基于LSTM-RNN模型的铁水硅含量预测

doi:10.11949/j.issn.0438-1157.20171534

摘要/Abstract

摘要：

针对高炉炼铁是一个动态过程，具有大延迟，工况复杂的特性。采用LSTM-RNN模型进行硅含量预测，充分发挥了其处理时间序列时挖掘前后关联信息的优势。首先根据时间序列趋势及相关系数选择自变量，并采用复杂工况的实际生产数据进行验证。然后用程序自动求解最优参数进行硅含量预测。最后将LSTM-RNN模型与PLS模型及RNN模型的结果进行对比，验证该方法的优势。研究发现LSTM-RNN模型预测误差稳定，预测精度较高，比传统的统计学及神经网络方法取得了更好的预测精度。

关键词: 预测, 动态建模, 神经网络, 高炉炼铁, 硅含量

Abstract:

The ironmaking in blast furnace, with large delay and complex conditions, is a dynamic process. The traditional methods for prediction of silicon content in hot metal are mostly based on the statistics or the simple neural networks, leading to lower accuracy. However, a model based on the long short-term memory-recurrent neural network (LSTM-RNN) is proposed to exploit the characteristics of the mutual information before and after the time series in this paper. The independent variables are selected according to the time series trend and the correlation coefficient. After that, the silicon content is predicted according to the input variables by optimizing the parameters automatically. In order to verify the constructed model, the extremely complex production data is used to compare the LSTM-RNN and simple RNN models. Remarkably, the result shows that the prediction error of LSTM-RNN model is stable and the prediction accuracy is high.

Key words: prediction, dynamic modelling, neural network, ironmaking, silicon content

中图分类号:

TQ028.8

李泽龙, 杨春节, 刘文辉, 周恒, 李宇轩. 基于LSTM-RNN模型的铁水硅含量预测[J]. 化工学报, 2018, 69(3): 992-997.

LI Zelong, YANG Chunjie, LIU Wenhui, ZHOU Heng, LI Yuxuan. Research on hot metal Si-content prediction based on LSTM-RNN[J]. CIESC Journal, 2018, 69(3): 992-997.

