658 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 21. NO.

4, JULYIAUGUST 1991

Battery Protection- Where Do We Stand?

Richard L. Nailen, Senior Member, IEEE

Abstract-Growing emphasis on operating reliability is lead- 2) Whereas care is taken to keep other electrical equip-
ing both utility and industrial/commerciaI power system opera- ment free from chemical contaminants or corrosive
tors to confront a long-neglected system protection issue: proper
fault protection for large batteries. Full-load current from a
liquids, the battery inherently requires those potentially
large UPS battery may reach thousands of amperes, with fault destructive agents to do its work. Fumes, corrosion,
currents 10-12 times as high. Battery destruction during a major and possibly explosive conditions all are essential parts
fault can acid-contaminate an entire building, escalating damage of the electrical installation, rather than being carefully
far beyond battery cost or equipment downtime alone. Etisting excluded
IEEE standards, the NEC, other published standards, and sup- 3) The battery is a distributed power source. Its many
plier literature offer little help in applying fuses (seldom mar-
keted with specific dc current or voltage ratings) or circuit
interconnected components take up a wide area, rather
breakers td battery circuits. Often, no protective devices at all than being concentrated within a single compact assem-
are proqided because of a fear of their nuisance operation. Of bly like an alternator or engine. Those components
special concern is the “internal” fault that may go undetected must be readily separable, meaning that numerous “in-
by devices at a battery’s main terminals. This paper will outline ternal” connections are fully exposed to the environ-
some u,ser experience, bring together existing recommendations,
ment and to careless handling and not encased by a
and perhaps stimulate development of useful standard practices.
protective insulation system that can be left undis-
turbed.
WHYBATTERIES
AREDIFFERENT 4) Cell action cannot be shut off by throwing a switch,
closing a valve, or braking some machine to a stop.
A MONG THE energy sources available to modem indus-
trial and commercial power systems, a large storage
battery is unique. Until recently, batteries having terminal
Voltage-and the possibility for destructive short-cir-
cuit current-remains present at the terminals of each
cell, whatever switches are opened or connectors re-
voltages of 250 or more, with capacities of thousands of moved.
ampere hours, were confined to major utility substations or
large switchgear installations in industrial plants such as In addition, unlike other power sources, batteries will
refineries. Today, however, these power sources are becom- usually cause major, widespread damage to other equipment,
ing widely used in data processing facilities. They may be by as well as endangering personnel, when a major failure does
far the highest energy equipment on the premises. The poten- occur.
tial for catastrophic damage from battery failure is often
unappreciated by system designers or users. Within the IEEE Consequences of Battery Failure
and other standards-making agencies, no comprehensive
guidelines exist for the protection of battery circuits from Batteries supplying major loads up to several thousand
such damage. kilowatts at 400-500 V are no longer found only in heavy
Because of their predominance in the United States, the industrial facilities with competent plant engineering and
batteries considered in this paper are of the lead-acid type electrical maintenance staffs. They are going into office com-
[l]. But, similar protective measures are equally important to plexes where the big selling point is that the battery’ is
the nickel-cadmiud type. “maintenance free. (Users neglected their batteries enough
”

Large battery circuit protection is rendered both more as it was; now, their neglect seems to be encouraged by
important and more difficult by the following ways in which product marketing. Put the battery in place, says one sup-
plier, “and put it out of your mind.” Prospective users of
this source of electrical energy differs from others:
large battery systems may find that advice pleasant to hear
1) When it’s needed, it must perform. The control or UPS but would be wise not to take it too seriously [2].) Personnel
battery is the “last resort.” There’s no time to warm it capability in such facilities may be quite limited.
up; it must carry full load at once because nothing else Available short-circuit current can be tens of thousands of
is available. amperes. Faults are not in the “spark” category; rather,
vaporization of large conductors is possible. This current
Paper IPSD 90-11, approved by the Power Systems Protection Committee depends not only upon external circuitry, and battery termi-
of the IEEE Industry Applications Society for presentation at the 1990
Industrial and Commercial Power Systems Annual Technical Conference, nal voltage at the time of fault, but also upon such internal
Detroit, MI, April 29-May 1. Manuscript released for publication January battery conditions as electrolyte level and the state of charge.
21, 1991. The basis for calculating ‘ ‘available” short-circuit current is
The author is with the Electrical Engineering Division, Wisconsin Electric
Power Co., Milwaukee, WI 53201. the so-called “one-minute current” rating of the battery-the
IEEE Log Number 9144793. discharge amperes that will lower cell voltage to 1.75 within

