Converter efficient smelting method based on big data steelmaking model under low molten iron ratio condition

Abstract

The invention discloses a converter efficient smelting method based on a big data steelmaking model under a low molten iron ratio condition. Obtaining converter smelting historical data; training by using historical data to obtain a first model; acquiring material data during actual converter smelting; inputting the material data into a first model to obtain first charging data; feeding is conducted according to the first feeding data, TSC measurement is conducted after the smelting stage is completed, and TSC measurement data are obtained; inputting the material data, the first charging data and the TSC measurement data into a first model to obtain second charging data; charging according to the second charging data, continuing blowing, and inputting the material data, the first charging data, the TSC measurement data and the second charging data into a first model to obtain TSO measurement data; judging whether the TSO measurement data meet target requirements or not; and if the target requirement is met, unequal sample tapping is executed. According to the scheme, the steel smelting process is assisted by utilizing the advantages of the large model, and the production efficiency and the product quality can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of steel smelting, and particularly to a converter high-efficiency smelting method based on a big data steelmaking model under the condition of a low hot metal ratio. Background Art

[0002] In the traditional heavy industry field of steel smelting, the introduction of technologies such as network models and artificial intelligence large models is driving a profound intelligent transformation. Steel production involves complex processes such as blast furnace ironmaking, converter steelmaking, continuous casting and rolling, and the complexity of its process optimization, energy consumption control, and quality management is extremely high.

[0003] Currently, existing methods all rely on empirical formulas and manual regulation, while large models can integrate massive production data (such as temperature, pressure, and composition parameters) to build a high-precision digital twin system and achieve dynamic simulation and prediction of the entire process. Therefore, how to utilize the advantages of large models to assist the steel smelting process and improve production efficiency and product quality is a technical problem that urgently needs to be solved currently.

[0004] Currently, with the development of the steel industry, a low-carbon production method has emerged. Adding the amount of scrap steel during the smelting process helps to reduce the dependence on iron ore and lower carbon emissions. However, with the increase in the amount of scrap steel during the smelting process, the content of unknown elements in the steel increases, prolonging the smelting time. At the same time, it is also not applicable to the existing steelmaking model, causing great trouble to the existing smelting process. Summary of the Invention

[0005] To solve the above technical problems, an embodiment of the present invention provides a converter high-efficiency smelting method based on a big data steelmaking model under the condition of a low hot metal ratio.

[0006] The technical solution of the embodiment of the present invention is realized as follows: An embodiment of the present invention provides a converter high-efficiency smelting method based on a big data steelmaking model under the condition of a low hot metal ratio. The method includes: Obtain historical data of converter smelting; the historical data includes material data, first feeding data, TSC measurement data, second feeding data, and TSO measurement data; wherein, the material data includes hot metal weight, hot metal temperature, hot metal Si content, hot metal Mn content, hot metal P content, hot metal S content, converter furnace age, heavy scrap weight, light scrap weight, pig iron weight, ladle scrap steel weight, and ladle scrap steel temperature; the first feeding data includes lime addition amount, light burned magnesia ball addition amount, ferrosilicon addition amount for heating, coal-based carburizer addition amount, and oxygen supply amount before TSC measurement; the TSC measurement data includes TSC measurement temperature, C content measured by TSC, and P content measured by TSC; the second feeding data includes the oxygen supply amount after TSC measurement; the TSO measurement data includes TSO measurement temperature, C content measured by TSO, O content measured by TSO, and P content measured by TSO; Obtain an initial model, and train the initial model using the historical data to obtain a first model; the first model includes a first sub-model, a second sub-model, and a third sub-model; the input of the first sub-model is material data; the output of the first sub-model is first feeding data; the input of the second sub-model is material data, first feeding data, and TSC measurement data; the output of the second sub-model is second feeding data; the input of the third sub-model is material data, first feeding data, TSC measurement data, and second feeding data; the output of the third sub-model is TSO measurement data; Obtain the material data during actual converter smelting; input the material data into the first sub-model to obtain the first feeding data output by the first sub-model; Feed materials according to the first feeding data output by the first sub-model, and perform TSC measurement after the smelting stage is completed to obtain TSC measurement data; Input the material data, the first feeding data, and the TSC measurement data into the second sub-model to obtain the second feeding data output by the second sub-model; Continue blowing after feeding materials according to the second feeding data output by the second sub-model, and input the material data, the first feeding data, the TSC measurement data, and the second feeding data into the third sub-model to obtain the TSO measurement data output by the third sub-model; Judge whether the TSO measurement data output by the third sub-model meets the target requirements; if it meets the target requirements, perform unequal-sample tapping.

