Professional Documents
Culture Documents
2010a Novel Faul Locator
2010a Novel Faul Locator
1, JANUARY 2010 35
Abstract—This paper presents a novel fault-locator system for plemented in practice, for estimating the location of distribution
radial distribution feeders based on the feeder architecture and faults, consist of using voltages and currents measured at one
voltage and current measurements at a subtransmission substa- or some points of a feeder. Fault-location methods can be di-
tion (63/20 kV). The proposed algorithm determines the fault lo-
cation precisely by considering the variable characteristic of the vided into three categories: 1) methods that are based on trav-
load, unbalanced condition, and unsymmetrical nature of distri- eling waves [2], [3], [21]; 2) methods that use higher frequency
bution feeders. Load variations affect the accuracy of fault-locator components (harmonics) [4] of currents and voltages; and 3)
systems significantly. Real-time load estimation is implemented to methods that use the fundamental frequency voltages and cur-
reduce the differences between the calculated fault location and rents [5] measured at the terminals of a line. The last method,
the actual fault location. In addition, self-supervisory characteris-
tics are added to the system to increase the accuracy of the collected
also classified as the impedance-based method, consists of cal-
data. Actual test results demonstrate the effectiveness of the system culating line impedances as seen from the line terminals and es-
in practical applications. timating distances of the faults. Impedance-based methods are
Index Terms—Calculus of variations, distribution feeder, fault
more popular among utilities, because of their ease of imple-
diagnosis, fault locator. mentation. The methods in this category can be further clas-
sified into two subcategories: methods that use measurements
from one terminal of the transmission line and methods that use
I. INTRODUCTION measurements taken from both terminals.
The apparent impedance that is calculated by using a funda-
2) This novel fault locator algorithm is derived from the cal- Fig. 2. Prefault equivalent circuit.
culus of variations and is implemented in load-flow anal-
ysis for the first time.
3) Since constant impedance model is used for load modeling,
the variation of the load in the actual condition needs to
be calculated and exerted in the during-fault calculations.
Load variations will be determined based on the prefault
data and the configuration of the distribution feeder. Fi-
nally, the real-time impedances are applied in the fault-lo-
cator algorithm.
4) The output of “Trip Start” contact is employed as an aux-
iliary means to check the accuracy of fault discrimination.
The rest of this paper is organized as follows.
Theoretical and mathematical principles are presented in Sec-
tion II where the proposed method is thoroughly explained. In
Sections II and III, the proposed fault diagnosis, fault classifica-
tion, and fault-locator algorithms are explained and clarified by
using simple examples and illustrations. The required hardware
configuration for collecting data from the distribution feeder is
described in Section IV. The software analysis including Elec-
Fig. 3. One section of a hypothetical distribution feeder.
tromagnetic Transient Program (EMTP) simulation results and
sensitivity analysis is depicted and compared in Section V. The
result of field tests is presented in Section VI. Both simulation In addition, the impedances used in the during-fault equiv-
and practical results were interpreted as well as compared with alent circuit of each section of the distribution feeder are cal-
the results of other studies in Section VII. culated based on the prefault condition. Therefore, the prefault
load variation must be exerted in the postfault equivalent circuit.
The equivalent circuit for section is illustrated for the pre-
II. FAULT-LOCATION PRINCIPLES fault condition in Fig. 2. Thevenin equivalent impedances of the
rest of the feeder toward the end of the feeder are measured from
Consider the feeder shown in Fig. 1 as an example of a radial point of the section .
distribution feeder. In this feeder, sections and buses are speci- In Fig. 3, the ideal three-phase voltage sources represent the
fied with unique numbers starting from 1 to for the sections input voltage at point . The voltage values are extracted from
and from 1 to for the buses. the prefault load flow information. The equivalent Y model of
the load is shown for each section in Fig. 2 where is the
load variation.
A. Prefault Load Estimation The objective of the load estimation is to determine the
amount of changes in impedances ( , , and )
Loads vary unpredictably in distribution feeders. These vari- based on the prefault values of voltages and currents in this
ations affect the performance of fault locator systems signifi- section.
cantly. A fault-locator system is a real-time device and needs KVL equations at the source side can be written in the matrix
to measure the load variation somehow. In this study, the con- form
stant impedance model is used. The amount of load impedance
varies in each phase during the normal operation of the distribu-
tion feeder before the fault takes place. The variations should be
incorporated in the during-fault equivalent circuit because it is
assumed that the load remains unchanged during the short fault (1)
interval.
MIRZAI AND AFZALIAN: NOVEL FAULT LOCATOR SYSTEM 37
(2)
Vector can be easily worked out by
solving (1) and (2) simultaneously. Finally, during-fault equiv- (4)
alent Impedance is achieved and will be used in the
during-fault calculation
KVL equations for loop 1 in Fig. 3 can be written as
Loop 1
(3)
Fig. 6. Waveform patterns associated with different faults (a) A conductor was broken and immediately eliminated by CB. (b) Fuse. (c) CB. (d) Coordinated fuse
and Autorecloser operation waveform.
(9)
(17)
Let us define
(10)
or
(11)
Then, we can write
(18)
(12)
(13) (19)
(14)
The fault location (D) can be easily calculated based on (9). Equations (18) and (19) are used in the fault-locator algo-
Thereafter, R is achieved easily. rithm. As shown in (18) and (19), D and R are strongly de-
To do this, (9) is rewritten in the complex form pendent on and . Considering the fact that the “prefault”
voltage is approximately constant, the “prefault” current varia-
(15) tion determines the “prefault” load variation in the distribution
feeder.
