Method for evaluating lag time of influence of blast furnace ironmaking operation parameters on state parameters

Abstract

The invention discloses a method for evaluating lag time of influence of blast furnace ironmaking operation parameters on state parameters, which comprises the following steps of: constructing a mapping relation between time delay sequence data of operation variables and blast furnace state parameters, and processing blast furnace state parameter data by adopting modes such as wavelet filtering and difference so as to construct a drift term, a diffusion term and a random process model. And then, evaluating the operation time delay through the established stochastic differential equation simulation result. Constraint conditions are added for optimization of the model according to the blast furnace smelting characteristics and the hierarchical clustering result, and the optimization efficiency and precision are further improved. Finally, the importance of the parameters is evaluated based on the gain or SHAP algorithm, the regulation and control priorities of the operation variables are ranked, and a basis is provided for regulation and control of the blast furnace ironmaking process. The method has the advantages that the delay time of the influence of the operation variables on the state parameters of the blast furnace can be accurately evaluated by fusing the production rule of blast furnace smelting, the production cost is reduced while the quality and the efficiency of the ironmaking process are improved, and the method has a wide application prospect.

Description

Technical Field

[0001] The present invention relates to the technical field of steel smelting, and in particular to a method and system for evaluating the lag time of the influence of blast furnace ironmaking operation parameters on blast furnace state parameters. Background Art

[0002] Blast furnace ironmaking is the core process in steel production, and the state of the ironmaking blast furnace affects the entire steel production process. For example, the utilization rate of coal gas is directly related to the reduction reaction efficiency and heat utilization efficiency in the blast furnace; the coke ratio is an important technical and economic indicator that measures the raw materials, fuels, equipment and technical operation level of blast furnace ironmaking; the precise control of the content of molten iron components, such as silicon, phosphorus, sulfur, manganese and other elements, determines the physical properties and chemical properties of steel, and affects the composition control and quality management in the subsequent steelmaking process, which is crucial to improving the quality of steel products and reducing production costs. However, the blast furnace ironmaking process is complex and has significant nonlinear and multi-factor coupling characteristics. The operating parameters such as charging system, air supply system, and charge composition have a large lag in the impact on the blast furnace state.

[0003] During the blast furnace smelting process, changes in blast furnace state parameters such as gas utilization, coke ratio, and molten iron composition usually do not immediately reflect the adjustment of ironmaking operations, but there is a certain time delay. This lag effect not only affects the accuracy and timeliness of the production process, but also poses a challenge to the stable operation of the blast furnace, the quality control of molten iron, and the economic benefits of the entire plant. If the lag time of the state parameter change cannot be accurately grasped, it may cause fluctuations in the blast furnace production state and no longer be stable and smooth, or even cause blast furnace production accidents, which will affect the overall benefits of the entire steel production line and factory. In addition, the estimation of operation delay is directly related to the energy consumption in the smelting process, the reasonable input of raw materials, and the control of production costs. Therefore, its accurate estimation is of great significance for the optimization of the blast furnace ironmaking process.

[0004] At present, the estimation of lag time usually relies on empirical formulas or simplified mathematical models. However, these methods usually fail to effectively consider the nonlinear characteristics existing in the blast furnace smelting process and the complex interactions between the operating variables. For example, although the traditional time series correlation matrix or gray correlation analysis can estimate the lag time to a certain extent, these methods ignore the nonlinear relationship between variables and the coupling effect of multiple factors, resulting in the inability to accurately reflect the true lag effect of changes in blast furnace smelting state parameters such as molten iron composition, gas utilization rate and coke ratio during the smelting process. In addition, the existing technology often lacks constraints on the operating laws of blast furnace smelting in the optimization process of solving the problem, which may cause the obtained lag time results to be inconsistent with the actual production, thereby affecting the optimization of operating parameters and the fine control of blast furnace smelting.

[0005] Therefore, how to accurately estimate the lag time of state parameters responding to changes in operating parameters and consider the nonlinear and multi-factor coupling characteristics in the blast furnace smelting process has become the key to improving blast furnace smelting efficiency, reducing costs and improving product quality. Summary of the invention

[0006] In view of the defects of the prior art, the present invention provides a method for evaluating the lag time of the influence of blast furnace ironmaking operation parameters on state parameters.

