Page 491 - Decision Making Applications in Modern Power Systems

Pattern-recognition methods for decision-making Chapter | 17 451

example, “Is there a fault or not?” and “Is a stable swing or an unstable?”. In
regression problems, the pattern-recognition model predicts continuous out-
comes. For example, “What is the distance between the fault location and
relaying point?”. The same as classification, some data should be used to
train the algorithm, and then the learned algorithm can make a prediction. In
prediction problems, prediction error of input test data set is the criterion
that can be employed to evaluate how good the learned algorithm is. In pre-
diction of continues target, prediction error can be computed based on abso-
lute or relative error. Some known predictors that are used to estimate value
in pattern-recognition-based relays are ANN, SVMs, k-NN, etc.

17.3 Pattern recognition application on protection
of transmission line
Generally, protection of transmission line employs two subroutines including
fault detection and fault-type classification. The pattern-recognition-based
methods for fault detection and classification in transmission lines have been
comprehensively deliberated over the last three decades. This section pre-
sents a comprehensive survey of the pattern-recognition methods for
decision-making in protection of transmission lines. With the progression of
the present electrical power grid to a smart grid, the significance of design-
ing a smart relay capable of rapidly detecting and classifying the fault in a
transmission line is growing interest of several researchers. The develop-
ments in signal processing and feature extraction techniques, PMU, global
positioning system, and communications have facilitated the protection engi-
neers to develop smart and adaptive relays. Recently, few review articles
related to the transmission lines protection schemes [23 27] have been pub-
lished. The major challenges in the protection of transmission lines are the
detection of high-impedance fault (HIF), power swings, and fault during
power swings, wherein the conventional distance protection scheme fails.
Understanding the concept of these phenomena would assist the protection
engineers to cultivate an intelligent fault detection and classification techni-
ques based on fault pattern recognition to assist the conventional relaying
scheme.

17.3.1 Fault detection, classification, and location
Fault in transmission lines is an inevitable event either due to environmental
stress such as thunder storm, lightning, fog, snow fall, dust contamination, or
electrical stress, for instance, internal or partial discharge in insulators caus-
ing its failure. The status of the transmission line, whether it is healthy or
faulty, is usually predicted by distinctive characteristic variation in the asso-
ciated voltage and current waveforms during faulty or healthy conditions.
Therefore the task of protective relaying can be treated as pattern recognition