Prévision de l'envasement des barrages par la méthode des réseaux de Neurones (cas du bassin hydrographique Algérois Hodna Soummam.

Abstract

This work introduces a concept of neural network to estimate suspended sediment in catchments witch a hydrometric data are absents for forecasting a silting of dams. The adopted methodology consists to estimate and to predict a suspended sediment fort the best dealt with of the inheritance water –Ground. Two methods are used, the first one is parametric, linear regression and PCA, the second one is not parametric: a method of neural network, this last one is used to predict the specify erosion on the watershed of Algerois Hodna Soummam, supporting not only the hydrometric and pluviometric data, but also on the hydromorphometric characteristic and vegetable cover of catchments in the site of dams and small dams of the study space.