This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON SMART GRID 1

An Innovative Measurement Approach for Load

Flow Analysis in MV Smart Grids
Antonio Cataliotti, Member, IEEE, Valentina Cosentino, Dario Di Cara, Pierluca Russotto, Enrico Telaretti,
and Giovanni Tinè, Member, IEEE

Abstract—This paper proposes a novel measurement proce- not only on the political support [3] but also on several tech-
dure for load flow analysis in medium voltage (MV) smart nical issues, which can be solved with the development of
grids. The new approach is based on power measurements the smart grids (SGs). Such networks are characterized by an
at low voltage (LV) side of MV/LV distribution substations.
Such measurements are carried out by means of power qual- increased number of monitoring, communication, and man-
ity analyzers. If compared with the load flow methods based agement systems, which can guarantee many functionalities
on measurements at the MV side, the proposed approach has such as: voltage and frequency regulation, anti-islanding pro-
a lower cost and it offers an equally reliable measurement sys- tection, fault location and isolation, service restoration, power
tem. The proposed algorithm takes the basic concept of the quality monitoring, system diagnostic, and so on [4]–[6].
current summation backward/forward method; some modifica-
tions are introduced to use LV load powers measurements. To At distribution level, key elements of a SG are control
this aim, the MV/LV power transformer is properly modeled and automation systems of primary and secondary substa-
and taken into account in the algorithm. The proposed approach tions, as well as of DG systems connected to the grid.
is validated by means of both simulations and experimental The loads control can also improve the power flow man-
measurements; the test system is the MV distribution grid of agement. All these functionalities can be achieved only with
Ustica Island, Italy. First, the proposed algorithm is compared
with the MATLAB/Simulink load flow tool, which is based on a suitable communication system, with proper reliability and
the traditional Newton–Raphson method. Second, a comparison real-time features [7]–[9]. Traditionally, supervisory control
is carried out between the proposed method and the measure- and data acquisition (SCADA) systems are used to ensure
ment data collected in the actual MV network of Ustica. Such proper networks operation, by means of continuous electri-
comparison takes into account the uncertainties of both pro- cal quantities monitoring and distributed actuators managing.
posed method, and experimental active and reactive MV power
measurements. A fundamental role of the monitoring system is load flow anal-
ysis, which evaluates a lot of network electrical data, such
Index Terms—Advanced metering infrastructure, load flow as node voltages and power flows on the distribution feed-
analysis, power quality analyzer (PQA), power system measure-
ments, power system monitoring, smart grid (SG). ers. To perform load flow analysis in medium voltage (MV)
distribution networks, two types of measurement instruments
can be used: 1) phasor measurement units (PMU) [9]; and
2) power quality analyzers (PQAs). These devices are gen-
I. I NTRODUCTION erally connected to the MV network by means of proper
HE MODERN power grids are undergoing fundamental
T changes in their structure, thanks to the integration of
distributed generation (DG) plants, distributed energy storage,
voltage and current transducers. However, such instruments
are not widespread diffused over distribution networks; thus
high instrumentation and installation investments are needed.
and smart metering devices [1], [2]. Within this context, the In the last years, state estimation (SE) algorithms have been
diffusion and integration of renewable energy sources depends proposed in literature, in order to reduce the number of MV
measurement points. Such methods require an accurate model
Manuscript received September 30, 2014; revised January 12, 2015 and of the network and some historical information on the powers
April 2, 2015; accepted May 2, 2015. This work was supported by the drawn by the loads (pseudo-measurements) [11], [12].
Research Project 2012-ATE-0395 “Sviluppo di dispositivi di misura innovativi In this paper, the authors present an innovative measurement
di potenza, energia e power quality per la gestione delle Smart Grid attraverso
un’opportuna infrastruttura di comunicazione” (Development of new power, approach for load flow analysis in MV distribution networks,
energy and power quality measurement devices and communication infrastruc- in which one voltage measurement is performed at the MV bus
tures for smart grids applications) through the University of Palermo under bar of the primary substation while the loads powers are mea-
Grant FFR 2012-2013. Paper no. TSG-00973-2014.
A. Cataliotti, V. Cosentino, P. Russotto, and E. Telaretti are with the sured at low voltage (LV) level. The measurements of loads
Department of Energy, Information Engineering and Mathematic Models, active and reactive powers are performed by PQAs, which are
University of Palermo, Palermo 90128, Italy (e-mail: acataliotti@ieee.org; connected at the LV side of each secondary substation. The
cosentino@dieet.unipa.it; pierluca.russotto@unipa.it; telaretti@dieet.unipa.it).
D. Di Cara and G. Tinè are with the National Research Council-Institute use of LV measurements instead of MV ones allows to avoid
of Intelligent System for Automation, Palermo 90141, Italy (e-mail: or reduce the measurement points installation costs. In fact, in
dicara@pa.issia.cnr.it; tine@pa.issia.cnr.it). many cases PQAs are already installed in the secondary sub-
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. stations to compute the total amount of energy absorbed by
Digital Object Identifier 10.1109/TSG.2015.2430891 LV loads supplied by the power transformer. In such cases,
1949-3053 c 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE TRANSACTIONS ON SMART GRID

