Beruflich Dokumente
Kultur Dokumente
1, FEBRUARY 2014
251
I. INTRODUCTION
OTIVATED by the recent advances in communication and measurement systems, utilities have been
seeking efficient solutions for improving distribution systems
monitoring and automation. These advances include two-way
communication of the meter, data-management system, and
real-time access to information, such as consumption, voltage
sags, power outages, etc. [1], [2]. As a result, smart feeder and
consumer meters can achieve potential applications beyond
meter reading for billing purposes. For example, in [3], the
usage of feeder meters, which are installed in primary networks, is discussed as a cost-effective solution for distribution
monitoring and automation.
An important potential application of those meters is related
to fault management, since faults result in outages and lead to
reliability and power-quality (PQ) problems such as voltage
sags, temporary and sustained interruptions, and high operational costs. In this context, one of the main issues related to
Manuscript received January 28, 2013; revised May 17, 2013; accepted June
25, 2013. Date of publication July 16, 2013; date of current version January
21, 2014. This work was supported by FAPESP, CAPES, Brazil, and CAPES,
Brazil. Paper no. TPWRD-00127-2013.
F. C. L. Trindade and W. Freitas are with the Department of Electrical Energy
Systems, University of Campinas (UNICAMP), Campinas 13083-852, Brazil
(e-mail: fernanda@ieee.org; walmir@ieee.org).
J. C. M. Vieira is with the Department of Electrical and Computer Engineering, Sao Carlos School of Engineering/USP, So Carlos 13566-590, Brazil
(e-mail: jose.vieira@ieee.org).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TPWRD.2013.2272057
fault management are fault-location techniques [4] since reasonably high-accuracy fault-location schemes enable reducing
costs and time for energy supply restoration.
Traditionally, in distribution systems, fault-location techniques are based on manual outage mapping using consumers
calls, fundamental voltage, and current components measured
only at the substation and network electrical parameters. However, several methods have been recently proposed to take
advantage of advanced metering systems. There are methods
based on automated outage mapping [5], [6] or voltage sag
source detection [7][10], which indicate a probable area
(region) that may contain the faulted bus instead of pointing
out a strict fault location. There are also invasive methods.
For instance, in [11], the fault-location method requires the
injection of sinusoidal signals with two different frequencies
into the faulted distribution line. Thus, by using signal extraction techniques and mathematical processing, permanent
single-phase-to-ground faults can be located.
Another class of methods uses the voltage and current
measured at the substation, sparse voltage measurements, and
the network electrical parameters to estimate the fault location
[12][14]. Their performance can be considerably affected by
the fault resistance, and additional solutions are necessary in
order to deal with this uncertainty. For instance, an iterative
process dedicated to fault resistance estimation is used in [13].
These additional techniques can increase the complexity of the
fault-location methods.
In order to overcome some of the drawbacks observed in previous fault-location methods and aggregate value to the smart
feeder meters, this work proposes a robust fault-location technique that explores the voltage sag measurement capability of
these devices and fundamentals of the short-circuit theory. Once
a fault is detected and classified, the proposed method can be
used to locate faults at the distribution system by using measurements from smart meters installed at the primary network.
These smart feeder meters, which will be referred to simply as
feeder meters in this paper, can be polled (triggered) everytime a fault is detected to provide the information demanded by
the proposed method under request [15]. Sensitivity study results have shown that the proposed method presents good performance for different values of fault resistance, quantity, and
location of the feeder meters as well as measurement and load
estimation errors.
This paper is organized as follows. Section II presents the
proposed method as well as a general view of how the feeder
meters can be used to improve the fault-location methods.
Sections IIIV provide simulation results and sensitivity
0885-8977 2013 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.
