No records which did not of themselves indicate multi­ centage of total faults (and disturbances) which developed

ple fault conditions were considered as such in this study, from one to multiple line-to-ground faults at different
even though the correlation data indicated that more points is given below. This is a summation of the per­
than one phase wire may have been in trouble. The centage figures given in table VI for items 1.22, 1.32,
reason for this was that the correlation data in many and 2.22. Also items 1.23, 1.33, and 2.23 of table VI,
instances were not complete enough to base a classifica­ probably include some cases where the multiple fault
tion entirely on them. developed at different locations.
No classification of faults not involving ground could
Percentage of Total Records Indicat­
be made, since the oscillographs, in most instances, were ing the Development of a Fault From
not arranged to operate unless ground was involved. 1 or 2 Phase Wires to a Greater N u m -
The classification was not extended beyond the first change Neutral ber of Phase Wires and Involving
in the number of phase wires involved. Thus, if an Resistance, More T h a n One Point on the Power
Power System Ohms System
oscillogram indicated that a disturbance initiated as one
line-to-ground and subsequently involved 2 and then 3
phase wires, the classification shows only the initiation A—26 kv 75 5.8
of the disturbance and its development from a one line to £—33 kv 30 or 60 0.3
a 2 line-to-ground fault. E—44 kv 150 1.2
An important reason for making this type of classifica­
tion was to obtain some indication of the extent of double
The authors take this opportunity to thank, in behalf
faults* due to overvoltage. It is felt that the develop­
of the Joint Subcommittee on Development and Research
ment of a fault from one or 2 phase wires to a greater
of the Edison Electric Institute and Bell Telephone
number of phase wires at more than one point on the
System, all of the companies and individuals therein who
power system is more indicative of the effect of over-
co-operated in securing the data on which this paper is
voltage than is a fault which initiates as multiple line-to
based.
ground and remains so throughout, even though more than
one point on the power system is involved. In the latter
case the faults might well be caused by an electrical References
storm or storms extending over a large area. The per- 1. DISTRIBUTION OF THUNDERSTORMS IN THE UNITED STATES, William H.
Alexander. Monthly Weather Review, volume 52, July 1924, page 337.
* Two line-to-ground faults in which the fault on one phase wire is at a diflFerent 2. N E W SEQUENCE SYSTEM OF POLYPHASE M E T E R S , R. D. Evans. Electrical
point on the system from that on the other phase wire. World, volume 81, September 10, 1923, page 333.

Power System Faults to Ground

Part II: Fault Resistance

By C. L. GILKESON A. JEANNE E. F. VAAGE

MEMBER AIEE MEMBER AIEE Member Am. Math. Soc.

Synopsis to assume zero fault impedance in computing fault cur­

rents. For certain other purposes, however, such as

A
N ALLOWANCE for fault resistance in fault inductive co-ordination problems, it seems preferable to
current computations is desirable in certain types include in the calculations a representative value for
of problems. The present paper gives the results fault impedance or to otherwise allow for its effect on
of a study made to determine reasonable values of fault fault currents.
resistance to use in computing line-to-ground fault cur­ For faults involving ground, fault impedance is generally
rents, particularly in inductive co-ordination studies. understood to include the impedance in the path of the
The sources and nature of the data used in the study are fault current between a phase wire at the point of fault
described in a companion paper. 1 and ground, including local ground resistance. Thus
A paper recommended for publication by the AIEE committee on power trans­
mission and distribution. Manuscript submitted January 5, 1937; released
Introduction for publication February 25, 1937.
C. L. GILKESON is assistant engineer with the Edison Electric Institute, New
York, N . Y.; P. A. JEANNE and E. F. VAAGB are engineers with the Bell Tele­
For many purposes it is sufficient and even desirable phone Laboratories, Inc., New York, N. Y.
as in the case of certain relaying problems, for example, 1. For all numbered references see list at end of paper.