no further installation costs are needed. On the other hand, In order to improve the accuracy of load flow analysis with-
in the case of new installations, the costs are lower than out increasing the cost/benefits ratio, the proposed solution
those for MV installations. Starting from LV measurements, herein proposed is based on an innovative approach, which
a backward/forward (B/F) algorithm is used to evaluate the makes use of PQAs active and reactive power measurements
power system load flow. This new measurement approach was installed at LV level. As already mentioned, the use of LV
introduced in [13] for the MV distribution network voltage measurements, instead of MV ones, allows a significant cost
profile analysis, by considering only the lines longitudinal reduction for measurement instruments and installation. This
impedances. The performances of the load flow algorithm were feature can be a strong stimulus for the implementation of the
tested by means of a preliminary simulation study, which was proposed solution in the real networks. In fact, in many cases
carried out on one of the two radial branches of the MV dis- PQAs are already installed in the secondary substations at LV
tribution network of Ustica Island (Italy, Mediterranean Sea). level. On the other hand, in the case of new installations volt-
In this paper, the complete load flow algorithm is presented. age transducers are not required, LV current transducers are
It allows to evaluate both amplitudes and phases of the nodal cheaper than MV ones and they can be more easily connected
voltages and the active and reactive powers in each branch. The to the LV cables or bus bars; moreover, the energy interruption
lines are modeled considering both longitudinal impedances is shorter (because the installation at the LV level is faster) and
and shunt admittances. The method is validated by mak- it involves a lower number of customers.
ing a comparison with a traditional Newton–Raphson (NR) Furthermore, the proposed approach is based on the use
resolution method. The simulations are carried out on the com- of experimental measurements. Thus, in comparison with the
plete model of the Ustica’s MV distribution grid. The results pseudo-measurements estimations of traditional SE methods,
obtained with the proposed method are also compared with the proposed measurement procedure offers the advantage of
the actual measurements on the MV network under test, taking higher accuracy and lower computational efforts.
into account the measurement uncertainties. On the other hand, the proposed approach requires the
This paper is organized as follows. Section II presents an acquisition of all the PQAs measurements. For this purpose,
analysis of the existing load flow methods and techniques, the PQAs should be connected to a central monitoring sys-
comparing their features and performances with those of the tem by means of a proper communication system, which
proposed approach. Section III describes the proposed mea- allows transmitting the measured data. In the perspective of
surement approach and the algorithm used for load flow SG development, such communication infrastructure is a gen-
analysis. Section IV presents the validation of the proposed eral requirement. Since a higher number of measurements
approach, which is carried out by means of both simulations have to be acquired, the updating time of the proposed algo-
and experimental measurements. In detail, the comparison rithm is longer than the one of the SE algorithms. This means
between the proposed algorithm and the NR method is car- that the use of PMUs and SE algorithms can be preferred
ried out in simulation and the obtained results are reported in when high real-time performances are needed. In conclusion,
Section IV-A. Section IV-B presents the results of the exper- depending on the specific application, the choice of the most
imental comparison between the algorithm results and the suitable solution shall be based on a tradeoff between real-time
measurement data collected in the MV network under test. features, accuracy requirements, and costs.
As regards the load flow algorithm for distribution systems,
the choice is very challenging, due to the networks intrinsic
II. BACKGROUND OF THE P ROPOSED features, such as radial topology, high number of nodes, and
L OAD F LOW S OLUTION high resistance/reactance (R/X) ratio of the feeders [14], [15].
As already mentioned, PMUs and PQAs can be used in Specifically, due to the high R/X ratio, distribution networks
MV distribution networks to perform load flow analysis. In behave as “ill conditioned systems,” thus the classical algo-
actual MV networks, very few instruments (PMUs or PQAs) rithms (such as fast-decoupled and Newton methods) are not
are already installed at MV level. Thus, in order to apply suitable for most of the distribution power flow programs.
traditional load flow methods, new MV measurement points B/F methods are mostly used in radial distribution systems,
should be installed. This entails the costs of measurement thanks to their robust convergence characteristics, simplicity
instruments, MV voltage and current transducers (one for each of implementation, low memory, and computational require-
phase), MV cabinet and connections, and energy interruption ments. Such methods were originally developed for radial
for installation. Such high costs and installation requirements networks and they have been extended to weakly meshed
have limited the wide use of both PMUs and PQAs in MV networks [16], using a multiport compensation technique [17].
distribution networks. To overcome these problems, SE algo- In fact, for radial networks the B/F algorithms convergence is
rithms have been proposed in literature, which are based on always guaranteed. On the other hand, the convergence char-
few MV measurements and the availability of load powers acteristics of B/F methods deteriorate when the distribution
pseudo-measurements. Nevertheless, such load models provide grid becomes more meshed [16]. To cope with meshed net-
only an approximate estimation of the load and this affects works, a loop-analysis-based power flow method was proposed
the accuracy of SE methods. As regards this, several studies in [18], whose convergence features remain stable even when
have been carried out in order to identify the proper number the network becomes more meshed. Several variant of the B/F
and placement of measurement instruments for a reliable and method have been analyzed in the last two decades [19], [20],
enough accurate SE in distribution networks. with the aim to enhance the model performances in terms of
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

