Information providing device, information providing method, and information providing program

Abstract

To easily verify the effect of AI control introduced into a system.SOLUTION: A server device 10 generates N simulation models, based on an operation data collected from a system; determines a simulation model having the largest simulation model evaluation value for evaluating, based on the operation data, the simulation model; generates M AI control models, based on the operation data and the determined simulation model; and determines an AI control model having the largest AI control model evaluation value for evaluating, based on the operation data, the AI control model.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an information providing device, an information providing method, and an information providing program. [Background technology]

[0002] When introducing artificial intelligence (AI) control into a plant process, an AI control introduction effect report is known to be used to verify in advance whether or not AI control can be introduced and to verify the effects of AI control. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2023-042919 Summary of the Invention [Problem to be solved by the invention]

[0004] However, it is difficult to easily verify the effects of AI control introduced into systems such as plants. For example, when creating a report on the effects of introducing AI control into a plant process, a skilled engineer must manually create a plant simulator model and adjust the setting parameters, create an AI control model and adjust the setting parameters, and then use the created plant simulator model and AI control model to create the report on the effects of introducing AI control.

[0005] The present invention has been made in view of the above, and aims to easily verify the effects of AI control introduced into a system. [Means for solving the problem]

[0006] An information providing device according to one embodiment of the present invention includes a first generation unit that generates, based on driving data collected from a system, a plurality of first models that output predictive data of the driving status of the system in response to input of the driving data; a first evaluation unit that determines, based on the driving data, the first model with the largest first evaluation value for evaluating the first models; a second generation unit that generates, based on the driving data and the determined first models, a plurality of second models that output control data related to control of the system in response to input of the driving data; and a second evaluation unit that determines, based on the driving data, the second model with the largest second evaluation value for evaluating the second models.

[0007] In an information provision method according to one embodiment of the present invention, a computer executes the following processes: based on driving data collected from a system, a plurality of first models are generated, each of which outputs predictive data on the driving status of the system in response to input of the driving data; based on the driving data, a first model having the largest first evaluation value for evaluating the first models is determined; based on the driving data and the determined first models, a plurality of second models are generated, each of which outputs control data related to control of the system in response to input of the driving data; and based on the driving data, a second model having the largest second evaluation value for evaluating the second models is determined.

[0008] An information provision program according to one embodiment of the present invention causes a computer to execute the following processes: based on driving data collected from a system, generate a plurality of first models that output predictive data on the driving status of the system in response to input of the driving data; based on the driving data, determine the first model with the largest first evaluation value for evaluating the first models; based on the driving data and the determined first models, generate a plurality of second models that output control data related to control of the system in response to input of the driving data; and based on the driving data, determine the second model with the largest second evaluation value for evaluating the second models. [Effects of the Invention]

[0009] According to the present invention, it is possible to easily verify the effectiveness of AI control introduced into a system. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a diagram illustrating a configuration example and a processing example of an effect report providing system according to an embodiment; [Figure 2] 1 is a block diagram showing an example of the configuration of each device in an effect report providing system according to an embodiment. [Figure 3] 1 is a block diagram showing an example of processing of the entire effect report providing system according to an embodiment; [Figure 4] 4 is a diagram illustrating an example of a first setting value storage unit of the server device according to the embodiment. FIG. [Figure 5] FIG. 2 is a diagram illustrating an example of a first model storage unit of the server device according to the embodiment. [Figure 6] 4 is a diagram illustrating an example of a second setting value storage unit of the server device according to the embodiment. FIG. [Figure 7] FIG. 4 is a diagram illustrating an example of a second model storage unit of the server device according to the embodiment. [Figure 8] FIG. 10 is a diagram illustrating an example of an effect report storage unit of the server device according to the embodiment. [Figure 9] FIG. 10 is a diagram illustrating an example of a training data storage unit of the operator terminal according to the embodiment. [Figure 10] FIG. 10 is a diagram illustrating an example of an evaluation data storage unit of the operator terminal according to the embodiment. [Figure 11] FIG. 10 is a diagram illustrating an example of a definition information storage unit of the operator terminal according to the embodiment. [Figure 12] FIG. 2 is a diagram showing a specific example 1 of a display screen of an operator terminal according to the embodiment. [Figure 13] FIG. 10 is a diagram showing a specific example 2 of a display screen of an operator terminal according to the embodiment. [Figure 14] 1 is a flowchart showing an example of the overall flow of an effect report providing system according to an embodiment. [Figure 15]10 is a flowchart showing an example of the flow of a simulation model generation management process of the effect report providing system according to the embodiment. [Figure 16] 10 is a flowchart showing an example of the flow of an AI control model generation and management process of the effect report providing system according to the embodiment. [Figure 17] 10 is a flowchart showing an example of the flow of an AI control introduction effect report generation and management process of the effect report providing system according to the embodiment. [Figure 18] FIG. 2 is a diagram illustrating an example of a hardware configuration according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0011] An information providing device, an information providing method, and an information providing program according to embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Note that the present invention is not limited to the embodiments described below.

[0012] The following describes the configuration and processing of the effect report providing system 100 according to the embodiment, the configuration and processing of each device of the effect report providing system 100, the processing flow of the effect report providing system 100, the effects of the embodiment, and applications of the embodiment.

[0013] 1. Configuration and Processing of the Effect Report Providing System 100 The configuration and processing of the effect report providing system 100 according to the embodiment will be described in detail using Figure 1. Figure 1 is a diagram showing an example of the configuration and processing of the effect report providing system 100 according to the embodiment. Below, an example of the overall configuration of the effect report providing system 100, an example of the processing of the effect report providing system 100, and the effects of the effect report providing system 100 will be described. Note that in the embodiment, the creation of an AI control introduction effect report in a plant will be described as an example, but the field of use is not limited thereto, and the system can also be applied to air conditioning control in convenience stores, supermarkets, etc., for example.

[0014] (1-1. Example of the overall configuration of the effect report providing system 100) The effect report providing system 100 includes a server device 10 and an operator terminal 20. The server device 10 and the operator terminal 20 are connected to each other via a predetermined communication network (not shown) so as to be able to communicate with each other via wired or wireless communication. The predetermined communication network may be any of various communication networks such as the Internet or a dedicated line.

[0015] (1-1-1. Server device 10) The server device 10 is an information providing device that generates a simulation model SM and an AI-controlled model CM, and generates an AI-controlled model introduction effect report (simply referred to as an "effect report" as appropriate). For example, the server device 10 may be realized in a cloud environment, an on-premise environment, an edge environment, or the like. Note that the effect report providing system 100 shown in FIG. 1 may include multiple server devices 10.

[0016] (1-1-2. Operator terminal 20) The operator terminal 20 is a manager terminal used by an operator O who is a manager of the plant. Note that the effect report providing system 100 shown in FIG.

[0017] (1-2. Example of the overall processing of the effect report providing system 100) The following describes the overall processing of the effect report providing system 100. Note that the processing of steps S1 to S11 below may be executed in a different order. Also, some of the processing of steps S1 to S11 below may be omitted.

[0018] (1-2-1. Data input process for generating simulation model) First, the operator O inputs data for generating a simulation model to the server device 10 via the operator terminal 20 (step S1). For example, the operator O transmits to the server device 10, by operating the operator terminal 20, training data and evaluation data to be input to the simulation model (appropriately, "first model") SM, which are operation data for the plant stored in the operator terminal 20.

[0019] Here, the simulation model SM is a machine learning model that has the function of dynamically predicting the next state of a process to be controlled from the previous state and operation, and is, for example, a machine learning model that outputs predicted data of the operating status of a plant in response to input of operating data collected from the plant.

[0020] Furthermore, the operating data is time-series data indicating the process values ​​of each process collected from a system such as a plant, and is a data set that associates, for example, sensor values ​​indicating the state of the process, such as temperature and pressure, with operation values ​​indicating operations on the process, such as the opening degree of a valve.

[0021] Furthermore, the operator O transmits the input / output definition information for generating a simulation model (appropriately referred to as "first definition information") stored in the operator terminal 20 to the server device 10 by operating the operator terminal 20.

[0022] Here, the input / output definition information for generating a simulation model refers to setting values ​​that define the input data and output data of the simulation model SM, such as an input tag, which is the name or identifier of a process value that changes the state of a process such as a control valve, an output tag, which is the name or identifier of a process value that indicates the state of the process such as temperature or pressure, and a delay time between input and output that indicates the time it takes for the output data to change when the input data changes.

[0023] (1-2-2. Simulation model generation setting parameter creation process) Second, the server device 10 creates setting parameters for generating a simulation model ("first setting values" as appropriate) (step S2). For example, the server device 10 creates N patterns of setting parameters for generating a simulation model by grid research.

[0024] Here, the setting parameters for generating a simulation model are candidates for hyperparameters of the simulation model SM, and are candidates for internal parameters used to generate the simulation model SM in a system such as a plant, for example.

[0025] (1-2-3. Simulation model generation process) Third, the server device 10 generates a simulation model SM (step S3). For example, the server device 10 inputs a data set of sensor values ​​and operation values ​​of the plant as training data to the simulation model SM, inputs N patterns of setting parameters for generating a simulation model to the simulation model SM, and inputs / output definition information for generating a simulation model to the simulation model SM, thereby generating N simulation models SM.

[0026] (1-2-4. Simulation model evaluation process) Fourth, the server device 10 evaluates the simulation model SM (step S4). For example, the server device 10 inputs a data set of the plant's sensor values ​​and operation values ​​as evaluation data into the simulation model SM, and calculates a simulation model evaluation value (appropriately referred to as a "first evaluation value"). At this time, the server device 10 calculates, as the simulation model evaluation value, for example, the agreement rate between the sensor values ​​collected from the plant and the sensor values ​​output as prediction data, and determines the simulation model SM with the highest agreement rate as the best simulation model SM (Best) with the best performance, out of the N simulation models SM generated in the processing of step S3.

[0027] (1-2-5. Data input processing for generating AI control models) Fifth, the operator O inputs data for generating an AI control model to the server device 10 via the operator terminal 20 (step S5). For example, the operator O transmits to the server device 10, by operating the operator terminal 20, the training data and evaluation data to be input into the AI ​​control model (appropriately, the "second model") CM, which are operating data for the plant stored in the operator terminal 20.

[0028] Here, the AI ​​control model CM is a machine learning model that has the function of dynamically manipulating a process to be controlled from a previous state to a next state, and is, for example, a machine learning model that outputs plant control data in response to input of operation data collected from a plant. The control data is, for example, a control instruction indicating an operation to be instructed to a plant operator O or worker, or a control signal indicating an operation to be performed by an actuator.

[0029] Furthermore, the operator O transmits the input / output definition information for generating an AI control model (appropriately referred to as "second definition information") stored in the operator terminal 20 to the server device 10 by operating the operator terminal 20.

[0030] Here, the input / output definition information for generating an AI control model refers to setting values ​​that define the input data and output data of the AI ​​control model CM, such as an observation tag, which is the name or identifier of a process value, such as temperature or pressure, that the AI ​​control model CM observes when operating; a control tag, which is the name or identifier of a process value, such as a control valve, that the AI ​​control model CM operates; and the operation period when controlling.

