1632 IEEE Transactions on Power Apparatus and Systems, Vol.PAS-98, No.
5 Sept/Oct 1979
PRESSURE WAVES DUE TO ARCING FAULTS IN A SUBSTATION
Michel G. Drouet Frangois Nadeau
Direction Sciences de base
Institut de Recherches de l'Hydro-Quebec
Varennes, Quebec, Canada.
ABSTRACT: The collapse of the building housing a large an arcing fault and we will give on expression for the
substation following an arcing fault motivated the mea- dependence of the magnitude of the pressure wave on two
surements of the amplitude of the pressure wave genera- characteristics of the substation: power available and
ted by an AC arc with current from 10 A to 80 kA and response time of the protection.
with arc length from 8 mm to 15 m, burning in air for
up to one second. At low arc power there is a corres- VARIATION OF THE ARC LENGTH AS A FUNCTION OF TIME
pondance between the pressure amplitude and the rate of
change of the power which is not true for high arc pow- The tests were conducted in the high power labo-
er due to absorption. However, the pressure is very ratory at IREQ. The AC arc was produced between two
large and should be taken into account in the design of vertical copper electrodes located at a distance of one
substations. An empirical formula is given relating the meter from each other. It was initiated using a small
pressure variation amplitude to the short-circuit cur- copper wire short-circuiting the electrodes. The arc
rent and the response time of the protection system. length, iniatially one meter, increased with time to
reach approximately 15 meters after one half second. A
INTRODUCTION high speed camera was used to photograph the arc at 500
frames per second. These recordings of the shape of the
Large energy dissipation can lead to the produc- arc were used to determine the time evolution of the
tion of intense pressure waves. In substations a high arc length. Measurements were performed for arc current
rate of energy release is possible under fault condi- between 1 and 20 kA and arc voltage up to 18 kV. The
tions. Several investigations of the pressure generated high speed movie film revealed that the arc is burning
by faults in electrical equipment, such as oil and gas in the hot luminous gases produced in the electrode re-
filled transformers, have been reportedl2. The growth gion and that there is very little change in the shape
of pressure in a small enclosure containing an electric of the arc during a half-cycle; this means that the JxB
arc has also been studied3'4. In addition, it has been force acting on the arc does not control the arc length
shown recently5 that for low values of the arc power, even for a current as high as 20 kA; on the contrary,
the amplitude of the sound wave generated by a free bur- it appears that the arc extends upwards only as far as
ning arc is proportional to the rate of change of the the rising hot conducting gases: the arc length is
electrical power dissipated in the arc. therefore mainly controlled by the convection of these
gases. In agreement with this observation, we found
The destruction of the structure of a large sub- that we may consider,as a first approximation,that the
station following the operation of an open-air-switch total arc length is a function of time only over the
in a circuit carrying a current of tens of kA motivated range of current studied. For our conditions of a free
the present study; the collapse of the building was at- burning AC arc on copper electrodes and for an arc du-
tributed not to a pressure build up in the building but ration of about 0.5 second, we found that the total arc
rather to the effect of successive pressure waves ori- length L can be expressed approximately as follows:
ginating in the electrical arc over several 60 Hz pe- L (in meters) 30. t (in seconds)
riods. During that time interval, the length of the arc
would have increased and with it the amplitude of the AMPLITUDE OF THE PRESSURE WAVE
pressure wave produced by this arc. For these reasons,
we had to determine the time variation of the length of For the determination of the amplitude of the
a free burning arc for values of the current between pressure wave produced by the electric arc, we used a
1 and 8OkA and the amplitude of the sound pressure wave sound level meter, Bruel and Kjaer type 2209. The dyna-
generated by the arc for the same range of current and mic range of the microphone used (B and K 4138) extends
for arc lengths between 1 and 15 meters. The magnitude to 175 db with a flat frequency response from 20 Hz to
of the pressure, obtained here, seems to indicate that 70 kHz (± 1.5 db).
the arc could have caused the damage to the substation.
Measurements of the amplitude of the sound wave
were performed for arc length of 1.5, 3 and 6 meters
The results presented here, should help the desi- while the arc current was varied from 1 to 80 kA. The
gner evaluate the magnitude of the pressure wave which duration of the arc in this series of tests was only of
could be generated in a given substation in the case of one and a half cycles. Measurement of the pressure wave
for longer periods would have been complicated by the
contribution, to the sound pressure at the position of
the detector, of waves reflected by the walls of the
laboratory: i.e. one and a half cycle corresponds to
the time of arrival at the sound detector of the first
reflection from the wall.
