炼化一体智能化工厂计划优化模型设计及应用实践

炼化企业供应链协同优化对实现智能炼厂至关重要,计划优化系统是供应链协同优化实现的核心系统。本文详细介绍了规划总院自主研发的新一代计划优化软件整体框架及主要功能,在该平台基础上结合炼化一体化企业流程数据、原油数据库及产品销售等边界条件进行了炼化一体化模型设计及建设,用于开展各类优化分析,并基于线性规划原理,通过边际效益分析及多方案对比分析,结合原油保本价法、缩减成本法及等量替代法等不同的原油选购方法,开展了原油选购测算及应用分析,并对炼化一体化企业加工原油提出了相关建议,指导企业原油选购。     

基于特征选择的常减压装置模型及在计划优化中的应用

常减压装置将原油切割为不同中间产品,其作为炼油工艺的龙头装置,对炼油过程生产计划排产与效益提升至关重要。通过建立一种高精度且具有良好求解效率的常减压模型,用以求解模型关键指标实沸点（TBP）曲线,即综合考虑切割产品的实沸点（TBP）与原油TBP、流量、温度等变量影响,构建非线性方程组模型来表征输入输出间的关系;利用特征选择方法遴选相关变量（包括进料性质、相邻TBP及其二次项等）,并采用鲸鱼优化算法优化方程组系数。仿真结果表明,该多输出相互关系模型与已有文献工作相比,在兼顾求解效率基础上,常减压装置各蒸馏切割产品TBP曲线预测上有更高的精度,将此模型应用到炼厂计划优化中,与传统的悬摆切割模型对比,优化结果优于传统悬摆切割模型。     

原油作业过程优化的研究进展

原油作业过程是石油供应链的重要组成环节,包括炼油企业生产过程中的原油采购、原油分配、原油输送、原油储存和原油调和等多个工业流程。原油作业过程优化具有很高的学术理论价值与工业应用价值,与其相关的研究工作是目前学术界与工业界共同关注的热点。首先简要描述了原油作业过程,并对其优化问题的难点进行分析;其次,分别从优化模型、优化算法以及不确定性优化方法三个研究角度,重点阐述了原油采购优化、原油储运优化、原油调和优化以及不确定性条件下原油作业优化四个主要研究方向的学术进展,并对当前已有的研究成果进行了归纳总结;最后,对原油作业过程优化当前存在的一些问题提出了建议,并对该领域未来的发展方向与趋势进行了展望。     

基于EGO算法的常压塔能量优化

常减压装置能量消耗约占炼厂总用能的25%～30%,在保证产品产量与质量的条件下,优化常减压蒸馏塔操作条件,可有效降低能耗.为了避免随机优化算法对常压塔机理模型进行操作优化时,存在计算资源消耗大、效率低的问题,文中采用基于代理模型的全局优化方法优化常压塔的余热回收过程,在优化迭代过程中用Kriging代理模型来代替耗时的精确模型评估.实验表明模型调用次数相较于粒子群优化算法减少了90%,优化时间减少了85%,实现了能量优化并且保证了侧线产品之间的分离精度.     

Experimental measurement and modeling of saturated reservoir oil viscosity

A novel mathematical-based approach is proposed to develop reliable models for prediction of saturated crude oil viscosity in a wide range of PVT properties. A new soft computing approach, namely least square support vector machine modeling optimized with coupled simulated annealing optimization technique, is proposed. Six models have been developed to predict saturated oil viscosity, which are designed in such a way that could predict saturated oil viscosity with every available PVT parameter. The constructed models are evaluated by carrying out extensive experimental saturated crude oil viscosity data from Iranian oil reservoirs, which were measured using a “Rolling Ball viscometer.” To evaluate the performance and accuracy of these models, statistical and graphical error analyses were used simultaneously. The obtained results demonstrated that the proposed models are more robust, reliable and efficient than existing techniques for prediction of saturated crude oil viscosity.     