252
meaning
(1)
where the subscript is related to a feeder meter installed at
and
are the voltage magnitudes measured
bus ;
during and before the fault (prefault), respectively; and superscript
represents each phase, so that one has the voltage deviation on phase , , and . If the bus is under fault, one
can estimate the fault current at each phase by using the voltage
deviation determined by meter based on the following expression:
for
(2)
where
is the
3
3 submatrix from the system
, as discussed in
three-phase bus impedance matrix
Appendix A, and
is the fault current calculated by
using the voltage measurement from meter and considering
a fault at bus . Note that the loads, represented by constant
impedance models, must be included into the bus impedance
matrix to improve the method accuracy. In order to include
the loads in the bus impedance matrix, all of the loads are represented by constant impedance models. Then, the equivalent
shunt impedances associated with the bus are included at the
matrix elements [16]. The load estimation can be done by using
typical load curves or based on the information provided by
smart consumer meters. The influence of the loads in method
performance is analyzed in Section IV-A. and in Appendix A.
Therefore, if one has
meters, there are
estimated
fault currents based on the assumption that the bus under fault
is bus . If the fault really occurred at bus , all of the estimated
currents must have practically the same value, which is close to
the real value. On the other hand, if the fault occurred at any
other bus, there will be an error on the fault current estimated
based on measurements from each meter . In addition, the error
(3)
where
is the fault current calculated for phase
with
measurements from the feeder meter at bus using (2), and
is the average of all the fault current values calculated
using the voltage measured at each feeder meter for bus . As
mentioned before, if the bus under analysis is the faulted bus, all
of the estimated currents will be practically the same. Thus, the
bus associated with the minimum
is selected as the faulted
bus.
Typically, feeder meters supply only voltage magnitudes;
however, in (2) and (3), currents are written as phasor quantities. In order to numerically calculate these expressions, it was
assumed that phases , , and are simply shifted by 120 ,
although distribution systems are unbalanced. Note that such
an assumption was used only for calculation purposes. In the
results presented in this paper, the system was unbalanced so
that the phase currents are not really shifted by 120 ; even so,
the proposed method shows good performance. The influence
of this assumption is investigated in Section V. When only
voltage magnitude is used, there is no need for synchronization;
since all meters are simultaneously polled by the same fault
event, decreasing the costs and complexity.
For certain situations, numerically, the minimum may be
associated with more than one bus , indicating multiple estimation. In order to deal with such cases and improve the method
performance, automated outage mapping can be used, as described in the following section.
A. Automated Outage Mapping
The communication among the feeder and consumer meters
and the control center allows monitoring de-energized meters
and, consequently, determining the part of the network that is
de-energized. By incorporating data from the geographic information system (GIS), the outage mapping can speed up and automate the typical manual process of fault location, which depends on consumer calls.
With smart meters, one can automatically build the outage
mapping by monitoring the de-energized transponders. As a result, the chances of indicating a wrong bus as the faulted bus
and multiple estimation occurrences can be minimized or even
eliminated. An example of such an application can be observed
in the feeder of Fig. 2. In this system, if a fault occurs downstream the recloser R1 (fault 1), it will cause its opening, and
the feeder meters M3 to M6 would report the outage in Zone
1. Therefore, the fault-location method could consider this information in order to decrease the search space as well as the
chances of indicating the wrong bus as the faulted bus. In case
of fault 2, if the fuse F3 blows, outage mapping would indicate
a fault occurrence in an even reduced zone (Zone 2), since M3
would be the only meter to report the outage.
The outage mapping can be combined with the proposed
fault-location method using two different approaches as
follows.
The first option consists in exploring the outage mapping
to decrease the search space in the fault-location process.
In this case, the computing demand is reduced. However,
this alternative requires polling several meters in order to
identify the affected zone, increasing the data traffic.
The second option is based on applying outage mapping
only in cases of multiple estimation, so that specific meters
are polled in order to support the decision of which local
refers to the affected one. In comparison to the first option,
even though the entire feeder is used as the search space
of the fault-location method (increasing the computational
effort), the data traffic is reduced.
