Method, device, equipment and medium for constructing molten iron model

Abstract

The present invention relates to the technical field of industrial big data analysis, and discloses a method, device, equipment and medium for constructing a molten iron model. The method includes obtaining historical molten iron production data, including raw material parameters, process parameters, and molten iron quality parameters; based on the obtained molten iron quality parameters, according to the molten iron quality requirements and the principle of minimum coke ratio, and using Pareto optimization and / or industrial standard filtering, screening out the optimal process parameter combination corresponding to different raw material parameters; using an integrated learning algorithm to construct a regression model, using the obtained raw material parameters as input data, using the screened optimal process parameter combination as output data, training the regression model, inputting the raw material parameters to be produced into the trained regression model, and obtaining the optimal process parameter combination. The method provided by the present invention can obtain process parameters that take into account molten iron quality, coke ratio and equipment safety based on the raw material parameters to be produced, thereby achieving process optimization for molten iron production.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial big data analysis, and in particular to a method, device, equipment and medium for constructing a molten iron model. Background Art

[0002] The setting of molten iron process parameters (such as air temperature and oxygen enrichment rate) directly affects molten iron quality, coke consumption and production safety. Traditional process optimization mainly relies on operator experience or single mechanism models, which have significant limitations. For example, due to factory or market reasons, production may face complex raw material parameter fluctuations. Manual adjustment of process parameters has a lag phenomenon and is difficult to quantify to obtain the optimal solution. The single mechanism model lacks a system coordination mechanism and is difficult to coordinate multiple conflicting goals such as meeting molten iron quality standards, minimizing coke ratios, and equipment safety boundaries. Therefore, although there are many years of molten iron production experience and a wealth of historical production data, this valuable historical production data has not been fully utilized to optimize existing molten iron production. Summary of the Invention

[0003] The purpose of the present invention is to provide a method, device, equipment and medium for constructing a molten iron model, which can obtain process parameters that take into account molten iron quality, coke ratio and equipment safety based on the raw material parameters to be produced, and realize process optimization of molten iron production, in order to address the problems in related technologies such as difficulty in achieving efficient process parameter adjustment and difficulty in coordinating multiple conflicting production goals.

[0004] In order to achieve the above-mentioned purpose, in a first aspect of the present invention, a method for constructing a molten iron model is provided, comprising: S1, obtaining historical molten iron production data, including raw material parameters, process parameters, and molten iron quality parameters; S2, based on the obtained molten iron quality parameters, in accordance with the molten iron quality requirements and the principle of minimum coke ratio, and using Pareto optimization and / or industrial standard filtering, screening out the optimal process parameter combination corresponding to different raw material parameters; S3, using an integrated learning algorithm to construct a regression model, using the obtained raw material parameters as input data and the screened optimal process parameter combination as output data, training the regression model, and adding thermodynamic equilibrium constraints, equipment safety boundaries, and process procedures as regularization terms to the loss function of the regression model; S4, inputting the parameters of the raw materials to be produced into the trained regression model, and using a sequential quadratic programming algorithm to fine-tune the output results of the regression model based on the thermodynamic equilibrium constraints, equipment safety boundaries, and process procedures, to obtain the optimal process parameter combination, continuously collect production data, and update the regression model through incremental learning.

[0005] In a second aspect of the present invention, a device for constructing a molten iron model is provided, comprising: an acquisition module for acquiring historical molten iron production data, including raw material parameters, process parameters, and molten iron quality parameters; a screening module for screening out optimal process parameter combinations corresponding to different raw material parameters based on the acquired molten iron quality parameters, in accordance with molten iron quality requirements and the principle of minimum coke ratio, and by adopting Pareto optimization or industrial standard filtering; a training module for constructing a regression model using an integrated learning algorithm, taking the acquired raw material parameters as input data and the screened optimal process parameter combination as output data, training the regression model, and adding thermodynamic equilibrium constraints, equipment safety boundaries, and process procedures as regularization terms to the loss function of the regression model; an application update module for inputting the parameters of the raw materials to be produced into the trained regression model, and using a sequential quadratic programming algorithm to fine-tune the output results of the regression model based on the thermodynamic equilibrium constraints, equipment safety boundaries, and process procedures, to obtain the optimal process parameter combination, continuously collect production data, and update the regression model through incremental learning.

