Converter oxygen consumption optimization method and system based on random forest increment prediction model

Abstract

The invention discloses a converter oxygen consumption optimization method and system based on a random forest increment prediction model. The converter oxygen consumption optimization system comprises a machine learning module, an increment model module and a selection calculation module. The machine learning module builds a machine learning model, the incremental model module builds an incremental model, and the selection calculation module selects reference furnaces meeting requirements, calculates oxygen consumption values of the reference furnaces according to the machine learning model and calculates oxygen consumption of the current furnace according to the incremental model. According to the method, complex changes in the smelting process can be captured more accurately, so that the accuracy of oxygen consumption prediction is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of converter smelting, and in particular to a converter oxygen consumption optimization method and system based on a random forest incremental prediction model. Background Art

[0002] In the field of steel smelting, converter smelting is an important steelmaking process, and the oxygen consumption in the process is a key indicator for evaluating production efficiency and cost control. At present, the calculation of the total oxygen consumption in converter smelting mainly adopts an empirical formula based on the composition of molten steel and slag. This formula estimates the oxygen consumption by comprehensively considering the content of elements such as carbon, silicon, phosphorus, sulfur and manganese in molten steel, as well as the content of silicon dioxide in slag.

[0003] Although this method has been widely used in actual production, there are still some technical shortcomings that cannot be ignored. First, due to the complex and changeable composition of molten steel and slag, the accuracy of the empirical formula is greatly affected, and it is difficult to accurately predict the oxygen consumption of different batches of molten steel. Secondly, this method ignores the influence of other factors in the smelting process such as molten iron temperature, scrap steel composition, molten iron weight, auxiliary material dosage, etc. on oxygen consumption, so there are certain limitations in practical application. In addition, with the advancement of smelting technology and the optimization of production processes, the existing oxygen consumption calculation method has been difficult to meet the increasingly stringent energy conservation, emission reduction and cost control requirements, especially in the current global advocacy of green and low-carbon development. How to more accurately calculate and control the oxygen consumption in the converter smelting process has become an important issue facing steel companies. Summary of the invention

[0004] Purpose of the invention: The present invention provides a converter oxygen consumption optimization method and system based on a random forest incremental prediction model, which can more accurately capture the complex changes in the smelting process, thereby ensuring the accuracy of oxygen consumption prediction.

[0005] Technical solution: The present invention discloses a method for optimizing converter oxygen consumption based on a random forest incremental prediction model, comprising the following steps:

[0006] Step 1: Build a machine learning model;

[0007] Step 2: Establish an incremental model;

[0008] Step 3: Select reference furnaces that meet the requirements and calculate the oxygen consumption values ​​of these reference furnaces according to the machine learning model;

[0009] Step 4: Calculate the oxygen consumption of the current furnace according to the incremental model.

[0010] Furthermore, in step 1, establishing a machine learning model specifically includes the following steps:

[0011] Step 11: Get the original data, eliminate anomalies and noise, select useful features for data division, and use them for training set, validation set and test set;

[0012] Step 12: Use random forest regression algorithm;

[0013] Step 13: Model training and evaluation;

[0014] Step 14: Package the model;

[0015] Step 15: Calculate the oxygen consumption value Y through the machine learning model 机器 .

[0016] Furthermore, in step 13, the random model is trained and evaluated. The evaluation is calculated based on the training results. The smaller the sum of squares of the difference between the predicted value and the actual value, the better. Common indicators of random forest regression include mean square error (MSE) and R-squared (R 2 ), the smaller the MSE value, the better the model fits the data. 2 The closer the value is to 1, the better the model fits the data, and vice versa;

[0017]

[0018] In the formula, n represents the number of samples, y i represents the true value of the i-th sample, Y i Represents the predicted value of the i-th sample;

[0019]

[0020] Where: Represents the average of all true sample values.

