Estimating the daily inflow of Sefidroud dam using meta-heuristic algorithms combined with fuzzy neural inference system

Estimating the water inflows to water resources systems is one of the essential measures for awareness of planning and allocating optimal water resources in different sectors of consumption. In this study, a combination of Meta Heuristic algorithms including the Water Cycle Algorithms (WCA), Grey Wolf Optimizer (GWO), Whale Optimization Algorithm (WOA), Neural Network Algorithm (NNA) and Grasshopper Optimization Algorithm (GOA) for Neural-Fuzzy System training and update parameters was used. Finally, the best models were developed to predict the daily inflow of Sefidrood reservoir dam. This method does not suffer from the problems of training gradient-based algorithms. Four features including dam lake area, reservoir volume and reservoir level of the dam in the 7 lags and inlet flow in the 1 lag were selected according to the AutocVarious statistical indicators were used to evaluate the performance of utilized models. In the test stage, ANFIS-WCA model presents the lowest values of SI, MAE and NRMSE equal to 0.0736, 0.05048 and 0.0736, respectively, and the maximum value of R^2equals to 0.9840, which indicates its superiority over other models. Based on GPI index, ANFIS-WCA model was selected as the best model and then ANFIS-NNA, ANFIS-GOA and ANFIS-WOA models were ranked, while the worst accuracy was obtained through the ANFIS-GOA model. The high accuracy of the ANFIS-WCA model compared to other hybrid models indicates the performance of the WCA for escaping local optimization in combination with the ANFIS model which has enabled this algorithm as a powerful tool for estimating the inflow of Sefidrood dam.