[0007] In one embodiment, if the TSO measurement data output by the third sub-model does not meet the target requirements, then perform: Use the TSO measurement temperature in the TSO measurement data as the latest TSC measurement temperature, the C content measured by TSO as the latest C content measured by TSC, the P content measured by TSO as the latest P content measured by TSC, and use the current oxygen supply amount as the oxygen supply amount before the latest TSC measurement; Input the updated data into the second sub-model to obtain the second feeding data output again by the second sub-model; continue blowing after feeding materials according to the second feeding data output again by the second sub-model, and input the updated material data, the first feeding data, the TSC measurement data, and the second feeding data into the third sub-model again to obtain the TSO measurement data output by the third sub-model; Judge again whether the TSO measurement data output by the third sub-model meets the target requirements; if it meets the target requirements, perform unequal-sample tapping; if it does not meet the target requirements, perform the above operations again until the TSO measurement data output by the third sub-model meets the target requirements.

[0008] In one embodiment, the initial model is a gradient boosting classifier algorithm model.

[0009] In one embodiment, the range of parameters in the gradient boosting classifier algorithm model is as follows: the range of the number of weak classifiers is 800 - 1200, the range of the learning rate is 0.01 - 0.02, the range of the maximum depth of the decision tree is 4, the range of the minimum number of samples for decision tree splitting is 2, and the range of the minimum number of samples for decision tree leaf nodes is 1.

[0010] In one embodiment, the process of determining the range of parameters in the gradient boosting classifier algorithm model is as follows: Obtain the initial value range of each parameter of the gradient boosting classifier algorithm model; Use grid search or random search to traverse the initial value range of each parameter to form multiple parameter combinations; Perform cross-validation on each parameter combination to obtain the model performance evaluation results; Determine the optimal parameter combination according to the model performance evaluation results; use the optimal parameter combination as the final value range of the parameters of the gradient boosting classifier algorithm model.

[0011] In one embodiment, after smelting is completed, detect the slag composition and judge whether the data of this heat is updated to the converter smelting historical data: Determine whether the data of this heat meets the basic requirements, where the basic requirements are: 120 t ≤ molten iron weight ≤ 200 t, 1000 °C ≤ molten iron temperature ≤ 1550 °C, 0.1% ≤ Si content in molten iron ≤ 0.9%, 0.05% ≤ Mn content in molten iron ≤ 0.6%, 0.06% ≤ P content in molten iron ≤ 0.13%, S content in molten iron ≤ 0.1%, converter campaign life ≤ 10000 heats, heavy scrap weight ≤ 60 t, light scrap weight ≤ 10 t, pig iron weight ≤ 70 t, 5 t ≤ scrap weight in ladle ≤ 25 t, 0.5 t ≤ lime addition amount ≤ 13 t, light burned magnesia balls addition amount ≤ 3 t, ferrosilicon addition amount for heating ≤ 3 t, coal-based carburizer addition amount ≤ 3 t, 6000 m³ ≤ oxygen supply before TSC measurement ≤ 12000 m³, 1520 °C ≤ TSC measurement temperature ≤ 1700 °C, 0.005% ≤ C content measured by TSC ≤ 0.5%, 0.002% ≤ P content measured by TSC ≤ 0.1%, oxygen supply after TSC measurement ≤ 6000 m³, 1540 °C ≤ TSO measurement temperature ≤ 1700 °C, 0.02% ≤ C content measured by TSO ≤ 0.08%, 0.03% ≤ O content measured by TSO ≤ 0.13%, 0.002% ≤ P content measured by TSO ≤ 0.08%; If the data of this heat meets the second requirement under the condition of meeting the basic requirements, then determine that the data of this heat can be updated to the converter smelting historical data; Among them, the second requirement includes: tapping temperature of steel grade - 30 °C ≤ slag melting point ≤ tapping temperature of steel grade - 20 °C, FeTOT ≤ 25%, 1% ≤ P2O5 ≤ 3%, CaO / SiO2 > CaO / SiO 2(中间值) ; Among them, the process of determining whether the data of this heat meets the second requirement includes: According to the slag composition detection results, use Factsage software to calculate the CaO - SiO2 - FeO - Fe2O3 - MgO five - component phase diagram of the slag of this heat; among them, MgO is the detected value of the slag composition, Fe2O3 is 0.25FeTOT, FeO is 0.75FeTOT, FeTOT is the detected value of the slag composition, find the area in the five - component phase diagram where the slag melting point is between tapping temperature of steel grade - 30 °C and tapping temperature of steel grade - 20 °C, calculate the CaO / SiO2 value range and the intermediate value of the area, and record the intermediate value as CaO / SiO 2(中间值) ; and calculate the slag melting point of the slag of this heat using the five - component phase diagram.

