基于BP神经网络的COREX铁水硅含量预测模型
|
摘要
COREX铁水硅含量偏高且易波动一直是生产过程中面临的难题,而精准预测COREX铁水硅含量可为稳定并降低铁水硅含量提供理论依据和技术参考。利用BP神经网络建立了COREX铁水硅含量预测模型,通过相关分析法确定模型的输入参数,采用计算邓氏关联度的方式确定各参数对应的滞后炉次。并利用某钢厂COREX实际生产数据分别进行学习和验证,结果表明预测误差为±0.1%时,其命中率为80%。为提高模型的预测精度,在该模型的基础上,采用时间序列推移法,实时更新训练样本,优化模型。研究结果表明,改进后的模型预测误差为±0.1%时,命中率是90%,提高了模型预测精度。该模型可为判断铁水硅含量变化以及后续操作提供理论依据。
The high and fluctuation property of [Si] content in hot metal(HM) is always a problem in COREX process. The precise prediction of [Si] content in HM from COREX process can provide a theoretical basis and technical reference for stabilizing and reducing the [Si] content in HM. A back propagation(BP) neural network was established to predict the [Si] content in HM from COREX process. The input parameters of the model were determined by correlation analysis, and the hysteretic heats corresponding to each parameter were determined by calculating the Deng′s relevancy. The results show that when the prediction error is ±0.1%, the hit rate is 80%. The method of continuous updating the training samples was used to improve the prediction accuracy of the model. The prediction results show that the hit rate is 90% in absolute error range of ±0.1%, and the prediction accuracy has been greatly improved compared with previous model. The improved model can provide a theoretical basis for judging the change of [Si] content in HM and subsequent operations.
The high and fluctuation property of [Si] content in hot metal(HM) is always a problem in COREX process. The precise prediction of [Si] content in HM from COREX process can provide a theoretical basis and technical reference for stabilizing and reducing the [Si] content in HM. A back propagation(BP) neural network was established to predict the [Si] content in HM from COREX process. The input parameters of the model were determined by correlation analysis, and the hysteretic heats corresponding to each parameter were determined by calculating the Deng′s relevancy. The results show that when the prediction error is ±0.1%, the hit rate is 80%. The method of continuous updating the training samples was used to improve the prediction accuracy of the model. The prediction results show that the hit rate is 90% in absolute error range of ±0.1%, and the prediction accuracy has been greatly improved compared with previous model. The improved model can provide a theoretical basis for judging the change of [Si] content in HM and subsequent operations.
引文
[1] 许海川, 齐渊洪. COREX炼铁煤气生产热压块流程的模拟分析[J]. 钢铁, 2007, 42(4):16.(Xu H C, Qi Y H. Simulation analysis of shaft production of HBI with COREX offgas[J]. Iron and Steel, 2007, 42(4):16.)
[2] 曲迎霞, 王臣, 邹宗树, 等. COREX工艺模型及应用分析[J]. 过程工程学报, 2008, 8(S1):68.(Qu Y X, Wang C, Zou Z Z, et al. Process modeling of COREX and its application analysis[J]. The Chinese Journal of Process Engineering, 2008, 8(S1):68.)
[3] Wu S L, Xu J, Yagi J I, et al. Prediction of pre-reduction shaft furnace with top gas recycling technology aiming to cut down CO2 emission[J]. ISIJ International, 2011, 51(8):1344.
[4] Shen W, Wu S L, Kou M Y, et al. Establishment of a static model based on measured heat loss for COREX process[J]. Journal of Iron and Steel Research International, 2015, 22(3):200.
[5] 张军红, 金永龙, 沈峰满, 等. 应用优化BP神经网络建立铁水硅含量的预测模型[J]. 钢铁研究学报, 2007, 19(11):60.(Zhang J H, Jin Y L, Shen F M, et al. Prediction model of silicon content in hot metal using optimized BP Network[J]. Journal of Iron and Steel Research, 2015, 22(3):200.)
[6] Kumar P P, Dasu A, Ranjan M, et al. Influence of operational parameters on silicon in hot metal from COREX[J]. Ironmaking and Steelmaking, 2008, 35(2):108.
[7] Wu B, Han S J, Xiao J, et al. Error compensation based on BP neural network for airborne laser ranging[J]. Optik-International Journal for Light and Electron Optics, 2016, 127(8):4083.
[8] 吕庆. 熔融气化炉内Si的反应机理与分配规律[J]. 金属学报, 1998, 34(7):763.(Lü Q. Reaction mechanism and distribution rule of silicon in melting gasifier[J]. Acta Metallurgica Sinica, 1998, 34(7):763.)
[9] Han L H, Huang W Q, Liu Y X. Effects of operation parameters on cohesive zone in COREX melter gasifier[J]. ISIJ International, 2016, 58(9):1559.
[10] 刘金琨, 邓守强, 苏士权. 高炉铁水硅含量的神经网络时间序列预报[J]. 钢铁研究学报, 1996, 8(3):63.(Liu J K, Deng S Q, Su S Q. Time-series prediction of silicon content in hot metal by neural network[J]. Journal of Iron and Steel Research, 1996, 8(3):63.)
