Prediction system, prediction method, and prediction program

Abstract

A prediction system according to the present invention: acquires operation data for each of a plurality of past operation cycles and a target operation cycle of a chemical processing apparatus; for each of the plurality of past operation cycles and the target operation cycle, generates a first mathematical model on the basis of the operation data, and acquires a coefficient for each of the operation cycles from the first mathematical model; selects one or more past operation cycles that each have a coefficient satisfying a selection condition which is set on the basis of the coefficient of the target operation cycle; generates a second mathematical model on the basis of the group of operation data of the one or more selected past operation cycles and the target operation cycle; and inputs, to the second mathematical model, an operation condition at a time that is later than any time indicated in the operation data of the target operation cycle, and predicts a state in the target operation cycle at said later time.

Description

Prediction system, prediction method, and prediction program

[0001] One aspect of this disclosure relates to a prediction system, a prediction method, and a prediction program.

[0002] Mechanisms for predicting the state of chemical process equipment using mathematical models have been known for some time. For example, Patent Document 1 describes an ethylene production and cracking furnace operation support system that accepts eigenvalue information of an ethylene production and cracking furnace as input parameters, calculates the surface temperature of the radiating coil at an arbitrary future time using a coil surface temperature estimation model that incorporates a physicochemical model, and obtains a predicted coil surface temperature.

[0003] Patent Document 2 describes a data-driven model that can use one or more key performance indicators (KPIs) to predict both short-term and long-term degradation processes of a chemical production plant, as a function of input parameters including one or more expected operating parameters representing the planned operating conditions of at least one chemical process apparatus and process data derived from sensors available in the production plant.

[0004] International Publication 2021 / 070804, U.S. Patent No. 11860617

[0005] A system is needed to accurately predict the state of chemical process equipment.

[0006] A prediction system relating to one aspect of this disclosure comprises at least one processor. The at least one processor acquires operating data showing the time changes in the operating conditions and state of the chemical process equipment for each of the multiple past operating cycles of the chemical process equipment and a target operating cycle of the chemical process equipment which is an operating cycle later than the multiple past operating cycles. For each of the multiple past operating cycles and the target operating cycle, it generates a first mathematical model based on the operating data, with operating conditions as explanatory variables and state as the dependent variable. For each operating cycle, it acquires coefficients of the first mathematical model showing the relationship between operating conditions and state. It selects one or more past operating cycles having coefficients that satisfy selection criteria set based on the coefficients of the target operating cycle. Based on the set of operating data for the selected one or more past operating cycles and the target operating cycle, it generates a second mathematical model with operating conditions as explanatory variables and state as the dependent variable. It inputs the operating conditions at a time later than any time indicated by the operating data of the target operating cycle into the second mathematical model to predict the state at that later time in the target operating cycle.

[0007] In this respect, coefficients for a first mathematical model showing the relationship between operating conditions and state are obtained for each of the multiple past operating cycles of the chemical process equipment and the target operating cycle. Then, one or more past operating cycles having coefficients that satisfy selection criteria based on the coefficients of the target operating cycle are selected. Based on the set of operating data from the selected past operating cycles and the target operating cycle, a second mathematical model showing the relationship between operating conditions and state is generated, and the state of the chemical process equipment in the target operating cycle is predicted using this second mathematical model. Since the one or more selected past operating cycles have a predetermined relationship with the target operating cycle, it can be said that the operating data of these operating cycles have similar trends. Therefore, the second mathematical model generated using this set of similar operating data is expected to be able to predict the state of the chemical process equipment in the target operating cycle with high accuracy. In other words, the state of the chemical process equipment can be predicted with higher accuracy using this second mathematical model.

[0008] According to one aspect of this disclosure, the state of chemical process equipment can be predicted with high accuracy.

[0009] This is a diagram illustrating an example of the functional configuration of a prediction system. This is a flowchart illustrating an example of the operation of a prediction system. This is a diagram schematically showing an example of a cracking reactor. This is a diagram illustrating an example of the operating trends of a cracking reactor. This is a diagram illustrating an example of selecting past operating cycles that satisfy the selection criteria. This is a diagram illustrating another example of selecting past operating cycles that satisfy the selection criteria.

[0010] The following describes various examples in this disclosure in detail with reference to the attached drawings. In the description of the drawings, identical or equivalent elements are denoted by the same reference numeral, and redundant descriptions are omitted.

[0011] [System Overview] The prediction system described in this disclosure is a computer system that uses mathematical models to predict the state of chemical process equipment. Chemical process equipment refers to equipment used in the chemical industry to carry out processes such as chemical reactions, separation or mixing of substances, and transfer. Chemical process equipment repeatedly operates to perform predetermined processes and is shut down for maintenance. In other words, chemical process equipment operates periodically. In this disclosure, the period from the start to the end of one operation is referred to as an "operation cycle." The prediction system predicts the state of chemical process equipment in a specific operation cycle. In this disclosure, the operation cycle that is the subject of the prediction is referred to as the "target operation cycle."

