Research On Multi-objective Optimization Algorithm Based On Evolutionary Mechanism

Optimization problem is a common problem in engineering practice and scientificresearch and multi-objective optimization problem (MOP) is the problem whichoptimizes more than one objective simultaneously. Generally speaking, theseobjectives are conflicting with each other; therefore, optimal solutions of MOP are theset of optimal solutions which are different from the single objective optimizationproblem. How to obtain the optimal set of solution is always the focus attention ofacademic and engineering. Multi-objective optimization algorithm based onevolutionary mechanism is proved to be one of the most effective and widely usedalgorithms. This intelligent optimization method applies the calculate technology ofevolutionary algorithm in multi-objective optimization field, which obtains a set ofsolution in a run, has a high efficiency and also can avoid the local optima. However,the quality of the solutions obtained by the existing algorithms is still difficult to meetthe actual needs. Therefore, this paper researches on the multi-objective optimizationalgorithm based the evolutionary mechanism and main work focuses on theconstruction of the elitist population and the methods producing offspring which aretwo key elements of evolutionary mechanism:(1)In order to improve the quality of solutions of MOP, in the constructingmethod of elitist, we propose the enhanced elitist mechanism based on methodmaintaining diversity of entropy. This enhanced elitist mechanism selects the set ofnon-dominated solutions as the elitist population, when the number of thenon-dominated solutions in population is no more than the size of elitist population;when the size of the non-dominated solutions is more than the size of elitistpopulation, the method maintaining diversity of entropy is applied to filter the set ofthe non-dominated solutions. This maintaining method takes use of the regions takingevery solution as the center and the information entropy to select the most crowd anduneven region. The individual with the least desirable distribution will be deletedfrom this region. Combining the enhanced elitist mechanism and the traditionalmethod producing offspring, multi-objective evolutionary algorithm based on anenhanced elitist mechanism is obtained. Experimental results demonstrate that thenew algorithm not only has the better convergence, but also improves the diversity ofpopulation, compared with several state of art EAs. (2)To solve the multi-objective optimization problem with variables associatedbetter, focusing on the method producing used in the estimation of distributionalgorithm, the multi-objective estimation of distribution algorithm based on chaosoptimization and grid selection is proposed. In this algorithm, the population isinitialized by use of chaos and an adaptive strategy to produce offspring is applied,which speed up the convergence early in the algorithm and producing thenon-dominated solutions, and improves the modeling quality when estimating ofdistribution. Secondly, local search based on chaos will be used to optimize locally fornon-dominated individuals in offspring, which can make full use of the advantages ofevery individual to enhance the accuracy of the solution. Finally, a simple gridselection is employed to keep a uniformed distribution and enhance the diversity ofthe elite population. Experimental results show that compared with the existingalgorithms, the proposed algorithm can obtain better solutions.This paper researches mainly on the multi-objective optimization algorithm basedon evolutionary mechanism. The work not only enriches the theoretical study ofmulti-objective optimization algorithm, but also enhances the ability to solve the MOPof the multi-objective optimization algorithm based on evolutionary mechanism.