WO2023190457A1 - Support device, statistical model generation device, support method and support program - Google Patents

Abstract

A support device (10) for supporting the operation of a plant comprises: a data acquisition unit for acquiring simulation data pertaining to the operation state of a plant and calculated by entering data that includes a predetermined condition list into a physical model simulator, the predetermined condition list being configured so that factors that determine the operation state of the plant and an index related to the factors are associated; a statistical model generation unit (14c) for generating a first statistical model on the basis of simulation data; and an operation state determination unit (14f) for referencing the first statistical model to determine the operation state of the plant on the basis of process data for the plant.

Description

Support device, statistical model generation device, support method and support program

The present invention relates to a support device, a statistical model generation device, a support method, and a support program for supporting plant operation.

The central monitoring room in the plant constantly monitors process data and alarms issued by the DCS (distributed control system). The operating state of the plant is determined by evaluating process data acquired by sensors installed in the plant based on a statistical model, and if an abnormality is determined as a result, the operator is notified. The statistical model used for determination can be generated by learning process data obtained by actually operating a plant.

JP2011-253275

However, if a sufficient amount of process data has not been accumulated, such as immediately after the plant starts operating, learning cannot be performed sufficiently, and the accuracy of determining the operating state by the statistical model decreases.

One exemplary object of an embodiment of the present invention is to provide a support device, a statistical model generation device, a support method, and a support program that determine the operating state of a plant with high accuracy.

In order to solve the above problems, a support device according to an aspect of the present invention is a support device for supporting plant operation, and is calculated by inputting data including a predetermined condition list into a physical model simulator. A data acquisition unit that acquires simulation data regarding the operating state of a plant, the predetermined condition list includes a data acquisition unit that is associated with factors that determine the operating state of the plant and indicators for the factors, and simulation data. The present invention includes a statistical model generating section that generates a first statistical model based on the first statistical model, and an operating state determining section that refers to the first statistical model and determines the operating state of the plant based on process data of the plant.

According to the above aspect, since the data used for learning the statistical model is obtained through physical model simulation, the statistical model can determine the operating state with high accuracy even if sufficient process data actually obtained from the plant has not been accumulated. can be generated.

Yet another aspect of the present invention is a statistical model generation device. This device is a device that generates a statistical model for determining the operating state of a plant, and acquires simulation data regarding the operating state of the plant, which is calculated by inputting data including a predetermined list of conditions into a physical model simulator. The predetermined condition list includes a data acquisition unit in which factors that determine the operating state of the plant and indicators for the factors are associated with each other, and a first statistical model based on the simulation data. and a statistical model generation unit that generates the .

Yet another aspect of the present invention is a support method. This method is a support method for supporting the operation of a plant, and is a data acquisition method for obtaining simulation data regarding the operating state of the plant calculated by inputting data including a predetermined list of conditions into a physical model simulator. a data acquisition step in which the predetermined condition list is associated with factors that determine the operating state of the plant and indicators for the factors; and generating a first statistical model based on the simulation data. The method includes a step of generating a statistical model, and a step of determining an operating state of the plant based on process data of the plant with reference to the first statistical model.

Another aspect of the present invention is a support program. This program is a support program for supporting the operation of a plant, and uses a computer to acquire simulation data regarding the operating state of the plant, which is calculated by inputting data including a predetermined list of conditions into a physical model simulator. The predetermined condition list includes a first statistical model based on the simulation data; and operating state determining means that refers to the first statistical model and determines the operating state of the plant based on the process data of the plant.

Note that arbitrary combinations of the above constituent elements and mutual substitution of constituent elements and expressions of the present invention among methods, apparatuses, systems, computer programs, data structures, recording media, etc. are also aspects of the present invention. It is valid as

According to the present invention, the operating state of a plant can be determined with high accuracy.

1 is a diagram showing the configuration of a support device 10 according to an embodiment of the present invention. 1 is a diagram showing an example of a hardware configuration of a support device 10. FIG. FIG. 3 is a diagram showing an example of a condition list. FIG. 3 is a diagram showing an example of a condition list. 3 is a flowchart illustrating an example of a support method by the support device 10. FIG. 3 is a flowchart illustrating an example of a support method by the support device 10. FIG. FIG. 3 is a diagram illustrating an example of a statistical model generation process. FIG. 3 is a diagram illustrating an example of a statistical model generation process. 3 is a diagram for explaining a support method by the support device 10. FIG. 3 is a diagram for explaining a support method by the support device 10. FIG. 3 is a diagram for explaining a support method by the support device 10. FIG.

Hereinafter, the present invention will be explained through embodiments of the invention with reference to the drawings, but the following embodiments do not limit the claimed invention, and the embodiments described in the embodiments do not limit the claimed invention. Not all combinations of features are essential to the solution of the invention. Identical or equivalent components, members, and processes shown in each drawing are designated by the same reference numerals, and redundant explanations will be omitted as appropriate.

