基于规范和多Agent的煤炭供应链建模与仿真研究

发布时间：2019-01-14 15:22

【摘要】：近年来,煤炭供应链出现了库存增加、供求不平衡、对顾客需求反应迟钝等问题。面对动态开放的市场竞争环境,若想了解并快速满足消费者的需求,煤炭供应链上的各主体需要紧密联系彼此,在整条供应链上实现对煤炭资源的调运,即实现企业主体之间的良好协商。煤炭供应链中各个主体的行为变化影响着煤炭供应链的演化发展,整个系统形成了一个复杂适应系统。目前,研究复杂适应系统的一种有效途径是结合多Agent的理论对其进行建模和仿真。在该理论指导下,煤炭供应链中的各企业以及企业内部的各生产单元可看作动态变化的Agent智能主体,企业之间以及企业内部各生产单元之间的协商实际为Agent智能主体之间的动态协商。 Agent之间的协商机制主要通过Agent行为及资源的动态选择实现。合同网是实现Agent之间资源分配及复杂任务求解的一种经典协商策略。而对于MAS中主体行为协商过程的规范约束则依赖于一种有效的语义描述工具。组织符号学中的规范扩充了规则与惯例的语义表述范围,可以很好的描述Agent的行为特性,实现对系统中Agent行为的规范约束。Norm作为Agent掌握的一种知识,需要适应动态变化的复杂环境。而对煤炭供应链中企业Agent知识的学习和发现机制则依赖于一种有效的学习算法。GALCS是一种并行的、基于规则的、可自动更新的智能系统,可以实现对企业Norm的学习。论文的研究课题来源于国家自然科学基金项目—基于规范和多Agent的企业演化建模与仿真。本文在其基础上,分析其相关领域的国内外研究现状,结合规范和多Agent技术对企业Agent的协商机制进行了改进,主要内容如下： (1)结合基于遗传算法的学习分类器系统设计一种基于Norm学习的企业决策框架,通过对供应链采购环节Norm的相关举例,证明在该框架下设计的企业模型能够对企业Norm进行有效的学习和监督。 (2)结合对遗传算法和合同网的改进提出基于Norm的全局信任协商模型。通过对煤炭供应链内外部环境下协商模型的举例,证明模型可以从局部到全局对企业涉及到的各类协商议题进行任务求解。 (3)在仿真实验中利用了之前的研究。通过仿真实验,证明本文研究的Agent协商模型的有效性。
[Abstract]:In recent years, the coal supply chain has some problems, such as increasing inventory, unbalanced supply and demand, slow response to customer demand and so on. In the face of dynamic open market competition environment, if we want to understand and quickly meet the needs of consumers, the main bodies in the coal supply chain need to connect closely with each other to realize the transfer of coal resources in the whole supply chain. That is, to achieve good consultation between the main body of the enterprise. The behavior change of each main body in the coal supply chain affects the evolution and development of the coal supply chain, and the whole system forms a complex adaptive system. At present, an effective way to study complex adaptive systems is to model and simulate them based on the theory of multiple Agent. Under the guidance of this theory, the enterprises in the coal supply chain and the production units within the enterprises can be regarded as dynamic Agent intelligent agents. The negotiation between enterprises and production units is actually a dynamic negotiation between Agent agents. The negotiation mechanism between Agent is mainly realized by Agent behavior and dynamic selection of resources. Contract net is a classical negotiation strategy for resource allocation and complex task solving between Agent. The canonical constraints for the negotiation process of agent behavior in MAS depend on an effective semantic description tool. The specification in organizational semiotics expands the scope of semantic representation of rules and conventions, and can well describe the behavior characteristics of Agent and realize the canonical constraint on Agent behavior in the system. Norm is a kind of knowledge that Agent grasps. Complex environments that need to adapt to dynamic changes. However, the learning and discovery mechanism of Agent knowledge in coal supply chain depends on an effective learning algorithm. GALCS is a parallel, rule-based and automatic updating intelligent system, which can realize the learning of enterprise Norm. The research topic of this paper comes from the project of National Natural Science Foundation-Enterprise Evolution Modeling and Simulation based on Specification and multiple Agent. Based on it, this paper analyzes the current situation of domestic and foreign research in related fields, and improves the negotiation mechanism of enterprise Agent with the combination of specification and multiple Agent technology. The main contents are as follows: (1) combined with the learning classifier system based on genetic algorithm, an enterprise decision-making framework based on Norm learning is designed, and an example of Norm in supply chain procurement is given. It is proved that the enterprise model designed under this framework can effectively learn and supervise the enterprise Norm. (2) combined with the improvement of genetic algorithm and contract net, a global trust negotiation model based on Norm is proposed. Through the example of negotiation model in coal supply chain, it is proved that the model can solve all kinds of negotiation issues involved in enterprises from local to global. (3) the previous research is used in the simulation experiment. The effectiveness of the Agent negotiation model studied in this paper is proved by simulation experiments.
【学位授予单位】：宁夏大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F407.21;F274;TP18

【参考文献】