CATALIOTTI et al.: INNOVATIVE MEASUREMENT APPROACH FOR LOAD FLOW ANALYSIS IN MV SGs 3

precision and computational efficiency. In brief, B/F meth- and transformers data were provided by the Ustica’s DSO
ods can be mainly divided in three different classes [21]: (see the next section).
1) current summation methods; 2) power summation methods;
and 3) admittance summation methods. One of the most known B. Load Flow Algorithm
current summation methods was developed by Kersting [21].
The proposed algorithm (LV-B/F) for the load flow cal-
It is based on the ladder iterative technique and it involves
culation of the MV system starts from the aforesaid LV
forward and backward sweeps through the network, by using
measurements and it employs an iterative B/F sweep algorithm
Kirchoff’s voltage and current laws. The backward sweep
to update bus voltages and branch currents.
starts from the last node and ends with the source node, while
A flow chart of the LV-B/F algorithm is shown in Fig. 1.
the forward sweep calculate the voltage drops. This method is
The first step is the acquisition of the powers Pload, i and
the most suitable for radial distribution systems; with proper
Qload,i measured by the PQAs at each substation and the
modifications, it can also be used to solve weakly meshed
mean value Vm of the voltages measured at the MV bus bars
networks. In [22], a power summation method is proposed
of the generation power plant (see the next section, Fig. 2).
for meshed networks with more than one feeding node. First,
The MV active, reactive and apparent powers (Pi , Qi , and S̄i ,
a conversion process is performed to obtain an equivalent sin-
respectively) at each node i are calculated starting from the
gle source network with radial configuration. This conversion
measured quantities
allows using conventional equations to solve the load flow
 
problem. The network conversion is performed by creating S̄load,i  2
break points and dummy buses, in which appropriate complex Pi = Pload,i + Po + Pcc (1)
Sr,i
power are injected to preserve the original system characteris-
 
