In order to address the problem that the harmony search and particle swarm optimization algorithms are easy to fall into local optimum when solving high-dimensional complex problems, a particle swarm optimization algorithm based on multi-subgroup harmony search (i.e. PSO-LHS) is proposed. With respect to the PSO-LHS, we introduce the Levy flight and give the parameter adaptive adjustment method in the harmony search algorithm. By the aid of a hierarchical search strategy for enhancing the global search ability of PSO-LHS, the bottom layer is composed of a series of sub-populations with the Levy flight harmony search algorithm. Moreover, the upper layer consists of the optimal individuals of each sub-population to form an elite group, and the particle swarm algorithm is used for improving the accuracy of local search. In the process of searching, the sub-population exchanges information with the upper elite to improve the diversity and search efficiency of the population. The experimental results demonstrate that our PSO-LHS algorithm has better efficiency and global convergence compared with HS, PSO and improved HS algorithm.