428 Gilkeson, Jeanne, Vaage—Power System Faults ELECTRICAL ENGINEERING

Fig. 1 . Cumulative percent­
NUMBER RANGE OF
age curves and average fre­ POWER OF CASES APPARENT FAULT
CODE SYSTEM CC.M5ΞDERED RESISTANCE-OHMS
quency distribution curve of
—CD B- 33 KV· Ibl -9 TO 8 7
apparent fault resistances
C-33 KV. 89 -9 TO 160
80 E-44KV. 1031 -10 TO 720

—x— 1 -I32KV. 14 8 TO 68
—o — J-220KV. 60 5 TO 132
fault impedance includes 5 2 70 AvroΒr.r CD c α 11 r κι r v M C T D I D-riAU
II moue
the resistance in the arc, in
the contact between con­ z
o 60
NOTE*. THE COMPUTED VALUES OF APPARENT
ductor and ground (in the FAULT RESISTANCE WERE CLASSIFIED IN 5-OHM
case of wires falling on the STEPS,AND SMOOTH CURVES FITTED TO THE RESULTING
DATA
50
earth), in tower footings, in
a foreign object between
< ©
conductor and tower (or : <-> 4 0 r 2.0

ground) or in some com­ AVERAGE FREQUENCY DISTRIBUTION CURVE

bination of these factors. 30 IS
Fault impedance may be
constant throughout the SCALE B

duration of a fault or it 20 I0

may vary continuously, as

in the case of an arc which is I0r 5
blown out to gradual extinc­
tion by the wind. Fault im­
pedance, in general, appears I0 20 30 40 50 60 70 80 100
to be predominantly resis­
APPARENT FAULT RESITANCE -OHMS
tive and in this paper will be
treated as though entirely so.
In this paper the approach to the problem of fault where
resistance is largely statistical and is based on an oscillo-
graphic study of power system fault currents, made by IF = fault current
E = normal generated voltage to neutral
the Joint Subcommittee on Development and Research of Zi = positive-sequence impedance of power system
the Edison Electric Institute and Bell System, in co­ Z 2 = negative-sequence impedance of power system
operation with a number of power companies. The Zo = zero-sequence impedance of power system inclusive of neutral
power systems from which the data were obtained are, in impedance but exclusive of fault resistance and capacitance
to ground
most instances, referred to by a letter, i.e., system A,
RF = fault resistance (used here in preference to ZF to simplify
system B, etc. Information as to the characteristics of explanation)
these systems and the nature and extent of the study on
which this paper is based is given in a companion p a p ^ , 1 In case of accidental faults to ground during routine
The method of attack on the problem of fault resistance operation, IF cannot be measured directly, but in its place
has, in general, been along 3 different lines: a measurement of neutral current at one or more points
can be obtained. If there is but one neutral ground on
1. Measurements of neutral current, supplemented by calculations, the system IN = IF (except for charging current); if
to determine the "apparent fault resistance" for one line-to-ground there is more than one ground the division of the computed
faults occurring under actual operating conditions.
value of IF between the neutrals can be determined.
2. Measurements of phase currents and voltages on a particular Assuming that values of the sequence impedances of the
line to determine more directly the fault resistance under actual
operating conditions.
power system are available or can be computed, that
system conditions at the time of fault including fault
3. Measurements (staged tests) and calculations of certain of the
individual elements entering into or affecting fault resistance.
location are known, and that measured values of one or
more neutral currents under fault conditions are available,
a value of RF can be determined which will make the
Determination of Apparent Fault Resistance computed value of IN equal to the measured value of
THEORY AND LIMITATIONS
neutral current.
The value of RF so obtained is of the nature of a "cor­
For grounded systems, the formula for computing one rective resistance," since it includes in addition to fault
line-to-ground fault currents by the method of symmetrical resistance, the effects of differences between the actual
components is: and assumed constants of the system. Therefore, values
of RF determined in this way are referred to herein as
3E "apparent fault resistances." While this method may
I, = (1)
Z\ -f- Z% -\- ZQ + SRF not give a true fault resistance, it provides a statistical