tics. In [23], an admittance summation method is proposed for S̄load,i  2
radial and weakly meshed distribution network. This method is Qi = Qload,i + Qo + Qcc (2)
Sr,i
suitable when all active and reactive node loads have constant
admittance; in this case, the method is noniterative and much S̄i = Pi + jQi (3)
faster than the other methods, especially with a high number of
where Po , Qo , Pcc , Qcc , and Sr,i are the no-load active and
loads. All the proposed methods require an accurate network
reactive powers, the short-circuit active and reactive power
model and the load data (measured or estimated).
sand the rated apparent power of the MV/LV transformer con-
In this paper, a current summation method is chosen because
nected at the node i, respectively; S̄load,i = Pload,i + jQload,i is
it is the most widely used and it can be easily applied to
the total apparent power absorbed by the loads connected at
radial networks (such as the distribution grid of Ustica). The
the LV side of the distribution transformer at node i. For each
proposed algorithm takes the basic concept of the current
radial feeder, the N nodes are numbered from the last node
summation method; in order to use the LV measurements of
(i = 1) to the slack bus (i = N).
the load powers, the MV/LV power transformer is properly
After this, the voltage at each node is set equal to the value
modeled and take into account in the algorithm.
Vm and the iteration process for load flow calculation begins.
At each iteration, both backward and forward sweep are per-
III. D ESCRIPTION OF THE P ROPOSED M EASUREMENT formed. In the backward sweep, both branch currents and node
A PPROACH AND A LGORITHM voltages are calculated; in the forward sweep only these last
A. Measurements and Network Model are updated. The iteration process ends when the differences
between the voltage values obtained in the backward sweep
The proposed approach requires the following measurement
and those obtained in the forward sweep are less than or equal
instruments, which provide the input data for the load flow
algorithm (see the next section): three voltmeters installed at to a preassigned tolerance. Generally, a tolerance of 10−4 V
the MV bus bars of the primary substation and connected by is acceptable.
means of three voltage transformers (VTs); one three phase The backward sweep begins calculating the branch current
PQA for each secondary substation, which is installed at the I r,1 injected at the terminal node (i = 1)
 ∗
LV side of the MV/LV power transformer and connected by S1 Ẏ1 V̄1
means of three LV current transformers (CTs). I r,1 = √ + ·√ . (4)
3V 1 2 3
The network model is also needed for load flow analysis.
In the model, each secondary substation represents a node of In (4), V̄1 is the line to line voltage phasor, Ẏ1 is the shunt
the MV network; the MV lines between the substations are admittance of the line and S̄1 is the complex apparent power
the branches. The lines are modeled considering both longitu- absorbed at the node 1. A FOR cycle is then performed from
dinal impedances and shunt admittances. The power MV/LV the node i = 2 to the slack bus (i = N), which allows to both
transformers and their LV loads are modeled by their equiva- update the node voltages and evaluate the branch currents.
lent MV loads. The transformers losses are evaluated starting More in detail, the voltage phasor V̄i at node i is expressed
from the rated data of the short circuit and no-load tests [24]. as the sum of the voltage phasor V̄i−1 at node i − 1, and the
All the aforesaid parameters can be obtained starting from voltage drop between node i and node i − 1
the data provided by the distributor system operator (DSO) of √
the network. For the study presented in this paper, the lines V̄i = V̄i−1 + 3 · Īr,i−1 · Żr,i−1 (5)
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE TRANSACTIONS ON SMART GRID

(i.e., the branch between the nodes i − 1 and i), is evaluated

as the sum of the current of the branch i − 1, Īr,i−1 , the
current absorbed by the equivalent MV loads connected to
the node i and the currents absorbed by the shunt admittances
fed by the node voltage V̄i
 ∗  
Si Ẏi + Ẏi−1 V̄i
I r,i = I r,i−1 + √ + ·√ . (6)
3V i 2 3
At the end of the backward sweep, the voltages are memo-
rized in a vector (V̄b ) and the forward sweep begins. It starts
from the node nearer to the slack bus (i = N − 1) and stops
at the terminal node (i = 1). The node voltages are updated
as follows:
√
V̄i = V̄i+1 − 3 · Īr,i · Żr,i (7)
where Īr,i is the current calculated in the backward sweep. The
new node voltages are memorized in a second vector (V̄f ). The
algorithm calculates the difference between the two voltage
vectors (|V̄b − V̄f |); if this difference is lower than the spec-
ified tolerance (Toll = 10−4 V), the sequence of backward
and forward sweeps ends. Otherwise, a new iteration begins.
Unlike many traditional B/F methods, in the proposed algo-
rithm this condition should be verified not only in the slack
bus but also in all the nodes. Once the iteration process ends,
the power fluxes are calculated
√ ∗
S̄r,i = Pr,i + jQr,i = 3 · Īr,i · V̄i . (8)

