1

Optimal operational characteristics of

directional earth fault protection
Pedro J. L. Aleixo, José L. C. Pinto de Sá, Instituto Superior Técnico, Universidade de Lisboa, Lisbon
e-mail: pedro.aleixo@ist.utl.pt pintosa@tecnico.ulisboa.pt

different subsystems, namely, the instrument transformers, the

Abstract— The function of the directional earth ground fault protection unit and the analog-to-digital conversion. All these
protection is to protect against ground faults by selecting the sources of error are studied empirically. In this work, the
output at the substation. Nowadays, ground faults are covered by analysis of the errors is possible through a probabilistic
this function protection, that may have low sensitivity. Given the approach, that was used in previously works. This approach
large number of variables that influence the system, it is necessary
allows the quantification of the precision of the directional
to analyze their impact in order to reach an optimal setting. Highly
resistive ground faults result in very low residual voltages and ground fault protection. In turn, this opens up opportunities for
currents, but also an angle between them for different ground optimization. [2] [3].
connections. This results in a drawback in the normal exploration
of the MV network. II. OPERATING PRINCIPLES
The measuring devices’ errors and their electrical schemes used A multi-function relay with directional ground fault
in the calculation of residual voltage and currents were the subject
protection operates if the following conditions exist
of study in this work. Moreover, a protection unit was tested in the
laboratory to measure the loads of the voltage and current analog simultaneously during a setting time interval: Minimum of
inputs. zero-sequence current, 𝐼0 zero-sequence voltage, 𝑉0 and an
The adopted neutral regime and the variations between its angle between them, 𝛿𝑉0 − 𝛿𝐼0 [4].
characteristics overhead, underground and mixed lines were The first two conditions are intended to detect a defect in the
analyzed to define voltage and current of natural asymmetries. network, while the third condition indicates the directionality,
For the correct operation, several network topologies and forward or reverse. To calculate the angle between the voltage
voltage levels of MV were simulated in the ATP software. The and current, the zero-sequence complex power, 𝑆0 is used,
study considers the different characteristics of normal solutions, given by:
such as frames, cables, lines and height. The simulation conditions * j ( V   I 0 )
are the steady state operation for the calculation of the natural S0  V0 I 0  V0 I 0 e 0
(1)
asymmetries of the MV networks.
The operational characteristic of a directional earth fault
Finally, based on the detailed information, a probabilistic
analysis of the zones of operation was made conducted some error protection usually is present in Fig.1. It is characterized by two
ellipses calculated in MATLAB. zones, one for the operation and one for the non-operation. The
position of them in complex plan is the characteristic angle
Index Terms— Directional ground fault protection, ground called α.
fault, resistive fault, MV networks, zero-sequence complex power

I. INTRODUCTION
owadays directional earth fault protection has the function
N of detecting phase-to-ground faults in MV and HV
networks. However, this protection has some limitations
on the high sensitivity. A big advantage of this protection is to
ensure selectivity and to release communications in the lines.
The main problem of this protection is to give it a degree of
sensitivity, i.e., to establish its operating parameters, a
minimum of voltage and current is required for its operation as
a function of a fault resistance. In order to increase his Fig. 1 – Operation characteristic of the directional earth-fault
protection's sensibility, an analysis of the errors that influence protection, Adapted from [4]
its operating parameters (residual voltage, residual current and
Directional ground fault protection is affected by
the angle between them) is necessary.
measurement errors committed throughout this process. These
Currently the solution adopted for this problem is the inverse
measurement errors have their origin in measuring devices such
time protection based on the time for the selectivity which can
as voltage transformers, current transformers and protection
make this protection in the elimination of ground faults quite
units. The impact of these errors is not the same in all network
slow [1].
topologies and their ground system [1].
These measurements are affected by errors that stem from
 2

