Research Of Multi-objective Optimization Based On Improved Particle Swarm Optimization Algorithm

There are a lot of multi-objective optimization problems in scientific research andengineering practice. The fields involve city transportation、 urban layouts、 energydistribution、 capital operation and so on. Thus, designing effective algorithms formulti-objective optimization problems is not only the great importance in scientific research,but also the great value in applications. As the implement of the objectives often conflict witheach other, no unique best solution may exist, no solution is superior to other solutions whenconsidering all the objectives, so it brings some difficulties to solve the multi-objectiveoptimization problems. Despite the considerable diversity of techniques developed in theOperations Research field to tackle these problems, their intrinsic complexity calls forhigh-efficiency approaches.Particle swarm optimization algorithm (PSO) is a non-linear function optimizationtechnique of recent development and a swarm intelligence algorithm to simulate the behaviorof birds’ predation which is frequently used to solve multi-objective optimization problemsbecause of its simplicity, fast convergence and less parameters. Based on the theory ofmulti-objective optimization and PSO, the paper focuses on the principle of PSO which isused to solving multi-objective optimization problems. The main contents include:Firstly the paper introduces the purpose and significance of studying modified PSO andmulti-objective optimization problems and development process and research status ofmulti-objective optimization algorithm. Then the analysis on the basic theory ofmulti-objective optimization problems and the basic principles of PSO is given. In order toconquer particle swarm optimization algorithm easily converge to a local optimum andimprove PSO’s convergence performance, after comparing several operators which are incommon use, a particle swarm optimization algorithm based on multi-scale self-adaptiveescape is proposed in this paper. The special multi-scale Gaussian mutation operators areintroduced to make particles explore the search space more efficiently. The large-scalemutation operators can be utilized to quickly localize the global optimized space at the earlyevolution, then the novel strategy produces a scale mutation operator which gradually reduced according to the change of the fitness value, the small-scale mutation operators can implementlocal accurate minima solution search at the late evolution. Convergence principle is given atthe same time. Then in order to solve multi-objective optimization problems, a new velocityupdate equation is used to maintain the distribution of the solutions. The algorithm maintainsthe archived optimal solutions based on crowding distance sorting technique.The Comparison of the performance of the proposed approach with other PSOalgorithms with mutation escape on four typical Benchmark functions is experimented. Theexperimental results show the proposed method can not only effectively solve the prematureconvergence problem, but also significantly speed up the convergence. Then in order to solvemulti-objective optimization problems, comparison of the performance of the proposedapproach with the other algorithms on five typical two objective functions and two threeobjective functions are experimented. The experimental results show that the proposedapproach has better performance.