[0012] The prediction system generates a mathematical model to predict the state of chemical process equipment using not only operating data from the target operating cycle, but also operating data from past operating cycles with similar trends to the target operating cycle. Operating data refers to data showing the operating history of chemical process equipment during an operating cycle. A mathematical model refers to a method of representing real-world phenomena using mathematical formulas or algorithms. The prediction system uses the generated mathematical model to predict the future state of chemical process equipment during the target operating cycle. A mathematical model generated using sets of similar operating data is expected to predict the future state of chemical process equipment during the target operating cycle with greater accuracy than a mathematical model generated using a data set that includes operating data with different trends from the target operating cycle. Therefore, the prediction system can accurately predict the state of chemical process equipment by using its mathematical model.

[0013] [System Configuration] The prediction system consists of one or more computers. When multiple computers are used, these computers are connected via a communication network such as the Internet or an intranet to logically construct a single prediction system.

[0014] A computer comprising a prediction system generally includes a processor, storage device (memory), and communication interface as hardware components. The processor is, for example, a CPU, and the storage device consists of flash memory, a hard disk, etc. The communication interface consists of a network card, a wireless communication module, etc. Each functional module of the prediction system is realized when the processor executes a program stored in the storage device.

[0015] A prediction program for enabling a computer to function as a prediction system includes program code for implementing each functional module of the prediction system. This prediction program may be provided on a non-temporary recording medium such as a CD-ROM, DVD-ROM, or semiconductor memory. Alternatively, the prediction program may be provided via a communication network as a data signal superimposed on a carrier wave. The provided prediction program is then recorded, for example, on a storage device.

[0016] Figure 1 shows the functional configuration of a prediction system 10 in one example. In this example, the prediction system 10 connects to a database 20 via a communication network. The communication network is typically constructed by the internet, an intranet, or a combination thereof. The communication network can be constructed by a wired network, a wireless network, or a combination thereof.

[0017] The database 20 is a storage device that stores operating data for individual operating cycles of chemical process equipment. The operating data for each operating cycle shows the time-dependent changes in the operating conditions and state of the chemical process equipment during that cycle. In one example, each data record of the operating data includes an identifier that uniquely identifies the operating cycle, time, operating conditions, and state as data items. The database 20 stores operating data for at least several past operating cycles and for the target operating cycle, which is an operating cycle that is later than those past operating cycles. The database 20 may be a component of the prediction system 10 or it may be located outside of the prediction system 10.

[0018] The prediction system 10 includes a processor 101 that functions as an acquisition unit 11, a first model generation unit 12, an operation cycle selection unit 13, a second model generation unit 14, and a prediction unit 15.

[0019] The acquisition unit 11 is a functional module that acquires operating data of chemical process equipment from the database 20.

[0020] The first model generation unit 12 is a functional module that generates a first mathematical model based on the operation data of each of a plurality of past operation cycles and the target operation cycle, and obtains the coefficients of the first mathematical model for each of these operation cycles. The first mathematical model is a mathematical model that uses operation conditions as explanatory variables and states as objective variables. That is, the first mathematical model calculates the state based on the operation conditions.

[0021] The operation cycle selection unit 13 is a functional module that selects one or more past operation cycles having coefficients that satisfy the selection conditions set based on the coefficients of the target operation cycle.

[0022] The second model generation unit 14 is a functional module that generates a second mathematical model based on the set of operation data of each of the selected one or more past operation cycles and the target operation cycle. Similar to the first mathematical model, the second mathematical model is also a mathematical model that uses operation conditions as explanatory variables and states as objective variables. That is, the second mathematical model calculates the state based on the operation conditions.

[0023] The prediction unit 15 is a functional module that inputs the operation conditions at a time later than any time indicated by the operation data of the target operation cycle into the second mathematical model, and predicts the state at the later time in the target operation cycle.

[0024] [System operation] The prediction method executed by the prediction system 10 will be described while referring to FIG. 2. FIG. 2 is a flowchart showing an example of the operation of the prediction system 10 as a processing flow S1.

[0025] In step S11, the acquisition unit 11 acquires operation data for each of the multiple past operation cycles of the chemical process equipment and the target operation cycle. For example, the acquisition unit 11 identifies the current operation cycle, which has started but has not yet finished, as the target operation cycle. The acquisition unit 11 then acquires the operation data for the identified target operation cycle and the operation data for each of the multiple past operation cycles prior to the target operation cycle. The acquisition unit 11 may acquire operation data for all past operation cycles from the database 20, or it may acquire operation data for some of the past operation cycles.

