چكيده به لاتين
One of the main challenges in hydrologic engineering is to determine the response of catchment to a specified precipitation. We need an appropriate rainfall-runoff model to optimize its parameters to find the responses of the catchment. The process of model calibration is done either manually or by using computer-based automatic procedures. Manual trial and error calibration is time consuming and depends on the modeler’s experience, skill, and knowledge of the model’s processes and dynamics. So automatic calibration procedure accepted in this study. Calibration based on single objective function cant considered all aspect of hydrograph, So multi-objective calibration chooses. In this research, the efficiency of Non-dominated Sorting Genetic Algorithm (NSGA-II), Strength Pareto Evolutionary Algorithm (SPEA-II) and Multi-objective Particle Swarm Optimization (MOPSO) were compared by implementing the ARNO rainfall-runoff model. The case study was accomplished on the Saroogh Chay river catchment, up stream of Shahid Kazemi (Bookan) until Safakhaneh hydrometry station. The ARNO model inputs are represented by rainfall and potential evapotranspiration data series. Hypsometric curve was derived from GIS software to calculate the average temperature in catchment and the regression model was proposed to find relationship of altitude and temperature. ARNO rainfall-runoff model (were written in FORTRAN code) linked to evolutionary algorithm (which were written in MATLAB). In this study, two objective functions such as: a root mean square of errors (RMSE) and LOGE considered and minimize of them were determined. Where one of the objective functions emphasizes on fitting the peak flow and the other stresses on fitting low flow values of the watershed responses considered. The multi-objective optimization of the ARNO Model was carried out and the effectiveness of multi-objective optimization approach for performance evaluation of rainfall-runoff models was investigated. Pareto front from three multi-objective optimization achived. Three point of Pareto front include the best of two amounts of objective functions and the solution of kalay-smorodinsky were selected. The result showed that NSGA-II algorithm has better performance than two other algorithms from the point of maximum distribution and number of nondominated solution. For three point in pareto front in each algorithm the simulation performance of the model was evaluated on the basis of Efficiency coefficient and coefficient of determination which were reached value between – during calibration period, – during validation period. The multi-objective optimization of the ARNO Model was carried out and the effectiveness of multi-objective optimization approach for performance evaluation of rainfall-runoff models was investigated.
