You are on page 1of 8

Arcing-Fault Models for Low-Voltage Power Systems

Tammy Gammon, Member, IEEE John Matthews, Senior Member, IEEE


NC State Engineering Program at UNCA John Matthews and Associates, Inc.
P. 0. Box 18071, Asheville, NC 28814 P. 0. Box 3229, Cookeville, TN 38502

Abstract - An arcing fault is a dangerous form of short I. INTRODUCTION


circuit that may have a low current magnitude. In the
case of such faults, the magnitude of the current is Arcing faults have been recognized as a potential
limited by the resistance of the arc and may also be hazard in low-voltage systems as far back as the 1920s, but
limited by the impedance of a ground path. This lower the number of arcing-fault events began to rise in the
level fault current is often insufficient to immediately 1960s as electrical systems changed to meet greater load
trip overcurrent devices, resulting in the escalation of demands. At service entrances, voltages shifted from
the arcing fault, increased system damage, tremendous 208Y/120 V to 480Yl277 V systems. A typical trip rating
release of energy, and threat to human life. Despite for an overcurrent device at a building’s main distribution
modem advances in system protection, people are panel shifted from a lower current rating, such as 600 A, to
critically injured or killed each year when they are in a higher rating, such as 4,000 A. The 1971 National
the vicinity of an arcing fault that is either accidentally Electric Code Section 230-95 attempted to reduce the
physically initiated or initiated through a glow-to-arc number of arcing-fault disasters by requiring ground-fault
transition. The initial phase of an ongoing arcing-fault protection for all service entrance equipment rated at 1,000
research project was to review the pioneering work, A or more, when operating more than 150 V to ground [2].
dating back to the 1920s. The historical evolution of arc Unfortunately, the adoption of NEC Section 230-95 did
modeling for low-voltage systems and a summary of the not eradicate arcing faults and the devastating fires and
electrical aspects and the physics involved in arcing human tragedies that result from arcing faults. The article
faults were reviewed in a companion paper 111. After a applied to new construction; buildings constructed prior to
comprehensive literature search was completed, today’s the change were not required to install ground-fault
better analytical tools facilitated the development of protection. Some building service entrances require less
new arc models with current-dependent arc voltages. A than 1,000 A. Although arcing most commonly occus on
current-dependent arc voltage better represents the 480Y1277 V systems, arcs have been known to cause
arcing phenomenon than the assumed arc voltage devastating fires on 208Y/120 V systems [3]. Furthermore,
associated with previous instantaneous arc models. when arc tracking occurs, arcing may occur in single-phase
120 V systems [4]. Therefore, improvements in arc
Key Words-arcing, ground fault, arcing-ground fault modeling for low-voltage systems deepen the electrical
understanding of the phenomenon.
NOMENCLATURE
Gap width (inches) between electrodes. IL TYPICAL MEDIUM INDUSTRIAL SYSTEM
Instantaneousarc current.
RMS arc current. For the purposes of this work, the electrical system of a
RMS available short-circuit current. typical medium-size industrial building with a 480Y1277 V
Arc voltage. service will be used to illustrate the phenomenon of arcing
Supply voltage. in low-voltage power systems. Figure 1 shows a single-line
Peak of sinusoidal supply voltage. diagram of a small industrial power system for fault-
Ratio of arc voltage to peak voltage. current analysis. In Figure 1, locations A, B, and C
Angle of impedance. correspond to the main distribution panel, a sub-panel, and
Conductionangle. a branch circuit, respectively.
Time that arc current begins to flow. The available short-circuit current at locations A, B,
Time. and C within the building system depend on the sum of the
System fresuency (radianslsecond). impedances seen up to the location of the fault. The
impedance magnitude, X / R ratio, and short-circuit current

0-7803-5843-0/00/$10.0002000 IEEE
119
I

UtilayImpdaaoe
The general solution for the arc current is:

MainFeeder I -a ( t - r a ) (2)
(%-sin(ota- 0)). I.

The conduction angle, y, can be determined with the FTND


function in Mathcad from the following expression:

Figure 1 - Typical Electrical System for


Medium-Size Industrial Building

In the first half-cycle, the arc current begins to flow at b,


available at the MDP, A, the Panel, B, and the Branch, C, when the supply voltage equals the restrike voltage. The
are listed in Table 1. Arcing fault currents will be current continues to flow until the conduction angle has
determined for the locations A, Byand C by the Matthews been reached. At this point, the current becomes zero and
model and by the currentdependent model and the results the stored magnetic energy has been dissipated. The arc
will be listed in Section V. current in the second half-cycle is a mirror image of the
current in the first half-cycle. The restrike voltage is
assumed to be 375 V and fhe flat-topped arc voltage is
Table 1 - Short-circuit Currents assumed to be 140 V, but Merent values could easily be
Branch substituted.
20.98 179.25 In Figure 2, a simple circuit illustrates a phase-to-
ground arc at the main distribution panel; the impedance
0.810 0.103 reflects the sum of impedances from the system,
13,205 1,545 transformer, and all conductors up to the location of the
fault, which serve to lower the available fault current. The
predicted arc current and voltage waveforms at the main
distribution panel, a sub-panel, and a branch circuit are
IIL THE MATTHEWS MODEL shown in Figures 3, 4, and 5, respectively. The circuit in
Figure 2 could also represent a phase-to-ground arc at a
Earlier models were reviewed in the companion paper sub-panel or a branch circuit, if 16.30 + j13.20 mn were
[l] and presented in [5]. The Matthews model will be
reviewed here since it is the only previously published
instantaneous resistive-inductive model. In 1993, Z = 0.702+J6.512mSZ
Matthews developed a generalized instantaneousarc model v, = VaK=
for a resistive-inductive system, based on a method for . I I,"
determining a power electronic conduction angle [6]. The
method was described in the companion paper [11, and will
be more thoroughly explored here. An arcing fault in a ( P O V
complex circuit can be described by the following I I
merentid equation:
Figure 2 - Complex Circuit with Arcing Fault

120
3

substituted for the impedance at the panel or 178.30 + solved using a Runge-Kutta numerical method. The first
j18.43 m were substituted for the impedance at the model incorporates a currentdependent arc voltage based
branch. The ground impedance is assumed to be zero. The on an equation formulated by Stokes and Oppedander [7]
system impedance seen at the main distribution panel is for instantaneous arc voltage. By Kirchhoff s voltage law,
typically highly inductive and much smaller in magnitude the following equation holds:
than the highly resistive branch circuits that typically have
dim,
much larger impedancevalues. ,,V sin(ot) = Rim +L-dt
+ (20 + 0.534 -
25.4 g)i,, .
0.12

For a specific system voltage, a given restrike voltage, (4)


and a known arc voltage, the impact of a changing X/R
ratio may be observed from Figures 3,4, and 5. As the X/R
ratio decreases, the arc current becomes less sinusoidal in
nature and the extinction time between arcing in
consecutive half cycles increases. The lengthened arc
extinction time reduces the likelihood that the arc will be
self-sustaining. Although this model is limited by an arc
voltage assumed to be a known constant., it is an insighthl
tool for aualyzing electrical systems in buildings. Table 2
lists arc duration and short-circuit ratio as functions of X/R
ratio. Arc duration, expressed as the fraction of each cycle
that arcing occurs, increases as the X/Rratio increases and
the system becomes more inductive. The magnitude of
system impedance determines the magnitude of the short-
circuit current. The rms arc current may be determined by
multiplying the short-circuit ratio listed in Table 2 by the
available short-circuit current.

JY. NEW MODELS


0.00000 0.00417 0.00833 0.01250 0.01667
'IIme in Seconds
Two arc models have been developed that define arc
voltage as a function of arc current. The models are -
Figure 4 Arc Current & Voltages at Panel
expressed as differential equations; the arc currents are

-100 -

-300 -
-400 - ResMke Voltage/
1 . 1 . a . i .
O.OOOO0 0.00417 0.00833 0.01250 0.01667
O.OOOO0 0.00417 0.00833 0.01250 0.01667
"e in Seconds TTme in Seconds
-
Figure 3 Arc Current & Voltages at MDP Figure 5 - Arc Current & Voltages at Branch

