Dynamic, data-driven prediction model of molten steel temperature in the second blowing stage of a converter
-
摘要: 轉爐鋼水溫度是轉爐終點控制的工藝參數之一，精確的鋼水溫度預測對轉爐終點控制具有重要的指導意義。然而，以往的大多數轉爐終點預測模型屬于靜態模型，只能夠實現對轉爐吹煉終點鋼水溫度的預測，無法實現動態預測，導致模型的作用有限。針對該問題，提出了一種基于數據驅動的轉爐二吹階段鋼水溫度動態預測模型。模型先通過新案例主吹階段的工藝參數，基于案例推理算法找到歷史案例庫中相似案例。再利用相似案例的二吹階段工藝參數并基于長短期記憶網絡（Long short-term memory，LSTM）算法訓練工藝參數與鋼水溫度的變化關系。然后利用訓練好的LSTM模型，計算新案例二吹階段的鋼水溫度變化。最后，利用某鋼廠實際生產數據，研究了不同重用案例個數及神經元個數對模型預測精度的影響，實驗結果表明：模型在重用案例個數為4，神經元個數為10時模型的預測精度最高，此時模型對鋼水溫度的預測誤差在[?5 ℃, 5 ℃]、[?10 ℃,10 ℃]和[?15 ℃,15 ℃]的命中率分別達到40.33%、68.92%和88.33%，模型的性能高于傳統二次方模型和三次方模型。Abstract: Molten steel temperature is a parameter in converter end-point control. Accurate prediction of molten steel temperature is crucial for converter end-point control. However, most of the previous end-point prediction models are static models, which can only predict the molten steel temperature at the end-point of converter blowing and cannot realize dynamic prediction, affording a limited role for these models. To solve this challenge, a data-driven prediction model of molten steel temperature in the second blowing stage in a converter is proposed. First, the model retrieves the similar cases in the historical case base through the process parameters in the main blowing stage of the new case, such as carbon content and temperature of TSC measurement, based on the case-based reasoning (CBR) algorithm. Second, the process parameters in the second blowing stage of the similar cases, such as oxygen flow, lance position, and argon flow, are used to train the relationship between the process parameters and the molten steel temperature based on the long short-term memory (LSTM) algorithm. Third, the trained LSTM model is used to dynamically calculate the molten steel temperature in the second blowing stage of the new case. Finally, the actual production data is divided into five sets for cross-validation, and the model prediction accuracy changes are tested when the number of reuse cases ranges from 1 to 10, and the number of neurons is 5, 10, 15, and 20. The results show that, on the one hand, the prediction accuracy of the model first increases and then decreases with an increasing number of cases, and when the number of reused cases is 4, the prediction accuracy of the model is the highest, indicating that the number of cases is increased when training the model. Improving the prediction accuracy of the model is beneficial; however, the reference value of the case decreases with the similarity of the case, reducing the prediction accuracy of the model. Conversely, when the number of neurons is 10, the prediction accuracy of the model reaches it’s the highest value. The hit rate of the prediction error in the range of [?5 ℃, 5 ℃], [?10 ℃, 10 ℃], and [?15 ℃, 15 ℃] reached 40.33%, 68.92%, and 88.33%, respectively. This paper also establishes the traditional quadratic model and cubic model as well as further proves the effectiveness of the model by comparing the three indicators of these models, namely, the RMSE, MSE, and hit rate.