0093-9994/91/0700-0658$01.00 0 1991 IEEE

NAILEN: BATTERY PROTECTION-WHERE DO WE STAND 659

60 sec. Assume a 250-V (nominal) battery capable of sustain- protection available. ” When that “cable” becomes a 0.5 by
ing a 100-kw load for 1 hr. The 1-hr ampere rating of the 5 in busbar, prudence dictates that it not be given the role of
battery will be 100000/250 or 400 A. For one typical make a fuse.
of cell, the 1-min current will be about twice that, or 800 A. Chapter 6 of the Orange Book goes on to indicate how
Leaving aside the minor corrections for state of charge or large battery short-circuit currents can be. But no guidelines
temperature and ignoring intercell connection resistance, the are offered for selection of either fuses or circuit breakers to
calculated short-circuit current would typically be 7000 to protect against such current. The only circuit diagram pro-
10000 A. Large UPS loads today range as high as several vided shows a breaker protecting the battery charger and not
thousand kilowatts so that short-circuit currents of 100 000 A the battery itself. None of the numerous bibliographical
or more are theoretically possible. references touches on the subject.
In practice, as a battery manufacturer has pointed out, Several IEEE standards, such as 450, 484, and 485, are
“one or more intercell connectors will go open circuit within concerned with design and installation of large storage batter-
10 to 15 s.” Unfortunately, that might not interrupt the flow ies. Only one, No. 946 [4], deals with circuit protection for
of fault current any more than vaporized busbars or cables battery systems (written for nuclear generating station appli-
will interrupt an arcing burndown fault in 480-V switchgear. cations, and then only in general terms, it is now being
In battery circuits, as in ac distribution systems, the high-cur- revised to cover other power plants as well). Other docu-
rent “bolted fault” is less likely than the low-current sus- ments, such as NFPA 70B on electrical equipment mainte-
tained burndown. Just as destruction is prolonged and en- nance, NFPA 110 on emergency power sources, or the
larged by the heat and sprayed conductive particles during various NEMA battery standards dealing almost entirely with
the ac fault, in the battery system the scattering of conductive testing, are silent concerning fault protection.
electrolyte escalates the damage. Article 480 of the National Electrical Code applies to “all
What’s more, the rapid spread of acidic smoke can quickly stationary installations of storage batteries. It says nothing
”

bring about catastrophic damage to electronic equipment far about battery circuit protection. At least one pertinent UL
from the fault itself. As an example, a manufacturing plant apparatus standard is in preparation but offers little specific
basement contained a UPS and its battery supply. The batter- advice (UL 1778, drafted in March 1989 for public com-
ies were in a separate, cut-off room, equipped with exhaust ment, titled “Uninterruptible Power Supply Equipment”). A
ventilation, smoke detectors, and automatic sprinklers. Above few paragraphs of this 215-page draft dealt with battery
that, on the first floor, was a large computer room plus other supply circuit protection for UPS installations. Either circuit
offices. breakers or fuses were allowed. Paragraph 25.22 stated “The
At 9:05 one Saturday evening, a smoke detector alarm was battery supply circuit shall be provided with acceptable rated
recorded, followed 3 min later by a sprinkler waterflow dc overcurrent protection to reduce the risk of fire and
alarm. The Fire Department arrived within 4 min to find that electric shock resulting from overload or short-circuit condi-
a short circuit in one battery string had caused a small tions.” The “acceptable rating” was to protect against ex-
explosion and fire, ultimately damaging 15 cells, that the plosion, flame, molten metal, and cracking or bursting of
sprinklers had already extinguished. But the entire basement battery cases, but just how such ratings were to be arrived at
was filled with smoke-necessitating the subsequent cleaning was not specified. Test procedures were not given.
of all the electrical equipment at a cost of $100000. Without
the prompt alarm, plus the battery room separation from the FUSES
IN BATTERY
CIRCUITS
computers, that loss would have been far greater-and this The only definitive book on fuse application [5] devotes
was an exceptionally well-designed and well-equipped instal- nearly 200 pages to the behavior and application of fuses in
lation. all sorts of electric circuits-except battery circuits. The
The extensive application literature available from large word “battery” does not even appear in the book’s index.
UPS suppliers usually offers much detail on battery/charger Whether or not fuses are used at all to protect a battery,
sizing, installation, and maintenance. The following is typical and if so, how the fuses are selected, are matters for individ-
user information: ‘‘The battery represents a significant por- ual engineering judgement in each application. But the deci-
tion of the cost of a UPS package. . . since the battery is the sion can be much more complex in a battery circuit than in an
sole source of power during an outage no shortcuts should be ac power system. That complexity stems from the basing of
taken regarding maintenance and care for the battery. Yet
” modem power fuse technology primarily on ac circuit behav-
these publications are silent concerning circuit protection for ior. Fuses do not work the same way in a dc circuit. Because
this essential back-up power source. low-voltage, high-current dc circuits were so seldom encoun-
tered in industrial or commercial facilities until recently, little
Lack of Protection Guidelines market demand existed to support the testing needed to
Do existing standards offer guidelines concerning battery establish dc fuse ratings or to develop fuses best suited to dc
protection? Almost nothing is published on the subject. As usage. Although the rapid growth in power electronics is
the IEEE Orange Book [3] says in Chapter 6 (“Protection”), changing that, much work remains to be done.
“an area of battery protection often overlooked is overcur-
rent protection. . . a designer should be aware that without dc versus ac Fuse Ratings
overcurrent protection a battery can be damaged and, in some Cautions the “Overcurrent Protection Handbook” issued
cases, the battery cable may inadvertently be the only fuse by one international fuse manufacturer, “A-c rated fuses
660 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 21, NO. 4, JULYIAUGUST 1991