Regardless of computational effort and data traffic, both options give the same results, and, in the following sections, the
first option will be employed.
B. Method Guidelines
The steps below summarize the method application after a
fault has been detected:
Step 1) Construct the bus impedance matrix
(this step
can be done offline).
253
(4)
254
Fig. 3. Distribution test feeder used for evaluating the fault-location algorithm (data are given in Appendix B).
The faults were applied at each bus of the circuit, and the
errors were calculated and grouped into classes, in order to simplify the results analyses. These error classes were divided from
0 to 400 m into intervals of 100 m, resulting in four classes. In
addition, if the distance error is higher than 400 m, it is considered that the method has failed to locate the fault region.
NUMBER
OF
TABLE I
BUSES WITHIN EACH ERROR CLASS WITHOUT
CONSIDERING OUTAGE MAPPING
TABLE II
NUMBER OF BUSES WITHIN EACH ERROR
CLASS CONSIDERING OUTAGE MAPPING
bus 36 (120 m of distance from bus 40), while when outage mapping is considered, bus 40 was indicated correctly as the faulted
bus. This happens because the search space is reduced to buses
40 to 45. The analysis of the results in Table II allows concluding
that the results can be improved by using outage mapping.
In the next section, the fault-location method robustness will
be assessed by running several sensitivity studies. Unless otherwise specified, all the results are obtained using outage mapping
and considering a fault resistance of 10.0 as the base case.
255
TABLE III
IMPACT OF LOAD VALUE ESTIMATION ERRORS
TABLE IV
IMPACT OF FAULT RESISTANCE VARIATION
TABLE V
IMPACT OF MEASUREMENT ERRORS
Measurement errors may degrade the quality of the fault current estimation. Therefore it is necessary to analyze the method
sensitivity to inaccuracies in voltage measurements. Errors can
be caused by measurement noise or measurement device inaccuracy, and in this case the deviations are relatively small, lower
than 1%, but likely to occur in all measurements. On the other
By comparing Tables V and II (ideal condition without measurement errors), it is possible to conclude that errors in voltage
measurement can decrease the method performance. However,
the results are still acceptable since for most of the presented
cases, the buses were located with distance errors equal or lower
B. Fault Resistance
Typically, fault-location methods are affected by the fault
resistance, which influences the fault current value. In order
to determine this influence on the proposed method, Table IV
presents the results considering different fault resistance values.
For the fault resistance values tested in this paper, it can be observed that the proposed method performance is not deeply affected by the range of fault resistance values. Therefore, fault
resistance uncertainty does not have a deep impact on method
accuracy.
256
TABLE VI
IMPACT OF DIFFERENT QUANTITIES OF VOLTAGE MEASUREMENTS
than 400 m. As a result, it must be pointed out that: the more accurate the voltage measurements, the better the method performance. This finding can be used to support the definition of the
technical requirements of smart feeder meters and their communication system infrastructure.
D. Feeder Meters Quantity and Location
All of the previous results had considered that voltage measurements were available from feeder meters installed in 13 out
of the 134 buses of the distribution system (i.e., about 10% of
the total number of buses). The feeder meters quantity and location were chosen without previous studies.
Since the proposed formulation is dedicated to the fault location through voltage measurements, it is reasonable to conclude
that if more feeder meters are installed, more accurate results
can be achieved. On the other hand, it is necessary to consider
that the costs and the amount of data grow if the quantity of
measurements increases.
In this context, the method performance was assessed by
using 10 and 5 feeder meters in the system of Fig. 3. For the
case with 10 feeder meters, their respective locations are buses:
3, 20, 30, 60, 100, 111, 118, 127, 129, and 134; and for the case
with five buses, they are: 3, 30, 100, 129, and 134. The results
are presented in Table VI, where 13 feeder meters refer to the
base case (Section III-B). It can be observed that the use of 5 m
did not degrade the quality of the solution significantly, since
most of the buses were located with distance errors lower than
100 m. Therefore, the method is robust enough to cope with a
reduced number of meters. This is an important characteristic
since economical factors may limit the installation of feeder
meters with voltage sag monitoring capability in distribution
systems.
