近年來，隨著商業環境的快速變遷，使得企業在經營上為了保持市場競爭力，必須提供充足的貨源以隨時能滿足市場的需求變動。因此，若能實施供應商管理庫存系統，且建立良好的補貨水準控管策略，將可以更有效率地進行存貨管理，進而降低整體成本及提升業者營運效率。目前實務上在對於存貨水準的控管仍依賴相關作業人員之經驗與主觀判斷，如此難以瞭解規劃結果之整體最佳性。緣此，本研究以系統最佳化觀點，考量系統總成本最小化為目標，針對存貨水準上下限控管與隨機性需求之問題，發展確定性需求存貨水準上下限控管模式、隨機性需求存貨水準上下限控管模式，以期能提供給決策者進行最佳的補貨水準控管的規劃。本研究利用網路流動技巧及數學規劃方法以構建所有模式，並根據問題特性加上適當的目標式及限制式，以滿足實務的營運條件，此二模式可定式為含額外限制之整數網路流動問題，屬NP-hard問題。因此本研究利用問題分解及合併策略，且配合CPLEX軟體，發展適合的啟發式解法。最後，本研究以連鎖零售業者為立場，針對單一供應商對多個零售商為例，進行範例測試與敏感度分析，進而提出結論與建議。;Faced with rapid changes and a competitive environment, in order for enterprises to maintain their competitiveness, they must ensure a stable supply of products to satisfy fluctuations in market demand. If a good replenishment strategy is built with the Vendor Management Inventory (VMI), it will not only lead to efficient inventory management, but will also reduce the overall system cost and increase operation efficiency. In practice, the inventory level is controlled manually by experienced planning personnel. Consequently, the planning results may not reflect overall optimality. Therefore, in this study, based on the system optimization perspective, we develop the deterministic-demand upper-and-lower inventory level control model, the stochastic-demand upper-and-lower inventory level control model, with the objective of minimizing the total cost, considering the upper-and-lower inventory level control problem and stochastic daily demands. The models are expected to be useful tools for the planner to decide on the optimal upper-and-lower inventory level planning. We employed the mathematical programming and network flow techniques to develop all the models with suitable objective functions and constraints based on the problem characteristics to comply with real operating requirements. These two models are expected to be formulated as integer network flow problems with side constraints, which are characterized as NP-hard. We developed a solution algorithm for each model by adopting a problem decomposition strategy, coupled with the use of the CPLEX software. Finally, case studies and sensitivity analyses are performed the operating data for one supplier and retailers from the chain retailers. Conclusions and suggestions then are given.