Processing apparatus, processing method, and program

Abstract

To provide a processing apparatus capable of accurately estimating a heating state from a combustion chamber to a carbonization chamber of a coke oven.SOLUTION: A processing apparatus 300 calculates a heating state quantity as a physical quantity corresponding to a heating state from a combustion chamber to a carbonization chamber, based on a carbonization state quantity as a physical quantity corresponding to a carbonization state of coke, a time from a start of a coal charging operation to an end of a coke pushing operation or an operation time linked to the time, and a heating state estimation model 410.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to a processing device, a processing method, and a program. [Background technology]

[0002] When producing coke, it is desirable to estimate the heating state from the combustion chamber to the coke chamber. For example, this is because it is possible to calculate operating conditions under which estimated values ​​of physical quantities corresponding to the heating state from the combustion chamber to the coke chamber approach (preferably coincide) target values, and to produce coke based on the calculated operating conditions. Patent Document 1 discloses a technique for predicting furnace temperature (smoothed combustion chamber temperature) as a physical quantity corresponding to the heating state from the combustion chamber to the coke chamber. Specifically, Patent Document 1 discloses predicting furnace temperature based on the current measured value of the furnace temperature, past actual values ​​of the flow rate of fuel gas used to heat the furnace from the combustion chamber, and a prediction error of the furnace temperature. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 5-255668 [Patent Document 2] Japanese Patent Application Publication No. 2023-039670 Summary of the Invention [Problem to be solved by the invention]

[0004] However, in the technology described in Patent Document 1, the carbonization process in the operation for which the furnace temperature is to be predicted is taken into account using the prediction error of the furnace temperature. This prediction error of the furnace temperature is due to thermal load disturbances and uncertainties in the mathematical model, and therefore the carbonization process itself is not explicitly evaluated. Therefore, with the technology described in Patent Document 1, it is not easy to improve the calculation accuracy of the furnace temperature.

[0005] The present invention has been made in consideration of the above problems, and aims to accurately estimate the heating state from the combustion chamber to the coke chamber of a coke oven. [Means for solving the problem]

[0006] The processing device of the present invention is a processing device that performs processing to estimate a heating state quantity, which is a physical quantity corresponding to the heating state from the combustion chamber of a coke oven to the carbonization chamber, and is equipped with a heating state calculation unit that calculates the heating state quantity based on a carbonization state quantity, an operating time, and a heating state estimation model, wherein the carbonization state quantity is a physical quantity corresponding to the carbonization state of coke, the operating time is the time from the start of the loading operation of loading coal into the carbonization chamber to the end of the discharge operation of discharging coke from the carbonization chamber, or a time linked to that time, and the heating state estimation model is a model that represents the relationship between the carbonization state quantity and the operating time and the heating state quantity.

[0007] The processing method of the present invention is a processing method that performs processing to estimate a heating state quantity, which is a physical quantity corresponding to the heating state from the combustion chamber of a coke oven to the carbonization chamber, and includes a heating state calculation process that calculates the heating state quantity based on a carbonization state quantity, an operating time, and a heating state estimation model, wherein the carbonization state quantity is a physical quantity corresponding to the carbonization state of coke, the operating time is the time from the start of the loading operation of loading coal into the carbonization chamber to the end of the discharge operation of discharging coke from the carbonization chamber, or a time linked to that time, and the heating state estimation model is a model that represents the relationship between the carbonization state quantity and the operating time and the heating state quantity.

[0008] The program of the present invention causes a computer to function as a heating state calculation unit of the processing device. [Effects of the Invention]

[0009] According to the present invention, the heating state quantity is calculated based on the carbonization state quantity, operation time, and heating state estimation model. Therefore, the process of coal carbonization in the operation for which the furnace temperature is to be predicted can be evaluated by using the carbonization state quantity and operation time. Therefore, the heating state from the combustion chamber to the coke chamber of the coke oven can be accurately estimated. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 is a diagram showing a schematic configuration of a coke oven. [Figure 2A] FIG. 1 is a diagram showing an outline of a coke oven during carbonization. [Figure 2B] FIG. 1 is a diagram showing an outline of a coke oven during the unloading (extrusion) operation. [Figure 3] FIG. 2 illustrates an example of a functional configuration of a processing device. [Figure 4] FIG. 1 is a diagram illustrating an example of an outline of an estimation model. [Figure 5] FIG. 1 is a scatter diagram showing an example of the relationship between coke temperature and furnace temperature. [Figure 6A] FIG. 10 is a scatter diagram showing an example of the relationship between the dry distillation time and the furnace temperature in the first period. [Figure 6B] FIG. 10 is a scatter diagram showing an example of the relationship between the carbonization time and the furnace temperature in the second period. [Figure 6C] FIG. 10 is a scatter diagram showing an example of the relationship between the carbonization time and the furnace temperature in the third period. [Figure 6D] FIG. 10 is a scatter diagram showing an example of the relationship between the carbonization time and the furnace temperature in the fourth period. [Figure 6E] FIG. 10 is a scatter diagram showing an example of the relationship between the carbonization time and the furnace temperature in the fifth period. [Figure 6F] FIG. 10 is a scatter diagram showing an example of the relationship between the carbonization time and the furnace temperature in the sixth period. [Figure 6G] FIG. 10 is a scatter diagram showing an example of the relationship between the carbonization time and the furnace temperature in the seventh period. [Figure 7] FIG. 10 is a diagram showing an example of the relationship between the dry distillation time, the regression coefficient, and the intercept. [Figure 8]FIG. 1 is a diagram showing an example of a linear regression equation calculated using a regression coefficient and an intercept as constants, and the linear regression equation of equation (1). [Figure 9] 10 is a flowchart illustrating an example of a method for creating an estimation model. [Figure 10] 1 is a flowchart illustrating an example of a method for calculating an optimum value of a target carbonization state quantity (target coke temperature). [Figure 11] FIG. 4 is a diagram showing an example of the relationship between target coke temperature and time. [Figure 12] FIG. 10 is a diagram showing an example of the relationship between DI (drum strength index) and time. DETAILED DESCRIPTION OF THE INVENTION

[0011] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In addition, the term "compared objects" including length, position, size, spacing, etc., being the same includes not only cases where they are strictly the same, but also cases where they are different within the scope of the gist of the invention (for example, cases where they differ within the tolerance range determined at the time of design).

[0012] [Outline of Coke Oven 1 and Coke Manufacturing Process] First, with reference to FIGS. 1, 2A, and 2B, the schematic configuration of a coke oven 1 and an outline of a coke manufacturing process will be described. In the coke oven 1, carbonization chambers (kilns) 2 and combustion chambers 3 are arranged alternately with furnace walls 4 interposed therebetween. The carbonization chambers 2 carbonize the coal charged therein to obtain coke. The combustion chambers 3 keep the carbonization chambers 2 at a high temperature by burning fuel gas. In the following description, charging coal into the carbonization chambers 2 will be referred to as "charring" as necessary.

[0013] In the coke production process using a coke oven 1, the so-called block unloading method is used for the unloading coal loading operation. The unloading coal loading operation consists of the operation of pushing coke out of the carbonization chamber 2 using an extruder (unloading operation), and the subsequent operation of supplying coal to the carbonization chamber 2 (charging operation). In this case, the unloading operation is an example of a discharge operation, and the charging operation is an example of a charging operation. In the following explanation, unloading will be referred to as pushing out as necessary. In the block unloading method, the entire carbonization chamber 2 is divided into multiple blocks (in the following explanation, these blocks will be referred to as "blocks" as necessary). When all the carbonization chambers 2 are divided into five patterns, for example, the carbonization chambers 2 are divided at intervals of five, such as pattern 1 (carbonization chambers No. 1, 6, 11, 16, etc.), pattern 2 (carbonization chambers No. 2, 7, 12, 17, etc.), pattern 3 (carbonization chambers No. 3, 8, 13, 18, etc.), pattern 4 (carbonization chambers No. 4, 9, 14, 19, etc.), and pattern 5 (carbonization chambers No. 5, 10, 15, 20, etc.). By setting the order of unloading and loading the carbonized materials, for example, as pattern 1, pattern 3, pattern 5, pattern 2, and pattern 4, it is possible to prevent a sudden drop in temperature in each carbonization chamber 2. Furthermore, within each pattern, the unloading and loading work is carried out in order from the lowest number (carbonization chamber 2 with the lowest carbonization chamber number). The time from the time when the unloading coal loading work is completed in one way to the time when the unloading coal loading work is completed in the next way (for example, in the above example, way 2, which follows way 1) is called the running time. The running time is generally about 3 to 6 hours. Note that this embodiment is not limited to the block unloading method. For example, if the following explanation treats a way (block) as an individual coking chamber 2, it can also be applied to cases where the unloading coal loading work is performed in units of one coking chamber 2.

[0014] Here, in this embodiment, the average value of the amount (e.g., mass) of coal charged into the coking chambers 2 belonging to one run is taken as the coal charging amount (run average value). The run average value is an arithmetic mean value (the sum of the coal amounts charged into the coking chambers 2 belonging to one run divided by the number of coking chambers 2 belonging to that run). Note that the method of determining the coal charging amount itself may be, for example, one adopted in a coke plant, and is not limited to the above. For example, the coal charging amount may be the average value for one kiln unloading operation.

