Professional Documents
Culture Documents
Sponsored by
POWER ENGINEERING EDUCATION COMMITTEE
in cooperation with
POWER SYSTEM RELAYING COMMITTEE
TRANSMISSION AND DISTRIBUTION COMMITTEE
SUBSTATIONS COMMITTEE
of the
IEEE POWER ENGINE,E,RING SOCIETY
Available from
Publication Sales Department
IEEE Service Center
445 Hoes [-ane
P.O. Box 1331
Piscataway, NJ 08855-1331
Copyright O 1989 by The Institute of Electrical and Electronics Engineers, Inc. For copying, reprint, or republication
permission, write to Staff Director, Publishing Services, IEEE, 345 East 47th Street, New York, NY 10017-2394 USA
Foreword
Utilitie.s strive to provide electric service to the puAHc in a safe and reliable manner. However, the very
nature of distribution systems means that accidents will occur resulting in downed electrical conductors. Th!
detection of these downed conductors is a vexing problem for which a solution has long been sought.
_. P_rotection enginee!\, using various methods, successfully detect the vast majority of faults on electric
distribution systems. However, a certain percentage of these faults, typically refeired to as high impedance
faults, cannot be detected by conventional detection devices. For many years, specifically thE last'decade,
researchers have sought improved protection methods in an attempt to detect a trlgtrer percentage of the
downed conductors, thereby reducing t -q4r to the public. These investigations hive mor-e clearly
.
documented the nature and characteristics ofhigh impedance, low current faults and have resulted in several
theoretical methods for improved detection.
The- prrrpose of this tutorial is to inform and educate engineers and utility leaders concerning the current
state-of-the-art of research in downed conductor detection. The results of-many years of inveitigation are
documented_and,presented including attempts by utilities, manufacturers, and university reseirchers to
describe fault behavior and discover new detection approaches. Additionally, non-electrical means and
improved operating methods for mitigating the problem of downed conductors are discussed.
All faults will never be detected. Yet, it is the hope of the authors of this document that the information
presented.will encourage indepth investigations and discussion among technical experts resulting in further
advances in the detection of downed conductors.
A document of this maglritude is not prepared without considerable sacrifice of time by dedicated
volunteers. The following individuals have selflessly given their time and experience to organize, write, edit,
and.present this material.- Their.only reward,is the knowledge that the utility industryls weil served by
solving this problem and the public is made safer if more downed conductors can be deiected. The efforti
of the following individuals have greatly contributed to these goals.
Technical Committee
Contributors: High Impedance Fault Detection Working Group
Power System Relaying Committee
CHAPTER PAGE
FOREWORD 2
REFERENCES .. 39
The subject of this tutorial is the detection of downed From the above example, it should be obvious that even
conductors on distribution systems. More specifically, the the most intelligent use of overcurrent protection devices by
discussion will corcentrate on detection of downed conductors electric utilitiei will not solve the high impedance fault
which do not make continued contact with other power problem. It will also be shown that some high impedance
conductors or which draw very low fault current. iaults draw virtually no current and there is currently no
commercial or even anticipated research method which will
The distribution system primarily considered here is the detect these faults. In short, some faults will always remain
four wire, multigrounded system serving distribution load undetected. However, the last ten years has brought much
directly. The distribution circuit main lines include three research into the characteristics and nature of high impedance
phase conductors and a neutral conductor which is grounded faults. From this research, new detection approaches have
iegularly along its entire length and is connected to the been theorized and are currently under study.
distribution substation transformer neutral. On these systems,
a significant portion of the load is connected from phase to The use of non-fundamental current signals such as the
neutral and is fed from single-phase laterals which connect to higher frequencies generated by arcing faultsand other unique
the main line. characteristics of these faulted conductors offers potential for
new protective relays in the future. While none of these
Detection of short circuits on distribution systems is techniques have been developed sufficiently for safe use by
typically done by relays, fuses, and other devices which electriC utilities, research is ongoing and researchers are ever
respond to excess current flow. These "overcurrent" hopeful for breakthrough solutions to this vexing problem.
techniques are virtually the only methods available for
protection of existing power systems. Their characteristics will Downed conductors on distribution systems are the subject
be discussed in detail in Chapter 3. of the tutorial primarily because of public safety. People
become involved with such faults through vehicular accidents
The longstanding problem for electric utilities is that short with power poles and through incidental contact or proximity
circuits on distribution systems not involving normal power in a wide variety of situations. This tutorial is not being
carrying conductors frequently cannot be reliably detected. presented to broadcast a solution to the problem; however, it
These short circuits or faults are characterized by high is motivated to facilitate the education of engineers into the
impedance at the point of fault. This impedance is often technical aspects of the problem, to focus on ongoing fault
sufficient to restrict current flow to values which correspond detection research, and to encourage discussion of ways in
to "normal load" rather than to fault levels. which incidents involving the public can be minimized.
Figure 1-1 illustrates the problem ofprotecting distribution The Power System Relaying Committee of the IEEE Power
feeders from low current faults using conventional overcurrent Engineering Society has been vitally interested in this subject
protection devices. When viewed from substation phase for many years. Recognizing this interest, the PES Technical
ielays, the feeder is protected for all fault currents whose level Council authorized the preparation of a document intended
exceeds the settings of the phase relays. These relays must be to discuss all aspects of this problem in terms reasonably
set higher than the maximum anticipated loads on the feeder. educated non-technical persons could understand. This
Obviously, faults which draw lower current than the phase document was to educate utility personnel, teachers,
relay settings will not be detected by these devices. This legislators, and the judiciary with the ultimate goal of
leaves an unprotected region for the feeder which is typically improving public safety. "Downed Power Lines: Why They
visited by high impedance faults. Can't Always be Detected," published February 22, 1989,was
a result of this assignment [i]. The conclusions reached in
Maximm this report are as follows:
Anticipated Phxe RelaY
Load Settings
Feeder
Curent "Detection of high impedance faults on overhead
Levels distribution systems continues to be an unsolved problem for
electric utilities. This is true in spite of the continuing efforts
of some of the foremost engineers and scientists in the
country. The root problem is that a high impedance fault
does not cause a detectable change in current flow on a
circuit. The much more common low impedance faults cause
substantial fault current to flow and are easily detectable.
Figure L-1. Feeder protection levels.
Downed conductors are of major concern to electric utilities
While the addition of ground overcurrent devices mitigates because they may result in public involvement in a hazardous
the above problem, many utilities have found that wide situation. Downed conductors may not contact a conductive
variations in system load and extreme system load imbalance object and, therefore, have good probability of remaining
make it impossible to use very sensitive ground protection. energized. When lying on certain surfaces, they may look
Hence, ground relays are typically set at 5070 of phase relay quite harmless. A person touching an energized power line
settings or higher. This still leaves the feeder exposed to the cbnductor faces substantial risk, since no detection device
high impedance, low current faults. known today can react fast enough to prevent injury.
!he^onfy available solurion to this problem today is an alert
and informed public. Electric utilitiet continue toemphasize
this danger in their public affairs campaigns. public ipirited
citizens must call the police when tliey see a dbwned
conductor and wnrn anyone approaching to itay away from the
wire."
O_ne can clearly see the need for wide exposure of this
problem to the public and the continued need for research
toward improved protection by concerned engineers. It is the
intent of this tutorial to summarize and preient the current
understaadings of high impedance, low currlnt faults, including
their nature and characteristics, their physical behavior, ani
their effect on the power system. Tlie-inability of existing
protection systems to detect such faults will be -discussed ii
detail along with the efforts by electric utility engineers to
mitigate the downed conductor problem. pait research on
detection methods will be described including the proposed
use of new time domain and frequency domairisolutibns. I't e
use of mor-e sophisticated, computer ielalng devices and the
potential for algorithms implemented on ihese devices to
detect the faults is discussed.
It is appropriate to begin an analysis of the high impedance The majority of feeders monitored in the study were
fault piobtem with a discussion of experiences and primarily ii ttre tS kV voltage class-. Circuit voltages-ranged
observations offaults on distribution feeders' This discussion irom 4.16 to 34.5 kV. The average feeder was 10.8 miles long
of distribution faults will provide information on the with 93 percent of the circuit length consisting of overhead
characteristics of these faults to help identifu the design goals construciion. The typical feeder served loads characterized
of a fault detection system which is developed. Much of the as indicated in Table 2-1.
information provided in this chapter has been gained from
extensive studies of distribution faults, for high current as well Type of l-oad Percent
as low current faults.