能量集成常减压系统的简捷模型及优化

针对常减压装置,提出基于Fenske-Underwood-Gilliland方程和焓衡算的简捷模型对现有的常减压系统复杂塔进行操作型计算和优化,并利用夹点分析确定系统的热回收目标。以常减压系统的年净利润最大为目标,同时考虑系统的产品收益和能量集成,建立了基于PSO随机粒子群算法的优化方法。以现有的常减压装置为算例,得到了年净利润最大条件下的各工艺参数。通过与严格模拟对比,表明常减压系统简捷模型能够得到较准确的计算结果。     

Intelligent control of a grain drying system using a GA-SVM-IMPC controller

The grain drying is difficult to control because of its characteristics of long delay, strong nonlinearity, and uncertainty parameters. The aim of the paper is to design a suitable controller for a newly designed grain dryer. First, a nonlinear math model of the wheat mixed flow drying was established and analyzed based on the fundamental laws of simultaneous heat and mass transfer. The simulations for the batch circulating drying process and the continuous grain drying process have been made and the simulation results show that it fits well with the actual drying process. Second, an internal model proportional integral derivative (PID) controller (IMPC) based on the support vector machine (SVM) algorithm and the genetic algorithm (GA-SVM-IMPC) was proposed from the view of the energy loss and the dried grain quality. The structure of the GA-SVM-IMPC controller consists of a SVM prediction model, a SVM inverse model controller, a PID controller and a genetic optimization algorithm. Finally, the effectiveness of this controller was demonstrated by computer simulations, and the comparative study with the other controllers further confirmed the superiority of the proposed dryer controller.     

水泥工厂电耗预测及优化

针对当前水泥工厂面临的高电耗问题,提出了基于支持向量机（SVM）的电耗预测。通过预测未来时刻的电耗,在优化控制上做相应的调整,在不影响生产质量的前提下达到高效生产的目的。首先分析了解水泥的生产工艺及其数据特点,确定影响电耗的关键指标;对于水泥工艺变量存在的滞后性、多工况等难点,在预测模型中加入了自适应的策略;经离线验证并在某水泥现场实施应用了两个月,证实了预测的可靠性,并在节能方面取得了不错的效果。     

基于智能预测和机理模型的换热网络清洗决策

原油石化装置的换热器常受到结垢影响,导致换热器性能衰退严重,换热网络性能也会随之衰退,同时换热器之间的耦合关系,导致不同换热器的性能衰退对换热网络整体性能变化的影响不同。以往的清洗决策主要是根据单台换热器性能衰退到一定程度来制定的,这会导致换热网络整体可能处在较差的运行状态。因此,本文提出一种基于智能预测和机理模型的换热网络清洗决策方法,基于换热器的运行数据建立智能预测模型,获得换热器性能变化趋势,结合换热网络的性能模拟模型,进一步获得换热网络的性能变化趋势,从而从换热网络整体性能变化的角度来制定清洗方案。研究表明,对收集到的原油换热器运行数据,建立神经网络预测模型,具有较好的预测精度。通过对原油精馏装置换热网络的案例分析,当HE1、HE2和HE5三台换热器同时发生性能衰退时,换热网络年度公用工程能耗费用将增加12.1%。与传统基于单台换热器性能衰退情况制定的清洗方案相比,从换热网络整体性能衰退角度制定的清洗方案,年度额外公用工程费用减少13.1%,损失费用减少14.1%,年度总费用减少13.8%,而清洗次数仅增加3台次。     

Fault prediction of non-linear multi-mode industrial process based on MKPCA model

The actual industrial process is regarded as a non-linear multi-mode process. If the system is described according to the model of the single-mode or the linear multi-mode, it will inevitably lead to a large number of false alarms or missed alarms. The fault detection, fault reconstruction, fault amplitude estimation, and prediction for non-linear multi-mode process based on the multi-kernel principal component analysis (MKPCA) model are studied in this paper. First, the MKPCA model will be applied to fault detection in the steady-state process of different modes, and a weighted algorithm will be adopted for fault detection in the transition process. Then, the fault degree will be described quantitatively, and the fault amplitude in the form of fault reconstruction will be solved by the optimization method, and the consistent estimation algorithm of fault amplitude under different modes will also be studied. Finally, the prediction model of support vector machine (SVM) prediction model will be applied to predict the development trend of the fault amplitude. Furthermore, the Tennessee Eastman (TE) process will be taken as an application object to verify the effectiveness and superiority of the method.     

