Shaft furnace production process parameter prediction method and device

Abstract

The invention provides a shaft furnace production process parameter prediction method and device, and the method comprises the steps: obtaining a current inlet parameter of a direct reduction shaft furnace, inputting the current inlet parameter into a shaft furnace body prediction model, and obtaining a predicted outlet parameter of the direct reduction shaft furnace, inputting the predicted outlet parameters into a reducing gas forward cycle calculation model to obtain predicted inlet and outlet parameters of each equipment node in the top gas recovery system, predicted inlet and outlet parameters of each equipment node in the feed gas pretreatment system and predicted inlet and outlet parameters of each equipment node in the feed gas conversion system; taking the predicted outlet parameters and all the predicted inlet and outlet parameters as parameter prediction results of the shaft furnace production process; the prediction accuracy of the production process of the direct reduction shaft furnace and the control accuracy of the production process of the direct reduction shaft furnace are improved.

Description

Technical Field

[0001] The present application relates to the field of metallurgical technology, and in particular to a method and device for predicting parameters of a shaft furnace production process. Background Art

[0002] The environmental pressures of the long steelmaking process, typically represented by blast furnaces, are increasing dramatically. There is an urgent need for a new, more environmentally friendly, greener non-blast furnace ironmaking technology. This presents an opportunity for the shortened steelmaking process of "direct reduced iron and scrap steel + electric furnaces." Currently, gas-based shaft furnace direct reduction is the mainstream direct reduction process worldwide, accounting for ≥75% of sponge iron produced globally using this process. Gas-based shaft furnaces offer high production output per unit (typically 500,000 to 2.5 million tons / year), do not consume coking coal, and are energy-efficient, environmentally friendly, low in energy consumption, and low in CO2 emissions.

[0003] For the vertical furnace (or "black box"), which operates continuously under high-temperature, high-pressure, and sealed conditions, real-time understanding of the temperature and pressure fields within the furnace and prediction of the metallization rate of sponge iron are crucial for ensuring stable product quality, improving production efficiency, and increasing resource utilization. These insights are also key to achieving high efficiency, high quality, low energy consumption, and environmental protection. However, the working environment of the vertical furnace (or "black box") presents significant challenges in predicting its production process parameters, leading to inaccurate control of the "black box" process. Summary of the Invention

[0004] In view of the above-mentioned shortcomings of the prior art, the present application provides a method and device for predicting parameters of a shaft furnace production process to solve the above-mentioned technical problems.

[0005] According to one aspect of an embodiment of the present application, a method for predicting parameters of a shaft furnace production process is provided, comprising: obtaining current inlet parameters of a direct reduction shaft furnace; inputting the current inlet parameters into a shaft furnace body prediction model to obtain predicted outlet parameters of the direct reduction shaft furnace; the shaft furnace body prediction model is obtained by training a pre-established shaft furnace prediction model with first sample data; the first sample data is determined by historical inlet parameters of the direct reduction shaft furnace and a shaft furnace body simulation model; inputting the predicted outlet parameters into a reduction gas forward circulation calculation model to obtain predicted inlet and outlet parameters of each equipment node in a top gas recovery system, raw gas pre-treatment parameters, and the like. The predicted inlet and outlet parameters of each equipment node in the treatment system and the predicted inlet and outlet parameters of each equipment node in the raw gas conversion system; the reducing gas forward circulation calculation model is obtained by training the graph neural network model with the second sample data; the second sample data includes: the historical inlet parameters and historical outlet parameters of each equipment node in the top gas recovery system, the historical inlet parameters and historical outlet parameters of each equipment node in the raw gas pretreatment system, and the historical inlet parameters and historical outlet parameters of each equipment node in the raw gas conversion system; the predicted outlet parameters and all the predicted inlet and outlet parameters are used as the parameter prediction results of the vertical furnace production process.

[0006] In one embodiment of the present application, after obtaining the predicted outlet parameters, the method further includes: obtaining the calculated inlet and outlet parameters of each device node in the top gas recovery system, the calculated inlet and outlet parameters of each device node in the raw gas pretreatment system, and the calculated inlet and outlet parameters of each device node in the raw gas conversion system based on the predicted outlet parameters, the energy conservation equation, and the mass conservation equation; comparing the predicted outlet parameters of the direct reduction shaft furnace with the measured outlet parameters of the direct reduction shaft furnace to obtain a first comparison result; comparing the calculated inlet and outlet parameters of each device node in the top gas recovery system with the measured inlet and outlet parameters of the corresponding device node in the top gas recovery system to obtain a second comparison result; comparing the raw gas The calculated inlet and outlet parameters of each device node in the pretreatment system are compared with the measured inlet and outlet parameters of the corresponding device nodes in the raw gas pretreatment system to obtain a third comparison result; the calculated inlet and outlet parameters of each device node in the raw gas conversion system are compared with the calculated inlet and outlet parameters of the corresponding device nodes in the raw gas conversion system to obtain a fourth comparison result; based on the first comparison result, the second comparison result, the third comparison result and the fourth comparison result, the working status of the measuring device is determined; the working status includes: normal status and abnormal status; the measuring device is installed at the inlet and outlet of each node device; when the working status is an abnormal status, an alarm is issued and a prompt is given that the working status of the measuring device is abnormal.

[0007] In one embodiment of the present application, after obtaining the vertical furnace body prediction model, the method further includes: obtaining the export parameter constraints of the direct reduction vertical furnace; inputting the export parameter constraints into a multi-objective optimization evolutionary model to obtain multiple inlet parameter sets of the direct reduction vertical furnace; inputting each inlet parameter set into the vertical furnace body prediction model to obtain the export parameter set corresponding to each inlet parameter set, and inputting each inlet parameter set into a reducing gas reverse circulation calculation model to obtain the control inlet and outlet parameters of each device node in the top gas recovery system, the control inlet and outlet parameters of each device node in the raw gas pretreatment system, and the control inlet and outlet parameters of each device node in the raw gas conversion system; the reducing gas reverse circulation calculation model is obtained by training the graph neural network model with third sample data; the third sample data includes: the historical export parameters and historical inlet parameters of each device node in the top gas recovery system, the historical export parameters and historical inlet parameters of each device node in the raw gas pretreatment system, the historical export parameters and historical inlet parameters of each device node in the raw gas pretreatment system inlet parameters and historical inlet parameters and historical outlet parameters and historical inlet parameters of each equipment node in the raw gas conversion system; any inlet parameter set and the outlet parameter set and control inlet and outlet parameters calculated by the arbitrary inlet parameter set are taken as an optimization control scheme, and the operating cost of each optimization control scheme is calculated; the operating cost includes energy consumption cost and environmental protection cost; the energy consumption cost is determined based on the gas consumption and liquid consumption of each optimization control scheme; the environmental protection cost is determined based on the gas emission, electricity consumption and the output of pellets within a preset time unit of each optimization control scheme; according to the screening conditions of the operating cost, the target optimization control scheme is determined to control the vertical furnace production through the target optimization control scheme; the screening conditions include: the deviation between the preset operating cost and the operating cost of the target optimization control scheme is less than or equal to the preset deviation threshold, and the confidence of the target optimization control scheme is greater than or equal to the preset confidence threshold.

[0008] In one embodiment of the present application, the process of controlling the production of a vertical furnace by the target optimization control scheme includes: generating operating control parameters based on the target optimization control scheme; the operating control parameters include: a temperature control setting value of a heating furnace, a setting value of an oxygen-coal ratio of a converter, and a load value of a decarbonization device; and, in the process of controlling the production of the direct reduction vertical furnace, each device node in the top gas recovery system, each device node in the raw gas pretreatment system, and each device node in the raw gas conversion system according to the operating control parameters, measuring the inlet and outlet parameters of the direct reduction vertical furnace, the inlet and outlet parameters of each device node in the top gas recovery system, the inlet and outlet parameters of each device node in the raw gas pretreatment system, and the inlet and outlet parameters of the device nodes in the raw gas conversion system to obtain measurement parameters; and based on the inlet parameters of the direct reduction vertical furnace, The inlet and outlet parameters of each equipment node in the top gas recovery system, the inlet and outlet parameters of each equipment node in the raw gas pretreatment system, and the inlet and outlet parameters of the equipment nodes in the raw gas conversion system are predicted to obtain predicted parameters; the adjustment strategy of the operation control parameter is determined according to the deviation between the measured parameter and the predicted parameter; the adjustment strategy includes: proportional-integral-differential adjustment strategy and feedforward regulation strategy; based on the adjustment strategy, the operation control parameter is adjusted to obtain a parameter adjustment amount; and the sum of the operation control parameter and the parameter adjustment amount is used as the adjusted operation control parameter to control the direct reduction vertical furnace, each equipment node in the top gas recovery system, each equipment node in the raw gas pretreatment system, and the equipment node in the raw gas conversion system to produce according to the adjusted operation control parameters.

[0009] In one embodiment of the present application, before determining the first sample data through the historical inlet parameters of the direct reduction shaft furnace and the shaft furnace body simulation model, the method further includes: performing geometric modeling according to the size, position and direction of the direct reduction shaft furnace to obtain a physical model of the direct reduction shaft furnace; setting the inlet, outlet, wall and heat flux density of the physical model, and discretizing the physical model to obtain grid units of the physical model; configuring reaction equations of the grid units, and running the reaction equations to obtain the shaft furnace body simulation model, the reaction equations including a continuity equation, an energy equation and a momentum equation, the energy equation using the enthalpy of the reaction, the heat exchange of the ore and the heat exchange of the gas as energy source terms, and using the temperature of the grid unit as the solution result; the continuity equation using the rate of the chemical reaction as the energy source term, and using the gas composition of the grid unit as the solution result; the momentum equation using the gas inertial resistance, the iron ore inertial resistance, the gas viscous resistance and the iron ore viscous resistance as energy source terms, and using the gas flow rate of the grid unit as the solution result.

[0010] In one embodiment of the present application, after obtaining the vertical furnace simulation model, the method further includes: performing stratified sampling on the historical inlet parameters of the direct reduction vertical furnace according to the parameter design range of the direct reduction vertical furnace to obtain an inlet parameter sample; the historical inlet parameters include: reducing gas composition, temperature, flow rate, pressure and flow rate of pellets; the reducing gas composition includes hydrogen and carbon monoxide; the inlet parameter sample is input into the vertical furnace body simulation model to obtain an outlet parameter sample of the direct reduction vertical furnace; and the ratio of hydrogen to carbon monoxide, gas reduction rate, gas oxygen-coal ratio and ton iron gas volume are added as input characteristic parameters to the inlet parameter sample to obtain an expanded inlet parameter sample; and the expanded inlet parameter sample and the outlet parameter sample are used as the first sample data.

[0011] In one embodiment of the present application, the process of training a pre-established vertical furnace prediction model with the first sample data to obtain the vertical furnace body prediction model includes: normalizing the first sample data and dividing the normalized first sample data into training data and test data; training the pre-established vertical furnace prediction model with the training data to obtain the trained vertical furnace prediction model; the pre-established vertical furnace prediction model includes: an input preprocessing layer, a first residual unit layer, a transition layer, a second residual unit layer, a self-attention layer, a feature compression layer and a linear layer; the input preprocessing layer is used to perform a linear transformation on the training data to obtain linear transformation feature data, map the linear transformation feature data to a first preset dimensional feature space, and normalize, activate and apply a first random deactivation weight to the mapped linear transformation feature data; the first residual unit layer is used to calculate based on the first random deactivation weight and the normalized linear transformation feature data to obtain first residual data; the transition layer is used to map the first residual data to a second preset dimensional feature space, and normalize, activate and apply a second random deactivation weight to the mapped first residual data inactivation weight; the second residual unit layer is used to calculate based on the second random inactivation weight and the normalized first residual data to obtain second residual data; the self-attention layer is used to adjust the feature weights in the second residual data through the query matrix, the key matrix and the value matrix, and output the weighted sum of the second residual data; the feature compression layer is used to compress the weighted sum of the second residual data into a third preset dimensional feature space, and normalize, activate and apply the mapped weighted sum to the third random inactivation weight; the linear layer is used to perform a linear transformation on the mapped weight based on the third random inactivation weight to obtain prediction parameters; the trained shaft furnace prediction model is tested using the test data to obtain a test result; if the test result is greater than or equal to a preset test accuracy, the trained shaft furnace prediction model is used as the shaft furnace body prediction model; if the test result is less than the preset test accuracy, the parameters in the trained shaft furnace prediction model are adjusted until the test result output by the adjusted shaft furnace prediction model is greater than or equal to the preset test accuracy, and the adjusted shaft furnace prediction model is used as the shaft furnace body prediction model.

