An Improved Self-Adaptive Particle Swarm Optimization Algorithm with Simulated Annealing

A parameter automation strategy for particle swarm optimization (PSO) is introduced to enhance the performance to solve high dimensions objects. Initially, to maintain the diversities of the population, the concept of ¿individual coefficients¿ (IC) is employed, where each particle has the individual inertia weight and social acceleration coefficient. From the basis of IC, The ¿individual coefficients¿ particle swarm optimization with simulated annealing (PSO-ICSA) is proposed, where two new strategies are discussed to adjust the coefficients self-adaptively. First, the inertia weights and social acceleration coefficients are adjusted by evaluating the adaptive values of the just passed evolution at each iteration step, while the cognitive acceleration coefficient varies linearly with time. Second, a simulated annealing mutation strategy (SA) is combined to enhance the global convergence ability. The test on benchmark problems shows that the proposed method is more effective, reliable and insensitive to dimensions than the existed time-varying coefficients methods especially of high dimensions objects.