In order to better identify the parameters of the fractional-order system, a modified particle swarm optimization (MPSO) algorithm based on an improved Tent mapping is proposed. The MPSO algorithm is validated with eight classical test functions, and compared with the POS algorithm with adaptive time varying accelerators (ACPSO), the genetic algorithm(GA), and the improved PSO algorithm with passive congregation(IPSO). Based on the systems with known model structures and unknown model structures, the proposed algorithm is adopted to identify two typical fractional-order models. The results of parameter identification show that the application of average value of position information is beneficial to making full use of the information exchange among individuals and speeds up the global searching speed. By introducing the uniformity and ergodicity of Tent mapping, the MPSO avoids the extreme value of position information, so as not to fall into the local optimal value. In brief, the MPSO algorithm is an effective and useful method with a fast convergence rate and high accuracy.