化工学报 ›› 2016, Vol. 67 ›› Issue (3): 729-735.DOI: 10.11949/j.issn.0438-1157.20151847

• 研究论文 • 上一篇    下一篇

改进型EMD-Elman神经网络在铁水硅含量预测中的应用

宋菁华, 杨春节, 周哲, 刘文辉, 马淑艳   

  1. 浙江大学控制科学与工程学院, 浙江 杭州 310027
  • 收稿日期:2015-12-07 修回日期:2015-12-20 出版日期:2016-03-05 发布日期:2016-01-12
  • 通讯作者: 杨春节
  • 基金资助:

    国家自然科学基金项目(61290321);国家高技术研究发展计划项目(2012AA041709)。

Application of improved EMD-Elman neural network to predict silicon content in hot metal

SONG Jinghua, YANG Chunjie, ZHOU Zhe, LIU Wenhui, MA Shuyan   

  1. College of Control Science and Engineering, Zhejiang University, Hangzhou 310027, Zhejiang, China
  • Received:2015-12-07 Revised:2015-12-20 Online:2016-03-05 Published:2016-01-12
  • Contact: 67
  • Supported by:

    supported by the National Natural Science Foundation of China (61290321) and the National High Technology Research and Development Program of China (2012AA041709).

摘要:

针对高炉炼铁过程的多尺度和动态特征,建立了基于经验模态分解(empirical mode decomposition, EMD)和Elman神经网络的铁水硅含量预测模型。该模型先采用EMD将硅含量序列分解成有限个、相对平稳的本征模函数(intrinsic mode function, IMF)和剩余分量;然后,分别对每个IMF和剩余分量建立Elman神经网络子模型;为了进一步提高预测精度,将子模型的结果进行加权融合,并利用粒子群算法进行权值的寻优,最终获得硅含量的预测结果。将该模型用于某钢厂铁水硅含量的预报,实验结果证实了该方法的有效性。

关键词: 硅含量, 预测, 多尺度, 动态建模, 经验模态分解, 神经网络

Abstract:

To handle the multiscale and dynamic characteristics of blast furnace ironmaking process, a soft sensor model based on empirical mode decomposition (EMD) and Elman neural network is proposed. First, the original silicon content dataset is decomposed into a finite collection of intrinsic mode functions (IMFs) and a residue by EMD, obtaining relatively stationary sub-series from original data set. Second, each IMF and the residue are utilized to establish the corresponding Elman neural network model. To further improve the accuracy of prediction, the result of each sub-series is multiplied by a weight and then summed up to obtain the final silicon content. Here, all the weights are optimized by particle swarm optimization (PSO). The model was applied to the prediction of silicon content of blast furnace in a steel mill, and the result proved the effectiveness of the proposed method.

Key words: silicon content, prediction, multiscale, dynamic modeling, empirical mode decomposition, neural networks

中图分类号: