不确定需求下考虑协议库存的电网物资供应优化建模  

Modeling and Optimization of Material Supply in the Power Grid Considering Capacity Reservation Contracts with Uncertain Demand

作　　者：俞春华 陈国华[2] 喻鸿 余建新 王庆阳 YU Chunhua;CHEN Guohua;YU Hong;YU Jianxin;WANG Qingyang(Material Branch,State Grid Jiangsu Electric Power Co.,Ltd.,Nanjing 210036,China;School of Management and Engineering,Nanjing University,Nanjing 210093,China)

机构地区：[1]国网江苏省电力有限公司物资分公司,江苏南京210036 [2]南京大学工程管理学院,江苏南京210093

出　　处：《工业工程》2024年第6期82-92,共11页Industrial Engineering Journal

基　　金：国家自然科学基金资助项目(72074110)。

摘　　要：如何综合挖掘既有仓储网络和协议供应商资源,实现彼此互补和协同、按需备货和分配,是电网企业供应链管理降本增效的难点,其中优化模型的选择是关键。本文采用集成管理的思想,考虑需求的不确定性,基于电网现有物资供应模式,在构建确定性模型的基础上,提出两类物资供应的鲁棒优化模型,利用Bender分解方法设计模型的求解算法。基于国网案例,收集数据,设计实验,对模型和算法的性能进行分析;并针对不同的物资供应模式,探讨需求未满足偏差的惩罚成本参数和预算不确定参数的变化对总成本的影响。研究发现如下。1)Bender分解算法对于求解大规模问题效率更高。2)基于期望需求的确定性模型虽然总成本最低,但抵御不确定需求风险方面不如两类鲁棒模型;基于随机情景的鲁棒模型可以平衡总成本和需求未满足偏差的惩罚成本;考虑预算不确定集的鲁棒模型可以将需求不确定性控制在特定范围内。3)考虑实物库存优先分配与配送的物资供应模式在总成本控制方面更好。4)考虑预算不确定集的鲁棒模型具有很好的需求满足能力,但成本更大;基于随机情景的鲁棒模型在成本上优于考虑预算不确定集的鲁棒模型,但可能部分需求无法满足。Comprehensively exploring resources from the existing warehouse network and contractual suppliers to achieve mutual complementarity and collaboration,coordinated inventory and allocation,is a challenge faced by power grid enterprises to reduce cost and increase efficiency in supply chain management,where the selection of optimization models is critical.From the perspective of integrated management,this paper considers demand uncertainty based on current material supply modes in the power grid.It proposes robust optimization models for two types of material supplies based on a deterministic model.A solution algorithm based on the Bender decomposition method is designed to solve the above models.Using data collected from a State Grid case,experiments are designed to analyze the performance of the models and the algorithm.Additionally,the impacts of changes in the penalty cost parameter of unmet demand deviation and the budget uncertainty parameter on total cost are explored under different material supply modes.Results show that 1)the Bender decomposition algorithm in this paper is more efficient in solving large-scale problems.2)The deterministic model based on expected demand has the lowest total cost but is less effective in addressing uncertain demand risks compared to the two robust models:the robust model based on stochastic scenarios can balance the total cost and the penalty cost of unmet demand deviation,while the robust model based on budget uncertainty can control demand uncertainty within a specific range.3)Material supply modes considering physical inventory prioritization for allocation and distribution offer better total cost control.4)The robust model based on budget uncertain sets possesses a good demand satisfaction capability but reveals a greater cost;the robust model based on stochastic scenarios is better than that based on budget uncertain sets in terms of cost,but may cause some unmet demand.

关 键 词：电网 协议库存 物资供应 鲁棒模型 Bender分解 

分 类 号：F406.14[经济管理—产业经济] O221.4[理学—运筹学与控制论] F252[理学—数学]