高炉透气性指数的改进多层超限学习机预测模型

Prediction model of improved multi-layer extreme learning machine for permeability index of blast furnace

中文关键词  高炉  透气性指数  建模  多层超限学习机  预测

英文关键词  blast furnace  permeability index  model  multi-layer extreme learning machine  prediction

中文摘要

      高炉透气性指数是高炉操作者衡量高炉顺行状态的指标之一. 针对传统透气性指数测量模型的缺陷, 本文 提出了一种基于改进的多层超限学习机(multi-layer extreme learning machine, ML–ELM)的高炉透气性指数预测模 型. 首先分析影响高炉透气性指数的相关操作参数, 考虑到高炉生产数据含有大量噪声, 运用小波去噪方法消除数 据的噪声干扰. 然后建立高炉透气性指数预测模型. 在建模过程中,将偏最小二乘(partial least square, PLS)与多层超 限学习机算法结合,消除多层超限学习机最后一层隐藏层的多重共线性,提高了模型预测精度.并且所提出的改进 算法称为PLS–ML–ELM. 最后使用现场生产数据对该模型训练和测试,预测结果表明所提出模型能够快速、精确地 预测高炉透气性指数, 并且为高炉的后续操作提供有效的决策与支持.

英文摘要

      Permeability index of blast furnace is one of signi?cant indicators of measuring the anterograde state of blast furnace for operators. Aiming at the defects of traditional permeability index measurement model, this paper proposes a prediction model for permeability index based on improved multi-layer extreme learning machine algorithm (ML–ELM). Firstly, relevant operation parameters are chosen through analyzing the mechanism of blast furnace. Given to blast furnace production data contain noise, wavelet transform is adopted to get rid of interference. Secondly, the prediction model of permeability index is established. Multi-layer extreme learning machine and partial least square method (PLS) are combined to overcome output matrix multicollinearity of the last hidden layer for ML–ELM and prediction accuracy is improved. And the improved algorithm is named as PLS–ML–ELM. Finally, practical production data are used to train and test this model. Simulation results indicate that the model can quickly and accurately predict permeability index and can offer ef?cient decision for sequent blast furnace operation