[0031] (1-2-6. AI control model generation setting parameter creation process) Sixth, the server device 10 creates setting parameters for generating an AI control model ("second setting values" as appropriate) (step S6). For example, the server device 10 creates setting parameters for generating an M-pattern simulation model by grid research.

[0032] Here, the setting parameters for generating an AI control model are candidates for hyperparameters of the AI ​​control model CM, and are candidates for internal parameters used to generate the AI ​​control model CM in a system such as a plant.

[0033] (1-2-7. AI control model generation process) Seventh, the server device 10 generates the AI ​​control model CM (step S7). For example, the server device 10 inputs a data set of the plant's sensor values ​​and operation values ​​as training data to the AI ​​control model CM, inputs M patterns of setting parameters for generating the AI ​​control model to the AI ​​control model CM, inputs input / output definition information for generating the AI ​​control model to the AI ​​control model CM, and generates M AI control models CM using the best simulation model SM (Best) with the best performance determined in the processing of step S4.

[0034] (1-2-8. AI control model evaluation process) Eighth, the server device 10 evaluates the AI ​​control model CM (step S8). For example, the server device 10 inputs a data set of the plant's sensor values ​​and operation values ​​as evaluation data into the AI ​​control model CM, and calculates an AI control model evaluation value (appropriately referred to as a "second evaluation value"). At this time, the server device 10 calculates, as the AI ​​control model evaluation value, a reward value when controlling the plant process using a reward function, for example, and determines the AI ​​control model CM with the highest reward value as the best AI control model CM (Best) with the best performance, out of the M simulation models SM generated in the processing of step S7.

[0035] (1-2-9. Data input process for generating a report on the effects of introducing AI control) Ninth, the operator O inputs data for generating an AI control introduction effect report to the server device 10 via the operator terminal 20 (step S9). For example, the operator O operates the operator terminal 20 to transmit evaluation data, which is operational data for the plant stored in the operator terminal 20 and is to be used to generate an AI control introduction effect report, to the server device 10.

[0036] Here, the AI ​​control introduction effect report is a report that shows the effect of introducing an AI control model CM into a system, and is a report that visualizes, for example, the sensor values ​​when an AI control model CM is not introduced in a plant with the sensor values ​​when an AI control model CM is introduced, using a graph or the like that allows comparison.

[0037] (1-2-10. AI control introduction effect report generation process) Tenth, the server device 10 generates an AI control introduction effect report (step S10). For example, the server device 10 generates an AI control introduction effect report that displays, side by side, a bar graph showing the average value and standard deviation of the sensor values ​​output by inputting past operation values ​​for the plant process into the best simulation model SM (Best) and a bar graph showing the average value and standard deviation of the sensor values ​​output by inputting control data output by the best AI control model CM (Best) into the best simulation model SM (Best). The server device 10 also generates an AI control introduction effect report that displays, side by side, a time series graph showing the time series changes in the sensor values ​​output by inputting past operation values ​​for the plant process into the best simulation model SM (Best) and a time series graph showing the time series changes in the sensor values ​​output by inputting control data output by the best AI control model CM (Best) into the best simulation model SM (Best).

[0038] (1-2-11. AI control introduction effect report notification processing) Eleventh, the server device 10 notifies the operator O of the AI ​​control introduction effect report (step S11). For example, the server device 10 transmits the generated AI control introduction effect report to the operator terminal 20, and causes the monitor of the operator terminal 20 to display a bar graph or a time series graph.

[0039] (1-2-12.Other) In the effect report providing system 100, the AI ​​control model CM is used not only to create an AI control introduction effect report but also when introducing the AI ​​control model CM into an actual plant site. Therefore, among the processes of steps S1 to S11 described above, step S10, which executes an AI control introduction effect report generation process, and step S11, which executes an AI control introduction effect report notification process, can be omitted.

[0040] (1-3. Effects of the effect report providing system 100) Below, an overview and problems of the effect report providing system 100P according to the reference technology will be explained, and then the effects of the effect report providing system 100 will be explained.

[0041] (1-3-1. Overview of the 100-page effectiveness report system) The effect report providing system 100P according to the reference technology creates an AI control introduction effect report that is used to verify the feasibility of introducing AI control and to verify the effects of AI control in advance when introducing AI control into a plant process. The following describes the simulation model creation process, AI control model creation process, and AI control introduction effect report creation process executed by the effect report providing system 100P.

[0042] (1-3-1-1. Simulation model creation process) An engineer inputs the training data, evaluation data, and simulation model setting parameters into a simulation model generator to create a simulation model SM. At this time, the accuracy of the simulation model SM is evaluated by comparing the actual operating data with the simulation data. The engineer also manually adjusts the simulation model setting parameters to create a simulation model SM with high prediction accuracy.

[0043] (1-3-1-2. AI control model creation process) Engineers input training data, evaluation data, simulation model SM, and AI control model setting parameters into the AI ​​control model generator to create the AI ​​control model CM. At this time, the algorithm for the AI ​​control model CM uses reinforcement learning technology, allowing the AI ​​to learn through trial and error the policy that maximizes the reward value, thereby creating the AI ​​control model. Engineers manually adjust the AI ​​control model setting parameters to create a model with high control performance.

[0044] (1-3-1-3. AI control introduction effect report creation process) The engineer creates an AI control introduction effect report using the simulation model SM and AI control model CM created using the procedure described above. At this time, the engineer compares the simulation data of the simulation model SM with the control data of the AI ​​control model CM, and creates an AI control introduction effect report that describes the effect of introducing the AI ​​control model CM.

[0045] (1-3-2. Problems with the 100P Effectiveness Report System) The effect report providing system 100P related to the reference technology has the problem that in order to create an AI control introduction effect report, it is necessary for an engineer who is familiar with the plant's simulation model SM and AI control model CM to manually perform complex system design and model adjustment.In addition, the effect report providing system 100P has the problem that it takes time and costs money to create an AI control introduction effect report.

[0046] (1-3-3. Overview of the effect report providing system 100) The following processes are executed in the effect report providing system 100. Below, we will explain the simulation model generation management process, AI control model generation management process, and AI control introduction effect report generation management process executed by the effect report providing system 100.

[0047] (1-3-3-1. Simulation model generation management process) The effect report providing system 100 executes the following simulation model generation management process. First, an operator O inputs data for generating a simulation model to the server device 10 via the operator terminal 20. Second, the server device 10 creates N patterns of setting parameters for generating a simulation model by grid research. Third, the server device 10 generates N simulation models SM using the data for generating a simulation model and the N patterns of setting parameters for generating a simulation model. Fourth, the server device 10 evaluates the simulation models SM using the match rate and determines the best simulation model SM (Best) from the N simulation models SM.

[0048] (1-3-3-2. AI control model generation management process) The effect report providing system 100 executes the following AI control model generation management process. First, the operator O inputs data for generating an AI control model to the server device 10 via the operator terminal 20. Second, the server device 10 creates M patterns of setting parameters for generating an AI control model by grid research. Third, the server device 10 generates M AI control models CM using the data for generating an AI control model and the M patterns of setting parameters for generating an AI control model. Fourth, the server device 10 evaluates the AI ​​control models CM using a reward value and determines the best AI control model CM (Best) from the M simulation models SM.

[0049] (1-3-3-3. AI control introduction effect report generation and management process) The effect report providing system 100 executes the following AI control introduction effect report generation management process. First, the operator O inputs data for effect report generation to the server device 10 via the operator terminal 20. Second, the server device 10 generates an AI control introduction effect report. Third, the server device 10 notifies the operator O of the AI ​​control introduction effect report.

[0050] (1-3-4. Effects of the effect report providing system 100) The effect report providing system 100 has the following effects. First, in the effect report providing system 100, even an operator O who is a non-AI engineer with no knowledge of the simulation model SM or the AI ​​control model CM in the plant can verify the effects of introducing the AI ​​control model CM. Second, in the effect report providing system 100, the operator O can automatically generate the simulation model SM and the AI ​​control model CM and automatically generate an AI control introduction effect report simply by inputting the plant's process values. Therefore, the contents of the AI ​​control introduction effect report can be used to verify the effects of introducing AI control by comparing them with existing control. Third, in the effect report providing system 100, the server device 10 can output, as an AI control introduction effect report, graphs of statistical values ​​(average value, standard deviation, etc.) of the output results of AI control, offset (deviation), rise time, etc., and can output a graph that allows comparison of the output results of AI control with those of not implementing AI control.

[0051] As described above, the effect report providing system 100 makes it easy to verify the effect of AI control introduced into the system.

[0052] 2. Configuration and Processing of Each Device of the Effect Report Providing System 100 2 to 13, the configuration and processing of each device included in the effect report providing system 100 shown in Fig. 1 will be described. Below, an example of the configuration and processing of the entire effect report providing system 100 according to the embodiment will be described, followed by a detailed description of an example of the configuration and processing of the server device 10 and an example of the configuration and processing of the operator terminal 20.

[0053] (2-1. Example of the overall configuration of the effect report providing system 100) An example of the overall configuration of the effect report providing system 100 shown in Fig. 1 will be described using Fig. 2. Fig. 2 is a block diagram showing an example of the configuration of each device of the effect report providing system 100 according to the embodiment. As shown in Fig. 2, the effect report providing system 100 has a server device 10 and an operator terminal 20. Furthermore, the server device 10 and the operator terminal 20 are communicatively connected via a communication network N realized by the Internet, a dedicated line, or the like.

[0054] The server device 10 is installed in a cloud environment, an on-premise environment, an edge environment, etc. The operator terminal 20 is installed in a monitoring room of a plant managed by an operator O, etc.

[0055] (2-2. Example of the overall processing of the effect report providing system 100) An example of the overall processing of the effect report providing system 100 shown in Fig. 1 will be described using Fig. 3. Fig. 3 is a block diagram showing an example of the overall processing of the effect report providing system 100 according to the embodiment. A simulation model generation management process, an AI control model generation management process, and an AI control introduction effect report generation management process executed by the effect report providing system 100 will be described.

[0056] (2-2-1. Simulation model generation management process) First, a description will be given of an example of a simulation model generation management process executed by the effect report providing system 100. In the simulation model generation management process, a first setting value storage unit 14a and a first model storage unit 14b of the storage unit 14 of the server device 10, a first acquisition unit 15a, a first adjustment unit 15b, a first generation unit 15c, a first evaluation unit 15d, and the like, which will be described later, of the control unit 15, execute the process.

[0057] (2-2-1-1. Data input process for generating simulation models) In the simulation model generation data input process, the operator terminal 20 inputs the training data, evaluation data, and simulation model input / output definition information, which are data for generating a simulation model, to the first acquisition unit 15a of the server device 10. The first acquisition unit 15a then inputs the input training data, evaluation data, and simulation model input / output definition information to the first generation unit 15c. The first acquisition unit 15a then inputs the input evaluation data to the first evaluation unit 15d.

[0058] (2-2-1-2. Simulation model generation setting parameter creation process) In the simulation model generation setting parameter creation process, the first adjustment unit 15b reads the simulation model generation setting parameters before adjustment by referring to the first setting value storage unit 14a, and creates N patterns of simulation model generation setting parameters by grid research. The first adjustment unit 15b also stores the created N patterns of simulation model generation setting parameters in the first setting value storage unit 14a.

