BULETINUL INSTITUTULUI POLITEHNIC IAŞI

TOMUL LV (LIX), FASC. 1, 2009

ELECTROTEHNICĂ, ENERGETICĂ, ELECTRONICĂ

FAST METHOD FOR ACCURATE FAULT TYPE

DETECTION IN HIGH VOLTAGE NETWORKS
BY

*F. MUNTEANU and *C. NEMES

Abstract. The papers focuses on a fast and reliable method for fault type detection
based on the initial value of the d.c. components of the short-circuit current in every
phase. To avoid some ambiguous cases due to initial voltage phase at the fault
initiation, the negative and zero sequence fault components are to be considered. Fast
and accurate are the main attributes of this method due to lack of iterative calculations
and to the qualitative analysis of the initial d.c. components.
Keywords: short-circuit current, fault type detection, DC component.

1. Introduction

Detecting and solving faults in power systems actually is of a great

importance related mainly but not only to the new smart grids concept
including here special aspects like renewable sources development, distributed
generation, availability, power quality and load–source compatibility. Fast and
accurate fault location plays an important role in accelerating system
restoration, reducing outage time and involved costs, diminishing associate
transients, and significantly improving system reliability.

2. State of art in fault location and type detection methods

There are many methods for fault location usually considering the
heterogeneity of lines, unbalances due to non-transposed lines, presence of
single, double and three phase loads, presence of laterals along a main feeder,
etc. All these elements introduce errors when calculating fault locations.
The methods for fault location are based on [1]:
- impedance and other fundamental frequency components;
- high frequency components and traveling waves;
- knowledge: artificial intelligence and statistical analysis, distributed
device and hybrid techniques.
Concerning the fault type detection methods, the literature is not so rich in
presenting the subject. Only during the last years the problem was increasing in
importance due to the new and very fast circuit-breakers and to the intelligent
switching technology [2], [3] in high voltage networks.
Fault type detection can be very useful as an initial sequence in solving a
fault, followed immediately by the fault location which can be, consequently,
easier to implement.
 One of these methods was published in 2004 by Kiyoshi Yoshinaga based
on an algorithm for fault type detecting near generators considering the
influence of the time constant difference between the presumed and the actual
values for current detection as well as the generator transient and sub-transient
time constants affecting the fault current wave shape [4].
In 1999, A. Pöltl and K. Frölich published a method for fault type detection
by means of parameter fitting and artificial neural networks [5].
Some papers referring to the fault location methods implicitly consider the
fault type even this is very important not only for theoretical reasons but
specially for practical industrial activity: we need to know first what happened
and after that where!

3. A new method for type fault detection

This method is based on the qualitative relationship between the initial

values of DC component of the short-circuit phase currents. This relationship is
specific for every fault type as it will be demonstrated here.
A simplified diagram of a power circuit with a fault located in k is shown in
fig.1.
P, Q
eg
x sis
e sys
”
xd xt dxl k (1-d)xl
Fig.1 The simplified circuit for fault type detection

In fig.1 eg is the machine internal voltage, xd’’ its sub-transient direct axis
reactance, xt the power transformer reactance, d the fault distance from the
source, xsys the equivalent system reactance and esys the system voltage. It was
noted:
(1)

x g = x + xt + d ⋅ xl "
d
(2) x s = (1 − d ) xl + x sys
In the followings it was considered:
- the voltage at the fault location and before the fault is 1 p.u.;
- the circuit resistances are zero (they were neglected);
- the initiation fault moment is given by voltage angle θ.
For accurate and fast fault type detection during the first or even the first
half of a time period of the short-circuit current it is strongly necessary that:
- the method principle to be a comparative one knowing the main
 BUL. INST. POLIT. IAŞI, TOMUL LIV (LVIII), FASC. 5

advantages of this kind of measurement techniques: independence of the

measurement device errors, high measurement precision and practically
instantaneous;
- full independence by the fault moment initiation;
- the primary information has to be taken from high class current
transformers; Rogowski type coils could be a suitable solution.
Having in view that P and Q is active and reactive power respectively
flowing through the line at the fault point before the fault initiation, the phase
load currents are expressed by:

 ia = P s i nω (t + θ ) − Q c o sω (t + θ )

(3)  ib = P s i nω (t + θ − 2π / 3) − Q c o sω (t + θ − 2π / 3)
 i = P s i nω (t + θ − 4π / 3) − Q c o sω (t + θ − 4π / 3)
c
The Clarke component currents are given by:

 iα = P s i nω(t + θ ) − Q c o sω (t + θ )

(4)  iβ = − P c o sω(t + θ ) − Q s i nω(t + θ )

 i0 = 0
The currents are expressed in p.u. while the base current is the crest value
of the rated current enabling to remove the coefficient 2 .
To use the well known overlapping effects theorem, the currents variations
are necessary to be known. The following equation gives these variations where
DC transient components are contained since the currents in inductive circuits
do not change suddenly:

 ∆ iα = M α [ − c o sω (t + θ ) + c o θs ]