[0012] In one embodiment of the present application, after obtaining the vertical furnace body prediction model, the method further includes: if any inlet parameter of the direct vertical furnace body exceeds the corresponding parameter design range, the arbitrary inlet parameter, the inlet parameter in the same group as the arbitrary inlet parameter, and the newly added outlet parameter are added as a group of newly added sample data to the first sample data to obtain updated sample data; and the parameter interval of all parameters in the newly added sample data is used as the updated parameter interval; the newly added outlet parameter includes: the arbitrary inlet parameter and the inlet parameter in the same group as the arbitrary inlet parameter as the input parameter of the direct reduction vertical furnace body, and the outlet parameter output by the direct reduction vertical furnace body; counting the number of groups of the newly added sample data, and if the number of groups of the newly added sample data reaches a preset number threshold, the pre-established vertical furnace prediction model is trained by the updated sample data.

[0013] In one embodiment of the present application, the process of training the graph neural network model with the second sample data to obtain the reducing gas forward circulation calculation model includes: preprocessing the second sample data to obtain preprocessed feature data; the preprocessing operation includes: data cleaning, removal of outliers, null value marking, average value calculation at different time points, and normalization calculation; training the graph neural network model with the preprocessed feature data to obtain the reducing gas forward circulation calculation model; the graph neural network model includes: a feature masking layer, a two-order attention propagation layer, a dynamic regularization layer and a regression output layer; the feature masking layer is used to selectively mask the preprocessed feature data to obtain the masked features; the two-order attention propagation layer is used to screen and extract the masked features, and weighted aggregate the screened and extracted features, and redistribute the random inactivation probability in the aggregated features to correct or enhance the aggregated features through the redistributed attention weights, the dynamic regularization layer is used to regularize the corrected or enhanced aggregated features, and the regression output layer is used to map the regularized features to the parameter range of the inlet and outlet parameters.

[0014] According to one aspect of an embodiment of the present application, a device for predicting parameters of a vertical furnace production process is provided, comprising: a parameter acquisition module for acquiring current inlet parameters of a direct reduction vertical furnace; an outlet parameter output module for inputting the current inlet parameters into a vertical furnace body prediction model to obtain predicted outlet parameters of the direct reduction vertical furnace; the vertical furnace body prediction model is obtained by training a pre-established vertical furnace prediction model with first sample data; the first sample data is determined by historical inlet parameters of the direct reduction vertical furnace and a vertical furnace body simulation model; an inlet and outlet parameter output module for inputting the predicted outlet parameters into a reduction gas forward circulation calculation model to obtain predicted outlet parameters of each device node in a top gas recovery system. inlet parameters, predicted inlet and outlet parameters of each equipment node in the raw gas pretreatment system, and predicted inlet and outlet parameters of each equipment node in the raw gas conversion system; the reducing gas forward circulation calculation model is obtained by training the graph neural network model with the second sample data; the second sample data includes: the historical inlet parameters and historical outlet parameters of each equipment node in the top gas recovery system, the historical inlet parameters and historical outlet parameters of each equipment node in the raw gas pretreatment system, and the historical inlet parameters and historical outlet parameters of each equipment node in the raw gas conversion system; a prediction result output module is used to use the predicted outlet parameters and all predicted inlet and outlet parameters as the parameter prediction results of the vertical furnace production process.

[0015] The beneficial effects of the present application are as follows: the present application obtains the current inlet parameters of the direct reduction vertical furnace, inputs the current inlet parameters into the vertical furnace body prediction model, obtains the predicted outlet parameters of the direct reduction vertical furnace, inputs the predicted outlet parameters into the reduction gas forward circulation calculation model, obtains the predicted inlet and outlet parameters of each equipment node in the top gas recovery system, the predicted inlet and outlet parameters of each equipment node in the raw gas pretreatment system and the predicted inlet and outlet parameters of each equipment node in the raw gas conversion system, and uses the predicted outlet parameters and all the predicted inlet and outlet parameters as the parameter prediction results of the vertical furnace production process. In the above process, the vertical furnace body prediction model is obtained by training the pre-established vertical furnace prediction model with the historical inlet parameters of the direct reduction vertical furnace and the vertical furnace body simulation model, and the reduction gas forward circulation calculation model is obtained by training the pre-established vertical furnace prediction model with the historical inlet parameters of the direct reduction vertical furnace and the vertical furnace body simulation model, and the reduction gas forward circulation calculation model is obtained by training the pre-established vertical furnace prediction model with the historical inlet parameters of the equipment nodes in the top gas recovery system. The graph neural network model is trained with the historical inlet parameters and historical outlet parameters, the historical inlet parameters and historical outlet parameters of each equipment node in the raw gas pretreatment system, and the historical inlet parameters and historical outlet parameters of each equipment node in the raw gas conversion system. When the predicted outlet parameters of the direct reduction vertical furnace are predicted by the vertical furnace body prediction model and the predicted inlet and outlet parameters of each equipment node in the top gas recovery system, the predicted inlet and outlet parameters of each equipment node in the raw gas pretreatment system, and the predicted inlet and outlet parameters of each equipment node in the raw gas conversion system are predicted by the reducing gas forward circulation calculation model, the accuracy of the prediction of the direct reduction vertical furnace production process is improved; and based on the vertical furnace body prediction model and the reducing gas forward circulation calculation model, the target optimization control scheme is determined, thereby improving the accuracy of control of the direct reduction vertical furnace production process.

[0016] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings are incorporated into and constitute a part of the specification, illustrating embodiments consistent with the present application and, together with the specification, serving to explain the principles of the present application. It is obvious that the drawings described below are merely some embodiments of the present application, and a person of ordinary skill in the art can derive other drawings based on these drawings without inventive effort. In the drawings:

[0018] Figure 1 is a process flow chart of a vertical furnace production process shown in an exemplary embodiment of the present application;

[0019] Figure 2 is a schematic diagram of an exemplary system architecture shown in another exemplary embodiment of the present application;

[0020] Figure 3 is a flow chart of a method for predicting parameters of a shaft furnace production process, shown in an exemplary embodiment of the present application;

[0021] Figure 4 is a structural diagram of a pre-established shaft furnace prediction model shown in an exemplary embodiment of the present application;

[0022] Figure 5 is a flowchart illustrating an exemplary embodiment of the present application showing how the self-attention layer adjusts the feature weights in the second residual data;

[0023] Figure 6 is a system architecture diagram for predicting parameters of a shaft furnace production process, shown in an exemplary embodiment of the present application;

[0024] Figure 7 A block diagram of a shaft furnace production process parameter prediction device suitable for implementing an embodiment of the present application is shown;

[0025] Figure 8 A schematic diagram of the structure of a computer system suitable for implementing an electronic device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0026] Exemplary embodiments will be described in detail herein, examples of which are illustrated in the accompanying drawings. When the following description refers to the drawings, identical numerals in different figures represent identical or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0027] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically separate entities. That is, these functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0028] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, while others may be combined or partially combined. Therefore, the actual execution order may vary depending on the actual situation.

[0029] In this application, "plurality" refers to two or more. "And / or" describes the relationship between related objects, indicating that three possible relationships exist. For example, "A and / or B" can mean: A exists alone, A and B exist simultaneously, and B exists alone. The character " / " generally indicates that the related objects are in an "or" relationship.

[0030] The technical solutions of the embodiments of the present application involve metallurgy and other related technologies, which are specifically described through the following embodiments:

[0031] Figure 1 is a process flow chart of a shaft furnace production process shown in an exemplary embodiment of the present application, such as Figure 1 As shown, the process flow chart of the vertical furnace production process includes: top gas recovery system, raw gas pretreatment system, raw gas conversion system and vertical furnace main body system, among which, the top gas recovery system includes: dry bag, washing tower, compressor, heating furnace and decarbonization device, the raw gas pretreatment system includes: natural gas pressure reducing equipment, natural gas pressurizing equipment, natural gas preset equipment, the raw gas conversion system includes: conversion furnace, and the vertical furnace main body system includes: direct reduction vertical furnace.

[0032] In one embodiment of the present application, the process flow of the vertical furnace production process includes: (1) natural gas enters the natural gas pressure reducing device from the pipeline network and is pressure-reduced, then enters the natural gas pressure increasing device and pressurizes the pressure-reduced natural gas by spraying water and filling it with nitrogen, and then the pressurized natural gas sprayed with water and filled with nitrogen is sent to the converter; (2) the pressurized natural gas sprayed with water and filled with nitrogen is incompletely burned in the converter to generate reformed gas mainly composed of H2 and CO; (3) the reformed gas output from the converter is mixed with the circulating gas output from the heating furnace to obtain a mixed reducing gas; (4) the mixed reducing gas is introduced into the direct reduction vertical furnace through the tuyere, The direct reduction of iron ore is completed at 850-1050℃; (5) the high-temperature gas discharged from the top of the direct reduction vertical furnace is sequentially subjected to dry bag dust removal and washing in a washing tower to form purified gas; (6) a portion of the flow in the purified gas is supplied to the combustion chamber of the heating furnace as fuel gas, and a portion of the flow is pressurized by the top gas compressor to obtain increased gas; if there is additional surplus flow, it is released; (7) the pressurized gas is subjected to two-stage adsorption to remove CO2 in the decarbonization device; (8) the circulating gas after decarbonization is heated to 850-1050℃ in the heating furnace, and then mixed with the reforming gas output by the converter for a second time to form a closed loop.

[0033] In another embodiment of the present application, the chemical reaction of the natural gas after pressurized water injection and nitrogen filling in the converter includes:

[0034] CH4+0.5O2→CO+2H2 Formula (1)

[0035] 2H2+3O2→2H2O Formula (2)

[0036]

[0037] Figure 2 FIG. 1 is a schematic diagram of an exemplary system architecture shown in another exemplary embodiment of the present application.

[0038] Reference Figure 2 As shown, the system architecture may include a storage device 201 and a computer device 202. Computer device 202 may be at least one of a desktop graphics processing unit (GPU) computer, a GPU computing cluster, a neural network computer, and the like. Relevant technicians can use computer device 202 to obtain current inlet parameters of the direct reduction shaft furnace, input the current inlet parameters into a shaft furnace prediction model, obtain predicted outlet parameters of the direct reduction shaft furnace, input the predicted outlet parameters into a reduction gas forward circulation calculation model, and obtain predicted inlet and outlet parameters for each device node in the top gas recovery system, each device node in the raw gas pretreatment system, and each device node in the raw gas conversion system. The predicted outlet parameters and all predicted inlet and outlet parameters are used as parameter prediction results for the shaft furnace production process. Storage device 201 is used to store the current inlet parameters of the direct reduction shaft furnace. In this embodiment, storage device 201 uses random access memory (RAM) or the like to store the current inlet parameters of the direct reduction shaft furnace and provide them to computer device 202 for processing.

[0039] Schematically, after the computer device 202 obtains the current inlet parameters of the direct reduction shaft furnace from the storage device 201, the current inlet parameters are input into the shaft furnace body prediction model to obtain the predicted outlet parameters of the direct reduction shaft furnace, and the predicted outlet parameters are input into the reduction gas forward circulation calculation model to obtain the predicted inlet and outlet parameters of each device node in the top gas recovery system, the predicted inlet and outlet parameters of each device node in the raw gas pretreatment system, and the predicted inlet and outlet parameters of each device node in the raw gas conversion system, and the predicted outlet parameters and all the predicted inlet and outlet parameters are used as the parameter prediction results of the shaft furnace production process. In the above process, the shaft furnace body prediction model is obtained by training the pre-established shaft furnace prediction model with the historical inlet parameters of the direct reduction shaft furnace and the shaft furnace body simulation model, and the reduction gas forward circulation calculation ... The graph neural network model is trained based on the historical inlet parameters and historical outlet parameters of the points, the historical inlet parameters and historical outlet parameters of each equipment node in the raw gas pretreatment system, and the historical inlet parameters and historical outlet parameters of each equipment node in the raw gas conversion system. When the predicted outlet parameters of the direct reduction vertical furnace are predicted by the vertical furnace body prediction model and the predicted inlet and outlet parameters of each equipment node in the top gas recovery system, the predicted inlet and outlet parameters of each equipment node in the raw gas pretreatment system, and the predicted inlet and outlet parameters of each equipment node in the raw gas conversion system are predicted by the reducing gas forward circulation calculation model, the accuracy of the prediction of the direct reduction vertical furnace production process is improved; and based on the vertical furnace body prediction model and the reducing gas forward circulation calculation model, the target optimization control scheme is determined, thereby improving the accuracy of control of the direct reduction vertical furnace production process.