The influence of the feeder meters placement was analyzed
as well. The results for three scenarios considering different locations of 10 feeder meters are shown in Table VII. These scenarios are arbitrarily divided in:
Case 1) feeder meters installed at buses 3, 20, 30, 60, 100,
111, 118, 127, 129, and 134;
Case 2) feeder meters installed at buses 15, 25, 44, 56, 66, 85,
89, 102, 112, and 131;
Case 3) feeder meters installed at buses 7, 36, 47, 66, 76, 84,
91, 104, 112, and 125.
TABLE VII
IMPACT OF FEEDER METERS ALLOCATION
TABLE VIII
IMPACT OF VOLTAGE MAGNITUDE AND ANGLE MEASUREMENTS
Note: in this table, the results were the same for all of the tested fault
resistances.
257
TABLE X
NUMBER OF BUSES WITHIN EACH ERROR CLASS, CONSIDERING OUTAGE
MAPPINGMOTOR LOAD OF 100 HP
TABLE XI
NUMBER OF BUSES WITHIN EACH ERROR CLASS, CONSIDERING OUTAGE
MAPPINGMOTOR LOAD OF 200 HP
TABLE XII
NUMBER OF BUSES WITHIN EACH ERROR CLASS, CONSIDERING OUTAGE
MAPPINGMOTOR LOAD OF 100 HPPHASOR MEASUREMENTS
TABLE IX
IMPACT OF ERRORS IN PHASOR MEASUREMENTS
observed that they change in the same proportion from one bus
to another and, as a result, the minimum is always associated
with the same bus.
A. Sensibility to Voltage Measurement Accuracy
The results from Section IV show that one of the main factors affecting the method performance is the accuracy of voltage
measurement. Therefore, Table IX presents the results considering voltage phasor measurements with outage mapping and
random noise (errors) in addition to all measurements according
to Case 1, as described in Section IV-C. In addition, the random
noises were included simultaneously in the voltage magnitude
and angle. Again, one can see this is an important factor affecting the method performance, but by comparing Tables V
with IX, it can be stressed that the method accuracy has been
improved significantly when angle measurements are considered, even in the presence of measurement errors.
258
is not iterative. Both methods require system parameters; however, the method proposed in [12] runs several load flows while
the proposed one applies a simple equation. On the other hand,
the advantage of the method proposed in [12] is that typically
fewer meters are needed.
The fault-location method from [13] uses voltage phasors
from sparse measurements along the feeder to locate the fault
in distribution systems. The fault-location process explores a
short-circuit program, a technique for fault resistance estimation and another for load estimation. According to the authors,
the method can precisely indicate the fault location and does
not suffer from multiple fault-location estimation imprecision.
Since this method needs a stage for the estimation of fault resistance before its application, it is expected that the computational
effort be higher and that the fault resistance values have more
influence on the results than the proposed method. In addition,
the method presented in [13] demands the usage of global positioning system (GPS) information, which is not typically available in distribution systems.
Finally, the method from [14] uses a formulation similar to
the method proposed here; however, in [14], estimating the fault
current by measuring all of the phasors of the currents injected in
the system (substation and distributed generation, for example)
is proposed and the total sum of the measured currents is assumed to be the fault current. After the estimation of the fault
current value, the voltage deviation measured is compared to the
calculated voltage deviation. The disadvantage of this method is
that it needs synchronized measurements of voltage and current
phasors at every distributed generator and substation, increasing
the costs and complexities.
VIII. CONCLUSION
In this paper, the voltage measurement capability of feeder
meters and the system bus impedance matrix were used to develop a robust fault-location method. The main findings of this
research are summarized as follows.
