A particle swarm optimizer with dynamic balance of convergence and diversity for large-scale optimization

Particle swarm optimization is found ineffective in large-scale optimization. The main reason is that particle swarlarge-scalem optimization cannot effectively balance convergence and diversity. This paper proposes a particle swarm optimizer with a dynamic balance of convergence and diversity (PSO-DBCD). In the proposed algorithm, a competitive multi-swarm mechanism is put forward, based on which a convergence-guiding learning strategy is proposed for the management of convergence pressure. Furthermore, an entropy-based local diversity measurement is proposed to measure the local diversity of particles. Afterwards, a diversity-guiding learning strategy is proposed based on the local diversity information to further improve the diversity preservation ability of the algorithm. Theoretical analyses are presented to investigate the characteristics of PSO-DBCD. Comprehensive experiments are conducted based on the benchmarks posted on CEC 2013 and several state-of-the-art algorithms to test the performance and scalability of the proposed algorithm. The PSO-DBCD exhibits evident advantages over the compared algorithms in the optimization results with respect to the statistical test results. The proposed strategies are demonstrated to be effective in managing the convergence speed and the swarm diversity. Lastly, a case study of centralized electric vehicle charging optimization shows that PSO-DBCD can reduce the cost of charging for people who use electric vehicles.