[0059] (Example of setting parameters for generating a simulation model) Here, a specific example of the setting parameters for generating a simulation model created by the first adjustment unit 15b will be described. For example, if there are two types of setting parameters A and B for generating a simulation model before adjustment, and the candidate values ​​of setting parameter A are "0.1, 0.01, 0.001" and the candidate values ​​of setting parameter B are "1, 0.1, 0.01," the first adjustment unit 15b combines the candidate values ​​of each setting parameter to create nine patterns of setting parameters for generating a simulation model, namely, (A:0.1,B:1), (A:0.1,B:0.1), (A:0.1,B:0.01), (A:0.01,B:1), (A:0.01,B:0.1), (A:0.01,B:0.01), (A:0.001,B:1), (A:0.001,B:0.1), and (A:0.001,B:0.01).

[0060] (2-2-1-3. Simulation model generation process) In the simulation model generation process, the first generation unit 15c reads adjusted setting parameters for generating a simulation model by referring to the first setting value storage unit 14a via the first adjustment unit 15b. The first generation unit 15c also inputs training data to the simulation model SM, inputs N patterns of setting parameters for generating a simulation model to the simulation model SM, and inputs input / output definition information for generating a simulation model to the simulation model SM, thereby generating N simulation models SM. The first generation unit 15c also stores the generated N simulation models SM in the first model storage unit 14b.

[0061] (2-2-1-4. Simulation model evaluation process) In the simulation model evaluation process, the first evaluation unit 15d refers to the N simulation models SM stored in the first model storage unit 14b, inputs evaluation data into each of the N simulation models SM, calculates a simulation model evaluation value for each, and determines the simulation model SM with the largest simulation model evaluation value among the N simulation models SM as the best simulation model SM (Best) with the best performance. Furthermore, the first evaluation unit 15d stores only the best simulation model SM (Best) in the first model storage unit 14b and deletes the other simulation models SM from the first model storage unit 14b.

[0062] (2-2-2. AI control model generation management process) Second, we will explain an example of the AI-controlled model generation management process executed by the effect report providing system 100. In the AI-controlled model generation management process, the second setting value storage unit 14c and second model storage unit 14d of the storage unit 14 of the server device 10, the second acquisition unit 15e, second adjustment unit 15f, second generation unit 15g, second evaluation unit 15h of the control unit 15, and the like, which will be described later, execute the process.

[0063] (2-2-2-1. Data input processing for generating AI control models) In the AI-controlled model generation data input process, the operator terminal 20 inputs training data, evaluation data, and AI-controlled model input / output definition information, which are data for generating an AI-controlled model, to the second acquisition unit 15e of the server device 10. The second acquisition unit 15e then inputs the input training data, evaluation data, and AI-controlled model input / output definition information to the second generation unit 15g. The second acquisition unit 15e then inputs the input evaluation data to the second evaluation unit 15h.

[0064] (2-2-2-2. AI control model generation setting parameter creation process) In the AI-control-model-generating setting parameter creation process, second adjustment unit 15f reads the pre-adjustment AI-control-model-generating setting parameters by referencing second setting value storage unit 14c, and creates M-pattern AI-control-model-generating setting parameters by grid research. Second adjustment unit 15f also stores the created M-pattern AI-control-model-generating setting parameters in second setting value storage unit 14c.

[0065] (2-2-2-3. AI control model generation process) In the AI-controlled model generation process, the second generation unit 15g reads adjusted setting parameters for generating an AI-controlled model by referencing the second setting value storage unit 14c via the second adjustment unit 15f. The second generation unit 15g also inputs training data to the AI-controlled model CM, inputs M patterns of setting parameters for generating an AI-controlled model to the AI-controlled model CM, and inputs input / output definition information for generating an AI-controlled model to the AI-controlled model CM, thereby generating M AI-controlled models CM. The second generation unit 15g generates the AI-controlled model CM using the best simulation model SM (Best) stored in the first model storage unit 14b. The second generation unit 15g also stores the generated M AI-controlled models CM in the second model storage unit 14d.

[0066] (2-2-2-4. AI control model evaluation process) In the AI-controlled model evaluation process, the second evaluation unit 15h references the M AI-controlled models CM stored in the second model storage unit 14d, inputs evaluation data into each of the M AI-controlled models CM, calculates an AI-controlled model evaluation value for each, and determines the AI-controlled model CM with the largest AI-controlled model evaluation value among the M AI-controlled models CM as the best AI-controlled model CM (Best) with the best performance. Furthermore, the first evaluation unit 15d stores only the best AI-controlled model CM (Best) in the second model storage unit 14d, and deletes the other AI-controlled models CM from the second model storage unit 14d.

[0067] (2-2-3. AI control introduction effect report generation and management process) Thirdly, we will explain an example of the AI ​​control introduction effect report generation and management process executed by the effect report providing system 100. In the AI ​​control introduction effect report generation and management process, the effect report storage unit 14e of the storage unit 14 of the server device 10, the third acquisition unit 15i, the third generation unit 15j, the notification unit 15k, etc., which will be described later, execute the process.

[0068] (2-2-3-1. Data input process for generating a report on the effects of introducing AI control) In the data input process for generating an AI control introduction effect report, the operator terminal 20 inputs evaluation data, which is data for generating an AI control introduction effect report, to the third acquisition unit 15i of the server device 10. In addition, the third acquisition unit 15i inputs the input evaluation data to the third generation unit 15j.

[0069] (1-2-3-2. AI control introduction effect report generation process) In the AI ​​control introduction effect report generation process, the third generation unit 15j generates an AI control introduction effect report using the best simulation model SM (Best) stored in the first model storage unit 14b and the best AI control model CM (Best) stored in the second model storage unit 14d. At this time, the third generation unit 15j inputs evaluation data into the best simulation model SM (Best) to generate predicted data for when AI control is not executed. The third generation unit 15j also inputs control data output by inputting the evaluation data into the best AI control model CM (Best) into the best simulation model SM (Best) to generate predicted data for when AI control is executed. The third generation unit 15j also generates an AI control introduction effect report that includes bar graphs that allow comparison of statistical values ​​(average values, standard deviations) of each predicted data, time series graphs that allow comparison of time series changes in each predicted data, and the like. The third generation unit 15j also stores the generated AI control introduction effect report in the effect report storage unit 14e.

[0070] (1-2-3-3. AI Control Introduction Effect Report Notification Processing) In the AI ​​control introduction effect report notification process, the notification unit 15k acquires the AI ​​control introduction effect report stored in the effect report storage unit 14e and transmits the AI ​​control introduction effect report to the operator terminal 20. At this time, the notification unit 15k causes the monitor of the operator terminal 20 to display the AI ​​control introduction effect report.

[0071] (2-3. Configuration Example and Processing Example of Server Device 10) 2, a description will be given of an example of the configuration and processing of the server device 10. The server device 10 is an information providing device, and includes an input unit 11, an output unit 12, a communication unit 13, a storage unit 14, and a control unit 15.

[0072] (2-3-1. Input section 11) The input unit 11 controls input of various information to the server device 10. For example, the input unit 11 is realized by a mouse, a keyboard, etc., and accepts input of various information to the server device 10.

[0073] (2-3-2. Output section 12) The output unit 12 controls the output of various information from the server device 10. For example, the output unit 12 is realized by a display or the like, and displays various information stored in the server device 10.

[0074] (2-3-3. Communications Department 13) The communication unit 13 controls data communication with other devices. For example, the communication unit 13 performs data communication with each communication device via a router, etc. The communication unit 13 can also perform data communication with an operator's terminal (not shown).

[0075] (2-3-4. Storage section 14) The storage unit 14 stores various information referenced by the control unit 15 when it operates and various information acquired when the control unit 15 operates. The storage unit 14 includes a first setting value storage unit 14a, a first model storage unit 14b, a second setting value storage unit 14c, a second model storage unit 14d, and an effect report storage unit 14e. Here, the storage unit 14 may be realized by, for example, a semiconductor memory element such as a random access memory (RAM) or a flash memory, or a storage device such as a hard disk or an optical disk. Note that, in the example of FIG. 2, the storage unit 14 is installed inside the server device 10, but it may also be installed outside the server device 10, or multiple storage units may be installed.

[0076] (2-3-4-1. First set value storage unit 14a) The first setting value storage unit 14a stores setting parameters for generating a simulation model. For example, the first setting value storage unit 14a stores setting parameters for generating a simulation model created by a first adjustment unit 15b of the control unit 15, which will be described later. Here, an example of data stored in the first setting value storage unit 14a will be described with reference to FIG. 4. FIG. 4 is a diagram illustrating an example of the first setting value storage unit 14a of the server device 10 according to the embodiment. In the example of FIG. 4, the first setting value storage unit 14a has items such as "simulation model" and "setting parameters."

[0077] The "simulation model" indicates identification information for identifying the simulation model SM, such as an identification number or identification symbol of the simulation model SM. The "setting parameters" are candidates for hyperparameters to be input to the simulation model SM, such as a combination of candidate values ​​for each hyperparameter generated by grid research.

[0078] That is, FIG. 4 shows an example in which, for a simulation model SM identified by "simulation model SM-1," data in which setting parameters for generating N patterns of simulation models are "setting parameter SMP-1-1," "setting parameter SMP-1-2," "setting parameter SMP-1-3," ..., "setting parameter SMP-1-N," etc., are stored in the first setting value memory unit 14a.

[0079] (2-3-4-2. First model storage unit 14b) The first model storage unit 14b stores a simulation model SM. For example, the first model storage unit 14b stores the simulation model SM, which is a machine learning model generated by a first generation unit 15c of the control unit 15, which will be described later. The first model storage unit 14b also stores a simulation model evaluation value calculated by a first evaluation unit 15d of the control unit 15, which will be described later. Here, an example of data stored in the first model storage unit 14b will be described with reference to FIG. 5. FIG. 5 is a diagram illustrating an example of the first model storage unit 14b of the server device 10 according to the embodiment. In the example of FIG. 5, the first model storage unit 14b has items such as "simulation model" and "score."

[0080] The "simulation model" is model data of the trained simulation model SM, and is data including, for example, execution data for executing the algorithm of the simulation model SM, hyperparameters, etc. The "score" is an evaluation value indicating the performance of the trained simulation model SM, and is, for example, the rate of agreement between the sensor values ​​collected from the plant and the sensor values ​​output as prediction data.

[0081] That is, FIG. 5 shows an example in which N trained simulation models SM and simulation model evaluation values, which are {simulation model: "simulation model SM-1-1", score: "score SM-1-1"}, {simulation model: "simulation model SM-1-2", score: "score SM-1-2"}, {simulation model: "simulation model SM-1-3", score: "score SM-1-3"}, ..., {simulation model: "simulation model SM-1-N", score: "score SM-1-N"}, are temporarily stored in the first model storage unit 14b, and data other than {simulation model: "simulation model SM-1-3", score: "score SM-1-3"} is deleted.