[0021] In the actual smelting process, oxygen consumption is related to not only the composition of molten iron, but also the amount of molten iron, the amount of scrap steel, the type of scrap steel, the auxiliary materials added during the smelting process, the life of the oxygen lance nozzle, the oxygen flow rate, etc. In the present invention, in addition to considering the molten iron components C, Si, Mn, P, and S, machine learning also adds the weight of molten iron, the weight of scrap steel, the amount of smelting lime, the amount of smelting light burning, the amount of smelting ore, the life of the nozzle, and the average oxygen flow rate.

[0022] Furthermore, in step 2, establishing the incremental model specifically includes the following steps:

[0023] Step 21, reference furnace selection and calculation: select the N furnaces closest to the current furnace that meet the requirements as the reference furnace, and then calculate the target deviation value ΔY of the N furnaces reference furnace. The calculation formula is as follows:

[0024]

[0025] where Y 参考炉实际i is the actual target value of the reference furnace, Y 参考炉理论i It is the theoretical calculated value of the reference furnace target.

[0026] Step 22: Build an incremental model: Set the current furnace target theoretical calculation value to Y 理论 , the current furnace target Y value is calculated as follows:

[0027] Y=Y 理论 +ΔY.

[0028] In the actual smelting process, in addition to the above-mentioned parameters that can be obtained, oxygen consumption is also related to equipment, such as the status of the oxygen lance nozzle, the effect of re-blowing, and the condition of the converter. These equipment status changes will continuously affect the data of the recent furnaces until the equipment returns to normal or the equipment is replaced. In order to accurately calculate the oxygen consumption, it is necessary to consider the short-term impact caused by equipment abnormalities. When equipment abnormalities occur, the actual oxygen consumption is often more or less than normal. The difference between the oxygen consumption of recent furnaces and the theoretical value is calculated. This difference can be considered to be caused by equipment abnormalities. The actual calculation process will correct this difference to the calculation results, which can reduce the impact of equipment abnormalities on the results.

[0029] Further, in step 3, a reference furnace that meets the requirements is selected, the endpoint carbon is within the target range, the endpoint oxygen is within the range of 0.02-0.06%, and there is heat information for molten iron and scrap steel.

[0030] Furthermore, in step 4, the oxygen consumption of the current furnace is calculated according to the incremental model, and the calculation result of the machine learning model is substituted into the incremental model calculation formula to calculate the oxygen consumption γ of the current furnace:

[0031]

[0032] Where γ is the oxygen consumption of the current furnace, γ 机器学习 Machine learning calculates oxygen consumption for the current furnace, γ 参考炉实际 is the actual oxygen consumption of the reference furnace, γ 参考炉机器学习 Machine learning calculation of oxygen consumption for a reference furnace.

[0033] Correspondingly, a converter oxygen consumption optimization system based on a random forest incremental prediction model includes: a machine learning module, an incremental model module and a selection calculation module; the machine learning module establishes a machine learning model, the incremental model module establishes an incremental model, the selection calculation module selects reference furnaces that meet the requirements, calculates the oxygen consumption values ​​of these reference furnaces according to the machine learning model, and calculates the oxygen consumption of the current furnace according to the incremental model.

[0034] Further, a reference furnace that meets the requirements is selected, the endpoint carbon is within the target range, the endpoint oxygen is within the range of 0.02-0.06%, and the furnace has molten iron information and scrap steel information.

[0035] Furthermore, the oxygen consumption γ of the current furnace is calculated:

[0036]

[0037] Where γ is the oxygen consumption of the current furnace, γ 机器学习 Machine learning calculates oxygen consumption for the current furnace, γ 参考炉实际 is the actual oxygen consumption of the reference furnace, γ 参考炉机器学习 Machine learning calculation of oxygen consumption for a reference furnace.

