Directional Overcurrent Relaying (67) Concepts

John Horak, Member, IEEE

Basler Electric

Abstract Directional overcurrent relaying (67) refers to lagging current conditions rather then 1.0 power factor
relaying that can use the phase relationship of voltage and conditions. One approach, seen in Fig. 1, is to phase shift the
current to determine direction to a fault. There are a variety of voltage signal so that the relays internal voltage signal
concepts by which this task is done. This paper will review the (VPolarity, abbreviated as VPol) is in phase with current when
mainstream methods by which 67 type directional decisions are
made by protective relays. The paper focuses on how a numeric
current lags the 1.0 power factor condition by some setting,
directional relay uses the phase relationship of sequence typically between 300 and 900. The angle setting is commonly
components such as positive sequence (V1 vs. I1), negative referred to as the maximum torque angle, MTA. In some
sequence (V2 vs. I2), and zero sequence (V0 vs. I0) to sense fault designs of this concept, the current signal is skewed rather
direction, but other concepts such as using quadrature voltage than the voltage signal. In some designs, other phase voltages
(e.g., VAB vs IC) are included. are used. For instance, IA could be compared to VAB, VCA,
VBN, or VCN, and the detection algorithm would work, though
Index Terms: directional relaying, sequence component, the quadrature voltage VBC gives the most independence of the
negative sequence, zero sequence, 67, 32, quadrature voltage. voltage signal from the effects of an A-N, A-B, or A-C fault.

I. INTRODUCTION
In some medium voltage distribution lines and almost all high
voltage transmission lines, a fault can be in two different
directions from a relay, and it can be highly desirable for a
relay to respond differently for faults in the forward or reverse
direction. The IEEE device number used to signify a
directional element is either a 21 (impedance element, based
on Z=V/I, and having a distance to fault capability) or a 67
(directional overcurrent, generally based on the phase
relationship of V and I, with no distance to fault capability).
Some applications also might use a 32 (power element, based
on P=Re[VxI*]) for directional control, though in some
circumstances a 32 element may not be a good indication of
direction to fault. This paper will review some of the various
implementations of 67 elements as found in
electromechanical, solid state, and numeric (i.e., multifunction
programmable logic microprocessor based) relays.

II. CLASSICAL CONCEPTS FOR DIRECTIONAL ANALYSIS

The classic electromechanical and solid state relay, as well as
some common numeric relays, determines the direction to
fault by comparing the phase angle relationship of phase
currents to phase voltages. If only per phase watt flow (32
element) is to be considered, the basic concept would be that if
IPh is in phase with VPh-N (0, 90), then power flow on that
phase is indicated as forward (or reverse, depending on ones
perspective). However, for a phase to ground fault, the VPh-N
may collapse to 0, and I may be highly lagging, so that VPh-N x
Fig. 1. Classic VQuadrature Directional Element
IPh may be mostly VAR flow, and thus prevent the relay from
making a correct directional decision. To resolve the low The response of the design to a phase to ground and phase to
voltage issue, quadrature voltages (i.e., VBC vs. IA) are phase fault is shown in Fig. 2. The response to a phase to
commonly used. To resolve the issue that fault current is ground fault is fairly apparent because the quadrature voltages
typically highly lagging, the relay current vs. voltage detection are assumed to be relatively unaffected by the faulted phase
algorithm is skewed so that the relay is optimized to detect currents. However, for a phase to phase fault, the quadrature
Prepared by Basler Electric for Presentation to the
Georgia Tech Protective Relaying Conference, April 2005
Revision Date 02/15/06
voltages are affected. The effect is difficult to give in text. One conditions as reverse fault current, as seen in Fig. 3. An
should study the diagram to develop an understanding. approach to addressing this condition is to set the MTA to 300
Basically, in the ph-ph fault, relative to ph-ground fault, note or less, so that the reverse zone reaches minimally into the
that both Vquadrature and Ifault have shifted by 300, so there is no forward zone.
net change in tendency of the element to operate.

Fig. 3. Power Flow vs. MTA

III. SYMMETRICAL COMPONENTS FOR DIRECTIONAL

ANALYSIS
Many modern microprocessor relays use the angular
relationships of symmetrical component currents and voltages
and the resultant angular nature of Z1, Z2, and Z0 as calculated
from Vphase/Iphase to determine direction to fault. These three
impedances are used to create, respectively, three directional
assessments, 67POS, 67NEG, and 67ZERO, that are used in relay
logic in various ways by each manufacturer. There are
variations among manufacturers on of how one senses the
angular relationships and, in most cases since the angular
relationship is the only concern, the magnitude is not
calculated. The common concept is that in faulted conditions
there is an approximate 1800 difference of calculated Z1, Z2
and Z0 for faults in the two directions from the relay location.
This high variation in phase angle is a reliable indication of
direction to fault.

As described in detail in reference [1], the three phase voltage

drop equation for a system that can be represented by voltages
at two defined locations, (VSys and VFault in this example) is

VA, Sys VA, Fault Z AA Z AB Z AC I A

Fig. 2. Phasors in Classical VQuadrature Directional Element VB , Sys - VB , Fault = Z BA Z BB Z BC I B . (1)
VC , Sys VC , Fault Z CA Z CB Z CC I C
The MTA setting is commonly thought of in terms of the
forward-looking line impedance angle. This would be
particularly true if the relay simply compared voltage and Again, as discussed in [1], when the impedances are highly
current from a common phase for a line to ground fault (e.g., balanced (i.e., the diagonal self impedance elements ZAA, ZBB,
IA is compared to VAN). In this case, the relay is sensing ZA and ZCC are all one value, and all off diagonal mutual
between the relay and the fault. However, when quadrature impedance elements are another value), (1) can be restated in
voltage is used, then VPol is somewhat independent of the fault symmetrical component quantities by the equation
current, especially for a phase to ground fault. The angle by
which current lags quadrature voltage is a factor of both V0, Sys V0, Fault Z 0 0 0 I0
source impedance as well as forward-looking line impedance,
so a compromise value is utilized. An MTA in the range of 300 V1, Sys - V1, Fault = 0 Z1 0 I1 . (2)
to 750 is common. When setting MTA, if an overcurrent V2, Sys V2, Fault 0 0 Z 2 I 2

element is to be set below reverse direction load current, there
is a risk of the element seeing abnormal forward load

