Improvement And Application Of Particle Swarm Optimization Algorithm Based On Gradient Information

In 1995, J. Kennedy and R. C. Eberhart proposed the Particle Swarm Optimization algorithm. It has a simple concept, and it is easy to calculate, and it has few parameters, which is all favored by experts and scholars in various fields in the nearly 20 years. Meanwhile, with the deepening of Particle Swarm Optimization algorithm research, the disadvantages are also shown, which can easily prone to some of the best in the iterative process of particles, as well as the slowly convergence speed in the later period. So the improvement of Particle Swarm Optimization algorithm has been the focus of attention of many experts and scholars in recent years.In unconstrained optimization problems, Particle Swarm Optimization algorithm has been widely used. For unconstrained optimization problems, combining the steepest descent method with Particle Swarm Optimization algorithm, we can get a new algorithm. For constrained optimization problems, we can take advantage of augmented Lagrange multiplier method to make constrained optimization problems into unconstrained optimization problems. As we all know, the steepest descent method is fast and careful. After the particle has a velocity iteration, we take the best position of the group to have a three times iteration of the steepest descent method until we get the optimal solution. We take this improvement algorithm as LA_PSO_SD algorithm. Then finding the middle point and the highest point, and taking the same action, we take this improvement algorithm as ZK_PSO_SD algorithm. By using a test question to compare the LA_PSO_SD algorithm with the ZK_PSO_SD algorithm, then we can conclude that ZK_PSO_SD algorithm has a precision, and fewer iterations, which is a successful improved optimization algorithm.There are many ways that can deal with multi-objective optimization problems that may be encountered in location problems. The method in this thesis is to take the multi-objective optimization in location problems into single objective optimization in location problems, then the problem will be converted to the constrained optimization problems, and used the steepest descent method to improve the particle swarm optimization algorithm, which not only provides a new solution to solve the multi-objective location problem, but also increases the PSO algorithm in practical applications and makes the PSO algorithm been widely used in different aspects and fields.