区域危险废物管理网络规划的改进多目标优化

Improved Multi-objective Optimization for Network Planning in Regional Hazardous Waste Management

为提高区域危险废物管理效率、保障区域危险废物处置安全,探讨了具有多类危险废物和多类废物设施的区域危险废物管理网络规划问题,同时选择了回收、处理和处置中心的位置(含处理技术),并确定了危险废物和废物残渣从源点到各类废物设施及在各类废物设施间的分配和运输方案。以运输和选址过程中的总费用和总风险最小为目标,引入废物设施能力与路段风险承载能力约束,构建了可同时优化设施选址、废物分配和废物运输决策的多目标混合整数线性规划模型。结合原问题具有多目标的特点,定制了增广加权切比雪夫算法,对总费用和总风险两目标进行了折中,并近似估计了原问题的非支配前沿。最后,设计了包含45批危险废物、14组候选中心的案例,对所提出多目标优化方法的效果和效率进行了测试。计算结果表明:相比于既有模型,所提出模型可更全面地刻画区域危险废物管理的全过程及其网络规划的特点;同时,所提出算法可在平均耗时低于3 min的情况下找到6个不同的代表性非支配解,且在非支配解的数量及其区分度方面明显优于既有文献普遍采用的线性加权求和算法。

To improve the efficiency of regional hazardous waste management and ensure the disposal security of regional hazardous wastes, we investigated the network planning problem in regional hazardous waste management system with multiple hazardous waste types and multiple waste facility types. We also simultaneously selected the locations of recycling centers, treatment centers （including the treatment techniques） and disposal centers, and determined the allocation and transport schemes of hazardous wastes and waste residues from waste origins to these waste facilities and between these waste facilities. Aiming at minimizing the total cost and total risk in the transport and location processes, and introducing the constraints of waste facility capacity and link risk tolerance capacity, we formulated a multi-objective mixed integer linear programming model to incorporate optimize the facility location, waste allocation and waste transportation decision-making simultaneously. By exploiting the multi-objective characteristics of the original problem, we customized an augmented weighted Tchebycheff algorithm to balance the total cost and total risk objectives, and approximated the non-dominated frontier of the original problem. Finally, we designed a realistic case which contains 45 hazardous wastes and 14 candidate center groups to test the effectiveness and efficiency of the proposed multi-objective optimization approach. The computational result shows that （1） compared with existing models, the proposed model can more comprehensively describe the whole procedure of regional hazardous waste management and the characteristics of the associated network planning problem; （2） meanwhile, the proposed algorithm can find 6 distinct representative non-dominated solutions with the average time lower than 3 min, and it is obviously better than the linear weighted-sum algorithm widely explored by existing works in terms of the number and uniqueness of distinct non-dominated solutions.