[0026] In step S12, the first model generation unit 12 generates a first mathematical model for each of the multiple past operating cycles and the target operating cycle based on the operating data of the operating cycle and obtains the coefficients of the first mathematical model. The coefficients of the first mathematical model indicate the relationship between operating conditions and states. In this disclosure, the coefficients of the first mathematical model for an operating cycle are also simply referred to as "operating cycle coefficients".

[0027] For example, the first model generation unit 12 may generate a linear regression model as the first mathematical model. In the case of a simple regression model, the first mathematical model is expressed as y = ax + b, where x represents the operating conditions, y represents the state, a represents the regression coefficient, and b represents the intercept. In the case of a multiple regression model, the first mathematical model is expressed as y = a １ x １ +a ２ x ２ +...+a ｎ x ｎ This is expressed as +b. The first model generation unit 12 performs linear regression analysis on the operating data for each operating cycle and generates a linear regression model for that operating cycle as the first mathematical model. For example, the first model generation unit 12 may generate a linear regression model with an intercept of 0. In the case of a simple regression model, the first model generation unit 12 obtains the regression coefficient a for each operating cycle as the coefficient of the first mathematical model. In the case of a multiple regression model, the first model generation unit 12 obtains the regression coefficient a of one explanatory variable of interest for each operating cycle. ｉ You may obtain this as the coefficient of the first mathematical model.

[0028] As another example, the first model generation unit 12 may generate a machine learning model as the first mathematical model. In this case, for each operation cycle, the first model generation unit 12 executes machine learning using the operation data in the operation cycle, and generates a machine learning model that accepts an input of operation conditions and predicts a state as the first mathematical model. For each operation cycle, the first model generation unit 12 obtains, as the coefficient of the first mathematical model, a SHAP (SHapley Additive exPlanations) value indicating how much a feature amount (operation condition) input as an explanatory variable to the machine learning model contributes to the prediction (prediction of state) of the machine learning model. The first model generation unit 12 obtains the SHAP value of one explanatory variable of interest as the coefficient of the first mathematical model.

[0029] In step S13, the operation cycle selection unit 13 selects one or more past operation cycles based on the coefficient of the first mathematical model of the target operation cycle. The operation cycle selection unit 13 selects one or more past operation cycles having coefficients that satisfy the selection conditions set based on the coefficient of the target operation cycle.

[0030] For example, the selection condition may be defined using a lower limit value L that is a value smaller than the coefficient a of the target operation cycle. ｔｇｔ In this case, a selection condition that the coefficient of the past operation cycle is equal to or greater than the lower limit value L is used. The operation cycle selection unit 13 selects one or more past operation cycles having coefficients that are equal to or greater than the lower limit value L. ｌｏｗ ｌｏｗ ｌｏｗ

[0031] Alternatively, the selection condition may be defined using an upper limit value L that is a value greater than the coefficient a in addition to the lower limit value L. In this case, a selection condition that the coefficient of the past operation cycle is within the range from the lower limit value L to the upper limit value L is used. The range is such that the difference between the coefficient a and the upper limit value L is ｌｏｗ ｔｇｔ ｈｉｇｈ ｌｏｗ ｈｉｇｈ ｔｇｔ ｈｉｇｈ ｔｇｔ ​​​​​​​​​​and lower limit L ｌｏｗ The operating cycle selection unit 13 may be set to be greater than the difference between the two. ｌｏｗ From upper limit L ｈｉｇｈ Select one or more past driving cycles that have a coefficient within the specified range.

[0032] In step S14, the second model generation unit 14 generates a second mathematical model based on the set of operating data for one or more selected past operating cycles and the target operating cycle. Hereinafter, this set will also be simply referred to as the "data set". Similar to the first mathematical model, the second mathematical model may be a linear regression model or a machine learning model. The types of models in the first mathematical model and the second mathematical model may be the same or different. For example, if the first mathematical model is a linear regression model, the second mathematical model may be either a linear regression model or a machine learning model. If the first mathematical model is a machine learning model, the second mathematical model may be either a machine learning model or a linear regression model.

[0033] For example, the second model generation unit 14 may generate a linear regression model as the second mathematical model. In this case, the second mathematical model will be similar to the first mathematical model, with y = ax + b or y = a １ x １ +a ２ x ２ +...+a ｎ x ｎ This is expressed as +b. The second model generation unit 14 reads a data set from the database 20 and performs linear regression analysis on that data set to generate a linear regression model as the second mathematical model. The second model generation unit 14 may generate a linear regression model with an intercept of 0, or it may generate a linear regression model with an intercept of non-zero.

