Document Type : Research Paper

Authors

• Seyed Ahmad Hosseini ¹
• Ahmad Tabatabaei ²

¹ SCWMRI

² Iran_ SCWMRI

Abstract

Abstract
Correct estimation of the suspended sediment content of rivers plays an important role in studies of erosion and sediment, hydrology, and watershed management. Simulation of suspended sediment in hydrological systems with a high degree of uncertainty and yet, our understanding of the components and processes within them is faced with uncertainties, causing many applications of intelligent models, including artificial neural networks. However, the use of these intelligent models is also facing challenges. Determining the appropriate network structure requires optimizing the parameters used (such as the optimal number of neurons and layers, weight and bias, and the type of activation functions) so that their proper calibration by trial and error, while low efficiency, results in time-consuming. In the present study, in order to simulate the daily suspended sediment load in selected watersheds of Ardabil province, including the Sarab Gharasu watershed (Ghorchai and Hirchai Rivers), a multilayer perceptron artificial neural network was used. A particle swarm optimization (PSO) algorithm was used to train the neural network model, in addition to the conventional error propagation method, and to optimize the weight and bias values of neural network model neurons. Also, the self-organizing map clustering method was used to increase the generalization power of the models. The results of the present study showed that training of neural network models with the PSO algorithm in all selected rivers was more efficient than neural network models which use only the error propagation method. Since evolutionary algorithms (such as PSO algorithm) can provide suitable solutions for the optimization of neural network parameters, their application in training neural network models can be a good solution to improve the efficiency of smart models in simulating suspended sediment of rivers and using its results in the progress of watershed structures and water resources operations.

Keywords

• Artificial Neural Network
• Clustering
• Particle Swarm Optimization Algorithm
• Self-Organizing Map
• Suspended Sediment