[0007] In order to achieve the above invention object, the technical solution adopted by the present invention is as follows:

[0008] A method for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters comprises the following steps:

[0009] S1: Construct the time delay sequence data of the operation variable according to the delay list Γ = [τ1, τ2, ... τ i ,…π n ], where τ i The time delay set for each variable is used to construct the sequence data of the blast furnace state parameters As input data, the corresponding relationship between the blast furnace state parameter x0 and the corresponding operation variable sequence is constructed:

[0010]

[0011] In the formula, The actual value of the blast furnace status parameters at the current moment.

[0012] The values ​​of multiple operating variables at different previous times indicate that there is a time delay in the impact of these variables on the blast furnace state.

[0013] X: The input data set consisting of the blast furnace state variables at the current moment and the operation variables at multiple previous moments, used for model training and prediction.

[0014] S2: Filter the blast furnace state parameters, and use the wavelet filtering method to filter the sequence data of the blast furnace state parameters to obtain the "change trend" sequence data of the target sequence;

[0015] S3: Perform differential processing on the “changing trend” sequence data to obtain the incremental data set Set_drift for each step;

[0016] S4: Construct the drift term model A, with the differential incremental data set Set_drift as the output and the input containing Sequential data from the dataset are used to learn and establish the relationship between the drift term and the operating variable using a nonlinear machine learning model.

[0017] S5: Construct diffusion term model B. First, use drift term model A to predict the input data, calculate the residual, and further perform differential processing on the residual to obtain the residual incremental data set Set_ diffusion ; Set_ diffusion For output, including The sequence data of is taken as input, and the relationship between the diffusion term and the operating variable is established through model B;

[0018] S6: Construct a random process model C, which will also include The sequence data is taken as input, combined with the drift term Set_drift and the diffusion term Set_ diffusion Calculate the random perturbation term dW at each step t , using the Euler-Maruyama method to simulate the random process; dW t As output, the random disturbance term in the blast furnace smelting process is further learned and simulated by model C.

[0019] S7: Based on SDE simulation of blast furnace state parameter sequence, the drift term, diffusion term and random disturbance term in the prediction results of models A, B and C are used, that is, f(X t ), g(X t ), dW t , construct the stochastic differential equation (SDE):

[0020] Y t+1 =Y t +f(X t )dt+g(X t )dW t

[0021] Where Y t is the state or variable value of the system at time t;

[0022] The Euler-Maruyama method is used for numerical solution, and the simulation error loss is calculated based on MSE (mean square error). SDE ;

[0023] S8: Build an optimization model. The optimization object is the total error of all parts of the world. The error calculation formula is Loss = loss A +loss B +loss C +loss SDE , the goal is to minimize the global error;

[0024] S9: Add optimization constraints. According to the production rules of blast furnace smelting, add major and minor constraints to ensure that the estimated results of operation delay meet the actual production needs. Major constraints limit the delay range of the charging system and the air supply system, and minor constraints classify the delay of variables through cluster analysis.

[0025] S10: Execute the optimization operation, and control the initial state and sampling point of each optimization by the random number determined by the timestamp through the optimization strategy.

[0026] S11: When the optimization is completed, the parameter set that minimizes the global loss is the optimal parameter set, from which the operation variable delay list is extracted, which is the lag time of the influence of the blast furnace ironmaking operation parameters on the blast furnace state parameters. According to the importance of the gain or SHAP output parameters, they are sorted to obtain the priority of parameter regulation, thereby providing a basis for blast furnace operation optimization.

[0027] Furthermore, the sequence data of blast furnace status parameters in step S1 include gas utilization rate, coke ratio, or molten iron composition data such as molten iron silicon content, molten iron manganese content, molten iron phosphorus content, molten iron sulfur content and other related status parameters in the smelting process.

[0028] Furthermore, the model A is an extreme gradient boosting tree model (Extreme Gradient Boosting, XGBoost), which is trained using a gradient-based weighted Boosting algorithm.