should not be applied in d-c voltage circuits unless d-c TABLE I

application ratings are provided by the fuse manufacturer COMPOSITE
OF DC VERSUSAC CURRENT-INTERRUPTING
VALUES
PUBLISHEDBY VARIOUS
FUSESUPPLIERS
[and they seldom are]. Except for some special purpose
fuses, d-c ratings are not usually shown on fuse labels.” A Type of fuse ac amperes dc amperes Ratio, dc to ac
literature search made as background for this paper, admit- K-5 50 OOO 20 OOO 1:2.5
tedly not an exhaustive one, involved (besides the standards L 200 OOO 40 OOO 1:5
RK-1 200000 “contact factory” ?
mentioned earlier) 18 technical articles, two papers, several RK-5 200 OOO 20 OOO 1:lO
books and reports, and eight manufacturer handbooks. Of
them all, only one even mentioned dc application of power
fuses. And you will find that those fuses, even when offered as
Many questions arise concerning dc interpretation of famil- protection for dc drives and for battery chargers, usually
iar fuse data and application criteria for ac-rated fuses. The carry a published interrupting rating only in ac amperes, such
answers to those questions vary with the fuse supplier. For as “200,000 RMS,” or “235,999 peak.” This simply in-
example: vites questions.
1) How will peak let-through current compare with ac Application literature on semiconductor fuses tends to be
based largely on the influence of various rectifier/inverter
values? (Answer: Usually less, because current rises
waveforms. That information does not aid the engineer seek-
more slowly with time, and no current offset exists. But
ing to apply fuses to a “pure dc” battery circuit. (Even in the
numbers are hard to get. One supplier suggests a
type of UPS application involving considerable harmonic
typical ratio of 1/2, dc to peak ac.)
ripple current through the battery itself, a matter now becom-
2) Will time-current melting curves based on 60-Hz ac
ing of concern to users, the ripple will be a minor component
conditions still apply? (Generally, yes, for dc or for
of battery fault current.)
other waveforms, provided an RMS value is definable.)
3) Can the well-known “up-over-down” method, using H o w Fuses Behave
curves published for ac operation, be applied to a dc
To understand some of the uncertainty regarding use of
circuit? (Probably not. One manufacturer says “All you
fast-acting, current-limiting (CL) fuses in dc circuits, we
can do is ‘up-over,’ to find the peak let-through cur-
rent, and divide it by two for a dc value.” Others may must examine fuse behavior. A fusible link overheats and
offer different rules of thumb.) melts after carrying a specific overcurrent for a specific time.
Melting opens the metallic path through which the current
All manufacturers of power fuses are being asked these had flowed. But, that gap in the circuit does not interrupt
and other questions about dc fuse behavior. Few have stock current flow. Rather, the flow continues via the arc devel-
answers. Most suppliers offer some dc products. But almost oped between the open ends of the fuse link.
none of the ratings are published. Few of the offerings are Arcing continues until the “arc voltage” across it rises
UL listed, or the equivalent. When comparing published high enough to equal the “driving voltage” of the power
information for a particular fuse (such as the RK-5), you may source. At that point, the arc will be extinguished, and
find that one or another of the following conditions may current flow ceases.
apply : In commercial power fuses, that process is aided by vari-
ous heat transfer mechanisms (such as a sand filling within
1) Both ac and dc current interrupting ratings are consid-
the fuse tube) that siphon off heat energy from the arc itself.
ered the same, but the voltage rating must be less for
In relatively slow-acting fuses, of course, another essential to
dc .
ac arc extinguishment is the periodic appearance of the
2) Both ac and dc voltage ratings are considered the same
“current zero.” Twice during each cycle of alternating cur-
-but the dc interrupting rating must be less (see Table
rent, the ampere flow passes through zero. Given sufficient
I, for example).
cooling or dispersion of the arc, that current zero will
3) Both current and voltage ratings are the same, whether
extinguish it, and it will not subsequently restrike.
the circuit is ac or dc.
Such is not the case for fast-acting CL fuses because arc
4) No dc current or voltage rating is offered.
extinction must take place long before a current zero point
Yet the basic fuse construction, and its behavior in an ac appears. So we might expect a CL fuse to behave the same
circuit, may be the same for most of the suppliers. way in a dc circuit as in the ac. However, that does not
As a convenient way around such seemingly irreconcilable happen. The reason is this: In an ac circuit, the effect of
alternatives, a prospective user may do well to select semi- circuit inductance is to cause an offset between voltage and
conductor fuses. These have been developed to protect power current waves (expressed as power factor). But the rate of
electronic devices in systems involving at least a pulsating dc rise of a fault current, derived from its power frequency, will
current [6]. Although the terminology and the rating structure be essentially the same as that of the associated voltage.
may seem unfamiliar at first, suitable choices can generally Given the system frequency, and knowing the magnitude of
be made. Despite the apparent suitability of semiconductor expected ac fault current, we will then know the rate of rise
fuses to battery protection, though, manufacturers’ literature of the initial fault current. The arc voltage within the fuse
on such fuses does not mention such use. responds to that rate of rise, expressed as d i l d t . The CL
NAILEN: BATTERY PROTECTION-WHERE DO WE STAND 66 1