IV. VALIDATION OF THE P ROPOSED

M EASUREMENT A PPROACH
A. Simulations
In order to validate the proposed LV-B/F algorithm, a simu-
lation study was carried out on the Ustica Island’s distribution
network model (see Fig. 2). The network is composed by
two MV radial feeders departing from the diesel power plant.
The first feeder (upper feeder, “Vittorio Emanuele”) sup-
plies 13 secondary substations; the second one (lower feeder
“AUSL”) supplies 11 secondary substations. In Fig. 2, caps
letters and numbers indicate network nodes and branches,
respectively. Each substation is equipped with a MV/LV power
transformer, which is coupled to the MV bus bar through
a bypass connection. A PQA is connected on the LV side of
the power transformer. Two PQAs are installed at the MV bus
bars of the generating station; each of them is connected by
means of three CTs and three VTs. For the implementation of
the LV-BF algorithm, such PQAs provide the Vm measurement.
All the measurements are collected and transmitted to the DSO
control center, by means of a HiperLAN communication sys-
tem. The lines characteristics and transformers parameters are
detailed in the Appendix.
The simulation study was carried out by reproducing some
Fig. 1. Flow chart of the LV-B/F load flow algorithm. actual conditions of power consumption at the LV side of
each secondary substation. The data were provided by the
Ustica DSO. The proposed LV-B/F method was validated by
where Żr,i−1 and Īr,i−1 are the longitudinal impedance comparing its simulation results with those provided by the
and the branch current between node and node MATLAB/Simulink load flow tool, based on the NR method.
i − 1, respectively. The current of the branch i For example, Figs. 3 and 4 show the comparison for the nodal
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

CATALIOTTI et al.: INNOVATIVE MEASUREMENT APPROACH FOR LOAD FLOW ANALYSIS IN MV SGs 5

Fig. 2. Scheme of the Ustica MV network (with the detail of PQAs connection scheme).

Fig. 3. Simulation results. Voltage magnitude and phase at the MV nodes of the upper feeder. Comparison between the NR method and the LV-B/F algorithm.

Fig. 4. Simulation results. Voltage magnitude and phase at the MV nodes of the lower feeder. Comparison between the NR method and the LV-B/F algorithm.

voltages (in magnitude and phase) in the upper and the lower B. Comparison With Measured Values
feeder, respectively. Figs. 5 and 6 show the same comparison The active and reactive powers were evaluated with the
for the active and reactive powers. These results refer to the LV-B/F algorithm at the beginning of each MV feeder depart-
load condition of Table I, which reports the active and reac- ing from the MV bus bars of the generation power plant. These
tive power values measured in August 18, 2012, 12:30 A . M. simulation results were compared with the active and reactive
For the Ustica’s network, this is a typical load condition powers measured by two PQAs, which are actually installed
of the summer period, when the highest power flows occur. in the Ustica network in the aforesaid metering sections. In
It can be seen that the nodal voltages and the active and reac- order to verify the compatibility between the simulation results
tive powers are correctly evaluated by the proposed algorithm. of the LV-B/F method and the measured values, the uncer-
The maximum and average deviations in the voltage magni- tainty on the active and reactive MV powers measurements
tude are equal to 1.1 and 0.120 V for the upper feeder and was estimated (type B evaluation), starting from PQAs, CTs,
1 and 0.121 V for the lower feeder; respectively. The max- and VTs accuracy classes and considering a rectangular dis-
imum and average deviations in the voltage phase are lower tribution. The PQAs have an accuracy class of 0.5 and 2
than 10−3 degree, for both feeders. for active and reactive power measurement, respectively.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE TRANSACTIONS ON SMART GRID

Fig. 5. Simulation results. Active reactive power at each branch of the upper feeder. Comparison between the NR method and the LV-B/F algorithm.

Fig. 6. Simulation results. Active reactive power at each branch of the lower feeder. Comparison between the NR method and the LV-B/F algorithm.

TABLE I
L OAD C ONDITIONS —AUGUST 18, 2012, 12:30 A . M . on the active power measurement can be expressed as
ηVT % − 100εVT tan(θ)
uP_VT % ≈ √ (11)
3
(ηVT % and εVT are the VT ratio error and phase displacement,
respectively).
Starting from the aforesaid uncertainty contributions, the
total relative standard uncertainty on the active power mea-
surements (up %) and the expanded absolute uncertainty
(Up , confidence level of 95%, k = 2) can be obtained as

