A novel bad-experience-enhanced Particle Swarm Optimization for metabolic model parameter identification

Particle Swarm Optimization (PSO) is a popular and bionic algorithm for optimization problems based on the social behavior associated with bird flocking. However, the standard PSO algorithm can easily get trapped in the local optima when solving complex multimodal problems, a novel bad experience enhanced particle swarm optimization was proposed in this study to improve the performance of PSO. The proposed approach has modified the velocity updating section with adding bad experience component and devised the position updating section with integrating evolution strategies. The proposed algorithm has comprehensively been compared with classical PSO and other famous bionic algorithms through evaluation on five famous multimodal benchmark optimization functions. Simulation results show that it can effectively enhance the searching efficiency and greatly improve the searching quality. Moreover, the proposed approach showed superior performance in metabolic model parameter identification during closed-loop glycemic control for Type 1 diabetes mellitus. Hence, it could be a valuable asset for metabolic model parameter identification.