A New Multi-objective Evolutionary Algorithm Based On Similarity

In recent years, MOPs also become focus on the domain of EAs. How to further improve EAs, and in MOPs how to combine effectively local search strategies with optimization techniques for ultimately improving quality of solutions, should be central in this work. The research in this work extends techniques used in MOEAs.In this paper, we firstly introduce research background and significance, then we introduce the basic conceptions and definitions of multi-objective optimization, and we also introduce main strategies of multi-objective optimization techniques, the current MOEAs, hybrid MOEAs. In the third chapters, aiming at the drawbacks of MOEAs, we propose a new MOEAs based on similarity. In the forth chapter, aiming at most of engineer optimization problems, we do numerical experiments with three measure metrics, via properly setting up the parameter —distribution index of crossover operator, and make the comparison with classical algorithm NSGA-II. The numerical experiment results indicate that, the proposed algorithm can deal with this kind of test functions highly effective. In the fifth chapter, focusing on the general MOPs, we reset distributed index of crossover operator, the numerical experiments indicate that, it still outperforms NSGA-II, this also shows the proposed algorithm had the strong versatility. Therefore this dissertation proposes the algorithm which has the certain actual value, and it has the very strong attraction and the practical value for project domain that timely and optimization are the same important.The main innovations are as follows: 1. Constructing a kind of new crossover operator; 2. Introducing into the self-adapted neighborhood contraction strategy and the local optimization method; 3. Improved crowded operation of NSGA-II; 4. Reflecting individuals' difference degree with individual's fitness which is together determined by the rank value and the crowded distance, describing individuals' similar degree with similarity, so realizes population classification. 5. Proposing a new MOEA based on similarity.