Procedure creation assistance device, procedure creation assistance method, and procedure creation assistance program

Abstract

The present invention assists in determining an operation condition in non-steady operation of a plant. This procedure creation assistance device includes a processing unit that: acquires process data outputted in past non-steady operation of a plant; creates a trained model in which a relationship with a prescribed objective variable is learned by regression analysis using a feature amount based on the process data as an explanatory variable; performs a simulation using the trained model and an assumed value of the explanatory variable; outputs a condition serving as a predetermined constraint with respect to a regression coefficient of the trained model, a simulation result, and the explanatory variable; and corrects the trained model in cases where an operation for correcting the explanatory variable is received from a user, or performs a simulation using a corrected assumed value in cases where an operation for correcting the assumed value is received from the user.

Description

Procedure creation support device, procedure creation support method, and procedure creation support program

[0001] The present disclosure relates to a procedure creation support device, a procedure creation support method, and a procedure creation support program.

[0002] A system for optimizing plant operating conditions has been proposed (see Patent Document 1). In this system, a measurement data recording unit records, in a measurement database, a series of measurement data in which the operating state data acquired by the operating state data acquisition unit and the operating index data calculated by the operating index data acquisition unit are associated based on predetermined items. A regression model creation unit creates a regression model by performing a predetermined multivariate analysis using operating state variables representing the operating state data as explanatory variables and operating index variables representing the operating index data as objective variables. An operation index variable optimization unit calculates operating state variables that optimize the operation index variables based on the regression model.

[0003] JP 2012-74007 A

[0004] For example, in unsteady operation such as during the start-up of a plant from the start of operation to steady operation, the causal relationships between the effects of operating conditions on product quality and cost are complex, making it difficult to find desirable operating conditions. Therefore, an object of the present disclosure is to provide a technology for supporting the determination of operating conditions during unsteady operation of a plant.

[0005] The procedure creation support device according to the present disclosure can be realized by the following aspects. (Aspect 1) A procedure creation support device comprising a processing unit that performs the following steps: acquires process data output during past unsteady operation of a plant; creates a trained model that learns a relationship between a predetermined objective variable and feature quantities based on the process data as explanatory variables through regression analysis; performs a simulation using the trained model and expected values ​​of the explanatory variables; outputs regression coefficients of the trained model, simulation results, and conditions that serve as predetermined constraints on the explanatory variables; and modifies the trained model when a user operation to modify the explanatory variables is received, or modifies the expected values ​​when a user operation to modify the expected values ​​is received. (Aspect 2) In the procedure creation support device according to Aspect 1, the processing unit may output the regression coefficients of the trained model or the absolute values ​​of the regression coefficients as an index of the magnitude of the influence of the explanatory variables, by which the regression coefficients are multiplied in the trained model, on a change in the objective variable. (Aspect 3) In the procedure creation support device according to Aspect 1 or 2, the constraint may include being within a range of an explanatory variable based on process data output during past unsteady operation. (Aspect 4) In the procedure creation support device according to any one of Aspects 1 to 3, when a user specifies the sign of a regression coefficient, the trained model performs regression analysis using a regularization term that increases the cost when the sign of the regression coefficient differs from the user's specification, to minimize a cost function including the regularization term. The constraint may include causal relationship information regarding a direction in which one explanatory variable should be increased or decreased in order to change the target variable to a value in steady operation of the plant. (Aspect 5) In the procedure creation support device according to Aspect 4, the causal relationship information may include multiple pieces of information indicating directions in which the explanatory variable should be increased or decreased based on different perspectives.

[0006] The content of the means for solving the problem can be provided as a device such as a computer or a system including multiple devices, a method executed by a computer, or a program executed by a computer. A recording medium storing the program may also be provided.

[0007] The disclosed technology can provide a technology for supporting the determination of operating conditions during unsteady operation of a plant.

[0008] FIG. 1 is a diagram illustrating an example of a system. FIG. 2 is a block diagram illustrating an example of the configuration of an information processing device. FIG. 3 is a schematic diagram illustrating an example of a process performed by equipment provided in a plant. FIG. 4 is a diagram illustrating an example of process data in a batch process. FIG. 5 is a diagram illustrating an example of process data in a continuous process. FIG. 6 is a process flow diagram illustrating an example of a scenario creation support process executed by an information processing device. FIG. 7 is a diagram illustrating an example of a schematic scenario. FIG. 8 is a diagram illustrating an example of operating conditions. FIG. 9 is a diagram illustrating an example of feature quantities.