121
4

-
Table 2 Short-circuit Ratio and Arc Duration produced sustained arcing for rms supply voltages as low
as 146 V (206 V peak) [SI. The exact voltage required
Power
depends on many factors. These include: the surface
Factor Duration condition of the cathode and the presence of conductive
9.950 0.576 0.827 dusts (or other conductive material) to help bridge the gap
0.2 4.899 0.53 1 0.799 [l]. The exact voltage also depends on the time of arc
extinction since arcing last occurred [113. Since voltages
I 0.3 I 3.180 I 0.495 I 0.774 I less than about 350 V are not capable of initiating arcs at
I 0.4
I
I 2.291 I 0.468 I 0.751 I
I
room temperature and atmospheric pressure [12], a 350 V
10.5
I
I 1.732 I 0.448 I 0.728 I
I
restrike voltage has been adopted for the new models.
The Merentid equations in (4) and (5) produce fairly
I 0.6
I
I 1.333
I
I 0.433
I
I 0.704
I
I
I
similar arc current and voltage waveforms. Since the arc
0.7 1.020 0.424 0.678 voltage expression developed by Stokes and Oppenlander
was determined from over 200 digitally sampled arc tests
0.8 0.750 0.422 0.647 with more than two million arc current and voltage data
0.9 0.484 0.43 1 0.605 points [7], the arc current and arc voltage will be predicted
by (4). Figure 6 shows the relative Waveform shapes and
0.95 0.329 0.444 0.571 current magnitudes for arcing faults at the main
0.995 0.100 0.474 0.508 distributionpanel, a sub-panel, and a branch circuit for the
electrical system illustrated in Figure 1. The predicted arc
current and voltage waveforms at the main distribution
panel, sub-panel, and branch circuit are shown in Figures
The second model incorporates an arc voltage based on the 7,8, and 9, reWstiVelY.
product of arc current and Fisher's equation for arc The arc Voltage is now a function Of arc current;
resistance [SI. ~y Kirchhoffs voltage law, the following therefore, it increases and decreases with the magnitude of
equation holds: the arc current. Neglecting the fluctuations associated with
the random and dynamic nature of arcing, the arc current
and arc voltage feach peak values simul&wusly because
V- ~in(ot)
= Rim +L G
dt
+ (25-.&)itt5. (5)

The arc voltage expression incorporated in (4) was derived


from the pseudoanstant portion of an ac signal, i.e., near
the peak The arc currents determined from the differential
40000
30000
7
equations (4) and (5) are significantly more accurate than
the currents predicted by the Matthews model and the 20000
validation results will be given in Section V.
The arc voltage expression incorporated in (5) is
* 10000
actually derived from a method used to determine the rms $
h
arc current [SI. The voltage waveforms graphed in Figures
7,8, and 9 are fairly accurate representations of the stable, 5 0

self-sustaining arcs examined in this paper. The arc


voltage of a stable arc with a fairly large current magnitude -10000
8 I
experiences fairly small changes in voltage magnitude. 8
: I

Although the arc voltages have been shown to be fairly -20000 - I8


1
8
8
t :
#
I

flat-topped [9,10], a larger potential, or voltage transient, \\ I

is sometimes visible at the moment of arc initiation.


8
8.
*/.
':

-30000 m

However, due to the complexity of the transient, the


absence of transient data, and its limited impact on the
-40000
complete arc current waveform, it has been neglected in 0.00000 0.00417 0.00833 0.01250 0.01667
this work. Time in Seconds
In truth, the magnitude of the restrike voltage is not a
constant value. Arcing faults have initiated devastating
fires on 208Y/120 V systems, and laboratory tests have Figure 6 - Arc Currents at Three Building Locations

122
5

f140000

-100 -

-400 - Restlikevoltage
-40000
0.00000 0.00417 0.00833 0.01250 0.01667
Time in Seconds O.OOOOO 0.00417 0.00833 0.01250 0.01667
Time in Seconds
-
Figure 7 Arc Current & Voltages at MDP
-
Figure 9 Arc Current & Voltages at Branch

13000 When the arc voltage is greater than the source voltage, the
magnetic energy stored in the system begins dissipating.
Once the magnetic energy has dissipated, the arc becomes
extinct.
In addition to the current magnitude and not addressed
by the Matthews model, a close relationship exists between
the arc voltage and the distance from the cathode to the
anode (gap width). The arc length is often longer than the
gap width, and the difference tends to increase with
increasing gap width. Unlike the flat 140 V arc voltages
shown in Figures 3, 4, and 5, the arc voltages shown in
Figures 7,8, and 9 have significantly different peak values
of 209.8 V, 143.0 V, and 82.1 V which correspond to 3-
inch, 2-inch, and 1-inch gap widths, respectively.
-400 - Restrike Voltage
. . .
- -11000
I I m m *
-13000 V. COMPARISON OW SYSTEM RESULTS
0.00000 0.00417 0.00833 0.01250 0.01667
Time in Seconds A summary of system calculations predicted by the
Figure 8 - Arc Current & Voltages at Panel Matthews 140 V arc model and the new arc model
incorporating the voltage expression based on the work of
Stokes and Oppenlander are listed in Table 3. At the MDP,
an arc can be considered a purely resistive component. the Matthews model predicts a larger arc current and a
However, the arc resistance is highly nonlinear; large smaller arc voltage. An overcurrent device would react to a
increases in current magnitude produce much smaller larger current magnitude more quickly. The overcurrent
corresponding increases in arc voltage. One improvement device would take longer to react to the smaller arc current
associated with the currentdependent arc voltage is that predicted by the new model. Consequently, the arc is likely
after the arc current has reached its peak value and is to propagate for a longer period of time than indicated by
decreasing, the arc voltage is also decreasing. Unlike the the Matthews model. Arcing is typically extremely
Matthews model, the decreasing arc voltage delays the dangerous and explosive at the MDP because of the high
moment that the arc voltage exceeds the source voltage. level of current magnitudes involved. At both the MDP and