F 79 172-8 A paper recommended and approved by the The initial arc lengths of 1.5, 3 and 6 meters
IEEE Substations Committee of the IEEE Power Engi- were obtained using a small copper wire 1.5 to 6 meters
neering Society for presentation at the IEEE PES Winter in length and mounted on an insulating support of a
Meeting, New York, NY, February 4-9, 1979. circular shape. As illustrated in one of the inserts in
Manuscript submitted September 5, 1978; made available Figure 1, the sound detector was located on the axis of
for printing November 15, 1978. this circular arc and at a distance of 3, 6 or 9 meters
0018-9510/79/0900-1632$00.75SD 1979 IEEE
1633
from the plane of the arc as indicated on the diagram. tantaneous values of A and P were recorded and we found
The product of the distance d by the amplitude A of the that the low (120 Hz) and high frequency components of
r.m.s. value of the pressure variation is plotted in both A and P correspond; this correspondance is well
Figure 1 against P, the corresponding r.m.s. value of illustrated in the oscillogram of Figure 2, which was
the rate of change of the electrical power dissipated recorded for an 8 mm, 20 A, 100 V AC arc.
in the arc, Prms = w * Prms w . (U.I)rms.
These last results obtained at very low arc power
DISCUSSION emphasized the departure from linearity observed for
high electrical input ti the arc. This departure from
We observe that the amplitude of the pressure wa- the relation A = AO(P/Po) is attributed to non-linea-
ve is approximately proportional to the rate of change rity in both the production and the propagation of the
of the arc electrical power. Furthermore, it is inter- sound wave. With regard to the production of the sound
esting to note that the measurements corresponding to wave Fitaire et al. have shown that, for the same rate
the different values of d are such that the product A.d of change of the electrical input, the amplitude of the
corresponding to the same P has the same value; this sound wave varies inversely with the arc temperature.
indicates that, over the range of distance studied Thus the increasing departure from linearity shown in
1.5 m to 6 m the amplitude varies as l/d, this is due Figure 1, for increasing values of the electrical in-
to a spherical expansion of the wave and thus it indi- put, could be explained partly by an increase of the
cates that the absorption of the wave in the air is arc temperature with increasing arc power. Howeyer we
negligible. We note also that the value of the standard believe that the departure from linearity must be main-
atmospheric pressure being 105 N.m-2, the maximum pres- ly attributed to absorption of the sound wave near the
sure variation monitored, which corresponds to 104 source. As noted earlier, this absorption phenomenon is
N.m-2 at one meter from the arc, is an appreciable not observed over the range of distance of propagation
fraction of the atmospheric pressure. studied, however, near the source, the amplitude of the
pressure variation will be very much larger and as the
The relation between the amplitude of the pressu- absorption coefficient is given by a = w . A/ao , we
re variation A generated by a variation P of the elec- expect the absorption near the source to be larger for
trical power in an arc has been studied by Fitaire and higher value of the electrical power thus accounting
his co-workers. For a pulsed arc they found, in agree- for the dependence A = AO(P/Po)) 9 observed. The ex-
ment with their theory, that A is proportional to P pression given above for a indicates that the absorp-
for small variation of A relative to the atmospheric tion depends not only on the amplitude A but also on
pressure. We verified this relation in the case of a the frequency w of the wave. This is well illustrated
low power AC arc, burning between electrodes separated by the recordings of Figure 3 and 4 which correspond to
by only 8 mm, as shown in the second insert in Figure 1. the time variation of the sound pressure at a distance
We found that indeed A is proportional to P; the ins- d of 3 meters for an arc of 6 meters in length with a
E
z
4Q42
A (W S1) No
Fig. 1. Value of A.d, the product of the amplitude A of the pressure wave by d the distance arc-mi-
crophone, as a function of P, the rate of change of the electrical power in the arc for
two differents experimental conditions.
1634
I I I is less powerful and in agreement with the expression
given above the attenuation is less; therefore, the
800 frequency content of the variations of pressure monito-
t 400 red in this case correspond more to those present at
the source.
