自然血亲排斥策略对遗传算法的改进及其在剩余静校正中的应用

The improvements of natural cognation exclusion strategy for genetic algorithm and its application in residual static correction

现有遗传算法在处理剩余静校正问题时往往出现早熟收敛现象,难以实现全局寻优,究其原因主要在于现有遗传算法中的选择模型父本的方法过于简单,导致用于交叉操作的模型父本群体多样性过低。本文针对这一问题,提出以模拟退火概率函数生成模型父本群体为基础,使用自然血亲排斥策略的选择算子替代现有遗传算法中优次配对的选择算子。理论模型和实际数据试验结果表明,此法优于常规遗传算法,主要表现为自然血亲排斥策略具有保证群体多样性,抑制早熟收敛,扩大搜索空间,实现全局寻优的特性。

The premature convergence often appears when using existing genetic algorithm in residual static correction,resulting in difficult realization of the global optimization,the key reason is that the selection method of model male parent is too simple,leading to the population diversity of the model male parent which was used in the cross-operation is too low.Regarding the problem,based on the model male parent group which was generated by the simulated annealing probability function the selection operator which uses natural cognation exclusion strategy replaces the matching priority selection operator in existing genetic algorithm.The theoretical model and field data experiments show that the method is superior than the conventional genetic method,its main representation is that the natural cognation exclusion strategy has the characteristics of the assurance of group population diversity,the inhibition of premature convergence,the expansion of the search space,and the realization of the global optimization.