[0040] It should be noted that the shaft furnace production process parameter prediction method provided in the embodiment of the present application is generally executed by the computer device 202 , and accordingly, the shaft furnace production process parameter prediction device is generally provided in the computer device 202 .

[0041] The following is a detailed description of the implementation details of the technical solution of the embodiment of the present application:

[0042] Figure 3 This is a flow chart of a method for predicting parameters of a shaft furnace production process according to an exemplary embodiment of the present invention. The method for predicting parameters of a shaft furnace production process can be executed by a computing and processing device. The computing and processing device can be Figure 1 The computer device 202 shown in FIG. Figure 3 As shown, the method for predicting parameters of the vertical furnace production process includes at least steps S310 to S340, which are described in detail as follows:

[0043] In step S310, current inlet parameters of the direct reduction shaft furnace are obtained. In one embodiment of the present application, the current inlet parameters of the direct reduction shaft furnace include: inlet flow rate, inlet temperature, and inlet pressure of the direct reduction shaft furnace. The current inlet parameters of the direct reduction shaft furnace can be measured using a flow meter, thermocouple, pressure gauge, etc. installed at the inlet of the direct reduction shaft furnace, or calculated based on the energy conservation equation and the mass conservation equation.

[0044] In step S320, the current inlet parameters are input into the shaft furnace prediction model to obtain predicted outlet parameters for the direct reduction shaft furnace. In one embodiment of the present application, the shaft furnace prediction model is obtained by training a pre-established shaft furnace prediction model using first sample data; the first sample data is determined by historical inlet parameters of the direct reduction shaft furnace and a shaft furnace simulation model. The pre-established vertical furnace prediction model includes: an input preprocessing layer, a first residual unit layer, a transition layer, a second residual unit layer, a self-attention layer, a feature compression layer and a linear layer. The first residual unit layer is used to calculate based on the first random inactivation weight and the normalized linear transformation feature data to obtain the first residual data; the transition layer is used to map the first residual data to the second preset dimensional feature space, and normalize, activate and apply the second random inactivation weight to the mapped first residual data; the second residual unit layer is used to calculate based on the second random inactivation weight and the normalized first residual data to obtain the second residual data; the self-attention layer is used to adjust the feature weights in the second residual data through the query matrix, the key matrix and the value matrix, and output the weighted sum of the second residual data; the feature compression layer is used to compress the weighted sum of the second residual data to the third preset dimensional feature space, and normalize, activate and apply the third random inactivation weight to the mapped weighted sum; the linear layer is used to perform a linear transformation on the mapped weight based on the third random inactivation weight to obtain the prediction parameters.

[0045] In step S330, the predicted outlet parameters are input into the reducing gas forward circulation calculation model to obtain the predicted inlet and outlet parameters of each device node in the top gas recovery system, the predicted inlet and outlet parameters of each device node in the raw gas pretreatment system, and the predicted inlet and outlet parameters of each device node in the raw gas conversion system. In one embodiment of the present application, the reducing gas forward circulation calculation model is obtained by training a graph neural network model using second sample data; the second sample data includes: historical inlet parameters and historical outlet parameters of each device node in the top gas recovery system, historical inlet parameters and historical outlet parameters of each device node in the raw gas pretreatment system, and historical inlet parameters and historical outlet parameters of each device node in the raw gas conversion system. The graph neural network model includes: a feature masking layer, a two-order attention propagation layer, a dynamic regularization layer and a regression output layer; the feature masking layer is used to selectively mask the preprocessed feature data to obtain the masked features; the two-order attention propagation layer is used to screen and extract the masked features, and weightedly aggregate the screened and extracted features, and redistribute the random inactivation probability in the aggregated features to correct or enhance the aggregated features through the redistributed attention weights, the dynamic regularization layer is used to regularize the corrected or enhanced aggregated features, and the regression output layer is used to map the regularized features to the parameter range of the entry and exit parameters.

[0046] In step S340, the predicted outlet parameters and all predicted inlet and outlet parameters are used as parameter prediction results of the vertical furnace production process. In one embodiment of the present application, after obtaining the current inlet parameters of the direct reduction shaft furnace, a shaft furnace body prediction model is obtained by training a pre-established shaft furnace prediction model using the historical inlet parameters of the direct reduction shaft furnace and a shaft furnace body simulation model, and a reducing gas forward circulation calculation model is obtained by training a graph neural network model using the historical inlet parameters and historical outlet parameters of each device node in the top gas recovery system, the historical inlet parameters and historical outlet parameters of each device node in the raw gas pretreatment system, and the historical inlet parameters and historical outlet parameters of each device node in the raw gas conversion system. When the predicted outlet parameters of the direct reduction shaft furnace are predicted by the shaft furnace body prediction model and the predicted inlet and outlet parameters of each device node in the top gas recovery system, the predicted inlet and outlet parameters of each device node in the raw gas pretreatment system, and the predicted inlet and outlet parameters of each device node in the raw gas conversion system are predicted by the reducing gas forward circulation calculation model, the accuracy of the prediction of the direct reduction shaft furnace production process is improved; and based on the shaft furnace body prediction model and the reducing gas forward circulation calculation model, a target optimization control scheme is determined, thereby improving the accuracy of control of the direct reduction shaft furnace production process.

[0047] In one embodiment of the present application, after obtaining the predicted outlet parameters, the method for predicting parameters of a shaft furnace production process further includes:

[0048] Based on the predicted outlet parameters, the energy conservation equation, and the mass conservation equation, the calculated inlet and outlet parameters of each device node in the top gas recovery system, the calculated inlet and outlet parameters of each device node in the raw gas pretreatment system, and the calculated inlet and outlet parameters of each device node in the raw gas conversion system are obtained. In one embodiment of the present application, the device nodes in the top gas recovery system include: a dry bag filter, a scrubber, a pressurizer, a heating furnace, and a decarbonization device; the device nodes in the raw gas pretreatment system include: a natural gas pressure reducing device, a natural gas pressurizing device, and a natural gas preset device; and the device nodes in the raw gas conversion system include: a converter.

[0049] In some embodiments of the present application, reference is made to Figure 1 As shown, taking the gas pressure parameter as an example, based on the predicted outlet parameters (i.e., the parameters of node 6), the energy conservation equation, and the mass conservation equation, the calculated outlet gas pressure of dry dust removal (i.e., the parameters of node 7) is obtained. The calculation formula for the calculated inlet gas pressure of dry dust removal is as follows:

[0050] P i =P j -P loss Formula (5)

[0051] Among them, P i represents the predicted outlet gas pressure of the direct reduction shaft furnace, P j Indicates the calculated outlet gas pressure of dry dust removal, P loss It indicates the resistance loss of the equipment between the outlet of the direct reduction shaft furnace and the outlet of the dry dust collector. The resistance loss is a constant value. If there is no equipment between the outlet of the direct reduction shaft furnace and the outlet of the dry dust collector, there is no resistance loss.

[0052] exist Figure 1In the embodiment, the calculated gas pressure value of node 8 is calculated based on the calculated gas pressure value of node 7 and formula (5), the calculated gas pressure value of node 9 is calculated based on the calculated gas pressure value of node 8 and formula (5), the calculated gas pressure value of node 10 is calculated based on the calculated gas pressure value of node 8 and formula (5), the calculated gas pressure value of node 12 is determined by the target output gas pressure of the compressor, the calculated gas pressure value of node 13 is calculated based on the calculated gas pressure value of node 12 and formula (5), the calculated gas pressure value of node 14 is calculated based on the calculated gas pressure value of node 13 and formula (5), or, based on the calculated gas pressure value of node 9 The calculated gas pressure value of node 0 is obtained by pressure gauge detection, the calculated gas pressure value of node 1 is determined based on the target output pressure value of the natural gas pressure reducing device, the calculated gas pressure value of node 2 is determined based on the target output pressure of the natural gas compressor, the calculated gas pressure value of node 3 is calculated based on the calculated gas pressure value of node 2 and formula (5), the calculated gas pressure value of node 4 is calculated based on the calculated gas pressure value of node 3 and formula (5), the calculated gas pressure value of node 5 is calculated based on the calculated gas pressure value of node 4 and formula (5), or, is calculated based on the calculated gas pressure value of node 14 and formula (5).

[0053] In some embodiments of the present application, for natural gas compression equipment and compressors, it is also necessary to calculate the saturated water vapor partial pressure of the natural gas compression equipment based on the gas outlet pressure value of the natural gas compression equipment, and it is also necessary to calculate the saturated water vapor partial pressure of the compressor based on the gas outlet pressure value of the compressor.

[0054] Taking natural gas compression equipment as an example, the calculation formula for the saturated water vapor partial pressure of natural gas compression equipment is as follows:

[0055]

[0056] Among them, P′ i Indicates the saturated water vapor partial pressure of natural gas compression equipment, P i Indicates the calculated outlet gas pressure value of the natural gas compression equipment.

[0057] In some embodiments of the present application, for a compressor and a natural gas compression device, the formula for calculating the percentage of outlet H2O is as follows:

[0058]

[0059] in, Indicates the percentage of H2O at the outlet of natural gas compression equipment, Pi Indicates the calculated outlet gas pressure value of the natural gas compression equipment, P' i Indicates the saturated water vapor partial pressure of natural gas compression equipment.

[0060] Taking the gas flow parameters as an example, based on the predicted outlet parameters, energy conservation equation and mass conservation equation, the calculated outlet gas flow of dry dust removal is obtained. The calculation formula for the calculated outlet gas flow of dry dust removal is as follows:

[0061] Q i =Q j +Q k Formula (8)

[0062] Among them, Q i represents the predicted outlet gas flow rate of the direct reduction shaft furnace, Q j Indicates the calculated outlet gas flow rate of dry dust removal, Q k Indicates other flow lost or gas flow diverted to other equipment.

[0063] exist Figure 1 In the equation (8), the calculated gas flow value of node 8 is calculated based on the calculated gas flow value of node 7 and formula (8), the calculated gas flow value of node 9 is calculated based on the product of the calculated gas flow value of node 8 and the diversion ratio of the diverter valve and formula (8), the calculated gas flow value of node 10 is calculated based on the product of the calculated gas flow value of node 8 and the diversion ratio of the diverter valve and formula (8), the calculated gas flow value of node 12 is calculated based on the calculated gas flow value of node 10 and formula (8), the calculated gas flow value of node 13 is calculated based on the calculated gas flow value of node 12 and formula (8), and the calculated gas flow value of node 14 is calculated based on the calculated gas flow value of node 13. The calculated gas flow value of node 0 is obtained by flow meter detection, the calculated gas flow value of node 1 is calculated based on the calculated gas flow value of node 0 and formula (8), the calculated gas flow value of node 2 is calculated based on the calculated gas flow value of node 1 and formula (8), the calculated gas flow value of node 3 is calculated based on the calculated gas flow value of node 2 and formula (8), the calculated gas flow value of node 4 is calculated based on the calculated gas flow value of node 3, the flow conversion rate and formula (1), and the calculated gas flow value of node 5 is calculated based on the calculated gas flow value of node 4, the calculated gas flow value of node 14 and formula (8). The calculated gas flow value of node 11 is the remaining flow value obtained by subtracting the calculated gas flow value of node 9 and the calculated gas flow value of node 10 from the calculated gas flow value of node 8.

[0064] In one embodiment of the present application, the calculation formula for calculating the gas flow value of node 14 is as follows:

[0065] Q 14 =Q 13 +Q9+Q k Formula (9)

[0066] Among them, Q 14 Indicates the calculated gas flow value of node 14, Q 13 represents the calculated gas flow value of node 13, Q9 represents the calculated gas flow value of node 9, Q k Indicates other flow lost or gas flow diverted to other equipment.