[0029] Furthermore, the model B is a lightweight gradient boosting machine model (Light Gradient Boosting Machine, LGBM), which is trained using a leaf node splitting strategy and a histogram bucketing technique.

[0030] Furthermore, the model C is an adaptive boosting model (Adaptive Boosting, AdaBoost), which is trained by an additive model and a forward step-by-step algorithm.

[0031] Furthermore, the model B is subjected to differential processing, as shown in the following formula:

[0032]

[0033] Tesiduals = x0-pred A *dt

[0034] Perform differential processing on the residuals to obtain the incremental data set Set_ for each step of the residuals diffusion .

[0035] Furthermore, in step S6, the drift term Set_drift and the diffusion term Set_ diffusion Calculate the random perturbation term dW at each step t , use the Euler-Maruyama method to simulate the random process, the formula is as follows:

[0036]

[0037] Where Y i+1 -Y i : The state change of the system from time i to i+1 includes the total effect of drift and diffusion.

[0038] Furthermore, the major constraints in step S9 include the time delay range constraints of the charging system and the air supply system, which are [4h, 8h] and [1h, 4h] respectively, and the minor constraints determine the operation parameter groups with similar time delays through hierarchical clustering analysis.

[0039] Furthermore, when the optimization operation is performed in step S10, it is performed through one of particle swarm optimization, genetic algorithm, Bayesian optimization or evolutionary algorithm, the random number is set by the timestamp of the current moment to ensure the difference of the random number, and the operation delay is optimized through a multi-step optimization strategy.

[0040] The present invention also discloses a system for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters. The system can be used to implement the above-mentioned method for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters, specifically comprising:

[0041] Data acquisition module: used to collect various operating parameters and blast furnace status parameters during blast furnace operation in real time.

[0042] Data preprocessing module: responsible for filtering and differential processing of the collected blast furnace status data: using the wavelet filtering method to smooth the blast furnace status parameters to obtain the change trend of the target sequence; performing differential processing on the change trend sequence to generate an incremental data set to provide a data basis for subsequent model training.

[0043] Drift term model building module: Based on the incremental data set after differentiation, a nonlinear machine learning model is used for training to learn the lag influence relationship between the operating variables and the blast furnace state parameters, thereby building a drift term model A. The input is the incremental data set after differentiation, and the output is the drift term.

[0044] Diffusion term model building module: Based on the output of drift term model A, the prediction residual is first calculated and the residual is differentiated. Then, the diffusion term model B is trained using a nonlinear machine learning algorithm to establish the relationship between the operating variable and the residual increment. The input is the residual data set and the operating variable lag data, and the output is the diffusion term.

[0045] Stochastic process model building module: The output of drift term model A and diffusion term model B is used to calculate the random disturbance term in combination with the Euler-Maruyama method. The random disturbance term in the blast furnace smelting process is further studied using the stochastic process model C. The input is the lag value of the operating variable, the drift term, and the diffusion term, and the output is the random disturbance term.

[0046] SDE simulation module: Based on drift terms, diffusion terms and random disturbance terms, the Euler-Maruyama method is used to numerically solve stochastic differential equations, simulate the change process of blast furnace state parameters, calculate the simulation error, and use the error for subsequent optimization operations.

[0047] Optimize model module: The goal of this module is to minimize the global error, and the optimization object is the total error of all global parts. Use the error calculation formula combined with the global minimization method to optimize the hyperparameters of Model A, Model B, and Model C and the lag time of the operating variables.

[0048] Constraint processing module: According to the production rules of blast furnace smelting, add major and minor constraints to ensure that the estimated results of operation delay meet the actual production needs. Major constraints limit the delay range of the charging system and the air supply system, and minor constraints classify the delay of variables through cluster analysis.

[0049] Optimization module: This module performs optimization operations. Through the optimization algorithm, it uses timestamp random numbers to control the initial state and sampling points of each optimization to obtain the optimal set of operating parameters.