A. Protection System substation. A classical way to analyze the transformers is

To guarantee the proper operation of the directional earth through Steinmetz scheme present as depicted in the Fig. 2.
fault protection, it is necessary to allow for the different
possibilities of obtaining the components: zero-sequence
voltage, zero-sequence current and the delta of their angles.
In MV (medium voltage) or HV (high voltage) networks it is
necessary to consider the wiring diagram of the circuits of the
measurement transformers. The measure of the zero-sequence
voltage in the HV and MV networks is different. For MV
networks the open delta configuration is most common, which
consists of a circuit in the secondary of the voltage transformers
that reduces the voltage level in order to add all phases.
Whereas in HV networks the measurement of the zero- Fig. 2 – Steinmetz equivalent circuit, Taken from [5]
sequence voltage is made through the internal sum in the
protection. In this case the three phases are analyzed separately Regarding the voltage transformer, the secondary circuit is
from the voltage transformer to the inside of the protection unit, considered to be open. As for the current transformer the
then summed internally the three phases. On the measuring secondary circuit is in short circuit characterized by a load
side, it is common to have access of the phases through a cable, impedance [5].
contrary to what happens in HV networks, due to the higher A. Voltage Transformer
voltages, there is no access to the three phases together in cable The voltage transformers have the function of transforming
at the outputs of substation. The zero-sequence current in MV primary voltages, which require large insulation distances, into
is obtained through a core transformer (donut type), whose voltages that can be treated by the protection units easily, and
magnetic core envolves the three phases in the primary. In are divided into two types: inductive and capacitive. The
contrast in the HV networks the zero-sequence current is voltage transformers studied in this work were the inductive
normally measured by internal sum in the protection unit like type. To thoroughly analyze the constitution of a voltage
the zero-sequence voltage measurement. There is another transformer we use the Steinmetz scheme [5].
possibility that it is through the Holmgreen circuit that it
U s  ( I 0  I s )  R ' p  X ' s   I s  Rs  X s  (2)
considers the sum of all the currents in the secondary of the
np
U 'p 
current transformers. In this way Holmgreen is less precise than ns
the internal sum, having an advantage of the error being only in The voltage error has two components: ratio and phase. The
an analog input protection as opposed to being in three inputs. ratio error (Δu) is the difference between the voltage modules
In this way, the measurements in HV networks are clearly lower 𝑈′𝑝 and 𝑈𝑠 , while the phase displacement corresponds to the
in terms of accuracy due to the different readings of the zero-
phase angle between the primary voltage and the secondary
sequence voltage and current.
voltage rotated 180 ° due to the transformer [5] and [1].
Directional earth fault protection is strongly influenced by
the neutral grounding connection. As shall be later explained, B. Current Transformer
the grounding regime influences the angle between the zero- The main purpose of the current transformer is to provide an
sequence voltage and zero-sequence current, 𝛿𝑉ℎ − 𝛿𝐼ℎ [1]. image of the current (ratio and phase) that passes through the
primary circuit to a secondary circuit galvanically separated.
B. MV distribution and HV transmission networks
These transformers are used due to the high voltage levels. The
In the HV networks directional earth fault protection has reduction of the current thereof provides an easy interface
several issues, namely the fact that the grounding system is between the primary circuit and the protection units situated at
solidly grounded, which makes it difficult to detect the zero- lower levels of isolation. To measure current, the primary
sequence voltage. A factor conditioning the operation is the circuit is placed in series with the network, resulting in a smaller
measurement of the zero-sequence current in the HV networks current in the secondary circuit proportional to the turns ratio.
where the access to the outputs in cables for the use of a toroidal With this reduced current, the measuring devices and protection
current transform are not feasible. In this way, the errors made units can operate more easily and safely [1].
in the current measurement can be higher, which impairs the There is also a special type of current transformers called
correct obtaining of the directionality. Therefore, this work toroidal current transformers used in MV networks. In this
focuses on the analysis of this protection function in the MV work, the study of these transformers has the purpose of
networks of 10kV, 15kV and 30kV. At those voltage levels, the obtaining an image in the secondary of the residual current, i.e.,
grounding connection can be of various types which becomes 3 times zero-sequence current used specifically in the analysis
advantageous for the analysis and comparison. Additionally, of the ground faults [1].
since phase-to-ground faults have a higher incidence at MV From the Steinmetz equivalent circuit in Fig. 2 it is verified
levels, it is more valuable to analyze the MV networks [1]. that the main source of error is the magnetization current of the
transformer, which means that the current in the secondary is
III. “THE BACKBONE” OF THE SYSTEM ANALYZE not exactly the current that crosses the primary in both ratio and
Voltage and current transformers are key components phase. The impedance of the secondary (load) is also not zero,
because they are responsible for direct measurements at the which causes a voltage drop and is responsible for the
 3

appearance of the magnetizing current. The composition of the Zh  3Z N  jX ht (5)

