Second International Conference on Advances In Civil, Structural and Environmental Engineering- ACSEE 2014
Author(s) : MAHYAR ABOUTALEBI, OMID BOZORGHADDAD
In this paper a novel tool, support vector machine (SVM) based on Non-dominated sorting genetic algorithm (NSGAII), is proposed for prediction of the monthly inflow stream in the hydropower reservoir system. The two objectives which are considered in NSGAII are minimizing the error of the prediction by SVM and minimizing the number of variables which are selected for SVM as the input variables. The statistical indicator which is considered for the evaluation of the error is root mean square error (RMSE) and the hydropower reservoir of Karoon-4 which is located in Iran is considered as the case study. In this optimization problem, the decision variables of NSGAII have two parts. The First part is the names of the input variables as predictors and the other part is the values of the SVM parameters. In order to create the data base of SVM, the input variables (monthly inflow and monthly precipitation) in the previous periods and monthly inflow of reservoir in the current period as the target variable are considered as the data base for SVM. Results showed that the SVM-NSGAII tool can achieved a wide alternative that provides the different selection input variables and different RMSE.