In this paper a novel approach for nonlinear system identification is proposed based on adaptive particle swarm optimization. Particle swarm optimization is demonstrated as efficient global search method for complex surfaces, and in order to quick the convergence speed, an adaptive particle swarm optimization strategy was introduced. The proposed method formulates the nonlinear system identification as an optimization problem in parameter space, and then adaptive particle swarm optimization are used in the optimization process to find the estimation values of the parameters respectively. Application to Hammerstein model, in which the nonlinear static subsystems and linear dynamic are separated in different order, is studied and compared with other methods and the simulation results show the identification by adaptive particle swarm optimization is very effective and superior accuracy.