uP % = uP_PQA %2 + uP_VT %2 + uP_CT %2 (12)
2uP %P
UP = (13)
100
where P is the active power measured by the PQA at the MV
bus bars of the generation power plant. Similar considerations
can be made for the reactive power uncertainty evaluation.
The uncertainty of the proposed LV-B/F algorithm was eval-
As regards the active power measurement, the PQA relative uated in [27], by means of a Monte Carlo analysis. In brief,
standard uncertainty can be calculated as follows: the LV load power measurements uncertainties were estimated
(as previously described but in the absence of VTs). After this,
EP_PQA % the load flow algorithm was run 105 times; at each iteration,
uP_PQA % = √ (9)
3 the LV load powers were randomly changed within their uncer-
tainty ranges. In this way, the uncertainties on the estimated
where EP_PQA % = 0.5 (percentage error of the PQA active MV power flows were obtained.
power measurement). The compatibility between the simulation results and the
CTs and VTs have an accuracy class of 0.5. As regards measurements at the MV bus bars (with their uncertainties) is
the CT, the ratio error ηCT % and the phase displacement shown in Table II, in the reference load condition of Table I.
CT are very small [25]. Thus, cos(εCT ) ≈ 1, sin(εCT ) ≈ εCT , A further load condition is reported in Table III. It represents
ηCT %  100 and the CT relative standard uncertainty on the a typical load condition of the winter period, when the lowest
active power measurement can be expressed as power flows occur. For this new condition, the compatibility
ηCT % + 100εCT tan(θ) between the simulations and the measured data (with their
uP_CT % ≈ √ (10) uncertainties) is shown in Table IV.
3
The aforesaid results were obtained considering extremely
where θ is the phase shift between the voltage and current different load conditions and they confirm the reliability and
phasors [26]. Similarly, the VT relative standard uncertainty feasibility of the developed algorithm to evaluate the power
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

CATALIOTTI et al.: INNOVATIVE MEASUREMENT APPROACH FOR LOAD FLOW ANALYSIS IN MV SGs 7

TABLE II TABLE V
C OMPARISON B ETWEEN S IMULATION R ESULTS AND M EASUREMENT T RANSFORMER PARAMETERS —U PPER AND L OWER F EEDERS
DATA —S UMMER L OAD C ONDITION (TABLE I)

TABLE III
L OAD C ONDITIONS M EASURED IN N OVEMBER 1, 2013, 12:30 A . M .

TABLE VI
MV L INE PARAMETERS

of each secondary substation and one voltage measurement at

the MV bus bars of the power plant. The proposed LV-B/F
algorithm takes the basic concept of the current summation
method; some modifications are introduced to use the LV
TABLE IV measurements of the load powers. To this aim the MV/LV
C OMPARISON B ETWEEN S IMULATION R ESULTS AND M EASUREMENT power transformer is modeled and taken into account in the
DATA —W INTER L OAD C ONDITION
algorithm. Starting from the collected measurement data, the
LV-B/F algorithm is used to evaluate the power flows. It is
simpler than traditional load flow algorithms based on NRs
method, so it can be easily implemented in a SCADA system.
The proposed approach has been tested in the MV distribu-
tion grid of Ustica Island. The method has been validated by
flows in MV distribution systems. The comparison between the comparing the simulation results of the LV-B/F algorithm with
algorithm simulation results and the measured powers verify those obtained using the MATLAB/Simulink load flow tool.
the correctness of the implemented models of the MV lines and Furthermore, the compatibility has been shown between the
transformers and the weak sensitivity to the LV load measure- simulation results and the actual active and reactive powers,
ment uncertainties. In the algorithm, the rated data are used which were measured by two PQAs installed at the MV bus
for the model, but the real-network parameters can be differ- bars of the Ustica’s power plant.
ent (i.e., drifts from the rated data can occur). On the other Further studies are in course concerning the sensitivity anal-
hand, the measurement input data (from LV PQAs and MV ysis of the proposed method with respect to the unavailability
voltmeter) are intrinsically affected by the instruments uncer- of some LV load power measurements.
tainty. Despite to these occurrences, the obtained results show
the compatibility between the proposed method and the exper- A PPENDIX
imental on-field measurements. This is an indirect proof of the
feasibility and robustness of the proposed method with respect The MV/LV transformers and MV lines data of the Ustica’s
to bad measurements and/or inaccurate model. On the other network are detailed in Tables V and VI, respectively. It can
hand, further simulations were carried out in [27] to estimate be seen that the open load and short circuit powers are small
the influence of the uncertainty of the network model param- (in comparison with the load powers, see Table I). Thus an
eters. More in detail, the Monte Carlo analysis was repeated error in the identification of such parameters will not have
by considering a 10% error in the branches parameters; in high influence on the load flow calculation.
such condition a maximum increment of 0.5% of the standard
uncertainty on the power flow was obtained. R EFERENCES
[1] J. Fan and S. Borlase, “The evolution of distribution,” IEEE Power
Energy Mag., vol. 7, no. 2, pp. 63–68, Mar. 2009.
V. C ONCLUSION [2] K. Moslehi and R. Kumar, “A reliability perspective of the smart grid,”
In this paper, an innovative measurement approach has been IEEE Trans. Smart Grid, vol. 1, no. 1, pp. 57–64, Jun. 2010.
[3] A. Campoccia, L. Dusonchet, E. Telaretti, and G. Zizzo, “An analysis of
proposed for MV distribution network load flow analysis. The feed’in tariffs for solar PV in six representative countries of the European
method is based on the power measurements at the LV level Union,” Solar Energy, vol. 107, no. 3, pp. 530–542, 2014.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE TRANSACTIONS ON SMART GRID

