关键词: 连锁零售企业, 供应链, 智能补货, 需求预测, 库存管理, 整数非线性规划

Abstract: Effective supply chain management in retail enterprises entails complex challenges. Among these, proficient inventory management is crucial—not only does it reduce inventory costs and meet customer demands, but also it significantly enhances the efficiency of collaborative management across the supply chain. At the heart of inventory management lies the process of inventory replenishment, which requires a detailed understanding and precise prediction of future product demand. This process must consider various factors, including current inventory levels, order volumes, minimum order requirements and so on. There are problems with traditional manual replenishment methods while widely used, when applied on a large scale. They are susceptible to the influence of personal bias and preference, leading to suboptimal replenishment decisions.
This study presents an intelligent replenishment system that employs statistical and optimization models. It consists of two core components: a demand prediction module and a proactive replenishment module. The demand prediction model incorporates long-term historical sales data to account for irregular demand fluctuations, forecasting future demands through the sales distribution patterns to stabilize inventory levels. We outline two distinct forecasting strategies to support varying replenishment approaches. An aggressive strategy prioritizes demand fulfillment, while the conservative strategy considers the risk of inventory excess. The choice of strategy is user-dependent, allowing for adaptability to the specific conditions of each product.
On this foundation, our study introduces a proactive replenishment optimization model that handles multiple replenishment constraints, ranging from demand constraints to limits on replenishment amounts, quantities, and units. The model aims to achieve dual optimization goals: minimizing replenishment costs and maximizing inventory turnover rate. The relative importance of these goals is adjustable through the assignment of different weights, allowing for a balanced approach to evaluation. To solve the associated integer nonlinear programming challenge effectively, we design a bespoke heuristic algorithm.
The effectiveness of this system is evaluated through simulations and practical trials at the PetroChina Group gas station convenience stores. We test various combinations of demand forecasting and proactive replenishment strategies to facilitate replenishment decisions. These tests allow us to monitor and analyze dynamic shifts in inventory and sales at both the individual store and central warehouse levels over the trial period. Subsequently, we assess key performance indicators, including demand fulfillment rates and inventory turnover rates. The findings suggest that the intelligent replenishment system proficiently meets targeted demand fulfillment objectives while effectively controlling inventory costs. Notably, the system exhibits a low dependence on initial inventory levels and demonstrates the capacity of compensating for initial inventory variances over successive replenishment cycles.
In conclusion, this study offers an advanced framework for decision-making in supply chain management, leveraging operations research optimization models. It facilitates precise demand predictions and intelligent replenishment decisions, reducing the risk of inventory surplus, cutting costs, and enhancing resource utilization efficiency, thereby improving the overall operational performance of the supply chain. Moreover, the study highlights the importance of data-driven technologies in modern business and logistics, emphasizing the critical role of statistical modeling and optimization methods in complex decision-making scenarios.

Key words: chain retailer, supply chain, intelligent replenishment, demand prediction, inventory management, integer nonlinear programming