Classifications

• • C—CHEMISTRY; METALLURGY
  • C21—METALLURGY OF IRON
  • C21B—MANUFACTURE OF IRON OR STEEL
  • C21B5/00—Making pig-iron in the blast furnace

Definitions

• This disclosure relates to a molten iron temperature control method, a molten iron temperature control device, an operation guidance method, a blast furnace operation method, and a molten iron manufacturing method.
• Patent Document 1 discloses a method for controlling the molten iron temperature based on a predicted value in order to compensate for the error between the calculated value and the actual measured value of at least one of the reducing agent ratio, solution loss carbon amount, ironmaking rate, and gas utilization rate.
• the technology in Patent Document 1 adjusts the gas reduction equilibrium parameters or the coke rate parameters at the furnace top in the physical model, while calculating a predicted value for the molten iron temperature using the physical model, assuming that the current operating variables are maintained.
• Patent Document 3 also discloses a molten iron temperature control method that uses a one-dimensional convolutional neural network to construct a behavioral model in which an operator determines operational actions, and then uses the constructed behavioral model to determine operational actions to bring the molten iron temperature in a blast furnace to a target temperature.
• the behavioral model uses at least one of the molten iron temperature, tuyere embedding temperature, and coke ratio as an observed variable (input data), and at least one of the pulverized coal ratio, blast moisture, and coke ratio as an manipulated variable (output data).
• the method described in Patent Document 1 can predict changes in molten iron temperature over a long period of time (e.g., 4 to 12 hours). However, it is difficult to use the method described in Patent Document 1 to predict sudden changes in molten iron temperature due to hanging or slippage caused by a sudden deterioration in permeability, or to predict sudden changes in molten iron temperature associated with these furnace abnormalities.
• the method described in Patent Document 2 can input various observational information, such as the results of reactions in the furnace. Therefore, the method described in Patent Document 2 can predict sudden changes in molten iron temperature that are difficult to predict using physical models, but the prediction range is limited to a short period of time (e.g., 1 to 4 hours).
• Patent Document 3 can learn and imitate past exemplary operational actions performed by operators based on experience, even under conditions where the prediction accuracy of physical or statistical models decreases.
• the method relies on the learning data used for machine learning, and the prediction accuracy of the generated machine learning model decreases when there is little data similar to the situation to be addressed.
• each model has its own advantages and disadvantages, and when a single model is used, appropriate operational actions may not be suggested due to decreased prediction accuracy.
• the pulverized coal ratio is the amount of pulverized coal per ton of molten iron.
• the blast moisture is the amount of moisture added to the hot blast.
• the blast temperature is the temperature of the hot blast.
• Manipulating the blast moisture is also known to be a highly effective way of controlling the molten iron temperature.
• the decomposition of water vapor is an endothermic reaction. Therefore, by increasing the blast moisture content above the atmospheric humidity, the amount of reducing agent required to maintain a constant molten iron temperature increases.
• the amount of reducing agent required to maintain a constant molten iron temperature can be reduced.
• the gas temperature at the tuyere tip drops, and it takes about five hours for the molten pig iron temperature to rise.
• adjusting the pulverized coal ratio has a less immediate effect than adjusting the blast moisture. Therefore, in order to achieve both a reduction in the reducing agent rate and highly accurate control of the molten pig iron temperature, it is necessary to appropriately use the control variables of the operational actions.
• the purpose of this disclosure is to provide a molten iron temperature control method, molten iron temperature control device, operation guidance method, blast furnace operation method, and molten iron manufacturing method that can achieve both a reduction in the reducing agent rate and highly accurate molten iron temperature control.
• a molten iron temperature control method includes: a model calculation step of calculating the molten iron temperature or an action determination value for an operational action for controlling the molten iron temperature using a plurality of models that input operational data of the blast furnace; and an operational action determination step of selecting one of the plurality of models using the calculated molten iron temperature or the action determination value based on a determination criterion based on trends in the prediction accuracy of the plurality of models, and determining an operation variable and an operation amount as the operational action.
• the plurality of models are a physical model capable of expressing a transient state by calculating heat transfer, chemical reactions, and fluid flow inside the blast furnace; and a statistical model for extracting similar operating conditions based on past operating data of the blast furnace and predicting the molten iron temperature.
• the plurality of models are The system further includes a machine learning model that is generated by machine learning using learning data that links past operation data of the blast furnace with exemplary operation actions performed by operators, and that takes the operation data of the blast furnace as input and outputs operation actions that are determined to be appropriate.
• the operational action determination step determines an operation variable and an operation amount as the operational action based on the predicted long-term molten iron temperature calculated using the physical model, the predicted short-term molten iron temperature calculated using the statistical model, the action determination value calculated using the machine learning model, a blast moisture control range determined in consideration of a reduction in reducing agent rate, and a pulverized coal use restriction based on operational constraints.
