Performance of novel neutral admittance criterion in MVfeeder earthfault protection
ABSTRACT This paper describes the fundamentals of the neutral admittance based earthfault protection. First, the theory is briefly introduced. Secondly, certain improvements to the traditional measuring principle and operation characteristics are suggested. Finally, the performance is evaluated and compared with traditional earthfault protection schemes using simulated and recorded data. The results show that the neutral admittance criterion has potential to become a standard and widely used earthfault protection function in high impedance earthed networks.
 Citations (1)
 Cited In (0)

Conference Paper: Admittance criteria for earth fault detection in substation automation systems in Polish distribution power networks
[Show abstract] [Hide abstract]
ABSTRACT: Recent experience in the exploitation of Polish mediumvoltage power networks shows that prevailing method used to determine distribution line earth faults, based on directional criteria, is often unreliable and poorly effective. This paper describes how current earth fault detection automation proves to be particularly ineffective during high resistance or arcing in the vicinity of the earth faultElectricity Distribution. Part 1: Contributions. CIRED. 14th International Conference and Exhibition on (IEE Conf. Publ. No. 438); 02/1997
Page 1
? ? ? ??????
??????????
?????????
????????
????
?????????????????????????????????????????????????????????????????????????????????
?
??????? !"??
??#???????$??????? ? ??????%???? !"?????
PERFORMANCE OF NOVEL NEUTRAL ADMITTANCE CRITERION IN MVFEEDER
EARTHFAULT PROTECTION
Ari WAHLROOS Janne ALTONEN
ABB Oy Distribution Automation – Finland ABB Oy Distribution Automation – Finland
ari.wahlroos@fi.abb.com janne.altonen@fi.abb.com
ABSTRACT
This paper describes the fundamentals of the neutral
admittance based earthfault protection. First, the theory is
briefly introduced. Secondly, certain improvements
to the traditional measuring principle and operation
characteristics are suggested. Finally, the performance is
evaluated and compared with traditional earthfault
protection schemes using simulated and recorded data. The
results show that the neutral admittance criterion has
potential to become a standard and widely used earthfault
protection function in high impedance earthed networks.
INTRODUCTION
Earthfault (EF) protection is traditionally based on
directional residual overcurrent criterion with zerosequence
voltage as the polarizing quantity. An example of such is the
I0cos(ϕ)criterion that is commonly used in compensated
medium voltage distribution systems. However, e.g. in
Poland, the neutral admittance (Y0) criterion has become
popular and is today a standard EF protection function
required by the local utilities [1].
THEORY
The following analysis assumes that all the measured
quantities are fundamental frequency phasors. The equations
are valid for the phase L1toearth fault, but similar
equations can be derived for phase L2 or L3toearth faults.
The theory of the Y0 criterion can be explained with the aid
of a simplified equivalent circuit of a 3phase distribution
network illustrated in Fig. 1. The network consists of two
feeders, one representing the protected feeder (Fd) and the
other the background network (Bg). The background
network represents the rest of the feeders in the substation.
The line series impedances are neglected as their values are
very small compared with the shunt admittances. Also the
loads and phasetophase capacitances are disregarded as
they do not contribute to the zerosequence current.
Notations used in Fig. 1:
U0 = (UL1+UL2+UL3)/3 = Zerosequence voltage of the network
3I0Fd = (IL1Fd+IL2Fd+IL3Fd) = Residual current of the protected feeder
3I0Bg = (IL1Bg+IL2Bg+IL3Bg) = Residual current of the background network
ICC = Current through the earthing arrangement
YCC = Admittance of the earthing arrangement
EL1 = Source voltage, phase L1 (e.g. 20/√3 kV∠0o)
ULx = Phase voltage of phase L1, L2 or L3 at the substation
ILxFd = Phase current of phase L1, L2 or L3 of the protected feeder
ILxBg = Phase current of phase L1, L2 or L3 of the background network
YFFd = Fault admittance when the fault is in the protected feeder
YFBg = Fault admittance when the fault is in the background network
YLxFd = Admittance of phase L1, L2 or L3 of the protected feeder
YLxBg = Admittance of phase L1, L2 or L3 of the background network
Fig. 1 Simplified equivalent circuit for high impedance earthed three phase
distribution network with a singlephase earth fault in phase L1.
