Key Technology Research And Application Of Constrained Optimization Algorithm

Constrained optimization problems widely exist in the field of scientific research and engineering applications.At present,they have become hot research topic in the field of intelligent information processing.The disadvantages of current constrained single-objective optimization algorithms include:easily falling into local optimum,low convergence accuracy,difficulties of setting the parameters.Especially for constrained multi-objective optimization problems with nonlinear,strong constraint and multi-peak,distribution and convergence of the Pareto solution set obtained by constrained mutil objective optimization algorithms are weak.Therefore,it has important theoretical and practical significance to explore the more efficient constrained optimization algorithms.Aiming at the shortcomings of constrained optimization algorithms for dealing with constrained single-objective optimization and constrained multi-objective optimization,the key technologies such as constraint handling techniques,evolutionary strategies,diversity maintenance strategies and elite selection strategies in constrained optimization algorithms are deeply studied.Many improvement measures are put forward to increase the overall performance of the presented algorithms.Meanwhile,the proposed algorithms are applied to the practical engineering problems for improving the optimization effect compared to the available methods.The main research contents of this dissertation include the following six aspects.Firstly,aiming at the problem of low convergence precision in dual population storage technique,a dual population constrained optimization algorithm based on mix strategy is presented.First of all,the constraint domination and the optimal constraint domination are proposed to update infeasible solution set for strengthening both exploration ability and evolutionary efficiency.And second,the hybrid strategy is employed to evolve population:in the early stage of evolution,Deb criterion is used to generate feasible solutions;in the late stage of evolution,the global optimal solution and second-best solution are applied to guide evolution for quick convergence rate.Last but not least,the good point set method is introduced to produce initial population for improving the diversity.Numerical simulation results verify the effectiveness of the improved algorithm.Secondly,aiming at the problems that ?-constraint could easily lead to local optimum and it's difficult to set parameters,a self-adaptive ? constrained single optimization algorithm is proposed to avoid these.On the one hand,the improved individual comparison criterion can make full use of effective information carried by infeasible solutions,then exploration scopeis enlarged and population diversity is increased.On the other hand,an adaptive adjustment strategy is suggested to produce suitable e for balancing relationship of objective function and constraint violation degree,so it makes more reasonable comparison between individuals.Numerical simulation results verify the effectiveness of the improved algorithm.Thirdly,aiming at the problem that the distribution of obtained Pareto solution set is poor,a constrained multi-objective optimization algorithm based on double population storage is presented.On the one hand,the improved Harmonic distance is used to eliminate adverse effect which is caused by weak grade individuals and distant individuals.It can assess the distribution of the feasible solution set more accurately,and the amount of computation can be reduced to some extent.On the other hand,the update mode of the infeasible solution set closely links the feasible solution set,and it can preserve the infeasible solutions with the better objective value and constraint violation,consequently the diversity is strengthened.In addition,the improved mutation strategy makes full use of the optimal solutions and excellent infeasible solutions,so both exploration and exploitation are balanced.Numerical simulation results verify the effectiveness of the improved algorithm.Fourthly,aiming at the problem that the convergence of obtained Pareto solution set is weak,a constrained multi-objective optimization algorithm based on adaptive a truncation strategy is presented.First,through the proposed adaptive e truncation selection strategy,the Pareto optimal solutions and excellent infeasible solutions are retained to improve diversity,and convergence is also coordinated.Second,the exponential variation is introduced to further enhance the local exploitation ability.Third,a part of the Pareto optimal individuals and the near individuals are chosen by using the improved crowding density estimation to take part in calculation,thus it not only assesses the distribution of the solution set more accurately,but also reduces the amount of computation.Numerical simulation results verify the effectiveness of the improved algorithm.Fifthly,aiming at the problem that distribution and convergence of Pareto scolution set obtained MOEA/D are weak,a ?-constraint multi-objective decomposition optimization algorithm based on re-matching strategy is suggested.First,the Chebycheff decomposition strategy is analyzed,and we get two theorems about diversity and convergence for providing a theoretical basis.Next,in order to effectively solve the diversity loss for weight vector caused by randomly distribute individual,the re-matching strategy is proposed,and it improves the diversity of population.Last,the improved ?-constraint can further enhance the overall performance of the algorithm.Numerical simulation results verify the effectiveness of the improved algorithm.Finally,five proposed algorithms are employed to optimize three practical constrained optimization problems,such as software project scheduling,phased array radar parameter optimization design and large ship total factor optimization design.These validate the effectiveness of the presented algorithms,whilst these could enrich their application fields.