To overcome the disadvantage of particle swarm optimization(PSO) algorithm such as easily trapping into local
optimal solution, this paper proposes a kind of hybrid particle swarm optimization algorithm. The maximum speed linear degressive method can effectively compromise between global searching capability and algorithm convergence precision. The random weight can effectively balance the global and local searching ability of the algorithm. Second-order oscillative learning factor can maintain the population diversity under the condition of invariable particle number. At the same time, natural selection principle can improve the convergence speed. The function test experimental results show that the proposed algorithm can avoid premature convergence problem, and effectively improve its optimization ability.