Electric Power Systems Research 120 (2015) 136149

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

The Italian smart grid pilot projects: Selection and assessment of the
test beds for the regulation of smart electricity distribution
M. Coppo c , P. Pelacchi b , F. Pilo a, , G. Pisano a , G.G. Soma a , R. Turri c
a
University of Cagliari, Italy
b
University of Pisa, Italy
c
University of Padova, Italy

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

Article history: Nowadays grid connected renewable energy sources have reached a level that was not even conceiv-
Received 11 February 2014 able just few years ago with serious concerns on power systems stability and security. The Italian TSO
Received in revised form 6 May 2014 and the Italian regulatory authority for electricity, gas and water system (AEEGSI) pioneered this eld,
Accepted 24 June 2014
and imposed the participation of distributed generation (DG) into voltage and frequency regulation for
Available online 11 July 2014
reducing the risk of loosing signicant power generation during frequency transients. Smart grids are
the next step towards DG integration. In 2010, the AEEGSI demonstration phase on smart grids was
Keywords:
designed according to an input based incentive scheme and a competitive selection process in order to
Smart grid
Demonstration projects
award a limited number of projects. According to AEEGSI resolution ARG/elt 39/10, the selected smart
Smart grid regulatory framework grid demonstration projects did benet from an extra remuneration of capital cost (a 2% extra WACC in
Distribution planning addition to the ordinary return) for a period of 12 years. All projects are now in the rollout phase and
Performance assessment of smart grid the rst data are going to be collected from the eld in a real world environment. This paper gives a
Use cases for smart grid brief description of the projects and adds information about the selection criteria that allowed identify-
ing what are the benets that the system stakeholders should expect by the innovation of distribution
operation and planning. Furthermore, the procedure used to dene a regulatory environment suited for
smart grids is described. The main idea is to dene the fundamentals of a fair and transparent regulation
mechanism based on different levels of smartness and on the identication of some suitable indicators
that allow the AEEGSI to establish the expected performances of novel smart grids and, consequently, to
dene penalties and/or rewards for DSOs. With the aid of some examples the role of indicators is showed.
Finally, it must be remarked again that the SG projects are in progress and eld data are not yet
disposable. This is the reason why the numerical simulations reported in this paper have been developed
on a reference network that can represent the behaviour of some of the real SG networks relevant to
the projects; such simulations have been performed in order to better quantify the benets that the
stakeholders of the system should expect. When the projects will enter the operation phase the eld data
collected after the new operational approach will be collected and can be used to verify the simulations
hypothesis assumed.
2014 Elsevier B.V. All rights reserved.

1. Introduction energy storages in order to balance the non programmable renew-

During the last decade the development of DG (distributed gen-
eration), particularly due to the energy efciency improvement of
fossil fuel generation and to the deployment of new renewable
sources, has increasingly stressed the operation of distribution net-
works, originally conceived to be operated as passive systems for
delivering energy to loads. In the next future the need to install
Corresponding author. Tel.: +39 0706755883; fax: +39 0706755900.
E-mail address: pilo@diee.unica.it (F. Pilo).
able generation, the growing up of electrical vehicle recharging
infrastructures for electromobility, and the increasing of demand
response/awareness of customers will create new challenges for
the technological innovation of such networks requiring them to
become smarter by developing innovative operating procedures
in this new framework under increasing security and efciency
constraints [1].
This innovation path can be undertaken by developing pilot
(or demonstration) projects that apply the knowledge and results
obtained by research studies carried out in the last years to
some real distribution electricity networks. Recently, following the

http://dx.doi.org/10.1016/j.epsr.2014.06.018
0378-7796/ 2014 Elsevier B.V. All rights reserved.
 M. Coppo et al. / Electric Power Systems Research 120 (2015) 136149 137

experience in output based incentive regulation applied to quality

Nomenclature of service [5,6], but the development of an output based regulation
applied to smart grids has proven rather challenging [7].
AEEGSI Italian regulatory authority for electricity, gas and In this paper the demonstration pilot projects promoted by
water system the resolution ARG/elt 39/10 are described with emphasis on the
AMI advanced metering infrastructures adopted input based selection mechanism. Lessons learnt useful
ASSD: automatic sectionalizing switching device for the development of new possible output based co-founding cri-
AVDG: average cost for connecting DER teria are discussed with the aid of some quantitative examples to
AVR: automatic voltage regulator address market, policy, and regulation issues related to the practical
BAU: business as usual implementation of smart grid (SG).
CAIDI: customer average interruption duration index
CAIFI: customer average interruption frequency index
CAIFIDG: DG customer average interruption frequency 2. The Italian distribution context
CAPEX capital expenditure
CHP combined heat and power In 2010, the electricity demand in Italy was above 300 TWh,
CEMIn : customers experiencing multiple interruptions 86% was covered by national production, while imports accounted
DER distribute energy resources for the rest. Out of these 300 TWh, 75 TWh were produced by
DG distributed generation renewable energy sources (RES), which recorded a 9% increase with
DMS: distribution management system respect to 2009. A remarkable growth was registered in wind pro-
DSO distribution system operator duction (+29.1%), biomass/waste (+21.6%) and photovoltaic (almost
EMS: energy management system 1600 GWh, compared to 677 GWh in 2009). About 5 million house-
EnTSO: European network of transmission system opera- holds and 2.7 million small businesses have chosen their supplier in
tors the liberalized market. As for the electricity distribution sector, Enel
GPRS general packet radio service Distribuzione is the largest operator, with 86% of the total volume,
FLISR: fault location isolation and service restoration followed by A2A Reti Elettriche (4%), Acea Distribuzione (3.4%) and
HSDPA high speed downlink packet access Aem Torino Distribuzione (1.3%). The other operators have marginal
HV: high voltage quotas only.
ICT information & communication technology AEEGSI has introduced a series of input based incentives, in the
LV: low voltage form of an increase in the WACC, aimed at promoting strategic
MV: medium voltage investments, in electricity transmission rst and in electricity dis-
NTO: network topology optimization tribution later. As of today, regulatory policies need to focus on new
OLTC: on load tap changer problems, stemming from the need to meet long-term, European-
PI: performance indicator wide objectives. In order of urgency:
PS primary substation
PSB RPC: rated power capacity of the primary substation integration of RES, especially as distributed generation (DG), in
PSMC: pseudo sequential Monte-Carlo electrical networks: these intermittent production units are also
PV: photo voltaic expected to contribute not only to the European targets for RES
QoS quality of service penetration but also to the security of the overall electrical sys-
RES renewable energy source tem;
SAIDI: system average interruption duration index introduction of techniques for load control, thanks to intelligent
SAIDIDG: system average duration of DG interruptions systems and meters located at the point of connection with the
SAIFI: system average interruption frequency index customer and its end-users;
SAIFIDG: system average frequency of DG interruptions introduction of opportunities for all customers to become active
SG smart grid participants in the electricity market, using electronically con-
SGO smart grid operator veyed information related to electricity prices and adopting new
TETRA terrestrial trunked radio technologies (e.g. electro-mobility).
TSO transmission system operator
WACC weighted average capital cost It should be reminded that in the above-described processes ICT
3G 3rd generation technology standard systems play a crucial role.