load also contributes to the ratio error and phase displacement
[5]. It should be noted that the zero-sequence impedance of the
transformer only exists if the grounding connection is directly
np connected to the neutral of the transformer (star point). On the
Is  I p  I0 (3)
ns other hand, an artificial neutral should be used whenever the
winding transform is connected in delta, in which case the zero-
The ratio error, in absolute, is derived from the magnetizing sequence impedance of the transformer does not exist. The total
current 𝐼0 . This magnetizing current results from the magnetic
admittance 𝑌ℎ corresponds to the sum of this last impedance
flux in the core that induces an electromagnetic force in the
with the parallel of the total earth capacitances 𝐶𝑇𝑜𝑡𝑎𝑙 , which
secondary. The secondary circulating current, secondary
also includes the zero-sequence impedance of the transformers
impedance and load type should be reduced to the minimum
to LV (low voltage). This is a result from the π equivalent
value, ideally should be zero (short circuit) to reduce errors, but
scheme which characterizes a line [1].
technically impossible due to the cost of devices and sensitivity
1 1
of the measure devices [5]. Yh  j  Ch,k   jC  (6)
C. Protection Unit Zh Zh
The directional earth fault protection is in the protection Equation 7 corresponds to the generic expression for the
units. These devices that process observations of network data calculation of the fault current [1].
and its components, such as current, voltage, phase shifts, also 3Ea
I def  3I h  (7)
command the circuit breakers among other more specific 1
functions. The protection unit receives the network information 3R  j 2 X T 
1
through the secondary of the CTs and VTs installed in the jC 
output of the substations. The protection unit processes and Zh
calculates the desired parameters [1]. In case the grounding system is isolated, the currents are
In this work, the digital protections were studied unlike the closed only by the capacities of all lines. Since there is no
old electromechanical protections that today are only used in connection to the neutral, then the neutral impedance is infinite,
some specific places. The analog inputs of the protection units, which means that the total admittance is the result of all
specially designed for zero-sequence current and voltage capacities to earth between conductors [1].
measurement and associated errors were analyzed. The study of 1
errors in protection units can be divided into four parts: internal Yh  jC   jC (8)
voltage and current transformers; sampling resistors; analog-to- Zh
digital converter; sampling by DFT [1]. In the resonant grounded neutral is aimed to obtain a zero
ground fault current. The fault current corresponds to the
D. The grounding system
resistive current by the neutral point in the substation
The grounding system consists of the connection of the (accessible neutral or artificial neutral) and by the capacities of
neutral to the ground in a substation. This connection can be the line with fault. The reactance to be installed at the substation
made in several ways, all of them have advantages and neutral is named as Petersen's coil, 𝑋𝑃 . This reactance is rated
disadvantages. Grounding connection can be divided into two for a current equal to the capacitive current from all earth
parts: small fault currents, and large fault currents. The capacities of the lines [1].
grounding systems corresponding to the small fault currents are
the isolated and resonant grounded neutral. For the large fault
Z N  RP  jX P (9)
currents the grounding systems are: grounding with limiting The reactance is not perfect that´s why present a resistance.
resistance, grounding with limiting reactance and solidly Quality factor named for Equation 10 [1]:
grounded systems. The last is not used in sub transmission XN
(MV) but in transmission (HV) networks [1]. Q (10)
A simplification that can be applied in this case is the
RN
generators are far away from earth fault. With this we can admit The resonant grounded neutral is rarely tuned, because tuning
the positive sequence impedance approximately equal to the causes high imbalance which leads to stress in the insulation in
negative sequence impedance 𝑍1 ≈ 𝑍2 , neglecting the real part the network. This way, the Petersen coil is not in tune, the
of the upstream network, 𝑋𝑚 and the transformer 𝑋𝑇𝑑 . Petersen coil is not really in tune, but more in an over-
Considering that the ground fault is close to the substation, then compensation state δ, so that in the case of an output being
the longitudinal impedances between the substation transformer disconnected, the resonant grounded neutral does not return to
and the ground fault point are neglected. Equation 4 is a direct tune [1].
result of this assumption [1]. 1
XP  1    (11)
Zi  Z d  j  X m  X Td   jX T (4) 3Ctotal
The zero-sequence impedance 𝑍ℎ is constituted by the In large fault currents, the grounding system could be a
impedance of the neutral 𝑍𝑁 and the zero-sequence of the resistance or reactance limitation. There are two extremely
transformer of substation 𝑋ℎ𝑡 [1]. important points that imply the use of an impedance between
the neutral of the distribution network and the earth [1].
 4

Firstly, to obtain reduced currents, there are hence extremely important. In a healthy network, the protection must
overvoltages, i.e., the level of insulation that the lines must not operate. Since the values of the asymmetries are measured
support for a safe operation of the network. In this case it is also through the protections that obtain the values of the secondary
relevant to add that the insulation levels of the insulators have ones of the measurement transformers, these devices have an
associated costs. As such, if the MV network can operate with associated error. By analyzing the natural asymmetries and
lower overvoltages, its dimensioning will be more economical. knowing the range of values of zero-sequence currents and
Secondly, if the fault current is high then the overvoltages are voltages, it is possible to obtain the zone of non-operation
attenuated. The use of a reactance in the grounding system related to the natural asymmetries of the network.
rather than a resistance is based on the substations where the To achieve high sensitivity for this protection function,
neutral point is not accessible due to the MT winding of the natural asymmetries should be considered. Since these natural
transformer being in a delta. The neutral with reactance is no asymmetries essentially depend on the characteristics of the
more than the creation of an artificial neutral taking advantage network, namely position of conductors. The Fig. 3 represents
of the dispersion reactance of its own transformer zig-zag type the self and mutual capacities of three phase system and
[1]. respective Equations 13 [6].
Z N  RN  jX N (12)

