DC Arc Hazard Mitigation Design

at a Nuclear Research Facility

Copyright Material IEEE - Paper No. ESW2014-29

Walter P. Mendenhall, PE
Senior Member, IEEE
Wiley|Wilson, 127 Nationwide Drive, Lynchburg, VA 24502
wmendenhall@wileywilson.com

Though potential arc heat and blast energy from dc sources electrical workers under-protected.
may be significant, the risk assessment for exposure to dc NFPA 70E – 2012, Annex D.8, includes a validated,
energy sources is often not performed because the requirement conservative method for computing incident energy from
to do so is not understood. Making the task of quantifying dc arc flash where the available dc bolted fault current is
dc arc energies more intimidating is that the professional known. Quantified warning labels for equipment can now
engineer often assumes liability for any errors or omissions be provided, where workers are exposed to live dc so that
that may result in injury — or worse. Compounding the issue the appropriate personal protective equipment (PPE) can be
even further is that the analysis work is often based on tools, worn. It is anticipated that OSHA and state safety agencies
methods, and verifications for dc arc heat and blast energies will soon enforce compliance of arc flash responsibilities for
that are incomplete and immature. dc equipment as they have for ac.
The emergence of a method to determine dc arc flash
Explained here is one approach taken to perform an arc hazard is a significant step forward. Unfortunately, the
heat and blast analysis for a research facility, specifically a developments of dc arc flash software tools and dc short
concentrated load supplied by eight 1 MW SCR rectifiers circuit analysis tools that are fully integrated with ac tools
regulating +/-300 Vdc outputs to tightly arranged terminations lags behind the new standard. This gap presents challenges to
at a load. During the course of the research, the connections engineers but also represents an opportunity for manufacturers
were to be reconfigured, followed by a planned retuning of to introduce tools and software that can fill this need.
the regulated dc output using hand-held multi-meters at the Until such items are introduced, however, it is necessary
load terminations, without wearing PPE. After assessing the for engineers to use other methods to determine dc arc flash
risk, the research engineers were easily convinced that a full hazards. This summary is an example of a process that was
analysis needed to be performed with risk mitigation design. used to evaluate dc arc hazard for multiple rectifiers supplying
an array of resistive loads. The purpose is to encourage others
Index Terms — arc flash analysis, dc arc, dc arc hazard, to recognize dc arc hazards and to suggest resources that
mitigation may be helpful to those involved in electrical systems hazard
analysis.
I. INTRODUCTION

The arc hazard from ac electrical

systems has become well known over 600A 600A
the past decade thanks in large part to the 880 kVA, dry, isol. Tx 880 kVA, dry, isol. Tx
12kSCA @ Secondary 12kSCA @ Secondary
widespread application and enforcement 600A 600A
of NFPA 70E, the Standard for Electrical
Safety in the Workplace®. NFPA 70E Two SCR Drives Positive dc 1M Power W Negative dc 1MW Power
& dc Bus Cabinet Supply (Drive) Cabinet Supply (Drive) Cabinet
was developed using several incident (Typ. of 4)
Paper No. ESW2014-29
+300Vdc Common Bus -300Vdc
energy quantification methods. One of Bus Cabinet
dc supply cables dc supply cables
these, included in its annex, is IEEE 1584, terminations terminations
dc return
“Guide for Performing Arc-Flash Hazard resistive elements cables resistive elements
Calculations.” Until now a validated
method of calculating the arc hazard COM
from dc sources has been conspicuously RING
missing from this standard. Therefore,
dc arc flash hazard has been estimated at
Figure 1. DC Arc Hazards
best and ignored at worst, which leaves