3. Selection criteria for the demonstration pilot projects

experience gained during the development of smart metering
thanks to ENEL demonstration project [2,3], a number of smart
grids pilot projects have been proposed by some Italian distribution
system operators (DSOs) within the framework of the resolution
ARG/elt 39/10 of the Italian authority for electricity, gas and water
system (AEEGSI) [4]. This resolution denes which characteristics a
distribution grid must have in order to obtain incentives, and pro-
vides the rules for preparing and selecting demonstration projects.
It is worth noting that the resolution provides mechanisms for the
co-founding process too.
The assessment and progress monitoring activity of the demon-
stration projects which is being carried out by the AEEGSI is also a
chance for trying to implement an output based regulation mecha-
nism, i.e. based on the effects of a given activity or service, for future
incentivizing actions. In recent years Italy has gained a signicant
In general, demonstration (pilot) projects represent a crucial and
necessary step between laboratory tests and full industrial deploy-
ment; the experience gained in developing pilot projects can be
extremely valuable for the subsequent application and integration
in real systems, estimate of realization and management costs in
the industrial phase, denition of parameters and factors to be con-
sidered for economic compensation policy, taking also into account
input/output based regulation philosophy.
The uncertainties regarding costs, efciency and efcacy for car-
rying out these pilot projects require a careful selection procedure
for incentivising only those expected to provide quantiable ben-
ets. According to the resolution ARG/elt 39/10, the selected smart
grid demonstration projects are granted an extra remuneration
of capital cost (2% yearly extra WACC in addition to the ordinary
138 M. Coppo et al. / Electric Power Systems Research 120 (2015) 136149

