This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON POWER DELIVERY 1

On Human Life Risk-Assessment and Sensitive

Ground Fault Protection in MV Distribution Networks
José L. Pinto de Sá, Senior Member, IEEE, and Miguel Louro

Abstract—The minimum phase-to-ground fault current re- some limitations [5]. Meanwhile, many ordinary zero overcur-
quired to be sensed by protection systems in medium-voltage rent relays, either directional or nondirectional, are in the field.
(MV) networks can be as low as 0.7 A in a few countries, leading Utilities make different assessments about the safety risk of
to a lot of undesired relay trips and poor service quality to cos-
tumers. However, these settings raise the protection threshold high-resistance ground faults, but they all are pressed to pro-
above the minimum fault current that concerns network operators vide the best possible service, which brings necessary com-
regarding human safety, although they cannot be practiced when promises between protection sensitivity and security [6]. While
grounding is distributed. The purpose of this paper is to present a some utilities are comfortable with settings of 10% of rated cur-
risk assessment foundation to determine the required protection rent and above for ground fault protection in distribution net-
sensitivity to ensure human safety in MV distribution networks.
The proposed approach is based on a biophysical model included works, others have a tradition of more sensitive protection. In
in IEC standards, the consideration of current paths models for Europe, where there is no neutral conductor in MV feeders and
typical faults and Monte Carlo methods to deal with nonlinearity, sensitive ground relaying is common, a number of Distribution
and the many involved random variables. Downed conductors utilities require the detection and elimination of ground faults
and line-to-concrete pole faults are investigated and sensitivity with as low as 1 A [7] and even less [8]. To cope with non-
analysis performed to highlight some important determinants of
the results. transposed lines asymmetry and charging current unbalances,
which also produce nonfault zero currents, this high sensitivity
Index Terms—Monte Carlo methods, overcurrent protection,
was traditionally the role of a relay dealing with the current
safety.
of the substation grounding connection, which is nonselective
by nature. Zero overcurrent relays in feeders were set above
I. INTRODUCTION 5–7.5 A to avoid tripping with nonfault zero currents [8], and
the relay at the substation grounding connection was set to 1
A or less and timed to 2 to 3 min. Usually, this last relay trig-
H IGH resistance phase-to-ground faults are dangerous
events, because they can be very hard to detect. Of these
faults, downed conductors have motivated a lot of investigation
gers an automatism, which works by sequentially disconnecting
the feeders until the relay resets and the faulted feeder is found.
[1]. Don Russell and his team [2] developed a very detailed re- For insulated grounding systems, a zero overvoltage relay at the
search to find that a number of detection methods were efficient substation bus is used instead. However, excessive use of this
but also that it could be most difficult to discriminate a fault practice led to many consumer complaints and with time, new
situation from “normal” electrical events on the distribution protection approaches have been developed which are more se-
feeders. As those authors noted, if an algorithm regularly iden- lective [9].
tifies normal system activity as a “fault,” it is virtually useless Anyway, many utilities around the world still face a dilemma
in practical relay applications. between guaranteeing sensitive ground faults relaying and
On the other hand, in overhead line networks, faults with avoiding power-quality (PQ) complaints, especially for feeders
very high resistance can appear due to trees leaning against a which are disconnected too often. Although the quest for sen-
conductor, for instance. VTT researchers in Finland have devel- sitive and yet secure ground fault protection keeps deserving
oped a very sensitive algorithm [3] capable of coping with these a lot of research, to these authors’ knowledge, a quantitative
evaluation of how dangerous a high resistance ground fault can
faults, but field monitoring has shown that a number of nonfault
events could also produce the voltage disturbances upon which really for human life in MV distribution networks was never
the authors’ methods were based. published. Exceptions exist only in some particular industries
Some of the aforementioned protection methods have been where risk conditions are well characterized, as in coal mines
[10].
refined and are now suitable for the processing power of modern
microprocessor-based protective relays [4], but they all present At this point, we should stress that a direct contact between a
person and an ungrounded live part of an MV power system will
probably lead to the appearance of an arc flash or to tissue elec-
Manuscript received December 18, 2009; revised May 23, 2010; accepted
troporation [11]. In either case, it is unlikely that a person will
June 14, 2010. Paper no. TPWRD-00944-2009. survive without severe injuries. Electroporation occurs in just
J. L. Pinto de Sá is with the Instituto Superior Técnico, CIEEE, Lisbon 1049- 5–40 ms and, therefore, it is impossible for a relaying system
001, Portugal (e-mail: [email protected]). to prevent its consequences to the victim [12]. However, it is
M. Louro is with EDP Distribuição, Lisbon 1050-044, Portugal (e-mail:
[email protected]). possible to minimize the amount of energy transmitted to an arc
Digital Object Identifier 10.1109/TPWRD.2010.2053564 flash by means of fast tripping [13]. Anyway, for arc flash and
0885-8977/$26.00 © 2010 IEEE
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE TRANSACTIONS ON POWER DELIVERY

