DIRECTIONAL COMPARISON PROTECTION

FOR LINES, BUSES, AND TRANSFORMERS

Roberto Cimadevilla Gonzlez
ZIV P+C

Ainhoa Fernndez Valdivieso

ZIV P+C

Oscar E. Bolado
ZIV USA

Parque Tecnolgico, Edificio 210

48170 Zamudio, Vizcaya, Espaa
r.cimadevilla@ziv.es

Parque Tecnolgico, Edificio 210

48170 Zamudio, Vizcaya, Espaa
a.fernandez.v@ziv.es

2340 River Rd, Suite 210

Chicago, IL 60018
o.bolado@zivusa.com

This paper describes a directional comparison protection unit that can be applied to power lines, buses or
transformers either as main protection or backup for the differential protection. Without settings, this unit
provides excellent dependability and security. The protection algorithms have been tested with a RTDS
system simulating very adverse internal and external faults.
Keywords: Differential Protection, Directional Comparison Protection, Outfeed, Dependability, Security,
RTDS (Real Time Digital Simulator).
1. INTRODUCTION
In general, differential protection is the main unit of choice for power lines, buses, or transformers. Such
protection usually includes a percentage restraint characteristic, with settings (sensitivity, restrain slopes,
slope start point, etc.) that are not always easy to calculate. The complexity of these calculations is due to the
multiple factors involved such as CT measuring errors (with or without saturation), CT saturation level, or
factors pertaining to the protected element. Power transformers bring in the magnetizing current, or the
differences between the central and the end taps. Power lines introduce the capacitive current and errors in
the communications channel delay compensation. In substation buses, settings calculation requires
consideration of the differential current generated by the opening of one of the bay CTs during maximum
load conditions. These settings always compromise the balance between dependability and security. A steep
restrain slope adds security for external faults but it may compromise the dependability for internal faults. It
is difficult to maintain the stability of the differential protection during external faults with severe saturation
in one of the CTs, while guaranteeing dependability for internal faults also with severe saturation or with
heavy load conditions and low fault current level.
Therefore, the differential unit needs to be complemented with other units easier to set that help to maintain
the dependability and security in extreme cases as the ones mentioned above. This paper describes a
directional comparison unit, without settings, that increases the dependability and security of the differential
unit.
2. DIFFERENTIAL UNIT
This section reviews the operation principles of a differential unit with percentage restrain. To simplify the
explanations, a two terminal element is selected (line, bus, or transformer), as represented in figure 1. The
differential unit operates with the currents I-1 and I-2. There can be discrepancies between those currents
introduced by different CT ratios, or in case of the transformers, its own transformation ratio, the connection
group, and the zero sequence filters. These differences are considered to be compensated.

978-1-4244-4183-9/09/$25.00 2009 IEEE

513

The differential unit operates based on two parameters:

-

Differential current: IDIF I-1+I-2

Restraint Current: There are various restraint current formulas; this paper considers the following:
IRES1 = I-1  I-2 ; IRES2 =( I-1  I-2 ) / 2; IRES3 =( I-1  I-2  IDIF ) / 2

I-2p

I-1p
PROTECTED
ELEMENT

I-1

I-2
DIFFERENTIAL
RELAY

Figure 1. Circuit Diagram of the protected element by the differential unit

The differential unit will operate when the point ( IDIF , IRES ) is above the corresponding characteristic. In
modern relays, such characteristic is similar to the one represented in figure 2, with either one or two
restraint slopes.

IDIF
ZONE 1

ZONE 2

ZONE 3
=SLOPE 2

=SLOPE 1

I0

I2

I1

IRES

Figure 2. Differential Characteristic

The following section analyzes the geometrical location of the point ( IDIF , IFRE ) for internal and external
faults considering several factors.
2.1.

