【作者】 石成华；

【导师】 王雅鹏；

【作者基本信息】 华中农业大学 ， 农业经济管理， 2014， 博士

【摘要】 提供农资送货上门服务,是目前农资企业争取客户的重要竞争手段,直接关系到企业的农资销量和市场占有率。面对我国农户地理位置分散、需求量少、配送时间不集中等现状,“最后一公里”的车辆配送占农资企业配送成本和配送时间的一半以上,如何安排农资车辆配送路线直接影响到配送成本和服务水平。同时,由于农资客户需求的不确定性,受到农民的群体购买行为、左邻右舍的比较因素、以及作物的种植面积、天气因素等的影响,在进行农资配送过程中会经常发生潜在需求。如果将这些潜在农资需求提前装载在物流配送的车辆上,可以减少配送车辆的发货次数,进而减少配送成本,提高企业的配送效率,增加企业的经济收益。另外,在农资车辆配送过程中,往往存在各种干扰事件：农资需求量方面,如农户增加或减少需要的农资数量、农户因某种原因取消原来农资订单、出现新的农资需求用户等；配送时间及地址方面,如农户因临时有事等原因改变接货时间、农户因运输需要改变配送的地址等；配送车辆方面,存在车辆损坏、道路堵塞等干扰事件。如何在这些干扰事件发生后,生成新的农资车辆配送路线是一个非常有挑战的问题。基于以上分析,本文拟研究考虑潜在需求的农资车辆调度及其干扰管理,主要包括潜在农资需求预测方法、考虑潜在需求的农资车辆调度优化模型与算法,以及考虑潜在农户的农资车辆调度干扰管理模型三个方面。本文的贡献和可能的创新之处有四点：(1)针对农资配送过程中潜在农资需求的影响因素复杂、各影响因素维度不一和预测指标数据难以获得等难点,本研究首先根据全面性原则、目的性原则、实用性原则和可操作性原则,从农资产品、配送区域农户和农资企业三个方面识别出了影响潜在农资需求的16个具体指标,然后利用粗糙集的遗传算法约简法则,提出基于粗糙集的潜在农资需求预测指标约简方法,把潜在需求预测指标约简到6个指标,最后在此基础上,构建了基于RS-SVM的潜在农资需求预测方法,可以对农资配送过程中的潜在需求进行有效预测,进而能够为考虑潜在农资需求的车辆路线优化提供了基础。(2)考虑到在实际农资车辆配送过程中由于农户间的购买行为相互影响,本研究首先对考虑潜在农资需求的农资车辆调度问题进行描述,并给出模型假设,构建出了带有软时间窗的考虑农资需求车辆路线优化模型,接着通过采用遗传算法对约束满足算法的优化机制进行优化,提出了基于CSGA的模型求解算法,并通过数值试验验证了模型和算法的有效性,能够使得农资企业提前将潜在农资需求装载在配送车辆上,可以减少配送车辆的发货次数,进而减少配送成本,提高企业收益。(3)针对农资配送过程中可能发生的各种干扰事件,本研究首先构建了新增农户十扰事件识别与度量方法,从服务农户的时间干扰、农资配送司机的路线干扰和农资配送商的成本干扰三个方面并对新增农户干扰事件对原计划带来的负面影响进行定量化,并提出其他类型干扰事件转换成新增农户干扰事件的方法；接着提出了农资配送车辆最优出发时间的确定方法,在此基础上,并根据对干扰事件识别与度量的基础上,建立了考虑潜在农户的农资车辆调度干扰恢复模型,并设计了模型求解的遗传算法和嵌套分割算法,最后数据实验验证了构建模型与算法的有效性。(4)以邯郸市永年县某农资有限公司为实际应用研究对象,按照产品类型和重量选取邯科玉1号、邯682、沃尔森玉米施用肥和银棉棉花施用肥四种农资产品作为实际研究对象,首先分别采用提出的基于RS-SVM的潜在农资需求预测方法对该四种农资产品的100次配送过程中的潜在需求进行了预测,通过与实际发生的新增需求对比,验证了方法的有效性；接着根据该企业每次配送农户需求数据,并考虑潜在农资需求量,利用建立的考虑潜在农资需求量的车辆调度优化模型及算法对该企业的50次配送路线进行优化,为该企业的配送路线进行了有效优化；最后,通过对该企业100次配送过程中的干扰事件进行统计,选择出5次邯科玉1号配送过程为例,为该企业进行实时配送路线调整,并为该企业提出了冗余装载策略、车辆差异化策略、延迟发货策略和实时调整策略,可以有效降低企业配送成本。本文的主要内容安排如下：第一章首先给出本论文的研究背景及意义、研究问题和目标,介绍了本文的研究思路、技术路线和主要研究内容,以及研究方法与实验手段。第二章详细地论述和评价了相关理论基础：粗糙集理论、支持向量机原理和支持向量机回归、遗传算法原理。全面、系统地对的农资物流配送与客户挖潜、粗糙集与支持向量机、车辆调度优化、干扰管理文献进行回顾与评述。第三章分别从农资本身、农户和配送企业方面构建农资潜在需求预测的初始指标体系,根据预测指标体系建立了基于RS-SVM的农资物流配送过程中潜在农资需求预测模型,根据实际调查数据,使用粗糙集获得了约简后的潜在农资需求预测指标体系,并使用实际数据验证了基于RS-SVM潜在农资需求量预测方法的有效性。第四章考虑到在实际车辆配送过程中由于客户间的购买行为相互影响,建立考虑潜在客户需求的农资车辆调度优化数学模型,并针对该模型在标准遗传算法和约束满足技术的基础上,提出了一种新的约束满足-遗传算法用于求解所建立多约束数学模型,最后通过仿真实验验证所提出模型的有效性和算法的可行性。第五章针对配送过程中可能发生的各种干扰事件,首先从服务农户的时间干扰、配送司机的路线干扰和农资配送商的成本干扰三个方面并对新增农户干扰事件对原计划带来的负面影响进行定量化,并提出其他类型干扰事件转换成新增农户干扰事件的方法；接着提出了农资配送车辆最优出发时间的确定方法,进而建立了考虑潜在农户的农资车辆调度干扰恢复模型,设计了模型求解的遗传算法和嵌套分割算法。数据实验验证了构建模型与算法的有效性。第六章以邯郸市永年县某农资有限公司为实际应用研究对象,按照产品类型和重量选取邯科玉1号、邯682、沃尔森玉米施用肥和银棉棉花施用肥四种农资产品作为实际研究对象,进行实际应用研究,为实际企业提出具体配送策略。第七章总结论文研究结论。更多还原

【Abstract】 Providing door-to-door services in agricultural materials delivery is an important competition means for agricultural materials enterprise to acquire customers, which is directly related to the agricultural materials sales and market share of enterprises. Due to the current situations in farmers in China, for example, dispersed geographically, less demand, different delivery time, it make "the last kilometer" vehicle distribution of agricultural materials account for the half of the shipping cost and delivery time. Therefore, how to arrange for agricultural materials vehicle delivery route directly affects the distribution cost and service level. At the same time, because of the demand uncertainty of the agricultural materials, the affect from the farmer group purchasing behavior, the comparison of neighbors, the crop planting area, and the influence of weather factors, the potential demand often occur in the process of agricultural materials distribution. If these potential agricultural materials requirements can be loaded in advance on the logistics distribution vehicles, the number of shipments for delivery vehicle and the distribution cost may reduce, improving the distribution efficiency of enterprises and increasing enterprise economic benefits. In addition, in the process of agricultural materials vehicle distribution, there are often various kinds of interference events:firstly, in the demand of agricultural materials, such as farmers increase or decrease the amount of agricultural materials demand, farmers cancel the original agricultural materials orders for some reason and the new customers emerges, etc.; secondly, in the delivery time and address, such as farmers change the receiving time