基于大偏差技术的子集排序最优仿真预算分配方法

Optimal simulation budget allocation for subset ranking from a large deviation perspective

排序选优方法已广泛应用于求解离散事件动态系统中的仿真优化问题,但该类方法鲜有研究聚焦于子集排序问题的高效求解,而子集排序问题广泛存在于智能制造、电气工程、供应链管理等众多领域.本文针对k个备选方案的子集排序问题,构建了以最大化子集正确排序概率为目标的仿真预算优化分配模型,推导了该优化问题的渐进最优条件,并提出了相应的序贯仿真算法来实现仿真预算的渐进最优分配规则.数值实验结果表明,本文所提出的算法显著地提高了子集排序问题的仿真优化效率.

The ranking and selection method has successfully been used to solve simulation optimization problems for many discrete event dynamic systems.However,little research has focused on optimizing the simulation efficiency of the subset ranking problem,which has wide applications in areas such as intelligent manufacturing,electrical engineering,and supply chain manufacturing.This research studies the simulation budget problem of dividing k simulated alternatives into r subsets and formulates this problem using the optimal computing budget allocation framework.The asymptotic optimality condition that maximizes the probability of correct subset ranking is derived,and a corresponding sequential simulation algorithm for implementing the budget allocation rule is suggested.The results of numerical experiments show that the proposed simulation algorithm significantly improved the simulation optimization efficiency for subset ranking.