[0012] In an embodiment, obtaining an initial model and training the initial model using the historical data to obtain a first model includes: Divide the historical data into multiple categories according to the grades of steel types in the historical data; use the historical data of each category to train the initial model respectively to obtain the first model corresponding to the historical data of each category.

[0013] The solution of this embodiment has the following beneficial effects: The solution of this embodiment uses the advantages of the large model to assist the steel smelting process, which can improve production efficiency and product quality. Description of the Drawings

[0014] Figure 1 It is a schematic flow chart of the converter high-efficiency smelting method based on the big data steelmaking model under the condition of low hot metal ratio in the embodiment of the present invention; Figure 2 It is a schematic diagram of multiple linear programming in the embodiment of the present invention; Figure 3 It is a schematic diagram of the CaO-SiO2-FeO-MgO-Fe2O3 five-component phase diagram calculated by Factsage in the embodiment of the present invention; Figure 4 It is a schematic diagram of the comparison between the predicted value and the actual value of metallurgical lime in the embodiment of the present invention; Figure 5 It is a schematic diagram of the comparison between the predicted value and the actual value of light-burned magnesium balls in the embodiment of the present invention; Figure 6 It is a schematic diagram of the comparison between the predicted value and the actual value of coal-based carburizer in the embodiment of the present invention; Figure 7 It is a schematic diagram of the comparison between the predicted value and the actual value of the TSC measured temperature in the embodiment of the present invention; Figure 8 It is a schematic diagram of the comparison between the predicted value and the actual value of the C content measured by TSC in the embodiment of the present invention; Figure 9 It is a schematic diagram of the comparison between the predicted value and the actual value of the P content measured by TSC in the embodiment of the present invention; Figure 10 It is a schematic diagram of the comparison between the predicted value and the actual value of the oxygen supply amount after TSC measurement in the embodiment of the present invention; Figure 11 It is a schematic diagram of the comparison between the predicted value and the actual value of the TSO measured temperature in the embodiment of the present invention; Figure 12 It is a schematic diagram of the comparison between the predicted value and the actual value of the C content measured by TSO in the embodiment of the present invention; Figure 13 It is a schematic diagram of the comparison between the predicted value and the actual value of the O content measured by TSO in the embodiment of the present invention; Figure 14 It is a schematic diagram of the comparison between the predicted value and the actual value of the P content measured by TSO in the embodiment of the present invention. Detailed Embodiments

[0015] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments.

