基于改进PLS算法的延迟焦化装置建模研究与应用

发布时间：2018-02-09 01:00

本文关键词： 延迟焦化软测量偏最小二乘算法混合PI-SIGMA模糊神经网络核偏最小二乘算法　出处：《华东理工大学》2015年硕士论文　论文类型：学位论文

【摘要】：随着轻质原油资源的逐渐消耗,迫使石油化工企业向原油重质化、产品轻质化转型。为了充分利用有限的石油资源,满足轻质油的大量需求,重油加工技术迅速发展。重油加工技术中以延迟焦化工艺发展最为快速,使用最为普遍。延迟焦化装置本身投资低、费用少,而且能够加工各种重质渣油,提升汽油和柴油产量,因而如何使延迟焦化技术更加成熟显得尤为重要。本文针对延迟焦化装置进行了建模研究,具体内容可分为以下几个方面： (1)本文利用偏最小二乘法提取有效信息、去噪能力强等优点,结合混合PI-SIGMA模糊神经网络算法极强的处理非线性问题能力,提高了偏最小二乘算法处理非线性问题的能力。将该算法运用于焦炭塔生焦速率建模,并根据生焦速率预测生焦周期焦层高度。经检验生焦高度的预测精度具有较高的准确性,满足生产要求。 (2)为了实现延迟焦化分馏塔柴油95%点的软测量预测,减少人工分析柴油95%点的次数,提高装置调整的效率,通过研究延迟焦化装置的工艺流程,分析了影响柴油95%点的因素,提出了一种基于改进核偏最小二乘法的柴油95%点的软测量方法。结果表明该方法能够有效地、快速地处理各变量之间的非线性关系,软测量结果明显优于偏最小二乘方法等方法。 (3)针对延迟焦化生产装置,设计开发了延迟焦化分析测算软件。该软件的功能包括实时数据通信、分馏塔柴油95%点在线计算、焦炭塔生焦高度预测、结果在线显示和报警、电子生产报表生成等。该应用软件自实际投运以来取得了较好的效果。
[Abstract]:With the gradual consumption of light crude oil resources, petrochemical enterprises are forced to transform to heavy crude oil and light products. In order to make full use of the limited petroleum resources and to meet the large demand for light oil, Heavy oil processing technology is developing rapidly. In heavy oil processing technology, delayed coking process is the most rapid development and most widely used. The delayed coking plant itself has low investment, low cost, and can process all kinds of heavy residuals. It is very important to improve gasoline and diesel production, so it is very important to make delayed coking technology more mature. In this paper, the modeling of delayed coking plant is studied, which can be divided into the following aspects:. In this paper, we use the partial least square method to extract effective information and have strong denoising ability, and combine with the hybrid PI-SIGMA fuzzy neural network algorithm to deal with nonlinear problems. The ability of partial least square algorithm to deal with nonlinear problems is improved. The algorithm is applied to modeling coke generation rate of coke tower, and the height of coking period is predicted according to coke generation rate. The prediction accuracy of coking height is higher than that of other methods. Meet production requirements. 2) in order to realize the soft-sensing prediction of 95% points of diesel oil in delayed coking fractionator, reduce the frequency of 95% points of manual analysis of diesel oil, and improve the efficiency of equipment adjustment, the factors affecting 95% points of diesel oil are analyzed by studying the technological process of delayed coking unit. A soft sensing method for diesel oil 95% points based on improved kernel partial least square method is proposed. The results show that this method can deal with the nonlinear relationship between variables effectively and quickly. The results of soft sensing are obviously superior to those of partial least square method and so on. The software for analysis and calculation of delayed coking is designed and developed for delayed coking production unit. The functions of the software include real-time data communication, on-line calculation of 95% points of diesel oil in fractionator, prediction of coke height in coke tower, and on-line display and alarm of the results. The application software has achieved good results since it was put into operation.
【学位授予单位】：华东理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TE96;TP18

【参考文献】