[0015] In this embodiment, the carbonization time (average value per run) is the average value of the time from the start of loading of coal into one coking chamber 2 to the end of unloading (extrusion). The average value per run is the arithmetic mean value (the sum of the time from the start of loading of coal into the coking chambers 2 belonging to one run to the end of unloading (extrusion) by the number of coking chambers 2 belonging to that run).

[0016] The time from the start of loading one coke chamber 2 to the end of unloading (pushing) is expressed, for example, as the sum of the flame-out time (the time required for carbonizing the coal) and the storage time (the time the coke after carbonization is retained in the coke chamber 2). The method of determining the carbonization time itself may be, for example, a method adopted in a coke plant, and is not limited to the above. For example, instead of the time from the start of loading one coke chamber 2 to the end of unloading (pushing) (the time the coal and the coke produced from that coal are actually present in the coke chamber 2), the carbonization time may be defined using the time from the start of loading work for one coke chamber 2 to the end of unloading work. Furthermore, the carbonization time may be an average value for one unloading work.

[0017] In this embodiment, an example is shown in which the carbonization time is a time linked to the time from the start of the loading operation in (one) carbonization chamber 2 to the end of the unloading operation. However, such a time is not limited to the carbonization time. For example, the fire-out time (average value of the passing time) or the passing time may be used instead of the carbonization time. In the following description, such a time will be referred to as the operating time as necessary. Note that the operating time may be the time itself from the start of the loading operation in (one) carbonization chamber 2 to the end of the unloading operation.

[0018] 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. Specifically, the heat input to the coke oven 1 is controlled by operating a single 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 operated via an actuator (not shown) under the control of a control device for the coke manufacturing process. The control device for the coke manufacturing process may be realized, for example, by the control device described in Patent Document 2. A representative value of the temperature of all combustion chambers 3 is referred to as the furnace battery temperature. For example, 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 where the thermometers 6 are installed is defined as the furnace battery temperature. The method of this embodiment is not limited to simultaneously adjusting the heat input to all combustion chambers 3. For example, when the unloading and loading work is performed for each carbonization chamber 2, an adjustment valve and an actuator may be installed in each combustion chamber 3 to control the carbonization state (input heat amount) for each carbonization chamber 2. In addition, the thermometer 6 may be installed in each of all combustion chambers 3 or only in some of the combustion chambers 3.

[0019] The furnace temperature can be obtained, for example, by a measurement value using a thermometer 6 shown in FIG. 2A. In this embodiment, the average value of the furnace battery temperature in one run is taken as the furnace temperature (run average value). The run average value is an arithmetic mean value. It is the sum of the average values ​​of the furnace battery temperature at each time in one run divided by the number of coke chambers 2 belonging to that run. The furnace temperature may be the temperature of the combustion chamber 3, and the thermometer for measuring the furnace temperature and the method of determining the furnace temperature may be, for example, one that is used in a coke plant, and are not limited to those described above. For example, the furnace temperature may be the average value for one unloading operation.

[0020] Furthermore, in this embodiment, an example is given in which the furnace temperature is a physical quantity corresponding to the heating state from the combustion chamber 3 to the coke chamber 2. However, the physical quantity corresponding to the heating state from the combustion chamber 3 to the coke chamber 2 is not limited to the furnace temperature, and may be, for example, the amount of heat input to the combustion chamber 3 or the flow rate of fuel gas used to heat the coke chamber 2 from the combustion chamber 3. In the following description, such physical quantities will be referred to as heating state quantities as necessary. Note that the heating state from the combustion chamber 3 to the coke chamber 2 indicates the degree of heating being performed from the combustion chamber 3 to the coke chamber 2 during coke production (production).

[0021] As described above, coke is pushed out of the coke chamber 2 by the pusher. In the example shown in FIG. 2B , coke 10 pushed out of the coke chamber 2 by the pusher ram 7 of the pusher is discharged via a 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, and then the guide car 9 moves to the coke chamber 2 located at the top of FIG. 2B . This is indicated by the two-dot chain line after the guide car 9 has moved. In addition, in FIG. 2B , 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 path of the coke 10 inside the guide car 9 through a window provided in the guide car 9. As described above, in this embodiment, the temperature of the coke 10 is measured immediately after it leaves the coke chamber 2 during the unloading operation (at the time of extrusion). However, as long as the temperature of the coke leaving the coke chamber 2 is measured, the temperature of the coke 10 does not necessarily have to be measured in this manner. In addition, a thermometer (not shown) that measures the temperature of the oven wall 4 in a non-contact manner may be installed on the extrusion ram 7. The thermometer measures the temperature of the oven wall 4 (on both sides of the coke chamber 2) during the unloading operation (at the time of extrusion) of the coke 10. The temperature of the oven wall 4 may be measured instead of or in addition to measuring the temperature of the coke 10. In the following description, the temperature of the oven wall 4 will be referred to as the oven wall temperature as necessary.

[0022] The coke temperature can be obtained, for example, by a measurement value using a thermometer 8 shown in FIG. 2B. When the extrusion ram 7 is pushing the coke 10 out of the coke chamber 2, the temperature of the coke 10 sequentially discharged from the coke chamber 2 is measured using the thermometer 8, and the average value of the temperatures measured at each time and each position (the sum of the temperatures measured at each time and each position divided by the number of temperature measurements) is defined as the temperature of the coke 10 produced in the coke chamber 2. The average value of the temperatures of the coke 10 produced in the coke chambers 2 belonging to one run is defined as the coke temperature (run average value). The run average value is the arithmetic mean value (the sum of the temperatures of the coke produced in the coke chambers 2 belonging to one run divided by the number of coke chambers 2 belonging to that run). The coke temperature is preferably the temperature of the coke 10 immediately after it is discharged from the coke chamber 2, and therefore the coke temperature is determined as shown in Fig. 2B. However, the thermometer for measuring the coke temperature and the method for determining the coke temperature may be, for example, any of those employed in a coke plant, and are not limited to the above. For example, the coke temperature may be an average value for one unloading operation.

[0023] In addition, in this embodiment, a case where the coke temperature is the physical quantity corresponding to the carbonization state of the coke is exemplified. However, the physical quantity corresponding to the carbonization state of the coke is not limited to the coke temperature, and may be, for example, the furnace wall temperature. In the following description, such a physical quantity will be referred to as a carbonization state quantity as necessary. Note that the carbonization state of the coke indicates the degree to which coal has been carbonized (pyrolyzed) in the produced coke, and is used as an intermediate index of the quality of the coke.

[0024] [Processing device 300] FIG. 3 is a diagram illustrating an example of the functional configuration of a processing device 300. The processing device 300 includes, as hardware, one or more hardware processors, such as a central processing unit (CPU), and one or more memories, such as a random access memory (RAM) and a read-only memory (ROM). The processing device 300 executes one or more programs stored in the memories by using the one or more hardware processors to perform various calculations. An input device 301 and an output device 302 are communicatively connected to the processing device 300. Communication between the processing device 300 and the input device 301 and the output device 302 may be wired or wireless. The processing device 300 may also include the input device 301 and the output device 302. The processing device 300 may also be implemented by dedicated hardware, such as an application-specific integrated circuit (ASIC).

[0025] In this embodiment, the processing device 300 includes a creating unit 310 and a calculating unit 320. As shown in FIG.

[0026] The creation unit 310 creates an estimation model to be used by the calculation unit 320. The calculation unit 320 performs calculations using the estimation model created by the creation unit 310. In this embodiment, the processing device 300 includes the creation unit 310 and the calculation unit 320. However, this is not necessarily required, and the functions of the processing device 300 may be realized by a plurality of devices. For example, the functions of the processing device 300 may be realized by a device that realizes the functions of the creation unit 310 and a device that realizes the functions of the calculation unit 320.

[0027] <Estimation model> FIG. 4 is a diagram illustrating an example of an outline of an estimation model. 4, the heating state estimation model 410 is a model that represents the relationship between the carbonization state quantity and operation time (coke temperature and carbonization time in this embodiment) and the heating state quantity (furnace temperature in this embodiment), and is, for example, a trained model that has learned the relationship by machine learning, etc. As will be described later, this embodiment illustrates a case where the heating state estimation model 410 includes a regression equation.

[0028] The quality estimation model 420 is a model that represents the relationship between a first influencing factor, which is a factor that affects the quality of the coke, and a primary quality index (final quality index) of the coke, and is, for example, a trained model that has learned the relationship by machine learning, etc. The explanatory variable of the quality estimation model 420 is the first influencing factor, and the objective variable is the primary quality index of the coke.

[0029] Here, quality indices that indicate the quality of coke include intermediate quality indices such as coke temperature and oven wall temperature as well as final quality indices. The final quality indices are obtained by testing samples taken from the produced coke in a laboratory or the like. In this embodiment, the case where such a final quality index is the drum strength index (DI) is exemplified. However, the final quality index is not limited to the DI and may be, for example, other indices specified in JIS K2151 (e.g., SI (drop strength index) and TI (tumbler strength index)) or the CSR (coke strength after reaction) value. Although the coke temperature is a coke quality index (intermediate quality index), in the quality estimation model 420, the aforementioned final quality index is treated as the primary quality index of the coke, and other coke quality indices are not treated as primary quality indexes of the coke (i.e., unless otherwise specified, the quality index refers to the aforementioned final quality index). This also applies to the quality change amount estimation model 430 described later.