IV. METHODOLOGY
A. Laboratory
As the analysis of the load of the secondary circuit of the
current transformer is fundamental to know the accuracy of the
equipment, the laboratorial work is based on the study of the
external equipment placed in the secondary of the current
transformer, as is the case of the protection unit. However, other
equipment such as to measure or energy counters can be placed
Fig. 3- Earth capacitance of a three phase line with neutral
in the secondary circuit, which affects the secondary load. In connection, Taken from [6]
this way, an analysis of the load characteristic in the secondary
circuit of the voltage and current transformers was carried out.  jC AVA
The TPU-S420 protection unit was chosen to be analyzed in I A 
jC A Z N  1
 
 jC AB VA  VB  jCCA VA  VC 
laboratory. 
The board is responsible for the decoupling between the  jCB VB (13)
electronics circuits and instrument transformers. This board IB 
jCB Z N  1
 
 jC AB VB  VA  jCBC VB  VC 
contains several internal transformers to analog inputs. There 
are four voltage transformers dedicated to 3 normal and 1 
sensitive inputs and four current transformers, 3 normal and 1  I C  jCC VC  jCCA VC  VA  jCBC VC  VB
   
 jCC Z N  1
sensitive inputs.
The tests with a digital multimeter are aimed at collecting the In a healthy network, the neutral voltage ̅̅̅𝑉𝑁 is common to
primary winding resistance of the internal transformers of the all outputs and depends essentially on the grounding connection
protection unit. The tests with alternating current(AC) were used, but it is also influenced by the constant unbalance that
carried out with the aim of obtaining a complete characteristic relates the capacities of all lines in each phase. [6]
of the impedance of the transformer, not only in the primary VN 3 j kC
circuit, but also what influences the secondary in the primary  (14)
Vd 1
circuit.  3 jC
To control the voltage levels to be applied to the circuit, an ZN
autotransformer was used to regulate the voltage applied on the On the other hand, in the calculation of the residual current
secondary circuit. This transformer was chosen in a suitable of asymmetries, only each output of the substation and its
way for the tests. The voltage analog inputs was measured by imbalance are evaluated, because the voltages and currents
an oscilloscope. The waves of voltage and current were could be different in each output. [6]
visualized and used to calculate the phase angle.
For the current analog inputs measurement, a first attempt
 N

was by using an oscilloscope. However, since the measured
current was too small a multimeter was used to overcome the   9C i  k jC j 
low precision of the oscilloscope. The aim was to measure I N ,i   3 j ki Ci j 1
 Vd (15)
 1 N

current and voltage at the protection input terminals and
consequently the input impedance for each of the voltage
  3 j  Cj 
 ZN j 1 
inputs. The ATPDraw software (Alternative Transient Program) was
B. Natural asymmetries of networks used to study the natural asymmetries of the networks. Using
this software, the MV networks 10kV, 15kV and 30kV were
In order for directional earth fault protection to operate
correctly, the analysis of the network’s natural asymmetries is
 5

analyzed by simulation with the realistic characteristics   Re al  Re al 

commonly used, Table I.    Ar 
TABLE I
SIMULATION DATA
 TTA

J 210   (20)
Specification Type - Network Values  
Overhead, 15kV and 30kV Scc = 500 MVA
  Im ag  Im ag 
Upstream   TTA  Ai 
network Underground, 10kV Scc = 2000 MVA
The covariance matrix (𝐶𝑒 ) is a square matrix containing the
Sn =20 MVA
Substation Transformer variances and covariances associated to the different variables,
Vcc =8%
Sn =160 kVA errors. The elements of the main diagonal of the matrix contain
Outdoor Transformer the variances of the variables and the elements outside the main
Transformers Vcc =4%
to LV Sn =1 MVA diagonal contain the covariance between all possible pairs of
Indoor Transformer
Vcc =5% variables [6].
150 kW
Var ( x)  E  X    
Loads 2
300 kW (21)
 
W=Watt, VA=Volt Ampere, V=Volt
Cov ar iance( X , Y )  E  X   X Y  Y   (22)
The simulation was performed in the normal conditions of
operation of the network, that is, in a healthy network, its
purpose being to simulate the different arrangements of lines  Var (VTA ) Co var(VTA ,  Si ) 
and cables. Also in consideration is the different grounding Ce1010    (23)

connection.
Co var( Si , VTA ) Var ( Si ) 
C. Probabilistic method analysis This result serves only to calculate errors associated to zero-
In the analysis of the errors of these devices, only the worst- sequence measurement, but for this protection but for this
case scenario is considered for security reasons. That is to say, protection other parameters are necessary.
the probability of the error being maximum or being minimum The operation of the protection unit is based on the
is equal. The characteristic of this error probability is the calculation of the zero-sequence power measure. For example,
uniform distribution. To calculate the variance in this error associated in Equation 24. With this measure is possible
probability distribution the Equation 16 is used. [7] to obtain the angle between the zero-sequence voltage and the
(b  a ) 2 zero-sequence current without resorting to more complex
Var ( x)  (16) calculations [8].
12 Var ( X  Y )  Var ( X ) Var (Y )  E( X )2Var (Y )  E(Y ) 2Var( X ) (24)
Zero-sequence voltage measurement can be done in one of Zero-sequence power is obtained by two measures: zero-
two ways. The first is by using software (internal sum in sequence voltage and current. At this point we are missing the
protection) and then the error is considered for three inputs. The zero-sequence current. There exists two distinct ways of
second is the open delta method, which involves adding the measuring the zero-sequence current. The measure could be
three phases in the secondary of the voltage transformer (𝑉𝑟𝑒𝑠 = done using the software (like in zero-sequence voltage
3𝑉0 ). This way, the protection error resides in only one input measure), whose error is influenced by the three current
referring to the residual voltage [1] and [6]. transformers. The other option is measuring in the primary
Software  3  VT  3  VT int ern  3   sampling (17) through a toroidal current transform, using only one
transformer.
open _ delta  3  VT  VT int ern   sampling (18) To calculate the zero-sequence current error for the internal
The method used to determine errors consists in considering sum and for the toroidal transformer, respectively Equation 24
the measured errors as random variables according to the and Equation 25.
uniform probability distribution. The measure of the voltage I Software  3   CT  3   CT int ern  3   sampling (25)
affected by the errors of the three components, VT, unit
protection and sampling is given by [6]: I toroidal   CT   CT int ern   sampling (26)
VXm  1   VTX 1   VPX  e  VTX VPX  VX   SrX  j SiX (19)
j  
In MV networks it is common to use toroidal current
In order to analyze the propagation of the zero-sequence transformers, Equation 27.
voltage measurement by open-delta, must begin by applying the
Jacobian matrix, where the errors are divided into real and I 0* 
1
3
 
