Fluid catalytic cracking (FCC) kinetic models are characterized by high dimension, nonlinearity, discontinuity, and non-differentiability. Particle swarm optimization is easy to fall into local optima prematurely when it is applied to the optimization of kinetic models. To solve this problem, an improved two-swarm cooperative particle swarm optimization (ITCPSO) is proposed. Considering the reaction mechanism of FCC, an 8-lumps kinetic model was developed. According to the pilot data, nine PSO algorithms and ITCPSO are presented to estimate the parameters. The results demonstrate that better performance of global searching is gained by ITCPSO compared to other PSOs, thus, ITCPSO is expected to be implemented in the optimization of complex kinetic models.