Research On Adaptive Multi-objective Particle Swarm Optimization And Its Application

Multi-objective optimization problems are widely existed in many real-world applications.Multiple objectives,which are often conflicting with each other,are need to be optimized simultaneously when solving a multi-objective optimization problem.Therefore,unlike the single objective optimization problem that only has one global optimal solution,a set of solutions is often used as the best solutions for a multiobjective optimization problem.The performance of these optimal solutions on each objective may be good or bad,but none of them can achieve better performance than the other solutions on all objectives.Solving multi-objective optimization problem is one of the key research directions in the field of optimization.As a classical swarm-based intelligent optimization algorithm,particle swarm optimization has a strong potential to solve multi-objective optimization problems for its simple form,fast convergence and easy adjustment of parameters.In particle swarm optimization,the movement of particles is mainly guided by the global best solution of the swarm and the personal best solution of the particles.Therefore,reasonable selection strategies for global best solution and personal best solution are the most important part of a multi-objective particle swarm optimization.In the thesis,a global best solution selection algorithm based on adaptive virtual Pareto front and virtual generational distance is proposed to select a more reasonable global best solution of the swarm.And an adaptive personal best solution selection algorithm based on the the relative position of the particle,the global best solution and the personal best solution is presented to select a more suitable personal best solution for each particle according to the current evolution state.These two selection algorithms are then integrated into an adaptive multiobjective particle swarm optimization algorithm based on the virtual Pareto front(AMOPSO/v PF),and the performance of the AMOPSO/v PF is then compared with several well-known multiobjective evolutionary algorithms on several common benchmark problems.The experiment results show that whether using IGD or HV as the performance evaluation indicator,the AMOPSO/v PF both has the best comprehensive performance.In addition,a case study on the site selection problem for mobile earthquake monitoring stations is also illustrated the effectiveness of the proposed algorithm in the real-world application.The experiment results show that the mobile station layout scheme given by AMOPSO/v PF is more practical than other comparison algorithms.