引用本文:方一鸣,赵晓东,张攀,刘乐,王硕玉.基于改进灰狼算法和多核极限学习机的铁水硅含量预测建模[J].控制理论与应用,2020,37(7):1644~1654.[点击复制]
FANG Yi-ming,ZHAO Xiao-dong,ZHANG Pan,LIU Le,Wang Shuo-yu.Prediction modeling of silicon content in liquid iron based on multiple kernel extreme learning machine and improved grey wolf optimizer[J].Control Theory and Technology,2020,37(7):1644~1654.[点击复制]
基于改进灰狼算法和多核极限学习机的铁水硅含量预测建模
Prediction modeling of silicon content in liquid iron based on multiple kernel extreme learning machine and improved grey wolf optimizer
摘要点击 1517  全文点击 539  投稿时间：2019-07-16  修订日期：2019-12-10
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2020.90571
  2020,37(7):1644-1654
中文关键词  改进灰狼优化算法  最优–最差正交反向学习  多核极限学习机  铁水硅含量  预测建模
英文关键词  improved grey wolf optimizer  optimal-worst orthogonal opposition-based learning  multiple kernel extreme learning machine  silicon content in liquid iron  prediction modeling
基金项目  国家自然科学基金项目(61873226, 61803327), 河北省自然科学基金项目(F2017203304, F2019203090), 河北省人才工程培养经费资助科研项目 (A2016015002)资助.
作者单位E-mail
方一鸣* 燕山大学 fyming@ysu.edu.cn 
赵晓东 燕山大学  
张攀 燕山大学  
刘乐 燕山大学  
王硕玉 高知工科大学  
中文摘要
      针对高炉铁水硅含量难以在线检测的问题, 本文提出一种基于改进灰狼算法(IGWO)优化的多核极限学习 机(MKELM)高炉铁水硅含量预测建模方法. 首先, 针对灰狼算法(GWO)寻优能力的不足, 将最优–最差正交反向学 习(OWOOBL)策略应用于灰狼算法的位置更新, 得到一种改进灰狼优化算法. 通过10种标准函数对所提算法进行 仿真测试, 结果表明此算法具有更好的寻优能力. 其次, 针对单核极限学习机(KELM)回归能力不足, 将不同种类的 核函数加权组合, 并采用改进灰狼算法对多核极限学习机中的加权系数等参数进行优化. 最后, 基于某钢厂的实测 数据对高炉铁水硅含量进行预测建模, 仿真结果表明, 本文所提方法的预测效果优于反向传播神经网络(BP–NN)、 极限学习机(ELM)、KELM和GWO–MKELM, 对高炉炼铁具有较好的指导意义.
英文摘要
      Aiming at the difficulty of on-line detection of silicon content in liquid iron of blast furnace, a predictive modeling method of silicon content in liquid iron of blast furnace based on multiple kernel extreme learning machine (MKELM) optimized by improved grey wolf optimizer (IGWO) is proposed. Firstly, aiming at the shortage of the search ability of grey wolf optimizer (GWO), the optimal-worst orthogonal opposition-based learning (OWOOBL) is applied to the location update of grey wolf algorithm, and an IGWO is obtained. The simulation of 10 standard functions shows that the improved grey wolf optimizer has better optimization ability. Secondly, aiming at the insufficient regression ability of single kernel extreme learning machine (KELM), different kinds of kernel functions are weighted and combined, and the weighted coefficients and other parameters of the MKELM are optimized by IGWO. Finally, the prediction model of the silicon content in liquid iron of blast furnace is established based on the measured data of a steel plant. The simulation results show that the prediction effect of the proposed method is better than that of back propagation neural network (BP– NN), extreme learning machine (ELM), KELM andGWO–MKELM, and therefore the proposed method has a good guiding significance for blast furnace iron making.