Fig. 2. Conventional time/current zones of effects of ac currents (15 Hz to 100

Hz) on people for a current path corresponding to left hand to feet ([15]).
Fig. 1. Occurrence of the vulnerable period of ventricles during the cardiac
cycle (from [15]).
For a given contact time, ventricular fibrillation can be mod-
eled by a lognormal probability function that fits the three curves
electroporation, currents are so large that the issue is protection
of Fig. 2, as shown by animal experiences.
speed, not sensitivity.
It must also be reminded that the curves in Fig. 2 are for a
The origin of most high-resistance faults, according to the
current passing through the heart from the left hand to the feet.
experience of a few utilities, including the one for which one
For other body paths, the values for the current differ and are
of the authors works for, is the conductor touching the ground,
usually larger, which is accounted for by the IEC 60479 standard
MV/low-voltage (LV) transformer faults, and cable failures be-
through “heart factors” relating the fibrillation current
tween tower anchor clamps [9]. Therefore, the purpose of this
in a body path and this reference path
paper is to offer a human safety risk assessment for the most
typical high-resistance ground faults endangering human life in (1)
MV distribution networks in order to provide a quantitative basis
for the definition of an informed tradeoff between sensitivity and Besides ventricular fibrillation, muscle tetanic contraction
security requirements in ground-fault protection. may be a severe hazard when stepping on a conductor on the
ground, because it can cause knees to flex and the person to
II. REVIEW OF THE PROBABILISTIC LETHALITY fall. C. Dalziel has conducted a lot of research which is not in
OF THE ELECTRIC CURRENT contradiction with IEC 60479 assumptions and provides further
In high-resistance ground faults, small currents are present, data [16] about the current thresholds for tetanic contraction.
and that is why protection sensitivity is an issue. For these small Although Dalziel’s research was focused on hand muscles’
currents, the most important lethal effect is ventricular fibrilla- tetanization, and hence, on “let-go” current thresholds, recent
tion and, therefore, some established facts about this phenom- results point to the validity of extending those results to other
enon must be reminded [14]. body extremities [17].
As Fig. 1 illustrates, ventricular fibrillation can only occur
if an electrical current overlaps the nervous electric heart con- III. MODELING COMMON GROUND FAULTS INVOLVING
trol in the recovery period of the cardiac cycle, which lasts for HUMAN CONTACTS
only 10% of the entire cycle. This current has also to reach a In what follows, line-to-tree faults will not be addressed be-
minimum threshold, but if it persists in time, only 5% of it is cause they do not define an immediate danger for human safety,
required to disturb the cardiac cycle, reducing its duration and although it is known that they can ignite fires.
lowering the fibrillation threshold. This is the reason for the “S”
shape of the threshold-time functions illustrated in Fig. 2 (from A. Downed Conductors
[15]). Undetected downed conductors are a human safety issue be-
Since the number of heart cycles is about 70–80 for a healthy cause the distances between live parts of the power system and
person and can reach a hundred under physical effort, the ven- the public are diminished beyond acceptable levels. A downed
tricular fibrillation threshold starts lowering only after 0.1 s, to conductor may have a very low fault current even for low resis-
reach a steady 5% of its initial value in about 2 s. tivity soils; for example, if the open circuit is from the source
Another important fact is that the fibrillation threshold varies side and the line is back fed from MV distribution transformers,
from one person to another. In Fig. 2, there are three curves: or in resonant or insulated networks.
c1, c2, and c3. Under c1, the fibrillation probability is zero, c2 Step Voltage Near a Downed Conductor: When somebody
is reached for 5% of the people, and c3 for 50%. This means, touches a downed conductor with a foot, the current will flow
for example, that even though the absolute minimum fibrillation from one foot to the other (Fig. 3).
threshold may be 0.04 A, for a current of 1 A lasting less than The situation presented in Fig. 3 is represented in Fig. 4 to
0.1 s, approximately 95% of the people will survive. highlight the associated electrical phenomenon.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

PINTO DE SÁ AND LOURO: ON HUMAN LIFE RISK-ASSESSMENT AND SENSITIVE GROUND FAULT PROTECTION 3

Fig. 3. Step voltage in case there is foot contact with a downed conductor. Fig. 5. Electrical equivalent of step voltage near a downed conductor.

two layers model) and a constant that depends on the top

layer depth and the difference between top soil and bottom soil
resistivity [18]

(5)

These expressions follow what is usually derived for substa-

tions grounding design and are not new [19], [20].
The calculation of the grounding resistance for
a conductor laid on soil can be done through the formula ([19,
p. 167])

Fig. 4. Step voltage in case there is foot contact with a downed conductor (pro-
file view).
(6)

The ground potential rise (GPR) from the downed conductor where
is obtained by
radius of the conductor;
(2)
2L buried horizontal wire length (a portion of );
where s/2 depth of burial.

length of the downed conductor in contact In this formula, all of the variables are to be defined. It has
with the soil; been found that for a particular point of a downed conductor,
the fault current to soil tends to be random in magnitude be-
grounding resistance of the downed cause of varying contact effects of grass, roots, bushing, and
conductor; rocks that cause intermittent ground contact paths of variable
short-circuit current to ground. impedance. Arcing disturbs the ground path by rapid heating,
drying, and expansion of soil, charring of glass and the melting
The voltage applied to a person stepping on the downed con- of sand into glass, in a complex and dynamic way. However, at
ductor can be determined by the following equation (refer to least for conductors lying on soil for a few yards (or meters), it
Fig. 4): appears that experience confirms the findings in [21] in which
the arcing from the conductor to the ground is seen as equivalent
(3)
to a flat plate. This arcing contact is modeled in (6) by a depth
where of burial.
Substituting (5) in (4), the following equation is obtained:
step voltage (in relative value of GPR);
(7)
step length.
The step voltage can also be obtained from Fig. 5 and ex- For the case represented in Fig. 5, (8) can be derived by re-
pressed by placing (7) in (3)

(4) (8)

Of course, in (8), all of the involved variables are random,

The resistance of the foot contact to ground can be rep- namely, the length of the laid conductor, the step length, the soil
resented as a function of the top soil resistivity (assuming a resistivity, the type of shoes and their electrical resistance, the
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE TRANSACTIONS ON POWER DELIVERY

Fig. 7. Touch voltage across the human body.

Besides heart fibrillation, a current from hand to feet can

also cause tetanic muscle contraction in those body extremities,
for values much lower than those causing fibrillation. Anyway,
by observing (12), it is apparent that it also depends on many
random factors, such as soil resistivity, shoes electrical resis-
Fig. 6. Touch voltage when touching a faulty MV pole.
tance, body path for the current and its heart factor, applied
voltage, and the person fibrillation threshold.
body electrical resistance, and the person fibrillation threshold,
although they exist between some limits. IV. ASSUMPTIONS FOR A RISK ANALYSIS
B. Pole Faults A. Probabilistic Approach
Touch Voltage Near a Pole: Another common indirect A probabilistic approach is how these authors propose dealing
contact can occur when there is a fault to a metallic or steel-re- with the randomness of the variables involved in the studied
inforced concrete pole and a person accidentally touches it. events. Risk analysis as related to electric injuries resulting from
This hazard is more probable when a pole-mounted distribution ground faults is not a new idea [22], but to the authors’ knowl-
transformer has a fault to its grounded tank, particularly when it edge, it was never applied to protection practices.
is located near houses, but it also occurs with broken insulators. Traditional grounding theory uses pessimistic values for all
Touch voltage is the most limiting aspect because the proba- of the random variables present in danger assessment. This im-
bility of the electric current passing through the heart is higher plies the assumption that all of the unfavorable situations will
than for step voltage. occur simultaneously, which rarely occurs. This prudent prac-
The case of an indirect contact is illustrated in Fig. 6. It is tice makes sense when designing grounding systems for instal-
assumed that the pole in Fig. 6 does not have a ground return lations with an expected life of several decades and with opera-
wire, as is most usual in MV lines. Should the pole have a ground tional conditions not entirely foreseeable.
return wire, the results would probably be less constraining. However, the goal of this paper is to determine the fault cur-
Considering Fig. 6, the touch voltage to the human body can rent that is harmful for people stepping on downed conductors
be calculated through and touching poles under fault in real conditions. A determin-
(9) istic analysis based on the worst conditions would lead to a pes-
simistic result of almost null probability and requiring too much
The pole resistance depends on the placing of the protection sensitivity. On the other hand, it appears wise to relax
ground electrodes and the soil resistivity. The placing of the the protection sensitivity to determine the associated risk. Thus,
ground electrodes is uniform, so we have in what follows, we start by proposing probability distributions
for the involved random variables.
(10)
B. Probabilistic Assumptions for Common Random Variables
In Fig. 7, the touch voltage is presented as depending on the
resistance of the human body, the shoe resistance, the grounding Soil Resistivity : According to data from a local utility
resistance of the feet, and the current passing through the human (EDP), its most common soil resistivity in dry weather is 150
body. The touch voltage can be obtained by using m and the largest identified value was 37 k m. It was consid-
ered that the soil resistivity follows a lognormal-type distribu-
(11) tion and, fitting that data, a median value of 150 m with 95%
of the soils having a resistivity below 2000 m was found. Soil
resistivity depends a lot on seasonal moisture and regional vari-
Equation (12) establishes the relationship between the short- ations, and in other regions, other statistic parameters can fit the
circuit current to an MV pole and the current passing through a data better.
person’s heart Nonuniform Soil : The larger the soil resistivity is, the
worse it is. The top soil layer usually has a larger resistivity than
(12)
the average soil because it is more exposed to water evaporation.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

PINTO DE SÁ AND LOURO: ON HUMAN LIFE RISK-ASSESSMENT AND SENSITIVE GROUND FAULT PROTECTION 5

TABLE I
SHOE PROPERTIES AND PROBABILITY OF OCCURRENCE

Fig. 8. Electric potential near a downed conductor.

In reference [23], a 30% resistivity increase in the top soil layer

due to evaporation is presented.
According to the weather profile of Portugal, it was found that Conductor Lengths : This is a most determining
for 150 days of the year, the top layer resistivity is identical to variable for a conductor-to-ground resistance, which is in-
the average resistivity. In the remaining days of the year (dry versely proportional to the contact length defined in (6). On
weather), the top layer resistivity should be 30% above the av- one hand, the average span between MV poles depends
erage resistivity. on the voltage level. For the lowest voltage in MV overhead
Human Body Resistance : Statistics for the human body lines, it can be around a hundred yards (or meters), varying up
resistance are presented in [16], which follow a lognormal-type to 1/3 that value. On the other hand, when an overhead line
distribution [24]. However, its average value depends on the path falls on ground, only a portion of that maximum value
considered for the current: it is somewhat larger for foot to foot will be in touch with the soil—or close enough for arcing to
than for hand to feet. IEC 60479 provides means to evaluate the perform the grounding.
relative impedance for different current paths. Reports from maintenance crews point to two types of reasons
Heart Current : If the heart current surpasses a cer- for a line to break in regions where it is unusual to have lines
tain threshold, ventricular fibrillation is probable. The data pre- running on roadsides and, hence, driving accidents are not a
sented in [15] are used for the fibrillation current. A lognormal cause for line falling. The most usual reason is simply breaking
probability function is the most adequate for accessing the fib- on the points where tensile efforts are at maximum, which are
rillation current [24]. at the poles suspensions. Trees and human artifacts cause other
Shoe Resistance : Grounding design guides usually fractures, which have a higher occurrence at midspan, where the
assume barefooting in their models. However, for risk assess- conductors are at their minimum height. Considering all of these
ment in real life, barefooting is uncommon, at least in industri- aspects would lead to a complex probability distribution, but
alized regions. Data on shoe resistances can be found in [25] we consider that a uniform distribution is a reasonable model,
and are being used for grounding design in Australia and in the taking into account that it probably underestimates .
U.K. Table I shows those data, with minor adjustments to local Step Voltage : The step voltage near a
conditions. downed conductor depends on the step length and the potential
The voltage withstand thresholds of the shoes have to be near the conductor. Fig. 8 represents the potential near a downed
compared to the actual value calculated in the analysis. If the conductor. This data were obtained through a simulation ([26]).
threshold is violated, any considered algorithm must be ad-
justed for this circumstance. This, of course, makes the analysis Step Length : The standards for grounding design al-
of the statistics very nonlinear. ways take the step length at its maximum, 1 m. However, as for
barefooting, that model is too pessimistic for modeling reality;
C. Specific Probabilistic Assumption for Downed Conductors hence, a step length with a uniform distribution probability be-
tween 0 and 1 m was assumed. Probably a lognormal distribu-
Conductor Section and Burial Depth : To simulate a tion would fit reality the best, but in the end, the convergence of
downed conductor, a very low burial depth can be assumed (e.g., all random variables will always lead to results approximating a
0.002 m). For illustrations purposes, a 25-mm section was also lognormal distribution, regardless of the assumed distributions
assumed as typical of MV conductors. These parameters have a for each variable, thanks to the central limit theorem.
reduced effect on the conductor-to-ground resistance; thus, they
were not taken as random. For example, a 10-times increase in
D. Specific Probabilistic Assumptions for Faulted Poles
the depth burial (justifiable to model arcing into the soil near
the conductor contact) only reduces the grounding resistance in Pole Grounding Factor : This factor represents the re-
13%. lationship between the pole grounding resistance and the soil av-
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE TRANSACTIONS ON POWER DELIVERY

Fig. 10. Fault current for 0.5% probability of heart fibrillation when stepping
4
on a downed conductor: —with random shoes, —barefooted.

