Abstract
The traditional serial simulated annealing algorithm has low efficiency, and is high depending on the setup parameters. The author presents a new hybrid parallel simulated annealing algorithm based on multi-thread parallel computing. Algorithm generates a set of random initial points, and then calculates a number of Markov chains in parallel. Each Markov chain has its own setup parameter. At every parallel point, the groups of parallel Markov chains’ solutions are used to set the current global optimization strategies which will overcome premature convergence and improve the ability of searching global optima. A local search algorithms-Powell algorithm is used when the parallel annealing algorithm is finished at the approximate optimal point to improve the algorithm accuracy, and get the improved global optimal solution. The results show that the hybrid parallel simulated annealing algorithm can deal with multimodal function more effectively.