Current

Fig. 1. Fault current versus time in a battery circuit (dashed curves) Number o f time constants following fault
compared with the initial fault current in an ac circuit (solid curve). initiation
Fig. 2. The solid curve is a plot of the time constant function given in the
text. For each time constant period, the circled point represents the RMS
fuse is designed to interrupt when that rate has been present current reached up to the end of that period.
for a time normally much less than half a cycle (Fig. 1).
In the illustration, the shaded area expresses the energy tance, thus
passing on downstream from the fuse for the duration of the i = ( v / ~- )A E - ( R / L ) ~
fault-the "let-through" energy. When the same fuse is used
in a dc circuit, containing some inductance even when a in which A is a constant dependent upon the circuit current
battery is the energy source and the load includes neither flowing at the instant of fault initiation. From that equation,
transformers nor motors, the situation is quite different. short-circuit current should follow a rate-of-rise curve such
Current rise rate is now determined by the circuit L / R ratio. that at the end of one time constant, the current will have
When inductance L is relatively low, current will rise almost reached 63% of its eventual maximum; after two time con-
in step with voltage. The latter does not follow a sinusoidal stants, 87%; and so on-see Fig. 2.
waveform rising at a rate by fixed frequency because no Here is an example of one method that has been used to
alternating frequency exists. Therefore, di / dt can become select battery protection fuses, based on use of such a curve:
extremely high. Result: The fuse blows even faster than in
the ac situation. Battery ampere hours = 330
Expected continuous current load = 48 A
Obviously, that's good, because let-through energy will be
One-minute current rating = 335 A.
held at a relatively safe, low level. What happens, though,
when the circuit inductance is high? Then we find a low Battery short-circuit current [7], [8] is calculated as
value of d i l d t . Fault current rises with time much more
slowly (Fig. 1). The resultant di / d t does not soon create an
arc voltage high enough to begin the arc extinction process.
= ( ampere rating ) (worst-case
1-min initial ) (4.4/K)
cell volts
Worse yet, when time tl is reached, the dc circuit offers no
current zero to promote interruption. Even though actual in which K is a temperature correction factor from IEEE 450
current is well below the peak value, it is able to persist so (1 .O may be used for this example). Thus:
long that the 12t let-through energy to the fault becomes
unacceptably high. I = (335)(2.33)(4.4) = 3440 A.

This is quite close to 10 times the 1-min ampere rating-a

The Circuit Time Constant rule of thumb suggested by [9].
As a result, dc capability of CL fuses is always discussed To examine fuse melting behavior, based on heat input, we
in terms of the circuit time constant-the L / R ratio-ex- must consider the RMS value of current reached within each
pressed in milliseconds. The higher that ratio, the relatively successive time constant, rather than the peak value reached
lower the degree of protection (the slower the operation) to at the end of each period. If each time constant is 10 ms, that
be expected from the fuse. For battery circuits, typical time yields the individual points plotted on Fig. 3, and the follow-
constants range from 10 to 80 ms. Both battery and fuse ing results for this example:
manufacturers tend to accept 10 as most representative. (For
comparison, note that one cycle of a 60-HZ ac waveform is Elapsed time, second Amperes, RMS
16 ms long.) 0.01 1440
As stated in [5], fault currents rise exponentially towards a 0.02 2130
value equal to the circuit voltage divided by circuit resis- 0.03 2550
662 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 21, NO. 4, JULYIAUGUST 1991

v)
V
r
0

WI
G

rz
r
0 10 20 30 40
W
r
L/R in milliseconds
c
Fig. 5 . Voltage derating curves for dc use of certain ac-rated fuses, as
0 supplied by one fuse manufacturer.
100 1000
Current in amperes
Fig. 3. Using a battery circuit fault current versus time plot (dashed line) Should the formula involving L / R be applied to the ultimate
to select a suitable fuse rating [lo].
value or to the transient peak?
Furthermore, accurate calculation of L and R is a difficult
*.5 task. The circuit may involve many relatively short lengths of
large cable, bundled in various ways, in differing proximity
to magnetic material such as cable trays. Resistance is vari-
able with battery conditions. All the values will be small so
that slight errors can have major effect.
Voltage Derating of Fuses
The arc voltage that a fuse must withstand will typically be
much higher than the system or source voltage. In a dc
circuit containing both R and L , voltage across the various
inductive and resistive portions will vary as current rises
during a fault. If L / R is high so that current rises slowly,
high voltages will develop across inductive elements of the
circuit, acting in a direction such as to maintain the current
0 o 10o 20o flow. Hence, dc use may subject the fuse to higher voltages
Elapsed time, milliseconds relative to the system level than would be true in ac service.
Fig. 4. One battery manufacturer [9] has suggested this variation in battery That gives rise to a “voltage derating” curve often (though
fault current accompanied by a drop in battery terminal voltage as shown.
This current variation does not suit the simple formula, involving an L / R
not always) used to govern application of ac power fuses in
time constant, cited in the text. dc circuits. Fig. 5 is an example. One supplier uses this
general rule: Rate the fuse for a dc voltage one half its
published ac voltage rating.
When those values are compared with melting curves for a A different concept of voltage derating is offered by an-
typical family of semiconductor fuses, we see from Fig. 3 other fuse supplier. The increased let-through energy prior to
that a 250-A fuse is the highest usable rating. Because the fuse operation is compounded by another effect that is also
maximum continuous current is only 48 A, a 50-A fuse could peculiar to dc behavior. The higher the current-whatever
be used. However, the fuse manufacturer recommends a fuse the time constant-the faster the fuse will operate because the
rating at least double the continuous current to minimize arc tends to establish itself at multiple locations within the
thermal fatigue of the fuse. Therefore, a 100-A rating is fuse chamber. In the absence of ‘a current zero, dc interrup-
selected. (Note: At least one other fuse supplier draws en- tion depends entirely upon the fuse’s ability to divide and
tirely different conclusions from similar curves, emphasizing cool the arc. Therefore, because a lower fault current estab-
that no universally accepted practice exists.) lishes an arc at only one location, extinguishment is more
Unfortunately, this usage of the L / R constant in discus- difficult than with the higher current.
sions of dc fuse behavior is not a true picture because it Unfortunately, just as in the 480-V ac system where the
assumes a constant voltage acting across a constant imped- relatively low-current burndown fault is now widely recog-
ance; hence, an ultimate value of fault current that is also a nized as the greatest hazard, UPS designers are well aware
constant. But battery voltage actually decays rapidly during that low-current battery system faults are more common than
the initial 5 or 10 ms of a fault. Thus, actual current is not a the bolted fault short circuit. A fuse that interrupts relatively
true exponential function but varies with time as in Fig. 4. less predictably on low current can therefore be particularly
Current does indeed stabilize at a value dependent upon troublesome.
battery voltage and circuit resistance. . . but not until after The approach taken by one manufacturer-unpublished,
passing through a transient peak well above that value. but based upon information available upon request-is to
NAILEN: BATTERY PROTECTION- WHERE DO WE STAND 663