Fig. 9. Touch voltage dependence on arm length.

erages resistivity. In this study, some particular grounding prac-

tices for MV poles yield [27]

(13)

Arm Length : The IEEE [19] states that for project pur-
Fig. 11. Fault currents versus probability of a fall when stepping a downed
poses, an arm length of 0.8 m must be used, but this is too pes-
simistic for average real conditions. Therefore and in order to
4
conductor. —with random shoes, — barefooted, X- with only one type of
shoes (dry new black rubber).
obtain a realistic assessment of the arm length, a uniform distri-
bution probability between zero and 0.8 m is proposed.
Touch Voltage : The touch voltage (obtained from A. Case for Downed Conductors
[27]), which depends on the arm length, is represented in Fig. 9.
A first investigation was done for the fibrillation risk when
Heart Factor : The heart factor depends on the point stepping on the downed conductor. Fig. 10 shows the achieved
of entry of the electric current in the human body. If this point results.
is the left hand, then the factor is 1, but if it is the right hand, the Perhaps surprisingly, the fault currents causing immediate
heart factor is 0.8 [15]. We have assumed a probability for the fibrillation are quite large, even for barefooted people. It must
point of entry being the right hand and equal to that of the left be reminded that this is not the current through the person, but
hand (50%). the fault current from the conductor to ground, for a 0.5% prob-
ability of fibrillation with random variability of all the parame-
ters. For a typical step time of 2/3 of a second, more than 100 A
V. RISK ANALYSIS THROUGH MONTE CARLO METHODS are needed even for barefooted people. One of the reasons for
this is that the current that passes through the heart in case of
Monte Carlo methods work by repeated random sampling to the current entering one foot and leaving the other is only 4% of
compute results. With readily available inexpensive personal the total body current [15].
computers and powerful software, such as MATLAB, Monte Unfortunately, the current threshold that makes the leg mus-
Carlo methods are an easy approach to estimate risk for the situ- cles stop to respond to voluntary actions and causes knee flexing
ations under study, particularly considering the large amount of and loss of balance is much smaller [18]. Once a person falls,
random variables and the nonlinear behavior of the shoes. For there is a large chance that the fall will be on the conductor, and
example, for computing equations (8) and (12), the method was that is why this hazard is what must be considered for stepping
applied as follows. downed conductors.
1) In the Monte Carlo simulation, random numbers that Fig. 11 shows how the falling risk grows with the fault cur-
follow the presented random variables probability distri- rent. Since muscle contraction has nothing to do with the cardiac
bution were generated. cycle, now the contact time is irrelevant.
2) Five million random data were calculated for each random Even a fault current of 1 A can make 1% of the peopl fall,
variable (using each random variable probabilistic dis- and 5% of them if barefooted. The risk of falling for people with
tribution functions) by using MATLAB, followed by the dry new black rubber shoes is equal to the average shoes up to
final equations calculation for that amount of data. Finally, 10 A, but above this current, there are other shoes, with larger
the five million results for the equations were sorted resistances, which increase the amps required to cause a fall.
and the thresholds verifying particular probabilities were Fig. 12 illustrates the importance of the soil resistivity and the
determined. length of the conductor making contact with the ground, for the
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

PINTO DE SÁ AND LOURO: ON HUMAN LIFE RISK-ASSESSMENT AND SENSITIVE GROUND FAULT PROTECTION 7

Fig. 12. Fault currents versus fall probability when stepping a downed con-
4
ductor. —for a short portion of the conductor on the ground (5 m); X- for
very resistive soil (5  ); - all random (reference); —for low resistivity soil
( =5); 5—for a long portion of the conductor on the ground (95 m).
Fig. 14. Time to heart fibrillation versus fault current for: different probabili-
ties. From left to right—0.5%, 2.5%, 5%, and 25%.

Fig. 13. Fault currents versus heart fibrillation probability for touching a
downed conductor. —falling on it, with chest making contact, 4—with the
hand, standing 1.5 m from the point the conductor touches the ground.