[0050] Result output module: After the optimization is completed, the parameter set with the smallest global loss is output, from which the operation variable delay list is extracted to obtain the lag time of the influence of the blast furnace ironmaking operation parameters on the blast furnace state parameters. And according to the importance of the gain or SHAP output parameters, they are sorted to obtain the priority of parameter regulation, providing a basis for blast furnace operation optimization.

[0051] The present invention also discloses a computer device, including a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters is implemented.

[0052] The present invention also discloses a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters is implemented.

[0053] Compared with the prior art, the advantages of the present invention are:

[0054] The present invention can better handle the complex nonlinear relationship and strong coupling between the operating variables and the blast furnace state parameters in the blast furnace ironmaking process by introducing a nonlinear machine learning model. This modeling method is highly consistent with the actual working conditions of blast furnace smelting, thereby improving the accuracy of solving the operation delay and enhancing the scientificity and persuasiveness of the model results.

[0055] Based on the actual production rules of the blast furnace smelting process, the present invention adds optimization constraints to the optimization model, including major and minor constraints. The introduction of these constraints not only effectively reduces the optimization space and simplifies the optimization process, but also significantly reduces the complexity of the optimization calculation and improves the solution efficiency.

[0056] The present invention establishes a stochastic differential equation (SDE) based on the drift term, diffusion term and random process model constructed by the machine learning model, and uses the Euler-Maruyama method to numerically simulate the blast furnace state parameter sequence. Through the optimization algorithm, the operation delay can be accurately calculated, and the numerical simulation effect of SDE can evaluate the accuracy of the operation delay estimation, thus laying the foundation for optimizing the blast furnace smelting process.

[0057] Through the analysis of the importance of parameters by machine learning models, the present invention can provide the production site with the control priority of the operating variables. Sorting the importance of each operating variable according to the gain or SHAP (Shapley Additive Explanations) algorithm can provide a clear control basis for production management, helping the factory to more effectively adjust the operation strategy during the ironmaking process, thereby improving production efficiency and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 It is a flow chart of a method for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters according to an embodiment of the present invention.

[0059] Figure 2 It is a step-by-step operation flow chart of optimization in an embodiment of the present invention.

[0060] Figure 3 It is a group parameter constraint added by the embodiment of the present invention based on the clustering result.

[0061] Figure 4 This is a ranking diagram of the importance of various operating parameters in the embodiment of the present invention.

[0062] Figure 5 It is a simulation of the silicon content in molten iron based on SDE in an embodiment of the present invention. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and examples.

[0064] like Figure 1 As shown, the evaluation method of the lag time of the influence of blast furnace ironmaking operation parameters on state parameters comprises the following steps:

[0065] S1: Based on the operation variable delay list, construct the sequence data of the “variable to be estimated” at different times in the past. Operation delay list Γ = [τ1, τ2, ... τ i ,…τ n ], where τ i is the time delay temporarily set for each variable. At this time, a certain sequence data of blast furnace state parameters (silicon content of molten iron) x0 and the "variable to be estimated" [x1, x2, ... x i …x n The corresponding relationship of ] is:

[0066]

[0067] S2: The silicon content of molten iron in blast furnace is filtered to obtain the "change trend" sequence data of the target sequence. The filtering method generally adopts wavelet filtering.

[0068] S3: Perform differential processing on the "change trend" sequence data to obtain the incremental data set Set_drift for each step. The differential processing generally combines forward difference, central difference and backward difference.

[0069] S4: Construct the drift term model A, with Set_drift as output, including the time delay and no delay As input, this correspondence is learned through model A. Model A is a nonlinear machine learning model, and this embodiment adopts an extreme gradient boosting tree model.

[0070] S5: Construct diffusion term model B. First, transform the delayed and no delay As input, predictions are made based on model A, and the residuals are further calculated:

[0071]

[0072] residuals = x0 - pred A *dt

[0073] Perform differential processing on the residuals to obtain the incremental data set Set_ for each step of the residuals diffusion . With Set_ diffusion For output, including delay and no delay As input, this correspondence is learned through model B. Model B is also a nonlinear machine learning model, and this embodiment adopts an extreme gradient boosting tree model.