[0016] An embodiment of the present invention provides a converter high-efficiency smelting method based on a big data steelmaking model under low hot metal ratio conditions, as Figure 1 shown, the method includes: Step 101: Obtain converter smelting historical data; the historical data includes material data, first feeding data, TSC measurement data, second feeding data, and TSO measurement data; wherein, the material data includes hot metal weight, hot metal temperature, hot metal Si content, hot metal Mn content, hot metal P content, hot metal S content, converter furnace age, heavy scrap weight, light scrap weight, pig iron weight, ladle scrap weight, and ladle scrap temperature; the first feeding data includes lime addition amount, light burned magnesia ball addition amount, ferrosilicon addition amount for heating, coalification carburizer addition amount, and oxygen supply amount before TSC measurement; the TSC measurement data includes TSC measurement temperature, C content measured by TSC, and P content measured by TSC; the second feeding data includes oxygen supply amount after TSC measurement; the TSO measurement data includes TSO measurement temperature, C content measured by TSO, O content measured by TSO, and P content measured by TSO; Step 102: Obtain an initial model, and use the historical data to train the initial model to obtain a first model; the first model includes a first sub-model, a second sub-model, and a third sub-model; the input of the first sub-model is material data; the output of the first sub-model is the first feeding data; the input of the second sub-model is material data, first feeding data, and TSC measurement data; the output of the second sub-model is the second feeding data; the input of the third sub-model is material data, first feeding data, TSC measurement data, and second feeding data; the output of the third sub-model is TSO measurement data; Step 103: Obtain the material data during actual converter smelting; input the material data into the first sub-model to obtain the first feeding data output by the first sub-model; Step 104: Feed materials according to the first feeding data output by the first sub-model, and perform TSC measurement after the smelting stage is completed to obtain TSC measurement data; Step 105: Input the material data, the first feeding data, and the TSC measurement data into the second sub-model to obtain the second feeding data output by the second sub-model; Step 106: Continue blowing after feeding materials according to the second feeding data output by the second sub-model, and input the material data, the first feeding data, the TSC measurement data, and the second feeding data into the third sub-model to obtain the TSO measurement data output by the third sub-model; Step 107: Determine whether the TSO measurement data output by the third sub-model meets the target requirements; if it meets the target requirements, perform unequal-sample tapping.

[0017] In practical applications, if the TSO measurement data output by the third sub-model does not meet the target requirements, then perform the following: Use the TSO measurement temperature in the TSO measurement data as the latest TSC measurement temperature, the C content measured by TSO as the latest C content measured by TSC, the P content measured by TSO as the latest P content measured by TSC, and use the current oxygen supply amount as the oxygen supply amount before the latest TSC measurement; Input the updated data into the second sub-model to obtain the second feeding data output by the second sub-model again; perform feeding according to the second feeding data output by the second sub-model again and continue blowing, and input the updated material data, the first feeding data, the TSC measurement data, and the second feeding data into the third sub-model again to obtain the TSO measurement data output by the third sub-model; Determine again whether the TSO measurement data output by the third sub-model meets the target requirements; if it meets the target requirements, perform unequal-sample tapping; if it does not meet the target requirements, perform the above operations again until the TSO measurement data output by the third sub-model meets the target requirements.

[0018] TSC measurement and TSO measurement are the contents detected by the sublance. The sublance detection is an operation process for detecting whether the molten steel in the converter steelmaking process meets the requirements. The sublance mainly measures the carbon content of the molten steel in the molten bath during the blowing process when performing TSC measurement in the converter steelmaking process; the sublance mainly measures the carbon content and oxygen content of the molten bath at the end of blowing when performing TSO measurement in the converter steelmaking process.

[0019] The solution of this embodiment uses the advantages of the large model to assist the steel smelting process, which can improve production efficiency and product quality.

[0020] Next, a specific embodiment will be used to illustrate this solution.

[0021] Specifically, this solution includes the following steps: S1: Collect valid historical data and establish a database. The data includes hot metal weight, hot metal temperature, hot metal Si content, hot metal Mn content, hot metal P content, hot metal S content, converter campaign life, heavy scrap weight, light scrap weight, pig iron weight, scrap weight in ladle, scrap temperature in ladle, lime addition, light burned magnesia addition, ferrosilicon addition for heating, coal-based carburizer addition, oxygen supply before TSC (i.e., oxygen supply before TSC measurement), TSC temperature (i.e., TSC measurement temperature), TSC C (i.e., C content measured by TSC), TSC P (i.e., P content measured by TSC), oxygen supply after TSC (i.e., oxygen supply after TSC measurement), TSO temperature (i.e., TSO measurement temperature), TSO C (i.e., C content measured by TSO), TSO O (i.e., O content measured by TSO), TSO P (i.e., P content measured by TSO).

