Accueil > Science et Technologie > Revue des Energies Renouvelables > Volume 14 > Numéro 02

Using multilayered neural networks for determining global solar radiation upon tilted surface in Fianarantsoa Madagascar

Andriazafimahazo L.F.G. , Ramamonjisoa B.O.A. , Razafiarison I.A.J. , Zeghmati B.

The knowledge of the local solar radiation characteristics is indispensable in
the survey of any system exploiting solar energy in any location. The author is
particularly interested by the global solar radiance upon tilted surface per time unit to
help operators using solar energy in their work. The target is, among others, helping
solar drying operators that need while tuning drying system the knowledge of the global
solar radiation that could be received on inclined solar captors in implantation site. The
aim of this paper is to use neural network method to search for solar radiation upon a
tilted surface. Multilayered neural networks (MNN) trained by gradient back-propagation
are used to determine numeric values of monthly means and hourly variations of the
global solar radiation upon a titled surface per time unit. The numerical calculations are
made with the geographical and meteorological parameters (latitude, longitude and
clearness index) of the location of Fianarantsoa, Madagascar.