Coordinated charging of
plug-in hybrid electric vehicles to minimize distribution system losses, IEEE
Transactions on Smart Grid 2 (1) (2011) 198–205.
[53] S. Vandael, K. De Craemer, N. Boucké, T. Holvoet, G. Deconinck, Decentralized
coordination of plug-in hybrid vehicles for imbalance reduction in a smart
grid, in: Proc. of ACM AAMAS 2011, 2011, pp. 803–810.
[54] H. Lund, W. Kempton, Integration of renewable energy into the transport and
electricity sectors through V2G, Energy Policy 36 (9) (2008) 3578–3587.
[55] S. Galli, A. Scaglione, Z. Wang, For the grid and through the grid: the role of
power line communications in the smart grid, Proceedings of the IEEE 99 (6)
(2011) 998–1027.
[56] National Energy Technology Laboratory, Energy storage – a key enabler of the
smart grid, Tech. rep., U.S. Department of Energy, 2009.
[57] S. Suryanarayanan, F. Mancilla-David, J. Mitra, Y. Li, Achieving the smart grid
through customer-driven microgrids supported by energy storage, in: Proc. of
IEEE ICIT’10, 2010, pp. 884–890.
[58] J. Carr, J. Balda, H. Mantooth, A survey of systems to integrate distributed
energy resources and energy storage on the utility grid, in: Proc. of IEEE
Energy 2030 Conference, 2008, pp. 1–7.
[59] S. Yeleti, Y. Fu, Impacts of energy storage on the future power system, in:
Proc. of IEEE NAPS’10, 2010, pp. 1–7.
[60] C. Marnay, G. Venkataramanan, Microgrids in the evolving electricity
generation and delivery infrastructure, in: Proc. of IEEE Power Engineering
Society General Meeting’06, 2006.
[61] P. Asmus, Microgrids, virtual power plants and our distributed energy future,
The Electricity Journal 23 (10) (2010) 72–82.
[62] C. Colson, M. Nehrir, A review of challenges to real-time power management
of microgrids, in: Proc. of IEEE PES ’09, 2009.
[63] F. Katiraei, R. Iravani, N. Hatziargyriou, A. Dimeas, Microgrids management,
IEEE Power and Energy Magazine 6 (3) (2008) 54–65.
 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
[64] N. Hatziargyriou, H. Asano, R. Iravani, C. Marnay, Microgrids, IEEE Power and
Energy Magazine 5 (4) (2007) 78–94.
[65] D. Cornforth, Role of microgrids in the smart grid, Journal of Electronic
Science and Technology 9 (1) (2011) 9–16.
[66] R. Lasseter, Smart distribution: coupled microgrids, Proceedings of the IEEE
99 (6) (2011) 1074–1082.
[67] H. Qin, W. Zhang, Charging scheduling with minimal waiting in a network of
electric vehicles and charging stations, in: Proc. of ACM VANET’11, 2011, pp.
51–60.
[68] J. Fluhr, K.-H. Ahlert, C. Weinhardt, A stochastic model for simulating the
availability of electric vehicles for services to the power grid, in: Proc. of IEEE
HICSS’10, 2010.
[69] B. Jansen, C. Binding, O. Sundstrom, D. Gantenbein, Architecture and
communication of an electric vehicle virtual power plant, in: Proc. of IEE
SmartGridComm’10, 2010, pp. 149–154.
[70] V. Terzija, G. Valverde, D. Cai, P. Regulski, V. Madani, J. Fitch, S. Skok, M.
Begovic, A. Phadke, Wide-area monitoring, protection, and control of future
electric power networks, Proceedings of the IEEE 99 (1) (2011) 80–93.
[71] A. Johnson, J. Wen, J. Wang, E. Liu, Y. Hu, Integrated system architecture and
technology roadmap toward wampac, in: Proc. of IEEE PES ISGT’11, 2011.
[72] H. Qi, Y. Liu, F. Li, J. Luo, L. He, K. Tomsovic, L. Tolbert, Q. Cao, Increasing the
resolution of wide-area situational awareness of the power grid through
event unmixing, in: Proc. of IEEE HICSS’11, 2011.
[73] D. Tholomier, H. Kang, B. Cvorovic, Phasor measurement units: functionality
and applications, in: Proc. of PSC ’09, 2009.
[74] X. Xie, Y. Xin, J. Xiao, J. Wu, Y. Han, Wams applications in chinese power
systems, IEEE Power and Energy Magazine 4 (1) (2006) 54–63.
[75] Y. Yang, D. Divan, R. Harley, T. Habetler, Design and implementation of power
line sensornet for overhead transmission lines, in: Proc. of IEEE PES ’09, 2009.
[76] J. Wang, Y. Hu, A. Johnson, H. Tram, R. Nasri, System requirements of
visualization platform for wide area situation awareness system, in: Proc. of
IEEE Power and Energy Society General Meeting’10, 2010.
[77] A. Molderink, V. Bakker, M. Bosman, J. Hurink, G. Smit, Management and
control of domestic smart grid technology, IEEE Transactions on Smart Grid 1
(2) (2010) 109–119.
[78] P. Vytelingum, T. Voice, S. Ramchurn, A. Rogers, N. Jennings, Agent-based
micro-storage management for the smart grid, in: Proc. of AAMS’10, 2010, pp.
36–46.
[79] M. Albadi, E. El-Saadany, A summary of demand response in electricity
markets, Electric Power Systems Research 78 (11) (2008) 1989–1996.
[80] S. Singh, J. Ostergaard, Use of demand response in electricity markets: an
overview and key issues, in: Proc. of EEM’10, 2010.
[81] A. Barbato, A. Capone, G. Carello, M. Delfanti, M. Merlo, A. Zaminga, House
energy demand optimization in single and multi-user scenarios, in: Proc. of
IEEE SmartGridComm’11, 2011, pp. 345–350.
[82] Y.-S. Son, T. Pulkkinen, K.-D. Moon, C. Kim, Home energy management system
based on power line communication, IEEE Transactions on Consumer
Electronics 56 (3) (2010) 1380–1386.
[83] C.R.S. Keshav, Direct adaptive control of electricity demand, Tech. Rep. CS-
2010-17, CS Dept., Universty of Waterloo, ON, Canada, 2010.
[84] G. Xiong, C. Chen, S. Kishore, A. Yener, Smart (in-home) power scheduling for
demand response on the smart grid, in: Proc. of IEEE PES ISGT’11, 2011.
[85] H. Saboori, M. Mohammadi, R. Taghe, Virtual power plant (vpp), definition,
concept, components and types, in: Proc. of IEEE APPEEC’11, 2011.
[86] S. You, C. Traeholt, B. Poulsen, A market-based virtual power plant, in: Proc. of
Conf. on Clean Electrical Power’09, 2009, pp. 460–465.
[87] P. Lombardi, M. Powalko, K. Rudion, Optimal operation of a virtual power
plant, in: Proc. of IEEE PES ’09, 2009.
[88] M. Salmani, A. Anzalchi, S. Salmani, Virtual power plant: new solution for
managing distributed generations in decentralized power systems, in: Proc.
of IEEE MASS’10, 2010.
[89] M. Musio, P. Lombardi, A. Damiano, Vehicles to grid (v2g) concept applied to a
virtual power plant structure, in: Proc. of ICEM’10, 2010.