Industrial 13
Urban/commercial 18
CHARACTERISTICS OF HIGH CURRENT FAULTS Rural 12
Suburban 56
High current faults in distribution systems cause protective Other 1
Much has been learned about the nature of high impedance Current levels observed during high impedance faults may
faults in the last fifteen years because of exterisive iesearch range from zero up to the operation level of installed
efforts during this period. Much of the information discussed overcurrent devices. During the Texas A&M fault tests, fault
in this section is the result of numerous staged faults cuffent during most tests tended to remain between about 10-
conducted for research purposes by Texas A&M University. 50 amperes. Waveforms from typical staged fault tests are
The study of the characteristics of downed conductor faulis provided in Appendix C.
has led, to a greater awareness of their significance for utilities,
as well as to identify potential means for detecting these Research has shown that the predominant influences on
faults. fault current for a given location are local conditions such as
soil type, soil moisture, and the presence of serendipitous
ground paths, such as water pipes, metal fences or extensive
Rate of Occurrence tree roots. There is little correlation between available fault
current at the location and actual fault current when the fault
There is substantial disagreement in the utility industry on does not involve a solid contact between the phase conductor
the degreelo_which high impedance faults occur. Many relay and a solid electrical path such as neutral.
engineers believe that only a small percentage of-fallen
SURFACE FAULT CURRENT (A)
Dry asphalt 0
Concrete (non-reinforced) 0
Dry sand 0
Wet sand 15
Dry sod. 20
Dry grass 25
Wet sod 40
Wet grass 50
Concrete (reinforced) 75
CONCLUSIONS
INTRODUCTION
For understanding the problem of high impedance faults, Most load on four-wire distribution circuits consists of
it is essential to have knowledge of the existing protection single-phase distribution transformers, conaected between
one
practices on distribution feeders. This subject is discussed of the phase wires and the neutral conductor (Figure :-Z;.
through a description of typical distribution circuits and their The neutral conductor is connected to earth at the zubstation
protection, as well as a survey of utilities to identify the range and at frequert intervals throughout the length of the circuit.
of actual practices in use at these companies. The review of Becatse it is impossible to exaitly balance iLe load between
present protection practices provides a framework for the three phases, there is usually a continuous flow of residual
discussing the improvements in fault detection which can be current in,the neutral, but this normally presents no hazard,
provided by sensitive high impedance fault detection. because this conductor is maintain.d it o. near ground
potential. AJthough the load unbalance may be quitE Ugn,
and even approach the loading on one o, ,o." ofihe ph;;
TYPICAL DISTRIBUTION CIRCUIT DESCRIPTION conductors, the neutral conductor is frequently sized at ibout
one-half the current-carrying rating ;f tne phase wires,
The majority of electric utility customers in the United because a large amount oi current i,ill return in the
earth.
States receive service from single-phase transformers The size of the neutral conductor should theoretically have
connected to three-phase, four-wire, solidly grounded some effect on the ability of protective devices to detect
a
distribution circuits. These circuits usually have only one downed phase wire; as a praciical matter, however. it does
source and are fed by substation transformers connected as not appear to be a significant factor in most instances.
shown in Figure 3-1. The phase-to-phase voltage on these
circuits generally ranges from 4160 to 13,800 volts, with a
large number operating at 12,470 volts. Other distribution
circuits operate at higher voltages, including particularly
24,900 and 34,500 volts. Also, ungrounded delta circuits still
exist in certain areas, most often at the 2400 volt level, but
this type represents a very small percentage of the total.
TRANSFOBMER
CURRENT
TRANSFORMERS
GURRENT IN
GROUND RELAY
AFTER OCR
OPERATES
103
INTRODUCTION
This chapter focuses on research and development efforts By 1979, several conclusions concerning
on high impedance fault detection initiated by utility
.harmonics were formed:
the behavior of
companies. . M?lV companies have experienced problemi
associated with high impedance faults on their systbms, and
have .closely examined and studied the probiem. Two l. Harmonic currents as high as 5 percent are
investigations are presented here. The first ii research begun common under normal operating conditions,
by Rochester Gas and Electric which was sutrsequeitly particularly during light load periods.
continued by Nordon Research and Development. The iecond
is an investigation originating with Pennsylvania power and 2. The phase angle of the third harmonic current
Light Company which led to a collaborative project with relative to the substation bus voltage on a 1g0
Westinghouse. Hz basis varies over a wide range, but varies
slowly under normal conditions except during
capacitor switching.
RESEARCH BY ROCHESTER GAS AND ELECTRIC
3. Distribution transformers generate differing
amounts of third harmonic current; however, the
Rochester Gas and Electric Corporation has had an interest harmonic angle seems to remain stable.
in the detection of high impedance ground faults for a number
of years. Statistics on the number and types of uncleared 4. Although instrumentation was not available to
ground faults are difficult to obtain; however, informal measure the absolute phase angle shift of the
interviews with line people make it clear that the problem is third harmonic, data indicated that during a high
significant. .The problem may or may not be significant from impedance ground fault, the angle of th; thi;d
an economic standpoint. It is certainly significant from a harmonic shifts rapidly as much as 100 degrees
moral standpoint. Unless it can be proven that it is impossible on a 180 Hz base and the average shift is about
to.detect high impedance faults, the electric industry has an 60 degrees.
obligation to its employees and the public to searth for a
solution.
Study of Downed Conductor Faults To assist in the research on high impedance faults, PP&L
developed a Distribution Fault Analysis Program [17,18] which
PP&L proceeded to gather additional information provided a computer model of a distribution circuit. The
concerning the nature of the problem on their distribution p.og.a- allowed complete modelling of a feeder, and allowed
system. High impedance fault data on the PP&L distribution calculation of steady-state circuit Parameters under a variety
system was collected between April 1974 and December 1975 of faulted and unfaulted conditions. Using a "typical line"
[14]. In total, data on 390 faults was recorded. The 390 cases model, a variety of calculations were performed to gain
of broken conductor faults represented about '1,1,% of 3605 experience with the program, investigate sensitivity to fault
total faults during an equivalent period. Downed conductors impedance and various circuit parameters, as well as to
remained energized in about 113 or 123 out of 390 cases of acquire a feel for characteristic circuit changes during fault
broken conductor faults recorded. These 123 cases conditions. Operation of the model indicated that obvious
represented about 3.4Vo of all faults. changes were present in the negative and zero sequence
voltage beyond the open conductor fault, and negative and
PP&L s overhead distrilrution system, at the time, consisted ,ero iequence current at the protective device responsible for
of approximately 957o bare conductor and 57o XLP covered detecting the open conductor fault.
conductor. Bare conductors were involved in 83.57o and XLP
conductors in1.6.57o of the 390 high impedance faults. Of the
uncleared faults, bare conductor accounted for 62Vo and XLP Collaborative Research with Westinghouse
covered conductor for 38Vo. This documented an existing
problem with bare as well as XLP covered conductors. There ln 1977 PP&L began a joint study with the Advanced
was no apparent correlation between the ability of a protective Systems Technology Division of Westinghouse Electric
device to operate and factors such as contacted surface, Corporation (AST) to assess various relay schemes for the
proximity of pole grounds, size and type of conductor, and detection of high impedance faults and in 1979 additional
available fault current. Fault current was proven to be an work was started to investigate the mechanism of XLP
unreliable fault characteristic to detect high impedance faults' covered conductor burndown [19,20,211.
More than 50% of the cases of covered conductor Based on the results of the 1974-75 data on 390 cases of
burndown were caused by lightning, with tree contact the fallen conductors, it was apparent that the problem with
second leading cause. Ughtning and tree contact accounted clearing high impedance faults was concentrated on the three
for over 75Vo of. all burndown cases. No single cause in the phase portion of the feeder. Efforts were focused on
remaining 2570 was accountable for more than 5Vo. developing techniques to detect faults in this area.
Several criteria were identified for use in evaluating various
Results of Relay Evaluation
technologies for high impedance fault detection. It was
identified that a high impedance fault detection system should
The conclusions of the overall evaluation were:
be: sensitive, discriminative, reliable, secure, economical,
rugged, relatively simple, easy to apply, capable of fast .
installation, and easily maintained. Phase overcurrent devices are ineffective in
detecting high impedance faults.
Six relay schemes and ground overcurrent relays were . Solidly grounded phase conductors, broken or
evaluated against phase overcurrent relays for their ability to
detect high impedance faults. Each scheme was evaluated, by
not, produce detectable voltage unbalance
downstream of the fault location.
computer simulation in terms of its technical performance
under simulated fault conditions and its estimated installed .
cost. Comparison with PP&L's existing protective scheme was
Ungrounded broken conductors produce
detectable voltage unbalance downstream of the
made to estimate the cost effectiveness of the proposed relay
schemes.
fault location.
. Zero and Negative Sequence Overcurrent Relays A Ratio Ground Relay design was initiated as an
electromechanical device. The selected scheme had to be
- a high impedance fault may cause heavy economical, permit construction in a standard relay case size,
current unbalance which could be detected by
present low burden to existing substation current trinsformeri
zero ot negative sequence overcurrent relays
located at the protective device.
and be simple to. apply, test and set. As with any relay, it
must be dependable, secure, selective and sensitivl. In'ihe
Evaluation of Downstream Voltage Techniques
PP&L .application of the concept, high speed is not a
constraint. provided reliable operation takis place with proper
coordination with existing overcurrent deviies.
A relay evaluation program evaluated the behavior of the
six proposed relay schemes and ground overcurrent relays.
Studies were conducted on four possible Ratio Ground
Negative and zero sequence overvoltage sensing at the ends
Relay designs. Each scheme used the induction disc concept
of the three phase line sections and branches would be the
with.operating and restraint windings. Contact closing torque
most effective method to detect high impedance faults. in all schemes,.is produced by residual current (3t) in ihe
However, the end-of-the-line sequence overvoltage schemes
operating.rvinding. Contact opening torque is produled by a
have several disadvantages:
combination of phase, positive aia/or' negaiiue sequence
. Multiple sensing locations are required, each
current in the restraint winding. The designi were modelled
with digital techniques and prototypes weie constructed for
with a communications system.
testing on the Westinghouse ANACbM III analog compurer
. Receiver equipment will be necessary at the
which modelled one of PP&Us circuits.
protective device.
