A New Strategy of Acceleration Coefficients for Particle Swarm Optimization

Acceleration coefficients controlled the impact of the particle´s own experiences and the other particles´ experiences on the trajectory of each particle. The setting of acceleration played a key role in the performance of particle swarm optimization. To efficiently control the local search and convergence to the global optimum solution, a good investigation to the key role of the setting of acceleration coefficients was made. A new recommended value for acceleration coefficients and corresponding PSO algorithm were presented basing on the investigation. In the new automation strategy, an unsymmetrical transfer range of acceleration coefficients were considered and it obtained preferable effects when changing c₁ from 2.75 to 1.25 and changing c₂ from 0.5 to 2.25. Four benchmark functions were selected as test functions. All four functions were tested with different dimensions and different population sizes. The experimental results illustrate that the new method of setting acceleration coefficients speeds up the convergence of PSO and improves the performance of PSO. Furthermore, it also can reserve more diversity of the swarm at the beginning period of the algorithm and thus have more ability to escape from local minimum