[90] A. Raab, M. Ferdowsi, E. Karfopoulos, I. Unda, S. Skarvelis-Kazakos, P.
Papadopoulos, E. Abbasi, L. Cipcigan, N. Jenkins, N. Hatziargyriou, K. Strunz,
Virtual power plant control concepts with electric vehicles, in: Proc. of IEEE
ISAP’11, 2011.
[91] F. Baker, D. Meyer, Internet protocols for the smart grid, RFC 6272, June 2011.
[92] V. Gungor, D. Sahin, T. Kocak, S. Ergut, C. Buccella, C. Cecati, G. Hancke, Smart
grid technologies: communication technologies and standards, IEEE
Transactions on Industrial Informatics 7 (4) (2011) 529–539.
[93] IEC, Power systems management and associated information exchange.
<http://tc57.iec.ch/index-tc57.html>.
[94] IEC, Electrical energy measurement, tariff and load control. <http://
www.iec.ch>.
[95] IEC, Core IEC Standards. <http://www.iec.ch/smartgrid/standards/>.
[96] IEC, Power system control and associated communications – reference
architecture for object models, services and protocols, IEC/TR 62357.
[97] IEC 61970-301, Energy management system application program interface
(EMS-API) – Part 301: Common information model (CIM) base, 2011.
[98] IEC/TR 61968-1, Application integration at electric utilities – System
interfaces for distribution management – Part 1: Interface architecture and
general requirements, 2003.
[99] IEC 62056-21, Electricity metering – data exchange for meter reading, tariff
and load control – Part 21: Direct local data exchange, 2002.
1693
[100] IEEE, IEEE 2030 Draft Guide for Smart Grid Interoperability of Energy
Technology and Information Technology Operation with the Electric Power
System (EPS), and End-Use Applications and Loads, September 2011.
[101] IEEE Smartgrid. <http://smartgrid.ieee.org/>.
[102] IEEE 1815, IEEE Standard for Electric Power Systems Communications –
Distributed Network Protocol (DNP3) (2010).
[103] IEEE C37.118.2, IEEE Standard for Synchrophasor Data Transfer for Power
Systems, 2011.
[104] IEEE 1547, IEEE Guide for Monitoring, Information Exchange, and Control of
Distributed Resources Interconnected With Electric Power Systems, 2007.
[105] IEEE 1901, IEEE Standard for Broadband over Power Line Networks: Medium
Access Control and Physical Layer Specifications, September 30 2010.
[106] A. Tanenbaum, M. Steen, Distributed Systems: Principles and Paradigms,
second ed., Prentice Hall, Inc., 2006.
[107] E. Lua, J. Crowcroft, M. Pias, R. Sharma, S. Lim, A survey and comparison of
peer-to-peer overlay network schemes, IEEE Communications Surveys &
Tutorials 7 (2) (2005) 72–93.
[108] A. Passarella, A survey on content-centric technologies for the current
Internet: CDN and P2P solutions, Computer Communications 35 (1) (2012)
1–32.
[109] Y.-J. Kim, M. Thottan, V. Kolesnikov, W. Lee, A secure decentralized data-
centric information infrastructure for smart grid, IEEE Communications
Magazine 48 (11) (2010) 58–65.
[110] U.S. Department of Energy (DOE), Communication requirements of smart grid
technologies, 2010. <http://energy.gov/gc/downloads/communications-
requirements-smart-grid-technologies>.
[111] M. Daoud, X. Fernando, On the communication requirements for the smart
grid, Energy and Power Engineering 3 (2011) 53–60.
[112] IEEE 1646, Ieee standard communication delivery time performance
requirements for electric power substation automation, 2005.
[113] M. Chenine, L. Nordstrom, P. Johnson, Factors in assessing performance of
wide area communication networks for distributed control of power systems,
in: Proc. of IEEE Lausanne Power Tech’07, 2007, pp. 1682–1687.
[114] D. Niyato, P. Wang, Z. Han, E. Hossain, Impact of packet loss on power
demand estimation and power supply cost in smart grid, in: Proc. of IEEE
WCNC’11, 2011, pp. 2024–2029.
[115] J. Zhou, R. Hu, Y. Qian, Scalable distributed communication architectures to
support advanced metering infrastructure in smart grid, IEEE Transactions on
Parallel and Distributed Systems 23 (9) (2012) 1632–1642.
[116] D. Popa, J. Jetcheva, N. Dejean, R. Slazar, K. Hui, J. amd Monden, Applicability
statement for the routing protocol for low power and lossy networks (RPL) in
AMI Networks, Internet Draft, 2011
[117] H. Khurana, M. Hadley, N. Lu, D. Frincke, Smart-grid security issues, IEEE
Security Privacy 8 (1) (2010) 81–85.
[118] P. Serrano, A. de la Oliva, P. Patras, V. Mancuso, A. Banchs, Greening wireless
communications: status and future directions, Computer Communications 35
(14) (2012) 1651–1661.
[119] S. Galli, O. Logvinov, Recent developments in the standardization of power
line communications within the IEEE, IEEE Communications Magazine 46 (7)
(2008) 64–71.
[120] M. Bauer, W. Plappert, C. Wang, K. Dostert, Packet-oriented communication
protocols for smart grid services over low-speed plc, in: Proc. of IEEE
ISPLC’09, 2009, pp. 89–94.
[121] HomePlug powerline alliance, 2013. <https://www.homeplug.org>.
[122] IEEE P905.1, Convergent digital home network for heterogeneous
technologies, IEEE Draft Standard P1905.1/D02, December 2011.
[123] J. Buysse, M. De Leenheer, B. Dhoedt, C. Develder, Providing resiliency for
optical grids by exploiting relocation: a dimensioning study based on ILP,
Computer Communications 34 (12) (2011) 1389–1398.
[124] IEEE 1138, Testing and Performance for Optical Ground Wire (OPGW) for Use
on Electric Utility Power Lines, 2009.
[125] M. Maier, Fiber-wireless sensor networks (fi-wsns) for smart grids, in: Proc.
of IEEE ICTON’11, 2011, pp. 1–4.
[126] L. Jianming, Z. Bingzhen, Z. Zichao, The smart grid multi-utility services
platform based on power fiber to the home, in: Proc. of IEEE CCIS’11, 2011,
pp. 17–22.
[127] M. McGarry, M. Reisslein, M. Maier, WDM Ethernet passive optical networks,
IEEE Communications Magazine 44 (2) (2006) 15–22.
[128] IEEE, IEEE 802.15.5: Mesh Topology Capability in Wireless Personal Area
Networks (WPANs), IEEE Std 802.15.5-2009, 2009.
[129] C. Yoon, H. Cha, Experimental analysis of IEEE 802.15.4a CSS ranging and its
implications, Computer Communications 34 (11) (2011) 1361–1374.
[130] K.-H. Chang, B. Mason, The IEEE 802.15.4g standard for smart utility
networks, in: Proc. of IEEE SmartGridComm, 2012, pp. 476–480.
[131] G. Anastasi, M. Conti, M. Di Francesco, A comprehensive analysis of the MAC
unreliability problem in IEEE 802.15.4 wireless sensor networks, IEEE
Transactions on Industrial Informatics 7 (1) (2011) 52–65.
