【作者】 李明；

【导师】 李歧强；

【作者基本信息】 山东大学 ， 控制理论与控制工程， 2011， 博士

【摘要】 炼油行业是连续过程工业的典型代表,是国民经济的支柱行业,具有举足轻重的作用。生产调度作为炼油企业综合自动化系统的中间环节,是企业生产经营的指挥中心,提高生产调度的质量和效率对于提高企业的经济效益和社会效益具有重要的作用。因而,对炼油过程生产调度的研究具有重要意义。炼油过程具有生产规模大,工艺复杂,物流连续运作,缓冲余地小,生产连续、平稳,装置多且衔接紧密等突出特点。生产调度模型的建立是炼油过程调度优化问题的研究重点,也是当前优化领域的研究热点之一。由于炼油过程上述特点的存在,其生产调度问题在数学上呈现高度耦合,求解困难,某些因素难以用传统的数学模型表达。鉴于已有调度建模方法存在的种种不足,目前的炼油过程生产调度对炼油生产的特有性质和生产中的逻辑规则往往难以进行全面、有效地描述,建立的调度模型较复杂,超结构性强,直观性、实用性较差,难以满足实际调度需求。这一现状使得我国炼油行业目前大多采用基于经验规则的人工调度,不但费时费力,而且难以保证调度优化性。因此,如何在现有调度理论和研究成果的基础上,针对炼油生产的特有性质,寻找适合于炼油过程的生产调度建模优化方法,辅助企业调度人员实现优化排产,是目前调度优化研究中的一个热点也是一个难点问题,也是本课题的研究目的所在。本文在介绍了炼油过程生产调度问题及其研究现状之后,针对炼油过程特点及其生产调度需求,提出了几种调度建模新方法和调度优化模型以解决炼油过程的静态调度和动态调度优化问题。在此基础上设计开发了炼油过程动态调度优化软件。该软件已在中石化齐鲁分公司胜利炼油厂的生产实际中试用,获得了较好的效果。具体研究内容和成果概括如下：(1)针对基于逻辑的建模方法在模型表达的灵活性和求解优化性等方面的不足,为有效表达生产中的启发式逻辑规则,提出了一种集成启发式规则的混合整数规划调度优化模型。在分析启发式规则逻辑关系的基础上,用布尔逻辑表达启发式规则,给出了启发式规则的基本逻辑表达式及其等价代数表达。将启发式规则的代数表达与混合整数规划相结合,实现了启发式规则的集成。集成启发式规则的混合整数规划调度模型在数学优化的基础上能够方便、灵活地表达启发式逻辑规则,利用启发式规则有效地描述调度问题,简化调度模型,提高求解效率,保证调度优化性。(2)针对炼油过程生产特性,研究了其静态调度优化问题,建立了集成启发式规则的混合整数规划调度优化模型。提出了平稳性、连续性和长期性等生产特性定义及其性能指标。基于炼油生产中的逻辑规则,对生产特性进行了模型化描述,并对生产特性性能进行了定量化表达和评价,以实现生产利润和生产特性性能的综合利益最大化,满足生产调度实际需求。以某炼厂生产调度为例,对比有无生产特性表达的优化调度仿真结果表明：考虑炼油过程生产特性的调度优化模型能够满足生产特性要求,提高生产特性性能,得到的调度方案更为可行、实用。(3)炼油过程以生产的连续运行为主,同时也具有装置启停、方案切换、设备故障等离散生产事件,其本质上是一个混杂生产系统。系统中的生产事件体现了炼油生产过程的运行演变。调度专家通过对生产事件的合理安排和处理实现生产调度。据此,针对炼油过程的静态调度问题,提出了一种基于预知事件的调度优化建模方法。首先定义和描述了炼油生产调度中的预知事件,进而定义了一种嵌套混合自动机,在此基础上建立了基于预知事件的调度优化模型。为实现模型的有效求解,将基于预知事件的调度模型转换为等价的混合整数规划模型,从而能够方便地应用成熟的优化算法,保证调度优化性。基于预知事件的调度优化模型具有直观、模块化、超结构性弱的特点,为炼油过程静态调度问题提供了一种新的模型描述方法。(4)针对集成启发式规则的混合整数规划调度模型表达较复杂、直观性差、具有较强超结构性的不足,利用基于预知事件的调度优化建模方法,对考虑炼油过程生产特性的调度优化问题进行了重新表达,建立了基于预知事件的调度优化模型。模型表达更直观、模块化性强,易于建立。针对同一仿真实例,获得了与集成启发式规则的混合整数规划调度模型相同生产特性性能的调度方案,满足调度实际需求。(5)针对炼油过程动态调度的实时性和优化性需求难以满足的问题,基于反应式局部重调度策略,提出了一种基于突发事件的调度优化建模方法。定义和描述了炼油生产中的突发事件及其事件逻辑,通过事件逻辑可以体现并灵活选择调度专家处理突发事件的经验规则,体现突发事件和专家经验对于生产过程的影响。在此基础上,给出了一种分层受控混合自动机,建立了基于突发事件的炼油过程动态调度优化模型。通过两阶段事件触发机制响应突发事件,利用生产流程仿真模拟确定动态调度的流程范围,生成动态调度子模型；通过对子模型的求解最终得到响应突发事件的动态调度优化方案。基于突发事件的炼油过程动态调度有效结合了仿真方法、启发式方法和严格的数学优化,在满足动态调度实时性的同时保证了调度优化性。在上述研究基础上开发了炼油过程动态调度优化软件。该软件已在中石化齐鲁分公司胜利炼油厂试用并获得好评。通过软件的实际应用案例验证了基于突发事件的动态调度优化模型的有效性和实用性。更多还原