[0022] S2: Train the historical data in the database. Use parameters such as hot metal weight, hot metal temperature, hot metal Si content, hot metal Mn content, hot metal P content, hot metal S content, converter campaign life, heavy scrap weight, light scrap weight, pig iron weight, scrap weight in ladle, scrap temperature in ladle for multiple linear programming of lime addition, light burned magnesia addition, ferrosilicon addition for heating, coal-based carburizer addition, oxygen supply before TSC. Then use parameters such as hot metal weight, hot metal temperature, hot metal Si content, hot metal Mn content, hot metal P content, hot metal S content, converter campaign life, heavy scrap weight, light scrap weight, pig iron weight, scrap weight in ladle, scrap temperature in ladle, lime addition, light burned magnesia addition, ferrosilicon addition for heating, coal-based carburizer addition, oxygen supply before TSC for multiple linear programming of TSC temperature, TSC C, TSC P. Next, use parameters such as hot metal weight, hot metal temperature, hot metal Si content, hot metal Mn content, hot metal P content, hot metal S content, converter campaign life, heavy scrap weight, light scrap weight, pig iron weight, scrap weight in ladle, scrap temperature in ladle, lime addition, light burned magnesia addition, ferrosilicon addition for heating, coal-based carburizer addition, oxygen supply before TSC, TSC temperature, TSC C, TSC P for multiple linear programming of oxygen supply after TSC. Finally, use parameters such as hot metal weight, hot metal temperature, hot metal Si content, hot metal Mn content, hot metal P content, hot metal S content, converter campaign life, heavy scrap weight, light scrap weight, pig iron weight, scrap weight in ladle, scrap temperature in ladle, lime addition, light burned magnesia addition, ferrosilicon addition for heating, coal-based carburizer addition, oxygen supply before TSC, TSC temperature, TSC C, TSC P, oxygen supply after TSC for multiple linear programming of TSO temperature, TSO C, TSO O, TSO P. After the data training is completed, a model is obtained.

[0023] S3: Production application. Refer to Figure 2, in the first step, input the real-time molten iron weight, molten iron temperature, Si content of molten iron, Mn content of molten iron, P content of molten iron, S content of molten iron, converter campaign life, heavy scrap weight, light scrap weight, pig iron weight, scrap weight in ladle, and scrap temperature in ladle into the model, and use multiple linear programming to calculate the lime addition amount, light burned magnesia addition amount, ferrosilicon addition amount, coal-based carburizer addition amount, and oxygen supply amount before TSC. In the second step, add materials according to the results of multiple linear programming. After this smelting stage, measure the temperature and composition (TSC measurement) to obtain the TSC temperature, TSC C, and TSC P. In the third step, input the parameters of molten iron weight, molten iron temperature, Si in molten iron, Mn in molten iron, P in molten iron, S in molten iron, converter campaign life, heavy scrap weight, light scrap weight, pig iron weight, scrap weight in ladle, scrap temperature in ladle, lime addition amount, light burned magnesia addition amount, ferrosilicon addition amount, coal-based carburizer addition amount, oxygen supply amount before TSC, TSC temperature, TSC C, and TSC P into the model, and use multiple linear programming to calculate the oxygen supply amount after TSC. In the fourth step, supply oxygen according to the calculation results and continue blowing. In the fifth step, input the parameters of molten iron weight, molten iron temperature, Si content of molten iron, Mn content of molten iron, P content of molten iron, S content of molten iron, converter campaign life, heavy scrap weight, light scrap weight, pig iron weight, scrap weight in ladle, scrap temperature in ladle, lime, light burned magnesia, ferrosilicon, coal-based carburizer, oxygen supply amount before TSC, TSC temperature, TSC C, TSC P, and oxygen supply amount after TSC into multiple linear programming to calculate the TSO temperature, TSO C, TSO O, and TSO P. Based on the calculation results, tap the steel when the temperature and composition are qualified. If not, repeat the third step, where the oxygen supply amount before TSC is the cumulative value of the oxygen supply amount. After this stage of smelting, measure the temperature, carbon, and oxygen (TSO measurement). To save smelting time, adopt the method of tapping the steel without waiting for samples and directly tap the steel.

[0024] S4: If there is no abnormality in smelting and the data meets the preset requirements, import the smelting data of this furnace into the database by category as historical data and conduct training regularly.

