Research On Diversity Maintenance Strategy Of Multi-objective Evolutionary Algorithm Based On Decomposition

There are many problems to be optimized in real life.Such problems have more than one goal to be optimized,and each target conflicts and affects each other.Such problems can be defined as multi-objective optimization problems.In order to solve such problems better,some excellent multi-objective optimization algorithms have been proposed by scholars.The optimization algorithms have gradually become an important topic in the field of intelligent computing.Among them,based on the evolutionary algorithm of population evolution,it can obtain multiple Pareto optimal solutions for each run,so it is often used to solve multi-objective optimization problems and achieve better results.Among many evolutionary algorithms,the decomposition-based multi-objective evolutionary algorithm(MOEA/D)is one of the best.MOEA/D introduces the idea of decomposition into the evolutionary algorithm,so that the multi-objective problem can be decomposed into several single-object sub-problems to be optimized for parallel optimization.Therefore,MOEA/D is more advantageous in terms of diversity and convergence speed when dealing with complex optimization problems.Therefore,further improving the performance of the MOEA/D algorithm is of far-reaching significance and practical value.In this thesis,the shortcomings of the fixed neighborhood size and the unrestricted replacement neighborhood update strategy in the decomposition-based multi-objective evolutionary algorithm are insufficient,and the diversity is not enough to carry out targeted research.Innovatively proposed dynamic neighborhood size setting strategy and adaptive neighborhood updating strategy,and these two strategies are respectively integrated into the MOEA/D algorithm framework,and the effectiveness of the strategy is verified by simulation experiments.The main innovations of this thesis are as follows:1.A strategy for dynamic neighborhood size setting is proposed.Through the selfexcavation operation,the offspring individuals obtain the size of the ability of the offspring to update the neighborhood.Based on this,a judgment mechanism that can accurately reflect the evolutionary range of individuals and the evolutionary state of populations is designed.Combined with the above judgment mechanism,a strategy for dynamic neighborhood size setting(ANS)is proposed for the shortage of population size caused by the fixed neighborhood size.According to the evolutionary range of the individual and the evolutionary state of the population,the strategy can adjust the neighborhood size of the population and the offspring in real time,so that the population diversity can be better maintained during the evolution of the population.Finally,a series of comparative experiments were conducted to verify the effectiveness of the dynamic neighborhood size setting strategy.2.An adaptive neighborhood update strategy is proposed.In the process of population evolution,an adaptive neighborhood update strategy(ENU)is proposed for the unrestricted neighborhood update strategy in MOEA/D algorithm,which leads to the lack of population diversity.The strategy is mainly divided into two parts.The first part is the self-excavation strategy of the offspring individuals.Through the self-excavation of the offspring individuals,more useful information about the algorithm is obtained.The second part is a new neighborhood update strategy,which consists of four different neighborhood update strategies,which enables different types of offspring individuals to adaptively select the corresponding neighborhood update strategy so that the population can evolve during the evolution process.Maintain the diversity of the population well.Finally,a series of comparative experiments were carried out to verify the effectiveness of the adaptive neighborhood update strategy.