1902

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 47, NO. 4, JULY/AUGUST 2011

Arc-Flash Analysis Approaches for Medium-Voltage Distribution

Tom A. Short, Senior Member, IEEE
AbstractArc ash is an important consideration for personnel safety. This paper explores available ways to evaluate arc-ash severity for primary distribution systems, both in equipment and open air. Results from staged tests and from utility monitoring data suggest the need to use longer arc lengths than normal as input into open-air arc-ash calculations. Results from tests also show how fast arcs can lengthen and evolve to multiple phases, and the performance of conductor covers was evaluated. Index TermsArc ash, personnel protection, power distribution, safety.

I. I NTRODUCTION RC FLASH from faults on distribution circuits is a safety issue that can impact work practices, protection requirements for line and substation workers, and relay and other overcurrent protection settings and practices. In this paper, we discuss approaches to analyzing arc ash on medium-voltage utility distribution equipment. The severity of an arc-ash event depends on many factors, including the worker position relative to the fault, the fault duration, the fault current magnitude, and the gap spacing and the arc length as it bridges the gap through the air. Fig. 1 shows a staged test for a fault initiated on the left side of the pole. Heat from the arc is released as radiation, and that creates a very hot reball. Burning hot metallic pieces are also expelled from the arcing site. Arc voltage is an important component of the energy that develops during an arcing fault. The voltage across an arc remains relatively constant over a wide range of currents and arc lengths. The voltage across an arc ranges between 25 and 40 V/in (1016 V/cm) over the current range of 10080 kA [1], [2]. The arc voltage is somewhat chaotic and varies as the arc length changes. More variation exists at lower currents. As an illustration of the energy in an arc, consider a 3-in (7.6-cm) arc that has a voltage of about 100 V. If the fault current is 10 kA, the power in the arc is P = V . I = 100 V. 10 kA = 1 MW. The severity of an arc-ash event is normally quantied as the incident energy that would reach a worker, normally given
Manuscript received January 9, 2009; accepted July 17, 2009. Date of publication May 12, 2011; date of current version July 20, 2011. Paper 2008-REPC232, presented at the 2009 IEEE Rural Electric Power Conference, Fort Collins, CO, April 2629, and approved for publication in the IEEE T RANSACTIONS ON I NDUSTRY A PPLICATIONS by the Rural Electric Power Committee of the IEEE Industry Applications Society. This work was supported by the Electric Power Research Institute under Work Order 067148. The author is with the Electric Power Research Institute, Burnt Hills, NY 12027 USA (e-mail: tshort@epri.com; t.short@ieee.org). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TIA.2011.2153810

Fig. 1.

Staged arc-ash test.

in terms of calories per cubic centimeter. Flame-resistant clothing systems have an arc thermal performance value (ATPV) rating, based on the American Society for Testing and Materials test standards [3]. This rating is the incident energy in calories per cubic centimeter on the clothing surface that has a 50% probability of causing a second-degree skin burn. The goal of an arc-ash analysis is to ensure that workers have an ATPV protection sufcient to handle the incident energy that might be expected in a given work scenario. Out of 14 responses to an Electric Power Research Institute (EPRI) survey, utility minimum ATPV ratings ranged from 4 to 8.7 cal/cm2 with a median value of 5.4 cal/cm2 . Arc-ash protection can be treated as an overcurrent coordination problem. We want the protective device to clear the fault before a fault arc could cause incident energy in excess of the rating of the clothing. Fig. 2 shows an example of coordinating open-air calculation results from ARCPRO for 4- and 8-cal/cm2 clothing against a circuit-breaker clearing curve with the given relay settings. Zahalka and Hoagland [4] provide more information and examples of arc-ash analysis and protection. See [5] for a 12-kV case study. As an overcurrent protection problem, two related assumptions are made. 1) The incident energy linearly increases with time. If you double the duration, the incident energy doubles. All of the modeling approaches make this same assumption. 2) At lower currents and longer durations, the main impact is due to lower current. Most models have an almost linear

0093-9994/$26.00 2011 IEEE

SHORT: ARC-FLASH ANALYSIS APPROACHES FOR MEDIUM-VOLTAGE DISTRIBUTION

1903

Fig. 2.