[0067] In one embodiment of the present application, the calculation formula for calculating the gas flow value of node 4 includes:

[0068] Q4=Q 3* a+Q k Formula (10)

[0069] Among them, Q4 represents the calculated gas flow value of node 4, Q3 represents the calculated gas flow value of node 3, a represents the flow conversion rate, Q k Represents other lost flow or gas flow diverted to other equipment. The flow conversion rate is obtained by pre-simulating and calculating the converter's conversion process. This process includes: pre-determining the converter's output parameters, using the gas flow measured at the converter's inlet as input, and inputting the oxygen flow rate as a boundary condition into the reforming reaction kinetics model. The oxygen flow rate is varied while simultaneously measuring the converter's output parameters in real time until the error between the converter's real-time output parameters and the pre-determined converter output parameters is less than 5%. The ratio of the output parameter flow value to the input parameter flow value is then used as the flow conversion rate.

[0070] Taking the gas component content parameters as an example, based on the predicted outlet parameters, energy conservation equation and mass conservation equation, the calculated outlet gas component flow rate of dry dust removal is obtained. The calculation formula for the calculated outlet gas component flow rate of dry dust removal is as follows:

[0071] Q j *V j,m =Q i *V i,m Formula (11)

[0072] Among them, Q i is the predicted outlet gas flow rate of the direct reduction shaft furnace, V i,mIt represents the percentage of gas component m in the predicted outlet gas flow of the direct reduction shaft furnace, Q j Indicates the calculated outlet gas flow rate of dry dust removal, V j,m It indicates the percentage of gas component m in the calculated outlet gas flow of dry dust removal, where m represents H2, CO, CO2, N2, CH4 or H2O, etc.

[0073] In one embodiment of the present application, within the allowable error range, the percentage content of the inlet and outlet gas flow of the direct reduction vertical furnace, the percentage content of the inlet and outlet gas flow of the dry bag, the percentage content of the inlet and outlet gas flow of the washing tower, the percentage content of the inlet and outlet gas flow of the compressor, the percentage content of the inlet and outlet gas flow of the heating furnace, the percentage content of the inlet and outlet gas flow of the natural gas pressure reducing equipment, and the percentage content of the inlet and outlet gas flow of the natural gas pressure reducing equipment are consistent.

[0074] exist Figure 1In the figure, each calculated gas component flow rate of node 8 is calculated based on the calculated gas flow value of node 8 and the percentage content of each gas component at node 8, each calculated gas component flow rate of node 9 is calculated based on the calculated gas flow value of node 9 and the percentage content of each gas component at node 9, each calculated gas component flow rate of node 10 is calculated based on the calculated gas flow value of node 10 and the percentage content of each gas component at node 10, each calculated gas component flow rate of node 12 is calculated based on the calculated gas flow value of node 12 and the percentage content of each gas component at node 12, each calculated gas component flow rate of node 13 is calculated based on the calculated gas flow value of node 13 and the percentage content of each gas component at node 13, each calculated gas component flow rate of node 14 is calculated based on the calculated gas flow value of node 13, the calculated gas flow value of node 9, the percentage content of each gas component at node 13, the percentage content of each gas component at node 13, the percentage content of each gas component at node 14, the percentage content of each gas component at node 14, the percentage content of each gas component at node 14, the percentage content of each gas component at node 14, the percentage content of each gas component at node 14, the percentage content of each gas component at node 14, the percentage content of each gas component at node 14, the percentage content of each gas component at node 14, the percentage content of each gas component at node 14, the percentage content of each gas component at node 14, the percentage content of each gas component at node The calculated gas component flow rates of node 0 are calculated based on the calculated gas flow value of node 0 and the percentage content of each gas component at node 0. The calculated gas component flow rates of node 1 are calculated based on the calculated gas flow value of node 1 and the percentage content of each gas component at node 1. The calculated gas component flow rates of node 2 are calculated based on the calculated gas flow value of node 2 and the percentage content of each gas component at node 2. The calculated gas component flow rates of node 3 are calculated based on the calculated gas flow value of node 3 and the percentage content of each gas component at node 3. The calculated gas component flow rates of node 4 are calculated based on the calculated gas flow value of node 4 and the percentage content of each gas component at node 4. The calculated gas component flow rates of node 5 are calculated based on the calculated gas flow value of node 5, the percentage content of each gas component at node 5, the calculated gas flow value of node 14, and the percentage content of each gas component at node 14.

[0075] In one embodiment of the present application, the component flow rate of H2 at node 13 is calculated based on the calculated gas flow rate value of node 13 and the percentage content of H2 at node 13, the component flow rate of N2 at node 13 is calculated based on the calculated gas flow rate value of node 13 and the percentage content of N2 at node 13, the component flow rate of H2O at node 13 is calculated based on the calculated gas flow rate value of node 13 and the percentage content of H2O at node 13, the component flow rate of CO at node 13 is based on the multiplication of the calculated gas flow rate value of node 13 and the percentage content of CO at node 13, multiplied by the CO removal efficiency, the component flow rate of CO2 at node 13 is based on the multiplication of the calculated gas flow rate value of node 13 and the percentage content of CO2 at node 13, multiplied by the CO2 removal efficiency, and the component flow rate of CH4 at node 13 is based on the multiplication of the calculated gas flow rate value of node 13 and the percentage content of CH4 at node 13, multiplied by the CH4 removal efficiency.

[0076] In some embodiments of the present application, taking H2O as an example, the calculated gas component flow rate of H2O at node 14 is as follows:

[0077]

[0078] in, represents the calculated gas component flow rate of H2O at node 14, Q 13 Indicates that node 13 calculates the gas flow value, Indicates the percentage of H2O at node 13, Q9 indicates the gas flow value calculated at node 9, Indicates the percentage of H2O at node 9.

[0079] In some embodiments of the present application, taking H2O as an example, the calculated gas component flow rate of H2O at node 5 is as follows:

[0080]

[0081] in, It represents the calculated gas component flow rate of H2O at node 5, Q4 represents the calculated gas flow value at node 4, Indicates the percentage of H2O at node 4, Q 14 Indicates that node 14 calculates the gas flow value, Indicates the percentage of H2O at node 14.

[0082] Taking the gas temperature parameter as an example, based on the predicted outlet parameters, energy conservation equation and mass conservation equation, the calculated outlet gas temperature of dry dust removal is obtained. The calculation formula for the calculated outlet gas temperature of dry dust removal is as follows:

[0083] C p *Q i *T i =C p *Q j *T j -T loss Formula (13)

[0084] Among them, C p represents the specific heat capacity of gas, Q i is the predicted outlet gas flow rate of the direct reduction shaft furnace, T i represents the predicted outlet gas temperature of the direct reduction shaft furnace, Q j Indicates the calculated outlet gas flow rate of dry dust removal, T j Indicates the gas temperature of dry dust removal, T loss The temperature loss between the outlet of the direct reduction shaft furnace and the outlet of the dry dust collector is calculated as follows:

[0085] T loss =L i *t Formula (14)

[0086] Among them, T loss It represents the temperature loss between the outlet of the direct reduction shaft furnace and the outlet of the dry dust removal, L i It represents the length of the pipeline between the outlet of the direct reduction shaft furnace and the outlet of the dry dust removal system, and t represents the temperature loss per unit length.

[0087] In one embodiment of the present application, the outlet of the direct reduction vertical furnace is connected to the inlet of the dry bag, the outlet of the dry bag is connected to the inlet of the washing tower, the outlet of the washing tower is connected to the inlet of the compressor and the inlet of the heating furnace, the outlet of the compressor is connected to the inlet of the decarbonization device, the outlet of the decarbonization device is connected to the inlet of the heating furnace, the outlet of the heating furnace is connected to the inlet of the direct reduction vertical furnace, the inlet of the natural gas pressure reducing device inputs natural gas, the outlet of the natural gas pressure reducing device is connected to the inlet of the natural gas pressure reducing device, the outlet of the natural gas pressure reducing device is connected to the inlet of the natural gas preheating device, the outlet of the natural gas preheating device is connected to the inlet of the converter, and the outlet of the converter is connected to the inlet of the direct reduction vertical furnace.

[0088] In some embodiments of the present application, after obtaining the calculated gas flow rate and calculated gas temperature of node 7, the calculated gas flow rate of node 8, and the temperature loss between node 7 and node 8, the calculated gas temperature of node 8 can be calculated by formulas (13) and (14). Similarly, the calculated gas temperatures of nodes 9, 10, 12, and 13 can be calculated. After obtaining the calculated gas flow rate and calculated gas temperature of node 0, the calculated gas flow rate of node 1, and the temperature loss between node 0 and node 1, the calculated gas temperature of node 8 can be calculated by formulas (13) and (14). Similarly, the calculated gas temperatures of nodes 2, 3, and 4 can be calculated.

[0089] In some embodiments of the present application, the calculation formula for the calculated gas temperature of the node 14 is as follows:

[0090] C p14 Q 14 T 14 =C p13 Q 13 T 13 +C p9 Q9T9 formula (15)

[0091] Among them, C p14 represents the specific heat capacity of the gas at node 14, Q 14 represents the calculated gas flow rate of node 14, T 14 represents the computational gas temperature at node 14, C p13 represents the specific heat capacity of the gas at point 13, Q 13 represents the calculated gas flow rate of node 13, T 13 represents the computational gas temperature at node 13, C p9 represents the specific heat capacity of the gas at node 9, Q9 represents the calculated gas flow rate at point 9, and T9 represents the calculated gas temperature at node 9.

[0092] In some embodiments of the present application, the calculation formula for the calculated gas temperature of node 5 is as follows:

[0093] C p5 Q5T5=C p14 Q 14 T 14 +C p4 Q4T4 formula (16)

[0094] Among them, C p5 represents the gas specific heat capacity of node 5, Q5 represents the calculated gas flow rate of node 5, T5 represents the calculated gas temperature of node 5, C p14 represents the specific heat capacity of the gas at point 14, Q 14 represents the calculated gas flow rate of node 14, T 14represents the computational gas temperature at node 14, C p4 represents the specific heat capacity of the gas at node 4, Q4 represents the calculated gas flow rate at node 4, and T4 represents the calculated gas temperature at node 4.

[0095] The predicted outlet parameters of the direct reduction vertical furnace are compared with the measured outlet parameters of the direct reduction vertical furnace to obtain a first comparison result; the calculated inlet and outlet parameters of each equipment node in the top gas recovery system are compared with the measured inlet and outlet parameters of the corresponding equipment nodes in the top gas recovery system to obtain a second comparison result; the calculated inlet and outlet parameters of each equipment node in the raw gas pretreatment system are compared with the measured inlet and outlet parameters of the corresponding equipment nodes in the raw gas pretreatment system to obtain a third comparison result; the calculated inlet and outlet parameters of each equipment node in the raw gas conversion system are compared with the calculated inlet and outlet parameters of the corresponding equipment nodes in the raw gas conversion system to obtain a fourth comparison result. In one embodiment of the present application, the first comparison result includes: the comparison is consistent and the comparison is inconsistent within the allowable error range (the comparison difference is greater than the allowable error range), the second comparison result includes: the comparison is consistent and the comparison is inconsistent (the comparison difference is greater than the allowable error range) within the allowable error range, the third comparison result includes: the comparison is consistent and the comparison is inconsistent (the comparison difference is greater than the allowable error range) within the allowable error range, and the fourth comparison result includes: the comparison is consistent and the comparison is inconsistent (the comparison difference is greater than the allowable error range) within the allowable error range.

[0096] The operating state of the measuring device is determined based on the first comparison result, the second comparison result, the third comparison result, and the fourth comparison result. In one embodiment of the present application, if the first comparison result, the second comparison result, the third comparison result, and the fourth comparison result are all consistent within the allowable error range, the operating state of the measuring device is normal; if the first comparison result, the second comparison result, the third comparison result, and the fourth comparison result are inconsistent, the operating state of the measuring device is determined to be abnormal. The operating state includes: normal state and abnormal state; the measuring device is installed at the entrance and exit of each node device. The measuring device includes: flow meter, thermocouple, pressure gauge, etc.