[0038] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: (1) The present invention introduces an advanced machine learning algorithm in the field of converter oxygen consumption calculation. By comprehensively considering multi-dimensional parameters such as steel type, scrap steel, molten iron information and auxiliary material information, the accuracy of calculation is significantly improved. The parameters considered are more comprehensive and detailed, and can more accurately capture the complex changes in the smelting process, thereby ensuring the accuracy of oxygen consumption prediction; (2) The present invention innovatively adopts an incremental model architecture, which effectively reduces the fluctuation effect caused by factors such as equipment, measurement and auxiliary material scrap steel quality in the current furnace of smelting in the recent period of time. This design not only enhances the robustness of the model, but also further improves the accuracy and stability of oxygen consumption calculation; (3) The hit rate of converter oxygen consumption calculation error within the range of 350Nm3 is significantly improved by 10%. This significant improvement not only reflects the technical advantages of the present invention, but also provides steel companies with a more accurate and reliable oxygen consumption prediction tool, which helps companies achieve more efficient and energy-saving production operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a schematic diagram of the method flow of the present invention.

[0040] Figure 2 A schematic diagram of the process of establishing a machine learning model for the present invention.

[0041] Figure 3 A schematic diagram of the incremental model flow chart for the present invention. DETAILED DESCRIPTION

[0042] like Figure 1 As shown, a converter oxygen consumption optimization method based on a random forest incremental prediction model comprises the following steps:

[0043] Step 1: Build a machine learning model;

[0044] Step 2: Establish an incremental model;

[0045] Step 3: Select reference furnaces that meet the requirements and calculate the oxygen consumption values ​​of these reference furnaces according to the machine learning model;

[0046] Step 4: Calculate the oxygen consumption of the current furnace according to the incremental model.

[0047] like Figure 2 As shown, in step 1, establishing a machine learning model specifically includes the following steps:

[0048] Step 11: Get the original data, eliminate anomalies and noise, select useful features for data division, and use them for training set, validation set and test set;

[0049] Step 12: Use random forest regression algorithm;

[0050] Step 13: Model training and evaluation;

[0051] Step 14: Package the model;

[0052] Step 15: Calculate the oxygen consumption value Y through the machine learning model 机器 .

[0053] In step 13, the random model is trained and evaluated. The evaluation is calculated based on the training results. The smaller the sum of squares of the difference between the predicted value and the actual value, the better. Common indicators of random forest regression include mean square error (MSE) and R-squared (R 2 ), the smaller the MSE value, the better the model fits the data. 2 The closer the value is to 1, the better the model fits the data, and vice versa;

[0054]

[0055] In the formula, n represents the number of samples, y i represents the true value of the i-th sample, Y i Represents the predicted value of the i-th sample;

[0056]

[0057] Where: Represents the average of all true sample values.

[0058] In the actual smelting process, oxygen consumption is related to not only the composition of molten iron, but also the amount of molten iron, the amount of scrap steel, the type of scrap steel, the auxiliary materials added during the smelting process, the life of the oxygen lance nozzle, the oxygen flow rate, etc. In the present invention, in addition to considering the molten iron components C, Si, Mn, P, and S, machine learning also adds the weight of molten iron, the weight of scrap steel, the amount of smelting lime, the amount of smelting light burning, the amount of smelting ore, the life of the nozzle, and the average oxygen flow rate.

[0059] like Figure 3 As shown, in step 2, establishing the incremental model specifically includes the following steps:

[0060] Step 21, reference furnace selection and calculation: select the N furnaces closest to the current furnace that meet the requirements as the reference furnace, and then calculate the target deviation value ΔY of the N furnaces reference furnace. The calculation formula of ΔY is as follows:

[0061]

[0062] where Y 参考炉实际i is the actual target value of the reference furnace, Y 参考炉理论i It is the theoretical calculated value of the reference furnace target.

[0063] Step 22: Build an incremental model: Set the current furnace target theoretical calculation value to Y 理论 , the current furnace target Y value is calculated as follows:

[0064] Y=Y 理论 +ΔY.

[0065] In the actual smelting process, in addition to the above-mentioned parameters that can be obtained, oxygen consumption is also related to equipment, such as the status of the oxygen lance nozzle, the effect of re-blowing, and the condition of the converter. These equipment status changes will continuously affect the data of the recent furnaces until the equipment returns to normal or the equipment is replaced. In order to accurately calculate the oxygen consumption, it is necessary to consider the short-term impact caused by equipment abnormalities. When equipment abnormalities occur, the actual oxygen consumption is often more or less than normal. The difference between the oxygen consumption of recent furnaces and the theoretical value is calculated. This difference can be considered to be caused by equipment abnormalities. The actual calculation process will correct this difference to the calculation results, which can reduce the impact of equipment abnormalities on the results.