risk of a fall. These results confirm how high-resistance soil and

a small length of contacting conductor greatly increase the risk
for the same fault current, which must be taken into account if
the probability for these occurrences is considerable. Fig. 15. Time-to-heart fibrillation versus fault current for different assump-
tions, all for a 0.5% risk of death.
Once a person falls, there is a chance to fall on the downed
conductor. The risk of fibrillation can be estimated considering
as the worst condition the chest touches the conductor and both
bare hands touch the ground, at a random distance between 0 the related “let-go” current is much smaller; therefore, once a
and 1 m from the conductor. Fig. 13 shows the probability for hand touch is made to a faulted pole, it will likely remain until
this hazard, and the risk of fibrillation for touching the hanging heart fibrillation occurs, unless the current is small enough or
portion of the conductor with a hand when standing, instead of the fault is eliminated by some protection. For standing faults,
stepping on it. the asymptotic current is what matters, for example, 4 A for a
The heart fibrillation probability when falling on the con- death risk of 0.5%.
ductor is very similar to the probability of muscle tetanization Of course, these results are for random variables in which
for stepping on it, but the probability of both occurring is ap- shoes have a determining importance. Fig. 15 illustrates how
proximated by their product; hence, it can be said that for the barefooting as well as considering only a maximum distance
considered statistics, the risk of dying from stepping on the con- (0.8 m) from the feet to the pole will affect the results (all of the
ductor will be only 0.5% if the fault current is less than around remaining variables are still random). For touches lasting longer
4 A. However, for this current, the risk of dying from touching than 2 s, barefooting reduces the current required for heart fib-
the hanging conductor with a hand will be 4 times greater. rillation from 4 A to 2.9 A, with the same risk of 0.5%, under
the considered statistics.
These illustrative results can be compared as shown in
B. Case for Faulted Poles
Table II.
The application of the same analytic methodology to the case From the obtained results, it can be said that to reach a very
of a faulted pole provides the results of Fig. 14 for a growing high safety level, fault currents to ground above 3 to 5 A shall
risk of heart fibrillation. The amount of time until heart fibril- not be allowed, but below these values, the risk of death is very
lation occurs depends on the touch duration, but unfortunately, low.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE TRANSACTIONS ON POWER DELIVERY

TABLE II low zero-sequence currents and still provide a measurement ac-

MINIMUM FAULT-TO-GROUND CURRENTS FOR RISK OF DEATH curate enough for the use of sensitive wattmetric relays. This
accuracy usually demands a CT with a primary current vector
addition or flux summation, in which the conductor cables pass
through the center hole of the core balance CT. For overhead
lines, the use of this CT is only possible if the line has a cable
portion between the substation bus and its first pole, which is a
usual design in many European regions.
Nonresonant Networks With Neutral Grounded Only in the
VI. EXPORTING THE RESULTS OF THE RISK ANALYSIS TO THE Substations: In networks that are ungrounded or grounded only
DESIGN OF SENSITIVE ROUND PROTECTION SCHEMES in the substations, without a neutral conductor in the feeders,
phase-to-ground faults usually have considerable currents,
The results reached in the previous section define fault cur- which are easy to detect by protective relaying, particularly if
rents that can endanger human life for a calculated level of prob- flux-summation CTs are available to provide good measure-
ability. The analysis leads to values for currents because electric ments of zero-sequence currents. However, there are two very
current is the lethal agent in the human body, once electropora- different situations to consider:
tion (which has more to do with voltage) and arc flash are ex- Backfed Downed Conductors and Faulted Poles: A downed
cluded. Of course, if different assumptions are made about the conductor broken near the suspension point in the source-side
random behavior of the independent variables, particularly the pole can have a considerable length on soil with good resis-
soil resistivity and the features of the shoes, the same risk levels tivity and still be characterized by small fault currents. A pole-
will be yielded by different currents. However, it is unlikely that mounted transformer with a phase fault to its inner grounded
realistic parameters however different will change these values tank creates a similar situation if the fuse of the faulted phase is
in orders of magnitude. blown.
On the other hand, the evaluation of the total risk depends on It can be shown that the phase-to-ground backfed fault cur-
the probability of the events themselves (people moving under rent is limited to 1/3 of the prefault load current in the line,
overhead lines, how the lines are maintained, and how they are unless the fault resistance is really large, and that the residual
clamped to their poles, the materials used to manufacture MV current ( the zero-sequence current) on the source side of
poles, how these poles are grounded, etc.). In addition, of course, the line is equal to that fault current [26]. Therefore, sensitive
it is up to each community to decide on what the risk is that can zero-sequence overcurrent relays can guarantee human safety
be afforded. against the hazards previously investigated if they can be set to
Notwithstanding, if a very low risk of casualties is desired, currents as small as 3–5 A, about the same amps to which reso-
(e.g., 0.5%), phase-to-ground faults yielding more than a few nant grounding usually limits any phase-to-ground fault current.
amps cannot be tolerated. Under this constraint, the suitable pro- However, if the region served by the network is populated
tection schemes depend very much on the type of network and by barefooted people, has poorly maintained equipment, and
particularly on its grounding philosophy. In what follows, some moderate power-quality (PQ) requirements, a sensitivity as low
typical scenarios will be addressed. as 1 or 2 A may be a better option at the substation neutral
Networks With Resonant Grounding: Networks with reso- connection.
nant grounding (through arc-suppression coils, also named Pe- Nonbackfed Phase-to-Ground Faults: If zero-sequence over-
tersen coils), can limit the current of any phase-to-ground fault current relays are set to 3–5 A, they will be very sensitive to
to very small values. A perfect resonance would eliminate that faults fed by the source, unless a very short length of conductor
current, but real systems cannot do it, first because the coils are makes contact with very high-resistance soil. However, setting
not perfect inductances and always have some resistance, and an overcurrent relay to 3–5 A can make it sensitive to charging
second because a perfect tuning would create a large voltage un- zero currents from the feeder to a fault somewhere else. Means
balance even for very small phase asymmetries. Real systems to guarantee selectivity include varimetric directionality in un-
can reduce the current of a phase-to-ground fault to an extent grounded networks and an inverse-time behavior for the other
capable of suppressing most arcs, and that is precisely its best grounding options.
value, but cannot eliminate the fault current from a downed con- Networks With Distributed Grounding: In networks with
ductor or to a pole from a pole-mounted faulted transformer. distributed grounding, it may be difficult to distinguish by the
However, from the risk analysis presented before, it can be amount of amps between currents flowing to ground through
said that if a persistent fault current to ground is less than 3 to 5 planned ground connections, or those from a downed con-
A, the risk of human casualties may be really small. Sustained ductor. For low-resistance faults, this is not an issue because
operation with this fault is precisely the usual practice in Aus- currents are large and easy to sense by usual relays. However,
trian and German utilities [28], where grounding through Pe- faults involving short lengths of conductors contacting high-re-
tersen coils is an old and established tradition and fault currents sistance soils can yield low fault currents and be precisely the
are most often reduced to a few amps with 3 A being a typical most dangerous (see Fig. 12).
reference. Hence, there is no doubt that high-resistance faults in net-
For a phase-to-ground fault to be detected in a feeder in those works with distributed grounding will be better identified
networks, the current transformers (CTs) have to cope with very through the signature of their waveforms, not through their
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

