Research On The Multi-Objective Optimization Problem Based On Differential Evolution Algorithm

In reality,multi-objective optimization problem can be seen everywhere,it is more in line with the actual development of studying it,which has important research significance.these optimization problems often have multiple design objectives that often conflict with each others, So achieving an optimum for all objectives simultaneously is very difficult. There are many shortcomings that using the traditional multi-objective optimization methods to solve this problem.Recently, evolutionary algorithm has gradually become the popular method to solve the multi-objective optimization problem,it has been widely used in solving multi-objective optimization problem. As an important component of EA community, differential evolution has a simple structure, high robustness,etc. It is easy to cobine with other algorithms for constructing efficient hybrid algorithm to solve practical multi-objective optimization problem. This thesis aims at solving multi-objective optimization problems based on DE. The main contents are as follows:First of all, Because single mutation operator based traditional multi-objective differential evolution algorithm easily make the algorithm appear premature of convergence phenomenon, this paper introduces an adaptive mutation operator by adaptively tunning the mutation scale factor according to the algorithm search process. At the same time, algorithm also adopts an external archive to save the nondominated solutions, which can prevent the loss of outstanding individuals and speed up the convergence, and make the non-dominated solutions constantly approach to optimal boundary. Experiments based on five widely used multiple objective functions are conducted. Simulation results demonstrate that our improved algorithm has some advantages compared to several other multi-objective algorithm.Next,Differential evolution algorithm is applied to solve the multi-objective job shop scheduling problems. in order to successfully apply it in practical engineering optimization problem, First discrete the differential evolution operation, change the coding to make it suitable for solving discrete problems. Discrete differential evolution algorithm inherits the advantages ofDE algorithm that has a rapid convergence rate, however,compared with clonal selection algorithm, discrete differential evolution algorithm has a poor local search ability for JSSP, which makes discrete differential evolution algorithm trap in local optima rapidly. So,An effective hybrid algorithm based on discrete differential evolution algorithm and clonal selection algorithm is proposed, the clonal selection algorithm is introduced in order to improve the local search ability.Take full advantages of them and make up for their defects, we combine them for solving multi-objective job shop scheduling problems, Experiments based on a large number of shop scheduling instances are conducted, Simulation results verify the effectiveness of the algorithm,and get the good non-dominated solutions. Finally, the research of this thesis is summarized and prospected.