The equivalent circuit of Fig. 1 is equally valid during
healthy and faulty states. During the healthy state the fault
resistances equal infinity i.e. the fault admittances YFFd and
YFBg are zero. In case of an earth fault inside the protected
feeder, then YFFd > 0 and YFBg = 0. Further, if an earth fault
occurs outside the protected feeder i.e. somewhere in the
background network, YFBg >0 and YFFd = 0.
In compensated networks, the admittance of the earthing
arrangement equals YCC = GCC  j⋅BCC = 1/RCC  j⋅K⋅Btot,
where K is the degree of compensation, Btot is the total
susceptance of the network and RCC is the resistance of the
parallel resistor of the compensation coil. It should be noted
that the value of YCC is affected by the connection status of
the parallel resistor according to the applied Active Current
Forcing (ACF) scheme. Typical ACF schemes are:
I. The resistor is continuously connected during the healthy state,
and then momentarily disconnected and again reconnected
during the fault. The purpose of disconnecting is to improve the
conditions for selfextinguishment of the fault arc.
II. The resistor is disconnected during the healthy state, and then
connected during the fault until the protection operates.
III. The resistor is permanently connected. The primary purpose is
to limit the healthy state U0.
In all ACF schemes the feeder EF protection is typically set
to operate based on the resistive current increased by the
parallel resistor during the fault.
The applied ACF scheme also affects the behavior of the
Y0 protection algorithms as shown in the following. In
addition it is important to note that in the schemes I and III
the admittance YCC is the same prior to the fault and when
the feeder EF protection operates. In the scheme II, the
admittance YCC is different prior to the fault and when the
feeder EF protection operates.
Page 2
? ? ? ??????
??????????
?????????
????????
????
?????????????????????????????????????????????????????????? ???????????????????????
?
??????? !"??
(6)
From Fig. 1, general equations for the zerosequence
voltage U0 and the residual current of the protected feeder
3I0Fd can be derived:
?
++
Bgtot FdtotCC
YYY
??#???????$??????? ???????%???? !"?????
??
?
?
??
?
⋅
++
+++
⋅−=
FBg FFd
FBgFFduBg uFd
L
YY
YYYY
EU
1
0
(1)
)()(3
1
0
0FFduFdLFFd FdtotFd
YYEYYUI
+⋅++=
(2)
where
Y
Fd
,
LFdLFdL uFd
YaYaY
32
2
1
++=
=
,
Y
FdL
,
Fd
Y
+
L Fd
+
LFdtot
YYYY
321
++=
Y
,
BgLBg
120
LBgLuBg
=
Y
a
)
Y
sin(
⋅
aY
120
Y
a
Admittances YuFd and YuBg represent the asymmetrical part of
the corresponding total phase admittance, YFdtot and YBgtot. In
an ideally symmetrical network, YuFd and YuBg equal zero. In
practice, there is always some difference between the
phases, and according to Eq. 12 this asymmetry creates a
healthystate zerosequence voltage and residual current.
Neutral admittance protection is based on evaluating the
quotient between the residual current and zerosequence
voltage. According to the simplest approach the neutral
admittance is calculated utilizing the residual current and
zero sequence voltage phasors during the fault (at time t2).
)/(3
2_0
2_00
t
t Fd
UIY
−=
32
2
1
++
)
BgL BgL Bg1LBgtot
Y
32
=
cos(
oo
j
+
(3a)
An equivalent method is the use of phasors S1, S2, S3 and S4:
−⋅==
)(/ )
1
(5 . 0)Re(
2_02
0
0t
U absSSYG
(3b)
)(/ )
4
(5 . 0)Im(
2_03
0
0
t
U absSSYB
−⋅==
(3c)
where
=
S
=
)3( ),3
j
(
(
2_
3
0
2_0
U
2
2_0
3
2_01
t
I
Fd
j
t
tFd
I
t
IUabs
=
S
),
2
IU
U
abs
abs
S
−=
S
+
+
,
)(
2_0
2_04
_0
2_03
tFd
t
tFd
t
abs
⋅−⋅
Inserting Eq. 12 into Eq. 3a gives the following:
a)
Assuming an earth fault inside the protected feeder,
YFFd > 0, YFBg = 0:
)(YYY
BgtottCCin
⋅+=
where
=
,
2
k
=
YYYk
++=
3
b)
Assuming an earth fault outside the protected feeder,
i.e. in the background network, YFFd = 0 and YFBg > 0:
kkYY
Fdtotout
+⋅−=
where
/ )(
kYYk
FBg
=
,
5
k
=
YYYk
++=
6
From Eq. 45 it can be seen that the calculated neutral
admittance utilizing Eq. 3a is not singlevalued neither
during inside nor outside faults, but it is affected by e.g. the
degree of asymmetry of the network. Also the fault
resistance affects the measured admittance, which is not
desirable.