[132] M. Di Francesco, G. Anastasi, M. Conti, S. Das, V. Neri, Reliability and energy-
efficiency in IEEE 802.15.4/ZigBee sensor networks: an adaptive and cross-
layer approach, IEE Journal on Selected Areas in Communications 29 (8)
(2011) 1508–1524.
[133] International Society of Automation, 2013. <http://www.isa.org/>.
[134] IEC 62591, Industrial communication networks – Wireless communication
network and communication profiles – WirelessHART, 2010.
[135] ZigBee Alliance. <http://www.zigbee.org/>.
1694 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
[136] N. Fourty, A. van den Bossche, T. Val, An advanced study of energy
consumption in an IEEE 802.15.4 based network: Everything but the truth
on 802.15.4 node lifetime, Computer Communications 35 (14) (2012) 1759–
1767.
[137] ZigBee Alliance, ZigBee smart energy – profile specification, Version 1.1,
March 2011.
[138] IP for Smart Objects (IPSO) Alliance 2013. <http://www.ipso-alliance.org/>.
[139] T. Winter, P. Thubert, A. Brandt, T. Clausen, J. Hui, R. Kelsey, P. Levis, K. Pister,
R. Struik, J. Vasseur, RPL: IPv6 routing protocol for low-power and lossy
networks, IETF RFC 6550, March 2012.
[140] IEEE, IEEE Standard for Local and Metropolitan Area Networks – Part11:
Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY)
Specifications, IEEE Std 802.11-2007, 2007 (Revision of IEEE Std 802.11-
1999).
[141] IEEE, Part11: Wireless LAN Medium Access Control (MAC) and Physical Layer
(PHY) Specifications Amendment 5: Enhancements for Higher Throughput,
IEEE Std 802.11n-2009, 2009.
[142] E. Ancillotti, R. Bruno, M. Conti, Design and performance evaluation of
throughput-aware rate adaptation protocols for IEEE 802.11 wireless
networks, Performance Evaluation 66 (12) (2009) 811–825.
[143] S.-L. Tsao, C.-H. Huang, A survey of energy efficient MAC protocols for IEEE
802. 11 WLAN, Computer Communications 34 (1) (2011) 54–67.
[144] IEEE, Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer
(PHY) Specifications Amendment 8: Medium Access Control (MAC) Quality of
Service Enhancements, IEEE Std 802.11e-2005, 2005.
[145] IEEE, Part11: Wireless LAN Medium Access Control (MAC) and Physical Layer
(PHY) Specifications Amendment 10: Mesh Networking, EEE Std 802.11s-
2011, 2009.
[146] IEEE, Part11: Wireless LAN Medium Access Control (MAC) and Physical Layer
(PHY) Specifications Amendment 6: Wireless Access in Vehicular
Environments, IEEE Std 802.11p-2010, 2010.
[147] C. Eklund, R. Marks, K. Stanwood, IEEE Standard 802.16: a technical overview
of the WirelessMA air interface for broadband wireless access, IEEE
Communications Magazine (2002) 98–107.
[148] IEEE, IEEE Standard for Local and Metropolitan Area Networks – Part 16: Air
Interface for Fixed and Mobile Broadband Wireless Access System, IEEE Std
802.16-2009, 2009.
[149] IEEE, Air Interface for Fixed Broadband Wireless Access Systems – Multihop
Relay Specification, IEEE Std 802.16j-2009, 2009.
[150] IEEE, Air Interface for Broadband Wireless Access Systems Amendment 3:
Advanced Air Interface, IEEE Std 802.16m-2011, 2011.
[151] X. Zhang, Y. Gao, G. Zhang, G. Bi, CDMA2000 cellular network based SCADA
system, in: Proc. of IEEE PowerCon’02, vol. 2, 2002, pp. 1301–1306.
[152] H. Tan, C. Lee, V. Mok, Automatic power meter reading system using GSM
network, in: Proc. of IEEE IPEC’07, 2007, pp. 465–469.
[153] A. Ghosh, R. Ratasuk, N. Mondal, B., Mangalvedhe, T. Thomas, LTE-advanced:
next-generation wireless boradband technology, IEEE Wireless
Communications 17 (3) (2010) 10–22.
[154] Y. Hu, V. Li, Satellite-based Internet: a tutorial, IEEE Communications
Magazine 39 (3) (2001) 154–162.
[155] P. Baronti, P. Pillai, V. Chook, S. Chessa, A. Gotta, Y. Hu, Wireless sensor
networks: a survey on the state of the art and the 802.15.4 and zigbee
standards, Computer Communications 30 (7) (2007) 1655–1695.
[156] M. Gharbaoui, L. Valcarenghi, R. Bruno, B. Martini, M. Conti, P. Castoldi, An
advanced smart management system for electric vehicle recharge, in: Proc. of
IEEE IEVC’12, 2012, pp. 1–8.
[157] J. Marquez-Barja, C. Calafate, J.-C. Cano, P. Manzoni, An overview of vertical
handover techniques: algorithms, protocols and tools, Computer
Communications 34 (8) (2011) 985–997.
[158] M. Zekri, B. Jouaber, D. Zeghlache, A review on mobility management and
vertical handover solutions over heterogeneous wireless networks, Computer
Communications 35 (17) (2012) 2055–2068.
[159] R. Bruno, M. Conti, E. Gregori, Mesh networks: commodity multihop ad hoc
networks, IEEE Communications Magazine 43 (3) (2005) 123–131.
[160] N. Psimogiannos, A. Sgora, D. Vergados, An IMS-based network architecture
for WiMAX-UMTS and WiMAX-WLAN interworking, Computer
Communications 34 (9) (2011) 1077–1099.
[161] D. Corujo, M. Lebre, D. Gomes, R. Aguiar, MINDiT: A framework for media
independent access to things, Computer Communications 35 (15) (2012)
1772–1785.
[162] Y. Gobena, A. Durai, M. Birkner, V. Pothamsetty, V. Varakantam, Practical
architecture considerations for Smart Grid WAN network, in: Proc. of IEEE/
PES PSCE’11, 2011.
[163] S. Keshav, C. Rosenberg, How internet concepts and technologies can help
green and smarten the electrical grid, in: Proc. of ACM SIGCOMM workshop
on Green networking, 2010, pp. 35–40.
[164] Y. Xu, J. Zhang, W. Wang, A. Juneja, S. Bhattacharya, Energy router:
Architectures and functionalities toward Energy Internet, in: Proc. of IEEE
SmartGridComm’11, 2011, pp. 31–36.
[165] P. Nyeng, J. Ostergaard, Information and communications systems for
control-by-price of distributed energy resources and flexible demand, IEEE
Transactions on Smart Grid 2 (2) (2011) 334–341.
[166] M. Conti, S. Chong, S. Fdida, W. Jia, H. Karl, Y.-D. Lin, P. Mahonen, M. Maier, R.
Molva, S. Uhlig, M. Zukerman, Research challenges towards the Future
Internet, Computer Communications 34 (18) (2011) 2115–2134.
[167] S. Paul, J. Pan, R. Jain, Architectures for the future networks and the next
generation internet: a survey, Computer Communications 34 (1) (2011) 2–
42.
[168] D. Bakken, A. Bose, K.M. Chandy, P. Khargonekar, A. Kuh, S. Low, A. von Meier,
K. Poolla, P. Varaiya, F. Wu, GRIP – Grids with intelligent periphery: Control
architectures for Grid2050, in: Proc. of IEEE SmartGridComm’11, 2011, pp. 7–
12.
