Predicción de caudales en la cuenca del río Chalpi Norte mediante la aplicación de redes neuronales

Abstract

The present degree work aims to estimate the daily flows that the micro-basin of Chalpi Norte flows downstream through the selection of the optimal model using artificial neural networks. For the construction of models, data series of 4 pluviometric stations and 1 hydrometric station provided by EPMAPS were used, from which precipitation and flow values were collected as input data. Each of the models were carried out through supervised learning in training and validation stages. 24 models were built combining different parameters such as: operating algorithms, output and activation functions, neural network architecture and variation of the input data to the models. The models developed include input using normalized precipitation and flow values for entry into the neural network, evaluating the approximation of the values measured by the hydrometric station and the values estimated by the models by calculating error coefficients. Finally concluding that there is an optimal model that approximates the estimates to the flow values measured in the micro-basin, since it presents values of the error coefficients very close to optimal values.