FIGS. 1 to 4 are diagrams for explaining a support device 10 according to an embodiment of the present invention. Specifically, FIG. 1 is a diagram showing the configuration of a support device 10 according to an embodiment of the present invention, and FIG. 2 is a diagram showing an example of the hardware configuration of the support device 10. 3 and 4 are diagrams showing examples of condition lists stored in the condition list storage section 18c.

The support device 10 is a device that supports the operation of the plant 1. Examples of the plant 1 include a power generation plant, an incineration plant, a chemical plant, and the like. In plant 1, arbitrary process data is used. The process data is, for example, data on process values such as temperature, pressure, air volume, concentration, or components.

The support device 10 is connected to the plant 1 via a DCS (distributed control system) 2, and acquires process data from sensors installed in the plant 1. According to the support device 10, even if sufficient process data is not accumulated, the operating state can be determined with high precision using a statistical model.

The support device 10 has a predetermined program necessary for executing the support method according to the present embodiment installed in advance, and an example of its hardware configuration is shown in FIG. 2. Specifically, the support device 10 can be a general-purpose or dedicated computer that includes a CPU 100, a ROM 102, a RAM 104, an external storage device 106, a user interface 108, a display 110, and a communication interface 112. The CPU 100 performs calculations based on information input by the plant operator through the user interface 108 and outputs the calculation results to the display 110, and while the operator recognizes the output, the CPU 100 outputs the calculation results to the support device 10 through the user interface 108. You can enter the necessary information.

The support device 10 may be composed of a single computer or may be composed of multiple computers distributed on a network. In the support device 10, for example, the CPU executes a predetermined program (a program specifying the support method according to the present embodiment) stored in the above-mentioned ROM, RAM, external storage device, etc. or downloaded via a communication network. By doing so, the support device 10 can function as various functional blocks or various steps described below.

Hereinafter, various functional blocks of the support device 10 will be explained using FIG. 1.

The support device 10 according to the present embodiment includes an input section 12 having an input receiving section 12a and an operation receiving section 12b, a processing section 14, a display section 16 having a display 16a, and a storage section 18. Here, the processing section 14 includes a control section 14a, a process data acquisition section 14b, a statistical model generation section 14c, a random number generation section 14d, a physical model simulator section 14e, an operating state determination section 14f, and a display control section 14g. The storage unit 18 also includes a process data storage unit 18a, a simulation data storage unit 18b, a condition list storage unit 18c, a distinction information storage unit 18d, a statistical model storage unit 18e, a physical model storage unit 18f, and a judgment It has a result storage section 18g.

The control unit 14a is composed of a microcomputer including a CPU and a semiconductor memory, and performs general arithmetic processing that is not executed by other functional blocks included in the processing unit 14. For example, time synchronization with an external device via the communication interface 110, name resolution using DNS (Domain Name System), etc. are performed.

The process data acquisition unit 14b acquires process data of the plant 1 from the DCS 2. Process data is data representing the operating state of a plant, and can also be referred to as operating data. The process data is, for example, data indicating changes over time in measured values of a sensor. In this case, the process data may be changes in continuous sensor measurements at predetermined time intervals. Process data may be multidimensional data. The process data acquired by the process data acquisition section 14b is stored in the process data storage section 18a.

The statistical model generation unit 14c generates a statistical model by learning predetermined data. The predetermined data is, for example, process data or simulation data output from a physical simulation described later. The process data used for learning may be different from the process data for evaluation used for determining the operating state. The generated statistical model is stored in the statistical model storage section 18e.

Note that generating a statistical model includes generating a machine learning model that outputs a predetermined determination when certain data is input. This machine learning model may be generated by any method. As the algorithm of the machine learning model, for example, support vector machine, logistic regression, neural network, deep neural network, k-means method, etc. can be used, but the type thereof is not particularly limited.

The random number generation unit 14d generates random numbers according to a predetermined probability distribution. The predetermined probability distribution is, for example, a normal distribution with a mean M and a standard deviation σ. Note that M and σ are arbitrary real numbers. In addition, the probability distribution may be a uniform distribution, an exponential distribution, a gamma distribution, or the like.

The physical model simulator section 14e executes a physical model simulation based on the physical model stored in the physical model storage section 18f and the condition list stored in the condition list storage section 18c, and obtains simulation data. The simulation data is stored in the simulation data storage section 18b.

Simulation data is pseudo process data calculated by physical model simulation. The simulation data may have a similar data structure and dimensions to the process data obtained from the plant. Specifically, simulation data can be subjected to similar calculation processing in the statistical model generation unit 14c without being distinguished from process data. More specifically, the statistical model generation unit 14c can generate a statistical model corresponding to either simulation data or process data by learning either of them.

Note that the physical model simulation by the physical model simulator section 14e can be performed regardless of the operating status of the plant. For example, it can be executed before the plant starts operating, while the plant is operating, or even during a period when the plant is temporarily stopped.

The condition list stored in the condition list storage unit 18c includes information regarding the execution conditions of the physical model simulation. An example of the condition list will be described later using FIGS. 3 and 4.

The operating state determination unit 14f refers to the statistical model and determines whether the process data to be determined is within the range of the threshold value. The threshold value is, for example, a value determined when the statistical model is generated, and is information stored in the statistical model storage unit 18e. The determination result is stored in the determination result storage section 18g.