978-1-4799-2098-3/14/$31.00 ©2013 IEEE

II. EVALUATING DC ARC HAZARDS FROM RECTIFIERS

This project involved architecture and engineering The objective was to develop a documented standard
services for the design and construction of a nuclear research method to compute a fault value for the exposed terminations
facility. The work included an arc flash hazard analysis for the of the core simulator heater resistors, the buses in the dc
ac electrical system components that were to be installed. The distribution enclosure, and the output terminations of the
system included eight 1050 HP dc drives with SCRs to regulate rectifiers (drives). It was decided to evaluate the outputs of
their variable outputs to a nominal 300 Vdc, which originally dc fault currents from existing dc system analysis software [1]
was not part of the scope of analysis. The drives supplied against values obtained from manual spreadsheet calculations
300 Vdc resistive heater loads that could generate nearly 1.4 based on ANSI [2] and IEC [3] methods. a Initial discussions
MW of heat. Because personnel, contractors, and vendors with the analysis software vendors at the time indicated that
would be exposed to live dc voltages during commissioning several companies were in the process of developing dc arc
and startup system testing, the project included analyzing the hazard analysis tools; unfortunately, they were not available
dc system arc hazard. until after this project was completed. Because there were
Models of the electrical system were built in a widely used some unanswered questions about the meanings of some
electrical system analysis tool, but the short circuit analyses of the available beta version tool’s dialog entries, manual
for the ac and dc portions of the electrical system had to be calculations of the dc fault current were performed based on
performed separately due to the lack of integration between the IEC Standard 61660-1 [3] method. The referenced work
these tools. For instance, the software model for rectifiers by J.C. Das [4] provides an example of this method.
required manual entry for values of the ac source system
impedance and available fault energy into the component
editor dialog box. Therefore, if the input parameters changed B. Calculations Used During the Process
due to reconfiguration of the ac system, any updated values The peak short circuit current i1pD was computed with:
from the ac short circuit analysis results had to be found and
copied into the dc system analysis application manually. With i1pD = κ D I k D (1)
multiple rectifiers in the system, each with different length of
supply conductors, iterations of this process quickly became where constant κ D was calculated by the equation from [3]:
labor intensive and subject to errors. π
2 − ( +ϕ D ) cot φ D π L  (2)
κ D = 1+ e 3
sin φ D  − arctan DBr 
A. Solving the dc Fault Current Problem π LN
2 
The first step in computing the dc arc hazard is to where
determine the available current at the fault. The dc system is LDBr is the inductance on the load side of the rectifier
composed of eight dc drives, each with a rated input of 690
Vac and rated output of 1410 A at 750 Vdc. This rating was LN is the inductance on the line side and
used to determine the “effective impedance” of the rectifier
[1]. The eight drive outputs were configured as four pairs Ik D is the quasi steady-state current is defined in terms of
of plus and minus supplies to four bus cabinets collocated equivalent circuit diagram of the rectifier defining parameters
with each pair of drives. The controllers were programmed used in multiple defining equations (8) through (13) in [3].
to respond to a remote master controller to supply a variable The rise time constant τ1D was calculated from:
voltage regulated output centered about a nominal +/- 300
Vdc. Highly stranded flexible cables were routed in trays   LDBr 
distributing the dc power to an array of resistive elements.
τ 1D = 2 + (κ D − 0.9 ) 2.5 + 9  ms (3)
  LN 
Initially, half of these elements were terminated with +300 Vdc
and an adjacent element termination of -300 Vdc was located and the decay time constant τ2D from:
Paper No. ESW2014-29
only a few inches away. All of the elements had common
returns to the associated bus through the tray cabling. With the 2
τ 2D = ms
array as originally constructed, the alternating plus and minus RN  R  (4)
terminations of the resistors with rows of similarly alternating  0.6 + 0.9 DBr 
terminations immediately above and below, supplied by the XN  RN 
other three pairs of dc drives could easily produce a high- The variables in these equations in this summary are
energy cascading failure. documented in [3] and [4].

978-1-4799-2098-3/14/$31.00 ©2013 IEEE

 Published parameters for resistive and inductive
impedances of dc conductors were not readily available from C. Converting dc Fault Current into Incident Energy
vendors, which also explained why dc conductors were not
found in the vendor’s software library. Only typical values for In his recent paper on arc flash energy conversion of dc
relatively low strand count conductors were found in sources faults [6], Dan Doan developed the following formula for
like the Standard Handbook for Electrical Engineers [5]. incident energy:
Only through verbal contact with vendor engineers was this  Vsys 2  Tarc
information gathered. IE max power = 0.005  (5)
The fault current formula was tested for sensitivity to
 Rsys  D2
 
the minimum and maximum variations in conductor lengths
used in the distribution of ac to each of the eight power where
supply rectifiers as well as within the dc distribution to the
bus enclosure and resistors. Because conductors were fairly IEmax power estimated incident energy at maximum
large and the lengths short, there was no significant variation power point (in cal/cm2);

in results so an average length was used. Vsys system voltage (in volts);
There were no dc protective devices in this system so the
calculated peak current and rise and fall time duration was Rsys system resistance (in ohms);

plotted on time-current curves for the upstream electronic trip Tarc arcing time (in seconds); and
circuit breaker and the current limiting URS fuses on three-

phase 480Vac (source) side of the rectifiers for each of the D distance from arc (in centimeters)
three fault locations. In each case, the magnitude and duration
of the peak fault current was insufficient to part the fuse before The formula has been validated by several years of
the breaker opened. Therefore the quasi-steady state value for empirical measurements made of dc arcs [7].
dc fault current was used to determine an equivalent ac fault
current to determine the trip delay time needed to compute
incident arc energy.

