Research And Application Of P Referenceinspired Multi-Objective Co-Evolutionary Algorithm

The multi-objective co-evolution algorithm can effectively improve insufficient selection pressure in the traditional multi-objective evolutionary algorithm by using the cooperation and competition relationship between multiple populations.However,with the increase of the number of objectives,the performance of the algorithms decreases.The reason for above phenomenon is that the proportion of non-dominated solutions in the population increases rapidly,and the diversity selection mechanism is lacking,so it is difficult to identify the quality of the solution set in the same dominant relationship.In addition,with the increase of the object number of optimization problems and the complexity of the problem characteristics,the number of solutions needed to approach the Pareto front increases rapidly,which makes it difficult to balance the convergence and diversity among populations.In order to solve above issues,this paper proposes the following method,which is based on preference-inspired multi-objective co-evolutionary algorithm and form the perspective of diversity selection and resource allocation.Firstly,this paper proposes a preference-inspired multi-objective co-evolutionary algorithm based on based on differentiated resource allocation strategy(PICEAg-DS),which mapping multiobjective spaces to two-objective spaces and designs spatial distance operators to dynamically allocate the goal vectors.Secondly,the dynamic adjustment of goal vector strategy under differential space,which improved the performance of PICEAg-DS in many-objective optimization problem,is designed.It uses reference vectors to divide objective space,and assesses the search ability of subspace by the number of nondominated solution in subspaces.Due to the lack of diversity selection mechanism,a neighbor selection strategy is designed to judge the neighbor relationship based on the angle between individuals,and uses fitness value to screen the superior neighbor individuals to maintain the diversity of population.Thirdly,considering the conflict between convergence and diversity,a preference-inspired multi-objective coevolutionary algorithm based on grouping strategy is proposed.It designs the index of evolutionary potential,which uses generations as influence factor,the density of non-dominated solutions in subspaces to evaluate the difficulty of evolution,and uses the breadth of non-dominated solutions to evaluate the diversity of subspaces.Then this paper generates the random goal vectors to whole objective space,preventing the degradation of individuals in spaces with high evolutionary potential.Lastly,the proposed algorithm is applied to the electric vehicle charge-discharge scheduling optimization problem,which can verify the performance of the proposed algorithm and provide a reasonable choice for decision makers.