Specification of tap settings for the CGR require an
. The total reliability of the scheme may be lower
examination of normal load unbalance. Tripping thL circuit
than required. on an unbalance because of a blown single phasJ tap fuse is
undesirable. Therefore, the unbalance setiings allowed normal
. The ends of three phase line sections or load unbalance with the tap fuse blown thatireated the worst
circuit unbalance.
branches are dynamic positions that may require
regular relocation of the relays and related
From these studies one CGR scheme was chosen to fit the
equipment.
various requ.irements. A restraint element responsive to
ll, l,
- I,I, l'. cotrp.led with a l3lolr operating elemint on a singl'e
. Total cost per circuit is presently high.
induction disc was chosen for prototype construction. fle
prototype CGR was an electromechanical induction disc unit The covering increased the fault impedance for a contact with
with operating and restraint elements to be installed in grass substantialiy, yielding a value of about 270 ohms versus
existing distribution substation switchgear and driven by the 80 ohms for bare conductor.
existing phase current transformers andwill not require any
additional substation equipment. The prototype was
constructed and installed in a standard directional overcurrent Results of Relav Tests
relay case. Based on tests on the Westinghouse ANACOM III
computer with simulated faults and circuit operation, the relay Observations were noted on the operation of the relay
operation substantiated the previous analytical work. methods during the above fault tests. The results of these
relay tests are as follows:
Tests of Relays with Staged Faults . The Ratio Ground Relay detected 9 out of 19
single phase-to-ground faults. It was unable to
In 1981, staged fault tests were conducted to substantiate detect the phase-to-tree fault with either bare or
the analytical work that had been completed through exposure XLP covered conductors. It did detect all of the
of the several relay schemes to high impedance fault open conductor faults.
conditions on an operating 12 kV circuit [22]. The test
configurations included: . The conventional Ground Overcurrent Relav did
not detect any of the phase-to-ground faulti. It
. Open conductor tests detected all but one of the open conductor
faults.
. Single phase-to-ground faults using bare and
XLP covered conductors in contacfwith grass, . The downstream Negative Sequence Ovewoltage
gravel, asphalt, concrete and an evergreen tree Relay was unable to detect any of the faults
because of voltage support equipment installed
. Open conductor tests with the source or load to support customer voltage during the tests.
side conductor on the ground
. The downstream Zero Sequence Overvoltage
Relay detected 13 out ofthe 15 open conductor
The test data is shown in Table 4-1. faults, even though the voltage support
equipment substantially reduced the zero
sequence voltage on the load side of the open
No. of Avg. Fault phase.
Conductor and Surface Cases Current (A)
. No voltage activated relay detected any single
Covered to grass 4 26 phase-to-ground fault.
Covered to gravel 3 55
Covered to asphalt J 0
Covered to concrete 3 61 Fault current and voltage data recorded during the tests
Covered to tree 1 0 compared favorably with output from the Distribution Fault
Bare to grass 4 86 Analysis Program and ANACOM III studies. The operation
Bare to gravel 3 87 of the several relay schemes performed as expected within the
Bare to asphalt 4 0 restrictive requirements of the test scenario and further
Bare to concrete 3 97 substantiated earlier analvses.
Bare to tree 1 1
ffi
\ ir/ 1.5 Amp.ro Srtllng
o
o
o.
Advantages of the Ratio Ground Relav ol
o
Figure 4-1 illustrates the typical load variations experienced
(?
durirlg the lifetime of a feeder- Because of these wide Ol 234
variatlons at different times in the feeder's life, a conventional
3l.,-only relay must be set high enough to ignore maximum
la (Load Currsnt, Amperes)
n<i.ma[ excursions. Thus, for the rest of the time, a
substantial peritld, the feeder is underprotected. (a)
PEAK ULTIMATE
3 (FULLY MATUBE)
o 100
x= llt,
-o
e5 75
o,
lr
o 5eo n3'u"
ut
o Eo
E@ 50
F
z
LIFETIM€ AVEFAOE Frr
ul E'O
o 6f 25
tr
IJJ OFF.PEAX J@
Qu
o. sEAgoil
Etc o
123
ao la (Load Current, AmPeres)
M 6
nME oF DAY
(b)
22
switching; others were caused by blown capacitor bank fuses.
10 One utility experienced problems in maintaining balance on
^.=
L
a mixed residential/industrial feeder on weekends, although
f the feeder was balanced on weekdays.
o
gE Due to an unacceptatrle number of false operations, many
of the utilities have completely removed their CGRS relays.
!, The City of l^akeland has placed all of their CGRS's in an
o
o alarm-only mode; several other utilities are also continuing
o their evaluations of the CGRS in an alarm-only mode.
6o
o Based on the experience gained by trial utilities, it can be
o
(, said that use of this relay, even as a monitor, requires: a)
c,
6 frequent balancing of phase currents, b) aggressive tree
fr4
o
trimming schedules, and c) a very structured and effective
response by operators when a trip or alarm occurs. Since
o these practices have other benefits to a utility, however, their
t implementation should not be charged entirely against down-
=H2 31o Relay wire protection. It should also be noted that this device is
l.l. most effective when used as part of a composite protection
x(0 plan.
=
12 3
!L (Load Current, Amperes)
(a)
86
o
CIE
E
q,
lnprov.m.il Dy CGRS
!,
c3
o
o
o
o2
o
2
car tre
12
l1 (Load Current, Amperea)
(b)
.
The problem of maintaining phase balance caused many
misoperations of the CGRS relays, both at the City oi
I-akeland and at other installations. These unbalances had a
variety of causes. Some were caused by single phase
23
CHAPTER 5
THIRD HARMONIC MAGNITUDE AND ._ Under EPRI Project 1285-1, Power Technologies, Inc.
PHASE ANGLE MONITORS (PTI) examined the behavior of low order odd iarmonic
current [27]. Based on their examination of the behavior of
Hughes Aircraft Company investigated the use of third these harmonics, both during staged downed conductor faults
harmonic phase current. as a possible indicator of high and during normal system operating conditions, pTI suggested
impedance faults. This work was done under Epi.I detecting broken conductor faults by monitoring the
sponsorship on Project 1285-2 [26). non-characteristic sequence components of the firsi (i.e.,
fundamental), third and fifth current harmonics.
After analyzing. data from a series of staged downed
-involving
conductor faults both grounded-wyJ and delta Under balanced conditions, the fundamental component is
connected distribution systems, Hughes' researchers decided a positive sequence quantity, the third harmonic is a zero
that two separate detectors were needed, one for use on sequence quantity, and the fifth harmonic is a negative
grounded-wye systems, and the other for use on sequence quantity. Thus, the PTI researchers decided that the
delta-connected systems. non-characteristic sequence components of the first, third, and
fifth current harmonics, which are zero under perfectly
The detector designed for use on wye-connected systems balanced conditions, would be good indicators of a bioken or
monitored the phase angle relationshipi between each-of the downed conductor. To monitor all of these quantities
three third harmonic phase currents. A sudden shift in one of simultaneously, they decided to combine t[em all
these.three^with respect to the other two indicated a high mathematically to form a single number for use as a relaying
impedaace fault. To lessen the possibility of normal systein quantity. This number, which they called the Z statistii, ii
events incorrectly being identified as faulti, the fault/no-fault calculated as Z = (X - Yf Fl' 6- Y). Here, X is a IZX l
decision also considered the magnitude of the fundamental vector whose components are the real and imaginary parts of
current component. A fault was indicated if the relative phase the non-characteristic sequence components of the firit three
angle between-one phase and the other two phases suddenly odd harmonics, and lP/ is the covariance matrix of the
changed by at least 15 degrees concurrent with an increase in elements in the X vector. Also, Y is the average of many
single.phase load current of at least 15 amperes, and this previous samples of the X vector, and the superscript 7
condition persisted for at least 5 seconds. denotes a matrix transpose operation. Because of the
computational burden of calculating the necessary quantities,
After initial laboratory testing using staged fault data the PTI algorithm sampled one 60 Hz cycle each 30 seconds.
yielded positive results, two of these phase shift detectors were Initial results indicated "spikes" in the Z statistic which would
built, tested, and installed at utilitierfor long-term evaluation. require the threshold level to be set quite high in order to
These detectors worked well for staged faulti, but experienced avoid false trips. Making the threshold this high, however,
false trips at a tate of three per year. There were no naturally meant that the detector would miss many of the lower grade
occurring downed conductor faults during the evaluation faults. To overcome this, since most of these spikes lasted
period. only one or two sample periods, a four sample delay time was
added to the algorithm (i.e., no fault was indicated unless four
Because the ambient level of third harmonic is verv low on successive samples exceeded the Z threshold).
delta-conaected circuits, a phase relationship measurement
bgcomq quite difficult and somewhat meaningless. However, PTI built a prototype detector which implemented
-They its
since this normal level is so low, a substantial increase in iti harmonic sequence component algorithm [28]. used an
magnitude on such delta circuits could be used as an indicator Intel 8086-based host microprocessor for this implementation
of an abnormal condition, e.g., a downed conductor fault. and programmed the data acquisition and processing code in
24
PL/M-86 and PASCAL 86. They increased the number of other normal system events produced similar increases, these
cycles checked from one cycle per 30 seconds to one cycle per increases were always time-limited, Iasting only a few cycles,
one second. indicating that monitoring for an increase in high frequency
activity could be used as an indicator of a downed conductor
Based on the performance of the detector when tested fault, with time being used as a discriminating factor to
against recorded fault and normal system data, and on the prevent false trips on normal system transients.