[0009] Hereinafter, an embodiment of an information processing device (procedure creation support device) will be described with reference to the drawings.

[0010] <Embodiment> FIG. 1 is a diagram illustrating an example of a system according to this embodiment. The system 100 includes an information processing device 1, a control station 2, and a plant 3. The system 100 is, for example, a distributed control system (DCS) and includes multiple control stations 2. That is, the control system of the plant 3 is divided into multiple sections, and each control section is controlled in a distributed manner by the control station 2. The control station 2 is an existing facility in the DCS, and receives status signals output from sensors and the like provided in the plant 3 and outputs control signals to the plant 3. Actuators such as valves and other devices provided in the plant 3 are then controlled based on the control signals. The system 100 also includes multiple plants 3. The plants 3 are, for example, chemical plants, production plants that produce products, etc.

[0011] The information processing device 1 acquires status signals (process data) of the plant 3 via the control station 2. The process data includes the temperature, pressure, flow rate, etc. of the processing targets, such as raw materials, intermediates, products, waste liquids, and drainage, as well as set values ​​that define the operating conditions of the equipment included in the plant 3. The information processing device 1 then creates a model for predicting the quality, etc. of the product based on at least a portion of the acquired process data, and assists in determining desirable operating conditions. For example, a regression model is created using process data that serves as a quality evaluation standard as the objective variable and feature quantities that are candidates for control targets during a specified operation as explanatory variables. Then, based on the created regression model, information is output for determining operations to be performed in the operation of the plant 3, desirable operating values, target values, etc.

[0012] <Device Configuration> FIG. 2 is a block diagram showing an example of the configuration of the information processing device 1. The information processing device 1 is a computer and includes a communication interface (I / F) 11, a storage device 12, an input / output device 13, and a processing unit (processor) 14. The communication I / F 11 may be, for example, a network card or a communication module, and communicates with other computers based on a predetermined protocol. The storage device 12 may be a main storage device such as a random access memory (RAM) or a read-only memory (ROM), or an auxiliary storage device (secondary storage device) such as a hard-disk drive (HDD), a solid-state drive (SSD), or a flash memory. The main storage device temporarily stores programs read by the processor 14 and information exchanged with other computers, and secures a working area for the processor 14. The auxiliary storage device stores programs executed by the processor 14 and information exchanged with other computers. The input / output device 13 is a user interface, such as an input device such as a keyboard or a mouse, an output device such as a monitor, or an input / output device such as a touch panel. The processor 14 is an arithmetic processing device such as a CPU (Central Processing Unit), and performs each process according to this embodiment by executing a program. In the example of Fig. 13, functional blocks are shown within the processor 14. That is, the processor 14 functions as a process data acquisition unit 141, a scenario creation support unit 142, and a control support unit 143 by executing a predetermined program.

[0013] The process data acquisition unit 141 acquires process data from sensors provided in the plant 3, for example, via the communication I / F 11 and the control station 2, and stores the acquired process data in the storage device 12. The scenario creation support unit 142 creates a prediction model for predicting a predetermined target variable based on the past operating performance of the plant. The prediction model can support the creation of a scenario representing the operating conditions at the start of operation of the plant 3. The control support unit 143 outputs information to an operator and controls equipment provided in the plant 3 based on the operator's operation. At this time, the scenario creation support unit 142 may predict the target variable using the process data and prediction model acquired from the plant 3. Furthermore, the control support unit 143 automatically controls the operation of the plant 3 according to a predetermined scenario, and also controls the operation of the plant 3 in response to manual operation by a user, particularly during unsteady operation.

[0014] The above-mentioned components are connected via a bus 15 .

[0015] Each of the plants 3 outputs a plurality of pieces of process data. FIG. 3 is a schematic diagram showing an example of a process performed by equipment included in the plant 3. In this embodiment, the process may include a batch process 31 and a continuous process 32. In the batch process 31, the processing target is sequentially processed for each predetermined processing unit, and processing such as receiving, holding, and discharging raw materials into each piece of equipment is performed in sequence. In the continuous process 32, the processing target stored in a buffer tank or the like is continuously processed, and processing such as receiving, holding, and discharging raw materials is performed in parallel. Furthermore, the process may include a plurality of series 33 performing the same processing in parallel.