[169] Q. Yang, J. Barria, C. Aramburo, A communication system architecture for
regional control of power distribution networks, in: Proc. of IEEE INDIN’09,
2009, pp. 372–377.
[170] Z. Fadlullah, M. Fouda, N. Kato, A. Takeuchi, N. Iwasaki, Y. Nozaki, Toward
intelligent machine-to-machine communications in smart grid, IEEE
Communications Magazine 49 (4) (2011) 60–65.
[171] D. Niyato, L. Xiao, P. Wang, Machine-to-machine communications for home
energy management system in smart grid, IEEE Communications Magazine
49 (4) (2011) 53–59.
[172] C. Bennett, D. Highfill, Networking AMI smart meters, in: Proc of IEEE
ENERGY 2030 Conference, 2008, pp. 1–8.
[173] J.-Y. Cheng, M.-H. Hung, J.-W. Chang, A ZigBee-based power monitoring
system with direct load control capabilities, in: Proc. of Conf. on Networking,
Sensing and Control’07, 2007, pp. 895–900.
[174] S.-W. Luan, J.-H. Teng, S.-Y. Chan, L.-C. Hwang, Development of a smart power
meter for AMI based on ZigBee communication, in: Proc. of IEEE PEDS’09,
2009, pp. 661–665.
[175] M. LeMay, R. Nelli, G. Gross, C. Gunter, An integrated architecture for demand
response communications and control, in: Proc. of IEEE HICCS’08, 2008, pp.
174–183.
[176] M. Erol-Kantarci, H. Mouftah, Wireless sensor networks for cost-efficient
residential energy management in the smart grid, IEEE Transactions on Smart
Grid 2 (2) (2011) 314–325.
[177] B. Lichtensteiger, B. Bjelajac, C. Muller, C. Wietfeld, RF mesh systems for
smart metering: system architecture and performance, in: Proc. of IEEE
SmartGridComm’10, 2010, pp. 379–384.
[178] P. Kulkarni, S. Gormus, F. Fan, F. Ramos, AMI Mesh Networks – a practical
solution and its performance evaluation, IEEE Transactions on Smart Grid 3
(3) (2012) 1469–1481.
[179] B. Heile, Smart grids for green communications [Industry Perspectives], IEEE
Wireless Communications 17 (3) (2010) 4–6.
[180] Mobile Ad Hoc Networking: the Cutting Edge Technologies, IEEE Press and
John Viley & sons, Inc., New York, NY, USA, 2013.
[181] R. Bruno, M. Conti, A. Pinizzotto, Routing Internet traffic in heterogeneous
mesh networks: analysis and algorithms, Performance Evaluation 68 (9)
(2011) 841–858.
[182] Y. Ding, L. Xiao, Channel allocation in multi-channel wireless mesh networks,
Computer Communications 34 (7) (2011) 803–815.
[183] E. Ancillotti, R. Bruno, M. Conti, A. Pinizzotto, Load-aware routing in mesh
networks: models algorithms and experimentation, Computer
Communications 34 (8) (2011) 948–961.
[184] A. de la Oliva, A. Banchs, P. Serrano, Throughput and energy-aware routing for
802.11 based mesh networks, Computer Communications 35 (12) (2012)
1433–1446.
[185] A. Jabbar, J. Rohrer, V. Frost, J.P. Sterbenz, Survivable millimeter-wave mesh
networks, Computer Communications 34 (16) (2011) 1942–1955.
[186] J. Al-Karaki, A. Kamal, Routing techniques in wireless sensor networks: a
survey, IEEE Wireless Communications 11 (6) (2004) 6–28.
[187] K. Akkaya, M. Younis, A survey on routing protocols for wireless sensor
networks, Ad Hoc Networks 3 (2005) 325–349.
[188] M. Di Francesco, S. Das, G. Anastasi, Data collection in wireless sensor
networks with mobile elements: a survey, ACM Transactions on Sensor
Networks 8 (1) (2011) 1–31.
[189] R. Mini, A. Loureiro, Energy-efficient design of wireless sensor networks
based on finite energy budget, Computer Communications 35 (14) (2012)
1736–1748.
[190] G. Anastasi, M. Conti, M. Di Francesco, A. Passarella, Energy conservation
in wireless sensor networks: a survey, Ad Hoc Networks 7 (3) (2009)
537–568.
[191] V. Sharma, U. Mukherji, V. Joseph, S. Gupta, Optimal energy management
policies for energy harvesting sensor nodes, IEEE Transactions on Wireless
Communications 9 (4) (2010) 1326–1336.
[192] J. Zhang, X. Shen, S. Tang, G. Dai, Energy efficient joint data aggregation and
link scheduling in solar sensor networks, Computer Communications 34 (18)
(2011) 2217–2226.
[193] W. Seah, Y. Tan, A. Chan, Research in Energy Harvesting Wireless Sensor
Networks and the Challenges, Springer, Verlang, 2012.
[194] L. Hanzo, R. Tafazolli, A survey of QoS routing solutions for mobile ad hoc
networks, IEEE Communications Surveys & Tutorials 9 (2) (2007) 50–70.
[195] M. Campista, P. Esposito, I. Moraes, L. Costa, O. Duarte, D. Passos, C. de
Albuquerque, D. Saade, M. Rubinstein, Routing metrics and protocols for
wireless mesh networks, IEEE Network 22 (1) (2008) 6–12.
[196] F. Cadger, K. Curran, J. Santos, S. Moffett, A survey of geographical routing in
wireless ad-hoc networks, IEEE Communications Surveys & Tutorials (2012)
1–33.
[197] M. Biagi, L. Lampe, Location assisted routing techniques for power line
communication in smart grids, in: Proc. of IEEE SmartGridComm’10, 2010,
pp. 274–278.
 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
[198] G. Iyer, P. Agrawal, E. Monnerie, R. Cardozo, Performance analysis of wireless
mesh routing protocols for smart utility networks, in: Proc. of IEEE
SmartGridComm’11, 2011, pp. 114–119.
[199] B. Karimi, V. Namboodiri, V. Aravinthan, W. Jewell, Feasibility, challenges,
and performance of wireless multi-hop routing for feeder level
communication in a smart grid, in: Proc. of IFIP Conf. on E-energy’11, 2001.
[200] J.-C. Chiu, K.-M. Yang, Y.-C. Huang, M.-S. Wu, Dynamic multi-channel multi-
path routing protocol for smart grid, in: Computer Science and Convergence,
Lecture Notes in Electrical Engineering, vol. 114, 2012, pp. 725–733.
[201] J.-S. Jung, K.-W. Lim, J.-B. Kim, Y.-B. Ko, Y. Kim, S.-Y. Lee, Improving IEEE
802.11s wireless mesh networks for reliable routing in the smart grid
infrastructure, in: Proc. of IEEE ICC’11, 2011, pp. 1–5.
[202] H. Gharavi, B. Hu, Multigate communication network for smart grid,
Proceedings of the IEEE 9 (6) (2011) 1028–1045.
[203] T. Iwao, K. Yamada, M. Yura, Y. Nakaya, A. Cardenas, S. Lee, R. Masuoka,
Dynamic data forwarding in wireless mesh networks, in: Proc. of IEEE
SmartGridComm’10, 2010, pp. 385–390.