APRIL 1937 Gilkeson, Jeanne, Vaage—Power System Faults 429

Fig. 2. Cumulative percent­ 100

age curves of apparent fault

resistance (or steel-tower line,
wood-pole line, and sub­
station faults <

method of arriving at a
value for fault resistance for
use in practical calculations
of fault current. This 0 . 0
method has been used in < oCL
most of the fault resistance
determinations made for < o
this paper.
In a few of the cases a
larger value of neutral cur­
rent was measured than was
computed under the assump­
tion of a zero-ohm fault re­
sistance, indicating that the
conditions existing on the
power system at the time
of the fault, or some of the
constants, were not accu­ 30 40 SO 60 70 80 90 100 110
rately known. On directly APPARENT FAULT RESISTANCE-OHMS
grounded systems it will be
seen, by reference to equation 1, that the largest fault current Since these curves are similar in shape and lie fairly
magnitude that can be calculated by varying RF (a real close together, an average cumulative percentage curve
quantity), with other things fixed, is that which corresponds wa$-determined from these curves and from it the average
to a negative value of RF equal in absolute value to one third: frequency distribution curve shown in figure 1 was derived.
of the sum of the resistance components of the sequence The. abscissa corresponding to the peak of a frequency
impedances. Any assumed negative value of RF largei distribution curve is its mode, or most frequently occur­
in absolute magnitude than this would decrease the com­ ring value. Similar distribution curves were also pre­
puted fault current. In the cases used for this paper, pared for the individual systems. The modal, median,
one third of the sum of the resistance components of the and average values of apparent fault resistance applying
sequence impedances seldom exceeded 5 ohms for the to the individual systems and the values from the average
directly grounded systems. Of course, in cases where a curve offigure 1 are shown in table I.
neutral resistance is used, a negative value of RF much Systems B and E contain a considerable amount of
larger in absolute magnitude becomes mathematically both steel tower and wood pole line construction. The
possible. Negative values of apparent fault resistance, wood pole lines are not generally of the type recently
wherever indicated by the calculations, have been in­ advocated to take advantage of the wood insulation and
cluded and given equal weight in the analysis, although it some of them carry ground wires. The apparent fault
is realized that physically they do not exist. The meas­ resistance data for these systems were classified in accord­
ured current for 2 cases from system C, one case from ance with the type of line to which the individual cases
system / and 2 cases from system / so greatly exceeded applied; also all apparent fault resistances deduced from
the computed values that no figure for apparent fault substation faults were tabulated separately. Figure 2
resistance could be deduced and these cases were excluded shows the resultant cumulative percentage curves based
in presenting the results. on these classifications. For the respective systems, the
curves applying to substations are much lower than those
RESULTS OF APPARENT
applying to lines of either type. While the curves for the
FAULT-RESISTANCE DETERMINATIONS
steel tower lines on both systems indicate somewhat
During some 6 years of oscillographic observations on smaller apparent fault resistances than those for wood
several power systems, a total of 1,375 records were ob­ pole lines, particularly at the higher values, the modal
tained for which the conditions existing on the power values for these 2 types of lines are not very different for
system at the time of fault were known with sufficient the respective systems.
accuracy to permit computations of apparent fault re­ Figure 3 shows average curves, based on a classification
sistance. The results of these computations for the 5 between apparent fault resistances deduced for faults on
systems for which the data were sufficient to warrant lines and at substations for systems for which such classi­
curves are shown in figure 1. fication seemed warranted. Due to the fact that wood

