Automatic Classification of Voltage Disturbances using the Support Vector Machine Method

grizzlybearcroatianΤεχνίτη Νοημοσύνη και Ρομποτική

16 Οκτ 2013 (πριν από 3 χρόνια και 7 μήνες)

88 εμφανίσεις

Automatic Classification of Voltage Disturbances using the Support Vector
Machine Method

P.G.V. Axelberg, I.Y.H Gu and M.H.J. Bollen

Abstract
Analyzing voltage disturbances and other power quality related phenomena is an important step towards
automatic power system network diagnostics. Efficient methods for automatic classification of voltage
disturbances and other power quality related abnormalities are therefore highly in demand. This paper
proposes a novel method for classifying the underlying causes of voltage disturbances using soft-margin
Support Vector Machines (SVMs). Support vector machine classification is a statistic learning theory-
based method which offers significant advantages over the conventional methods such as artificial neural
networks. In the proposed method, a set of features are extracted from various domains (e.g. time domain,
frequency domain, and time-frequency domain) for classifying the underlying causes of 5 types of power
system disturbances. The proposed SVM classifier then uses (a) training data and testing data measured
from different power networks in two countries; (b) training data were synthetically generated and testing
data were from the measurements. Our experimental results have shown a high accuracy in classification
(>98%) for case (a), and a slightly lower rate for case (b). Details of the method and further performance
evaluations will be described in the full paper.


Accepted and selected for presentation in 19th International Conference on Electricity
Distribution (CIRED 2007), Vienna, Austria, 21-24 May 2007