Research On Differential Evolution Algorithm To Solve Constrained Optimization Problems

Constrained optimization problems are a type of optimization problems that are common in scientific research and engineering practice.Therefore,solving them has great theoretical significance and practical value.The traditional optimization methods are used to solve constrained optimization problems,which need to use the gradient information of functions,and are only suitable for differentiable functions.In addition,these methods are easy to fall into the local optimum and cannot meet the actual needs.Evolutionary algorithms are the population-based search algorithms,which have been widely used in many fields.They have the advantages of high search efficiency,not easy to fall into the local optimum,and strong robustness.In recent years,evolutionary algorithms have attracted increasing attention for dealing with constrained optimization problems.This paper develops a systematic research on solving constrained optimization problems.The main works are as follows:1.An evolutionary algorithm based on multiobjective optimization is proposed to solve constrained optimization problems.Firstly,a constrained optimization problem is converted into a biobjective optimization problem.Then,the local and global search strategies based on decomposition are presented to solve the transformed optimization problem,and the direction vector adjusting strategy is utilized to guide the population to find the feasible optimum.During the evolution,two improved differential evolution algorithms are used to produce offspring.In this way,we can strike a balance between diversity and convergence.Experiments on two sets of test problems,namely,IEEE CEC2010 and IEEE CEC 2017,have shown that the proposed method performs better than or competitive with other compared methods.2.An evolutionary algorithm based on the penalty function method is proposed to solve constrained optimization problems.Firstly,the penalty function method is used to transform a constrained optimization problem into an unconstrained optimization problem.Then,a two-phase(the exploration phase and exploitation phase)search strategy is presented to solve the transformed optimization problem.In these two optimization phases,the exterior and interior penalty function methods are used to select individuals,respectively.During the evolution,two differential evolution algorithms are used to produce offspring.In this way,we can strike a balance between diversity and convergence.In addition,the dynamic penalty coefficient strategy is adopted to guide the population to find the feasible optimum.Experiments on two sets of test problems,namely,IEEE CEC 2010 and IEEE CEC 2017,have demonstrated that the proposed method is competitive with other popular methods.