Training sample dimensions impact on artificial neural network optimal structure

Author

Manusov, V.Z. ; Makarov, I.S. ; Dmitriev, S.A. ; Eroshenko, S.A.

Author_Institution

Dept. of Power Supply of the Enterprises, Novosibirsk State Tech. Univ., Novosibirsk, Russia

fYear

2013

fDate

5-8 May 2013

Firstpage

156

Lastpage

159

Abstract

The paper addresses the problem of electric load forecasting, using artificial neural networks mathematical apparatus, subject to error minimization on the long forecasting interval. Balanced artificial neural network architecture gives the possibility to maintain small deviation between forecasted and real values simultaneously with constrained squared error variation maintenance. Proposed methodology was verified using real data.

Keywords

learning (artificial intelligence); load forecasting; neural net architecture; power engineering computing; artificial neural network mathematical apparatus; artificial neural network optimal structure; balanced artificial neural network architecture; constrained squared error variation maintenance; electric load forecasting problem; error minimization; sample dimension impact training; Artificial neural networks; Biological neural networks; Energy consumption; Forecasting; Training; Vectors; artificial neural network; forecasting; neural network training;

fLanguage

English

Publisher

ieee

Conference_Titel

Environment and Electrical Engineering (EEEIC), 2013 12th International Conference on

Conference_Location

Wroclaw

Print_ISBN

978-1-4673-3060-2

Type

conf

DOI

10.1109/EEEIC.2013.6549608

Filename

6549608