data available about the frequency of occurrence of various
tlpes of high impedance faults, the PTI researchers estimated Under normal (non-faulted) system conditions, the Energy
that the algorithm would detect approximately 40Vo of allhigh algorithm maintains a system status of "Normal." During suiir
impedance faults. They also noted that the algorithm would times, it calculates and stores an average ambient energy level
detect a blown phase fuse as a high impedance fault, since for the particular frequency of interest, the energlr being
each event has the effect of altering the balance between the calculated as the sum of the squared sample values over one
three phases. Because of the low detection efficienry of the fundamental frequency cycle. As each new energy value is
detector, the technique was not pursued further. calculated, it is compared to the previously stored average. If
it is not more than a predefined percentage greater than this
avetage, the system status remains "Normal," the average is
THIRD HARMONIC PHASOR DETECTOR re-calculated including the effect of this new energy value, and
execution continues with the next cycle's energy. If, however,
Nordon Research and Development recently developed a the cycle's energy is greater than the allowable percentage
downed conductor detector based on the phase angle of the increase over the average, the system's status becomes
third harmonic fault current with respect to phase voltage "Disturbance," and the next few cycles are checked for similar
129,301. increases in energy. Ifthe check of these next cycles indicates
that the energy increase was very short-lived (i.e., only one or
Nordon's development is based on the premise that the two cycles contained increased high frequency activity), the
current drawn by a downed conductor fault produces third system's status is returned to "Normal" and execution
harmonic current with a unique phase relationship to the continues as before. If the check shows that the increase was
faulted phase voltage. To find the third harmonic fault not so short-lived, however, the system status becomes "Event"
current only, without a contribution from the normal system and the energy values are checked for the next few seconds.
third harmonic current, an average third harmonic phasor is Again, if only a few (or none) of the energy values in these
calculated and stored; then, when each new third harmonic next few seconds show increased energy, the system status
phasor is calculated, the stored average is subtracted returns to "Normal" and monitoring continues. If, however, a
(vectorially) from it. The angle ofthis resulting phasor is then large number of these cycles show high energy levels, the
checked. If it matches the direction defined for a downed system status is changed to "Fault."
conductor fault and the mqgnitude of the phasor is greater
than some threshold level, alfault is indicated. This threshold As previously mentioned, the Energy algorithm was initially
is set so that gradual norm4l system harmonic level changes applied to high frequency current. To test the algorithm,
will not be incorrectly identified as faults. several prototype high frequency Energy detectors were built
and installed for long-term evaluation at various cooperating
While normal system changes (e.g., capacitor bank utilities (grounded-wye circuits) 132,331. These detectori
open/close operations, voltage regulator tap changes, etc.) performed well for staged fault tests. No false trips occurred
may cause substantial sudden changes in the third harmonic during the evaluation installations; there were no
current, Nordon asserts that these changes will produce naturally-occurring downed conductor faults during this period
different phasor changes than will a downed conductor fault. either.
Because of this, the two types of changes will be
distinguishable. The Energy algorithm was later applied to odd harmonics
(180, 300, 42A, ... Hz), even harmonics (120, 240, 360, ... Hr),
A patent has only recgntly been issued on Nordon's and non-harmonics (30, 90, 150, ... Hz) in the frequency range
development [31]. As of puhlication of this tutorial, a detailed below 1200 Hz. The Energy algorithm's performance when
description of the technique,was not available. tested with recorded staged fault data was good, especialty
when applied to even harmonics and non-harmonics.
Problems can be seen if the Energy algorithm were to be
. ENERGY ALGORITHM applied to the odd harmonics in the field, since the level of
these odd harmonics are generally higher and less predictable
In the late 1970's, Texas A&M University researchers than the levels of the other components. While the high
began an investigation of the high impedance downed frequency current component provides a very good indicator
conductor fault problem. Under EPRI Project 1285-3, they of a high impedance fault in many cases, the presence of an
developed the microcomputer based Enerry algorithm, which energized grounded capacitor bank sometimes shunts the high
detects a sudden increase in the level of a given frequency, or frequency signal to ground, making the use of the lower
frequency band, of current [3]. The Energy algorithm was frequency components necessary.
originally applied to high frequency (2 - 70kHz) current; later,
its applicability to "low" frequency harmonic and non-harmonic
current components was also investigated. RANDOMNESS ALGORITHM
Based on observations made during staged fault tests, In most downed conductor faults, arc current is drawn
capacitor bank operations, voltage regulator tap changes, etc., intermittently. Most faults arc for a short period of time
the A&M researchers noticed that arcing downed conductor (f.9-.-u few 60 Hz cycles to a few rens of ryclds), followed by
faults produced a marked increase in the level of high a similar period in which no current is drawn. Further, even
frequency (2 - 10 kHz) current. They also noticed that while during periods of arcing, there are often large variations from
25
one cycle to the next. This phenomenon hinders the extracting the fault current from the total line current' Their
performance of many of the proposed algorithms. Seeing this techniques work on a sample-by-sample basis rather than on
problem, Texas A&M University researchers developed a an RMS current signal.
micro-computer based algorithm which would not be adversely
affected by this random behavior, and which would actually The first of the two techniques simply subtracts, point by
use this type of behavior as an indicator of a fault [6,34]. point, the previous cycle from the preselt cycle and looks at
the resrlt. Under normal system conditions, the output of this
The Randomness algorithm uses the same energy values subtraction should be very small, since loading tends to change
that the Energy algorithm uses, and even keeps a normal quite slowly. Whenever the subtraction yields a substantial
system average like the Energy algorithm. Whenever a cycle output, however, something has happened to the system.
energy is found which deviates from the average by a certain Foliowing this, time can be used as a discriminatory factor; if
predefined percentage, the system status is changed from the curreflt goes from one stable constant level to another
normal to "Disturbance." At this point, the Randomness stable constant level, no fault is indicated since this was
algorithm does not look for a sustained increase, however. probably a normal system switching event. If, however, the
Instead, it counts the number of transitions from arcing (high butput of the subtractor continues to fluctuate to a large
level of energy) to non-arcing (low level of energy) during a degree, a fault is indicated.
short period of time (e.g., 30 cycles). During this time, it also
counts the number of adjacent cycles which have energies A second, very closely related technique calculates the
which differ from each other by more than a certain threshold frequency spectrum of the current signal at each moment in
amount (to detect faults which do display longer bursts of fault time; then, using this spectral information, it attempts to
current). After checking these cycles, the two counters are predict what the samples of current in the next cycle should be
checked. If either counter exceeds its threshold number, a [:S1. tf there is close agreement between the two, the system
fault is signalled. is assumed to be normal; if there is a large discrepancy,
however, and the discrepancies continue for some nominal
The Randomness algorithm has been tested with high amount of time, a fault is indicated.
frequency (2 - 6 kJlz) current as its input quantity. It has
been tested in the laboratory only, using recorded staged fault
tests and recorded normal system events (e.g., capacitor bank ARCING BURST PATTERN ANALYSIS
switchings, voltage regulator tap changes, etc.). It performed
well in most cases, missing only the small percentage of faults Fault current in a downed conductor fault is driven by the
which exhibited extremely low levels of high frequency activity. fault point voltage between the downed line and the ground.
It correctly distinguished the normal system events as only Because of this, the fault current will follow a somewhat
system "Disturbances." repeatable pattern [3,36]. It begins to flow when the physjcal
briakdown voltage of the air gap between the downed line
The Randomness algorithm has only been tested on high and the ground is reached and continues to flow until the fault
frequency curreot to date. Testing its applicability to other point voltage can no longer maintain the current flow across
frequencies is planned. ihe gap, somewhere close to the voltage zero-crossing. Then,
this paltern is repeated in the negative half-cycle of the phase
voltage.
RATE.OF.CHANGE DETECTOR
When the arc begins, it generally sparks multiple times
The problem with relaying based on conventional 60 Hz before becoming well established; this sparking causes high
current information is that normal pick-up levels must be set frequency current activity. A smaller burst of high frequelcy
high enough to allow for maximum load and for other events activity is produced when the current flow extinguishes at the
such as cold load pickup (following an outage). With the relay voltage zero. This sequence of events produces high
pick-ups set this high, downed conductor faults, which typically frequency activity which can be matched fairly well to the
draw RMS currents of only a few tens of amperes and only for phaie voltage. Thus, Texas A&M Unirersity researchers
very short periods of time, will never be seen by conventional iealized thai by looking at the level of increased activity in
relays. Realizing this and realizing that normal load currents various portions of the voltage period, they could distinguish
tend to change slowly over time, J. Carr suggested the use of arcing faults from other events which might cause a sudden
a ratg-of-change detector for downed conductor faults [9]. increase in high frequency activity.
Although never built, Carr's detector would monitor a A second effect of the predictable phase voltage angles at
signal proportional to the RMS value of the ground current. which current will begin to flow and cease to flow can be seen
This signal would be passed through a "high-pass filter," which by isolating the fault component from the load current (e.g.,
would pass only quickly changing signals, blocking the slowly as suggested above in the section Extracted Fault Current
changing load current signal. The output of this filter could Techniques), and comparing the fault component to the phase
be monitored, and a fault signalled if this output exceeded a voltage. A positive correlation of the fault component and the
threshold. Carr suggested applying this technique to a circuit's phase voltage is a good indicator that the current is actually
ground current; it would probably work equally well if applied flowing into an arc.
to phase current.
26
By averaging the frequency spectrum of a number (e.g., recloser. Following this, the indication of arcing concurrent
300) of power frequency cycles, a //distribution can be seen with the reduction in the original load current level can be
during an arcing fault. While certain normal system loads may securely identified as a downed, arcing conductor.
display broad band noise, the noise is repetitive on a
cycle-by-cycle basis, which causes certain discrete frequencies By identifying patterns for the load analysis algorithm to
to be displayed, rather than the smooth frequency distribution recognize as situations in which a conductor is broken and
characteristic of downed conductor, arcing faults, other pattern for it to classify the conductor as intact, the
overall security of any downed conductor fault detector can be
By calculating this average frequency spectrum in the greatly enhanced. Given the utility position on maintaining
background while running a detection algorithm, the frequency the security of their systems, Texas A&M researchers feel that
spectrum can be examined whenever the detection algorithm such an algorithm must be used in any eventual detector for
indicates a fault. In this case, if the frequency spectrum it to meet with success.
indicates that the event which caused the detection algorithm
to operate is not a true arcing fault, the trip signal can be
inhibited. In this way, system security against false tripping CONCLUSIONS
can be improved.
