Intelligent online optimization method of converter steelmaking process for low-carbon and low-cost smelting

Abstract

The invention belongs to the technical field of ferrous metallurgy, and particularly relates to an intelligent online optimization method of a converter steelmaking process for low-carbon low-cost smelting, which applies field manufacturing standards and historical furnace production data and comprehensively considers indexes such as converter molten iron conditions, smelting targets, upper and lower limits of manufacturing conditions, smelting cost and carbon emission intensity. By combining intelligent algorithms such as local online modeling and multi-objective optimization, intelligent online optimization of converter steelmaking smelting and operation process parameters is achieved, the intelligent control level of a field converter is improved, and the dual purposes of energy conservation, emission reduction, cost reduction and efficiency improvement of converter smelting are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of iron and steel metallurgy, and in particular to an intelligent online optimization method for a converter steelmaking process for low-carbon and low-cost smelting. Background Art

[0002] Low-carbon and low-cost converter smelting is the key path for the current steel industry to pursue green transformation and economic benefits. As the global climate change problem becomes increasingly serious, reducing carbon emissions has gradually become a consensus of the international community. As a major carbon emitter, the steel industry faces tremendous pressure to reduce emissions. At the same time, rising raw material prices and increased energy costs have also prompted steel companies to seek low-cost production methods. Therefore, low-carbon and low-cost converter smelting has become an important direction for the transformation and upgrading of the steel industry. Through technological innovation and process optimization, it aims to achieve the dual goals of energy conservation and emission reduction and cost reduction and efficiency improvement.

[0003] At present, the optimization of converter smelting and operation process parameters of steel enterprises is mainly based on on-site manufacturing procedures and manual experience, which is difficult to adapt to refined production control and diversified smelting goals. To this end, the present invention proposes an intelligent online optimization method for converter steelmaking process for low-carbon and low-cost smelting, which uses on-site manufacturing standards and historical furnace production data in the big data platform of steel enterprises, comprehensively considers multiple indicators such as converter molten iron conditions, smelting goals, upper and lower limits of manufacturing conditions, smelting costs and carbon emission intensity for the converter smelting process, and combines local online modeling and multi-objective optimization and other intelligent algorithms to perform intelligent online optimization of the converter steelmaking process, achieve converter low-carbon emissions and low-cost smelting production, and guide and improve the converter smelting level of on-site operators. Summary of the invention

[0004] In order to solve the problems existing in the prior art, the main purpose of the present invention is to propose an intelligent online optimization method for converter steelmaking process for low-carbon and low-cost smelting. Based on the converter on-site manufacturing standards and historical furnace production data, it comprehensively considers multiple indicators such as converter molten iron conditions, smelting goals, upper and lower limits of manufacturing conditions, smelting costs and carbon emission intensity, and adopts local online modeling and multi-objective optimization and other intelligent algorithms to perform intelligent online optimization of converter steelmaking and operating process parameters, and guide the on-site realization of converter low-carbon and low-cost smelting production. Specifically,

[0005] An intelligent online optimization method for converter steelmaking process for low-carbon and low-cost smelting, characterized in that, for the converter smelting target of the current furnace, the on-site manufacturing standards and historical furnace production data are applied to optimize the converter production process parameters of the current furnace online and in real time to meet the requirements of low-carbon and low-cost converter smelting; the optimization model used comprehensively considers indicators such as converter molten iron conditions, smelting targets, upper and lower limits of manufacturing conditions, smelting costs and carbon emission intensity, and flexibly adjusts the importance of different indicators in the optimization process through weight coefficients; combined with intelligent algorithms such as local online modeling and multi-objective optimization, the optimal smelting and operating process parameters of the current furnace are output to achieve intelligent online optimization of the converter steelmaking process.

[0006] Furthermore, the optimization model construction includes three modules:

[0007] 1) Initial value selection module: Taking into account the converter molten iron conditions, smelting targets, smelting costs and carbon emission intensity, select the furnace samples with similar smelting targets to the current furnace and lower costs and carbon emissions from the historical furnaces as the initial values ​​for optimization;

[0008] 2) Local weighted regression modeling module: Based on the obtained optimal initial value, local weighted regression modeling is performed in the vicinity. The solution process includes sample weight determination, fitting equation construction, equation parameter optimization, and loss function evaluation.