[204] S. Cespedes, A. Cardenas, T. Iwao, Comparison of data forwarding
mechanisms for AMI networks, in: Proc. of IEEE PES ISGT’12, 2012, pp. 1–8.
[205] T. Watteyne, A. Molinaro, M. Richichi, M. Dohler, From MANET To IETF ROLL
standardization: a paradigm shift in WSN routing protocols, IEEE
communications Surveys & Tutorials 13 (4) (2011) 688–707.
[206] E. Ancillotti, R. Bruno, M. Conti, The role of the RPL routing protocol for
smart grid communications, IEEE Communications Magazine 51 (1) (2013)
75–83.
[207] N. Bressan, L. Bazzaco, N. Bui, P. Casari, L. Vangelista, M. Zorzi, The
deployment of a smart monitoring system using wireless sensor and
actuator networks, in: Proc. of IEEE SmartGridComm’10, 2010, pp. 49–54.
[208] N. Accettura, L. Grieco, G. Boggia, P. Camarda, Performance analysis of the rpl
routing protocol, in: Proc. of IEEE ICM’11, 2011, pp. 767–772.
[209] J. Tripathi, J. de Oliveira, J. Vasseur, A performance evaluation study of rpl:
routing protocol for low power and lossy networks, in: Proc. of IEEE CISS’10,
2010, pp. 1–6.
[210] E. Ancillotti, R. Bruno, M. Conti, RPL routing protocol in advanced metering
infrastructures: an analysis of the unreliability problems, in: Proc. of IFIP
SUSTAINIT’12, 2012, pp. 1–10.
[211] D. Wang, Z. Tao, J. Zhang, A. Abouzeid, RPL based routing for advanced
metering infrastructure in smart grid, in: Proc. of IEEE ICC’10, 2010, pp. 1–6.
[212] V. Borges, M. Curado, M. Monteiro, Cross-layer routing metrics for mesh
networks: current status and research directions, Computer Communications
34 (6) (2011) 681–703.
[213] J. Tripathi, J. de Oliveira, J. Vasseur, Applicability study of RPL with local repair
in smart grid substation networks, in: Proc. of IEEE SmartGridComm’10,
2010, pp. 262–267.
[214] G. Piro, L. Grieco, G. Boggia, P. Camarda, QoS in Wireless LAN: a comparison
between feedback-based and earliest due-date approaches, Computer
Communications 35 (3) (2012) 298–308.
[215] S.-J. Wu, S.-Y. Huang, K.-F. Huang, Efficient quality of service scheduling
mechanism for WiMAX networks, Computer Communications 35 (8) (2012)
936–951.
[216] R. Braden, D. Clark, S. Shenker, Integrated services in the internet
architecture: an overview, IETF RFC 1633, June 1994.
[217] S. Blake, D. Black, M. Carlson, E. Davies, Z. Wang, W. Weiss, An architecture
for differentiated services, IETF RFC 2475, December 1998.
[218] E. Rosen, A. Viswanathan, R. Callon, Multiprotocol Label switching
architecture, IETF RFC 3031, December 2001.
[219] K. Hopkinson, G. Roberts, X. Wang, J. Thorp, Quality-of-service considerations
in utility communication networks, IEEE Transactions on Power Delivery 24
(3) (2009) 1465–1474.
[220] Y. Ting, Z. Zhidong, L. Ang, W. Jiaowen, W. Ming, New IP QoS algorithm
applying for communication sub-networks in smart grid, in: Proc. of IEEE
APPEEC’10, 2010, pp. 1–4.
[221] A. Varela, T. Vazao, G. Arroz, Poviding service differentiation in pure IP-based
networks, Computer Communications 35 (1) (2012) 33–46.
[222] R. Garroppo, S. Giordano, L. Tavanti, A survey on multi-constrained optimal
path computation: exact and approximate algorithms, Computer Networks
54 (2010) 3081–3107.
[223] H. Li, W. Zhang, QoS Routing in smart grid, in: Proc. of IEEE GLOBECOM’10,
2010, pp. 1–6.
[224] C.-L. Chuang, Y.-C. Wang, C.-H. Lee, M.-Y. Liu, Y.-T. Hsiao, J.-A. Jiang, An
adaptive routing algorithm over packet switching networks for operation
monitoring of power transmission systems, IEEE Transactions on Power
Delivery 25 (2) (2010) 882–890.
[225] D. Bertsekas, R. Gallager, Data Networks, Prentice Hall, 1992.
[226] T. Zourzouvillys, E. Rescorla, An introduction to standards-based VoIP: SIP,
RTP, and friends, IEEE Internet Computing 14 (2) (2010) 69–73.
[227] Y.-J. Kim, V. Kolesnikov, H. Kim, M. Thottan, Sstp: a scalable and secure
transport protocol for smart grid data collection, in: Proc. of IEEE
SmartGridComm’11, 2011, pp. 161–166.
[228] T. Khalifa, K. Naik, M. Alsabaan, A. Nayak, N. Goel, Transport protocol for
smart grid infrastructure, in: Proc. of IEEE ICUFN’10, 2010, pp. 320–325.
[229] M. Allalouf, G. Gershinsky, L. Lewin-Eytan, J. Naor, Data-quality-aware
volume reduction in smart grid networks, in: Proc. of IEEE
SmartGridComm’11, 2011, pp. 120–125.
[230] H. Karl, A. Willig, Protocols and Architectures for Wireless Sensor Networks,
Wiley, 2005.
1695
[231] D. Germanus, I. Dionysiou, H. Gjermundrod, A. Khelil, N. Suri, D. Bakken, C.
Hauser, Leveraging the next-generation power grid: Data sharing and
associated partnerships, in: Proc. of IEEE PES ISGT Europe’10, 2010, pp. 1–8.
[232] D. Villa, C. Martí, F. Villanueva, F. Moya, J. López, Middleware-based
management for smart grids, in: Proc. of ICCE’11, 2011, pp. 545–546.
[233] A. Rogers, N. Jennings, Intelligent agents for the smart grid, PerAda Magazine.
[234] K. Tomsovic, D. Bakken, V. Venkatasubramanian, A. Bose, Designing the next
generation of real-time control, communication, and computations for large
power systems, Proceedings of the IEEE 93 (5) (2005) 965–979.
[235] C. Hauser, D. Bakken, A. Bose, A failure to communicate: next generation
communication requirements, technologies, and architecture for the electric
power grid, IEEE Power and Energy Magazine 3 (2) (2005) 47–55.
[236] C. Hauser, D. Bakken, I. Dionysiou, K. Gjermundrod, V. Irava, J. Helkey, A. Bose,
Security, trust, and QoS in next-generation control and communication for
large power systems, International Journal of Critical Infrastructures 4 (1/2)
(2008) 3–16.
[237] P. Eugster, P. Felber, R. Guerraoui, A.-M. Kermarrec, The many faces of
publish/subscribe, ACM Computer Survey 35 (2003) 114–131.
[238] H. Gjermundrod, D. Bakken, C. Hauser, A. Bose, Gridstat: a flexible QoS-
managed data dissemination framework for the power grid, IEEE
Transactions on Power Delivery 24 (1) (2009) 136–143.
[239] A. Liakopoulos, A. Zafeiropoulos, P. Gouvas, Smart grid evolution through the
exploitation of future internet networking concepts, in: Proc. of IEEE ICIT’12,
2012, pp. 415–420.
