【作者】 李朝春；

【导师】 钱锋；

【作者基本信息】 华东理工大学 ， 控制科学与工程， 2014， 博士

【摘要】 二氯乙烷裂解炉是氯乙烯乃至聚氯乙烯工业生产中的重要设备和用能大户。因此,研究如何降低二氯乙烷裂解炉中的能耗,提高裂解产物质量,有着非常重要的作用。由于现场缺乏足够的检测设备,二氯乙烷裂解炉现场操作水平不高,唯一标识二氯乙烷裂解性能的控制指标为炉管出口温度。因此,工业二氯乙烷裂解炉的操作优化还有很大的提升的空间。为了加深对二氯乙烷裂解炉本质的认识,定量分析现场无法实际测量的关键变量信息,实现二氯乙烷裂解炉的操作优化,建立准确的二氯乙烷裂解炉模型显得非常关键。为此,本文从不同层面出发,建立多个不同复杂度的二氯裂解炉模型,并设计一种高效的多目标算法,实现二氯乙烷裂解炉的多目标操作优化。工作主要的创新点总结如下：(1)设计了一种带约束处理机制的自适应差分进化多目标优化算法(SMODE-εCD)。算法在差分进化的变异和交叉过程中引入了自适应调整策略,从而有利于进化前期保持种群的多样性,进化后期收敛到全局最优。此外,算法引入ε约束处理机制,用于处理多目标问题的约束,并且设计了外部存档集,保存非支配关系的最优解,用于引导多目标算法的进化。通过对9个标准约束多目标问题以及4个工程设计优化问题的仿真测试,并与典型的多目标优化算法NSGA-Ⅱ和CMODE算法作对比,有效验证了SMODE-εCD算法在处理各种多目标约束问题的有效性,为二氯乙烷裂解炉性能指标的多目标优化奠定了算法基础。(2)建立二氯乙烷裂解炉清洁管时期的简易机理模型。炉膛采用一维Lobo-Evans法描述炉膛烟气传热过程,管内二氯乙烷的裂解以一维平推流假设、裂解反应动力学采用一阶串联的分子反应模型。在该模型的基础上,通过定义燃料气负荷分配因子a,分析不同的燃料气负荷分配策略对二氯乙烷裂解炉中关键的变量以及重要性能指标的影响。此外,通过调整燃料气用量,得到炉管出口温度与裂解性能指标如转化率、选择性、单耗之间的对应关系,用于指导现场操作。(3)建立了二氯乙烷裂解炉全周期模型。在清洁管时期模型的基础上,通过引入自由基反应机理和裂解结焦模型,建立二氯乙烷裂解炉全周期模型。乙炔作为唯一的结焦母体,结焦过程作拟稳态假设,炉管外壁最高温度为裂解炉全周期模拟结束的制约条件。通过二氯乙烷裂解炉全周期模拟,定量分析结焦过程对于二氯乙烷裂解炉关键信息的影响,包括炉管内径,炉管外壁温度,炉管传热速率,炉膛烟气温度,二氯乙烷裂解转化率、选择性、单耗随着运行时间的推进的变化情况。此外,由于二氯乙烷原料的纯度以及炉膛内燃料气负荷分配策略对二氯乙烷裂解结焦过程有着重要的影响,本文在全周期的模型的基础上,定量分析二氯乙烷原料中典型的反应促进剂CCl4的浓度以及燃料气负荷分配因子a对于二氯乙烷结焦过程以及全周期裂解性能指标的影响,并给出合理的CCl4浓度以及燃料气负荷分配策略建议。(4)建立了二氯乙烷裂解炉CFD模型。由于简易的机理模型忽略了烟气流动情况对于二氯乙烷裂解过程的影响,为了准确描述炉膛内烟气的速度场、温度场、浓度场,以及烟气对管内反应的影响,文章依据计算流体力学的原理,建立复杂的二氯乙烷裂解炉CFD模型。炉膛烟气流动模型用雷诺平均的守恒方程描述,其中的湍流项用k-ε方程封闭；炉膛燃气的燃烧采用一阶串联的简易动力学模型,燃烧过程作全预混假设,其中湍流作用以结合有限速率模型的涡耗散模型描述；烟气的辐射模型采用离散坐标模型(DO模型)。通过二氯乙烷裂解炉的CFD模拟,给出了炉膛详细的炉膛烟气流速、温度、浓度在炉膛中的三维分布信息,有助于加深人们对二氯乙烷裂解炉炉膛燃料气燃烧、烟气流动传热传值过程的本质特征的了解,从而辅助裂解炉的工艺设计。此外,二氯乙烷CFD模拟虽然非常耗时,但其结果更为准确,可用于辅助验证简易模型的准确性。(5)完成了二氯乙烷裂解炉多目标优化。考虑计算代价,二氯乙烷裂解炉的操作优化选择简易的机理模型作为多目标优化模型,优化的目标为二氯乙烷裂解的转化率最大化,选择性最大化,单耗最小化。优化的决策变量则为二氯乙烷进料流量,燃料气进料量,燃料气负荷分配因子、以及炉管出口压力等四个重要现场操作变量。通过将整体的多目标优化问题进一步分解,确定了转化率-选择性,转化率-单耗,选择线-单耗三个双目标优化问题。利用所建立的SMODE-εCD多目标优化算法,实现二氯乙烷裂解炉性能指标的多目标优化,找到了转化率-选择性,转化率-单耗,选择线-单耗的近似Pareto边界,并给出了典型边界点所对应的操作工况,用于辅助现场决策者做出正确的判断。更多还原