_. 0 25 E
APPLICATION TO THE DESIGN OF A SUBSTATION
i -400 12.5
- 800 0 < The results of our study show that the applica-
.*L IE tion of the linear relation between A, the amplitude of
-12.5 ;z the pressure wave, and P, the rate of change of the po-
-25 4 wer dissipated in the arc following a fault, would lead
I I I I I I I the designer to overestimate the magnitude of the pres-
sure. As an example for the 100 kA, 10 kV arc of 6 me-
2ms/div. ters in length (B on figure 1) i a 3.6 101l W S-1 and
Fig. 2. Simultaneous time variations of the rate of
', the sound pressure at a distance d & 1 m is approxima-
change of the electrical power in a small AC tely 2.104 N. m-2 while the application of the linear
vertical arc (8 mm in length, 20 A, 100 volt) relation A - AO(P/P,)l yields a value of approximately
and of A the amplitude of the pressure wave
monitored 20 mm from the arc axis.
3.105 N. m-2 i.e., one order of magnitude larger than
the measured value. As discussed earlier the difference
between the two values is mainly attributed to a large
absorption in the vicinity of the arc.
None the less, the magnitude of the pressure is
considerable;for example, the force applied on a wall,
4 x 4m2 in area, located 1 m away from the arc would be
approximately 3.105 N (60,000 lb). It is also of inte-
rest to compare the magnitude of the sound pressure at
one meter resulting from this high power arc, 2.104
N. m-2 (i.e., a 400 lb/ft2), to a typical value for the
wind pressure considered in the design of building
which is of 35 lb/ft2. Furthermore, and in contrast
with the pressure exerted on a structure by the wind,
the pressure resulting from the arc varies with time at
120 Hz and in position, the wavelength being a 3 me-
ters; these rapid variations and non-uniformity of the
5ms/div.- pressure should lead to a dislocation of the structure.
Fig. 3. Time variations of the amplitude of the pres- Due to the wave nature of the pressure variation gene-
sure wave monitored 3 meters from a 6 m circu- rated by the electric arc, it is difficult to derive an
lar AC, 1 kA r.m.s. arc. exact relation giving the magnitude of the pressure as
a function of the short-circuit power in a substation.
However, we found that the following expression,
A (kN. mi2) 1.5 I (kA) . t
R (mn)
(s)
relates approximatively, in our case, the amplitude A
of the pressure wave at a distance R from the source
(R > 1 m) to the value I of the arc current and to t,
the arc duration, i.e., the response time of the pro-
tection.
ACKNOWLEDGEMENTS
Special mention is made of M. Roy, who directed
the test at the IREQ high power laboratory, and of J.G.
Bibeau, for the high speed recording of the arc motion.
5 ms /div.
Fig. 4. Time variations of the amplitude of the pres- REFERENCES
sure wave monitored 3 meters from a 6 m circu-
lar AC, 70 kA r.m.s. arc. 1. M.A. Nettleton, Electrical Review p.116, July 1976.
2. R.J. Ringlee and Roberts N.W. Trans. AIEE TP. 59,
current of 1 kA r.m.s. and 70 kA r.m.s. respectively. 1168 (1959).
The frequency content of the electrical power input was
observed to be very similar in both cases; we believe 3. K.K. Schwarz, Inst. Elec. Eng. Part A Power Eng.
that the frequency content of the sound wave at the 109, 151 (1959).
source are also similar, the observed difference in the
two recordings being attributed to higher attenuation 4. D.K. Holway, MS Thesis, Mass. Inst. Techn. (1947).
in the case of the more powerful arc, Figure 4. The at-
tenuation of the high frequency components being high- 5. M. Fitaire, T.D. Montei, Phys. Fluids,15,464(1972).
est in this case - Figure 4 we obtain a variation of
-
H. Dadgar, M.S. Thesis, Orsay, France (1977).
H. Dadgar, M. Fitaire, A.Pilorget, to be published.
pressure corresponding to the lower end of the frequen-
cy of spectrum, 120 Hz, as it is the least attenuated 6. R.B. Lindsay, Mechanical Radiation, Mc Graw Hill
component. In the case of Figure 3, the pressure wave (1960).
1635
Francois Nadeau, photograph and biography not available at the time
of publication.
Michel G. Drouet graduated in Mechanical
Engineering from the "Ecole Superieure d'In-
genieur Arts et Metiers" in Paris, France, in
1961. He received the M.Sc. from MIT in 1962,
the Ph.D. in physics from the University of
IA
Montreal in 1967 and the D.Sc. in physics from
the University of Paris, France, in 1978.
