Abstract: To avoid the prematurity phenomenon of particle swarm optimization,a novel hybrid genetic particle swarm optimization(PSO) algorithm is proposed.An eliminative mechanism is introduced into the iteration of PSO.In addition to multi-offspring competition crossover between the eliminated particles and the best particle,the mutation of the best particle is also preformed to obtain new particles which have better fitness values.Test experiments of four classic benchmark functions indicate that the proposed algorithm can avoid prematurity effectively,and the algorithm possesses better ability in finding global optimum than PSO.The improved algorithm is subsequently used in parameter estimation of Lorenz chaotic systems.Numerical simulation shows that the algorithm can estimate the system unknown parameters effectively even in the presence of measurement noises.