Therefore In order to use (4), the variation must be small as shown
in Fig. 4 [13].
As a result, the terms and
(16)
are neglected in (1).
MIRZAI AND AFZALIAN: NOVEL FAULT LOCATOR SYSTEM 39
(20)
(21)
On the other hand, for the faults to take place in the beginning V : faulted line voltage;
sections of the feeder, (22) and (23) are used but for the other I : prefault current;
sections, (18) and (19) are strongly recommended [7] V : zero-sequence voltage;
I : faulted line current;
V : unfaulted line voltage;
(22) I : zero-sequence current;
V : nominal phase voltage.
(23)
[14], which is classified under “pattern recognition.” In order to
The flowchart of the algorithm for finding the fault location is apply the method for the fault diagnosis purpose, four main pat-
shown in Fig. 5. Due to the radial structure of distribution sys- terns are defined and called basic pattern. A fault current wave-
tems, the fault-location algorithm can obtain multiple choices form in the distribution feeder is then compared with each basic
for the fault locations when measurements are only available at pattern. The patterns are defined over three or four intervals (i.e.,
the substation. Thus, it is necessary to identify the actual fault prefault, during-fault, and postfaults time intervals. These wave-
location from the calculated multiple choices by using an addi- form patterns are shown in Fig. 6. Distribution feeders are al-
tional fault diagnosis algorithm. ways liable to be affected by dangerous situations of various
types of faults. There are varieties of protective devices, such
C. Underground Cable as cutout fuses, autoreclosers, and protective relays. Protective
relays (OC/EF) are installed in the substation in order to protect
The algorithm and the system are designated to find the fault the main branch of the feeder while cutout fuses and autore-
location in underground and overhead distribution feeders. The closers are mainly erected in the lateral branches. In the pro-
underground fault-location estimation formula is slightly dif- tection expert point of view, faults are classified into three main
ferent than the overhead line because the parallel capacitors in groups: 1) low impedance fault along the main branch; 2) broken
the section model (Fig. 4) cannot be neglected and the amount conductor faults; and 3) the operation of the protective devices
of this capacitor is significant. in the lateral branches [18].
In the case of underground cables, (10) will be substituted by The overcurrent relay operates immediately when low
impedance faults occur along the main branch. Consequently,
(24) C.B. operates to prevent further damage. The waveform of this
class of fault is illustrated in Fig. 6(c). The class of fault (low
where impedance) can be identified by using Table I [15].
Sometimes, a severe accident causes a breakdown in con-
(25) ductors of overhead lines. If broken conductors fell on other
live conductors, the fault can be easily detected because of
and the current increase in the affected phases. However, the main
branch current differs slightly when the conductors fall on
(26) the ground. This type of fault cannot be detected easily if the
earthing transformer connected the secondary side of power
transformer (63/20) to the ground. Indeed, the fault detection in
this type of fault strongly depended on the network grounding
III. FAULT DIAGNOSIS ALGORITHM
method.
The voltage and current signals are being recorded in the sub- The waveform changes when the overhead line conductors
station when a fault occurs along the main or lateral branches of are broken and falling on the ground, the second interval is di-
a distribution feeder. These recorded signals contain important vided into two parts (i.e., “open-line” and “during-fault.” On the
information which can be used for diagnosing the fault. other hand, in addition to the three fundamental parts of the fault
The first piece of evidence which can be applied to the fault current waveform, “line-open” interval is added to the sequen-
diagnosis process, involves the sequence of events extracted tial events of the waveform. The current pattern of this kind of
from the waveforms. The method is called “waveform pattern” fault is illustrated in Fig. 6(a)
40 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 1, JANUARY 2010
Fig. 8. Fault diagnosis in the scheme that the autorecloser and fuse were coor-
dinated.
(27)
In branch II
(28)
Fig. 7. Fault-locator algorithm. In branch III
(29)
The fault-locator algorithm suggests multiple choices for the
fault location in lateral branches (Fig. 7). As an example of how Only is the right answer (Fig. 8).
MIRZAI AND AFZALIAN: NOVEL FAULT LOCATOR SYSTEM 41
as load, the place of fault, and the resistance of fault are changed
intentionally and the error of this variation is measured. Both of
these results are very important.
Simulation signals are prepared for the test in the EMTP soft-
ware [20]. A set of generated voltages and currents signals is
shown in Fig. 14 and 15, respectively. The fault resistance in
this case is .
the proposed system is less than or equal to the one in [7] in five That is because; two different algorithms were applied for the
cases. There is only one case where our result is worse than [7]. different part of the distribution in this study.
44 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 1, JANUARY 2010
Fig. 14. Generated fault voltage waveform in EMTP for the fault in phase C Fig. 16. Error in D versus load variation in three different fault resistances
R = 2
. ( )
R along the 18th section.
Fig. 17. Error in D versus load variation in three R along the 12th section.
Fig. 15. Generated fault current waveform in the EMTP for a fault in phase C
R = 2
.
TABLE II
ALGORITHM ACCURACY TEST RESULT
Fig. 18. Error in D versus load variation in three R along the 25th section.
TABLE III
SUMMARY OF THE TEST RESULT