[0034] As another example, the second model generation unit 14 may generate a machine learning model as the second mathematical model. In this case, the second model generation unit 14 performs machine learning using the data set read from the database 20 to generate a machine learning model as the second mathematical model that accepts input of operating conditions and predicts the state.

[0035] In step S15, the prediction unit 15 inputs the operating conditions at a time later than any time indicated by the operating data of the target operating cycle into the second mathematical model to predict the state of the chemical process equipment at that later time in the target operating cycle. Hereinafter, "later time" will also be referred to as "target time". If the target operating cycle is the current operating cycle that has not yet finished, the target time is a future time. For example, the prediction unit 15 accepts user input indicating the operating conditions at the target time. The prediction unit 15 inputs these operating conditions into the second mathematical model and obtains the state calculated by the second mathematical model as the state of the chemical process equipment at the target time. The prediction unit 15 may input the operating conditions at each of two or more target times into the second mathematical model to predict the state at each target time. The prediction unit 15 outputs the predicted state. The prediction unit 15 may display the predicted state on a display device, store the predicted state in a given database, or transmit the predicted state to another computer or computer system. The manager or operator of chemical process equipment can adjust operating conditions as needed based on predicted conditions.

[0036] [Application to Decomposition Furnaces] Chemical process equipment can be decomposition furnaces. An example of predicting the state of such a decomposition furnace will be explained with reference to Figures 3 to 6. Figure 3 is a schematic diagram showing an example of a decomposition furnace. Figure 4 is a diagram showing an example of the operating trend of a decomposition furnace. Figures 5 and 6 are diagrams that illustrate an example of selecting a past operating cycle that satisfies selection criteria.

[0037] As shown in Figure 3, the decomposition furnace 200 comprises a reaction tube 210 and a heating unit 220 for heating the reaction tube 210. In the decomposition furnace 200, raw materials and steam are introduced into the reaction tube 210, and the introduced raw materials are thermally decomposed near the heating unit 220 to produce decomposition gas. The produced decomposition gas passes through the outlet 211 of the reaction tube 210 and is supplied to the downstream process. The raw materials may be, for example, hydrocarbon compounds such as naphtha. The decomposition gas may be, for example, ethylene, propylene, butadiene, benzene, toluene, xylene, etc.

[0038] As the raw materials are thermally decomposed, carbon deposits on the inside of the reaction tube 210, gradually forming a coke layer. Because the coke layer has low thermal conductivity, heat conduction is inhibited, causing the surface temperature of the reaction tube 210 to gradually increase during the decomposition gas generation process. To prevent damage to the reaction tube 210 due to this temperature rise, an upper limit is set for its surface temperature. In the operation of the decomposition furnace 200, before the surface temperature reaches this upper limit, the furnace 200 is stopped, and a decoking operation is performed to burn off the deposited carbon and remove the coke layer. Conventionally, the operating conditions of the decomposition furnace 200 are adjusted to efficiently produce decomposition gas before the decoking operation. By using the prediction system 10, the state of the decomposition furnace 200 can be predicted with high accuracy, making it possible to adjust the operating conditions to produce decomposition gas even more efficiently. The following describes an example of predicting the state of the decomposition furnace 200 using the prediction system 10.

[0039] The database 20 stores operating data for each operating cycle of the reaction tube 210. The operating conditions of the cracking furnace 200, as indicated by the operating data, may include at least one of the following: the flow rate of the raw materials, the flow rate of the steam, the amount of cracking gas produced, the heating temperature of the reaction tube 210 by the heating unit 220, and the integrated value of the coil outlet temperature, which is the temperature at the outlet 211 of the reaction tube 210. The state of the cracking furnace 200, as indicated by the operating data, may include at least one of the pressure and flow rate at the inlet 212 of the reaction tube 210. The pressure at the inlet 212 of the reaction tube 210 can be measured by a Venturi gauge. The length of each operating cycle of the cracking furnace 200 is, for example, 5 to 100 days. In one example, data records of the operating data are generated every hour and stored in the database 20.

[0040] In one example, the prediction system 10 generates mathematical models (first mathematical model and second mathematical model) with the integrated value of the coil outlet temperature as the explanatory variable and the pressure at the inlet 212 of the reaction tube 210 as the dependent variable. Hereinafter, the integrated value of the coil outlet temperature will be referred to as the "COT integrated value," and the pressure at the inlet 212 of the reaction tube 210 will simply be referred to as the "inlet pressure." Since the inlet pressure increases as the coke layer thickens, it can be said that it changes over time in the same way as the surface temperature of the reaction tube 210. Therefore, by looking at the inlet pressure instead of the surface temperature, the operating conditions of the decomposition furnace 200 can be adjusted for the efficient production of decomposition gas.