PINTO DE SÁ AND LOURO: ON HUMAN LIFE RISK-ASSESSMENT AND SENSITIVE GROUND FAULT PROTECTION 9

magnitude. In other words, those are the scenarios where the [14] G. Biegelmeier and W. R. Lee, “New considerations on the threshold
solutions developed by Don Russell and others and now com- of ventricular fibrillation for AC shocks at 50–60 HZ,” Proc. Inst. Elec
t. Eng. A., Phys. Sci. Meas. Instrum. Manage. Educ. Rev., vol. 127, no.
mercialized by a number of manufacturers are most valuable. 2, pp. 103–110, 1980.
[15] Effects of Current on Human Beings and Livestock—Part 1: General
Aspects, IEC 60479-1, IEC, 2007.
[16] C. F. Dalziel and W. R. Lee, “Re-evaluation of Lethal Electric cur-
VII. CONCLUSION rents,” IEEE Trans. Ind. Appl., vol. IA-4, no. 5, pp. 467–467, Sep. 1968.
[17] L. A. Geddes and R. M. Fish, “Conduction of electrical current to and
through the human body: A review,” Open Acess J. Plastic Surgery,
Beginning with a biophysical model for human physiology, a Oct. 2009.
set of random variables was identified as determining the risk of [18] Guide for Safety in AC Substation Grounding, IEEE Std. 80, IEEE,
2000.
heart fibrillation for hazards related to downed conductors and [19] Recommended Practice for Grounding of Industrial and Commercial
faults to poles. By applying Monte Carlo methods, it was found Power Systems, IEEE Std. 142, IEEE, (Green Book), 2007.
that to guarantee a very low risk of death, fault currents above a [20] AIEE Committe Report, “Voltage gradients trough the ground under
fault conditions,” Trans. Amer. Inst. Elect. Eng.s: Power Apparatus and
few amps shall not be sustained. Downed conductors and faulted Systems, Part III, vol. 77, no. 3, pp. 669–685, Oct. 1958.
poles yield similar fault current limits. [21] D. Jeerings and J. Linders, “Ground resistance revisited,” IEEE Trans.
Fulfilling the requirement to limit ground fault currents to Power Del., vol. 4, no. 2, pp. 949–956, Apr. 1989.
[22] W. D. Carman, “Probabilistic comparison and application of interna-
a few amps depends on how a network is grounded. Special tional electrical safety criteria,” in Proc. Int. Conf. Power System Tech-
relays are most valuable where other methods cannot be applied, nology PowerCon, Australia, 2000, pp. 1347–1352.
as in multigrounded networks or in lines where sensitive zero- [23] T. Klügel, G. Harnisch, and M. Harnisch, “Measuring integral soil
moisture variations using a geoelectrical resistivity meter,” Bull. d’Inf.
sequence overcurrent relays cannot be used. Marées Terrestres, vol. 141–142, pp. 11369–11375, Aug. 2006.
[24] J. M. Nahman, “Assessment of the risk of fatal electric shocks inside a
substation and in nearby exposed areas,” IEEE Trans. Power Del., vol.
5, no. 4, pp. 1794–1801, Oct. 1990.
REFERENCES [25] W. D. Carman and D. J. Woodhouse, “Performance evaluation of series
impedance insulation used as earthing system safety mitigation mea-
[1] PSRC Working Group D15, High Impedance Fault Detection Tech- sures,” in Proc. Int. Conf. Power System Technology, Australia, 2000,
nology, IEEE. New York, 1996. pp. 1353–1358.
[2] B. Don Russell and C. L. Benner, “Arcing fault detection for distribu- [26] J. Pinto de Sá and M. Louro, “Evaluation of protection approaches to
tion feeders: Security assessment in long term field trials,” IEEE Trans. detect broken conductors in distribution networks,” presented at the
