Method for predicting production and quality of product gas of carbon monoxide and tempering co-production factory

Abstract

The method for predicting the yield and quality of carbon monoxide (CO) product gas includes: obtaining multiple sets of historical data from the steelmaking co-production pilot plant; training a machine learning model based on multiple cumulative intake volumes, multiple compressor pressures, multiple adsorption times, multiple tail gas CO concentrations, and multiple CO product gas yields from these multiple sets of historical data, thereby predicting the yield of CO product gas; and training a quality classification model based on multiple intake CO concentrations, multiple spray tower outlet temperatures, multiple cumulative intake volumes, multiple compressor pressures, multiple hydrolysis heater outlet temperatures, multiple adsorption times, multiple adsorption operating pressures, multiple tail gas CO2 concentrations, multiple tail gas CO concentrations, and multiple CO product gas yields from these multiple sets of historical data, thereby predicting whether the concentration and quality of CO product gas are up to standard.

Description

Technical Field

[0001] This invention relates to a method for predicting the yield and quality of carbon monoxide product gas, and particularly to a method for predicting the yield and quality of carbon monoxide product gas and a pilot plant for combined steel production. Prior Technology

[0002] To achieve the goals of low-carbon production and carbon recycling, a pilot steelmaking cogeneration plant has been proposed. This plant primarily utilizes an adsorption process to capture and reuse carbon monoxide (CO) and carbon dioxide (CO2) from the byproduct fuel gas produced during the steelmaking process. This not only reduces carbon emissions but also allows the collected CO to be supplied to downstream plants, saving them the cost of importing carbon feedstocks. However, due to the significant variability in the CO concentration of the byproduct fuel gas and the current lack of research on the relative relationship between the output of the pilot steelmaking cogeneration plant and various parameters in the adsorption process, it is impossible to predict in advance the yield and quality of CO product gas produced by the pilot plant under different operating conditions. Summary of the Invention

[0003] The purpose of this invention is to propose a method for predicting the yield and quality of carbon monoxide (CO) product gas, applicable to a steel co-production pilot plant. The prediction method includes: obtaining multiple sets of historical data from the steel co-production pilot plant, including: multiple historical inlet CO concentrations, multiple historical spray tower outlet temperatures, multiple historical cumulative inlet gas volumes, multiple historical compressor pressures, multiple historical hydrolysis heater outlet temperatures, multiple historical adsorption times, multiple historical adsorption operating pressures, and multiple historical tail gas carbon dioxide (CO) concentrations. 2) Concentration, multiple historical tail gas CO concentrations, and multiple historical CO product gas yields; a machine learning model is trained based on the multiple historical cumulative intake volumes, multiple historical compressor pressures, multiple historical adsorption times, multiple historical tail gas CO concentrations, and multiple historical CO product gas yields to predict the CO product gas yield of the steelmaking co-production pilot plant; and a quality classification model is trained based on the multiple historical intake CO concentrations, multiple historical spray tower outlet temperatures, multiple historical cumulative intake volumes, multiple historical compressor pressures, multiple historical hydrolysis heater outlet temperatures, multiple historical adsorption times, multiple historical adsorption operating pressures, multiple historical tail gas CO2 concentrations, multiple historical tail gas CO concentrations, and multiple historical CO product gas yields to predict whether the concentration and quality of the CO product gas in the steelmaking co-production pilot plant are up to standard.

[0004] In some embodiments, the steel co-production pilot plant sequentially includes a spray tower, a compressor, a hydrolysis heater, a desulfurization unit, a pressure swing adsorption decarbonization unit, a deoxygenation tower, and a CO purification unit.

[0005] In some embodiments, the plurality of historical inlet CO concentrations are taken from the inlet of the spray tower, the plurality of historical spray tower outlet temperatures are taken from the outlet of the spray tower, the plurality of historical cumulative inlet volumes are taken from the inlet of the compressor, the plurality of historical compressor pressures are taken from the compressor's set value, the plurality of historical hydrolysis heater outlet temperatures are taken from the hydrolysis heater's outlet, the plurality of historical adsorption times are the plurality of adsorption times set by each of the plurality of adsorption towers included in the pressure swing adsorption decarbonization unit and the CO purification unit, the plurality of historical adsorption operating pressures are the operating pressure set values ​​of the pressure swing adsorption decarbonization unit and the CO purification unit, the plurality of historical tail gas CO2 concentrations are taken from the outlet of the deaerator, the plurality of historical tail gas CO concentrations are taken from the adsorption tail gas outlet of the CO purification unit, and the plurality of historical CO product gas yields are taken from the CO product gas outlet of the CO purification unit.