Many techniques have been proposed to solve the problem
of undetected downed conductor faults. Several of these
PROPORTIONAL RELAYING techniques have been tested, either in a laboratory setting or
in the field, with mixed results.
On a grounded-wye system, Carr shows that the fault
current can be calculated by summing properly proportioned Each of the mentioned techniques has known deficiencies
neutral and ground currents [9]. If the fault current only is in that it will miss some subset of the true downed conductor
calculated in this way, it would be much easier to detect a arcing faults; some also suiier irom the relatively worse
fault than if any raw current signal is used. problem of false tripping, either from unknown causes or from
normal system conditions such as blown phase fuses or phase
The proportioning constants can be calculated as ratios of unbalance. While none of the techniques provide the "magic"
the line and ground impedances. Carr reasons that while the solution which will detect lNVo of the dorvned conductors
individual impedances used for the calculations may change while providing complete immunity from false trips, some
due to moisture, temperature, etc., each of the impedances combination of a subset of the techniques would greatly
should,change proportional to the others, making their ratio increase the probability of detecting a large number of
unchanged. If this assumption does not hold, he suggests the downed conductors.
use of a feedback path which would automatically adjust the
proportioning constants based on ambient conditions.
LOAD ANALYSIS
Utilities representatives have stressed that one of their
main concerns is the minimization of false detections. Their
position is that they want a detector which will detect only a
broken conductor on the ground. Any other condition (e.g.,
broken conductor dangling out of the reach of the public,
intact or broken conductor in contact with a tree, etc.) ihoutd
be left energized. The reason for this position is that, while
energized downed conductors are a public safety hazard,
frequent unnecessary service interruptions can pose safety
problems of their own. For instance, during a power outage
traffic signals are off, lighting is lost in homes and businessel,
etc. Because of this concern over maintaining system security
against false tripping, Texas A&M researchers are developing
a load analysis algorithm to monitor patterns in load current.
27
CHAPTER 6
device is included in the logic which determines a down_wire We must take a hard look at construction standards and
condition. Therefore, fused taps, or any portion of the main practices used over the last fifty years with an eye for things
feeder beyond a fuse or singie phase recloser can not be which contribute to failure of a conductor; and when a wire
protected with this method. does fall, things which allow it to miss contact with the
grounded neutral.
The ultimate integration is found in development of expert Use larger conductor, even when smaller wire will meet
systems for protection against down-wires. Many detection and load current requirements. A common scenario for a down-
venllcatlon lnputs can be combined with non_electrical wire incident begins with a "normal" fault (from lightning
information, such as time of day and weather conditions, to flashover or tree contact) which melts the primary wire. After
leld a much better basis for circuit interruption decisions. the circuit recloses, the primary is left on the ground
F*p".t systems are also adaptive and dynamic; able to learn energized. Bigger wire dissipates much more heat from the
from mistakes. Such systemscpuld signi"ficantly improve fault arc without melting. Using ACSR (aluminum conductor, steel-
detection. However, until didtribution automation is more reinforced) instead of all-aluminum conductor is a related
widespread and sophisticated, expert relay systems are practice. I-arge wire is also more resistant to mechanical
available, we should turn to improved practicis to help solve failure from fatigue or the weight of ice and trees.
this problem.
Use tighter configurations to improve chances of neutral
contact by a falling phase wire. This rule precludes framing
COORDINATED EMERGENCY RESPONSE SYSTEMS with crossarms, standoff brackets, or anything else which
pushes the conductors out in the horizontal plane. The tightest
. ,Utilities can greatly improve the response to reports and design available is vertical post, with the neutral spool on the
rndrcatlons of down-wire. There are major problems with same side of the pole as the phase conductors. A subset of
existing response procedures, but many can be oue.come if the this practice is single phase construction on a post insulator-
ttnqrunjtf, public safety agencies, uiility operators, and the
orstnbutton plant are viewed as one macrosystem. This
on the side (over the neutral) of the pole, instead of the
common pin insulator centered on top of the pole. A centered
discussion must remain fairly vague, because each utility and phase wire has a 507o chance of falling down the side away
community are unique; howevei, examples of this ',system" from the neutral, placing 8 to 10 inches ofwood between the
approach come to mind: phase and metallic neutral contact.
Avoid ncompressed neutral' construction. When the neutral Increase tree trimming. There are many reasons, other than
spacing is "compressed" (ultimately up to the crossarm, as in down-wire risk, for enhanced trimming; so add public safety
one REA standard), it is difficult for a falling phase wire to to the 1ist. Chronic arcing to branches weakens and melts
make metallic contact. conductors, branches can cause phase-to-phase faults, and
trees often physically tear down the line.
Avoid covered conductor, unless needed for an over-riding
safety concern like clearance to a building. Even when it falls Convert 4160 volt systems to the higher voltage. There are
on a fairly conductive surface, covered wire often doesn't draw capacity and economic incentives to make the conversions we
enough current to blow a fuse. Covered wire may also be have been planning for so many years. A
public safety
more likely to fall. A power arc tends to concentrate at the incentive is added when we see how difficult it is to detect a
point offailure in the covering, instead of"motoring" along, as 2400 volt fault to ground.
is common with bare conductor. The resulting localized
heating often melts the wire. Reduced tree trimming Repair crossarm framing with post construction when
requirements may not be sufficient cause to run the risks of failures occur. The tighter configuration is preferred, and risk
covered conductor. of decayed crossarms is eliminated. Though this "inconsistent"
framing may be traumatic to the utility culture, it has been
Avoid unfused taps, especially with small wire size, and proven that the public doesn't notice such construction
even when the tap load is insignificant. The reliable fused features of overhead lines.
cutout is our first line of defense against "normal" faults.
When a liberal use of fuses is combined with the changes in Evaluate reclosing philosophy. Try to reduce the number
fusing philosophy proposed later in this discussion, the cutout of reclosures and shorten the elapsed time between fault
becomes very effective against down-wires. Fortunately, initiation and final reclosure. Each shot is a full arcing and
cutouts are usually easy to retrofit onto all but the most wire-melting event at the fault location. Long cycle times give
crowded poles. curious bystanders time to approach the temporarily dead
conductor. Present practice should be examined in both
Avoid mid.span connections, or "double up" the jumpers. quantitative and qualitative ways. Ask questions - What is
The flying tap is a handy construction technique, but can be really gained by a 4th reclose that a 3rd misses? What is
a source of localized heating and mechanical stress on the really gained by waiting 60 seconds to reclose that is lost by
wire. waiting only a few seconds? Isn't "no intentional delay'' of the
first reclose as upsetting to digital-age customers as a 3 second
Avoid floated-wye capacitor banks out on the distribution delayed reclose (which helps arc products dissipate)? Are
feeder. This configuration can be a dangerous backfeed source there times or circumstances when no reclose should occur?
during load-side down situations. When a wire breaks at the [3e]
pole [where most do], and only one span falls, the odds are
5G50 for the load side going down. Backfeed situations are a Evaluate fusing philosophy with a focus towards minimum
particularly insidious hazard, because the equivalent source sizing. The present emphasis seeks the maximum rating which
impedance may keep the down-wire from sparking, smoking, will still coordinate with upstream devices, hoping to reduce
or giving other signs of its energized state. nuisance blowing during peak load periods. This has resulted
in the simple stepped schemes (eg 600,4-100,4'-65,4.) common
Avoid grounded-rvye high, delta low side distribution to the industry. A more rational approach asks the minimum
transformer connections. This presents a backfeed hazard for size possible given a generous overload factor. The result is
Ioad-side down incidents similar to that caused by a floated- a wide range of sizes, so this scheme really wasn't practical
wye capacitor bank. before computerized databases and computer generated fusing
diagrams became available. Refering to Figure 6-3, why put a
Improve lightning protection in lightning prone areas. IJse 40A link in series with a 6,{ transformer fuse, when we can
more arresters, better arresters, lower grounding impedance put a solid link in the transformer jack and move the 6,4 link
and larger downleads (low impedance to high frequency out to the road to replace the bigger fuse? [39]
content of lightning current waveshape) for arresters.
Boost public education programs, especially for children,
Reduce rear lot line construction. In overgrown rear and especially focused on the down-wire problem. Stress the
easements, maintenance is more difficult and occurs less dangers even if the wire isn't sparking and even if it is covered
frequently, so these lines are more likely to fall. When they do with insulation and looks like an extension cord.
fall, they often go down in hedgerows and are generally more
accessible, but less visible to the public. Provide scenario simulation for operators. No amount of
"what-if?" discussion will give operators all the answers, but
Boost incentives for underground development. This can every exercise helps provide a structure for thinking through
even be economically favorable if maintenance and storm situations when they occur.
repair costs are evaluated over the life of the system.
Developers will often participate for the aesthetic benefits of DEVICES
underground distribution.
Every fault protection device can be characterized by the
location it senses the fault and the location it clears the fault.
OPERATING PRACTICES Down-wire protection should follow this same assessment
scheme. Sensing the down-wire at the first upstream protective
The same kind of hard look at how we operate our systems device, such as a fused cutout, is said to be local sensing (15).
has revealed a number of ways to minimize down-wire events, If this detection occurs two or more protective levels
and improve things when they do happen. These upstream, the protection method is source sensing (SS). De-
recommendations can be as debatable as the changes in energizing of the down-wire can be grouped likewise as either
construction standards, and require equally careful analysis of local or source clearing.
risk and benefit. Some recommendations:
30
tr
$t
tt
U
o
u
u
sl
to
z< is
:< $t
ti
uI
@Q
$i
<I :itl
Ig $i
s!
