Device, method, and program

Abstract

To provide a device, a method, and a program.SOLUTION: A device comprises: an acquisition unit for acquiring status data of a facility; a selection unit for selecting a recommended learning model recommended to be used for controlling a control target in order to adapt the KPI of the facility to a reference condition on the basis of the status data acquired by the acquisition unit, from among multiple learning models for outputting control parameters to be applied to the control target of the facility, according to the supply of the status data of the facility; and an output unit for outputting the identification information of the recommended learning model.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

[0002] Patent Document 1 and other documents state that the system "includes a candidate model storage unit that stores a plurality of candidate models that are generated by reinforcement learning and that can output behavior according to the state of the equipment, and a model selection unit that selects a target model for controlling the controlled object from among the plurality of candidate models based on the plurality of indicators" (Claim 1 of Patent Document 1). [Prior art document] [Patent documents] [Patent Document 1] JP 2023-173459 A [Patent Document 2] JP 2021-174259 A [Patent Document 3] JP 2014-2579 A [Patent Document 4] JP 2022-117730 A [Patent Document 5] JP 2021-72049 A Summary of the Invention

[0003] In a first aspect of the present invention, an apparatus is provided that includes an acquisition unit that acquires status data related to equipment, a selection unit that selects, based on the status data acquired by the acquisition unit, a recommended learning model from a plurality of learning models that output control parameters to be applied to a controlled object in the equipment in response to the status data related to the equipment being supplied, which is recommended to be used to control the controlled object in order to adapt a KPI related to the equipment to a reference condition, and an output unit that outputs identification information of the recommended learning model.

[0004] In the above device, the selection unit may select the recommended learning model based on a history of the status data acquired by the acquisition unit.

[0005] The above-mentioned device in which the recommended learning model is selected based on the history of the status data further includes a prediction unit that predicts status data for each of multiple future points in time based on the history of the status data acquired by the acquisition unit, a determination unit that determines the start timing for starting use of the recommended learning model based on the predicted status data for each point in time, and a generation unit that generates a report including identification information of the recommended learning model and the start timing, and the output unit may output the report.

[0006] In the above device, the prediction unit further predicts status data for each of multiple time points after the start timing when use of the recommended learning model starts at the start timing, the selection unit further selects another recommended learning model from the multiple learning models that is different from the recommended learning model and is recommended for use in controlling the control object in order to make the KPI conform to the reference conditions in the state indicated by the status data predicted for a time point after the start timing, the determination unit further determines another start timing at which use of the other recommended learning model should start based on the status data predicted for each time point after the start timing, and the generation unit may generate a report further including identification information of the other recommended learning model and the other start timing.

[0007] In any of the above devices in which the recommended learning model is selected based on the history of the status data, the selection unit may select the recommended learning model further based on the history of the past status data, the history of the KPI, and the history of the learning model used to control the control object.

[0008] In the above device, the selection unit may have a model that performs a learning process using learning data including status data at each past point in time, the KPI, and identification information of the learning model used to control the control object, and outputs identification information of the recommended learning model in response to being supplied with the history of the status data acquired by the acquisition unit and the identification information of the learning model used to control the control object.

[0009] In the above device, the acquisition unit further acquires control parameters applied to the control object, and the model undergoes a learning process using learning data that further includes control parameters at each past point in time, and outputs identification information of the recommended learning model in response to being supplied with the history of the state data and the history of the control parameters acquired by the acquisition unit, and identification information of the learning model used to control the control object.

[0010] The above-mentioned device, in which the selection unit selects the recommended learning model further based on the past history of the status data, the history of the KPI, and the history of the learning model used to control the controlled object, further includes a memory unit that stores history data that corresponds, each time the learning model used to control the controlled object is switched, the identification information of the learning model that started to be used, the history of the status data before the switching, and the value of the KPI after the switching, and the selection unit selects one history of the status data included in the history data based on the similarity between the history of the status data acquired by the acquisition unit and each history of the status data included in the history data, and selects the learning model associated with the one history of the status data as the recommended learning model, provided that the value of the KPI associated with the one history of the status data in the history data satisfies the standard condition.