Arc-ash timecurrent coordination example.

relationship between current and incident energy; if the current doubles, the incident energy doubles. At longer durations, these assumptions are uncertain. As in Fig. 2, incident energies often become more of a concern where fault currents are lower and durations are longer. Arc-ash models have been mostly tested with durations less than 0.5 s. At longer durations, several factors can come into play, i.e., the arc can move and/or elongate (increasing energy), the arc may involve additional phases (increasing energy), and the arc may self-extinguish (decreasing energy). We evaluated monitoring data and performed tests to better consider these. When treating arc ash as a protection problem, several questions need to be addressed. Do we use bolted fault current or some reduced value? Do we use line-to-ground, line-to-line, or three-phase fault current, or all of the above? Do we assume only a single-phase arc and fault? If so, in what situations do we do this? Can we assume the fault self-clears at some point? Should a bolted fault or an arc impedance be used? The 30- to 40- impedance given by the Rural Electrication Association [6] is too large (see also [7] and [8]). The arc in an arc-ash scenario involves relatively low arc resistances. A 3-ft (1-m) arc has a voltage of about 1400 V. If the fault current at that point in the line was 1000 A, then the arc resistance is about 1.4 . A 1-ft (0.3-m) arc with the same fault current has a resistance of 0.47 . Most fault arcs have resistances of 02 ; thus, that can be used as guidance to nd the minimum fault current. II. OVERVIEW OF A NALYSIS A PPROACHES A number of approaches are available for estimating arc ash. Of the most commonly cited methods, three are based on

a single arc in open air, i.e., the Lee method [9], the Privette Electric Arc Heat Flux calculator (also known as the Duke Heat Flux calculator) [10], and the ARCPRO program [11]. The IEEE 1584 method [12] is based on curve-t regressions to mainly three-phase arc-in-a-box tests. Lee rst brought arc ash to the attention of the IEEE Industry Application Society with his 1982 paper [9]. A single open-air arc is modeled. The arc voltage is assumed to be 70.7% of the system voltage. This is the point of maximum power in the arc, and the voltage drop evenly splits between the system (assumed as all reactive impedances) and the arc. The arc energy is contained within a sphere with a diameter of a few inches (centimeters) with the diameter of the sphere increasing with the square root of the arc power. For an arc power of 5 MW, the sphere diameter is 1.62 in (4.11 cm). All energy is assumed to be released as radiation. The Lee method has a number of weaknesses. It has not been corroborated by measurements. It does not account for arc-ina-box effects; thus, it is most suited for open-air calculations. Because the incident energy is tied to the system voltage, the incident energy dramatically increases with voltage. Whether in open air or for faults in equipment, the large increase in incident energies with voltage is unreasonable. Arc lengths and voltages are primarily a function of gap spacings, not the system driving voltage. This was also pointed out within the IEEE Industry Applications Society (IAS) community in [13] and [14]. For a 12.47-kV system with a bolted fault current of 5 kA, the Lee method predicts an arc voltage of 8.8 kV. Using a typical voltage gradient of 40 V/in, that translates into an arc length of 18.3 ft (5.6 m). Yet, that same 18-ft (5.5-m) 44-MW arc is supposed to be conned to a sphere with a radius of 5 in (13 cm). At 34.5 kV and 5 kA, a similar analysis predicts a 51-ft (15.5-m) arc t inside an 8-in (20-cm) diameter sphere. Both scenarios show inconsistent answers, and the arc lengths assumed are well beyond what is reasonable for mediumvoltage systems, either in open air or in equipment. IEEE 1584-2002 [12] was developed by the IEEE IAS, which is a society focusing on industrial and commercial power. IEEE 1584 is the most widely adopted approach to arc-ash analysis. The method for estimating arc-ash incident energies is based on tests performed at several short-circuit laboratories. From this test set, regression was used to nd equations to best t the test data. IEEE 1584 assumes a three-phase fault and is mainly geared toward arc-in-a-box evaluations. Above 15 kV, the IEEE 1584 guide and companion spreadsheet default to the Lee method. The Privette Heat Flux calculator assumes that incident energy drops with the distance squared. It does not account for arc-in-a-box effects. It assumes that all energy is released as radiation. The energy received at a target is a function of what is called as shape or transfer function. The shape factor used is that between a cylinder (the arc) and a rectangle (the person) at some distance. The shape factor is a complicated equation involving the geometry of the arc and the receiving shape that must be integrated. The system voltage is only used to estimate if the arc will sustain. It assumes that about 150 V/in (60 V/cm) is needed to sustain an arc.