The display control unit 14g displays information including the determination result stored in the determination result storage unit 18g and the distinction information stored in the distinction information storage unit 18d on the display 16a. An example of the display contents will be described later using FIG. 10.

The input accepting unit 12a and the operation accepting unit 12b accept input and operations from the user via the user interface 108. The display control unit 14g may change the content displayed on the display 16a based on the received input and operation. An example of changing the display content will be described later using FIG. 10. Alternatively, the condition list may be created or changed based on received input or operation.

Next, an example of the condition list will be explained using FIGS. 3 and 4.

A condition list is a set of execution conditions for physical model simulation. Specifically, the condition list includes information that includes one or more factors and one or more indicators associated with each factor. Here, the factor is, for example, an element that affects the operating state of the plant. As an example, the condition list shown in FIG. 50% + PKS50%, etc.), two indicators 214 (summer average temperature, year-round average temperature) are associated with the factor 212, and five indicators 216 (100%, 90%, etc.) are associated with the factor 222. There is. The condition list is input to the physical model simulator to specify conditions for executing the simulation.

Alternatively, the condition list may be information including multiple records as shown in FIG. Here, a record includes one in which one index is associated with each of one or more factors. For example, the record 322 includes an index 322b (100% pellets), an index 322c (average summer temperature), and an index 322d (100%) for a factor 310b (fuel), a factor 310c (outside temperature), and a factor 310d (boiler load). are associated with each other, and the condition list is information consisting of eight records 320 including record 322.

The physical model simulation is performed based on the physical model simulation execution conditions included in the condition list. A specific example of the execution method will be described later using FIG. 6.

Some of the information included in the condition list may not substantially affect the results of the physical model simulation. For example, in the condition list of FIG. 4, the condition number 310a column only represents a serial number for each record, and does not affect the results of the physical model simulation.

The condition list may include accepting input of sensor data from the user via the user interface 108, receiving from another device via the communication interface 112, or reading via the RAM 104 or external storage device 106. The condition list may be stored in the condition list storage section 18c.

Furthermore, at least some of the indicators in the condition list may be set based on random numbers generated by the random number generation unit 14d and according to a predetermined probability distribution. For example, each index associated with the factor 310d (boiler load) may be set based on a value that does not exceed 100% of random numbers that follow a normal distribution with an average of 70% and a standard deviation of 5%.

The condition list is not limited to the list format shown in FIG. 3 or the table format shown in FIG. 4. For example, it may be represented by a plurality of tables that can create a table equivalent to the table in FIG. 4 by appropriately combining the tables. Other formats include tree structure, JSON (JavaScript Object Notation) format, YAML format (YAML Ain't Markup Language), and XML (Extensible Markup Language) format. etc., may be represented by a data structure other than a table.

Hereinafter, as an embodiment of the support device 10 of the present invention, an example of the operation of the support device 10 will be described using FIGS. 5 to 8. The flowchart in FIG. 5 is an example of the entire operation of the support device 10 according to this embodiment. The flowchart in FIG. 6 is an example of a method for generating a statistical model by the statistical model generating unit 14c using simulation data. 7 and 8 are examples of the process of generating a statistical model by learning process data and simulation data, respectively.

FIG. 5 is a flowchart showing an example of the operation of the support device 10.

In FIG. 5, first, the process data acquisition unit 14b acquires the process data of the plant 1, and stores it in the process data storage unit 18a (S10). Process data may be acquired directly from the DCS 2, or may be performed by accepting input of sensor data from a user via the user interface 108, or by receiving sensor data from another device via the communication interface 112. Alternatively, the data may be read out via the RAM 104 or the external storage device 106. Further, the process data may be acquired sequentially (for example, every second), or process data for a certain period of time (for example, one day) may be acquired all at once.

Next, it is determined whether a first statistical model, which is a statistical model generated based on simulation data, has been generated and is already usable (S12). If the first statistical model has not been generated (S12 NO), the statistical model generating unit 14c generates a first statistical model, and information regarding this is stored in the statistical model storage unit 18e (S14). An example of the method for generating the first statistical model will be described later using the flowchart of FIG.

Note that since the first statistical model is generated based on simulation data, it can be generated regardless of the operating state of the plant and used to determine the operating state. The first statistical model can be generated within the support device 10, for example, before the plant starts operating, while the plant is operating, or even during a period when the plant is temporarily stopped.

If the first statistical model has already been generated (S12: YES) or if the first statistical model has been generated in step S14, the operating state of the plant is determined in the operating state determination unit 14f (S16). The operating state is determined based on the process data stored in the process data storage section 18a, with reference to the first statistical model stored in the statistical model storage section 18e. The determination result is stored in the determination result storage section 18g.

The case where the first statistical model has already been generated means that, in addition to the case where the first statistical model has already been generated by the statistical model generation unit 14c of the support device 10, for example, the first statistical model is generated via the communication interface 112 before the start of operation of the plant. This also includes when information is received or when it is read out via the RAM 104 or external storage device 106.

Next, it is determined whether the second statistical model, which is a statistical model generated based on the process data, is ready for use (S18). Information regarding the second statistical model is stored in the statistical model storage unit 18e separately from information regarding the first statistical model.

The accuracy of the second statistical model is improved by sufficiently learning the process data, so for example, the process data in the relevant operating state, such as immediately after the start of plant operation or immediately after a significant change in the operating conditions of the plant, is used. In a situation where a sufficient amount of fuel is not accumulated, the operating state cannot be determined with high accuracy.

In the above situation, if it is not practical to use the second statistical model (NO in S18), the statistical model generation unit 14c learns the process data stored in the process data storage unit 18a to generate the second statistical model. 2. The statistical model is updated (S20). Through the process of step S20, information regarding the second statistical model stored in the statistical model storage unit 18e is updated.

Note that even if the second statistical model is updated (S20), it is not necessarily the case that a sufficient amount of process data has been learned, so the second statistical model is not used to determine the operating status of the plant. It does not necessarily mean that it is ready for use. Therefore, in the present embodiment, when the second statistical model is updated (S20), the driving state is not determined with reference to the second statistical model (S22), and the driving state determination result is displayed (S24). ).

On the other hand, if the second statistical model has learned a sufficient amount of process data and is already in a usable state (S18 YES), the operating state determination unit 14f determines the operating state of the plant (S22). . The operating state is determined based on the process data stored in the process data storage section 18a with reference to the second statistical model. The determination result is stored in the determination result storage section 18g, distinguishing it from the determination result stored in step S16.

The types of determination results are, for example, "normal" and "abnormal", and "abnormal" may have stages such as "caution", "warning", and "monitoring required". Abnormality is not limited to a state in which the operation of the plant needs to be stopped, but also includes a state in which the plant can continue to operate but is not in a good operating state.

Finally, the display control unit 14g displays information regarding the driving state on the display 16a based on the determination result stored in step S16 or step S22 (S24). The displayed information can be changed according to the input from the user to the input reception section 12a or the operation reception section 12b. An example of the display screen will be explained separately using FIG. 10. After step S24, the support device 10 ends its operation.

In an example of the present embodiment, the operating state of the plant can be determined with high accuracy even during a period when process data is not sufficiently accumulated. Specifically, the second statistical model is not practical for determining operating conditions until the process data is sufficiently learned, but if the first statistical model is generated based on simulation data, highly accurate operating conditions can be determined. can be determined. More specifically, even if the second statistical model is not ready for use and further learning of the process data is required (S20), the first statistical model determines the operating state (S16). The determination result can be presented to the operator (S24).

As a result of this effect, for example, in the past, it was necessary to perform a test run of a plant in order to accumulate process data to be learned by a statistical model over a long period of time. For example, since the operating state can be determined with reference to a statistical model generated from simulation data immediately after the start of operation, there is no need for a trial run.

Furthermore, as another effect, the validity of process data learning can be determined by comparing the first statistical model and the second statistical model. Specifically, a statistical model generated from process data may learn that process data when an abnormality has originally occurred is assumed to be normal, for example due to a temporary malfunction of equipment within the plant. In some cases, when an operator compares the statistical model generated from simulation data with the data, it is possible to discover the data and exclude it from the learning target.

The operation of the embodiment of the present invention described above is only an example, and is not limited thereto. For example, this operation may be repeatedly executed while the support device 10 is in operation. That is, the operation may be restarted after the operation is completed. Further, the order of each step may be changed as long as there is no contradiction in operation, and some of the steps may be repeatedly executed (for example, after step S24, the process may return to step S10 without ending).

Additionally, some steps may be performed in parallel. For example, after acquiring process data in step S10, processing related to the first statistical model (corresponding to S12 to S16) and processing related to the second statistical model (corresponding to S18 to S22) are executed in parallel. It's okay. In this case, the user may be prompted to select which model, the first statistical model or the second statistical model, to perform processing on.

In addition, the determination of the driving state may not be performed temporarily in either the first statistical model or the second statistical model. For example, if the second statistical model has learned a sufficient amount of process data and is already usable, the processing related to the first statistical model (corresponding to S12 to S16) will be temporarily stopped. You may.

In addition, the process data used for determining the operating state (corresponding to S16 and S22) and learning the second statistical model (corresponding to S20) is process data obtained in the immediately preceding process corresponding to step S10. The process data is not limited to this, and may be process data stored in the process data storage unit 18a before the process. For example, while process data acquisition (corresponding to S10) is performed every second, the process data used to update the second statistical model (corresponding to S20) was acquired one day before the update process. There may be.

In addition, process data may be handled separately into evaluation process data and learning process data according to an arbitrary rule. For example, the process data storage unit 18a distinguishes process data acquired within the past week as evaluation process data and previous process data as learning process data, and uses only the learning process data to perform second statistics. The model may be learned (corresponding to S20) and the operating state may be determined using only evaluation process data (corresponding to S16 and S22).

Alternatively, one statistical model may be created by learning both process data and simulation data. For example, the first statistical model may be generated by performing a physical model simulation before the plant starts operating, and the first statistical model may be made to learn the process data after the plant starts operating.

Next, an example of the first statistical model generation method (S14) involving physical model simulation will be described using an example of the flowchart of FIG. 6 and the condition list of FIG. 4.

First, in the control unit 14a, the initial value of the variable N is set to "1" (S14a). The variable N is a variable that is incremented according to repeated processing.

Next, a record in which the condition number 310a has the same value as the variable N is read from the condition list stored in the condition list storage unit 18c (S14b). For example, when the variable N is 1, the record 322 whose condition number is "1" is read.

Next, the read record 322 is input to the physical model simulator section 14e, and a physical model simulation is executed based on the factors and indicators of the record (S14c). In this case, the fuel (factor 310b) is "100% pellets (index 322b)", the outside temperature (factor 310c) is "average summer temperature (index 322c)", and the boiler load (factor 310d) is "100% (index 322d)". ” to perform physical model simulations based on the factors and indicators. Simulation data, which is pseudo process data calculated as an execution result, is stored in the simulation data storage section 18b.

Note that the physical model simulation is, for example, inputting predetermined parameters and reproducing the operating state of the plant under the parameters in a computer. A physical model simulator that performs physical model simulation includes a program used when designing a plant.

Thereafter, the simulation data stored in the simulation data storage section 18b is input to the statistical model generation section 14c, and the first statistical model is made to learn the data. Accordingly, the information regarding the first statistical model stored in the statistical model storage unit 18e is updated (S14d).

The above processes from step S14b to step S14d are executed based on each of the plurality of records. Specifically, it is determined whether the value stored in the variable N is the last number of the condition number 310a (S14e), and if not (S14e NO), 1 is added to the variable N (S14f). Return to step S14b. For example, if the variable N is "1", this is not the last number "8" of the condition number 310a, so the variable N is incremented to "2" and the process returns to step 14b.

In step 14b, in order to read the record whose condition number 310a has the same value as the variable N from the condition list stored in the condition list storage section 18c, the record 324 whose condition number is "2" is read.

By repeating the above process in the same way, the generation of the first statistical model is completed after the processes of steps S14b to S14d are executed based on all records.

The method for generating the first statistical model described above is only an example, and is not limited to this. For example, physical model simulations based on multiple records may be performed in parallel.

Next, an example of the process of learning the first statistical model and the second statistical model will be described using FIGS. 7 and 8.

FIG. 7 is an example of the process of setting the ideal line and threshold of the second statistical model based on the process data.

Graph 416 in FIG. 7 is a scatter diagram in which a plurality of process data are plotted, with process variable A on the horizontal axis and process variable B on the vertical axis. Both the process variable A and the process variable B are variables included in the process data, and are, for example, the amount of heat absorbed, the boiler load, etc. A plurality of dots represented by white circles are plotted, each corresponding to one piece of process data. As shown in the graph 416, if a sufficient amount of process data is not accumulated, a practical ideal line and threshold values cannot be determined and cannot be used to determine the operating state.

On the other hand, the graph 418 is a scatter diagram when a certain period of time has passed and a certain amount of process data has been accumulated. By learning the accumulated data, the ideal line 418a in the second statistical model can be set. The ideal line is set, for example, by a mathematical model generation method such as the nonlinear least squares method.

Graph 420 is a scatter diagram when a certain period of time has passed and sufficient process data has been accumulated. In addition to the ideal line, a threshold value 420a and a threshold value 420b can also be set, and if the process data exceeds these values, it can be determined that there is an abnormality in the operating state. Note that the threshold value may be set, for example, so that 95% of the process data used for learning is determined to be "no abnormality". Alternatively, for example, it may be assumed that the process data follows a predetermined distribution, and the threshold value may be set based on the distribution.

Since the second statistical model sets the ideal line and threshold values by learning process data, it is not possible to determine the operating state of the plant immediately after the start of operation (for example, the state of graph 416), and to some extent after the start of operation. The operating state of the plant can be determined with high accuracy only after a period of (for example, the state of graph 420) has elapsed.

FIG. 8 is an example of the process of setting the ideal line and threshold of the first statistical model based on simulation data.

Graph 516 in FIG. 8 is a scatter diagram in which a plurality of simulation data are plotted, with process variable A on the horizontal axis and process variable B on the vertical axis. A plurality of dots represented by squares are plotted, each corresponding to one piece of simulation data. Here, since one piece of simulation data is calculated based on one record in the condition list, one dot can also be said to correspond to one record in the condition list. For example, it can be said that the dot 516a corresponds to the record 322 and the dot 516b corresponds to the record 324.

Furthermore, either the process variable A or the process variable B may be a variable set as an input to a physical model simulator. For example, if the input of a physical model simulator includes a value related to "boiler load" and the output simulation data includes a value related to "endothermic amount", process variable A may be the boiler load, and process variable B may be the endothermic amount. .

Graph 518 is a scatter diagram in which more physical model simulations are performed than in graph 516, and the calculated simulation data is plotted. By learning the simulation data, the ideal line 518a in the first statistical model can be set.

The graph 520 is a scatter diagram in which more physical model simulations are performed than in the graph 518, and the calculated simulation data is plotted. By learning the simulation data calculated based on a sufficient combination of execution conditions, it is possible to set not only the ideal line but also the threshold 520a and the threshold 520b, and if the process data exceeds this, the operating state will be abnormal. It can be determined that there is. Note that the threshold value may be set, for example, so that 95% of the learning data is determined to be "no abnormality". Alternatively, for example, it may be assumed that the process data follows a predetermined distribution, and the threshold value may be set based on the distribution.

When calculating simulation data, some of the execution conditions may be common in some physical model simulations. For example, the dot group 520c consisting of three dots is the simulation data of a simulation performed with the process variable A in common (for example, "boiler load" being 70% in common) and other execution conditions changed. It is.

In general, it is difficult to plan physical model simulations by strictly predicting the actual situation of the plant. Therefore, the simulation data of the physical model simulation often differs from the actual process data, and it is not always possible to accurately determine the operating state of the plant depending on the execution results of the simulation.

Therefore, as shown in this embodiment, by executing a physical model simulation based on a condition list, which is a set of execution conditions corresponding to various operating conditions, and accumulating simulation data, it is possible to It is possible to generate statistical models that take into account variations in various factors.

The generation process of the first statistical model and the second statistical model and the process data determination method described using FIGS. 7 and 8 are merely examples, and do not limit the scope of application of the present invention. That is, the present invention can be implemented using any statistical model as long as it is a statistical model that can learn data and make decisions on the data. For example, the statistical model may be one that determines the operating state using three or more process variables, or may be one that determines the operating state based on a model generated by deep learning.

Additionally, a plurality of threshold values for each statistical model may be set depending on the type of determination result. For example, a threshold value that makes the determination result "caution" and a threshold value that makes the determination result "monitoring required" may be set, respectively.

Next, an example of the display screen of the support device 10 will be explained using FIGS. 9 to 11. FIG. 9 is an example of a screen that displays the determination result of the driving state, and FIGS. 10 and 11 are examples of screens that simultaneously display a plurality of statistical models. Each screen is displayed on the display 16a by the display control section 14g. The display contents are displayed based on information stored in the storage unit 18, and may be changed based on inputs and operations received by the input reception unit 12a and the operation reception unit 12b.

First, the driving diagnosis screen 600 will be explained using FIG. 9. The driving diagnosis screen 600 includes a time series graph display area 610, a driving state determination result display area 620, a model comparison button 630, a legend display area 640, a guide display area 650, and an alarm list display area 660. .

The time series graph display area 610 displays a time series graph in which the horizontal axis is the measurement time 614 and the vertical axis is the measured value 612 of the process variable B, together with a scroll bar 616. The process variable B may be a process variable that is particularly important in determining the operating state of the plant.

The legend display area 640 displays a legend for each piece of information displayed in the driving state determination result display area 620.

The driving state determination result display area 620 displays the ideal line and threshold of the statistical model (first statistical model) generated based on the simulation data, along with the learning data and evaluation data. Distinction information 624 displays the type of statistical model currently being referred to. In this example, the referenced statistical model is generated based on simulation data, so the learning data is simulation data and the evaluation data is process data.

The display content of the driving state determination result display area 620 may be changed based on the user's operation. For example, by receiving a mouse operation from the user via the user interface 108 and pressing the learning data 642 part of the legend display area 640, display or non-display of the learning data in the driving state determination result display area 620 can be switched. Good too. Alternatively, the display may be switched to a statistical model (second statistical model) generated from process data by a method described later using FIGS. 10 and 11.

The guide display area 650 displays, in accordance with the display content of the time series graph display area 610, how to view the graph 654 and how to respond when the driving state is determined to be abnormal 656. Although omitted in FIG. 9, sentences to be presented to the user are displayed in the graph view 654 and response method 656, respectively.

The alarm list display area 660 displays the date and time when the most recent alarm (notification indicating that there is an abnormality in the driving state) was issued and the date and time when the abnormality that caused the alarm was resolved. Among these, for example, those for which the abnormality has not been resolved after the alarm has been issued may be highlighted and displayed.

Next, an example of a method of comparing the first statistical model and the second statistical model according to the user's operation and switching the statistical model displayed in the driving state determination result display area 620 will be explained using FIGS. 9 to 11. do.

The first example is a method of comparing the first statistical model and the second statistical model side by side. Specifically, by pressing the model comparison button 630 on the driving diagnosis screen 600 of FIG. 9, the screen changes to the first statistical model comparison screen 700 of FIG.

FIG. 10 shows a display setting area 710, a statistical model comparison display area 740 including a driving condition determination result graph 742 and a driving condition determination result graph 746, a legend display area 720 corresponding to the driving condition determination result area 742, and a driving condition determination result graph 742 and a driving condition determination result graph 746. A legend display area 730 corresponding to the status determination result area 746 is included.

The operating state determination result graph 742 is a scatter diagram when the operating state is determined with reference to a statistical model (second statistical model) generated based on process data. On the other hand, the driving state determination result graph 746 is a scatter diagram when the driving state is determined with reference to a statistical model (first statistical model) generated based on simulation data.

The second example is a method in which the first statistical model and the second statistical model are overlapped and compared. Specifically, by pressing the model comparison button 630 on the driving diagnosis screen 600 in FIG. 9, the screen changes to the second statistical model comparison screen 800 in FIG.

FIG. 11 includes a display setting area 810, a statistical model comparison display area 830, and a legend display area 820.

The statistical model comparison display area 830 displays a scatter diagram when the operating state is determined with reference to the statistical model (second statistical model) generated based on the process data, and a statistical model (second statistical model) generated based on the simulation data. A scatter diagram in which the driving state is determined with reference to the first statistical model) is displayed in an overlapping manner.

The statistical model displayed in the driving state determination result display area 620 can be switched by pressing the actual model setting button 714d and the actual model setting button 716d. For example, in the initial state, a statistical model generated based on simulation data is displayed in the driving state determination result display area 620 as shown in FIG. 9, but by pressing the actual model setting button 714d, a graph is displayed. can be switched to a statistical model generated based on process data.

By comparing and displaying statistical models, as in the statistical model comparison display area 740 in FIG. 10 and the statistical model comparison display area 830 in FIG. 11, the user can determine when to switch from the first statistical model to the second statistical model. can do. Specifically, the comparison display is used as follows.

First, for a period when process data is not sufficiently accumulated, as described above, the statistical model (first statistical model) generated from simulation data can determine the operating state with higher accuracy. On the other hand, if sufficient process data has been accumulated, it is more accurate to use a statistical model (second statistical model) generated from process data that includes effects that cannot be expressed by physical model simulation, such as aging of the plant. The operating status can be determined. Therefore, it is necessary to switch the statistical model used from the first statistical model to the second statistical model at an appropriate timing, and this timing can be easily determined by comparing and displaying the statistical models.

The present invention is not limited to the above-described embodiments, and can be modified and applied in various ways. The operations of the support device 10 are not limited to those in which all operations are automated by computer processing, but also include operations in which at least some of the operations are manually performed by an operator. Furthermore, the display screens described in FIGS. 9 to 11 in the above embodiment are merely examples, and the present invention is not limited thereto.

The embodiments described through the embodiments of the invention above can be used in combination or with changes or improvements as appropriate depending on the application, and the present invention is not limited to the description of the embodiments described above. do not have.

The present invention also includes the following embodiments.

[Additional note 1]
A support device 10 according to one aspect of the present invention is a support device 10 for supporting the operation of a plant 1, and is a support device 10 for supporting the operation of a plant 1. A data acquisition unit that acquires simulation data regarding the operating state, the predetermined condition list is based on the data acquisition unit and the simulation data, in which factors that determine the operating state of the plant 1 and indicators for the factors are associated. It includes a statistical model generating section 14c that generates a first statistical model, and an operating state determining section 14f that refers to the first statistical model and determines the operating state of the plant 1 based on the process data of the plant 1.

[Additional note 2]
In the support device 10 of Supplementary Note 1, the predetermined condition list may include one or more factors, and one or more indicators may be associated with each factor.

[Additional note 3]
In the support device 10 of Appendix 2, the predetermined condition list may include a plurality of records, and one index may be associated with each of one or more factors.

[Additional note 4]
The support device 10 according to any one of Supplementary Notes 1 to 3 further includes an input receiving unit 12a that receives input of indicators from the user, and the predetermined condition list is such that at least some of the indicators are based on the input. May be set.

[Additional note 5]
In the support device 10 described in any one of Supplementary Notes 1 to 4, the predetermined condition list may be set based on random numbers in which at least some of the indicators follow a predetermined probability distribution.

[Additional note 6]
The support device 10 described in any one of Supplementary Notes 1 to 5 may further include a display unit 16 that displays the first determination result based on the first statistical model determined by the driving state determination unit 14f.

[Additional note 7]
In the support device 10 described in Appendix 6, the statistical model generation unit 14c further generates a second statistical model based on process data regarding the operating state of the plant 1, and the operating state determination unit 14f further generates a second statistical model. The operating state of the plant 1 is determined with reference to the model, and the display unit 16 displays at least one of the first determination result and the second determination result based on the second statistical model determined by the operating state determination unit 14f. It's okay.

[Additional note 8]
In the support device 10 described in Appendix 7, the display unit 16 displays discrimination information that distinguishes each of the first determination result and the second determination result in association with at least one of the first determination result and the second determination result. You may.

[Additional note 9]
The support device 10 described in Appendix 7 further includes an operation reception unit 12b that receives an operation from the user regarding switching the display method of the determination result, and the display unit 16 displays the first determination result and the first determination result in accordance with the user's operation. Both the second determination result and the second determination result may be displayed.