by reason of temporary something and farmers change the delivery address because of transport need, etc.; in the delivery vehicles, such as vehicle fails, road blockage, etc.. How to generate new agricultural materials vehicle distribution route after these interference events occur is a very challenging problem.Based on the above analysis, this paper studies the agricultural materials vehicle scheduling and interference management considering the potential demand, which mainly includes the forecasting method of potential agricultural materials demand, the agricultural materials vehicle scheduling model and algorithm considering the potential demand, and the agricultural materials vehicle scheduling interference model based on the potential demand. The contributions of this paper and the possible innovation places are as follows:(1) Aiming at some difficulties in the process of agricultural materials distribution, such as the complex factors affecting the potential agricultural materials demand, different influence dimensions and the forecast indicators data is difficult to obtain, this study is based on overall principle, objective principle, the principles of practicability and operability, identifies16specific indicators affecting the potential agricultural materials demand from products, customers and enterprises three aspects. Then using the reduction rule of rough set (RS), the reduction method of potential agricultural materials demand forecasting index based on rough set is presented.6indicators is obtained based on the proposed reduction method and finally, the agricultural materials potential demand forecasting method based on RS-SVM (Rough Set Support Vector Machine) is constructed to effectively predict the potential demand in the process of agricultural materials distribution that provides the basis for the vehicle route optimization considering potential agricultural materials demand.(2) Considering the buying behavior influence between farmers each other in the process of the actual agricultural materials vehicle distribution, this study first describes the potential agricultural materials vehicle scheduling problem by considering the potential demand, then the model assumptions are given, and the agricultural materials vehicle route optimization model with soft time windows considering the potential demand is established. The CSGA algorithm (Constraint Satisfaction-Genetic Algorithm) is developed by optimizing the optimization mechanism of constraint satisfaction algorithm using genetic algorithm. Finally, Numerical examples verify the effectiveness of the proposed model and algorithm that make the agricultural materials enterprises load the potential demand in advance on the distribution vehicles such that it helps to decrease the number of shipments for delivery vehicles and the distribution cost, improve enterprise benefits.(3) For all kinds of possible interference events in the process of agricultural materials distribution, this study constructs the new farmers interference event recognition and measurement method, then quantifies the negative effects of the new farmers interference event on the original plan in farmers service time interference, agricultural materials distribution route interference and the cost of agricultural materials distributors three aspects, and presents the method converting other types of interference events into the new farmers interference event. In the following, the method to find the optimal departure time of the agricultural materials distribution vehicle is presented, based on which the agricultural materials vehicle scheduling interference recovery model is established and the genetic algorithm and nested segmentation algorithm is designed for solving the model. Finally, the numerical simulation illustrates the effectiveness of the proposed model and algorithm.