[0016] The equipment for performing each process includes, for example, a reactor, a distillation apparatus, a heat exchanger, a compressor, a pump, a tank, etc., which are connected via piping. In addition, sensors, valves, etc. are provided at predetermined positions in the equipment and piping. The sensors may include thermometers, flow meters, pressure gauges, level gauges, concentration meters, etc. The sensors monitor the operating status of each equipment and output status signals. The sensors provided in the plant 3 are each assigned a "tag," which is identification information for identifying the sensor. The information processing device 1 and the control station 2 manage input / output signals to each equipment based on the tag.

[0017] FIG. 4 is a diagram illustrating an example of process data in a batch process. The left column of FIG. 4 shows a portion of the process of the batch process 31 shown in FIG. 3. Specifically, the plant 3 includes a shredder 301, a cyclone 302, a pretreatment 303, a precooler 304, and a reactor 305. These processes are classified into a pretreatment process, a precooling process, and a reaction process. The right column of FIG. 4 shows an example of process data acquired in each process. In the pretreatment process, time-series data is acquired from sensors with tags 001 and 002. In the precooling process, time-series data is acquired from sensors with tags 003 and 004. In the reaction process, time-series data is acquired from sensors with tags 005, 006, and 007. In addition, in the batch process, processing objects associated with serial numbers (also referred to as "production numbers") are intermittently processed. In other words, the serial numbers are identification information for identifying processing objects that are processed collectively in the batch process. As shown in FIG. 4, time-series data relating to the processing object associated with subsequent serial numbers is obtained as time passes.

[0018] FIG. 5 is a diagram illustrating an example of process data in a continuous process. The left column of FIG. 5 shows a portion of the process of the continuous process 32 shown in FIG. 3. Specifically, the plant 3 includes a tank 311 and a pump 312. The right column of FIG. 5 shows an example of process data acquired in each process. In the continuous process 32, time-series data associated with tags but not with serial numbers is continuously acquired from sensors. In the example of FIG. 5, time-series data is acquired from sensors with tags 102 and 103. In a continuous process, equipment continuously accepts and processes objects to be processed. When a batch process and a continuous process are performed consecutively, the objects to be processed in the batch process may be linked to the objects to be processed in the continuous process. For example, the serial number in the batch process may be associated with the process data in the continuous process based on the sampling interval of the sensor and the time the objects to be processed reside between the batch process and the continuous process.

[0019] 6 is a processing flow diagram showing an example of a scenario creation support process for non-steady operation executed by the information processing device 1. As an example of non-steady operation, this embodiment will explain a start-up process from the start of operation of the plant 3 to steady operation in which products are stably manufactured. In the start-up process, the plant 3 is controlled so that the concentration of unreacted raw materials in the products manufactured satisfies a predetermined standard, and once the concentration of unreacted raw materials in the products satisfies the predetermined standard, the plant transitions to steady operation.

[0020] FIG. 7 is a diagram illustrating an example of a schematic scenario. A scenario is, for example, a procedure manual created by a user. A program for controlling the plant 3 according to the scenario may be created for at least some of the processes, or the scenario may be used as a reference for an operator during operation. The scenario in FIG. 7 includes processes such as setting a target value according to a measured value (concentration) of the target substance, changing an operating value at a predetermined rate until the target value reaches the predetermined target value, changing the target value or operating value according to the measured value, and waiting for a predetermined period. The framework of the scenario can be created based on the principles of the reaction process. However, particularly in unsteady operation, it is not easy to determine which measured value should be used for control or to find desirable values, such as the underlined feedback control target value and the change amount per unit time of the operating value, in FIG. 7 . The scenario creation support process shown in FIG. 6 presents information useful for scenario creation. Specifically, a user can use the regression model described above to determine desirable values ​​of explanatory variables that bring a target variable, such as the concentration of unreacted raw materials in a product, closer to a value that satisfies a predetermined condition, or to identify explanatory variables that have a significant impact on the fluctuation of the target variable based on the magnitude of the explanatory variable's coefficient.

