管理评论

›› 2017, Vol. 29 ›› Issue (12): 185-194.

• 运作管理 • 上一篇    下一篇

部分需求信息下分布式鲁棒车辆配置优化

冉伦1,2, 吴东来1,2, 焦子豪1,2, 袁书宁1,2   

  1. 1. 北京理工大学管理与经济学院, 北京 100081;
    2. 北京经济社会可持续发展研究基地, 北京 100081
  • 收稿日期:2017-07-29 出版日期:2017-12-28 发布日期:2017-12-20
  • 通讯作者: 吴东来(通讯作者),北京理工大学管理与经济学院博士研究生。
  • 作者简介:冉伦,北京理工大学管理与经济学院教授,博士;焦子豪,北京理工大学管理与经济学院博士研究生;袁书宁,北京理工大学管理与经济学院硕士研究生。
  • 基金资助:

    国家自然科学基金重大研究计划重点支持项目（91746210）；北京市自然科学基金面上项目（9172016）；国家自然科学基金面上项目（71672011）；北京市社会科学基金一般项目（15JGB040）；北京市教委共建项目专项资助。

Distributionally Robust Optimization for Shared Vehicles Allocation under Partial Demand Information

Ran Lun1,2, Wu Donglai1,2, Jiao Zihao1,2, Yuan Shuning1,2   

  1. 1. School of Management and Economics, Beijing Institute of Technology, Beijing 100081;
    2. Sustainable Development Research Institute for Economy and Society of Beijing, Beijing 100081
  • Received:2017-07-29 Online:2017-12-28 Published:2017-12-20

PDF (PC)

468

摘要：

近年来，在共享经济及供给侧改革的时代背景下，车辆共享服务成为解决城市交通拥堵问题的创新服务模式，然而在实际运营过程中诸多不确定性因素降低其运营效率。本文以电动车为主要构成的车辆共享服务为研究对象，在初始车辆配置阶段，考虑不确定需求的部分矩信息，结合最小化最坏情形的思想，建立分布式鲁棒车辆配置模型。同时分别考虑“确定租赁需求”、“随机租赁需求”、“部分不确定租赁需求矩信息已知”三类情况，建立相应的车辆共享服务车辆配置模型，并结合北京市15个租赁点信息，对三类模型算例决策结果进行对比，通过蒙特卡洛模拟比较相应的配置量及车辆配置成本。结果表明本文提出的方法相对于另外两类模型具备更好鲁棒性，同时可较好地在不确定环境下节约运营成本，具备较强实际应用价值，运营商可通过部分需求信息对初期车辆配置进行决策。

关键词: 车辆配置, 分布式鲁棒优化, 需求不确定

Abstract:

In recent years, under the background of sharing economy and supply-side reform, vehicle sharing service becomes an innovative service model to mitigate the urban traffic congestion. However, in the process of vehicle sharing services operation, there are many uncertain factors which make the normal operation less efficient. To explore the approaches to solve this problem, this paper takes the electric vehicle as the main component of vehicle sharing service. In the initial stage of vehicle configuration, considering the partial moment information under the uncertain requirement, and combining the minimizing the worst-case theory, we propose distributionally robust vehicle allocation model. Further, partly considering three categories of situations namely "determined rental requirement", "stochastic rental requirement" and "partial uncertain rental requirement with known moment information", we also put forward corresponding vehicle sharing service vehicle configuration model, compare the results of the three models combining information of the 15 rental point in Beijing, and compare the corresponding allocation quantities and vehicle allocation cost through Monte Carlo simulation. The results show that the proposed method has better robustness relative to the other two models. At the same time, it can better reduce operating costs in an uncertain environment. It has strong practical application value for operators to make decisions on initial vehicle allocation through partial demand information.

Key words: vehicles allocation, distributionally robust optimization, demand uncertainty