[0041] The cumulative COT value at a specific time in an operating cycle is expressed as the sum of the coil outlet temperatures (COT) from the start of the operating cycle to that specific time. For example, if the COT changes from 10, 20, 10, 30, ... at each point in time from the start of an operating cycle, the cumulative COT value for that operating cycle will change as follows: 10, 30, 40, 70, ...

[0042] Figure 4 shows a graph illustrating the relationship between the time change of the inlet pressure 310 and the time change of the COT integrated value 320, as a trend in the operation of the decomposition furnace 200. The horizontal axis, left vertical axis, and right vertical axis of the graph represent elapsed time, inlet pressure, and COT integrated value, respectively. As shown in this graph, the inlet pressure 310 and the COT integrated value 320 are correlated with each other. Therefore, by preparing a mathematical model to calculate this correlation, it becomes possible to accurately predict the inlet pressure from the COT integrated value. The operator of the decomposition furnace 200 can adjust the operating conditions of the decomposition furnace 200 by referring to this prediction.

[0043] In one example, the prediction system 10 predicts the state of the cracking furnace 200 as follows. In this example, it is assumed that the operating data for each operating cycle shows the time change of the COT integrated value and inlet pressure of the cracking furnace 200 during that operating cycle.

[0044] In step S11, the acquisition unit 11 acquires operation data from the database 20 for each of the multiple past operation cycles of the decomposition furnace 200 and the target operation cycle.

[0045] In step S12, the first model generation unit 12 generates a first mathematical model for each of the multiple past operating cycles and the target operating cycle based on the operating data of the operating cycle and obtains the coefficients of the first mathematical model. In this example, the first mathematical model includes at least the COT integrated value as an explanatory variable, and the objective variable is the inlet pressure. The explanatory variables may further include other elements such as the flow rate of the raw material. The first model generation unit 12 obtains the coefficient of the COT integrated value as the coefficient of the first mathematical model.

[0046] In step S13, the operation cycle selection unit 13 selects one or more past operation cycles based on the coefficients of the first mathematical model of the target operation cycle. This selection process will be explained with reference to Figures 5 and 6. Both Figures 5 and 6 show the coefficients of the first mathematical model (coefficients of the COT integrated value) for each of the multiple operation cycles as bar graphs. In both graphs, the horizontal axis represents the individual operation cycle, and the vertical axis represents the coefficients.

[0047] Figure 5 shows the coefficient for the target operating cycle 410 and the coefficients for multiple past operating cycles 420. In this example, the operating cycle selection unit 13 selects a lower limit value L that is smaller than the coefficient for the target operating cycle 410. ｌｏｗ This defines the lower limit value L, for example, the operating cycle selection unit 13 sets a value corresponding to 80% of the coefficient of the target operating cycle 410. ｌｏｗ This is defined as follows. In this example, the operating cycle selection unit 13 has an upper limit value L that is greater than the coefficient of the target operating cycle 410. ｈｉｇｈ The operating cycle selection unit 13 does not define the lower limit L. ｌｏｗ Select a past operating cycle having a coefficient greater than or equal to the above, and set the lower limit L. ｌｏｗ Past operating cycles with coefficients less than a certain value are not selected. In Figure 5, selected operating cycles are shown with black bars, and non-selected operating cycles are shown with hatched bars.

[0048] Figure 6 shows the coefficient for the target operating cycle 510 and the coefficients for multiple past operating cycles 520. In this example, the operating cycle selection unit 13 selects a lower limit value L that is smaller than the coefficient for the target operating cycle 510. ｌｏｗAnd the upper limit L is a value greater than the coefficient. ｈｉｇｈ This defines the lower limit value L, for example, the operating cycle selection unit 13 sets a value equivalent to 80% of the coefficient of the target operating cycle 510. ｌｏｗ Defined as such, the upper limit value L is the value equivalent to 200% of the coefficient. ｈｉｇｈ This is defined as follows. As a result, the coefficient and upper limit L of the target operating cycle 510 ｈｉｇｈ The difference between the coefficient and the lower limit L is the coefficient and the lower limit L. ｌｏｗ The difference is greater than that. The operating cycle selection unit 13 has a lower limit value L ｌｏｗ From upper limit L ｈｉｇｈ Past operating cycles with coefficients within a certain range are selected, and past operating cycles with coefficients outside that range are not selected. In Figure 6, selected operating cycles are shown with black bars, and unselected operating cycles are shown with hatched bars. In the operating data of past operating cycles with relatively large coefficients, the error due to the progression of coking in the reaction tube 210 as the operating time elapses is small. In one example, the upper limit L ｈｉｇｈ By using a range set to provide a larger margin on the other side, it becomes possible to select more driving data with smaller errors.