430 Gilkeson, Jeanne, Vaage—Power System Faults ELECTRICAL ENGINEERING

Table I—Modal, Median, and Average Values of Apparent tance relaying.2 In order to obtain measurements of re­
Fault Resistance quired currents and voltages, at least one 6-element oscil­
lograph is required at each end of the line.
Apparent Fault Resistance Considerable difficulty was experienced in finding lines
Number suitable and available for fault impedance measurements
System of Cases Range Mode* Median** Average
of the type described above. As a consequence, only one
B— 33 kv . . . 181 . . - 9 to 87 8 22 . . .25 oscillograph installation which would provide data for this
C— 33 kv ... 89 .. - 9 to 160 24 28 ...30
£ — 44 kv . . . . .1,031 . . - 1 0 to 720 13 23 .. .57 phase of the study was made, namely, that on system D.
/—132 kv 14 8 to 68 15 28 ...30 The oscillographs were located at each end of a 64-mile
J—220 kv . .. 60 5 to 132 25 31 ...34
.. - 10 to 720 19 25 . . .35 line. During the 30 months this system was under
observation, 78 records of faults were obtained but only 7
* Value expected to occur most frequently.
** The median of a series of values is the midpoint of the series when arranged were suitable for fault resistance determinations. The
in order of magnitude.
results are given in table II.
From the tabulation it is seen that the fault resistances
determined from the 2 equations check reasonably well but
pole lines were not of special design and frequently carried that the fault locations are much less reliable, particularly
ground wires, and also to the relatively small differences for the higher values of fault resistance.
between the curves for wood pole and steel tower lines While these data are not extensive enough to add much
just referred to in the case of systems B and E, values to the information on fault resistance, the results are
deduced for faults on all types of lines were lumped to­ comparable with those given in figure 1. The observa­
gether in the average curves of figure 3. They show 18 tions indicated certain practical limitations in this method
ohms as the most frequently occurring value of apparent of measuring fault resistance. One difficulty encoun­
fault resistance for lines, and 5 ohms for substations. tered was that of adjusting the sensitivity of the oscillo­
Most of the data on apparent fault resistance were graphs to give proper deflections for faults at any point
computed from values of neutral current measured 3 or 4 on such a long line. Another was that of obtaining
cycles after the fault started, which, in the majority of sufficiently accurate measurements of phase angles from the
cases, allowed sufficient time for the disappearance of the oscillograms. Also, with the sensitivity required to record
unsymmetrical transient, if present. However, fault fault currents, it was practically impossible to determine
resistance frequently varied from cycle to cycle and in the magnitude of the load currents in the non-faulted phases,
some cases, particularly on systems operating with which may be a source of some error in the calculations.
neutral resistance, the fault
cleared before breaker
operation.

Direct Measurement
of Fault Resistance
A method of measuring
fault resistance during acci­
dental faults on a particu­
lar line has been suggested
by W. A. Lewis as the re­
sult of his studies of dis­
tance relaying problems.
The method requires the
measurement, at each end
of the transmission line
under study, of the phase
currents and the line-to-line
or line-to-ground voltages,
and the phase relationships
among these quantities.
From these measurements
fault resistance and the lo­
cation of the fault can be
determined by solving, after 30 40 50 60 70
some rearrangement, equa­ APPARENT FAULT RESISTANCE-OHMS
tions developed in connec­
tion with the theory of deter­ <
mining fault location in dis- Fig. 3. Average curves of apparent fault resistance; lines and substations