1904

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 47, NO. 4, JULY/AUGUST 2011

TABLE I A RC -F LASH C ALCULATION C OMPARISONS

ARCPRO is a commercial program for analyzing arc-ash incident energies, developed by Kinectrics. The ARCPRO algorithm is based on the work of Bingwu and Chengkang [15], but it is not completely described in any peer-reviewed paper. The ARCPRO model assumes the following [16]: a vertical free-burning arc in air; an arc length much greater than the arc diameter; a one-arc column, which is either phasephase or phase-ground; no electrode region heat transfer; and an optically thin plasma and gas. Cress [16] reported that ARCPRO was veried with over 300 test points for arc energy and incident energy for currents from 3 to 25 kA, arc durations from 4 to 35 cycles, distances from 8 to 24 in (20 to 60 cm), and with gaps from 1 to 12 in (2.530 cm). The 2007 National Electrical Safety Code (NESC) [17] requires that an arc-ash assessment be performed on systems above 1000 V. They do not provide specics in general but do offer a table with default assumptions based on an ARCPRO analysis for open-air single-phase-to-ground faults. The NESC table 410-1 footnotes specify a 15-in (38-cm) separation distance from the arc to the employee for glove work and arc gaps as follows: 115 kV = 2 in (5 cm); 15.125 kV = 4 in (10 cm); and 25.136 kV = 6 in (15 cm). Proposed Occupational Safety and Health Administration (OSHA) rules provide a similar table and assumptions [18]. III. C OMPARISON OF A NALYSIS A PPROACHES Table I compares results for the main arc-ash calculation tools for a single-phase open arc at several voltages and for arcs in switchgear for several system voltages. ARCPRO, the Privette Heat Flux program, and the Lee method were all designed for single arcs. For the open-air cases, IEEE 1584 assumes a three-phase fault. The arc gaps for the open mediumvoltage cases were chosen based on the NESC/OSHA table assumptions. The arc gaps for the switchgear cases at 480 V and 12.5 kV were chosen based on the IEEE 1584 defaults; at voltages above that, estimates were used. Fig. 3 compares incident-energy predictions for three of the tools as a function of distance for a 12.5-kV switchgear fault. ARCPRO and Privette Heat Flux use a single phase-to-phase

Fig. 3.

Comparisons of arc-ash analyses.

arc with an arc gap of 6 in, i.e., the default for switchgear in IEEE 1584. Neither ARCPRO nor the Privette model include a multiplier factor to account for multiphase arcs or arc-in-a-box effects. For distances in the range of 1518 in (3846 cm), it appears that no such multiplier is necessary for these assumptions. At a 36-in (0.9-m) working distance (the IEEE 1584 typical distance for medium-voltage switchgear), a multiplier of 2.5 is necessary to make ARCPRO values equal to IEEE 1584 values. It is difcult to compare tools, as each tool can have different assumptions. Different arc gap spacings, system voltages, and fault currents will change relative performances. Variations in parameters are as follows: 1) DurationAll of the tools linearly vary with fault duration, meaning that, if the duration doubles, the incident energy from an arc doubles. In actual practice, longer duration arcs may be more severe due to arc elongation and movement. 2) Fault current magnitudeThe Privette Heat Flux and Lee approaches both assume that the incident energy linearly increases with magnitude. Both ARCPRO and 1584 show energy increasing somewhat more than linearly (multiplying current by 10 increases energy by about a factor of 11.5). 3) Arc gapIn the Privette Heat Flux program, incident energy almost linearly increases with arc length. ARCPRO is close to that, depending on geometries. Both the Privette and ARCPRO tools have the arc length as inputs, and IEEE 1584 uses arc gap internally. These are often treated as the same, but the arc length can be longer than the arc gap. IEEE 1584 equations predict a shallow exponential relationship. Moving from 1- to 2-in (2.5- to 5-cm) arc gap adds about 9% to the incident energy, and for gaps up to 10 in (25 cm), it linearly continues. Note that the arc gap is not a direct input in the 1584 spreadsheet as it uses default gaps for different