[0025] The converter smelting data this time includes material data, first feeding data, TSC measurement data, second feeding data, TSO measurement data, and slag detection data; the acquisition method of the slag detection data is as follows: After smelting is completed, detect the slag composition, and use Factsage software to calculate the CaO-SiO2-FeO five-component phase diagram (MgO is the detected value of the slag, take Fe2O3 = 0.25FeTOT) (here, the CaO-SiO2-FeO five-component phase diagram can be seen Figure 3 as shown), take FeO = 0.75FeTOT, find the region in the phase diagram where the slag melting point is in the range of steel tapping temperature - 30°C to steel tapping temperature - 20°C, calculate the CaO / SiO2 value range in this region and calculate the intermediate value, denoted as CaO / SiO 2(中间值)Calculate the melting point of the current heat of slag using the phase diagram.

[0026] The preset requirements include basic requirements and desired conversion requirements.

[0027] In this embodiment, the converter specification is 210 t, and the basic requirements are: 120 t ≤ hot metal weight ≤ 200 t, 1000 °C ≤ hot metal temperature ≤ 1550 °C, 0.1% ≤ Si in hot metal ≤ 0.9%, 0.05% ≤ Mn in hot metal ≤ 0.6%, 0.06% ≤ P in hot metal ≤ 0.13%, S in hot metal ≤ 0.1%, converter campaign life ≤ 10000 heats, heavy scrap weight ≤ 60 t, light scrap weight ≤ 10 t, pig iron weight ≤ 70 t, 5 t ≤ scrap in ladle weight ≤ 25 t, 0.5 t ≤ lime addition amount ≤ 13 t, light burned magnesia balls addition amount ≤ 3 t, ferrosilicon addition amount for heating ≤ 3 t, coal-based carburizer addition amount ≤ 3 t, 6000 m³ ≤ oxygen supply before TSC ≤ 12000 m³, 1520 °C ≤ TSC temperature ≤ 1700 °C, 0.005% ≤ TSC C ≤ 0.5%, 0.002% ≤ TSC P ≤ 0.1%, oxygen supply after TSC ≤ 6000 m³, 1540 °C ≤ TSO temperature ≤ 1700 °C, 0.02% ≤ TSO C ≤ 0.08%, 0.03% ≤ TSO O ≤ 0.13%, 0.002% ≤ TSO P ≤ 0.08%.

[0028] The desired conversion requirements are: tapping temperature of steel grade - 30 °C ≤ slag melting point ≤ tapping temperature of steel grade - 20 °C, FeTOT ≤ 25%, 1% ≤ P2O5 ≤ 3%, CaO / SiO2 > CaO / SiO 2(中间值) .

[0029] In addition, the function used in multiple linear programming is the gradient boosting classifier algorithm model (GBM). The value range of each parameter in the gradient boosting classifier algorithm model can be optimized by the following steps; First, determine the value range of each parameter according to experience and data characteristics; then, traverse the parameter space using grid search (GridSearchCV) or random search (RandomizedSearchCV); after that, perform cross-validation on each parameter combination to evaluate the model performance; finally, select the optimal parameter combination according to the cross-validation results.

[0030] Considering that the model data set is small, to prevent overfitting, and at the same time through the cross-validation of 2000 groups of historical smelting data, the verification results show that: Set random_state = 46 to control the randomness parameter and ensure the reproducibility of the experiment; to maintain the model complexity and reduce the risk of overfitting, the minimum number of samples for decision tree splitting (min_samples_split) is set to 2; similar to min_samples_split, to maintain the model complexity and reduce the risk of overfitting, the minimum number of samples in the decision tree leaf nodes (min_samples_leaf) is set to 1; the sample ratio (subsample) is set to 1, and the number of weak classifiers (n_estimators) can take values from 800 to 1200; the learning rate (learning_rate) can take values from 0.01 to 0.02; the maximum depth of the decision tree (max_depth) can take values up to 4.

[0031] In addition, refer to Figures 4 - 14 , for the result comparison between the model prediction value and the actual value by adopting the above parameter range. According to the comparison result, after adopting the above parameter range, the hit rate of the model prediction value can reach more than 90%, and it can better realize the prediction of the converter smelting process under the condition of low hot metal ratio.

