Use of a new trnsys type using neural network for non linear models creation

Laetitia Adelard (1) , Franck Lucas (1) , Philippe Lauret (1) , Eric Fock (1)

1 PIMENT - Physique et Ingénierie Mathématique pour l'Énergie, l'environnemeNt et le bâtimenT

Model determination must sometimes be rapidly made because of missing data in a building simulation. This paper will deal about the creation of a new trnsys type using neural network method to create new non linear models. The new type 111 integrates both the models creation process, and the use of this model immediately in simulations. This new type will be used in Hvac building simulation in trnsys environment but it can also be used in a new weather data generator. Illustration of such a type is made in HVAC simulation, for the determination of the cooling power, or for missing data in meteorological data files like long wave radiation estimation.

Mots clés: HVAC simulations; Non linear stochastic models; neural network; Trnsys platform; HVAC simulations.

Domaines: Thermique [physics.class-ph]; Thermique [physics.class-ph]

Laetitia Adelard, Franck Lucas, Philippe Lauret, Eric Fock. Use of a new trnsys type using neural network for non linear models creation. 15-18 September 2001, Sep 2001, Naples, Italy. pp.1-9. ⟨hal-01068560⟩