[0097] If the working state is abnormal, an alarm is issued and a prompt is given that the working state of the measuring device is abnormal. In one embodiment of the present application, if the first comparison result, the second comparison result, the third comparison result, and the fourth comparison result are inconsistent, then the specification indicates that the measuring device installed on a certain device node is abnormal and cannot work normally, and therefore an alarm prompt is required.

[0098] In one embodiment of the present application, after obtaining the shaft furnace body prediction model, the shaft furnace production process parameter prediction method further includes:

[0099] Obtaining the export parameter constraints of the direct reduction shaft furnace. In one embodiment of the present application, the export parameter constraints include: target sponge iron production, metallization rate (≥92%), carbon content (1.5-4.5%), and unit energy consumption threshold.

[0100] The export parameter constraints are input into a multi-objective optimization evolutionary model to obtain multiple inlet parameter sets for the direct reduction shaft furnace. In one embodiment of the present application, the multi-objective optimization evolutionary model is implemented using an improved Non-dominated Sorting Genetic Algorithm II (NSGA-II). The improved NSGA-II (Non-dominated Sorting Genetic Algorithm II) processing process includes: taking the export parameter constraints as input, setting the crossover probability to 0.6-0.8, and the mutation probability to 0.05-0.1, and outputting multiple inlet parameter sets for the direct reduction shaft furnace when termination conditions are met. The termination conditions include: the Pareto front change rate is less than 0.1% for 10 consecutive generations or the maximum number of iterations is 1000.

[0101] Each inlet parameter set is input into the vertical furnace body prediction model to obtain the outlet parameter set corresponding to each inlet parameter set, and each inlet parameter set is input into the reducing gas reverse circulation calculation model to obtain the control inlet and outlet parameters of each equipment node in the top gas recovery system, the control inlet and outlet parameters of each equipment node in the raw gas pretreatment system, and the control inlet and outlet parameters of each equipment node in the raw gas conversion system. In one embodiment of the present application, the reducing gas reverse circulation calculation model is obtained by training the graph neural network model with the third sample data; the third sample data includes: the historical outlet parameters and historical inlet parameters of each equipment node in the top gas recovery system, the historical outlet parameters and historical inlet parameters of each equipment node in the raw gas pretreatment system, and the historical outlet parameters and historical inlet parameters of each equipment node in the raw gas conversion system; the graph neural network model includes: a feature masking layer, a two-order attention propagation layer, a dynamic regularization layer and a regression output layer; the feature masking layer is used to selectively mask the preprocessed feature data to obtain the masked features; the two-order attention propagation layer is used to screen and extract the masked features, and weightedly aggregate the screened and extracted features, and redistribute the random inactivation probability in the aggregated features to correct or enhance the aggregated features through the redistributed attention weights, the dynamic regularization layer is used to regularize the corrected or enhanced aggregated features, and the regression output layer is used to map the regularized features to the parameter range of the inlet and outlet parameters.

[0102] Any set of inlet parameters, the set of outlet parameters calculated from any set of inlet parameters, and the control inlet and outlet parameters are taken as an optimized control scheme, and the operating cost of each optimized control scheme is calculated. In one embodiment of the present application, each optimized control scheme corresponds to a different operating cost, which includes energy consumption cost and environmental protection cost; the energy consumption cost is determined based on the gas consumption and liquid consumption of each optimized control scheme, and the calculation formula of the energy consumption cost is as follows:

[0103] C1=S p1 P product (Q 天然气 e1+Q H2 e2+Q O2 e3+Q 水 e4+Q 蒸汽 e5) Formula (17)

[0104] Among them, P product represents the daily output of sponge iron, Q 天然气 Indicates the consumption of natural gas, Q H2 Indicates the consumption of hydrogen, Indicates oxygen consumption, Q 水 Indicates water consumption, Q 蒸汽 Indicates steam consumption, S p1 represents the cost per ton of standard coal, e1 represents the standard coal conversion coefficient of natural gas consumption, e2 represents the standard coal conversion coefficient of hydrogen consumption, e3 represents the standard coal conversion coefficient of oxygen consumption, e4 represents the standard coal conversion coefficient of water consumption, e5 represents the standard coal conversion coefficient of steam consumption, and C1 represents the energy consumption cost.

[0105] The environmental cost is determined based on the gas emission, power consumption, and pellet output within a preset time unit for each optimized control solution. The calculation formula for the environmental cost (CO2 emission tax) is as follows:

[0106] C2=S p P product (Q9V 9,CO2 +Ec e +Q 11 V 11,CO2 ) Formula (18)

[0107] Among them, P product represents the daily output of sponge iron, Q9 represents the flow rate of top gas used as fuel gas (i.e. the flow rate of top gas flowing into the heating furnace inlet), V 9,CO2 Indicates the CO2 content in the top gas flow at the hot furnace inlet, Q 11 Indicates the flow rate of top gas release (i.e. the remaining flow rate of top gas flowing out of the scrubber outlet after supplying the heating furnace and compressor), V11,CO2 represents the CO2 content in the top gas release flow, E represents the electricity consumption per ton of sponge iron, C e represents the carbon dioxide emission factor, S p It represents the carbon emission price per ton of CO2, and C1 represents the environmental protection cost.

[0108] According to the screening conditions of the operating costs, the target optimization control scheme is determined so as to control the vertical furnace production through the target optimization control scheme. In one embodiment of the present application, the screening conditions include: the deviation between the preset operating cost and the operating cost of the target optimization control scheme is less than or equal to the preset deviation threshold, and the confidence of the target optimization control scheme is greater than or equal to the preset confidence threshold. The preset operating cost is set according to the actual situation, the preset deviation threshold is set according to the actual situation, for example, 5%, and the preset confidence threshold can be set according to the actual situation, for example, 9%. The confidence of the target optimization control scheme is consistent with the confidence of the input parameter set corresponding to the target optimization control scheme, and the confidence of the input parameter set is obtained through the output of the multi-objective optimization evolution model.

[0109] In one embodiment of the present application, the process of controlling shaft furnace production through a target optimization control scheme includes:

[0110] Based on the target optimization control scheme, operating control parameters are generated. In one embodiment of the present application, the process of generating operating control parameters based on the target optimization control scheme can be implemented by referring to the process of generating operating control parameters based on the target optimization control scheme in the related art. The operating control parameters include: the heating furnace temperature control set value, the converter furnace oxygen-coal ratio set value, and the decarburization device load value.

[0111] Furthermore, during the production process of controlling the direct reduction shaft furnace, each device node in the top gas recovery system, each device node in the raw gas pretreatment system, and each device node in the raw gas conversion system according to the operating control parameters, the inlet and outlet parameters of the direct reduction shaft furnace, the inlet and outlet parameters of each device node in the top gas recovery system, the inlet and outlet parameters of each device node in the raw gas pretreatment system, and the inlet and outlet parameters of the device nodes in the raw gas conversion system are measured to obtain measured parameters; and based on the inlet parameters of the direct reduction shaft furnace, the inlet and outlet parameters of each device node in the top gas recovery system, the inlet and outlet parameters of each device node in the raw gas pretreatment system, and the inlet and outlet parameters of the device nodes in the raw gas conversion system are predicted to obtain predicted parameters. In one embodiment of the present application, the inlet and outlet parameters of the direct reduction shaft furnace, the inlet and outlet parameters of each device node in the top gas recovery system, the inlet and outlet parameters of each device node in the raw gas pretreatment system, and the inlet and outlet parameters of the device nodes in the raw gas conversion system are measured by measuring equipment installed at each device node.

[0112] According to the deviation between the measured parameter and the predicted parameter, the adjustment strategy of the operation control parameter is determined. In one embodiment of the present application, the adjustment strategy of the operation control parameter is determined based on the comparison result between the deviation and the deviation threshold. For example, when the deviation is less than or equal to 50% of the deviation threshold, the adjustment strategy of the operation control parameter is determined to be a proportional-integral-differential adjustment strategy. When the deviation is greater than 50% of the deviation threshold and less than or equal to the deviation threshold, the adjustment strategy of the operation control parameter is determined to be a feedforward adjustment strategy. When the deviation is greater than the deviation threshold, emergency intervention (such as load reduction, switching of standby control parameters) is triggered, and a re-optimization process (i.e., re-determining the target optimization control scheme) is started. The adjustment strategies include: a proportional-integral-differential adjustment strategy and a feedforward adjustment strategy.

[0113] In one embodiment of the present application, after the deviation is obtained, the size of the deviation and the adjustment strategy may be determined by using a pre-established fuzzy set of temperature / pressure deviations.

[0114] Based on the adjustment strategy, the operating control parameters are adjusted to obtain parameter adjustment amounts. The sum of the operating control parameters and the parameter adjustment amounts is used as the adjusted operating control parameters to control the direct reduction shaft furnace, each device node in the top gas recovery system, each device node in the raw gas pretreatment system, and each device node in the raw gas conversion system to produce according to the adjusted operating control parameters. In one embodiment of the present application, the parameter adjustment amount is calculated using the following formula:

[0115]

[0116] Among them, u 补偿 Indicates the parameter adjustment amount, K p represents the proportional adjustment coefficient, e(t) represents the function of the operation control parameter changing with time, K i Indicates the integral adjustment coefficient, K d Denotes the differential adjustment coefficient, M 预测 represents the model prediction gain, Δ 前馈 Represents the feedforward deviation, which is a key concept in feedforward control. It is used to compensate for known disturbances or setpoint changes in advance in the control system to reduce the dynamic error of the system.

[0117] In one embodiment of the present application, before determining the first sample data using the historical inlet parameters of the direct reduction shaft furnace and the shaft furnace body simulation model, the method further includes:

[0118] Geometric modeling is performed based on the size, position, and orientation of the direct reduction shaft furnace to obtain a physical model of the direct reduction shaft furnace. In one embodiment of the present application, the bottom of the direct reduction shaft furnace is used as the origin, the upward direction is the y-axis, and the x-axis is established in a right-handed manner. The x and y coordinates of each contour point are obtained clockwise.

[0119] The inlet, outlet, wall, and heat flux of the physical model are set, and the physical model is discretized to obtain mesh units of the physical model. In one embodiment of the present application, the inlet of the direct reduction shaft furnace is set as a velocity inlet, the top is set as a pressure outlet, and the remaining parts are set as walls and heat flux. The contour formed by the physical model of the direct reduction shaft furnace is discretized to obtain multiple mesh units.

[0120] Configure the reaction equation of the grid unit and run the reaction equation to obtain the vertical furnace body simulation model. In one embodiment of the present application, the reaction equation includes a continuity equation, an energy equation, and a momentum equation. The energy equation uses the enthalpy of the reaction, the heat exchange of the ore, and the heat exchange of the gas as energy source terms, and the temperature of the grid unit as the solution result; the continuity equation uses the rate of the chemical reaction as the energy source term, and the gas composition of the grid unit as the solution result; the momentum equation uses the gas inertial resistance, the iron ore inertial resistance, the gas viscous resistance, and the iron ore viscous resistance as energy source terms, and the gas flow rate of the grid unit as the solution result. Taking the reaction equation as an example, the calculation formula of the reaction equation is as follows:

[0121]

[0122] Where x represents the axial coordinate (unit: m), r represents the radial coordinate (unit: m), and ρ i Indicates the density of the gas or solid phase, u i Indicates the axial velocity (unit: m / s), ψ i Indicates the variable to be solved, v i represents radial velocity (unit: m / s), i represents gas phase or solid phase, Represents the source term, which is set according to different solution variables.

[0123] In some embodiments of the present application, during the operation of the reaction equation, the reducing gas composition (CO, CO2, H2, H2O, CH4, N2), reducing gas flow rate, reducing gas temperature, reducing gas pressure, and pellet flow rate at the inlet of the direct reduction shaft furnace are used as boundary conditions for simulation to obtain indicators such as simulated gas composition, simulated flow rate, simulated temperature, simulated pressure, and simulated metallization rate at the furnace top (i.e., the outlet of the direct reduction shaft furnace). The simulated gas composition, simulated flow rate, simulated temperature, simulated pressure, and simulated metallization rate are compared with the gas composition, flow rate, temperature, pressure, and metallization rate of the field measured values. If the relative error of all indicators is within 5%, the requirements are met. If the number of operating conditions that meet the requirements reaches the preset number of operating conditions, the calibration of the shaft furnace body simulation model is completed.

