Multiple Improvements to the Particle Swarm Optimization Algorithm

To address the tendency of the particle swarm optimization algorithm to fall into local minima as well as to speed up its final stages of convergence, this paper proposes a combination of several improvements to the algorithm. First, the convergence factor is introduced as a special case of the inertia weight and transboundary particles are reset. Second, to increase the diversity of the particles and break the stagnation states of the particle swarm, a random number is introduced into the speed and position of the particles. The variation of the particles' position and velocity is hence optimized, and the search space for the optimal value is expanded. We use the single-mode sphere and Rosenbrock functions as well as the multimodal Rastrigrin and Griewank functions to verify the algorithm. The maximum, mean, variance, stability, and convergence accuracy of the proposed algorithm are improved.