[0006] In some embodiments, the machine learning model is implemented using a gradient boosting machine (GBM).

[0007] In some embodiments, the quality classification model sets a CO product gas concentration greater than 98.5% as the quality qualification standard.

[0008] In some embodiments, the quality classification model divides the modeling data into a training group and a test group, and uses the training group in conjunction with a neural network model and sets the loss function to binary cross entropy.

[0009] In some embodiments, the multiple sets of historical data are hourly averages.

[0010] The purpose of this invention is to provide a pilot plant for combined steel production, comprising: a spray tower, a compressor located downstream of the spray tower, a hydrolysis heater located downstream of the compressor, a desulfurization unit located downstream of the hydrolysis heater, a pressure swing adsorption decarbonization unit located downstream of the desulfurization unit, a deoxygenation tower located downstream of the pressure swing adsorption decarbonization unit, a CO purification unit located downstream of the deoxygenation tower, and a computing module. The computing module communicates with the pressure swing adsorption (PSA) decarbonization unit and the CO purification unit, and is used to: acquire multiple sets of historical data, including: multiple historical inlet CO concentrations, multiple historical spray tower outlet temperatures, multiple historical cumulative inlet gas volumes, multiple historical compressor pressures, multiple historical hydrolysis heater outlet temperatures, multiple historical adsorption times, multiple historical adsorption operating pressures, multiple historical tail gas CO2 concentrations, multiple historical tail gas CO concentrations, and multiple historical CO product gas yields; train a machine learning model based on the multiple historical cumulative inlet gas volumes, the multiple historical compressor pressures, the multiple historical adsorption times, the multiple historical tail gas CO concentrations, and the multiple historical CO product gas yields, thereby predicting the CO product gas yield of the steelmaking co-production pilot plant; and based on the multiple historical inlet CO concentrations, multiple historical spray tower outlet temperatures, multiple historical cumulative inlet gas volumes, multiple historical compressor pressures, multiple historical hydrolysis heater outlet temperatures, multiple historical adsorption times, multiple historical adsorption operating pressures, and multiple historical tail gas CO2 concentrations, the multiple historical CO2 concentrations, the multiple historical CO2 concentrations, the multiple historical CO2 concentrations, and the multiple historical tail gas CO2 yields, a machine learning model is generated to predict the CO product gas yield of the steelmaking co-production pilot plant; and based on the multiple historical inlet CO concentrations, multiple historical spray tower outlet temperatures, multiple historical cumulative inlet gas volumes, multiple historical compressor pressures, multiple historical hydrolysis heater outlet temperatures, multiple historical adsorption times, multiple historical adsorption operating pressures, and the multiple historical tail gas CO2 concentrations, the multiple historical ... The concentration of CO in the exhaust gas, the historical CO concentration, and the historical CO product gas output are used to train a quality classification model to predict whether the concentration and quality of CO product gas in the steelmaking co-production pilot plant are up to standard.

[0011] In some embodiments, the machine learning model is implemented using a gradient boosting machine.

[0012] In some embodiments, the quality classification model sets the concentration of CO product gas greater than 98.5% as the quality qualification standard. The quality classification model divides the modeling data into a training group and a test group, and uses a neural network model with the training group and sets the loss function to binary cross-entropy.

[0013] To make the above features and advantages of the present invention more apparent and understandable, specific embodiments are described below in conjunction with the accompanying drawings for detailed explanation. Simple Explanation of the Diagram

[0014] A better understanding of the invention can be obtained from the following detailed description taken in conjunction with the accompanying drawings. It should be noted that, according to industry standard practice, the features are not drawn to scale. In fact, the dimensions of the features can be arbitrarily increased or decreased for clarity of discussion. [Figure 1] is a schematic diagram of a pilot plant for combined steel production according to an embodiment of the present invention. [Figure 2] is a flowchart of a method for predicting the yield and quality of CO product gas according to an embodiment of the present invention. Implementation

[0015] The embodiments of the present invention will be discussed in detail below. However, it will be understood that the embodiments provide many applicable concepts that can be implemented in a wide variety of specific contexts. The discussed and disclosed embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0016] Figure 1 is a schematic diagram of a steelmaking co-production pilot plant according to an embodiment of the present invention. The steelmaking co-production pilot plant sequentially includes a spray tower 110, a compressor 120, a hydrolysis heater 130, a desulfurization unit 140, a pressure swing adsorption (PSA) decarbonization unit 150, a deaerator 160, and a carbon monoxide (CO) purification unit 170. The steelmaking co-production pilot plant uses a mixture of blast furnace gas (BFG) and converter gas (LDG) as feed gas.