2
In the typical power system, we can usually assume that, at the V0, Relay
remote system, voltage has very low V0 and V2, and V1 is 1.0, Z 0,Relay = = -Z 0, Sys (6)
I 0, Relay
or at least very close to 1.0. At the other end, the fault
location, every type of fault will have differing values of V0, V2, Relay
V1, and V2 and will need to be calculated via means that will Z 2, Relay = = -Z 2, Sys . (7)
I 2, Relay
not be covered here (see [1]), but we know that some value
exists. Hence, (2) reduces to
Note that in (6) and (7) the equations for Z0,Relay and Z2,Relay,
the impedance seen by the relay will be dependent solely upon
0 V0, Fault Z 0 0 0 I0 the source impedance. (The dependency on source impedance
V
0 I1 .
1, System - V1, Fault = 0 Z1 (3) might be counter-intuitive to engineers accustomed to setting
0 V2, Fault 0 0 Z 2 I 2 impedance relays in terms of line impedances.) The angle of
Z0,Relay and Z2,Relay is the source of determining the direction to
a fault. For instance, in Fig. 1, a CT polarity orientation can
If Z0, Z1, and Z2 are divided into two impedances as seen from cause the apparent Z0 and Z2 at the relay to either match the
the relay location (line impedance and source impedance), the source impedance angle or to be inverted by 1800. The current
net system and associated voltage drop has the appearance of polarity would be the signature of a fault that is either forward
Fig. 4. or reverse from the relays location.

The apparent Z1 at the relay will be dependent on the fault

type. First, we need to define the impedance between the relay
and the fault location:

Z1, Line , Flt = Z1 from relay to line fault location (8)

For a three phase fault

V1, Relay V1, Sys - Z1, Sys I1, Relay

Z1, Relay ,3 ph = =
I1, Relay V1, Sys
. (9)
Z1, Sys + Z1, Line , Flt
= Z1, Line , Flt
Fig. 4. Single Source System with Relay
Phase to ground and phase to phase faults are more
In this application, (3) can be restated as complicated, but have a similar derivation to (9):

0 V0, Fault Z1, Relay , Ph -GND =

(10)
V1, Sys - V1, Fault = Z 0, Sys + Z 0, Line , Flt + Z1, Line, F + Z 2, Sys + Z 2, Line, Flt
0 V2, Fault (4)
Z1, Relay , Ph- Ph = Z 2,Sys + Z1, Line, Flt + Z 2, Line, Flt . (11)
Z 0, Sys 0 0 Z 0, Line 0 0 I 0, Relay

0 Z1, Sys 0 + 0 Z1, Line 0 I1, Relay
0 In these cases, the Z1 measurement as seen at the relay is a
0 Z 2, Sys 0 0 Z 2, Line I 2, Relay
mix of the various system and line impedances, but it tends to
be a measure of forward looking line impedance more than
The voltage division of (4) allows us to calculate the voltage at source impedance, most clearly seen in (9). The impedance
the relay by starting at the fault location and working back to angle of the various components of Z1,Relay tend to be similar,
the system or starting at the system and working toward the in the area of 500 - 850. The calculated angle of Z1,Relay again
fault. Since we do not know the fault voltages, we need to take has a 1800 phase angle reversal depending on direction to the
the latter approach, so we can calculate relay voltage from fault, which is the signature of a fault that is either forward or
reverse.

V0, Relay 0 Z 0, Sys 0 0 I 0, Relay If a relay uses Z1,Relay for sensing direction to fault, the Z1,Relay

V1,Relay = V1, Sys - 0 Z1, Sys 0 I1, Relay . (5) measurement will see balanced load flow as an indication of
V2, Relay 0 0 0 Z 2, Sys I 2,Relay the direction to fault and, hence, to turn on overcurrent

elements (67/51) that are set to look in the direction of present
load flow. The Z1 that is sensed during balanced load flow
If we solve (5) for the impedances, since V2,Sys = 0 and V0,Sys = conditions is a minor modification of (9):
0, then
Z1, Relay = Z1, Line + Z1, Load . (12)
3
 would fall into either the forward or reverse zone, depending
The angle of Z1 can be a poor indicator of fault location. For on relay setup and CT connections. Note that in an impedance
instance, when a customers DG (distributed generator) exists plot, MTA is counterclockwise from the reference R axis, as
to peak shave, it has the ability to control the power factor at compared to the classical approach of showing I relative to V
the PCC (point of common coupling). Power swings that where MTA is clockwise from the reference VA axis.
occur in post fault conditions give transient VA flow at almost
any angle. On a more steady state basis, a DG that runs to
keep power at the PCC near zero could cause net power factor,
and hence the angle of Z1, to be almost any value.

Directional overcurrent relaying would not be useful in a

system with only one source. A system with two sources is
shown in Fig. 5. We can apply the same concepts as above to
analyze the circuit. We have a fault location where there is a
calculable level of sequence voltages, and system and line
impedances in two directions, looking back toward the system
source voltage. The same approach as in (1) to (7) can be
applied to find the impedance seen by the relays at either end
of the line.

