考虑动态客户需求的物资配送问题求解方法

A Strategy for Solving Material Distribution Considering Dynamic Customer Demand

为明确需求预测方向和减少预测偏差,提出了一种考虑动态客户需求的物资配送问题求解方法。依据客户历史需求,以泊松分布模拟需求变化情况,建立需求预测模型,得到客户初始预测需求,并在此基础上建立了需求不确定的物资配送模型。为求解该模型,设计了一种预测需求可调节的遗传算法。在遗传算法局部搜索阶段,提出了需求调节算子。该算子以一定概率对客户初始预测需求进行调节,以符合泊松分布的需求变化量减少预测需求,同时对车辆间的顾客进行调整,顾客以调节后的预测需求挑选最小需求客户,移动该客户至其他最低载货车辆中并调整该车辆服务的顾客顺序,整个调节过程以最低配送成本为标准,保留最优配送路径。从标准车辆路径问题测试数据库中挑选10个典型样例进行测试,结果表明,与经典的最近邻算法和遗传算法对比,所提算法在所有算例中均取得了总成本最小值,在90%的算例中取得了运输成本最小值,在70%的算例中取得了车辆成本最小值。

To clarify the direction of demand forecast and reduce the forecast deviation,a strategy considering dynamic customer demand is proposed for material distribution.Following the historical demand of the customers,a Poisson distribution is used to simulate the changes in demand,and a demand forecasting model is established to obtain the initial forecasted demand for the customers.Then a material distribution model with uncertain demand is constructed.To solve the distribution model,a genetic algorithm with adjustable forecasting demand(GAAFD)is designed.In the local search stage of GAAFD,a demand adjustment operator is introduced.This operator adjusts the initial forecasted demand with a certain probability to obey Poisson distribution of the demand changes so that the forecasted demand can be reduced,and it simultaneously adjusts the customers between vehicles.The customers are moved to the other lowest cargo vehicle and sequenced for the vehicle service.The lowest delivery cost is treated as the purpose and the optimal delivery routes are saved.To verify the proposed GAAFD,ten typical test instances are selected from the benchmark test database of the vehicle routing problem,and compared with the representative algorithms,i.e.the classic nearest neighbor algorithm and genetic algorithm without the adjusting operator.The results show that for the ten test instances,GAAFD reaches the minimum total cost obtained in all instances,and in 90%of the instances the minimum transportation cost is achieved,in 70%of the instances the minimum vehicle cost is achieved.