VI
600 U
c
0
U
W
VI

575 --

rz,
0

550
L

-g
W
U.
a
+2

525 /

: L + L L J
500
0.01 0.02 0.05 0.1 0.2 0.5 1.0

difdt, amperes per microsecond

Fig. 6 . Differing from the approach shown in Fig. 5, this voltage derating 10 ioo lob0 10 ,boo
method used by another fuse supplier is based on the expected di / d t , which Amperes
is the rate of fault current change with time.
Fig. 7. For dc fuse tests, UL Standard 198L prescribes minimum circuit
time constants versus ampere ratings that lie on the solid curve. The dashed
line expresses test range UL specifies for mine duty fuses in 198M.
allow the full ac voltage rating in dc circuits, provided di / dt
exceeds a stated minimum. For smaller d i l d t values, a
voltage derating curve is to be followed; Fig. 6 is an exam-
ple.
No matter how they are arrived at, dc voltage ratings may
contain little leeway. If the value given is 250, for example,
the fuse may not be usable in a battery circuit where equaliz-
ing voltage becomes 280.
So much for voltage ratings. What about current? One
might intuitively expect that interrupting capacity ought to be
lower in a dc circuit than for ac. The higher let-through
energy with long time constants, as well as lack of current
b
Fig. 8. Battery protection using (a) a single midpoint fuse versus (b)
zero, supports that expectation. And that is often confirmed conventional line fuses.
by published supplier data. Table I is representative of such
data. For these fuses, the published voltage ratings are fuses having interrupting ratings over loo00 A. The fuse
generally identical, ac or dc. Confusing? Certainly. must not permit peak let-through current exceeding 80% of
Keep in mind, however, that whatever theory of fuse its interrupting rating.
behavior may be propounded based upon considerations such Some ac power fuses on the market today have been tested
as Fig. 1, published ratings-and particularly third-party against UL 198L. A few have been tested for dc mine circuit
listings such as those of UL-must depend upon testing. To use, per UL 198M, “Mine-Duty Fuses.” But most fuses do
quote one supplier engineer, “Tests to verify dc capability not carry specific, tested ratings for listed dc use. Be pre-
are extremely costly and the market does not support doing pared, even when dc ratings do appear in a catalog, to
much.” negotiate particular battery circuit application with the fuse
What test standards exist? Only one-UL 198. Most of it manufacturer. A local jobber, or manufacturer’s representa-
less than ten years old, this standard is divided into a number tive, may not be able to answer your questions.
of sections dealing with specific fuses. Of greatest general In any event, find out what dc testing has been done. Look
interest is UL 198L, “D-c Fuses for Industrial Use” [ll]. for specific data supporting dc current and voltage ratings.
The document includes three principal test requirements. Don’t expect, however, to find a strong experimental founda-
First, and apparently most difficult to perform, is the “over- tion underlying the final choice.
load test.” This requires single fuses to interrupt 200% of
rated continuous current at rated dc voltage, the voltage to be Fuse Location and Coordination
“maintained for 1 minute after circuit interruption to deter- Assuming a properly rated fuse to be available, how
mine that the fuse has permanently cleared the circuit.” should it be used in the circuit? Fig. 8 illustrates two com-
During this test, the L / R time constant of the test circuit mon methods of applying battery fuses. For a short circuit at
must not be less than the values expressed by Fig. 7. (or on the load/charger side of) the battery terminals, these
Second is the “interrupting ability” test. With a circuit two connections are electrically equivalent.
time constant of at least 10 ms, the fuse must clear its But, the version at Fig. 8(a) is a connection unique to the
interrupting rating at rated dc voltage, with voltage remaining battery system as contrasted with ac power supplies. Because
across the fuse for 30 s after interruption. large batteries are distributed energy sources, faults can
Finally, UL 198L calls for a “maximum energy test” on occur involving only portions of the entire battery. This
664 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 27, NO. 4, JULYIAUGUST 1991

Fig. 10. Fusing of individual cell strings in a series-parallel battery.