-
圖 9 相似案例對應的二吹階段工藝參數圖.（a）爐次2；（b）爐次3；（c）爐次4；（d）爐次5
Figure 9. Process parameters for the second blowing stage of similar cases: (a) Heat 2; (b) Heat 3; (c) Heat 4; (d) Heat 5
圖 12 CBR_LSTM模型預測結果. (a) k=5; (b) k=10; (c) k=15; (d) k=20
Figure 12. Prediction result of the model CBR_LSTM: (a) k = 5; (b) k = 10; (c) k = 15; (d) k = 20
表 1 轉爐主吹階段單值型工藝參數統計結果
Table 1. Statistical results of single-value type process parameters in main blowing stage of converter
Influence factors Symbols Maximum Minimum Mean Standard deviation Carbon content of hot metal/% X1 4.5016 3.9496 4.2273 0.0992 Silicon content of hot metal/% X2 0.49606 0.03189 0.2644 0.08119 Manganese content of hot metal/% X3 0.14347 0.07155 0.10811 0.01291 Phosphorus content of hot metal/% X4 0.07953 0.0527 0.065721 0.004375 Temperature of hot metal/℃ X5 1441 1267 1360.3 31.3 Weight of hot metal/t X6 282 263 272.57 3.39 Weight of scrap/t X7 71 40 60.262 6.207 Amount of lime/t X8 20.605 2.063 9.4865 3.6389 Amount of dolomite/t X9 7.991 2.001 4.7244 1.0092 Oxygen amount of main blowing stage/m3 X10 15022 11298 12931 560 TSC[C]/% X11 0.559 0.038 0.24437 0.10683 TSC[T]/℃ X12 1690 1545 1617.3 24.9 TSO[C]/% Y1 0.066 0.018 0.04225 0.00782 TSO[T]/℃ Y2 1705 1620 1663.42 33.71 
下載: 導出CSV
表 2 轉爐二吹階段工藝參數統計結果
Table 2. Statistical results of process parameters in the second blowing stage of conventer
Influence factors Maximum Minimum Maximum length of time-series/min Minimum length of time-series/min Oxygen flow 66282 m3·h?1 29998 m3·h?1 2 1 Lance position 5553 mm 1598 mm 2 1 Argon flow 2155 m3·h?1 465 m3·h?1 2 1
下載: 導出CSV
表 3 相似案例檢索結果
Table 3. Similarity between the new case and similar cases
Heat No. X1/% X2/% X3/% X4/% X5/℃ X6/t X7/t X8/t X9/t X10/ m3 X11/% X12/℃ Similarity 1 4.13317 0.14899 0.09142 0.06122 1377 273 63 5.981 5.276 12884 0.241 1584 2 4.10874 0.09697 0.08969 0.06242 1376 275 63 4.219 4.755 12659 0.207 1577 94.25% 3 4.05288 0.13592 0.0925 0.0605 1394 273 61 5.53 4.545 12878 0.226 1603 93.18% 4 4.04647 0.15276 0.09448 0.05964 1381 272 65 6.081 4.906 13373 0.299 1597 91.95% 5 4.1307 0.14709 0.0927 0.05653 1354 273 65 6.074 4.596 13101 0.303 1586 91.41%
下載: 導出CSV
表 4 模型擬合結果
Table 4. Model fitting results
Model Fitting formula R2 Quadratic model $T = - {10^{ - 5} } \cdot {x^2} + 0.03 \cdot x + {\rm{TS}}{{\rm{C}}_{[{\rm{T}}]} }$ 0.994 Cubic model $T = 5 \cdot {10^{ - 9} } \cdot {x^3} - 2 \cdot {10^{ - 5} } \cdot {x^2} + {\rm{TS}}{{\rm{C}}_{[{\rm{T}}]} }$ 0.9987
下載: 導出CSV
表 5 各模型預測精度對比
Table 5. Prediction accuracies of each model
Model MAE RMSE Hit rate/% Quadratic model 10.74 13.61 76.50 Cubic model 9.37 11.69 80.67 CBR_LSTM 7.54 9.38 88.33
下載: 導出CSV
www.77susu.com<span id="fpn9h"><noframes id="fpn9h"><span id="fpn9h"></span> <span id="fpn9h"><noframes id="fpn9h"> <th id="fpn9h"></th> <strike id="fpn9h"><noframes id="fpn9h"><strike id="fpn9h"></strike> <th id="fpn9h"><noframes id="fpn9h"> <span id="fpn9h"><video id="fpn9h"></video></span> <ruby id="fpn9h"></ruby> <strike id="fpn9h"><noframes id="fpn9h"><span id="fpn9h"></span> -