[0030] The first influence factor may be any factor that affects the quality of the coke, but preferably includes an influence factor that is expected to have a greater effect on the final quality index of the coke (than the second influence factor described later). The number of types of first influence factors may be one or more. In this embodiment, an example is shown in which the first influence factor includes a blending ratio, which is the content ratio of each component contained in the coal charged into the coke chamber 2.

[0031] The blending ratio of each component may be a mass ratio or a volume ratio. Furthermore, when multiple types (multiple brands) of coal are charged into the same coke chamber 2 in one charging operation, the amount of each component may be calculated by calculating a weighted average value for each component, with the weight coefficient being the amount of each type of coal divided by the amount of all types of coal charged. For example, when a coal having a total amount M1 (= Ma1 + Mb1) where the amount of component a is Ma1 and the amount of component b is Mb1, and a coal having a total amount M2 (= Ma2 + Mc2) where the amount of component a is Ma2 and the amount of component c is Mc2, the amount of component a is expressed as M1 ÷ (M1 + M2) × Ma1 + M2 ÷ (M1 + M2) × Ma2. In this case, the content ratio of component a is calculated by dividing the amount of component a by the total amount of coal (= M1 + M2). In this case, components b and c are contained in only one type of coal, so weighted averaging is not performed. The amounts of components b and c are Mb1 and Mc2, respectively, and the content ratios of components b and c are Mb1 ÷ (M1 + M2) and Mc2 ÷ (M1 + M2), respectively.

[0032] The present inventors have found that it is not easy to improve the estimation accuracy of the final coke quality index, whether the number of explanatory variables in an estimation model in which influential factors affecting coke quality are used as explanatory variables and a final coke quality index is used as a target variable, is small or large. Specifically, the present inventors have found that if explanatory variables include influential factors that significantly affect the final coke quality index as influential factors on coke quality indicators and influential factors that have a non-negligible effect on the estimation accuracy of the final coke quality index as influential factors on the final coke quality index, there is a risk that an estimation model will be created in which the estimation result of the final coke quality index is more influenced by the latter than the former. Furthermore, the present inventors have found that even if only influential factors that significantly affect the final coke quality index as factors affecting coke quality are used as explanatory variables, it is not possible to improve the estimation of the final coke quality index.

[0033] Therefore, the inventors have found that the accuracy of estimating the quality index can be improved by calculating the amount of change in the primary quality index (DI) of the coke based on a second influencing factor that is different from the first influencing factor as a factor that affects the quality of the coke, and by changing the primary quality index (DI) calculated using the quality estimation model 420 with the calculated amount of change.

[0034] The quality change amount estimation model 430 is a model that calculates the amount of change in such a primary quality index of coke. The quality change amount estimation model 430 is a model that represents the relationship between a second influencing factor, which is a factor that affects the quality of coke, and the amount of change in the primary quality index of coke, and is, for example, a trained model that has learned this relationship by machine learning or the like. The explanatory variables of the quality change amount estimation model 430 include the second influencing factor, and the objective variable includes the amount of change in the primary quality index of coke.

[0035] The second influence factor may be any influence factor that affects the quality of the coke and is different from the first influence factor, but is preferably an influence factor that is expected to have a smaller effect on the final quality index (primary quality index) of the coke than the first influence factor based on past knowledge, etc. However, this is not necessarily the case. For example, as long as the first influence factor includes an influence factor that is expected to have the greatest effect on the final quality index (primary quality index) of the coke, the second influence factor may include an influence factor that is expected to have a greater effect on the final quality index (primary quality index) of the coke than the first influence factor.

[0036] Furthermore, in this embodiment, an example is given in which the second influencing factors include operational factors, which are physical quantities that can be obtained as the operational performance of the coke oven after the start of coal loading work and before the start of a test to obtain a final coke quality index. Before the start of a test to obtain a final coke quality index may be, for example, before the start of coke transportation from the coke oven 1 or before a sample is taken from the coke. Specifically, in this embodiment, an example is given in which the second influencing factors (operational factors) include the oven temperature, coke temperature, coal loading amount, and carbonization time.

[0037] Here, the carbonization state quantity (e.g., coke temperature) and the heating state quantity (e.g., furnace temperature) are closely related quantities. Therefore, when the carbonization state quantity (e.g., coke temperature) and the heating state quantity (e.g., furnace temperature) are included in the second influencing factor, if these values ​​are determined independently without considering each other's values, the estimation accuracy of the quality change amount estimation model 430 may be reduced. Therefore, as described above, in this embodiment, the heating state quantity (e.g., furnace temperature) corresponding to the carbonization state quantity (e.g., coke temperature) is calculated using the heating state estimation model 410. Note that, for example, if it is not necessary to calculate the final quality index of the coke, the quality estimation model 420 and the quality change amount estimation model 430 may be omitted. For example, the heating state estimation model 410 may be used to calculate the flow rate of fuel gas when determining the input heat quantity based on the difference between the predicted value and the target value of the furnace temperature. In this case, the quality estimation model 420 and the quality change amount estimation model 430 may be omitted.

[0038] As described above, in this embodiment, an example is shown in which the estimated quality index of the coke is obtained by modifying the primary quality index (DI) of the coke calculated using the quality estimation model 420 by the amount of change in the quality index (DI change amount) of the coke calculated using the quality change amount estimation model 430 (the sum of the two). Note that the primary quality index and the estimated quality index are the same type of final quality index (DI in this embodiment).

[0039] <Creation Department 310> In this embodiment, Figure 3 illustrates an example in which the creation unit 310 has a learning data acquisition unit 311, a quality estimation model creation unit 312, a quality change amount estimation model creation unit 313, a heating state estimation model creation unit 314, and a model memory unit 315.

[0040] <<Learning Data Acquisition Unit 311>> The learning data acquisition unit 311 acquires learning data that is data used to create the heating state estimation model 410, the quality estimation model 420, and the quality change amount estimation model 430. In this embodiment, a case will be illustrated in which the heating state estimation model 410, the quality estimation model 420, and the quality change amount estimation model 430 are created by performing supervised learning.

[0041] In this embodiment, a case will be exemplified in which the learning data includes first learning data, second learning data, and third learning data. The first learning data for creating the quality estimation model 420 includes pairs of measured coke quality indicators (measured values ​​of final quality indicators) and measured first influencing factors (measured values ​​of first influencing factors) obtained during operations to produce the coke from which the measured quality indicators were obtained. In this embodiment, a case is illustrated in which the measured coke quality indicators are obtained three times a day and the measured first influencing factors are obtained once a day. In this case, the measured coke quality indicators and the measured first influencing factors obtained on the same day form one pair. The number of pairs included in the first learning data may be any number that allows supervised learning of the quality estimation model 420. In this embodiment, a case is illustrated in which the first influencing factors include the blending ratio and the final coke quality indicator (primary quality index) is DI. The measured coke quality indicators included in the first learning data are used as correct labels. The timing at which the measured quality index of the coke and the measured first influencing factor (blending ratio) are obtained may be, for example, any timing adopted in a coke plant, and is not limited to the above.

[0042] The second learning data for creating the quality change amount estimation model 430 includes pairs of a measured quality index of the coke, a measured first influencing factor obtained in an operation for producing the coke from which the measured quality index was obtained, and a measured second influencing factor (a measured value of the second influencing factor) obtained in an operation for producing the coke from which the measured first influencing factor was obtained. The number of pairs included in the second learning data may be any number that allows the quality change amount estimation model 430 to be learned in a supervised manner.

[0043] As will be described later, the quality change amount estimation model 430 is created after the quality estimation model 420 is created. In this embodiment, a case is illustrated in which a value obtained by subtracting an estimated quality index of coke calculated by applying the measured first influencing factor included in the second learning data to the trained quality estimation model 420 from the measured quality index of coke included in the second learning data (= measured quality index of coke - estimated quality index) is used as the correct label (the correct value of the change amount of the primary quality index of coke). Note that this embodiment illustrates a case in which the estimated quality index is the value of the final quality index at a stage before coke is produced (i.e., a predicted value). However, the estimated quality index may be a value after coke is produced, or a value calculated after the final quality index is actually measured.

[0044] As described above, this embodiment illustrates a case where the second influencing factors (operation factors) include the furnace temperature, coke temperature, coal loading amount, and carbonization time, and also illustrates a case where the furnace temperature, coke temperature, coal loading amount, and carbonization time are expressed as run averages. Therefore, this embodiment illustrates a case where the arithmetic mean of run averages of multiple runs corresponding to the timing at which the measured DI (measured value of DI) was obtained is used as the measured furnace temperature, measured coke temperature, measured coal loading amount, and measured carbonization time (measured values ​​of furnace temperature, coke temperature, coal loading amount, and carbonization time) included in the second learning data. The multiple runs corresponding to the timing at which the measured DI was obtained are, for example, runs that ended during a period that predates the time from the time at which the measured DI was obtained to the time at which the next measured DI was obtained by a time that is estimated as the time from the end of the kiln unloading operation to the time at which the DI was obtained. Furthermore, the multiple runs corresponding to the timing at which the measured DI was obtained may be runs that started in a period going back an amount of time that is estimated as the time from when the kiln unloading operation was completed to when the DI was obtained, relative to the period from when the measured DI was obtained to when the next measured DI was obtained.