SHORT: ARC-FLASH ANALYSIS APPROACHES FOR MEDIUM-VOLTAGE DISTRIBUTION

1905

Fig. 5. IEEE 1584 variation with arc gap. Fig. 4. Predicted versus measured incident energy for the IEEE 1584 data set.

equipment. The Lee method does not consider the arc length; it assumes the length that creates maximum arc energy. 4) Distance to the arcARCPRO, the Privette Heat Flux, the Lee method, and 1584 for open air all predict that incident energy inversely varies with the distance squared. For faults in equipment, IEEE 1584 has differing coefcients of distance with the incident energy varying in the range of 1/d1/d1.5 . 5) System voltageARCPRO and 1584 use the system voltage to determine the arcing current (important at secondary voltages). ARCPRO and the Privette Heat Flux use the system voltage to determine if the arc will sustain. Only the Lee method predicts signicant direct effects of system voltage on the incident energy. 6) Arc movement and plasma effectsNone of the methods account for the arc movement. Only IEEE 1584 accounts for plasma effects (and that accuracy has been questioned [13], [19]). IV. M EDIUM -VOLTAGE A RC - IN - A -B OX C ALCULATIONS For failures in three-phase equipment such as circuit breakers or switches, IEEE 1584 is widely used. It is based on tests that reproduce arcs in equipment. These arc-in-a-box tests included plasma effects. For medium-voltage switchgear, IEEE 1584 provides a typical working distance of 36 in (91 cm). Fig. 4 compares the predicted incident energies compared with the maximum measured incident energies for the mediumvoltage tests in the IEEE 1584 test set. The two different groupings represent data from tests from different test laboratories. The Lee method predicts unreasonably high increases in the incident energy with increasing system voltage, due to Lees assumption of an arc resistance that maximizes the arc energy. Spacings in higher voltage switchgear are not considerably higher than 15-kV systems. Energy is a function of gap spacings and arc bowing, not system voltage. Fig. 5 shows the variation predicted by IEEE 1584 with the gap length. Several gaps are highlighted; at 5 and 15 kV, the 1584 arc gaps of 4 and 6 in

(10 and 15 cm) are shown. For 25 and 35 kV, typical spacings of 9 and 12 in (23 and 30 cm) are shown. The difference between 15-kV class results and higher voltages are as follows: 1) 25 kV: For a gap of 9 in (23 cm), results are higher by 21%. 2) 35 kV: For a gap of 12 in (30 cm), results are higher by 47%. Another more conservative approach is to simply increase the IEEE 1584 results by a multiplier based on the gap length relative to the 6-in (15-cm) gap assumed for 15-kV systems as follows: 1) 25 kV: For a gap of 9 in (23 cm), multiply results by 1.5. 2) 35 kV: For a gap of 12 in (30 cm), multiply results by 2.0. V. A RC C HARACTERISTICS BASED ON M ONITORING The arc length and voltage are important arc characteristics impacting medium-voltage open-air arc-ash hazard estimation. To a rst approximation, arcs have a constant voltage drop. If we know the arc voltage, we know the arc length. The arc length is a key input into ARCPRO and the Privet method. In both of these tools, the arc length determines the arc voltage. The Privette calculator assumes a constant arc voltage of 30 V/in (11.8 V/cm). ARCPRO calculates a voltage that somewhat increases with the current; based on a 10-in (25-cm) arc, at 5 kA, the arc voltage is 39.5 V/in (15.6 V/cm), and at 10 kA, the arc voltage is 42.6 V/in (16.8). The arc voltage (along with the arc current) determines the arc power. We have two data sets with power quality monitors on overhead distribution circuits. These record voltages and currents on all three phases. When a monitor captures a deep voltage sag during a fault directly upstream, it is seeing the arc voltage. If the fault is not directly upstream but off of a tap upstream, there can be a sinusoidal component mixed in with the arc voltage. To best estimate the arc voltage, we developed an arcvoltage estimator. The estimator algorithm uses the following approach: 1) Find the third, fth, and seventh harmonics.