[0124] In one embodiment of the present application, after obtaining the vertical furnace simulation model, the method further includes:

[0125] Stratified sampling of historical inlet parameters of the direct reduction shaft furnace is performed according to the parameter design range of the direct reduction shaft furnace to obtain inlet parameter samples. In one embodiment of the present application, the historical inlet parameters include: reducing gas composition, temperature, flow rate, pressure, and pellet flow rate; the reducing gas composition includes hydrogen and carbon monoxide. Stratified sampling of the historical inlet parameters of the direct reduction shaft furnace can employ Latin hypercube sampling, proportional stratified sampling (PSS), equal size stratified sampling (ESSS), or other methods. By performing stratified sampling on the historical inlet parameters of the direct reduction shaft furnace, 1,000 to 5,000 sets of inlet parameter samples are obtained, thereby ensuring uniform coverage and non-overlapping of the multidimensional parameter space.

[0126] The inlet parameter samples are input into the vertical furnace simulation model to obtain the outlet parameter samples of the direct reduction vertical furnace; and the ratio of hydrogen to carbon monoxide, the gas reduction rate, the gas-oxygen-coal ratio, and the gas per ton of iron are added as input characteristic parameters to the inlet parameter samples to obtain an expanded inlet parameter sample. In one embodiment of the present application, the gas reduction rate is the ratio of the sum of carbon dioxide and water vapor in the gas at the furnace top (i.e., the outlet of the direct reduction vertical furnace) to the sum of carbon monoxide, carbon dioxide, hydrogen, and water vapor, the gas-oxygen-coal ratio is the ratio of the natural gas flow rate at the converter inlet to the oxygen flow rate at the inlet, and the gas per ton of iron is the ratio of the gas flow rate at the inlet of the direct reduction vertical furnace to the ore flow rate. Before inputting the inlet parameter samples into the vertical furnace simulation model, the inlet parameter samples need to be normalized and mapped to the interval [0,1], and abnormal inlet parameter samples that exceed the safe operation limit of the equipment are eliminated from the inlet parameter samples, thereby improving the accuracy of the outlet parameter samples.

[0127] The expanded inlet parameter sample and outlet parameter sample are used as the first sample data. In one embodiment of the present application, the ratio of hydrogen to carbon monoxide, the coal gas reduction rate, the coal gas oxygen-coal ratio, and the amount of gas per ton of iron are added as input feature parameters to the inlet parameter sample, thereby enhancing the features of the first sample data using process experience.

[0128] In one embodiment of the present application, a process of training a pre-established shaft furnace prediction model using first sample data to obtain a shaft furnace body prediction model includes:

[0129] The first sample data is normalized, and the normalized first sample data is divided into training data and test data. In one embodiment of the present application, before the first sample data is normalized, it is necessary to perform null value marking, outlier sliding window filtering, timestamp alignment, etc. on the first sample data, wherein the process of performing outlier sliding window filtering on the first sample data includes: determining the size and step size of the sliding window, for example, setting the sliding window size to 100 samples and the step size to 1; calculating the data statistics in each sliding window, for example, the median, the interquartile range, and the variance; and determining whether the data in the sliding window is an outlier based on the comparison result of the data in the sliding window with the set outlier threshold (for example, the median ± 1.5 * interquartile range, 3 times the variance). The process of aligning the timestamps of the first sample data includes: converting the timestamps of different type parameters in the first sample data into a unified format, unifying the time zone to determine the target timestamp sequence, for example, selecting the timestamp of one type parameter as a benchmark, or generating a uniform sequence covering all time points, resampling each type parameter, and aligning the type parameter to the target timestamp using an interpolation method (such as linear interpolation). All type parameters are merged into a single dataset to ensure that each timestamp has values ​​for all parameters.

[0130] The pre-established vertical furnace prediction model is trained by training data to obtain a trained vertical furnace prediction model. In one embodiment of the present application, the training data includes export parameter training data and import parameter training data. The pre-established vertical furnace prediction model includes: an input preprocessing layer, a first residual unit layer, a transition layer, a second residual unit layer, a self-attention layer, a feature compression layer and a linear layer; the input preprocessing layer is used to perform a linear transformation on the training data to obtain linear transformation feature data, map the linear transformation feature data to a first preset dimensional feature space, and normalize, activate and apply a first random deactivation weight to the mapped linear transformation feature data; the first residual unit layer is used to calculate based on the first random deactivation weight and the normalized linear transformation feature data to obtain the first residual data; the transition layer is used to map the first residual data to a second preset dimensional feature space, and normalize, activate and apply a second random deactivation weight to the mapped first residual data; the second residual The unit layer is used to calculate based on the second random inactivation weight and the normalized first residual data to obtain the second residual data; the self-attention layer is used to adjust the feature weights in the second residual data through the query matrix, key matrix and value matrix, and output the weighted sum of the second residual data; the feature compression layer is used to compress the weighted sum of the second residual data into a third preset dimensional feature space, and normalize, activate and apply the mapped weighted sum to the third random inactivation weight; the linear layer is used to perform a linear transformation on the mapped weight based on the third random inactivation weight to obtain the prediction parameter. The first preset dimensional feature space is a 512-dimensional feature space, the second preset dimensional feature space is a 256-dimensional feature space, and the third preset dimensional feature space is a 128-dimensional or 64-dimensional feature space. The first random inactivation weight, the second random inactivation weight, and the third random inactivation weight can be the same data or different data. The input preprocessing layer, transition layer, feature compression layer and linear layer are all composed of a linear transformation layer, a batch normalization layer, a Swish activation layer (i.e., Sigmoid-Weighted Linear Unit) and a random dropout layer.

[0131] The trained shaft furnace prediction model is tested using test data to obtain test results. In one embodiment of the present application, the test data includes inlet parameter test data and outlet parameter test data. The test results are used to characterize the accuracy of the outlet parameter prediction.

[0132] If the test result is greater than or equal to the preset test accuracy, the trained shaft furnace prediction model is used as the shaft furnace body prediction model. In one embodiment of the present application, the preset test accuracy can be set according to actual conditions. If the test result is greater than or equal to the preset test accuracy, it means that the trained shaft furnace prediction model meets the accuracy requirement.

[0133] If the test result is less than the preset test accuracy, the parameters of the trained shaft furnace prediction model are adjusted until the test result output by the adjusted shaft furnace prediction model is greater than or equal to the preset test accuracy, and the adjusted shaft furnace prediction model is used as the shaft furnace body prediction model. In one embodiment of the present application, if the test result is less than the preset test accuracy, it indicates that the accuracy of the trained shaft furnace prediction model has not met the required accuracy, and further adjustment of the parameters of the trained shaft furnace prediction model is required until the trained shaft furnace prediction model meets the required accuracy.

[0134] In one embodiment of the present application, after obtaining the shaft furnace body prediction model, the shaft furnace production process parameter prediction method further includes:

[0135] If any inlet parameter of the direct reduction shaft furnace exceeds the corresponding parameter design range, the arbitrary inlet parameter, the inlet parameters in the same group as the arbitrary inlet parameter, and the newly added outlet parameter are added to the first sample data as a set of newly added sample data to obtain updated sample data; and the parameter range of all parameters in the newly added sample data is used as the updated parameter range. In one embodiment of the present application, the newly added outlet parameter includes: using the arbitrary inlet parameter and the inlet parameters in the same group as the arbitrary inlet parameter as the input parameters of the direct reduction shaft furnace body, and using the outlet parameter output by the direct reduction shaft furnace body. Any inlet parameter may be reducing gas temperature, reducing gas pressure, reducing gas flow rate, direct reduction shaft furnace top pressure, ore flow rate, reducing gas composition (CO, CO2, H2, H2O, CH4, N2), etc., wherein the parameter design range of the reducing gas temperature is 700-1050°C, the parameter design range of the reducing gas pressure is 300-500 kPa, the parameter design range of the reducing gas flow rate is 0.3-0.5 kg / s, the parameter design range of the direct reduction shaft furnace top pressure is 290-490 kPa, the parameter design range of the ore flow rate is 0.45-0.5, the CO content in the reducing gas is 15-30%, the CO2 content in the reducing gas is 1-4%, the H2 content in the reducing gas is 40-70%, the H2O content in the reducing gas is 2-6%, the CH4 content in the reducing gas is 2-4.5%, and the N2 mole fraction in the reducing gas is 2-10%.

[0136] The number of newly added sample data groups is counted. If the number of newly added sample data groups reaches a preset threshold, the pre-established shaft furnace prediction model is trained by updating the sample data. In one embodiment of the present application, the preset threshold is set based on actual conditions, for example, 200 groups. Training the pre-established shaft furnace prediction model with updated sample data can adaptively update the pre-established shaft furnace prediction model, thereby improving the accuracy of the trained shaft furnace prediction model.

[0137] In one embodiment of the present application, the reducing gas pressure, reducing gas flow rate, and ore flow rate are determined according to the model, working environment, etc. of the direct reduction shaft furnace equipment. If the model, working environment, etc. of the direct reduction shaft furnace equipment changes, the reducing gas pressure, reducing gas flow rate, and ore flow rate will also change.

[0138] In one embodiment of the present application, the process of training the graph neural network model using the second sample data to obtain the reduction gas forward circulation calculation model includes:

[0139] The second sample data is preprocessed to obtain preprocessed feature data. In one embodiment of the present application, the preprocessing operation includes: data cleaning, outlier removal, null value marking, average value calculation at different time points, and normalization calculation. By preprocessing the second sample data, the accuracy of the preprocessed feature data is improved.

[0140] A graph neural network model is trained using preprocessed feature data to obtain a reduced gas forward circulation calculation model. In one embodiment of the present application, the graph neural network model includes: a feature masking layer, a two-order attention propagation layer, a dynamic regularization layer, and a regression output layer; the feature masking layer is used to selectively mask the preprocessed feature data to obtain masked features; the two-order attention propagation layer is used to screen and extract the masked features, and then weighted aggregate the screened and extracted features, redistributing the random inactivation probability in the aggregated features to correct or enhance the aggregated features using the redistributed attention weights; the dynamic regularization layer is used to regularize the corrected or enhanced aggregated features; and the regression output layer is used to map the regularized features to parameter ranges of inlet and outlet parameters. After obtaining the reduced gas forward circulation calculation model, the predicted outlet parameters are input into the reduced gas forward circulation calculation model to obtain the predicted inlet and outlet parameters of each device node in the top gas recovery system, the predicted inlet and outlet parameters of each device node in the raw gas pretreatment system, and the predicted inlet and outlet parameters of each device node in the raw gas conversion system.

[0141] Figure 4 is a structural diagram of a pre-established shaft furnace prediction model shown in an exemplary embodiment of the present application, such as Figure 4As shown, the pre-established vertical furnace prediction model includes: an input preprocessing layer, a first residual unit layer, a transition layer, a second residual unit layer, a self-attention layer, a feature compression layer and a linear layer; the input preprocessing layer is used to perform a linear transformation on the training data to obtain linear transformation feature data, map the linear transformation feature data to a first preset dimensional feature space, and normalize, activate and apply a first random inactivation weight to the mapped linear transformation feature data; the first residual unit layer is used to calculate based on the first random inactivation weight and the normalized linear transformation feature data to obtain first residual data; the transition layer is used to map the first residual data to a second preset dimensional feature space, and perform a random inactivation weight on the mapped linear transformation feature data. The mapped first residual data is normalized, activated and applied with a second random inactivation weight; the second residual unit layer is used to perform calculations based on the second random inactivation weight and the normalized first residual data to obtain the second residual data; the self-attention layer is used to adjust the feature weights in the second residual data through the query matrix, key matrix and value matrix, and output the weighted sum of the second residual data; the feature compression layer is used to compress the weighted sum of the second residual data to a third preset dimensional feature space, and normalize, activate and apply the third random inactivation weight to the mapped weighted sum; the linear layer is used to perform a linear transformation on the mapped weights based on the third random inactivation weight to obtain prediction parameters.

[0142] Figure 5 This is a flowchart of an exemplary embodiment of the present application showing that the self-attention layer adjusts the feature weights in the second residual data. Figure 5 As shown, the process of the self-attention layer adjusting the feature weights in the second residual data includes: inputting the second residual data; generating a query matrix, a key matrix and a value matrix; calculating the attention score using the query matrix, the key matrix and the value matrix; normalizing the attention score; and performing weighted feature fusion based on the normalized score to obtain the weighted sum of the second residual data.