Fig. 5. Two Source System

The impedance as seen by relay A will vary according to the

direction to the fault. For faults on the two different sides of
the breaker, the relay will sense two completely different
impedances. For fault FA,F and FA,R the impedance seen by
relay RA will be:

Z 0, Relay , Fault A, For = -Z 0, Sys , A

Z 0, Relay , Fault A, Rev = 11800 ( -Z 0, Sys , B - Z 0, Line )
= ( Z 0, Sys , B + Z 0, Line )
. (13)
Z 2, Relay , Fault A, For = -Z 2, Sys , A
Z 2, Relay , Fault A, Rev = 11800 ( -Z 2, Sys , B - Z 2, Line )
= ( Z 2, Sys , B + Z 2, Line )

In (13), Z#,Line refers to the entire line impedance. The 11800 Fig. 6. Forward and Reverse Impedance Angles
factor in (13) accounts for the effective change in CT polarity As previously mentioned, the various relay manufacturers
for faults in the reverse direction. The positive sequence have differing ways of sensing angular relationships. The most
impedance does not lend itself to simple equations such as obvious process is to measure the impedance angle and
(13), but for 3 phase faults and unfaulted load flow conditions compare it to a window of the MTA +/-900 as forward or
reverse. There are alternate processes in use. For instance, as
Z1,Re lay , Forward , Faulted = Z1, Line, Flt seen in Fig. 7, one manufacturer configures its Z2,Relay and
Z1,Re lay ,Re veverse , Faulted = Z1, Sys , A,to Fault Z0,Relay directional elements to subtract the MTA from the
. (14) calculated impedance, and then find the real portion of this
Z1,Re lay , Forward ,Unfaulted = Z1, Line + Z1, Load , B resultant impedance, ZReal, which can be a positive or negative
Z1,Re lay ,Re verse ,Unfaulted = Z1, Sys , A + Z1, Load , A value. Then, ZReal is compared to user settings for the decision
points for forward and reverse. For the great majority of cases
this gives the same result as the phase angle window.
A graphical representation of the forward and reverse zones of
protection can be seen in Fig. 6. The MTA is a user setting
that effectively defines forward and reverse phase angles.
Sensed impedance angles that are +/- 900 from the MTA
4
 and 67/51Q (negative sequence) elements and similar 67/50
elements, and that each has a forward or reverse looking mode
with different settings for each direction. There are three
directional elements called the 67POS (positive sequence),
67NEG (negative sequence), and 67ZERO (zero sequence) that
control the 67/51 and 67/50 elements. The protective elements
and their directional controls are:

A given relay may have more than one copy, or no copy, of

the indicated element, and a given relay may or may not give
the user direct access to 67POS, 67NEG, and 67ZERO.
Fig. 7. Alternate Process to Using Phase Angle

TABLE 2 - TYPICAL DIRECTIONAL ELEMENTS

IV. VARIATIONS OF ZERO SEQUENCE DIRECTIONALITY Directionally Controlled by

67/51 Elements 67/50 Elements (typically)
Zero sequence directionality has several variations. The V0
67/51P - Forward 67/50P - Forward 67POS or 67NEG = Forward
and I0 used in a Z0 measurement each can be obtained from
various inputs: 67/51P - Reverse 67/50P - Reverse 67POS or 67NEG = Reverse
67/51G - Forward 67/50G - Forward 67ZERO or 67NEG = Forward
V0 as calculated from the 3 phase VT inputs 67/51G - Reverse 67/50G - Reverse 67ZERO or 67NEG = Reverse
V0 as seen on a 4th auxiliary VT input on the relay (VX 67/51Q - Forward 67/50Q - Forward 67NEG = Forward
below). This VX input can be connected to a variety of 67/51Q - Reverse 67/50Q - Reverse 67NEG = Reverse
sources, such as:
o a broken delta VT, or
o the neutral of an impedance grounded generator. VI. OTHER ISSUES
I0 as calculated from the 3 phase CT inputs. There is a number of subtleties involved in the forward/reverse
I0 as seen on a 4th CT input on the relay (IG below). This direction decision and element operation that will not be
auxiliary CT input can be connected to a variety of covered here. One should refer to the various relay
sources, such as: manufacturers instruction manuals for details on their relays
o a window CT that wraps all 3 phases, algorithms. Some issues that need to be understood that can
o a window CT that wraps all 3 phases as well as a vary by manufacturer implementation:
power carrying neutral conductor,
o the neutral of a transformer, or A. Memory Polarization
o the neutral of a generator. For close-in three phase faults, the voltage at the relay may
The result is that there are 5 different combinations of currents fall to near 0. Due to the low voltage, the relays 67 logic
and voltages that can be used to create a directional 67ZERO cannot be relied upon to make a correct directional analysis
decision, and in some relay configurations, if the relay cannot
The last item in Table 1 uses only current for the directional determine forward or reverse, it does not trip at all. To address
decision and is sometimes referred to as zero sequence current this issue, numeric relay manufacturers create a memory
polarization. The MTA is always 00. If the two currents are in polarization scheme. The relay constantly is reading the
phase (+/-900), the fault is forward. present voltage and using it to create a voltage vector (V1). If a
fault occurs that suddenly drives voltage too low to be used for
TABLE 1 - VARIATIONS OF ZERO SEQUENCE POLARIZATION directional analysis, the relay reaches back to its memory and
67ZERO projects the past voltage vector into the present. The V1
Type Quantity 1 Quantity 2 voltage vector change very slowly in the normal power
V0I0 Calculated V0 from phase VTs Calculated I0 from phase CTs system, so a past V1 voltage vector is a good indication of the
V0IG Calculated V0 from phase VTs Current on 4th CT input voltage vector that would exist if a fault had not occurred, and
VXI0 Voltage on 4th VT input Calculated I0 from phase CTs
it is a reliable backup for directional analysis.
VXIG Voltage on 4th VT input Current on 4th CT input B. Close in to Fault Logic
I0IG Calculated I0 from phase CTs Current on 4th CT input When a breaker is closed into a three phase fault (i.e.,
grounding chains), the memory polarization scheme will not
work because there is no pre-event V1 vector for the relay to
V. OVERCURRENT AND DIRECTIONAL ELEMENT NAMES AND work with. The Close In To Fault logic monitors for a breaker
CONTROL close and enables a high set three phase non-directional
One needs to understand which directional decision controls overcurrent sensing circuit for a short period of time. The
which overcurrent element. There is no standard way to name setting of the 50 element must be above maximum load inrush
all of the overcurrent elements that are involved. Assume for in either direction.
the discussion that there are 67/51P (phase), 67/51G (ground),