1   CT 1   CP  e jCT CP    SP  j SP I res (27)
imaginary parts [6]. To obtain the zero-sequence power complex that includes the
zero-sequence voltage (appendix equation) equation and the
conjugate of zero-sequence current (Equation 27).
The method is identical to the measure of zero sequence
voltage previously discussed.
 6

The result of the covariance matrix will be a 2 × 2 matrix that V. RESULTS

allows to determine the measurement error of zero-sequence
A. Laboratory
power in the complex plane [6].
 Var (Re) Co var(Re, Im)  The laboratory tests aimed at measuring the primary winding
Matrixerrors  CS 0   (28) resistance of the analog input of the protection in DC, unlike the
Co var(Im, Re) Var (Im)  AC test where the impedance seen from the primary is obtained.
The process of computing error ellipse can start. In this two- The voltage analog inputs of this protection unit are all the
dimensional space, the probability boundary has the shape of an same, i.e. the sensitive input is equal to the normal inputs. The
ellipse. There are two error variables, the real part and the results show that the resistance values are quite close, except
imaginary part. In general terms, a contour with probability 1- for the voltage analog input UC that has a slightly higher
α of a multivariate distribution can be expressed by Equation resistance than the others. The results of the AC tests for the
29 [6]. voltage inputs show that the input impedance is quite high,
 X   C 1  X      p 2   considering that the secondary is open. Since there is no
T
(29)
information on current and voltage waves, it is not possible to
2
In the chi-square distribution ( 𝜒𝑝 ), p corresponds to the conclude whether the voltage analog input is of the inductive or
number of variables or degrees of freedom and in this case takes capacitive type.
the value of 2, real and imaginary. This distribution defines the For analog current inputs, the AC tests are considered
probability boundary of the ellipse for a given confidence level, because the secondary ones of the internal CTs are short-
which allows the observation to be accepted. circuited. The results of the tests are purely resistive. The
Next step is the calculation of eigenvalues and eigenvectors current analog inputs have an almost purely resistive
through Equations 30 and 31 [6]. characteristic, with an angle of 1.8 ° at the sensitive input and
C  I  0 (30) 1.26 ° at the normal inputs.

C  e i   i e i
B. Instrument transformers tests
(31)
In order to analyze the errors of the measuring instruments,
Then by means of the previously calculated eigenvectors the the results were compared with some laboratory tests assigned
directions of the new axes of the ellipse are defined. Associated and performed by ABB for the same purpose. The instrument
with these vectors are the eigenvalues that influence the transformers are tested for various conditions such as for the
amplitude of the ellipse vertically and horizontally according to load and its type, resistive or inductive.
the chi-square distribution, as shown in Fig. 4. This distribution The voltage transformer usually has two windings, one of
causes the ellipse area to increase with increasing confidence which is dedicated to the measurement in each phase while the
level. Fig. 4 is the representation of the error ellipse as a other one is aimed at the protection, usually for the connection
function of the previously calculated parameters [9]. to the open delta. The very resistive ground-faults have equal
phase voltages in the order of the rated voltage, so the residual
voltage measurement by the open delta has a smaller error ratio
and phase displacement. As far as the error ratio is concerned,
it is not possible to conclude its nature, for it may be positive or
negative. This result is due to the compensation of the number
of turns in the transformer that can make the signal of the error
ratio vary. It has been experimentally verified by the tests that
the reduction of the VT load causes the increase of the error
ratio in the positive direction, which can increase or decrease
the error. Regarding the phase displacement, it is observed that
by decreasing the transformer secondary load the error
decreases, that is, the phase displacement is reduced in the
negative direction.
Fig. 4 – Example of error ellipse, Adapted from [10]
In the protection class CTs it is more difficult to conclude
because they are tested for rated currents. Thus, the plausible
From this probabilistic method, it was possible to comparison of the errors of the 5P transform will be made with
characterize all the errors of the zero-sequence power. Errors as the class 1 current transformer. Since no class 1 current
they have a uniform distribution have a mostly pessimistic transformer tests were obtained, those of class 0.5 were used,
weight, which makes the results with low risk. In this way, it where the errors are roughly half of those of class 1. In the first
was finally summed up in only two errors, error of the real part place, it was observed that the errors in the tests of the CTs have
and error of the imaginary part. The area of this ellipse no standard. Regarding 2 separate CTs 5P10 it was observed
conforms to the desired confidence level. that even though they possess the same characteristics, the error
ratio differed, with one being positive and the other negative.
For the phase displacement, although it is positive in both CTs
the results are not the same. The phase displacement is of the
same magnitude of the ratio error. This is due to the test
conditions, namely that the load is not purely resistive.
 7