[0017] The spray tower 110 is used to cool the feed gas. Since the temperature of the feed gas is higher than the ambient temperature, it needs to be cooled before compression. The cooling method is water spraying. The temperature of the feed gas after being cooled by the spray tower 110 is about 40˚C.

[0018] The compressor 120 is used to compress the feed gas. The compressor 120 is a water-injected screw compressor, which is used to pressurize the feed gas to a set pressure (e.g., 5 kg / cm2G).

[0019] The hydrolysis heater 130 includes a hydrolysis tower, which uses the moisture carried by the feed gas itself to convert carbon oxysulfide (COS) into hydrogen sulfide (H2S).

[0020] The desulfurization unit 140 includes a desulfurization tower, which removes hydrogen sulfide (H2S) from the feed gas using a desulfurizing agent. The principle is based on the reaction of hydrogen sulfide (H2S) in the feed gas with trace amounts of oxygen (O2) to generate elemental sulfur or sulfate, which is then deposited in the micropores of the desulfurizing agent. Specifically, the hydrolysis tower and desulfurization tower are designed for feed gas with low sulfur content, purifying sulfur impurities in the feed gas.

[0021] The pressure swing adsorption (PSA) decarbonization unit 150 is used to remove impurities such as carbon dioxide (CO2) and water (H2O) from the desulfurized feed gas. The PSA decarbonization unit 150 comprises multiple (e.g., six) adsorption towers that operate alternately to achieve continuous CO2 removal. Specifically, in one cycle, each adsorption tower sequentially undergoes the following steps: adsorption, pressure equalization and depressurization, forward depressurization, reverse depressurization, rinsing, pressure equalization and pressurization, and final pressurization. The desulfurized and decarbonized feed gas is sent to the deoxygenation tower 160, while the remaining gas in the aforementioned adsorption towers is discharged through the desorbed gas (DES) outlet.

[0022] The deoxygenation tower 160 is used to catalyze the reaction of O2 and CO to produce CO2, thereby achieving deoxygenation. Because the presence of oxygen can deactivate the catalyst in the CO purification unit 170, a high-efficiency deoxygenation catalyst PU-5 is used in the deoxygenation tower 160, and the gas is heated to 100˚C via a deoxygenation heat exchanger. Specifically, the feed gas enters the deoxygenation tower 160 after desulfurization and adsorption decarbonization. Under the action of the deoxygenation catalyst, O2 and CO react to produce CO2, thus achieving deoxygenation.

[0023] CO purification unit 170 is used to separate and purify CO components. CO purification unit 170 contains multiple (e.g., five) adsorption towers that operate alternately to achieve continuous CO purification. Specifically, in one cycle, each adsorption tower sequentially undergoes the following steps: adsorption, pressure equalization and depressurization, displacement, reverse release, vacuuming, pre-adsorption, pressure equalization and pressurization, and final pressurization. CO product gas from the steelmaking co-production pilot plant is discharged from the CO product gas outlet OUT1 of CO purification unit 170, while the adsorption tail gas outlet OUT2 of CO purification unit 170 discharges tail gas CO. Additionally, a portion of the CO product gas is compressed (represented by the symbol COM in Figure 1) and then returned to CO purification unit 170.

[0024] As shown in Figure 1, the pilot plant for combined steel production further includes a computing module 180, which is used to execute methods for predicting the yield and quality of CO product gas. The computing module 180 can be a personal computer, laptop computer, server, industrial computer, or various electronic devices with computing capabilities, including a central processing unit, microprocessor, microcontroller, application-specific integrated circuit, etc.

[0025] Figure 2 is a flowchart of a method for predicting the yield and quality of CO product gas according to an embodiment of the present invention. As shown in Figure 2, in step S1, the calculation module 180 obtains multiple sets of historical data from the steelmaking co-production pilot plant. The multiple sets of historical data include: multiple historical inlet CO concentrations, multiple historical spray tower outlet temperatures, multiple historical cumulative inlet gas volumes, multiple historical compressor pressures, multiple historical hydrolysis heater outlet temperatures, multiple historical adsorption times, multiple historical adsorption operating pressures, multiple historical tail gas CO2 concentrations, multiple historical tail gas CO concentrations, and multiple historical CO product gas yields.

