In view of the characteristics of multivariate strong coupling, long lag, and dynamic fluctuation of gas utilization in blast furnace, a TPE-TCN-SAM dynamic prediction model based on multi-time scale fusion was proposed. First, noise, missing values, and temporal misalignment in the raw data were repaired through multi-strategy cleaning and interpolation, and variables were aligned by combining lag correlation analysis. Then, 13 key influencing variables were selected using an improved random forest (IRF) algorithm. In model construction, a tree-structured Parzen estimator (TPE) was used to optimize hyperparameters of temporal convolutional networks (TCN), and input weights were adjusted by self-adaptive modulation (SAM) according to working conditions to improve the capability of long-period lag modeling and non-stationary feature mining. The measured results of the blast furnace indicate that MAE, RMSE, and MAPE of the model reach 0.403%, 0.504%, and 0.894%, respectively, and its accuracy and robustness are superior to those of comparison models. The model was integrated into the steel enterprise’s intelligent system, with a one-hour early warning capability, providing support for gas regulation and energy management.
ZhangL, YuanY X, XiJ X, et al. Synergistic enhancement for energy-saving, emission reduction and profit improvement in iron and steel manufacturing system: strategies for parameter regulation and technologies integration [J]. Energy Conversion and Management, 2024, 322: 119101.
[2]
KimJ, HanS S, KimJ, et al. Vacuum pressure swing adsorption for efficient off-gas recycling: techno-economic and CO2 abatement study [J]. Energy, 2023, 264: 126281.
[3]
NathN K. Simulation of gas flow in blast furnace for different burden distribution and cohesive zone shape [J]. Materials and Manufacturing Processes, 2002, 17(5): 671-681.
ChuMan-sheng, WangGuo-dong, TangJue, et al. Research progress on development and application of digital blast furnace ironmaking technology [J]. Journal of Northeastern University (Natural Science), 2025, 46(7): 113-130.
LiYang-long, ChengShu-sen, ChenChuan. Mathematical model of adjusting blast volume of blast furnace tuyeres [J]. Journal of Northeastern University (Natural Science), 2016, 37(3): 357-362.
[8]
SpirinN A, RybolovlevV Y, LavrovV V, et al. Scientific problems in creating intelligent control systems for technological processes in pyrometallurgy based on industry 4.0 concept [J]. Metallurgist, 2020, 64(5): 574-580.
[9]
MaL, LiY N, ZhangT L, et al. Trustworthy battery state of charge estimation enabled by multi-task deep learning [J]. Energy, 2025,326: 136264.
[10]
DettoriS, MatinoI, CollaV, et al. Neural network-based modeling methodologies for energy transformation equipment in integrated steelworks processes [J]. Energy Procedia, 2019, 158: 4061-4066.
[11]
SantoG C M, GarciaC, ChavesC R. System identification with historical data: a multivariable approach applied to a 50 MW petrochemical furnace [J]. Computers & Chemical Engineering, 2022, 157: 107623.
[12]
ShinS J, WooJ, RachuriS. Energy efficiency of milling machining: component modeling and online optimization of cutting parameters [J]. Journal of Cleaner Production, 2017, 161: 12-29.
[13]
XuH W, QinW, SunY N, et al. Attention mechanism-based deep learning for heat load prediction in blast furnace ironmaking process [J]. Journal of Intelligent Manufacturing, 2024, 35(3): 1207-1220.
[14]
ShiQ, TangJ, ChuM S. Key issues and progress of industrial big data-based intelligent blast furnace ironmaking technology [J]. International Journal of Minerals, Metallurgy and Materials, 2023, 30(9): 1651-1666.
[15]
ShiQ, TangJ, ChuM S. Process metallurgy and data-driven prediction and feedback of blast furnace heat indicators [J]. International Journal of Minerals, Metallurgy and Materials, 2024, 31(6): 1228-1240.
ShiLin, LiuWen-hui, CaoFu-jun, et al. Combined forecast of blast furnace gas utilization rate based on CEEMDAN-SVM-LSTM [J]. China Measurement & Testing Technolgy, 2023, 49(1): 86-91.
[18]
AnJ Q, ShenX L, WuM, et al. A multi-time-scale fusion prediction model for the gas utilization rate in a blast furnace [J]. Control Engineering Practice, 2019, 92: 104120.
JinYa-tao, ZhangYu-jie, LiuXiao-jie, et al. Prediction for utilization rate of blast furnace gas based on EEMD-BAS-RBF [J]. Journal of Iron and Steel Research, 2023, 35(11): 1330-1338.
GuoYun-peng, AnJian-qi, ZhaoGuo-yu. Time-series prediction model of gas utilization rate in a blast furnace considering smelting intensity classification [J]. Metallurgical lndustry Automation, 2024, 48(2): 60-73.
AnJian-qi, ZhaoGuo-yu, HeYong, et al. Data-driven multi-time scale model predictive control for blast furnace gas utilization rate [J]. Control Theory & Applications, 2025, 42(1): 189-201.
DangXiao-jing, ShiLin, ZhaoNa, et al. Prediction of utilization ratio of blast furnace gas based on parameter optimized by SVR method [J]. Journal of Iron and Steel Research, 2021, 33(4): 279-283.
[27]
ZhaoY F, ZhuW W, WeiP P, et al. Classification of Zambian grasslands using random forest feature importance selection during the optimal phenological period [J]. Ecological Indicators, 2022, 135: 108529.
LiDeng-hui, ZhaoYan, LeiHong, et al. Intelligent prediction for endpoint mass fraction of carbon in molten steel of RH [J]. Journal of Northeastern University(Natural Science), 2025, 46(10): 51-58.
JiYing-jun, YongXiao-yue, LiuYing-lin, et al. Random forest based quality analysis and prediction method for hot-rolled strip [J]. Journal of Northeastern University(Natural Science), 2019, 40(1): 11-15.
[32]
TaoS T, PengP, LiY F, et al. Supervised contrastive representation learning with tree-structured parzen estimator Bayesian optimization for imbalanced tabular data [J]. Expert Systems with Applications, 2024, 237: 121294.
[33]
IslamR, SultanaA, TuhinM D N. A comparative analysis of machine learning algorithms with tree-structured parzen estimator for liver disease prediction [J]. Healthcare Analytics, 2024, 6: 100358.
[34]
BiJ, ZhangX, YuanH T, et al. A hybrid prediction method for realistic network traffic with temporal convolutional network and LSTM [J]. IEEE Transactions on Automation Science and Engineering, 2022, 19(3): 1869-1879.
[35]
DudukcuH V, TaskiranM, Cam TaskiranZ G, et al. Temporal convolutional networks with RNN approach for chaotic time series prediction [J]. Applied Soft Computing, 2023, 133: 109945.