A dynamic multi-swarm cooperation particle swarm optimization with dimension mutation for complex optimization problem

Particle swarm optimization (PSO) has been used to solve numerous real-world problems because of its strong optimization ability. However, PSO still has some shortcomings in solving complex optimization problems, such as premature convergence and poor balance between exploration and exploitation. To overcome these drawbacks of PSO, a dynamic multi-swarm cooperation PSO with dimension mutation (MSCPSO) is proposed in this paper. There are two contributions in MSCPSO, which are the adaptive sample selection strategy (ASS) and the adaptive dimension mutation strategy (ADM). Firstly, in ASS, particles in each sub-swarm are sorted into three states (elite, ordinary and inferior) according to their fitness. Three samples pool are used to save elite, ordinary and inferior particles. Particles in each sub-swarm can select their learning samples in their sample pools adaptively according to their fitness. Therefore, ASS can facilitate information interaction among the sub-swarms and increase the diversity of the population. Secondly, ADM generates the mutation positions for the whole population according to the information and knowledge acquired by particles during the evolution. In this case, ADM is used to enhance the exploitation ability of DMS-PSO without losing population diversity. Finally, two test suites (CEC2013 and CEC2017) and four practical engineering problems are used to verify the performance of MSCPSO. Experimental results verify that MSCPSO has a remarkable performance compared with 7 recent state-of-the-art PSO variants in most complex and multimodal conditions.