[0026] In embodiments of the present invention, the multiple sets of historical data are hourly averages. For example, the unit of historical intake CO concentration is cubic meters per hour.

[0027] In an embodiment of the present invention, the plurality of historical inlet CO concentrations are taken from the inlet of the spray tower 110, the plurality of historical spray tower outlet temperatures are taken from the outlet of the spray tower 110, the plurality of historical cumulative inlet volumes are taken from the inlet of the compressor 120, the plurality of historical compressor pressures are taken from the set value of the compressor 120, the plurality of historical hydrolysis heater outlet temperatures are taken from the outlet of the hydrolysis heater 130, the plurality of historical adsorption times are the plurality of adsorption times set by each of the plurality of adsorption towers included in the pressure swing adsorption decarbonization unit 150 and the CO purification unit 170, the plurality of historical adsorption operating pressures are the operating pressure set values ​​of the pressure swing adsorption decarbonization unit 150 and the CO purification unit 170, the plurality of historical tail gas CO2 concentrations are taken from the outlet of the deaerator 160, the plurality of historical tail gas CO concentrations are taken from the adsorption tail gas outlet OUT2 of the CO purification unit 170, and the plurality of historical CO product gas yields are taken from the CO product gas outlet OUT1 of the CO purification unit 170.

[0028] In an embodiment of the present invention, a plurality of sensors (not shown in Figure 1) are installed in the steelmaking co-production pilot plant. These sensors can be positioned at corresponding locations within the pilot plant to sense relevant values ​​(e.g., two CO detectors to sense the CO concentration in the inlet gas and the CO concentration in the exhaust gas, a CO2 detector to sense the CO2 concentration in the exhaust gas, two temperature sensors to sense the outlet temperature of the spray tower and the outlet temperature of the hydrolysis heater, and two inlet sensors to sense the cumulative inlet gas volume and the CO product gas output). The calculation module 180 can be connected to these sensors via wired or wireless means, so that these values ​​are transmitted to the calculation module 180. The operation of the sensors will be understood by those skilled in the art and will not be described in detail here. In addition, the calculation module 180 can also be connected to the compressor 120 via wired or wireless means to obtain the compressor pressure. Furthermore, the calculation module 180 can also be connected to the pressure swing adsorption decarbonization unit 150 and the CO purification unit 170 via wired or wireless means to obtain the adsorption time and adsorption operating pressure.

[0029] In an embodiment of the present invention, after step S1 and before step S2, data preprocessing can be performed by the calculation module 180, for example, removing abnormal values ​​of flow rate that are still present even when the outlet valve is closed from the historical data.

[0030] As shown in Figure 2, in step S2, the calculation module 180 trains a machine learning model based on the multiple historical cumulative intake volumes, multiple historical compressor pressures, multiple historical adsorption times, multiple historical tail gas CO concentrations, and multiple historical CO product gas yields, thereby predicting the CO product gas yield of the steelmaking co-production pilot plant.

[0031] In embodiments of the present invention, the machine learning model is implemented using a gradient boosting machine (GBM), but the present invention is not limited thereto. For example, the machine learning model used in step S2 can also be a decision tree, random forest, multi-level neural network, convolutional neural network, support vector machine, etc.

[0032] Specifically, the purpose of the machine learning model in step S2 is to predict the CO product gas output based on the cumulative intake air volume, compressor pressure, adsorption time, and tail gas CO concentration. In other words, during the training phase, historical cumulative intake air volume, historical compressor pressure, historical adsorption time, and historical tail gas CO concentration are used as inputs to the machine learning model, and historical CO product gas output is used as the output. During the testing phase (or prediction phase), the current cumulative intake air volume, compressor pressure, adsorption time, and tail gas CO concentration of the steel co-production pilot plant are input into the trained machine learning model, causing the model to output the predicted CO product gas output of the steel co-production pilot plant accordingly.

[0033] As described above, this invention can predict in advance the output of CO product gas produced in a pilot plant under different operating conditions, providing operators with a reference. This allows operators to know the impact of changes in inlet CO concentration on the output of CO product gas, solving the problem that previously, the correlation between changes in CO product gas output and changes in inlet CO concentration could only be known after the fact. In other words, this invention can avoid the problem of unstable CO product gas output.

