Congress Resources: Papers, posters and presentations

< Return to abstract list

Study of the calibration, the validation and the performances of the Formal Neurons Networks for hydrological modelling starting from the hydro-climatic data of Bandama Blanc (Côte d’Ivoire)

Author(s): Study of the calibration, the validation and the performances of the Formal Neurons Networks for hydrological modelling starting from the hydro-climatic data of Bandama Blanc (Côte d’Ivoire)
Congress: 2008
Author(s): 1. KOFFI Yao Blaise (Université de Cocody, Cote d’Ivoire, Laboratoire des Sciences et Techniques de l’Eau et de l’ Environnement)
1. KOFFI Yao Blaise (Université de Cocody, Cote d’Ivoire, Laboratoire des Sciences et Techniques de l’Eau et de l’Environnement) 2. AYRAL Pierre Alain (Ecole des Mines d’Alès, France, Laboratoire, Génie de l’Environnement Industriel et des Risques Industriels et Naturels) 3. BIEMI Jean (Université de Cocody, Cote d’Ivoire, Laboratoire des Sciences et Techniques de l’Eau et de l’ Environnement)

Keyword(s): Conceptual models, Formal Neurons Networks, Multi-layer Perceptrons, GR2M, Simulation, Forecast, Hydrology, Bandama Blanc, Ivory Coast.
AbstractThe present study relates to the modelling of the flows of Bandama Blanc at the limnimetric stations of Bada, Marabadiassa, Tortiya and Bou with the Formal Neurons Networks (FNN). It projects to provide more robust tools to the African hydrologists for the simulation and the forecast of the flows of the measured rivers when the quality and the quantity of the data are missing. To achieve this objective, two neuronal models (Multi-layer Perceptron Not Buckled Directed (PMCNB) and Multi-layer Perceptron Buckled and Directed (PMCBD)), involved with the algorithm of the retro propagation of the error, were built. The first model is used only in simulation and the second in simulation and forecast. Conceptual model GR2M, is also used to validate the results obtained with the Formal Neurons Networks. The weather data, in particular the temperatures and the rains used in this work result respectively from the data base of the basic Integrated Information system of Data (IDIS) and of the services of the Development company and Development Airport, Aeronautic and Meteorology (SODEXAM). With regard to the data of flows, they come from Direction of Human hydraulics and Management water of the Ivory Coast. The results obtained are strong satisfactory and definitely higher than those obtained with conceptual model GR2M. The Formal Neurons Networks manage to explain more than 70% of the variation of the flows, with coefficients of correlation of Pearson which exceed 0,80 between the simulated flows and the measured flows on the one hand and between the predicted flows and the flow measured on the other hand. However, these models manage with difficulty to simulate and make the forecast of the extreme flows (low water levels and raw) because of the reduced number of data we have.
© 2011 IWRA - International Water Resources Association office@iwra.org - http://www.iwra.org - Admin