Sistemas inteligentes para el ajuste de modelos hidrológicosaplicación al río Paraná

Abstract

The aim is the implementation of intelligent systems to adjust hydrological models comparing time series and neural networks which allow learning and setting parameters for models that make optimal predictions of the Paraná river heights in flood periods. The interest lies in its implementation in the province of Corrientes, Argentina, hit by floods causing losses in regional economy. We performed a time-series analysis to discover the variables and factors that influence the hydrometric height in flood periods in the town of Corrientes. Subsequently we present a short-term prediction for flood periods, which predicts the hydrometric heights three days in advance, using neural networks with a modified penalty function. Then we obtain a medium-term forecast for flood periods, seven days in advance, using neural networks with different architectures.