炼钢厂生产实时调度的动态网络建模及调度策略优化方法

The real-time scheduling problem in a steel plant is one of the critical technologies of manufacturing execution in a steel corporation. The modeling and optimization methods of the scheduling problem are also a challenging scientific problem in the scheduling theory. In this project, the real-time scheduling problem is described a progressive scheduling problem which contains three stages: scheduling,scheduling instructions determining and making scheduling strategy under disturbances. In order to realize the automation of scheduling in a steelmaking plant, the project focuses on the methods of modeling and scheduling strategy optimization with the comprehensive consideration of the objectives and constraints of manufacturing technique and scheduling, especially the constraint of hot metal supply and the influence of uncertainty in production process, based on the established scheduling model which can make a schedule by considering the a snapshot current production environment as initial constraints. With the core of establishing a dynamic network model of cellular automation based on scheduling task, evolutionary mechanism of self-organization and other-organization of the dynamic network are designed to depict the characteristics in real-time scheduling process. The cloud model based on data and the inference engine are provided to determine the processing time of scheduling task in a stochastic production environment. The research work will identify and classify mechanism of production disturbances, reveal the influence regularity of production disturbances, and establish the criterion to judge the invalid scope and degree of the schedule during the period of execution, to realize the automatic adjustment and human-computer collaborative optimization of scheduling strategy, which contains scheduling fine tuning, scheduling repair and re-planning, in real-time production process. The research is expected a great deal to enrich the scheduling theory and boost the core competition of iron and steel corporation in our country.

炼钢厂生产实时调度问题既是钢铁企业制造执行的关键技术，相关建模与优化方法也是先进制造领域调度理论的科学难题。项目将炼钢生产实时调度问题描述为调度计划排程、调度指令确定和扰动下调度策略的递进式调度问题，以实现炼钢生产调度"自动化"为目标，以已建立的基于调度实景"快照"的调度计划编制模型为基础，研究综合考虑生产工艺和调度目标及约束，特别是铁水资源约束、生产过程不确定性影响的建模与调度策略优化方法；以建立基于调度任务的元胞自动机动态网络模型为核心，设计表达实时调度过程特征的模型自组织和他组织演化机制；提出基于数据的云模型及推理机，实现调度任务作业时间的有限随机确定；研究生产扰动的实时识别及分类处理机制，揭示扰动影响规律，建立调度计划执行的"失效"范围和程度判据，进行调度微调、调度修复和重计划的重调度策略"自动"调整及人机协同优化决策。研究对丰富生产调度理论、提高我国钢铁企业核心竞争力有重要意义。

针对炼钢生产实时调度关键技术问题和实时调度问题优化建模求解的科学难题，项目将炼钢生产实时调度问题考虑为调度排程、调度指令确定和扰动下调度策略的递进式调度问题。以实现炼钢生产调度自动化为目标，在考虑生产工艺和调度目标及约束特别是铁水资源约束以及生产过程不确定性影响条件下，研究了炼钢全流程生产实时调度建模理论与重调度策略优化方法。在炼钢生产运行动态调控特征分析基础上，提出了实时调度过程中不确定性问题的描述方法；建立了能表达生产过程不确定性（如铁水随机到达、加工和运输时间不确定性、设备突发故障等）的调度排程与重调度优化模型；建立了基于调度任务的元胞自动机网络演化模型，设计了具有实时调度功能的网络演化模型的自组织和他组织演化机制，研究了上述模型和方法的实现技术及在炼钢厂 MES 系统中的应用方法，在此基础上实现了一套面向复杂钢厂生产实时调度优化系统的软件原型。该软件可给出调度排程、调度指令确定和扰动下重调度策略优化的递进式调度问题求解方案；研究了生产扰动的实时识别及分类处理机制，确定了调度计划执行的失效范围和程度判据，实现了调度微调、调度修复和重计划的重调度策略自动调整及人机协同优化决策。以典型复杂钢厂的探索性应用实例对实时调度理论方法及模型软件系统进行了检验。结果表明：模型方法和软件系统可有效给出炼钢实时调度及优化方案。另外，为获得未来模型方法的更好应用效果，项目还拓展研究了与实时调度问题相关的炼钢-连铸-热轧生产批量计划制定以及针对中厚板的组板组坯计划编制问题。项目工作已发表期刊学术论文15篇，参加国内外学术交流16次，申请发明专利7项，其中授权5项，获软件著作权1件。培养了4名硕士和2名博士, 另有在读博士研究生4人，硕士研究生2人。实现了项目预期研究目标，超出预期计划完成任务。研究成果对丰富生产调度理论、发展钢铁智能制造、提高我国钢铁企业核心竞争力具有重要的理论和实践价值。