[0009] 3) Optimal condition derivation module: Combining multiple objective conditions such as converter smelting objectives, upper and lower limits of manufacturing conditions, smelting costs and carbon emission intensity, the optimization model is iteratively solved until convergence is reached, and the optimal smelting and operating process parameters for the current furnace are output.

[0010] Furthermore, the initial value selection module can be expressed as:

[0011]

[0012] In the formula, y n is the molten iron condition and smelting end point of the historical furnace, y 0 For the current furnace hot metal conditions and smelting targets, is a diagonal matrix of appropriately scaled output variables, x is the converter smelting and operating process parameters, q is the weight of the regression model fitting results and the smelting target deviation, r is the weight of the smelting cost, c≡[c1c2…c M ] T represents the cost value corresponding to each process parameter, p represents the weight of carbon emission intensity, e≡[e1e2…e M ] T Indicates the carbon emission intensity value corresponding to each process parameter.

[0013] Furthermore, the local weighted regression modeling module can be expressed as:

[0014] 1) Sample weight

[0015]

[0016] In the formula, is the Euclidean distance between the query sample x and the i-th sample; σ is a hyperparameter that determines the degree of fit of the model.

[0017] 2) Fitting equation

[0018]

[0019] In the formula, is the predicted value of the i-th output; b (i) is the constant term of the mth input; a m(i) is the i-th parameter of the m-th input; x m is the mth input.

[0020] 3) Equation parameters

[0021] θ (i) =[b (i) a 1(i) a 2(i) …a M(i) ] T

[0022] Where b (i) is the constant term of the mth input; a M(i) is the i-th parameter of the M-th input.

[0023] 4) Loss Function

[0024]

[0025] In the formula, X is the matrix form of the input,

[0026] Furthermore, the optimal condition derivation module can be expressed as:

[0027]

[0028] x LB ≤x≤x UB

[0029] y LB ≤y≤y UB

[0030] In the formula, x LB Indicates the lower limit of the process parameter, x UB Indicates the upper limit of the process parameter, y LB Indicates the lower limit of the smelting target, y UBIndicates the upper limit of the smelting target.

[0031] Furthermore, the cost value corresponding to each process parameter is provided by the ERP management system of the steel enterprise, the upper and lower limits of the process parameters and smelting targets are specified by the on-site manufacturing standards, and the carbon emission intensity value is calculated based on the energy consumption and material consumption corresponding to each process parameter.

[0032] Furthermore, the weight value range is: 0.5≤q≤1, 0≤c≤1, 0≤e≤1.

[0033] Furthermore, the smelting and operating process parameters include scrap steel ratio, lime addition amount, dolomite addition amount, oxygen blowing amount, gun position, smelting time, deoxidation alloy addition amount, etc.

[0034] The beneficial effects of the present invention are as follows:

[0035] The present invention proposes an intelligent online optimization method for converter steelmaking process for low-carbon and low-cost smelting. It utilizes on-site manufacturing standards and historical furnace production data in the big data platform of steel enterprises, comprehensively considers multiple indicators such as converter molten iron conditions, smelting goals, upper and lower limits of manufacturing conditions, smelting costs and carbon emission intensity for the converter smelting process, combines local online modeling and multi-objective optimization and other intelligent algorithms, and performs intelligent online optimization of the converter steelmaking process, thereby improving the intelligent control level of converter smelting, which is conducive to achieving the dual goals of energy conservation and emission reduction and cost reduction and efficiency improvement of converter smelting. DETAILED DESCRIPTION

[0036] The following will be described clearly and completely in conjunction with the technical solutions in the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] The technical solution of the present invention is further described below in conjunction with specific embodiments.

[0038] Example

[0039] Taking the silicon steel converter steelmaking production of a steel plant as an example, the historical production data of 2620 furnaces were collected for modeling. At the beginning of a furnace converter smelting, the molten iron conditions were [C] = 4.32%, [Si] = 0.41%, [Mn] = 0.3%, [P] = 0.016%, T = 1278°C. The carbon temperature target at the end of converter smelting was [C] = 0.06%, T = 1665°C.

[0040] According to the initial value selection formula (1), a heat sample with a smelting target close to the current heat and lower cost and carbon emission is selected from the historical heats.

[0041]

[0042] In the formula, y n is the molten iron condition and smelting end point of the historical furnace, y 0 For the current furnace hot metal conditions and smelting targets, is a diagonal matrix of appropriately scaled output variables, x is the converter smelting and operating process parameters, q is the weight of the regression model fitting results and the smelting target deviation, r is the weight of the smelting cost, c≡[c1c2…c M ] T represents the cost value corresponding to each process parameter, p is the weight of carbon emission intensity, e≡[e1e2…e M ] T Indicates the carbon emission intensity value corresponding to each process parameter.