[0021] FIG. 8 shows an example of operating conditions. Such operating conditions can be created in advance by a user based on the principles of the reaction process, the operator's knowledge, and past operating results. The table in FIG. 8 includes attributes such as "No.", "Feature," "Optimal Value," "Expected Value," and "Condition." The "No." field contains the identification number of the feature. The "Feature" field contains the definition of the feature that can affect the change in the target variable, which is the concentration of unreacted raw materials in the product. The feature is a value that can be used as an explanatory variable in a regression model. Note that the feature may change over time, such as process data for a predetermined period or a value corresponding to the process data. Furthermore, the "Optimal Value" field contains an optimal value based on, for example, principles of chemical reactions. Note that for feature values ​​that include conditions that are trade-offs with other feature values, values ​​used in the simulation may be set within a range of values, ratios, etc., from the perspective of reaction stability, etc. The "Expected Value" field contains the expected value used in the simulation. That is, even if individual feature values ​​are optimal, if the combination of multiple feature values ​​represents a new operating condition, it may be desirable to set an expected value within the range of past operating experience or gradually approach the optimal value. Therefore, a simulation using a regression model is performed using the expected value. The "Condition" field contains conditions that act as constraints when performing control. The conditions may include trade-offs between multiple feature values. A trade-off refers to a relationship in which the direction in which one explanatory variable should be increased or decreased is not uniquely determined by its relationship with other explanatory variables in order to bring the target variable, which is the concentration of unreacted raw materials in the product, closer to the value during steady-state operation of plant 3. The user sets the expected value based on the conditions. For example, if the theoretical optimal values ​​are a minimum feed rate of raw material 1 and a maximum feed rate of raw material 3, a condition may be set based on operating experience to limit the ratio of the feed rate of raw material 1 to the feed rate of raw material 3, as shown in FIG. 8 (No. 1 and No. 3), and the expected value may be set in accordance with the limit.Furthermore, as shown in FIG. 8 (No. 6) for the condition of the liquid level at the start of the reaction in the reactor 1, an upper limit may be set with some leeway based on the operational record, and an estimated value that satisfies this may be set.

[0022] FIG. 9 is a diagram showing an example of feature quantities stored in the storage device 12. For example, records in the table shown in FIG. 9 are prepared for the feature quantities shown in FIG. 8. The table in FIG. 9 includes the attributes "feature quantity," "sign constraint," "relationship with reaction influencing factors" ("temperature," "concentration," "residence time"), and "trade-off." Information corresponding to the feature quantities in FIG. 8 is registered in the "feature quantity" field. Information indicating the sign of the coefficient of each feature quantity (explanatory variable) to be set in the regression equation is registered in the "sign constraint" field. It is understood that, when there is sufficient training data, the coefficient of the explanatory variable in the regression equation can be estimated (converged) with an appropriate sign according to the relationship between the target variable and the explanatory variable. However, when there is little training data, it may not be possible to obtain a coefficient with an appropriate sign. By performing regression analysis with a sign constraint, an appropriate regression equation can be created even when there is little training data. That is, the information processing device 1 performs regression analysis based on the information registered in the "sign constraint" field. Regression analysis with a sign constraint can be performed using a known method disclosed in, for example, International Publication No. WO 2021 / 157669. In other words, when a sign constraint is specified, the regression model performs regression analysis using a regularization term that increases the cost when the sign of the regression coefficient differs from the sign specified in the sign constraint, and minimizes the cost function including the regularization term.

[0023] The "Relationship with Reaction Influence Factors" field stores a sign indicating whether the objective variable will increase or decrease when the feature is increased. "+" indicates an increase, and "-" indicates a decrease. Conversely, when the feature is decreased, "-" indicates an increase in the objective variable, and "+" indicates a decrease. The signs registered in the "Relationship with Reaction Influence Factors" field are set by the user based on the principles of chemical reactions, broken down into temperature, concentration, and residence time. If the same sign is registered for these three aspects, it can be determined that the direction of increase or decrease in the objective variable relative to the increase or decrease in the feature represented by the record (i.e., the sign registered in the "Sign Constraint" field) is determined. The regression equation in regression analysis with a sign constraint is expressed, for example, as follows: Furthermore, w ｋ is the regression coefficient, w ０ is a constant term. ｋ is determined according to a predetermined constraint code. The regression coefficients and constant term can be determined using the cost function expressed by the following equation (2). The coefficients w that minimize the cost function E(w) are ｋ The regression equation is determined by selecting αR is a regularization term (penalty term), and its coefficient α is a parameter that represents the strength of the constraint. If "+" is registered in "Relationship with reaction influence factors", R ＋ (w) value, and if "-" is registered, R － In this way, the regularization term αR according to this embodiment imposes a sign constraint by L1 regularization on either the positive or negative side. That is, the regularization term is a coefficient w ｋ However, the cost is increased according to the sum of the absolute values ​​of the coefficients in either a positive or negative interval depending on the constraint sign.