[0066] In step 3, a reference furnace that meets the requirements is selected, the endpoint carbon is within the target range, the endpoint oxygen is within the range of 0.02-0.06%, and the furnace has molten iron information and scrap steel information.

[0067] In step 4, the oxygen consumption of the current furnace is calculated according to the incremental model, and the calculation result of the machine learning model is substituted into the incremental model calculation formula to calculate the oxygen consumption γ of the current furnace:

[0068]

[0069] Where γ is the oxygen consumption of the current furnace, γ 机器学习 Machine learning calculates oxygen consumption for the current furnace, γ 参考炉实际 is the actual oxygen consumption of the reference furnace, γ 参考炉机器学习 Machine learning calculation of oxygen consumption for a reference furnace.

[0070] Correspondingly, a converter oxygen consumption optimization system based on a random forest incremental prediction model includes: a machine learning module, an incremental model module and a selection calculation module; the machine learning module establishes a machine learning model, the incremental model module establishes an incremental model, the selection calculation module selects reference furnaces that meet the requirements, calculates the oxygen consumption values ​​of these reference furnaces according to the machine learning model, and calculates the oxygen consumption of the current furnace according to the incremental model.

[0071] Embodiment 1:

[0072] The five most recent reference furnaces were selected for calculation. The oxygen consumption values ​​of the reference furnace and the current furnace and the oxygen consumption values ​​calculated by machine learning are shown in Table 1.

[0073] Table 1 Comparison of oxygen consumption values ​​of reference furnace and current furnace and oxygen consumption values ​​calculated by machine learning

[0074] Heat Machine learning to calculate oxygen consumption / Nm3 Actual oxygen consumption / Nm3 Current furnace 7404.2 γ Reference furnace 1 7490.9 7472 Reference furnace 2 7576.5 7358 Reference furnace 3 7427.4 7163 Reference furnace 4 7398.1 7572 Reference furnace 5 7451.2 7486

[0075]

[0076] 5+7163-7427.4+7572-7398.1+7486-7451.2)=7345.6Nm3. The actual oxygen consumption of the current furnace is 7345Nm3, and the error between the calculated value and the actual value of the oxygen consumption of the current furnace is 0.6Nm3, which is within the range of 350Nm3.

[0077] Embodiment 2:

[0078] The most recent 10 reference furnaces are selected for calculation. The oxygen consumption values ​​of the reference furnaces and the current furnace and the oxygen consumption values ​​calculated by machine learning are shown in the following table.

[0079] Table 2 Comparison of oxygen consumption values ​​of reference furnace and current furnace and oxygen consumption values ​​calculated by machine learning

[0080] Heat Machine learning to calculate oxygen consumption / Nm3 Actual oxygen consumption / Nm3 Current furnace 7516.8 γ Reference furnace 1 7525.3 7387 Reference furnace 2 7485.2 7297 Reference furnace 3 7564.7 7264 Reference furnace 4 7472.8 7466 Reference furnace 5 7587.3 7614 Reference furnace 6 7585.8 7335 Reference furnace 7 7438.4 7698 Reference furnace 8 7533.0 7334 Reference furnace 9 7414.8 7178 Reference furnace 10 7538.4 7746

[0081] Substituting the data into calculation, the oxygen consumption of the current furnace is 7434.1Nm3, and the actual oxygen consumption of the current furnace is 7457Nm3. The error between the calculated value and the actual value of the oxygen consumption of the current furnace is 22.9Nm3, and the error is within the range of 350Nm3.

Claims

1. A converter oxygen consumption optimization method based on a random forest incremental prediction model, characterized in that: The steps include: Step 1: Build a machine learning model; Step 2: Establish an incremental model; Step 3: Select reference furnaces that meet the requirements and calculate the oxygen consumption values ​​of these reference furnaces according to the machine learning model; Step 4: Calculate the oxygen consumption of the current furnace according to the incremental model.