[0045] The third learning data for creating the heating state estimation model 410 includes pairs of measured heating state quantities (measured values ​​of heating state quantities), measured carbonization state quantities (measured values ​​of carbonization state quantities) and measured operation times (measured operation times) obtained during the operation to produce the coke from which the measured heating state quantities were obtained. The number of pairs included in the third learning data may be any number that allows supervised learning of the heating state estimation model 410. In this embodiment, a case is illustrated in which the heating state quantity is the furnace temperature, the carbonization state quantity is the coke temperature, and the operation time is the carbonization time. The measured furnace temperature included in the third learning data is used as a ground truth label. In this embodiment, a case is illustrated in which the arithmetic mean value of the mean values ​​of multiple runs corresponding to the timings at which the measured DIs were obtained is used as the measured furnace temperature, measured coke temperature, and measured carbonization time included in the third learning data. Note that the furnace temperature, coke temperature, and carbonization time are also included in the second influence factors. Therefore, the measured oven temperature, the measured coke temperature, and the measured carbonization time included in the second learning data may be used instead of the third learning data.

[0046] The learning data acquisition unit 311 may, for example, input learning data (first to third learning data) from the input device 301, or may calculate learning data based on the data input from the input device 301. The input device 301 may be a storage medium, a user interface, a receiving device, or a device including two or three of these.

[0047] <<Quality estimation model creation unit 312>> The quality estimation model creation unit 312 creates a quality estimation model 420 and stores it in the model storage unit 315. The quality estimation model creation unit 312 creates (learns) the quality estimation model 420 by, for example, performing supervised learning using the first learning data acquired by the learning data acquisition unit 311 as learning data. For example, the quality estimation model creation unit 312 may create the quality estimation model 420 by using gradient boosting such as CatBoost (Category Boosting). Note that the quality estimation model 420 is not limited to this and may be, for example, a neural network. The method of creating the quality estimation model 420 itself can be realized by known technology, and therefore a detailed description thereof will be omitted here.

[0048] <<Quality change amount estimation model creation unit 313>> The quality change amount estimation model creation unit 313 creates a quality change amount estimation model 430 and stores it in the model storage unit 315. In this embodiment, a case where the quality change amount estimation model 430 is created after the quality estimation model 420 is created will be illustrated as an example.

[0049] In this case, for example, the quality change amount estimation model creation unit 313 calculates the estimated quality index (DI) of the coke by providing the measured first influencing factor (blending conditions) included in the second learning data acquired by the learning data acquisition unit 311 to the quality estimation model 420. The quality change amount estimation model creation unit 313 calculates, as the change amount of the primary quality index of the coke, a value obtained by subtracting the estimated quality index (DI) of the coke from the measured quality index (DI) of the coke paired with the measured first influencing factor (blending conditions) in the second learning data.

[0050] The quality change amount estimation model creation unit 313 uses the change amount of the primary quality index of the coke calculated in this way and the measured second influencing factors (furnace temperature, coke temperature, coal charging amount, and carbonization time) paired with the measured first influencing factor (blend condition) in the second learning data as one piece of learning data. The quality change amount estimation model creation unit 313 creates multiple pieces of learning data as this learning data using each of the measured first influencing factors (blend condition) included in the second learning data acquired by the learning data acquisition unit 311.

[0051] The quality change amount estimation model creation unit 313 creates (learns) the quality change amount estimation model 430 by performing supervised learning using these learning data. For example, the quality change amount estimation model creation unit 313 may create the quality change amount estimation model 430 by using gradient boosting such as CatBoost (Category Boosting). Note that the quality change amount estimation model 430 is not limited to this and may be, for example, a neural network. The quality estimation model 420 and the quality change amount estimation model 430 may be the same type of model or different types of models. The method of creating the quality change amount estimation model 430 itself can be realized by known technology, so a detailed description thereof will be omitted here.

[0052] <<Heating State Estimation Model Creation Unit 314>> The heating state estimation model creation unit 314 creates a heating state estimation model 410 and stores it in the model storage unit 315. The heating state estimation model creation unit 314 creates (learns) the heating state estimation model 410 by performing supervised learning using the third learning data acquired by the learning data acquisition unit 311 as learning data.

[0053] The heating state estimation model 410 is a model that represents the relationship between the carbonization state quantity (coke temperature in this embodiment) and the operating time (carbonization time in this embodiment) and the heating state quantity (furnace temperature in this embodiment), and is, for example, a trained model that has learned the relationship by machine learning or the like.

[0054] As described above, the carbonization state quantity (coke temperature, etc.) and the heating state quantity (furnace temperature, etc.) are closely related quantities. Therefore, the present inventors considered using a model in which the carbonization state quantity (coke temperature in this embodiment) is an explanatory variable and the heating state quantity (furnace temperature in this embodiment) is an objective variable as the heating state estimation model 410. In this case, since the heating state estimation model 410 has one explanatory variable and one objective variable, respectively, the present inventors considered expressing the heating state estimation model 410 as a regression equation. FIG. 5 is a scatter diagram showing an example of the relationship between the coke temperature and the furnace temperature. ro1 , T ro2 The difference between them is 100°C. In FIG. 5, a point determined by the measured coke temperature and the measured oven temperature on the same street is given as one point (plot) on the scatter diagram. The inventors realized that it would not be easy to find the correlation between the coke temperature and the oven temperature from the scatter diagram shown in FIG. 5. Therefore, the inventors created scatter diagrams by dividing the scatter diagram shown in FIG. 5 by carbonization time. FIGS. 6A to 6G are scatter diagrams showing an example of the relationship between the coke temperature and the oven temperature created in this way. Specifically, FIGS. 6A to 6G are obtained by dividing the measured carbonization time into seven periods T1 to T7, each of which has the same duration (length), and creating a scatter diagram for each period T1 to T7.

[0055] The inventor then performed linear regression analysis using each of the seven scatter plots (coke temperatures and furnace temperatures indicated by ) in Figures 6A to 6G as learning data, with the regression coefficient and intercept as constants, respectively. Figure 7 is a diagram showing an example of the relationship between carbonization time and the regression coefficient and intercept. In Figure 7, times t1, t2, t3, t4, t5, t6, and t7 correspond to the aforementioned periods T1, T2, T3, T4, T5, T6, and T7, respectively, and the values ​​of the slope (regression coefficient) 710 and intercept 720 at times t1, t2, t3, t4, t5, t6, and t7 are values ​​calculated using the scatter plots for the periods T1, T2, T3, T4, T5, T6, and T7, respectively.

[0056] 7, it can be seen that the slope (regression coefficient) 710 is approximately constant at a value close to the minimum value in the first interval (the interval from time t1 to t2), increases with increasing carbonization time in the second interval (the interval from time t2 to t5) following the first interval (i.e., the interval following the first interval), and remains approximately constant at a value close to the maximum value in the third interval (the interval from time t5 to t7) following the second interval (i.e., the interval following the second interval). It can also be seen that the intercept 720 has an inverse relationship to the slope (regression coefficient) 710, i.e., is approximately constant at a value close to the maximum value in the first interval (the interval from time t1 to t2), decreases with increasing carbonization time in the second interval (the interval from time t2 to t5) following the first interval, and remains approximately constant at a value close to the minimum value in the third interval (the interval from time t5 to t7) following the second interval.

[0057] From this, the inventors have found that by expressing the regression coefficients and intercepts as a function of carbonization time and using a broadly monotonically increasing function in which the slopes in the first and second sections of the carbonization time are smaller than the slope in the third section of the carbonization time, a regression equation having the carbonization state quantity (coke temperature in this embodiment) as an explanatory variable and the heating state quantity (furnace temperature in this embodiment) as a response variable can be made into a regression equation that can appropriately express the relationship between the carbonization state quantity and the heating state quantity. Note that here, an example is given in which the operating time (the time linked to the time from the start of the coal loading operation in one coke chamber 2 to the end of the unloading operation) is the carbonization time. However, the fact that a regression equation that can appropriately express the relationship between the carbonization state quantity and the heating state quantity can be obtained is also true for operating times other than the carbonization time (for example, the fire-out time and the passing time).

[0058] Various functions can be used as such a function, for example, a function based on the standard logistic function, which is a function obtained by transforming the standard logistic function by at least one of enlarging (along at least one of the horizontal and vertical axes), shrinking (along at least one of the horizontal and vertical axes), translating (along at least one of the horizontal and vertical axes), and rotating.

[0059] In this case, the heating state estimation model 410 is, for example, ro is used as the objective variable, and the carbonization state quantity T co It is expressed by the following linear regression equation (1) with explanatory variables. In equation (1), K is an example of the above-mentioned function, and is expressed by the following equation (2). w1 to w6 are variables determined by regression analysis. In this embodiment, the case where w1 to w6 are positive values ​​is exemplified. t k is the operation time such as the carbonization time.