Analysis of the Geometrical Location for (IDIF, IRES)

2.1.1. External Faults

2.1.1.1. External Errors Influence
In theory, the differential current for external faults is zero, therefore the geometrical point ( IDIF , IRES ) will
be in the horizontal axis. Nevertheless, there are external factors that can be a source of differential current,
such as CT errors; in case of power transformers, the magnetization current or the operation of the tap
changer; or in case of lines, the capacitive current and the errors from the communication channel delay
compensation. Hence, the point will be located in a line with a slope dependent of the errors introduced by
such factors. The first slope of the differential characteristic is set to compensate for such errors.

514

2.1.1.2. CT Saturation Influence

CT saturation introduces errors in the phasor estimation, decreasing the phasor magnitude and leading its
phase [1]. During external faults, such errors increase the differential current while decreasing the restraint
current, pushing the point ( IDIF , IRES ) towards the operation zone. Note that the effect on the restraint current
depends on the formula selected. Reference [2] analyzes how these errors in phasor estimation due to CT
saturation influence the dependability and security of percentage differential characteristics during internal
and external faults. This analysis requires defining the equations of the differential characteristic as a
function of the mentioned errors. For a fault with saturation of the CT-2 (the CT measuring the current I-2),
I-1 and I-2 can be expressed as I-1=I 0 y I-2=k*I 180+, where I is the fault current magnitude, and k and
represent the errors in the phasor I-2 magnitude and angle introduced by the CT-2 saturation. Considering
that IDIF =I-1+I-2 and IRES =  I-1  I-2  / 2 it is possible to obtain the following equations defining the zone 2
and 3 on the characteristic represented in figure 2:
2

2
IRESI2: cos G ! 1  k  1 (1  k ) D  I1 (1)
2k
2k 2
I
2

2
IRES>I2: cos G ! 1  k  1 I 2 (D  E )  (1  k ) E  I1 (2)
2k
2k I
2
I

The k and values generated by the CT saturation are a function of the fault current I. Such values can be
obtained if the CT parameters are known. A PSCAD simulation was used in reference [2] to obtain such
values. Once the k and curves as a function of I are known, it is possible to check the security of the
differential characteristic for a given group of settings. This method allows identifying the settings that
provide the highest dependability while maintaining the security of the differential unit.
This is a very effective method to set the differential unit but it can be time consuming.
2.1.2. Internal Faults
When currents I-1 and I-2 are in phase, IDIF IRES1 , IDIF IRES2 / 2 , IRES3 0 , the point ( IDIF , IRES ) can be
defined by lines with slopes of 100%, 200% and infinity, respectively. Nevertheless, the following factors
are source of phase differences between I-1 and I-2 that modify the theoretical geometric locations.
2.1.2.1. Load Influence
During internal faults with load flow through the protected element, currents I-1 and I-2 become out of
phase. Figure 3 depicts the circuit diagrams for an internal fault in the protected element represented by
impedance Z. Using the superposition principle; it is possible to split the circuit in a prefault circuit and a
pure fault circuit. Based on these two circuits we can obtain the relation between I-1 and I-2, calculated as:
I-1=I-1fp+I-1pf y I-2=I-2fp+I-2pf, where the subindices fp and pf indicate pure fault and prefault,
respectively. Defining Z1 ZS1+n Z , Z2 ZS2+(1-n) Z , ZT ZS1+Z+ZS2 , it is possible to easily
calculate the following magnitudes, represented in the phasor diagram in figure 4:

I-1pf

ES1-ES2
; I-2pf
ZT

-I-1pf ; Vpf

ES1 - I-1pf Z1 ; I-1fp

515

Vpf
Z1
Z1+RF (1+ )
Z2

; I-2fp

Vpf
Z2+RF (1+

Z2
)
Z1

Figure 3. Circuits for fault, prefault and pure fault

The system under consideration is homogeneous, where Z1 and Z2 have the same angle, causing no phase
difference between the pure fault currents. Also, the relation of the pure fault currents considered is
I-2pf= -I-1pf. In case of power lines the current derived by the capacitances in deemed negligible.