[0143] Figure 6 This is a system architecture diagram for predicting parameters of a shaft furnace production process, as shown in an exemplary embodiment of the present application. Figure 6In the present invention, the system architecture for predicting parameters of the vertical furnace production process includes: in the production process of the direct reduction vertical furnace, collecting on-site production data, fusing the collected data from multiple sources, and storing the fused data as process node measurement data in the process node measurement database; inputting the collected input parameters of the direct reduction vertical furnace into the reduction gas forward circulation calculation model, predicting the predicted inlet and outlet parameters of each equipment node in the top gas recovery system, the predicted inlet and outlet parameters of each equipment node in the raw gas pretreatment system, and the predicted inlet and outlet parameters of each equipment node in the raw gas conversion system, and storing the predicted inlet and outlet parameters of each equipment node in the top gas recovery system, the predicted inlet and outlet parameters of each equipment node in the raw gas pretreatment system, and the predicted inlet and outlet parameters of each equipment node in the raw gas conversion system in the process node calculation database; comparing the measurement data of each process node with the corresponding predicted inlet and outlet parameters, and determining whether an early warning is needed based on the comparison result; inputting the current inlet parameters of the direct reduction vertical furnace into the vertical furnace An ontology prediction model is used to obtain the predicted outlet parameters of the direct reduction shaft furnace; the outlet parameter constraints of the direct reduction shaft furnace are obtained; the outlet parameter constraints are input into the multi-objective optimization evolution model to obtain multiple inlet parameter sets of the direct reduction shaft furnace; each inlet parameter set is input into the shaft furnace ontology prediction model to obtain the outlet parameter set corresponding to each inlet parameter set, and each inlet parameter set is input into the reducing gas reverse circulation calculation model to obtain the control inlet and outlet parameters of each equipment node in the top gas recovery system, the control inlet and outlet parameters of each equipment node in the raw gas pretreatment system, and the control inlet and outlet parameters of each equipment node in the raw gas conversion system; any inlet parameter set and the outlet parameter set and control inlet and outlet parameters calculated by any inlet parameter set are used as an optimization control scheme, and the operating cost of each optimization control scheme is calculated; according to the screening conditions of the operating cost, the target optimization control scheme is determined to control the shaft furnace production through the target optimization control scheme.

[0144] The following describes an apparatus embodiment of the present application, which can be used to implement the shaft furnace production process parameter prediction method described in the above-mentioned embodiment of the present application. For details not disclosed in the apparatus embodiment of the present application, please refer to the embodiment of the shaft furnace production process parameter prediction method described in the above-mentioned embodiment of the present application.

[0145] Figure 7 This is a block diagram of a vertical furnace production process parameter prediction device shown in an exemplary embodiment of the present application. The device can be applied to Figure 2 The implementation environment shown in FIG2 is specifically configured in the computer device 202. The apparatus may also be applicable to other exemplary implementation environments and specifically configured in other devices. This embodiment does not limit the implementation environment to which the apparatus is applicable.

[0146] like Figure 7 As shown, the exemplary vertical furnace production process parameter prediction device 700 includes:

[0147] The parameter acquisition module 701 is used to obtain the current inlet parameters of the direct reduction shaft furnace.

[0148] The outlet parameter output module 702 is used to input the current inlet parameters into the vertical furnace body prediction model to obtain the predicted outlet parameters of the direct reduction vertical furnace.

[0149] The inlet and outlet parameter output module 703 is used to input the predicted outlet parameters into the reducing gas forward circulation calculation model to obtain the predicted inlet and outlet parameters of each equipment node in the top gas recovery system, the predicted inlet and outlet parameters of each equipment node in the raw gas pretreatment system, and the predicted inlet and outlet parameters of each equipment node in the raw gas conversion system.

[0150] The prediction result output module 704 is used to use the predicted outlet parameters and all predicted inlet and outlet parameters as parameter prediction results of the vertical furnace production process.

[0151] In one embodiment of the present application, the current inlet parameters of the direct reduction shaft furnace include: the inlet flow rate of the direct reduction shaft furnace, the inlet temperature of the direct reduction shaft furnace, and the inlet pressure of the direct reduction shaft furnace. The current inlet parameters of the direct reduction shaft furnace can be measured by a flow meter, thermocouple, pressure gauge, etc. installed at the inlet of the direct reduction shaft furnace, or can be calculated based on the energy conservation equation and the mass conservation equation.

[0152] In one embodiment of the present application, the shaft furnace body prediction model is obtained by training a pre-established shaft furnace prediction model with first sample data; the first sample data is determined by historical inlet parameters of the direct reduction shaft furnace and the shaft furnace body simulation model. The pre-established vertical furnace prediction model includes: an input preprocessing layer, a first residual unit layer, a transition layer, a second residual unit layer, a self-attention layer, a feature compression layer and a linear layer. The first residual unit layer is used to calculate based on the first random inactivation weight and the normalized linear transformation feature data to obtain the first residual data; the transition layer is used to map the first residual data to the second preset dimensional feature space, and normalize, activate and apply the second random inactivation weight to the mapped first residual data; the second residual unit layer is used to calculate based on the second random inactivation weight and the normalized first residual data to obtain the second residual data; the self-attention layer is used to adjust the feature weights in the second residual data through the query matrix, the key matrix and the value matrix, and output the weighted sum of the second residual data; the feature compression layer is used to compress the weighted sum of the second residual data to the third preset dimensional feature space, and normalize, activate and apply the third random inactivation weight to the mapped weighted sum; the linear layer is used to perform a linear transformation on the mapped weight based on the third random inactivation weight to obtain the prediction parameters.

[0153] In one embodiment of the present application, a reduction gas forward circulation calculation model is obtained by training a graph neural network model with second sample data; the second sample data includes: historical inlet parameters and historical outlet parameters of each device node in the top gas recovery system, historical inlet parameters and historical outlet parameters of each device node in the raw gas pretreatment system, and historical inlet parameters and historical outlet parameters of each device node in the raw gas conversion system. The graph neural network model includes: a feature masking layer, a two-order attention propagation layer, a dynamic regularization layer, and a regression output layer; the feature masking layer is used to selectively mask the preprocessed feature data to obtain masked features; the two-order attention propagation layer is used to screen and extract the masked features, and weightedly aggregate the screened and extracted features, redistribute the random inactivation probability in the aggregated features, so as to correct or enhance the aggregated features through the redistributed attention weights, the dynamic regularization layer is used to regularize the corrected or enhanced aggregated features, and the regression output layer is used to map the regularized features to the parameter range of the inlet and outlet parameters.

[0154] In one embodiment of the present application, after obtaining the current inlet parameters of the direct reduction shaft furnace, a shaft furnace body prediction model is obtained by training a pre-established shaft furnace prediction model using the historical inlet parameters of the direct reduction shaft furnace and a shaft furnace body simulation model, and a reducing gas forward circulation calculation model is obtained by training a graph neural network model using the historical inlet parameters and historical outlet parameters of each device node in the top gas recovery system, the historical inlet parameters and historical outlet parameters of each device node in the raw gas pretreatment system, and the historical inlet parameters and historical outlet parameters of each device node in the raw gas conversion system. When the predicted outlet parameters of the direct reduction shaft furnace are predicted by the shaft furnace body prediction model and the predicted inlet and outlet parameters of each device node in the top gas recovery system, the predicted inlet and outlet parameters of each device node in the raw gas pretreatment system, and the predicted inlet and outlet parameters of each device node in the raw gas conversion system are predicted by the reducing gas forward circulation calculation model, the accuracy of the prediction of the direct reduction shaft furnace production process is improved; and based on the shaft furnace body prediction model and the reducing gas forward circulation calculation model, a target optimization control scheme is determined, thereby improving the accuracy of control of the direct reduction shaft furnace production process.

[0155] It should be noted that the shaft furnace production process parameter prediction device provided in the above-mentioned embodiment and the shaft furnace production process parameter prediction method provided in the above-mentioned embodiment are based on the same concept. The specific manner in which each module and unit performs operations has been described in detail in the method embodiment and will not be repeated here. In actual applications, the shaft furnace production process parameter prediction device provided in the above-mentioned embodiment can, as needed, allocate the above-mentioned functions to different functional modules, that is, divide the internal structure of the device into different functional modules to complete all or part of the functions described above, and this is not limited here.

[0156] An embodiment of the present application also provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by one or more processors, the electronic device implements the vertical furnace production process parameter prediction method provided in the above-mentioned embodiments.

[0157] Figure 8 The following is a schematic diagram showing the structure of a computer system suitable for implementing an electronic device according to an embodiment of the present application. Figure 8 The computer system 800 of the electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0158] like Figure 8 As shown, the computer system 800 includes a central processing unit (CPU) 801, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 802 or the program loaded from the storage part 808 into the random access memory (RAM) 803, such as executing the method in the above embodiment. Various programs and data required for system operation are also stored in the RAM 803. The CPU 801, ROM 802 and RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0159] The following components are connected to the I / O interface 805: an input section 806 including a keyboard, a mouse, and the like; an output section 807 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and a speaker; a storage section 808 including a hard disk and the like; and a communication section 809 including a network interface card such as a LAN (Local Area Network) card or a modem. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the I / O interface 805 as needed. Removable media 811, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 810 as needed, so that computer programs read therefrom can be installed into the storage section 808 as needed.

[0160] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 809, and / or installed from a removable medium 811. When the computer program is executed by the central processing unit (CPU) 801, the various functions defined in the system of the present application are executed.

[0161] It should be noted that the computer-readable medium shown in the embodiments of the present application can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, which carries a computer-readable computer program. This propagated data signal can take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. A computer program embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, or any suitable combination thereof.

[0162] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. Among them, each box in the flowchart or block diagram can represent a module, program segment, or part of the code, and the above-mentioned module, program segment, or part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0163] The units involved in the embodiments described in this application may be implemented by software or hardware, and the units described may also be set in a processor. In some cases, the names of these units do not constitute limitations on the units themselves.

[0164] Another aspect of the present application provides a computer-readable storage medium having computer-readable instructions stored thereon. When executed by a computer processor, the computer executes the shaft furnace production process parameter prediction methods provided in the aforementioned embodiments. The computer-readable storage medium may be included in the electronic device described in the aforementioned embodiments, or may exist independently and not be incorporated into the electronic device.

[0165] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiment of the application, the features and functions of two or more modules or units described above can be concretized in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be concretized.

[0166] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the embodiments of the present application.

[0167] Those skilled in the art will readily conceive of other embodiments of the present application after considering the specification and practicing the embodiments disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present application that follow the general principles of this application and include common knowledge or customary techniques in the art that are not disclosed herein.

[0168] It should be understood that the above content is only a preferred exemplary embodiment of the present application and is not intended to limit the implementation scheme of the present application. Ordinary technicians in this field can easily make corresponding changes or modifications based on the main concept and spirit of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection required by the claims.

Claims

1. A method for predicting parameters of a shaft furnace production process, characterized in that: include: Get the current inlet parameters of the direct reduction shaft furnace; Inputting the current inlet parameters into a shaft furnace body prediction model to obtain predicted outlet parameters of the direct reduction shaft furnace; the shaft furnace body prediction model is obtained by training a pre-established shaft furnace prediction model using the first sample data; The first sample data is determined by the historical inlet parameters of the direct reduction shaft furnace and a simulation model of the shaft furnace body; Inputting the predicted outlet parameters into the reducing gas forward circulation calculation model to obtain the predicted inlet and outlet parameters of each device node in the top gas recovery system, the predicted inlet and outlet parameters of each device node in the raw gas pretreatment system, and the predicted inlet and outlet parameters of each device node in the raw gas conversion system; The reducing gas forward circulation calculation model is obtained by training the graph neural network model with the second sample data; The second sample data includes: historical inlet parameters and historical outlet parameters of each device node in the top gas recovery system, historical inlet parameters and historical outlet parameters of each device node in the raw gas pretreatment system, and historical inlet parameters and historical outlet parameters of each device node in the raw gas conversion system; The predicted outlet parameter and all predicted inlet and outlet parameters are used as parameter prediction results of the shaft furnace production process.