If feeder meters with voltage sag measurement capability
are available, the proposed fault-location method can be
easily applied due to its simplicity, increasing the aggregated value of such devices.
The results have shown that the method is robust since
it presented good performance when several parameters
were varied, even for different fault resistance values.
Automated outage mapping can improve the method performance by reducing the search space and the possibility
of multiple estimation.
The sensitivity studies showed that the method performance was satisfactory even with a reduced number of
meters. This is an important finding since it may not be
economically feasible to install many feeder meters in the
distribution system.
Some topics related to the proposed method require further
investigation, such as the possibility of optimal placement of the
feeder meters dedicated to the proposed fault-location method,
and the possibility of considering distributed generators, based
on their representation in the bus impedance matrix.
APPENDIX A
Different from traditional short-circuit analysis, the proposed
fault-location method considers the loads in the construction of
the bus impedance matrix. If the load is ignored in the proposed
fault-location method, the fault-location index , calculated
for many different buses, could assume the same value. Consequently, it would be challenging to identify the faulted bus.
One illustrative example of this multiple estimation problem is
shown, based on a simple single-phase distribution system presented in Fig. 5.
The bus impedance matrix
, without considering the load
impedances of the circuit shown in Fig. 5 is given by the equation shown at the bottom of the page.
Considering the illustrative case with two feeder meters: one
at bus 3 and another at bus 5, and a fault at bus 1, the indices
and
are the same if the loads impedances are neglected,
as shown. Therefore, the algorithm would not be able to distinguish the actual bus under fault.
Following the proposed algorithm, the bus associated with
the minimum
is selected as the faulted bus. And
is given
by the total sum of
[as shown in (3)], where is associated
with the feeder meters and with the candidate bus. The indices
calculated using voltage measurements from the feeder meters
at buses 3 and 5 for a short circuit at buses 1 and 2 (not simultaneously) are given by the following equations of and .
TABLE XIII
TEST SYSTEM DATA
259
TABLE XIII
(Continued.)
The indices
and
have the same value, consequently
yielding multiple identification. However, in the proposed faultlocation method, the inclusion of loads in the bus impedance
matrix turns possible to distinguish between both buses, as the
elements in the matrix become different from each other. However, it is important to mention that accurate information about
the load is not necessary as discussed in Section IV-A.
APPENDIX B
The different conductor types of the system in Fig. 3 were
represented by different colors. The impedance associated
with each conductor type is presented and
is the
impedance of the external Thevenin equivalent connected to
bus 1, as shown in the equation at the bottom of the next page.
The test system data are presented in Table XIII, where and
represent the initial and final bus of each branch, respectively,
260
and the loads are always connected to the final bus and the loads
power factor is 0.92.
REFERENCES
[1] V. C. Gngr, D. Sahin, T. Kocak, S. Ergt, C. Buccella, C. Cecati, and
G. P. Hancke, Smart grid technologies: Communication technologies
and standards, IEEE Trans. Ind. Inf., vol. 7, no. 4, pp. 529539, Nov.
2011.
[2] M. Popa, Data collecting from smart meters in an advanced metering
Infrastructure, in Proc. 15th IEEE Int. Conf. Intell. Eng. Syst., 2011,
pp. 137142.
[3] W. Luan, Low cost feeder monitoring solution in support of utility operations, in Proc. CIGRE Conf. Power Syst., Vancouver, BC, Canada,
2010.
[4] M. Kezunovic, Smart fault location for smart grids, IEEE Trans.
Smart Grid, vol. 2, no. 1, pp. 1122, Mar. 2011.
[5] P. Deepal Rodrigo, A. Pahwa, and J. E. Boyer, Location of outages
in distribution systems based on statistical hypotheses testing, IEEE
Trans. Power Del., vol. 11, no. 1, pp. 546551, Jan. 1996.
[6] S. T. Mak, A synergistic approach to using AMR and intelligent electronic devices to determine outages in a distribution network, in Proc.