[0006] In a third aspect of the present invention, an electronic device is provided, comprising a memory and a processor coupled to the memory, wherein the processor is configured to execute the method of the first aspect based on instructions stored in the memory.

[0007] In a fourth aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the method of the first aspect is implemented.

[0008] The technical solution of the present invention uses historical data to construct a regression model from raw material parameters to optimal process parameters. It can capture the nonlinear relationship in complex smelting processes that is difficult to describe with a precise mechanism model. It is more reliable than traditional experience or a single model. Through systematic solution optimization, it can intelligently generate an optimal process parameter combination for a specific raw material ratio that meets the molten iron quality requirements, significantly reduces the coke ratio, and meets the safety of equipment operation. It effectively solves the problem of collaborative optimization of multiple objectives such as quality, cost, efficiency, and safety in smelting. The constructed regression model has the ability to adapt to changes in raw materials and continuously improve itself, and can achieve stable, efficient, low-consumption and safe operation of molten iron production. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application, a brief introduction will be given below to the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0010] Figure 1 is a flow chart of the method for constructing the molten iron model described in the embodiment;

[0011] Figure 2 This is a flow chart of obtaining parameters of raw materials to be produced based on market prices according to the embodiment;

[0012] Figure 3 Schematic diagram of the structure of the device for constructing the molten iron model described in the embodiment;

[0013] Figure 4 Schematic diagram of the structure of the electronic device described in the embodiment. DETAILED DESCRIPTION

[0014] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0015] Since ironmaking involves a large number of process parameters and process indicators, some existing molten iron production models predict the optimal molten iron quality based on the input raw material ratio to obtain the optimal process parameters. However, in actual application, the prediction effect is often not achieved. However, in this embodiment, through the design of a regression model, the optimal process parameters obtained are based on the optimal molten iron quality verified by historical data, so it can achieve a good optimization effect.

[0016] The present application is described below with reference to specific embodiments in conjunction with the accompanying drawings.

[0017] like Figure 1 As shown, a method for constructing a molten iron model in this embodiment includes: S1, obtaining historical molten iron production data, including raw material parameters, process parameters, and molten iron quality parameters.

[0018] Raw material parameters include ore type and its chemical composition, and coke composition. For example, ore types can include sintered ore, pelletized ore, lump ore, and miscellaneous ore, and the corresponding chemical compositions can include specific contents of TFe, SiO2, Al2O3, CaO, MgO, etc. The coke composition mainly includes the specific contents of fixed carbon, ash, and sulfur. Process parameters may include fuel ratio (mainly the proportion of coke and pulverized coal), solvent ratio (such as the proportion of limestone and dolomite), coal injection amount (mainly the amount of pulverized coal or pulverized coal substitutes), air supply parameters (mainly air temperature, air volume, wind speed, and oxygen enrichment), etc. Molten iron quality parameters include molten iron temperature, silicon content, and sulfur content. Each set of raw material parameters corresponds to at least one set of process parameters and corresponding molten iron quality parameters.

[0019] S2, based on the obtained molten iron quality parameters, in accordance with the molten iron quality requirements and the principle of minimum coke ratio, and using Pareto optimization and / or industrial standard filtering, screen out the optimal process parameter combination corresponding to different raw material parameters.

[0020] After acquiring historical data and before screening, the historical data can be preprocessed, such as cleaning the data and eliminating abnormal operating conditions, performing consistency checks, etc., to ensure the physical rationality and logical consistency of the historical data, and avoid distortion of the optimal process parameters for screening due to erroneous data. At the same time, some data can be standardized (such as air volume, oxygen enrichment) or normalized to eliminate dimensional effects and improve the convergence of the algorithm.

[0021] The quality requirements for molten iron can be determined based on the actual operating conditions of the factory. The principle of minimum coke ratio is the core energy efficiency goal of ironmaking. It means minimizing the coke consumption required to produce each ton of qualified molten iron through systematic optimization. It is to minimize the dependence on expensive coke while ensuring the smooth operation of equipment and the quality of molten iron, thereby effectively reducing production costs and carbon emissions, and achieving dual optimization of economic benefits and environmental protection and emission reduction.