[0049] In step S14, the second model generation unit 14 generates a second mathematical model based on the set of operating data for one or more selected past operating cycles and the target operating cycle. In this example, the explanatory variable of the second mathematical model is the COT integrated value, and the dependent variable is the inlet pressure.

[0050] In step S15, the prediction unit 15 inputs the operating conditions at a time later than any of the times indicated by the operating data of the target operating cycle (target time) into the second mathematical model to predict the state of the decomposition furnace 200 at that later time (target time) in the target operating cycle. For example, for each of the one or more target times, the prediction unit 15 receives user input indicating the operating conditions including the COT integrated value at that target time, inputs those operating conditions into the second mathematical model, and predicts the inlet pressure at that target time using the second mathematical model.

[0051] [Variations] The technology relating to this disclosure has been described in detail above based on various examples. However, this disclosure is not limited to the above examples. Various modifications are possible to the technology relating to this disclosure without departing from its essence.

[0052] The processing steps for a method executed by at least one processor are not limited to the examples above. For example, some of the steps described above may be omitted, or each step may be performed in a different order. Also, any two or more of the steps described above may be combined, or some of the steps may be modified or deleted. Alternatively, other steps may be performed in addition to each of the steps described above.

[0053] In comparing the relative magnitudes of two numerical values ​​in this disclosure, either of the two criteria, "greater than or equal to" and "greater than," may be used, or either of the two criteria, "less than or equal to" and "less than," may be used.

[0054] In this disclosure, the expression "at least one processor executes a first process, a second process, ... and the nth process," or a corresponding expression, refers to a concept that includes cases where the entity executing the n processes from the first process to the nth process, i.e., the processor, changes along the way. In other words, this expression refers to a concept that includes both cases where all n processes are executed by the same processor and cases where the processor changes at an arbitrary rate for the n processes.

[0055] [Note] As can be seen from the various examples above, this disclosure includes the following aspects: (Note 1) A prediction system comprising at least one processor, wherein the at least one processor acquires operating data showing the time changes in the operating conditions and state of the chemical process equipment for each of the multiple past operating cycles of the chemical process equipment and a target operating cycle of the chemical process equipment which is an operating cycle later than the multiple past operating cycles; generates a first mathematical model based on the operating data for each of the multiple past operating cycles and the target operating cycle, with the operating conditions as explanatory variables and the state as the dependent variable; acquires coefficients of the first mathematical model showing the relationship between the operating conditions and the state for each operating cycle; selects one or more of the past operating cycles having coefficients that satisfy selection conditions set based on the coefficients of the target operating cycle; generates a second mathematical model based on the set of operating data for each of the selected one or more past operating cycles and the target operating cycle, with the operating conditions as explanatory variables and the state as the dependent variable; and inputs the operating conditions at a time later than any time shown by the operating data of the target operating cycle into the second mathematical model to predict the state at a later time in the target operating cycle. (Note 2) The prediction system according to Note 1, wherein the chemical process equipment comprises a reaction tube, and the state includes at least one of the pressure and flow rate at the inlet of the reaction tube. (Note 3) The prediction system according to Note 2, wherein the operating conditions include the integrated value of the coil outlet temperature, which is the temperature at the outlet of the reaction tube. (Note 4) The prediction system according to Note 2 or 3, wherein the selection condition is defined using a lower limit that is smaller than the coefficient of the target operating cycle, and at least one processor selects one or more past operating cycles having the coefficient that is greater than or equal to the lower limit.(Note 5) The prediction system according to Note 4, wherein the selection condition is further defined using an upper limit which is greater than the coefficient of the target operating cycle, and at least one processor selects one or more past operating cycles having the coefficient which is within the range from the lower limit to the upper limit. (Note 6) The prediction system according to Note 5, wherein the difference between the coefficient of the target operating cycle and the upper limit is greater than the difference between the coefficient of the target operating cycle and the lower limit. (Note 7) The prediction system according to any one of Notes 1 to 6, wherein the first mathematical model is a linear regression model with an intercept of 0. (Note 8) A prediction method performed by a prediction system comprising at least one processor, comprising: acquiring operating data showing the time changes in the operating conditions and state of a chemical process equipment for each of a plurality of past operating cycles of the chemical process equipment and a target operating cycle of the chemical process equipment which is an operating cycle later than the plurality of past operating cycles; generating a first mathematical model based on the operating data for each of the plurality of past operating cycles and the target operating cycle, with the operating conditions as explanatory variables and the state as the objective variable, and acquiring coefficients of the first mathematical model showing the relationship between the operating conditions and the state for each operating cycle; selecting one or more of the past operating cycles having coefficients that satisfy selection conditions set based on the coefficients of the target operating cycle; generating a second mathematical model based on the set of operating data for each of the selected one or more past operating cycles and the target operating cycle, with the operating conditions as explanatory variables and the state as the objective variable; and inputting the operating conditions at a time later than any time shown by the operating data of the target operating cycle into the second mathematical model to predict the state at a later time in the target operating cycle. A prediction method that includes this.(Note 9) A prediction program that causes a computer to perform the following steps: acquiring operating data showing the time changes in the operating conditions and state of a chemical process equipment for each of several past operating cycles of the chemical process equipment and a target operating cycle of the chemical process equipment which is an operating cycle later than the several past operating cycles; generating a first mathematical model based on the operating data for each of the several past operating cycles and the target operating cycle, with the operating conditions as explanatory variables and the state as the dependent variable, and acquiring coefficients of the first mathematical model showing the relationship between the operating conditions and the state for each operating cycle; selecting one or more of the past operating cycles having coefficients that satisfy selection conditions set based on the coefficients of the target operating cycle; generating a second mathematical model based on the set of operating data for each of the selected one or more past operating cycles and the target operating cycle, with the operating conditions as explanatory variables and the state as the dependent variable; and inputting the operating conditions at a time later than any of the times shown by the operating data of the target operating cycle into the second mathematical model to predict the state at a later time in the target operating cycle.

