Abstract:

In order to solve constrained optimization problem, a dynamic multi-swarm particle swarm optimizer (DMCPSO) is proposed. In DMCPSO, dynamic multi-swarms and comprehensive learning strategy are applied to improve the swarm diversity. On the basis of the idea “no talent is to be wasted” from human society, each sub-swarm selects its member to the maximum utility of each particle, and the dynamic mutation operator is adopted for best performing particle (Gbest) to improve the ability of escaping from local optima. Experimental simulation results of benchmark functions show that DMCPSO achieves better solutions than other algorithms.

Key words: constrained optimization; dynamic multiswarm; particle swarm optimizer