【作者】 张劲松；

【导师】 李歧强；

【作者基本信息】 山东大学 ， 控制理论与控制工程， 2008， 博士

【摘要】 流程工业是国民经济的支柱产业,具有举足轻重的作用,因而国家投入了大量人力、物力、财力,对现有流程工业企业进行现代化改造,积极推广建立流程工业的CIPS(Computer Integrated Processing System,计算机集成过程系统)。而生产调度位于流程工业CIPS结构的中间层,是流程工业企业生产运行的指挥中心,提高生产调度的质量和效率对于提高流程工业企业的经济效益和社会效益起着重要的作用,因而对流程工业生产调度建模和优化方法进行研究具有重要的意义。流程工业总体上可以分为连续过程和间歇过程两种生产方式。对于连续过程而言,其原材料、半成品和产品通常是流体,加工过程长时间连续不间断地进行,如炼油、化肥、硝酸工业等。由于其生产对象不仅有物理变化而且有化学反应,因而其生产过程具有复杂性、不确定性、非线性和多约束、多资源相互协调等特点,其调度问题在数学上呈现高度耦合,求解非常困难。同时由于连续过程生产物流的连续性,缓冲余地小,这就对其生产调度的实时性、协调性、可靠性提出了更高的要求。因此本文以流程工业中连续过程的生产调度建模和优化方法为主要研究对象。对于复杂的连续过程调度问题,应用单一的模型往往无法加以正确的描述和分析,需要同时采用多种模型。正是基于这一思想,本文将基于受控混杂Petri网的过程仿真模型和基于数学规划的优化模型相结合,为连续过程的静态调度提出了一种“仿真+优化”的建模结构;将基于受控混杂Petri网的过程仿真模型、事件逻辑网、逻辑规划和经验规则相结合提出了一种基于受控混杂Petri网和事件逻辑网的连续过程动态调度建模方法。基于受控混杂Petri网的优化调度模型既能实现过程的直观显示和实验仿真,又能进行各种分析和计算,为解决连续过程的生产调度问题找到了一条有效的途径。为此,本文主要从以下几个方面进行了研究:1.连续过程以生产的连续运行为主,同时也具有方案切换、设备故障等离散事件,因而本质上是一个混杂系统。连续过程的仿真模型不但要反映生产过程中的混杂特性,还要能体现调度方案的实施对生产过程的影响。因而对现有混杂Petri网结构进行了扩展,并将对连续变迁和离散变迁的控制作用同时引入混杂Petri网,提出了一种既能反映连续过程生产特点又能满足生产调度建模需要的新型受控混杂Petri网。在受控混杂Petri网中,连续库所和连续变迁分别表示连续的物料流和加工装置的连续运行;离散库所和离散变迁分别表示加工装置的状态和离散事件的发生,连续控制库所和离散控制库所分别表示对连续变迁和离散变迁的优化控制。2.给出了该受控混杂Petri网的定义,对受控混杂Petri网的使能和激发规则及其动态特性进行了研究。详细介绍了在Matlab Simulink/Stateflow环境下建立受控混杂Petri网模型的方法。以受控混杂Petri网为工具,可以对流程工业中的连续过程建立仿真模型。该模型既能反映连续过程中生产连续的特性,又能体现出设备故障、设备修复等离散事件对生产的影响,不但为千差万别的连续过程提供了统一的模型表达形式,为调度优化模型的建立奠定了基础,而且对变迁控制作用的引入也为调度方案的实施和调度结果的验证提供了条件。3.提出了一种基于受控混杂Petri网的静态调度建模方法。在利用受控混杂Petri网对生产装置和存储装置分别建立仿真模型的基础上,建立整个连续过程生产系统的仿真模型。连续过程中的生产装置加工能力、存储装置存储能力、生产工艺、生产计划以及原料供应等约束条件都可以通过分析受控混杂Petri网模型得到,再结合一定的优化目标,便得到连续过程的静态调度模型。通过受控混杂Petri网对连续过程的仿真,既简化了对连续过程生产系统中复杂约束关系的分析过程,又能快速检验调度方案执行的效果,实现了连续过程的可视化建模。4.针对连续过程调度模型中普遍存在非线性的特点,提出了一种采用局部混沌搜索的混合粒子群优化算法。该算法以基本粒子群优化算法的运算流程作为主体流程,把混沌搜索机制引入其中,以此来增强全局搜索能力,摆脱局部极值点的吸引,提高了粒子群算法在非线性规划问题中求解的精度和收敛的速度,在对连续过程调度模型的求解中取得良好的效果。5.采用基于受控混杂Petri网的静态调度建模方法建立了某炼油厂的氢气平衡静态调度模型,考虑了氢气价格变动和装置检修计划对氢气利用的影响,并利用局部混沌搜索的混合粒子群算法对某个月内氢气平衡的静态调度问题进行了求解,既实现了氢气的生产和使用的平衡关系又能最大限度的降低氢气使用成本。6.针对在连续过程中经常出现的设备故障停机等突发事件,构建了基于受控混杂Petri网和事件逻辑网的动态调度模型结构。该模型结构采用受控混杂Petri网建立连续过程的仿真模型,以该仿真模型为基础建立起动态调度优化模型的约束条件,并根据生产过程中的经验规则,在约束条件中增加了逻辑变量,表示对约束条件的分枝选择,从而建立起逻辑规划形式的动态调度优化模型。为了实现从生产过程仿真模型中的突发事件向逻辑规划模型中逻辑变量的转换,定义了一种特殊的Petri网—事件逻辑网,它以受控混杂Petri网中的事件输出库所作为输入库所,以受控混杂Petri网的离散控制库所作为输出库所,中间通过逻辑变迁,实现逻辑命题的推理过程,从而可以响应受控混杂Petri网中出现的突发事件,并根据经验规则,对受控混杂Petri网中的受控离散变迁进行控制。同时事件逻辑网中的逻辑输出库所与逻辑规划模型中的逻辑变量相关联,可以实现对逻辑规划模型中的逻辑约束条件进行选择,实现优化模型的动态更新。逻辑规划模型的求解结果作为受控混杂Petri网连续控制库所的标识,实现对受控连续变迁的控制。受控混杂Petri网在离散控制库所和连续控制库所的共同作用下,在新的调度方案下稳定运行。7.以一个典型的化工生产过程为例,对基于受控混杂Petri网和事件逻辑网的动态调度建模方法进行了说明。为了使得该生产过程在有突发事件出现的情况下,尽量减少对其它装置的影响,使得整个生产系统可以继续稳定运行,我们总结了16条经验规则,并将其转换成事件逻辑网的形式,从而实现“事件触发—逻辑推理—控制输出”的自动过程。在逻辑规划模型优化目标的选取中,综合考虑了利润最大和系统长期稳定运行的要求。通过对不同调度模型的对比表明,该模型虽然使得短期生产效益略有下降,但在保证生产长期稳定运行方面具有明显的优势,这对于连续过程具有更重要的意义,因而该模型所得到的结果更具有可行性。最后,在总结本论文研究情况的基础上,提出了需要更进一步探索和研究的若干问题。更多还原

【Abstract】 Process industry is the mainstay of the national economy and acts as a very important function. So the nation has invested a lot of money, manpower and material resources to modernize the process industry enterprise in existence and the implementation of CIPS (Computer Integrated Processing System) is popularized. Production scheduling lies in the middle level of the structure of CIPS and it is the command center of the process industry production. Improving the quality and efficiency of production scheduling is very important to increase the economic and social benefit. So the research of the modeling and optimization method of production scheduling for process industry has significant meanings.Process industry is composed of two kinds of production process—continuous process and batch process. In the continuous process, such as oil refining, chemical fertilizer and nitric acid industry, the raw material, semi-product and final product are almost liquid and the production process is longtime continuous. Its production process has physical and chemical transformations and it has the characteristics of complexity, uncertainty, nonlinearity, multi-constrain and multi-resource collaboration. Its scheduling problem has highly coupling characteristic in mathematics and can be hardly solved. At the same time, the material flow in continuous process