[0043] The weights of the regression model fitting results and the smelting target deviations are set to q=1, the weight of the smelting cost is set to r=1, and the weight of the carbon emission intensity is set to e=1. The initial value selection module is used to determine that the initial value furnace sample is molten iron [C]=4.35%, [Si]=0.44%, [Mn]=0.31%, [P]=0.017%, T=1290°C, the converter smelting end point carbon [C]=0.05%, and the end point temperature is 1663°C.

[0044] The local weighted modeling module is used to build a local prediction model. By prioritizing past data near the initial value, a linear prediction model with sufficient accuracy can be built around the initial value. By using the generated model, the output corresponding to the target value is calculated.

[0045] The local weighted linear regression model is used to assign weights to samples. The weight calculation formula is shown in formula (2).

[0046]

[0047] In formula (2) ‖xx (i) ‖2 is the weighted Euclidean distance, x is the query function (initial value), x (i) is the i-th training sample (historical sample), and σ is a hyperparameter that determines the degree of model fit. If the hyperparameter σ is too small, the model will be overfitted; if σ is too large, the model will be underfitted. In practical applications, there are few cases where the output variables and feature vectors of the data are linearly related, so the pure linear regression model is not very applicable, but the local weighted linear regression can make a better fit for nonlinear data distribution. Randomly extract n historical samples from the database for full cross-validation to determine the appropriate σ value.

[0048] For a given feature vector as input to predict the output variable, the weighted loss function of formula (3) is used. Formula (3) is the summation equation with weights added to the loss function. By minimizing the loss function, the algorithm can find the value of θ when the sample distance is close to x and the loss is minimized. In other words, the algorithm pays more attention to those points close to x, which is also conducive to more accurate prediction of y. Formula (3) can be expressed in the form of a matrix using formula (4). The local linear regression model expression is shown in formula (5).

[0049]

[0050] In the formula, is the predicted value of the ith output, b (i) is the constant term of the mth input, a m(i) is the i-th parameter of the m-th input, x m is the mth input, b (i) is the constant term of the mth input; a M(i) is the i-th parameter of the M-th input.

[0051]

[0052] Finally, the module for deriving the optimal manufacturing conditions uses a quadratic programming method to calculate the objective function to minimize under given constraints. Since the derived optimal manufacturing conditions are different from the initial conditions, it is necessary to continuously iterate in the local model to obtain the local weighted regression model and the optimal manufacturing conditions derivation until the optimal manufacturing conditions converge. This iteration is necessary to ensure that the accuracy of the local linear model is improved around the derived optimal manufacturing conditions. The role of the module for selecting initial values ​​is to provide initial values ​​for the convergence calculation.

[0053] Through equations (2) to (6), the model parameters θ can be obtained (i) =[b (i) a 1(i) a 2(i) …a M(i) ] T The values ​​of various manufacturing parameters in the model are summarized in Table 1, and the model parameter values ​​corresponding to the smelting targets (carbon content, temperature) are summarized in Table 1.

[0054] Table 1 Model parameter values ​​corresponding to operating parameters

[0055]

[0056] Next, the operating parameter x is optimized through a quadratic programming problem, and the optimization problem is shown in formula (7).

[0057]

[0058] x LB ≤x≤x UB

[0059] y LB ≤y≤y UB

[0060] In the formula, x LB Indicates the lower limit of the process parameter, x UB Indicates the upper limit of the process parameter, y LB Indicates the lower limit of the smelting target, y UB Indicates the upper limit of the smelting target.

[0061] The experiment has obtained the fitted prediction equation during the modeling process The expression of y 0 、c T The upper and lower limits of the smelting parameters and the end-point composition in formula (7) are determined based on actual historical data, and the upper and lower limits of the relevant parameters are shown in Table 2.

[0062] Table 2 Upper and lower limits of smelting process parameters

[0063]

[0064] After the second optimization iteration of formula (7), the predicted optimal manufacturing parameters are shown in Table 3.

