Abstract: Particle swarm optimization (PSO) algorithm, as a classical algorithm of swarm intelligence optimization, has been widely used in practice. However, in the face of different problems, it cannot make real-time adjustments according to the group status, and lacks certain flexibility, therefore, a particle swarm optimization algorithm (DVPSO) based on fusion density peak decision is proposed. For initialization, design the binary mapping of elite point set in order to improve the distribution quality of different types of particles. For velocity update, an information exchange mechanism based on density peak is constructed to balance the particle search tendency. For position updating, a dynamic two-neighborhood search strategy of step size search operator is proposed, which combines population state and optimization range changes to regulate particle movement and give consideration to search flexibility. 12 test functions were compared with PSO and 5 new swarm intelligent optimization algorithms, and 2 engineering problems were compared with 5 new intelligent algorithms. The results show that the DVPSO algorithm has better search accuracy and stability, which verifies the adaptability and good performance of the algorithm.

Key words: particle swarm optimization algorithm, peak density, elite point set, information interaction, dynamic neighborhood search