基于PCA-RBF层叠泛化网络的硝酸生产工艺历史数据建模

详细信息    查看全文 | 推荐本文 |

英文篇名：Historical data modeling of nitric acid production process based on PCA RBF stacked generalization network 作者：李爱民 ; 代敏 ; 杨福胜 ; 侯建民 ; 何瑞明 英文作者：LI Aimin;DAI Min;YANG Fusheng;HOU Jianmin;HE Ruiming;Shanxi Applied Research Institute of Rock and Mineral Testing;College of Chemical Engineering,Xi'an Jiaotong University;Xi'an Senwei Auto-control Engineering Co.,Ltd.; 关键词：硝酸工艺 ; PCA(主成分分析) ; 层叠泛化 ; 预测模型 英文关键词：nitric acid process;;PCA(principal component analysis);;cascade generalization;;prediction model 中文刊名：SDHW 英文刊名：Shanxi Chemical Industry 机构：山西省岩矿测试应用研究所;西安交通大学化工学院;西安森威自控工程有限公司; 出版日期：2018-04-15 出版单位：山西化工 年：2018 期：v.38;No.174 语种：中文; 页：SDHW201802028 页数：4 CN：02 ISSN：14-1109/TQ 分类号：77-80

摘要

硝酸工业对国民经济、国防工业的发展起着重要的支持作用,而成品酸流量是硝酸生产过程重要的技术指标。提出了一种基于层叠泛化策略的成品酸流量的预测建模方法。首先,用主成分分析(PCA)对原始数据降维,提取前5个主成分作为影响因子集合;然后,构建两级RBF网络学习层,第一级针对影响因子空间,对一组RBF神经网络进行交互验证式训练得到输出值和对应的真实值所组成的的特征空间;第二级由单个RBF网络将第一级网络得到的特征空间进行非线性组合;最后,将模型用于预测出口成品酸流量,并与相应实测值进行比较以验证其合理性。结果表明,本文建立的层叠泛化网络模型具有较高的预测精度和泛化性能,对于硝酸生产工艺的分析和优化具有显著的指导意义。
        The nitric acid industry plays an important role in the development of national economy and national defense industry,and finished acid flow is an important technical index in the production process of nitric acid.A predictive modeling method for finished acid flow based on cascading generalization strategy is proposed.First,the principal component analysis is used(PCA)to reduce the original data and extract the first 5 principal components as the set of influence factors.Then,the RBF network learning layer with two levels is constructed.The first level is directed against the influence factor space,and a group of RBF neural networks is trained by interactive validation to get the feature space of the output value and the corresponding real value.The second level is a nonlinear combination of the feature space obtained by the first level network by a single RBF network.Finally,the model is used to predict the exported acid flow,and it is compared with the measured values to verify the rationality of the model.The results show that the cascaded generalization network model established in this paper has high prediction accuracy and generalization performance,and has great guiding significance for the analysis and optimization of nitric acid production process.

引文

[1]唐文骞,张友森.我国硝酸工业生产现状分析及发展建议[J].化肥工业,2013(1):31-35,42.
    [2]张夕专.硝酸生产工艺研究[J].科技视界,2013(19):203-204.
    [3]靳小桂.我国硝酸、硝酸铵行业现状及发展趋势---在2011年硝酸硝铵行业工作会上的讲话[J].化肥工业,2011(4):1-3,8.
    [4]陈超.智能建模中冗余问题的分析与处理及其应用[D].上海:华东理工大学,2015.
    [5]张丹.硝酸生产过程操作参数优化的研究[D].沈阳:沈阳工业大学,2005.
    [6]朱玉华.硝酸生产过程操作参数的智能优化[D].沈阳:沈阳工业大学,2003.
    [7]黄帅栋,卫志农,丁恰,等.基于层叠泛化策略的母线负荷预测模型[J].电力系统及其自动化学报,2013,(3):8-12,55.
    [8]Guo Hongwei,Su Buxin,Zhang Jianliang,et al.Extracting the core indicators of pulverized coal for blast furnace injection based on principal component analysis[J].International Journal of Minerals Metallurgy and Materials,2013(3):246-252.
    [9]苏兴.基于神经网络的化工建模预测方法研究[D].大连:大连理工大学,2011.
    [10]乔弘.火电厂热工参数软测量关键技术和方法研究[D].北京:华北电力大学,2009.
    [11]Cerny B A,Kaiser H F.A Study Of A Measure Of Sampling Adequacy For Factor-Analytic Correlation Matrices[J].Multivariate Behav Res,1977,12(1):43-47.
    [12]董忠哲.双加压法硝酸生产工艺流程的模拟与优化[D].天津:河北工业大学,2013.
    [13]江纪海.硝酸装置中一氧化氮氧化反应速度的计算[J].山西化工,1983(4):3-9,19.
    [14]刘萍.发烟硝酸漂白塔的技改[J].中氮肥,2010(4):52-53.