Parameter identification for a water quality model using two hybrid swarm intelligence algorithms

Parameter identification or estimation is important to model simulations. This paper firstly carried out a sensitivity analysis of a water quality model using the Monte Carlo method. Then, two hybrid swarm intelligence algorithms were proposed to identify the parameters of the model based on the artificial bee colony and quantum-behaved particle swarm algorithms. One hybrid strategy is to use sequential framework, and the other is to use parallel adaptive cooperative evolving. The results of sensitivity analysis reveal that the average velocity and area of the river section are well identified, and the longitudinal dispersion coefficient is difficult to identify. The velocity is the most sensitive, followed by the dispersion and area parameters. Furthermore, the posterior parameter distribution and the collaborative relationship between any two parameters can be gotten. To verify the effectiveness of the proposed hybrid algorithms, this paper compared performances of the artificial bee colony, quantum-behaved particle swarm, their sequential combinations, and parallel adaptive dual populations. The experimental results demonstrate that the parallel dual population method is more effective than the original algorithms, when the data has added noise.