Publication detail

Model Predictive Controller Based on Neural Network Used for Multi-Dimensional Control

P. Nepevný, P. Pivoňka

Original Title

Model Predictive Controller Based on Neural Network Used for Multi-Dimensional Control

English Title

Model Predictive Controller Based on Neural Network Used for Multi-Dimensional Control

Type

conference paper

Language

en

Original Abstract

This paper presents a solution of multi-dimensional Model Predictive Control (MPC) based on feed-forward Neural Network (NN) model. Autoregressive NN model with back-propagation learning algorithm is used for system output prediction. It is able to observe system changes and adapt itself, therefore adaptive MPC controller is obtained. MPC is a kind of optimal controller, because a control action is always optimal according to the given criterion. There is shown, how to create multi-dimensional predictive controller. Possibilities of multi-dimensional MPC were tested on laboratory physical model – hot-air tunnel. Two quantities of hot-air tunnel were controlled – the air flow and the temperature. The algorithm was implemented in MATLAB-Simulink and tested on a physical model. Communication between PC and hot-air tunnel was provided by PLC (connected via Ethernet).

English abstract

This paper presents a solution of multi-dimensional Model Predictive Control (MPC) based on feed-forward Neural Network (NN) model. Autoregressive NN model with back-propagation learning algorithm is used for system output prediction. It is able to observe system changes and adapt itself, therefore adaptive MPC controller is obtained. MPC is a kind of optimal controller, because a control action is always optimal according to the given criterion. There is shown, how to create multi-dimensional predictive controller. Possibilities of multi-dimensional MPC were tested on laboratory physical model – hot-air tunnel. Two quantities of hot-air tunnel were controlled – the air flow and the temperature. The algorithm was implemented in MATLAB-Simulink and tested on a physical model. Communication between PC and hot-air tunnel was provided by PLC (connected via Ethernet).

Keywords

Predictive Controllers, Neural Networks for Identification, Multi-Dimensional Control

RIV year

2006

Released

02.10.2006

Publisher

Rektor der Hochschule Zittau/Gorlitz

Location

Zittau

ISBN

3-9808089-8-X

Book

East West Fuzzy Colloquium

Pages from

69

Pages to

74

Pages count

6

BibTex


@inproceedings{BUT19681,
  author="Petr {Nepevný} and Petr {Pivoňka}",
  title="Model Predictive Controller Based on Neural Network Used for Multi-Dimensional Control",
  annote="This paper presents a solution of multi-dimensional Model Predictive Control (MPC) based on feed-forward Neural Network (NN) model. Autoregressive NN model with back-propagation learning algorithm is used for system output prediction. It is able to observe system changes and adapt itself, therefore adaptive MPC controller is obtained. MPC is a kind of optimal controller, because a control action is always optimal according to the given criterion. There is shown, how to create multi-dimensional predictive controller. Possibilities of multi-dimensional MPC were tested on laboratory physical model – hot-air tunnel. Two quantities of hot-air tunnel were controlled – the air flow and the temperature. The algorithm was implemented in MATLAB-Simulink and tested on a physical model. Communication between PC and hot-air tunnel was provided by PLC (connected via Ethernet).",
  address="Rektor der Hochschule Zittau/Gorlitz",
  booktitle="East West Fuzzy Colloquium",
  chapter="19681",
  institution="Rektor der Hochschule Zittau/Gorlitz",
  year="2006",
  month="october",
  pages="69",
  publisher="Rektor der Hochschule Zittau/Gorlitz",
  type="conference paper"
}