APRIL 1937 Gilkeson, Jeanne, Vaage—Power System Faults 431

Table II—Fault Resistances Determined by Direct Measure­ mining voltage per foot of arc, as no special means of
ments determining arc length were used. The conditions under
which certain of these incidental tests were made and
Computed Fault Location—
the over-all arc resistances measured at various stages of
Miles From Station G Fault Resistance—Ohms the arc are shown in figure 4. Some additional data on
Case arc resistance and of the arc and contact resistance of a
(Avg.) (2)* (Avg.)
No. (D* (2)* (D* conductor dropped on the ground are given in table I I I .
Some arc resistance data were also derived from a series
(a) ...40... ...25... . .33 .. ..16.. .12
(6) ...65... ...22.. . ...44 . . . .45.. .45 of relay tests made on system J. The faults were initiated
to ...44... . . 4 3 . . . ...44 .. . . 7.. . 7
by pulling a string, saturated with a salt solution, across
(«0 ...20... ...22... ...21 . . . . 7.. .22
teV** ...28....
...28....
,36
. .32 .
the arcing rings of an insulator string. The approximate
if): **
ω ...28.... .23 lengths of the arcs were determined from moving pictures
* These values were obtained from 2 equations, each of which could be resolved taken with one camera. The voltage gradient derived
into real and imaginary components and solved for fault location and fault
resistance and the averages taken. The values under (1) and (2) give the
from the measurements is shown in figure 5 by the small
results for the individual solutions. squares in the range between 2,000 and 3,000 amperes.
** Stub-end feed, fault location known. (/) and (g) are reclosures on fault of
case («). An average arc voltage gradient of 360 peak volts per
foot was deduced from these tests.
A more detailed study of the characteristics of arcs was
Table III —Miscellaneous Arc and Fault-Impedance Tests made in another set of tests in which representatives of the
joint subcommitee participated, the results of which have
Minimum Length Fault Computed Fault been given in an AIEE paper. 3 In those tests, arc length
of Arcing Current- Resistance—
Type of Test Path—Inches RMS Amps Ohms was determined from moving pictures of the arc taken
with 2 cameras set at right angles to each other. It was
(1) From a 132-kv system—tests of 1928 found that for arc currents between 100 and 800 amperes
Arc across insulator 52 ( a p p r o x . ) . . . . 359 8.7
Arc across insulator 52 ( a p p r o x . ) . . . . 348 9.4 peak, 67 per cent of the measurements of arc voltage fell
Length in Contact
With Ground—Feet
Conductor dropped on CURVE NUMBER ~|
wet ground 500 (approx.) 356 9.0
Conductor dropped on 1 Zl 3 4 45 6
wet ground 500 (approx.) 365 9.6 ARC ACROSS ARC ACROSS
Resistances computed from readings taken at sending end of line. NATURE OF TEST INSULATOR ARC BETWEEN PIN
STRING CONDUCTORS INSULATOR
(2) 33-kv system—tests of 1927
Conductor dropped on ARC STARTED BY*. FUSE FUSE FUSE
dry ground 5 180 (approx.) 45 SYSTEM KV: no 33 33
Conductor dropped on MIN. ARCING DISTANCE 74" 36" 10"
dry ground 45 200 (approx.) 36
Conductor dropped on CURRENT RANGE
dry ground 100 200 (approx.) 33 300-440 40-150 390-450
(ΡΕΑΚ-Γ>/1Γ)

TO 260 OHMS
AT 42. CYCLES

Direct Measurement of Certain Items

Entering Into Fault Resistance
In the course of the fault resistance study information
on certain of the items entering into or related to fault
resistance has been secured. While data on all items are
not available, those given may be of interest in connection
with certain types of problems.

ARC RESISTANCE

Certain investigations have indicated that for fairly

large currents, the voltage gradient in an arc is independ­
ent of current. Consequently, if the voltage per foot of
arc can be determined and the length of the arc path is
known or can be approximated, the arc resistance is given
immediately by :

arc voltage gradient X length of arc

arc resistance =
fault current
8 12 16 20 24 2Θ 32
TIME IN CYCLES FROM INITIATION OF ARC
From time to time advantage has been taken of oppor­
tunities to determine arc resistances and arc voltage Fig. 4. A r c resistance

gradients experimentally. Certain of these arc measure­ Results based on arc-voltage and fault-current measurements
ments, of an incidental nature, were not suitable for deter- during staged tests on transmission lines

432 Gilkeson, Jeanne, Vaage—Power System Faults ELECTRICAL ENGINEERING

Fis- 5. Volts per foot of arc CODE SOURCE OF DATA |
versus arc amperes a SYSTEM J - 220 KV
*Total initial arc voltage across pin type • TABLE I - ACKERMAN PAPER^
insulators for currents ranging From 7 0 to + TABLE E - ( I * * CYCLE) ACKERMAN PAPER W