參考文獻
[1] Gao F, Bao Y P, Wang M, et al. Prediction model of end-point phosphorus content of converter based on FA-ELM. Iron Steel, 2020, 55(12): 24高放, 包燕平, 王敏, 等. 基于FA-ELM的轉爐終點磷含量預測模型. 鋼鐵, 2020, 55(12):24 [2] Feng K, Xu A J, He D F, et al. An improved CBR model based on mechanistic model similarity for predicting end phosphorus content in dephosphorization converter. Steel Res Int, 2018, 89(6): 1800063 doi: 10.1002/srin.201800063 [3] Gu M Q, Xu A J, Yuan F, et al. An improved CBR model using time-series data for predicting the end-point of a converter. ISIJ Int, 2021, 61(10): 2564 doi: 10.2355/isijinternational.ISIJINT-2020-687 [4] Gao C, Shen M G, Liu X, et al. End-point static control of basic oxygen furnace (BOF) steelmaking based on wavelet transform weighted twin support vector regression. Complexity, 2019, 2019: 7408725 [5] Sala D A, Jalalvand A, Van Yperen-De Deyne A, et al. Multivariate time series for data-driven endpoint prediction in the basic oxygen furnace // 2018 17th IEEE International Conference on Machine Learning and Applications (ICMLA). Orlando, 2018: 1419 [6] He F, Zhang L Y. Prediction model of end-point phosphorus content in BOF steelmaking process based on PCA and BP neural network. J Process Control, 2018, 66: 51 doi: 10.1016/j.jprocont.2018.03.005 [7] Cheng J, Wang J. Endpoint prediction method for steelmaking based on multi-task learning. J Comput Appl, 2017, 37(3): 889 doi: 10.11772/j.issn.1001-9081.2017.03.889程進, 王堅. 基于多任務學習的煉鋼終點預測方法. 計算機應用, 2017, 37(3):889 doi: 10.11772/j.issn.1001-9081.2017.03.889 [8] Yang X M, Zhao Y, Zhong L C, et al. Endpoint prediction of converter steelmaking based on XGBoost algorithm. Steelmaking, 2021, 37(6): 1楊曉猛, 趙陽, 鐘良才, 等. 基于XGBoost算法的轉爐吹煉終點預報. 煉鋼, 2021, 37(6):1 [9] Xuan M T, Li J J, Wang N, et al. Endpoint prediction of basic oxygen furnace steelmaking based on FOA-GRNN model. J Mater Metall, 2019, 18(1): 31鉉明濤, 李嬌嬌, 王楠, 等. 基于FOA-GRNN模型的轉爐煉鋼終點預報. 材料與冶金學報, 2019, 18(1):31 [10] Han M, Zhao Y, Yang X L, et al. Endpoint prediction model of basic oxygen furnace steelmaking based on robust relevance-vector-machines. Control Theory Appl, 2011, 28(3): 343韓敏, 趙耀, 楊溪林, 等. 基于魯棒相關向量機的轉爐煉鋼終點預報模型. 控制理論與應用, 2011, 28(3):343 [11] Liu C, Han M, Wang X Z. Endpoint prediction model for basic oxygen furnace steelmaking based on membrane algorithm evolving extreme learning machine. J Dalian Univ Technol, 2014, 54(1): 124 doi: 10.7511/dllgxb201401019劉闖, 韓敏, 王心哲. 基于膜算法進化極限學習機的氧氣轉爐煉鋼終點預報模型. 大連理工大學學報, 2014, 54(1):124 doi: 10.7511/dllgxb201401019 [12] Yan L T, Li M, Yang D Y. Prediction of carbon content at end point based on GA-KPLSR in converters. Control Eng China, 2017, 24(5): 923 doi: 10.14107/j.cnki.kzgc.150355嚴良濤, 李鳴, 楊大勇. 基于GA-KPLSR的轉爐終點碳含量的預測研究. 控制工程, 2017, 24(5):923 doi: 10.14107/j.cnki.kzgc.150355 [13] Zhang C J, Han Y, He S Y, et al. Furnace mouse flame spectrum driven steelmaking end control. Chin J Sci Instrum, 2018, 39(1): 24張彩軍, 韓陽, 何世宇, 等. 爐口火焰光譜驅動的煉鋼終點控制. 