解约束最优化问题的一个新的多目标进化算法

A Novel Multiobjective Evolutionary Algorithm for Constrained Optimization Problems

把约束函数作为目标函数,将约束优化问题转化为多目标规划问题。对这个多目标规划,根据带权极小极大策略构造了一个同进化代数有关的变适应值函数。利用广义球面坐标变换和均匀设计法来选择权重,使得由此权重确定的适应值函数能使种群中的容许解逐渐增加并且保持其多样性。用均匀设计法构造的带有自适应性的变异算子增强了算法的局部搜索能力。该方法能有效处理约束,特别是紧约束。计算机仿真显示了该方法是有效的。

By treating the constraints as objectives,the proposed algorithm transforms the constrained optimization prob-lem into a multiobjective optimization problem,and self-adaptive fitness functions depending on the number of genera-tions are proposed based on a min-max-weighted fitness strategy.Using generalized sphere coordinate transformation and uniform design to determine weights increase the number of feasible solutions and keep the diversity of the population.Furthermore,the self-adaptive mutation operator constructed by uni form design enhancs its power of local search.As a result,the pro posed algorithm can handle constraints effectively,especially the active constraints.Simulation results indi-cate the efficacy of the proposed algorithm.