He worked as a research Engineer at MIT in
1962-63 and as a senior teaching assistant at the .
University of Montreal from 1963 to 1967. In
1968-69, he was on a research fellowship at the
University of York, England, when he joined the Hydro-Quebec In-
stitute of Research, Varennes, Canada, where he has been engaged,
since 1970, in the laser triggering of high voltage multi-spark gaps and
the investigation of high current arcs in circuit breakers. 5)S
Discussion
I' l
Michael A. Nettleton (Central Electricity Research Laboratories, Fig. I High Frequency Oscillations of Current in an Arc Powered by
Leatherhead, Surrey, U. K.): Energy dissipated in an electrical fault can a Capacitor Discharge
produce either a blast wave or a presure wave. It is important to be able C = 0.1jF V1 = 4.0 kV
to predict which type of wave is formed, since the resultant damage to Writing Speed 20 m ps-'
surrounding structures depends on the nature of the wave. The present
and very interesting paper applies the Fitaire & Montei [1] analysis of
the production and amplification/attenuation of pressure waves, cylindrically expanding fronts. With the much larger arc, most of which
resulting from a number of cycles of power, in a plasma. The authors is well away from the influence of the electrodes, the microphone is like-
quite properly justify this by an appeal to the type of damage produced ly to be in a region where the fronts are expanding spherically.
in a substation. Furthermore, they show a convincing experimental rela-
tionship between the amplitude of the pressure wave and P, based on REFERENCES
the slow oscillations of power over an approximate range of 108 to 10"
W s-' m -'. [1] M. Fitaire and T. D. Montei, Phys. Fluids, 15, 464 (1972)
Nettleton [2] has proposed that blast-waves are the source of [21 M. A. Nettleton, Electrical Review, p. 116, July 1976
damage from electrical faults. In support of this view are the predic- [3] M. N. Plooster, Phys. Fluids, 14, 2111 and 2124 (1971)
tions of Plooster [3] on the strengths of shock produced by a linear [4] M. A. Uman, A. H. Cookson and J. B. Moreland, J. App. Phys.,
discharge of various energy releases per unit length of discharge. Such 41, 3148 (1970)
predictions were shown by Uman, Cookson & Moreland [4] to describe
adequately the strength of shocks from 4 m arc, dissipating 2 x 104 J Manuscript received January 25, 1979.
with P initially -10'4 rising to -10'1 W s-' m-', provided that the tor-
tuosity of the arc were taken into account.
Thus it would be interesting to find more accurately the value of Michel G. Drouet and Francois Nadeau: We appreciate very much the
for the onset of shock formation (10" < P.,,, < 10'1). This need is efforts of the discussor in clarifying and expanding on a few points in
emphasized by the presence of high frequency oscillations on the history our paper; his comments are a welcome contribution.
of P shown in Fig. 2 of Drouet & Nadeau. Such oscillations appear to As pointed out by Dr. Nettleton, the expansion of the acoustic
be a common feature of electrical discharges, probably due to temporal wave from the 8 mm arc was assumed to be spherical a few cms from
variations in the capacity, inductance and resistance of the arc. Fig. I the source; this assumption was based on an observed 1/R dependence
shows the very high frequency oscillations of current observed by the of the amplitude of the wave.
author, in an undamped arc powered by a capacitor. They could well The objective of the present study was the measurement of the
represent sufficiently high values of P for the formation of a blast- pressure wave generated by a 60 Hz high power arc. However, as il-
wave. There must be some doubt about whether the frequency response lustrated in our fig. 3, high frequency components are generated in the
of the microphone used by Drouet & Nadeau is sufficiently high to arc in addition to the expected 120 Hz main wave. As indicated by Dr.
record satisfactorily such blast-waves. Nettleton, these high frequency fluctuations of pressure originate in the
One further point concerns an alternative explanation of the high frequency oscillations of P in the arc. Our present study is being
discontinuity in the Ad versus P plot (Fig. 1, Drouet & Nadeau), bet- extended to include the systematic study of these much higher values of
ween the results for an 8 mm discharge and those for the much longer P using high voltage capacitive discharges and high frequency modula-
one. The arc represents a line source from which the resultant compres- tions of the arc current.
sion fronts will initially expand cylindrically. With the microphone only
a few cms from an 8 mm arc, it is probable that it is still responding to Manuscript received April 2, 1979.