[0074] S6: Constructing the random process dW t , based on the sequence data of furnace temperature and the numerical solution of random SDE "Euler-Maruyama method":

[0075] Y t+1 =Y t +f(X t )dt+g(X t )dW t

[0076] dW per step t :

[0077]

[0078] Get Set_dW, with Set_dW as output, including delay and no delay As input, this correspondence is learned through model C. Model C is also a nonlinear machine learning model, and this embodiment adopts an extreme gradient boosting tree model.

[0079] S7: Simulate the silicon content sequence x0 of molten iron in blast furnace based on SDE. Models A, B, and C correspond to f(X t ), g(X t ), dW t , take the first value in x0 as the SDE simulation value Y sim The initial value Y sim [0], based on the “Euler-Maruyama method”, simulate the entire sequence to get Y sin , and calculate the simulation error loss based on MSE SDE .

[0080] S8: Build an optimization model, and the optimization object is the total error of all parts. The error mainly comes from nonlinear machine learning models A, B, C and SDE. The errors of the three machine learning models are also calculated by MSE, that is, loss A , loss B , loss C . Global total error:

[0081] Loss = loss A +loss B +loss C +loss SDE

[0082] The optimization goal is to minimize the global error.

[0083] S9: Add optimization constraints. The constraints are mainly composed of large-category constraints under the blast furnace smelting system and small-category constraints under cluster analysis. Large-category constraints: The change of the charging system on the smelting effect is often longer than the change of the air supply system. If the operating parameter belongs to the charging system, the time delay range of the parameter is usually [4h, 8h], that is, it must be longer than half a smelting cycle; if the operating parameter belongs to the air supply system, the time delay range of the parameter is usually [1h, 4h], that is, it does not exceed half a smelting cycle. Small-category constraints: Through hierarchical clustering analysis, the parameters are divided into several groups. Parameters belonging to the same group often have the same characteristics in production, and their time delays should be equal or similar. For example Figure 3 As shown, for example, the coal quantity and oxygen enrichment rate belong to the same cluster in the clustering results. However, in production, with the increase of coal injection amount, in order to achieve full combustion of the newly introduced coal powder, the oxygen enrichment rate also needs to be increased.

[0084] S10: Figure 2 As shown, the optimization operation is performed. The optimization is implemented in batches and multiple steps, and each batch is optimized in multiple steps. The random numbers of each batch are different, and the optimal target value and optimization parameters of the previous batch will be used as the starting point and starting parameters of the next optimization. The random number is set by the timestamp of the current moment to ensure the difference of the random number; the random number usually affects the initial state of the optimization, such as the initial sampling point or the initial state of the population; the optimization algorithm can be an optimization algorithm such as particle swarm optimization, genetic algorithm, Bayesian optimization and evolutionary algorithm.

[0085] S11: When the optimization is completed, the parameter set that minimizes the global loss is the optimal parameter set, from which the operation variable delay list Γ = [τ1, τ2, ... τ i ,…π n ] is the lag time of the influence of blast furnace ironmaking operation parameters on the silicon content of blast furnace hot metal. The best hyperparameters of models A, B, and C are extracted from the optimal parameter set, and a model corresponding to the training is constructed. And based on the importance of the gain or SHAP output parameters, the priority of parameter regulation can be obtained by sorting them. The simulation results of the silicon content of blast furnace hot metal based on the SDE established by models A, B, and C are used to evaluate the required operation delay.

[0086] Taking the blast furnace state parameter molten iron silicon content (w[Si]) as an example, the lag time of the influence of each operating parameter on the molten iron composition is obtained through this solution mode, as shown in Table 1. The input variables are ranked in order of importance, such as Figure 4 As shown in the figure, it can be seen from the sorting results that air volume and air temperature occupy a dominant position among the operating variables and should be regulated first in production. At the same time, the simulation results based on SDE show that the constructed SDE can better characterize the silicon content sequence of molten iron, and the simulation error is small, such as Figure 5 As shown, the manipulated variable delay results have a high degree of convincingness.