[0011] Any of the above devices may further include a generation unit that generates a report including identification information of the recommended learning model and the reason why the selection unit selected the recommended learning model, and the output unit may output the report.

[0012] In the above device, the selection unit may select multiple recommended learning models, and the generation unit may generate the report for each recommended learning model, including identification information of the recommended learning model and the reason for selecting the recommended learning model.

[0013] In any of the above devices, which further includes a generation unit that generates a report including the reason why the selection unit selected the recommended learning model, the reason may include information indicating the predicted change in the KPI when the recommended learning model is used.

[0014] In any of the above devices, the selection unit may include a setting unit that switchably sets any one of a plurality of KPIs related to the equipment as the KPI to be used in selecting the recommended learning model.

[0015] In the above device, the setting unit may set a KPI designated by a user from among the plurality of KPIs as the KPI to be used for selecting the recommended learning model.

[0016] In any of the above devices in which the selection unit has the setting unit, the setting unit may set a KPI among the multiple KPIs that can be improved from the current state as the KPI to be used for selecting the recommended learning model.

[0017] Any of the above devices may further include a switching unit that switches the learning model used to control the control object to the recommended learning model.

[0018] In a second aspect of the present invention, a method is provided comprising: an acquisition step of acquiring status data related to equipment; a selection step of selecting, based on the status data acquired in the acquisition step, a recommended learning model from among a plurality of learning models that output control parameters to be applied to the controlled object in the equipment in response to the status data related to the equipment being supplied, which is recommended to be used for controlling the controlled object in order to adapt the KPI related to the equipment to a reference condition; and an output step of outputting identification information of the recommended learning model.

[0019] In a third aspect of the present invention, a program is provided that, when executed by a computer, causes the computer to function as an acquisition unit that acquires status data related to equipment, a selection unit that, in response to the supply of status data related to the equipment, selects, based on the status data acquired by the acquisition unit, a recommended learning model from among a plurality of learning models that output control parameters to be applied to the controlled object in the equipment, which is recommended to be used to control the controlled object in order to adapt the KPI related to the equipment to a reference condition, and an output unit that outputs identification information of the recommended learning model.

[0020] The above summary of the invention does not list all of the necessary features of the present invention, and subcombinations of these features may also constitute inventions. [Brief explanation of the drawings]

[0021] [Figure 1] 1 shows a system 1 according to the present embodiment. [Figure 2] The operation of the control assistance device 4 is shown. [Figure 3] 4 shows changes in the status data acquired by the acquisition unit 401. [Figure 4] This shows the difference in profit depending on the timing of switching the target learning model 420(T). [Figure 5] The graph shows the trends in "profits from AI selection" and "profits from user selection." [Figure 6]12 illustrates an example computer 1200 in which aspects of the present invention may be embodied, in whole or in part. DETAILED DESCRIPTION OF THE INVENTION

[0022] The present invention will be described below through embodiments of the invention, but the following embodiments do not limit the scope of the invention according to the claims. Furthermore, not all of the combinations of features described in the embodiments are necessarily essential to the solution of the invention.

[0023] (Configuration of System 1) 1 shows a system 1 according to this embodiment. The system 1 includes a facility 2 and a control assistance device 4.

[0024] ((Equipment 2)) The facility 2 is equipped with a plurality of devices 20. For example, the facility 2 may be a plant, or a composite device that combines a plurality of devices 20. Examples of plants include industrial plants such as chemical and bio plants, plants that manage and control wellheads and surrounding areas of gas fields and oil fields, plants that manage and control power generation such as hydroelectric, thermal, and nuclear power, plants that manage and control environmental power generation such as solar and wind power, and plants that manage and control water supply and sewage systems, dams, etc. In this embodiment, as an example, the facility 2 has one or more devices 20 and one or more sensors 21.