• the operational action determination step determines an operation variable and an operation amount as the operational action based on the action determination value when the long-term future predicted molten iron temperature and the short-term future predicted molten iron temperature are both within a target range.
• the operational action determination step determines an operation variable and an operation amount as the operational action based on the predicted molten iron temperature in the short time future when both the predicted molten iron temperature in the long time future and the predicted molten iron temperature in the short time future deviate from their target ranges and the corresponding operational actions match, or when the predicted molten iron temperature in the long time future is within the target range and the predicted molten iron temperature in the short time future deviates from the target range.
• any one of (1) to (6) In the operational action determination step, when the long-term future predicted molten iron temperature deviates from a target range and the respective operational actions corresponding to the long-term future predicted molten iron temperature and the short-term future predicted molten iron temperature do not match, and the operational action corresponding to the long-term future predicted molten iron temperature is at least not in conflict with the operational action corresponding to the action determination value, an operation variable and an operation variable are determined as an operational action based on the long-term future predicted molten iron temperature, and when the operational action corresponding to the long-term future predicted molten iron temperature does not match with the operational action corresponding to the action determination value, no proactive operational action is executed.
• the operational constraints include at least one of an upper limit of tuyere gas temperature, a lower limit of tuyere gas temperature, an upper limit of reducing agent rate, a lower limit of reducing agent rate, and a lower limit of furnace top gas temperature.
• the control range of the blast humidity is determined by adding a value to the atmospheric humidity that allows reduction of the blast humidity by an amount determined as a one-time operation amount at least once.
• a molten iron temperature control device includes: a model calculation unit that calculates the molten iron temperature or an action determination value for an operational action for controlling the molten iron temperature using a plurality of models that input operational data of the blast furnace; and an operational action determination unit that uses the calculated molten iron temperature or the action determination value to select one of the plurality of models based on a determination criterion that is based on trends in the prediction accuracy of the plurality of models, and determines an operation variable and an operation amount as the operational action.
• the plurality of models are The system further includes a machine learning model that is generated by machine learning using learning data that links past operation data of the blast furnace with exemplary operation actions performed by operators, and that takes the operation data of the blast furnace as input and outputs operation actions that are determined to be appropriate.
• the operational action determination unit determines operation variables and operation amounts as the operational action based on the predicted long-term molten iron temperature calculated using the physical model, the predicted short-term molten iron temperature calculated using the statistical model, the action determination value calculated using the machine learning model, a blast moisture control range determined in consideration of a reduction in reducing agent rate, and a pulverized coal use restriction based on operational constraints.
• An operation guidance method includes: The operational action determined by any one of the molten iron temperature control methods (1) to (9) is presented so that the operator of the blast furnace can confirm it.
• a method for operating a blast furnace includes: The method includes a step of controlling the blast furnace in accordance with the operational action determined by the molten iron temperature control method according to any one of (1) to (9).
• a method for producing molten iron according to an embodiment of the present disclosure includes: (15) A step of controlling the blast furnace according to the blast furnace operation method to produce molten iron.
• This disclosure provides a molten iron temperature control method, a molten iron temperature control device, an operation guidance method, a blast furnace operation method, and a molten iron manufacturing method that can achieve both a reduction in the reducing agent ratio and highly accurate molten iron temperature control.
• FIG. 1 is a block diagram showing the configuration of a molten iron temperature control device according to an embodiment of the present disclosure.
• FIG. 2A is a flowchart showing the flow of processing by the operational action determination unit.
• FIG. 2B is a flowchart showing the flow of processing by the operational action determination unit.
• FIG. 3 is a diagram showing changes in the molten iron temperature, coke rate, pulverized coal rate, blast humidity, and atmospheric humidity after four days of operation using a blast furnace operating method using a molten iron temperature control device according to an embodiment of the present disclosure.
• FIG. 3 is a diagram showing changes in the molten iron temperature, coke rate, pulverized coal rate, blast humidity, and atmospheric humidity after four days of operation using a blast furnace operating method using a molten iron temperature control device according to an embodiment of the present disclosure.
• FIG. 4 is a diagram showing a comparison of the molten iron temperature variation and the reducing agent ratio when a blast furnace operation method using a molten iron temperature control device according to an embodiment of the present disclosure is carried out and a conventional operation method carried out by an operator.
• FIG. 5 is a diagram illustrating an example of a system configuration including a molten iron temperature control device according to an embodiment of the present disclosure.