(4) Taking an agricultural materials co., LTD in Yongnian handan city as the study object, Hankeyu1, Han682, Warsun Corn applied fertilizer and Silver cotton applied fertilizer four kinds of agricultural materials products are chosen as the actual research objects in this study according to the product type and weight. Firstly, the potential agricultural materials demand forecasting method based on RS-SVM is used to predict the potential demand of these four agricultural materials products during100times distribution process, and the method is proved to be effective by comparing the predicting results with the actual demands. Then, in terms of the farmer demand data in the distribution process and considering the potential demands, the50times distribution route in this case is optimized by adopting the agricultural materials vehicle route optimization model and algorithm; finally, by the interference event statistics in100times distribution process and choosing5times distribution process of Hankeyu1as an example, the real-time distribution route adjustment is shown, also the redundant loading strategy, differentiation strategy, late delivery vehicle and real-time adjustment strategy are proposed to reduce the enterprise distribution costs.The main contents of this paper are organized as following:Chapter1firstly introduces the research background and significance, research questions and objectives of this work, then the research thought, the technical route and the main research contents, research method and experimental means are also introduced briefly.Some related theoretical foundations are reviewed in detailed in Chapter2, for example, rough set, support vector machine and support vector machine regression, and genetic algorithm. Moreover, the review of the previous work on agricultural materials logistics distribution and finding the new customers, rough set and support vector machine, vehicle scheduling optimization and interference management are made comprehensively and systematically.In Chapter3, the initial index system of the agricultural materials potential demand prediction is constructed in agricultural materials itself, farmers and distribution enterprises three aspects; then in the light of the index system, the agricultural materials potential demand forecasting model based on RS-SVM is developed. According to the real-world data, the agricultural materials potential demand forecasting index system after reduction by rough set is obtained and the effectiveness of the agricultural materials potential demand forecasting method based on RS-SVM is illustrated by simulating the collected real-world data.Chapter4considers the buying behavior influence between farmers each other in the process of the actual agricultural materials vehicle distribution, establishes the agricultural materials vehicle route optimization model considering the potential demand. Based on standard genetic algorithm and constraint satisfaction technique, a novel CSGA algorithm is developed to solve the multi-constrains mathematical model and finally, simulation results show that the proposed model and algorithm is effective.In Chapter5, by aiming at all kinds of possible interference events in the process of agricultural materials distribution, this study quantifies the negative effects of the new farmers interference event on the original plan in farmers service time interference, agricultural materials distribution route interference and the cost of agricultural materials distributors three aspects, and presents the method converting other types of interference events into the new farmers interference event. In the following, the method to find the optimal departure time of the agricultural materials distribution vehicle is given, also the agricultural materials vehicle scheduling interference recovery model is then established; and the Genetic algorithm and nested segmentation algorithm is designed for solving the model. Finally, the numerical simulation illustrates the effectiveness of the proposed model and algorithm.Chapter6takes an agricultural materials co., LTD in Yongnian handan city as the study object, Hankeyu1, Han682, Warsun Corn applied fertilizer and Silver cotton applied fertilizer four kinds of agricultural materials products are chosen as the actual research objects in this study according to the product type and weight. The case study is finished based on the proposed methods and algorithms which provide the concrete the distribution strategy for enterprises.The conclusions are remarked in Chapter7.更多还原