[4] A. Cataliotti et al., “Hybrid passive and communications-based meth- [26] A. Cataliotti, D. Di Cara, A. E. Emanuel, and S. Nuccio, “Current trans-
ods for islanding detection in medium and low voltage smart grids,” formers effects on the measurement of harmonic active power in LV and
in Proc. IEEE Int. Conf. Power Eng. Energy Elect. Drives, Istanbul, MV networks,” IEEE Trans. Power Del., vol. 26, no. 1, pp. 360–368,
Turkey, May 2013, pp. 1563–1567. Jan. 2011.
[5] A. Kulmala, S. Repo, and P. Javentausta, “Coordinated voltage control [27] A. Cataliotti, V. Cosentino, D. Di Cara, E. Telaretti, and G. Tinè,
in distribution networks including several distributed energy resources,” “Uncertainty evaluation of a backward/forward load flow algorithm for
IEEE Trans. Smart Grid, vol. 5, no. 4, pp. 2010–2020, Jul. 2014. a MV smart grid,” in Proc. IEEE Int. Instrum. Meas. Technol. Conf.,
[6] H. Farag, E. El-Saadany, and R. Seethapathy, “A two ways Pisa, Italy, May 2015, pp. 1279–1284.
communication-based distributed control for voltage regulation in
smart distribution feeders,” IEEE Trans. Smart Grid, vol. 3, no. 1,
pp. 271–281, 2012.
[7] A. Cataliotti, D. Di Cara, R. Fiorelli, and G. Tinè, “Power-line com- Antonio Cataliotti (M’01) received the M.S. and the Ph.D. degrees in elec-
munication in medium-voltage system: Simulation model and onfield trical engineering from the University of Palermo, Palermo, Italy, in 1992 and
experimental tests,” IEEE Trans. Power Del., vol. 27, no. 1, pp. 62–69, 1998, respectively.
Jan. 2012. Since 2005, he has been an Associate Professor of Electrical and Electronic
[8] G. Artale et al., “Medium voltage smart grid: Experimental analysis Measurements with the Department of Energy, Information Engineering
of secondary substation narrow band power line communication,” IEEE and Mathematic Models, University of Palermo. His current research inter-
Trans. Instrum. Meas., vol. 62, no. 9, pp. 2391–2398, Sep. 2013. ests include power quality measurements, power line communications, and
[9] Y. Deng et al., “Communication network modeling and simulation for smart grids.
wide area measurement applications,” in Proc. IEEE PES Innov. Smart
Grid Technol. (ISGT), Jan. 2012, pp. 1–6.
[10] P. Castello, P. Ferrari, A. Flammini, C. Muscas, and S. Rinaldi, “A new
IED with PMU functionalities for electrical substations,” IEEE Trans. Valentina Cosentino received the M.S. and Ph.D. degrees in electrical engi-
Instrum. Meas., vol. 62, no. 12, pp. 3209–3217, Dec. 2013. neering from the University of Palermo, Palermo, Italy, in 2001 and 2005,
[11] M. E. Baran and T. E. McDermott, “Distribution system state estimation respectively.
using AMI data,” in Proc. IEEE/PES Power Syst. Conf. Expo. (PSCE), She is currently an Assistant Professor of Electrical and Electronic
Seattle, WA, USA, Mar. 2009, pp. 1–3. Measurements with the Department of Energy, Information Engineering
[12] M. Karimi et al., “Impact of load modeling in distribution state estima- and Mathematic Models, University of Palermo. Her current research inter-
tion,” in Proc. IEEE Int. Power Eng. Optim. Conf. (PEOCO), Melacca, ests include power energy and power quality measurements, detection of
Malaysia, 2012, pp. 67–71. disturbances sources in power systems, digital signal processing, virtual
[13] A. Cataliotti, D. Di Cara, P. Russotto, E. Telaretti, and G. Tinè, instrumentation, and smart grids.
“New measurement procedure for load flow evaluation in medium volt-
age smart grids,” in Proc. IEEE Int. Instrum. Meas. Technol. Conf.,
Minneapolis, MN, USA, May 2013, pp. 517–522.
[14] K. Balamurugan and D. Srinivasan, “Review of power flow studies on Dario Di Cara received the M.S. and Ph.D. degrees in electrical engineering