1906

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 47, NO. 4, JULY/AUGUST 2011

Fig. 7. Fig. 6. Histogram of arc voltages in the DPQ study.

Example of an increasing arc voltage with the arc-voltage estimation. TABLE II A RC VOLTAGES AND E QUIVALENT A RC L ENGTHS F ROM VARIOUS M ONITORING DATA S ETS

2) Divide the third, fth, and seventh harmonics by 0.3, 0.18, and 0.129, respectively. Each of these is an estimate of the arc voltage if the arc were an ideal square wave. 3) Take the median of these scaled square-wave estimates. 4) Find the root-mean-square (RMS) voltage. 5) Take the minimum of the RMS voltage and the median of the square-wave estimates. This algorithm is applied on a rolling basis, sweeping along the voltage wave. To make automatic identication of arc voltages easier, we concentrated on voltage sags with the following characteristics: single-phase voltage sags, sags to below 40% of the nominal, and sags to below 5% or those with the third harmonic above 20% of the fundamental and the fth harmonic above 10%. The EPRIs Distribution Power Quality (DPQ) project was the basis for one fault data set. In the DPQ study, power quality monitors were installed on distribution primaries at voltages from 4.16 to 34.5 kV [20], [21]. Two hundred seventy seven sites resulted in 5691 monitor months of data from 1993 to 1995. In most cases, three monitors were installed for each randomly selected feeder, i.e., one at the substation and two at randomly selected places along the feeder. We used the downline feeder results for these analyses. A similar data set was provided by a DPQ participant, referred to as utility A, who kept their monitors in place after the study ended, and the data covered from 1995 until 2008. We used the subset of 3465 DPQ events captured as RMS events where the voltage dropped below 40% of the nominal. This made processing the events more manageable. From this set of voltage sags, we applied the criteria given above for likely arc voltages followed by the manual removal of events that were not faults (circuit interruptions or monitor errors sometimes satised the criteria). This left a data set of 209 events. See Fig. 6 for a histogram of arc voltages estimated from this data. For these events, we only have a wave shape for the rst 0.2 s of the event; thus, our arc-voltage and arc-length estimates are for the rst 0.2 s of the event. Fig. 7 shows an example arc voltage recorded. For the rst 3.5 cycles, the fault looks fully sinusoidal, probably because it is a bolted fault. Then, it transitions to having a more traditional

arcing wave signature. Fig. 7 shows the rolling RMS voltage superimposed upon the arc-voltage estimate. The 2007 NESC table 410-1 and OSHA 1910 table assume certain arc lengths for different distribution voltage classes, i.e., 2 in for 15-kV class systems, 4 in for 25-kV systems, and 6 in for 35-kV systems. The results from utility monitoring are summarized in Table II. Equivalent arc lengths are determined based on using a constant arc-voltage gradient of 40 V/in (15.7 V/cm). We also tried another approach to estimate arc voltages based on substation measurements. This allows a different estimation approach, and it opens up more data for review as more utilities have substation data than downline feeder data. Radojevic et al. [22][24] initially developed this arc-voltage estimation. Their approach relied on the approximation that the arc voltage was a square wave. A modied version of this method was developed to evaluate arc characteristics and also to improve fault locating algorithms [25], [26]. Results from one utility are similar to those found with more direct measurements, as shown in Table II. Other ndings on the arc voltage and length from analysis of the monitoring data include: 1) Impact of system voltageData from both the DPQ study and from utility-A data suggest that longer arc lengths are appropriate at higher voltage systems. In the DPQ data set, the median arc voltage for 15-kV systems was 578 V (n = 183), as compared with a median of 833 V for 25-kV systems (n = 26).