[0025] (((equipment 20))) Each device 20 is an instrument, machine, or apparatus, and may be, for example, an actuator, i.e., an operating element, such as a valve, pump, heater, fan, motor, or switch that controls at least one physical quantity, such as pressure, temperature, pH, speed, or flow rate, in the process of the facility 2, and performs a given operation according to the operating quantity. However, the device 20 is not limited to this, and may also be, for example, a controller that controls the operating element.

[0026] In this embodiment, as an example, the facility 2 is equipped with a plurality of devices 20. The devices 20 may be of different types, or at least some of the devices 20 may be of the same type.

[0027] Each device 20 may be controlled externally via a network (not shown) in a wired or wireless manner, or may be controlled manually. Note that the term "control" used in this specification may be broadly interpreted to include not only direct control of an operating element, but also indirect control of an operating element via a controller.

[0028] At least some of the devices 20 may be control targets (also referred to as control target devices 20(T)) controlled by the control assistance device 4. When the system 1 is equipped with multiple control target devices 20(T), these multiple control target devices 20(T) may have a relationship in which they are controlled in conjunction with each other (for example, a master-slave relationship, a relationship in which they are not controlled independently). Furthermore, the control target devices 20(T) may be devices 20 of the same type or different types.

[0029] (((Sensor 21))) Each sensor 21 measures a state related to the equipment 2. The state related to the equipment 2 may be the state of the equipment 2 itself or the state of a product of the equipment 2. The state of the equipment 2 itself may be, for example, the internal pressure, flow rate, temperature, pH, speed, power consumption, or concentration of the equipment 2, or may be a state indicating a process variable, for example. The state related to the equipment 2 may be the temperature, humidity, sunlight, wind direction, wind volume, or precipitation of the external environment of the equipment 2, the amount of energy or raw material consumed by the equipment 2, the amount of greenhouse gas emissions, the yield, or the production cost. The state of the product may be a predetermined quality (for example, purity, concentration, composition, viscosity, or color) or the production volume. Each sensor 21 may supply state data obtained by measurement to the control assistance device 4.

[0030] ((Control Support Device 4)) The control assistance device 4 assists in the control of the facility 2. The control assistance device 4 may be one or more computers, and may be configured with a PC or the like. The control assistance device 4 has an acquisition unit 401, an AI control unit 402, an input unit 403, a prediction unit 404, a selection unit 405, a determination unit 406, a generation unit 407, an output unit 408, a switching unit 409, and a storage unit 410.

[0031] (((Acquisition unit 401))) The acquisition unit 401 acquires various information related to the facility 2. For example, the acquisition unit 401 may acquire status data related to the facility 2. The acquisition unit 401 may acquire the status data for each reference interval. The acquisition unit 401 may acquire the status data from each sensor 21.

[0032] The acquiring unit 401 may further acquire one or more KPIs (Key Performance Indicators) related to the equipment 2. The acquiring unit 401 may acquire the KPIs for each reference interval. Here, the KPI may be a value indicated by any of the status data related to the equipment 2, or may be a value calculated using at least one of the status data. For example, the KPI may be the production volume of a product within a reference period, the yield of the product, the quality, the profit from production, or the amount of carbon dioxide emissions. When the KPI is different from the status data, the acquiring unit 401 may calculate the KPI using at least one of the status data and a preset value (for example, material cost or electricity cost), etc.

[0033] The acquisition unit 401 may further acquire control parameters applied to the control-target device 20(T). When multiple devices 20 are control-target devices 20(T), the acquisition unit 401 may acquire control parameters for each control-target device 20(T). The control parameters may indicate instruction values ​​for manipulated variables or may indicate target values ​​(also referred to as set values). The acquisition unit 401 may acquire each control parameter for each reference interval. The acquisition unit 401 may acquire each control parameter from an AI control unit 402, which will be described later.

