Abstract: As sustainable supply chains evolve and innovate,existing sustainable supply chain networks with outdated indicators that do not take full advantage of the latest sustainability indicators to accurately measure and optimize the network.Based on the CITI Evaluation Guide 7.0,a five-in-one sustainable closed-loop supply chain network model was proposed,which integrated economic cost,compliance reform and energy saving and emission reduction,green supply chain,promoting public green choices,supply chain communication and transparency.A branch delimitation algorithm MCTS_BB using Monte Carlo tree search improvement was designed for efficient solution.The five indicators were normalized by their interrelated influence and affiliation and a multi-objective decision matrix was constructed.Then,a mixed-integer linear planning model was constructed by linearly combining the multi-objective functions using the eigenvectors corresponding to the largest eigenvalues in the decision matrix.The model was solved with branch selection,stochastic simulation search and pruning strategies in MCTS_BB.The validity of the model and algorithm was verified through different scale cases,and the experimental results showed that the five-in-one model achieved an effective optimal balance between multiple objective functions and provided decision guidance to decision makers at all levels for the integrated management of model indicators and parameters.

Key words: sustainable supply chain, combinatorial optimization, Monte Carlo tree search, branch and bound

摘要： 针对现有可持续供应链网络中指标陈旧,且未能充分利用最新可持续指标对网络进行准确衡量与优化的问题,基于《CITI评价指南7.0》,提出一种将经济成本、合规整改与节能减排、绿色供应链、推动公众绿色选择、供应链沟通与透明5项指标进行层次分析加权整合的五位一体可持续闭环供应链网络模型,并设计了一种采用蒙特卡洛树搜索改进的分支定界算法MCTS_BB进行高效求解。首先,对5项指标按相互关联影响及隶属度关系进行归一化并构造多目标决策矩阵。然后,使用决策矩阵中最大特征值对应的特征向量对多目标函数进行线性组合,建立混合整数线性规划模型。最后,通过MCTS_BB中的分支选择、随机模拟搜索和剪枝策略求解模型。通过不同规模算例验证了模型和算法的有效性,实验结果表明,五位一体模型实现了多目标函数间的有效最优平衡,为各层级决策者对模型指标与参数进行统筹管理提供决策指导。

关键词: 可持续供应链, 组合优化, 蒙特卡洛树搜索, 分支定界