【Abstract】 As representative of continuous process industry, refinery production enterprises are the mainstay of national economy and act very important role. Production scheduling is the intermediate link of Computer Integrated Processing System and is the core of business operations. Improving quality and efficiency of production scheduling is very important to increase economic and social benefit of refinery companies. So, research on refinery production scheduling has significant meanings.Refinery process not only has large production scale, but also has complex production technology and little cushion, whose logistics operates continuously. There are many salient features of refinery process, such as continuity, stability, and close connection between refining units, etc. Establishing scheduling model is the emphasis of refinery scheduling research, which is also one of the hot topics of combinatorial optimization research. Due to the above features, refinery production scheduling problem presents highly coupling in mathematics, which is hard to describe and solve by traditional mathematic models. By existing scheduling modeling methods, it is difficult to represent the unique natures and heuristic rules of refinery process comprehensively and efficiently. The scheduling models are usually very complex and have strong super structure, poor intuition and weak practical applicability. Due to the current situation of refinery scheduling research, refineries in our nation usually adopt artificial scheduling base on expert experience, which is not only time-consuming and taxing but also hard to guarantee scheduling optimization. So, there have profound academic and realistic meanings to find suitable scheduling modeling methods in view of the unique natures of refinery process based on current scheduling theory and research results. The research and development of such methods is a difficult and hot point in scheduling research area and also is the purpose of our research topic.This paper proposes several new modeling methods and detail models for static and dynamic scheduling of refinery process in view of its features and production scheduling requirements. Based on the proposed methods, a refinery dynamic scheduling system had been developed and utilized in Sinopec Shengli refinery, which had obtained well effects. Detailed research contents and outputs are summarized as follows:(1) In view of the shortcomings of logic-based modeling method on flexible expressing and solving optimality, a heuristic rule-integrated mixed integer programming model is proposed for scheduling optimization problem. Heuristic rules are expressed by logic variables, and their basic logic expressions and corresponding equivalent algebra expressions are presented, by which heuristic rules can be formulated. Combining the algebra expressions with mixed integer programming, integration of heuristic rules is realized. The proposed model can desceibe heuristic rules easily and flexibly to facilitate formulating and simplifying scheduling problem based on mathematical optimization, which can improve solution efficiency and guarantee optimization performance.(2) The refinery static scheduling problem in view of production characteristics is then studied. A heuristic rule-integrated mixed integer programming model is set up based on given definitions of stability, continuity and chronicity as well as their performance criterions which evaluate production characteristics quantitatively. The model describes production characteristics by logic rules of refinery process, and optimizes the trade-off between production profit and characteristic performances. The proposed formulation was used to address scheduling of a certain refinery. Numerical results obtained from the scheduling considering production characteristics whether or not show that schedule of the production characteristic-based scheduling model can improve the performance of production characteristics and is more feasible and practical.(3) Although refinery process mainly runs continuously, but it also has discrete production events such as unit start and stop, scheme switch and equipment failure, so it is a hybrid system in nature. Production events of the system reflect operations of refinery process, which are arranged and handled by scheduling experts to realize production scheduling. Based on the cognition, a prescient event-based modeling method for refinery static scheduling problem is proposed. By defining and describing prescient events of production scheduling, a coupled hybrid automaton is defined, based on which a prescient event-based scheduling model is set up. To realize optimal scheduling, the proposed model is converted to mixed integer programming model equivalently, by which mature optimization algorithms can be utilized conveniently and scheduling optimality is guaranteed. The prescient event-based scheduling model has advantages of intuition, modularization and weak super structure, which provides a new modeling method for refinery static scheduling problem.(4) Based on the prescient event-based modeling method, the scheduling problem considering refinery production characteristics is reformulated, and a prescient event-based scheduling model is set up. The model is easy to build and is more concise in contrast to the heuristic rule-integrated mixed integer programming model. Same schedule was obtained by dealing with refinery scheduling example of the heuristic rule-integrated mixed integer programming scheduling model, which satisfies practical scheduling requirements.(5) To deal with difficulties in satisfying refinery dynamic scheduling requirements of real-time and optimality, an emergency event-based scheduling modeling method is proposed on the grounds of reactive local rescheduling strategy. Emergency event and its event logics are defined, by which expert experience can be chosen flexibly to affect production process. A hierarchical controlled hybrid automaton is given, based on which an emergency event-based refinery dynamic scheduling model is built up. The final dynamic scheduling sub model responding to emergency event can be obtained by a proposed two stage event trigger mechanism, by which dynamic scheduling region can be determined through process simulation. Solving the sub model, dynamic scheme is got. The emergency event-based dynamic scheduling utilizes simulation method, heuristic method and rigorous mathematical optimization effectively, which satisfies scheduling requirements of real-time and optimality. A refinery dynamic scheduling system developed based on the above research had been utilized in Sinopec Shengli refinery and received high praise. Finally, the effectiveness and practicality of the proposed dynamic scheduling model was validated by a realistic scheduling case.更多还原