Investigation for Application of Adaptive Neuro-Fuzzy Inference System (ANFIS) in Babolroud Suspended sediment Load Estimation

Sediment load estimation is one of the most important issues in rivers & dam reservoirs management and generally in water projects. Various empirical equations show that proper analytical or empirical method is not suggested for correct estimation of suspended sediment, yet. In the present study, to assessment of closer estimation to actual data of transported sediment in Ghoran Talar station located in Babolroud River, the adaptive neuro-fuzzy inference system (ANFIS) technique is used as an Artificial Intelligence method. At first, various combinations of discharges based on time delay is considered as input parameters and the suspended sediment load is applied as output of the model. After Learning the network and assessment of the best structure according to type, number of membership function and related rules by use of MATLAB software, appropriate model is obtained based on statistical indices viz. mean square error, model efficiency and determination coefficient. As a result, one dimensional input according to Sugeno inference system with two triangular membership functions is introduced as an appropriate model and is compared with sediment rating curve values. Finally, the results showed that ANFIS method (MSE=0.08, EF=0.78, R2=0.72) has higher accuracy than sediment rating curve (MSE=0.16, EF=0.57, R2=0.73) and has better efficiency in suspended sediment load estimation.