ES1-ES2 ZT

ES1

Vpf

I-1pf

I-1

I-2pf
I-1fp

ES2

I-1fp

I-1pf
D

I-2fp

I-2fp
I-2

I-2pf
Figure 4. Phasor diagrams associated to the circuits in figure 3
For the same pure fault current, the larger the load current, the larger the phase difference between the
currents I-1 and I-2. Under extreme conditions, such as faults with outfeed, I-1 and I-2 will present more than
90 difference, with a tendency for being in counterphase. Note that the differential current is solely
dependent on the pure fault currents, since the differential current during prefault is basically zero. On the
other hand the restraint current depends on the total currents (pure fault plus prefault). Therefore the load
current increases the restraint current without increasing the differential current. In fact, there is a slight
decrease in the differential current since the prefault voltage is decreased then reducing the pure fault
currents. This behavior tend to shift the point ( IDIF , IRES ) towards the no operation zone of the differential
characteristic.

516

2.1.2.2. CT Saturation Influence

During internal faults, the error in phasor estimation introduced by CT saturation decreases both the
differential and restraint currents. The restraint current will vary more or less depending on the selected
formula. Depending on the ratio in the reduction of the ( IDIF , IRES ) coordinates, the point could shift towards
the no operation zone.
Figure 5 represents the geometrical locations for ( IDIF , IRES ) for a restraint current IRES =  I-1  I-2  / 2 .

IDIF

IRES

200%
INTERNAL FAULTS

EXTERNAL FAULTS WITHOUT SATURATION

EXTERNAL FAULTS WITH SATURATION

Figure 5. Geometrical locations for (IDIF, IRES)

Reference [2] analyzes the influence of the CT saturation over the dependability of the differential
characteristic. For an internal fault with saturation on the CT-2, I-1 and I-2 are expressed as
I-1=I 0 and I-2=M*k*I +, where I is the magnitude of the fault current, and k and are the errors in the
phasor magnitude and angle, introduced by the CT saturation. M represents the ratio between the I-1 and I-2
magnitudes and the phase difference between them. These differences are caused by the difference between
the sources 1 and 2 and the differences between the fault loops. In a similar fashion as for external faults, the
curves are obtained for I, k, M, and , guaranteeing the dependability of the differential characteristic for a
given group of settings.
In case of a transformer differential, the CT saturation produces the operation of the harmonic restraint or
blocking units, caused by the harmonic content of saturated currents, implying a possible lack of
dependability.
The lack of dependability caused by CT saturation is usually of transient character, since the saturation
diminishes with time. The result is an increase in the trip times that could be unacceptable in certain
instances.
Usually, the differential unit settings are calculated to ensure its stability. Such settings could limit the
differential sensitivity under certain conditions, such as faults with severe CT saturation or with outfeed. The
directional comparison unit described below increases the dependability and security of the differential
characteristic without settings.

517

3. DIRECTIONAL COMPARISON UNIT

The directional comparison unit described contrasts the phases of I-1 and I-2. Ideally, during internal faults
to the protected element I-1 and I-2 are in phase, while during external faults the currents are in
counterphase. This behavior rationalizes a directional comparison or phase comparison that easily determines
the fault location. Yet, based on the cases described in previous sections, during internal faults, the load and
the CT saturation could cause an important phase shift between I-1 and I-2. Also, during external faults, the
influence of external errors and the CT saturation cause phase differences other than 180.
A 90 threshold has been considered to differentiate between internal and external faults. The fault will be
considered internal when arg(I-2)-arg(I-1) <90 , and it will be considered external for all other cases. Figures
6.a and 6.b represent an internal and an external fault, respectively.
The directional comparison unit operates with I-1 and I-2. To evaluate these currents the design includes two
directional comparison units, one operating on pure fault current and other operating on fault current.
Independent on the current type selected, these must be matched, and therefore the method requires the same
compensations as for differential schemes.

EXTERNAL FAULT

INTERNAL FAULT

EXTERNAL FAULT

INTERNAL FAULT

I-2
I-1

I-1

I-2

a. Internal Fault

b. External Fault
Figure 6. Directional Criteria

3.1.

