Recurrent fuzzy-neural network with fast learning algorithm for predictive control

Yancho Todorov, Margarita Terzyiska, Michail Petrov

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

7 Citations (Scopus)

Abstract

This paper presents a Takagi-Sugeno type recurrent fuzzy-neural network with a global feedback. To improve the predictions and to minimize the possible model oscillations, a hybrid learning procedure based on Gradient descent and the fast converging Gauss-Newton algorithms, is designed. The model performance is evaluated in prediction of two chaotic time series - Mackey-Glass and Rossler. The proposed recurrent fuzzy-neural network is coupled with analytical optimization approach in a Model Predictive Control scheme. The potentials of the obtained predictive controller are demonstrated by simulation experiments to control a nonlinear Continuous Stirred Tank Reactor.

Original languageEnglish
Title of host publicationArtificial Neural Networks and Machine Learning, ICANN 2013 -
Subtitle of host publication23rd International Conference on Artificial Neural Networks
PublisherSpringer
Pages459-466
ISBN (Electronic)978-3-642-40728-4
ISBN (Print)978-3-642-40727-7
DOIs
Publication statusPublished - 8 Oct 2013
MoE publication typeA4 Article in a conference publication
Event23rd International Conference on Artificial Neural Networks, ICANN 2013 - Sofia, Bulgaria
Duration: 10 Sept 201313 Sept 2013

Publication series

SeriesLecture Notes in Computer Science
Volume8131
ISSN0302-9743

Conference

Conference23rd International Conference on Artificial Neural Networks, ICANN 2013
Country/TerritoryBulgaria
CitySofia
Period10/09/1313/09/13

Keywords

  • Gauss-Newton method
  • Gradient descent
  • momentum learning
  • optimization
  • predictive control
  • recurrent fuzzy-neural networks
  • Takagi-Sugeno

Fingerprint

Dive into the research topics of 'Recurrent fuzzy-neural network with fast learning algorithm for predictive control'. Together they form a unique fingerprint.

Cite this