Abstract:

The real-time transshipment problem of in-transit inventory in replenishment management systems of gas stations needs to consider not only transshipment time and uncertain future demand but also real-time response to transshipment demand. In response to this problem, with the goal of maximizing the overall profit of the remaining decision time period, this paper proposes a method to divide the overall profit into two parts by combining the demand forecasting model and the stochastic demand distribution. First, the remaining decision time period is divided into two parts. Then, the demand for the first period is quantified according to the demand forecasting model, and the total profit for this period and the inventory level at the beginning of the second period are determined. Next, the uncertainty of the demand is measured using the stochastic distribution, and the value function of the inventory level is solved based on the approximate dynamic programming algorithm to measure the total profit of the second period. Afterwards, the selection rule of stations where some in-transit inventory can be transshipped is designed, which can reduce the decision space and improve the response efficiency of the strategy. Finally, the comparative experiment verifies the effectiveness of the transshipment strategies generated by the proposed method under different station sizes and initial inventory ranges, as well as obtains the managerial implications. This paper can provide decision-making support for the real-time transshipment of in-transit inventory in the replenishment management system of gas stations, and enlighten significance for similar real-time transshipment problems that need to consider both transshipment time and future demand.

Key words:

transshipment, in-transit inventory, demand forecasting, approximate dynamic programming

摘要：

加油站补货管理系统中的在途库存实时转运问题需要考虑转运时间以及未来不确定的需求，且需要实时响应转运需求。针对该难题，以最大化剩余决策时段内整体收益为目标，提出一种结合需求预测模型和随机需求分布以分段衡量整体收益的方法，解决了加油站补货管理系统中在途库存的实时转运难题。首先，将剩余决策时段划分为两个时段。其次，基于需求预测模型量化第1个时段内的需求，以衡量该时段的利润总和，并确定第2个时段初站点的库存水平；采用随机分布衡量需求的不确定性，基于近似动态规划算法求解库存水平的值函数，以衡量第2个时段的利润总和；设计拟被转运走部分在途库存的站点的选择规则，缩小决策空间以提高策略的响应效率。最后，对比实验验证了本文方法产生的转运策略在不同站点规模和初始库存范围下的有效性，并得出管理启示。本研究可为加油站补货管理系统中在途库存的实时转运提供决策支持，对类似需同时考虑转运时间和未来需求的实时转运问题具有启发意义。

关键词:

转运, 在途库存, 需求预测, 近似动态规划