• FIG. 1 is a block diagram showing the configuration of a molten iron temperature control device 10 according to an embodiment of the present disclosure.
• the molten iron temperature control device 10 is configured by an information processing device such as a computer.
• an arithmetic processing device such as a CPU (Central Processing Unit) executes a program to function as a model calculation unit 11 and an operational action determination unit 12.
• the processing performed by the model calculation unit 11 may be referred to as a model calculation step.
• the processing performed by the operational action determination unit 12 may be referred to as an operational action determination step. The functions of each unit will be described later.
• An operation database 20 is connected to the molten iron temperature control device 10 in a data-readable form.
• the operation database 20 stores operation factors such as the furnace top coke ratio, blast flow rate, oxygen enrichment amount, blast temperature, blast moisture, and pulverized coal flow rate.
• the operation database 20 also stores actual values of process variables calculated based on the volume fractions of CO and CO2 in the discharged furnace top gas.
• the operation database 20 also stores historical data on temperatures and pressures measured at at least one of the furnace body, furnace top, and furnace bottom. Examples of process variables include the molten iron temperature, iron-making rate, solution loss carbon amount, and gas utilization rate.
• the molten iron temperature control device 10 executes each process of the molten iron temperature control method using the model calculation unit 11 and operational action determination unit 12 to identify operational variables for molten iron temperature control and output the optimal action as guidance.
• the model calculation unit 11 calculates an action determination value for the molten iron temperature or an operation action for controlling the molten iron temperature using multiple models that use blast furnace operation data as input. The action determination value will be described later.
• the model calculation unit 11 executes processing using a physical model, a statistical model, and a machine learning model when a command to execute model calculation is input to the molten iron temperature control device 10.
• three models are used in this embodiment, the number of models is not limited to three and may be, for example, two or four or more.
• the multiple models used by the model calculation unit 11 are generated in advance and stored in a storage device accessible to the molten iron temperature control device 10. The model calculation unit 11 reads the models from the storage device before executing the calculation processing.
• the physical model is composed of a group of partial differential equations that consider multiple physical phenomena, such as iron ore reduction, heat exchange between iron ore and coke, and iron ore melting, and is capable of calculating variables (output variables) that represent the state inside the blast furnace in a non-steady-state.
• the physical model can predict the reactions that occur within the furnace, it is capable of long-term predictions (4 to 12 hours) into the future, including when operating variables are changed recently.
• the long period (4 to 12 hours) in the physical model's predictions takes into account the time it takes for the raw materials charged from the top of the blast furnace to descend.
• the statistical model predicts the molten iron temperature by extracting similar operating conditions based on past operating data of the blast furnace.
• the statistical model in this embodiment is a model that uses the just-in-time method described in Reference 2 (Yamamoto Shigeru, "Just-In-Time Predictive Control: Predictive Control Based on Accumulated Data," Measurement and Control, Vol. 52, No. 10, p. 878).
• the statistical model predicts the molten iron temperature a short time (1 to 4 hours) into the future using measured values of temperature and pressure measured at the blast furnace body, top, and bottom, as well as actual values of process variables, as explanatory variables.
• the statistical model sequentially constructs local linear functions using data near the explanatory variables at the current time.
• the explanatory variables at the current time are linked to the molten iron temperature 1 to 4 hours into the future in order to take into account the residence time of the molten iron at the bottom of the furnace.
• the machine learning model is generated by machine learning using learning data that links past operational data of the blast furnace with exemplary operational actions performed by operators.
• the generated machine learning model takes operational data of the blast furnace as input and outputs operational actions that are determined to be appropriate.
• the machine learning model may use a one-dimensional convolutional neural network technique, similar to the method described in Patent Document 3, for example.
• the machine learning model in this embodiment outputs a value (hereinafter referred to as the action judgment value) obtained by subtracting the judgment probability of the molten iron temperature lowering action from the calculated judgment probability of the molten iron temperature raising action.
• the molten iron temperature raising action is an operational action for raising the molten iron temperature.
• the molten iron temperature lowering action is an operational action for lowering the molten iron temperature.
• the machine learning model in this embodiment includes an action judgment value as the operational action that is determined to be appropriate to output.
• machine learning using a one-dimensional convolutional neural network as a machine learning model may involve learning not only the direction of action (increase or decrease in pulverized coal ratio, blast moisture, etc.) but also the amount of operation (amount to increase or decrease).
• the first is the relationship between model selection and the time delay until changes in the manipulated variables are reflected in the hot metal temperature.
• the first is the relationship between model selection and the time delay until changes in the manipulated variables are reflected in the hot metal temperature.
