Abstract System identification can be formulated as a multimodal optimization problem with high dimension. The original particle swarm optimization (PSO) usually suffers from premature convergence tending to get stuck to local optima and low solution precision while solving these complex multimodal problems. In order to solve this problem, a comprehensive learning particle swarm optimization (CLPSO) method was utilized to estimate parameters of structural systems. This variant of PSO enables the diversity of the swarm to be preserved to discourage premature convergence. The effectiveness of the proposed method is evaluated through the numerical analysis and an application to a real building under conditions including limited measurement data, noise polluted signals, and no prior knowledge of mass, damping, or stiffness.