is continuous and the buffer is limited. So the production scheduling has more demand for the immediateness, coordination and reliability. Therefore this paper takes the modeling and optimization of continuous process production scheduling as the main research object.To solve the scheduling problem of complex continuous process, single model can not get correct description and analysis, so multi-model is needed. Based on this idea, this paper proposes a kind of "simulation plus optimization" modeling structure for continuous process static scheduling and proposes a dynamic scheduling modeling method based on controlled hybrid Petri nets and event logic nets which combines controlled hybrid Petri nets, logic programming and experience rules together. The scheduling modeling based on controlled hybrid Petri nets not only can realize the visualization and simulation of the production process, but also can do some analysis and computation. Therefore an effective method to solve the problems of continuous process production scheduling is found. This paper did some research from the following respects.1. Although continuous process mainly runs continuously, but it also has discrete events such as scheme switch and equipment failure, so it is a hybrid system in nature. The simulation modeling of continuous process not only need to reflect the hybrid characteristic in the production process but also need to embody the influence of the scheduling scheme for the production process. Therefore, this paper proposes a new kind of controlled hybrid Petri nets, which extend the existing hybrid Petri nets and introduce control places for continuous transitions and discrete transitions. This kind of controlled hybrid Petri nets not only can embody the characteristics of the continuou process, but also can fulfill the modeling requirements of the production scheduling. In the controlled hybrid Petri nets, the continuous places and the continuous transitions represent the continuous material flow and the continuous running of the equipments respectively. The discrete places and the discrete transitions represent the states of the equipments and the occurrence of discrete events respectively. The continuous control places and the discrete control places represent the optimal control for the corresponding continuous transitions and discrete transitions.2. The concept of this kind of controlled hybrid Petri nets is given. The rules of enabling and firing and the dynamic characteristic are researched. The method of constructing the controlled hybrid Petri nets under the environment of Matlab Simulink/Stateflow is introduced in detail. The simulation model of continuous process can be constructed by using controlled hybrid Petri nets. This simulation model not only can reflect the continuous characteristic of continuous process, but also can embody the impact of discrete events such as equipment failure and equipment restore for the production. This simulation model offers a uniform expression form for different kinds of continuous process and establishes the base for the scheduling model. The introduction of the control for the transitions offers conditions for the implement of the scheduling scheme and the validation for the scheduling results.3. A static scheduling modeling method based on controlled hybrid Petri nets is proposed. After constructing the simulation model of production equipments and storing equipments by using controlled hybrid Petri nets, the simulation model of the whole continuous process production system can be got. The constrains of the production ability of production equipments, the storing ability of storing equipments, the processing recipes, the production plan and the supply of raw materials can be got by analyzing controlled hybrid Petri nets. Combined with certain objective, the static scheduling model of continuous process can be got. Simulating the continuous process by using controlled hybrid Petri nets not only can simplify the analyzing procedure of the complex constrain relation in continuous process, but also can testify the implement effect of the scheduling scheme and visualize