SHORT: ARC-FLASH ANALYSIS APPROACHES FOR MEDIUM-VOLTAGE DISTRIBUTION

1907

2) Fault causeThe data from utility B where cause code information was available showed longer arc lengths for some fault types, particularly conductors on crossarms. 3) Fault durationMost arcing faults stayed constant during the rst ten cycles captured by these power quality monitors. Some events had arcs that lengthen at up to 12 in/cycle (60 ft/s or 18 m/s). 4) Faults evolving to multiple phasesFor utilities A and B, where we had access to several fault events longer than 0.5 s, one-third to one-half of these faults evolved to multiple phases. For utility B, the median time to involve another phase was 0.2 s, but less than 25% of faults evolved to another phase within 0.1 s. VI. OVERHEAD A RC -F LASH T EST R ESULTS Exploratory arc-ash tests were done at the EPRI Lenox facility, MA, to establish realistic arc lengths and voltages, nd how fast arcs can lengthen and evolve to multiple phases, and evaluate the performance of conductor covers. The setup used a 4160-V supply with a three-phase available fault current of about 3500 A. The main ndings of the testing are: 1) Arc length and voltageThe arc lengths assumed in the NESC tables are unrealistically short. In most work scenarios that could lead to a fault being initiated, the arc length quickly grows to many inches and sometimes a few feet. 2) MovementOnce initiated, a fault arc can quickly grow and move. Magnetic forces from the fault current dominate the arc movement. The reball is normally pushed in the same direction as the arc, and the reball also vertically rises from thermal forces. In a phase-to-phase event, velocities of tens of feet per second (several meters per second) were observed. 3) DirectionalityEven in open air, some fault scenarios are directional, meaning that the reball is focused in one direction (like a amethrower), similar to arc-in-a-box type faults. 4) Conductor cover-upConductor covers are effective at reducing the likelihood of ashovers to protected conductors from a fault initiated elsewhere. Even if arcs start across a short gap, the arc length and voltage can rapidly increase. This was observed with several tests, i.e., a jumper wire shorting phase to phase or phase to ground, a solid pipe bridging conductors, and a mock tool bridging a bushing. In nine similar tests, the median arc voltage over the rst 0.1 s (6 cycles) was 446 V, corresponding to an 11-in (28-cm) arc length. Fig. 8 shows results from a test with a phase-to-phase jumper solidly tied at one end and through a #12 wire at the other. During the test, the #12 wire acts as a fuse and burns away within a short time, and the arc is free to move. We attached the jumper to a stirrup to minimize damage to the conductor. The main goal of this test is to mimic a case where a worker accidentally jumpers an energized phase to a grounded conductor or to another energized phase. The side with the connection through the #12 wire is the side where the arc occurs right when the worker touches the jumper to that conductor.

Fig. 8. Video snapshots and voltage waveform for a jumper pulling away.

The video frames on the right in Fig. 8 are from a dark lter that passes infrared and some red light. This allows us to see the hottest part of the arc, i.e., the arc channel. Fig. 8 also shows the voltage between phases for the line-to-line fault initiated. In this particular event, the #12 wire took several cycles to burn away because more than one wire was jumpered to the stirrup. Once the connector wire burned away, the voltage and the arc length increased to several feet (more than 1 m) until the fault selfcleared. The 4160 V was unable to sustain an arc that long. For 15- to 35-kV systems, much longer arcs could be possible. A number of tests were performed to evaluate the effectiveness of conductor covers at preventing faults from escalating to multiphase faults. This is important if you assume only lineto-ground faults in your arc-ash analysis for glove work. We