1 0 , 0 0 0 amperes as given in Ackerman's o TABLE Π - ( 6 l * CYCLE) M

* |
paper were:
1 NOTES'. APPARENT ARC LENGTHS WERE DETERMINED FROM MOVING PICTURES* 1
2.2-kv pin insulator—250 root-mean- VALUES OF ARC VOLTAGES AND CURRENTS ARE PEAK VALUES DIVIDED BY ^ Ί Γ .
700
-square volts +
27-kv pin insulator—500 root-mean- o
square volts
600
o < ^
60-kv pin insulator—900 root-mean- o o
square volts 500 " l¥
+
+ %V +
within =*= 20 per cent of 300 peak 2 6 400 + «9*
troc
volts per foot of arc length. Also U J < f l ° + P «
Γ o 0
c
53 per cent of the cases were above Q. + r,

1 °+
ω£ 300 + o +
300 peak volts per foot and 47 per o OH-
1 a +
cent were below this value. Con­ g 2: 200 +
sequently, the average voltage + 4
o
gradient in an arc for currents too τ σ
from 100 to 800 amperes peak o

is concluded to be about 300

50 I00 200 300 500 1000 2000 3000 5000 10000
peak volts per foot.
ARC CURRENT-AMPERES
Investigations by Paul Acker-
man 4 have given valuable test
data regarding the relationship between arc voltage, on lines having ground wires, may be neglected without
arc current, and arc length, the latter being determined introducing appreciable error in fault current calculations
from moving pictures of the arc (taken in one plane only). if /, the length in miles of the section of line under con­
Arc lengths determined in this manner are generally sideration, is greater than
greater, particularly in the case of small pin type insula­ f v average tower footing resistance \
tors, than the striking distance across the insulator. \ number of towers per mile /
Results from these tests, in so far as they relate to the
voltage per foot of arc, have been plotted as a function of This conclusion is based on a study of the effect of tower
current in figure 5. On the basis of his tests Ackerman footing resistance on the zero-sequence impedance of
reaches the following conclusion regarding arc voltage: power lines by a method given in an engineering report
"The arc voltage is thus chiefly dependent on the arc of the Joint Subcommittee of Development and Research,
length. . . . For practical purposes it may be assumed that E.E.I, and Bell System. 6
for long stretched arcs, the arc voltage will be approxi­
mately 400 root-mean-square volts per foot length of arc, Conclusions
irrespective of short-circuit current.''
The data from the various investigations cited above (a) The most frequently occurring values of ' 'apparent
are in reasonably close agreement with regard to the fault resistance" for the systems studied ranged from 5 to
voltage gradient of an arc, considering the necessary 25 ohms. Those determined for faults at substations
approximations in arc length. The range in average were generally less than for faults on lines; those deter­
value of gradient is from 200 to 400 root-mean-square mined for faults on steel tower lines tended, for the 2
volts per foot of arc. (While root-mean-square values systems for which such separation could be made, to be
are quoted, it should be realized that arc current or somewhat smaller than for faults on wood pole lines, but
voltage waves are usually distorted. Peak values divided taking the results as a whole there appears to be no
by \/2 were used in analyses.) marked difference for these 2 types of construction.
The data indicate that, where fault resistance is to be
TOWER FOOTING RESISTANCE allowed for in fault current computations, 20 ohms for
For faults to towers on lines having no ground wires, a line and 5 ohms for substation ground-faults are reason­
single tower or pole footing resistance enters into the able values to use.
fault resistance, and unless data are available on footing (b) The direct measurement of fault resistance during
resistances, the amount to allow is problematical. For accidental faults, while possible, is subject to numerous
lines provided with ground wires, the shunting effect of practical difficulties which make the accumulation of data
the other towers along the line becomes a factor and the by this method slow and rather expensive.
importance of tower footing resistance in influencing fault (c) Fairly comprehensive information on arc resistance
current diminishes. It appears that, in a majority of indicates that, above about 100 amperes arc current, the
cases encountered in practice, tower footing resistance (Concluded on page 474)

APRIL 1937 Gilkeson, Jeanne, Davenport—Power System Faults 433