[0056] According to appendices 1, 8, and 9, coefficients for a first mathematical model showing the relationship between operating conditions and state are obtained for each of the multiple past operating cycles of the chemical process equipment and the target operating cycle. Then, one or more past operating cycles having coefficients that satisfy selection criteria based on the coefficients of the target operating cycle are selected. Based on the set of operating data from the selected past operating cycles and the target operating cycle, a second mathematical model showing the relationship between operating conditions and state is generated, and the state of the chemical process equipment in the target operating cycle is predicted using this second mathematical model. Since the one or more selected past operating cycles have a predetermined relationship with the target operating cycle, it can be said that the operating data of these operating cycles have similar trends. Therefore, the second mathematical model generated using this set of similar operating data is expected to be able to predict the state of the chemical process equipment in the target operating cycle with high accuracy. In other words, the state of the chemical process equipment can be predicted with higher accuracy using this second mathematical model.

[0057] According to Appendix 2, for chemical process equipment equipped with a reaction tube, at least one of the pressure and flow rate at the inlet of the reaction tube can be predicted with greater accuracy.

[0058] According to Appendix 3, at least one of the pressure and flow rate at the reaction tube inlet can be predicted based on operating conditions including the integrated coil outlet temperature. The integrated coil outlet temperature and the inlet pressure tend to correlate with each other, and the flow rate changes with pressure. Therefore, by using the integrated coil outlet temperature as an explanatory variable in the mathematical model, at least one of the pressure and flow rate at the reaction tube inlet can be predicted with greater accuracy.

[0059] According to Appendix 4, one or more past operating cycles are selected based on a lower limit set using the coefficient of the target operating cycle as a reference. Considering that the reaction in the reaction tube progresses as the operating time of the chemical process equipment increases, the coefficient of the target operating cycle is likely to increase as the operating time increases. Therefore, by increasing the margin for selection for past operating cycles with a larger coefficient than the target operating cycle, operating data from past operating cycles with similar trends to the target operating cycle can be selected more appropriately. As a result, the second mathematical model can be generated with greater accuracy, and the state of the reaction tube can be predicted with greater accuracy.

[0060] According to Appendix 5, one or more past operating cycles are selected based on lower and upper limits set using the coefficient of the target operating cycle as a reference. By defining the range in this way, operating data from past operating cycles with similar trends to the target operating cycle are selected, allowing for the more accurate generation of the second mathematical model. As a result, the state of the reaction tube can be predicted with greater accuracy.

[0061] According to Appendix 6, the range is set by lower and upper limits to provide a greater margin for selection of past operating cycles with larger coefficients than the target operating cycle. As mentioned above, considering that the reaction in the reaction tube progresses as the operating time of the chemical process equipment increases, the coefficient of the target operating cycle is likely to increase as the operating time increases. Therefore, by providing a greater margin for selection of past operating cycles with larger coefficients than the target operating cycle, it is possible to more appropriately select operating data from past operating cycles that have similar trends to the target operating cycle. As a result, the second mathematical model can be generated with greater accuracy, and the state of the reaction tube can be predicted with greater accuracy.