In the various tests of the measurement CTs, the error ratio because different arrangements of overhead conductors
is not consistent with changing loads. Both increase and produce different ground capacities of the lines. Thus, it was
decrease, either positive or negative. In the various tests, it was necessary to adjust the compensation on the Petersen coil. The
found that the phase displacement was positive and increased results obtained are found in Table IV.
slightly in module. In general, the errors are much smaller than These results are for the worst-case scenario among the three
the standard IEC 61689-3. positions of conductors. In practice, the value of the
asymmetries will be between the clover geometry and the
C. Natural asymmetries of networks
horizontal and vertical mats geometries.
To simulate the networks, the software ATPDraw was used. TABLE IV
The simulation is based on the lines / cables. A conductive cable RESULTS OF WORST CASE SCENARIO OF NATURAL ASYMMETRIES
per phase was used in this case and the Bergeron model was Confidence Voltage Characteristic
Vres
chosen since lines and cables could be characterized with the level level
[%]
Ires [A] angle
distances between conductors, heights, among other 99,5% [kV] [°]
parameters. The ground resistivity was assumed to be 20 10 0,184 5,85E-05 undefined
ohm.m, the fundamental operating frequency 50 Hz and the Isolated 15 4,907 1,26E-04 undefined
30 4,935 3,15E-04 undefined
respective distance. The cables and lines were then
10 4,378 6,06E-01 -92,344
dimensioned according to the desired characteristics.
Resonant 15 18,8 2,3E-01 -92,257
TABLE II 30 37,9 1,65E+00 -92,222
NETWORKS CHARACTERISTICS
10 0,096 2,21E-02 -179,243
Outputs Resistance to
15 0,235 4,97E-02 -152,913
Network 300A
Utilization Distance 30 0,409 2,00E-01 -164,778
[kV] Number Type
[km] 10 0,095 3,85E-02 -102,411
Reactance to
Underground 15 0,289 5,15E-02 -102,433
10 Urban 10 6 300A
cable 30 0,459 2,14E-01 -102,048
Overhead 10 0,028 2,95E-02 -104,043
15 Short distance rural 10 30 Reactance to
line 15 0,159 5,02E-02 -103,295
1000A
Overhead 30 0,197 2,03E-01 -102,946
30 Long distances rural 5 60
line
D. Results of probabilistic method analysis
TABLE III
CONDUCTORS CHARACTERISTIC The results of equiprobable zones were obtained by
calculating the residual power vector of the faulty line and the
Type Overhead - ACSR Underground - Al
healthy line using the probabilistic method described in
Area
90 240 Chapter IV. To these areas, in this case they are called the error
[mm2]
ellipse of the measure of the residual complex power.
Total radius
[mm]
6 17 The following simulations take into account some aspects:
• Positive-sequence component, in this case of voltage, is the
Conductor insulation - PEX
single voltage of each network
Sheath insulation - PVC • Negative-sequence component is relatively low, therefore
assumed negligible
Resistivity 2,82×108
[Ohm m]
- • The desired confidence level is 99.5% as it is an acceptable
risk value
When using the ATPDraw software function line check, the
TABLE V
zero-sequence shunt admittance line is obtained. In this case, CT AND VT ERRORS
the values of the line capacities are analyzed [11]. Phase
The dimensioning of the impedances for each grounding Ratio error
displacement 𝝈𝑹𝒂𝒕𝒊𝒐 𝝈𝑷𝒉𝒂𝒔𝒆
connection is based on the equations described in Chapter III. [%]
[min]
In the simulation of the resonant grounding connection of the 1% 9 540 5,196 9,032
10kV underground network, the Petersen coil was first used de- 7,5 450 4,330 7,536
2%
tuned, and the asymmetries were reduced. In this way, the
simulation was repeated considering the Petersen coil in tune. %𝑰𝑵 3% 6 360 3,464 6,035
Later this neutral impedance is used in simulations of the 5% 3 180 1,732 3,022
asymmetries in the MV networks. 20% 1,5 90 0,866 1,511
Due to the strong contribution of capacities to ground in the
%𝑽𝑵 100% 3 120 1,732 2,015
exclusively underground network, the maximum fault current
is 5A. In the exclusively underground 10kV MV network, the Residual voltage through open delta at the sensitive input of
asymmetries are not significantly high, thus allowing a the protection. Residual current is obtained by a toroidal CT.
comparison of the de-tuned resonant grounding connection in The limit of the minimum residual current value allowed in
the 15kV and 30kV networks. In the 15kV and 30kV overhead the protection is 10%IN. Assuming a sensitive current input of
MV networks, the maximum fault current is 3A. In the resonant 0.2A and the secondary circuit with the rated current of 1A then
grounding connection, some adjustments were necessary the minimum possible parameter setting is 2%IN primary value.
 8

The limit of the minimum value of residual voltage in the

protection is 0.5%Vres. The analog voltage sensitive input
dedicated to the open delta, must have a setting of 0.5% of the
maximum residual voltage. The characteristic angle is between
0 ° and 90 °.
For the isolated neutral regime, the vectors of the residual
powers of the healthy lines and the faulty line are in opposition.