return) for a period of 12 years [4]. The incentive is funded with the Table 1
Benet score (Aj ).
network tariff. The selection of these projects was carried out, on
behalf of AEEGSI, by a committee of experts1 who worked according A1 Size N. generation plants/storage 6
to the input based regulation philosophy provided by the resolution Increase of electricity production injected into 12
ARG/elt 39/10. the grid
Increase of ratio electricity 8
In order to be admitted to the selection process, the project production/electricity consumption
proposals had to satisfy the following minimum requirements: N. primary substations involved in the project 4
MAX A1 30
the demonstration project has to be implemented and tested on A2 Innovation Participation of disperse generation to voltage 6
a real existing MV network connecting both passive users (end regulation
customers) and active users (dispersed generators); Presence of control system (SCADA) 6
in the electricity distribution area covered by the demonstration Bidirectional communication and demand 6
response
project, a reverse active power ow from distribution network to Presence of storage systems and active power 12
transmission grid must occur for at least 1% of the time in a year; modulation
the selected MV network has to be equipped with real-time mon- Participation of DSO to ancillary service market 10
MAX A2 40
itoring systems able to record all data needed for the correct
operation of the network; A3 Feasibility Project schedule 4
only open and non-proprietary communication protocols are to Quality improvements 6
MAX A3 10
be used for any communication applications involving network
users in order to minimize customer costs at the network inter- A4 Replicability % of costs on not regulated subjects (DG and 2
face. storage)
Standard protocols 8
Consistency between investment costs and 10
The rst requirement was introduced to focus on innovation expected benets of the project
opportunities for real networks; the MV level was intentionally MAX A4 20
Max Project 100
chosen because 75% of installed RES power is currently connected
to MV networks. The second requirement was aimed at giving pri-
ority to the projects that addressed the problem of inverse power As regards the technical score, it is given by four components,
ows which, in the context of the existing MV networks, can gener- namely:
ate problems to the protection system and to the voltage regulation
scheme currently adopted [8], being also an index of high penetra- A1. Size considers the number of active users involved, the size
tion of dispersed generation in absence of control system at the of the area involved in the pilot project and the attitude of the
interface between TSO and DSO. The third requirement ensured project on increasing production from DG;
the availability of data relevant to the operation of the system in A2. Innovation considers the degree of innovation that the pilot
real time, which can be suitably used and elaborated by a distribu- project will introduce in the distribution system, with reference
tion management system (DMS). The fourth requirement focuses to the ability of aggregating DG, regulating the voltage and man-
on a key issue for smart grids: only standard protocols, of an open aging the production diagram, by making use of communication
and non-proprietary type, allow implementing a free market and systems, for a better control and management of distribution
minimizing the costs as well as the technological complexities that networks;
users of the intelligent network will have to face in exploiting the A3. Feasibility considers the timing of the project and the impact
benets of smart grids. on quality of supply. A project that could lead to a depletion of the
In addition to minimum requirements, other functionalities levels of continuity is considered not feasible or poorly valuable;
have been positively considered, such as provision of bidirec- A4. Replicability on a large scale considers the requirement of
tional communication with nal customers for demand response reproducibility on a large scale of the technical solutions pro-
strategies, including recharging infrastructures for electric vehicles, posed in the pilot project, looking at economies of scale and
and/or storage systems. peculiarities of solutions.
By referring to the input based regulation philosophy mentioned
above, the selection of the projects was based on a comparative The benets considered and the assigned maximum scores are
assessment according to a key performance indicator (PI) based on a listed in Table 1, whereas the rank-ordered projects, selected on
combination of quantitative and qualitative indicators and dened the basis of the relevant PI, are reported in Table 2.
as The eight selected projects have common characteristics (e.g.,
innovative fault location and service restoration, voltage control
Psmart 
4
PI = Aj (1) in presence of reverse ow, standard communication protocols,
C
j=1
Table 2
where C is the total cost of the project. The quantitative indicator Selected smart grid projects.
(Psmart ) is dened as the increase of DG power that can be connected Position rank Primary Substation (PS) (Aj ) C [kD] PI
to the grid with few or none network reinforcement without the involved in the pilot project
violations of voltage, current and frequency constraints thanks to
1 A2A PS Lambrate 65 733 4715
the smart investments in the network. The qualitative indicator 2 ASM Terni 68 800 1375
(Aj ) is a technical score attributed by a panel of independent eval- 3 A2A PS Gavardo 65 755 663
uators, coordinated by the committee of experts, on the basis of 4 ACEA Distribuzione 73 4970 660
their engineering judgement. 5 ASSM Tolentino 66 689 595
6 ENEL Distribuzione PS 96 6242 569
Carpinone
7 Deval PS Villeneuve 68 545
1 8 A.S.SE.M. San Severino Marche 64 642 365
The experts are P. Pelacchi, F. Pilo and R. Turri.
 M. Coppo et al. / Electric Power Systems Research 120 (2015) 136149
etc.), and specic features with respect to the type of environment
(i.e., urban or rural), the project size, and the technological solu-
tions developed. With reference to the project size, some projects
include few MV feeders supplied by a single HV/MV transformer,
other projects consider all MV feeders supplied by a PS; one project
considers two PS with the possibility of network reconguration.
About the technologies adopted, all projects make use of a bidi-
rectional communication system; almost every project considers
the future participation of the DSO to the system services market;
few projects use a storage system and a recharging infrastructure
for electric vehicles; only one project includes demand response
strategies.
An important condition for the selection of demonstration
projects was the publicity of the results; for this reason all results
will be public through the website of the AEEGSI, in order to allow
the dissemination of experiences and a more realistic evaluation
of the real outcomes of the projects in the perspective of full scale
smart grid deployment.
4. Current development of the demonstration pilot projects
Seven of the eight selected pilot projects have been developed.
In the following a concise description of their main features is pre-
sented, with reference to Table 3 for a qualitative comparison of
the main functions considered in each project.
Enel Distribuzione carries out the most comprehensive project,
whose aim is to combine the generation from distributed energy
resources with a reliable and safe management of the system under
real operating conditions. The project offers a new approach to
the distributed generation management, which monitors the active
involvement of both distributors and customers, recognized as
prosumers of energy. Monitoring occurs through a broadband con-
nection, based on a Wi-Max communication protocol as well as
a bre optics communication infrastructure. The project includes
the installation of nearly 8000 Smart Info devices for customers
connected to the low voltage grid. These devices will supply infor-
mation regarding changes in the price of energy based on time
slots, promoting efcient use and increasing active customer par-
ticipation in the management of the system. The project also
includes the installation of a charging station to power a eet of ve
electric vehicles, integrated with a photovoltaic plant and a multi-
functional storage system. Finally, the project will also represent a
real eld test of an innovative model for the protection, automation
and management of power generation in the distribution network
according to the principles of smart grids.
The project carried out by ACEA, a medium/large size utility
located in Central Italy, has the ambition of realizing a real smart
grid prototype, which could be replicable on the entire network of
the utility. For that reason, the network involved in the pilot project
is a signicant part of the real MV network, which includes 2 pri-
mary sub-stations, 68 secondary cabins, 70 km overall length of MV
lines, 28 MVA of distributed generation and more than 1000 LV end
users. The project focusses on advanced automation of MV network
(i.e. advanced fault line automatic localization, management of dis-
tributed generation plants, a novel monitoring system of transient
faults for localizing the trigger point), acquisition of electrical and
environmental variables of the MV and LV grids using TETRA/GPRS
network enabling the new management techniques to be applied
such as optimization of feeder voltage proles, power ow man-
agement, minimization of energy losses and diagnosis of primary
substation components (power transformers and circuit breakers).
The project proposed by DEVAL, a small utility located in the
Alpine region, is aimed at restructuring an existing MV network
by employing innovative technologies enabling the active net-
work operation with particular attention to standardization and
139
unication requirements as well as costs minimization. The main
features are advanced network automation, Volt/var control under
normal operating conditions, active power limitation/regulation
under emergency conditions, real-time monitoring of both load and
distributed generation, and real-time control/dispatching of local
resources. The proposed architecture is a sort of extended sub-
station organized into three functional levels (primary substation
level, secondary distribution and user substation level, active user
level). The communication system respects the centralized hierar-
chical control of the extended substation concept and employs
a heterogeneous infrastructure (wireless, HSDPA, 3G, bre optics,
communicating each other through the IEC 61850 protocol) in
order to better cope with the specic territorial constrains.
A2A, a large size utility of Northern Italy, is carrying out two
companion projects, trying to apply the same concepts to different
network contexts. The main features of the projects are: advanced
network automation, tripping of DGs remote controlled by the
DSO in case of network faults, logical selectivity between DSO and
end users (active and passive) protection systems, Volt/var control
of DGs under normal operating conditions, active power limita-
tion/regulation under emergency conditions, DG monitoring and
real time data provision to TSO, and optimal management of DGs
based on production forecast and real-time control. These projects
aim at fully exploiting the regulation potentialities of DG through a
local dispatch, suitably coordinated with the TSO, which may also
enhance the network hosting capacity.
The last projects are led by two small utilities located in Central
Italy. The ASSEM San Severino Marche project represents a further
application of the extended substation concept, i.e. electric sys-
tem organized into logical levels, centralized hierarchical real-time
control of remote distributed generation and use of different trans-
mission vectors, whereas the ASM Terni smart grid project is more
focused on services and the participation of active users to network
management; more specically the realization of the fundamental
infrastructure to implement proper services for increasing energy
efciency, use of renewables and e-mobility is the most important
aspect of the project.
All projects reached an advanced level of development, having
installed and tested in the eld most of the electric and ICT equip-
ment, after completing the necessary administrative procedures
for having permission to build the new installations on the public
network, selecting and acquiring the components on the market.
The nal stage of the projects will be devoted to the real-life
experimentation. The assessment of the results will be performed
with metrics capable of quantifying the level of smartness reached
and the overall achieved performances (in terms of quality of ser-
vice, hosting capacity, loss minimization, etc.).
Although many of the functions are common to all projects, their
design and realization are done independently on network portions
of different size and characteristics. A comparison of the perfor-
mances would thus provide valuable information on the level of
applicability and standardization of the various functions tested.
5. Input/output based regulation for smart grids
The described pilot projects will enter in their operation phase
during 2014. The nal tests will be performed by the end of the
same year. After the test phase the projects will be evaluated in
order to verify the validity of the technical choices set up and the
replicability of the design solutions.
The ranking of the pilot projects was carried out according to
an input based regulation as explained above. The on-going assess-
ment activity of the demonstration projects will represent a good
opportunity to study and tune an output based regulation mech-
anism. In fact a good regulation should be based on outputs, i.e.
140 M. Coppo et al. / Electric Power Systems Research 120 (2015) 136149
Table 3
Main functions developed in the smart grid pilot projects.
Function ENEL ACEA
Bi-directional communication X X
DG monitoring and real time data provision to TSO X X
Participation of active users to network regulation X X
Advanced automation and logical selectivity X X
Storage X X
EV re-charging infrastructure X X
Demand response X
on the effects of a given activity or service, rather than on inputs,
which means imposing some predened choices to the regulated
company. Because of the clear difculties, the existing regulatory
frameworks are generally input based in the sense that the incen-
tive is applied to incurred project costs (input) rather than output
based, that is by providing penalties and rewards related to certain
criteria and performance targets achieved. Indeed, performance
measurement requires clear and fair indicators strictly related to
the pursued objectives and cleansed of external effects outside the
control of network operators. For this reason, in order to rational-
ize the selection process, it has been assumed that an eligible smart
grid demonstration project should include, from a technical point
of view, functions of voltage regulation and DG reactive power dis-
patching, load management and DG active power dispatching and
on-line reconguration with dynamic protection settings. All these
features require real time high performance communication sys-
tems; a distribution grid having such features would also enable
the nal customers to take part to the electricity markets and the
TSO to exchange data with the DSO.
The following considerations come from the hypothesis that the
networks eligible for smart grid reward have to be active and the
generation power installed is relevant with respect to the load.
This means, for example, that the generation power installed is
at least higher than the minimum annual level of the load. This
assumption is relevant to the consideration that economic incen-
tives should be primarily given to projects trying to solve critical
network conditions.
In this way and under the assumptions stated above it is possible
to dene levels of smart grid according to the increasing smart-
ness of their characteristics:
- smart grid basic level: the smart grid operator (SGO) can use the
communication system for the ultra rapid DG disconnection dur-
ing line disturbances. The SGO exchanges data with TSO for power
systems control and security;
- smart grid advanced level: in addition to the features of the pre-
ceding level, SGO adopts Volt/var regulation systems;
- smart grid full operation level: in addition to the features of the
preceding levels, SGO adopts Volt/var centralized/decentralized
regulation systems that enable all nal customers to access the
system services market, possibly by exploiting also storage facil-
ities.
It is worth noting that at any of the above smart grid lev-
els, the presence of a reliable ICT infrastructure enables also the
implementation of advanced grid automation systems for fault
location isolation and service restoration (FLISR), which signif-
icantly improve the continuity of supply of active distribution
networks. Advanced FLISR systems positevely impact the quality
of service (QoS) for all customers and, since Italy has an advanced
QoS regulation, the smart grid effect on QoS will be part of that
specic regulation and will not considered in this paper.
For an output-based regulation, the main variables involved
should be identied. In order to simplify the problem the
set of such variables should be limited in number and easily
A2A (1) A2A (2) ASM Terni A.S.SE.M. Deval
X X X X X
X X X X X
X X X X X
X X X X X
X X
measurable. Moreover, variables should synthesize as better as pos-
sible the system behaviour and highlight the increase of benets
for nal customers. In this work, the set of variables identied as
the most representative of the smartness level of a distribution
system is [912]:
- the capacity to connect new DER with a small CAPEX rate,
- the QoS for producers and consumers,
- the DER energy production,
- the voltage prole, and
- the integration with TSO control systems.
5.1. DER connection costs
One of the reasons because in recent years the stakeholders
of the distribution system moved towards SG is the fact that SG
technologies allow the integration of novel DER with less capital
expenditures if compared to the traditional approach applied in
distribution business. This is often referred to as the increasing of
hosting capacity achieved with, for instance, voltage regulation and
the implementation of local markets for ancillary services. Even
though the increase of hosting capacity is a clear follow up of SG
implementation, it should be recognized that its usage for output
based regulation in neither easy nor straightforward. Indeed, distri-
bution systems, particularly in Italy, have recently been drastically
revamped to allow the interconnection of RES. New underground
cables and overhead lines were built and some primary substations
were refurbished with new transformers. As a consequence, there
are portions of the network that can now easily accept more RES and
DER with really few investments; however, these networks cannot
be considered SG in the sense used in this work. Indeed, if the rate
of investments was used as one of the leading indicators for the
SG output based regulation, there could be a clear risk to reward
as SG a network previously revamped according to the traditional
t&forget policy simply because it can accept more generation
than before the upgrade. In order to avoid (or limit) such risk, the
amount of CAPEX in the MV/LV distribution network during the
prexed time horizon of regulation framework (e.g., three years)
has to be taken into account.
A possible way to simply assess the impact of SG on the reduc-
tion of the connection costs is given by Eq. (2). ADVG [D/MW] is the
average cost for connecting DER in areas similar to the one where
the examined primary substation is located. In (2), the CAPEX in the
regulation period (current CAPEX) is increased by a term that takes
into account how much the previous expenditures (past CAPEX)
were related to DER connections. The ratio between the installed
DER rated power capacity (DER RPC) and the rated power capacity
of the PS (PSB RPC) is used to weight the past CAPEX.
DER RPCPast
CAPEXPast + CAPEXCurrent
PSB RPCPast
AVDG = (AVDG) (2)
DER RPCCurrent
DER RPCCurrent > 0
Eq. (2) allows calculating the SG impact in the reduction of
CAPEX caused by DER connections. If (AVDG) is greater than zero,
 M. Coppo et al. / Electric Power Systems Research 120 (2015) 136149
the performance of the SG is positive, and this can be used to reward
a premium. The formula is simple and it is based on parameters
that can be easily gathered from distribution companies databases.
AVDG and (AVDG) should be compared with national bench-
mark data applicable to areas similar for network typology and
customers concentration. The performance indicator (AVDG) is
expected to be small or negligible for the SG basic level, since a
SG can increase the Hosting Capacity only if some regulation capa-
bility is obtained with network active management. Anyway, this
index can be associated to any level of smartness and will be more
effective only in those cases with the highest capability of control.
In order to assess the sensitivity of the index, an example of
application is provided in Appendix I, on the basis of the expected
evolution of costs in an area covered by a primary substation that
can be considered representative of distribution systems, in a rural
ambit (the distribution network used for the test is one of the
representative networks proposed in the ATLANTIDE Project [13];
the evolution of load demand as well as the expected growth of
photovoltaic (PV) generation has been derived by the ATLANTIDE
project reports, summarized in Appendix I. The scope of the follow-
ing example is to highlight the sensitivity of the proposed metric
and to prove its applicability to assess the performances of Smart
Grid operation. Since the example refers to one network only,
the AVDG used for the application of (2) is the ratio between the
DER capacity and the network capital expenditures calculated with
reference to the observation timeframe prior to the current one.
Table 4 shows the results with reference to the capital expendi-
tures (all costs are annualized to 2010 for comparisons) supposing
that all operational issues, namely voltage regulation, were solved
with a traditional passive distribution system capable to cope with
all worst-case events (i.e., t&forget). It clearly emerges that in
20102014 (particularly in 20102012), when in Italy the feed-in
tariffs for PV were particularly generous, the level of investments
would have been enough to face also the expected growth of PV
generation in the examined area without any investment. It clearly
appears that by using the reduction (AVDG) without any cor-
rection, the DSO of the examined area would expose a positive
indicator but it is also evident that the network would have had
no smart installations at all. Conversely, with the addition of the
suitable correction term proposed in (2) to track previous network
investments, the (AVDG) is negative, which indicates more cor-
rectly that no improvement related to SG can be identied and then
rewarded.
Table 5 gives the same information of Table 6 by assuming that
starting from 2014 a smart grid is in place. It is straightforward to
observe that the level of network expenditures for the connection
of novel DG is extremely low and that the indicator expressed by
(2) provides a quantitative identication of the smart grid invest-
ments by properly weighting the impact of the previous network
investments.
5.2. QoS for producers and consumers
The performance of increasingly complex power distribution
systems is often described by reliability indices for the overall sys-
tems as well as performance indices or other metrics for particular
technologies and resources [14]. IEEE Standard 13662012 denes
several standard indices for sustained interruptions (i.e., outages). A
sustained interruption is basically any interruption of power longer
than a predetermined period of time that generally requires human
intervention to restore power. The time period used by utilities for
reporting such interruptions ranges from 1 to 5 min in different
countries or regulatory jurisdictions. The reported indices usually
exclude major storms that damage extensive parts of the power
delivery system. The indexes that are more useful for smart grid
regulation are briey recalled.
141
- SAIFI system average interruption frequency index
This index is computed by summing the number of customers
interrupted for each event and dividing (normalizing) by the total
number of customers.
total number of customer interruptions
SAIFI =
total number of customers served
- SAIDI system average interruption duration index
This index is computed by summing the restoration time for
each event times the number of customers interrupted for each
event and dividing (normalizing) by the total number of customers.

customer interruptions durations
SAIDI =
total number of customers served
- CAIDI customer average interruption duration index
This index is computed by rst computing SAIDI and SAIFI and
dividing as shown.

customer interruptions durations SAIDI
CAIDI = =
total number of customers interruptions SAIFI
- CAIFI customer average interruption frequency index
This is the average frequency of interruption for those customers
who actually experienced an interruption. The customer is counted
but once regardless of the number of times interrupted. This index
is computed like one would be computing SAIFI but dividing by
the total number of customers who have experienced a sustained
interruption. According to the standard, each individual customer
should be counted only once regardless of how many times inter-
rupted during the reporting period.
total number of customer interruptions
CAIFI =
total number of customers interrupted
CEMIn customers experiencing multiple interruptions
This index is designed to track the number, n, of sustained inter-
ruptions to a specic customer. Averages tend to obscure particular
trouble spots that need attention. This is an example of an index that
is becoming more important for reliability-centred maintenance
(RCM) analysis.
The SG can positively impact all these indices depending on the
smartness level introduced. With a basic SG, that has in place a
communication system for remote disconnection of generators, the
number of unnecessary DER disconnections decreases. Unneces-
sary disconnections are those caused by the stringent thresholds
of the loss of mains relay or due to impossibility to identify those
generators that can stay connected during network faults until the
faulted area is cleared. In Italy, the loss of mains relay thresholds
were recently relaxed for the integration with TSO control and secu-
rity policies and similar specications are going to be implemented
with the European Grid Code, recently approved by EnTSO, also in
Europe [15]. Nevertheless, the absence of a direct DER control still
causes some unnecessary DER and RES disconnections that can gen-
erate requests for economic compensation and do not comply with
the idea of a low carbon environment. The interruptions suffered
by DER for network problems are seldom considered in reliability
calculations, and the improvement achievable with SG with dif-
ferent level of complexity is often jeopardized by indexes that do
not explicitly consider DER. Indeed, the availability of communica-
tion signals with small latency (100200 ms) improves reliability
142 M. Coppo et al. / Electric Power Systems Research 120 (2015) 136149

Table 4
Capital expenditures in the Italian rural representative network with the expansion plan guided by the t and forget paradigm. It is evident the paradox caused by the simple
reduction of capital expenditures. Indeed, the network is not smart but the connection cost per MW installed is reduced thanks to previous investments. The performance
indicator expressed by (2) allows identifying such situation.

Regulation timeframe 20102014 20142018 20182022

Primary substation 25 MVA 25 MVA 25 MVA

DER RPC 17.74 MW 13.50 MW 5.56 MW
CAPEX 929.43 kD 178.12 kD 22.80 kD
DER RPC/CAPEX 52.40 kD/MW 13.20 kD/MW 4.10 kD/MW
(AVDG) without the correction from Eq. (2) 39.20 kD/MW 9.10 kD/MW
(AVDG) with the correction from Eq. (2) 9.65 kD/MW 8.20 kD/MW

Table 5
Capital expenditures in the Italian rural representative network with the expansion plan guided by the t and rely upon paradigm (e.g., smart grid). The performance indicator
in (2) is capable to discriminate between the role of past investments and the impact of smart grid.

Regulation timeframe 20102014 20142018 20182022

Primary substation 25 MVA 25 MVA 25 MVA

DER RPC 17.74 MW 13.50 MW 5.56 MW
CAPEX 929.43 kD 0.00 kD 0.00 kD
DER RPC/CAPEX 52.40 kD/MW 0.00 kD/MW 0.00 kD/MW
(AVDG) without the correction from Eq. (2) 52.40 kD/MW 0.00 kD/MW
(AVDG) with the correction from Eq. (2) 3.55 kD/MW 0.00 kD/MW

