13-08-2012, 12:37 PM
Application of RBF neural network to fault classification and location in transmission lines
[attachment=31247]
Abstract:
The application of radial basis function (RBF) neural networks for fault classification
and location in transmission lines is presented. Instantaneous current/voltage samples have been
used as inputs to artificial neural networks (ANNs). Whereas, for fault classification, prefault and
postfault samples of only the three-phase currents are sufficient, for fault location, postfault
samples of both currents and voltages of the three phases are necessary. To validate the proposed
approach simulation studies have been carried out on two simulated power-system models: one in
which the transmission line is fed from one end and another, in which the transmission line is fed
from both ends. The models are subjected to different types of faults at different operating
conditions for variations in fault location, fault inception angle and fault point resistance. The
results of the simulation studies which are presented confirm the feasibility of the proposed
approach.
Introduction
Atransmission line is an important component of the
electric power system and its protection is necessary for
ensuring system stability and to minimise damage to
equipment due to short circuits that may occur on
transmission lines. Transmission-line relaying involves three
major tasks, namely detection, classification and location of
transmission line faults. Fast detection of a transmissionline
fault enables quick isolation of the faulty line from
service and, hence, protecting it from the harmful effects of
the fault. Classification of faults means identification of the
type of fault, and this information is required for fault
location and assessing the extent of repair work to be
carried out. Accurate fault location is necessary for
facilitating quick repair and restoration of the faulty line,
to improve reliability and availability of the power supply.
RBF neural network
The radial basis function network (RBFN) has a feedforward
structure consisting of three layers, an input layer, a
nonlinear hidden layer and a linear output layer, as shown
in Fig. 1. The hidden nodes are the radial basis function
units and the output nodes are simple summations. The
number of input, output and hidden nodes are nl, no and nh,
respectively. This particular architecture of RBFN has
proved to directly improve training and performance of the
network [9]. Any of the functions, namely spline, multiquadratic
and Gaussian function, may be used as a transfer
function for the hidden neurons. The Gaussian RBF, which
is the most widely used one, has been considered for the
proposed fault classification and location applications.
Fault classifier
The radial-basis-function ANN-based scheme for classification
of transmission line faults is shown in Fig. 4 [17]. The
fault classifier consists of two separate ANNs: one for
faults involving earths and another for faults not involving
earth. Hence, the prerequisite of the proposed scheme
is that the fault should be detected and, also, it should
be known whether the fault involves earth or not. In Fig. 4,
F, D and G are the outputs of a fault detector representing
the presence of fault, the direction of fault and
the involvement of ground in the fault. On the basis of
these data the appropriate ANN is selected for fault
classification, as indicated in Fig. 4. The ANN outputs have
been termed as A, B, C and G, which represent the three
phases and ground. Any one of the outputs A, B, C
approaching 1 indicates a fault in that phase.
Conclusion
Methodologies for classification and location of transmission-
line faults based on RBF neural networks have been
presented, while, for fault classification prefault and
postfault samples of three-phase currents are sufficient, for
fault location postfault samples of three-phase voltages and
currents are necessary. The use of separate ANNs, for faults
involving earth and not involving earth, has proved to be a
convenient way of classifying faults. For fault location,
two ANNs have been used for each type of fault to
obtain accurate estimates. Simulation studies carried out
considering wide variations in fault location, fault inception
angle, fault resistance and prefault load, show that the
proposed methods are suitable for classification and
location of transmission-line faults including the highimpedance
ones.
[attachment=31247]
Abstract:
The application of radial basis function (RBF) neural networks for fault classification
and location in transmission lines is presented. Instantaneous current/voltage samples have been
used as inputs to artificial neural networks (ANNs). Whereas, for fault classification, prefault and
postfault samples of only the three-phase currents are sufficient, for fault location, postfault
samples of both currents and voltages of the three phases are necessary. To validate the proposed
approach simulation studies have been carried out on two simulated power-system models: one in
which the transmission line is fed from one end and another, in which the transmission line is fed
from both ends. The models are subjected to different types of faults at different operating
conditions for variations in fault location, fault inception angle and fault point resistance. The
results of the simulation studies which are presented confirm the feasibility of the proposed
approach.
Introduction
Atransmission line is an important component of the
electric power system and its protection is necessary for
ensuring system stability and to minimise damage to
equipment due to short circuits that may occur on
transmission lines. Transmission-line relaying involves three
major tasks, namely detection, classification and location of
transmission line faults. Fast detection of a transmissionline
fault enables quick isolation of the faulty line from
service and, hence, protecting it from the harmful effects of
the fault. Classification of faults means identification of the
type of fault, and this information is required for fault
location and assessing the extent of repair work to be
carried out. Accurate fault location is necessary for
facilitating quick repair and restoration of the faulty line,
to improve reliability and availability of the power supply.
RBF neural network
The radial basis function network (RBFN) has a feedforward
structure consisting of three layers, an input layer, a
nonlinear hidden layer and a linear output layer, as shown
in Fig. 1. The hidden nodes are the radial basis function
units and the output nodes are simple summations. The
number of input, output and hidden nodes are nl, no and nh,
respectively. This particular architecture of RBFN has
proved to directly improve training and performance of the
network [9]. Any of the functions, namely spline, multiquadratic
and Gaussian function, may be used as a transfer
function for the hidden neurons. The Gaussian RBF, which
is the most widely used one, has been considered for the
proposed fault classification and location applications.
Fault classifier
The radial-basis-function ANN-based scheme for classification
of transmission line faults is shown in Fig. 4 [17]. The
fault classifier consists of two separate ANNs: one for
faults involving earths and another for faults not involving
earth. Hence, the prerequisite of the proposed scheme
is that the fault should be detected and, also, it should
be known whether the fault involves earth or not. In Fig. 4,
F, D and G are the outputs of a fault detector representing
the presence of fault, the direction of fault and
the involvement of ground in the fault. On the basis of
these data the appropriate ANN is selected for fault
classification, as indicated in Fig. 4. The ANN outputs have
been termed as A, B, C and G, which represent the three
phases and ground. Any one of the outputs A, B, C
approaching 1 indicates a fault in that phase.
Conclusion
Methodologies for classification and location of transmission-
line faults based on RBF neural networks have been
presented, while, for fault classification prefault and
postfault samples of three-phase currents are sufficient, for
fault location postfault samples of three-phase voltages and
currents are necessary. The use of separate ANNs, for faults
involving earth and not involving earth, has proved to be a
convenient way of classifying faults. For fault location,
two ANNs have been used for each type of fault to
obtain accurate estimates. Simulation studies carried out
considering wide variations in fault location, fault inception
angle, fault resistance and prefault load, show that the
proposed methods are suitable for classification and
location of transmission-line faults including the highimpedance
ones.