[0022] S3 uses an integrated learning algorithm to build a regression model, takes the obtained raw material parameters as input data, and the screened optimal process parameter combination as output data. The regression model is trained, and thermodynamic equilibrium constraints, equipment safety boundaries, and process procedures are added as regularization terms to the loss function of the regression model.

[0023] In some optional implementations, the ensemble learning algorithm can specifically use LightGBM (Light GradientBoosting Machine) to construct regression model training. Compared with using XGBoost (Extreme GradientBoosting) or deep neural networks, the use of LightGBM can quickly perform experimental iterations, has good robustness, is insensitive to missing values ​​and outliers in the data, and is particularly suitable for the characteristics of molten iron production data, combining efficiency and ease of use.

[0024] The regularization term is to add an additional penalty term to the basic loss function to penalize predictions that violate the constraints. The regularization term is added to the loss function during the training phase to embed constraints. Specifically, the thermodynamic equilibrium equation is introduced as a soft constraint, and the equipment safety boundary and process procedures are used as hard constraints. This can effectively improve the model's ability to learn the historical optimal operating rules, so that the predicted value is close to the theoretically optimal process parameters, and can also ensure that the predicted value is within the safe range of the process parameter implementation. In particular, the weight coefficient corresponding to the constraint in the regularization term can be increased according to the working conditions, so that the output results of the regression model can focus more on safety. In addition, after imposing constraints in the training phase, the optimization burden of the regression model in the inference phase can also be reduced.

[0025] S4, input the parameters of the raw materials to be produced into the trained regression model, and use the sequential quadratic programming algorithm to fine-tune the output results of the regression model based on thermodynamic equilibrium constraints, equipment safety boundaries and process regulations to obtain the optimal process parameter combination, continuously collect production data and update the regression model through incremental learning.

[0026] After completing the training of the regression model, the factory's raw material parameters to be produced can be input into the regression model to obtain the optimal process parameter combination output by the regression model. Then, based on the actual needs of the factory, a set of process parameters can be selected from the optimal process parameter combination and selectively input into the control systems of some equipment such as rotary kilns, vertical furnaces, and blast furnaces to guide the on-site production of molten iron. For example, a set of process parameters with the lowest coke ratio can be selected, or a set of process parameters with a slightly higher coke ratio but the lowest silicon content in the molten iron can be selected.

[0027] During the inference stage of the regression model, the data constraints embedded in the sequential quadratic programming algorithm can ensure the feasibility and safety of the optimal process parameters predicted by the regression model. The fine-tuning of the sequential quadratic programming algorithm has a real-time response at the millisecond level, which can quickly complete refined adjustments to ensure that the output results strictly meet all constraints. That is, the sequential quadratic programming algorithm is used as a fine-tuner for the output results of the regression model; by dynamically updating the model parameters through incremental learning (such as iForest and online gradient descent), the training data of the regression model can be continuously enriched, thereby continuously improving the training effect of the regression model and enhancing the prediction accuracy of the regression model. In addition, it should be noted that the previous preprocessing method of historical data should also be used in incremental learning to ensure that the distribution of new data is consistent with that of historical data.

[0028] The method for constructing a molten iron model of the present invention uses historical data to construct a regression model from raw material parameters to optimal process parameters. It can capture nonlinear relationships in complex smelting processes that are difficult to describe with precise mechanism models, and is more reliable than traditional experience or single models. Among them, for the input historical data, high-quality training data can be screened out through molten iron quality requirements and the minimum coke ratio principle, which is a key step to ensure the practicality of regression model prediction. Pareto optimization and industrial standard filtering can achieve efficient screening under multiple conditions. In the training and prediction stages of the regression model, constraints such as thermodynamic equilibrium, equipment safety boundaries, and process regulations are embedded to ensure that the output process parameters comply with basic physical and chemical principles and ensure the feasibility of the scheme in engineering practice. Therefore, through systematic scheme optimization, it is possible to intelligently generate an optimal process parameter combination for a specific raw material ratio that meets the molten iron quality requirements and significantly reduces the coke ratio, effectively solving the problem of collaborative optimization of multiple objectives such as quality, cost, efficiency, and safety in smelting. It has the ability to adapt to raw material changes and continuous self-improvement, and can achieve stable, efficient, low-consumption and safe operation of molten iron production.