Directional Comparison Unit with Pure Fault Current

This unit is intended to clear faults with low pure fault current (high fault loop impedance) and with high
load flow. The unit can be based on phase or positive sequence pure fault currents, on zero sequence current,
or on negative sequence current. In the first two cases it is necessary to subtract the prefault current. In the
last two cases, subtracting prefault current is only required for high unbalance during load conditions. This
behavior must be considered in the line relays always that a single pole recloser is used. For transformer
applications, it is not recommended to select the zero sequence current, since it can be filtered by the zero
sequence filters.
The use of the pure fault current, allows detecting any type of internal fault, without CT saturation, even
under outfeed conditions. CT saturation modifies the fault current measuring, without changing the prefault
current. Under severe saturation, this effect can miscalculate the directionality. Therefore, it is recommended
to block this unit under CT saturation conditions. A saturation detector is described in section 3.2.1.

518

3.2.

Directional Comparison Unit with Fault Currents

This unit employs phase currents, without subtracting any prefault current. The unit will operate correctly for
external faults if the phase errors introduced by CT saturation plus the consequences of the factors described
in section 2.1.2.1 do not exceed 90. Note that the errors described in 2.1.2.1 are usually quite small. Also,
only extremely severe saturations will cause phase errors exceeding 90.
Regarding internal faults, load flow and CT saturation with the addition of the factors in 2.1.2.1, can cause
phase shifts between Iph-1 and Iph-2 (phase currents 1 and 2) larger than 90. Load flow influence is
considerable for low fault currents. Considering that CT saturation occurs under high fault current, when the
influence of the saturation factor is high, the influence of the load factor will be low and vice versa. A
current threshold is introduced to resolve misoperation under outfeed conditions (phase shift between Iph-1
and Iph-2 larger than 90 during internal faults). Outfeed faults occur with a weak or zero infeed in one of the
ends, with a very small measured current. The directional comparison only considers those currents above
the threshold.
In case of faults with CT saturation, the selected 90 threshold provides an ample safety margin to determine
the directionality. Even with the low influence of the load, for severe saturation conditions it is possible to
find internal faults causing phase shifts in the phase currents exceeding the threshold, losing the
dependability of the unit. To ensure proper operation the design includes two phase current directional
comparison units. The first unit employs phasors calculated by a one-cycle DFT and the second estimates the
phasors with a least squares algorithm as described in the following section. This last method is immune to
CT saturation.
3.2.1. Phasor Estimation based on Least Squares Method
To eliminate CT saturation influence, this phasor estimation algorithm only utilizes samples from the
portions of the wave that do not present saturation. This method requires a detector to discern between the
periods of saturation and no saturation on the CT. Reference [3] describes several methods that are usually
reliable to determine when the CT becomes saturated but are less reliable to detect when the saturation
disappears. Reference [4] describes a method to effectively detect the start point of the non-saturated portion
of the wave by obtaining a reference point right after the fault inception. Note that this method is not valid
for evolving faults since it is not possible to determine the reference point.
The selected method during this research to detect the CT saturation is based on the derivative of the
measured current. At the moment of CT saturation, the wave presents an inflexion point that causes a
discontinuity in its derivative.
The fault current derivative significantly reduces its exponential components due to the high time constants
[5]. Therefore the first derivative can be considered as a pure sinusoidal. Then, it is possible to estimate the
next value in the wave by taking two values of such wave and using a sinusoidal extrapolation. If the
estimated value has a noticeable difference with the measured value (taking both percentage and absolute
thresholds) the CT is deemed saturated at that precise instant. In the transition from saturation to nonsaturation, the first derivative also presents discontinuities that could be used to determine the instant to start
reading non-saturated samples. This method has two back draws. First these discontinuities are usually much
smaller than the ones when transitioning from non-saturated to saturated. Second, during the saturation
period it is possible to experience multiple discontinuities. The discontinuity of the derivative will be only
monitored from a peak value of the wave to the zero crossing. If saturation is detected during this period, the
value of the derivative is not considered until the next peak of opposite sign. This method only detects
transitions from non-saturated to saturated. To record the non-saturated samples, the method considers a
number of samples previous to the point of detection of saturation and equal to the minimum time that the
CT requires to be without saturation for the correct operation of the protection unit. For example if the
required time is 1/8 of a cycle, considering a sampling rate of 32 s/c, the number of samples to consider will
be 4. Note that the saturation detector will only operate when the fault detector is active.