[0082] (2-3-4-3. Second set value storage unit 14c) The second setting value storage unit 14c stores setting parameters for generating an AI control model. For example, the second setting value storage unit 14c stores setting parameters for generating an AI control model created by a second adjustment unit 15f of the control unit 15, which will be described later. An example of data stored in the second setting value storage unit 14c will now be described with reference to FIG. 6. FIG. 6 is a diagram illustrating an example of the second setting value storage unit 14c of the server device 10 according to the embodiment. In the example of FIG. 6, the first setting value storage unit 14a has items such as "control model" and "setting parameters."

[0083] "AI-controlled model" refers to identification information for identifying the AI-controlled model CM, such as an identification number or symbol of the AI-controlled model CM. "Configuration parameters" refer to candidate hyperparameters to be input to the AI-controlled model CM, such as a combination of candidate values ​​for each hyperparameter generated by grid research.

[0084] That is, Figure 6 shows an example in which, for the AI ​​control model CM identified by "AI control model CM-1," data in which the setting parameters for generating M patterns of AI control models are "setting parameter CMP-1-1," "setting parameter CMP-1-2," "setting parameter CMP-1-3," ..., "setting parameter CMP-1-M," etc. are stored in the second setting value memory unit 14c.

[0085] (2-3-4-4. Second model storage unit 14d) The second model storage unit 14d stores an AI-controlled model CM. For example, the second model storage unit 14d stores the AI-controlled model CM, which is a machine learning model generated by a second generation unit 15g of the control unit 15, which will be described later. Here, an example of data stored in the second model storage unit 14d will be described with reference to FIG. 7. FIG. 7 is a diagram illustrating an example of the second model storage unit 14d of the server device 10 according to the embodiment. In the example of FIG. 7, the second model storage unit 14d has items such as "AI-controlled model" and "score."

[0086] An "AI control model" is model data of a trained AI control model CM, and is data including, for example, execution data for executing the algorithm of the AI ​​control model CM, hyperparameters, etc. A "score" is an evaluation value that indicates the performance of the trained AI control model CM, and is, for example, a reward value calculated using a reward function when controlling a plant process.

[0087] That is, Figure 7 shows an example in which M trained AI-controlled models CM and AI-controlled model evaluation values, which are {AI-controlled model: "AI-controlled model CM-1-1", score: "score CM-1-1"}, {AI-controlled model: "AI-controlled model CM-1-2", score: "score CM-1-2"}, {AI-controlled model: "AI-controlled model CM-1-3", score: "score CM-1-3"}, ..., {AI-controlled model: "AI-controlled model CM-1-M", score: "score CM-1-M"}, are temporarily stored in the first model storage unit 14b, and data other than {AI-controlled model: "AI-controlled model CM-1-2", score: "score CM-1-2"} is deleted.

[0088] (2-3-4-5. Effect report storage unit 14e) The effect report storage unit 14e stores an AI control introduction effect report. For example, the effect report storage unit 14e stores an AI control introduction effect report generated by a third generation unit 15j of the control unit 15, which will be described later. Here, an example of data stored in the effect report storage unit 14e will be described with reference to FIG. 8. FIG. 8 is a diagram showing an example of the effect report storage unit 14e of the server device 10 according to the embodiment. In the example of FIG. 8, the effect report storage unit 14e has items such as "Plant" and "Effect Report."

[0089] "Plant" indicates identification information for identifying the plant, which is a system, such as the plant's identification number or symbol. "Effectiveness Report" is a report showing the effect of introducing the AI-controlled model CM into the system, such as a report that visualizes the sensor values ​​when the AI-controlled model CM is not introduced in the plant and the sensor values ​​when the AI-controlled model CM is introduced, using a graph or the like that allows comparison.

[0090] That is, FIG. 8 shows an example in which data for a plant identified by "Plant #1" in which the AI ​​control introduction effect report is "Effect Report #1" is stored in the effect report storage unit 14e.

[0091] (2-3-5. Control unit 15) The control unit 15 controls the entire server device 10. The control unit 15 has a first acquisition unit 15a, a first adjustment unit 15b, a first generation unit 15c, a first evaluation unit 15d, a second acquisition unit 15e, a second adjustment unit 15f, a second generation unit 15g, a second evaluation unit 15h, a third acquisition unit 15i, a third generation unit 15j, and a notification unit 15k. Here, the control unit 15 can be realized by, for example, an electronic circuit such as a CPU (Central Processing Unit) or an MPU (Micro Processing Unit), or an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

[0092] (2-3-5-1. 1st acquisition part 15a) The first acquisition unit 15a acquires various types of information. The first acquisition unit 15a may store the acquired various types of information in the storage unit .

[0093] The first acquisition unit 15a acquires operation data collected from the system. For example, the first acquisition unit 15a acquires, as operation data transmitted from the operator terminal 20, training data and evaluation data, each including a data set of sensor values ​​indicating the state of a plant process and operation values ​​indicating operations on the process. The first acquisition unit 15a also acquires first definition information defining input and output signals of the simulation model SM, i.e., input / output definition information for generating a simulation model. For example, the first acquisition unit 15a acquires, as the input / output definition information for generating a simulation model transmitted from the operator terminal 20, input tags, which are names or identifiers of process values ​​that change the state of the process, output tags, which are names or identifiers of process values ​​that indicate the state of the process, and delay times between inputs and outputs, which indicate the time it takes for output data to change when input data changes.

[0094] To explain a specific example, the first acquisition unit 15a acquires {Plant: "Plant #1", training data: "Training data #1"}, {Plant: "Plant #1", evaluation data: "Evaluation data #1"}, {Plant: "Plant #1", definition information 1: "Definition information #1-1"}, etc. as data for generating a simulation model sent by the operator terminal 20, and outputs them to the first generation unit 15c.

[0095] (2-3-5-2. 1st adjustment section 15b) The first adjustment unit 15b adjusts various pieces of information. The first adjustment unit 15b may store the adjusted various pieces of information in the storage unit 14. The first adjustment unit 15b may also refer to the various pieces of information stored in the storage unit 14.

[0096] The first adjustment unit 15b generates a plurality of first setting values, i.e., a plurality of setting parameters for generating a simulation model. For example, the first adjustment unit 15b refers to candidate values ​​of hyperparameters before adjustment of the simulation model SM and generates combinations of candidate values ​​of the hyperparameters by grid research.

[0097] To explain a specific example, the first adjustment unit 15b creates N patterns of setting parameters for generating a simulation model, namely, "setting parameters SMP-1-1," "setting parameters SMP-1-2," "setting parameters SMP-1-3," ..., "setting parameters SMP-1-N," for the simulation model SM identified by "simulation model SM-1," and stores them in the first setting value memory unit 14a.

[0098] (2-3-5-3. 1st generation section 15c) The first generating unit 15c generates various types of information. The first generating unit 15c may store the generated various types of information in the storage unit 14. The first generating unit 15c may also refer to the various types of information stored in the storage unit 14.

[0099] Based on operation data collected from the system, the first generation unit 15c generates a plurality of first models, i.e., a plurality of simulation models SM, that output prediction data of the operation status of the system in response to input of the operation data. For example, the first generation unit 15c uses, as training data, a data set of sensor values ​​indicating the state of the system's process and operation values ​​indicating operations on the process, as the operation data, a plurality of hyperparameters generated by grid research as setting parameters for generating the plurality of simulation models, and further uses input / output definition information for generating simulation models that defines input / output signals of the simulation models SM, thereby generating the plurality of simulation models SM.

[0100] To explain a specific example, the first generation unit 15c inputs {plant: "plant #1", training data: "training data #1"} as the dataset of the training data output by the first acquisition unit 15a, {plant: "plant #1", definition information 1: "definition information #1-1"} as the input / output definition information for generating the simulation model output by the first acquisition unit 15a, and "setting parameter SMP-1-1," "setting parameter SMP-1-2," "setting parameter SMP-1-3," ..., "setting parameter SMP-1-N" as setting parameters for generating N patterns of simulation models stored in the first setting value storage unit 14a into the simulation model SM identified by "simulation model SM-1," and generates "simulation model SM-1-1," "simulation model SM-1-2," "simulation model SM-1-3," ..., "simulation model SM-1-N" as N simulation models SM, and stores them in the first model storage unit 14b.

[0101] (2-3-5-4. First evaluation unit 15d) The first evaluation unit 15d evaluates various pieces of information. The first evaluation unit 15d may store the evaluation results in the storage unit 14. The first evaluation unit 15d may also refer to the various pieces of information stored in the storage unit 14.

[0102] The first evaluation unit 15d determines a first evaluation value for evaluating the multiple generated simulation models SM, i.e., the simulation model SM with the largest simulation model evaluation value, based on the operation data. For example, the first evaluation unit 15d uses, as the operation data, a data set of sensor values ​​indicating the state of the system's process and operation values ​​indicating operations on the process as evaluation data, calculates, as the simulation model evaluation value, the agreement rate between the sensor values ​​collected from the system and the sensor values ​​output as prediction data, and determines, from the multiple generated simulation models SM, the simulation model SM with the largest calculated agreement rate.

[0103] At this time, the first evaluation unit 15d calculates the coincidence rate between the operational data, which is the process value during actual control, and the process value simulated by the simulation model SM as a simulation model evaluation value S SM It is calculated as follows.

[0104]

number

[0105] To explain a specific example, the first evaluation unit 15d inputs {plant: "Plant #1", evaluation data: "Evaluation data #1"} as a data set of the evaluation data output by the first acquisition unit 15a into N simulation models SM, namely "simulation model SM-1-1", "simulation model SM-1-2", "simulation model SM-1-3", ..., "simulation model SM-1-N", and calculates N simulation model evaluation values, namely "score SM-1-1", "score SM-1-2", "score SM-1-3", ..., "score SM-1-N"}, and stores them in the first model memory unit 14b. In addition, the first evaluation unit 15d identifies "score SM-1-3" as the maximum simulation model evaluation value and "simulation model SM-1-3" as the corresponding best simulation model SM (Best), and deletes all data stored in the first model memory unit 14b except for {simulation model: "simulation model SM-1-3", score: "score SM-1-3"}.

[0106] (2-3-5-5.Second acquisition part 15e) The second acquisition unit 15e acquires various types of information. The second acquisition unit 15e may store the acquired various types of information in the storage unit .

[0107] The second acquisition unit 15e acquires operation data collected from the system. For example, the second acquisition unit 15e acquires, as operation data transmitted from the operator terminal 20, training data and evaluation data, each including a data set of sensor values ​​indicating the state of a plant process and operation values ​​indicating operations on the process. The second acquisition unit 15e also acquires second definition information defining input / output signals of the AI-controlled model CM, i.e., input / output definition information for generating an AI-controlled model. For example, the second acquisition unit 15e acquires, as input / output definition information for generating an AI-controlled model transmitted from the operator terminal 20, an observation tag, which is the name or identifier of a process value observed when the AI-controlled model CM is operated, a control tag, which is the name or identifier of a process value operated by the AI-controlled model CM, an operation period during control, and the like.

[0108] To explain a specific example, the second acquisition unit 15e acquires {Plant: "Plant #1", training data: "Training data #1"}, {Plant: "Plant #1", evaluation data: "Evaluation data #1"}, {Plant: "Plant #1", definition information 2: "Definition information #1-2"}, etc. as data for generating an AI control model sent by the operator terminal 20, and outputs them to the second generation unit 15g.