123
6

the panel, the new model predicts that more energy has Fisher. Peak arc currents were used to assess the models’
been released through the arc than the Matthews model ability to predict arc currents, since the true rms currents
predicts. were not measured at the time. Table 4 lists the average
In general, the models predict more congruous results deviation of the peak currents predicted by Matthews
at the panel. h4atthews’ 140 volt arc voltage is only slightly model and the new models with the peak arc currents
lower than the new model’s peak 143.0 V, but the new measured by Stanback and Fisher. Although the model
model predicts a larger arc current and a longer arc based on Fisher’s work more accurately predicted arc
duration. The arc duration contributes to the likelihood currents for the test data, the model based on the work of
that the arc will remain self-sustaining; the longer arc Stokes and Oppenlander is preferred at this point in the
duration predicted by the new model more strongly research because Fisher’s test data is limited and arcing-
indicates that a self-sustaining arc will be produced at the fault data from Fisher, as well as Stanback, was used to
panel than the Matthews model suggests. validate the models. In addition, an: voltages were also
The results of the models deviate widely at the branch. validated in [13], and the new models are better able to
The rms arc current and peak arc voltage predicted by the predict arc voltage [131.
new model are at least 34% higher and lower, respectively,
than those values predicted by the Matthews model. The
new model also predia a longer arc duration than the
Wtthews model. Although arcing is not as critical and Model Stanback’s Data Fisher’s Data
hazardous at the branch, the magnitude of the arc current Matthews 19.82 % 30.61 %
and the duration of arcing contributesto the likelihood that
New #1 (*) 6.59 % 14.81%
the arc will restrike and that it will initiate a fire.
The models were validated in [13]. Arcing-fault test New #2 (**) 5.39 % 12.04 %
data published in [8] and [9] were used to validate the
models. The rms shortcircuit currents for the test systems
ranged from 650 to 41,600 A, power factors ranged from
0.09 to 0.44 lagging, and gap widths ranged from 0.5 to 4
inches [8,9]. Predicted peak arc currents were compared
with the data from 24 self-sustaining tests performed by VI. HARMONICS
Stanback and from 10 self-sustaining tests performed by
At the present time, arcing-fault currents in building
systems are detected through a ground-fault protective
Table 3 - Summary of System Cal ulations system, which detects current flowing to ground. The arc-
fault circuit interrupter (AFCI) cau detect arcing faults in
I MDP residential 20 A circuits [4]. Since much interest has been
20.98 179.25 generated about arcing-fault detection from harmonic
9.272 0.810 components, the harmonic contents of the arc currents at
the MDP, panel, and branch for the typical building system
Isc (A) 42,290 13,205
have been identified from the arc currents calculated with
2 the new model based on the work of Stokes and
Matthews’ model: Oppenlander. The coefficient, c, of the Fourier Transform
IarCASC 0.572 0.421 0.473 was calculatedby the following quation [141:
24,206 5,565 731
140 140
KW 2,771.4 572.9 66.8
0.655 0.509
New model:
0.482 0.658 In equation (6), n is the harmonic number and N is the
total number of discrete current data. The third, fifth,
6,370 1,017
seventh, ninth, eleventh, and thirteenth harmonic
143.0 82.1 percentages at the MDP, panel, and branch are graphed in
679.4 58.7 Figures 10, 11, and 12, respectively. The harmonic
ArcDuration I 0.795 0.723 0.627 percentageswere found by the following equation:

124
7

I MDP h, = 2IC. 1I 0 0 % .
IC1 I
(7)

The branch harmonic magnitudes were generally


sigdicantly larger than those at the MDP and panel. The
MDP and panel harmonics were closer in magnitude than
the panel and branch harmonics. Lower frequency
harmonic currents generated larger harmonic percentages
than higher frequency harmonics.
Total harmonic distortion is a measure of the harmonic
content. The total harmonic distortion, THD, was
determinedby:

I Harmonics

-
Figure 10 MDP Harmonic Percentages

Mathematically,the summation in equation (8) reaches the


I Panel infinite harmonic. However, the fiftieth harmonic was used
for the results listed in Table 5 and is usually satisfactory.
The arc current's total harmonic distortion at the MDP,
panel, and branch for the typical system is listed in Table
5. Arc currents with longer durations had a lower total
harmonic distortion. The THD at the MDP and panel was
in closer proximity than the THD at the panel and branch.
The total harmonic distortion was significantly larger at
the branch, which corresponded to a notable decrease in
arc duration.

Table 5 - THD and Arc Duration

I Harmonics MDP Panel Branch


THD 22.095 29.245 41.934
Figure 11 - Panel Harmonic Percentages 0.795 0.723 0.627
Arc Duration

I Branch
Va CONCLUSIONS
Instantaneousarc current models have been explored in
this work. Arc currents predicted by the Matthews model
are independent of the gap width, since the arc voltage is a
constant value. The Matthews model has been used to
generate a table for predicting rms arc currents based on
the X/R ratio. Although the new models are more
laborious, they can more accuately predict arc currents
since the arc voltage is a currentdependent function.
Harmonics are also calculated for the arc currents
predicted by one of the new models to show the harmonic
I Harmonics contents of the arc currents. Eventually harmonic analysis
may be a viable method for arcing-fault detection in
Figure 12 - Branch Harmonic Percentages

125
8

industrial and commercial buildings. However, such a


device would need to be able to distinguish arcing faults [5] 0.R Schurig “Fault voltage drop and impedance at short-circuit
currents m low-voltage circuits,” ALEE Trans., vol. 60, 1941, pp. 479-
from transients and the harmonics generated by loads. 487.
Furthermore, the random and dynam~cnature of arcing [6] S. B. Dewan and A straughen,Power Semiconductor Circuits, New
will also profoundly impact the harmonic content of the arc YOrtc: J o b Wiley & Sons,1975.
[7] A D. Stokes and W. T. Oppenlander, ‘‘Electric afcs in open ak,”Journal
current. OfPhysiCsD,Vol. 24, NO. 1, 14,1991,pp. 26-35.
[8] Lawrence E. Fisher, “Resistance of low-voltage ac arcs,” LEE.!? Trans.
Ind. Gen. Appl., vol. IGA-6, NovJDec. 1970, pp. 607-616.
[9] H. I. Stanback, ‘‘Predicting damage linm 277 volt single phase to ground
arcing faults,” Conference Record of 1975 Industrial and Commercial
REFERENCES Power Systems Technical Conference,May 5-8, 1975, pp. 15-25.
[lo] Richard R Conrad and Derio Dalasta, “A new ground fault prdeaive
[l] Tammy Gammon and John Maithew “The historical evolution of system for electrical distribution c h i t s , ” LEEE Trans. Ind. Gen. AppL,
arcing-fault models for low-voltage ysk”’ W i c e Record of the vol. IGA-3, May/June 1967, pp. 217-227.
1999 IEEE Industrial and Commercial Power Sy;tems Technical [111 J. Slepian and A P. Strom, “Arcs in low-voltage a-c. networks,” ALEE
Werence, Sparks, Nevada, May 1999. Trans.,vol. 50, Sept. 1931, pp. 847-853.
[2] J. R Dunki-Jacobs, ‘The effects of arcing ground faults on low-voltage [12] Bemard Beland, “Examination of electrical conductors following a fire,’’
system design,” IEEE Trans. Ind. Applicat., vol. IA-8, MaylJune 1972, Fire Technology, Vol. 16, No. 4, November 1980, pp. 252-258.
p. 223-230. [13] Tammy Lea Gammon, Improved Arcing-Fault Current Models for
[3] R H. Kaufinann and J. C. Page, “Arcing fault protection for low-voltage Low-Voltage Power Systems (<l kv, Dissertation at the Georgia
power distribution systems- nature of the problem,” ALEE Trans., June Instituteof Technology, August 1999.
1960, pp. 160-167. [14] John H. Matthew Introduction to the Design and Analysis of
[4] George D. Gregory and Gary W. Scott, “The arc-fault circuit interrupter: Building Electrical Systems, New York: Van Nostrand Reinhold, 1993.
an emerging product,” IEEE Trans. Ind. Applicat., vol. 34, Sept./Oct
1998, pp. 928-933.

126

You might also like