[0087] Table 1 Lag time of the effect of operating parameters on state parameters

[0088]

[0089] In another embodiment of the present invention, a system for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters is provided. The system can be used to implement the above-mentioned method for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters, specifically, including:

[0090] Data acquisition module: used to collect various operating parameters (such as air volume, air temperature, batch size, coke batch size, etc.) and blast furnace status parameter data (such as blast furnace gas utilization, coke ratio, and the proportion of main elements in molten iron such as silicon, manganese, and phosphorus) in real time during blast furnace operation. Data is collected through sensors, PLC control systems, SCADA systems, etc.

[0091] Data preprocessing module: This module is responsible for filtering and differential processing of the collected blast furnace state parameter data: using the wavelet filtering method to smooth the blast furnace state parameter sequence to obtain the change trend of the target sequence; performing differential processing on the change trend sequence to generate an incremental data set, providing a data basis for subsequent model training.

[0092] Drift term model building module (model A): This module uses a nonlinear machine learning model (such as decision tree, random forest, etc.) for training based on the incremental data set after differentiation to learn the lag influence relationship between the operating variables and the blast furnace state parameters, thereby building the drift term model A. The input is the incremental data set after differentiation (time series data of operating variables and state variables), and the output is the drift term (i.e., the influence of the operating variables on the increment of the blast furnace state parameters).

[0093] Diffusion term model building module (model B): Based on the output of drift term model A, the prediction residual is first calculated and the residual is differentially processed. Then, the diffusion term model B is trained using nonlinear machine learning algorithms (such as decision trees, random forests, etc.) to establish the relationship between the operating variables and the residual increments. The input is the residual data set and the operating variable lag data, and the output is the diffusion term (i.e., the influence of the operating variables on the random disturbance of the blast furnace state parameters).

[0094] Stochastic process model building module (model C): The output of drift term model A and diffusion term model B is used to calculate the random disturbance term in combination with the Euler-Maruyama method. Stochastic process model C is used to further study the random disturbance term in the blast furnace smelting process. The input is the lag value of the operating variable, the drift term, and the diffusion term, and the output is the random disturbance term (stochastic process simulation).

[0095] SDE simulation module: Based on drift, diffusion and random disturbance terms, the Euler-Maruyama method is used to numerically solve the stochastic differential equation (SDE), simulate the change process of blast furnace state parameters, and calculate the simulation error (loss SDE ), and use the error in subsequent optimization operations.

[0096] Optimize model module: The goal of this module is to minimize the global error, and the optimization object is the total error of all parts of the world. Use the error calculation formula and the global minimization method to optimize the hyperparameters of Model A, Model B, and Model C and the lag time of the operating variables.

[0097] Constraint processing module: According to the production rules of blast furnace smelting, add major and minor constraints to ensure that the estimated results of operation delay meet the actual production needs. Major constraints limit the delay range of the charging system and the air supply system, and minor constraints classify the delay of variables through cluster analysis.

[0098] Optimization module: This module performs optimization operations. Through the random strategy optimization algorithm, it uses timestamp random numbers to control the initial state and sampling points of each optimization to obtain the optimal set of operating parameters.

[0099] Result output module: After the optimization is completed, the parameter set with the smallest global loss (i.e. the optimal parameter set) is output, from which the operation variable delay list is extracted to obtain the lag time of the influence of the blast furnace ironmaking operation parameters on the blast furnace state parameters. And according to the importance of the gain or SHAP output parameters, they are sorted to obtain the priority of parameter regulation, providing a basis for blast furnace operation optimization.

[0100] In another embodiment of the present invention, a terminal device is provided, the terminal device includes a processor and a memory, the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions to implement the corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used to evaluate the operation of the method for evaluating the influence of blast furnace ironmaking operating parameters on state parameters on the lag time.

[0101] In another embodiment of the present invention, the present invention further provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a terminal device for storing programs and data. It can be understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and the extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory, or a non-volatile memory (non-volatile memory), such as at least one disk storage.

[0102] One or more instructions stored in a computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the method for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters in the above embodiment; one or more instructions in the computer-readable storage medium are loaded and executed by the processor.

