22-09-2012, 02:32 PM
An ANN Based Approach to Improve the Speed of a Differential Equation Based Distance Relaying Algorithm
1An ANN Based Approach.pdf (Size: 760.54 KB / Downloads: 20)
Abstract
This paper presents an artificial neural network
(ANN) based approach to improve the speed of a differential
equation based distance relaying algorithm. As the differential
equation used for the transmission line protection is valid only at
low frequencies the distance relaying algorithm requires a
lowpass filter removing frequency components higher than those
for relaying. However, the lowpass filter causes the time delay of
the components for relaying. Thus, the calculated resistances
and reactances do not converge directly to the fault distance
even after data window occupies post fault data. Faults with the
same fault inception angle have similar shapes of impedance loci.
If an ANN is trained with the shape of various impedance loci
for fault distances and fault inception angles, it can predict the
fault distance with some values of calculated resistances and
reactances before they converge to the fault distance. Therefore,
the ANN can improve the speed of the distance relaying
algorithm without affecting its accuracy.
INTRODUCTION
In a differential equation based distance relaying algorithm
for transmission line protection, the fundamental assumption
is that the transmission line can be represented by a lumped
series impedance. In other words, it is assumed high
frequency components are filtered out from the original
faulted waveforms[l]. Thus, it requires a lowpass filter
rejecting frequency components higher than those for
relaying. However, the lowpass filter causes time delay of the
components for relaying. Due to the transient components as
well as the time delay caused by the Lowpass filter, the
impedances calculated by the relaying algorithm do not
converge directly to the fault distance even after data window
occupies post fault values.
The impedance loci under various conditions
The 345 (kV) transmission system configuration studied is
shown in Fig. 2 and its power frequency is 60 (Hz). A
distance relay protects the 120 (km) transmission line and its
trip zone covers 100 (km) from the P bus.
Fig. 3 shows typical impedance loci after a fault occurs,
where circles represent the trip zone. After a fault, resistance
and reactance loci calculated by the method described in
Section I1 progress from an initial point to the points
corresponding to the fault distances. Fig. 3a and Fig. 3b
represent impedance loci in case of 10 (deg) and 90 (deg)
faults, respectively, at some fault distances. Each impedance
loci converge to the different points corresponding to the fault
distances. However, the shapes of impedance loci for faults
with the same fault inception angle are similar though
convergent points are different. While a 10 (deg) fault has
monotonous impedance loci, a 90 (deg) fault has somewhat
complex impedance loci which have a vertex tuming its
direction. Fig. 3c indicates impedance loci for faults at the
same distance with different fault inception angles. Like a 90
(deg) fault, the impedance loci for 50 (deg) faults also have a
tuming point.
CONCLUSION
This paper presents an approach to improve the speed of a
differential equation based distance relaying algorithm using
ANNs. For impedance calculation, an integral approximation
method is used and then ANNs predict the fault distance with
some input pairs of calculated resistances and reactances.
Though a lowpass filter causes the delay of the components
for relaying, the ANNs can predict the fault distance only
with a few values of calculated impedances before they
converge to the impedance corresponding to the fault
distance. Three cases of sampling rate such as 24, 48 and 96
(s/c) were tested. Four, six, and nine input pairs of resistance
and reactance are necessary for ANNs to predict the fault
distance in sampling rates of 24, 48, and 96 (SIC),
respectively. It is clearly shown from the results of the case
studies that the proposed approach can improve the speed of
the conventional distance relaying algorithm without affecting
its accuracy.