基于种群分类排序的约束优化遗传算法

针对约束优化问题,提出了一类将种群中的个体分类排序的思想.算法的特点在于:先将种群中的解分为可行解和不可行解两类,然后分别按照不同的标准排序.由于很多约束优化问题的最优解位于可行域的边界上或附近,所以排序时并不认为可行解一定优于不可行解.基于此分类排队思想,特别设计了只允许同等级个体进行交叉的新的交叉算子,称之为同等级交叉算子,以及基于一维搜索的变异算子.算法同时采用了保证固定比例不可行解的自适应策略.4个标准测试函数的数值仿真结果验证了算法的有效性.

An Evolutionary Algorithm Based on Classification and Sorting for Solving Constrained Optimization Problem

A novel evolutionary algorithm based on classification and sorting of the population is proposed for solving constrained optimization problems.The primary features of the algorithm proposed are stated as follows:Firstly,all the individuals should be divided into two classes,which are feasible solutions and infeasible solutions.These tow cases are hence sorted with different category standard respectively.Considering that the global optimal solutions locate on or near the boundary of the feasible region for many constrained optimization problems,the feasible solutions are not always considered to be better than infeasible ones;Secondly,based on the idea mentioned above,a special crossover operator named ranked crossover operater is designed and this operator is performed only when the father individuals are in the same level.The mutation operator takes good use of the advantages of one-dimensional research properly.Finally,an adaptive strategy of keeping a fix number of infeasible solutions is introduced in the paper.The numeral results of 4 benchmark test functions demonstrate that the algorithm is effective.