[0103] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0104] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks 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 processes in the flowchart and / or block diagram. 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.

[0105] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate 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 A function specified in one or more boxes.

[0106] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0107] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the implementation methods of the present invention, and should be understood that the protection scope of the present invention is not limited to such special statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A method for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters, characterized in that: The following steps are involved: S1: Construct the time delay sequence data of the operation variable according to the delay list Γ = [τ1, τ2, ... τ i ,…π n ], where τ i The time delay temporarily set for each variable is used to construct the sequence data of the blast furnace state parameters As input data, and construct the corresponding relationship between the blast furnace state parameter x0 and the corresponding operation variable sequence: In the formula, The actual value of the blast furnace state parameters at the current moment; The values ​​of multiple operating variables at different previous moments indicate that there is a time delay in the impact of these variables on the blast furnace status; X: Input data set consisting of the blast furnace state variables at the current moment and the operation variables at multiple previous moments, used for model training and prediction; S2: Filter the blast furnace state parameters, and use the wavelet filtering method to filter the sequence data of the blast furnace state parameters to obtain the "change trend" sequence data of the target sequence; S3: Perform differential processing on the "change trend" sequence data to obtain the incremental data set Set_drift for each step; S4: Construct the drift term model A, with the differential incremental data set Set_drift as the output and the input containing Sequential data of the model are used to learn and establish the relationship between the drift term and the operating variable. S5: Construct diffusion term model B. First, use drift term model A to predict the input data, calculate the residual, and further perform differential processing on the residual to obtain the residual incremental data set Set_ diffusion ; Set_ diffusion For output, including The sequence data of is taken as input, and the relationship between the diffusion term and the operating variable is established through model B; S6: Construct a random process model C, which will also include The sequence data is taken as input, combined with the drift term Set_drift and the diffusion term Set_ diffusion Calculate the random perturbation term dW at each step t , using the Euler-Maruyama method to simulate the random process; dW t As output, the random disturbance term in the blast furnace smelting process is further learned and simulated through model C; S7: Based on SDE simulation of blast furnace state parameter sequence, the drift term, diffusion term and random disturbance term in the prediction results of models A, B and C are used, that is, f(X t ), g(X t ), dW t , construct the stochastic differential equation (SDE): Y t+1 =Y t +f(X t )dt+g(X t )dW t Where Y t is the state or variable value of the system at time t; The Euler-Maruyama method is used for numerical solution, and the simulation error loss is calculated based on MSE (mean square error). SDE ; S8: Build an optimization model. The optimization object is the total error of all parts of the world. The error calculation formula is Loss = loss A +loss B +loss C +loss SDE , the goal is to minimize the global error; S9: Add optimization constraints. According to the production rules of blast furnace smelting, add major and minor constraints to ensure that the estimated results of operation delay meet the actual production needs. Major constraints limit the delay range of the charging system and the air supply system, and minor constraints classify the delay of variables through cluster analysis. S10: Execute the optimization operation, and control the initial state and sampling point of each optimization by the random number determined by the timestamp through the optimization strategy; S11: When the optimization is completed, the parameter set that minimizes the global loss is the optimal parameter set, from which the operating variable delay list is extracted, that is, the lag time of the influence of the blast furnace ironmaking operating parameters on the blast furnace state parameters; according to the importance of the gain or SHAP output parameters, they are sorted to obtain the priority of parameter control, thereby providing a basis for the optimization of blast furnace operation.

2. The method for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters according to claim 1, characterized in that: The sequence data of blast furnace status parameters in step S1 include gas utilization rate, coke ratio, or molten iron composition data such as molten iron silicon content, molten iron manganese content, molten iron phosphorus content, molten iron sulfur content and other related status parameters in the smelting process.

3. The method for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters according to claim 1, characterized in that: The model A is an extreme gradient boosting tree model, which is trained using a gradient-based weighted Boosting algorithm.

4. The method for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters according to claim 1, characterized in that: The model B is a lightweight gradient boosting machine model, which is trained using a leaf node splitting strategy and a histogram bucketing technique.