indexes referred to DER. Particularly, CAIFI and CEMIn can be used
to measure the effect of the SG basic level, whereas all reliability
indexes are positively affected by more complex SG as listed in the
following.
Basic level SG improvement of CAIFI and CEMIn for DER are
expected. Improvements should be compared with the situation
before the SG implementation and referred to homogenous data
for similar areas. If FLISR is implemented, then the novel protec-
tion system allows identifying within 100200 ms a faulted area
and reconguring the network so that only generators and cus-
tomers in the faulted portion are disconnected. A SG with advanced
FLISR can thus improve service continuity to all customers since the
number of customers affected by faults during the fault location
stage is reduced. This results, for instance, in a SAIFI reduction. This
improvement is perceived by DER only if FLISR allows the prompt
disconnection of the generators involved in the fault.
Advanced level SG with this level of implementation, the SG
is supposed to have both advanced FLISR and Volt/VAR in place.
The communication system is used for voltage regulation, namely
achieved with the optimal control of the reactive power produced
by DER (particularly at MV). Since the vast majority of DER is inter-
faced with power electronic converters, the provision/absorption
of reactive power can be easily achieved without the curtailment of
active power. The third SG tier again positively impacts both CAIFI
and CEMIn for DER since there is less or none need to disconnect
generators for sustained overvoltage.
Full operation level SG also the third tier of SG implementation
is supposed to use FLISR and volt/VAR regulation. In this case DER
and aggregated active demand, eventually with distribution stor-
age devices, can participate to a local market for ancillary services
and there are systems in place that allow the optimal exploitation
of all resources. Again, in this case the optimal management of DER
can reduce CAIFI and CEMIn for DER but, since power congestions
can be solved without tripping line overload breakers, a positive
impact has to be expected also for the SAIFI.
In order to demonstrate the validity of the above assertions,
the expected reliability indexes have been calculated for the Ital-
ian rural representative network described in Appendix I. The fault
location and isolation time has been assumed conventionally equal
to 1 h for manually operated networks and to 5 min in remote con-
trolled networks with advanced FLISR; the repair stage is assumed
equal to 5 h.
The results reported in Table 7 show the impact of different SG
levels of implementation on the reliability of the active distribu-
tion network. The calculations have been preformed with a pseudo
sequential Monte-Carlo (PSMC) algorithm that enables calculating
single load/generation point indices and system reliability metrics
[16]. In the simulations the SG reliability has been evaluated by
modelling and simulating the reliability of the power system as
well as the communication and control infrastructure. The com-
munication system simulated in the paper is based on a backbone
with directional (point-to-point) WiMax antennas. WiMax point-
to-point is particularly well suited for rural applications and can
guarantee communication latency as small as 100 ms, which is nec-
essary for FLISR applications. The reliability indexes refer only to
DGs since at present the power quality improvement to all cus-
tomers is not considered a matter of SG regulation. In this case, the
system average frequency of DG interruptions (SAIFIDG), the aver-
age duration of DG interruptions (SAIDIDG) and the DG customer
average interruption frequency index (CAIFIDG) have been exam-
ined. The choice of these indexes, that reect the capability of the
system to accept the power produced by DER, is related to the oper-
ation of distribution systems. Normally, the generators installed in
the network are disconnected when voltage and/or frequency are
out of permissible bands or for line faults. The active management
of the system has the duty to reduce disconnections with a proper

Table 6
Reliability indexes for different SG levels.

SG level SAIDIGD [hours/year] SAIFIGD [occurrences/year] CAIDIGD [hours/occurrences]

Passive mode (without SG) 146.8 30.1 4.87

SG basic level (only FILSR) 142.2 29.7 4.78
SG advanced level (FILSR and Volt/VAR) 129.0 28.1 4.59
SG full operation level (FILSR, Volt/VAR and active power control) 115.9 26.2 4.43
 M. Coppo et al. / Electric Power Systems Research 120 (2015) 136149 143

Table 7
Results for the voltage index evaluations in the four scenarios.

Scenario Year Monitored busbar Max over-voltage [pu] Zone 1 (5%Vn) Zone 2 (10%Vn)

% time VOOB % time VOOB

Base case 2014 N 030 0.088 15.60% 3.153 0.00% 0.000

2015 N 077 0.158 27.81% 17.152 10.22% 10.916
2016 N 077 0.173 28.83% 19.813 12.44% 13.948
2017 N 077 0.175 28.85% 20.063 12.53% 14.215

Smart grid basic 2014 N 030 0.075 9.95% 0.978 0.00% 0.000
2015 N 077 0.138 22.93% 9.678 3.59% 3.859
2016 N 077 0.154 24.37% 11.914 6.40% 6.562
2017 N 077 0.153 24.39% 11.909 5.85% 6.239

Smart grid advanced 2014 N 030 0.053 0.67% 0.010 0.00% 0.000
2015 N 077 0.068 6.10% 0.451 0.00% 0.000
2016 N 077 0.078 6.87% 0.827 0.00% 0.000
2017 N 077 0.077 6.89% 0.825 0.00% 0.000

Full smart grid 2014 N 030 0.050 0.00% 0.000 0.00% 0.000
2015 N 077 0.061 0.03% 0.003 0.00% 0.000
2016 N 077 0.065 0.06% 0.004 0.00% 0.000
2017 N 077 0.053 0.24% 0.002 0.00% 0.000

