Abstract: Owing to the problem that particle swarm optimization algorithm is easily falling into local optima in optimization of high-dimensional and complicated functions，an improved particle swarm optimization algorithm based on two sub-swarms multi-phase exchange is proposed.The whole particle swarm is divided into two sub-swarms of same size.The models of updating the position and velocity of each population particles are different.The number of exchange particles is different in different searching phases of the algorithm，and the amount is constantly decreasing.With these methods，the population diversity can be improved and the vitality of particles can be enhanced.Results show that TSME-PSO can avoid trapping into local optima effectively and has good ability of searching for global optima，and the overall optimization performance is also better than other comparison algorithms.