[0062] According to Appendix 7, a linear regression model with an intercept of 0 is generated as the first mathematical model. By unifying the intercept to 0 among the first mathematical models for each operating cycle, the coefficients of the first mathematical models being compared are standardized between the target operating cycle and multiple past operating cycles. Therefore, one or more past operating cycles can be selected more accurately based on the selection criteria, using the coefficients of the target operating cycle. As a result, the second mathematical model is generated with greater accuracy, and the state of chemical process equipment can be predicted with greater accuracy.

[0063] 10...Prediction system, 11...Acquisition unit, 12...First model generation unit, 13...Operation cycle selection unit, 14...Second model generation unit, 15...Prediction unit, 20...Database, 200...Decomposition furnace, 210...Reaction tube, 211...Reaction tube outlet, 212...Reaction tube inlet, 220...Heating unit.

Claims

1. A prediction system comprising at least one processor, wherein the at least one processor acquires operating data showing the time changes in the operating conditions and state of a chemical process equipment for each of a plurality of past operating cycles of the chemical process equipment and a target operating cycle of the chemical process equipment which is an operating cycle later than the plurality of past operating cycles; generates a first mathematical model based on the operating data for each of the plurality of past operating cycles and the target operating cycle, with the operating conditions as explanatory variables and the state as the dependent variable; acquires coefficients of the first mathematical model showing the relationship between the operating conditions and the state for each operating cycle; selects one or more of the past operating cycles having coefficients that satisfy selection conditions set based on the coefficients of the target operating cycle; generates a second mathematical model based on the set of operating data for each of the selected one or more past operating cycles and the target operating cycle, with the operating conditions as explanatory variables and the state as the dependent variable; and inputs the operating conditions at a time later than any time shown by the operating data of the target operating cycle into the second mathematical model to predict the state at a later time in the target operating cycle.

2. The prediction system according to claim 1, wherein the chemical process equipment comprises a reaction tube, and the state includes at least one of the pressure and flow rate at the inlet of the reaction tube.

3. The prediction system according to claim 2, wherein the operating conditions include the integrated value of the coil outlet temperature, which is the temperature at the outlet of the reaction tube.

4. The prediction system according to claim 2 or 3, wherein the selection condition is defined using a lower limit which is a value smaller than the coefficient of the target operating cycle, and at least one processor selects one or more past operating cycles having the coefficient which is greater than or equal to the lower limit.

5. The prediction system according to claim 4, wherein the selection condition is further defined using an upper limit which is greater than the coefficient of the target operating cycle, and at least one processor selects one or more past operating cycles having the coefficient which is within the range from the lower limit to the upper limit.

6. The prediction system according to claim 5, wherein the difference between the coefficient of the target operating cycle and the upper limit is greater than the difference between the coefficient of the target operating cycle and the lower limit.

7. The prediction system according to any one of claims 1 to 3, wherein the first mathematical model is a linear regression model with an intercept of 0.

8. A prediction method performed by a prediction system comprising at least one processor, comprising: acquiring operating data showing the time changes in the operating conditions and state of a chemical process equipment for each of a plurality of past operating cycles of the chemical process equipment and a target operating cycle of the chemical process equipment which is an operating cycle later than the plurality of past operating cycles; generating a first mathematical model based on the operating data for each of the plurality of past operating cycles and the target operating cycle, with the operating conditions as explanatory variables and the state as the dependent variable, and acquiring coefficients of the first mathematical model showing the relationship between the operating conditions and the state for each operating cycle; selecting one or more of the past operating cycles having coefficients that satisfy selection conditions set based on the coefficients of the target operating cycle; generating a second mathematical model based on the set of operating data for each of the selected one or more past operating cycles and the target operating cycle, with the operating conditions as explanatory variables and the state as the dependent variable; and inputting the operating conditions at a time later than any time indicated by the operating data of the target operating cycle into the second mathematical model to predict the state at a later time in the target operating cycle.

9. A prediction program that causes a computer to perform the following steps:

1. Acquire operating data showing the time changes in the operating conditions and state of a chemical process machine for each of several past operating cycles of the chemical process machine and a target operating cycle of the chemical process machine which is an operating cycle later than the several past operating cycles; 2. For each of the several past operating cycles and the target operating cycle, generate a first mathematical model based on the operating data, in which the operating conditions are explanatory variables and the state is the dependent variable, and for each operating cycle, acquire coefficients of the first mathematical model showing the relationship between the operating conditions and the state; 3. Select one or more of the past operating cycles having coefficients that satisfy selection conditions set based on the coefficients of the target operating cycle; 4. Generate a second mathematical model based on the set of operating data for each of the selected one or more past operating cycles and the target operating cycle, in which the operating conditions are explanatory variables and the state is the dependent variable; 5. Input the operating conditions at a time later than any of the times shown by the operating data of the target operating cycle into the second mathematical model to predict the state at a later time in the target operating cycle.

Images