[0034] The acquisition unit 401 may supply each acquired state data to the AI ​​control unit 402 and the prediction unit 404. The acquisition unit 401 may supply each acquired state data, each KPI, and each control parameter 0 to the selection unit 405 and the storage unit 410. The state data supplied from the acquisition unit 401 to the AI ​​control unit 402 and the state data supplied to the selection unit 405 may be of the same type, or at least some of the state data may be of different types.

[0035] (((AI control unit 402))) The AI ​​control unit 402 controls the control-target device 20(T) using one of a plurality of learning models 420. The AI ​​control unit 402 according to this embodiment may have a plurality of learning models 420.

[0036] Each learning model 420 outputs a control parameter to be applied to the control-target device 20(T) in response to the supply of at least one type of status data related to the facility 2. When multiple devices 20 are control-target devices 20(T), each learning model 420 may output a control parameter for each control-target device 20(T). Each learning model 420 may be generated using a conventionally known learning algorithm, a common learning algorithm, or different learning algorithms. As an example, the learning model 420 may be generated by reinforcement learning and may be trained using a reward value corresponding to the value of at least one KPI. The reward value may be a value determined by a preset reward function. The reward function may be a function that increases the reward value as the value of the KPI to be evaluated approaches the target value. Note that the function is a mapping having a rule that associates each element of a set with each element of another set in a one-to-one relationship, and may be a mathematical formula or a table.

[0037] The learning data used in the learning process of the learning model 420 may include control parameters applied to each control-target device 20(T) and at least one type of state data related to the equipment 2 before or after application of the control parameters. The learning data may further include at least one KPI value before or after application of the control parameters. The learning data may be read from the storage unit 410 described below. At least one of the learning algorithm, learning data, or reward function in reinforcement learning may be different between the multiple learning models 420. "Different learning data" may mean that the types of state data or control parameters included in the learning data are different, or that the acquisition times of the state data or control parameters are different.

[0038] The AI ​​control unit 402 may supply the status data supplied from the acquisition unit 401 to a learning model 420 (also referred to as a target learning model 420(T)) set as the target for use among the multiple learning models 420, and may supply the control parameters output from the target learning model 420(T) to the control-target device 20(T). When multiple devices 20 are control-target devices 20(T), the AI ​​control unit 402 may supply the corresponding control parameters for each control-target device 20(T). The AI ​​control unit 402 may supply to the target learning model 420(T) status data of the same type as the status data included in the learning data of the target learning model 420(T), among the status data acquired by the acquisition unit 401.

[0039] (((input unit 403))) The input unit 403 acquires a signal in response to an input operation by a user. The input unit 403 may acquire the signal via an input device such as a keyboard. The input unit 403 may acquire a query signal requesting selection of a learning model 420 (also referred to as a recommended learning model 420(S)) recommended for use in controlling the control-target device 20(T), and may supply the query signal to the selection unit 405 and the prediction unit 404. The query signal may include identification information (also referred to as a KPI ID) of a KPI designated by the user from among multiple KPIs related to the equipment 2, and may further include information indicating a reference condition that the KPI must satisfy. For example, the reference condition may be that the KPI value becomes better than the current value, or that the KPI value falls within a predetermined reference range. The input unit 403 may acquire a signal approving the use of the selected learning model 420, and may supply the signal to the switching unit 409.

[0040] (((Prediction unit 404))) The prediction unit 404 predicts state data at each of a plurality of future time points based on the history of state data acquired by the acquisition unit 401. The history of state data may include state data in chronological order at a plurality of time points, and in the present embodiment, as an example, includes state data in chronological order and time information (for example, the time at which the state data was acquired). The prediction unit 404 may predict the state data using a conventionally known method. For example, the prediction unit 404 may predict future state data by performing regression analysis on past state data, or may predict future state data by a learning process using learning data including past state data. The prediction unit 404 may make a prediction for each state data acquired by the acquisition unit 401. The prediction unit 404 may make a prediction in response to a query signal supplied from the input unit 403. The prediction unit 404 may supply each predicted state data to the determination unit 406.