distribution network with distributed generation,” in Proc. IEEE PEDS, from the University of Palermo, Palermo, Italy, in 2005 and 2009, respectively.
Singapore, Dec. 2011, pp. 411–417. He is currently a Researcher with the National Research Council-Institute
[15] A. B. Eltantawy and M. M. A. Salama, “A novel zooming algorithm for of Intelligent System for Automation, Palermo. His current research interests
distribution load flow analysis for smart grid,” IEEE Trans. Smart Grid, include power quality measurements, characterization of current transduc-
vol. 5, no. 4, pp. 1704–1711, Jul. 2014. ers in nonsinusoidal condition, harmonic power measurements, power line
[16] D. Shirmohammadi, H. W. Hong, A. Semlyen, and G. X. Luo, communications, and smart grids.
“A compensation-based power flow method for weakly meshed distribu-
tion and transmission networks,” IEEE Trans Power Syst., vol. 3, no. 2,
pp. 753–762, May 1988.
[17] G. Gross and H. W. Hong, “A two-step compensation method for solving Pierluca Russotto received the M.S. and Ph.D. degrees in electrical engi-
short circuit problems,” IEEE Trans. Power App. Syst., vol. PAS-101, neering from the University of Palermo, Palermo, Italy, in 2011 and 2015,
no. 6, pp. 1322–1331, Jun. 1982. respectively.
[18] W. Wu and B. Zhang, “A three-phase power flow algorithm for distri- His current research interests include power systems modeling, measure-
bution system power flow based on loop-analysis method,” Int. J. Elect. ment and communication systems for smart grids, and smart metering.
Power Energy Syst., vol. 30, no. 1, pp. 8–15, 2008.
[19] G. W. Chang, S. Y. Chu, and H. L. Wang, “An improved
backward/forward sweep load flow algorithm for radial distribution
systems,” IEEE Trans. Power Syst., vol. 22, no. 2, pp. 882–884, Enrico Telaretti received the M.S. and Ph.D. degrees in electrical engineering
May 2007. from the University of Palermo, Palermo, Italy, in 2004 and 2008, respectively.
[20] J. Yuntao, W. Wenchuan, Z. Boming, and S. Hongbin, “An extension of Since 2008, he has been a Postdoctoral Fellow with the Department
FBS three-phase power flow for handling PV nodes in active distribu- of Energy, Information Engineering and Mathematic Models, University of
tion networks,” IEEE Trans. Smart Grid, vol. 5, no. 4, pp. 1547–1555, Palermo. His current research interests include electrical power system anal-
Jul. 2014. ysis and simulations, fault analysis in distribution networks, and distributed
[21] W. H. Kersting, Distribution System Modeling and Analysis, 3rd ed. generation and its impact on electrical systems.
Boca Raton, FL, USA: CRC Press, 2012.
[22] H. Haque, “A general load flow method for distribution systems,” Elect.
Power Syst. Res., vol. 54, no. 1, pp. 47–54, Apr. 2000.
[23] D. Rajicic and R. Taleski, “Two novel methods for radial and weakly Giovanni Tinè (M’04) received the M.S. degree in electronic engineering and
meshed network analysis,” Elect. Power Syst. Res., vol. 48, no. 2, the Ph.D. degree in electronics, computer science, and telecommunications
pp. 79–87, Dec. 1998. engineering from the University of Palermo, Palermo, Italy, in 1990 and 1994,
[24] T.-H. Chen and Y.-L. Chang, “Integrated model of distribution trans- respectively.
former and their loads for three-phase power flow analyses,” IEEE Trans. He is currently a Researcher with the National Research Council-Institute
Power Del., vol. 11, no. 1, pp. 507–513, Jan. 1996. of Intelligent System for Automation, Palermo. His current research inter-
[25] Instrument Transformers—Part 1: Current Transformers, IS/IEC ests include electromagnetic compatibility of power electrical drive systems,
Standard EN 60044-1, 1999. power-line communications, and smart grids.