[0060]

number

[0061] In equation (1), the carbonization state quantity T co Not only the operating time t k is also included in the explanatory variables, but the heating state quantity T ro is used as the objective variable, and the carbonization state quantity T co The relationship between the two is expressed in the form of a linear regression equation with the explanatory variables, and the operating time t k is included in the regression coefficient and intercept. That is, as shown in equation (1), the heating state quantity T ro and the carbonization state quantity T co and operation time t k The relationship between and is expressed as the heating state quantity T ro is used as the objective variable, and the carbonization state quantity T co When expressed in the form of a simple regression equation with the explanatory variables, the operating time t k is included in the regression coefficients and intercepts and is not represented as a variable by which the regression coefficients are multiplied.

[0062] Furthermore, equation (1) illustrates the case where the regression coefficient is K (the function described above). Furthermore, equation (1) illustrates the case where the intercept is w1(1-Kw2). In this way, equation (1) illustrates the case where the regression coefficient and intercept of the linear regression equation are expressed using K (the function described above). Furthermore, equation (1) illustrates the case where the value of the regression coefficient has a positive correlation with the value of K (a relationship in which the value of the regression coefficient increases as the value of K increases). Furthermore, equation (1) illustrates the case where the value of the intercept (w1(1-Kw2)) of the linear regression equation has a negative correlation with the value of K (a relationship in which the value of the intercept decreases as the value of K increases). Note that in the following equations (1) and (2), when the carbonization time is within the range expected in the operation of a coke oven, the heating state quantity T ro and the carbonization state quantity T co is assumed to be linearly related.

[0063] Fig. 8 shows an example of a linear regression equation 810 calculated using the regression coefficient and intercept as constants, and a linear regression equation 820 of equation (1). The linear regression equations 810 and 820 shown in Fig. 8 are calculated by using, as learning data, values ​​in approximately the first half of period T1 in the scatter diagram of Fig. 6A (the coke temperature and oven temperature indicated by ). As shown in Fig. 8, when the regression coefficient and intercept are expressed using K shown in equation (2), the regression coefficient and intercept can be calculated to correspond to the scatter diagram, compared to when the regression coefficient and intercept are constants.

[0064] The heating state estimation model creation unit 314 creates (learns) the heating state estimation model 410 by, for example, performing supervised learning using the third learning data acquired by the learning data acquisition unit 311 as learning data. At this time, the heating state quantity (furnace temperature) included in the third learning data is used as a correct label. The heating state estimation model creation unit 314 may create the heating state estimation model 410 by, for example, using a regression analysis technique such as Bayesian optimization. The method of creating the heating state estimation model 410 itself can be realized by known technology, and therefore a detailed description thereof will be omitted here.

[0065] According to the knowledge obtained by the present inventors in the process of expressing the heating state estimation model 410 as a regression equation (the knowledge that it is necessary to further consider not only the carbonization state quantity but also the operation time), it can be said that not only the carbonization state quantity but also the operation time are explanatory variables that explain the heating state quantity. Therefore, for example, as the heating state estimation model 410, a model such as a neural network model in which the carbonization state quantity (coke temperature) and the operation time (carbonization time) are explanatory variables and the heating state quantity (furnace temperature) is a response variable may be used.

[0066] <Calculation section 320> In this embodiment, FIG. 3 illustrates a case where the calculation unit 320 includes an estimation data acquisition unit 321, a target calculation unit 322, and an output unit 323. As described above, the calculation unit 320 may perform calculations using the estimation models (in this embodiment, the heating state estimation model 410, the quality estimation model 420, and the quality change amount estimation model 430) created by the creation unit 310. However, this embodiment illustrates a case where the target value of the carbonization state quantity is calculated using the estimation models created by the creation unit 310. In the following description, the target value of the carbonization state quantity will be referred to as the target carbonization state quantity as needed, and the target value of coke, which is an example of the carbonization state quantity, will be referred to as the target coke temperature as needed. In addition, this embodiment illustrates a case where the target value of the carbonization state quantity is calculated using metaheuristics such as a pollination algorithm or a genetic algorithm. Note that the calculation unit 320 may calculate a predicted value of the coke final quality index (DI in this embodiment) without calculating (searching) the target carbonization state quantity. In this case, in the following description, the calculation for calculating (searching) the target carbonization state quantity (target coke temperature) may be omitted. Also, instead of the target carbonization state quantity (target coke temperature) calculated (searched) as described below, for example, a planned value of the carbonization state quantity (coke temperature) may be used. In this way, the estimated value of the carbonization state quantity may be a target value, a planned value, or another set value.

[0067] <<Estimation Data Acquisition Unit 321>> The estimation data acquisition unit 321 acquires estimation data, which is data necessary for the calculation by the calculation unit 320. In this embodiment, the estimation data includes a planned value or an actual measurement value of the first influencing factor, a planned value or an actual measurement value of the second influencing factor (coal loading amount and carbonization time) other than the carbonization state quantity and the heating state quantity (coke temperature and furnace temperature), a plurality of candidate initial values ​​of the target carbonization state quantity, and a target value of the estimated coke quality index. Note that an assumed value other than the planned value may be used instead of the planned value.

[0068] In addition, in this embodiment, a case is illustrated in which the estimation data acquisition unit 321 acquires multiple candidate initial values ​​for each of the target carbonization state quantities at each timing (one or more timings) after the estimation start timing, which is the timing to start operation to bring the carbonization state quantity (coke temperature) closer to (preferably coincide with) the target carbonization state quantity (target coke temperature). Also, as described above, in this embodiment, a case is illustrated in which the operation time is the carbonization time. Therefore, in this embodiment, a case is illustrated in which the carbonization time included in the second influence factor is used as the operation time (given to the heating state estimation model 410).

[0069] In this embodiment, the estimation data acquisition unit 321 acquires the actual measured values ​​of the first influencing factor and the second influencing factor other than the carbonization state quantity and the heating state quantity, if the actual measured values ​​are paired with multiple candidates for the target carbonization state quantity, and acquires the planned values ​​if not. However, this is not necessarily required, and the estimation data acquisition unit 321 may acquire the planned values ​​instead of the actual measured values, for example.

[0070] The values ​​paired with the multiple candidates for the target carbonization state quantity at each timing after the estimation start timing are, for example, values ​​in the operation performed at each timing. As described above, in this embodiment, the arithmetic mean value of the average values ​​of multiple runs corresponding to the timings at which the measured DIs are obtained is used as the measured furnace temperature, measured coke temperature, measured coal amount, and measured carbonization time included in the second learning data and the third learning data. Therefore, in this embodiment, the target carbonization state quantity (target coke temperature) is also expressed as the arithmetic mean value of the average values ​​of multiple runs corresponding to the timings at which the DIs are obtained. In this case, the values ​​paired with the multiple candidates for the target carbonization state quantity at each timing after the estimation start timing refer to, for example, the arithmetic mean value of the average values ​​of multiple runs including the same run as the run performed at each timing.

[0071] In the present embodiment, the estimation data acquiring unit 321 acquires a target value of DI as the target value of the estimated quality index of coke. In the following description, the target value of the estimated quality index of coke will be referred to as a target estimated quality index as necessary.

[0072] The estimation data acquisition unit 321 may input estimation data from the input device 301, or may calculate estimation data based on data input from the input device 301, for example.

[0073] <<Target Calculation Unit 322>> The target calculation unit 322 calculates the target carbonization state quantity based on the estimation data acquired by the estimation data acquisition unit 321. In this embodiment, the target calculation unit 322 includes a heating state calculation unit 322a, a quality index calculation unit 322b, an evaluation unit 322c, and a target determination unit 322d. In this embodiment, the target calculation unit 322 calculates the target carbonization state quantity at one of multiple timings after the estimation start timing of the target carbonization state quantity for each of the multiple timings. That is, the following descriptions of the <<<heating state calculation unit 322a>>>, the <<<quality index calculation unit 322b>>>, the <<<evaluation unit 322c>>>, and the <<<target determination unit 322d>>> are for one timing, and the following description illustrates a case where the processing described in these sections is performed for each of multiple timings.

[0074] <<<Heating State Calculation Unit 322a>>> The heating state calculation unit 322a calculates the heating state quantity corresponding to the dry distillation state quantity using the heating state estimation model 410. In this embodiment, the heating state calculation unit 322a calculates the heating state quantity (furnace temperature) for each of a plurality of candidates for the target carbonization state quantity (coke temperature) by providing candidates for the target carbonization state quantity (coke temperature) and the actual measured value or planned value of the operation time (carbonization time) to the heating state estimation model 410. Note that, for example, if an actual measured value of the operation time (carbonization time) of the operation to be estimated for the target carbonization state quantity is obtained at the timing when the estimation of the target carbonization state quantity is started, the actual measured value may be used. However, for example, if an actual measured value of the operation time (carbonization time) of the operation to be estimated for the target carbonization state quantity is not obtained at the timing when the estimation of the target carbonization state quantity is started, the planned value (estimated value) of the operation time (carbonization time) of the operation may be used. In this embodiment, the planned value of the operation time (carbonization time) is used as an example. In the following description, the planned value of the operation time (carbonization time) will be referred to as the planned operation time as necessary.

