Comparison Between Soft Computing Methods for Prediction of Sediment Load in Rivers: Maku Dam Case Study

Abstract

The most important problem threatening dams are the sediment inputs to the dam reservoir. Due to various problems, estimating the amount of sediments is a complicated process. So some methods have been created by researchers to overcome these problems. Among these methods, three methods, namely artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS), and genetic algorithm (GA), are used and evaluated in this study. They are used to predict the sediment load in the Maku dam reservoir, Maku City, Iran. Mazra_e station on Gizlarchay River is selected for this study. The data of temperature, discharge, and CM (three-section method of sediment sampling) are utilized as input parameters, which have been harvested from 12 consecutive years (2002–2013). Sediment data are used as output parameter. Input parameters in ANN and ANFIS have been normalized with two methods: first between − 1 and + 1 range and second between − 2 and + 2 range. Input parameters for GA were without normalization. Output was natural data for all three approaches. Internal percentage error (PE) is applied to evaluate the error of performances between approaches. Results revealed that “logsig” membership function (MF) with five neurons has the best performance in ANN approach. Second normalization method had better performance for ANN, while the first one had better results in ANFIS. Results for ANFIS indicated that “gaussmf” MF had the best performance. The number of 100 and 1200, respectively, for individual populations and generations produced better performance in GA approach. Finally, it is concluded that ANFIS with the average 0.968% PE had the least error and ANN with the average 5.63% PE was in the second position. Although GA with an average 10% PE had the third place, considering that it did not need any normalization at input stage, it can be said that it had superior advantage in comparison with the other two approaches.