• blast moisture has a more immediate effect (in other words, better responsiveness). Therefore, if a statistical model predicts a sudden rise or fall in the hot metal temperature in the short term, manipulation of blast moisture is prioritized. If a physical model predicts a long-term deviation of the hot metal temperature from the target, manipulation of either the pulverized coal ratio or blast moisture is selected.
• the machine learning model is used to proactively execute appropriate operational actions when both the physical model and the statistical model determine that no action is necessary, or when the prediction accuracy of the physical model and the statistical model deteriorates due to disturbances.
• Physical and statistical models can experience a significant drop in prediction accuracy when disturbances that are difficult to measure are large. This can lead to conflicting operational actions corresponding to the long-term future molten iron temperature based on the physical model and the short-term future molten iron temperature based on the statistical model. In such cases, it becomes difficult to determine which operational action to take or when to switch operational actions. Furthermore, if the long-term future molten iron temperature based on the physical model is predicted to deviate from the target range, but the short-term future molten iron temperature based on the statistical model is predicted to be within the target range, it becomes difficult to determine the timing to take the operational action corresponding to the long-term future molten iron temperature. However, even in such cases, experienced operators look at the trends in the operational data and take the appropriate operational action at the appropriate time.
• Machine learning models can be used to mimic the operational actions of exemplary operators, allowing appropriate operational actions to be taken in advance and at the appropriate time, even in such cases. Machine learning models can also be used to determine the appropriateness of operational actions based on physical and statistical models.
• the second point is a guideline for determining the blast moisture control amount based on the desired blast moisture range, taking into account the realization of low RAR operation and response to sudden changes in molten iron temperature.
• the blast moisture is reduced as an action to raise the molten iron temperature. Reducing the blast moisture reduces the RAR.
• the control variable for molten iron temperature control to raise the molten iron temperature is limited to the pulverized coal ratio, and the action is limited to increasing the pulverized coal ratio.
• the responsiveness of operational actions deteriorates when the molten iron temperature suddenly changes.
• the predetermined value may be determined by adding a value (see ⁇ in Figures 2A and 2B) to the atmospheric humidity that allows a reduction in blast humidity by an amount determined as a single operation amount to be performed at least once.
• the predetermined value may also have a range so that the blast humidity is maintained within a desired range. That is, if a long-term decrease in the molten iron temperature is predicted based on a physical model, and the blast humidity exceeds the desired range, an action to decrease the blast humidity is prioritized.
• control range the desired range of blast humidity
• the third point is the restriction on the use of pulverized coal based on operational constraints. For example, if the pulverized coal ratio is increased when the reducing agent ratio is high, the amount of fines in the lower furnace may increase, potentially worsening permeability. Furthermore, when the tuyere gas temperature is low, pulverized coal has poor combustibility, so increasing the pulverized coal ratio may result in the generation of unburned pulverized coal and worsening permeability. Therefore, when the reducing agent ratio is at the predetermined upper limit or the tuyere gas temperature is at the predetermined lower limit, it is preferable to reduce the blast moisture content in order to increase the molten iron temperature.
• the manipulated variables based on the physical model and statistical model may be calculated, for example, using a known model predictive control technique or a known method of defining an influence coefficient, or may be set as constants.
• the manipulated variables based on the machine learning model may be calculated by defining an influence coefficient for the action determination value, or may be set as constants.
• the operational action decision unit 12 is constructed based on these concepts of determining manipulated variables and manipulated variables.
• 2A and 2B are flowcharts showing the processing flow of the operational action determination unit 12.
• 2A and 2B constitute one flowchart.
• the processing of the flowchart selects one of a plurality of models based on a determination criterion based on the trend of the prediction accuracy of the plurality of models.
• the operation of the operational action determination unit 12 will be described with reference to 2A and 2B.
• the flowcharts of 2A and 2B start when an execution command for operational action determination is input to the molten iron temperature control device 10, and the processing proceeds to step S1.
• the operational action decision unit 12 determines the difference between the predicted value of the hot metal temperature based on the physical model calculated by the model calculation unit 11 (physics-model-predicted hot metal temperature, long-term predicted hot metal temperature) and the target value of the hot metal temperature (target hot metal temperature).
• the target range of the difference between the physical-model-predicted hot metal temperature and the target hot metal temperature is set to be greater than or equal to - ⁇ 1 and less than or equal to ⁇ 1 .
• step S11 If the physical-model-predicted hot metal temperature ⁇ 1 hour into the future is less than - ⁇ 1 compared to the target hot metal temperature, the operational action determination process proceeds to step S11 so that an operational action to increase the hot metal temperature is taken. If the physical-model-predicted hot metal temperature ⁇ 1 hour into the future is greater than ⁇ 1 compared to the target hot metal temperature, the operational action determination process proceeds to step S21 so that an operational action to decrease the hot metal temperature is taken. If the physical model-predicted hot metal temperature ⁇ 1 hour ahead is greater than or equal to - ⁇ 1 and less than or equal to ⁇ 1 compared to the target hot metal temperature, it is determined that no action is required in the prediction based on the physical model, and the operational action determination process proceeds to step S31.