[0034] As shown in Figure 2, in step S3, the calculation module 180 trains a quality classification model based on the multiple historical inlet CO concentrations, multiple historical spray tower outlet temperatures, multiple historical cumulative inlet air volumes, multiple historical compressor pressures, multiple historical hydrolysis heater outlet temperatures, multiple historical adsorption times, multiple historical adsorption operating pressures, multiple historical tail gas CO2 concentrations, multiple historical tail gas CO concentrations, and multiple historical CO product gas yields. This model predicts whether the CO product gas concentration quality of the steelmaking co-production pilot plant is up to standard. In other words, when training the quality classification model, the output of the model is whether the CO product gas concentration quality is up to standard or not, and the inputs to the model are the multiple historical inlet CO concentrations, multiple historical spray tower outlet temperatures, multiple historical cumulative inlet air volumes, multiple historical compressor pressures, multiple historical hydrolysis heater outlet temperatures, multiple historical adsorption times, multiple historical adsorption operating pressures, multiple historical tail gas CO2 concentrations, multiple historical tail gas CO concentrations, and multiple historical CO product gas yields.

[0035] In an embodiment of the present invention, the quality classification model sets a CO product gas concentration greater than 98.5% (this is only an example and the present invention is not limited to this) as the quality qualification standard. In an embodiment of the present invention, the quality classification model divides the modeling data into a training group and a test group, and uses a neural network model with a loss function of binary cross entropy in the training group, and uses the test group to verify the classification accuracy of the quality classification model.

[0036] Specifically, the quality classification model is trained based on the multiple historical inlet CO concentrations, multiple historical spray tower outlet temperatures, multiple historical cumulative inlet gas volumes, multiple historical compressor pressures, multiple historical hydrolysis heater outlet temperatures, multiple historical adsorption times, multiple historical adsorption operating pressures, multiple historical tail gas CO2 concentrations, multiple historical tail gas CO concentrations, and multiple historical CO product gas yields. The trained neural network model can be used to predict whether the quality of the CO product gas produced at the current steelmaking co-production pilot plant is up to standard.

[0037] As described above, this invention can predict in advance whether the quality of CO product gas produced in the pilot plant of the steelmaking co-production facility is up to standard under different operating conditions, providing operators with a reference. This allows operators to know the impact of changes in the inlet CO concentration on the quality of the CO product gas, solving the problem that previously, the correlation between changes in CO product gas quality and changes in inlet CO concentration could only be known after the fact. In other words, this invention can avoid the problem of unstable CO product gas quality.

[0038] It is worth noting that each step in Figure 2 can be implemented as multiple pieces of code or circuits, and this invention is not limited thereto. Furthermore, the method in Figure 2 can be used in conjunction with the above embodiments, or it can be used alone. In other words, other steps can also be added between the steps in Figure 2.

[0039] The foregoing has outlined the features of several embodiments, thus enabling those skilled in the art to better understand the nature of the invention. Those skilled in the art should recognize that the invention can be readily used as a basis to design or modify other processes and structures, thereby achieving the same objectives and / or advantages as the embodiments described herein. Those skilled in the art should also understand that these equivalent constructions do not depart from the spirit and scope of the invention, and that various changes, substitutions, and modifications can be made without departing from the spirit and scope of the invention.

[0040] 110: Spray Tower 120: Compressor 130: Hydrolysis heater 140: Desulfurization Unit 150: Pressure Swing Adsorption Decarbonization Unit 160: Deaerator 170: CO purification unit 180: Computing Module BFG: Blast Furnace Gas COM: Compression DES: Desorption gas discharge port LDG: Converter Gas OUT1: CO product gas outlet end OUT2: Adsorption exhaust gas outlet end S1, S2, S3: Steps