[240] Y.-J. Kim, J. Lee, G. Atkinson, H. Kim, M. Thottan, SeDAX: a scalable, resilient,
and secure platform for smart grid communications, IEEE Journal on Selected
Areas in Communication 30 (6) (2012) 1119–1136.
[241] S. Rusitschka, C. Gerdes, K. Eger, A low-cost alternative to smart metering
infrastructure based on peer-to-peer technologies, in: Proc. of IEEE EEM’09,
2009, pp. 1–6.
[242] O.M. Group, Common Object Request Broker Archictecture, 2002.
[243] M. Henning, A new approach to object-oriented middleware, IEEE Internet
Computing 8 (1) (2004) 66–75.
[244] D. Villa, C. Martin, F. Villanueva, F. Moya, J. Lopez, A dynamically
reconfigurable architecture for smart grids, IEEE Transactions on Consumer
Electronics 57 (2) (2011) 411–419.
[245] K.J.S., S. Kim, An object-based middleware for home network supporting the
interoperability among heterogeneous devices, in: Proc. of IEEE ICCE’11,
2011, pp. 585–586.
[246] C.-F. Lai, Y.-W. Ma, S.-Y. Chang, H.-C. Chao, Y.-M. Huang, OSGi-based services
architecture for cyber-physical home control systems, Computer
Communications 34 (2) (2011) 184–191.
[247] H. Kim, S. Lee, H. Kim, H. Kim, Implementing home energy management
system with UPnP and mobile applications, Computer Communications 36
(1) (2012) 51–62.
[248] K. Gama, L. Touseau, D. Donsez, Combining heterogeneous service
technologies for building an Internet of Things middleware, Computer
Communications 35 (4) (2012) 405–417.
[249] M. Adamiak, D. Baigent, R. Mackiewicz, Iec 61850 communication networks
and systems in substations: an overview for users, The Protection & Control
Journal.
[250] Y. Yaqin, W. Zhongxi, H. Chunming, A Novel implementation of IEC 61970 CIS
based on Ice middleware, in: Proc. of IEEE/PES Transmission and Distribution
Conference and Exhibition: Asia and Pacific, 2005, pp. 1–4.
[251] C. Xue, X. Huang, Z. Zhang, The information integration scheme of intelligent
substation, in: Proc. of IEEE CSAE’11, 2011, pp. 361–365.
[252] M. Amin, Toward self-healing energy infrastructure systems, IEEE Computer
Applications in Power 14 (1) (2001) 20–28.
[253] F. Brazier, J. Kephart, H. Van Dyke Parunak, M. Huhns, Agents and service-
oriented computing for autonomic computing: a research agenda, IEEE
Internet Computing 13 (3) (2009) 82–87.
[254] C. Rehtanz, Autonomous Systems and Intelligent Agents in Power System
Control and Operation, Springer, 2003.
[255] S. McArthur, E. Davidson, V. Catterson, A. Dimeas, N. Hatziargyriou, F. Ponci,
T. Funabashi, Multi-agent systems for power engineering applications – Part
I: Technologies, standards, and tools for building multi-agent systems, IEEE
Transactions on Power Systems 22 (4) (2007) 1753–1759.
[256] T. Nagata, H. Sasaki, R. Yokoyama, Power system restoration by joint usage of
expert system and mathematical programming approach, IEEE Transactions
on Power Systems 10 (3) (1995) 1473–1479.
[257] T. Nagata, H. Sasaki, A multi-agent approach to power system restoration,
IEEE Transactions on Power Systems 17 (2) (2002) 457–462.
[258] J. Hossack, S. McArthur, J. McDonald, J. Stokoe, T. Cumming, A multi-agent
approach to power system disturbance diagnosis, in: Proc. of Conf. on Power
System Management and Control’ 02, 2002, pp. 317–322.
[259] J. Hossack, J. Menal, S. McArthur, J. McDonald, A multiagent architecture for
protection engineering diagnostic assistance, IEEE Transactions on Power
Systems 18 (2) (2003) 639–647.
[260] S. McArthur, E. Davidson, J. Hossack, J. McDonald, Automating power system
fault diagnosis through multi-agent system technology, in: Proc. of IEEE Conf.
on System Sciences, 2004, pp. 1–4.
[261] A. Deshmukh, F. Ponci, A. Monti, M. Riva, L. Cristaldi, Multi-agent system for
diagnostics, monitoring and control of electric systems, in: Proc. of IEEE Conf.
on Intelligent Systems Application to Power Systems’05, 2005, pp. 1–6.
[262] Z. Jiang, Agent-based control framework for distributed energy resources
microgrids, in: Proc. of IEEE/ACM IAT ’06, 2006, pp. 646–652.
1696 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
[263] M. Baran, I. El-Markabi, A multiagent-based dispatching scheme for
distributed generators for voltage support on distribution feeders, IEEE
Transactions on Power Systems 22 (1) (2007) 52–59.
[264] M. Lu, C. Chen, The design of multi-agent based distributed energy system, in:
Proc. of IEEE SMC’09, 2009, pp. 2001–2006.
[265] S. Ramchurn, P. Vytelingum, A. Rogers, N. Jennings, Agent-based control for
decentralised demand side management in the smart grid, in: Proc. of
AAMAS’11, 2011, pp. 5–12.
[266] C. Wei, H. Hu, Q. Chen, G. Yang, Learning agents for storage devices
management in the smart grid, in: Proc. of IEEE Conf. on CiSE’10, 2010, pp.
1–4.
[267] S. Vandael, N. Boucke, T. Holvoet, G. Deconinck, Decentralized demand side
management of plug-in hybrid vehicles in a smart grid, in: Proc. of ACM
AAMAS’10, 2010, pp. 67–74.
[268] S. Vandael, T. Holvoet, G. Deconinck, A multi-agent system for managing
plug-in hybrid vehicles as primary reserve capacity in a smart grid, in: Proc.
of AAMAS’11, 2011.
[269] S. Kamboj, W. Kempton, K. Decker, Deploying power grid-integrated electric
vehicles as a multi-agent system, in: Proc. of AAMAS’11, 2011, pp. 13–20.
[270] D. Buse, P. Sun, Q. Wu, J. Fitch, Agent-based substation automation, IEEE
Power and Energy Magazine 1 (2) (2003) 50–55.
[271] M. Pipattanasomporn, H. Feroze, S. Rahman, Multi-agent systems in a
distributed smart grid: design and implementation, in: Proc. of IEEE/PES
Power Systems Conference and Exposition, 2009, pp. 1–8.
[272] T. Nagata, H. Fujita, H. Sasaki, Decentralized approach to normal operations
for power system network, in: Proc. of IEEE Conf. on Intelligent Systems
Application to Power Systems’05, 2005, pp. 1–6.
[273] L. Cristaldi, A. Monti, R. Ottoboni, F. Ponci, Multiagent based power systems
monitoring platform: a prototype, in: Proc. of IEEE Power Tech Conference,
vol. 2, 2003.
[274] P. Vytelingum, S. Ramchurn, T. Voice, A. Rogers, N. Jennings, Trading agents
for the smart electricity grid. Autonomous Agents And MultiAgent Systems,
in: Proc. of ACM AAMAS’10, vol. 1, 2010, pp. 897–904.
