Comparison of three multi-objective optimization algorithms for hydrological model

Abstract

In our research of this article, the efficiency of Multi-objective Particle Swarm Optimization (MOPSO), Non-dominated Sorting Genetic Algorithm (NSGA-II), and Multi-objective Shuffled Complex Evolution Metropolis (MOSCEM-UA) algorithms were compared by implementing the Hydrological Model (HYMOD) and the related observed daily precipitation, evaporation and runoff data. High flow Nash-Sutcliffe efficiency and Low flow Nash-Sutcliffe efficiency were used to optimize the model parameters as two criterions; the time consumption, the dominating rate and, the quality of Pareto set (distance, distribution, and extent) were used to analyze the performance of the three algorithms. Compared with NSGA-II and MOSCEM-UA, the MOPSO algorithm performed most efficiently to complete each trail. The non-dominant solutions derived from MOPSO algorithm were seldom dominated by those from the other two algorithms, while a high proportion of solutions drawn from NSGA-II are dominated by the other two algorithms. When we come to the three optimization goal of multi-objective optimization, there is a complication. The shortest distance of the resulting non-dominated set to the Pareto-optimal front was from the NSGA-II algorithm the most uniform distribution of the solutions was derived from the MOSCEM-UA algorithm; and the maximal extent of the obtained non-dominated front was stemmed from the MOPSO algorithm. The results demonstrated that all three algorithms were able to find a good approximation of the Pareto set of solutions, but differed in the rate of convergence to the optimal solutions. © 2012 Springer-Verlag.