XF: T1
Name: XF: T1
880 kVA SWBD: IST
480 V -- 480V
5.80%Z CB: T1
Inrush: 12.0x

Name: FU: HTR_PS_Group1A

SIBA URS SQB3 F: XF T1
SIBA
Trip: 1400.0 A

Name: CB: T1 BUS: XF T1 Primary

CUTLER-HAMMER Magnum DS, RMS 520 P
MDS-616 XF: T1
Trip: 1600.0 A S
Plug: 1600.0 A
Settings: Phase
BUS: XF T1 Secondary
LTPU (0.4-1.0 x P) 0.9 (1440A)
LTD (2-24 Sec.) 4
STPU (2-10 x LTPU) 4 (5760A) F: HTR 1_2a
STD (0.1-0.5 Sec.) 0.3 (I^2t Out)
INST (2-12 x P) M1(12) (19200A) HTR: 1_2a
Name: CB: HTR_PS_Group1A
ALLEN-BRADLEY 140U-N
CB: HTR_PS_Group1A
Paper No. ESW2014-29
140U-N
Trip: 1200.0 A
Plug: 1200.0 A
TX Inrush FU: HTR_PS_Group1A
Settings: Phase
LTPU (1.0 x P) Fixed (1200A)
LTD (Fixed) Fixed
STPU (2-8 x P) 2 (2400A) REC: PS_Group1A
STD (Fixed) Fixed (I^2t In) 11807 A
INST (14000A) Fixed (14000A)

Fig. 2. Time Current Characteristics for DC Power Supply Protective Devices

978-1-4799-2098-3/14/$31.00 ©2013 IEEE

D. Single Fault Results F. System and Procedures Redesigned to Reduce Arc Hazard

By applying (5) to the system voltage, system resistance, This study caused immediate and significant changes in
and arc duration at an arm’s length working distance of 46 cm the planned procedures for testing and commissioning the dc
(18 in), the incident energy was calculated and the results are system. As a result, measurements were taken with remote
shown in Table 1. sensors rather than handheld meters. Because the terminals of
the resistors are not enclosed and subject to exposure, water
Table 1. System DC Arc Hazard Results leakage, and thermal cycling, an unexpected fault could occur
exposing unprotected personnel within the area to arc flash
Incident Arc
Prot Equiv Trip hazard. This led to discussions of the design for an effective
Bus Energy @ Flash
Dev SCac Delay protective barrier surrounding these terminations.
Name 46cm Boundary
Name (kA) (sec) The barrier design objectives were to achieve safety from
(cal/cm2) cm (in)
arc heat and blast pressures immediately outside the barrier
Rectifier CB:
2.45 0.601 18.20 178 (70) and had to be flexible enough to be disassembled for resistor
Output HTR_PS
reconfigurations and routine maintenance when the equipment
Heater Dist CB:
2.45 0.606 17.79 175 (69) was powered down. Because of the space constraints, arc
Bus HTR_PS
suppression blankets supported by steel frameworks were
CB: initially proposed as a shield design. However, the intensity
Heater 2.43 0.633 8.61 135 (53)
HTR_PS
and duration of the cascaded fault heat energy at a distance

