近年来,随着市场竞争的日益加剧和环保要求的不断提高,迫切要求企业从有限的资源中不断挖掘潜力,提高经济效益,这给过程控制和过程优化提出了新的要求,从而也对过程建模提出了更高的要求。
一切工业生产的目的都是为了获得合格的产品,质量控制成为所有控制的核心。为了实现良好的质量控制,就必须对产品质量或与之相关的重要过程变量进行严格控制。由于在线分析仪表不仅价格昂贵,而且分析仪表滞后大,最终将导致控制系统的性能下降,从而难以满足生产要求。例如石油化工生产中精馏塔产品成分、塔板效率、反应器中反应物浓度、转化率等参数。为解决这类变量的测量问题,软测量技术得到了很大的发展。
本文对软测量技术进行了深入的研究与探讨,在此基础上利用多元线性回归、多元线性逐步回归、人工神经网络等方法来建立数学模型,并结合扬子石化丁二烯装置优化改造项目,在Fisher-Rosemount公司的DeltaV DCS上成功实现了基于软测量模型的先进控制系统。
本文的主要研究工作概括如下:
(1)首先对过程控制的发展和研究现状做了论述。对现有的研究成果进行了分析和阐述,并指出了理论研究与实际应用中所存在的困难。
(2)扬子石化公司丁二烯生产装置配套的DeltaV DCS没有选择与管理计算机系统的通讯接口。以Fisher-Rosemount公司的DeltaV DCS与PC LAN的数据传输为例,介绍了相应技术、实现方法及其应用。我们设计一套DeltaV DCS与PC机通讯方案,用于实时采集生产过程现场数据,并把DCS中的生产工艺数据存入数据库中。
(3)针对数据库中的数据我们采用多元线性回归、多元线性逐步回归、人工神经网络等方法来建立相应的数学模型。
(4)OPC(OLE for Process Control)是微软公司的对象链接和嵌入技术在过程控制方面的应用,为工业自动化软件面向对象的开发提供一项统一的标准。OPC的目的是为工厂底层设备或控制室数据库中大量数据源之间的通信提供一种标准的通信机制。
(5)利用DeltaV DCS这个开发平台,进行组态和下装模型,从而达到先进控制要求。
关键词 软测量 线性回归 人工神经网络 DeltaV DCS OPC (OLE for Process Control)
ABSTRACT
Recently, with the becoming severely market competition and environment requirements force companies to improve their productivity and efficiency, which poses new requirements on process control and process optimization, as a result, more severe requirements are posed on process modeling.
All industries’ purposes are getting the eligible product. The core of all controls is quantity control. In order to achieve good quantity control, we should control product quantity or correlative process variable severely. On-line analysis instruments have several disadvantages including costly price, great lag etc., which lead to descend of control system’s performance and can’t meet production requirements. For example, in order to solve these variables measure such as distillation tower production component、tray efficient、reactant concentration、conversion and so on in petrochemical produce. Soft sensor technique has greatly developed.
Soft sensor technique has been studied and discussed in this dissertation deeply. Apply multivariable linear regression、multivariable step-wise regression and artificial neural networks to establish mathematical model, and realize the advanced process control system in Fisher-Rosemount DeltaV DCS.
The research works can be summarized as follows:
(1) Discuss the development and research of process control firstly. Analyze and expatiate the existing production.
(2) Without the communication interface between DeltaV DCS and Supervise Computer, we design a scheme of communication system between DeltaV DCS and PC, and then apply it to access process data, and then deposit them to database.
(3) Make use of multivariable linear regression、multivariable step-wise regression and artificial neural networks and so on to establish corresponding mathematical model.
(4) OPC(OLE for process control),which is the application of Microsoft’s Object Linking and Embedding technology in process control system, provides a unified standard for Object-Oriented Design in industrial automation software. A standard mechanism for communicating to numerous data sources, either devices on the factory floor, or a database in a control room is the purpose for OPC.
(5) Use the flat development of DeltaV DCS to configure、download and realize the model arithmetic in DeltaV DCS to achieve advanced process control.