,2 AAC 6 AMP
r LINK 15 KVA
i
I
1 PHAgE LATERAL
INTRODUCTION
The application of high impedance fault detection on a Category II - Sagging or Fallen Conductors
distribution feeder has important implications for the
operation of the circuit. This type of fault detection has This category differs from the previous one in that the
similarities and significant diffelences with respect to conductors themselves are not broken so line current
.conventional.
,
continues to flow. This situation might typically result from
protectiory factors which must be iarefully
c,onsidered in implementation. High impedance fauli a.broken pole or pole hardware such as an insulator pin
detection adds several unique complications to system failure. The conductor may sag or contact the ground, trees,
operation which must also be considered. Because sensitive shrubs or some other foreign object. Two types of etectricai
fault detection ga1 implct the number and duration of outages conditions are therefore possible:
on a feeder, it is beneficial to develop high impedance fir.lt
detection in such a manner as to idenlif, and rimove known 5. no ground contact or very high impedance to ground
hazards while not adversely affecting service.
6. low impedance ground contact.
33
Group B - Loss of Voltage FAULT TYPE
'lt
o
a @
A break in a conductor could potentially be detected by the P: A Po g rop- 9
loss of voltage on the load side of the break. Taking action --6 oor=
such as deenergizing the line would require some form of iE; A1
:u
E?
aE
6=
O
9@ t o
:'iP
9
^
Y
communication link from the detection location back to the ='f 9
EEd ql
o o{ !2o f
e385 S
substation. DE:i
60
B 8pol9
ao E3 A E!E 3
9b d.6
S
I
Group C - Non 60 Hz Measurements iBgs
of
OF
50
Several investigations have been completed into the noise B-
I
and harmonic currents that are caused by some types of faults
involving downed conductors. There is both experimental oa a aa A- Sansitivo Ground Fauh
evidence and theoretical confirmation of the fact that certain o
types of faults have a harmonic or noise signature that could aoaa B. Loss ol Voltag€
{m
m
Detection Ootions
'Table GOAL OF HIGH IMPEDANCE FAULT DETECTION
7-1 shows that no sfingle measurement method will
detect all the electrical conjditions that can result from a The concept of high impedance fault detection has often
downed conductor. Even the Eonditions that can be measured
will. not always . be reliably detected due to sensitivity
been discussed as a form of protection, analogous to
limitations. The situation when a single measurement method
overcurrent protection. In actuality, the goal of high
impedance fault detection and overcurrent piotection aie
is used can,be depicted as in Figure 7-2. The heavy circle somewhat different, a fact which merits discussion.
represents the area in which pll possible electrical conditions
can occur and the light circlet represents the area covered by
The primary motivation for overcurrent protection is the
the measurement method. rlnsofar as the circles do noi protection of the electrical network. The need for overcurrent
coincide exactly, there will be both false indications of faults protection is based upon the electrical and physical properties
and failures to indicare genuine faults.
ofthe conductors which supply electricity. Ifthese conductors
carry too much current for too long a period, they will melt,
One approach to successful fallen conductor detection is to destroying part of the electrical supply system. A conventional
reduce the area of the light circle until there are no false
fault usually interferes with the normal operation of the
indications, and then accept a high probability of failure to system and if left energized could result in extensive
detect. Another approach is to use more than one equipment damage. It is straightforward to protect the system
measurement technique, as slown in Figure 7-3. By choosing against such an occurrence, and it is easy to measure
compiementary techniques and suitably adjusiing theii parameters which indicate when a potentially damaging
sensitivities, a much greate( range of condiiions witt Ue condition exists. In practice, an additional motivation fo;
detected. In some situations only one measurement element overcurrent detection is the removal of safety hazards caused
would detect the condition, while in others two or more might by faulted equipment.
simultaneously react. This is depicted in Figure 7-3 by having
the three light circles (measdrement meth-ods) coveiing thI In contrast, a fallen conductor quite often presents no risk
single heary circle (all possibfe electrical conditions). to system equipment and need not affect system operation.
It can, however create a serious safety hazard. The incentive
for applying detection methods is therefore to improve safety,
All Faults not to enhancg system operation or protect system equipment.
The electrical network is designed to carry the amount of
Urldetec{ed cxrrent common in high impedance faults. Although some
damage to c-onductors and equipment may occur during a high
impedance fault, this damage is typically localized to the polnt
Delected of the fault with few, if any, adverse effects on the rest of the
system. The primary motivation for high impedance fault
Falso detection is not to protect system equipment or to enhance
system operation, but rather to improve safety. The action
l taken to improve safety may be the same as used in
Figrre 7-2. Single detector covers only some faults. overcurrent detection, namely, to de-energize the affected
portion of the circuit.
D€t€ctor I
APPROXIMATION OF UNSAFE CONDITIONS
All Fau[! With regard to high impedance faults, the goal of operation
.
is to detect and remove unsafe conditions, and to iontinue
service otherwise. The major challenge with this type of fault
detectioa is to identify unsafe conditions. Unfortunately, this
task is impossible to achieve. "Safety" is not a measurable
quantity upon which one can design the operation of a
detector. As a result, one can only measure parameters which
an -safe
provide approximate indication of or unsafe
conditions.
The determination of whether a high impedance fault is Texas A&M organized the EPRI High Impedance Fault
imminently hazardous or temporarily acceptable is largely Detection Workshop in November, 1988 to provide a forum
in{luenced tly conditions at the fault site which may be for utility representatives to discuss different aspects of high
difficult to measure from a remote location. We desire that impedance fault detection. The utility representatives
a high impedance fault detector clear hazardous situations and indicated strongly that it is important for a high impedance
allow non-hazardous conditions. This choice is largely fault detector to operate correctly, especially that it not falsely
influenced by such information as: indicate the presence of a fault when there is none. They
indicated a preference for failing to trip for some high
. Is a conductor on or near the ground? impedance faults rather than accepting nuisance trips.
. Are people nearby a downed conductor, or is no The utility personnel also stated that they would like to trip
one near it? for a conductor on or near ground condition, and that they
would not trip for a high irnpedance fault in which the
. Is fire a possibility? conductor remained intact on the pole top, for example, in the
case of a tree limb in contact with an intact primary.
It is impossible to implement a fault detector which could
absolutely answer these and similar questions. One can only When a high impedance fault detector trips a circuit, utility
measure parameters or identiff Patterns which are considered personnel agree that reclosing should be disabled.
indicative of a hazardous situation.
Despite our best efforts, some unsafe conditions will remain RECOMMENDED IMPLEMENTATION
too subtle for detection. As an example, a pole crossarm may
break causing an intact primary to fall near, but not touching Given these guidelines, as part of research for EPRI
the ground. While everyone would agree that this is a Project RP1.472-4, Texas A&M has recommended the
dangerous condition, there are no measurable parameters implementationof a high impedance fault detector which
which would indicate the condition has
occurred. provides both tripping and alarming actions. It is
Notwithstanding such very subtle faults, new technologies, recommended that the unit trip only on the detection of the
once conmercial, promise to identif a high percentage of sequence of: (1) a significant, immediate loss of load, followed
high impedance faults and unsafe conditions. shortly by, (2) electrical activity indicative of arcing. This
sequence is the most reliable indication of a broken conductor
on the ground, the condition which is most easily measured
OPERATIONAL TRADEOFFS accurately to indicate a hazardous condition.
The implementation of sensitive fault detection will, in all Other faults can also be detected, but it is less definitive
liketihood, mean that there will be more outages. Because that the situation is immediately hazardous. For example, the
these faults may be hard to locate, it is likely that the outages presence of arcing without a loss of load is most likely tree
will have longer durations. It is important that utilities try to limb contact with an intact primary, or leakage current at a
trip a circuit only when necessary to remove a hazard and to cracked insulator. The same electrical signature could also be
preserve service continuity otherwise. caused by a single phase, intact lateral on the ground. Given
the utilities'suggested bias toward avoiding nuisance trips, it
Given that one can only approximate unsafe conditions would be recommended for this case to alarm the condition,
through measurement of various parameters, the choice of but not trip.
operating characteristics becomes critical and the
implementation must be sophisticated. It is difficult to weigh In practice, it is likely the most effective fault detection
the implications to safety of various approaches to the systemwould include a communications link from the detector
operation of a high impedance fault detector. to the utility system operator. The detector could provide an
alarm to th"e operatoi for evdnts which are likely io be high
If a fallen conductor is located in a residential impedance faults but for which the data is not sufficiently
neighborhood, one could reasonably assume that the condition definitive to warrant a trip. The operator can make a
is imminently hazardous and should be cleared at once. But determination for a course of action based upon established
if the fallen conductor is located in a remote area far from company procedures. The operator-may choose to manually
homes, there may be no immediatehazard, and there may be trip the feeder or send a trouble crew to investigate the event.
some benefit to customers and the public by keeping the The decision can be made based upon the data from the
circuit in service. detector as well as from other information such as an outage
call from a customer or weather-related data.
This last point complicates the whole matter of making
operational decisions about high impedance fault detection' Allowing utilities to set the operating characteristics for an
While we may improve safety on the one hand by de- automatic trip, a trip on confirrnation, or an alarm will provide
erergizirg a feeder at an indication of a downed conductor, a Eteat deal of flexibility to utilities. Not only would the
we may also diminish safety at the same time because other abitity to trip a fault be of significant help, but the information
types of accidents occur during electrical outages. One need provided byan alarm would also be a substantial improvement
only consider the danger encountered at a major intersection ior utilities in addressing high impedance fault detection.