519

Since a 4 sample window is short to perform phasor estimation, the algorithm will consider longer windows,
containing only sampled values from the non-saturated wave portions. The calculation windows, in these
cases, will include non-contiguous blocks of samples separated by the saturated periods. The angle accuracy
required to implement the directional comparison unit is achieved with 8-sample windows. The phasor
estimation will be done with a least squares filter similar to the one described in reference [4]. The equation
of the current defined by the samples in the calculation window is:
2S k
 M )  B eO k . Such equation can be expressed as follows:
N
I k C1 cos( wk )  C 2 sin( wk )  B  O k , where w 2S / N , C1 A cos M and C 2  A sin M . C1 and C2
represent the real and imaginary parts of the phasor corresponding to the fundamental component of the fault
current
Ik

A cos(

sin( w 0 't )
0 't
cos( w 0 't )
cos( w 1 't )
sin( w 1 't )
1 't

cos( w 2 't )
sin( w 2 't )
2 't

sin( w 3 't )
3 't
cos( w 3 't )
cos( w (n  0) 't ) sin( w (n  0) 't ) (n  0) 't

cos( w ( n  1) 't ) sin( w (n  1) 't ) (n  1) 't

cos( w ( n  2) 't ) sin( w (n  2) 't ) (n  2) 't

cos( w (n  3) 't ) sin( w (n  3) 't ) (n  3) 't

1
1
1 C1

1 C 2

1 O

1 B
1

Ii
I

i 1
Ii 2

I i 3 (3)
I

i 3 n
I i  4 n
I

i  5 n
I i  6 n

In equation (3), n-1 represents the number of samples between the two 4-sample blocks selected for the
phasor estimation, and 't represents the time between samples. If equation (3) is expressed as M P I , P
is obtained by the least squares method as P ( M T M ) 1 M T I . To perform the phasor estimation is only
required to calculate C1 and C2.
Figures 7 and 8 depict the measured current (signal I) during two faults causing CT saturation. Signal SAT
indicates the instants where CT saturation is detected. For the phasor estimation based on the least squares,
the relay stores the 4 samples previous to the saturation detection. Such samples are represented by the signal
LES_SAMPLES. As soon as the calculation window includes 8 samples, the phasor angle will be calculated.
The current in figure 7 has offset, therefore saturation only occurs during the positive semicycles. This effect
increases the required time to collect 8 unsaturated samples in the calculation window.

Figure 7. Current I, samples taken for phasor estimation by least squares method, and saturation detection
signal for fault with DC offset

520

Figure 8. Current I, samples taken for phasor estimation by least squares method, and saturation detection
signal for fault without DC offset.
Figure 9 represents, the same fault as in figure 7 including the evolution of the first derivative of the current
(signal I_DER) along the extrapolated derivative (signal I_DER_EXTR). The figure includes samples
between two peaks in the current wave (signal I). During the period without saturation, the signals I_DER
and I_DER_EXTR remain superimposed. At sample number 259, where CT saturation begins, there is a
significant difference between signals, activating the saturation detector (signal SAT).

Figure 9. Current I, current derivative, extrapolated current derivative, and saturation detection for fault with
offset.
The position of the calculation window will be determined by the saturated current if only one of the CTs
saturates. In such case the phase difference between both currents will be error free during the steady state,
since the start points for both calculation windows coincide, being the points of reference for the angles in
the phasor estimation. Also, in theory, there are no errors during the transient state, since the samples belong
to the same time interval. In cases with saturation of both CTs, each calculation window has a different time
position. This causes angle errors both for the steady estate, since start points do not coincide, and the
transient state, where one window will take samples earlier than the other. Angle errors during the steady
state are easily corrected, by measuring the time difference between the start points for each calculation
window. Transient state errors are difficult to correct. To avoid the transient state errors influence, the
method do not consider directionality decisions during such transient state, then the directional comparison
unit output is only considered when the phase difference between both currents is constant. This is achieved
by locating two consecutive values between two previously defined thresholds. Note that phasor estimation
based on least squares is only calculated with the fault detector activated, therefore it always will operate
with fault current samples, avoiding the prefault to fault transient. In case of evolving faults, the transient is
not avoided since the fault detector is already active.
In reality, the transient state will cause phasor estimation errors even when the calculation windows have the
same time position, also when using DFT, because the samples considered will belong to different waves.
Therefore, these methods should disregard the directional comparison unit results during transient state for
any phasor estimation included in this paper.