[0075] First, the heating state calculation unit 322a calculates the heating state quantity (furnace temperature) for each of the multiple candidates by providing the initial value of the target carbonization state quantity candidate and the planned operation time to the heating state estimation model 410. Thereafter, if the target determination unit 322d (described later) determines that the convergence condition is not satisfied, the heating state calculation unit 322a updates at least one of the multiple candidates for the target carbonization state quantity (coke temperature) and provides the updated value of the target carbonization state quantity (coke temperature) candidate and the planned operation time to the heating state estimation model 410 to calculate an estimated value of the heating state quantity (furnace temperature) for each of the multiple candidates. The candidate is updated, for example, according to a metaheuristics algorithm. The metaheuristics algorithm itself is realized by a known technique, and therefore a detailed description thereof will be omitted here. For convenience of explanation, this embodiment illustrates a case where the initial values ​​of multiple candidates for the target carbonization state quantity are included in the estimation data. However, this is not necessarily the case. For example, the heating state calculation unit 322a may calculate the initial values ​​of a plurality of candidates for the target amount of dry distillation state using, for example, random numbers.

[0076] <<<Quality index calculation unit 322b>>> The quality index calculation unit 322b calculates a final quality index of the coke based on the first influencing factor and the second influencing factor. In this embodiment, the quality index calculation unit 322b calculates an estimated quality index (DI) of the coke for each of a plurality of candidates for the target carbonization state quantity based on a candidate for the target carbonization state quantity, a heating state quantity (furnace temperature) calculated by providing the candidate to the heating state estimation model 410, a first influencing factor (blending ratio) paired with the candidate for the target carbonization state quantity, and a second influencing factor (coal loading amount and carbonization time) paired with the candidate for the target carbonization state quantity among second influencing factors other than the carbonization state quantity and the heating state quantity. In addition, in this embodiment, the quality index calculation unit 322b includes a first quality index calculation unit 322b1, a change amount calculation unit 322b2, and a second quality index calculation unit 322b3.

[0077] <<<<First quality index calculation unit 322b1>>>> The first quality index calculation unit 322b1 calculates the primary quality index of the coke based on the first influencing factor and the quality estimation model 420.

[0078] In this embodiment, an example is shown in which the first quality index calculation unit 322b1 calculates the primary quality index (DI) of coke for each of multiple candidates for the target distillation state quantity by providing the quality estimation model 420 with the actual value or planned value of the first influencing factor (compound ratio) that is paired with the candidate for the target distillation state quantity.

[0079] <<<<<Change amount calculation unit 322b2>>>> The change amount calculation unit 322b2 calculates the change amount of the primary quality index of the coke based on the second influencing factor and the quality change amount estimation model 430.

[0080] In this embodiment, an example is given in which the change amount calculation unit 322b2 calculates the change amount of the primary quality index (DI) of the coke for each of multiple candidates for the target distillation state quantity by providing the heating state quantity (furnace temperature) calculated by the heating state calculation unit 322a based on the candidate for the target distillation state quantity (coke damage degree) and the planned value or actual measured value of the second influence factor (coal loading amount and distillation time) that is paired with the distillation state quantity and the candidate other than the heating state quantity as second influence factors to the quality change amount estimation model 430.

[0081] <<<<Second quality index calculation unit 322b3>>>> The second quality index calculation unit 322b3 calculates an estimated quality index of the coke based on the primary quality index of the coke calculated by the first quality index calculation unit 322b1 and the change amount of the quality index of the coke calculated by the change amount calculation unit 322b2.

[0082] In this embodiment, an example is given in which the second quality index calculation unit 322b3 calculates an estimated quality index of coke for each of multiple candidates for target distillation state quantity by adding the change amount of the coke quality index calculated by the change amount calculation unit 322b2 for the candidate for target distillation state quantity to the primary quality index (DI) of coke calculated by the first quality index calculation unit 322b1 for the candidate for target distillation state quantity.

[0083] <<<Evaluation Unit 322c>>> The evaluation unit 322c evaluates the difference between the estimated quality index of the coke calculated by the quality index calculation unit 322b using the target carbonization state quantity and the target estimated quality index. In this embodiment, the evaluation unit 322c calculates an evaluation index for evaluating the difference between the estimated quality index (DI) of the coke calculated by the quality index calculation unit 322b (second quality index calculation unit 322b3) and the target estimated quality index for each of a plurality of candidates for the target carbonization state quantity. The evaluation index is, for example, the absolute value of the difference between the estimated quality index (DI) of the coke and the target estimated quality index.

[0084] <<<Goal determination section 322d>>> The target determination unit 322d determines a target quantity of state of pyrolysis based on the result of the evaluation by the evaluation unit 322c. In this embodiment, the target determination unit 322d determines whether the convergence condition is satisfied, and if the convergence condition is not satisfied, requests the heating state calculation unit 322a to change multiple candidates for the target carbonization state quantity (target coke temperature in this embodiment). In this case, the heating state calculation unit 322a updates at least one of the multiple candidates for the target carbonization state quantity based on the request, as described above. Then, using the updated multiple candidates, the processes described in the <<<heating state calculation unit 322a>>>, <<<quality index calculation unit 322b>>>, and <<<evaluation unit 322c>>> sections are executed again. The processes described in the <<<quality index calculation unit 322b>>> and <<<evaluation unit 322c>>> sections and the determination by the target determination unit 322d of whether the convergence condition is satisfied are repeated until the convergence condition is satisfied. In the following description, such repeated processing will be referred to as the "repeated processing" as necessary.

[0085] The convergence condition may be, for example, that the number of iterations reaches a predetermined number. Alternatively, the absolute value of the difference between the sum of absolute values ​​of the differences between the estimated coke quality index (DI) for each of the multiple candidates for the target carbonization state quantity calculated by the evaluation unit 322c in the current iteration and the target estimated quality index, and the sum of absolute values ​​of the differences between the estimated coke quality index (DI) for each of the multiple candidates for the target carbonization state quantity calculated by the evaluation unit 322c in the previous iteration, and the target estimated quality index, is less than or equal to a predetermined value. The convergence condition is not limited to these conditions and may be, for example, a convergence condition determined by a metaheuristic algorithm.

[0086] In this embodiment, when it is determined that the convergence condition is satisfied, the target determination unit 322d determines an optimal value of the target carbonization state quantity based on the latest multiple candidates of the target carbonization state quantity. For example, the target determination unit 322d determines, from the latest multiple candidates of the target carbonization state quantity, the candidate that minimizes the absolute value of the difference between the estimated quality index (DI) of coke and the target estimated quality index, as the optimal value of the target carbonization state quantity.

[0087] As described above, the target calculation unit 322 calculates the optimal value of the target dry distillation state quantity at one of multiple timings after the estimation start timing of the target dry distillation state quantity, for each of the multiple timings. <<Output section 323>> The output unit 323 outputs information indicating the optimal value of the target carbonization state quantity to the output device 302. The output device 302 may be a computer display, a storage medium, a transmission device, or a device equipped with two or three of these. The output device 302 may also be the control device described in Patent Document 2. In this case, the optimal value of the target coke temperature is used instead of the target temperature predetermined as the "coke temperature (i.e., the coke temperature at the time of extrusion)" described in Patent Document 2.

[0088] <Flowchart> Next, an example of a method for creating estimation models (heating state estimation model 410, quality estimation model 420, quality change amount estimation model 430) using the processing device 300 will be described with reference to the flowchart of FIG.

[0089] First, in step S901, the learning data acquisition unit 311 acquires learning data (first learning data, second learning data, and third learning data). The first learning data includes, for example, a set of a measured coke quality index and a measured first influencing factor in the operation at which the measured coke quality index was obtained. The second learning data includes, for example, a set of a measured coke quality index, a measured first influencing factor in the operation at which the measured coke quality index was obtained, and a measured second influencing factor in the operation at which the measured coke quality index was obtained. The third learning data includes, for example, a set of a measured heating state quantity, a measured carbonization state quantity, and a measured operation time.

[0090] Next, in step S902, the quality estimation model creation unit 312 creates (learns) the quality estimation model 420 by performing supervised learning using the first learning data as learning data.

[0091] Next, in step S903, the quality change amount estimation model creation unit 313 calculates an estimated quality index (DI) of the coke by providing the measured first influencing factor (blending conditions) included in the second learning data to the quality estimation model 420 created in step S902. The quality change amount estimation model creation unit 313 calculates a value obtained by subtracting the estimated quality index (DI) of the coke from the measured quality index (DI) of the coke paired with the measured first influencing factor (blending conditions) in the second learning data, as the change amount (correct label) of the primary quality index of the coke. The quality change amount estimation model creation unit 313 sets the change amount of the primary quality index of the coke calculated in this way and the measured second influencing factors (measured oven temperature, measured coke temperature, measured coal amount, and measured carbonization time) paired with the measured first influencing factor (blending conditions) in the second learning data as one piece of learning data. The quality change amount estimation model creation unit 313 creates a plurality of pieces of learning data as such learning data using each of the actually measured first influencing factors (combination conditions) included in the second learning data.