36
FAULT L,OCATION
Because high impedance faults are often subtle, there may made between sensitivity and discrimination, and quite
be little evidence providing the location of the fault. If a fault different performance accepted with respect to false deteciion
_
detector is used to trip an entire feeder, and ifit is difficult to and failure to detect.
locate the fault, customers may be without service for a
lengthy period. The implementation of high impedance fault detection on
a distribution feeder. presenta many challenges f-
Typically those faults which involve a broken conductor or effectiveness. While the goal of detection is to remove
other significant item of poleline hardware will be relatively hazardous conditions, measurement of unsafe conditions can
easy to find. But faults for which the conductor remains intact only be approximated. It is important to bias the operation
and fault contact is intermittent may be quite difficult to of a detector to trip only when iertain that a conducior is on
locate in a timely manner. It may be helpful to use a fault or near ground. The recommended implementation is a
locating device such as an RF direction finder which can detector,which.trips upon recognition of a pattern of loss of
Iocate RF noise associated with arcing.
l9T.,t"9lro*,.0
ty arcing, a condition which most accurately
ldentltres a
broken conductor on the ground. Other conditioni
which are potentially hazardous but-for which the data is not
conclusive can be alarmed.
coNCLvSTONS
Utility relay engineers have done an excellent job in shown to successfully detect certain types of faults, no single
designing systems which will automatically deenergize the approach can reliably detect all high impedance faults.
power system as a result of abnormal or fault conditions.
However, a significant number of faults occurring in electric In the last few years research has intensified on this
utility distribution systems due to downed conductors cannot problem. Research by Nordon Research and Development,
be detected by existing commercial protection devices. A Westinghouse, several foreign research teams, and intensive
public hazard results when an electrical conductor is downed, efforts at Texas A&M University have added to our body of
within reach, and energized. Realizing this, utilities have knowledge. The present conclusion generally agreed to by
taken extraordinary means to find a solution to this vexing most researchers is that no single detection technique can be
problem. To date, no solution exists, but research depended on to detect the majority of high impedance faults.
investigations continue. However, research has shown that a combination of several
techniques might be used to detect a significant percentage of
The fault problem can be described as follows. Certain the presently undetected faults. In order to implement a
downed conductors are partially "insulated" because of the combined set of detection schemes, a more intelligent relaying
surface conditions on which they fall. For example, a system will be needed, probably demanding that a
conductor falling on dry asphalt has a very high impedance microcomputer based system be utilized. Work on such
path to ground and will draw very little, if any, current' synergistic detection methods 4nd intelligent adaptive relays
Hence, it is possible to define the problem of high impedance continues.
faults as a detection problem wherein these faults draw
insufficient current to be detected by conventional overcurrent While we wait for a significant breakthrough in substation
protection devices and methods. Since overcurrent protection based protection, other steps cafl be taken to mitigate the high
utilizing relays, reclosers, and fuses represerts the primary tool impedance fault problem. The availability of distribution
ofprotection engineers, a certain percentage offaults become auiomation systems, including feeder communications, offers
simply undetectable. new possibilities for improved feeder protection. Improved
construction practices, including tighter construction
High impedance, low current faults have characteristics configurations, the use of iarger conductors, and avoidance of
significantly different than high current faults. The typical covered conductor can improve the detection of certain types
fault which occurs when a conductor touches another curent of fauls. However, there are obvious places where such
carrying conductor will draw many times the current which construction approaches are inappropriate. Improved
will flow under normal conditions. However, a high operating practices such as frequent tree trimming and
impedance fault may draw currents which are essentially equal improved lightning protectio4 can assist in reducing the
to or less than normal load currents on a feeder. It is number of downed conductors. New devices such as
impossible using existing commercial relays to disting-uish mechanical guards can also be of assistance by converting
berween normal load currents and these high impedance fault faults to high current faults before the conductors fall to earth.
currents. Other characteristics of high impedance faults
include random behavior which tends to confuse protection While all of these methods may offer some improvement,
devices. By this we mean that the fault path is unstable and none represent a direct solution to the problem of detecting
the current levels fluctuate widely. Such faults are extremely high impedance faults. If all of the proposed construction and
hard to reliably detect. operating practices were utilized, a certain percentage -of
dbwned ionductors would remain undetected. A method for
For many years, utility engineers and researchers have detecting these faults is still needed. However, the solution
sought a solution to this problem. Specifically during the last must be economical and practical to apply. Research by
decade, research projects sponsored by utilities, manufacturers, engineers dedicated to public safety continues. It is hoped
and universities have intensively looked for unique that through continued discussion, investigation, and further
characteristics of high impedance faults which could be used research, methods will be found to detect a higher percentage
to build new protective devices. Significant efforts by such ofdowned conductor faults. Hpwever, it should be noted that
utilities as Rochester Gas and Electric and Pennsylvania it is theoretically impossible to detect all such faults. Even
Power and Light have resulted in a clearer understanding of with an ideal protection system, some faults would remain
high impedance faults. These investigations along with undetected and the public must be conscious and cautious of
reiearch sponsored by the Electric Power Research Institute the danger.
have resulted in several theoretical concepts for new devices.
Researchers at PTI, Hughes I-aboratories, and Texas A&M It is the hope of the authors of this document that the
University investigated several new concepts during the 1970's. tutorial presentations and text will encourage discussion of
The early efforts of these five gtoups resulted in various the problem, self evaluation by electric utilities, and further
proposed detection methods ranging from sequence motivation by all researchers toward improved distribution
compooent and harmonic analysis to high frequency spectrum fault detection methods.
and noise analysis. While several of these methods were
38
REFERENCES
37. R.E. I-ee and C.L. Brooks, "A Method and Its
Application to Evaluate Automated Distribution
Control", IEEE Paper 87 WM 153-0.
40
APPENDIX A
SUGGESTED SOURCES FOR FURTHER READING
. This. appendix contains a bibliography of high impedance, 14. J. Carr, "Detection of High Impedance Faults on Mult!
downed conductor fault literature which is not citeddirecily in thti Grounded Primary Distribulion Systems,,IEEE Traruactions
text of the Tutorial. While most of the entries are technical in ort Apparatus and Systems,Vol. pAS_100, No. 4, April,
nature, several also deal wifh related, non-technical issues ^?ower
1981, pp. 2008-2016.
concerning the downed conductor problem.
15. C.:..: Fault Detection Using 60 FIz
.An.attempt has been made to be as complete as possible in this
{;
Methods," l,Higll"_,p"!gnce
IEEE pES Transmission and Disiribution
collection of literature. As with pny project of this nature, however, Conference, Anaheim, September, 19g6.
there may be omissions.
t6. R. .
P. Chinchali, ph.D. Dissertation of Texas A&M
University, College Station, Texas, August 19gg.
1. Y, Al-Dabbagh, R. Daoud, R. Coulter, ,Improved
Microprocessor Based Feeder Earth Fault protection Using 17. R. R. Conrad, D. Dalasta,'A New Ground Fault protective
Pattern Recognition,,, presented at the Fourth Internationa'i System for Electrical Distribution Circuits,, IEEE
Conference on Developments in power System protection, Trattsac-tions ott Industry and GeneralApplicatioru, Vol. IGA_
Edinburgh, UI( April Il-13, j989. 3, No. 3, May/June, '1967, pp.217_22i.
2. B. M. Aucoin, "Detection of High Impedance Faults Using R. F. Cook, R. V. Adams, I. A. Whyte, ,Developing a
Arcing-Generated Burst Noise Sienals,,, IEEE pES Commu.nicarion System for Automated OistribirtiJn,,
Transmission and Distribution Coiference, Anaheim, l|testingltouse Engineer, yol.35, No. l, January 197i, pp. li_
September, I986. 74.
3. B M: Aucoin, _"Operational, Economic and Legal "Downed Power Lines: Why They Can't Always Be
Considerations of High Impedance Fault Detection," IEEE Detected," IEEE power Engineering Society public Affairs
PES Transmission and Distribution Conference, Anaheim, Document, 1989.
September, l986.
20. "Distribution Fault Interruption Open Conductor Test,,,
4. B. M. Aucoin, B. D. Rusbell, ,,Algorithm for the Detection report by the Distribution Fault Curient Interruption Task
gf yojs_e_Iequencies Generated by High Impedance Force, Report# SA0 907041, pennsylvania powei and Light
Faults," IEEE iontrol of irower System; Co;fe."n"L, pup". Company, Ocrober 20, 1975.
No. 78CHi282-3, Oklahoma City, Oklahoma, March, 1978.
21, "Distribution Line Protection practices Industry Survey
5. B. M. Aucoin, B. D. Russell, "Distribution High Impedance Analysis." IEEE Committee Report, IEEE Traruictior* on
Fault Detection Utilizing High Freque-ncy 'Current P^ower Apparatus and Systeftrs: Vot. pAS-102, No. 10,
Components," IEEE Trar1sactiorx on powir Apparanx and October, 1983, pp. 3279-3287.
Systents, Vol. PAS-101, No. 6, June, 19g2. pp.-i596-1606.
Dbtribution Systen*, Westinghouse Electric Corp., East
6. B.M. Aucoin, B.D. Russell, ,Detection of Distribution High Pittsburgh, PA, 1965,567 pp.
Impedance Faults Using Burst Noise Sigtals Near 60 llz.,
IEEE Trattsactions on Power Delivery,yol. pWRD-2, No. 2, 23. "Distribution System Protection Manual", McGraw-Edison
April 1987, pp. 342-348. Co., Canonsburg pA, Bulletin No. 71021, 155 pp_
7. M. Aucoin, B. D. Russell; C. L. Benner,'High Impedance 24. S. D. Ebron, "A Neural Network processing Strategv
for the
Fault Detection for Industrial power Syslems,i' IEEE D erec{ i on_ of HigJr.Impeda nce Fa ulrs,, The jis, North
M.i.
Industrial_Applications Society Conference, San Diego, Carolina State University, Raleigh, North Carotina, 19gg.
October, 1989.