[0029] In some optional embodiments, in step S2, the molten iron quality requirements include limitations on molten iron temperature, silicon content, and sulfur content, and Pareto optimization is to use molten iron temperature, silicon content, sulfur content, and coke ratio as multiple optimization targets, and obtain the Pareto optimal solution on the premise of meeting the molten iron quality requirements to form an optimal process parameter combination corresponding to different raw material parameters.

[0030] The quality requirements for molten iron usually include molten iron temperature ≥ first target value, silicon content ≤ second target value, and sulfur content ≤ third target value as the conditions for qualified molten iron quality. This is also the production target of the factory. The specific corresponding target values ​​are determined according to on-site needs.

[0031] Pareto optimization specifically involves a trade-off between molten iron quality and coke ratio, and obtaining the Pareto optimal solution to form an optimal combination of process parameters while meeting the molten iron quality requirements. For example, among multiple Pareto optimal solutions, the molten iron quality parameters corresponding to the first set of process parameters meet the molten iron quality requirements and have the lowest coke ratio, but the sulfur content of the molten iron is so high that it is close to the third target value. The molten iron quality parameters corresponding to the second set of process parameters, while meeting the molten iron quality requirements, have a slightly higher coke ratio than the first set, but the sulfur content of the molten iron is significantly lower. In this case, both sets of process parameters belong to the Pareto optimal solution, that is, in the optimal process parameter combination formed by Pareto optimization, each set of process parameters has different advantages, which are not listed here one by one.

[0032] Industrial standard filtering directly screens multiple sets of process parameters corresponding to the molten iron quality parameters that meet the molten iron quality requirements, and then sorts and selects the multiple sets of process parameters according to the coke ratio, for example, selecting the top five sets of process parameters with the best coke ratio. Therefore, industrial standard filtering takes the coke ratio as the only optimization target, and the calculation to obtain the optimal process parameter combination is simple and easy to implement. Pareto optimization can discover the trade-off relationship between multiple optimization targets, retaining process parameters with slightly higher coke ratios but significantly better quality indicators. Therefore, the optimal process parameter combination obtained is more comprehensive, but compared with industrial standard filtering, the calculation and application of Pareto optimization are more complicated. Factories can choose one screening method or combine two screening methods to obtain multiple sets of training data according to actual needs.

[0033] In some optional embodiments, in step S3 and step S4, the thermodynamic equilibrium constraint is determined by the predicted value and the theoretical value of the silicon content.

[0034] The silicon content of molten iron is strongly positively correlated with the furnace hearth temperature. Using silicon content as the basis for thermodynamic equilibrium constraints allows for both high-precision and time-efficient balance verification, verifying the rationality of energy distribution. Furthermore, manganese and carbon contents can also be used as auxiliary indicators for thermodynamic equilibrium constraints.

[0035] In some optional embodiments, the theoretical value of silicon content is obtained by calculating the molten iron temperature based on the regression model output, and the predicted value of silicon content is obtained based on the raw material parameters and the process parameters output by the regression model, specifically by constructing a silicon content prediction sub-model.

[0036] The theoretical value of silicon content can be obtained by the formula k*e −a / T Calculated, where k is the theoretical coefficient, T is the theoretical molten iron temperature (which can be obtained through the output of the regression model), a is the fitting parameter (determined by regression of historical data), and e is a natural constant; when obtaining the predicted value of the silicon content, compared with selecting historical data with similar working conditions to the raw material parameters and process parameters to determine the predicted value of the silicon content, the output result is obtained by establishing a silicon content prediction sub-model, which can improve the accuracy of the silicon content prediction value and does not rely on the limitation of similar working conditions in the historical data. It should be noted that the silicon content prediction sub-model can use an existing mature model or be trained by establishing a regression model, and is trained with historical raw material parameters, historical process parameters and historical actual silicon content of molten iron as training data.

