Multi-Objective Particle Swarm Optimization Algorithm And Its Application

In real life, most of problems consist of multi-objective, which need to be optimized at the same time, and the problems are often conflicting. Particle swarm optimizaiton works as a heristic optimizaiton algorithm, that is simple, easy to realize and convergence quickly, which is used in solving optimizaiton problems, but it was firstly proposed as a single objective optimization problem. The parallel character is used in this paper, the PSO is updated to solve system identification and tunning, and modeling and optimizaiton of real industry.The main contents is as follows:Firstly, the basic particle swarm optimization (PSO) and multi-objective particle swarm algorithm (MOPSO) is briefly introduced, and study and analysis the research status in-depth, on the basis of these, the design main points of the exist improved particle swarm is summarized, finally, the general step of multi-objective evolutionary algorithm is given, and related multi-objective test functions and performance metrics index is introduced to prove the reasonableness and effectiveness of the improved algorithm. Secondly, An adaptive Multi-objective Particle Swarm Optimization algorithm based on dynamic link matrix is proposed. The link matrix is used to describle the swarm topology, and the link mode and evolution weight are adjusted based on the evolution state, when the swarm trap into the local optimum, the mutation is introduced to make the swarm search in the new scope to get the global optimum. And the uniform design is used to select the global best particle, which make sure the distribution of result, at last, Well-known benchmark functions are used to test the performance of the proposed algorithm, which the diversity and convergence are compared with other algorithms. The results show that the proposed algorithm has better performance in global search, the diversity and convergences are better as well.Third, For the problem that the input signal is strong related with output noise through feedback loop in closed control loop, which made idenfication difficult and even unidentify, a new system identification and tuning method based on forward channel is proposed, the forward channel is used to isolate the disturbance from output noise, and simulate the dynamic character of system, and the feasibility and reality is proved in the article. The proposed improved multi-objective PSO is used to tunning the forward channel model, the result was used to the real process, and the result show that the tunning method is better. Fourth, the proposed algorithm is used for modeling, the process model coefficient of the conversion of methanol to hydrocarbons is determined and the first order reaction coefficien of Kumar are optimized, all of these make the model available, and the application scope of the model is broaden.Finally, the total work of the article is summarized; the future direction is pointed out.