521

The directional comparison unit based on phasors estimated by the least squares method will only operate
when saturation in any CT is detected. Under this condition it will complement the unit based on phasor
estimation by one-cycle DFT. If any of both units detects an internal fault the trip will be activated. Note that
CT saturation detection under normal conditions does not have any negative effect. Even if the least squares
unit presents angle errors greater than the DFT unit, it never will trip for external faults. To trip for external
faults, errors should be very large, close to 90. On the other hand, if the least squares unit does not trip for
internal faults, the DFT unit will trip. Lack of detection of small CT saturations is not considered to be a
source for misoperation. Under such conditions the directional unit based on DFT will operate properly, due
to the small angle errors.
3.3.

Protected Element with More Than Two Terminals

During external faults, when the protected element includes more than two terminals, there will be multiple
incoming currents into the element and one outgoing. Due to the load effect, the outgoing current will not be
in perfect counterphase with the rest of the currents. Figure 3 represents an external fault to a three terminal
element. Note how I-1 is in counterphase with I-2+I-3, instead of being in counterphase with I-2 and I-3.

I-2
I-2+I-3

I-1
I-3

Figure 10. External fault for a three terminal element

For a protected element of m terminals, each current I-n will be in counterphase with the current

I-i

i=1

with

in. Then, the directional comparison unit should consider such currents. If any of the currents is in
counterphase with the sum of the rest, the unit will consider the fault to be external. Note that the unit
utilizing pure fault currents can perform a direct angle comparison for every measured current. Eliminating
the load influence, in figure 10, I-1fp would be in counterphase with I-2fp and I-3fp, since both are in phase.
The algorithm employed by the pure fault directional comparison unit would be: if the absolute value of the
phase difference between two currents is greater than 90, then the fault is external.
3.4.

RTDS Simulation

To test the directional comparison algorithms, a power transformer is simulated in a RTDS system with the
characteristics as shown in figure 11. The operation of the directional comparison unit is then compared to
the operation of a differential unit, with typical settings corresponding to the characteristic of figure 2: I0=0.3
pu , I1=1 pu, I2=8 pu, =20%, =100%.
The conclusions obtained for the power transformer can be extrapolated to a line or bus protection.

