Simulation du comportement d'un refroidisseur et du programme optimal de fonctionnement grâce à un réseau neuronal.

The simulation of chiller-running behaviour and operation-sequence optimal using neural network.


The main function of the chiller is to supply chilled water to the air conditioner to lower the temperature and dehumidify so the product yield is maintained. However, keeping a chiller running is very power consuming. Therefore, it is a worth trying innovation concept to search for a chiller system simulation and forecasting method on how to lower the running cost and the reliability on professionals, to provide the system the best running reference. This paper uses principal component analysis and covariance matrix analysis to screen the function parameters that affect the running of the chiller the most then establishes a simple and clear back-propagation neural network framework. Through training and comparison testing of the experimental data, it is proved that effective and accurate local model which is suitable for single chiller operation can be established under relatively low cost. In addition, the neural network model is applied to the simulations of optimal chiller sequence and optimal chiller loading distribution of the chiller unit while carrying out optimization selection and adjustment to the system and the unit, respectively. The authors conduct tests on the system's requirement cases and try different optimization techniques which show quite remarkable efficiency that improves the achievement. The results showed that by using neural network model to simulate the chiller unit, whether it is a single unit or an integrated system, not only the creditability is satisfactory, it also provides convenient, fast and effective optimization approach.

Documents disponibles

Format PDF

Pages : pp. 629-634


  • Prix public

    20 €

  • Prix membre*

    15 €

* meilleur tarif applicable selon le type d'adhésion (voir le détail des avantages des adhésions individuelles et collectives)


  • Titre original : The simulation of chiller-running behaviour and operation-sequence optimal using neural network.
  • Identifiant de la fiche : 2010-0893
  • Langues : Anglais
  • Source : ACRA-2009. The proceedings of the 4th Asian conference on refrigeration and air conditioning: May 20-22, 2009, Taipei, R.O.C.
  • Date d'édition : 20/05/2009


Voir d'autres communications du même compte rendu (102)
Voir le compte rendu de la conférence