[0037] In some optional embodiments, in step S3 and step S4, the equipment safety boundary requires that the wind temperature is ≤ the fourth threshold, the coal injection amount is ≤ the upper limit of the spray gun flow rate, and the process regulations require that the coke load ∈ [4.0, 5.5] and the oxygen enrichment amount is ≤ 30%.

[0038] Limiting wind temperature and coal injection amount can prevent equipment damage or avoid process parameters from being unable to be achieved. For example, according to the requirements of factory equipment, the wind temperature is limited to ≤1250℃ and the coal injection amount is limited to ≤210kg / t. The numerical limit of oxygen enrichment is also to ensure the safe operation of the equipment and the regulation of thermal balance, thereby achieving a balance between improving combustion efficiency and stable operation of the equipment. The constraint range of coke load can balance the heat and permeability of the equipment, avoiding the predicted optimal process parameters to have too little coke ratio, resulting in insufficient molten iron temperature and furnace cooling, and avoiding the risk of too thick coke layer reducing gas utilization and increasing suspended materials.

[0039] In some optional embodiments, after step S1, derived features are calculated based on the raw material parameters and process parameters. The derived features may include comprehensive charge basicity, magnesium-aluminum ratio, charge structure index, theoretical combustion temperature, direct reduction degree, slag basicity binary degree, slag aluminum saturation, etc. In step S3, the corresponding screened derived features and raw material parameters are used as input data to train the regression model. In step S4, the derived features to be produced are calculated in real time according to the raw material parameters to be produced, and are input into the trained regression model to output the optimal process parameters.

[0040] Since the interactions between raw material parameters (such as ore type and chemical composition) are difficult to be automatically captured by the regression model, the derived features (such as comprehensive charge alkalinity and slag alkalinity binary degree) can provide prior knowledge for the model and reduce the learning difficulty by quantifying the core constraints of metallurgical reactions. Therefore, by introducing multiple core derived features, the regression model's ability to fit metallurgical mechanisms can be significantly improved, especially in the scenario of raw material fluctuations to ensure the robustness of the optimization scheme. In addition, some features such as theoretical combustion temperature are directly related to thermodynamic equilibrium and safety margins. Some derived features (such as theoretical combustion temperature) can also be added as regularization terms to the loss function of the regression model, making it easier for the regression model to meet process constraints during the training stage.

[0041] In some optional embodiments, in step S4, the parameters of the raw materials to be produced are obtained based on factory inventory, channel supply or market price.

[0042] Factories face complex fluctuations in raw material parameters, that is, the types, chemical compositions, and coke compositions of various ores may change during production, mainly due to changes in factory inventory, changes in channel supply, and changes in market prices. For example, in order to avoid dependence on a single ore, factory inventory may proactively reserve multiple types of ores. When high-grade iron ore inventory is insufficient, it is necessary to temporarily switch to medium- and low-grade ores. For example, if upstream mine production capacity is insufficient due to natural disasters, equipment failures, and other reasons, resulting in a shortage of the original ore supply, the factory is forced to look for alternative minerals. For example, when the price of the original mineral rises sharply, the factory may turn to lower-priced alternative minerals.

[0043] Among them, when market prices fluctuate, purchasing raw materials with the best cost-effectiveness is an important means for factories to reduce production costs. Therefore, how to determine the optimal raw material parameters based on current market prices is also an issue that factories urgently need to solve.

[0044] In some optional embodiments, such as Figure 2 As shown, in step S4, the parameters of the raw materials to be produced are obtained based on the market price, including the following steps: A1, obtaining raw material composition data and corresponding raw material market price data; A2, based on the obtained raw material composition data and raw material market price data, constructing a linear programming model for minimizing the cost of raw materials per ton of molten iron; A3, setting constraints on the output results of the linear programming model, requiring that the output raw material ratio has a total iron content not lower than a first threshold, a total sulfur content not exceeding a second threshold, a total phosphorus content not exceeding a third threshold, and a total silicon dioxide content within a preset range; A4, using the simplex method to solve the linear programming model, and obtaining the parameters of the raw materials to be produced that can minimize the cost of raw materials per ton of molten iron while satisfying all the constraints in step A3.

