Abstract:
In this paper,two sub-swarms substituting particle swarm optimization algorithm(TSSPSO) is proposed.The algorithm parameters are analyzed and the iteration equations are amended.The new algorithm assumes that particles are divided into two sub-swarms.One sub-swarm flies toward the global best particle.and the other flies in the opposite direction.Not only its search experience and the best individual's position of its own sub-swarm,but also the best individual's position of the whole swarm can affect each particle's search during iterations.Each iteration,some bad particles of one sub-swarm are replaced with some good particles of another under a substituting probability.Then,both TSSPSO and particle swarm optimization algorithm(PSO)are used to resolve four well-known and widely used test functions' optimization problems.Results show that TSSPSO has greater efficiency,better performance and more advantages than PSO in many aspects.In addition,TSSPSO is applied to train artificial neural network to construct a practical soft-sensor of gasoline endpoint of crude distillation unit.The obtained results and comparison with actual industrial data indicate that the new method is feasible and effective in soft-sensor of gasoline endpoint.