USING PARTICLE SWARM OPTIMIZATION ALGORITHM FOR PARAMETER ESTIMATION IN HYDROLOGICAL MODELLING

Particle swarm optimization (PSO) is a global optimization technique from the group of swarm intelligence, which is based on iterative work with a population. It is used for finding an optimum of given objective function. The PSO was applied in many real life problems and it has also a wide use in hydrology. In general, optimization methods in hydrological modelling are using for calibration of models, estimation of rainfall-runoff relationships, meteorological forecasts, or runoff predictions. The optimization serves to find the best set of parameters of the model. The main goal of this paper is to present a new approach in searching of the best set of parameters of the Bilan rainfall-runoff model. The particle swarm optimization technique was used and we compared the results of the PSO algorithms with linearly decreasing inertia weight and PSO with constriction factor. The model was tested on 30 monitored catchments with a length of observation of 54 years. The input data to the model are precipitation, evaporation, and air temperature, and the model simulates the runoff from the basin. The optimization ability of each method was estimated by the Nash-Sutcliffe coefficient, and other accuracy criteria like mean squared error, or mean absolute error was calculated. The results show that the PSO algorithm modified by the parameter of linearly decreasing inertia weight gives better estimation of parameters than the algorithm with constriction factor. The findings of this paper increase the usage of the PSO algorithm in real-life optimization problems and they extend the program settings of the Bilan rainfall-runoff model.