Publication: Effect of Signal Decomposition in Power Quality Disturbances Classification
Date
2020-02
Authors
Kavines a/l Murugesu
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The Revolution of the present modern era demands an unprecedented and good day-today
supply of power quality. Suppliers of power utilities and power trade services face
difficult problems in recognizing and separating out the Disturbance of Power Quality
(PQD). This thesis presents a novel approach to identify and classify the power quality
disturbances signal based on Singular Spectrum Analysis (SSA) and two different
methods of dimension reduction which is Kernel PCA (KPCA) and Principle
Component Analysis (PCA) with a recommended Classifier for Classification which is
the K-Nearest Neighbors (K-NN) . Total of 16 PQDs waveform is designed on
MATLAB using the mathematical model as defined by customary IEEE 1159 and
IEC61000 specification and parameters. SSA is a non-parametric technique, does not
require any supposition to generate the observed signal, and provides an effective way
to decompose and understand the PQ signal. In this thesis, my dataset consists of 16000
generated signals of all 16 types of PQD which is sated in the Power quality Names list,
which is then divided into 70% for training and 30% for testing sets for each PQDs. The
main objective of this thesis is to improve the accuracy of the K-NN classifier after
applying dimension reduction technique. Different tests were carried out by teaching
the classifier to analyze and compare the results. The performance outputs varied in
reduction of dimension between the classifier k-NN. The dimensionally decreased
classifier k-NN succeeds in classifying the Power Quality Disturbance with satisfying
performance in both training and testing sets.
Description
FYP 2 SEM 2 2019/2020
Keywords
singular spectrum analysis , power quality disturbances , machines learning