[0109] (2-3-5-6. 2nd adjustment section 15f) The second adjustment unit 15f adjusts various pieces of information. The second adjustment unit 15f may store the adjusted various pieces of information in the storage unit 14. The second adjustment unit 15f may also refer to the various pieces of information stored in the storage unit 14.

[0110] The second adjustment unit 15f generates a plurality of second setting values, i.e., a plurality of setting parameters for generating an AI-controlled model. For example, the second adjustment unit 15f references candidate values ​​of hyperparameters of the AI-controlled model CM before adjustment and generates combinations of candidate values ​​of the hyperparameters by grid research.

[0111] To explain a specific example, the second adjustment unit 15f creates M patterns of setting parameters for generating an AI control model, namely, "setting parameter CMP-1-1," "setting parameter CMP-1-2," "setting parameter CMP-1-3," ..., "setting parameter CMP-1-M," for the AI ​​control model CM identified by "AI control model CM-1," and stores them in the second setting value memory unit 14c.

[0112] (2-3-5-7. 2nd generation part 15g) The second generating unit 15g generates various types of information. The second generating unit 15g may store the generated various types of information in the storage unit 14. The second generating unit 15g may also refer to the various types of information stored in the storage unit 14.

[0113] The second generation unit 15g generates, based on the operation data collected from the system and the determined simulation model SM, a plurality of second models that output control data related to the control of the system in response to the input of the operation data, i.e., a plurality of AI control models CM. For example, the second generation unit 15g uses, as the operation data, a data set of sensor values ​​indicating the state of the system's process and operation values ​​indicating operations on the process as training data, a plurality of hyperparameters generated by grid research as a plurality of AI control generation setting parameters, and further uses AI control model generation input / output definition information that defines the input / output signals of the AI ​​control model CM, thereby generating the plurality of AI control models CM using the determined simulation model SM.

[0114] To explain a specific example, the second generation unit 15g sets {Plant: "Plant #1", training data: "Training data #1"} as the data set of the training data output by the second acquisition unit 15e, {Plant: "Plant #1", definition information 2: "Definition information #1-2"} as the input / output definition information for generating the AI ​​control model output by the second acquisition unit 15e, and {Plant: "Plant #1", definition information 2: "Definition information #1-2"} as the setting parameters for generating the AI ​​control model of the M pattern stored in the second setting value storage unit 14c. -3", ..., "setting parameters CMP-1-M" are input to the AI-controlled models CM identified by "AI-controlled model CM-1" in the execution environment of the "simulation model SM-1-3", which is the best simulation model SM (Best) stored in the first model storage unit 14b, to generate M AI-controlled models CM, namely "AI-controlled model CM-1-1", "AI-controlled model CM-1-2", "AI-controlled model CM-1-3", ..., "AI-controlled model CM-1-M", and store them in the second model storage unit 14d.

[0115] (2-3-5-8. Second evaluation part 15h) The second evaluation unit 15h evaluates various pieces of information. The second evaluation unit 15h may store the evaluation results in the storage unit 14. The second evaluation unit 15h may also refer to the various pieces of information stored in the storage unit 14.

[0116] The second evaluation unit 15h determines a second evaluation value for evaluating the multiple generated AI-controlled models CM, i.e., the AI-controlled model CM with the largest AI-controlled model evaluation value, based on the operation data. For example, the second evaluation unit 15h uses, as the operation data, a data set of sensor values ​​indicating the state of the system's process and operation values ​​indicating operations on the process, calculates, as the AI-controlled model evaluation value, an average value of reward values ​​when the system is controlled using a reward function, and determines, from the multiple generated AI-controlled models CM, the AI-controlled model CM with the largest average value of the calculated reward values.

[0117] At this time, the second evaluation unit 15h calculates the average value of the reward value R as the AI ​​control model evaluation value S using the reward value R and the evaluation period n of the evaluation data, as shown in the following formula (2). CM It is calculated as follows.

[0118]

number

[0119] Here, the reward value R is a value calculated by a reward function to evaluate how an agent in reinforcement learning will behave in an environment. For example, the difference between the target control value of the controlled object and the current value of the controlled object when controlled is used as the reward function, but there is no particular limitation on the reward function used.

[0120] To explain a specific example, the second evaluation unit 15h inputs {Plant: "Plant #1", evaluation data: "Evaluation data #1"} as a dataset of evaluation data output by the second acquisition unit 15e into M AI-controlled models CM, namely, "AI-controlled model CM-1-1", "AI-controlled model CM-1-2", "AI-controlled model CM-1-3", ..., "AI-controlled model CM-1-M", and calculates M AI-controlled model evaluation values, namely, "Score CM-1-1", "Score CM-1-2", "Score CM-1-3", ..., "Score CM-1-M"}, and stores them in the second model memory unit 14d. In addition, the second evaluation unit 15h identifies "Score CM-1-2" as the maximum AI control model evaluation value and "AI control model CM-1-2" as the corresponding best AI control model CM (Best), and deletes all data stored in the second model memory unit 14d except for {AI control model: "AI control model CM-1-2", score: "Score CM-1-2"}.

[0121] (2-3-5-9. Third acquisition part 15i) The third acquisition unit 15i acquires various types of information. The third acquisition unit 15i may store the acquired various types of information in the storage unit .

[0122] The third acquisition unit 15i acquires operation data collected from the system. For example, the third acquisition unit 15i acquires evaluation data and the like including a data set of sensor values ​​indicating the state of a process in the plant and operation values ​​indicating operations on the process, as operation data transmitted from the operator terminal 20.

[0123] To explain a specific example, the third acquisition unit 15i acquires {Plant: "Plant #1", Evaluation data: "Evaluation data #1"} etc. as data for generating an AI control introduction effect report sent by the operator terminal 20, and outputs it to the third generation unit 15j.

[0124] (2-3-5-10. Third generation section 15j) The third generation unit 15j generates various types of information. The third generation unit 15j may store the generated various types of information in the storage unit 14. The third generation unit 15j may also refer to the various types of information stored in the storage unit 14.

[0125] The third generation unit 15j generates an effect report, i.e., an AI control introduction effect report, showing the effect of introducing the AI ​​control model CM into the system based on the operation data, the determined simulation model SM, and the determined AI control model CM. For example, the third generation unit 15j generates an AI control introduction effect report that displays a graph that allows comparison between the average value and standard deviation of sensor values ​​indicating the state of the process output by inputting, as operation data, operation values ​​indicating the operation of the system process into the determined simulation model SM, and the average value and standard deviation of sensor values ​​output by inputting, as operation data, control data output by the AI ​​control model CM into the determined simulation model SM. The third generation unit 15j also generates an AI control introduction effect report that displays a graph that allows comparison between the time-series changes in sensor values ​​indicating the state of the process output by inputting, as operation data, operation values ​​indicating the operation of the system process into the determined simulation model SM, and the time-series changes in sensor values ​​output by inputting, as operation data, control data output by the AI ​​control model CM into the determined simulation model SM.

[0126] To explain a specific example, the third generation unit 15j inputs {Plant: "Plant #1", evaluation data: "Evaluation data #1"} as a dataset of evaluation data output by the third acquisition unit 15i into the "simulation model SM-1-3", which is the best simulation model SM (Best) stored in the first model memory unit 14b, and generates predicted data "sensor value SM-1-3" for when AI control is not performed. The third generation unit 15j inputs the evaluation data dataset output by the third acquisition unit 15i ({plant: "Plant #1", evaluation data: "Evaluation Data #1") into the "AI-controlled model CM-1-2," which is the best AI-controlled model CM (Best) stored in the second model storage unit 14d, and then inputs the output control data "control data CM-1-2" into the "simulation model SM-1-3," which is the best simulation model SM (Best) stored in the first model storage unit 14b, to generate predicted data "sensor values ​​SM-1-3 / CM-1-2" when AI control is executed. The third generation unit 15j also generates "Effect Report #1" as an AI control introduction effect report that displays the predicted data "sensor values ​​SM-1-3" when AI control is not executed and the predicted data "sensor values ​​SM-1-3 / CM-1-2" when AI control is executed using comparable bar graphs, time-series graphs, etc., and stores the generated effect report in the effect report storage unit 14e.

[0127] (2-3-5-11.Notification section 15k) The notification unit 15k notifies various types of information. Note that the notification unit 15k may refer to various types of information stored in the storage unit 14.

[0128] The notification unit 15k notifies the operator O of the AI ​​control introduction effect report. For example, the notification unit 15k notifies the operator O of the AI ​​control introduction effect report generated by the third generation unit 15j.

[0129] To explain a specific example, the notification unit 15k obtains "Effect Report #1" from the effect report memory unit 14e as an AI control introduction effect report to be notified to the operator O, identifies the operator terminal 20 as the destination of the AI ​​control introduction effect report, sends "Effect Report #1" to the operator terminal 20, and displays bar graphs, time series graphs, etc. included in "Effect Report #1" on the display of the operator terminal 20.

[0130] (2-4. Configuration Example and Processing Example of Operator Terminal 20) 2, a description will be given of an example of the configuration and processing of the operator terminal 20. The operator terminal 20 is a posting device and a viewing device, and includes an input / output unit 21, a transmitting / receiving unit 22, a communication unit 23, and a storage unit 24.

[0131] (2-4-1. Input / output section 21) The input / output unit 21 controls the input of various information to the operator terminal 20. For example, the input / output unit 21 is realized by a mouse, a keyboard, a touch panel, or the like, and accepts input of various information to the operator terminal 20. The input / output unit 21 also controls the display of various information from the operator terminal 20. For example, the input / output unit 21 is realized by a display, or the like, and displays various information stored in the operator terminal 20.

[0132] The input / output unit 21 also displays the AI ​​control introduction effect report sent from the server device 10, which is an information providing device. Details of the display screen of the AI ​​control introduction effect report will be described later in (2-4-5. Specific example 1 of the display screen of the operator terminal 20) and (2-4-6. Specific example 2 of the display screen of the operator terminal 20).

[0133] (2-4-2. Transmitter / receiver 22) The transmitter / receiver 22 transmits various types of information. For example, the transmitter / receiver 22 transmits, as operation data collected from the system, training data including a data set of sensor values ​​indicating the state of the system's process and operation values ​​indicating operations on the process, evaluation data, etc. The transmitter / receiver 22 also transmits, as definition information defining input / output signals, input / output definition information for generating a simulation model, input / output definition information for generating an AI control model, etc. to the server device 10.

[0134] The transmitting / receiving unit 22 receives various types of information. For example, the transmitting / receiving unit 22 receives an AI control introduction effect report transmitted from the server device 10. The transmitting / receiving unit 22 also receives operating data, definition information defining input / output signals, and the like transmitted from a predetermined database.

[0135] (2-4-3. Communications Department 23) The communication unit 23 controls data communication with other devices. For example, the communication unit 23 performs data communication with each communication device via a router, etc. The communication unit 23 can also perform data communication with an operator's terminal (not shown).