[0024] The "trade-off" field stores information indicating whether or not there is a problem in which the desirable direction of change cannot be unequivocally determined in relation to other processes in the entire unsteady process (there is a trade-off). For features with a trade-off, the user may modify the sign constraint information and recreate the regression model. The information stored in the table in FIG. 9 can be determined in advance based on, for example, the principles shown in FIG. 8 or the operator's knowledge. It is assumed that the definitions of features shown in FIG. 9 are stored in advance in the storage device 12.

[0025] In the process of FIG. 6 , first, the process data acquisition unit 141 of the information processing device 1 reads process data output by the plant 3 during past operation ( S1 in FIG. 6 ). The read process data is the process data during the unsteady operation described above. In particular, it is preferable to use process data from cases in which the plant 3 operated without any abnormalities or malfunctions. Furthermore, the scenario creation support unit 142 of the information processing device 1 performs machine learning using the read process data ( S2 in FIG. 6 ). The machine learning may be a method using a regularization term, such as linear regression, Ridge regression, or Lasso regression, or a similar method. Furthermore, a regression model may be created based on the sign constraint described above. The objective variable of the regression model is, for example, the concentration of unreacted raw materials in the product manufactured in the plant 3. Furthermore, the explanatory variables of the regression model are at least some of the feature quantities shown in FIGS. 8 and 9 , and are set in advance by the user, for example. The explanatory variables may be the process data itself or some value calculated based on the process data, such as the integral value of the flow rate. The explanatory variables may be, for example, target values ​​or manipulated values ​​in an assumed unsteady operation procedure, or candidates for measured values ​​that serve as processing start conditions (triggers), standby times, etc., as shown in Fig. 7. The explanatory variables may also be integrals of manipulated values ​​or measured values, such as the flow rate and liquid level, as shown in Fig. 8. Based on a regression model with the integral values ​​as explanatory variables, the user can determine, for example, the speed at which the flow rate or liquid level should be changed in an unsteady operation procedure.

[0026] The scenario creation support unit 142 also evaluates the trained model (S3 in FIG. 6). For example, the scenario creation support unit 142 uses part of the process data acquired in S1 as verification data to calculate evaluation indices such as the coefficient of determination. The calculated evaluation indices are output via the input / output device 13.

[0027] The scenario creation support unit 142 also outputs information for creating a scenario representing the operating conditions of each device in the plant 3 during unsteady operation based on the trained model ( S4 in FIG. 6 ). The trained model is, for example, a regression equation in which regression coefficients (partial regression coefficients) and intercepts are determined by regression analysis. A simulation can be performed using the trained model and optimal or assumed values ​​of the explanatory variables to quickly reduce, for example, the concentration of unreacted raw materials in the product, which is the objective variable, to a desired threshold or lower, and output the desired values ​​of the explanatory variables. In S4, the scenario creation support unit 142 may perform predictions using the trained model, particularly for feature quantities (explanatory variables) including trade-off conditions, and present the results to the user. The simulation is performed, for example, based on the "assumed values" in FIG. 8 . The user can set assumed values ​​according to the "conditions" and "optimal values" shown in FIG. 8 and perform the simulation. For example, the assumed values ​​of the explanatory variables can be predicted by changing the assumed values ​​and repeating the simulation to change the objective variable, which is the concentration of unreacted raw materials in the product, to a value during steady operation of the plant 3. Based on the desired assumed values ​​obtained here, the user can determine the operation values, target values, etc. to be set in the scenario. Furthermore, the magnitude of the absolute value of the coefficients of the regression model can be said to represent the magnitude of the influence of changes in the dependent variable. Therefore, by outputting values ​​based on the regression coefficients, such as the regression coefficients or their absolute values, in S4, the user can select explanatory variables to be changed in the driving procedure based on a comparison of the regression coefficients. The user may change the assumed values ​​of the explanatory variables and recreate the regression model. Furthermore, the signs of the regression coefficients represent the correspondence between the control that increases or decreases the explanatory variable and the direction (increase or decrease) in which the dependent variable changes. By outputting the regression coefficients or their signs in S4, the user can understand this correspondence. Furthermore, based on the simulation results, the user may modify the sign constraint information and recreate the regression model, especially for features with a "trade-off" in Figure 9.

