Abstract: The traditional particle swarm optimization algorithm (PSO) trends to fall into local extremes and has slow convergence rate in the later stages of iteration when dealing with high-dimensional complex functions. To solve this problem, we propose a particle swarm optimization algorithm with growth property (GPPSO). According to the characteristics of human growth, the algorithm is divided into three stages. In the early stage, the speed update formula adds a rebellious item, so that the algorithm can reduce the probability of premature convergence. In the middle stage, in order to make a compromise between the global and local searches, a balance item is added to the speed update formula. In the later stage, the algorithm removes the velocity item in the speed update formula, which is conducive to rapid convergence. Simultaneously, the two bases for the growth stage division are given. GPPSO is applied to some well-known benchmark functions, and experimental results show that it has advantages for improving convergence performance.