Estimación de caudales medios en el área hidrográfica del caribe con información escasa, utilizando redes neuronales artificiales.

Abstract

In the following degree work, the calculation of average flows is carried out from artificial neural networks, taking into account a series of historical values provided by Ideam. These are obtained from a database requested from the Institute of Hydrology and Meteorology and Environmental Studies. After selecting the variables to be used, (precipitation, temperature and dates) these are historically organized from 1990 to 2015. Next, a set of representative input and output data is taken, which will later be introduced to the nntool application. from the matlab toolbox. There, the training and learning process of the rna takes place, taking into account the descending gradient mathematical model, the backpropagation algorithm with supervised learning. After processing the selector values, a series of conclusions and recommendations are reached in order to identify the possible arguments that could affect some of the results obtained.