[0028] Generally, when multicollinearity exists, the magnitude of the coefficient does not necessarily indicate the degree of influence on the dependent variable. On the other hand, in the regression equation according to this embodiment, the prediction accuracy was equivalent when a combination of explanatory variables with a correlation coefficient of 0.9 or higher was included and when one of the combinations was deleted. In other words, by using a regularization term that increases the cost when the sign of the regression coefficient differs from the sign specified in the "sign constraint" described above, it is possible to converge to a more appropriate coefficient, and it is understood that the problem of multicollinearity has been solved.

[0029] The scenario creation support unit 142 also determines whether to modify the model (S5 in FIG. 6). In this step, the scenario creation support unit 142 receives, via the input / output device 13, for example, a user's operation to determine whether to change the explanatory variables or sign constraints, and determines whether to modify the model based on that input. If it is determined that the model should be modified (YES in S5), the scenario creation support unit 142 receives a selection of explanatory variables to be used in the regression model and modifications to the sign constraints (S6 in FIG. 6), and returns to S1 to repeat the process.

[0030] <Effects> According to this embodiment, the scenario creation support unit 142 can assist the user in creating a regression model based on the magnitude of the influence of feature quantities (explanatory variables) on changes in the target variable. Furthermore, when changes in multiple explanatory variables have a trade-off relationship with each other and affect the overall results of the unsteady process, regression analysis can be performed using process data related to past operational performance to prioritize how to change (i.e., increase or decrease) the explanatory variables to improve the predictive accuracy of the regression model. For example, if the initial solvent charge in the reactor is increased to increase the initial liquid level in the reactor, which is one of the feature quantities, in order to reduce unreacted raw materials in the product, in principle, the raw material concentration in the reactor decreases, making the reaction more difficult, but the reaction time increases, making it easier to fully react. Using the learning model of this embodiment makes it possible to comprehensively determine and present which causal relationship has priority. Furthermore, based on the created regression model, the user can create a preferred operating scenario for unsteady operation. The control support unit 143 controls the plant 3 based on a program for operating in accordance with the scenario. During unsteady operation, the control support unit 143 may accept manual user input via the input / output device 13.

[0031] <Modifications> The configurations and combinations thereof in each embodiment are merely examples, and additions, omissions, substitutions, and other modifications of the configurations are possible as appropriate without departing from the spirit of the present disclosure. The present disclosure is not limited by the embodiments, but only by the scope of the claims. Furthermore, each aspect disclosed in this specification can be combined with any other feature disclosed in this specification.

[0032] Furthermore, at least some of the functions of the information processing device 1 may be distributed among multiple devices, or the same function may be provided in parallel by multiple devices. For example, the device that functions as the scenario creation support unit 142 may be different from the device that functions as the control support unit 143. At least some of the functions of the information processing device 1 may be provided on a so-called cloud.

[0033] In the embodiment, the processing at the time of plant start-up has been described, but the unsteady operation may be processing at the time of plant shutdown or processing to change the production amount per unit time. In this case, too, the processing can be performed according to a predetermined scenario.