[0045] Raw material composition data may include ores of different chemical compositions and cokes of different compositions. The corresponding raw material market prices are the selling prices of ores of different chemical compositions and cokes of different compositions. The raw material cost per ton of molten iron is the sum of the procurement costs of all raw materials involved in the proportioning, where the cost of each raw material is its current market price multiplied by its proportion usage in unit molten iron production. The specific thresholds and preset intervals of the constraint conditions can be set according to the on-site product requirements or process standards. For example, to ensure the quality of pig iron and smooth operation of equipment, the total iron content is required to be ≥60%, the total sulfur content ≤0.04%, the total phosphorus content ≤0.1%, and the total silica content ∈ [4.5%, 6%]. In step A4, the simplex method calculation can be implemented through Python's PuLP library or Excel planning solver, which has high computational efficiency and does not require processing real-time and complex constraints.

[0046] Through price correlation calculation, a linear programming model with the core goal of minimizing the cost of raw materials per ton of molten iron is established, and the simplex method is used to efficiently solve it. On the premise of meeting the molten iron quality requirements, the lowest cost raw material combination plan under the current market price system can be accurately calculated to form the parameters of the raw materials to be produced. Whenever the factory needs to purchase raw materials, the above algorithm can be used to guide the purchasing department to purchase the most cost-effective raw material varieties and quantities to achieve continuous cost optimization. Based on real-time market prices, key quality constraints are strictly incorporated to ensure that the optimized ratio plan absolutely meets the core quality indicators of molten iron production while pursuing low costs. In addition, only basic data tables and linear programming tools are required to obtain the most cost-effective parameters of raw materials to be produced, which is easy to implement in factories. According to the actual situation of the factory, users can customize the constraints of the model and reuse the existing model framework, which has good application prospects.

[0047] The following describes a device for constructing a molten iron model provided in an embodiment of the present application. The device described below and the method described above can be referenced to each other. Based on the above embodiment, Figure 3 It is a structural schematic diagram of a device for constructing a molten iron model provided in this embodiment.

[0048] like Figure 3 As shown, a device for constructing a molten iron model includes: an acquisition module 10, which is used to obtain historical molten iron production data, including raw material parameters, process parameters, and molten iron quality parameters; a screening module 20, which, based on the acquired molten iron quality parameters, according to the molten iron quality requirements and the principle of minimum coke ratio, and adopts Pareto optimization or industrial standard filtering to screen out the optimal process parameter combination corresponding to different raw material parameters; a training module 30, which adopts an integrated learning algorithm to construct a regression model, takes the acquired raw material parameters as input data, and takes the screened optimal process parameter combination as output data, trains the regression model, and adds thermodynamic equilibrium constraints, equipment safety boundaries, and process procedures as regularization terms to the loss function of the regression model; an application update module 40, which is used to input the raw material parameters to be produced into the trained regression model, and uses a sequential quadratic programming algorithm to fine-tune the output results of the regression model based on the thermodynamic equilibrium constraints, equipment safety boundaries, and process procedures to obtain the optimal process parameter combination, continuously collect production data, and update the regression model through incremental learning.

[0049] Based on the above method embodiment, please see Figure 4 , Figure 4 This is a structural diagram of an electronic device provided in an embodiment of the present application. An electronic device 300 provided in an embodiment of the present application includes: a processor 301 and a memory 302, the memory 302 stores machine-readable instructions executable by the processor 301, and the machine-readable instructions execute the above method when executed by the processor 301; the electronic device can be a physical device.

[0050] Based on the above method embodiments, a computer-readable storage medium is further provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method in any of the above embodiments are implemented.

