基于数据驱动的催化裂化过程反应—再生单元检测及评价

发布时间：2018-05-26 01:15

本文选题：故障诊断 + 积累和　；参考：《北京化工大学》2015年硕士论文

【摘要】：为了提高设备的安全性,经济效益和可靠性,工艺过程操作需要越来越先进的监测技术和故障诊断手段。早期的故障由于比较微弱所以不会被轻易发现,但是随着故障的加深会导致整个系统的失效。本文着重探究了工艺过程中这种缓慢的动态故障的监测方法及其数据处理方法。在炼油工业中,渣油催化裂化是非常重要的工艺装置。由于炼油产品的需求增加,加上严格的生产标准和环境要求,恰当正确的监测手段已不可或缺。渣油催化裂化装置操作中有许多问题,这其中的一些问题是由于人为错误操作,传感器失效等造成的。反应-再生装置是渣油催化裂化的核心部分,反应器参数的变化直接影响产品的质量和产量。在本文中,第一段再生量、中压蒸汽、原料油、浆料、贫气、第二段再生量作为其关键变量。对于这六个变量,利用分布式控制系统而得到。本文还考虑了引起反应器性能变差的条件。通常,任何极大影响焦炭和催化剂循环速率的条件会导致对产品转化和选择性的损失。这些变量的变化可以反应缓慢的动态故障。对于具有真实性的数据进行获取和采样后利用PCA和CUSUM-PCA进行数据分析。基于数据驱动的方法就是采用移动窗口对过程中的数据均值进行积累,然后利用传统的PCA方法对数据进行分析。在PCA分析法中,将正常的数据作为标准来判断与其他操作数据的偏离程度,然后利用平方预测误差(SPE)和Hotelling T2控制图来检测数据是否正常。通过对故障数据的监测分析表明,CUSUM-PCA可以更早更有效的监测早期故障。对中压蒸汽和贫气的异常水平研究案例结果表明,和常规PCA相比,CUSUM-PCA能提前一小时检测出故障。对于复杂性相同的相似工业操作,建议使用该研究中的方法。
[Abstract]:In order to improve the safety, economic benefit and reliability of the equipment, the process operation needs more and more advanced monitoring technology and fault diagnosis means. Early faults are not easy to find because they are weak, but as the faults deepen, the whole system will fail. This paper focuses on the monitoring method and data processing method of this slow dynamic fault in the process. Residuum catalytic cracking is a very important process unit in oil refining industry. Due to the increasing demand for refined oil products and strict production standards and environmental requirements, proper and correct monitoring tools are indispensable. There are many problems in the operation of residuum FCCU, some of which are caused by human error and sensor failure. Reaction-regeneration unit is the core part of residue catalytic cracking. The change of reactor parameters directly affects the quality and output of the product. In this paper, the first stage regeneration, medium pressure steam, feedstock, slurry, lean gas are the key variables. For these six variables, the distributed control system is used to obtain. The conditions that cause the reactor performance to become worse are also considered in this paper. Generally, any conditions that greatly affect the rate of coke and catalyst cycling will result in loss of product conversion and selectivity. Changes in these variables can respond to slow dynamic failures. PCA and CUSUM-PCA are used for data acquisition and sampling. The data-driven method is to use the moving window to accumulate the mean value of the data in the process, and then analyze the data by using the traditional PCA method. In PCA analysis, the normal data is used as the standard to judge the degree of deviation from other operational data, and then the square prediction error (SPE) and Hotelling T2 control chart are used to detect whether the data is normal or not. The analysis of fault data shows that CUSUM-PCA can detect early faults earlier and more effectively. A case study of abnormal levels of steam and lean gas at medium pressure shows that CUSUM-PCA can detect faults one hour earlier than conventional PCA. For similar industrial operations with the same complexity, it is recommended that the methods in this study be used.
【学位授予单位】：北京化工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TE96

【相似文献】