考虑原油性质波动的炼厂氢气网络集成优化

针对原油性质的不确定性,提出了一种基于质量传递机理的随机规划建模框架,以实现炼厂氢气网络在经济效益和抗扰能力上的同步优化。该框架耦合了常减压蒸馏、加氢精制以及闪蒸分离等过程单元,从微观上解析原油性质波动对网络运行的影响;采用了代理模型技术增设脱硫模块,并利用了二阶段随机规划方法改造管网,从宏观上优化氢气网络以满足生产要求。为验证所提方法的有效性和适用性,对某一现有的炼厂氢网络进行了改造设计研究。结果表明,集成过程单元的多场景优化策略能够有效提升网络的经济性能,并且能使其灵活应对因原油性质波动引起的操作场景的改变。     

基于PLS的原油选择与混合优化研究

为降低原油采购成本和生产符合质量要求的产品,将原油的选择与混合看作一个整体,基于原油属性数据信息,采用偏最小二乘法建立原油选择与混合优化的非线性混合整数规划模型,应用分支定界算法对模型进行求解,并通过具体实例分析证实了模型和算法的有效性。     

LSSVM过程建模中超参数选取的梯度优化算法

基于结构风险最小的最小二乘支持向量机（least squares support vector machine, LSSVM）为标准支持向量机（SVM）的约简;训练简易;性能良好。其模型精度受超参数影响;常规的网络搜索法很难搜得最佳超参数。在快速留一法的基础上;以全样本留一预测误差平方和最小化为目标;导出基于梯度的最优化算法;用以优选为LSSVM超参数;进而构建G-LSSVM模型。以柠檬酸发酵过程为算例对G-LSSVM进行检验;结果表明G-LSSVM的超参数选取耗时少;模型稳定性良好;且拟合和预报性能都优于标准SVM和神经网络。有望适用于机理不明、高度非线性、小样本的化工过程建模。     

LS-SVM模型在线校正的替代法及其软测量应用

软测量技术是现代化工过程控制系统的重要组成部分,化工过程的时变性要求软测量模型及时进行在线校正,以适应新工况,保持模型的预测精度。基于矩阵计算理论,提出了最小二乘支持向量机模型在线校正的替代法,它无需重新训练,校正速度快。通过分析还提出以距离标准选择被替代个体,可提高预测精度。将该法用于某炼油厂常压塔塔顶汽油干点的软测量,效果良好,表明该法可以克服工况变化对模型预测精度的影响。     

Soft-sensing modeling and intelligent optimal control strategy for distillation yield rate of atmospheric distillation oil refining process

It is a challenge to conserve energy for the large-scale petrochemical enterprises due to complex production process and energy diversification. As critical energy consumption equipment of atmospheric distillation oil refining process, the atmospheric distillation column is paid more attention to save energy. In this paper, the optimal problem of energy utilization efficiency of the atmospheric distillation column is solved by defining a new energy efficiency indicator — the distillation yield rate of unit energy consumption from the perspective of material flow and energy flow, and a soft-sensing model for this new energy efficiency indicator with respect to the multiple working conditions and intelligent optimizing control strategy are suggested for both increasing distillation yield and decreasing energy consumption in oil refining process. It is found that the energy utilization efficiency level of the atmospheric distillation column depends closely on the typical working conditions of the oil refining process, which result by changing the outlet temperature, the overhead temperature, and the bottom liquid level of the atmospheric pressure tower. The fuzzy C-means algorithm is used to classify the typical operation conditions of atmospheric distillation in oil refining process. Furthermore, the LSSVM method optimized with the improved particle swarm optimization is used to model the distillation rate of unit energy consumption. Then online optimization of oil refining process is realized by optimizing the outlet temperature, the overhead temperature with IPSO again. Simulation comparative analyses are made by empirical data to verify the effectiveness of the proposed solution.     

