Estimativa de escoamento em captação urbana utilizando modelos de rede neural artificial
Open Access
- 25 August 2020
- journal article
- research article
- Published by Laikos Servicos Ltda in Sistemas & Gestão
- Vol. 15 (2), 170-180
- https://doi.org/10.20985/1980-5160.2020.v15n2.1657
Abstract
Muitos tipos de modelos físicos foram desenvolvidos para estimativa de escoamento superficial com resultados bem sucedidos. Entretanto, a estimativa precisa do escoamento continua sendo um problema desafiador devido à falta de dados de campo e à complexidade de seu processo hidrológico. Neste artigo, um método de aprendizagem de máquina para a estimativa de escoamento é apresentado como uma abordagem alternativa ao modelo físico. Vários tipos de variáveis de entrada e arquiteturas de Rede Neural Artificial (RNA) foram examinados neste estudo. Os resultados mostraram que uma rede de duas camadas com a função de ativação de tansig e o algoritmo de aprendizagem Levenberg-Marquardt tinha o melhor desempenho. Para esta arquitetura, o vetor de entrada mais eficaz consiste em um perímetro de captação, comprimento do canal, inclinação, coeficiente de escoamento e intensidade da chuva. No entanto, os resultados da análise multivariada de variância indicaram o efeito de interação significativo dos dados de entrada e da arquitetura RNA. Assim, para criar um modelo RNA adequado para a estimativa de escoamento superficial, é necessária uma determinação sistemática do vetor de entrada. Many types of physical models have been developed for runoff estimation with successful results. However, accurate runoff estimation remains a challenging problem owing to the lack of field data and the complexity of its hydrological process. In this paper, a machine learning method for runoff estimation is presented as an alternative approach to the physical model. Various types of input variables and Artificial Neural Network (ANN) architectures were examined in this study. Results showed that a two-layer network with the tansig activation function and the Levenberg–Marquardt learning algorithm had the best performance. For this architecture, the most effective input vector consists of a catchment perimeter, canal length, slope, runoff coefficient, and rainfall intensity. However, results of multivariate analysis of variance indicated the significant interaction effect of input data and the ANN architecture. Thus, to create a suitable ANN model for runoff estimation, a systematic determination of the input vector is necessary.Keywords
This publication has 15 references indexed in Scilit:
- Modeling of urban growth dynamics and its impact on surface runoff characteristicsComputers, Environment and Urban Systems, 2013
- The economic benefits of rainwater-runoff reduction by urban green spaces: A case study in Beijing, ChinaJournal of Environmental Management, 2012
- Comparison of FFNN and ANFIS models for estimating groundwater levelEnvironmental Earth Sciences, 2010
- Comparing Sigmoid Transfer Functions for Neural Network Multistep Ahead Streamflow ForecastingJournal of Hydrologic Engineering, 2010
- Urban runoff modelling uncertainty: Comparison among Bayesian and pseudo-Bayesian methodsEnvironmental Modelling & Software, 2009
- Artificial neural network analysis for reliability prediction of regional runoff utilizationEnvironmental Monitoring and Assessment, 2009
- Development of analytical models for estimation of urban stormwater runoffJournal of Hydrology, 2007
- Application of neural approaches to one-step daily flow forecasting in Portuguese watershedsJournal of Hydrology, 2007
- Input determination for neural network models in water resources applications. Part 1—background and methodologyJournal of Hydrology, 2005
- Comparison of static-feedforward and dynamic-feedback neural networks for rainfall–runoff modelingJournal of Hydrology, 2004