• ⁇ 1 is the future time predicted using the physical model.
• ⁇ 1 is preferably set to 4 to 12 hours, taking into account the fact that it takes approximately 8 hours for raw materials charged from the top of the blast furnace to descend to the lower part of the furnace (tuyere) and the time delay until the combustion of pulverized coal in the tuyere is reflected in the hot metal temperature.
• ⁇ 1 is preferably set to 1 to 30°C depending on the hot metal temperature control target.
• the operational action determination unit 12 determines the relationship between the difference between the blast humidity at the current time and the atmospheric humidity and the blast humidity control range determined in consideration of a reduction in the reducing agent rate. If the blast humidity at the current time is greater than atmospheric humidity + ⁇ (g/Nm 3 ), an action to reduce the blast humidity is presented in order to increase the molten iron temperature, and the series of operational action determination processes ends. If the blast humidity at the current time is equal to or less than atmospheric humidity + ⁇ (g/Nm 3 ), the operational action determination process proceeds to step S12.
• ⁇ is a value that allows the blast humidity to be reduced by an amount determined as a single operation amount at least once.
• step S12 the operational action determination unit 12 makes a determination regarding the reducing agent ratio and tuyere gas temperature in relation to pulverized coal use restrictions based on operational constraints. If the reducing agent ratio is at the upper limit or the tuyere gas temperature is at the lower limit, the operational action determination process proceeds to step S15 because reducing the blast moisture content is preferable to increasing the pulverized coal ratio in order to raise the molten iron temperature. If the reducing agent ratio is below the upper limit and the tuyere gas temperature is above the lower limit, the operational action determination process proceeds to step S13.
• a case where the reducing agent ratio is at the upper limit or the tuyere gas temperature is at the lower limit corresponds to Yes in step S12.
• a case where the reducing agent ratio is below the upper limit and the tuyere gas temperature is above the lower limit corresponds to No in step S12.
• the upper limit of the reducing agent ratio and the lower limit of the tuyere gas temperature may be determined in advance based on, for example, past performance data or experimental data.
• the operational action decision unit 12 determines the difference between the predicted value of the hot metal temperature based on the statistical model calculated by the model calculation unit 11 (statistical model-predicted hot metal temperature, short-time-ahead predicted hot metal temperature) and the target value of the hot metal temperature (target hot metal temperature).
• the target range of the difference between the statistical model-predicted hot metal temperature and the target hot metal temperature is set to be equal to or greater than - ⁇ 2 and equal to or less than ⁇ 2 .
• Step S15 If the statistical model-predicted hot metal temperature ⁇ 2 hours into the future is less than - ⁇ 2 compared to the target hot metal temperature, it is highly likely that the hot metal temperature will suddenly drop, and it is determined that an action to reduce the blast moisture, which has good responsiveness, is desirable, and the operational action determination processing proceeds to step S15. That is, if the long-term future predicted hot metal temperature (Step S1) and the short-term future predicted hot metal temperature (Step S13) both deviate from the target range and the corresponding operational actions are consistent, i.e., the operational actions to increase the hot metal temperature, the process proceeds to Step S15, where the manipulated variables and manipulated variables for the responsive action to reduce the blast moisture are determined based on the short-term future predicted hot metal temperature.
• ⁇ 2 is the future time predicted using the statistical model.
• ⁇ 2 is preferably set to 1 to 4 hours, taking into account the prediction accuracy of the statistical model.
• ⁇ 2 is preferably set to 1 to 30°C depending on the hot metal temperature control target.
• the operational action decision unit 12 determines the action determination value (machine learning model action determination value) based on the machine learning model calculated by the model calculation unit 11. If the machine learning model action determination value is less than ⁇ 3 , the hot metal temperature increase action based on the prediction of a hot metal temperature decrease based on the physical model conflicts with the hot metal temperature decrease action based on the machine learning model, so the operational action decision unit 12 does not execute any proactive action. In other words, if the operational action of decreasing the hot metal temperature corresponding to the machine learning model action determination value conflicts with the operational action of increasing the hot metal temperature corresponding to the predicted hot metal temperature long into the future based on the physical model, no proactive action is executed. Therefore, the series of operational action determination processes ends without any action.
• the action determination value machine learning model action determination value
• the operational action decision unit 12 proposes an action to increase the pulverized coal ratio based on the prediction of the long-term future decrease in hot metal temperature based on the physical model, and the series of operational action judgment processes ends.
• the machine learning model action judgment value is a value obtained by subtracting the calculated probability of judgment of the hot metal temperature increase action from the probability of judgment of the hot metal temperature decrease action, as described above.