[0041] (((Selection unit 405))) The selection unit 405 selects a recommended learning model 420(S) that is recommended for use in controlling the control-target device 20(T) from among the multiple learning models 420 in the AI ​​control unit 402. Using a learning model 420 to control the control-target device 20(T) may mean that control parameters to be applied to the control-target device 20(T) are acquired from the learning model 420. In the present embodiment, as an example, the learning model 420 may be used as the target learning model 420(T) in the AI ​​control unit 402. The selection unit 405 may select the recommended learning model 420(S) in response to a query signal being supplied from the input unit 403. The selection unit 405 includes a setting unit 4051 and a model selection unit 4052.

[0042] ((((setting unit 4051)))) The setting unit 4051 switchably sets one of a plurality of KPIs related to the equipment 2 as a KPI (also referred to as a target KPI) used to select a recommended learning model 420(S). The setting unit 4051 may set a KPI indicated by a KPI ID included in the query signal, that is, a KPI specified by a user, as the target KPI. The setting unit 4051 may further set a reference condition for the target KPI included in the query signal. If the reference condition is not included in the query signal, a reference condition may be set in advance for the target KPI. The preset reference condition may be common regardless of the type of KPI, or may differ depending on the type of KPI. The setting unit 4051 may supply the KPI ID and the reference condition of the set target KPI to the model selection unit 4052.

[0043] ((((model selection unit 4052)))) The model selection unit 4052 selects the learning model 420 that is recommended for use to adapt the KPI for equipment 2 to the reference conditions as the recommended learning model 420(S). The model selection unit 405 may select the learning model 420 that is recommended for adapting the target KPI set by the setting unit 4051 to the reference conditions as the recommended learning model 420(S). The recommended learning model 420(S) may be different from the target learning model 420(T) currently being used.

[0044] The model selection unit 4052 may select a recommended learning model 420(S) based on at least one type of status data acquired by the acquisition unit 401, and may select a recommended learning model 420(S) based on the history of the status data acquired by the acquisition unit 401.

[0045] The model selection unit 4052 may select the recommended learning model 420(S) further based on the history of at least one type of past state data, the history of each KPI, and the history of the target learning model 420(T). In addition, the model selection unit 4052 may select the recommended learning model 420(S) further based on at least the model ID of the current target learning model 420(T). The model selection unit 4052 may acquire the model ID of the new target learning model 420(T) from the switching unit 409 described below each time the target learning model 420(T) is switched.

[0046] Here, the KPI history may include KPIs in chronological order at multiple points in time, and in this embodiment, as an example, includes KPIs in chronological order and time information (for example, the time when the KPI was acquired). The history of the target learning model 420(T) may include the model ID of the target learning model 420(T) in chronological order, and in this embodiment, as an example, includes the model ID of the target learning model 420(T) in chronological order and time information (for example, the time when use of the target learning model 420(T) began).

[0047] The model selection unit 4052 according to this embodiment may be a model that has undergone a learning process using learning data including at least one type of status data (i.e., the status data history), each KPI (i.e., the KPI history), and the model ID of the target learning model 420(T) (i.e., the history of the target learning model 420(T)) at each point in time in the past. The model of the model selection unit 4052 may be subjected to a learning process so as to output the model ID of the recommended learning model 420(S) in response to the supply of at least one type of status data history acquired by the acquisition unit 401 and at least the current model ID of the target learning model 420(T).