Claims

1. A method for predicting the yield and quality of carbon monoxide product gas, applicable to a steelmaking co-production pilot plant, executed by a computing module, wherein the computing module is a personal computer, laptop computer, server, industrial computer, or various electronic devices with computing capabilities, the prediction method comprising: Multiple sets of historical data were obtained from the pilot plant for the combined steel production, including: multiple historical intake carbon monoxide (CO) concentrations, multiple historical spray tower outlet temperatures, multiple historical cumulative intake volumes, multiple historical compressor pressures, multiple historical hydrolysis heater outlet temperatures, multiple historical adsorption times, multiple historical adsorption operating pressures, multiple historical tail gas carbon dioxide (CO2) concentrations, multiple historical tail gas CO concentrations, and multiple historical CO product gas yields. A machine learning model is trained based on historical cumulative intake air volume, historical compressor pressure, historical adsorption time, historical tail gas CO concentration, and historical CO product gas yield to predict the CO product gas yield of the steelmaking co-production pilot plant. A quality classification model is trained based on historical intake CO concentration, historical spray tower outlet temperature, historical cumulative intake air volume, historical compressor pressure, historical hydrolysis heater outlet temperature, historical adsorption time, historical adsorption operating pressure, historical tail gas CO2 concentration, historical tail gas CO concentration, and historical CO product gas yield to predict whether the CO product gas concentration quality of the steelmaking co-production pilot plant is up to standard. The quality classification model divides the modeling data into a training group and a test group, and uses the training group in conjunction with a neural network model with a loss function set as binary cross entropy.

2. The prediction method as described in claim 1, wherein the steel co-production pilot plant sequentially includes a spray tower, a compressor, a hydrolysis heater, a desulfurization unit, a pressure swing adsorption decarbonization unit, a deoxygenation tower and a CO purification unit.

3. The prediction method as described in claim 2, wherein the historical inlet CO concentrations are taken from the inlet of the spray tower, the historical spray tower outlet temperatures are taken from the outlet of the spray tower, the historical cumulative inlet gas volume is taken from the inlet of the compressor, the historical compressor pressure is taken from the set value of the compressor, the historical hydrolysis heater outlet temperatures are taken from the outlet of the hydrolysis heater, the historical adsorption times are multiple adsorption times set by each of the plurality of adsorption towers included in the pressure swing adsorption decarbonization unit and the CO purification unit, the historical adsorption operating pressures are the operating pressure set values ​​of the pressure swing adsorption decarbonization unit and the CO purification unit, the historical tail gas CO2 concentrations are taken from the outlet of the deaerator, the historical tail gas CO concentrations are taken from the adsorption tail gas outlet of the CO purification unit, and the historical CO product gas output is taken from the CO product gas outlet of the CO purification unit.

4. The prediction method as described in Request 1, wherein the machine learning model is implemented using a gradient boosting machine (GBM).

5. The prediction method as described in Request 1, wherein the quality classification model sets the concentration of CO product gas greater than 98.5% as the quality qualification standard.

6. The prediction method as described in claim 1, wherein the multiple sets of historical data are hourly averages.

7. A pilot plant for combined steel production, comprising: A spray tower; A compressor is located downstream of the spray tower; A hydrolysis heater is located after the compressor; a desulfurization unit is located after the hydrolysis heater; a pressure swing adsorption (PSA) decarbonization unit is located after the desulfurization unit; a deoxygenation tower is located after the PSA decarbonization unit; a CO purification unit is located after the deoxygenation tower; and a computing module is communicatively connected to the PSA decarbonization unit and the CO purification unit, and is used to: obtain multiple sets of historical data, including: multiple historical inlet carbon monoxide (CO) concentrations, multiple historical spray tower outlet temperatures, multiple historical cumulative inlet gas volumes, multiple historical compressor pressures, multiple historical hydrolysis heater outlet temperatures, multiple historical adsorption times, multiple historical adsorption operating pressures, multiple historical tail gas carbon dioxide (CO2) concentrations, multiple historical tail gas CO concentrations, and multiple historical CO product gas yields; A machine learning model is trained based on historical cumulative intake air volume, historical compressor pressure, historical adsorption time, historical tail gas CO concentration, and historical CO product gas yield to predict the CO product gas yield of the steelmaking co-production pilot plant. A quality classification model is trained based on historical intake CO concentration, historical spray tower outlet temperature, historical cumulative intake air volume, historical compressor pressure, historical hydrolysis heater outlet temperature, historical adsorption time, historical adsorption operating pressure, historical tail gas CO2 concentration, historical tail gas CO concentration, and historical CO product gas yield to predict whether the CO product gas concentration quality of the steelmaking co-production pilot plant is up to standard. The quality classification model divides the modeling data into a training group and a test group, and uses the training group in conjunction with a neural network model with a binary cross-entropy loss function.

8. The steel co-production pilot plant as described in claim 7, wherein the machine learning model is implemented using a gradient boosting machine.

9. The steel co-production pilot plant as described in claim 7, wherein the quality classification model sets the concentration of CO product gas greater than 98.5% as the quality qualification standard.

Images