5
C. Superimposed Components phase element. Assume for simplicity of the figure that the
When heavy load flow occurs at the same time as a low level fault is at a point such that half the system impedance is
fault, it can confuse a directional element. This situation will between the facility and the fault location.
be seen in the application discussed in section VIII. Some
manufacturers have implemented a scheme that tries to
separate out load flow currents from fault currents, using
schemes referred to as superimposed components. It is similar
in application to memory polarization but involves both
current and voltage from the past into the present, rather than
just the voltage. Assume steady state load flow conditions.
Assume a sudden change, due to a fault, is seen. The relay can
take the voltage and current from the past several cycles,
before the fault, and project it into the present. This projected
voltage and current is compared to actual faulted voltage and
current. This scheme allows the fault current to be separated
from the overriding load current and, hence, improve the
decision about where the fault is located. The algorithms need
to include intelligence and logic to differentiate normal
switching events from fault events.
D. Minimum Sensitivity
A relay has limits to its sensitivity. There must be sufficient
quantities of current and voltage for a directional decision.
The minimum quantity varies by manufacturer. The response
of the relay to low voltage or current varies, but typically, the
Fig. 8. System One-Line
relay will default to a neither forward nor reverse status if
either currents or voltages are very low. In this case, each
manufacturer and even each relay from a manufacturer may The symmetrical component network that describes the
have a different logic on how the relay responds. There may application, for an A phase to ground fault, is shown in Fig. 9.
be settings to define minimum quantities the relay needs to see Note that the low negative sequence impedance of the motors
and settings for how the relay responds when values are below means that the facility can be a negative sequence current
the minimum. source for the fault. Compare the polarity of forward current at
the CT in Fig. 8 to the polarity of current flow at the dot
E. Positive vs. Negative Torque Angles. between Z2T and Z2SYS in Fig. 9. The current flow at the CT in
Each manufacturer has its own way of presenting certain data. the two cases is opposite. Hence, the 67NEG sees this reverse
For instance, it is counter-intuitive to some users to have fault and turns on. The heavy current flow into the facility,
negative angles as forward and positive angles as reverse, as meanwhile, makes the 67/51R time out to a reverse trip, even
seen in (13) and (14), so some manufacturers effectively invert though power is flowing into the facility on all 3 phases and
the Z2 and Z0 forward impedance angle. Effectively, a -1 the generation is off-line.
factor is entered into the relays impedance angle calculations.
In one perspective this is a very good feature. Even in the
presence of large forward load flow, the fault still is seen and
VII. APPLICATION NOTE 1: FORWARD POWER SUPERIMPOSED cleared. This might be especially good if the generator was
ON REVERSE FAULT very small and incapable of feeding significant current into the
fault. However, on the other hand, with the generation off line,
To better understand the performance of directional relaying,
there was not any benefit in tripping the breaker at the PCC.
let us apply it to a real world situation.
One method to stop the trip is to block the 67/51R when the
generation is off line, but in some applications the generation
Assume the 67 element in Fig. 8 is an overcurrent relay that
might be quite remote from the PCC, making the control
derives its directional characteristics from negative sequence
concepts. The overcurrent element in the reverse direction is scheme difficult. Another possible resolution that does not
need to directly know the generator status will be seen in
set below load current in the forward direction, with the intent
Section VIII, where the directional power element, 32,
of being sensitive to remote utility faults. However, assume
the facility generator is off line, so the natural tendency of the supervises the 67 element, blocking the 67 when power is into
the facility on all three phases. However, there could be
operators of the facility is to believe that the 67/51Reverse
elements will not trip for a utility fault. However, assume that argument against this process: If the generation is in peak
shaving mode and power continues to flow into the facility on
the motors are a substantial percentage of the facility loads.
all three phases during the fault, the utility must trip its
Now, assume a remote system line to ground fault that is slow
to be cleared. Assume the fault is on a lateral and with some breaker before the relay at the PCC will ever see the fault. A
few utilities object to sequential tripping, where one relay will
fault impedance, such that the feeder still carries power to the
not see a fault until another breaker trips first.
facility on the faulted phase, above the pickup of the 67/51R
6
 generation may be off during times of high facility loading,
the overcurrent relaying at the Point of Common Coupling
(PCC) must be able to carry the peak loading of the facility, so
its overcurrent setting is, for the purpose of discussion, 1.25
per unit, where 1 per unit is the current at peak loading of the
facility. Assume a generator is obtained that is 1.0 per unit in
capability, though it is common for a DG to be sized much
smaller than the local load. The unit size relative to local load
means that, during facility light loading times, the generator
could easily carry the entire facility and the plant could even
export power. However, suppose the generation is designed as
a peak shaving no reverse power flow allowed facility;
therefore, there has to be a control package for the generator
that monitors the voltage, power, and VARs at the facility
PCC and takes appropriate action to turn down prime mover
power whenever the facility approaches the point of power
export. The control system is not a protective relay, so typical
engineers and utility personnel will not allow it to be relied
upon to prevent an island condition. Compared to facilities
where generation is always less than local load, this DG has an
increased risk of supporting an island or supporting the fault
condition for an extended period.

