Hybrid particle swarm optimization with simulated annealing

While solving the optimization problems of complex functions, particle swarm optimization (PSO) would be easy to fall into trap in the local optimum. Besides that, it has slow convergence speed and poor accuracy during the late evolutionary period. So a SA-PSO algorithm would be proposed in this paper. Classically, the probability to accept bad solutions is high at the beginning. It allows the SA algorithm to escape from local minimum. As the result of that, the improved algorithm, combined SA with PSO, would be given in this paper. The given algorithm owned the abilities of both increasing the diversity of particle swarm and jumping out of the local optimum. In this paper, several classic unimodal/multimodal functions were used to simulate the SA-PSO algorithm. The results illustrated that SA-PSO had a stronger ability to avoid prematurity and get rid of local optimum. Compared with traditional PSO, the SA-PSO has improvement over effectiveness and accuracy to some extent. And it has competitive potential for solving other complicated optimization problems.