为了寻求合理高效的索穹顶结构预应力优化方法，通过对模拟植物生长算法（PGSA）基本原理和算法机制的剖析，提出了大步长生长扩散机制及生长点筛选机制的新策略。在此基础上，建立了基于阶段生长的模拟植物生长算法（阶段生长PGSA），该算法将优化分为多个阶段，并在不同阶段引入相应的生长扩散或生长点筛选机制，先引入大步长生长扩散机制，以多倍步长一次性扩散生长的方式实现生长点的散播，再以中步长配合较宽松的筛选机制进行快速搜索，最后以精度要求的小步长与较严格的筛选机制进行收敛；由此，采用阶段生长PGSA对索穹顶结构进行预应力优化，并与其他算法进行对比。结果表明：与原PGSA相比，阶段生长PGSA可有效提高全局搜索能力、缩小生长空间且避免了生长点饱和的问题；与多岛遗传算法、自适应模拟退火算法及粒子群优化算法等优化方法相比，阶段生长PGSA的计算迭代次数最少，优化得到的结构初始应变能最小，表现出更高的优化效率及更好的优化效果，该算法适用于索穹顶结构的预应力优化。

In order to seek a reasonable and efficient prestress optimization method for cable domes, this paper proposed a new strategy for growth diffusion mechanism of large step and screening mechanism of growth points by analyzing the basic principle and algorithm mechanism of plant growth simulation algorithm (PGSA). On this basis, the plant growth simulation algorithm based on stage growth (stage growth PGSA) was established. This algorithm divides the optimization process into multiple stages and introduces the corresponding growth diffusion mechanism or screening mechanism of growth points in different stages. Firstly, the growth diffusion mechanism of large step was introduced to realize the spread of growth points by means of one-time diffusion growth with multiple steps. Then, the fast search was carried out with the medium step and the loose screening mechanism. Finally, the convergence was conducted with the small step for accuracy requirements and the strict screening mechanism. Thus, the prestress optimization of cable domes was done by using the stage growth PGSA and compared with other algorithms. The results show that the stage growth PGSA can effectively improve the global searching ability, reduce the growth space, and avoid growth points becoming saturated compared with the original PGSA. The number of calculation iteration of stage growth PGSA is the smallest and the initial strain energy of the structure after optimization is minimal in comparison with those of the multi-island genetic algorithm, adaptive simulated annealing algorithm, and particle swarm optimization algorithm. With its higher optimization efficiency and better optimization effect, therefore, this algorithm is applicable to the prestress optimization of cable domes.