To mitigate the risk of the DG supporting an island and being

slow to trip for a utility fault, the utility might ask that the DG
install overcurrent relaying that can sense all faults on the line
to which the DG is connected. The basic solution is to provide
two back-to-back directional overcurrent relays, one looking
into the DG facility (forward) and a second looking into the
utility lines (reverse). However, several things may cause a
determination that the reverse looking overcurrent relay will
need to be set fairly sensitively. 1) The utility distribution line
Fig. 9. Symmetrical Component Network continues many miles farther after the DG tap, so the DG sees
relatively high impedance to the fault location. 2) The utility
does not allow for sequential tripping (i.e., the DG must see
VIII. APPLICATION NOTE 2: APPLICATIONS THAT NEED the utility line fault even when the utility is still connected and
DIFFERENT FORWARD AND REVERSE FAULT RESPONSE TIME supporting the fault). 3) The utility discounts the use of
OR SENSITIVITY AT THE PCC generator sub-transient reactance (XD) and requests the use of
generator steady state reactance (XD) with no allowance for
A load-only facility, of course, has no need for directional excitation boosting, so the generator fault contribution is weak
current sensing. However, when generation is added to the on a steady state basis. These considerations, or others, result
facility, that facility, now a DG, can backfeed the utility lines. in a need to set the phase overcurrent element that looks
The fault current that is seen at the Point of Common toward the utility to below the pickup of the facilitys full load
Coupling (PCC) can vary hugely for faults in the DG facility current. A possible scenario is shown in Fig. 10 where the DG
that are fed by the utility and for faults in the utility that are is asked to trip at 0.25 pu for reverse phase-phase faults, which
fed by the DG. The differing current levels may require two is much less than load current into the facility.
67 relays, or two 67 functions in one relay, each designed to
respond to the appropriate fault in the two directions, each One concern that can be seen in this arrangement is whether
with differing overcurrent pickup and time dial settings. forward load current might ever be seen as a reverse condition
Further, two relays or relay functions can be used to determine by the 67POS element and lead to a subsequent misoperation.
appropriate response to the fault. For faults in the facility, the The generation in the facility might cause load current to be
response should be to trip the PCC breaker, but for faults in virtually any power factor. For example, suppose the
the utility, the response might be either to trip the generation generation output is set a little below present facility load
rather than the PCC breaker, or to trip the PCC and implement levels, so that net power into the facility is low, and since the
some immediate remedial action scheme to help generation facility is absorbing at least some power, there is a verification
survive any major difference between present generator output to the utility that the generator is not supporting an island.
and facility loading. Also suppose that the generator is supplying more VARs than
the local facility requires, so that the facility is shipping VARs
In many DG applications, the generator is intended for use in to the utility. In this mode the facility is seen by the utility as a
peak shaving purposes, not as a base load unit. Since leading power factor load. While this would not be a normal
7
excitation level for the generator, such a condition cannot be negligible power coming from the DG into a utility fault.
excluded from occurring. It also might occur during loss of an Faults tend to be limited by line inductance and, hence, may
inductive load in the facility. Figs. 3 and 11 show the involve large VAR flow with negligible Watt flow. Further,
condition. A positive sequence directional element, 67POS, is at the DG may be a weak source, further limiting the ability of
risk of seeing current in the reverse direction for this the generator to feed substantial power into a utility fault.
condition.

Fig. 12. 32 Supervision of Sensitive 67/51P-Reverse

IX. APPLICATION NOTE 3: USE OF THE 67/51 IN A LOSS OF

SYNCHRONISM DETECTION SCHEME
In the recent few years there has been some attention and
promotion given to requiring distributed generation to include
the loss of synchronism function (alternatively referred to as
an Out of Step relay or as Device number 78) in their
protective relaying packages. To many engineers the ideal 78
function should be done with the classical impedance
functions that monitor for relatively slow moving 3 phase
faults on an R-X plane. The cost of the required relay, the cost
of the engineering required to properly set the function, and
Fig. 10. Example DG One-Line Diagram various arguments that the function was not necessary,
eventually led to the function being left as a suggested or if
needed function in most regulatory and utility standards.
One approach to addressing the issue described above is to set However, if a generator does pull out of step, it is not a good
the MTA used by the 67POS element to about 100. This will situation, so it is good engineering practice to address the
skew the forward direction for faults to include this load flow concern if possible. Hence, given the restraint of the ideal 78
pattern as a forward condition, and since the angle of function is not available in the relaying packages being used at
maximum torque covers the zone of +/-900, the relay still will ones facility, let us see if the commonly available 27, 50,
be able to correctly distinguish forward and reverse fault 67/50, and 51VR/VC element can be configured to respond to
conditions. a loss of synchronism condition.

For a 27, 50, and 67/50 element to sense loss of synchronism,

we have to assume the event actually occurs; we cannot
predict it is about to occur, as might be possible with
impedance elements. To sense the condition, we want to
configure these elements to look for the low voltages, high
currents, and high currents in an abnormal direction that might
occur when the generator goes 1800 out of phase with the
system. There are notable issues, as we will see below, with
Fig. 11. DG S, V, Z1 at PCC for Leading PF Condition
using a 27 or non directional 50 element. A 67/50 element
An alternative approach to the issue described above is to may be the most capable method, with augmentation with
supervise the 67/51 Phase element with a 32 element, as voltage control/restrained functionality.
indicated in Fig. 12. Note the 32 element is set to 3 of 3
logic. In this logic, the block occurs only if all three phases are Voltage during Loss of Synchronism
supplying power into the DG facility. If there is a reverse fault
that interrupts power flow on only one phase, then we still Let us assume the simple system shown in Fig. 13. We will
want the 67/51P to operate. If the 32 element utilized total assume that the generator slips a pole and analyze the circuit
power, 1 of 3, or 2 of 3, logic, then the 32 might not drop to determine the voltages and currents at the PCC.
out if there is still some level of forward power into the
facility on the unfaulted phases. Also note that forward power
is monitored. There might be some temptation to use the logic
If there is reverse power flow, then enable the 51P-Reverse,
but this might not be a good idea because there may be Fig. 13. Simple System for Loss of Sync Analysis

8
We are ignoring load in this figure. To analyze the circuit reasonable considering a 5% impedance facility transformer
under the presence of load, see reference [2], which analyzes and a long line to the facility. In any case, the generator
Fig. 14 in detail and also includes facility loads. impedance is notably higher than the system impedance and k
has a value of about 0.14 in this case. If a generator is used
Let us use VSys as the fixed voltage and rotate the generator that could support the full facility load, k would have been
voltage phasor around by 3600. The voltage at the PCC will about 0.25. Per Fig. 7, k=0.25 indicates only a 50% voltage
be: dip at the PCC, and at k=0.14, the voltage will drop to about
78% of nominal.

