Aboalhasan Fathbabadi; Mahnaz Kohneshin; Ali Heshmatpour; Masome Farasati
Abstract
During last decades hydrological models were extesively used in rainfall-runoff modeling. These models contain some constant parameters that must be optimized through appropraite mthods. In addation to model structur, the efficieny of hydrological models depend on these optimized parameters. In this ...
Read More
During last decades hydrological models were extesively used in rainfall-runoff modeling. These models contain some constant parameters that must be optimized through appropraite mthods. In addation to model structur, the efficieny of hydrological models depend on these optimized parameters. In this study, the efficiency of three automatic optimizing algorithms including Dynamically Dimensioned Search (DDS), Shuffled Complex Evolution and Genetic algorithms in calibration lumped hydrological model HyMod in Ghorchay Ramian Catchment were investigated. For these mehods, convergence speed and variability of final optimized values were investigated. Results showed that Genetic algorithm converged faster than two other methods. Following, DDS algorithm converged faster than Shuffled Complex Evolution algorithm. Shuffled Complex Evolution and Genetic algorithms took shorter and longer time per each epock, respectively. Highest and the least variability of final results were obtained for DDS and Shuffled Complex Evolution algorithms, respectively. With respect to final results variability, Shuffled Complex Evolution algorithm was more satable and had better performance than other methods. Using analyse variance and comper means in Shuffled Complex Evolution algorithms for complexes less than 12, the model performance was increased as the number of complexe increased. As alpha value increased, the model performance decreased and model had the best performance at the value of 0.58. Conversely, model performance was increase as beta values increasd and the best perfromnce was obtained for beta equal to 1. For Genetic algorithm, the best performance was obtained when the value of values crossover, mutation and chromosome number was equal to 0.2 and 0.3 and 16, respectively.