In order to mitigate these effects Eq. 3a can be enhanced by
utilizing changes in the zerosequence quantities due to the
fault:
21
2_0
kk
+
(4)
31
/ )(kYYk
FFduFd+
3
/)(
kYYY
uBgFFd Fdtot
⋅+−
FFduBguFd
54
0
(5)
64
uBg+
62_
/ )(
kYYYY
FBgBgtot
t CC
uFd
++⋅
FBg uBg uFd
))(/()3 3 (
1_02_0
1_02_00
tt
tFdtFd
UUIIY
−−−−=
∆
where t1, t2 are time instances prior to and during the fault.
Inserting Eq. 12 into Eq. 6, and setting the fault resistances
to zero prior to the fault (at time t1), the following equations
are obtained:
a) Assuming an earth fault inside the protected feeder,
YFFd > 0, YFBg = 0:
YYY
+=
∆
2_0
Bgtott CCin
(7a)
FFdt CCt CCtCC
tCC
m
+
tCC
Y(
Fdtot
−
in
YYYm
YY
−
mYm
Y
⋅⋅
−⋅
)
⋅+
=
∆
)(
)(
2
1_1_2_
1
2_1_
10
0
(7b)
where
m
=
0
=
FFdBgtotBgtottCCt CCt CC
YY
,
Ymm
,
,
YY
,
Ym
⋅⋅−++⋅−⋅
))()((
45
1_2_1_
3
)(
1
uBguFd
YYm
+
aBgaFd
YYm
+=
2
=
aBguFd
YYm
−=
3
aFd
−
uBg
YYm
−=
4
,
Bgtot Fdtot
YYm
+
5
uFdFdtot aFd
YYY
=
,
uBgBgtot aBg
YYY
−=
Admittances YaFd and YaBg represent the symmetrical part of
the corresponding total phase admittance, YFdtot and YBgtot.
The result from Eq. 7a is valid when the admittance YCC is
the same at time t1 and t2, YCC_t1 = YCC_t2. This is the case
when the ACF scheme III is used or when ACF scheme I is
used and t2 equals time when the resistor is reconnected
during the fault. It is also theoretically valid with the ACF
scheme II, when t2 equals the time prior to the connection of
the resistor during the fault. It is important to notice that the
neutral admittance obtained in this way is not affected by the
network asymmetry or the fault resistance.
The result from Eq. 7b is valid when the admittance YCC is
different at time t1 and t2, YCC_t1 ≠ YCC_t2. This is the case
when the ACF scheme II is used, and t2 equals the time after
the connection of the resistor during the fault. The equation
is also theoretically valid with the ACF scheme I, when t2
equals the time when the resistor is disconnected during the
fault. The obtained neutral admittance is not singlevalued,
i.e. it is affected by the values of e.g. YCC_t1 and YCC_t2,
degree of asymmetry of the network and the fault resistance.
b) Assuming an earth fault outside the protected feeder,
YFFd = 0, YFBg > 0:
YY
−=
∆
0
According to Eq. 8, the calculated admittance always equals
the total neutral admittance of the protected feeder itself
with a negative sign, i.e. –YFdtot. It is important to note that
the result is valid regardless of the applied ACF scheme. In
addition, the obtained neutral admittance is not affected by
the network asymmetry or the fault resistance.
By comparing Eq. 7a to Eq. 4 and Eq. 8 to Eq. 5 it can be
seen that by calculating the neutral admittance utilizing
changes in the zerosequence quantities due to the fault, the
result can be made singlevalued by selecting the time
instance t2 during the fault so that YCC has the same value as
at time t1 prior to the fault. From this point of view, ACF
schemes with resistor switching should be avoided when
Eq. 6 is applied. In practice this means the parallel resistor
should preferably be constantly connected.
Fdtotout
(8)
Page 3
? ? ? ??????
??????????
?????????
????????
????
?????????????????????????????????????????????????????????????????????????????????
?
??????? !"??
