Effective inventorymanagement will directly affect the overall performance of an enterprise.

Effective inventory management can improvecustomers' satisfaction and sales, improve financial performance ofenterprises, further improve the efficiency of enterprises in the fields ofsite, production capacity, equipment, construction, capital and employees, andalso have an impact on the market area, helping enterprises to further expand .Themain purpose of inventory management is to analyze and control the cost ofinventory, to minimize the cost of inventory, in order to obtain more profits.

Inventory cost refers to the sum of allkinds of expenses incurred by inventory, which consists of purchase cost, orderor production preparation cost, inventory holding cost and out-of-stock cost.These are the four most important components of inventory cost, and the mainresearch object of this paper is also these four kinds of costs.

At present, the main research methods ofinventory control decision-making at home and abroad are quantitative analysis,that is, to determine the lead time and service level in advance, and todetermine the purchase cycle according to the actual situation. The mainmethods used are: ABC classification, economic order lot sizing, EOQ analysis,etc. These methods have been mature in studying single warehouse, but for thedistributed inventory system composed of multiple warehouses, due to poorinformation, uncertain demand and other reasons, there is often an overallinventory backlog or a partial inventory shortage. Inventory occupies a largeamount of capital, which reduces the rapid response ability of enterprises tothe market and the level of customer service. Therefore, reasonable, timely andeffective distributed inventory control is of great significance.

Most of the previous studies did notconsider the whole enterprise inventory system from a distributed perspective.The methods adopted include enumeration-based dynamic programming, branch andbound method, search-based heuristic method and simulation method. The formeris simple and feasible, but the calculation is large. Heuristic methods areeasy to fall into local optimum. The cost of applying simulation method is veryhigh, and the accuracy of simulation is limited by the judgement and skill ofprogrammers.

Thepurpose of this paper is to study a distributed inventory system consisting ofseveral warehouses based on coordinating centers. Coordinating centers play therole of joint inventory management. Customers send orders to the coordinatingcenters. Coordinating centers designate warehouses to supply their warehousesaccording to their location, delivery date, demand and inventory situation.When the total inventory drops to the total order point, warehouses jointlyorder from suppliers through the coordination center; when a warehouseinventory drops to the order point and the total inventory does not drop to thetotal order point, warehouses adjust each other under the unified scheduling ofthe coordination center. In the case of limited capital, storage capacity,supply capacity and variable cost, an improved genetic algorithm and stochasticsimulation method are used to determine the inventory ordering and allocationstrategy of each warehouse for a given user satisfaction rate, so as tominimize the total inventory cost. A more practical computer solving algorithmis proposed.

2. 研究的基本内容与方案

The basic content ofthis paper is to use the improved genetic algorithm to study the inventorymanagement of distributed warehouse: according to the inventory situation ofeach warehouse, when the total inventory falls to the total order point, howcan each warehouse jointly order from the coordination center to a supplier; whensome warehouse inventory falls to the order point, and the total inventory doesnot fall to the total order point, how can each warehouse adjust each other.

Genetic algorithm is an ordered globalstochastic optimization search algorithm, which avoids the possibility ofgeneral search algorithm falling into local optimum. By simulating biologicalevolution, the global optimal or approximate global optimal solution can beobtained, and the function to be optimized is basically unlimited. It requiresneither continuity nor differentiability. It can be an explicit functionexpressed by mathematical analytic expressions, a mapping matrix, or even animplicit function such as a neural network. It has strong adaptive and learningfunctions, and is also applicable to distributed inventory problem.

Inthis paper, genetic algorithm is used to construct and solve the distributedinventory problem. An improved genetic algorithm is proposed and its operationprocess and results are analyzed.

1. 张煜, 李文锋, 李斌. 基于动态联盟的虚拟企业的库存控制策略[J]. 计算机集成制造系统, 2008, 14(11).