State prediction device, state prediction method, and program

Abstract

To suppress the deterioration of the quality of a product caused by the change of operation conditions in a manufacturing process.SOLUTION: The state prediction device 400 calculates a predicted value of the coke temperature using the first coke temperature prediction model when an operation condition is changed from an operation condition in an unsteady operation to an operation condition in a steady operation, and then calculates a predicted value of the coke temperature by changing the coke temperature prediction model to be used from the first coke temperature prediction model to the second coke temperature prediction model when a predetermined condition under which the coke production process can be regarded as being in a steady state is satisfied.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] The present disclosure relates to a state prediction device, a state prediction method, and a program. [Background technology]

[0002] The state of a manufacturing process or the state of a product manufactured in the manufacturing process is predicted, and the manufacturing process is controlled based on the predicted state. Patent Document 1 discloses a technology for controlling a coke oven (coke production process) that produces coke by manipulating the heat input so that the coke temperature approaches a target temperature. Specifically, the technology described in Patent Document 1 calculates target oven battery temperatures for multiple future time periods based on an evaluation function including a term representing the difference between the target temperature and a coke temperature predicted by a coke temperature prediction model that predicts the coke temperature based on influencing factors including the oven battery temperature. Furthermore, the technology described in Patent Document 1 calculates the heat input amount using the target oven battery temperature based on an evaluation function including a term representing the difference between the target oven battery temperature and a oven battery temperature predicted by a oven battery temperature prediction model that predicts the oven battery temperature based on influencing factors including the heat input. In the technology described in Patent Document 1, for example, the oven battery temperature corresponds to the state of the production process, and the coke temperature corresponds to the state of the product. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2023-39670 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the present inventors have found that when the technology described in Patent Document 1 is applied to steady-state operation of a coke oven, if a change in operation is made that requires a change in operating conditions, such as when the operation is shifted from steady-state operation to non-steady-state operation for equipment maintenance and then shifted from non-steady-state operation to steady-state operation after the maintenance is completed, there is a risk that the actual coke temperature will deviate from the target temperature during steady-state operation (the details of this will be described later). Therefore, a technology is desired that can prevent a decrease in product quality due to a change in operating conditions in the manufacturing process.

[0005] The present disclosure has been made in consideration of the above-mentioned problems, and aims to prevent a decrease in product quality due to changes in operating conditions in the manufacturing process. [Means for solving the problem]

[0006] The state prediction device disclosed herein is a state prediction device that calculates a predicted value of a state factor that represents the state of a manufacturing process or the state of a product manufactured in the manufacturing process, and includes a calculation unit that calculates the predicted value of the state factor using a prediction model, and the prediction model includes a model that shows the relationship between a dependent variable expressed using the state factor and an explanatory variable expressed using factors that influence the state factor, and the prediction model includes a first prediction model that is used when the manufacturing process is in a transient state immediately after an operating condition is changed, and a second prediction model that is used when the manufacturing process is in a steady state, and when the operating conditions are changed from the first operating conditions to the second operating conditions, the calculation unit calculates the dependent variable using the first prediction model, and thereafter, when a predetermined condition is satisfied that indicates that the manufacturing process has reached a steady state, the calculation unit changes the prediction model that is used from the first prediction model to the second prediction model and calculates the dependent variable.

[0007] The state prediction method disclosed herein is a state prediction method for calculating a predicted value of a state factor that represents a state of a manufacturing process or a state of a product manufactured in the manufacturing process, and includes a calculation step of calculating the predicted value of the state factor using a prediction model, wherein the prediction model includes a model that indicates the relationship between a dependent variable expressed using the state factor and an explanatory variable expressed using factors influencing the state factor, and the prediction model includes a first prediction model that is used when the manufacturing process is in a transient state immediately after an operating condition is changed, and a second prediction model that is used when the manufacturing process is in a steady state, and the calculation step calculates the dependent variable using the first prediction model when the operating conditions are changed from the first operating conditions to the second operating conditions, and thereafter, when a predetermined condition is satisfied that indicates that the manufacturing process has reached a steady state, the prediction model that is used is changed from the first prediction model to the second prediction model to calculate the dependent variable.

[0008] The program disclosed herein causes a computer to function as a calculation unit of the state prediction device. [Effects of the Invention]

[0009] According to the present disclosure, it is possible to prevent deterioration in product quality due to changes in operating conditions in the manufacturing process. [Brief explanation of the drawings]

[0010] [Figure 1] 1 illustrates an example of a coke oven and coke making process. FIG. [Figure 2A] FIG. 2 is a diagram illustrating an example of furnace battery temperature. [Figure 2B] FIG. 10 is a diagram showing an example of coke being pushed out of a carbonization chamber. [Figure 2C] FIG. 1 is a diagram showing an example of the relationship between coke temperature, furnace battery temperature, input heat amount, and carbonization time and time. [Figure 3] FIG. 2 is a diagram showing an example of the relationship between coke temperature and carbonization time. [Figure 4] FIG. 2 is a diagram illustrating an example of a functional configuration of a state prediction device. [Figure 5A] 10 is a flowchart illustrating an example of processing by a target furnace temperature calculation unit. [Figure 5B] 10 is a flowchart illustrating an example of processing by an input heat amount calculation unit. [Figure 6] FIG. 1 is a diagram conceptually illustrating an example of a relationship between a response variable and an explanatory variable of a coke temperature prediction model. DETAILED DESCRIPTION OF THE INVENTION

[0011] Hereinafter, an embodiment of the present disclosure will be described with reference to the drawings. In addition, the comparison of items such as length, position, size, and spacing being the same includes not only items that are strictly the same, but also items that are different within the scope of the present disclosure (for example, items that are different within the tolerance range determined at the time of design).

[0012] [Overview of coke ovens and the coke-making process] Although the manufacturing process of the present disclosure is not limited as described below, in this embodiment, the manufacturing process is a coke manufacturing process as an example. Therefore, first, an overview of the coke oven and the coke manufacturing process will be described.

[0013] FIG. 1 is a diagram illustrating an example of a coke oven and a coke manufacturing process. FIG. 2A is a diagram illustrating an example of a furnace battery temperature. FIG. 2B is a diagram illustrating an example of a state in which coke is being pushed out of a coke chamber. Note that FIGS. 2A and 2B show a see-through view of the interior. As shown in Figures 1 and 2A, in a coke oven 1, carbonization chambers (kilns) 2 and combustion chambers 3 are arranged alternately with a furnace wall 4 interposed therebetween. The carbonization chambers 2 carbonize charged coal to obtain coke. The combustion chambers 3 burn fuel gas to keep the carbonization chambers 2 at a high temperature.

[0014] In the coke production process using a coke oven 1, the so-called block unloading method is used for the unloading coal loading work. The unloading coal loading work is a work in which coke is pushed out of the coke chamber 2 by an extrusion ram 7 as shown in FIG. 2B, and then coal is supplied to the coke chamber 2. In the block unloading method, all coke chambers 2 are divided into Da numbers (Da is an integer greater than or equal to 2), and the unloading coal loading work is carried out in units of the divided numbers. Each coke chamber 2 is assigned to one of the numbers so that multiple coke chambers 2 every Da numbers in the arrangement order of the coke chambers 2 belong to the same number. In this embodiment, an example is shown in which the unloading coal loading work is carried out using the block unloading method with Da being 5. In this case, for example, the carbonization chambers 2 No. 1, 6, 11, 16, etc. are assigned to street 1, the carbonization chambers 2 No. 2, 7, 12, 17, etc. are assigned to street 2, the carbonization chambers 2 No. 3, 8, 13, 18, etc. are assigned to street 3, the carbonization chambers 2 No. 4, 9, 14, 19, etc. are assigned to street 4, and the carbonization chambers 2 No. 5, 10, 15, 20, etc. are assigned to street 5. The carbonization chambers 2 are assigned in order of number, starting with the lowest number, and the carbonization chambers 2 are loaded into the kiln. For example, the unloading coal loading work for the coking chambers 2 assigned to Route 1 is carried out in the following order: coking chamber 2 of coking chamber No. 1, coking chamber 2 of coking chamber No. 6, coking chamber 2 of coking chamber No. 11, coking chamber 2 of coking chamber No. 16, etc. To prevent a sudden drop in temperature, the unloading coal loading order is, for example, Route 1, Route 3, Route 5, Route 2, Route 4. The time from the end of the unloading coal loading work on one route to the end of the unloading coal loading work on the next route is referred to as the "running time." The running time is generally about 3 to 6 hours. Note that the unloading coal loading work is not limited to the block unloading method. For example, if the following description treats each route (block) as an individual coking chamber 2, it can also be applied to cases where the unloading coal loading work is performed on a single coking chamber 2 basis.

[0015] In addition, in the coke oven manufacturing process, furnace battery control is performed to simultaneously adjust the heat input to all combustion chambers 3 and control the average carbonization state for each. That is, the heat input to the coke oven 1 is controlled by operating one regulating valve 5 installed for all combustion chambers 3. The regulating valve 5 is a valve for adjusting the flow rate of a mixture of fuel gas and combustion air. The regulating valve 5 is also operated via an actuator (not shown) under the control of the state prediction device 400. The representative value of the temperature of all combustion chambers 3 is called the furnace battery temperature. For example, as shown in FIG. 2A, thermometers 6 for measuring the ambient temperature of the combustion chambers 3 are installed in multiple combustion chambers 3, and the average temperature of the combustion chambers 3 where the thermometers 6 are installed is defined as the furnace battery temperature. Note that the heat input to all combustion chambers 3 does not necessarily have to be simultaneously adjusted. For example, when performing the unloading and loading work for each coke oven chamber 2, a regulating valve and actuator may be installed in each combustion chamber 3, and the carbonization state (heat input) may be controlled for each coke oven chamber 2.

[0016] Furthermore, the thermometer 6 may be installed in each of all combustion chambers 3, or may be installed in only some of the combustion chambers 3. For example, the thermometer 6 may be installed in all combustion chambers 3, and the temperature of each combustion chamber 3 may be taken as the temperature of that combustion chamber 3 (furnace temperature). As described above, the coke is pushed out of the coke chamber 2 by the pusher ram 7. In the example shown in FIG. 2B, the coke pushed out of the coke chamber 2 by the pusher ram 7 is discharged via the guide car 9 into a fire extinguishing car (not shown) located below the guide car 9, and then transported to a downstream process by the fire extinguishing car. The guide car 9 moves to the position of the coke chamber 2 where the unloading and loading work is performed. In FIG. 2B, the coke 10 produced in the coke chamber 2 located at the bottom of FIG. 2B is discharged via the guide car 9 into a fire extinguishing car (not shown) to complete the unloading and loading work. The guide car 9 moves to the coke chamber 2 located at the top of FIG. 1B, as indicated by the two-dot chain line. In addition, FIG. 2B illustrates a case where a thermometer 8 for non-contact measurement of the temperature of the coke 10 is installed inside the guide car 9. The thermometer 8 is installed so as to view the coke passage inside the guide car 9 through a window provided in the guide car 9. In this manner, this embodiment illustrates a case where the temperature of the coke is measured immediately after it leaves the coke chamber 2 during (at the time of extrusion) the coke extrusion operation (unloading operation). Note that the coke temperature does not necessarily have to be measured in this manner as long as the temperature of the coke extruded from the coke chamber 2 is measured. In the following description, the temperature of the coke that leaves the coke chamber 2 is also referred to as the coke temperature.

[0017] FIG. 2C is a diagram showing an example of the relationship between the coke temperature, the furnace battery temperature, the input heat amount, and the carbonization time. The coke temperature is the temperature of the coke pushed out of the coke chamber 2, and is calculated, for example, from the value measured by the thermometer 8 shown in FIG. 2B. When the extrusion ram 7 pushes the coke out of the coke chamber 2, the temperature of the coke sequentially pushed out of the coke chamber 2 is measured by the thermometer 8, and the representative value of the temperature measured at each time and each position is taken as the temperature of the coke produced in the coke chamber 2. Examples of the representative value include the arithmetic mean value (the sum of the temperatures measured at each time and each position divided by the number of temperature measurements), the median, the mode, and the minimum value. The representative value of the temperatures of the coke produced in the coke chambers 2 belonging to one set is taken as the coke temperature.

[0018] The carbonization time is equal to the time required to carry out the kiln unloading and charging operation once for all Da (e.g., Da=5) ways. As described above, in this embodiment, the case where the coke temperature and the carbonization time are expressed as representative values ​​for each run is exemplified. Therefore, the coke temperature and the carbonization time are obtained when the unloading and loading work for one run is performed. That is, the coke temperature and the carbonization time are obtained in a cycle of the run time. In the graphs of the coke temperature and the carbonization time shown in FIG. 2C (the top and bottom graphs), the interval on the time axis between two plots (●) adjacent to each other on the time axis corresponds to the run time t t In Figure 2C, at time t e As shown in time t t (t e ) is shown as an example. t is generally a constant time, but may vary.

[0019] In this embodiment, the case where Da=5 is illustrated. In this case, the carbonization time is equal to the time required to perform each of the five (=5) patterns of coal loading and unloading from the kiln. Therefore, in the graph of coke temperature and carbonization time shown in FIG. 2C, the distance between the two ends of the six adjacent plots (●) along the time axis is the carbonization time. In FIG. 2C, the time t e Dry distillation time t k (t e ) is shown below.

[0020] Furthermore, during downtime when coke production is temporarily halted, the coke temperature and carbonization time cannot be obtained because the coal loading operation is not performed (see Figure 2C, where there is no plot (●) on the graph of coke temperature and carbonization time during the downtime).

[0021] On the other hand, the oven battery temperature and input heat amount are obtained regardless of the coal unloading operation (see FIG. 2C , where plots (●) are added to the graphs of oven battery temperature and input heat amount even during idle periods). This embodiment illustrates a case in which actual values ​​of oven battery temperature and input heat amount are obtained in the control period of the coke oven 1 (the output period of the control signal from the control unit 514), actual values ​​of coke temperature are obtained in the pass-through period, and the scheduled and actual values ​​of carbonization time are obtained in the pass-through period. This embodiment also illustrates a case in which predicted values ​​of oven battery temperature, input heat amount, and coke temperature are calculated in the control period of the coke oven 1. For simplicity, this embodiment also illustrates a case in which the start or end time of the control period of the coke oven 1 coincides with the start time of one of the pass-through periods. The control period of the coke oven 1 is, for example, one hour, but is not limited to one hour and may be longer or shorter than one hour.

[0022] In FIG. 2C, time t s is an example of the start time of unsteady operation, and time t e is an example of the end time of unsteady operation. Specifically, in this embodiment, the end time of the coke extrusion work Db times (Db is an integer of 1 or more) before the start of the suspension of the coal loading work (charging into the coke chamber 2 and extrusion) is defined as the start time t s For example, when Db is 2 (it is to be noted that Db is not limited to 2), in FIG. 2C, the end time of the coke pushing operation in the second run before the start of the suspension period (the time of the second coke temperature plot counting backward from the start time of the suspension period) is the start time t sWhen the downtime is known in advance, such as when equipment maintenance is performed, Db may be 1 or an integer equal to or greater than 1. On the other hand, when the downtime is not known in advance, such as when an operational abnormality occurs, Db is preferably 1. Note that in FIG. 2C, for convenience of notation, some plots in the graph of coke temperature and carbonization time overlap with the downtime, but these plots are obtained by the kiln unloading and loading work on a run-by-run basis immediately before and after the downtime.

[0023] In this embodiment, the end time of the coke extrusion work in the Da+1th run after the end of the suspension of the coal loading work (charging into the coke chamber 2 and extrusion) is the end time of the unsteady operation t e Here, the start of coke extrusion work in a street refers to the start of coke extrusion work in the coke chamber 2 in which the coke extrusion work is performed first among the coke chambers 2 belonging to that street, and the end of coke extrusion work in a street refers to the end of coke extrusion work in the coke chamber 2 in which the coke extrusion work is performed last among the coke chambers 2 belonging to that street. When Da is 5 (note that Da is not limited to 5), in FIG. 2C, the end time of the coke extrusion work in the sixth street after the end of the suspension period (the time of the sixth coke temperature plot counting forward from the end time of the suspension period) is the end time t e After the rest period ends, in the first through Da (fifth) passes, coke is produced from the coal present in the coke chamber 2 during the rest period.

[0024] On the other hand, in the Da+1th (6th) case after the end of the halt period, coal is charged into the coke chamber 2 after the end of the halt period. It is preferable that the carbonization state of the coke charged into the coke chamber 2 after the end of the halt period approaches the carbonization state in the steady state as quickly as possible. Therefore, in this embodiment, the end time of the coke extrusion work in the Da+1th (6th) case after the end of the halt of the kiln unloading coal charging work (charging coal into the coke chamber 2 and extruding) is set to the end time te That is, the end time of the unsteady operation t e is the end time of the coke extrusion operation on the first run of unloading coal loading work after the end of the shutdown period. However, the end time of unsteady operation is not limited to the end time of the coke extrusion operation on the Da+1-th run after the end of the shutdown of the unloading coal loading work (loading coal into the coke chamber 2 and extrusion). For example, the end time of unsteady operation may be the end time of the coke extrusion operation on the Da+x-th run after the end of the shutdown of the unloading coal loading work (loading coal into the coke chamber 2 and extrusion), and the value of x may be selected from integers equal to or greater than 1. The values ​​of x and Db may be adjusted as appropriate, for example, so that coke of the desired quality is obtained as a result of actually controlling the input heat amount in accordance with the deviation of the actual value of the oven battery temperature from the target oven temperature trajectory, as described below. As described above, the period of non-steady operation (non-steady operation) is from time t s ~t e This will be the period.

[0025] [Issue and Overview] Using the above-described coke production process as an example, an example of a problem to be solved by the present disclosure and an example of an outline of this embodiment will be described. FIG. 3 is a diagram showing an example of the relationship between coke temperature and carbonization time. With reference to FIG. 3, an example of the problem with the technology described in Patent Document 1 will be described. The plots (●) shown in FIG. 3 have the same meaning as the plots (●) in FIG. 2C. In addition, in FIG. 3, the numbers shown beside the plots (●) indicate the number of cycles since the end of the rest period. That is, 1,...,20 shown beside the plots (●) respectively indicate the 1st,...,20th cycles since the end of the rest period. In FIG. 3, T co_a is the target temperature of the coke temperature. As described above, in this embodiment, the case where Da=5 is illustrated. In addition, in this embodiment, the end time of the coke extrusion work in the Da+1th (6th) way after the end of the suspension of the kiln unloading coal loading work (charging coal into the coke chamber 2 and extruding) is set as the end time t e The following example shows the case where:

[0026] The inventors have found that when the technology described in Patent Document 1 is applied during steady-state operation, the coke temperature may exceed the target temperature T due to the fact that the coke production process is in a transient state after transitioning from unsteady operation to steady-state operation. co_a It has been found that the deviation from the normal value increases during non-steady operation. In addition, during non-steady operation, for example, control using the technology described in Japanese Patent No. 2023-39669 may be performed, operation by operator operation may be performed, or a combination of both may be performed.

[0027] As described above, in the first through Da (fifth) sequences after the end of the quiescent period, coke is produced from coal present in the coke chamber 2 during the quiescent period. Furthermore, in the Da+1 (sixth) through Da+1+Da-1 (tenth) sequences after the end of the quiescent period, coke is produced from coal charged in the coke chamber 2 during non-steady operation after the end of the quiescent period. Furthermore, in the Da+Da+1 (eleventh) and subsequent sequences after the end of the quiescent period, coke is produced from coal charged in the coke chamber 2 during steady operation. Thus, in the coke production process, after transitioning from non-steady operation to steady operation, the process goes through a state in which coke is produced from coal charged in the coke chamber 2 during non-steady operation, and then transitions to a state in which coke is produced from coal charged in the coke chamber 2 during steady operation. In this way, when the so-called block unloading method is adopted as the unloading coal loading operation, the coke production process undergoes a transient state during steady operation in which coke is produced from coal loaded in the coke chamber 2 during non-steady operation, and then transitions to a steady state in which coke is produced from coal loaded in the coke chamber 2 during steady operation.

[0028] The inventors considered that when the technology described in Patent Document 1 is applied to steady-state operation of a coke oven, the accuracy of predicting the coke temperature during steady-state operation decreases because, when the coke production process is in a transient state, the coke temperature is predicted on the assumption that the coke production process is in a steady state. In the technology described in Patent Document 1, a coke temperature prediction model for predicting the coke temperature is created by using operational data during the steady state. The technology described in Patent Document 1 shows that the coke temperature prediction model is expressed by a linear regression equation. In this case, the regression coefficients of the linear regression equation are calculated by using operational data during the steady state. Note that, in the technology described in Patent Document 1, the objective variable of the coke temperature prediction model is the amount of change ΔT in the coke temperature from one cycle before (five cycles before) in the carbonization cycle of the same coke oven 2. co (n). The objective variable of the coke temperature prediction model may be expressed by using a physical quantity other than the coke temperature that indicates the carbonization state of the coke as an example of a state factor. An example of such a physical quantity is the temperature of the oven wall 4. The carbonization state of the coke indicates the degree to which the coal in the produced coke has been carbonized (pyrolyzed), and is an index that indicates the quality of the coke. In addition, in the technology described in Patent Document 1, the explanatory variable of the coke temperature prediction model is the oven battery temperature T ro , street time t t , dry distillation time t k , the amount of coal charged S, the change in coal moisture content W Δt t (n), Δt k (n), ΔS(n), and ΔW(n). An example of a coke temperature prediction model will be described later (see equation (1)).

[0029] During the transient state, the carbonization time is extended by the time of the rest period, so Δt k For example, as shown in FIG. 3, in the period n6 corresponding to the sixth run after the end of the rest period, the change in the dry distillation time of the sixth run from the dry distillation time of the first run is Δt k(n6) and becomes a large value (Δt k (n6)). The period n corresponds to the 11th time after the coke production process has transitioned from a transient state to a steady state and the rest period has ended. 11 In this case, the change in the distillation time of the 11th method from the distillation time of the 6th method is Δt k (n 11 ) (Δt in Figure 3) k (n 11 )). In this way, if the operating conditions do not change in a steady state, the coal discharge and charging process (carbonization cycle) is carried out at a roughly constant interval, so normally, Δt k (n) does not become so large, and the time fluctuation is not so large. On the other hand, in the transient state, Δt k (n) may become a large value. For this reason, the inventors considered that if a coke temperature prediction model specialized for the steady state is created, the prediction accuracy of the coke temperature in the transient state will decrease. For example, if a value lower than the actual value is calculated as the predicted value of the coke temperature in the transient state, the calculated manipulated variable (input heat amount) in the transient state will be small.

[0030] The coke production process is a process in which there is a large time delay (a large time constant) between when the manipulated variable is changed and when the change is reflected in the controlled variable. Therefore, the manipulated variable (input heat amount) in the transient state is reflected in the controlled variable (coke temperature) after the steady state is reached. Therefore, in the example shown in Figure 3, the target temperature T co_a The deviation of the coke temperature relative to the

[0031] As described above, the present inventors have discovered that one of the reasons for the reduced accuracy of coke temperature predictions when the technology described in Patent Document 1 is applied to steady-state operation is the use of a coke temperature prediction model specialized for steady-state operation, even in transient states. Therefore, the present inventors developed a first coke temperature prediction model (an example of a first prediction model) for use when the coke production process is in a transient state, and a second coke temperature prediction model (an example of a second prediction model) for use when the coke production process is in a steady state. They then calculated the predicted coke temperature using the first coke temperature prediction model when the coke production process is in a transient state, and calculated the predicted coke temperature using the second coke temperature prediction model when the coke production process is in a steady state. They found that this reduced the deviation of the predicted coke temperature from the target temperature. For example, in the example shown in FIG. 3, the first coke temperature prediction model may be used for the sixth through tenth runs after the end of the shutdown period, and the second coke temperature prediction model may be used for the eleventh and subsequent runs after the end of the shutdown period. In this case, if the current operation pattern is the 11th pattern or later since the end of the shutdown period, the coke temperature prediction model to be used may be changed from the first coke temperature prediction model to the second coke temperature prediction model, assuming that the specified condition is met that the manufacturing process has reached a steady state after steady operation has begun.

[0032] Here, in Figure 3, Δt k This example shows a case where the prediction accuracy of the coke temperature prediction model decreases due to an extremely large value of (n). However, factors other than these may also be involved in the decrease in the prediction accuracy of the coke temperature prediction model. For example, t Since (n) also increases, this may be a factor that reduces the prediction accuracy of the coke temperature prediction model. Therefore, the influencing factors that reduce the prediction accuracy of the coke temperature prediction model are not limited to the carbonization time.

[0033] FIG. 3 illustrates an example in which the prediction model is a coke temperature prediction model that expresses the relationship between a dependent variable expressed using coke temperature, which is an example of a state factor that expresses the state of a product produced in a manufacturing process, and an explanatory variable expressed using carbonization time, which is an example of an influencing factor for the state factor. However, the dependent variable in the prediction model is not limited to being expressed using a state factor that expresses the state of a product produced in a manufacturing process. The dependent variable in the prediction model may also be expressed using a state factor that expresses the state of the manufacturing process. For example, the prediction model may be a furnace battery temperature prediction model (a model for calculating a predicted value of the furnace battery temperature). In this case, the furnace battery temperature is an example of a state factor that expresses the state of the manufacturing process. Note that the furnace battery temperature prediction model created using operational data during steady operation may be, for example, a trained model described in Patent Document 1 (an example of a furnace battery temperature prediction model will be described later (see Equation (6))). Note that, as in the coke temperature prediction model and the furnace battery temperature prediction model described in Patent Document 1, the influencing factors (factors that affect the dependent variable) used as the explanatory variables may be the same type of physical quantity as the state factor used as the dependent variable. In this case, the value of the influential factor used as the explanatory variable may be, for example, a value at a time earlier than that of the state factor used as the objective variable.

[0034] FIG. 3 also illustrates an example in which the change in operating conditions is a change from operating conditions in unsteady operation to operating conditions in steady operation. Also, an example in which the transient state of the coke production process is a transient state of steady operation is illustrated. However, the change in operating conditions and the transient state are not limited to these. For example, when the type of coal (so-called brand) charged into the coke oven 2 is changed, a state may arise in which the coke oven contains both the coke oven 2 charged with the type of coal before the change and the coke oven 2 charged with the type of coal after the change, and then only the coke oven 2 charged with the type of coal after the change is present. In this way, the change in operating conditions may be a change in operating conditions due to a change in the type of coal (raw material). In this case, the state in which the coke oven 1 contains both the coke oven chamber 2 charged with the type of coal before the change and the coke oven chamber 2 charged with the type of coal after the change corresponds to a transient state, and the state in which the coke oven 1 contains only the coke oven chamber 2 charged with the type of coal after the change corresponds to a steady state. In addition to or instead of the type of coal, the moisture content of the coal may also be changed. This may also be a transient state during non-steady operation.

[0035] Also, in FIG. 3, the case where the production process is a coke production process is exemplified. However, the production process is not limited to a coke production process. For example, the production process may be a pig iron production process. In this case, for example, when the operation of a blast furnace is changed from an operation in which blasting is stopped (so-called shutdown operation) to an operation in which blasting is resumed (so-called normal operation), a transient state exists in the operation after the resumption, for example, until the temperature inside the blast furnace recovers to the temperature before the shutdown operation. Therefore, as models for predicting the temperature inside the blast furnace, a first blast furnace temperature prediction model is created as an example of a first prediction model by using operational data when the pig iron production process is in a transient state, and a second blast furnace temperature prediction model is created as an example of a second prediction model by using operational data when the pig iron production process is in a steady state. When the pig iron production process is in a transient state, the first blast furnace temperature prediction model is used to calculate a predicted value of the temperature inside the blast furnace, and when the pig iron production process is in a steady state, the second blast furnace temperature prediction model is used to calculate a predicted value of the temperature inside the blast furnace. In this case, for example, the first blast furnace temperature prediction model may be used until the absolute value of the change in the measured value of the temperature inside the blast furnace per unit time becomes equal to or less than a threshold, and the second blast furnace temperature prediction model may be used when the absolute value of the change in the measured value of the temperature inside the blast furnace per unit time becomes equal to or less than the threshold. In this way, the absolute value of the change in the measured temperature in the blast furnace per unit time falling below a threshold value may be an example of a predetermined condition under which the manufacturing process can be considered to have reached a steady state.

[0036] The manufacturing process may also be a manufacturing process, such as a coke manufacturing process or a pig iron manufacturing process, in which the production of a product in progress under pre-change operating conditions (first operating conditions) can be continued under changed operating conditions (second operating conditions). Operation under the first operating conditions is, for example, unsteady operation in a coke manufacturing process, or shut-off operation in a pig iron manufacturing process. Operation under the second operating conditions is, for example, steady operation in a pig iron manufacturing process, or normal operation in a pig iron manufacturing process. In this case, when the production of a product in progress under the first operating conditions is continued under the second operating conditions, the prediction model used may be changed from the first prediction model (the first coke temperature prediction model and the first blast furnace temperature prediction model in the above-mentioned example) to the second prediction model (the second coke temperature prediction model and the second blast furnace temperature prediction model in the above-mentioned example).

[0037] Furthermore, the manufacturing process does not have to be one in which the production of a product in the middle of being manufactured under the pre-change operating conditions (first operating conditions) can be continued under the post-change operating conditions (second operating conditions). For example, the manufacturing process may be a hot air production process. For example, in a hot air production process, hot air is not supplied (blows) from the hot air stove to the blast furnace when the blast furnace is shut down or during maintenance of the hot air stove. In this case, even in the hot air stove, for example, when the operation is changed from an operation in which blowing air to the blast furnace is stopped (so-called blowing operation) to an operation in which blowing air to the blast furnace is resumed (so-called normal operation), a transient state exists in which the temperature of the silica bricks after the restart recovers to the temperature before the blowing operation was stopped. Therefore, as models for predicting the temperature of silica bricks, a first brick temperature prediction model is created as an example of a first prediction model by using operational data when the hot air production process is in a transient state, and a second brick temperature prediction model is created as an example of a second prediction model by using operational data when the hot air production process is in a steady state. The first brick temperature prediction model is used to calculate a predicted value of the temperature of silica bricks when the hot air production process is in a transient state, and the second brick temperature prediction model is used to calculate a predicted value of the temperature of silica bricks when the hot air production process is in a steady state. This can improve the accuracy of the predicted value of the temperature of silica bricks. In this case, for example, the first brick temperature prediction model may be used until the absolute value of the change in the measured value of the temperature of silica bricks per unit time becomes equal to or less than a threshold, and the second brick temperature prediction model may be used when the absolute value of the change in the measured value of the temperature of silica bricks per unit time becomes equal to or less than the threshold. In this way, the absolute value of the amount of change per unit time of the measured temperature of the silica bricks being equal to or less than a threshold value may be used as a condition for satisfying a predetermined condition for determining that the manufacturing process has reached a steady state.

[0038] Furthermore, the product produced by the manufacturing process may be a solid, liquid, or gas, and the product whose state is to be predicted may be a manufactured product or a product in the middle of being manufactured.

[0039] In the above example, the coke temperature prediction model is a regression equation. However, for example, the coke temperature prediction model may be a trained model other than a regression equation. The trained model may be a trained model such as a neural network, or another prediction model (e.g., a physical model including mathematical expressions such as differential equations that represent physical phenomena). In a manufacturing process, the state of the product or the state of the manufacturing process can generally be obtained as operation data (e.g., actual values ​​or schedule values) (i.e., correct labels can be obtained), so supervised learning can be used. However, for example, when it is difficult to obtain such operation data or when it is difficult to use correct labels, learning may be performed using unsupervised learning or other methods. Note that when a trained model is used as the coke temperature prediction model, the first prediction model (first coke temperature prediction model) may be created by using operation data obtained when the manufacturing process is in a transient state as training data. Furthermore, the second prediction model (second coke temperature prediction model) may be created by using operation data obtained when the manufacturing process is in a steady state as training data. An example of the outline of this embodiment has been described above. Below, an example of the state prediction device and state prediction method of this embodiment will be described, taking as an example a case where the prediction model is a model based on the coke temperature prediction model described in Patent Document 1.

[0040] [State prediction device and state prediction method] FIG. 4 is a diagram showing an example of the functional configuration of a state prediction device 400 according to this embodiment. The state prediction device 400 has, as its hardware, one or more hardware processors such as a CPU (Central Processing Unit) and one or more memories such as a RAM (Random Access Memory) and a ROM (Read Only Memory), and performs various calculations by executing one or more programs stored in the memories using the one or more hardware processors. Furthermore, the state prediction device 400 also has, as its hardware, input devices and output devices. The hardware of the state prediction device 400 may be realized by a PLC (Programmable Logic Controller) or dedicated hardware such as an ASIC (Application Specific Integrated Circuit).

[0041] In FIG. 4 , in this embodiment, the state prediction device 400 includes an input unit 401, a calculation unit 402, and an input heat amount setting unit 403. In this embodiment, the calculation unit 402 further includes a target furnace temperature calculation unit 402a and an input heat amount calculation unit 402b. The state prediction device 400 of this embodiment differs from the technology described in Patent Document 1 mainly in its configuration and processing (specifically, part of the functions of the input heat amount calculation unit 402b) by using, as coke temperature prediction models, a first coke temperature prediction model used when the coke production process is in a transient state and a second coke temperature prediction model used when the coke production process is in a steady state. Therefore, detailed description of parts that can be realized by known technology, as described in Patent Document 1, will be omitted.

[0042] In this embodiment, a case is illustrated in which all information that the state prediction device 400 needs to acquire in advance for processing by the state prediction device 400 is input to the input unit 401. For example, operation data of the cokemaking process is input to the input unit 401. The operation data includes, for example, actual values ​​of operation from the present to the past and schedule values ​​of operation in the future. More specifically, the actual values ​​of operation include data indicating the control content and operation status of the cokemaking process, such as the measurement values ​​of each thermometer 8 and the oven battery temperature, actual values ​​of factors influencing the oven battery temperature such as the input heat amount, and information on the coke before being charged into the coke oven 1. In addition, in this embodiment, a case is illustrated in which the operation data also includes the coke temperature (run average value) obtained from the measurement value of the thermometer 6 as described above. In addition, the operation schedule values ​​include, for example, various schedule values ​​such as the target value (target temperature) of the coke temperature, the availability rate of the coke oven, and the amount of coal charged. Although the case where operation data is input from the storage unit 410 to the input unit 401 is exemplified here, the input unit 401 may receive operation data from an external device via a network, or an operator may directly input operation data to the input unit 401. The state prediction device 400 may include the storage unit 410.

[0043] The calculation unit 402 calculates a predicted value of a state factor that represents the state of the production process or the state of a product produced in the production process, using a prediction model. In this embodiment, a case where the calculation unit 402 calculates a predicted value of the coke temperature using a coke temperature prediction model is illustrated.

[0044] The target oven temperature calculation unit 402a calculates target oven battery temperatures for multiple future pass times (= N × pass times (N is an integer of 2 or more; N = 5 in Figures 1, 2C, and 3)) using a coke temperature prediction model for predicting the coke temperature based on factors influencing the coke temperature, so that the coke temperature measured as described above will be a value corresponding to the target temperature (target value). In this embodiment, the target oven temperature calculation unit 402a has a coke temperature prediction function and a oven battery temperature optimization function. The coke temperature prediction function predicts the coke temperature using the coke temperature prediction model. The oven battery temperature optimization function calculates the target oven battery temperature based on an evaluation function that includes a term representing the difference between the coke temperature predicted by the coke temperature prediction model and the target coke temperature. As described above, in this embodiment, the coke temperature prediction models include a first coke temperature prediction model used when the coke production process is in a transient state and a second coke temperature prediction model used when the coke production process is in a steady state.

[0045] The input heat amount calculation unit 402b calculates the input heat amount according to the target furnace battery temperature calculated by the target furnace temperature calculation unit 402a. In this embodiment, an example is shown in which the input heat amount calculation unit 402b has a furnace battery temperature prediction function and an input heat amount optimization function. The furnace battery temperature prediction function predicts the furnace battery temperature using a furnace battery temperature prediction model that predicts the furnace battery temperature based on factors that influence the furnace battery temperature, including the input heat amount. The input heat amount optimization function calculates the input heat amount using the target furnace battery temperature calculated by the target furnace temperature calculation unit 402a, based on an evaluation function that includes a term representing the difference between the furnace battery temperature predicted by the furnace battery temperature prediction model and the target furnace battery temperature.

[0046] The input heat amount setting unit 403 outputs the input heat amount calculated by the input heat amount calculation unit 402b to a control device of an actuator (not shown) so that the input heat amount is reflected in the coke production process. The control device of the actuator operates the regulating valve 5 (see FIG. 1) via the actuator (not shown) to set the opening of the regulating valve 5 to an opening corresponding to the input heat amount calculated by the input heat amount calculation unit 402b.

[0047] Next, an example of the processing of the target furnace temperature calculation unit 402a in this embodiment will be described with reference to the flowchart in FIG. 5A. In this embodiment, a case will be illustrated in which the flowchart in FIG. 5A is executed during steady operation of the coke production process but not during non-steady operation of the coke production process. The input heat amount in non-steady operation may be calculated, for example, as described in Japanese Patent Application Laid-Open No. 2023-39669, or may be determined by an operator. Whether the operation of the coke production process is steady operation or non-steady operation may be determined, for example, by an operator inputting information indicating that steady operation has started to an input device provided in the state prediction device 400. As described above, in this embodiment, the end time of the coke pushing operation in the Da+1th pass after the end of the suspension of the kiln unloading coal loading operation (charging coal into the coke chamber 2 and pushing) is set to the end time t of the non-steady operation. e The following example illustrates a case where:

[0048] The flowchart of Fig. 5A is executed at each new run time, for example, when a new run time arrives. The timing to start the flowchart of Fig. 5A is the timing when the last unloading operation for each run is completed. However, the timing to start the flowchart of Fig. 5A may also be the timing when the first loading operation for each run begins.

[0049] In step S501, the target furnace temperature calculation unit 402a inputs data on factors influencing the coke temperature for multiple periods from the present to the past for each coke chamber 2 via the input unit 401 (hereinafter referred to as carbonization information data). This carbonization information data may consist of actual values ​​only, or may include schedule values ​​if necessary, such as when actual values ​​are insufficient. The carbonization information data is data that represents the coal characteristics at the time of loading, the state of the coke at the time of extrusion, the furnace core temperature during carbonization, etc., and specifically, the carbonization information data includes, for example, the following data: Coke temperature, carbonization time ·Coal loading amount, coal moisture content Furnace temperature, passage time

[0050] In this embodiment, the coke temperature and carbonization time are obtained at or after the end of the coke unloading operation in each coke oven chamber 2 (the coke temperature is measured when the coke is pushed out, but the final coke temperature is obtained at or after the end of the unloading operation). The coal loading amount is obtained at or after the end of the coal loading operation in each coke oven chamber 2. The coal moisture content is obtained before the start of coal carbonization (before the start of the coal loading operation (e.g., before being transported to the coke oven 1)). The oven battery temperature is the average temperature per pass time of all combustion chambers in which thermometers 6 are installed. In addition, actual values ​​are used for the coke temperature, coal loading amount, coal moisture content, and oven battery temperature in the carbonization information data. However, for example, if the schedule values ​​are highly reliable, scheduled values ​​may be used instead of these actual values. In addition, both actual values ​​and scheduled values ​​are used for the carbonization time and pass time in the carbonization information data.

[0051] Next, in step S502, the target furnace temperature calculation unit 402a generates one or more initial values ​​of the target furnace battery temperature pattern as candidates for the target furnace battery temperature pattern (initial values ​​of candidate solutions) within predetermined constraints. The target furnace battery temperature pattern indicates the time progression of the target furnace battery temperature over multiple consecutive future run times. The first run of the multiple consecutive future run times is preferably the run time following the current run time in which coal loading and unloading operations are being performed. This is because the coke oven 1 has a large time constant (the time delay between when the input heat amount is changed and when that change affects the coke temperature). However, the first run of the multiple consecutive future run times may be the run time two or more times after the current run time. The aforementioned constraints include, for example, upper and lower limits of the furnace battery temperature and upper and lower limits of the amount of change in the furnace battery temperature.

[0052] Next, in step S503, the target furnace temperature calculation unit 402a determines whether a steady-state transition condition is satisfied. The steady-state transition condition is an example of a predetermined condition under which the coke production process can be considered to have reached a steady state after operation under the second operating conditions has started. In this embodiment, a case where the steady-state transition condition is a predetermined condition under which the coke production process can be considered to have reached a steady state after steady operation has started is illustrated. More specifically, in this embodiment, a case where the steady-state transition condition is a condition under which the current operation is the eleventh or subsequent run since the end of the downtime is illustrated. Note that the determination of whether the steady-state transition condition is satisfied may be performed, for example, by an operator inputting information indicating that the current operation is the eleventh or subsequent run since the end of the downtime into an input device provided in the state prediction device 400. Determining whether the steady-state transition condition is satisfied in this manner is preferable because it allows for more accurate determination of whether the coke production process has transitioned from a transient state to a steady state. However, if the first coke temperature prediction model is used during at least a part of the period when the coke production process is in a transient state, the prediction accuracy of the coke temperature during the transient state of the coke production process can be improved compared to the technique described in Patent Document 1. Therefore, for example, the first coke temperature prediction model may be used for a predetermined number of runs after the start of steady-state operation.

[0053] As a result of the judgment in step S503, if the steady state transition condition is not satisfied (No in step S503), the processing of step S504 is performed. In step S504, the target furnace temperature calculation unit 402a calculates a predicted value of the coke temperature of each coke chamber 2 using the first coke temperature prediction model. The predicted value of the coke temperature of each coke chamber 2 is calculated within the range of the period indicated by the target furnace battery temperature pattern (a plurality of consecutive passage times in the future).

[0054] On the other hand, if the result of the judgment in step S503 is that the steady state transition condition is met (Yes in step S503), the processing of step S505 is performed. In step S505, the target furnace temperature calculation unit 402a calculates a predicted value of the coke temperature of each coke chamber 2 using the second coke temperature prediction model. The predicted value of the coke temperature of each coke chamber 2 is calculated within the range of the period (plural consecutive future passage times) indicated by the target furnace battery temperature pattern.

[0055] In this embodiment, a case will be exemplified in which the first coke temperature prediction model and the second coke temperature prediction model are expressed by the following formulas (1) and (2), respectively.

[0056]

number

[0057] In equation (1), which is an example of the first coke temperature prediction model, Δt k The regression coefficient for (n) is (c-C'). On the other hand, in equation (2), which is an example of the second coke temperature prediction model, Δt k The regression coefficient for (n) is c. This is the only difference between equation (1) and equation (2). Note that equation (2) is a coke temperature prediction model exemplified in Patent Document 1.

[0058] The coke temperature prediction model exemplified in equations (1) and (2) is a regression model (regression equation) based on the physical phenomenon of a single coke chamber 2. In equations (1) and (2), T co is the coke temperature [℃], T ro is the furnace temperature [℃], t t is the time [hr], t krepresents the carbonization time [hr], S represents the amount of coal charged [tons], and W represents the coal moisture content [%]. Furthermore, nj is a variable specifying a period obtained by dividing the carbonization cycle by the number of passes, and each period is the same length as the pass time. In this embodiment, since coal is discharged from the kiln and charged at intervals of five kilns, the period from charging coal in one coke chamber 2 to extrusion is divided into five pass times (five periods from n to n-4 and five periods from n-5 to n-9). When the values ​​of the two variables nj are consecutive values, this indicates that the periods specified by the two variables are consecutive periods (for example, period n and period n-1 are consecutive periods). For consecutive periods n and n-1, the kiln loading sequence is 1, 3, 5, 2, and 4, and the kiln loading sequence for sequence 1 is carried out in period n. To explain this more specifically, period n is the period in which the kiln loading sequence for sequence 1 is carried out, and is the same length as the loading sequence for sequence 1. Furthermore, period n-1 is the period in which the kiln loading sequence for sequence 4 is carried out, and is the same length as the loading sequence for sequence 4. Hereinafter, the sequence in which the kiln loading sequence for sequence nj is carried out will be referred to as the sequence corresponding to period nj.

[0059] ΔT, which is the response variable in equations (1) and (2), co (n), ΔT as explanatory variable ro (n), Δt t (n), Δt k (n), ΔS(n), and ΔW(n) are the coke temperature T co , furnace temperature T ro , street time t t , dry distillation time t k , and is expressed as the change in the coal load S and coal moisture content W from the previous cycle (five cycles ago) in the carbonization cycle of the same carbonization chamber 2.

[0060] Table 1 shows the relationship between explanatory variables (○) and response variables (★) of the coke temperature prediction models (first coke temperature prediction model and second coke temperature prediction model) used in this embodiment. The response variables are calculated by the first coke temperature prediction model and the second coke temperature prediction model using operational data during carbonization as explanatory variables. In detail, the values ​​for each pass from coal loading to pushing are used for the furnace battery temperature and pass time. The values ​​for each carbonization cycle given at the time of the first coal loading in the carbonization cycle, etc., are used for the carbonization time, coal loading amount, and coal moisture content.

[0061] [Table 1]

[0062] Coefficient a in equation (2) j , b j , c, d, and e are explanatory variables, respectively. ro (nj), Δt t (nj), Δt k (n), ΔS(n), and ΔW(n). Regression coefficient a i , b i , c, d, and e are the coefficients when the form of equation (2) best matches the steady-state operation results of the coke oven 1 from among the past operation results of the coke oven 1. For example, a set of ΔT co (n), ΔT ro (nj), Δt t (nj), Δt k (n), ΔS(n), and ΔW(n) are used as training data to create a large number of training data, and multiple regression analysis is performed using the training data to obtain the coefficient a j , b j , c, d, e. In this case, ΔT co(n) is the correct label. In Figure 3, the operational results at the boundary between the end of the transient state and the start of the steady state are not included in the operational results during the steady state, but are included in the operational results during the transient state. Specifically, in the example shown in Figure 3, the operational results of the 10th sequence after the end of the downtime period are included in the operational results during the transient state.

[0063] In equation (1), Δt k The regression coefficient (c-C') for (n) is the Δt in equation (2). k (n) is the same as the regression coefficient c. k The regression coefficient (c-C') for (n) is -C', which is the Δt k corresponds to the amount of change to the regression coefficient c for (n).

[0064] Fig. 6 shows the objective variable ΔT co (n) and Δt, one of the explanatory variables of the coke temperature prediction model. k FIG. 10 is a diagram conceptually illustrating an example of the relationship between (n) and As mentioned above, during the transient state, the carbonization time is extended by the time of the rest period, so Δt k On the other hand, in a steady state, if the operating conditions do not change, the coal discharge and charging process (carbonization cycle) is carried out at a roughly constant interval, so Δt k (n) does not become a very large value, and the time fluctuation is not so large. Therefore, during the transient state, Δt k (n) and Δt k The relationship between Δt (n) and Δt (n) is as shown in graph 610. k (n) and Δt k The relationship with (n) is considered to be as shown in graph 620.

[0065] In this case, in the example shown in FIG. 6, the slope of the graph 620 is Δt kOn the other hand, the slope of graph 610 is smaller than the slope of graph 620, so in the transient state, Δt k The regression coefficient for (n) is (theoretically) smaller than c. This is one of the main reasons for the decrease in the accuracy of coke temperature prediction when the second coke temperature prediction model, which is created using operational data from constant-rate operation, is used in a transient state, as in equation (2).

[0066] Equation (1) is based on the idea of ​​minimizing the changes to the second coke temperature prediction model, and is used to estimate the coke temperature prediction error as follows: Δt k This is an example of a first coke temperature prediction model when it is assumed that the error is caused only by an error in (n) (carbonization time). When calculating C' shown in equation (1), for example, the regression coefficient C' is calculated as the coefficient when the form of the following equation (3) best matches the operation results during a transient state among the past operation results of the coke oven 1. For example, a set of ΔT obtained from the operation results during a transient state and the calculation results of equation (2) among the past operation results of the coke oven 1 and the calculation results of equation (2) is calculated as follows: co_2 (n)-ΔT co_r (n) and Δt k A large number of training data sets are created using the data in (n) as one training data set, and the coefficient C' is calculated by performing a simple regression analysis using the training data. In this case, the ΔT co_2 (n)-ΔT co_r (n) is the correct label.

[0067]

number

[0068] In equation (3), ΔT co_2 (n) is the ΔT calculated by equation (2) co (n). ΔT co_r (n) is ΔT co (n) is the actual value. coThe actual value of (n) is the change in the actual value of the coke temperature of the coke pushed out of a certain carbonization chamber 2 in the manner corresponding to the period n from the actual value of the coke temperature of the coke pushed out of the carbonization chamber 2 in the manner five manners before that (one cycle before in the carbonization cycle).

[0069] As mentioned in the section on [Issues and Overview], for example, Δt t (n) also becomes larger. Therefore, Δt t (n) can also be a factor in the deterioration of the prediction accuracy of the coke temperature prediction model when the second coke temperature prediction model is used in a transient state. co Instead of or in addition to (n), Δt shown in equation (2) t Regression coefficient b for (nj) j Change amount B j ' is calculated, and b in equation (2) is j (b j -B j '), the first coke temperature prediction model may be created.

[0070] In addition, for example, ΔT obtained from the operation results during a transient state among the past operation results of the coke oven 1 co (n), ΔT ro (nj), Δt t (nj), Δt k A set of data for (n), ΔS(n), and ΔW(n) is used as training data to create a large number of training data, and multiple regression analysis is performed using the training data to obtain ΔT ro (nj), Δt t (nj), Δt k The regression coefficients for (n), ΔS(n), and ΔW(n) may be calculated (all at once).

[0071] In step S504, the target furnace temperature calculation unit 402a may calculate the predicted value of the coke temperature of each coke chamber 2 by calculating the right side of the coke temperature prediction model of equation (1) using a candidate target furnace chamber temperature pattern (the initial value of the target furnace chamber temperature pattern generated in step S502 or the target furnace chamber temperature pattern generated in step S509 described below) and the carbonization information data (carbonization time, coal loading amount, coal moisture content, furnace chamber temperature, and pass time) imported in step S501, and calculating the value of the left side of the coke temperature prediction model of equation (1). Similarly, in step S505, the target furnace temperature calculation unit 402a may calculate the predicted value of the coke temperature of each coke chamber 2 by calculating the right side of the coke temperature prediction model of equation (2) using a candidate target furnace chamber temperature pattern (the initial value of the target furnace chamber temperature pattern generated in step S502 or the target furnace chamber temperature pattern generated in step S509 described below) and the carbonization information data (carbonization time, coal loading amount, coal moisture content, furnace chamber temperature, and pass time) imported in step S501, and calculating the value of the left side of the coke temperature prediction model of equation (2).

[0072] As described above in the section [Issues and Overview], the first coke temperature prediction model and the second coke temperature prediction model are not limited to regression equations. For example, one or both of the first coke temperature prediction model and the second coke temperature prediction model may be a neural network. The input unit to which the learning data is input and the model creation unit that creates a prediction model (e.g., the first coke temperature prediction model and the second coke temperature prediction model in this embodiment) using the learning data may be included in the state prediction device 400 or may be included in a device different from the state prediction device 400. The input unit to which the learning data is input may be implemented by the input unit 401. When a trained model is created by a device different from the state prediction device 400, the state prediction device 400 acquires information about the trained model. The input unit 401 may acquire the information about the trained model.

[0073] Next, in step S506, the target furnace temperature calculation unit 402a converts the predicted value of the coke temperature of each coke chamber 2 predicted in step S504 or S505 into a run average value, which is an average value for each run. In the following description, the run average value of the predicted value of the coke temperature of each coke chamber 2 will be referred to as the predicted value of the coke temperature of each coke chamber 2 for each run or the predicted value of the coke temperature for each run, as necessary. The predicted value of the coke temperature of each coke chamber 2 for each run is calculated by adding together the predicted values ​​of the coke temperature of each coke chamber 2 calculated in step S504 or S505 for multiple coke chambers 2 (13 in the example of FIG. 1) belonging to the run where extrusion is performed, and dividing the sum by the number of coke chambers 2 belonging to the run. Note that instead of the average value for each run, for example, the minimum value for each run may be used.

[0074] Next, in step S507, the target furnace temperature calculation unit 402a calculates the following evaluation function J1 of (4). J1 = (target coke temperature - predicted value for each coke temperature) + (lower limit of coke temperature) + (variation in target battery temperature) + (target battery temperature) (4)

[0075] The second term on the right side of equation (4) is provided to ensure that the coke temperature satisfies the preset lower limit. The third term on the right side of equation (4) is provided to prevent the target battery temperature from changing too much. The fourth term on the right side of equation (4) is provided to prevent the target battery temperature from becoming too high. For example, the first term on the right side of the evaluation function J1 may be calculated using the difference between the predicted value of the coke temperature of each coke chamber 2 calculated in step S504 or S505 and the target temperature of the coke temperature in each coke chamber 2.

[0076] Next, in step S508, the target furnace temperature calculation unit 402a determines whether or not a calculation termination condition has been reached. For example, the calculation termination condition may be that a predetermined number of repeated calculations has been reached. Alternatively, the calculation termination condition may be that the value of the evaluation function J1 has converged during the repeated calculations. If the determination in step S508 shows that the calculation termination condition has not been reached (No in step S508), the process proceeds to step S509.

[0077] In step S509, the target furnace temperature calculation unit 402a generates one or more new target furnace battery temperature patterns (candidate solutions for the target furnace battery temperature pattern) as candidates for the target furnace battery temperature pattern. Then, the processing of step S503 described above is performed again. Note that it is preferable that the number of initial values ​​of the target furnace battery temperature pattern generated in step S502 is the same as the number of target furnace battery temperature patterns newly generated in step S509. For example, based on the metaheuristic algorithm FPA (Flower Pollination Algorithm), a target furnace battery temperature pattern that reduces the evaluation function J1 is searched for, and 50 new target furnace battery temperature patterns are generated. Note that instead of FPA, optimization methods such as GA (Genetic Algorithm) and PSO (Particle Swarm Optimization) may be used as the optimization method (algorithm for solving the optimization problem).

[0078] The processing of steps S503 to S509 is repeated until it is determined in step S508 that the calculation termination condition has been met. If the determination in step S508 is that the calculation termination condition has been met (Yes in step S508), the processing of step S510 is performed. In step S510, the target furnace temperature calculation unit 402a outputs the target furnace battery temperature pattern that minimizes the evaluation function J1 from among the candidate target furnace battery temperature patterns to the input heat amount calculation unit 402b. Note that, for example, if the evaluation function J1 is obtained by multiplying each term on the right side of equation (4) by (-1), the target furnace temperature calculation unit 402a searches for the target furnace battery temperature pattern that maximizes the evaluation function J1.

[0079] Next, an example of the processing of the input heat amount calculation unit 402b in this embodiment will be described with reference to the flowchart of Fig. 5B. The flowchart of Fig. 5B is executed, for example, at a control period (here, for example, one hour period) of the input heat amount to the coke oven 1.

[0080] In step S521, the input heat amount calculation unit 402b inputs data on factors influencing the oven battery temperature for multiple consecutive periods of one hour from the present to the past for each coke chamber 2 via the input unit 401 (hereinafter referred to as combustion information data) (in the example shown in Table 2 below, each period is one hour). This combustion information data may consist of only actual values, or may include scheduled values ​​if necessary, such as when actual values ​​are insufficient. The combustion information data is data that represents the amount of heat input to the coke oven and the characteristics of the coal at the time of loading, and specifically includes, for example, the oven battery temperature, the input heat amount, the total amount of coal loaded in all coke chambers 2, and the coal moisture content.

[0081] Next, in step S522, the input heat amount calculation unit 402b calculates the input heat amount per hour that minimizes the evaluation function J2 of the following equation (5). J2 = (target furnace battery temperature - predicted furnace battery temperature) 2 + (Change in input heat) 2 ···(5)

[0082] In this embodiment, the furnace battery temperature prediction model can be expressed linearly as in the following equation (6), and an example is given in which the input heat amount per hour is calculated using generalized predictive control (GPC) to minimize the evaluation function J2 in equation (5).

[0083]

number

[0084] The furnace battery temperature prediction model is a model that predicts the furnace battery temperature based on factors that influence the furnace battery temperature. Factors that influence the furnace battery temperature include, for example, the input heat amount. In this embodiment, as shown in equation (6), a regression model is exemplified that includes a furnace battery temperature that is older than the furnace battery temperature to be predicted as an explanatory variable.

[0085] In equation (6), T ro represents the furnace temperature [℃], Q represents the input heat [GJ / h], S represents the amount of coal charged [tons], and W represents the coal moisture content [%]. Also, t represents the time a specified time ahead of the current time (in the following explanation, the specified time is assumed to be 1 hour). The objective variable ΔT ro (t), ΔQ(ti) and ΔT as explanatory variables ro (ti), ΔS(ti), and ΔW(ti) are the furnace battery temperatures T ro , input heat Q, furnace temperature T ro , the amount of coal charged S, and the amount of coal moisture W are expressed as the change from one hour before.

[0086] Table 2 shows the relationship between the explanatory variables (○) and the objective variables (★) of the furnace battery temperature prediction model used in this embodiment. In order to take into account the time delay, the actual values ​​for multiple consecutive periods in one-hour cycles (furnace battery temperature T ro For all other cases, the schedule value for one hour ahead is used as the explanatory variable, and the furnace battery temperature prediction model predicts the furnace battery temperature for one hour ahead, which is the target variable. In detail, each item marked with a circle from one hour ahead (t) to five hours ahead (t-6) is used as an explanatory variable. Since the furnace battery temperature is subject to periodic fluctuations caused by operation, it is preferable to apply a moving average filter as preprocessing to suppress periodic fluctuations.

[0087] [Table 2]

[0088] Coefficient a in equation (6) i , b i , c i , d i are the explanatory variables ΔQ(ti) and ΔT ro(ti), ΔS(ti), and ΔW(ti). Regression coefficient a i , b i , c i , d i The coefficients that best fit the form of equation (6) to the past operation results of the coke oven 1 are calculated separately. For example, a set of ΔT ro (t), ΔQ(ti), ΔT ro (ti), ΔS(ti), and ΔW(ti) are used as training data to create a large number of training data, and multiple regression analysis is performed using the training data to obtain the coefficient a i , b i , c i , d i In this case, ΔT included in the training data ro (t) is the correct label.

[0089] The coal amount S is the total value of the values ​​in all the coking chambers 2 included in the coke oven 1 to be predicted. The coal moisture content W is the average value of the values ​​in all the coking chambers 2 included in the coke oven 1 to be predicted. The coal moisture content W is expressed, for example, as the mass ratio (mass%) of coal. In addition, the explanatory variables ΔQ(ti) and ΔT ro When calculating ΔT(ti), ΔS(ti), and ΔW(ti), actual values ​​are used for past values, and for future values, schedule values ​​(values ​​determined in the operation schedule) or values ​​already calculated using equation (6) are used. ro (t) is the furnace battery temperature T ro (t) is used. By updating the value of t every hour, the furnace battery temperature T ro (t) is calculated.

[0090] The combustion information data acquired in step S521 is used as input data for the furnace battery temperature prediction model. The target value of the GPC is the target furnace battery temperature represented by the target furnace battery temperature pattern calculated by the target furnace temperature calculation unit 402a as described in the flowchart of Figure 5A.

[0091] The first term on the right side of the evaluation function J2 in equation (5) represents the difference between the predicted value of the furnace battery temperature calculated using the furnace battery temperature prediction model and the target furnace battery temperature represented by the target furnace battery temperature pattern calculated by the target furnace temperature calculation unit 402a as explained in the flowchart of Figure 5A. This term makes it possible to calculate the input heat amount that realizes the target furnace battery temperature pattern calculated by the target furnace temperature calculation unit 402a. Note that the second term on the right side of equation (5) is a term added to prevent the amount of change in the input heat amount controlled by operating the control valve 5 from becoming too large, in this case, between the input heat amount between adjacent time periods.

[0092] In GPC, the furnace battery temperature prediction model in equation (6) and the evaluation function J2 in equation (5) are written in vector form, and the vector-format furnace battery temperature prediction model is substituted into the vector-format evaluation function. By partially differentiating the evaluation function obtained in this way with respect to the input heat amount, the input heat amount that minimizes the evaluation function J2 in equation (5) is calculated.

[0093] In step S523, the input heat amount calculation unit 402b outputs the input heat amount that minimizes the evaluation function J2 to the input heat amount setting unit 403. In response to this, the input heat amount setting unit 403 outputs the input heat amount calculated by the input heat amount calculation unit 402b to a control device of an actuator (not shown) so that the input heat amount is reflected in the coke production process. The control device of the actuator operates the adjustment valve 5 (see FIG. 1) via an actuator (not shown) to set the aperture of the adjustment valve 5 to an aperture corresponding to the input heat amount calculated by the input heat amount calculation unit 402b. Note that, for example, when the evaluation function J2 is obtained by multiplying each term on the right side of equation (5) by (−1), the input heat amount calculation unit 402b calculates the input heat amount that maximizes the evaluation function J2.

[0094] As described above in the section [Issues and Overview], the furnace battery temperature prediction model may also include a first furnace battery temperature prediction model created using operational data when the coke production process is in a transient state, and a second furnace battery temperature prediction model created using operational data when the coke production process is in a steady state, similar to the coke temperature prediction model. In this case, for example, a set of ΔT ro (t), ΔQ(ti), ΔT ro (ti), ΔS(ti), and ΔW(ti) are used as training data to create a large number of training data, and multiple regression analysis is performed using the training data to obtain the coefficient a i , b i , c i , d i Similarly, a set of ΔT ro (t), ΔQ(ti), ΔT ro (ti), ΔS(ti), and ΔW(ti) are used as training data to create a large number of training data, and multiple regression analysis is performed using the training data to obtain the coefficient a i , b i , c i , d i The second furnace battery temperature prediction model may be created by calculating the following. Furthermore, the furnace battery temperature prediction models (first furnace battery temperature prediction model and second furnace battery temperature prediction model) are not limited to regression equations. For example, one or both of the first furnace battery temperature prediction model and the second furnace battery temperature prediction model may be a neural network or the like.

[0095] In this embodiment, the input heat amount calculation unit 402b uses GPC to control the input heat amount, but this is not limited to this. For example, as in the flowchart of Figure 5A, it is possible to provide candidates for input heat amounts and search for the input heat amount that minimizes the evaluation function J2. Alternatively, it is possible to perform PID control to change the input heat amount so that the actual furnace battery temperature becomes a value corresponding to the target furnace battery temperature pattern calculated in the target furnace temperature calculation unit 402a.

[0096] Table 3 shows an example of the coke temperature prediction results. In Table 3, in the example of the invention, during unsteady operation, the predicted coke temperature was calculated using the first coke temperature prediction model during the transient state, and the predicted coke temperature was calculated using the second coke temperature prediction model during the steady state. In the comparative example, the predicted coke temperature was calculated using the second coke temperature prediction model throughout the entire period of unsteady operation (during both the transient and steady states). The error mean (absolute value) represents the calculated average absolute value of the difference between the actual coke temperature for each coke temperature run from the 11th run to the 20th run after the end of the shutdown period and the target coke temperature. The error standard deviation represents the standard deviation of the difference between the actual coke temperature for each coke temperature run from the 11th run to the 20th run after the end of the shutdown period and the target coke temperature.

[0097] [Table 3]

[0098] As shown in Table 3, during transient operation, the predicted value of the coke temperature is calculated using the first coke temperature prediction model created using operational data when the coke production process is in a transient state, and during steady operation, the predicted value of the coke temperature is calculated using the second coke temperature prediction model created using operational data when the coke production process is in a steady state as training data.This shows that both the absolute value and variability of the prediction error of the coke temperature can be suppressed.

[0099] [summary] As described above, in this embodiment, when the operating conditions are changed from those for unsteady operation to those for steady operation, the state prediction device 400 calculates a predicted value of the coke temperature using the first coke temperature prediction model, and then, when a predetermined condition is satisfied that indicates that the coke production process has reached a steady state, the state prediction device 400 changes the coke temperature prediction model used from the first coke temperature prediction model to the second coke temperature prediction model and calculates a predicted value of the coke temperature. Therefore, it is possible to prevent a decrease in the quality of the coke, which is a product, due to a change in the operating conditions in the coke production process.

[0100] Furthermore, in this embodiment, when coke production in the middle of unsteady operation is continued in steady operation, the state prediction device 400 changes the coke temperature prediction model to be used from the first coke temperature prediction model to the second coke temperature prediction model to calculate the predicted value of the coke temperature. Therefore, it is possible to reduce prediction errors in the coke temperature that occur during steady operation due to the presence of coke in the coke chamber 2 during unsteady operation.

[0101] Furthermore, in this embodiment, when the production volume is increased during steady operation after temporarily reducing the production volume during a suspension period in unsteady operation, the state prediction device 400 changes the coke temperature prediction model used from the first coke temperature prediction model to the second coke temperature prediction model to calculate a predicted value of the coke temperature. Therefore, it is possible to reduce prediction errors in the coke temperature that occur during steady operation due to the temporary suspension of coke production during unsteady operation. Note that reducing the production volume includes both setting the production volume to a value greater than 0 and setting it to 0.

[0102] In this embodiment, the first prediction model is a trained model created by using, as training data, operation data obtained when the coke production process is in a transient state. The second prediction model is a trained model created by using, as training data, operation data obtained when the coke production process is in a steady state. Therefore, the first prediction model for a transient state and the second prediction model for a steady operation can be realized as prediction models with high prediction accuracy without using mathematical formulas that represent physical phenomena.

[0103] (Other embodiments) The above-described embodiments of the present disclosure can be realized by a computer executing a program. A computer-readable recording medium on which the program is recorded and a computer program product such as the program can also be applied as embodiments of the present disclosure. Examples of recording media that can be used include flexible disks, hard disks, optical disks, magneto-optical disks, CD-ROMs, magnetic tapes, non-volatile memory cards, and ROMs. Furthermore, the above-described embodiments of the present disclosure are merely examples of specific embodiments for carrying out the present disclosure, and the technical scope of the present disclosure should not be interpreted as being limited by these. In other words, the present disclosure can be embodied in various forms without departing from its technical concept or main features.

[0104] The disclosure of the above embodiment can be implemented as follows, for example. [Disclosure 1] A state prediction device that calculates a predicted value of a state factor that represents a state of a manufacturing process or a state of a product manufactured in the manufacturing process, a calculation unit that calculates a predicted value of the state factor using a prediction model, the prediction model includes a model showing a relationship between a dependent variable expressed using the state factor and an explanatory variable expressed using an influencing factor on the state factor, the prediction models include a first prediction model used when the manufacturing process is in a transient state immediately after an operating condition is changed, and a second prediction model used when the manufacturing process is in a steady state; the calculation unit calculates the dependent variable using the first prediction model when the operating conditions are changed from first operating conditions to second operating conditions, and thereafter, when a predetermined condition is satisfied that indicates that the manufacturing process has reached a steady state, the calculation unit changes the prediction model to be used from the first prediction model to the second prediction model and calculates the dependent variable. [Disclosure 2] The state prediction device described in Disclosure 1, wherein the calculation unit changes the prediction model to be used from the first prediction model to the second prediction model and calculates the dependent variable when production of a product in the middle of production under the first operating conditions is continued under the second operating conditions. [Disclosure 3] the period during which the operation is performed under the first operating conditions includes a period during which the production volume of the product is temporarily reduced; The state prediction device according to Disclosure 1 or 2, wherein the period during which operation is performed under the second operating conditions includes a period during which the production volume of the product is increased compared to the period during which operation is performed under the first operating conditions. [Disclosure 4] the first operating condition includes performing unsteady operation, The state prediction device according to any one of Disclosures 1 to 3, wherein the second operating condition includes performing steady operation. [Disclosure 5] The predictive model includes a trained model, the first prediction model is created by using operational data when the manufacturing process is in the transient state; The state prediction device according to any one of Disclosures 1 to 4, wherein the second prediction model is created by using operation data when the manufacturing process is in the transient state. [Disclosure 6] the manufacturing process comprises a coke manufacturing process; 6. The state prediction device according to any one of Disclosures 1 to 5, wherein the state factors include a physical quantity representing a carbonization state of coke or a physical quantity representing a state of a coke oven. [Disclosure 7] 1. A state prediction method for calculating a predicted value of a state factor representing a state of a manufacturing process or a state of a product manufactured in the manufacturing process, comprising: a calculation step of calculating a predicted value of the state factor using a prediction model, the prediction model includes a model showing a relationship between a dependent variable expressed using the state factor and an explanatory variable expressed using an influencing factor on the state factor, the prediction models include a first prediction model used when the manufacturing process is in a transient state immediately after an operating condition is changed, and a second prediction model used when the manufacturing process is in a steady state; the calculating step calculates the dependent variable using the first prediction model when the operating conditions are changed from first operating conditions to second operating conditions, and thereafter, when a predetermined condition is satisfied under which the manufacturing process can be considered to have reached a steady state, the prediction model used is changed from the first prediction model to the second prediction model to calculate the dependent variable. [Disclosure 8] A program for causing a computer to function as a calculation unit of the state prediction device according to any one of Disclosures 1 to 6. [Explanation of symbols]

[0105] 1. Coke oven 2. Carbonization chamber 3 Combustion chamber 4 Furnace wall 5. Regulating valve 6. Thermometer for measuring furnace temperature 7 Extrusion Ram 8. Thermometer for measuring coke temperature 9 Guide car 10. Coke 400 State Prediction Device 401 Input section 402 Calculation Unit 402a Target furnace temperature calculation section 402b Input heat amount calculation section 403 Input heat amount setting section 410 Storage section 610 Transient state graph 620 Steady state graph t k Dry distillation time t t Street time t s Start time of non-routine operation t e End time of non-routine operation T co_a Target coke temperature Δt k (n) Change in carbonization time (explanatory variable) ΔT co (n) Change in coke temperature (objective variable)

Claims

1. A state prediction device that calculates a predicted value of a state factor that represents a state of a manufacturing process or a state of a product manufactured in the manufacturing process, a calculation unit that calculates a predicted value of the state factor using a prediction model, the prediction model includes a model showing a relationship between a dependent variable expressed using the state factor and an explanatory variable expressed using an influencing factor on the state factor, the prediction models include a first prediction model used when the manufacturing process is in a transient state immediately after an operating condition is changed, and a second prediction model used when the manufacturing process is in a steady state; the calculation unit calculates the dependent variable using the first prediction model when the operating conditions are changed from first operating conditions to second operating conditions, and thereafter, when a predetermined condition is satisfied that indicates that the manufacturing process has reached a steady state, the calculation unit changes the prediction model to be used from the first prediction model to the second prediction model and calculates the dependent variable.

2. 2. The state prediction device according to claim 1, wherein, when production of a product in the middle of production under the first operating conditions is continued under the second operating conditions, the calculation unit changes the prediction model to be used from the first prediction model to the second prediction model and calculates the dependent variable.

3. the period during which the operation is performed under the first operating conditions includes a period during which the production amount of the product is temporarily reduced; 3. The state prediction device according to claim 1, wherein the period during which the operation is performed under the second operating condition includes a period during which the production volume of the product is increased compared to the period during which the operation is performed under the first operating condition.

4. the first operating condition includes performing unsteady operation, The state prediction device according to claim 1 or 2, wherein the second operating condition includes performing steady operation.

5. The predictive model includes a trained model, the first predictive model is created by using operational data when the manufacturing process is in the transient state; The state prediction device according to claim 1 or 2, wherein the second prediction model is created by using operational data when the manufacturing process is in the transient state.

6. the manufacturing process comprises a coke manufacturing process; The state prediction device according to claim 1 or 2, wherein the state factors include a physical quantity representing a carbonization state of coke or a physical quantity representing a state of a coke oven.

7. 1. A state prediction method for calculating a predicted value of a state factor representing a state of a manufacturing process or a state of a product manufactured in the manufacturing process, comprising: a calculation step of calculating a predicted value of the state factor using a prediction model, the prediction model includes a model showing a relationship between a dependent variable expressed using the state factor and an explanatory variable expressed using an influencing factor on the state factor, the prediction models include a first prediction model used when the manufacturing process is in a transient state immediately after an operating condition is changed, and a second prediction model used when the manufacturing process is in a steady state; the calculating step calculates the dependent variable using the first prediction model when the operating conditions are changed from first operating conditions to second operating conditions, and thereafter, when a predetermined condition is satisfied under which the manufacturing process can be considered to have reached a steady state, the prediction model used is changed from the first prediction model to the second prediction model to calculate the dependent variable.

8. A program for causing a computer to function as the calculation unit of the state prediction device according to claim 1 or 2.

Images