Author(s): Bryant Eduardo Sandoval Acevedo; Alejandro Mendoza Resendiz And Eliseo Carrozosa

Linked Author(s):

Keywords: Hydrograph routing; Neural networks; Hydraulic 1D model; Prediction accuracy

Abstract: This work focuses on the development of models to perform the routing of hydrographs in rivers using neural networks. A comprehensive comparison was made between the results of these models and those obtained from simulations with a traditional hydraulic 1D model, evaluating not only prediction accuracy, but also the computing time required for each approach. Adjustments were made to the hyperparameters, such as the number of layers, neurons and regularization techniques, to improve the accuracy of the predictions made by the neural networks. The results showed that both MLP-based and LSTM-based models can capture complex patterns in the data, with the LSTM model being superior in terms of stability and generalization capability for different hydrological conditions of the input hydrographs.

DOI: https://doi.org/10.64697/978-90-835589-7-4_41WC-P1955-cd

Year: 2025