[0065] Table 3 Model prediction of smelting process parameters

[0066]

[0067] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. An intelligent online optimization method for converter steelmaking process for low-carbon and low-cost smelting, characterized in that: According to the converter smelting target of the current furnace, the on-site manufacturing standards and historical furnace production data are applied to optimize the converter production process parameters of the current furnace online and in real time to meet the requirements of low-carbon emissions and low-cost converter smelting; the optimization model used comprehensively considers indicators such as converter molten iron conditions, smelting targets, upper and lower limits of manufacturing conditions, smelting costs and carbon emission intensity, and flexibly adjusts the importance of different indicators in the optimization process through weight coefficients; combined with local online modeling and multi-objective optimization and other intelligent algorithms, the optimal smelting and operating process parameters of the current furnace are output to achieve intelligent online optimization of the converter steelmaking process.

2. The intelligent online optimization method for converter steelmaking process for low-carbon and low-cost smelting according to claim 1 is characterized in that: The optimization model construction includes three modules: 1) Initial value selection module: Taking into account the converter molten iron conditions, smelting targets, smelting costs and carbon emission intensity, select the furnace samples with similar smelting targets to the current furnace and lower costs and carbon emissions from the historical furnaces as the initial values ​​for optimization; 2) Local weighted regression modeling module: Based on the obtained optimal initial value, local weighted regression modeling is performed near it. The solution process includes sample weight determination, fitting equation construction, equation parameter optimization and loss function evaluation. 3) Optimal condition derivation module: Combining multiple objective conditions such as converter smelting objectives, upper and lower limits of manufacturing conditions, smelting costs and carbon emission intensity, the optimization model is iteratively solved until convergence is reached, and the optimal smelting and operating process parameters for the current furnace are output.

3. According to the intelligent online optimization method of converter steelmaking process for low-carbon and low-cost smelting according to claims 1-2, it is characterized in that: The initial value selection module can be expressed as: In the formula, y n is the molten iron condition and smelting end point of the historical furnace, y 0 For the current furnace hot metal conditions and smelting targets, is a diagonal matrix of appropriately scaled output variables, x is the converter smelting and operating process parameters, q is the weight of the regression model fitting results and the smelting target deviation, r is the weight of the smelting cost, c≡[c1c2…c M ] T represents the cost value corresponding to each process parameter, p represents the weight of carbon emission intensity, e≡[e1e2…e M ] T Indicates the carbon emission intensity value corresponding to each process parameter.

4. The intelligent online optimization method for converter steelmaking process for low-carbon and low-cost smelting according to claim 1-2 is characterized in that: The local weighted regression modeling module can be expressed as: 1) Sample weight In the formula, is the Euclidean distance between the query sample x and the i-th sample; σ is a hyperparameter that determines the degree of fit of the model. 2) Fitting equation In the formula, is the predicted value of the i-th output; b (i) is the constant term of the mth input; a m(i) is the i-th parameter of the m-th input; x m is the mth input. 3) Equation parameters i (i) =[b (i) a 1(i) a 2(i) ...a M(i) ] T Where b (i) is the constant term of the mth input; a M(i) is the i-th parameter of the M-th input. 4) Loss Function In the formula, X is the matrix form of the input, 5. The intelligent online optimization method for converter steelmaking process for low-carbon and low-cost smelting according to claim 1-2 is characterized in that: The optimal condition derivation module can be expressed as: x LB ≤x≤x UB and LB ≤y≤y UB In the formula, x LB Indicates the lower limit of the process parameter, x UB Indicates the upper limit of the process parameter, y LB Indicates the lower limit of the smelting target, y UB Indicates the upper limit of the smelting target.

6. The intelligent online optimization method for converter steelmaking process for low-carbon and low-cost smelting according to claims 1-5 is characterized in that: The cost value corresponding to each process parameter is provided by the ERP management system of the steel enterprise. The upper and lower limits of the process parameters and smelting targets are specified by the on-site manufacturing standards. The carbon emission intensity value is calculated based on the energy consumption and material consumption corresponding to each process parameter.

7. The intelligent online optimization method for converter steelmaking process for low-carbon and low-cost smelting according to claims 1-5 is characterized in that: The weight value range is: 0.5≤q≤1, 0≤c≤1, 0≤e≤1.

8. The intelligent online optimization method for converter steelmaking process for low-carbon and low-cost smelting according to claims 1-5 is characterized in that: The smelting and operation process parameters include scrap steel ratio, lime addition amount, dolomite addition amount, oxygen blowing amount, smelting time, deoxidation alloy addition amount, etc.

Images