【Abstract】 Poly vinyl chloride (PVC), the second most used plastic worldwide, is produced by the polymerization of vinyl chloride monomers (VCMs). The pyrolysis of ethylene dichloride (EDC), is the main commercial route to form VCM. Given the high energy consumption of the EDC cracking process, the EDC cracker is considered as the heart of the entire VCM manufacturing process. Thus, it is very crucial to reduce the energy consumption and improve the quality of VCM production. Due to lack of on-site detecting equipments, only the coil outlet temperature (COT) is measured and controlled to evaluate the EDC cracking performance on site, which is very glib. In order to deepen the understanding of EDC cracking process, it is necessary to build an accurate model of the EDC cracker. This paper aims to establish multi-scale EDC models from different aspects. Besides that, multi-objective optimization strategy of the EDC cracker is developed based on built-up model. The main innovations of this work are summarized as follows:(1) An improved multi-objective optimization algorithm, named SMODE-εCD is proposed which integrates self-adaptive differential evolution multi-objective optimization algorithm with epsilon constrained-domination principle together. In SMODE-εCD, the trial vector generation strategies and the variation and cross-over parameters of DE are adaptively adjusted based on evolutionary progress knowledge. Furthermore, epsilon constrained domination principle is adopted to handle constraints in multi-objective problems. The advantageous performance of SMODE-εCD is tested and validated over nine standard test problems and four engineering design problems with comparisons of other two typical multi-objective problems named NSGA-II and CMODE. The performance indicators show that SMODE-εCD is an effective and feasible approach to solving all kinds of constrained multi-objective problems.Thus, the proposed algorithm is solid foundation and preparation for the multi-objective optimization of EDC cracker.(2)A simplified mechanistic model of EDC cracker at the clean-tube stage is established. Given the intense heat coupling between the furnace and the reactor, the cracker is divided into two parts:the furnace model and the reactor model, with heat flux and flue gas temperature profiles connecting the two models. The heat transfer process in the furnace is described using one-dimensional Lobo-Evans model. A one-dimensional plug flow reactor model and a two-step serried reaction kinetics model are assumed in the reactor. Based on the fist principle model, different fuel gas allocation strategies are investigated to improve performance indices such as selectivity, conversion, and fuel gas consumption (per vinyl chloride monomer production) with the definition of fuel gas allocation factorα. Based on that, the suitable coil outlet temperature (COT) is suggested to make good compromise among the performance indices. (3) Run length model of ethylene dichloride cracker is developed. Radical mechanism with coke formation is adopted to describe the EDC cracking reactions with24reaction equations and31components. Based on the run length model of EDC cracker, the impactions of coke deposition are analyzed with some core information variations with time being discussed. Two important aspects, namely, CCl4concentration and fuel gas allocation, are investigated to understand the overall benefits of the whole operation cycle. A comprehensive analysis shows that the concentration of the CCl4promoter should be controlled at100ppm wt%and the fuel gas allocation factor should be maintained at0.36to guarantee the overall economic benefits of the EDC cracker in the full operation cycle.(4) The computational fluid dynamic (CFD) model of EDC cracker is developed. Reynolds-averaged Navier-Stokes (RANS) model is used to describe the flue gas flow with a standard k-ε turbulence model to close that. A finite-rate/eddy dissipation model is used to model a pre-mixed combustion of the sidewall burners. The discrete ordinate model is applied to simulate the radiative heat transfer of a furnace in CFD simulation. The details of flue gas velocity, temperature, composition information are provided in the CFD model, which is effective supplement and verification for the simplified one dimensional Lobo-Evans model.(5) The multi-obejective optimization of EDC cracker is solved. The EDC cacker operation optimization problems are converted into standard mathematical constrained multi-objective problem with the definition of objective functions, decision variables and constaints based on the simplified EDC cracker model. In order to cut off the computational burden, the multi-objective optimization is divided into three cases named conversion-selectivity, conversion-consumption, selectivity-consumption. The propsed SMODE-sCD is applied to solve these problems. After that, a set of trade-off pareto optimal solutions are figured out, and some typical operational conditions related preferred solutions are selected and presented to help the on-site decision maker to make a more scientific choice of the on-site operation.更多还原