[275] P. Reddy, M. Veloso, Learned behaviors of multiple autonomous agents in
smart grid markets, in: Proc. of AAAI’11, 2011, pp. 1–8.
[276] R. Mallik, N. Sarda, H. Kargupta, S. Bandyopadhyay, Distributed data mining
for sustainable smart grids, in: Proc. of ACM SustKDD’11, 2011, pp. 1–6.
[277] A. Dimeas, N. Hatziargyriou, Operation of a multiagent system for microgrid
control, IEEE Transactions on Power Systems 20 (3) (2005) 1447–1455.
[278] F. Meng, R. Akella, M. Crow, B. McMillin, Distributed grid intelligence for
future microgrid with renewable sources and storage, in: Proc. of NAPS’10,
2010.
[279] Y. Xu, W. Liu, Novel multiagent based load restoration algorithm for
microgrids, IEEE Transactions on Smart Grid 2 (1) (2011) 152–161.
[280] M. Wooldridge, Intelligent agents, in: G. Weis (Ed.), Multiagent Systems: A
Modern Approach to Distributed Artificial Intelligence, MIT Press, 1999, pp.
27–77.
[281] A. Garcia, J. Oliver, D. Gosch, An intelligent agent-based distributed
architecture for smart-grid integrated network management, in: Proc. of
IEEE LCN’10, 2010, pp. 1013–1018.
[282] Foundation for Intelligent Physical Agents (FIPA), FIPA ACL message structure
specification, December 2002.
[283] V. Catterson, E. Davidson, S. McArthur, Issues in integrating existing multi-
agent systems for power engineering applications, in: Proc. of Conf. on
Intelligent Systems Application to Power Systems’05, 2005, pp. 1–6.
[284] S. McArthur, E. Davidson, V. Catterson, A. Dimeas, N. Hatziargyriou, F. Ponci,
T. Funabashi, Multi-agent systems for power engineering applications – Part
II: Technologies, standards, and tools for building multi-agent systems, IEEE
Transactions on Power Systems 22 (4) (2007) 1753–1759.
[285] F. Bellifemine, G. Caire, D. Greenwood, Developing Multi-Agent Systems with
JADE, Wiley, 2007.
[286] S. McArthur, S. Strachan, G. Jahn, The design of a multi-agent transformer
condition monitoring system, IEEE Transactions on Power Systems 19 (4)
(2004) 1845–1852.
[287] E. Davidson, S. McArthur, J. McDonald, T. Cumming, I. Watt, Applying multi-
agent system technology in practice: automated management and analysis of
scada and digital fault recorder data, IEEE Transactions on Power Systems 21
(2) (2006) 559–567.
[288] K. Hopkinson, X. Wang, R. Giovanini, J. Thorp, K. Birman, D. Coury, EPOCHS: a
platform for agent-based electric power and communication simulation built
from commercial off-the-shelf components, IEEE Transactions on Power
Systems 21 (2) (2006) 548–558.
[289] The network simulator NS-2, 2011. http://www.isi.edu/nsnam/ns/.
[290] G. Pagani, M. Aiello, Service orientation and the smart grid state and trends,
Service Oriented Computing and Applications 6 (3) (2012) 267–282.
[291] M. Irving, G. Taylor, P. Hobson, Plug in to grid computing, IEEE Power and
Energy Magazine 2 (2) (2004) 40–44.
[292] S. Rusitschka, K. Eger, C. Gerdes, Smart grid data cloud: a model for utilizing
cloud computing in the smart grid domain, in: Proc. of IEEE
SmartGridComm’10, 2010, pp. 483–488.
[293] X. Fang, S. Misra, G. Xue, D. Yang, Managing smart grid information in the
cloud: opportunities, model, and applications, IEEE Network 26 (4) (2012)
32–38.
[294] J. Zhu, Web services provide the power to integrate, IEEE Power and Energy
Magazine 1 (6) (2003) 40–49.
[295] P.-Y. Chen, S.-M. Cheng, K.-C. Chen, Smart attacks in smart grid
communication networks, IEEE Communications Magazine 50 (8) (2012)
24–29.
[296] X. Li, X. Liang, R. Lu, X. Shen, X. Lin, H. Zhu, Securing smart grid: cyber attacks,
countermeasures, and challenges, IEEE Communications Magazine 50 (8)
(2012) 38–45.
[297] J. Lin, W. Yu, X. Yang, G. Xu, W. Zhao, On false data injection attacks against
distributed energy routing in smart grid, in: Proc. of IEEE/ACM ICCPS’12,
2012, pp. 183–192.
[298] H. Beitollahi, G. Deconinck, Analyzing well-known countermeasures against
distributed denial of service attacks, Computer Communications 35 (11)
(2012) 1312–1332.
[299] F. Mámol, C. Sorge, O. Ugus, G. Pérez, Do not snoop my habits: preserving
privacy in the smart grid, IEEE Communications Magazine 50 (5) (2012) 166–
172.
[300] National Institute of Standards and Technology (NIST), Guidelines for smart
grid cyber security: smart grid cyber security strategy, architecture, and
high-level requirements, NISTIR 7628, vol. 1, August 2010.
[301] S. Sridhar, A. Hahn, M. Govindarasu, Cyber-physical system security for the
electric power grid, Proceedings of the IEEE 100 (1) (2012) 210–224.
[302] A. Metke, R. Ekl, Security technology for smart grid networks, IEEE
Transactions on Smart Grid 1 (1) (2010) 99–107.
[303] D. He, C. Chen, J. Bu, S. Chan, Y. Zhang, M. Guizani, Secure service provision in
smart grid communications, IEEE Communications Magazine 50 (8) (2012)
53–61.
[304] T. Baumeister, Adapting pki for the smart grid, in: Proc. of IEEE
SmartGridComm’11, 2011, pp. 249–254.
[305] D. Wu, C. Zhou, Fault-tolerant and scalable key management for smart grid,
IEEE Transactions on Smart Grid 2 (2) (2011) 374–381.
[306] J. Xia, Y. Wang, Secure key distribution for the smart grid, IEEE Transactions
on Smart Grid 3 (3) (2012) 1437–1443.
[307] S. Das, Y. Ohba, M. Kanda, D. Famolari, S. Das, A key management framework
for AMI networks in smart grid, Communications Magazine 50 (8) (2012) 30–
37.
[308] So, H.K.-H. and Kwok, S.H.M. and Lam, E.Y. and Lui, K.-S., Zero-configuration
identity-based signcryption scheme for smart grid, in: Proc. of IEEE
SmartGridComm’10, 2010, pp. 321–326.
[309] X. He, M.-O. Pun, C.-C. Kuo, Secure and efficient cryptosystem for smart grid
using homomorphic encryption, in: Proc. of IEEE/PES ISGT’12, 2012, pp. 1–8.
[310] M. Fouda, Z. Fadlullah, N. Kato, R. Lu, X. Shen, A lightweight message
authentication scheme for smart grid communications, IEEE Transactions on
Smart Grid 2 (4) (2011) 675–685.
[311] Q. Li, G. Cao, Multicast authentication in the smart grid with one-time
signature, IEEE Transactions on Smart Grid 2 (4) (2011) 686–696.
[312] K. Rogers, R. Klump, H. Khurana, A. Aquino-Lugo, T. Overbye, An
authenticated control framework for distributed voltage support on the
smart grid, IEEE Transactions on Smart Grid 1 (1) (2010) 40–47.