COMMON EF PROTECTION FUNCTIONS
As a common start criterion for EF protection functions a
U0 overvoltage criterion is typically used. Fig. 2 illustrates
U0 as a function of fault resistance RF in an example
network. The shaded areas represent the variation due to the
network asymmetry and the faulted phase with different
compensation degrees. In resonance condition (K = 1) the
healthy state U0 is 5% or 16% depending on whether or not
the parallel resistor of 5 A is connected. From Fig. 2 it can
be concluded that if the parallel resistor is not connected
during the healthy state, the considerably high healthystate
U0 limits the sensitivity of the protection. Therefore the
ACF scheme I or III should be applied in this example.
??#???????$??????? ???????%???? !"?????
Fig. 2 Behavior of U0 as a function of RF in an example network with
different compensation degrees when the 5 A parallel resistor is connected.
I0cos(ϕ ϕ ϕ ϕ) and phase angle criteria
In the I0cos(?) criterion operation is achieved when the
product: abs(3I0)*cos(?) exceeds the setting value.
Alternatively the phase angle criterion can be used, where
the operation is achieved, when the amplitude of 3I0 exceeds
the setting value and the phase angle ϕ between U0 and 3I0
is within the set limits, i.e. inside the operation sector. The
middle point of this sector is defined as the basic angle. In
compensated networks the basic angle equals 0o, and the
operation sector is typically either ±80o or ±88o wide.
Neutral admittance criterion
In the Y0 criterion, the neutral admittance is calculated using
e.g. Eq. 6. The result is compared to boundary lines in the
admittance plane. Examples of typical characteristics
available in modern feeder terminals are illustrated in Fig. 3.
The shaded area represents the nonoperation area i.e. the
operation of the protection is achieved, when the calculated
admittance moves outside the shaded area.
Fig. 3 Examples of Y0 characteristics.
The key result from analysis of Eq. 6 was that, when the
Y0 calculation is done utilizing changes in the zerosequence
quantities due to the fault, then in case of an outside fault,
the calculated admittance always equals –YFdtot i.e. the total
neutral admittance of the protected feeder itself with a
negative sign. This fact is utilized in the novel
characteristics presented in Fig. 4. The idea is to set the
nonoperation area around –YFdtot with a sufficient security
margin. The characteristic then becomes offset and
asymmetrical in the admittance plane. Such characteristic
improves the sensitivity of the protection and is valid also
when the compensation coil is temporarily disconnected.
Fig. 4 Examples of the novel Y0 characteristic.
The value of YFdtot is the primary setting base for the novel
characteristic. The imaginary part of YFdtot can be easily
calculated from the capacitive EF current of the protected
feeder itself:
Im(YFdtot) ≈ j⋅3I0Fd/Uphase
The value of 3I0Fd can be obtained directly from the utility
DMS, and can easily be updated, in case the feeder
configuration changes substantially. The real part of YFdtot
can be either determined by measurements or estimated to
be typically 20…30 times smaller than the imaginary part.
(9)
EARTHFAULT SIMULATIONS
The meaning of Eq. 3a and Eq. 6 is illustrated in Fig. 5
using simulated data. The simulated network represents a
simplified distribution network as illustrated in Fig. 1.
A singlephase earthfault is applied into each phase, and the
fault resistance is varied from 0 to 20 kΩ in 1 kΩ steps.
Two compensation degrees, K = 0.8 and 1.1 are analyzed.
The rated current of the parallel resistor (IACF) is assumed to
be 5 A. All ACF schemes are included in the simulation.
The degree of network asymmetry matches the values used
in the calculations represented in Fig. 2.
Fig. 5 Simulated behavior of the neutral admittance calculation algorithms.
From Fig. 5 the effect of fault resistance, network
asymmetry and the faulted phase on the different
Y0 calculation methods can clearly be seen:
• The best result is achieved, when the Y0 calculation is
based on Eq. 6 and when the admittance YCC is the same
prior to and during the fault, YCC_t1 = YCC_t2. The result is
then singlevalued and not affected by the value of RF,
the network asymmetry or the faulted phase. This allows
high sensitivity of the protection, especially if the novel
characteristic would be applied.
Page 4
? ? ? ??????
??????????
?????????
????????
????
?????????????????????????????????????????????????????????????????????????????????
?
??????? !"??