2. The method for predicting parameters of a shaft furnace production process according to claim 1, characterized in that: After obtaining the predicted export parameters, the method further includes: Based on the predicted outlet parameters, the energy conservation equation, and the mass conservation equation, the calculated inlet and outlet parameters of each device node in the top gas recovery system, the calculated inlet and outlet parameters of each device node in the raw gas pretreatment system, and the calculated inlet and outlet parameters of each device node in the raw gas conversion system are obtained; Comparing the predicted outlet parameters of the direct reduction shaft furnace with the measured outlet parameters of the direct reduction shaft furnace to obtain a first comparison result; comparing the calculated inlet and outlet parameters of each device node in the top gas recovery system with the measured inlet and outlet parameters of the corresponding device node in the top gas recovery system to obtain a second comparison result; comparing the calculated inlet and outlet parameters of each device node in the raw gas pretreatment system with the measured inlet and outlet parameters of the corresponding device node in the raw gas pretreatment system to obtain a third comparison result; comparing the calculated inlet and outlet parameters of each device node in the raw gas conversion system with the calculated inlet and outlet parameters of the corresponding device node in the raw gas conversion system to obtain a fourth comparison result; Determine the working state of the measuring device according to the first comparison result, the second comparison result, the third comparison result, and the fourth comparison result; the working state includes: normal state and abnormal state; the measuring device is installed at the entrance and exit of each node device; When the working state is abnormal, an alarm is issued to prompt that the working state of the measuring device is abnormal.

3. The method for predicting parameters of a shaft furnace production process according to claim 1, characterized in that: After obtaining the shaft furnace body prediction model, the method further includes: Obtaining the export parameter constraints of the direct reduction shaft furnace; Inputting the outlet parameter constraint conditions into a multi-objective optimization evolutionary model to obtain a plurality of inlet parameter sets of the direct reduction shaft furnace; Input each inlet parameter set into the vertical furnace body prediction model to obtain an outlet parameter set corresponding to each inlet parameter set, and input each inlet parameter set into the reducing gas reverse circulation calculation model to obtain control inlet and outlet parameters of each device node in the top gas recovery system, control inlet and outlet parameters of each device node in the raw gas pretreatment system, and control inlet and outlet parameters of each device node in the raw gas conversion system; the reducing gas reverse circulation calculation model is obtained by training the graph neural network model with third sample data; the third sample data includes: historical outlet parameters and historical inlet parameters of each device node in the top gas recovery system, historical outlet parameters and historical inlet parameters of each device node in the raw gas pretreatment system, and historical outlet parameters and historical inlet parameters of each device node in the raw gas conversion system; An arbitrary inlet parameter set, an outlet parameter set calculated from the arbitrary inlet parameter set, and control inlet and outlet parameters are used as an optimized control scheme, and an operating cost of each optimized control scheme is calculated; the operating cost includes energy consumption cost and environmental protection cost; the energy consumption cost is determined based on the gas consumption and liquid consumption of each optimized control scheme; the environmental protection cost is determined based on the gas emission, electricity consumption, and pellet output within a preset time unit of each optimized control scheme; According to the screening conditions of the operating cost, a target optimization control scheme is determined to control the vertical furnace production through the target optimization control scheme; the screening conditions include: the deviation between the preset operating cost and the operating cost of the target optimization control scheme is less than or equal to a preset deviation threshold, and the confidence of the target optimization control scheme is greater than or equal to a preset confidence threshold.

4. The method for predicting parameters of a shaft furnace production process according to claim 3, characterized in that: The process of controlling shaft furnace production through the target optimization control scheme includes: Based on the target optimization control scheme, operation control parameters are generated; the operation control parameters include: heating furnace temperature control set value, converter furnace oxygen-coal ratio set value, decarbonization device load value: and, in a process of controlling the direct reduction shaft furnace, each device node in the top gas recovery system, each device node in the raw gas pretreatment system, and each device node in the raw gas conversion system to produce according to the operation control parameters, measuring the inlet and outlet parameters of the direct reduction shaft furnace, the inlet and outlet parameters of each device node in the top gas recovery system, the inlet and outlet parameters of each device node in the raw gas pretreatment system, and the inlet and outlet parameters of the device nodes in the raw gas conversion system to obtain measured parameters; and predicting the inlet and outlet parameters of each device node in the top gas recovery system, the inlet and outlet parameters of each device node in the raw gas pretreatment system, and the inlet and outlet parameters of the device nodes in the raw gas conversion system based on the inlet parameters of the direct reduction shaft furnace to obtain predicted parameters; Determining an adjustment strategy for the operation control parameter according to the deviation between the measured parameter and the predicted parameter; the adjustment strategy includes: a proportional-integral-derivative adjustment strategy and a feedforward adjustment strategy; Based on the adjustment strategy, the operation control parameter is adjusted to obtain a parameter adjustment amount; and the sum of the operation control parameter and the parameter adjustment amount is used as the adjusted operation control parameter to control the direct reduction shaft furnace, each equipment node in the top gas recovery system, each equipment node in the raw gas pretreatment system, and each equipment node in the raw gas conversion system to perform production according to the adjusted operation control parameter.

5. The method for predicting parameters of a shaft furnace production process according to any one of claims 1 to 4, characterized in that: Before determining the first sample data using the historical inlet parameters of the direct reduction shaft furnace and the shaft furnace body simulation model, the method further includes: Performing geometric modeling according to the size, position and direction of the direct reduction shaft furnace to obtain a physical model of the direct reduction shaft furnace; Setting the inlet, outlet, wall and heat flux density of the physical model, and discretizing the physical model to obtain a grid unit of the physical model; A reaction equation for the grid unit is configured and run to obtain a simulation model of the vertical furnace body. The reaction equation includes a continuity equation, an energy equation, and a momentum equation. The energy equation uses the enthalpy of the reaction, the heat exchange of the ore, and the heat exchange of the gas as energy source terms, and the temperature of the grid unit as the solution result; the continuity equation uses the rate of the chemical reaction as the energy source term, and the gas composition of the grid unit as the solution result; the momentum equation uses the gas inertial resistance, the iron ore inertial resistance, the gas viscous resistance, and the iron ore viscous resistance as energy source terms, and the gas flow rate of the grid unit as the solution result.

6. The method for predicting parameters of a shaft furnace production process according to claim 5, characterized in that: After obtaining the vertical furnace simulation model, the method further includes: Performing stratified sampling of historical inlet parameters of the direct reduction shaft furnace according to the parameter design range of the direct reduction shaft furnace to obtain an inlet parameter sample; the historical inlet parameters include: reducing gas composition, temperature, flow rate, pressure, and flow rate of pellets; the reducing gas composition includes hydrogen and carbon monoxide; Inputting the inlet parameter sample into the vertical furnace body simulation model to obtain an outlet parameter sample of the direct reduction vertical furnace; and adding the ratio of hydrogen to carbon monoxide, the coal gas reduction rate, the coal gas oxygen-coal ratio, and the gas per ton of iron as input characteristic parameters to the inlet parameter sample to obtain an expanded inlet parameter sample; The expanded input parameter sample and the expanded output parameter sample are used as the first sample data.

7. The method for predicting parameters of a shaft furnace production process according to any one of claims 1 to 4, characterized in that: The process of training a pre-established shaft furnace prediction model using the first sample data to obtain the shaft furnace body prediction model includes: Normalizing the first sample data, and dividing the normalized first sample data into training data and test data; The pre-established vertical furnace prediction model is trained by the training data to obtain a trained vertical furnace prediction model; the pre-established vertical furnace prediction model includes: an input preprocessing layer, a first residual unit layer, a transition layer, a second residual unit layer, a self-attention layer, a feature compression layer and a linear layer; the input preprocessing layer is used to perform a linear transformation on the training data to obtain linear transformation feature data, map the linear transformation feature data to a first preset dimensional feature space, and normalize, activate and apply a first random deactivation weight to the mapped linear transformation feature data; the first residual unit layer is used to calculate based on the first random deactivation weight and the normalized linear transformation feature data to obtain first residual data; the transition layer is used to convert the first residual data Mapping to a second preset dimensional feature space, and normalizing, activating, and applying a second random inactivation weight to the mapped first residual data; the second residual unit layer is used to calculate based on the second random inactivation weight and the normalized first residual data to obtain second residual data; the self-attention layer is used to adjust the feature weights in the second residual data through the query matrix, the key matrix, and the value matrix, and output the weighted sum of the second residual data; the feature compression layer is used to compress the weighted sum of the second residual data to a third preset dimensional feature space, and normalize, activate, and apply a third random inactivation weight to the mapped weighted sum; the linear layer is used to perform a linear transformation on the mapped weight based on the third random inactivation weight to obtain a prediction parameter; Testing the trained shaft furnace prediction model using the test data to obtain a test result; If the test result is greater than or equal to a preset test accuracy, the trained shaft furnace prediction model is used as the shaft furnace body prediction model; If the test result is less than the preset test accuracy, the parameters in the trained shaft furnace prediction model are adjusted until the test result output by the adjusted shaft furnace prediction model is greater than or equal to the preset test accuracy, and the adjusted shaft furnace prediction model is used as the shaft furnace body prediction model.

8. The method for predicting parameters of a shaft furnace production process according to claim 7, characterized in that: After obtaining the shaft furnace body prediction model, the method further includes: If any inlet parameter of the direct reduction shaft furnace body exceeds the corresponding parameter design range, the arbitrary inlet parameter, the inlet parameters in the same group as the arbitrary inlet parameter, and the newly added outlet parameters are added to the first sample data as a group of newly added sample data to obtain updated sample data; and the parameter intervals of all parameters in the newly added sample data are used as the updated parameter intervals; the newly added outlet parameters include: the arbitrary inlet parameter and the inlet parameters in the same group as the arbitrary inlet parameter as the input parameters of the direct reduction shaft furnace body, and the outlet parameters output by the direct reduction shaft furnace body; The number of groups of the newly added sample data is counted, and if the number of groups of the newly added sample data reaches a preset threshold, the pre-established shaft furnace prediction model is trained using the updated sample data.

9. The method for predicting parameters of a shaft furnace production process according to any one of claims 1 to 4, characterized in that: The process of training the graph neural network model using the second sample data to obtain the reduction gas forward circulation calculation model includes: Preprocessing the second sample data to obtain preprocessed feature data; the preprocessing operations include: data cleaning, removing outliers, marking null values, calculating average values ​​at different time points, and normalizing; The graph neural network model is trained with the preprocessed feature data to obtain the reducing gas forward circulation calculation model; the graph neural network model includes: a feature masking layer, a two-order attention propagation layer, a dynamic regularization layer and a regression output layer; the feature masking layer is used to selectively mask the preprocessed feature data to obtain the masked features; the two-order attention propagation layer is used to screen and extract the masked features, and weightedly aggregate the screened and extracted features, and redistribute the random inactivation probability in the aggregated features to correct or enhance the aggregated features through the redistributed attention weights, the dynamic regularization layer is used to regularize the corrected or enhanced aggregated features, and the regression output layer is used to map the regularized features to the parameter range of the entry and exit parameters.

10. A device for predicting parameters of a shaft furnace production process, characterized in that: include: Parameter acquisition module, used to obtain the current inlet parameters of the direct reduction shaft furnace; an outlet parameter output module, configured to input the current inlet parameters into a shaft furnace body prediction model to obtain predicted outlet parameters of the direct reduction shaft furnace; the shaft furnace body prediction model is obtained by training a pre-established shaft furnace prediction model using the first sample data; The first sample data is determined by the historical inlet parameters of the direct reduction shaft furnace and a simulation model of the shaft furnace body; An inlet and outlet parameter output module is used to input the predicted outlet parameters into the reducing gas forward circulation calculation model to obtain the predicted inlet and outlet parameters of each device node in the top gas recovery system, the predicted inlet and outlet parameters of each device node in the raw gas pretreatment system, and the predicted inlet and outlet parameters of each device node in the raw gas conversion system; The reducing gas forward circulation calculation model is obtained by training the graph neural network model with the second sample data; The second sample data includes: historical inlet parameters and historical outlet parameters of each device node in the top gas recovery system, historical inlet parameters and historical outlet parameters of each device node in the raw gas pretreatment system, and historical inlet parameters and historical outlet parameters of each device node in the raw gas conversion system; The prediction result output module is used to use the predicted export parameter and all predicted inlet and outlet parameters as parameter prediction results of the vertical furnace production process.

Images