参考文献 33

[1]	李界家, 杨志宇, 曹阳. 铁水硅含量的集成模糊神经网络预测方法[J]. 计算机与应用化学, 2013, 30(10):1113-1116. LI J, YANG Z Y, CAO Y. Research on hot metal silicon content prediction based on integrated neural network[J]. Computers and Applied Chemistry, 2013, 30 (10):1113-1116.
[2]	TAHASHI H, KAWAI H, KOBAYASHI M, et al. Two dimensional cold model study on unstable solid descending motion and control in blast furnace operation with low reducing agent rate[J]. ISIJ International, 2005, 45(10):1386-1395.
[3]	CHU M S, YANG X F, SHEN F M. Numerical simulation of innovative operation of blast furnace based on multi-fluid model[J]. Journal of Iron and Steel Research, International, 2006, 13(6):8-15.
[4]	赵敏. 高炉冶炼过程的复杂机理及其预测研究[D]. 杭州:浙江大学, 2008. ZHAO M. Complexity mechanism and predictive research for BF ironmaking process[D]. Hangzhou:Zhejiang University, 2008.
[5]	NOGAMI H, CHU M S, YAGI J. Multi-dimensional transient mathematical simulator of blast furnace process based on multi-fluid and kinetic theories[J]. Computers & Chemical Engineering, 2005, 29(11):2438-2448.
[6]	郜传厚, 渐令, 陈积明, 等. 复杂高炉炼铁过程的数据驱动建模及预测算法[J]. 自动化学报, 2009, 35(6):725-730. GAO C H, JIAN L, CHEN J M, et al. Data-driven modeling and predictive algorithm for complex blast furnace iron -making process[J]. Acta Automatica Sinica, 2009, 35(6):725-730.
[7]	SAXEN H. Short-term prediction of silicon content in pig iron[J].Canadian Metallurgical Quarterly, 1994, 33(4):319-326.
[8]	LUO S H, GAO C H, ZENG J S, et al. Blast furnace system modeling by multivariate phase space reconstruction and neural networks[J]. Asian Journal of Control, 2013, 15(2):553. 561.
[9]	安剑奇, 陈易斐, 吴敏. 基于改进支持向量机的高炉一氧化碳利用率预测方法[J]. 化工学报, 2015, 66(1):206-214. AN J Q, CHEN Y F, WU M. A prediction method for carbon monoxide utilization ratio of blast furnace based on improved support vector regression[J]. CIESC Journal, 2015, 66(1):206-214.
[10]	曾燕飞, 李小伟. 基于BP神经网络的高炉铁水硅含量预测模型研究[J]. 控制与测量, 2006, 22(19):291-293. ZENG Y F, LI X W. Model prediction model of blast furnace hot metal Si-content based on BP neural network's study[J]. Control and Measurement, 2006, 22(19):291-293.
[11]	宋菁华, 杨春节, 周哲, 等. 改进型EMD-Elman神经网络在铁水硅含量预测中的应用[J]. 化工学报, 2016, 67(3):729-735. SONG J H, YANG C J, ZHOU Z, et al. Application of improved EMD-Elman neural network to predict silicon content in hot metal[J]. CIESC Journal, 2016, 67(3):729-735.
[12]	蒋朝辉, 董梦林, 桂卫华, 等. 基于Bootstrap的高炉铁水硅含量二维预报[J]. 自动化学报, 2016, 42(5):715-723. JIANG Z H, DONG M L, GUI W H, et al. Two-dimensional prediction for silicon content of hot metal of blast furnace based on Bootstrap[J]. Acta Automatica Sinica, 2016, 42(5):715-723.
[13]	宋贺达, 周平, 王宏, 等. 高炉炼铁过程多元铁水质量非线性子空间建模及应用[J]. 自动化学报, 2016, 42(21):1664-1679. SONG H D, ZHOU P, WANG H, et al. Non-linear subspace modeling of multivariate molten iron qualityin blast furnace ironmaking and its application[J]. Acta Automatica Sinica, 2016, 42(21):1664-1679.
[14]	ZHOU H, YANG C J, LIU W H, et al. A sliding-window T-S fuzzy neural network model for prediction of silicon content in hot metal[C]//20th IFAC World Congress. IFAC-PapersOnLine, 2017, 50(1):14988-14991.
[15]	SENIOR A. Context dependent phone models for LSTM RNN acoustic modeling[J]. IEEE Int. Conf. Acoust, Speech Signal Process, 2015, (1):4585-4589.
[16]	LIU C J, WANG Y Q, KSHITIZ K, et al. Investigations on speaker adaptation of LSTM-RNN models for speech recognition[C]//IEEE International Conference on ICASSP, 2016:5020-5024.
[17]	GRAVES A, JAITLY N, MOHAMED A. Hybrid speech recognition with deep bi-directional LSTM[C]//Proceedings of the IEEE Workshop on Automatic Speech Recognition and Understanding, 2013.
[18]	SAK H, SENIOR A, BEAUFAYS F. Long short-term memory recurrent neural network architectures for large scale acoustic modeling[C]//Annual Conference of the International Speech Communication Association (Interspeech), 2014:338-342.
[19]	ZACHARY C LIPTON. Learning to diagnose with LSTM recurrent neural networks[C]//ICLR 2016. Computer Science, 2017.
[20]	GRAVES A, SCHMIDHUBER J. Frame wise phoneme classification with bidirectional LSTM and other neural network architectures[J]. Neural Networks, 2005, 18(5/6):602-610.
[21]	SONG E, KANG H G. Multi-class learning algorithm for deep neural network-based statistical parametric speech synthesis[C]//Proc. 24th Eur. Signal Process Conf., 2016, 1951-1955.
[22]	ACHANTA S, GODAMBE T, GANGASHETTY S V. An investigation of recurrent neural network architectures for statistical parametric speech synthesis[C]//Proc. Interspeech, 2015:859-863.
[23]	ZEN H, SAK H. Unidirectional long short-term memory recurrent neural network with recurrent output layer for low-latency speech synthesis[C]//IEEE Speech Signal Process, 2015:4470-4474.
[24]	BYEON W, LIWICKI M, BREUEL T. Texture classification using 2D LSTM networks[C]//Pattern Recognition International Conference, 2014:1144-1149.
[25]	BYEON W, BREUEL T M. Supervised texture segmentation using 2D LSTM networks[C]//Processing 2014 IEEE International Conference, 2014:4373-4377.
[26]	PINHEIRO P, COLLOBERT R, JEBARAT, et al. Recurrent convolutional neural networks for scene labeling[C]//Proceedings of the 31 st International Conference on Machine Learning (ICML-14), 2014:82-90.
[27]	SOCHER R, HUVAL B, BATH B, et al. Convolutional recursive deep learning for 3D object classification[C]//Advances in Neural Information Processing Systems, 2012:665-673.
[28]	GONZALEZ-DOMINGUEZ J, LOPEZ-MORENO I, MORENO P J, et al. Frame by frame language identification in short utterances using deep neural networks[J]. Neural Networks, 2015, 64(C):49-58.
[29]	GERS F A, SCHMIDHUBER J, CUMMINS F. Learning to forget:continual prediction with LSTM[J]. Neural Computation, 2000, 12(10):2451-2471.
[30]	HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8):1735-1780.
[31]	BENGIO Y, SIMARD P, FRASCONI P. Learning long-term dependencies with gradient descent is difficult[J]. IEEE Transactions on Neural Networks, 1994, 5(2):157-166.
[32]	GREFF K, SRIVASTAVA R K, KOUTNÍK J, et al. LSTM:a search space Odyssey[J]. IEEE Transactions on Neural Networks & Learning Systems, 2015, 28 (10):2222.
[33]	GERS F A, SCHRAUDOLPH N N, SCHMIDHUBER J. Learning precise timing with LSTM recurrent networks[J]. Journal of Machine Learning Research, 2003, 3(1):115-143.