• the machine learning model action judgment value is not limited to a specific type of value such as probability, as long as it numerically indicates whether the hot metal temperature increase action or the hot metal temperature decrease action is appropriate.
• machine learning may be performed using the operation amount of the action performed by an exemplary operator obtained as past performance data, and the operation amount itself may be output as the machine learning model action judgment value.
• ⁇ 3 may be determined according to the machine learning model action judgment value, for example, based on past performance data or experimental data.
• the operational action determination unit 12 determines the relationship between the difference between the blast humidity at the current time and the atmospheric humidity and the blast humidity control range determined in consideration of a reduction in the reducing agent rate. If the blast humidity at the current time is smaller than atmospheric humidity + ⁇ (g/Nm 3 ), there is no room to reduce the blast humidity, so it is determined that no action should be taken, and the series of operational action determination processes ends.
• ⁇ is set to a value smaller than the above-mentioned ⁇ .
• the fact that the blast humidity at the current time is smaller than atmospheric humidity + ⁇ (g/Nm 3 ) indicates that the blast humidity is approximately the same as the atmospheric humidity. If the blast humidity at the current time is equal to or greater than atmospheric humidity + ⁇ (g/Nm 3 ), an action to reduce the blast humidity is proposed in order to increase the molten iron temperature, and the series of operational action determination processes ends.
• the operational action determination unit 12 determines the relationship between the difference between the blast humidity at the current time and the atmospheric humidity and the blast humidity control range determined in consideration of a reduction in the reducing agent rate. If the blast humidity at the current time is smaller than atmospheric humidity + ⁇ (g/Nm 3 ), an action to increase the blast humidity is presented in order to lower the molten iron temperature, and the series of operational action determination processes ends. If the blast humidity at the current time is equal to or greater than atmospheric humidity + ⁇ (g/Nm 3 ), the operational action determination process proceeds to step S22 in order to avoid excessive blast humidity.
• the operational action determination unit 12 makes a judgment regarding the reducing agent ratio, tuyere gas temperature, and top gas temperature in relation to the restrictions on pulverized coal use based on operational constraints. If the reducing agent ratio is at the lower limit, the tuyere gas temperature is at the upper limit, or the top gas temperature is at the lower limit, the action of increasing the blast moisture is preferable to lowering the pulverized coal ratio in order to lower the molten iron temperature, and therefore the operational action determination processing proceeds to step S25. If the reducing agent ratio is greater than the lower limit, the tuyere gas temperature is below the upper limit, and the top gas temperature is above the lower limit, the operational action determination processing proceeds to step S23.
• the reducing agent ratio is at the lower limit, the tuyere gas temperature is at the upper limit, or the top gas temperature is at the lower limit, this corresponds to Yes in step S22.
• the reducing agent ratio is greater than the lower limit, the tuyere gas temperature is below the upper limit, and the top gas temperature is above the lower limit, the answer to step S22 is No.
• the lower limit of the reducing agent ratio, the upper limit of the tuyere gas temperature, and the lower limit of the top gas temperature may be determined in advance based on, for example, past performance data or experimental data.
• step S23 the operational action decision unit 12 determines the difference between the statistical model-predicted hot metal temperature calculated by the model calculation unit 11 and the target hot metal temperature. If the statistical model-predicted hot metal temperature ⁇ 2 hours ahead is greater than ⁇ 2 compared to the target hot metal temperature, it is determined that a rapid rise in the hot metal temperature is likely, and therefore an action to increase the blast moisture content, which has good responsiveness, is desirable, and the operational action decision process proceeds to step S25.
• step S1 if the long-term future predicted hot metal temperature (step S1) and the short-term future predicted hot metal temperature (step S23) both deviate from the target range and the corresponding operational actions are consistent, that is, an operational action to decrease the hot metal temperature, the process proceeds to step S25, where the manipulated variables and manipulated variables for an action to increase the blast moisture content, which has good responsiveness, are determined as an operational action based on the short-term future predicted hot metal temperature.
• step S24 If the statistical model predicted hot metal temperature ⁇ 2 hours ahead is equal to or less than ⁇ 2 compared to the target hot metal temperature, that is, if the operational actions corresponding to the predicted hot metal temperature for the long time ahead and the predicted hot metal temperature for the short time ahead do not match, the operational action determination process proceeds to step S24.
• step S24 the operational action decision unit 12 determines the machine learning model action judgment value calculated by the model calculation unit 11. If the machine learning model action judgment value is greater than ⁇ 3 , the hot metal temperature reduction action based on the prediction of the hot metal temperature rise based on the physical model conflicts with the hot metal temperature increase action based on the machine learning model, and therefore the operational action decision unit 12 does not execute any proactive action. In other words, if the operational action corresponding to the machine learning model action judgment value conflicts with the operational action corresponding to the hot metal temperature predicted long-term in the future based on the physical model, no proactive action is executed. Therefore, the series of operational action determination processes ends without any action.