儀器儀表學報, 2018, 39(1):24 [14] Liu C, Liu H. Carbon content prediction of converter steelmaking end-point based on improved MTBCD flame image feature extraction. Comput Integr Manuf Syst, http://kns.cnki.net/kcms/detail/11.5946.TP.20210428.1806.020.html李超, 劉輝. 改進MTBCD火焰圖像特征提取的轉爐煉鋼終點碳含量預測. 計算機集成制造系統,http://kns.cnki.net/kcms/detail/11.5946.TP.20210428.1806.020.html [15] Zhu W Q, Zhou M C, Zhao Q, et al. End-point prediction of BOF steelmaking based on flame spectral feature selection using WCARS-ISPA. Spectrosc Spectr Anal, 2021, 41(8): 2332朱雯瓊, 周木春, 趙琦, 等. WCARS-ISPA 火焰光譜特征選擇的轉爐煉鋼終點預測. 光譜學與光譜分析, 2021, 41(8):2332 [16] Yue F, Bao Y P, Cui H, et al. Sub-lance control-based predication model for BOF end-point. Steelmaking, 2009, 25(1): 38岳峰, 包燕平, 崔衡, 等. 基于副槍控制的轉爐終點預測模型. 煉鋼, 2009, 25(1):38 [17] Wang X Z, Xing J, Dong J, et al. Data driven based endpoint carbon content real time prediction for BOF steelmaking // 2017 36th Chinese Control Conference (CCC). Dalian, 2017: 9708 [18] Hu Z G, He P, Liu L, et al. Continuous determination of bath carbon in BOF by off-gas analysis. Res Iron Steel, 2003, 31(3): 12 doi: 10.3969/j.issn.1001-1447.2003.03.004胡志剛, 何平, 劉瀏, 等. 利用爐氣分析進行轉爐鋼水連續定碳. 鋼鐵研究, 2003, 31(3):12 doi: 10.3969/j.issn.1001-1447.2003.03.004 [19] Hu Z G, Liu L, He P, et al. Continuous determination of bath temperature in BOF by off-gas analysis // Proceedings of 2003 China Iron & Steel Annual Meeting(3). Beijing, 2003: 613胡志剛, 劉瀏, 何平, 等. 利用爐氣分析進行鋼水連續定溫 // 中國金屬學會2003中國鋼鐵年會論文集(3). 北京, 2003: 613 [20] Liu K, Liu L, He P, et al. A new algorithm of endpoint carbon content of BOF based on of off-gas analysis. Steelmaking, 2009, 25(1): 33劉錕, 劉瀏, 何平, 等. 基于煙氣分析轉爐終點碳含量控制的新算法. 煉鋼, 2009, 25(1):33 [21] Wang X H, Li J Z, Liu F G. Technological progress of BOF steelmaking in period of development mode transition. Steelmaking, 2017, 33(1): 1王新華, 李金柱, 劉鳳剛. 轉型發展形勢下的轉爐煉鋼科技進步. 煉鋼, 2017, 33(1):1 [22] Liao D S, Sun S, Waterfall S, et al. Integrated KOBM steelmaking process control // Proceedings of the 6th International Congress on the Science and Technology of Steelmaking (Ⅰ). Beijing, 2015: 107 [23] Lin W H, Jiao S Q, Sun J K, et al. Modified exponential model for carbon prediction in the end blowing stage of basic oxygen furnace converter. Chin J Eng, 2020, 42(7): 854林文輝, 焦樹強, 孫建坤, 等. 轉爐吹煉后期碳含量預報的改進指數模型. 工程科學學報, 2020, 42(7):854 [24] Li N, Lin W H, Cao L L, et al. Carbon prediction model for basic oxygen furnace off-gas analysis based on bath mixing degree. Chin J Eng, 2018, 40(10): 1244李南, 林文輝, 曹玲玲, 等. 基于熔池混勻度的轉爐煙氣分析定碳模型. 工程科學學報, 2018, 40(10):1244 [25] Gu M Q, Xu A J, Wang H B, et al. Real-time dynamic carbon content prediction model for second blowing stage in BOF based on CBR and LSTM. Processes, 2021, 9(11): 1987 doi: 10.3390/pr9111987 [26] Hou Y M, Xu C Y. Review of case-based reasoning. J Yanshan Univ Philos Soc Sci Ed, 2011, 12(4): 102 doi: 10.15883/j.13-1277/c.2011.04.026侯玉梅, 許成媛. 基于案例推理法研究綜述. 燕山大學學報(哲學社會科學版), 2011, 12(4):102 doi: 10.15883/j.13-1277/c.2011.04.026 -
點擊查看大圖
圖(13) / 表(5)
計量
- 文章訪問數: 2320
- HTML全文瀏覽量: 200
- PDF下載量: 87
- 被引次數: 0