dispatch and the advanced FLISR can reduce the area of inuence of
faults and avoid the disconnection of those generators not directly
connected to a faulty area. Failures in the communication system
can jeopardize the benets of the active operation and make gen-
erators again prone to disconnections. For this reason the highest
attention has been paid to the reliability of generators, but similar
studies can be carried out, for example, to assess the impact of the
communication system on the modern protection system. In that
case, the reliability indexes have the classical expression referred
to all customers and not to generators only. The reliability with the
current passive management of distribution systems and without
a FLISR system capable to selectively isolate generators leads to a
SAIDIDG equal to 146.8 h and SAIFIDG as high as 30.1 interruptions
per year. In the SG basic level the adoption of an advanced FILSR
system, as reported in Table 6, does reduce the number of interrup-
tions occurred. Such a reduction is obtained with the FLISR system
that enables the disconnection of the DG in the faulty areas only.
Moreover, such FLISR system would also improve the service con-
tinuity of all customers, since the number of customers affected
by faults during the fault location stage is reduced. Table 6 clearly
shows that the impact would be not so relevant if only generators
are taken into account in assessing the advanced FLISR system and
the SG performances, and this fact conrms that FLISR systems are
much more suited to be dealt with the regulation for service quality
rather than for SG.
In the advanced SG level, both advanced FLISR and Volt/VAR reg-
ulation will be in place. The communication system is then used to
enable distributed voltage regulation, i.e. achieved with the opti-
mal control of the reactive power produced by DG units. Volt/VAR
regulation reduces the capital expenditures and improves voltage
prole across the network by limiting the curtailment of active
power generation. Table 6 again shows the improvement for DG
production.
The full operation SG level leads to the best compromise
between network expansion and the exploitation of RES. The dis-
patch of all distributed resources active demand, energy storage
devices and dispatchable generation relieves operation issues,
namely voltage regulation problems, particularly during the hours
with high generation and low demand. The use of the innovative
smart management with active and reactive power control pre-
serves the existing asset and requires capital expenditures mainly
for the communication infrastructure. It must be noticed that, even
if the cost for the communication infrastructure might be high, it
is possible to share the bandwidth capacity with other end users,
which signicantly helps to recover the cost of the installation. In
this SG level, the average number of interruptions suffered by DG
units is as small as 26 per year against 30 per year in the reference
case without SG.
In conclusion, it clearly emerges that all levels of SG can pos-
itively impact reliability indexes for DER as well as for all the
customers. The annual improvement of indexes should thus be con-
sidered to be used with an approach similar to the one presently
employed for the improvement of QoS, with penalties/rewards
related to the level of performances reached.
5.3. RES/DER energy production
RES/DER energy production depends on many different factors
as the availability of the primary source, the reliability/availability
of the power plant as a whole, and, nally, on the QoS of the
system. The QoS affects the capability to generate energy in two
ways. Indeed, there are cases where the generator is connected
to a portion of the system that suffered for a fault. In this par-
ticular situation the producers suffer for poor QoS as well as the
other consumers fed by the same portion of the system. All actions
that improve general QoS are positive for the producers too. Fur-
thermore, generators in network without SG are disconnected also
for faults in the feeder independently from the reciprocal posi-
tion of the fault and the generator itself. Indeed, the DG loss of
mains protection should command the generator disconnection
when perturbations on voltage and frequency are detected in order
to avoid unintentional islanding. For this reason, it clearly emerges
that the level of QoS for a producer is usually worse than the one
perceived by the ordinary consumers, and this fact often causes
to producers signicant economic damages, particularly because
the process of disconnection detection and reconnection can take
signicant time. Indeed, the DG/RES average size normally does
not justify the presence of personnel for operating the plant and
remote control might be not sophisticated enough. For this rea-
sons hours and sometimes days are necessary for DG/DER service
restoration.
The SG can relieve these issues and, again, a measure of the
improvements can be used for the performance-based regulation.
It is clear that a simple measurement of the energy not produced
by DER/RES cannot be used because there is the risk to charge
DSO DER power plant malfunctions or efciency decays. In order
to be more precise, it is necessary to register the disconnections
for each DER/RES caused by the opening of the loss of mains relay
144 M. Coppo et al. / Electric Power Systems Research 120 (2015) 136149
and the time of subsequent reconnection. This could be a measure
of the energy that is not produced due to the protection system
in the examined network, namely the network fed by a primary
substation. By summing the lost of energy production for the gen-
erators in the network the improvement achieved with the SG can
be assessed, but the use of the energy not produced cannot be
used as an SG performance index. Indeed, an index based on the
increasing of energy production will depend on the DG operation
and on the capability to fast recover after the disconnection. The
faster is the reconnection the better will be the SG, but this could
give the DSO a not justied reward if the reduction of restoration
was caused by improvements at the producers side. For these rea-
sons, the use of the number of interruptions per customer/producer
is better suited to SG regulation. The measure of the energy not
produced can be used as a performance indicator with the most
advanced SG implementations, which are based on a full control
of DER and DG that might require the curtailment of active pro-
duction. In this case, the smaller is the energy curtailment the
cheaper is the system operation and, as a consequence, the higher
is the SG performance. The energy not delivered due to SG con-
trol actions is a clear index for assessing the effectiveness and
optimality of the energy management system (EMS) used by the
SG. The role of the pilot projects in this context is fundamental to
understand the amount of power curtailment that is necessary in
distribution networks with high shares of DER. In future regula-
tion, the reduction of energy curtailment will be the simplest to be
applied.
5.4. Voltage prole
Voltage regulation is nowadays a major issue in the day-by-
day operation of distribution systems. RES and DER often cause
overvoltage, particularly in the point of common coupling with
the network. Recent Italian standards and the European Grid Code
require generators have voltage local control regulators but, for
sure, with SG the voltage quality and voltage prole can be fur-
ther improved with the optimal operation of at least DER for
reactive support and, in more advanced applications, with the
inclusion of active demand and distribution storage devices as
well as the control of active power. For voltage regulation pur-
poses, the communication infrastructure does not require very
small latencies since the time granularity for this regulation is
at least 1020 s or greater. The challenge here is represented
by the number of participants, particularly if LV customers and
prosumers will be involved, The communication noise, which
affects the communication among so many participants and the
reliability/availability of the communication system adopted, is
the major concern. The deployment of AMI can make it possi-
ble the acquisition of voltage measurement in the nodes where
generators are connected. The measurements can be grouped
by gravity of overvoltage, if any, and by the relevant dura-
tion.
The DSO involved in the projects have been addressed to char-
acterize their existing passive networks with a vast measurement
campaign in order to dene a starting situation that will be used
to dene a benchmark. Similar measurements will be performed in
the pilot projects networks before and after the implementation of
smart grid. The objective of this campaign will be the denition of a
reasonable starting situation as well as the quantitative denition
of the benets that could be achieved with modern voltage regu-
lation. Again it clearly emerges the importance of pilot projects to
identify the expected benets of SG.
A simple yet signicant system indicator, based on the occur-
rence of out-of-band network voltage conditions and named VOOB ,
is here proposed in expression (3). It is given by the sum of the
highest network bus voltage deviations from the allowed band
(either 5% or 10% of Vn referred to zone 1 and zone 2 respec-
tively), Vi , weighted by their corresponding time duration, ti , over
the total observation time period, T.

