基于.NET的流动腐蚀实时专家诊断监管系统开发及参数建模

发布时间：2019-04-24 04:36

【摘要】：随着我国原油开采进入中后期,石化炼油企业大量进口高硫、高酸、高氮、含氯等不同类型的劣质原油,由此引起炼油加工过程中设备工况的频繁变化,导致炼油高风险系统流动腐蚀失效、泄露、爆管频发,不仅影响石化企业的生产计划、经济效益,还严重威胁着安全、节能与环保,已成为制约行业可持续发展与企业安全的重要障碍。本文首先对石化企业炼油系统监管现状进行了研究,进而分析了实时监管系统的开发需求,包括业务需求、用户需求和功能需求;并针对现有系统监管数据分散、监管针对性不强、缺乏流动腐蚀分析等缺点,提出了基于.NET开发平台的流动腐蚀实时专家诊断监管系统整体构架。接着在介绍系统相关开发技术(包括MySQL数据库技术、Visual Basic(VB)的ActiveX数据交互技术、Aspen Plus工艺仿真技术、Windows Communication Foundation(WCF)与.NET Remoting通信技术等)的基础上,阐述了专家诊断监管系统的具体设计及实现。接下来,针对部分腐蚀相关影响因子数据监管实时性差的问题,提出了基于快速搜索剪枝极限学习机(Fast Search Pruned-Extreme Learning Machine,FSP-ELM)的数据驱动软测量模型,并将其用于空冷器pH值软测量建模,标准数据集及工业现场历史数据测试结果均表明:相比于BP(Back Propagation)、FSP-BP(Fast Search Pruned-Back Propagation)、SVR(Support Vector Regression)和ELM(Extreme Learning Machine)算法,FSP-ELM算法具有较高的预测精度和泛化能力。最后将流动腐蚀实时专家诊断监管系统应用到某炼化厂常压塔顶系统的实时监管过程中,实例结果证明该系统能够达到预期的效果,且性能良好。本文所提出的基于FSP-ELM的数据驱动软测量方法不仅可用于空冷器系统的pH值软测量建模,而且为石化生产过程中其它不可实时检测的参变量提供了一种新的软测量建模方法;所开发的实时专家诊断监管系统框架可应用于多种炼油系统腐蚀监控平台的开发,该监管平台能够实现装置数据的动态实时监测、后台工艺模拟、超限报警和措施指导、历史数据查询等功能,具有有效的腐蚀防控作用,能够为石化炼油企业的安全生产提供有效预警与防控。
[Abstract]:As China's crude oil exploitation enters the middle and late stages, petrochemical refining enterprises import a large number of low-quality crude oil with high sulfur, high acid, high nitrogen and chlorine, which results in frequent changes in equipment working conditions in the refining process. The failure of flow corrosion, leakage and frequent burst of pipes in high-risk oil refining systems not only affect the production plan and economic benefits of petrochemical enterprises, but also seriously threaten safety, energy conservation and environmental protection. It has become an important obstacle to restrict the sustainable development of the industry and enterprise safety. Firstly, this paper studies the present situation of petroleum refining system supervision in petrochemical enterprises, and then analyzes the development requirements of real-time supervision system, including business requirements, user requirements and functional requirements; Aiming at the shortcomings of the existing system, such as scattered supervision data, weak supervision pertinence and lack of flow corrosion analysis, the whole framework of real-time expert diagnosis and supervision system for mobile corrosion based on. Net development platform is proposed. Then, on the basis of introducing the development technology of the system, including the ActiveX data interaction technology of MySQL database technology, Visual Basic (VB), Aspen Plus process simulation technology, Windows Communication Foundation (WCF) and .NET Remoting communication technology, etc. The concrete design and realization of expert diagnosis supervision system are expounded. Next, a data-driven soft-sensing model based on fast searching pruning limit learning machine (Fast Search Pruned-Extreme Learning Machine,FSP-ELM) is proposed to solve the problem of poor real-time monitoring of partial corrosion-related factors. It is applied to the soft sensing modeling of air cooler pH value. The test results of standard data set and industrial field history data show that compared with BP (Back Propagation), FSP-BP (Fast Search Pruned-Back Propagation), SVR (Support Vector Regression) and ELM (Extreme Learning Machine) algorithm, the data of standard data set and industrial field history data show that, FSP-ELM algorithm has high prediction accuracy and generalization ability. Finally, the flow corrosion real-time expert diagnosis and supervision system is applied to the real-time supervision of atmospheric pressure tower top system in a refinery. The results show that the system can achieve the expected effect and the performance is good. The data-driven soft-sensing method based on FSP-ELM proposed in this paper can not only be applied to the soft-sensing modeling of air-cooler system, but also provide a new soft-sensing modeling method for other parameters which can not be detected in real-time in petrochemical production. The developed framework of real-time expert diagnosis and supervision system can be applied to the development of various corrosion monitoring platforms for oil refining systems. The monitoring platform can realize dynamic real-time monitoring of plant data, background process simulation, out-of-limit alarm and guidance of measures. The function of historical data query has effective corrosion prevention and control function, and can provide effective early warning and prevention for petrochemical refining enterprises' safety production.
【学位授予单位】：浙江理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F426.72;TP311.52

【相似文献】