数据挖掘技术在优化石化生产操作参数中的应用

Petrochemical Industry Production Operating Parameters Optimization Based on Data Mining Technology

10.6338/JDA.201012_5(6).0006

陆治荣(Zhi-Rong Lu)

数据挖掘 ； 工业优化 ； 数据分析 ； Data mining ； Industry optimization ； Data analyses

Journal of Data Analysis

5卷6期（2010 / 12 / 01）

91 - 107

簡體中文

由于石化工业生产机理的复杂性、生产模式的不断变化、设备状况的老化以及生产过程中的扰动等原因，常常使生产偏离优化状态。不少生产装置在产品质量、能耗和产量方面存在的问题，是生产操作参数设置不当造成的。许多流程业生产过程是多变量、非线性系统，变量之间具有强相关，生产数据呈非高斯分布，并伴有高噪声，又没有精确的机理模型，技术人员只能靠经验调整生产操作参数。我们应用数据挖掘技术，提出了一套分析数据的新方法和新思路，不是着力于寻求精确描述生产过程的函数表达式，而是设法寻求生产历史数据中存在的差异性，以及造成的原因；进而分析数据中存在的某种特定模式，以及使生产维持或避免该模式而采取的措施。应用案例证明，该数据分析方法是行之有效的，当今流程行业已经普遍使用DCS系统，有些企业建立了工厂级的实时数据库和关系数据库，这为数据挖掘技术的应用提供了宽广的舞台。

Production often deviated from the optimal conditions, because of the complexity of industrial production process mechanism, the changing of Production patterns, the status of aging equipment, the disturbance in the production process and the other reasons. Many production installations problems in aspects of product quality, energy consumption and output are caused by improper operation of the production parameters. Many process industry production process is a multi-variable, non-linear system and a strong correlation is between variable. Meanwhile, production data is non-Gaussian distribution, accompanied by high noise, and there is no accurate first principle model. Therefore, technical staffs have to adjust the production operation parameters based on their own experience. According to the data mining technology, we propose a new data analysis method. First, we are not focus on the search for function expression ,used for accurately describing the production process, but to seek the difference, existing in the historical data on the production, and the reasons. Then, we analyze the data to find some special model and measures to sustain production or avoid the pattern. The applications prove that the data analysis methods are effective. Today's process industry has already widespread use of DCS system. Some enterprises established a factory-level real-time database and relational database. All of them provide a broad stage to data mining application.

1. Chen、 Nianyi(1987).Pattern recognition applied to the optimization on chemical industry.Progress in Chemical Technology,2,7.
2. Chen、 Nianyi(1997).Pattern recognition applied to industrial optimization technology.Beijing=北京:Chinese Publisher of Petrochemical Technology=?石化出版社.
3. Chen、 Nianyi,Lu, Wencong,Ye, Chenzhou,Li, Guozheng(2002).Application of Support Vector Machine and Kernel Function in Chemometrics.Computers and Applied Chemistry,19(6),691.
4. Chen、 Nianyi,Qin, Pei,Chen, Ruiliang,Lu, Wencong(1999).Pattern recognition applied to chemistry and chemical industry.Beijing=北京:Science Press=科?出版社.
5. Hand, David(2003).Principles of Data Mining.机械工?出版社.
6. White, Douglas C..Emerson Process Management, the "Smart" Refinery: Economics and Technology.NPRA 2003 Annual Meeting
7. 陆治荣、陈念贻、陆文聪(2002)。DMOS—基于多种数据挖算法的工业优化软件系列。计算机与应用化学，19(6)，683。
8. 陈念贻、陆文聪、陆治荣(2002)。优化建模技术和机器学习理论的新发展。计算机与应用化学，19(6)，677。