[0136] (2-4-4. Storage section 24) The storage unit 24 includes a training data storage unit 24a, an evaluation data storage unit 24b, and a definition information storage unit 24c. Here, the storage unit 24 can be realized by, for example, a semiconductor memory element such as RAM or flash memory, or a storage device such as a hard disk or optical disk. Note that, in the example of Fig. 2, the storage unit 24 is installed inside the operator terminal 20, but it may be installed outside the operator terminal 20, or multiple storage units may be installed.

[0137] (2-4-4-1. Training data storage unit 24a) The training data storage unit 24a stores training data. For example, the training data storage unit 24a stores training data including a data set of sensor values ​​indicating the state of a plant process and operation values ​​indicating operations on the process, as operation data received by the transmitter / receiver 22. Here, an example of data stored in the training data storage unit 24a will be described with reference to FIG. 9. FIG. 9 is a diagram illustrating an example of the training data storage unit 24a of the operator terminal 20 according to the embodiment. In the example of FIG. 9, the training data storage unit 24a has items such as "Plant" and "Training Data."

[0138] "Plant" refers to identification information for identifying a plant, which is a system, such as the plant's identification number or symbol. "Training data" refers to a data set for training the simulation model SM or the AI ​​control model CM, such as a data set of sensor values ​​and operation values ​​collected in the actual environment of the plant.

[0139] That is, FIG. 9 shows an example in which data in which the data set of training data for a plant identified by "Plant #1" is "Training Data #1" is stored in the training data storage unit 24a.

[0140] (2-4-4-2. Evaluation data storage unit 24b) The evaluation data storage unit 24b stores evaluation data. For example, the evaluation data storage unit 24b stores evaluation data including a data set of sensor values ​​indicating the state of a plant process and operation values ​​indicating operations on the process, as operation data received by the transmitter / receiver 22. Here, an example of data stored in the evaluation data storage unit 24b will be described with reference to FIG. 10. FIG. 10 is a diagram showing an example of the evaluation data storage unit 24b of the operator terminal 20 according to the embodiment. In the example of FIG. 10, the evaluation data storage unit 24b has items such as "Plant" and "Evaluation Data."

[0141] "Plant" refers to identification information for identifying a plant, which is a system, such as the plant's identification number or identification symbol. "Evaluation data" refers to a data set for evaluating the simulation model SM or the AI ​​control model CM, such as a data set of sensor values ​​and operation values ​​collected in the actual plant environment, which are different from the training data.

[0142] That is, FIG. 10 shows an example in which, for a plant identified by "Plant #1", data in which the data set of evaluation data is "Evaluation Data #1" is stored in the evaluation data storage unit 24b.

[0143] (2-4-4-3.Definition information storage unit 24c) The definition information storage unit 24c stores definition information. For example, the definition information storage unit 24c stores input / output definition information for generating a simulation model, input / output definition information for generating an AI control model, etc. as definition information received by the transmission / reception unit 22. Here, an example of data stored in the definition information storage unit 24c will be described with reference to FIG. 11. FIG. 11 is a diagram illustrating an example of the definition information storage unit 24c of the operator terminal 20 according to the embodiment. In the example of FIG. 11, the definition information storage unit 24c has items such as "Plant," "Definition Information 1," and "Definition Information 2."

[0144] "Plant" refers to identification information for identifying the plant, which is a system, such as the plant's identification number or symbol. "Definition information 1" is input / output definition information for generating a simulation model that defines the input / output signals of the simulation model SM, and includes information such as an input tag, which is the name or identifier of the process value that changes the state of the process, an output tag, which is the name or identifier of the process value that indicates the state of the process, and the delay time between input and output, which indicates the time it takes for the output data to change when the input data changes. "Definition information 2" is input / output definition information for generating an AI control model that defines the input / output signals of the AI ​​control model CM, and includes information such as an observation tag, which is the name or identifier of the process value observed when the AI ​​control model CM operates, a control tag, which is the name or identifier of the process value operated by the AI ​​control model CM, and the operation period when controlling.

[0145] That is, FIG. 11 shows an example in which, for a plant identified by "Plant #1," data in which the input / output definition information for generating a simulation model is "Definition information #1-1," the input / output definition information for generating an AI control model is "Definition information #1-2," etc. are stored in the definition information storage unit 24c.

[0146] (2-4-5. Specific Example 1 of Display Screen of Operator Terminal 20) Here, an AI control introduction effect report showing a bar graph of sensor values ​​will be described with reference to Fig. 12 as a specific example 1 of a display screen output by the input / output unit 21 of the operator terminal 20. Fig. 12 is a diagram showing a specific example 1 of a display screen of the operator terminal 20 according to the embodiment. Below, "selection tag," "vertical axis," "without AI control," "with AI control," and "target value" will be described.

[0147] (2-4-5-1. Selection tag) As shown in Fig. 12(1), the operator terminal 20 displays a "selection tag" that displays a selectable system process. In the example of Fig. 12(1), the operator terminal 20 displays "LI001" as the plant process selected by the operator O.

[0148] (2-4-5-2. Vertical axis) As shown in Fig. 12(2), the operator terminal 20 displays a graph having a "vertical axis" indicating the numerical value of the sensor value of the system selected by the operator O. In the example of Fig. 12(2), the operator terminal 20 displays the sensor value of "LI001", which is a process of the plant.

[0149] (2-4-5-3. No AI control) As shown in Fig. 12(3), the operator terminal 20 displays a graph showing the average value and standard deviation of the sensor values ​​for "without AI control," which is the case when the AI-controlled model CM is not introduced into the system selected by the operator O. In the example of Fig. 12(3), the operator terminal 20 displays a bar graph corresponding to the average value of the sensor values ​​and error bars corresponding to the standard deviation of the sensor values ​​when AI control of "LI001," a plant process, is not executed.

[0150] (2-4-5-4. AI control) As shown in Fig. 12(4), the operator terminal 20 displays a graph showing the average value and standard deviation of the process values ​​"with AI control," which is the case when the AI ​​control model CM is introduced into the system selected by the operator O. In the example of Fig. 12(4), the operator terminal 20 displays a bar graph corresponding to the average value of the sensor values ​​and error bars corresponding to the standard deviation of the sensor values ​​when AI control is executed for "LI001," which is a plant process.

[0151] (2-4-5-5. Target value) As shown in Figure 12(5), the operator terminal 20 displays a graph having a "target value" that indicates the target numerical value of the sensor value of the system selected by the operator O. In the example of Figure 12(5), the operator terminal 20 displays the target numerical value of the sensor value of "LI001," a plant process, with a dashed line, making it visually easy to understand that the sensor value when the AI-controlled model CM is introduced into the system will be closer to the target numerical value than the sensor value when the AI-controlled model CM is not introduced into the system.

[0152] (2-4-5-6. Other) The operator terminal 20 may display bar graphs showing multiple types of sensor values ​​side by side. The operator terminal 20 may also display a legend that associates the type or name of the bar graph with the shape or color of the bar graph side by side with the bar graph.

[0153] (2-4-6. Specific Example 2 of Display Screen of Operator Terminal 20) Here, an AI control introduction effect report showing a time series graph of sensor values ​​will be described with reference to Fig. 13 as a specific example 1 of a display screen output by the input / output unit 21 of the operator terminal 20. Fig. 13 is a diagram showing a specific example 2 of a display screen of the operator terminal 20 according to the embodiment. Below, "selection tag," "vertical axis," "horizontal axis," "no AI control," and "with AI control" will be described.

[0154] (2-4-6-1. Selection tag) As shown in Fig. 13(1), the operator terminal 20 displays a "selection tag" that displays the processes of the system in a selectable manner. In the example of Fig. 13(1), the operator terminal 20 displays "LI001," "LI002," and "LI003" as the plant processes selected by the operator O.

[0155] (2-4-6-2. Vertical axis) As shown in Fig. 13(2), the operator terminal 20 displays a graph having a "vertical axis" indicating the numerical values ​​of the sensor values ​​of the system selected by the operator O. In the example of Fig. 13(2), the operator terminal 20 displays the sensor values ​​of "LI001," "LI002," and "LI003," which are processes of the plant.

[0156] (2-4-6-3. Horizontal axis) As shown in Fig. 13(3), the operator terminal 20 displays a graph having a "horizontal axis" showing the time series of the system selected by the operator O. In the example of Fig. 13(3), the operator terminal 20 displays the times of "LI001," "LI002," and "LI003," which are processes of the plant.

[0157] (2-4-6-4. No AI control) As shown in Figure 13(4), the operator terminal 20 displays a graph showing the time series changes in the sensor values ​​for "without AI control," which is the case when the AI ​​control model CM is not introduced into the system selected by the operator O. In the example of Figure 13(4), the operator terminal 20 displays, with dashed lines, time series graphs showing the time series changes in the sensor values ​​for the case when AI control is not executed for the plant processes "LI001," "LI002," and "LI003."

[0158] (2-4-6-5. AI control available) As shown in Fig. 13(5), the operator terminal 20 displays a graph showing the time series changes in the sensor values ​​when "with AI control," which is the case when the AI ​​control model CM is introduced into the system selected by the operator O. In the example of Fig. 13(5), the operator terminal 20 displays, with solid lines, time series graphs showing the time series changes in the sensor values ​​when AI control is executed for the plant processes "LI001," "LI002," and "LI003."

[0159] (2-4-6-6. Other) The operator terminal 20 may display multiple types of sensor values ​​superimposed on a time series graph. Furthermore, the operator terminal 20 may display a legend associating the type or name of the time series graph with the line type or line color superimposed on the time series graph. Furthermore, the operator terminal 20 may display a "target value" indicating a target numerical value for the sensor value of the system selected by the operator O, and a "target time" indicating a target time or target period when the sensor value will be equal to or less than the target numerical value or equal to or greater than the target numerical value, superimposed on the time series graph.

[0160] 3. Flow of each process in the effect report providing system 100 14 to 17, the process flow of the effect report providing system 100 according to the embodiment will be described. Below, the overall process flow of the effect report providing system 100 will be described, and then each process, namely, the simulation model generation management process, the AI ​​control model generation management process, and the AI ​​control introduction effect report generation management process, will be described.

[0161] (3-1. Overall processing of the effect report providing system 100) The overall processing flow of the effect report providing system 100 according to the embodiment will be described with reference to Figure 14. Figure 14 is a flowchart showing an example of the overall processing flow of the effect report providing system 100 according to the embodiment. Note that the processing of steps S101 to S103 below may be executed in a different order. Also, some of the processing of steps S101 to S103 below may be omitted.

[0162] (3-1-1. Simulation model generation management process) First, the server device 10 executes a simulation model generation management process (step S101). For example, the server device 10 executes the processes of steps S201 to S204 described below to manage the generation of a simulation model SM that predicts a process value of a plant.

[0163] (3-1-2. AI control model generation management process) Second, the server device 10 executes an AI control model generation management process (step S102). For example, the server device 10 executes the processes of steps S301 to S304 described below to manage the generation of an AI control model CM that controls a process value of a plant.

[0164] (3-1-3. AI control introduction effect report generation and management process) Third, the server device 10 executes an AI control introduction effect report generation management process (step S103), and ends the process. For example, the server device 10 executes the processes of steps S401 to S403 described below to manage the generation of an AI control introduction effect report that verifies the effect of introducing the AI ​​control model CM.