[313] E. Ayday, S. Rajagopal, Secure, intuitive and low-cost device authentication
for smart grid networks, in: Proc. of IEEE CCNC’11, 2011, pp. 1161–1165.
[314] H. Cheung, A. Hamlyn, T. Mander, C. Yang, R. Cheung, Role-based model
security access control for smart power-grids computer networks, in: Proc. of
IEEE/PES Power and Energy Society General Meeting, 2008, pp. 1–7.
[315] Y. Chengqun, Z. Ronggang, Access control for the smart meters based on ABE,
in: Proc. of IEEE CyberC’11, 2011, pp. 79–82.
[316] J. Naruchitparames, M. Gunes, C. Evrenosoglu, Secure communications in the
smart grid, in: Proc. of IEEE CCNC’11, 2011, pp. 1171–1175.
[317] C. Efthymiou, G. Kalogridis, Smart grid privacy via anonymization of smart
metering data, in: Proc. of IEEE SmartGridComm’10, 2010, pp. 238–243.
[318] J.-M. Bohli, C. Sorge, O. Ugus, A privacy model for smart metering, in: Proc. of
IEEE ICC’10, 2010.
[319] R. Lu, X. Liang, X. Li, X. Lin, X. Shen, EPPA: an efficient and privacy-preserving
aggregation scheme for secure smart grid communications, IEEE Transactions
on Paralled and Distributed Systems 23 (9) (2012) 1621–1631.
[320] S. Wang, L. Cui, J. Que, D.-H. Choi, X. Jiang, S. Cheng, L. Xie, A randomized
response model for privacy preserving smart metering, IEEE Transactions on
Smart Grid 3 (3) (2012) 1317–1324.
[321] G. Kalogridis, C. Efthymiou, S. Denic, T. Lewis, R. Cepeda, Privacy for smart
meters: towards undetectable appliance load signatures, in: Proc. of IEEE
SmartGridComm’10, 2010, pp. 232–237.
[322] S. McLaughlin, P. McDaniel, W. Aiello, Protecting consumer privacy from
electric load monitoring, in: Proc. of ACM CCS’11, 2011, pp. 87–98.
[323] S. Axelsson, Intrusion detection systems: a survey and taxonomy, Tech. rep.,
Chalmers University of Technology, 2000.
[324] P. Raj Kumar, S. Selvakumar, Distributed denial of service attack detection
using an ensemble of neural classifier, Computer Communications 34 (11)
(2011) 1328–1341.
[325] Y. Zhang, L. Wang, S. Sun, R. Green, M. Alam, Distributed intrusion detection
system in a multi-layer network architecture of smart grids, IEEE
Transactions on Smart Grid 2 (4) (2011) 796–808.
[326] P. Sangkatsanee, N. Wattanapongsakorn, C. Charnsripinyo, Practical real-time
intrusion detection using machine learning approaches, Computer
Communications 34 (18) (2011) 2227–2235.
[327] P. Casas, J. Mazel, P. Owezarski, Unsupervised network intrusion detection
systems: detecting the unknown without knowledge, Computer
Communications 35 (7) (2012) 772–783.
 E. Ancillotti et al. / Computer Communications 36 (2013) 1665–1697
[328] A. Ahmed, A. Jantan, T.-C. Wan, SLA-based complementary approach for
network intrusion detection, Computer Communications 34 (14) (2011)
1738–1749.
[329] Z. Fadlullah, M. Fouda, N. Kato, X. Shen, Y. Nozaki, An early warning system
against malicious activities for smart grid communications, IEEE Network 25
(5) (2011) 50–55.
[330] Z. Baig, On the use of pattern matching for rapid anomaly detection in smart
grid infrastructures, in: Proc. of IEEE SmartGridComm’11, 2011, pp. 214–
219.
[331] C.-W. Ten, J. Hong, C.-C. Liu, Anomaly detection for cybersecurity of the
substations, IEEE Transactions on Smart Grid 2 (4) (2011) 865–873.
[332] O. Linda, M. Manic, T. Vollmer, Improving cyber-security of smart grid
systems via anomaly detection and linguistic domain knowledge, in: Proc. of
IEEE ISRCS’12, 2012, pp. 48–54.
[333] R. Berthier, W. Sanders, Specification-based intrusion detection for
advanced metering infrastructures, in: Proc. of IEEE PRDC’11, 2011, pp.
184–193.
[334] P. Jokar, H. Nicanfar, V. Leung, Specification-based intrusion detection for
home area networks in smart grids, in: Proc. of IEEE SmartGridComm’11,
2011, pp. 208–213.
[335] X. Ou, S. Govindavajhala, A. Appel, MulVAL: a logic-based network security
analyzer, in: Proc. of ACM USENIX Security Symposium’05, vol. 14, 2005, pp.
8–23.
[336] Q. Liu, Y. Zhang, VRSS: a new system for rating and scoring vulnerabilities,
Computer Communications 34 (3) (2011) 264–273.
[337] M. Rahman, P. Bera, E. Al-Shaer, SmartAnalyzer: A noninvasive security threat
analyzer for AMI smart grid, in: Proc. of IEEE INFOCOM’12, 2012, pp. 2255–
2263.
1697
[338] D. Kundur, X. Feng, S. Liu, T. Zourntos, K. Butler-Purry, Towards a framework
for cyber attack impact analysis of the electric smart grid, in: Proc. of IEEE
SmartGridComm’10, 2010, pp. 244–249.
[339] Y. Wang, D. Ruan, J. Xu, M. Wen, L. Deng, Computational intelligence algorithms
analysis for smart grid cyber security, in: Lecture Notes in Computer Science,
Advances in Swarm Intelligence, vol. 6146, 2010, pp. 77–84.
[340] Z. Anwar, R. Shankesi, R. Campbell, Automatic security assessment of critical
cyber-infrastructures, in: Proc. of IEEE/IFIP DSN’08, 2008.
[341] A. Josang, Trust and reputation systems, in: A. Aldini, R. Gorrieri (Eds.),
Foundations of Security Analysis and Design IV, Lecture Notes In Computer
Science, vol. 4677, 2007, pp. 209–245.
[342] O. Pradhan, M. Awan, K. Newman, F. Barnes, Trust and reputation approach to
smart grid security, in: Proc. of IEEE ISRCS’11, 2011, pp. 101–104.
[343] J. Fadul, K. Hopkinson, C. Sheffield, J. Moore, T. Andel, Trust management and
security in the future communication-based ‘‘smart’’ electric power grid, in:
Proc. of IEEE HICSS’11, 2011, pp. 1–10.
[344] J. Borowski, K. Hopkinson, J. Humphries, B. Borghetti, Reputation-based trust
for a cooperative agent-based backup protection scheme, IEEE Transactions
on Smart Grid 2 (2) (2011) 287–301.
[345] C. Alcaraz, C. Fernandez-Gago, J. Lopez, An early warning system based on
reputation for energy control systems, IEEE Transactions on Smart Grid 2 (4)
(2011) 827–834.
[346] S. Misra, A. Vaish, Reputation-based role assignment for role-based access
control in wireless sensor networks, Computer Communications 34 (3)
(2011) 281–294.
[347] J. Bughin, M. Chiu, J. Manyika, Clouds, big data, and smart assets: ten tech-
enabled business trends to watch, McKinsey Quarterly.