[0032] Furthermore, in this embodiment, the steel grades are classified by brand, and each classification data is stored and linearly programmed separately.

[0033] This embodiment utilizes the advantages of the large model to assist the steel smelting process, which can improve production efficiency and product quality and achieve unequal-sample tapping.

[0034] It should also be noted that the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such process, method, commodity or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of another identical element in the process, method, commodity or device including the element.

[0035] The above are only embodiments of the present application and are not used to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A converter high-efficiency smelting method based on a big data steelmaking model under the condition of a low hot metal ratio, characterized in that, The method includes: Obtaining the historical data of converter smelting; the historical data includes material data, first feeding data, TSC measurement data, second feeding data, and TSO measurement data; wherein, the material data includes hot metal weight, hot metal temperature, hot metal Si content, hot metal Mn content, hot metal P content, hot metal S content, converter campaign life, heavy scrap weight, light scrap weight, pig iron weight, ladle scrap weight, and ladle scrap temperature; the first feeding data includes lime addition amount, light burned magnesia ball addition amount, ferrosilicon addition amount for heating, coal-based carburizer addition amount, and oxygen supply amount before TSC measurement; the TSC measurement data includes TSC measurement temperature, C content measured by TSC, and P content measured by TSC; the second feeding data includes oxygen supply amount after TSC measurement; the TSO measurement data includes TSO measurement temperature, C content measured by TSO, O content measured by TSO, and P content measured by TSO; Obtaining an initial model, training the initial model using the historical data to obtain a first model; the first model includes a first sub-model, a second sub-model, and a third sub-model; the input of the first sub-model is the material data; the output of the first sub-model is the first feeding data; the input of the second sub-model is the material data, the first feeding data, and the TSC measurement data; the output of the second sub-model is the second feeding data; the input of the third sub-model is the material data, the first feeding data, the TSC measurement data, and the second feeding data; the output of the third sub-model is the TSO measurement data; Obtaining the material data during actual converter smelting; inputting the material data into the first sub-model to obtain the first feeding data output by the first sub-model; Feeding according to the first feeding data output by the first sub-model, and performing TSC measurement after the smelting stage to obtain the TSC measurement data; Inputting the material data, the first feeding data, and the TSC measurement data into the second sub-model to obtain the second feeding data output by the second sub-model; Continuing blowing after feeding according to the second feeding data output by the second sub-model, and inputting the material data, the first feeding data, the TSC measurement data, and the second feeding data into the third sub-model to obtain the TSO measurement data output by the third sub-model; Judging whether the TSO measurement data output by the third sub-model meets the target requirements; if it meets the target requirements, perform unequal-sample tapping.

2. The converter high-efficiency smelting method based on the big data steelmaking model under the condition of low hot metal ratio according to claim 1, characterized in that, If the TSO measurement data output by the third sub-model does not meet the target requirements, perform: Taking the TSO measurement temperature in the TSO measurement data as the latest TSC measurement temperature, the C content measured by TSO as the latest C content measured by TSC, the P content measured by TSO as the latest P content measured by TSC, and taking the current oxygen supply amount as the latest oxygen supply amount before TSC measurement; Input the updated data into the second sub-model to obtain the second feeding data output again by the second sub-model; perform feeding according to the second feeding data output again by the second sub-model and then continue blowing, and input the updated material data, the first feeding data, the TSC measurement data, and the second feeding data into the third sub-model again to obtain the TSO measurement data output by the third sub-model; Judge again whether the TSO measurement data output by the third sub-model meets the target requirements; if it meets the target requirements, perform unequal-sample tapping; if it does not meet the target requirements, perform the above operations again until the TSO measurement data output by the third sub-model meets the target requirements.

3. The converter high-efficiency smelting method based on the big data steelmaking model under the condition of low hot metal ratio according to claim 1, characterized in that The initial model is a gradient boosting classifier algorithm model.

4. The converter high-efficiency smelting method based on the big data steelmaking model under the condition of low iron-water ratio according to claim 3, characterized in that, The ranges of the parameters in the gradient boosting classifier algorithm model are as follows: the range of the number of weak classifiers is 800 - 1200, the range of the learning rate is 0.01 - 0.02, the range of the maximum depth of the decision tree is 4, the range of the minimum number of samples for decision tree splitting is 2, and the range of the minimum number of samples for decision tree leaf nodes is 1.