[0165] (3-2. Simulation model generation management process) The flow of the simulation model generation management process of the effect report providing system 100 according to the embodiment will be described with reference to Fig. 15. Fig. 15 is a flowchart showing an example of the flow of the simulation model generation management process of the effect report providing system 100 according to the embodiment. Note that the processes of steps S201 to S204 below can also be executed in a different order. Also, some of the processes of steps S201 to S204 below may be omitted.

[0166] (3-2-1. Data input process for generating simulation model) First, the server device 10 executes a data input process for generating a simulation model (step S201). For example, an operator O operates the operator terminal 20 to transmit to the server device 10 training data, evaluation data, and simulation model input / output definition information to be input to the simulation model SM stored in the operator terminal 20.

[0167] (3-2-2. Simulation model generation setting parameter creation process) Second, the server device 10 executes a process of creating setting parameters for generating a simulation model (step S202). For example, the server device 10 creates N patterns of setting parameters for generating a simulation model by grid research.

[0168] (3-2-3. Simulation model generation process) Third, the server device 10 executes a simulation model generation process (step S203). For example, the server device 10 inputs training data, setting parameters for generating a simulation model, and input / output definition information for generating a simulation model into the simulation model SM, and generates N simulation models SM.

[0169] (3-2-4. Simulation model evaluation process) Fourth, the server device 10 executes a simulation model evaluation process (step S204), and ends the simulation model generation management process. For example, the server device 10 inputs evaluation data into the simulation model SM, calculates a simulation model evaluation value, and determines the best simulation model SM (Best) with the best performance among the N simulation models SM.

[0170] (3-3. AI control model generation management process) The flow of the AI ​​control model generation and management process of the effect report providing system 100 according to the embodiment will be described with reference to Figure 16. Figure 16 is a flowchart showing an example of the flow of the AI ​​control model generation and management process of the effect report providing system 100 according to the embodiment. Note that the processes of steps S301 to S304 below can also be executed in a different order. Also, some of the processes of steps S301 to S304 below may be omitted.

[0171] (3-3-1. Data input processing for generating AI control models) First, the server device 10 executes a data input process for generating an AI-controlled model (step S301). For example, the operator O operates the operator terminal 20 to transmit to the server device 10 the training data, evaluation data, and AI-controlled model input / output definition information to be input to the AI-controlled model CM stored in the operator terminal 20.

[0172] (3-3-2. AI control model generation setting parameter creation process) Second, the server device 10 executes a process for creating setting parameters for generating an AI control model (step S302). The server device 10 creates M-pattern setting parameters for generating an AI control model by grid research.

[0173] (3-3-3. AI control model generation process) Third, the server device 10 executes an AI-controlled model generation process (step S303). For example, in an execution environment of the best simulation model SM (Best), the server device 10 inputs training data, setting parameters for generating an AI-controlled model, and input / output definition information for generating an AI-controlled model into the AI-controlled model CM, and generates M AI-controlled models CM.

[0174] (3-3-4. AI control model evaluation process) Fourth, server device 10 executes an AI control model evaluation process (step S304), and ends the AI ​​control model generation management process. For example, server device 10 inputs the evaluation data into the AI ​​control model CM, calculates the AI ​​control model evaluation value, and determines the best AI control model CM (Best) with the best performance among the M AI control models CM.

[0175] (3-4. AI control introduction effect report generation and management process) The flow of the AI ​​control introduction effect report generation and management process of the effect report providing system 100 according to the embodiment will be described with reference to Figure 17. Figure 17 is a flowchart showing an example of the flow of the AI ​​control introduction effect report generation and management process of the effect report providing system 100 according to the embodiment. Note that the processes of steps S401 to S403 below can also be executed in a different order. Also, some of the processes of steps S401 to S403 below may be omitted.

[0176] (3-4-1. Data input process for generating a report on the effects of introducing AI control) First, the server device 10 executes a data input process for generating an AI control introduction effect report (step S401). For example, the operator O operates the operator terminal 20 to transmit evaluation data to be used for generating an AI control introduction effect report stored in the operator terminal 20 to the server device 10.

[0177] (3-4-2. AI control introduction effect report generation process) Second, the server device 10 executes a process for generating an AI control introduction effect report (step S402). For example, the server device 10 generates sensor values ​​when AI control is not performed by inputting the evaluation data into the best simulation model SM (Best), generates sensor values ​​when AI control is performed by inputting the control data output by the best AI control model CM (Best) into the best simulation model SM (Best), and generates an AI control introduction effect report that displays bar graphs and time series graphs for comparison.

[0178] (3-4-3. AI Control Introduction Effect Report Notification Processing) Third, the server device 10 executes an AI control introduction effect report notification process (step S403), and ends the process. For example, the server device 10 transmits the generated AI control introduction effect report to the operator terminal 20, and causes the monitor of the operator terminal 20 to display a bar graph or a time series graph.

[0179] 4. Effects of the embodiment Effects of the embodiment will be described. Effects 1 to 8 corresponding to the processing according to the embodiment will be described below.

[0180] (4-1. Effect 1) First, in the process according to the embodiment described above, the server device 10 generates N simulation models SM based on driving data collected from the system, each of which outputs predicted data for the system's driving status in response to input driving data; determines the best simulation model SM (Best) with the highest simulation model evaluation value for evaluating the simulation models SM based on the driving data; generates M AI-controlled models CM based on the driving data and the best simulation model SM (Best) and outputs control data for controlling the system in response to input driving data; and determines the best AI-controlled model CM (Best) with the highest AI-controlled model evaluation value for evaluating the AI-controlled models CM based on the driving data. Therefore, this process makes it easy to verify the effectiveness of AI control introduced into the system.

[0181] (4-2. Effect 2) Second, in the processing according to the above-described embodiment, the server device 10 generates an AI control introduction effect report that indicates the effect of introducing the best AI control model CM (Best) into the system based on the driving data, the best simulation model SM (Best), and the best AI control model CM (Best). Therefore, in this processing, the effect of the AI ​​control introduced into the system can be easily verified by creating an AI control introduction effect report.

[0182] (4-3. Effect 3) Third, in the processing according to the above-described embodiment, the server device 10 uses, as the training data, a data set of sensor values ​​indicating the state of the system's process and operation values ​​indicating operations on the process, as the operating data, uses N patterns of hyperparameters generated by grid research as setting parameters for generating N patterns of simulation models, and further uses input / output definition information for generating simulation models that defines the input / output signals of the simulation models SM, thereby generating N simulation models SM. Therefore, in this processing, the operator O can easily train the simulation models SM, and the effects of AI control introduced into the system can be easily verified.

[0183] (4-4. Effect 4) Fourth, in the processing according to the above-described embodiment, the server device 10 uses, as the operating data, a data set of sensor values ​​indicating the state of the system's process and operation values ​​indicating operations on the process as evaluation data, calculates the agreement rate between the sensor values ​​collected from the system and the sensor values ​​output as prediction data as the simulation model evaluation value, and determines the best simulation model SM (Best) among the N simulation models SM generated that has the highest calculated agreement rate. Therefore, in this processing, the operator O can easily evaluate the simulation models SM, and the effect of AI control introduced into the system can be easily verified.

[0184] (4-5. Effect 5) Fifth, in the processing according to the above-described embodiment, the server device 10 uses, as the training data, a data set of sensor values ​​indicating the state of the system's process and operation values ​​indicating operations on the process as operating data, uses M-pattern hyperparameters generated by grid research as setting parameters for generating the M-pattern AI control model, and further uses input / output definition information for generating the AI ​​control model that defines the input / output signals of the AI ​​control model CM, and generates M AI control models CM by using the best simulation model SM (Best). Therefore, in this processing, the operator O can easily train the AI ​​control model CM, making it easy to verify the effects of AI control introduced into the system.

[0185] (4-6. Effect 6) Sixth, in the processing according to the above-described embodiment, the server device 10 uses, as the operating data, a data set of sensor values ​​indicating the state of the system's process and operation values ​​indicating operations on the process as evaluation data, calculates the average value of reward values ​​when controlling the system using a reward function as the AI ​​control model evaluation value, and determines the best AI control model CM (Best) from among the M generated AI control models CM that has the largest average value of the calculated reward values. Therefore, in this processing, the operator O can easily evaluate the AI ​​control model CM, making it easy to verify the effectiveness of the AI ​​control introduced into the system.

[0186] (4-7. Effect 7) Seventh, in the processing according to the above-described embodiment, the server device 10 generates an AI control introduction effect report that displays a graph that allows a comparison between the average value and standard deviation of sensor values ​​indicating the state of the process output by inputting operation values ​​indicating the operation of the system process as operating data into the best simulation model SM (Best) and the average value and standard deviation of sensor values ​​output by inputting control data output by the AI-controlled model CM into the best simulation model SM (Best). Therefore, in this processing, the operator O can easily verify the effect of introducing the AI-controlled model CM using statistical data, and therefore the effect of AI control introduced into the system can be easily verified.

[0187] (4-8. Effect 8) Eighth, in the process according to the embodiment described above, the server device 10 generates an AI control introduction effect report that displays a graph that allows comparison of time-series changes in sensor values ​​indicating the state of the process output by inputting operation values ​​indicating the operation of the system process as operating data into the best simulation model SM (Best) with time-series changes in the sensor values ​​output by inputting control data output by the AI-controlled model CM into the best simulation model SM (Best). Therefore, in this process, the operator O can easily verify the effect of introducing the AI-controlled model CM using time-series data, and therefore the effect of AI control introduced into the system can be easily verified.

[0188] (4-9. Effect 9) Ninth, in the process according to the above-described embodiment, the system is a plant. Therefore, in this process, the operator O can easily verify the effect of introducing the AI-controlled model CM into the plant, and therefore the effect of the AI ​​control introduced into the system can be easily verified.

[0189] 5. Application Examples of the Embodiments Application Examples of the Embodiments will be Described below. Application Examples 1 and 2 of the embodiment will be described below.

[0190] (5-1. Application Example 1) In the effectiveness report providing system 100, the server device 10 can dynamically adjust process control. For example, the server device 10 can generate an AI control model CM that adapts to fluctuations in the production process by collecting operational data from a plant site in real time and continuously training the AI ​​control model CM. In other words, the server device 10 generates an AI control model CM that can maintain optimal control when there is a sudden fluctuation in demand or a change in material quality, allowing the introduction effects to be immediately confirmed.

[0191] (5-2. Application Example 2) In the effect report providing system 100, the server device 10 can generate an AI control introduction effect report for the KPI (Key Performance Indicator) values ​​of the entire plant. For example, the server device 10 can generate an effective AI control model CM for optimizing the performance of the entire plant by utilizing the KPI values ​​(e.g., production efficiency, energy usage efficiency) of the entire plant as training data and evaluation data. In this case, the server device 10 can visualize the changes in KPIs and the degree of improvement before and after the introduction of AI control in the generated AI control introduction effect report.

[0192] [6. System] The information including the processing procedures, control procedures, specific names, various data and parameters shown in the above documents and drawings can be changed arbitrarily unless otherwise specified.