• the operational action based on the machine learning model action judgment value is a hot metal temperature reduction action or a wait-and-see action (no proactive action is executed), and therefore is at least not contradictory to the hot metal temperature reduction action corresponding to the hot metal temperature rise predicted long-term in the future based on the physical model.
• the operational action decision unit 12 will present an action to reduce the pulverized coal ratio based on the prediction of the long-term future molten iron temperature rise based on the physical model, and the series of operational action judgment processes will end.
• step S25 the operational action determination unit 12 determines the relationship between the difference between the blast humidity at the current time and the atmospheric humidity and the blast humidity control range determined in consideration of a reduction in the reducing agent rate. If the blast humidity at the current time is greater than atmospheric humidity + ⁇ (g/Nm 3 ), this indicates that the blast humidity is being added in greater amounts than atmospheric humidity. Therefore, increasing the blast humidity would result in excessive moisture being introduced into the furnace, increasing the risk of a sudden drop in the molten iron temperature. Therefore, if the blast humidity at the current time is greater than atmospheric humidity + ⁇ (g/Nm 3 ), it is determined that no action is to be taken, and the operational action determination process is terminated.
• ⁇ is set to a value smaller than the above-mentioned ⁇ .
• ⁇ may be, for example, the same value as ⁇ . If the blast humidity at the current time is equal to or less than atmospheric humidity + ⁇ (g/Nm 3 ), an action to increase the blast humidity is proposed to lower the molten iron temperature, and the operational action determination process is terminated.
• step S31 the operational action decision unit 12 determines the difference between the statistical model-predicted hot metal temperature calculated by the model calculation unit 11 and the target hot metal temperature.
• step S31 if the physical model determines that the hot metal temperature will not change significantly in the long term (the predicted hot metal temperature in the long term is within the target range), but the statistical model predicts that the hot metal temperature will change suddenly in the short term (the predicted hot metal temperature in the short term deviates from the target range), the statistical model's prediction of the hot metal temperature in the short term is prioritized.
• step S15 If the statistical model-predicted hot metal temperature ⁇ 2 hours into the future is less than ⁇ 2 compared to the target hot metal temperature, it is determined that a sudden drop in the hot metal temperature is likely, and therefore a responsive action to reduce the blast moisture is desirable, and the operational action decision process proceeds to step S15. If the statistical model predicted hot metal temperature ⁇ 2 hours from now is greater than ⁇ 2 compared to the target hot metal temperature, it is determined that there is a high possibility that the hot metal temperature will rise sharply, and therefore it is desirable to take action to increase the blast moisture content, which has good responsiveness, and the operational action determination process proceeds to step S25. If the statistical model predicted hot metal temperature ⁇ 2 hours from now is greater than or equal to ⁇ 2 and less than or equal to ⁇ 2 compared to the target hot metal temperature, the operational action determination process proceeds to step S32.
• step S32 the operational action decision unit 12 determines the machine learning model action judgment value calculated by the model calculation unit 11.
• step S32 if it is determined that the future hot metal temperature will not change significantly in either the physical model or the statistical model, i.e., if the long-term future predicted hot metal temperature based on the physical model and the short-term future predicted hot metal temperature based on the statistical model are both within the target range, the machine learning model action judgment value is prioritized. If the machine learning model action judgment value is greater than ⁇ 3 , the operational action decision unit 12 proposes an action to increase the pulverized coal ratio, and the series of operational action decision processes ends.
• the operational action decision unit 12 proposes an action to decrease the pulverized coal ratio, and the series of operational action decision processes ends. If the machine learning model action judgment value is greater than ⁇ 3 and less than ⁇ 3 , the operational action decision unit 12 determines that no action is necessary, and the series of operational action decision processes ends.
• the operational action determination unit 12 determines the operation variables and operation amounts as operational actions based on the values calculated by the model calculation unit 11, the blast moisture control range determined in consideration of the reduction in reducing agent ratio, and the pulverized coal use restrictions based on operational constraints.
• the values calculated by the model calculation unit 11 include the predicted long-term molten iron temperature calculated using a physical model, the predicted short-term molten iron temperature calculated using a statistical model, and the action determination value calculated using a machine learning model.
• Figure 5 is a diagram showing an example system configuration including a molten iron temperature control device 10.
• the system may be configured to include the molten iron temperature control device 10, an operation data server 94, a control computer 74, and a display unit 95.
• the operation data server 94 accumulates data related to the molten iron production process.
• the molten iron temperature control device 10 acquires actual values and the like in the production process carried out in the molten iron production equipment from the operation data server 94.
• the storage device e.g., memory
• the control computer 74 controls the molten iron production equipment.
• the display unit 95 displays the guidance operation variables (operational actions determined in accordance with the above flowchart) output from the molten iron temperature control device 10, which functions as an operation guidance device.
