【作者】 李进；

【导师】 谭跃进；

【作者基本信息】 国防科学技术大学 ， 管理科学与工程， 2005， 博士

【摘要】 为了克服线性规划和单个处理单元非线性优化的内在缺点，本文研究了炼油过程系统操作优化的建模策略和优化算法，并应用于炼油厂全流程优化问题，主要研究内容与结论有： 1．在分析比较现有过程系统优化建模策略的基础上，借鉴大系统优化分解协调的思想，提出了一种新的过程系统优化建模与求解策略——两级分解协调法。该方法把模型分成两个层次，系统层和过程层，系统层决定各单元的物流和连接关系，过程层对每个处理单元进行优化。使用该方法，过程层每个过程单元的模型都可以根据实际需要更换，整个模型具有很好的柔性。 2．针对炼油过程系统优化自由度较小的特点，研究了自由度较小的非线性规划问题的求解算法，实现了一种基于简约空间序列二次规划和自动微分的优化算法。实验数值结果表明，该算法整体效率较高，适合求解大规模过程优化问题。 3．参考宣(Xuan)的原油蒸馏单元模型，建立了某燃料型炼油厂原油蒸馏单元的稳态模型，模型中包括常压塔和减压塔的详细模型，以及再蒸馏单元、渣油溶剂抽提单元、脱丁烷塔和石脑油分离器等原油辅助过程单元的简化模型。把该模型与严格机理模型的预测结果相比较，验证了该模型能够准确预测常压塔、减压塔侧线产品以及减压渣油的体积和性质。 4．参考埃利斯(Ellis)的催化裂化单元模型和塔斯卡(Taskar)的催化重整单元模型，针对具体的燃料型炼油厂进行修改，研究了催化裂化单元和催化重整单元的进料表征，并用炼油厂的工业数据对模型进行校准，使其更符合实际情况。总结了集总动力学模型建立的普遍方法与原则。 5．建立了气体处理单元、烷基化单元和柴油加氢处理单元的简化模型。研究了辛烷值调合模型、雷特蒸气压调合模型和馏出温度调合模型，并在此基础上建立了汽油调合单元模型。 6．研究并给出了炼油厂全流程优化问题的目标函数、决策变量和约束条件，并用实现的新算法进行求解，找到了最佳操作条件。与炼油厂现有操作条件相比，利润提高4.1％；与单处理单元优化相比，利润提高1.6％。对新算法和序列二次规划求解器rSQP的性能作了比较，计算结果表明，新算法可以高效求解炼油过程系统优化问题。更多还原

【Abstract】 In order to overcome the inherent disadvantages of line programming and single-unit nonlinear optimization, the modeling approach and optimization algorithm of refining system and their applications in refinery-wide optimization are studied in this dissertation. The main contents are as follows:1. Using the decomposition-coordination theory of large-scale system for reference, a novel decomposition strategy for process system optimization, namely two level decomposition-coordination approach, is presented based on the comparison and analysis of current modeling approaches. This approach decomposes the overall model into two levels, namely a system level and a process level. The system level determines common issues among processes; the process level optimizes individual processes. In this way, this approach provides the flexibility of replacing existing process models without affecting the optimization infrastructure.2. For refining process optimization has the characteristics of few degrees of freedom, an efficient optimization algorithm based on reduced sequential quadratic programming and automatic differentiation is presented in this paper. The experiment results show that the new algorithm calculate efficiently and is especially suitable for large-scale process optimization.3. Using Xuan’s model of crude distillation unit for reference, a first-principle, nonlinear, crude unit model was developed. The model, which is a steady-state model based on material balance and energy balance, include detailed model of Atmospheric Tower and Vacuum Tower and simplified model of some auxiliary process, such as Rerun unit, Residue Oil Solvent Extraction Unit, Debutanizer, and Naphtha Splitter. The comparison of the developed model and rigorous model shows that this model calculates the yields and properties of the products accurately based on the feed information.4. According to the specific fuel-oriented refinery, Ellis’ model of fluidized catalytic cracking (FCC) unit and Taskar’s model of catalytic reformer (CR) unit were used in this work after modification. The feed characterization of FCC and CR unit are studied. Each single-unit model was benchmarked against the industrial data obtained from the refinery. Common approaches and principles of lumping kinetic model research are also discussed.5. Simplified models were developed for gas plant, alkylation unit and diesel hydrotreater. Gasoline blending model was developed based on the studies on octane blending model, vapor pressure blending model and percent distilled model, which calculates the gasoline specifications of three grades of gasoline from the information of gasoline blending stocks.6. The objective, decision variables and constraints of refinery-wide optimization are given更多还原