[0051] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0052] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0053] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0054] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0055] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0056] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0057] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information, which can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0058] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0059] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0060] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A method for constructing a molten iron model, characterized in that: include: S1, obtain historical molten iron production data, including raw material parameters, process parameters, and molten iron quality parameters; S2, based on the obtained molten iron quality parameters, in accordance with the molten iron quality requirements and the principle of minimum coke ratio, and using Pareto optimization and / or industrial standard filtering, screen out the optimal process parameter combination corresponding to different raw material parameters; S3 uses an ensemble learning algorithm to build a regression model, taking the acquired raw material parameters as input data and the screened optimal process parameter combination as output data. The regression model is trained and thermodynamic equilibrium constraints, equipment safety boundaries, and process regulations are added as regularization terms to the loss function of the regression model. S4: Input the raw material parameters to be produced into the trained regression model. Based on thermodynamic equilibrium constraints, equipment safety boundaries, and process regulations, a sequential quadratic programming algorithm is used to fine-tune the output of the regression model to obtain the optimal process parameter combination. Production data is continuously collected and the regression model is updated through incremental learning. In step S3 and step S4, the thermodynamic equilibrium constraint is determined by the predicted value and theoretical value of the silicon content; the equipment safety boundary requires that the wind temperature ≤ the fourth threshold, the coal injection amount ≤ the upper limit of the spray gun flow rate, and the process regulations require that the coke load ∈ [4.0, 5.5] and the oxygen enrichment ≤ 30%.

2. The method for constructing a molten iron model according to claim 1, wherein: In step S2, the molten iron quality requirements include the limitations of molten iron temperature, silicon content, and sulfur content, and Pareto optimization takes molten iron temperature, silicon content, sulfur content, and coke ratio as multiple optimization targets, and obtains the Pareto optimal solution on the premise of meeting the molten iron quality requirements to form the optimal process parameter combination corresponding to different raw material parameters.

3. The method for constructing a molten iron model according to claim 1, wherein: The theoretical value of silicon content is obtained by calculating the molten iron temperature based on the regression model output, and the predicted value of silicon content is obtained based on the raw material parameters and the process parameters output by the regression model. Specifically, a silicon content prediction sub-model is constructed to obtain the value.

4. The method for constructing a molten iron model according to claim 1, wherein: In step S4, parameters of the raw materials to be produced are obtained based on factory inventory, channel supply or market price.

5. The method for constructing a molten iron model according to claim 4, characterized in that: In step S4, the parameters of the raw materials to be produced are obtained based on the market price, including the following steps: A1, obtain raw material composition data and corresponding raw material market price data; A2, based on the obtained raw material composition data and raw material market price data, build a linear programming model to minimize the raw material cost per ton of molten iron; A3 sets constraints on the output of the linear programming model, requiring that the output raw material ratio has a total iron content not less than a first threshold, a total sulfur content not exceeding a second threshold, a total phosphorus content not exceeding a third threshold, and a total silicon dioxide content within a preset range; A4. Solve the linear programming model using the simplex method to obtain the raw material parameters that can minimize the raw material cost per ton of molten iron while satisfying all the constraints in step A3.

6. A device for constructing a molten iron model, characterized in that: include: The acquisition module is used to obtain historical production data of molten iron, including raw material parameters, process parameters, and molten iron quality parameters; The screening module, based on the acquired molten iron quality parameters, selects the optimal process parameter combination corresponding to different raw material parameters according to the molten iron quality requirements and the principle of minimum coke ratio, and adopts Pareto optimization or industrial standard filtering; The training module uses an integrated learning algorithm to build a regression model, taking the acquired raw material parameters as input data and the screened optimal process parameter combination as output data to train the regression model. Thermodynamic equilibrium constraints, equipment safety boundaries, and process regulations are added as regularization terms to the loss function of the regression model. The application update module is used to input the parameters of the raw materials to be produced into the trained regression model. Based on thermodynamic equilibrium constraints, equipment safety boundaries, and process regulations, the sequential quadratic programming algorithm is used to fine-tune the output of the regression model to obtain the optimal process parameter combination. Production data is continuously collected and the regression model is updated through incremental learning. Among them, the thermodynamic equilibrium constraint is determined by the predicted value and theoretical value of the silicon content; the equipment safety boundary requires that the wind temperature ≤ the fourth threshold, the coal injection amount ≤ the upper limit of the spray gun flow rate, and the process regulations require that the coke load ∈ [4.0, 5.5] and the oxygen enrichment ≤ 30%.

7. An electronic device, characterized in that: The method comprises a memory and a processor coupled to the memory, wherein the processor is configured to execute the method according to any one of claims 1 to 5 based on instructions stored in the memory.

8. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images