distributed so simply. Rather than a single ‘ ‘midpoint, ’ ’

many may exist. More than one midpoint fuse could be used.
Fig. 9. Battery construction may allow “internal” faults to go undetected
by line fuses (upper). A midpoint fuse (lower) will interrupt such faults. However, each fuse and its associated clips adds two undesir-
able conditions to the circuit. One is added series resistance.
The other is the greater likelihood of unexpected fuse failure
‘‘midpoint fusing’’ scheme furnishes protection against many that would remove an unfaulted battery from the line.
such “intra-battery’ faults. One authority on battery usage has cited NEC Article 480
In battery rooms, cell groups rest on racks or tiers, one as a basis for midpoint fusing. He interprets the Code as
above the other-usually two high and sometimes three. requiring a maximum battery voltage of 250 within the
That’s not apparent in the simple electrical diagram of Fig. 8. jurisdiction of any one protective device. Neither battery nor
But examination of the physical arrangement itself, as in Fig. UPS manufacturers mention this in their literature as a
9, shows that any fault between the upper and lower requirement or a general practice, however. What the 1990
tiers will not be sensed by fuses at the main terminals. Code says (Section 480-6) is merely that lead-acid cells
Placing a fuse at the midpoint between tiers, though, will “shall be installed in groups having a total nominal voltage
result in fault interruption. of not over 250 volts.” (For rubber cell containers, the limit
Are such faults likely? Even in relatively well-maintained is 150 V.)Although the NEC says that “maximum protection
but unattended installations, cell leakage (a common cause of can be secured by sectionalizing high-voltage batteries into
battery trouble) has caused corrosion sufficient to partially groups,” no recommendation is made as to “group” size,
collapse a rack. The resulting localized fault current involv- nor is any mention made of protective device “jurisdiction”
ing some portion of the two tiers can be quite destructive. or selection.
The midpoint fuse prevents that. The combining of cells into convenient groups can also
Better maintenance, or a different sort of physical arrange- make parallel fusing desirable. In large batteries including
ment, can make the midpoint fuse ineffective or unnecessary. multiple strings of series cells that are then paralleled on a
But, many battery installations now employ midpoint fusing common bus, each string can, depending upon its physical
when it can be valuable. arrangement, be involved in a fault that will not appear on
Several utilities have adopted midpoint fusing as a conse- the bus [12].Some UPS suppliers therefore provide individ-
quence of major battery failures, and not just for substation ual string fusing; others do not. (See Fig. 10).
batteries. One large utility experienced total destruction by The circuit of Fig. 8(b) or Fig. 10 could use a single fuse
-
fire of a 576 A hr, 125-V battery some years ago, in an in one leg of the circuit rather than two fuses. However,
unattended location. What apparently happened was a rack except in the telecommunications field, large batteries are
collapse allowing part of the battery to spill and discharge normally operated ungrounded [ 131. Accidental grounds are
into an area that then caught fire. Fault current would have common in batteries because of electrolyte leakage to sup-
been detected by a midpoint fuse but not by main terminal porting structures. If one ground should develop on a portion
protection. of the battery at opposite polarity from an already grounded
Elsewhere, more recently, a substation battery fire de- leg, the result would be a short circuit that a single line fuse
stroyed 20 cells when a charger fault allowed short circuiting might not see. On any battery circuit lacking intentional
of the main battery terminals. (Depending upon the apparatus grounding, a ground detector is recommended to reveal the
configuration, that could occur in some installations even if first accidental ground so that it can be removed before a
fuse protection were provided within the charger itself.) The second one occurs. Some engineers feel, based on ac experi-
resulting damage totalled $250 0oO because the fire broadcast ence, that arcing bumdown faults are much less common in
conductive soot throughout the station control room, requir- ungrounded circuits. However, relevant battery system oper-
ing a great deal of electronic control and other equipment to ating data do not appear to be available.
be carefully cleaned or replaced. What about circuit breakers? Standard molded-case ac
This is not to say that such fusing is the answer in all breakers are usable in dc circuits. However, the dc interrupt-
situations; far from it. For the largest systems, cells are not ing rating will typically be “one-half to one-third” [14]of
 NAILEN: BATTERY PROTECTION-WHERE DO WE STAND 665

TABLE I1
MOLDED-CASE
CIRCUIT
BREAKERRATINGSLISTED
BY VARIOUS
MANUFACTURERS
SHOWINGTHE WIDEVARIATIONI N DC
VERSUS
AC CAPABILITY

Mfr. code Voltage rating Interrupting duty, amperes

ac dc ac dc dc lac
240 1251250 10000 5000 0.5
240 250 18000 10000 0.56
240 250 65000 10000 0.15
240 250 22000 10000 0.45
240 250 10000 5000 0.50
240 250 25000 10000 0.40
600 250 -various-
6ooto1000 300 -various-

