Abstract: On the basis of simulated annealing, a Particle Swarm Algorithm (PSA) was put forward for solving complicated constrained optimizations. In this algorithm the inertia weight was set to zero, and by simulated annealing algorithm it reproduces the positions of those particles whose evolution has been ceased. This algorithm does not require the penalty function. Instead, it uses the double population searching mechanism, one population storing the particles having feasible solution, the other storing the particles having no feasible solution. In a given probability, the feasible population accepts particles having no feasible solution to keep the diversity of the population. Simulation results show that, by this proposed algorithm, the particles reach the global optimum solutions located on or near the boundary of the feasible region quickly and precisely with good stability.

Key words: artificial intelligence, particle swarm optimization, simulated annealing, constrained optimization problems, double population, diversity