[0048] The model selection unit 4052 may be capable of calculating a predicted value of the target KPI at at least one point in the future when the recommended learning model 420(S) is set as the target learning model 420(T) based on the history of the supplied state data, and may output the model ID of the recommended learning model 420(S) on the condition that the predicted value of the target KPI satisfies a reference condition. If the reference condition is set to be that the KPI value will be better than the current value, the model selection unit 4052 may determine whether the predicted value of the target KPI satisfies the reference condition based on the value of the target KPI acquired by the acquisition unit 401. Note that the model selection unit 4052 may be capable of predicting the KPI in a steady state after switching the target learning model 420(T), and may further be capable of predicting the KPI in a transitional period due to the switching.

[0049] The model selection unit 4052 may select a recommended learning model 420(S) based on at least one type of state parameter and at least one type of control parameter acquired by the acquisition unit 401, or may select a recommended learning model 420(S) based on the history of at least one type of state data and the history of at least one type of control parameter acquired by the acquisition unit 401. In this case, the learning data for the model of the model selection unit 4052 may further include at least one type of control parameter at each time point in the past. The model of the model selection unit 4052 may undergo a learning process to output the model ID of the recommended learning model 420(S) in response to the supply of the history of the state data and the history of the control parameters acquired by the acquisition unit 401 and at least the model ID of the target learning model 420(T) at the current time.

[0050] Here, the control parameter history may include control parameters in chronological order at multiple points in time, and in the present embodiment, as an example, includes the control parameters in chronological order and time information (as an example, the time when the control parameters were applied to the control-target device 20(T)). The control parameters supplied to the model selection unit 4052 and the control parameters included in the learning data of the model selection unit 4052 may be control parameters of the control-target device 20(T) or may be control parameters of another device 20.

[0051] When outputting the model ID of the recommended learning model 420(S), the model selection unit 4052 may further output a reason for selecting the recommended learning model 420(S). The reason for selection may include information indicating a predicted change in KPI (in this embodiment, as an example, the target KPI) when using the recommended learning model 420(S). In this case, the reason for selection may include a predicted value of the target KPI at at least one point in the future. The reason for selection may include information that the user started using the recommended learning model 420(S) from a past state that is the same as the current situation. The reason for selection may include information indicating the reason why the user selected the recommended learning model 420(S) as the target learning model when the user started using the recommended learning model 420(S) as the target learning model in a past state that is the same as the current situation. In this case, the model selection unit 4052 may be trained using training data that includes the reason why the user changed the target learning model 420(T).

[0052] The model selection unit 4052 may supply the model ID of the selected recommended learning model 420(S) and the reason for selection to the determination unit 406.

[0053] (((Determining unit 406))) The determination unit 406 determines the start timing for starting use of the recommended learning model 420(S) based on the state data at each time point predicted by the prediction unit 404. The start timing may be the current time point or a time point later than the current time point. The determination unit 406 may acquire, from the model selection unit 4052, a predicted value of the target KPI at each time point in the future when the control-target device 20(T) is controlled using the control parameters acquired from the target learning model 420(T) at the current time point. The determination unit 406 may compare the KPI value at each time point in the future when the target learning model is continued to be used with the KPI value at each time point in the future when the recommended learning model 420(S) is used, and determine, as the start timing, the time when the KPI value is better when using the recommended learning model 420(S). Note that the KPI prediction by the model selection unit 4052 may include fluctuations in a transient state due to the start of use of the recommended learning model 420(S), and the determination unit 406 may determine the start timing taking into account the fluctuations in the transient state. The determination unit 406 may supply the generation unit 407 with information indicating the start timing, the model ID of the recommended learning model 420(S), and the reason for selecting the recommended learning model 420(S).

