Evolutionary Algorithms For Solving Two Types Of Complex Multi-Objective Optimization Problems

Multi-objective optimization problems are widely existed in scientific research and engineering practice,and have become a research hotspot in the field of intelligent information processing.In recent years,there are many types of optimization algorithms in the optimization field to solve multi-objective optimization problems.Among them,evolutionary algorithms show good performance in dealing with multi-objective optimization problems due to their global,parallel and high robustness.The multi-objective evolutionary algorithm based on intelligent algorithms to solve multi?objective optimization problems has attracted more and more researchers'attention.Although the multi-objective optimization algorithm has become the mainstream algorithm for solving multi-objective optimization problems,there are still some shortcomings in dealing with complex multi-objective optimization problems,such as complex decision space and restricted constraints.Therefore,this thesis is aimed at multi-objective optimization problem with complex decision space and constrained multi-objective optimization problem.An evolutionary algorithm based on multi-operator ensemble for multi-objective optimization and a new constrained dominance relation based evolutionary algorithm for constrained multi-objective optimization are proposed.The main research content of this thesis are divided into the following two parts:(1)For the multi-obj ective optimization problem with complex decision space,an evolutionary algorithm based on multi-operator ensemble for multi-objective optimization(EAMOE)is proposed.The core idea of EAMOE is to design a multi?operator ensemble strategy based on subpopulations,the performance of the operator is evaluated in the optimization process,and the size of the sub-population is adaptively adjusted at different stages.To be specific,a performance evaluation indicator is first constructed based on the ratio and the fitness improvement of each subpopulation.Then this indicator is used to update the size of each subpopulation in order to reward or punish the weights of operators.Experimental results on UF test suite demonstrate that EAMOE has better performance in comparison with existing algorithms for multi-obj ective optimization problems with complex decision space.(2)For the constrained multi-objective optimization problem,a new constrained dominance relation based evolutionary algorithm for constrained multi-objective optimization(C-ACEA)is proposed.C-ACEA designed a strategy based on angle-based constrained dominance and angle-based diversity assessment.The angle-'based constrained dominance is used to non-dominated sorting the population in the environment selection stage,and to assign the same dominance level to the infeasible solution with good diversity as the feasible solution,thus helping to explore more feasible areas.While in the diversity maintenance procedure,an angle-based density evaluation method is designed to give the infeasible solutions with good convergence a chance to survival that contribute to get across the large infeasible areas.Experimental results on CDTLZ and DCDTLZ constrained multi-objective optimization test suites demonstrate that the proposed algorithm has competitive performance in comparison with existing state-of-the-art constrained multi-objective evolutionary algorithms for constrained multi-objective optimization problems.