Improvements in the clustering validity indexes of the load profiling methodology

Author

Panapakidis, Ioannis P. ; Dagoumas, Athanasios S. ; Alexiadis, Minas C. ; Papagiannis, Grigoris K.

Author_Institution

Dept. of Electr. & Comput. Eng., Aristotle Univ. of Thessaloniki, Thessaloniki, Greece

fYear

2013

fDate

27-31 May 2013

Firstpage

1

Lastpage

8

Abstract

In the recent years the utilization of the load profiling tool for tracking the demand patterns is gathering momentum. There is variety of different clustering algorithms for the formation of daily load curve clusters. Their effectiveness is tested by a set of validity indexes or adequacy measures. This paper examines the behavior of all the adequacy measures that have been proposed in the related literature. We propose an alternative form of the measures that involves a weighting factor that refers to the variance of each element of the vector that represents the demand pattern. This fact increases the accuracy of the dissimilarity measures within and among the clusters. The data sample refers to the daily load curves of an existing high voltage industry within the Greek region and the period of study is the years 2003-2011. This vast amount of data is sufficient for assessing the load profiling methodology.

Keywords

electricity supply industry; pattern clustering; power engineering computing; unsupervised learning; Greek region; adequacy measures; clustering algorithms; clustering validity indexes; daily load curve cluster formation; demand pattern tracking; dissimilarity measures; high voltage industry; load profiling methodology; unsupervised machine learning; weighting factor; Artificial neural networks; Silicon; Clustering validation; Competative learning neural networks; Load profiles; Unsupervised machine learning;

fLanguage

English

Publisher

ieee

Conference_Titel

European Energy Market (EEM), 2013 10th International Conference on the

Conference_Location

Stockholm

Type

conf

DOI

10.1109/EEM.2013.6607330

Filename

6607330