• If the Y0 calculation is based on Eq. 3a or on Eq. 6 when
the admittance YCC is different prior to and during the
fault, YCC_t1 ≠ YCC_t2, then the result is not singlevalued.
Depending on the faulted phase, each phasetoearth
fault creates an individual admittance trajectory, the
course of which depends on RF and network asymmetry.
This reduces the dependability of the Y0 criterion,
especially if high sensitivity is required.
??#???????$??????? ???????%???? !"?????
FIELD TESTING AND EXPERIENCE
In recent years, ABB Oy, Distribution Automation, Finland
has undertaken intensive field testing in cooperation with
some Finnish power utilities in order to test and develop
new EF protection algorithms. Below, one field test series is
studied. These tests were made in a 20 kV rural distribution
network with overhead lines in Finland. The compensation
degree during the tests was K = 1.1 and the ACF scheme II
was applied. The faulted phase was L3 and RF was varied
from 0 to 10 kΩ. The capacitive EF currents of the network
match the values used in the simulation model, see Fig. 5.
The performance of traditional I0cos(?) criterion is
evaluated in Fig. 6. A twostage protection is applied using
the lowset stage I0cos(?)> for alarming, and the highset
stage I0cos(?)>> for tripping. In this case, the stages are set
to 2.5 % and 20 %. This corresponds to 0.5 A and 4.0 A
primary currents. Red color represents the results when the
parallel resistor is connected and blue color when the
parallel resistor is disconnected during the fault.
Fig. 6 Evaluation of the I0cos(?) criterion based on field test data.
From Fig. 6 it can be concluded that in order for the
highset stage to operate properly, the parallel resistor needs
to be connected. Also with the selected settings, the lowset
stage cannot detect faults with 10 kΩ fault resistance. More
sensitive operation could be achieved by the use of phase
angle criterion and ±88o wide operation sector.
As a comparison, the Y0 calculation methods are evaluated
in Fig. 7, prior to (left) and after (right) the 5 A parallel
resistor was connected during the fault. The results using
Eq. 3a are shown with red and magenta, whereas results
from Eq. 6 are highlighted in blue and cyan. Red and blue
color represents results when a cable core CT was used for
residual current measurement, whereas magenta and cyan
represent results with a Holmgreen connection.
From Fig. 7 it can be seen that utilization of changes in the
zerosequence quantities (Eq. 6) reduces the deviation in the
calculated admittance. With this measuring principle, the
maximum tested fault resistance of 10 kΩ can be detected
easily, provided that the novel Y0 characteristic is applied.
Fig. 7 Comparison of performance of the novel (red) and traditional (blue)
Y0 characteristics and Y0 calculation methods based on field test data.
Also the applied start criterion must have adequate
sensitivity without the risk of false starts, e.g. when the
switching state changes in the network. Further, the
traditional overconductance (G0>) method lacks sensitivity
compared with the novel approach, which can operate even
without the parallel resistor due to the natural losses of the
coil and the network. However, it is recommended to keep
the resistor constantly connected, as then the discrimination
between inside and outside faults is improved.
In case twostage protection is required, it can be achieved
by two independent Y0 protection instances (Y01>, Y02>) with
differently set start criteria. For example in case of network
studied in Fig. 2, the Y01>stage for alarming could have
U0start set to 10% corresponding to RF ≈ 6 kΩ (K=0.8) and
the Y02>stage for tripping could have U0start set to 30%
corresponding to RF ≈ 2 kΩ (K=0.8).
For both evaluated EF protection functions the measuring
principle of the residual current has a noticeable impact, and
cable core CTs should be used if high sensitivity is required.
CONCLUSIONS
The performance of the neutral admittance based EF
protection has been studied. The results show that
Y0 protection is a respectable alternative to traditional EF
protection functions. Benefits include e.g. good immunity to
fault resistance and easy setting principles. Based on the
theory and field tests, the admittance calculation should be
based on changes in the zerosequence quantities due to a
fault, the novel Y0 characteristic should be used and the
parallel resistor should be permanently connected. This
maximizes the performance of the Y0 protection. The neutral
admittance protection function together with the presented
algorithm enhancements will be implemented in the next
generation feeder terminals applied in distribution and
subtransmission networks.
REFERENCES
[1] J. Lorenc et. al, Admittance criteria for earth fault detection in
substation automation systems in Polish distribution power networks,
CIRED 1997 Birmingham.