Fig. 7 - Equiprobable zones of healthy line for currents from 1A, 5%

of CT- 10/1A

Although the quality factor is high, the Petersen coil is tuned,

causing the resistive component of the faulty line to be higher.
This decrease in the reactive component is due to non-
overcompensation in this neutral regime.
Fig. 5 – Equiprobable zones of faulty line for currents from 0,5A, 1%
of CT- 50/1A

Fig. 5 and Fig. 6 are the measured zones of residual power in

a faulty and healthy line. These simulations take into account
an error for the current about 1%IN. In case of faulty line, with
a characteristic angle of 90° there is no problem. The operation
and non-operation zones are distinct. The major limitation of
this neutral regime is essentially the minimum current
parameterization by the protection unit.

Fig. 8 - Equiprobable zones of faulty line for currents from 1A, 5% of

CT- 10/1A

Through the faulty line showed in Fig. 8, the characteristic

angle of 10° is sufficient to obtain directionality. For networks
of 15kV and 30kV with overhead lines and the
overcompensated of the Petersen coil with quality factor 20, the
vector of the residual power of the healthy line to the faulty line
is closer. Given this, it is reasonable to assume the current error
in the CT current range of 20%IN. Note that for this range of
Fig. 6 - Equiprobable zones of healthy line for currents from 0,5A, currents the error zone does not cross the characteristic line of
1% of CT- 50/1A 5º. This can be shown by Fig. 9.

For resonant grounded neutral, it was possible to use a

Petersen coil with a quality factor of 30 in an underground
network, 10kV. For this network, it was possible to explore the
Petersen coil in tune due to the low asymmetries verified in the
ATPDraw simulations. This network has a high value of the
total earth capacities which makes the reactive current of the
neutral is obligatorily high to compensate. This leads to the
output current also having a higher value of resistive component
that is closed the defect point. To obtain the directionality it is
reasonable to admit the current error in the CT current range of
5%IN. Fig. 7 is the zone measured of residual power in a
healthy line. Fig. 9 - Equiprobable zones of healthy line for currents from 2A, 20%
of CT- 10/1A

In Fig. 10 the error zone of a faulty line doesn’t overlap the

zone of a healthy line. The reactive component in a faulty line
 9

with overcompensation is greater than a grounding connection the faulty line overlaps the non-operating zone with
with a Petersen coil in tune. characteristic angle 8 °. This characteristic angle of healthy line
is in Fig. 13.

Fig. 10 - Equiprobable zones of faulty line for currents from 2A, 20%
of CT- 10/1A Fig. 13 - Equiprobable zones of healthy line for currents from 2,5A,
5% of CT- 50/1A
A grounding connection with limiting resistance of the fault
current at 300A the characteristic angle is favorable. The faulty By reducing the ratio X/R = 3, the residual power error zone
line has the phase a little over 90 ° from the healthy line, as Fig. of the faulty line is located outside the non-operating zone. As
11 represents. can be seen in Fig. 14, the residual power error zone of the
faulty line does not exceed the characteristic angle of 11 °.

Fig. 11 - Equiprobable zones of faulty line for currents from 1,5A, Fig. 14 - Equiprobable zones of faulty line for currents from 2,5A,
3% of CT- 50/1A 5% of CT- 50/1A

The detection of the directionality by the difference of the In the 10kV network, the angle associated to the residual
current in healthy line and a faulty line is directly observable. power is -106.78°, i.e., the angle is further away from the
The equiprobable zones of the residual power of a healthy line residual power of healthy line, in relation to the networks 15kV
take into account the error of the CT in the range of currents of and 30kV with overhead lines. The operating area of the faulty
3%IN. For this case, the error ellipses that were obtained by line in the 10kV network does not overlap the healthy line. The
simulation through Fig. 12 do not exceed the characteristic new characteristic angle is 9° Fig. 15.
angle of 14 °. Thus, it has no influence on the operating zone
characterized by the characteristic angle of 45 °.