the modeling procedure.4. To deal with the nonlinearity characteristic commonly existing in the scheduling model of continuous process, a hybrid particle swarm optimization algorithm with local chaos search is proposed. This hybrid algorithm takes the basic particle swarm optimization algorithm as the main body. It combines the chaos search mechanism to enhance the search ability of the overall situation and gets rid of the attraction of local extremum. As the precision of the resolution and the speed of convergent are improved, this algorithm gets better effect in resolving the scheduling model of continuous process.5. A static scheduling model of hydrogen balance in a refinery is constructed by using the static scheduling modeling method based on controlled hybrid Petri nets. This model takes the influence of the hydrogen price and the repair plan of the equipments for the hydrogen utilization into account. The hybrid particle swarm optimization algorithm with local chaos search is used to resolve the static scheduling model of hydrogen balance in one month. The balance of hydrogen production and utilization is achieved, and at the same time, the cost of hydrogen utilization is reduced as much as possible.6. As the sudden events such as equipment failure often occur in continuous process, a dynamic scheduling model based on controlled hybrid Petri nets and event logic nets is constructed. In this model, controlled hybrid Petri nets are used to construct the simulation model of continuous process. The constraints of the dynamic scheduling model can be got from this simulation model. According the experience rules in the production process, some logic variables are added into the constraints to represent the switch choice of the constraints, and then the logic programming model can be got. A kind of special Petri nets called event logic nets is defined to realize the conversion from the sudden event in the production process simulation model to the logic variable in the logic programming model. It takes the event output places in controlled hybrid Petri nets as the input places and the discrete control places in controlled hybrid Petri nets as the output places. The ratiocination of the logic proposition is achieved through the logic transitions in the event logic nets. It can response to the sudden events occurring in controlled hybrid Petri nets and control the discrete transition in controlled hybrid Petri nets according to experience rules. At the same time, the logic output places in event logic nets are associated with the logic variables in the logic programming model. The corresponding logic constraints in the logic programming model are chose and the dynamic updating of the scheduling model is realized. The resolution of the logic programming model is taken as the marking of the continuous control places in controlled hybrid Petri nets to realize the control for the controlled continuous transitions. Under the control of the discrete control places and the continuous control places, controlled hybrid Petri nets run steadily in the new scheduling scheme.7. A typical chemical production process is taken as an example to illustrate the dynamic scheduling modeling method based on controlled hybrid Petri nets. To reduce the influence for other equipments when sudden events occur in the production process and make the whole production system running continuously and steadily, we summarize 16 experience rules and convert them into the form of event logic nets. This can realize the automatic procedure of "event occur—logic ratiocination—control output". The objective in the logic programming model takes the maximum of the profit and the long steady running of the system into account altogether. By comparison of different scheduling models, although this model reduces the short-term profit a little, it has obvious advantage in keeping the production longtime steadily running. This is more important for continuous process, so the scheduling results of this model are more feasible.Finally, the research of this paper is summarized and the problems that need to research next are put forward.更多还原