of 14cm (5.5 inches) was greater than 250 cal/cm2 — well
E. Cascading Fault Results
beyond the capabilities of available validated multilayer glass
and ballistic-fiber blankets at the distance from the arc fault.
Because of the high possibility of a fault at one of the
This meant that additional arc fault energy reductions were
resistor terminations cascading to adjacent terminations, the
needed to ensure the blanket was not compromised by the arc
calculation was performed for the vectorial contribution of
blast heat or pressure.
adjacent faults up to the maximum for contributions from all
Reducing the arc energy by reducing fault clearing time of
four pairs of rectifiers. A spreadsheet was set up to determine
the ac protective devices was not practical without increased
the maximum heat and pressure contribution from cascaded
nuisance trips or preconditioning fuses. Space limitations of
arc faults on the surface sphere with an initial radius of 18
the system prevented the design from incorporating protective
inches beyond the initiating arc. The calculations considered
devices on the dc side of the rectifier. Changing the physical
the geometry occluding some of the terminations and that the
configuration of the resistor terminations from alternating the
initial arc would reduce the energy of cascaded arc energy
plus and minus 300 Vdc terminations for each resistor pair to
supplied from the same pair of rectifiers. The results of worst-
a new layout with all plus 300 Vdc resistors on one side of
case scenarios were well in excess of 40 cal/cm2 for heat
an electrical barrier material and all minus 300 Vdc resistors
energy.
on the other side reduced the driving arc voltage to 300V (+
or – to ground). Both arc heat and blast pressures are directly
Alternating Terminations Evenly proportional to the square of the voltage so the heat and
pressure were reduced by a factor of four.
Spaced Around Each Cylindrical Row Since arc heat and blast pressures are inversely
proportional to the square of the distance from the initiating
arc to the proposed shield blanket, doubling the distance to
28cm (11 inches) reduced the arc heat to a maximum of 44 cal/
cm2 — well within the validated testing of the seven layer arc
suppression blanket. The customized blanket was suspended
+ by carabiner clips through grommets to a steel structure and
Paper No. ESW2014-29
the lower portion of the blanket was cinched by straps below
the resistor array structure; thus it would deflect an arc blast
+ upward without rupturing and would be flexible enough to
maneuver for maintenance activities. The equipment space
where the resistors are located is no longer required to be
restricted to access during operation due to arc flash hazard.

Fig. 3. Cascaded Arc Energy

978-1-4799-2098-3/14/$31.00 ©2013 IEEE

 + 300Vdc
a
The power systems analysis application provider’s “ANSI method” of dc
+ 30
Vdc 0Vd fault current calculation resulted in a higher current with less clearing time
- 300 c for the fault. Without compelling rationale for the different results, the
conservative position was to assume the higher dc arc energy derived from

+
c
Vd

30
the IEC method of dc fault current calculations.
00

0V
b
The presence of an overcurrent limiting function in the control software
-3

dc
for the rectifier may significantly reduce the exposure to dc arc flash hazard

+ 30
c

for this system. However, there are numerous letters of interpretation for
0Vd

OHSA’s 29 CFR 1910.147, “The control of hazardous energy (lockout/

0Vd
- 30

tagout),” [8] that directly state that control hardware and software are NOT
Typical Element

c
normally considered effective, positive forms of isolating hazardous energy.

+ 300Vdc
- 300Vdc

REFERENCES

[1] DC Short Circuit (ANSI) Reference Manual, SKM Power

+ 30
c

Tools for WindowsTM DC Systems Analysis, Version 6.5.2.6

0Vd

0Vd
(Build 3), 2011 Help Document Library.
- 30

c
+ [2] IEEE 946, DC Auxiliary Power Systems for Generating
c
Vd

30

Stations, 1992, Standard 946.

0

0V
0
-3

dc

c + 30
0Vd 0Vd
- 30 - 300Vdc c [3] IEC, Short-Circuit Currents in DC Auxiliary Installations
in Power Plants and Substations - Part 1: Calculation of short-
circuit currents. Standard BS EN 61660-1:1997, incorporating
Non-Conductive Barriers (typ) corrigenda March 1999 and March 2000.
for each Cylinder
[4] J.C. Das, Power System Analysis: Short Circuit, Load
Fig. 4. Rearranged Load Termination Arrangement Flow, and Harmonics, p. 321-322, CRC Press, (2008).

[5] D. G. Fink and H. W. Beaty, Standard Handbook for

Electrical Engineers, 12th Edition, Section 4, McGraw Hill,
APPENDIX A: DC FAULT CURRENT VERIFICATION TESTING (1987).

As this paper was written, specifications are being [6] D. R. Doan, “Arc Flash Calculations for Exposures to DC
considered to test an exemplar rectifier for a bolted fault at Systems”, IEEE Transactions on Industry Applications, Vol.
the output terminals. If such a test is made, instrumentation 46, No. 6, November/December 2010.
for fault currents and voltages can be made that can be used
to verify the computations for this specific model. The effects [7] DC Arc Hazard Assessment Phase II Copyright Material
of overcurrent protection algorithms in the control software Kinectrics, Inc. Report No. K-012623-RA-0002-R00.
may limit the duration of the fault to significantly less than
0.6 second required to trip the breaker with the given source [8] US Department of Labor, 29 CFR 1910.147, Standard
system impedances.b Interpretation to Mr. Earl Reyes, January 25, 2008.

Paper No. ESW2014-29

978-1-4799-2098-3/14/$31.00 ©2013 IEEE