5. The method for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters according to claim 1, characterized in that: The model C is an adaptive enhancement model, which is trained by an additive model and a forward step-by-step algorithm.

6. The method for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters according to claim 1, characterized in that: The model B is subjected to differential processing, as follows: residuals=x0-pred A *dt Perform differential processing on the residuals to obtain the incremental data set Set_ for each step of the residuals diffusion .

7. The method for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters according to claim 1, characterized in that: In step S6, the drift term Set_drift and the diffusion term Set_ diffusion Calculate the random perturbation term dW at each step t , use the Euler-Maruyama method to simulate the random process, the formula is as follows: Where Y i+1 -Y i : The state change of the system from time i to i+1 includes the total effect of drift and diffusion.

8. The method for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters according to claim 1, characterized in that: The major constraints in step S9 include the time delay range constraints of the charging system and the air supply system, which are [4h, 8h] and [1h, 4h] respectively. The minor constraints determine the operation parameter groups with similar time delays through hierarchical clustering analysis.

9. The method for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters according to claim 1, characterized in that: When performing the optimization operation in step S10, it is performed through one of particle swarm optimization, genetic algorithm, Bayesian optimization or evolutionary algorithm, the random number is set by the timestamp of the current moment to ensure the difference of the random number, and the operation delay is optimized through a multi-step optimization strategy.

10. A system for evaluating the lag time of the influence of blast furnace ironmaking operating parameters on state parameters, characterized in that: The system can be used to implement the method for evaluating the lag time of the influence of blast furnace ironmaking operation parameters on state parameters as described in any one of claims 1 to 9, specifically comprising: Data acquisition module: used to collect various operating parameters and blast furnace status parameters during blast furnace operation in real time; Data preprocessing module: responsible for filtering and differential processing of the collected blast furnace status data: using the wavelet filtering method to smooth the blast furnace status parameters and obtain the change trend of the target sequence; performing differential processing on the change trend sequence to generate incremental data sets, providing a data basis for subsequent model training; Drift term model building module: Based on the incremental data set after differentiation, a nonlinear machine learning model is used for training to learn the lag influence relationship between the operating variables and the blast furnace state parameters, thereby building a drift term model A; the input is the incremental data set after differentiation, and the output is the drift term; Diffusion term model building module: Based on the output of drift term model A, the prediction residual is first calculated and the residual is differentiated; then, the diffusion term model B is trained using a nonlinear machine learning algorithm to establish the relationship between the operating variable and the residual increment; the input is the residual data set and the operating variable lag data, and the output is the diffusion term; Stochastic process model building module: Use the output of drift term model A and diffusion term model B, combined with the Euler-Maruyama method to calculate the random disturbance term; use the random process model C to further study the random disturbance term in the blast furnace smelting process; the input is the lag value of the operating variable, the drift term, and the diffusion term, and the output is the random disturbance term; SDE simulation module: Based on drift terms, diffusion terms and random disturbance terms, the Euler-Maruyama method is used to numerically solve stochastic differential equations, simulate the change process of blast furnace state parameters, calculate simulation errors, and use the errors for subsequent optimization operations; Optimization model module: The goal of this module is to minimize the global error, and the optimization object is the total error of all global parts; the error calculation formula is used in combination with the global minimization method to optimize the hyperparameters of model A, model B, and model C and the lag time of the operating variables; Constraint processing module: According to the production rules of blast furnace smelting, add major and minor constraints to ensure that the estimated results of operation delay meet the actual production needs; major constraints are limited by the delay range of the charging system and the air supply system, and minor constraints are used to classify the delay of variables through cluster analysis; Optimization module: This module performs optimization operations. Through the optimization algorithm, it uses timestamp random numbers to control the initial state and sampling points of each optimization to obtain the optimal set of operating parameters. Result output module: After the optimization is completed, the parameter set with the smallest global loss is output, from which the operating variable delay list is extracted to obtain the lag time of the influence of blast furnace ironmaking operating parameters on blast furnace state parameters; and according to the importance of gain or SHAP output parameters, they are sorted to obtain the priority of parameter regulation, providing a basis for blast furnace operation optimization.

Images