改进人工蜂群算法及其在应急调度优化问题中的应用

Improved artificial bee colony algorithm and its application on optimization of emergency scheduling

针对大规模灾难发生时首批生命物资的应急调度建模及优化求解问题进行了研究。将受灾点缺失损失评价函数由线性扩充到非线性,对一次性消耗类和连续性消耗类物资建立了多对多约束多目标调度模型。基于Pareto支配和拥挤距离的概念将人工蜂群基本算法应用到此问题的求解,并对算法进行了改进:基于反向食物源的定义提出融合反向学习的食物源初始化,以提高初始解的质量;将反向学习策略和广泛学习策略融合到蜜蜂搜索过程,以反向食物源和其他较好食物源信息来引导搜索方向。对三种规模下随机生成的调度问题数据的仿真实验表明,改进算法所求出的非支配前沿解集更具多样性,分布更加广泛和均匀,能够为首批应急物资调度决策进行支持。

This paper studied modeling and optimizing problems on first batch of emergency materials scheduling when large- scale disaster occurs. After extending loss evaluation function of affected point from linear to nonlinear, it constructed multi to multi constrained scheduling models with multiple objectives on disposable and consumable supplies. Then this paper applied artificial bee colony algorithm to solve this model based on Pareto dominance and crowding distance, and improved the algorithm by following policies： on the definition of backward food source, proposed foods initialization with backward learning to improve the quality of initial solutions; added backward learning and comprehensive learning into bee search procedure to af- fect searching direction by the information of backward and other better food source. Experiment results on randomly generated data of three scales scheduling problems show that non dominated front solutions set solved by the improved algorithm is more diverse, more extensive and more uniform, so it can be used to support for emergency scheduling decision on first batch of emergency supplies.