the ac value. (See Table 11). Many manufacturers publish no

dc application data at all for these breakers. Several makes
have been tested, but are not UL listed, for dc use. The
advantage of the breaker, of course, is that it functions as a Fig. 11. Typical wiring between battery cells and central control equip-
disconnection device as well as a fault interrupter. ment for one make of battery condition monitor.
Fuse-breaker coordination, when both are provided, is
simple enough in the ac circuit. The breaker handles over- Sensors themselves typically respond directly to cell volt-
loads, and the fuse handles short circuits. However, coordi- age. If an intercell connector is open, so that neither charging
nation is based on performance data for the devices used, nor load current can flow, cell voltage will change. Similar
such as the “up-over-down” fuse let-through curve. Such changes occur if cell sulfation takes place. Monitoring cir-
published information seldom exists for the dc situation. cuits are self-checking so that a monitor defect itself is
If provided on the charger output alone, neither fuses nor a translated into a condition alarm.
circuit breaker can be counted on to interrupt a battery fault. One monitoring system places a sensor within a group of
Chargers often contain internal current limiting that will hold cells totaling 12 V nominal. That voltage level best suits the
output current to 125- 150% of rating. Under that condition, unit’s resistance /voltage sensing circuits. The supplier rec-
the charger output voltage may simply “fold back” to essen- ommends shifting the position of the sensor month by month
tially zero. Low voltage will cause automatic disconnection to cover the entire battery in rotation. Each sensor comes
of the charger from the battery. In all these situations, a fault with 3-ft pigtail leads for attachment at the bolts holding
within the battery itself could persist without hindrance. intercell connectors in place. Normally closed dry contacts in
Battery technologists are quick to point out that fuse or all the sensors are connected in series. Thus, either operation
circuit breaker protection, wherever located, is of limited of any sensor or a break in the contact circuit will give the
value in ensuring battery reliability. That is certainly true. alarm.
Open-circuited intercell connectors, high terminal resistance, In a typical battery, proper cell voltage may range from a
sulfated plates, or any condition involving abnormally low low of 1.81 (above the usual l-min discharge level of 1.75)
(or high) terminal voltage will not be sensed by overcurrent to a float or equalizing high of 2.25 to 2.33. A cell monitor
devices. Estimates vary, but between 80 and 90% of all may therefore be set to a high limit of 2.4, a “minor low”
battery faults are said to involve “the conducting path”-the level of 1.81 for early warning, and a “major low” limit of
series resistance within and between cells, particularly the 1.75.
intercell connectors where looseness and corrosion are com- The largest single-cell monitoring system, involving two
mon. Nevertheless, when a relatively sudden short-circuit dozen central computers linked by fiber-optic cables, was
does involve the battery, no matter how infrequently, only recently completed for a government installation in the West;
fast and automatic circuit interruption can be counted on to wiring of the type illustrated in Fig. 11 was connected to
avoid widespread external damage. 8OOO cells. The entire battery, cell by cell, can be scanned
within 8 s.
BATTERY
CONDITION
MONITORS Monitor costs will increase with the number of cells in-
As a safeguard against unexpected battery failure caused volved, but not linearly, because the central control portion
by some of the nonfault conditions just mentioned, the “bat- of the system represents a single cost for a range of battery
tery monitor” is often recommended. Several makes are now sizes.
available. In general, these include a small local sensw that is The disadvantage of the battery monitor is in the sope-
installed at each cell or small group of cells. The sensors are times extensive sensor wiring needed. In the more sophisti-
wired to one or more electronic monitoring units that gener- cated units, the sensors supply information to a computer
ate alarm or relaying signals based on cell condition such as where battery condition is displayed on a monitor screen.
undervoltage, changes in connection resistance, or even low Another type of monitor, applied at the load terminals to
electrolyte level. Condition alarms may be local or remote. check the entire battery, measures circuit resistance by draw-

-
666 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 21, NO. 4, JULYIAUGUST 1991

ing repeated short-time (microseconds) current pulses from itself. When a fuse fails without load on the battery, how-
the battery. If a succession of such pulses is below a preset ever, the condition is likely to go undetected until an emer-
limit, the alarm operates. gency transfer to the battery fails to restore power. Whether
The choice of battery protection- particularly an either-or terminal or midpoint fuses are used, then, blown-fuse indica-
decision between fuses and monitors-is complicated for any tor switches should be provided to cause an alarm whenever
kind of standby system (such as an off-line UPS) by the any fuse opens. Such switches are a readily available fuse
normal absence of any load current. When a battery floats on accessory.
a charger, required to deliver load current only for infrequent Who is responsible for battery protection? Questioning
brief periods, neither thermal nor mechanical stress on com- those experienced in design or application of large battery
ponents is continually present. If it were, weak points would systems elicits conflicting answers. Says one, “In UPS sys-
show up gradually to be observed as they developed, and tem protection of the battery is largely left up to the user.”
timely corrective action could be taken while the problem Of the batteries installed in telecommunications or substation
was still minor. facilities, he added, “98% of them go in without any discon-
However, when the battery must instantly jump from float nect or fuse outside of the charger.” According to one
to peak overload, no one has any chance to respond to a trend battery supplier [7], “The designing engineer must review
alarm indicating that some problem is developing. If that the applicable codes and pertinent factors of the application
problem results in a fault, circuit interruption must be fast before deciding whether or not to install protective
and unequivocal. Whereas a fuse might not do that because devices. . . The designing engineer, however, generally
”