VSys VGen Let us call the undervoltage element that we want to sense the
VPCC = VSys ZSysISys = VSys ZSys out of step condition a 27T-3/3 (T indicates the 27 is a definite
Z +Z time delay element, and 3/3 indicates all 3 phases must be low
Sys Gen
for a trip). Some issues that need consideration when setting
(
= VSys k VSys VGen ) (15) the 27T-3/3 are:
where 1. The minimum voltage that will be seen at the PCC
during a pole slip needs to be determined. See (15) and
ZSys Ref. [2] as a starting point. The PU setting for 27T-3/3
k= element needs to be less than the worst case voltage
Z +Z
Sys Gen drop that will occur during load inrush at the facility.
Load inrush should not cause more than 20% voltage
A k value of 0.5 will indicate the PCC is at the electrical drop in most facilities, so a setting of 75% of nominal
impedance center. If 0.5<k<1 then the generator impedance is may be appropriate. Hence, the voltage during a loss of
smaller than the system impedance, and if 0<k<0.5 the sync condition might be insufficiently low to reliably
generator impedance is larger than the system impedance. differentiate from low voltage due to normal events.
This issue may prevent the ability of the 27 to detect
A graphical picture that helps give a feel for the significance loss of synchronism.
of this equation is seen in Fig. 14. Note in Fig. 14 that if the 2. The low voltage at the PCC for a pole slip will be seen
PCC is near the electrical impedance center, then a very low on all 3 phases, so to help differentiate the pole slip
voltage will be seen at the PCC for each slip cycle, but if the from a temporary fault, the element should monitor for
PCC is remote from the electrical impedance center, the all phases going low.
voltage drop seen at the PCC might be difficult to sense. 3. The 27T-3/3 should be time delayed only a matter of
cycles. If the PCC is at the electrical impedance center
where k=0.5, VPCC will transiently approach OV and
will be below 0.5pu for only about 1/3 of the slip cycle.
Assuming the slip is 1 pole slip per second, this gives a
low voltage for around 20 cycles. If the PCC is not at
the electrical center, the minimum low voltage will be
higher than 0V and the low voltage condition will last a
shorter time. If the slip was faster, there would be even
less time for the relay to respond. Time delay may need
to be as low as 5 cycles, if the voltage dip can be sensed
Fig. 14. VPCC for Varying k as Generator Slips Pole at all.
There is some difficulty in stating the apparent impedance of a 4. The 27-3/3 could be fooled by external events that
generator during a pole slip event. This paper takes the deliberately remove power at the PCC, so an input to
generally accepted view (which is in agreement with generator the relay to block operation for such conditions may be
simulation tests performed at the Basler factory) that the necessary.
transient reactance, XD, is a reasonable representation of the 5. If the breaker is opened at the moment of lowest PCC
generator impedance during a pole slip event, falling in the voltage, then the breaker will try to interrupt current
0.3pu range. During a slow pole slip, when a slip cycle with the generator and system 1800 out of phase and
exceeds twice the XD time constant (typically about 0.3 with twice the system voltage across the breaker at the
second time frame), it would be anticipated that the apparent moment after current is interrupted. It may be advisable
generator impedance will start to increase somewhat. to delay tripping until the 27-3/3 element drops out.

Let us look at our sample system in Fig. 13. The generator was Current during Loss of Synchronism
rated at half the system load, so its effective impedance at the
base of 1.0 PU will be XD = 0.6pu. A power system that is The current that will flow in the circuit in Fig. 6 during an out
capable of delivering 1pu current with good voltage regulation of step condition can be approximated by the equation:
likely has a fault duty on the order of 10 times the load current
level, so we might assume XSys of around 0.1pu, which is

9
 VSys VGen the relay to respond. A delay of 5-10 cycles may be
IPCC = appropriate.
Z +Z 5. If the breaker is opened at the moment of highest
Sys Gen (16) current, the breaker will try to interrupt current with the
2 generator and system 1800 out of phase and with twice
IPCC, Peak
ZSys + ZGen the system voltage across the breaker at the moment
after current is interrupted. It may be advisable to delay
This current does not reflect the load flow that will be tripping until the 50T-3/3 element drops out.
superimposed on top of these currents. Again, see [2] for a
more detailed analysis, and see earlier discussion on the use of The 67POS during Loss of Synchronism
XD. From the previous discussion, let us assume ZSys = 0.1pu
The above equations for current and voltage at the PCC can be
and ZGen = XD = 0.6pu after converting to our common base.
combined to create equations for the apparent impedance
This gives a peak current of around 2.9pu. If a generator had
during the loss of synchronism.
been used that could have supported the entire facility, XGen =
0.3, then the peak current would be around 5pu. Let us call an
overcurrent element that we want to sense the out of step VSys VGen
VSys ZSys
condition a 50T-3/3 (T indicates the 50 is a definite time delay
VPCC Z +Z
element, and 3/3 indicates all 3 phases must be high for a trip). ZPCC = = Sys Gen
Some issues that need consideration when setting the 50T-3/3 IPCC VSys VGen
are: (17)
1. The peak current for an out of step condition needs to be ZSys + ZGen
determined, and the peak load and/or transformer inrush VSys ZGen + VGen ZSys
as seen at the relay location needs to be determined. The =
pickup setting for 50T-3/3 element needs to be about VSys VGen
75% or less of the peak out of step current (= 2.1pu for
our small generator case and 3.75pu for our large The apparent ZPCC will be lowest when VGen is 1800 out of
generator case). We also need the pickup to be at least phase with VGen. The impedance takes a path shown in Fig. 15,
125% or more of the worst case load/transformer inrush seen in many resources. Fig. 15 includes the forward and
at the facility (maybe 2pu in this example). For the reverse zones for the 67POS element. At the point where the
small generator case, we have a marginal condition. On 67POS sees reverse current, the generator voltage will be just
a case-by-case basis, these pickup setting requirements past 1800 out of phase, so current will still be high. If this is a
may very well conflict with one another. If the conflict facility that should not be sending large amounts of current to
arises, either i) external logic must be used to block the the utility, then we have a signature for out of step that we can
relay element for load inrush conditions, or ii) the relay monitor with what we will call a 67/50T-3/3. Some issues that
should be put at the generator terminals and should be need consideration when setting the 67/50T-3/3 are:
set based upon peak load inrush as seen at the generator, 1. The peak current for an out of step condition needs to be
rather than at the PCC. determined, and the peak current that the facility will
2. Pole slip currents can be seen better at the generator ever send out to the utility needs to be determined.
terminals rather than at the PCC. In facilities with Similar to the 50T-3/3, the 67/50T-3/3 pickup setting
smaller generators, the load current flow at the PCC for 50T-3/3 element needs to be about 75% or less of
may mask the pole slip condition. Further, generators the peak out of step current but, in this case, it only
are not typically exposed to load currents above 1.5pu, needs to be about 125% or more of the worst case
which makes for easier discrimination between load outrush from the facility to the utility. Because the
current and loss of synchronism current. 67/50T-3/3 element is not turned on by the 67POS until
3. The high current for a pole slip will be seen on all 3 the pole slip is just past its peak, a more sensitive pickup
phases, so to help differentiate a pole slip from a fault, of 50% of peak current might be good, if possible.
the element should monitor for all phases going high. A 2. Pole slip currents can be seen better at the generator
positive sequence overcurrent current element, 50-I1, terminals rather than at the PCC. In facilities with
would not be appropriate since phase to phase and phase smaller generators, the load current flow at the PCC
to ground faults can cause high I1. may mask the pole slip condition.
4. The 50T-3/3 should operate in a matter of cycles. Just as 3. The high current for a pole slip will be seen on all 3
in the voltage dip discussions, the current level will phases, so to help differentiate a pole slip from a fault,
oscillate from 0 to IPCC,Peak and back to 0 in one slip the element should monitor for all phases going high. A
cycle and will be above about 2/3 of IPCC,Peak for only positive sequence overcurrent current element, 50-I1,
about 1/3 of the slip cycle. For a 1 second slip rate, a would not be appropriate since phase to phase and phase
pickup of 2pu current, and a peak current of 2.9pu, the to ground faults can cause high I1.
relay will be picked up for less than 20 cycles. If the slip 4. The 67/50T-3/3 should operate in a matter of cycles.
was faster, there would be proportionately less time for Because the element is not turned on by the 67POS until
the pole slip is just past its peak, high speed operation is