5. The converter high-efficiency smelting method based on the big data steelmaking model under the condition of low molten iron ratio according to claim 4, characterized in that, The process of determining the range of the parameters in the gradient boosting classifier algorithm model is as follows: Obtain the initial value ranges of the parameters of the gradient boosting classifier algorithm model; Use grid search or random search to traverse the initial value ranges of the parameters to form multiple parameter combinations; Perform cross-validation on each parameter combination to obtain the model performance evaluation results; Determine the optimal parameter combination according to the model performance evaluation results; take the optimal parameter combination as the final value range of the parameters of the gradient boosting classifier algorithm model.

6. The converter high-efficiency smelting method based on the big data steelmaking model under the condition of low hot metal ratio according to claim 1, characterized in that, After the smelting is completed, detect the slag composition and judge whether the data of this heat is updated to the converter smelting historical data: Judge whether the data of this heat meets the basic requirements, and the basic requirements are: 120 t ≤ hot metal weight ≤ 200 t, 1000 °C ≤ hot metal temperature ≤ 1550 °C, 0.1% ≤ hot metal Si content ≤ 0.9%, 0.05% ≤ hot metal Mn content ≤ 0.6%, 0.06% ≤ hot metal P content ≤ 0.13%, hot metal S content ≤ 0.1%, converter campaign life ≤ 10000 heats, heavy scrap weight ≤ 60 t, light scrap weight ≤ 10 t, pig iron weight ≤ 70 t, 5 t ≤ scrap weight in ladle ≤ 25 t, 0.5 t ≤ lime addition amount ≤ 13 t, light burned magnesia ball addition amount ≤ 3 t, ferrosilicon addition amount for heating ≤ 3 t, coal-based carburizer addition amount ≤ 3 t, 6000 m³ ≤ oxygen supply before TSC measurement ≤ 12000 m³, 1520 °C ≤ TSC measurement temperature ≤ 1700 °C, 0.005% ≤ C content measured by TSC ≤ 0.5%, 0.002% ≤ P content measured by TSC ≤ 0.1%, oxygen supply after TSC measurement ≤ 6000 m³, 1540 °C ≤ TSO measurement temperature ≤ 1700 °C, 0.02% ≤ C content measured by TSO ≤ 0.08%, 0.03% ≤ O content measured by TSO ≤ 0.13%, 0.002% ≤ P content measured by TSO ≤ 0.08%; When the basic requirements are met, if the data of this heat meets the expected conversion requirements, it is determined that the data of this heat can be updated into the converter smelting historical data; Among them, the expected conversion requirements include: the melting point of the slag satisfies -30°C ≤ melting point of the slag ≤ -20°C of the tapping temperature of the steel grade, FeTOT ≤ 25%, 1% ≤ P2O5 ≤ 3%, and CaO / SiO2 > CaO / SiO 2(中间值) ; Among them, the process of judging whether the data of this heat meets the expected conversion requirements includes: According to the detection results of the slag composition, the CaO-SiO2-FeO-Fe2O3-MgO five-component phase diagram of the slag for this heat is calculated using Factsage software; where MgO is the detected value of the slag composition, Fe2O3 is 0.25FeTOT, FeO is 0.75FeTOT, FeTOT is the detected value of the slag composition, find the region in the five-component phase diagram where the melting point of the slag is between the tapping temperature of the steel grade - 30°C and the tapping temperature of the steel grade - 20°C, calculate the value range and the intermediate value of CaO / SiO2 in the said region, and record the intermediate value as CaO / SiO 2(中间值) ; and calculate the melting point of the slag for this heat using the said five-component phase diagram.

7. The converter high-efficiency smelting method based on the big data steelmaking model under the condition of low molten iron ratio according to claim 1, characterized in that, Obtain an initial model, and use the historical data to train the initial model to obtain a first model, including: According to the grades of steel types in the historical data, divide the historical data into multiple categories; use the historical data of each category to train the initial model respectively to obtain the first model corresponding to the historical data of each category.

Images