• the display unit 95 may be a display device such as a liquid crystal display (LCD) or an organic electroluminescence panel (OLED).
• the display unit 95 may also be implemented by a display on a terminal device such as a smartphone or tablet.
• the operation data server 94, control computer 74, and display unit 95 can communicate with the molten iron temperature control device 10 via a network.
• the network may be, for example, the Internet.
• the molten iron temperature control device 10 may execute an operation guidance method that presents the operation actions determined by the above processing so that they can be confirmed by the blast furnace operator. Presentation includes display on the display unit 95 described above, but audio may also be used and is not limited to a specific presentation method.
• the operator may change the operating conditions of the molten iron production equipment based on the guidance operation amounts (operation variables and operation amounts) shown on the display unit 95. Such operation guidance for the molten iron production equipment may be executed as part of a production method for producing molten iron.
• the molten iron temperature control device 10 may output the determined operation variables and operation amounts to the control computer 74, and the control computer 74 may control the molten iron production equipment in accordance with these.
• the content of the operation actions determined by the molten iron temperature control device 10 may be automatically reflected in production or operation, for example, by using a communication function between the control computer 74 and the molten iron production equipment.
• the output and automatic change of such manipulated variables and manipulated quantities may be performed as part of the molten iron production method and operation method.
• the operation data server 94, control computer 74, and display unit 95 may each be located in the same place as the molten iron temperature control device 10 (e.g., in the same factory), or may be physically located separately.
• the molten iron temperature control method executed by the molten iron temperature control device 10 may be part of a blast furnace operation method.
• the blast furnace operation method may include a step of controlling the blast furnace in accordance with an operational action determined by the molten iron temperature control method.
• the molten iron temperature control method or the blast furnace operation method may be part of a molten iron production method.
• the molten iron production method may include a step of controlling the blast furnace in accordance with the blast furnace operation method to produce molten iron.
• Example Operation was carried out by manipulating the pulverized coal ratio and blast moisture using the hot metal temperature control device 10.
• ⁇ 1 and ⁇ 2 were set to 30°C.
• ⁇ 3 was set to 0.3.
• ⁇ was set to 7 g/Nm3.
• ⁇ was set to 0 g/ Nm3 .
• ⁇ was set to 12 g/ Nm3 .
• ⁇ 3 in this example is a probability and is a dimensionless number.
• FIG. 3 shows the changes in molten iron temperature, coke rate, pulverized coal rate, blast moisture, and atmospheric humidity during four days of operation of a blast furnace using the molten iron temperature control device 10 of this embodiment.
• the numbers on the horizontal axis indicate the number of days.
• the molten iron temperature graph in Figure 3 shows the difference from the target molten iron temperature at time zero.
• the other graphs have the average value for the period (average value over four days) set to zero.
• the dotted line represents the target value
• the solid line represents the actual value.
• the blast moisture and atmospheric humidity graphs in Figure 3 the dotted line represents the blast moisture and the solid line represents the atmospheric humidity.
• Figure 4 is a diagram comparing the molten iron temperature variation and reducing agent ratio when a blast furnace operating method using the molten iron temperature control device 10 according to this embodiment is carried out with a conventional operating method by an operator.
• the molten iron temperature variation and reducing agent ratio resulting from the conventional operating method are normalized to 1.
• Figure 4 shows, by using the molten iron temperature control device 10, it is possible to achieve both improved accuracy in molten iron temperature control and a reduced reducing agent ratio compared to the conventional method.
• the molten iron temperature control method, molten iron temperature control device 10, operation guidance method, blast furnace operation method, and molten iron manufacturing method are capable of achieving both a reduction in the reducing agent rate and highly accurate molten iron temperature control.
• by selectively using the execution results of multiple models related to blast furnace operation it is possible to compensate for the shortcomings of each model and implement optimal operational actions, including determining operational variables for molten iron temperature control.
• a process is performed in which hot metal temperature prediction using a physical model is used as a base, and this is supplemented with hot metal temperature prediction using a statistical model and hot metal temperature control action prediction using a machine learning model.
• the order of the flowchart may be rearranged to change the priority of the models that serve as the basis for presenting actions. Examples in which the priority of the models is changed are included in the scope of the technology disclosed herein.
• the above embodiment uses a physical model, a statistical model, and a machine learning model, for example, the machine learning model may be omitted.

Landscapes

• Engineering & Computer Science (AREA)
• Chemical & Material Sciences (AREA)
• Manufacturing & Machinery (AREA)
• Materials Engineering (AREA)
• Metallurgy (AREA)
• Organic Chemistry (AREA)
• Manufacture Of Iron (AREA)

Priority Applications (1)

Applications Claiming Priority (2)

Publications (1)

Family

ID=97063189

Family Applications (1)

Country Status (2)

Citations (2)

• 2025
  • 2025-01-21 WO PCT/JP2025/001767 patent/WO2025192039A1/ja active Pending
  • 2025-01-21 JP JP2025536413A patent/JP7772284B1/ja active Active

Patent Citations (2)

Also Published As

Similar Documents

Legal Events