(5)  ∆ iβ = M β [ − s i nω (t + θ ) + s i nθ ]
 ∆ i = M [ − c o sω (t + θ ) + c o θs ]
 0 0
In equation (5) coefficient M is the magnitude of the current variation
depending on type of fault.
The transformation equation from the Clarke components to phase
components allows determining the currents after the fault occurrence from
which we use only idca, idcb and idcc as the initial values of DC components of the
phase fault currents. We can express them depending on the fault type as it
follows considering also the sequence line parameters related to the
asymmetrical type faults.
a) Three phase fault (a-b-c)

1 1 2π 1 2π
(6) id = c c aθ o id = cs c b(θ - o ) ids = c c c(θ + o ) s
x g xg 3 x g 3
For this fault type the qualitative specific relation is:
(7) i dca + i dcb + i dcc = 0

b) The line b to line c fault (b-c)

3 3
(8)
id =c0 a id =c b s θ i n id =c- c s θ i n
2 xg 2 xg
It is defined by the qualitative relations:
(9) i dca = 0 and i dcb + i dcc = 0
c) The line a to ground fault (a-g) type has the initial DC phase
components as given by:
(10)
i dca = (
2 KY + K 0 Y g )cos θ
θ
g

i dcb = ( K0Y - K Y g )cos

θ
g

i dcc = ( K0Y g - K Y g )c
os
and is defined by
(11) i dcb = i dcc
d) The line b to line c to ground fault (b-c-g) has the initial DC phase
currents:

i dca = (K Y gg - K 0 Y gg ) cos θ
(12) i dcb = - Y gg ( K 0 + K / 2 ) cos θ + 3 / 2 KY sin θ
i dcc = - Y gg ( K 0 + K / 2 ) cos θ - 3 / 2 KY sin θ
and is defined by the qualitative condition
(13) i dca ≠ i dcb ≠ i dcc
The notations in the above equations are: K=x/xg; x=xgxs/(xg+xs); Yg=1/
(2x+x0); x0=xg0xs0/(xg0+xs0); K 0=x0/xg0; Ygg=1/((x+2x0) where 0 means the
 BUL. INST. POLIT. IAŞI, TOMUL LIV (LVIII), FASC. 5

zero sequence.
The equations (7), (9), (11) and (13) allow for a differential diagnosis of
faults by qualitative relationship.
There are some ambiguous cases due to the initial voltage phase given by
θ. All these are presented in table 1.
To preserve the method accurate, collecting information about the
symmetrical current components is strongly necessary (indices +, - and 0
refers to the positive, negative and zero sequence respectively):
(14) i + = i − = i0 = 0
for the line to ground fault type,
(15) i+ + i − = 0
for line to line fault type and
(16) i − = 0 and i0 = 0
for three phase fault type.
Table 1
Ambiguous cases due to angle θ
Three phase Phase b to Phase a to Phase b to phase
fault ground ground c to ground
Phase a 1/x g 0 Y g(2K+K 0 ) Y gg (K-K0 )
θ=0 Phase b -1/2x g 0 Y g(K 0 -K) -Y gg (K 0+K/2)
Phase c -1/2x g 0 Y g(K 0 -K) -Y gg (K 0+K/2)
Phase a 0 0 0 0
θ=π/2 Phase b 3/2x g 3/2x g 0 3/2x g
Phase c -3/2x g -3/2x g 0 -3/2x g
Phase a -1/x g 0 -Yg (2K+K 0) Y gg (K 0-K)
θ=π Phase b 1/2x g 0 Y g(K-K0) Ygg (K 0 +K/2)
Phase c 1/2x g 0 Y g(K-K0) Ygg (K 0 +K/2)
Phase a 0 0 0 0
θ=3π/2 Phase b -3/2x g -3/2x g 0 -3/2x g
Phase c 3/2x g 3/2x g 0 3/2x g

Attaching the conditions expressed by eq. (10), (12) and (14) the above
mentioned method can accurately detect the fault type and provide necessary
initial information for fault location as well for the intelligent device to
control the circuit breaker with a view to clear the short-circuit.
This method conducted to a hybrid structured relay for fault type
detection as presented in fig. 2.
 three phase
-1 fault validation
Σ -1
i dca a-b-c fault
-1
Σ
idcb -1
-1 b-c fault

idcc -1
Σ
i-
-1 -1
a-g fault
i0
Σ
-1
b-c-g fault

Fig.2 The block diagram of a hybrid structure relay for fault type detection

4. Conclusion

The authors’ method in this paper is a very suitable for fast and accurate
detection of fault type. Fast because no calculation is necessary and accurate
due to the qualitative relationship between the initial values of DC
components of short-circuit currents the method is based-on. This second
feature allows for a comparative, high precision measurement technique.
A hybrid structured relay was established for the fault type detection
having the initial DC components of the phase short-circuit currents as well
as their negative and zero sequence components with a view to avoid some
ambiguous situations related to the moment of the fault occurrence.
A future work is to be dedicated to the analysis of other line
configuration as the double-circuit transmission line influencing the fault
caracteristics.