[0092] Next, in step S904, the quality change amount estimation model creation unit 313 creates (learns) the quality change amount estimation model 430 by performing supervised learning using the learning data created in step S903. Next, in step S905, the heating state estimation model creation unit 314 creates (learns) the heating state estimation model 410 by performing supervised learning using the third learning data as learning data.

[0093] When the process of step S905 ends, the process according to the flowchart of FIG. 9 ends. Note that the process of step S905 does not necessarily have to be performed after step S904, as long as it is performed after the process of step S901. For example, the process of step S905 may be performed between step S901 and step S902.

[0094] Next, an example of a method for calculating an optimum value of the target carbonization state quantity (target coke temperature) using the processing device 300 will be described with reference to the flowchart in Fig. 10. Note that Fig. 10 illustrates an example of calculating an optimum value of the target carbonization state quantity at one timing. When calculating optimum values ​​of the target carbonization state quantity at multiple timings, for example, the flowchart in Fig. 10 may be repeated using estimation data corresponding to each timing.

[0095] First, in step S1001, the estimation data acquiring unit 321 acquires estimation data. The estimation data includes, for example, a planned value or an actually measured value of the first influencing factor, a planned value or an actually measured value of the second influencing factor (coal loading amount and carbonization time) other than the carbonization state quantity and the heating state quantity (coke temperature and furnace temperature), a plurality of candidate initial values ​​of the target carbonization state quantity, and a target estimated coke quality index.

[0096] Next, in step S1002, the heating state calculation unit 322a calculates a plurality of candidates for the target carbonization state quantity (coke temperature). In the first step S1002, initial values ​​of the plurality of candidates for the target carbonization state quantity (coke temperature) are selected.

[0097] Next, in step S1003, the heating state calculation unit 322a calculates the heating state quantity (furnace temperature) for each of the multiple candidates calculated in step S1002 by providing the candidate target distillation state quantity calculated in step S1002 and the planned operating time (distillation time) to the heating state estimation model 410.

[0098] Next, in step S1004, the first quality index calculation unit 322b1 calculates a primary quality index (DI) of the coke for each of the multiple candidates for the target carbonization state quantity calculated in step S1002 by providing the quality estimation model 420 with the actual measured value or planned value of the first influencing factor (blending ratio) that is paired with the candidate for the target carbonization state quantity calculated in step S1002. Note that the processing of step S1004 does not necessarily have to be performed after step S1004 as long as it is performed after step S1001. For example, the processing of step S1004 may be performed between step S1001 and step S1002.

[0099] Next, in step S1005, the change amount calculation unit 322b2 calculates the change amount of the coke primary quality index (DI) for each of the multiple candidates for the target distillation state quantity calculated in step S1002 by providing the candidate for the target distillation state quantity (coke damage degree) calculated in step S1002, the heating state quantity (furnace temperature) calculated in step S1003 based on the candidate, and the planned value or actual measured value of the second influence factor (coal loading amount and distillation time) that pairs with the candidate other than the distillation state quantity and the heating state quantity as second influence factors to the quality change amount estimation model 430.

[0100] Next, in step S1006, the second quality index calculation unit 322b3 calculates an estimated quality index of the coke based on the primary quality index of the coke calculated in step S1004 and the change amount of the primary quality index of the coke in step S1005.

[0101] Next, in step S1007, the evaluation unit 322c calculates an evaluation index that evaluates the difference between the estimated quality index (DI) of the coke calculated in step S1006 and the target estimated quality index for each of multiple candidates for the target distillation state quantity.

[0102] Next, in step S1008, the target determination unit 322d determines whether the convergence condition is satisfied. If the result of this determination is that the convergence condition is not satisfied (NO in step S1008), the process of step S1002 is performed again. Then, in step S1002, the heating state calculation unit 322a updates at least one of the multiple candidates for the target carbonization state quantity (coke temperature). Then, the processes of steps S1003 to S1007 are performed using the multiple updated candidates. In this way, the processes of steps S1002 to S1008 are repeated until it is determined in step S1008 that the convergence condition is satisfied.

[0103] Then, as a result of the determination in step S1008, if the convergence condition is satisfied (YES in step S1008), the process of step S1009 is performed. In step S1009, the target determination unit 322d determines an optimal value of the target carbonization state quantity based on the latest multiple candidates of the target carbonization state quantity (i.e., the multiple candidates of the target carbonization state quantity calculated last in step S1002).

[0104] Next, in step S1010, the output unit 323 outputs to the output device 302 information indicating the optimum value of the target amount of heat storage. When the process of step S1010 ends, the process according to the flowchart of FIG. 10 ends.

[0105] <Calculation results> Fig. 11 is a diagram showing an example of the relationship between time and the target coke temperature calculated using operational performance data. The operational performance data is performance data from operations performed with the target coke temperature at a constant value. Fig. 12 is a diagram showing an example of the relationship between time and DI (drum strength index). Note that the timings t1 to t9 on the horizontal axis of Fig. 11 and the timings t1 to t9 on the horizontal axis of Fig. 12 are the same, and the time interval between adjacent scales on the horizontal axis of Fig. 11 is two days.

[0106] Here, the target value of DI (target estimated quality index) was set to aim shown in Fig. 12, and the target coke temperature was searched for within the range of min to max shown in Fig. 11. In this case, in the metaheuristics algorithm described above, a constraint is imposed to exclude target coke temperatures that are not within the range of min to max from the optimal value.

[0107] In Fig. 11, an optimal value 1110 of the target coke temperature indicates the optimal value of the target coke temperature calculated by the method of this embodiment. In Fig. 12, an estimated DI 1210 indicates an estimated value of the DI (an estimated quality index of the coke) calculated by the quality index calculation unit 322b using the optimal value 1110 of the target coke temperature shown in Fig. 11. An actual measured DI 1220 is an actual measured value of the DI included in the above-mentioned operational performance data (performance data from operations performed with the target coke temperature at a constant value).

[0108] For example, in the technology described in Patent Document 2, the target coke temperature is set to a constant value based on past knowledge. Therefore, a difference occurs between the measured DI 1220 in FIG. 12 and the target value aim. In contrast, in the method of the present embodiment, as shown in FIG. 11, the optimal value 1110 of the target coke temperature is calculated by taking into account factors that affect the DI at each timing, and it is possible to search for the optimal value 1110 of the target coke temperature such that the estimated DI 1210 approaches (preferably coincides with) the target value aim. Therefore, by performing operations using such an optimal value 1110, the measured DI can be brought closer to the target value aim.

[0109] <Summary> As described above, in this embodiment, the processing device 300 calculates a heating state quantity, which is a physical quantity corresponding to the heating state from the combustion chamber to the coke chamber, based on the carbonization state quantity, which is a physical quantity corresponding to the carbonization state of the coke, the time from the start of the coal loading operation to the end of the unloading operation or the operation time linked to that time, and the heating state estimation model 410. Therefore, the process of coal carbonization in the operation for which the furnace temperature is to be predicted can be evaluated using the carbonization state quantity and the operation time. This allows for accurate estimation of the heating state from the combustion chamber to the coke chamber. Furthermore, the prediction range is not limited to the next operation (the next unloading operation), and it is possible to predict the heating state quantity for future operations. Furthermore, the type of heating state quantity to be calculated is not limited to a specific type of heating state quantity. Therefore, the heating state quantity can be calculated adaptively depending on the operating mode of the coke oven, etc.

[0110] In this embodiment, the processing apparatus 300 uses a regression equation as the heating state estimation model 410, which has the carbonization state quantity as an explanatory variable and the heating state quantity as a response variable, in which the regression coefficient multiplied by the carbonization state quantity and the intercept are expressed as a function of operation time. An estimation model that can improve estimation accuracy can be realized without using an estimation model that requires complex calculations to estimate the heating state quantity based on the carbonization state quantity. The estimation accuracy can be further improved by at least one of: making the value of the regression coefficient positively correlated with the value of the function; making the value of the intercept negatively correlated with the value of the function; or by having the function include a well-defined monotonically increasing function. Furthermore, when the function includes a well-defined monotonically increasing function, the estimation accuracy can be further improved by using a well-defined monotonically increasing function whose slope in the first and second operation time sections is smaller than the slope in the third operation time section. Furthermore, if a function based on the standard logistic function, such as that shown in K in equation (2), is used as such a broadly monotonically increasing function, the rate of increase (slope) of the monotonically increasing region (second section) can be easily adjusted.

[0111] In this embodiment, the processing device 300 calculates the heating state quantity based on the estimated value of the carbonization state quantity, the estimated value of the operation time, and the heating state estimation model 410. Therefore, the heating state quantity can be predicted early. For example, the heating state quantity can be predicted at least at any one of the following times: before the start of coal loading, between the start of coal loading and before the start of unloading, and between the start of unloading and before the end of unloading.