Fig. 15 - Equiprobable zones of faulty line for currents from 2,5A,

5% of CT- 50/1A
Fig. 12 - Equiprobable zones of healthy line for currents from 2A, 2%
of CT- 100/1A In the neutral system with limiting reactance at 1000A the
characteristic angle of the faulty line is similar to that of the
In the case of faulty lines, for a neutral reactance limited to limiting reactance of 300A, assuming that all the networks with
300A with the ratio X/R = 5, the residual power error zone of this neutral regime have the quality factor of 3. In this case the
 10

asymmetries are very low, thus with minimum residual voltage APPENDIX
of 0.5%, the corresponding residual current is also 0.5% of the
A. Zero-sequence voltage measure by open delta
maximum fault current of 1000A. Finally, Table VI considers
all the directional earth ground fault protection  1   VTA  e j VTA   1  VTB  e j VTB   
parameterizations analyzed in this chapter. To achieve those   
results, the instrument transformers considered are: voltage   1   VTC  e j VTC   
   
transformers with open delta connection for networks 10kV,
V0  e j VP  1  VP    Ar  j Ai  
15kV and 30kV. For current transformers, the values primary  
and secondary are 10/1A in neutral regime resonant, 100/1A  
j  VTB   
2 

 1   VTA  e  1  VTB  e 
j VTA 
Reactance to 1000A and for the others neutral regime 50/1A. 3 

  2 
  (32)
TABLE VI 1  j  VTC   
SETTINGS FOR DIRECTIONAL EARTH FAULT PROTECTION V0 m    1   VTC  e  3 

3
Confidence Voltage Vres Ires Characteristic 𝑅𝑑𝑒𝑓 𝑚𝑎𝑥  
V1  e 1   PV    Ar  j Ai 
j VP 
level level [%] [A] angle [Ω] 
99,5% [kV] [°]  
j  
2 

10 0,5 1 90 6350  1    e j VTA   1    e  VTB 3   
Isolated 15 5 1 90 9490  VTA VTB

30 5 1 90 16623  
j  VTC 
2 
 
10 6 1 10 18651   1   VTC  e  3 

Resonant 15 47 2 5 5037  
V2  e 1   PV    Ar  j Ai
j VP 
30 25 2 5 19338 
Resistance 10 0,5 1,5 45 4215
to 15 0,5 1,5 45 6319 VII. REFERENCES
300A 30 0,5 1,5 45 12638
10 0,833 2,5 9 2534
Reactance
15 0,833 2,5 11 3800
[1] J. L. Pinto de Sá, Textos de apoio à cadeira de
to 300A Proteções e Automação em Sistemas de Energia,
30 0,833 2,5 11 7600
Reactance 10 0,5 5 11 1267 Lisboa, 1993.
to 15 0,5 5 11 1902 [2] J. L. Pinto de Sá, "Stochastic analysis in the time
1000A 30 0,5 5 11 3805 domain of very high speed digital distance relays Part I:
Theory," IEE Proceedings - Generation, Transmission
and Distribution, vol. 141, no. 3, pp. 161-168, Maio
VI. CONCLUSION 1994.
Regarding the current analog inputs of protection unit, a [3] J. Pinto de Sá, J. Afonso, R. Rodrigues, "A probabilistic
valid conclusion of this study is that the secondary circuit of the approach to setting distance relays in transmission
CTs may have a resistive characteristic, which causes the errors networks," IEEE Transactions on Power Delivery, vol.
to be mostly of phase. This results from the current analog 12, no. 2, pp. 681-686, Abril 1997.
inputs being resistive, having also considered the low reactance [4] EFACEC Sistemas de Electrónica, S.A., TPU S420
observed in the secondary of the toroidal current transformers. Edição 1, Moreira Maia, 2005.
The network of underground lines showed reduced [5] Á. E. Jaime Berrosteguieta, Theory and technology of
asymmetries, whereas in the networks with overhead lines this instrument transformers, Arteche, Ed.
was not observed. Operation of the network in resonant regime [6] Migue Louro, O sistema de protecções na perspectiva
showed the highest asymmetries. The arrangement of the de segurança de pessoas em redes de MT, IST, 2008.
conductors in clover geometry showed the lowest asymmetries,
while the arrangement in vertical mat geometry showed the [7] M. C. Morais, Notas de apoio da disciplina de
highest asymmetries. The characteristic angle of the residual Probabilidades e Estatística, Lisboa, 2010.
power of all lines when there is no fault in the network between [8] EDP, On the limits of Directional ground fault
zero-sequence voltage and current is approximately equal to the protection, No prelo.
line in case of a fault. [9] M. Vermeer, "Statistical uncertainty and error
After the simulations in Matlab of the error zones of the propagation," Aalto University, 2014.
residual power complex it was possible to identify that the
[10] PennState Eberly College of Science, "4.6 - Geometry
quality factor of the coil/reactance is determinant in the neutral
of the Multivariate Normal Distribution," 2017.
regimes that have it. In order to optimize in some cases such as
[Online]. Available:
isolated neutral or with limiting impedance the protection unit
https://onlinecourses.science.psu.edu/stat505/node/36.
must have a lower minimum parameterization of the voltage
[Accessed 6 Março 2017].
and current sensitive analog inputs. In order to optimize the
resonant neutral regime, it is necessary to improve the obtaining [11] L. Prikler and H. K. Høidalen, ATPDraw Users'
of directionality, which consists of lowering the quality factor Manual, Preliminary Release No. 1.1 ed., 2002.
of the Petersen coil.