2. The converter oxygen consumption optimization method based on the random forest incremental prediction model according to claim 1, characterized in that: In step 1, building a machine learning model specifically includes the following steps: Step 11: Get the original data, eliminate anomalies and noise, select useful features for data division, and use them for training set, validation set and test set; Step 12: Use random forest regression algorithm; Step 13: Model training and evaluation; Step 14: Package the model; Step 15: Calculate the oxygen consumption value Y through the machine learning model 机器 .

3. The converter oxygen consumption optimization method based on the random forest incremental prediction model according to claim 2, characterized in that: In step 13, the random model is trained and evaluated. The evaluation is calculated based on the training results. The smaller the sum of squares of the difference between the predicted value and the actual value, the better. Common indicators of random forest regression include mean square error (MSE) and R 2 The smaller the MSE value, the better the model fits the data. 2 The closer the value is to 1, the better the model fits the data, and vice versa; In the formula, n represents the number of samples, y i represents the true value of the i-th sample, Y i Represents the predicted value of the i-th sample; Where: Represents the average of all true sample values.

4. The converter oxygen consumption optimization method based on the random forest incremental prediction model according to claim 1, characterized in that: In step 2, establishing the incremental model specifically includes the following steps: Step 21, reference furnace selection and calculation: select the N furnaces closest to the current furnace that meet the requirements as the reference furnace, and then calculate the target deviation value ΔY of the N furnaces reference furnace. The calculation formula is as follows: where Y 参考炉实际i is the actual target value of the reference furnace, Y 参考炉理论i It is the theoretical calculated value of the reference furnace target. Step 22: Build an incremental model: Set the current furnace target theoretical calculation value to Y 理论 , the current furnace target Y value is calculated as follows: Y=Y 理论 +ΔY。 5. The converter oxygen consumption optimization method based on the random forest incremental prediction model according to claim 1, characterized in that: In step 3, a reference furnace that meets the requirements is selected, the endpoint carbon is within the target range, the endpoint oxygen is within the range of 0.02-0.06%, and the furnace has molten iron information and scrap steel information.

6. The method for optimizing converter oxygen consumption based on the random forest incremental prediction model according to claim 1, characterized in that: In step 4, the oxygen consumption of the current furnace is calculated according to the incremental model, and the calculation result of the machine learning model is substituted into the incremental model calculation formula to calculate the oxygen consumption γ of the current furnace: Where γ is the oxygen consumption of the current furnace, γ 机器学习 Machine learning calculates oxygen consumption for the current furnace, γ 参考炉实际 is the actual oxygen consumption of the reference furnace, γ 参考炉机器学习 Machine learning calculation of oxygen consumption for a reference furnace.

7. A system for implementing the converter oxygen consumption optimization method based on the random forest incremental prediction model as claimed in claim 1, characterized in that: include: Machine learning module, incremental model module and selection calculation module; The machine learning module establishes a machine learning model, the incremental model module establishes an incremental model, the calculation module selects reference furnaces that meet the requirements, calculates the oxygen consumption values ​​of these reference furnaces according to the machine learning model, and calculates the oxygen consumption of the current furnace according to the incremental model.

8. The converter oxygen consumption optimization system based on the random forest incremental prediction model according to claim 7, characterized in that: Select a reference furnace that meets the requirements, with endpoint carbon within the target range, endpoint oxygen within the range of 0.02-0.06%, and furnaces with molten iron information and scrap steel information.

9. The converter oxygen consumption optimization system based on the random forest incremental prediction model according to claim 7, characterized in that: Calculate the oxygen consumption γ of the current furnace: Where γ is the oxygen consumption of the current furnace, γ 机器学习 Machine learning calculates oxygen consumption for the current furnace, γ 参考炉实际 is the actual oxygen consumption of the reference furnace, γ 参考炉机器学习 Machine learning calculation of oxygen consumption for a reference furnace.

Images