Power Systems Conf.e: Advanced Meter., Protect., Control, Commun.,
Distrib. Resources, Clemson, SC, 2006, pp. 447453.
[7] A. K. Pradhan and A. Routray, Applying distance relay for voltage sag
source detection, IEEE Trans. Power Del., vol. 20, no. 1, pp. 529531,
Jan. 2005.
[8] T. Tayjasanant, C. Li, and W. Xu, A resistance sign-based method for
voltage sag source detection, IEEE Trans. Power Del., vol. 20, no. 4,
pp. 25442551, Oct. 2005.
[9] A. C. Parsons, W. M. Grady, E. J. Powers, and J. C. Soward, A
direction finder for power quality disturbances based upon disturbance
power and energy, IEEE Trans. Power Del., vol. 15, no. 3, pp.
10811086, Jul. 2000.
[10] B. Wang, W. Xu, and Z. Pan, Voltage sag state estimation for power
distribution systems, IEEE Trans. Power Syst., vol. 20, no. 2, pp.
806812, May 2005.
[11] F. Han, X. Yu, M. Al-Dabbagh, and Y. Wang, Locating phase-toground short-circuit faults on radial distribution lines, IEEE Trans.
Ind. Electron., vol. 54, no. 3, pp. 15811589, Jun. 2007.
[12] R. A. F. Pereira, L. G. W. da Silva, M. Kezunovic, and J. R. S.
Mantovani, Improved fault location on distribution feeders based on
matching during-fault voltage sags, IEEE Trans. Power Del., vol. 24,
no. 2, pp. 852862, Apr. 2009.
in black
in red
[13] S. Lotfifard, M. Kezunovic, and M. J. Mousavi, Voltage sag data utilization for distribution fault location, IEEE Trans. Power Del., vol.
26, no. 2, pp. 12391246, Apr. 2011.
[14] S. M. Brahma, Fault location in power distribution system with penetration of distributed generation, IEEE Trans. Power Del., vol. 26, no.
3, pp. 15451553, Jul. 2011.
[15] T. Solvang, L. Aleixo, and H. Seljeseth, Power quality measurement
capabilities of Smart Energy Meters, presented at the Int. Conf.
Renew. Energies Power Qual., Granada, Spain, Mar. 2010.
[16] H. E. Brown, Solution of Large Networks by Matrix Methods. New
York: Wiley, 1975.
[17] R. A. F. Pereira, Localizao de faltas em alimentadores
de distribuio de energia eltrica usando medies esparsas
de tenses (in Portuguese) Ph. D. dissertation, Univ. Estadual Paulista, So Paulo, Brazil, 2007. [Online]. Available:
http://www.dee.feis.unesp.br/pos/teses/arquivos/031-tese_rodrigo_aparecido_fernandes_pereira.pdf
[18] Alternative Transients Program (ATP). Portland, OR: Rule Book,
1992, Canadian/American EMTP users Group.
Fernanda C. L. Trindade (S09) received the M.Sc. degree in electrical engineering from the University of Campinas, Campinas, Brazil, in 2009, where she
is currently pursuing the Ph.D. degree in electrical engineering.
Her research interests are power system automation, monitoring, and
protection.
Walmir Freitas (M02) received the Ph.D. degree in electrical engineering from
the University of Campinas, Campinas, Brazil, in 2001.
Currently, he is an Associate Professor with the University of Campinas,
Campinas, Brazil. His research interests are the analysis of distribution systems,
distributed generation, and power quality.
Jos C. M. Vieira (S98M07) received the M.Sc. and Ph.D. degrees in electrical engineering from the University of Campinas, Campinas, Brazil, in 1999
and 2006, respectively.
Currently, he is an Assistant Professor at the So Carlos School of Engineering, So Carlos, Brazil. His research interests are distributed generation,
power system control, dynamics, and protection.