Vi Vl up
i
Vi ti
VOOB = , Vi = 0 (3)
T
V V


i l low
where
Vn (1 0, 05) (zone 1)
Vl up , Vl low = ;
Vn (1 0, 1) (zone 2)
Using the case study described in Appendix I, supposed to evolve
through the years according to a business as usual (BAU) scenario
as described in the ATLANTIDE project [20], an example of the
evaluation of this system index is presented considering differ-
ent degrees of implementation of network controls, i.e. different
degrees of smartness. As smart grid application we have considered
the distributed network control described in [17], which has been
developed to control both the primary substation OLTC transformer
and the DGs injected active and reactive power. Traditionally,
the OLTC control behaves as an automatic voltage regulator (AVR)
based on the voltage measurement at the transformer secondary
busbar. In this procedure, the OLTC can be operated in coordina-
tion with network remote measures, dynamically setting the most
suitable voltage reference at the primary substation MV busbar
based on the downstream bus voltage requirements, thus facil-
itating the voltage regulation action performed by the DGs. The
network control strategy is based on an adaptive regulation area
selection which consists in the identication of a number of clus-
ters each one composed by MV network buses which are electrically
close and have similar working conditions and free participation
of DGs to the network regulation is enabled by a suitable price
mechanism.
The application considers the following smart-grid scenarios:
Base case: no remote measures and network control performed
only by the AVR operation for the OLTC based on the MV busbar
voltage at PS;
Smart grid basic: OLTC working as AVR as for the base case, dis-
tributed network control with reactive power provided only by the
DGs with size over 2 MW;
Smart grid advanced: OLTC coordinated with the remote voltage
measurements, distributed network control with reactive power
control provided only by the DGs with size over 2 MW;
Full smart grid: coordinated operation of OLTC, distributed net-
work control with reactive power control provided by all the DGs.
In Fig. 1 an example of a 24 h bus voltage prole computed for
the different scenarios is presented. As can be seen, the coordinated
control of OLTC allows limiting the voltage deviations inside zone 2
and for most of the time in zone 1, whereas the increase of available
reactive power contribution in the full smart-grid scenario allows
a more efcient voltage regulation with no voltage violation for the
entire day.
In Table 7, the numerical results of the simulations are shown
for the four-year period 20142017, reporting for each instance the
busbar identied with the highest voltage violation, the percent
time duration of out-of-band voltage conditions with respect to
the total observation period (1 year) as well as the resulting value
of VOOB , evaluated as for (3) with reference to both zones 1 and 2.
These latter values provide a clear quantitative indication of the
voltage prole performances achievable with the increasing level
of network smartness.
5.5. Integration with TSO control systems
In Italy the amount of RES and DG connected to the distribution
systems reached in few years 20 GW. The result is that DG can no
 M. Coppo et al. / Electric Power Systems Research 120 (2015) 136149
Fig. 1. Example of voltage prole under the four scenarios, resulted for a spring day of 2017.
longer be treated as a negative load and each primary substation
has to be considered as a traditional power generator with refer-
ence to the control capability offered to TSO. For this reason, each
20 s the expected production at the primary substation level aggre-
gated by photovoltaic, wind, CHP, and other DG should be provided
to TSO including estimates/data also LV generation. This means a
tremendous effort for DSO since MV and LV producers normally
are not equipped with measurement devices capable to transmit
data with such high frequency. Furthermore, there is also a serious
challenge for the communication system that has to take into con-
sideration a high number of participants. The solution of installing
millions of power measurement devices spread all over the MV and
LV networks is not feasible for the huge economic effort that is not
affordable by DSO in the present situation of nancial crisis. Then,
there is a general consensus on the fact that state estimators will
have to be used to integrate the data gathered from some strategic
points in the network. The quality of these estimators is fundamen-
tal for the proper operation of smart grids, but it is fundamental for
the global operation of power systems. The quality of estimates pro-
vided by DSO to TSO can be another performance parameter of the
SG. The uncertainty of estimates can be measured easily by compar-
ing estimates with what really happened in the system. Since TSO
started requiring these data before the full implementation of the
SG, a benchmark for the starting situation can be easily prepared by
comparing what happened before the smart grid with the situation
after the smart grid implementation in a given area. The per-
formance indicator is the year-by-year reduction of the achieved
uncertainties thanks to the integration of measurement devices, the
adoption of advanced state estimators and improvements in data
analytics.
As demonstrated in [18,19], the position of the measurement
devices plays a key role in the quality of estimates and, as a con-
sequence, on the effectiveness of an energy management system.
On the other hand, since the TSO requests to DSO at the primary
substation level the expected load demand and the expected pro-
duction aggregated by photovoltaic, wind, CHP, and other DG each
20 s, the DSO has to change its perspective and to exploit state
estimation techniques not only to improve the network opera-
tion (active management) but also to full the TSO requirements.
Indeed, this aspect should be guided by the performance regulation
since without any external drive the DSO is naturally motivated to
improve state estimation for achieving better and less expensive
network operation, although this does not necessarily help TSO
and the power system as a whole. The following example aims at
demonstrating that this is an essential point for both stability and
security of the power systems as well as for the reduction of oper-
ation costs. Starting from the results published in [18] where the
145
impact of state estimators on the energy management system is
highlighted, the following case study analyses to what extent the
estimates can affect the TSO operation. Three possible measure-
ment sets are analyzed. The default measurement conguration,
very similar to what can be currently found in distribution sys-
tems, and two optimized ones that further increase the accuracy
of the estimates. The default measurement conguration com-
prises:
- a voltage amplitude measurement at the primary substation,
- active and reactive power ow measurements at each MV feeder
head, and
- active and reactive power ow measurements at the MV photo-
voltaic generators connected to the test network (see Appendix
I, for the test network data).
The two different remeter placements allow accuracy as small
as 0.2 p.u. for the estimation of errors in 95% of the cases by also
using PMUs. Clearly the better the accuracy the better the energy
management system operates. In order to meet the TSO require-
ments, the DSO, even with the simplest (default) measurement
system, can disaggregate the (measured) MV DG power produc-
tion from the total load demand at the primary substation (i.e., the
sum of the measured power ows in the downstream branches).
Doing so, a virtual representation of the primary substation can be
used by the TSO based on an equivalent aggregated for PV, wind
and CHP generators and an equivalent aggregated load represen-
tation. All equivalent circuits should be updated every 20 s. Even
though the EMS worked properly thanks to the integration of mea-
surement devices and state estimators that give accurate enough
estimates of MV nodal voltages and line power ows, this was not
sufcient for the TSO whose prediction and control actions were
affected by a systematic error caused by a signicant amount of
LV generation. In order to force DSO to better estimate LV genera-
tion, the performance-based regulation should explicitly deal with
the quality of the data passed to the TSO. The following example
gives a quantitative idea of such systematic error. A representative
LV network, derived from the ATLANTIDE project database, con-
nected to a given number of MV nodes has been simulated [13].
In the LV network it has been assumed that 30% of customers are
equipped with PV generators. Fig. 2 shows, for a week of June
2020, the total load demand, obtained as sum of the measured
power ows at the starting end of each feeder minus the mea-
sured active power production of the MV generators (dotted line),
and the real power delivered to the customers (solid line) at the
substation. The residual power demand is smaller than the effec-
tive demand in the sunny hours of the day, when the PVs feed the
146 M. Coppo et al. / Electric Power Systems Research 120 (2015) 136149
Fig. 2. Real and estimated load demand at the substation of the test network.
load. It is clear that the data DSO provide to the TSO are affected
by an error that grows with the increase of PV in the LV networks.
The means for eliminating such systematic errors can be various;
measurements can be extended to the biggest LV generators, con-
sumption data analytics can be integrated with MV estimates (e.g.,
with 0.2 accuracy) to better estimate load, irradiance measurement
in each secondary substation can be used to assess the estimated
power production. Moreover, it is proved that without a specic
driver from a performance-based regulation there is little interest
to improve measurement/estimates at the level that is required by
TSO.
6. Conclusions
The Italian smart grid pilot projects are going to be completed
very soon. After a brief description of the most important fea-
tures of these projects, the paper identies the eld data needed
in order to prepare the regulation for smart grid. The regula-
tion should promote the innovation in the distribution system
so that future networks will have systems in place to handle the
novel paradigm of power generation. In order to avoid that smart
grid incentives do not give the expected benet, the regulation
should be based on the identication of few and simple indicators
and on a metric that allows the assessment of the performances.
The paper proposes ve indicators, easy to calculate, that can be
used to reward the DSOs that invest money in smart grids. The
most signicant indicators proposed in the paper are assessed
with simulations applied to an Italian representative case study
network.
Acknowledgments
The authors warmly thank Dr. L. Lo Schiavo and Dr. S. Larzeni
from AEEGSI for their valuable contribution to dene the theoreti-
cal basis for the smart grid regulation. The sharing of experience on
this specic eld and the continuous discussion helped the authors
propose a possible preliminary architecture for the novel regula-
tion, which will pave the way to the real large scale implementation
of innovation in the distribution business.
Appendix I. The Case Study: the ATLANTIDE rural network
The Case Study adopted in the paper has been selected in the
database made available by the Italian ATLANTIDE project [13].
The clustered rural network used in the simulation studies is
shown in Fig. 3. The network, characterized by long overhead lines
with r/x ratio close to unity, is constituted by 7 relatively long feed-
ers (mostly small cross section overhead conductors for a total
extension of about 160 km) with 103 MV nodes supplied by one
HV/MV substation. Five PV plants are connected at MV level for
a total of 7.7 MVA installed. The nominal load - a mix of agricul-
tural, residential and small industrial customers is about 18 MVA
at the peak. This is the starting situation for this network, refer-
ring to the year 2010. The used planning scenario, named in the
ATLANTIDE project as Business-As-Usual (BAU), assumes that no
signicant changes occur in the current political drivers (e.g. no
changes in the incentives for the renewables) and the actual growth
trends for demand and productions continue without discontinu-
ities [20]. The constant annual growth rates used in the studies
are reported in Table 8. The annual load growth has the typical
behaviour of developed countries and is inuenced by the eco-
nomic scenarios and the improvement of energy efciency. The
DG growth is inuenced by the current trends that envisage a 2
GVA/year integration rate for small, roof mounted PV generators in
the next years. While this representation is common for loads, with
a natural increment year by year of the existing customers demand,
the generators usually do not increase their nominal power during
their life. Therefore, the annual growth rates allotted to the genera-
tors have been turned into the appearance of new plants in specic
Table 8
Annual growth rates of loads and generators.
Annual growth rates Values
Loads Agricultural 0.4%
Residential, LV 1.6%
Rural MV customers 0.6%
Generators Photovoltaic 10%
Commercial/ofces 2.1%
Transportation 1.1%
 M. Coppo et al. / Electric Power Systems Research 120 (2015) 136149
Fig. 3. The ATLANTIDE representative rural network.
nodes of the distribution network and in given years within the
planning period. It is worth noting that the appearance of gener-
ators in the networks is completely random, in order to exalt the
system criticalities.
At the end of the studied/planning period (2030), according with
the BAU scenario, 17 further MV PV plants have been installed for
a total of about 52 MVA and the delivered power to the customers
has grown to 21.7 MVA at the peak [20].
In one of the proposed application some nodes of the represen-
tative rural MV network have been expanded down to the LV level.
The ATLANTIDE clustered LV representative network, used in the
simulation and shown in Fig. 4, is supplied by a 20/0.4 kV 250 kVA
secondary substation. The served customers adsorb 154.4 kVA
(consisting in 24 single-phase loads and 3 three-phase loads) at the
peak. The four feeders are long respectively 800 m, 600 m, 400 m
and 200 m. In the proposed study, it is assumed that 30% of LV cus-
tomers have a PV generator installed, whose size equals the rated
power of the load and is connected to the same phase conductor.
This network model is used by scaling the nominal MV/LV loads of
the rural MV network.
In another application proposed in the paper the test network
is used for a medium term planning study for the identication of
the optimal expansion of the rural network, with the tool described
below. The considered planning period that takes into account of
the natural growth of the loads and the connections of several new
generators, is 12 years, from 2010 to 2022, subdivided in four years
step,
In Table 9 the main parameters of the planning problem have
been summarized. For the sake of simplicity, the interest and ina-
tion rates have been assumed constant during the whole planning
period.
147
In the rural reference network the PV has been assumed the
only generation technology. The small electric demand does never
produce overload contingency. Conversely, the central hours of
the day (between 11:30 a.m. and 2:00 p.m.) are critical and some
nodes suffer for over-voltages beyond the maximum allowable
threshold. This happens because the feeders with the greatest
power production deliver to the nal users less power than the
one generated by the local PVs. Moreover, due to the length
of some feeders, in the evening, when the residential and agri-
cultural demand is at the peak, excessive voltage drops may
happen, especially in emergency conditions where some nodes
can nd themselves over 40 km far from the primary substa-
tion.
A.1. Planning of distribution systems
In the paper a planning software tool, developed in the last
decade at the University of Cagliari and able to support the expan-
sion and development of MV distribution networks with signicant
quantities of DG, has been used [21]. Through a user-friendly inter-
face, the planner has the possibility to perform different studies on
a given network, by resorting to specic optimization modules that
are interactive with each other. In the paper, the network topology
optimization (NTO) module has been used.
In the NTO module, that minimizes the generalized cost of the
network, taking into account building, maintenance, upgrading
costs and the cost of losses and interruptions [2224], the opti-
mization algorithm allows simultaneously nding spurious open
loop network topologies, with trunks and laterals, the normal state
network, the position of emergency connections and the optimal
allocation of automatic sectionalizing switching devices (ASSDs).
148 M. Coppo et al. / Electric Power Systems Research 120 (2015) 136149