522

ES1=400 kV

3000/5
TI-1

ZS1=285

2000/5
TI-2

ZT=0.189 pu

ES2=220 kV

ZS2=275

Figure 11. Model used in the RTDS simulation

Figure 12 represents the current measured by CT-1, during an external fault AG evolving to internal. Signal
SAT1 indicates the instants where saturation is detected in CT-1, while signal LES_SAMPLES indicates the
samples taken for phasor estimation based on least squares. Figure 13 represents, for the same fault, the
current measured by CT-2, which does not saturates. Note that signal SAT2 does not activate.
Figure 14 depicts the differential characteristic with the coordinates of (IDIF, IRES) for the steady state
conditions of external and internal faults, respectively. During the external fault, the point (IDIF, IRES) is
located inside the operation zone. Under such conditions the enabling of the harmonic restraint or blocking
avoids the trip of the unit, maintaining the security. Nevertheless, during internal faults, this enabling will
cause a lack of dependability.
Figure 15 displays the evolution of the phase difference between I-1 and I-2 calculated by one-cycle DFT
(ANG_DFT) and by the least squares method (ANG_LES). This last signal refreshes every time that the
calculation window contains 8 unsaturated samples. The signal ANG_DFT reaches values less than 90
during the prefault to fault transient, therefore, a time delay is required to avoid the directional unit trip. The
signal ANG_LES does not display any transient, since it is only calculated with fault current samples. As a
consequence of saturation in CT-1, the signal ANG_DFT, reaches values between 222 and 227 during the
fault steady state, while the signal ANG_LES is between 175 and 192. Note the better accuracy of this last
signal. When the fault evolves from external to internal, the signal ANG_DFT always reaches values greater
than 90, therefore the directional comparison unit based on DFT will never trip. Still, the signal ANG_LES,
takes values around 26, causing the unit based on least squares to trip. Considering that the real phase
difference between currents is 30, it can be concluded that under severe saturation in one of the CTs, the
least squares method presents much better accuracy than the DFT method.
Figure 16 displays the evolution of the point (IDIF, IRES) along the differential characteristic for two AG faults
with outfeed. In the second fault the load is larger than in the first one. Note, how in the second fault the
point (IDIF, IRES) has shifted to the right, caused by the increase on restraint current without an increase in
differential current due to the load flow. The consequence is a lack of dependability, since the point (IDIF,
IRES) does not reach the zone of operation.
Figure 17 shows the phase difference between positive sequence pure fault currents I-1 and I-2, after the
prefault values are subtracted, for the previous faults with outfeed. Note that negative or zero sequence
currents could also be used. Since pure fault currents are employed, the phase difference is independent of
the load. Both currents are not perfectly in phase due to the non-homogenous characteristic of the simulated
system. Nevertheless, the absolute value of the angle between currents is less than 90, causing the
directional comparison unit to trip. Both current magnitudes exceed the minimum threshold to enable the
directional unit. Note that the directional comparison unit based on pure fault currents present a higher
sensitivity than the differential unit.

523

Figure 12. Current IA-1, samples taken for phasor estimation by least squares method, and saturation
detection signal

Figure 13. Current IA-2 and saturation detection signal

b. Falta interna

a. Falta externa

Figure 14. Point (IDIF, IRES) in relation to the differential characteristic

for external to internal evolving AG fault

Figure 15. Phase difference between IA-1 and IA-2 measured by the least squares method and by DFT

524

a. Fault with outfeed 1

b. Fault with outfeed 2

Figure 16. Point (IDIF, IRES) in relation to the differential characteristic for faults with outfeed

Figure 17. Phase difference between pure fault I1-1 and I1-2 for faults with outfeed 1 and 2

4. CONCLUSIONS

This paper has described a directional comparison unit providing a higher security and dependability than
traditional differential units. The scheme employs pure fault currents to eliminate the load influence and to
provide higher sensitivity for faults with outfeed conditions. It also uses phase currents to properly determine
directionality under faults with CT saturation. Both currents are estimated by a one-cycle DFT and by a least
squares algorithm, effectively reducing the influence of CT saturation. Testing protocols include cases with
very unfavorable conditions where the unit displayed a secure and dependable behavior.

525

5. REFERENCES

[1] Zocholl, S.E., and Smaha, D.W., Current transformer concepts- 19th Annual Western Protective Relay
Conference, Spokane, Washington, Oct. 1992
[2] P.K. Gangadharan, T.S. Sidhu and A. Klimek, Influence of current transformer saturation on line current
differential protection algorithms IET Generation, Transmission & Distribution, Vol. 1, no. 2, Mar. 2007
[3] B. Kasztenny, E. Rosolowski, M. Lukowicz, and J. Izykowski, Current related relaying algorithms
immune to saturation of current transformers,in Proc. Conf. Developments in Power System Protection,
Mar. 1997.
[4] Jiuping Pan; Khoi Vu; Yi Hu, An efficient compensation algorithm for current transformer saturation
effects, IEEE Trans. Power Del. Vol. 19, no. 4, Oct. 2004
[5] Y. C. Kang , S. H. Ok ,and S. H. Kang , A CT saturation detection algorithm, IEEE Trans. Power
Del. , vol. 19, no. 1, Jan. 2004.

526