25. W.A. Elmore and R.M. Reedy, ,'Electromechanical
8. S. J. Balser, K. A. Clements, "Statistical Algorithm for Relay to
Detect Fallen Distribution -Conductors,, presentation to
Detection of High Impedance Faults,' submitted for 19g3 American Public power Association Engineering &
IEEE/PES Summer Power Meeting. Operations Workshop, February I9g3.
9. S.J. Balser, K.A. Clements, D.J. l.awrence, ,'A 26 A.8.. Ernan_uell
Microprocessor - Based Technique for Detection of High 9tj al., ',High.Imqedance Fault Arcing On
Sandy Soil In 15kV Distribution Feeders: Contributi,ons to
Impeclance Faults", IEEE Trarcactions on power Detivelry, the Evaluation of the Low Frequency Spectrum,,, IEEE
Vol. PWRD-1, No.3, July 1986. Paper 89 SM 784-0 PWRD.
31. C. L. Huang H. Y. Chu, M. T. Chen, "fualysis and 48. G. D. Rockefeller, "Fault Protection with a Digital
Detection Al"sorithms of High Impedance Fault in Primary Computer," IEEE Traruqctioru on Power Apparatus and,
bistribution Systems," National Science Council of Taiwan Systems, Vol. PAS-8E, Noi 4, April, 1969' pp. 438-464.
Final Repo*, Research Project NSC?4-0404-8006-06'
December, 1985. 49. B. D. Russell, "High Impedance Faults Can Now Be
Detected,n Transmission qnd Distributioa Vol. 34, No. 2,
C. L. Huang, H. Y. Chu, M. f' 'Algorithm February, 1982, pp.32-34, 60.
32.
Co-ouriron f6i ffish Impedance Fault -Chen,.
Detection Based on
stasJd Fault Tests,-- IEEE Traruactiorts on Power Delivery, B. D. Russell, B. M. Aucoin, C. Benner, "Computer
vo[ pwRo-1, No. 4, October 1988, pp. 1427-1435' Relaying Techniques for the Detection of High Impedance
Faulis Using Signal Processing and Pattern Recognition
JJ. A. O. Ibe, B. J. Cory, "A Travelling Wave-Based Fault Methods," International Conference on Power System
Lo"uto. for Two-and Three-Terminal Networks,* IEEE Protection, Singapore, September, 1989.
iiaioctiors on Power Delivery, Vol. PWRD-1, No' 2, April'
1986, pp. 283-28E.
51. B. D. Russell, R. P. Chinchali, 'A Digital Sigral Processing
Algorithm for Detecting Arcing Faults on Power
Diitribution Fe eders," IEEE Transactioru on Power Delivery,
34. "IEEE Guide for Interfacing Dispersed Storage and Vol. PWRD-4, No. 1, January 1989, pp. 132-140.
Generation Facilities with Electric Utility Systems"'
ANSI/IEEE Std 1001-1988.
52. B. D. Russell, K. Watson, 'Power Substation Automation
35. "IEEE Recommended Practices for Electric Systems in Using a Knowledge Baped System - Justification and
Health Care Faciities," ANSI/IEEE Standard 602-1986' Preliminary Field Experiments," IEEE Transactions on
Power Delivery, Vol. PWRD-2, No. 4, October, 1987, pp.
1090-1097.
36. "The Interruption of Downed Conductors on l-ow-Voltage
Power
OiitriUution'Systems", a report prepared by the IEEE
System Relaiing Committee for the Electric Power 53. B.D. Russell, R.P. Chinchali, C.J. Kim, 'Behaviour of Low
fiesearch Institute, October 1976. Frequency Spectra During Arcing Fault and SYi!!!!g
Events', IEEE Trarxactions on Power Delivery,Yol. PWRD-
3, No.4, October 1988, PP. 1485-1492.
3t. A. T. Johns, A. A. El-Alaily, "New Distance Protective Relay
*ittt f*p.ou"O Coverage ior High-Resistance Earth Fault"' B.D. Russell, K. Mehta, R.P. Chinchali, "An Arcing Fault
;;;;d;;i', t^int oinectrxai engineen, (GB), Yot' r24' 54.
Detection Technique Using Low Frequency Current
No. 4, Ap;il, 1977, PP.349'355.
Components-Performance Evaluation using RecoTlgd Field
H. L. Jou, 'Analysis and Detection Simulation of High Dat;', IEEE Transactiaru on Power Delivery, Vol PWRD-3'
38.
i-r"a.*.i paultJ in Distribution Systems," M' s' Thesis' No.4, October 1988, pp. 1493-1500.
lfuiion"t Cheng Kung University, Taiwan, June, 1984' 55. R. F. Shields, ''The Problem of Arcing Faults in Low-
DuringArcing Voltage Power Distribution Systems,"IEEE Tmnsactions on
39. C. J. Kim, B. D. Russell, 'Harmonic Behavior
oi-, Power Distribution Feeders,' Electric Power Indttstty and Geneml Applications, Vol. IGA-3, No. 3,
euritt January/February, 1967, pp. 15-25.
iii*r Research, Vol. 14, No.3, June, 1988, pp' 219-225'
43
APPENDIX B
PHOTOGRAPHS OF REPRESENTATIVE
DOWNED CONDUCTOR TESTS
PNM TESTS
TU ELE,CTRIC TESTS
The first two photographs are of tests conducted for Texas conducted for Texas
A&M by Public Service io-purry of New The faults The next photograph represents tests
-Mexico'
were stiged on Rodeo Feeder 2, a 12.5 kV line in Santa Fe
ll Fort Worth' Texas' Thesewing
p.r, h, Terar I hilities faults
^
f;.",";;Jo,; H;;;i; Feeder 54' a 12'5 kv line
which served primarily rurai 1oad. time of the tests was hot
suburban load' The weather at the
The faults were conducted at two different times' The first and dry.
was conducted in July of 1979, the weather was hot and dry,
and the soil was sandy. The second photograph was taken
during tests conducted in July 1981. Again, the weather was
hot and dry, and the soil was sandY.
RG&E TESTS
This appendix illustrates the range of fault currents possible The other current traces in $is appendix illustrate typical
during single line-to-ground faults. Figure C1 represents_ a fault currents experienced duriug downed conductors faults, as
phasJ C to ground fault which was cleared (properly) by demonstrated during- faults stagied by several utilities under
existing protection. various conditions. The procedure for these tests was, in
seneral. to attach a section of conductor to the intact primary,
ihrough a fuse cutout. Then, thd end of this attached line was
IowerEd to the ground using a rope and insulator arrangement
to allow a lineman in a bucket tquck to control the conductor'
The downed lines were always fused with a small (e.g.' 25T)
a.00 fuse in coordination with exisling upstream protection, to
assure that no customer service lintemrptions were caused by
l.o0 the tests.
U
a
$
o Several things can be seen and should be noted from these
0.00
plots. First, arl arc is charactetrized by very sporadic fault
tg iurrent behavior. Current sometimes will flow for only one-
-1.00 half of ore power frequenry cyqle, followed by a lo:g period
of quiescenfnon-arcing. In othdr cases, current will be drawn
-4. o0 for many tens of cycles; however, even in these cases, there
a. oo
often exists agreatdeal ofdifference in the magnitude, phase
angle, harmonic content, etc. from one arcing power frequency
F 1.00 cycle to the next.
r
-e
C
o. oo The second point of interest is that the magnitude of the
q
&
fault current varies greatly frorn one test to another, even if
E the tests are performed at the same test site under similar
I -r.oo conditions.
l
-e.00
e.00 Finally, it is important to rpoint out that none of the
followinj faults werL of sufficignt magnitude for a sufficient
amount of time to melt the fuse link.
o 1 .00
U
II
t 0 .00 TU ELECTRIC TESTS
a
f -1.00
A&$
The first seven plots represent tests conducted for Texas
by Texas Utilities in Fort fiorth, Texas dulng FeEary
1980. the faults were staged dn Randol Mill Feeder 253, a
-a- oo i2.5 kV feeder. The feeder serted primarily residential load'
1. OO
46
4
-ir ,:
4 c? I
o0
l
o 0
l
E E
z
o z
< -l o
-2
-5
-4
05 08 lo 12
nME (SECONOS)
Figure
- C2. Ten cycles of unfiltered and high frequency Figure C5. 100 cycles of phase current following those
(2 - 6 kHz) phase current during arcing fault shown in Figure C4.
on wet clay at TU Electric.
-i,r .-}
c ?
L
o0
f o
f
z
= E
< -t z
(,
oo ol o2 of 04 05
IIME (SECONDS)
oo2 00J o04
IIME (SECONDS)
Figure C3. 30 cycles ofphase current during arcing fault Figure C6. Three cycles of phase current, with phase
on wet clay at TU Electric. voltage shown as a reference (dashed line).
A /\ A A A A A A A
06 08 to rl
\/]/l/l/ \/vl/ I \/I
rlo o o. oo4 006 00a oro oll o14 016
IIMiE (SECONDS) TIME (SECONDS)
Figure C4. 100 cycles-ofphase current during arcing fault Figure C7. Ten cycles of high frequency (upper trace)
on wet clay at TU Electric. Note: tie30 and unfiltered (lower trace) fault current as
cycles shown in Figure C3 are included in this measured at the fault site (no load current
plot. i measured).
HI-&P TESTS
00 02 04 06 08 1.0 12 14 16
TIME (SECONDS)
l-T---l
o02 004 O06 o.oa o to a 12
lrME (SECONDS)
Fisure
" C8. Ten cvcles of unfiltered and high frequency
(2-6 kHz) phase current during arcing fault
to ground rod at HI-&P.
ir
c
oo
l
L
z
o
rIME (SECONDS)
48