10
 more important. At the turn-on point of the 67/50T-3/3,
and given a 1 second slip cycle, the current may decay
to below pickup in less than 10 cycles, and there would
be even less time if the slip rate were higher. A 3-5
cycle delay would be appropriate. Of course, the faster
one makes the relay, the more chance of a transient load
swing or CT error causing a misoperation of the
element.
5. If the breaker is opened at the moment of highest
current, the breaker will try to interrupt current with the
generator and system 1800 out of phase and with twice
the system voltage across the breaker at the moment
after current is interrupted. It may be advisable to delay
tripping until the 67/50T-3/3 element drops out.
6. The function can be augmented with voltage
supervision, enabling the 67/50-3/3 only under 3 phase Fig. 16. Ground Fault Sensing Schematic
low voltage. This might be viewed as a form of a 51VR
or 51VC relay. An element that uses a 51 may be too
slow to sense a pole slip, so a very high speed 51 or a XI. APPLICATION NOTE 5: DIRECTION TO FAULT ON HIGH
50T is appropriate. IMPEDANCE GROUNDED SYSTEMS
Some DG systems are run in an ungrounded or high
impedance grounded mode behind a transformer that isolates
the plant from the utility ground grid. In a system with a high
impedance ground, the existence of a ground somewhere in
the system is indicated by a high V0 voltage and sensed by
appropriate relaying, typically called a 59N relay. Once the
fault is detected, the faulted phase will be indicated by a low
VLG on the faulted phase. However, when multiple feeders or
generators are connected to the bus, the specific faulted feeder
or generator is unknown.

Though small, there will be some capacitive ground current

flow in the system through the fault. In Fig. 17, if feeder 3 has
a ground fault, phase to ground capacitance on the unfaulted
Fig. 15. ZPCC during Pole Slip Event phases of feeders 1 and 2 will result in current flow into the
X. APPLICATION NOTE 4: SENSITIVE GENERATOR GROUND ground fault. If enough capacitance is involved, the current
FAULT DETECTION may be detectable. A basic approach to determining which
feeder is faulted is to simply look at the magnitude of the
For the smaller generators found in the DG environment that current involved on each feeder, and the one with the most
lack a current differential (87G) element, Fig. 16 shows a current will be the faulted phase. This approach requires that
fairly fast and sensitive means of sensing a ground fault in a at least three feeders be on the faulted bus. With only two
generator. In this logic, the 50TN and 150TN are set non- feeders, the faulted feeder will see about the same current as
directional. Again, the T indicates that these are definite time the unfaulted feeder.
elements. The elements start to operate on any ground current,
but they are supervised by a directional element that picks up
whenever there is ground current flowing out of the generator.
The 50TN (the element on the phase CTs) needs to be time
delayed enough to ride through false residual current that
arises due to transient phase CT saturation that might be
caused by high DC offset in phase currents during normal load
energization or external faults. Both elements need to be
delayed long enough for the 67ZERO to make a correct
directional decision (e.g. ride through transient CT error). The
67ZERO needs to be any method that will sense out-of-generator
faults and needs to be sensitive enough to correctly see the
Fig. 17. Line to Ground Fault on High ZG System
same faults as the 50TN and 150TN elements.
A smarter method to sense the faulted feeder is the 67/50G
with the 67ZERO in the V0IG mode. By comparing the phase
relationship between the calculated V0 and the measured I0,

11
the relay can determine if there is a ground fault forward and concerned utility engineer that it is still possible and should be
hence on the feeder, or reverse and hence on another feeder. protected against.