[11] Renno C, Petito F, Gatto A. Artificial neural network models for predicting the solar radiation as input of a concentrating photovoltaic system[J]. Energy Conversion and Management, 2015, 106:999.
[12] 伍春香, 刘琳, 王葆元. 三层BP网隐层节点数确定方法的研究[J]. 武汉测绘科技大学学报, 1999, 24(2):85.(Wu C X, Liu L, Wang B Y. The study of the method to determining the number of hidden units of three-layer BP neural networks[J]. Journal of Wuhan Technical University of Surveying and Mapping, 1999, 24(2):85.)
[13] Hakim S J S, Noorzaei J, Jaafar M A, et al. Application of artificial neural networks to predict compressive strength of high strength concrete[J]. International Journal of the Physical Sciences, 2011, 6(5): 975.
[14] Esfe M H, Afrand M, Yan W M, et al. Applicability of artificial neural network and nonlinear regression to predict thermal conductivity modeling of Al2O3-water nanofluids using experimental data[J]. International Communications in Heat and Mass Transfer, 2015, 66:246.
[15] 周昌鸿, 张运陶. 一种改进BP神经网络算法的编程及应用[J]. 西华师范大学学报: 自然科学版, 2009, 30(3):1559.(Zhou C H, Zhang Y T. Programme of an improved BP neural network and its application[J]. Journal of China West Normal University: Natural Sciences, 2009, 30(3):1559.)
[2] 曲迎霞, 王臣, 邹宗树, 等. COREX工艺模型及应用分析[J]. 过程工程学报, 2008, 8(S1):68.(Qu Y X, Wang C, Zou Z Z, et al. Process modeling of COREX and its application analysis[J]. The Chinese Journal of Process Engineering, 2008, 8(S1):68.)
[3] Wu S L, Xu J, Yagi J I, et al. Prediction of pre-reduction shaft furnace with top gas recycling technology aiming to cut down CO2 emission[J]. ISIJ International, 2011, 51(8):1344.
[4] Shen W, Wu S L, Kou M Y, et al. Establishment of a static model based on measured heat loss for COREX process[J]. Journal of Iron and Steel Research International, 2015, 22(3):200.
[5] 张军红, 金永龙, 沈峰满, 等. 应用优化BP神经网络建立铁水硅含量的预测模型[J]. 钢铁研究学报, 2007, 19(11):60.(Zhang J H, Jin Y L, Shen F M, et al. Prediction model of silicon content in hot metal using optimized BP Network[J]. Journal of Iron and Steel Research, 2015, 22(3):200.)
[6] Kumar P P, Dasu A, Ranjan M, et al. Influence of operational parameters on silicon in hot metal from COREX[J]. Ironmaking and Steelmaking, 2008, 35(2):108.
[7] Wu B, Han S J, Xiao J, et al. Error compensation based on BP neural network for airborne laser ranging[J]. Optik-International Journal for Light and Electron Optics, 2016, 127(8):4083.
[8] 吕庆. 熔融气化炉内Si的反应机理与分配规律[J]. 金属学报, 1998, 34(7):763.(Lü Q. Reaction mechanism and distribution rule of silicon in melting gasifier[J]. Acta Metallurgica Sinica, 1998, 34(7):763.)
[9] Han L H, Huang W Q, Liu Y X. Effects of operation parameters on cohesive zone in COREX melter gasifier[J]. ISIJ International, 2016, 58(9):1559.
[10] 刘金琨, 邓守强, 苏士权. 高炉铁水硅含量的神经网络时间序列预报[J]. 钢铁研究学报, 1996, 8(3):63.(Liu J K, Deng S Q, Su S Q. Time-series prediction of silicon content in hot metal by neural network[J]. Journal of Iron and Steel Research, 1996, 8(3):63.)
[11] Renno C, Petito F, Gatto A. Artificial neural network models for predicting the solar radiation as input of a concentrating photovoltaic system[J]. Energy Conversion and Management, 2015, 106:999.
[12] 伍春香, 刘琳, 王葆元. 三层BP网隐层节点数确定方法的研究[J]. 武汉测绘科技大学学报, 1999, 24(2):85.(Wu C X, Liu L, Wang B Y. The study of the method to determining the number of hidden units of three-layer BP neural networks[J]. Journal of Wuhan Technical University of Surveying and Mapping, 1999, 24(2):85.)
[13] Hakim S J S, Noorzaei J, Jaafar M A, et al. Application of artificial neural networks to predict compressive strength of high strength concrete[J]. International Journal of the Physical Sciences, 2011, 6(5): 975.
[14] Esfe M H, Afrand M, Yan W M, et al. Applicability of artificial neural network and nonlinear regression to predict thermal conductivity modeling of Al2O3-water nanofluids using experimental data[J]. International Communications in Heat and Mass Transfer, 2015, 66:246.
[15] 周昌鸿, 张运陶. 一种改进BP神经网络算法的编程及应用[J]. 西华师范大学学报: 自然科学版, 2009, 30(3):1559.(Zhou C H, Zhang Y T. Programme of an improved BP neural network and its application[J]. Journal of China West Normal University: Natural Sciences, 2009, 30(3):1559.)