[Additional note 10]
A statistical model generation device according to another aspect of the present invention is a device that generates a statistical model for determining the operating state of a plant 1, and calculates the statistical model by inputting data including a predetermined condition list into a physical model simulator. a data acquisition unit that acquires simulation data regarding the operating state of the plant 1, in which the predetermined condition list is associated with factors that determine the operating state of the plant 1 and indicators for the factors;
A statistical model generation unit 14c that generates a first statistical model based on simulation data is provided.

[Additional note 11]
A support method according to another aspect of the present invention is a support method for supporting the operation of a plant 1, in which the support method for the plant 1 is calculated by inputting data including a predetermined condition list into a physical model simulator. The predetermined condition list is based on the data acquisition step and the simulation data, in which factors that determine the operating condition of the plant 1 and indicators for the factors are associated. The method includes a statistical model generation step of generating a first statistical model, and an operating state determination step of determining an operating state of the plant 1 based on process data of the plant 1 with reference to the first statistical model.

[Additional note 12]
The support program according to another aspect of the present invention is a support program for supporting the operation of the plant 1, and is calculated by inputting data including a predetermined condition list into a physical model simulator. A data acquisition means that acquires simulation data regarding the operating state of the plant 1, wherein the predetermined condition list includes a data acquisition means that is associated with factors that determine the operating state of the plant 1 and indicators for the factors, and simulation data. function as a statistical model generating means for generating a first statistical model based on the first statistical model; and an operating state determining means for determining the operating state of the plant 1 based on the process data of the plant 1 with reference to the first statistical model. .

DESCRIPTION OF SYMBOLS 1... Plant, 10... Support device, 12... Input unit, 14... Processing unit, 14a... Control unit, 14b... Process data acquisition unit, 14c... Statistical model generation unit, 14e... Physical model simulator unit, 14f... Operating state determination Part, 14g...Display control unit, 16...Display unit, 18...Storage unit, 106...External storage device, 108...User interface, 110...Display, 112...Communication interface, 600...Driving diagnosis screen, 660...Alarm list display Area, 700...First statistical model comparison screen, 800...Second statistical model comparison screen

Claims (12)

A support device for supporting plant operation,
A data acquisition unit that acquires simulation data regarding the operating state of the plant calculated by inputting data including a predetermined condition list into a physical model simulator, the predetermined condition list determining the operating state of the plant. a data acquisition unit, in which a determining factor and an index for the factor are associated;
a statistical model generation unit that generates a first statistical model based on the simulation data;
an operating state determination unit that refers to the first statistical model and determines the operating state of the plant based on process data of the plant;
A support device equipped with. The support device according to claim 1, wherein the predetermined condition list includes one or more of the factors, and one or more of the indicators are associated with each of the factors. The predetermined condition list includes a plurality of records,
The support device according to claim 2, wherein in the record, one of the indicators is associated with each of the one or more factors. further comprising an input reception unit that receives input of the index from the user,
The support device according to claim 1, wherein the predetermined condition list is set at least a part of the index based on the input. The support device according to claim 1, wherein the predetermined condition list is set based on random numbers in which at least some of the indicators follow a predetermined probability distribution. The support device according to any one of claims 1 to 5, further comprising a display unit that displays a first determination result based on the first statistical model determined by the driving state determination unit. The statistical model generation unit further generates a second statistical model based on process data regarding the operating state of the plant,
The operating state determining unit further determines the operating state of the plant with reference to the second statistical model,
The support device according to claim 6, wherein the display section displays at least one of the first determination result and the second determination result based on the second statistical model determined by the driving state determination section. 8. The display unit displays discrimination information that distinguishes each of the first determination result and the second determination result in association with at least one of the first determination result and the second determination result. Support equipment as described. further comprising an operation reception unit that receives an operation from the user regarding switching the display method of the determination result;
The support device according to claim 7, wherein the display unit displays both the first determination result and the second determination result in response to an operation from the user. A device that generates a statistical model for determining the operating state of a plant, the device comprising:
A data acquisition unit that acquires simulation data regarding the operating state of the plant calculated by inputting data including a predetermined condition list into a physical model simulator, the predetermined condition list determining the operating state of the plant. a data acquisition unit, in which a determining factor and an index for the factor are associated;
a statistical model generation unit that generates a first statistical model based on the simulation data;
A statistical model generation device comprising: A support method for supporting plant operation, comprising:
a data acquisition step of acquiring simulation data regarding the operating state of the plant calculated by inputting data including a predetermined condition list into a physical model simulator, the predetermined condition list including the operating state of the plant; a data acquisition step, in which a determining factor and an indicator for said factor are associated;
a statistical model generation step of generating a first statistical model based on the simulation data;
an operating state determination step of determining an operating state of the plant based on process data of the plant with reference to the first statistical model;
How we can help, including: A support program for supporting the operation of a plant, which uses a computer to
A data acquisition means for acquiring simulation data regarding the operating state of the plant calculated by inputting data including a predetermined condition list into a physical model simulator, the predetermined condition list determining the operating state of the plant. a data acquisition means, in which a determining factor and an indicator for the factor are associated;
statistical model generation means for generating a first statistical model based on the simulation data;
Operating state determining means for determining the operating state of the plant based on process data of the plant with reference to the first statistical model;
A support program that functions as a

Images