1908

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 47, NO. 4, JULY/AUGUST 2011

to 25- and 35-kV systems and open-air testing with calorimeter instrumentation to measure incident energies with longer arc lengths. ACKNOWLEDGMENT The author would like to thank Northeastern Utilities (NU) for performing arc-ash tests at the Electric Power Research Institute (EPRI) Lenox test facility. NU performed a number of circuit congurations to help provide a more isolated 23-kV supply to the laboratory and worked with the EPRI to coordinate fuses and relay settings. The author would also like to thank D. Thomas of NU and D. Childs, D. Crudele, C. Moore, M. Messana, and M. Clary of the EPRI for supporting the Lenox test work. R EFERENCES
[1] Y. Goda, M. Iwata, K. Ikeda, and S. Tanaka, Arc voltage characteristics of high current fault arcs in long gaps, IEEE Trans. Power Del., vol. 15, no. 2, pp. 791795, Apr. 2000. [2] A. P. Strom, Long 60-cycle arcs in air, AIEE Trans., vol. 65, pp. 113 118, 1946. [3] Standard Test Method for Determining the Arc Rating of Materials for Clothing, ASTM F1959, 2006. [4] G. Zahalka and H. Hoagland, Arc ash hazard assessment requirements, in Proc. IEEE Rural Elect. Power Conf., 2007, pp. C5-1C5-6. [5] A. Haman, Arc ash hazards: A 12.5 kV distribution system case study, presented at the Minnesota Power Conf., Minneapolis, MN, 2007. [6] Guide for Making a Sectionalizing Study on Rural Electric Systems, Rural Elect. Admin., U.S. Dept. Agriculture, REA Bull. 61-2, Washington, DC, 1978. [7] J. Dagenhart, The 40-ohm ground-fault phenomenon, IEEE Trans. Ind. Appl., vol. 36, no. 1, pp. 3032, Jan./Feb. 2000. [8] J. J. Burke, High impedance faults (40 ohms is a fallacy), presented at the IEEE Rural Electric Power Conf., Albuquerque, NM, 2006. [9] R. H. Lee, The other electrical hazard: Electric arc blast burns, IEEE Trans. Ind. Appl., vol. IA-18, no. 3, pp. 246251, May 1982. [10] A. Privette, Electric Arc Heat Flux Calculator, FLUX.EXE, 2009. [Online]. Available: http://www.arcwear.com/arcwear.com/downloads/ dukeprog.zip [11] Users Guide for ARCPRO, Version 2.0, Kinectrics, Toronto, ON, Canada, 2002. [12] IEEE Guide for Performing Arc-Flash Hazard Calculations, IEEE Std 1584-2002. [13] A. D. Stokes and D. K. Sweeting, Electric arcing burn hazards, IEEE Trans. Ind. Appl., vol. 42, no. 1, pp. 134141, Jan./Feb. 2006. [14] D. Sweeting and A. D. Stokes, Energy transfers within arcing faults in electrical equipment, in Proc. 8th ICEFA, 2007, pp. 169178. [15] G. Bingwu and W. Chengkang, The gasdynamic and electromagnetic factors affecting the position of arc roots in a tubular arc heater, Acta Mech. Sin., vol. 7, no. 3, pp. 199208, 1991. [16] S. L. Cress, Arc modeling and hazard assessment, presented at the EPRI Distribution Arc Flash Workshop, Chicago, IL, 2008. [17] National Electrical Safety Code, IEEE C2-2007. [18] Dept. Labor Occupational Safety Health Admin., 29 CFR Parts 1910 and 1926 Electric Power Generation, Transmission, and Distribution; Electrical Protective Equipment; Proposed Rule, vol. 70, 2005. [19] R. Wilkins, M. Allison, and M. Lang, Effect of electrode orientation in arc ash testing, in Conf. Rec. IEEE IAS Annu. Meeting, 2005, pp. 459465. [20] An assessment of distribution system power quality: Volume 2: Statistical summary report, Elect. Power Res. Inst., Palo Alto, CA, EPRI TR106294-V2, 1996. [21] An assessment of distribution system power quality: Volume 3: Library of distribution system power quality monitoring case studies, Elect. Power Res. Inst., Palo Alto, CA, EPRI TR-106294-V3, 1996. [22] Z. M. Radojevic, H.-J. Koglin, and V. V. Terzija, A novel approach to the distance protection, fault location and arcing faults recognition, in Proc. IEEE PES Power Syst. Conf. Expo., 2004, pp. 628634. [23] Z. M. Radojevic, C. J. Lee, J. R. Shin, and J. B. Park, Numerical algorithm for fault distance calculation and blocking unsuccessful reclosing

Fig. 9. Arc that attaches to an unfaulted phase at 90 ms.

Fig. 10. Arc test with a line hose protecting the unfaulted phase.

performed a number of tests where a fault was initiated about 2 ft (60 cm) under an unfaulted conductor. Of tests without any conductor covers, the unfaulted and unprotected phase ashed over at an average time of 0.1 s. Fig. 9 shows a ltered image overlaid on a background as the arc attaches to the unprotected phase above. In ve tests using line hoses to cover the unfaulted phase, the phase did not ashover in two of the cases (fault duration = 1.3 s). In the cases where it did ashover, the average time to ashover was 1.0 s. Based on the videos of the arcs and the examination of the line hoses after tests, all ashovers that occurred with cover-up were to exposed parts of conductors. We found no evidence that any events punctured through the coverings or snaked through the seams, although the line hoses were fully engulfed in the plasma reball from the arc (see Fig. 10 for a test without ashover). VII. D ISCUSSION For open-air medium-voltage evaluations, the monitoring and test results reported here suggest that longer arc lengths are suitable for inputs to arc-ash analyses than the default assumptions in NESC and OSHA. The NESC table gives lengths of 26 in (515 cm), and arc lengths of between 10 and 15 in (2538 cm) are more realistic. The analyses tools do need more evaluation and comparison with tests using longer more realistic arcs. For switchgear and other equipment, IEEE 1584 is commonly used for analysis. For 25- and 35-kV class voltages, the Lee method makes unrealistic assumptions; thus, using a multiplier to the 15-kV results in IEEE 1584 is a suitable option. Further work is warranted in the area of medium-voltage arc ash, particularly arc-in-a-box testing with wider electrode spacings to more condently extend the IEEE 1584 predictions

SHORT: ARC-FLASH ANALYSIS APPROACHES FOR MEDIUM-VOLTAGE DISTRIBUTION

1909

onto permanent faults, in Proc. IEEE Power Eng. Soc. Gen. Meeting, 2005, pp. 757762. [24] Z. M. Radojevic, V. V. Terzija, and N. B. Djuric, Numerical algorithm for overhead lines arcing faults detection and distance and directional protection, IEEE Trans. Power Del., vol. 15, no. 1, pp. 3137, Jan. 2000. [25] T. A. Short, D. D. Sabin, and M. F. McGranaghan, Using PQ monitoring and substation relays for fault location on distribution systems, in Proc. IEEE Rural Elect. Power Conf., 2007, pp. B3-1B3-7. [26] Distribution fault location: Field data and analysis, Elect. Power Res. Inst., Palo Alto, CA, EPRI 1012438. [Online]. Available: http://my.epri.com/portal/server.pt?Abstract_id=000000000001012438

Tom A. Short (M90SM98) received the M.S. degree in electrical engineering from Montana State University, Bozeman, in 1990. Starting in 2000, he was with Power Technologies, Inc. for ten years. He is currently with the Electric Power Research Institute, Burnt Hills, NY. He authored the Electric Power Distribution Handbook (CRC Press, 2004). In addition, he led the development of IEEE Standard 1410-1997, i.e., Improving the Lightning Performance of Electric Power Overhead Distribution Lines, as the Working Group Chair.