of uncertain selection criteria, a monitor can certainly never does not know what device application information is needed
do it. Even linking monitor output to a circuit breaker will or what is available. Says the battery system expert, “Here
not adequately limit fault let-through energy. are the types of fuses available; go ask the fuse manufacturer
what’s best to use.” The fuse manufacturer, in turn, says
PROTECTION PHILOSOPHY
what’s best to use depends upon the behavior of the battery
Either because of economics, uncertainty about the circuit.
prospects of failure, or the lack of any specific protection Probably few large UPS users are fully aware of the
guidelines in published industry standards, many large batter- overcurrent protection (if any) provided on their battery
ies are not being protected at all. Among these are two system. Nor do UPS manufacturers appear fully aware of the
experimental energy storage systems, one containing 8256 protective options. When asked how fuse or circuit breaker
cells in multiple 2OOO-V strings and the other containing 324 protection was selected, most manufacturers offered no reply.
-
cells, each of more than 2000 A hr capacity, able to supply When this paper was proposed, one of the engineers
0.5 MW. These, however, are carefully observed, well-con- consulted expressed the hope that solutions might be sug-
structed facilities that are still quite new. gested to battery problems that have generated “some real
There is precedent for omitting battery overcurrent protec- horror stories out there. ” Leaving large batteries unprotected
tive devices entirely on the premise that nuisance operation is no answer. For the present, the “real solution” appears to
(taking the battery out of circuit) is potentially more costly be the provision of carefully chosen battery protective de-
than a battery failure [7], [13]. Under NEC Section 430-31, vices to the maximum affordable extent, recognizing that a
for example, normal motor overload protection may be omit- major fault can spread damage far beyond the boundaries of
ted for fire pump drives because even though overloaded, the battery room or the battery circuit. In the long run, the
such a drive could be essential to the control of a fire that industry needs better standards for the selection and use of
could endanger lives and property if allowed to go unchecked. high-power dc protective equipment.
However, the two situations are not quite analogous, as In 1989, recognizing the growing importance of batteries
this comparison shows: in industrial and commercial power systems, the IEEE formed
Possible Consequences a new Standards Coordinating Committee, SCC-29, charged
Failure Nuisance Trip Only with overseeing and coordinating all standards written deal-
UPS Personnel hazard plus Little or no personnel or ing with installation, maintenance, and operation of station-
Battery extensive property loss property damage ary batteries. But although a number of such standards exist,
we clearly need a “protection standard” offering recommen-
Fire Personnel hazard plus Personnel hazard plus
dations concerning protective methods, and guidelines on
Pump extensive property loss extensive property loss
protective device selection, to fill an important gap in the
Thus, although great inconvenience and high monetary loss literature.
can result from a battery nuisance trip, the severity of the
consequences cannot normally be equated with those result- REFERENCES
ing from a serious battery fault-or from any unexpected [l] R. R. Fletcher, “Cell selection considerations,” Trans. IEEE En-
ergy Conv.,vol. EC-1, no. 4, pp. 69-72, Dec. 1986.
suspension in the operation of a fire pump. The more preva- [2] M. F. Migliaro, “Maintaining maintenance-free batteries,” in IEEE
lent opinion [4], [9], [ 151 is that battery systems should have Conf. Paper, Ind. Comm. Power Syst. Conf.(Chicago, IL), May
some sort of overcurrent protection. 1989.
[3] IEEE 446-1987, “Recommended practices for emergency and standby
Nuisance tripping that occurs while a battery carries load, power systems for industrial and commercial applications.”
inconvenient though that may be, will at least call attention to [4] IEEE 946-1985, “Recommended practice for the design of safety-
NAILEN: BATTERY PROTECTION-WHERE DO WE STAND 661

related dc auxiliary power systems for nuclear power generating and commercial switchgear installations,” Trans. IEEE Industry
stations.” Applications, vol. 1A-20, no. 3, pp. 667-671, May/June, 1984.
A. Wright and P. G. Newbery, Electric Fuses. London: Peter
Peregrinus, 1984.
T. M. Cmko, “Current-limiting fuse update-New developments: Richard L. Nailen (M’S-SM’68) was bom in
New style fuse for protection of semiconductor devices,” in IEEE San Jose, CA, on January 2, 1928. He received the
Conf. Paper Petrol. Chem. Industry Conf. (Dallas), Sept. 1977. B.E.E. degree with honors from the University of
“Stationary lead-acid battery systems,” Sect. 50.00, Exide Corp., Santa Clara, CA in 1950.
Form 7637, Jan. 1983. From 1953 to 1964, he was employed in the
F. L. Brennan, “DC distribution system,” Power Plant Electrical Motor Engineering Section, Westinghouse Electric
Reference Series, Electric Power Research Institute, vol. 9, 1987. Corp., Sunnyvale, CA, on electrical and mechani-
G. Walker, “Short circuit characteristics of lead acid batteries,” cal design of ac motors and generators through
unpublished Tech. Bull., C & D Battery Co., Jan. 1987. 19000 kVA. In 1964, he joined the Louis Allis
“Form 101 semiconductor fuse application guide,” Gould Shawmut, Co., Milwaukee, WI, from which he retired in
1981. 1985 as Engineering Specialist, Large Machinery
UL 198L, Standard f o r D-C Fuses f o r Industrial Use (2nd ed.), Engineering, where he was involved in design and application of induction
Mar. 1988. motors to 10000 hp. Now a Project Engineer with Wisconsin Electric Power
“Study of the auxiliaries for lead-acid battery systems for peaking Company in Milwaukee, he is concerned with application, maintenance,
power,” Final Rep. Westinghouse Electric Corp. for Energy Re- repair, and testing of all motors in the utility’s generating stations, as well as
search and Development Administration, U.S. Government Printing with utility customer problems involving motors and controls. He is the
Office, Dec. 1977. author of more than 300 published articles and 20 IEEE papers on motor
E. C. Korbeck, Jr. and J. W. Blankley, “Selection, use, and care of design and application, as well as many other subjects in electrical technol-
stationary batteries for paper mill service,” Trans. IEEE Ind. Gen. ogy, and the 1991 book, Managing Motors. He has presented numerous
Appl., vol. IGA-7, no. 6, pp. 742-749, Nov./Dec. 1971. University Extension courses, seminars, and workshops.
R. G. Brunner, “Auxiliary electrical equipment,” Power Plant Elec- Mr. Nailen is a member of the Petroleum and Chemical Committee of the
trical Reference Series, Electric Power Research Institute, vol. 7, IEEE, the Power Systems Engineering Committee, the Professional Commu-
1987. nications Society, Tau Beta Pi, the National Fire Protection Association, and
B. Bridger, Jr., “Control and auxiliary power systems for industrial is a registered Professional Engineer in the State of Wisconsin.