Through mechanism analysis of several modified particle swarm optimizations (PSO), a new uniform model of PSO is described, and the convergence is analysed with linear control theory. To improve the calculation efficiency, two enhanced global search capability self-adaptive PSOs, one-population self-adaptive PSO and two-population self-adaptive PSO, are proposed. The one-population self-adaptive PSO uses the diverse coefficients in the first evolutionary strategy. The two-population self-adaptive PSO uses two different populations: one owns global search capability, and the other owns local search, and through exchanging information the algorithm efficiency is improved. The simulation results show the correctness and efficiency of the presented methods.