原油库存调度优化技术研究

根据原有库存实际生产需要运用库存优化理论,在重点考虑热耗因素下,以此作为目标函数,建立数学模型,对所建立的模型进行分析,利用分支界限法进行求解,对原油库存优化后结果明显好于优化前。优化后调度方案满足原油库实际生产需要,按照库存优化后计算,所损失的热量为4,05×10^9kJ,而在原油库存优化前实际的损耗热量为4．4×10^9kJ,两者之差为3．5×10^8kJ,按照原料高位发热值计算,每天节省下来1．5t燃料,起到了很好的经济效益,降低了企业成本。     

自适应粒子群优化算法在聚丙烯熔融指数预报上的应用

针对丙烯聚合生产控制中聚丙烯熔融指数在线测量的控制要求,以及过程变量间相关性高的特点,提出一种基于自适应粒子群优化算法和径向基函数神经网络的聚丙烯熔融指数预报新方法。该方法采用变参数的自适应粒子群优化算法提高优化算法的效率和收敛性,并且融合了主成分分析、统计建模以及智能优化方法,从而降低了预报模型的复杂度。提出了一种基于径向基函数神经网络的统计预报模型的参数优化和结构优化方法。使用该统计模型对工厂实际生产过程进行预报,并与国内外相关研究报道相比较,表明了本文所提出的预报方法的有效性和更高的准确性。     

The optimization and prediction of properties for crude oil blending

Research into the properties of mixed crude oils has been the most important research for refineries and oil transportation. Because the desirable nature of mixed crude oils is a powerful guarantee for normal operations and corporate earnings. In this paper, the optimization and prediction model for Shanbei crude oil blending is established. The objectives were to improve the total yield of fractions and the compatibility and to reduce the viscosity of the mixed crude oils. After solving the model, two applicable proportions for mixed crude oils were given. With these proportions, two samples of mixed oils were prepared and distilled. The test results showed that the model was valid. The study of the viscosity and compatibility index was introduced to the optimization model because forecasting those properties in mixed crude oils has a certain significance.     

基于PSO-SVM逆的赖氨酸发酵过程软测量

针对赖氨酸发酵过程非线性、大滞后、多变量动态耦合,关键生化参数难以实时在线测量等问题,提出一种改进的粒子群-支持向量机(PSO-SVM)逆发酵过程软测量建模方法。首先分析逆系统的存在性,并结合赖氨酸发酵过程,引入发酵特征信息和舍弃次要信息构造逆扩展模型;然后利用支持向量机离线辨识初始逆扩展模型,并根据系统输入与模型输出的偏差信号,采用粒子群算法对初始逆扩展模型进行在线校正;最后将校正后的逆扩展模型串联在原发酵过程之后构成复合伪线性系统,实现不直接可测关键生化参数的在线预测。以L-赖氨酸流加发酵过程为例,验证了所提算法能够对发酵过程关键生物量参数进行较准确的在线预测,较普通的SVM逆建模方法具有更高的预测精度。     

基于变异CPSO算法的LSSVM蒸发过程软测量

在分析混沌粒子群优化算法(CPSO)和最小二乘支持向量机(SVM)理论基础上,以某氧化铝厂蒸发过程为对象,采用带有末位淘汰机制的混沌粒子群优化算法优化支持向量机的参数,建立了基于变异CPSO算法的LS-SVM的氧化铝蒸发过程软测量模型,并与PSO-LSSVM、LSSVM模型比较,研究表明,ICPSO-LSSVM模型预测准确,泛化性能好,且该模型预测结果中相对误差小于5%的样本达到92.5%,最大相对误差仅为8.1%,均方差MSE为0.05153,模型具有较高的精度,其现场实施结果表明基本可以实现出口浓度的实时在线预估。