Page 493 - Decision Making Applications in Modern Power Systems

Pattern-recognition methods for decision-making Chapter | 17 453

trial and error procedures to find the number of layers, neurons, and activa-
tion functions, which makes the overall design process a tedious and com-
plex task.
SVM is a statistical technique used for the purpose of computational
learning in Refs. [37 40], which overcomes the drawback of ANN by giving
a global solution rather than local minima. Finite impulse response filter
with SVMs has been used for fault detection, classification and support vec-
tor regression (SVRs) are utilized to locate the faults in Ref. [37]. EMD is
used for decomposing voltage signal into IMFs, and then Hilbert Huang
Transform is used to extract the features from IMFs. Finally, SVM is pro-
posed for fault classification in Ref. [38]. A new combination of hyperbolic
ST and SVMs/SVRs is applied for fault detection, classification, and location
in Ref. [10]. DWT is employed to yield the change in energy and standard
deviation (SD) of fault current and voltage signal for faulty phase identifica-
tion and fault location in Ref. [39].In Ref. [40] wavelet entropy criterion is
applied to detail wavelet coefficients of voltage and current signals to reduce
the size of the feature vector and fed to SVM for fault classification and
location estimation.
DWT-based algorithms in pattern recognition techniques are also popular
in recent times. An ANN classifies the fault based on the voltage and current
waveforms in the time domain using real oscillographic data in Ref. [41].
The entropies of the wavelet decomposed voltage signals have been fed to
the neural networks for fault classification and location in Ref. [42].An
exhaustive survey on the application of wavelet transforms-based fault iden-
tification in high voltage (HV) transmission line has been reported in Ref.
[43], which discusses different schemes utilizing DWT with different pattern
recognition schemes, useful for researchers working in this field. Another
scheme in Ref. [44] uses the wavelet packet transform, a generalization of
DWT wherein discrete time signal passes through a series of filters, for fea-
ture extraction and fed to SVM for fault classification and location. DWT is
used in conjunction with DT and k-NN classifiers as a means of the semisu-
pervised machine learning in Ref. [45].
FL provides an inference mechanism that enables approximate human
reasoning capabilities to be applied to knowledge-based systems such as pro-
tective relaying task. Fuzzy inference system (FIS) has been used to detect
the fault (both forward and reverse), locate and also identify the faulty phase
(s) in a multisection transmission system in Ref. [46]. Further, application of
FIS to detect the presence of a fault and its direction using phase angle of
positive sequence current only is proposed in Ref. [47].
DT-based methods have been employed for fault classification in a trans-
mission line using the odd harmonics up to 19th of the voltage and current
signals in Ref. [5]. Further, DWT-DT based fault classification scheme with
classification and regression tree analysis is proposed in Ref. [48]. ST and
PNN-based fault classification and section identification schemes have been