主讲人

（单位）

钟远光

（华南理工大学）

主持人

（单位）

何勇

（东南大学）

讲座时间

2026年4月17日上午10:00

讲座地点

经管楼B201

主讲人简介

钟远光，华南理工大学工商管理学院副院长，二级教授，兴华杰出学者，博士生导师。中国运筹学会行为运筹学与行为运营管理分会副理事长、管理科学与工程协会理事等。长期致力于供应链库存优化、库存共享与分配、优化理论方法与应用、平台运营等领域的研究，代表性论文发表在MS、OR、MSOM、POM等期刊上。主持国家自科青年A类项目，面上项目和青年C类项目（结题均为特优）以及南网技术、广州供电局、扬腾科技等企业委托项目。入选教育部高层次人才计划青年学者，获得教育部高等学校科学研究优秀成果三等奖，广东省哲学社会科学优秀成果奖一等奖和二等奖，安徽省科学技术三等奖，OMEGA Best Paper等奖励。

讲座内容摘要

We study a multi-period shared-inventory system in which a centralized warehouse replenishes inventory before demand realization and allocates inventory across multiple zones after observing the demand vector. The key challenge is that demand exhibits two distinct forms of dependence: intra-period dependence across zones and inter-period dependence over time. Under a one-step conditional full-information benchmark with homogeneous costs, we show that the period-$t$ joint replenishment and allocation problem admits an order-up-to policy governed by the conditional mean and variance of aggregate demand. This benchmark separates two dependence channels: intra-period dependence acts through aggregate-demand variability and risk pooling, whereas inter-period dependence acts through forecast updating and the evolution of expected demand. We then develop a residual-based multistage distributionally robust framework that preserves the predictive dependence structure of the fitted time-series model and yields a tractable linear-program approximation under affine policies, together with finite-sample out-of-sample performance guarantees. Numerical results show that, relative to sample-average, sample-path robust, and stagewise-robust benchmarks, preserving demand dependence improves out-of-sample cost and tail-risk performance, with gains especially pronounced in highly persistent demand settings. These findings suggest that even when ex post pooling is available, ignoring demand dependence can still materially misposition inventory ex ante.