[0112] Furthermore, in this embodiment, the processing device 300 calculates a primary quality index of the coke based on a first influencing factor, which is a factor affecting coke quality, and the quality estimation model 420; calculates a change amount of the primary quality index of the coke based on a second influencing factor, which is a factor affecting coke quality, and the quality change amount estimation model 430; and calculates an estimated quality index of the coke based on the primary quality index of the coke and the change amount of the primary quality index of the coke. Therefore, the coke quality index (primary quality index) calculated using the quality estimation model 420 can be made closer to (preferably identical to) the actual quality index by the change amount calculated using the quality change amount estimation model 430. This improves the estimation accuracy of the final quality index of the coke. In this case, an influencing factor with a relatively small effect on coke quality may be included in the explanatory variables of the model (quality change amount estimation model 430) that estimates the change amount of the primary quality index of the coke, rather than being included in the explanatory variables of the model (quality estimation model 420) that estimates the coke quality index itself. For example, if the final quality index of coke is DI, and operational factors such as the carbonization state quantity and the heating state quantity (physical quantities that can be obtained as the operational performance of the coke oven 1 after the start of coal loading work and before the start of testing to obtain the final quality index of coke) have a relatively smaller effect on DI than the blending ratio, which is the content ratio of each component contained in the coal, the former operational factor may be used as the explanatory variable (second influencing factor) of the quality change amount estimation model 430, and the latter blending ratio may be used as the explanatory variable (first influencing factor) of the quality estimation model 420. Furthermore, at least one of the carbonization state quantity and the heating state quantity may be adopted as the operational factor.

[0113] Furthermore, in this embodiment, the processing device 300 calculates an estimated quality index (DI) of the coke using a target carbonization state quantity (target coke temperature), and determines the optimal value of the target carbonization state quantity based on the results of evaluating the difference between the calculated estimated quality index of the coke and the target value (target estimated quality index). Therefore, it is possible to calculate the estimated quality index of the coke taking into account factors (influencing factors) that affect the quality of the coke, and to search for an optimal value of the target carbonization state quantity such that the estimated quality index approaches (preferably matches) the target value. Therefore, for example, it is possible to improve the calculation accuracy of the input heat amount when calculating the input heat amount so that the carbonization state quantity approaches (preferably matches) the target carbonization state quantity.

[0114] (Other embodiments) The above-described embodiments of the present invention 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 invention. 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. The embodiments of the present invention can also be realized by a programmable logic controller (PLC) or dedicated hardware such as an application-specific integrated circuit (ASIC). Furthermore, the above-described embodiments of the present invention are merely examples of specific embodiments for carrying out the present invention, and the technical scope of the present invention should not be construed as being limited by these. In other words, the present invention can be embodied in various forms without departing from its technical concept or main features.

[0115] The disclosure of the above embodiment can be implemented as follows, for example. [Disclosure 1] A processing device for performing processing to estimate a heating state quantity, which is a physical quantity corresponding to a heating state from a combustion chamber of a coke oven to a coke chamber, a heating state calculation unit that calculates the heating state quantity based on the carbonization state quantity, the operation time, and a heating state estimation model; The carbonization state amount is a physical amount corresponding to the carbonization state of the coke, The operating time is the time from the start of charging work of charging coal into the coke chamber to the end of discharge work of discharging coke from the coke chamber, or a time linked to that time, The heating state estimation model is a model that represents a relationship between the dry distillation state quantity and the operation time, and the heating state quantity. [Disclosure 2] the heating state estimation model includes a regression equation having the dry distillation state quantity as an explanatory variable and the heating state quantity as a response variable, The processing apparatus according to Disclosure 1, wherein a regression coefficient by which the dry distillation state quantity is multiplied in the regression equation and an intercept in the regression equation are expressed using a function of the operating time. [Disclosure 3] the value of the regression coefficient is positively correlated with the value of the function; 3. The processing device according to claim 2, wherein the value of the intercept is negatively correlated with the value of the function. [Disclosure 4] 4. The processing device according to claim 2 or 3, wherein the function includes a monotonically increasing function. [Disclosure 5] the function includes a broadly monotonically increasing function whose slope in the first interval of the operation time and the second interval of the operation time is smaller than the slope in the third interval of the operation time, the second section is a section following the first section, The processing device according to Disclosure 4, wherein the third section is a section following the second section. [Disclosure 6] The processing device according to any one of Disclosures 2 to 5, wherein the function includes a function based on a standard logistic function. [Disclosure 7] The processing apparatus according to any one of Disclosures 1 to 6, wherein the heating state calculation unit calculates the heating state quantity based on an estimated value of the dry distillation state quantity, an estimated value of the operation time, and the heating state estimation model. [Disclosure 8] A processing device described in any one of Disclosures 1 to 7, wherein the heating state quantity includes the temperature of the carbonization chamber during coke production, the amount of heat input to the combustion chamber during coke production, or the flow rate of fuel gas used to heat the carbonization chamber from the combustion chamber during coke production. [Disclosure 9] The processing device according to any one of Disclosures 1 to 8, wherein the carbonization state quantity includes a temperature of the coke obtained during the discharging operation or a temperature of a wall of the coke oven obtained during the discharging operation. [Disclosure 10] A processing method for estimating a heating state quantity, which is a physical quantity corresponding to a heating state from a combustion chamber of a coke oven to a coke chamber, comprising: a heating state calculation step of calculating the heating state quantity based on the carbonization state quantity, the operation time, and a heating state estimation model; The carbonization state amount is a physical amount corresponding to the carbonization state of the coke, The operating time is the time from the start of charging work of charging coal into the coke chamber to the end of discharge work of discharging coke from the coke chamber, or a time linked to that time, The heating state estimation model is a model that represents a relationship between the dry distillation state quantity, the operation time, and the heating state quantity. [Disclosure 11] 10. A program for causing a computer to function as a heating state calculation unit of the processing device according to any one of Disclosures 1 to 9. [Explanation of symbols]

[0116] 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 300 Processing Equipment 301 Input Device 302 Output Device 310 Creation Department 311 Learning Data Acquisition Unit 312 Quality Estimation Model Creation Department 313 Quality Change Estimation Model Creation Unit 314 Heating state estimation model creation unit 315 Model Memory Unit 320 Calculation Department 321 Estimation Data Acquisition Unit 322 Target Calculation Unit 322a Heating state calculation unit 322b Quality index calculation section 322b1 First quality index calculation section 322b2 Change amount calculation unit 322b3 Second quality index calculation section 322c Evaluation Department 322d Goal determination department 323 Output Section 710 Slope (regression coefficient) 720 Intersection 810 Linear regression equation calculated with regression coefficients and intercepts as constants 820 (1) Linear regression equation 1110 Optimum target coke temperature 1210 Measured DI value (when target coke temperature is constant) 1220 Estimated DI (when target coke temperature is optimal)

Claims

1. A processing device for performing processing to estimate a heating state quantity, which is a physical quantity corresponding to a heating state from a combustion chamber of a coke oven to a coke chamber, a heating state calculation unit that calculates the heating state quantity based on the carbonization state quantity, the operation time, and a heating state estimation model; The carbonization state amount is a physical amount corresponding to the carbonization state of the coke, The operating time is the time from the start of charging work of charging coal into the coke chamber to the end of discharge work of discharging coke from the coke chamber, or a time linked to that time, The heating state estimation model is a model that represents a relationship between the dry distillation state quantity and the operation time, and the heating state quantity.

2. the heating state estimation model includes a regression equation having the dry distillation state quantity as an explanatory variable and the heating state quantity as a response variable, The treatment apparatus according to claim 1 , wherein a regression coefficient by which the dry distillation state quantity is multiplied in the regression equation and an intercept in the regression equation are expressed using a function of the operation time.

3. the value of the regression coefficient is positively correlated with the value of the function; The processing device of claim 2 , wherein the value of the intercept is negatively correlated with the value of the function.

4. The processing device according to claim 2 or 3, wherein the function includes a monotonically increasing function.

5. the function includes a broadly monotonically increasing function whose slope in the first interval of the operation time and the second interval of the operation time is smaller than the slope in the third interval of the operation time, the second section is a section following the first section, The processing device according to claim 4 , wherein the third section is a section following the second section.

6. The processing device of claim 2 or 3, wherein the function comprises a function based on a standard logistic function.

7. The processing apparatus according to any one of claims 1 to 3, wherein the heating state calculation unit calculates the heating state quantity based on an estimated value of the dry distillation state quantity, an estimated value of the operation time, and the heating state estimation model.

8. A processing device described in any one of claims 1 to 3, wherein the heating state quantity includes the temperature of the carbonization chamber during coke production, the amount of heat input to the combustion chamber during coke production, or the flow rate of fuel gas used to heat the carbonization chamber from the combustion chamber during coke production.

9. The processing apparatus according to any one of claims 1 to 3, wherein the carbonization state quantity includes a temperature of the coke obtained during the discharging operation or a temperature of a wall of the coke oven obtained during the discharging operation.

10. A processing method for estimating a heating state quantity, which is a physical quantity corresponding to a heating state from a combustion chamber of a coke oven to a coke chamber, comprising: a heating state calculation step of calculating the heating state quantity based on the carbonization state quantity, the operation time, and a heating state estimation model; The carbonization state amount is a physical amount corresponding to the carbonization state of the coke, The operating time is the time from the start of charging work of charging coal into the coke chamber to the end of discharge work of discharging coke from the coke chamber, or a time linked to that time, The heating state estimation model is a model that represents a relationship between the dry distillation state quantity and the operation time, and the heating state quantity.

11. A program for causing a computer to function as the heating state calculation unit of the processing apparatus according to claim 1 or 2.

Images