Abstract:
In order to deal with the shortcoming of the local optima of the bird swarm algorithm (BSA), an improved bird swarm algorithm (IBSA) is proposed in this paper by introducing the migration strategy and the mutation strategy. In the stage of flight, the migration strategy is adopted to raise the ability of bird swarm migration and the convergence speed of BSA. In the later stage of convergence of the BSA, the mutation strategy is utilized to optimize the local searching of the bird swarm and improve the searching ability of the proposed algorithm. Six typical test functions are selected to perform the optimization experiments which are implemented by particle swarm optimization (PSO), bat algorithm (BA), BSA, and IBSA, respectively. It is shown from the above experimental results that IBSA has the highest convergence precision and the fastest searching speed. Finally, IBSA is used to estimate the parameters of the fermentation kinetic models. Compared with Gauss-Newton, GA and MAEA, IBSA can obtain the smallest value of square sum of deviations squares. Hence, IBSA has the highest model fitting precision and the highest model fitting accuracy of the four algorithms, which also means that IBSA is a reliable, fast and accurate optimization tool for complex optimization problems such as the non-convex and the non-differentiable.