一种半角距离变化的差分进化处理约束优化问题

A differential evolution method with half-angular distance variation to deal with constrained optimization problems

针对传统差分算法在求解约束优化问题时存在收敛速度慢、精度低等问题,本文提出了一种基于半角距离变化的改进差分进化算法(HDDE)﹒首先,由父代产生2个子代,并利用可行性规则选出最好的子代;其次,对所选最优子代进行约束违反度预处理,即当其满足半角距离变化时,计算出子代的真实约束违反度,否则子代的约束违反度为无穷;最后,若经过预处理的子代约束违反度小于父代的约束违反度,便使用可行性规则进行比较,否则保留父代﹒通过对12个基准约束优化问题进行仿真研究,结果发现:相较于对比算法,HDDE算法的收敛时间最短,为0.216 s;在精度方面,该算法有5个测试集的标准差为0,这说明所提算法具有更好的性能。

For the traditional difference algorithm has slow convergence and low accuracy in solving constraint optimization problems,an improved differential evolution algorithm based on half-angular distance variation(HDDE)is proposed.First of all,two offspring are generated by the parent,and the best offspring is selected by using the feasibility rules.Secondly,the constraint violation degree of the selected optimal offspring is preprocessed,that is,when it satisfies the change of half-angular distance,the real constraint violation degree of the offspring is calculated,otherwise the constraint violation degree of the offspring is infinite.Finally,if the constraint violation degree of the preprocessed offspring is less than that of the parent,the feasibility rules are used for comparison,otherwise the parent is retained.Through the simulation study of 12 benchmark constrained optimization problems,it is found that,compared with the comparison algorithm,the convergence time of HDDE algorithm is the shortest,which is 0.216 s,and in terms of accuracy,the standard deviation of 5 test sets is 0,which shows that the proposed algorithm has better performance.