In order to improve the performances of particle swarm optimization（PSO）on convergence rate and accurate, a hierarchical structure poly-particle swarm optimization（HSPPSO） approach using the hierarchical structure concept of control theory was presented. In the bottom layer, parallel optimization calculation was performed on poly-particle swarms. In the top layer, each particle swam in the bottom layer was treated as a particle of the single particle swarm. The best position found by each particle swarm in the bottom layer was regard as the best position of single particle of the top layer. The result of optimization on the top layer was fed back to the bottom layer. If some particles trended to local extremum in PSO algorithm implementation, the particle velocity was updated and re-initialized. The test of proposed method on four typical functions showed that HSPPSO performs better than PSO both on convergence rate and accurate solutions. The HSPPSO proposed in this paper provided a new idea for large scale system optimization problem.
