A Thermodynamical Selection Rule for the Particle Swarm Optimization

The particle swarm optimization, a stochastic, population-based optimization technique, suffers from a phenomenon called premature convergence. That is, the system often loses diversity of the population at an early stage of searching. In this paper, a novel method called the thermodynamical particle swarm optimization (TDPSO)is proposed, which adopts the concepts of the temperature and entropy in the selection rule, getting a hint from the method of simulated annealing to maintain diversity of the population. The performance of this algorithm is compared to the standard PSO algorithm and experiments indicate that it has better performance.