This approach to sensing direction to the fault has its issues. In

a high impedance grounded system, there can be some level
standing offset in V0 due to system unbalance ZLG; e.g., if
phase A is always strung closer to ground than phase B and C,
then XCap,LG is lower on phase A. This will add a confusion
factor to the unfaulted condition: Is this measured V0 due to a
fault or is this due to the normal offset in my system? If such
a problem shows up, one should attempt to find the source of
the standing unbalance and remove it, or one might mask it
with a small ground bank. A VT connected in a broken delta
arrangement with a resistor in the broken delta can act as a
very small ground bank. One should calculate the ground bank
impedance required before trying to use a broken delta VT so
one does not find the effect too small. Fig. 18. Phase Loss Condition with Generator on Line

Another issue is that if the phase to ground capacitance of the The negative sequence directional element (67/51Q) at the
system is small (XC large), there may not be enough current PCC in this case may see high current unbalance. The load
for the relay to work with. There are two issues: a) there needs current will appear as a reverse phase to ground fault. The
to be sufficient current for the overcurrent element to operate, current and voltage unbalance may be enough for the 67NEG or
and b) there must be sufficient current for the 67ZERO to 67ZERO to enable sensitive reverse looking overcurrent
determine direction to the fault. These issues are in part elements. Another element that will see the situation is a
addressed by CT selection, Also, for the current level to be reverse looking 32 element. A 32 element set to monitor on a
detectable, it will likely be necessary that the relay monitor a three phase basis will not see the situation, but a 32 element
window CT that wraps all phases, that the CT ratio be low, that is set to monitor power flow one phase at a time will sense
and the ground input (Ig) of the relay be configured as a 1A the problem. The 32 element is usually settable to be highly
input, rather than the more typical 5A input (U.S. market). sensitive, so it may be more able to sense the condition than a
Further, in some relays a sensitive earth fault (SEF) feature is 67/51Q.
available that makes the IG input highly sensitive to current in
the 10s of milliamps. One should study the relay
manufacturers instruction manuals for the minimum currents XIII. APPLICATION NOTE 7: UNBALANCED LOAD CONDITIONS
and voltages that are required for the 51/67G and 67ZERO CAUSE PICKUP OF DIRECTIONAL ELEMENT
elements to operate. In an industrial facility, there is a possibility of unbalanced
loading that could result in sufficient negative or zero
One more issue that some may be concerned about is ground sequence current flow and voltage for the 67NEG or 67ZERO
fault detection when the fault impedance itself is very high; directional element bit to set. If forward is into the facility as
possibly on the order of thousands of ohms of resistance. The in Fig. 6, the 67NEG or 67ZERO will be set as forward and this
fault resistance will make the V0 neutral shift small and make will turn on the forward version of the appropriate 67/51
the ground fault hard to sense, especially if there is any elements. It would be anticipated that the 67/51P-Forward and
standing V0 in the system due to unbalanced ZLG in the 67/51G-Forward element pickup will be set above highest
system. The fault resistance also will turn capacitive current expected forward current conditions, so only turning on the
flow into a capacitive/resistive flow, so that instead of I0 forward direction bit set should not be a problem. Further, in
leading V0 by 900, it will lead by, for instance, 300, and also such current conditions the 67POS element is already set to
reduce a small current into an even smaller current. If one is forward due to high facility load current, and hence the 67/51P
concerned about fault impedance, then the zero sequence forward looking element is already enabled, so setting the
MTA can be adjusted from 900 leading to some value closer to negative sequence directional has not affected the matter.
300 leading, but outside of this option, one will need to look Therefore, the effect of the unbalanced load on the directional
for specialized ground fault sensing relays or equipment. elements can be ignored in this case. One should think through
this issue for ones facility.
XII. APPLICATION NOTE 6: DETECTING HIDDEN PHASE LOSS
Examine Fig. 18. In this case one phase feeding the DG has XIV. APPLICATION NOTE 8: USING SETTING GROUPS TO
been lost, but due to generation on site, the matter is not easily CONFIGURE THE 67/51 FOR CHANGING SYSTEM CONDITIONS
detected simply from a voltage standpoint. It is likely that a In DGs with many possible modes of operation (e.g. Fig. 19),
fault occurred that caused this situation to arise, but somehow there are various conditions where it may be worthwhile to
the fault has cleared and the DG is left back feeding an consider different setting groups to change the performance of
unfaulted phase. One might make the argument that this a relay at the PCC.
situation would not likely occur, but that will not dissuade the
12
 A facility with multiple generators may operate the PCC
with more or less generators on line. With increased
generation on line, the current that the facility may feed
into a utility fault will increase. Higher sensitivity may
be required when only one generator is on line, and less
sensitivity when three generators are on line and
connected in parallel.
With increased generation and the PCC being run closer
to the float (0 power flow) point, the risk of transient
reverse power flow increases. If there is risk of a
sensitive reverse current relay from operation, some
allowance may be needed for desensitizing the current
relay when a large amount of generation is on line.
If there are one or two breakers or transformers that
connect the DG to the utility and the system might be run
with either one or two transformers, or generation tied to
either source, the 67/51 and relay logic may need to be
modified appropriately for system conditions. For
example, which breakers should be tripped for a reverse
fault? Which generators?

Fig. 19. DG with Many Possible Modes of Operation

IX. CONCLUSIONS
The paper discussed some of the variations in directional
control that can be found in the relays on the market. When
testing the relay, determining relay settings, or analyzing event
reports, some concept of how the relay determines direction to
a fault will be needed.
REFERENCES
[1] John Horak, A Derivation of Symmetrical Component Theory and
Symmetrical Component Networks, Georgia Tech Protective Relaying
Conference, April 2005. Available at www.basler.com.
[2] PoleSlip_R0.xls spreadsheet to model voltages, currents, and power
flow during a pole slip event. File is available at www.basler.com.

AUTHOR BIOGRAPHY
John Horak (M,1987) received his BSEE in 1987 from the University of
Houston and an MSEE in the field of power system analysis in 1995 from the
University of Colorado. He has worked for Houston Lighting and Power,
Chevron, and Stone and Webster Engineering, where he spent several years
on assignment in the System Protection Engineering offices of Public Service
Company of Colorado. In 1997 he began his present position as an
Application Engineer for Basler Electric.

13
 Route 143, Box 269, Highland, Illinois U.S.A. 62249
If you have any questions or need Tel +1 618.654.2341 Fax +1 618.654.2351
No. 1300 North Zhongshan Road
Wujiang Economic Development Zone
additional information, please contact e-mail: info@basler.com Suzhou, Jiangsu Province - P.R. China 215200
Tel +86(0)512 6346 1730 Fax +86(0)512 6346 1760
Basler Electric Company. e-mail: beichina@basler.com
Our web site is located at: P.A.E. Les Pins, 67319 Wasselonne Cedex FRANCE
Tel +33 3.88.87.1010 Fax +33 3.88.87.0808
http://www.basler.com e-mail: beifrance@basler.com
e-mail: info@basler.com