Fig. 4. The ATLANTIDE representative LV network.

Table 9
Main parameters used for planning calculations.

Planning parameters Values

Planning period 4 years

Unitary costs Overhead lines (length) 70 kD/km
Overhead lines (size) 413 D/(km mm2 )
Financial rates Interest 8.0%
Ination 1.0%
Operating Voltage Ordinary operating condition 5%
limits deviations Emergency operating condition 10%
Overload Ordinary operating condition 0%
Emergency operating condition +10%
Acceptable risks of Voltage deviations Any operating condition 5%
operating limits Overload Ordinary operating condition None (deterministic)
violations Emergency operating condition 10%

The comparison of planning strategies is made on the basis of the
overall cost of the network, composed of three terms: the invest-
ments (CAPEX) for upgrading existing network components or
installing new ones (lines, transformers, breakers, etc.), the energy
losses and the energy not supplied within prexed levels of sys-
tems availability. The CAPEX has been assessed with the following
equation:

Nelements

Nelements
Cinv = C0j = (B0j + M0j R0j )
j=1 j=1
where Nelements is the number of new or upgraded network equip-
ment, C0j is the net present value of the cost of the jth component,
and B0j , M0j and R0j are, respectively, the building cost, the main-
tenance cost and the residual value actualized at the beginning of
the planning period with suitable nancial equations that depend
on interest and ination rates.
References
[1] European Commission Joint Research Center, A view on smart grids from pilot
projects: lessons learned and current developments (July) (2011).
[2] F. Villa, Regulation of smart meters and AMM systems in Italy, CIRED pro-
ceedings, in: 19th International Conference on Electricity Distribution, Vienna
2124 May 2007, paper n, 2007, p. 437.
[3] E. Valigi, E. Di Marino, Networks optimization with advanced meter infrastruc-
ture and smart meters, CIRED proceedings, in: 20th International Conference
on Electricity Distribution, Prague, 811 June 2009, 2009.
[4] Resolution ARG/elt 39/10, Procedures and criteria for the selection of invest-
ments admissible for incentives pursuant to paragraph 11.4, letter d) of Annex
A to AEEGSI Authority for Electricity and Gas Resolution No. 348/07 of 29th
December, 2007.
[5] E. Fumagalli, L. Lo Schiavo, Regulating and improving the quality of electricity
supply: the Italian case, Eur. Rev. Electr. Mark. 3 (October (3)) (2009).
[6] L. Lo Schiavo, R. Vailati, The Italian incentive regulation for improving the con-
tinuity of electricity transmission, in: 10th IAEE European Conference, Paper
196, September, 2009.
[7] L. Lo Schiavo, M. Delfanti, E. Fumagalli, V. Olivieri, Changing the regulation for
regulating the change: Innovation-driven regulatory developments for smart
grids, smart metering and e-mobility in Italy, Energy Policy 57 (June) (2013)
506517, ISSN 0301-4215.
[8] M. Delfanti, L. Lo Schiavo, V. Olivieri, The Italian regulatory framework for devel-
oping smart distribution grids, Int. J. Emerg. Electr. Power Syst. 13 (5) (2012)
(Article 4).
[9] B. Dupont, L. Meeus, R. Belmans, Measuring the smartness of the electricity
grid, in: 7th International Conference on the European Energy Market (EEM)
2010, Madrid, 2010.
[10] The European Electricity Grid Initiative, Roadmap 201018 and Detailed
Implementation Plan 201012, 12 06 2012 (online), 2012, Available from:
www.smartgrids.eu
[11] I. Losa, M. De Nigris, T. Vu Van, I. Herold, W. Hribernik, Analysis of the on-
going research and demonstration efforts on smart grids in Europe, in: Proc.
22nd International Conference on Electricity Distribution, CIRED, paper 0836,
Stockholm, 1013 June, 2013.
[12] P. Caramia, G. Carpinelli, P. Verde, Power Quality Indices in Liberalized Markets,
John Wiley & Sons, Chichester, West Sussex, UK, 2009, ISBN: 978-0-470-03395-
1.
[13] http://www.progettoatlantide.it
[14] V. Giordano, I. Onyeji, G. Fulli, M.S. Jimnez, C. Filiou, Guidelines for Conducting
Cost-Benet Analysis of Smart Grid Projects, JRC Reference Reports, European
Commission (2012).
[15] Implementation Guideline for Network Code, Requirements for Grid Con-
nection Applicable to all Generators, European Network of Transmission
System Operators for Electricity (ENTSOE), 16 October, 2012, Available from:
http://www.entsoe.eu
[16] G. Celli, E. Ghiani, F. Pilo, G.G. Soma, Reliability assessment in smart
distribution networks, Electr. Power Syst. Res. 104 (2013) 164175,
http://dx.doi.org/10.1016/j.epsr.201307001, ISSN: 03787796.
[17] R. Caldon, M. Coppo, R. Turri, Coordinated voltage control in MV and LV distri-
bution networks with inverter-interfaced users, in: Proc. IEEE Powertech 2013,
Grenoble, 1620 June, 2013.
 M. Coppo et al. / Electric Power Systems Research 120 (2015) 136149 149

[18] G. Celli, P.A. Pegoraro, F. Pilo, G. Pisano, S. Sulis, DMS cyber-physical simulation
for assessing the impact of state estimation and communication media in smart
grid operation, IEEE Trans. Power Syst. (February) (2014).
[19] P.A. Pegoraro, F. Pilo, G. Pisano, S. Ruggeri, S. Sulis, Co-simulation of distribution
active management and distribution state estimation to reduce harmful effects
of inaccuracies, in: Proc. Powertech 2013, Grenoble (F), 1620 June, 2013.
[20] A. Bracale, R. Caldon, G. Celli, M. Coppo, D. Dal Canto, R. Langella, G. Petretto, F.
Pilo, G. Pisano, D. Proto, S. Scalari, R. Turri, Analysis of the Italian distribution
system evolution through reference networks, in: Proceedings IEEE PES ISGT
Europe Conf., Berlin 1417 October, 2012.
[21] G. Celli, F. Pilo, G. Pisano, G.G. Soma, Optimal planning of active networks,
in: Invited Paper in PSCC 2008, 16th Power Systems Computation Conference,
Glasgow, Scotland, July 1418, 2008.
[22] G. Celli, R. Cicoria, S. Mocci, F. Pilo, Optimization of MV distribution network
in presence of distributed generation, in: Proceedings 14th PSCC conference
Seville, Spain), 2002.
[23] F. Pilo, V. Allegranza, G. Celli, R. Cicoria, S. Mocci, An integrated tool for optimal
active network planning, Proc. CIRED Conf. (2003) 64.
[24] G. Celli, F. Pilo, Optimal sectionalizing switches allocation in distribution net-
works, IEEE Trans. Power Deliv. 14 (3) (1999 July) 11671172.

The author has requested enhancement of the downloaded file. All in-text references underlined in blue are linked to publications on ResearchGate.