A self-guided Particle Swarm Optimization with Independent Dynamic Inertia Weights Setting on Each Particle

In the standard PSO algorithm, each particle in swarm has the same inertia weight settings and its values decrease from generation to generation, which can induce the decreasing of population diversity. As a result, it may fall into the local optimum. Besides, the decreasing of weights values is restricted by the maximum evolutionary generation, which has an influence on the convergence speed and search performance. In order to prevent the algorithm from falling into the local optimum early, reduce the influence of the maximum evolutional generation to the decline rate of weights, A Self-guided Particle Swarm Optimization Algorithm with Independent Dynamic Inertia Weights Setting on Each Particle is proposed in the paper. It combines the changes of the evolution speed of each particle with the status information of current swarm. Its core idea is to set the inertia weight and accelerator learning factor dynamically and self-guided by considering the deviation between the objective value of each particle and that of the best particle in swarm and the difference of the objective value of each particle's best position in the two continuous generations. Our method can obtain a balance between the diversity and convergence speed, preventing the premature as well as improving the speed and accurateness. Finally, 30independent experiments are made to demonstrate the performance of our method compared with the standard PSO algorithm based on 9 standard testing benchmark functions. The results show that convergence accurateness of our method is improved by 30% compared with the standard PSO, and there are 4 functions obtaining the optimal value. And convergence accurateness is improved by more than 20% for 5 functions at the same evolution generation.