[0034] As shown in FIG. 8 , the feature quantities in the embodiment may include the definite integral of the feed flow rate, which is the flow rate at which the raw materials are transferred to a predetermined reactor; the start time of the transfer between two reactors; the liquid level in the reactor at the start of the reaction; the definite integral of the liquid level in the predetermined reactor; and the integral of the absolute temperature of the predetermined reactor. However, the feature quantities are not limited to the examples shown in FIG. 8 . For example, various values ​​based on process data may be used, such as the difference in the start time of the feeding of the first raw material and the second raw material, the definite integral of the flow rate of the cooling water for the reactor, and the integral of the absolute temperature of the cooling water for the reactor. Furthermore, the objective variable is not limited to the concentration of the unreacted raw material shown in FIG. 8 (i.e., the concentration of the unreacted raw material after a predetermined time from the start of the reaction). For example, the objective variable may be the time until the concentration of the unreacted raw material reaches a predetermined standard, the concentration of other impurities in the product, product physical properties, raw material usage, energy usage, out-of-system emissions, and other values.

[0035] The present disclosure also includes a method and a computer program for executing the above-described process, and a computer-readable recording medium having the program recorded thereon. The recording medium having the program recorded thereon enables the above-described process by causing a computer to execute the program.

[0036] Here, a computer-readable recording medium refers to a recording medium that stores information such as data and programs electrically, magnetically, optically, mechanically, or chemically and can be read by a computer. Among such recording media, those that can be removed from a computer include flexible disks, magneto-optical disks, optical disks, magnetic tapes, memory cards, etc. Furthermore, recording media that are fixed to a computer include HDDs, SSDs (Solid State Drives), ROMs, etc.

[0037] 100: System 1: Information processing device 2: Control station 3: Plant 11: Communication I / F 12: Storage device 13: Input / output device 14: Processing unit (processor) 141: Process data acquisition unit 142: Scenario creation support unit 143: Control support unit

Claims

1. A procedure creation support device having a processing unit that performs the following: acquiring process data output during past non-steady operation of a plant; creating a trained model that uses feature quantities based on the process data as explanatory variables and learns the relationship with a predetermined target variable through regression analysis; performing a simulation using the trained model and expected values ​​of the explanatory variables; outputting the regression coefficients of the trained model, the results of the simulation, and conditions that are predetermined constraints on the explanatory variables; and modifying the trained model when an operation to modify the explanatory variables is received from a user, or performing the simulation using the modified expected values ​​when an operation to modify the expected values ​​is received from the user.

2. The procedure creation support device of claim 1, wherein the processing unit outputs the regression coefficient of the trained model or a value based on the regression coefficient as an index of the magnitude of influence of the explanatory variable multiplied by the regression coefficient in the trained model on a change in the objective variable.

3. The procedure creation support device of claim 2, wherein, when the user specifies the sign of the regression coefficient, the trained model performs regression analysis using a regularization term that increases the cost when the sign of the regression coefficient differs from that specified by the user, so as to minimize a cost function including the regularization term, and the constraint condition includes causal relationship information regarding the direction in which one of the explanatory variables should be increased or decreased in order to change the target variable to a value in steady-state operation of the plant.

4. The procedure creation support device according to claim 3, wherein the causal relationship information includes a plurality of pieces of information that indicate directions in which the explanatory variables should be increased or decreased based on different viewpoints.

5. The procedure creation support device according to claim 1, wherein the constraint conditions include being within a range of explanatory variables based on process data output during the past non-steady operation.

6. A procedure creation support method executed by one or more computers, comprising: acquiring process data output during past non-steady operation of a plant; creating a trained model in which feature quantities based on the process data are used as explanatory variables and the relationship with a predetermined target variable is trained by regression analysis; performing a simulation using the trained model and expected values ​​of the explanatory variables; outputting the regression coefficients of the trained model, the results of the simulation, and conditions that are predetermined constraints on the explanatory variables; and modifying the trained model when an operation to modify the explanatory variables is received from a user, or performing the simulation using the modified expected values ​​when an operation to modify the expected values ​​is received from the user.

7. A procedure creation support program for causing a computer to execute the following steps: acquiring process data output during past non-steady operation of a plant; creating a trained model in which feature quantities based on the process data are used as explanatory variables and the relationship with a predetermined target variable is trained by regression analysis; performing a simulation using the trained model and expected values ​​of the explanatory variables; outputting the regression coefficients of the trained model, the results of the simulation, and conditions that are predetermined constraints on the explanatory variables; and modifying the trained model when an operation to modify the explanatory variables is received from a user, or performing the simulation using the modified expected values ​​when an operation to modify the expected values ​​is received from the user.

Images