Power Del., vol. 10, no. 2, pp. 676–683, Apr. 1995. 19th Int. Conf. Electricity Distribution, Vienna, Austria, 2007, paper
[3] S. Hanninen, M. Lehtonen, and E. Antilla, “A method for detection 023.
and location of high resistance earth faults,” in Proc. Int. Conf. Energy [27] J. Pinto de Sá and M. Louro, “Speed versus sensitivity in earth fault
Management and Power Delivery, 1998, vol. 2, pp. 495–500. protection regarding human safety in aerial MV networks,” presented
[4] M. Adamiak, C. Wester, M. Thakur, and C. Jensen, May 2009, High at the 19th Int. Conf. Electricity Distribution, Vienna, Austria, 2007,
impedance fault detection on distribution feeders. [Online]. Available: paper 345.
http://www.geindustrial.com/pm/pr/hiz_fault_dist.pdf [28] CIRED Working Group 03, Fault management in electrical distribution
[5] L. Li and M. A. Redfern, “A review of techniques to detect downed systems CIRED, 1998.
conductors in overhead distribution systems,” in Proc.7th Int. Conf.
Developments in Power System Protection, 2001, pp. 169–172. José L. Pinto de Sá (SM’90) was born in 1951. He is
[6] M. A. Bostwick, “Distribution protection as used on the portland a Professor of Electrical Engineering at the Instituto
general electric company system,” AIEE Trans., pp. 1081–1086, Dec. Superior Técnico, Lisbon, Portugal. He has worked in
1959. substations advanced automation since 1981. In the
[7] J. Curk, J. Gorisek, and D. Koncnik, “Improved protection scheme for 1990s, he led the development of multifunction pro-
selective high-resistance earth fault clearing in Slovenian distribution tective relays for EFACEC, and from 2001 to 2006,
networks,” presented at the PowerTech, Budapest, Hungary, 1999. he was with LABELEC as a Coacher and a Consul-
[8] D.E.R., La protection des réseaux aériens a moyenne tension. Elec- tant. His main research interests are in power sys-
tricité de France/Centre de Normalization, Spécification d’entreprise tems protection, protective relaying, and substations
74 H2 00219 1967 (in French). automation.
[9] L. Karsenti and P. Cabanac, “EDF’s field experience on MV networks
zero-sequence protection scheme,” presented at the CIRED, Vienna,
Italy, 2007, paper 129.
[10] T. Novak, L. A. Morley, and F. C. Trutt, “Sensitive ground-fault re-
laying,” IEEE Trans. Ind. Appl., vol. 4, no. 5, pp. 853–861, Sep./Oct. Miguel Louro was born in 1978. He received the
1988. M.Sc. degree in electrical and computer engineering
[11] H. W. Tinsley, M. Hodder, and A. M. Graham, “Arc flash hazard cal- from the Instituto Superior Técnico, Lisbon, Por-
culations: Myths, facts and solutions,” in Proc. Conf. Rec. IEEE Ind. tugal, in 2008
Appl. Soc. Pulp and Paper Conf., 2006, pp. 1–7. He was with LABELEC from 2002 to 2007. His
[12] R. C. Lee and W. Dougherty, “Electrical injury: Mechanisms, manifes- main tasks were to provide expertise toward power
tations, and therapy,” IEEE Trans. Dielectr. Elect. Insul., vol. 10, no. systems protection for power distribution and gener-
5, pp. 810–819, Oct. 2003. ation. He is now in the Network Analysis department
[13] Guide for Performing Arc-Flash Hazard Calculations, IEEE Std. 1584, of EDP Distribuição, where he focuses on large-scale
IEEE, 2002. incident analysis and protection systems.