[0193] Furthermore, the components of each device shown in the figure are functional concepts and do not necessarily have to be physically configured as shown. In other words, the specific form of distribution and integration of each device is not limited to that shown. In other words, all or part of them can be functionally or physically distributed and integrated in any unit depending on various loads, usage conditions, etc.

[0194] Furthermore, all or any part of the processing functions performed by each device may be realized by a CPU and a program analyzed and executed by the CPU, or may be realized as hardware using wired logic.

[0195] [7. Hardware] Next, an example of the hardware configuration of the server device 10, which is an information providing device, will be described. Note that other devices may also have a similar hardware configuration. FIG. 18 is a diagram illustrating an example of the hardware configuration according to an embodiment. As shown in FIG. 18, the server device 10 includes a communication device 10a, an HDD (Hard Disk Drive) 10b, a memory 10c, and a processor 10d. The components shown in FIG. 18 are connected to each other via a bus or the like.

[0196] The communication device 10a is a network interface card or the like, and communicates with other servers. The HDD 10b stores programs and databases that operate the functions shown in FIG.

[0197] The processor 10d reads out from the HDD 10b or the like a program that executes the same processes as the respective processing units shown in Fig. 2 and loads the program into the memory 10c, thereby operating a process that executes each function described in Fig. 2 or the like. For example, this process executes the same functions as the respective processing units of the server device 10. Specifically, the processor 10d reads out from the HDD 10b or the like a program that has the same functions as the first acquisition unit 15a, the first adjustment unit 15b, the first generation unit 15c, the first evaluation unit 15d, the second acquisition unit 15e, the second adjustment unit 15f, the second generation unit 15g, the second evaluation unit 15h, the third acquisition unit 15i, the third generation unit 15j, the notification unit 15k, and the like. Then, processor 10d executes a process that performs processing similar to that of first acquisition unit 15a, first adjustment unit 15b, first generation unit 15c, first evaluation unit 15d, second acquisition unit 15e, second adjustment unit 15f, second generation unit 15g, second evaluation unit 15h, third acquisition unit 15i, third generation unit 15j, notification unit 15k, etc.

[0198] In this way, the server device 10 operates as a device that executes various processing methods by reading and executing a program. The server device 10 can also realize functions similar to those of the above-described embodiment by reading the program from a recording medium using a media reader and executing the read program. Note that the program in these other embodiments is not limited to being executed by the server device 10. For example, the present invention can also be applied in the same way to cases where another computer or server executes the program, or where these execute the program in cooperation with each other.

[0199] This program can be distributed via a network such as the Internet. In addition, this program can be recorded on a computer-readable recording medium such as a hard disk, a flexible disk (FD), a CD-ROM, a magneto-optical disk (MO), or a digital versatile disk (DVD), and can be executed by being read from the recording medium by a computer.

[0200] [8. Other] Some examples of combinations of the disclosed technical features are set out below.

[0201] (1) An information providing device comprising: a first generation unit that generates, based on driving data collected from a system, a plurality of first models that output predictive data of the driving status of the system in response to input of the driving data; a first evaluation unit that determines, based on the driving data, the first model with the largest first evaluation value for evaluating the first models; a second generation unit that generates, based on the driving data and the determined first models, a plurality of second models that output control data for controlling the system in response to input of the driving data; and a second evaluation unit that determines, based on the driving data, the second model with the largest second evaluation value for evaluating the second models.

[0202] (2) The information providing device described in (1) further includes a third generation unit that generates an effect report indicating the effect when the determined second model is introduced into the system based on the driving data, the determined first model, and the determined second model.

[0203] (3) The information providing device described in (1) or (2), wherein the first generation unit generates the plurality of first models by using, as the operating data, a data set of sensor values ​​indicating the state of the system's process and operation values ​​indicating operations on the process as training data, using, as the plurality of first setting values, a plurality of hyperparameters generated by grid research, and further using first definition information defining the input / output signals of the first models.

[0204] (4) The information providing device according to any one of (1) to (3), wherein the first evaluation unit uses a data set of sensor values ​​indicating the state of the system's process and operation values ​​indicating operations on the process as evaluation data as the operating data, calculates the agreement rate between the sensor values ​​collected from the system and the sensor values ​​output as the prediction data as the first evaluation value, and determines the first model among the generated first models that has the highest calculated agreement rate.

[0205] (5) An information providing device described in any one of (1) to (4), wherein the second generation unit uses a data set of sensor values ​​indicating the state of the system's process and operation values ​​indicating operations on the process as training data as the operating data, uses a plurality of hyperparameters generated by grid research as the plurality of second setting values, and further uses second definition information defining input / output signals of the second model, and generates the plurality of second models by using the determined first model.

[0206] (6) The information providing device described in any one of (1) to (5), wherein the second evaluation unit uses a data set of sensor values ​​indicating the state of the system's process and operation values ​​indicating operations on the process as evaluation data as the operating data, calculates an average value of reward values ​​when controlling the system using a reward function as the second evaluation value, and determines the second model from the multiple generated second models that has the largest average value of the calculated reward values.

[0207] (7) The information providing device described in any one of (2) to (6), wherein the third generation unit generates the effect report displaying a graph that allows comparison between the average value and standard deviation of the sensor value indicating the state of the process output by inputting an operating value indicating the operation of the system's process as the operating data into the determined first model, and the average value and standard deviation of the sensor value output by inputting the control data output by the second model into the determined first model.

[0208] (8) The information providing device described in any one of (2) to (7), wherein the third generation unit generates the effect report that displays a graph that allows comparison between the time series change of the sensor value indicating the state of the process output by inputting an operating value indicating the operation of the system's process as the operating data into the determined first model and the time series change of the sensor value output by inputting the control data output by the second model into the determined first model.

[0209] (9) The information providing device according to any one of (1) to (8), wherein the system is a plant.

[0210] (10) An information provision method in which a computer executes a process in which, based on driving data collected from a system, a plurality of first models are generated, each of which outputs predictive data on the driving status of the system in response to input of the driving data; based on the driving data, the first model having the largest first evaluation value for evaluating the first models is determined; based on the driving data and the determined first models, a plurality of second models are generated, each of which outputs control data related to control of the system in response to input of the driving data; and based on the driving data, the second model having the largest second evaluation value for evaluating the second models is determined.

[0211] (11) An information provision program that causes a computer to execute the following processes: based on driving data collected from a system, generate a plurality of first models that output predictive data of the driving status of the system in response to input of the driving data; based on the driving data, determine the first model with the largest first evaluation value for evaluating the first models; based on the driving data and the determined first models, generate a plurality of second models that output control data for controlling the system in response to input of the driving data; and based on the driving data, determine the second model with the largest second evaluation value for evaluating the second models. [Explanation of symbols]

[0212] 10 Server device 10a Communication equipment 10b HDD 10c memory 10d processor 11 Input section 12 Output section 13 Communications Department 14 Storage section 14a First setting value storage unit 14b First model storage unit 14c Second setting value storage section 14d Second model memory section 14e Effect Report Memory Section 15 Control Unit 15a 1st acquisition part 15b 1st adjustment section 15c 1st generation part 15d First Evaluation Section 15e 2nd acquisition part 15f 2nd adjustment section 15g 2nd generation part 15h 2nd Evaluation Section 15i 3rd acquisition part 15j 3rd generation part 15k Notification Department 20 Operator terminal 21 Input / output section 22 Transmitter / Receiver 23 Communications Department 24 Memory section 24a Training data storage unit 24b Evaluation data storage unit 24c Definition information storage section 100 Effectiveness Report Providing System

Claims

1. a first generation unit that generates a plurality of first models based on driving data collected from the system, the first models outputting prediction data of a driving state of the system in response to input of the driving data; a first evaluation unit that determines the first model with the largest first evaluation value for evaluating the first model based on the driving data; a second generation unit that generates a plurality of second models based on the operating data and the determined first model, and outputs control data related to control of the system in response to input of the operating data; a second evaluation unit that determines the second model with the largest second evaluation value for evaluating the second model based on the driving data; An information providing device comprising:

2. a third generation unit that generates an effect report indicating an effect when the second model is introduced into the system based on the operating data, the determined first model, and the determined second model; The information providing device according to claim 1 , further comprising:

3. The first generation unit As the operation data, a data set of sensor values ​​indicating a state of a process of the system and operation values ​​indicating an operation on the process is used as training data; Using multiple hyperparameters generated by grid research, generating the plurality of first models by further using first definition information that defines input and output signals of the first models; The information providing device according to claim 1 .

4. The first evaluation unit As the operational data, a data set of sensor values ​​indicating the state of the process of the system and operation values ​​indicating operations on the process is used as evaluation data; calculating, as the first evaluation value, a coincidence rate between the sensor value collected from the system and the sensor value output as the predicted data; determining the first model that maximizes the calculated matching rate from among the plurality of generated first models; The information providing device according to claim 1 .

5. The second generation unit As the operation data, a data set of sensor values ​​indicating a state of a process of the system and operation values ​​indicating an operation on the process is used as training data; Using multiple hyperparameters generated by grid research, further using second definition information that defines input / output signals of the second model; generating the plurality of second models by using the determined first model; The information providing device according to claim 1 .

6. The second evaluation unit As the operational data, a data set of sensor values ​​indicating the state of the process of the system and operation values ​​indicating operations on the process is used as evaluation data; calculating, as the second evaluation value, an average value of reward values ​​when the system is controlled using a reward function; determining the second model that maximizes the average value of the calculated reward values ​​from among the plurality of generated second models; The information providing device according to claim 1 .

7. The third generation unit generating the effect report that displays a graph that allows comparison between the average value and standard deviation of the sensor values ​​that indicate the state of the process and are output by inputting, as the operating data, an operation value that indicates the operation of the system's process into the determined first model, and the average value and standard deviation of the sensor values ​​that are output by inputting, as the control data, output by the second model into the determined first model; The information providing device according to claim 2 .

8. The third generation unit generating the effect report that displays a graph that allows comparison between a time series change in the sensor value that indicates the state of the process and is output by inputting an operation value that indicates an operation of the system process as the operating data into the determined first model, and a time series change in the sensor value that is output by inputting the control data output by the second model into the determined first model; The information providing device according to claim 2 .

9. The system is a plant. The information providing device according to any one of claims 1 to 8.

10. The computer generating a plurality of first models based on the driving data collected from the system, the first models outputting prediction data of the driving status of the system in response to input of the driving data; determining the first model having the largest first evaluation value for evaluating the first model based on the driving data; generating a plurality of second models based on the operational data and the determined first model, the second models outputting control data relating to control of the system in response to input of the operational data; determining the second model with the largest second evaluation value for evaluating the second model based on the driving data; How information is provided to perform the process.

11. On the computer, generating a plurality of first models based on the driving data collected from the system, the first models outputting prediction data of the driving status of the system in response to input of the driving data; determining the first model having the largest first evaluation value for evaluating the first model based on the driving data; generating a plurality of second models based on the operational data and the determined first model, the second models outputting control data relating to control of the system in response to input of the operational data; determining the second model with the largest second evaluation value for evaluating the second model based on the driving data; An information providing program that executes processing.

Images