[0054] (((Generation unit 407))) The generation unit 407 generates a report regarding the recommended learning model 420(S). The generated report may include the model ID of the recommended learning model 420(S) and the determined start timing. The report may include the reason for selecting the recommended learning model 420(S) in addition to or instead of the start timing. For example, the report may include text in natural language. The generation unit 407 may generate a report based on the model ID, start timing, and selection reason provided by the determination unit 406. For example, if the model ID of the recommended learning model 420(S) is "Model B" and the target KPI is profit, the report may include the following content: "The temperature has increased by 5 degrees and the flow rate by 1 ton / h over the past three hours. Therefore, by immediately switching to Model B, profits may increase by approximately 10 to 20 million USD / h." The generation unit 407 may provide the generated report to the output unit 408. The generation unit 407 may further supply the output unit 408 with information indicating the start timing and the model ID of the recommended learning model 420(S).

[0055] In this embodiment, as an example, the selection unit 405, the determination unit 406, and the generation unit 407 may be integrated into a generation AI unit 430. In response to a query signal corresponding to a user operation, the generation AI unit 430 may output a natural language report including the model ID of the recommended learning model 420(S), a selection reason, and a start timing. The generation AI unit 430 may generate a natural language output sentence for a natural language input sentence input by a user via the input unit 403, and supply the output sentence to the output unit 408. The generation AI unit 430 may be generated by fine-tuning a known generation AI such as chat GPT using the learning data described above for the model selection unit 4052.

[0056] (((output unit 408))) The output unit 408 outputs the model ID of the recommended learning model 420(S). The output unit 408 may output the report generated by the generation unit 407. The output unit 408 may output the report to a display device (not shown). As a result, a signal indicating approval for using the recommended learning model 420(S) as the target learning model 420(T) may be acquired by the input unit 403 and supplied to the switching unit 409. The output unit 408 may supply information indicating the start timing and the model ID of the recommended learning model 420(S) to the switching unit 409.

[0057] (((Switching unit 409))) The switching unit 409 switches the target learning model 420(T) to the recommended learning model 420(S). In this embodiment, as an example, the switching unit 409 may switch the target learning model 420(T) in response to an approval signal being supplied from the input unit 403. The switching unit 409 may perform the switching at the start timing determined by the determination unit 406. The switching unit 409 may supply the model ID of the recommended learning model 420(S) to the AI ​​control unit 402, thereby setting the recommended learning model 420(S) as the target learning model 420(T). The switching unit 409 may also supply the model ID to the storage unit 410 and the selection unit 405.

[0058] (((storage unit 410))) The storage unit 410 stores various types of data. The storage unit 410 may store a history of each state data, a history of each control parameter, a history of the target learning model 420(T), and a history of each KPI. The storage unit 410 may store each state data, each control parameter, the target learning model, and each KPI in association with each other for each time. The information stored in the storage unit 410 may be used for learning by the model selection unit 4052 or may be used for learning by the generation AI unit 430.

[0059] According to the above-described control assistance device 4, a recommended learning model 420(S) that is recommended for use in controlling the control target device 20(T) to adapt the KPI to the reference conditions is selected from the multiple learning models 420 based on the acquired status data. Therefore, by acquiring control parameters using the recommended learning model 420(S), the KPI can be adapted to the reference conditions.

[0060] In addition, since the recommended learning model 420(S) is selected based on the history of acquired status data, it is possible to select the recommended learning model 420(S) that adapts the KPI to the reference conditions with high accuracy compared to when the recommended learning model 420(S) is selected based on status data at a single point in time.

[0061] Furthermore, based on the history of acquired status data, status data for each of a plurality of future time points is predicted, and based on the status data at each predicted time point, the start timing for starting use of the recommended learning model 420(S) is determined. Therefore, by obtaining control parameters using the recommended learning model 420(S) from the determined start timing, it is possible to reliably adapt the KPI to the reference conditions.

[0062] In addition, the recommended learning model 420(S) is selected based on the history of past state data, the history of KPI, and the history of the target learning model 420(T). Therefore, compared to when the recommended learning model 420(S) is selected without using the history of KPI and the history of the target learning model 420(T), the recommended learning model 420(S) that adapts the KPI to the reference conditions can be selected with high accuracy.

[0063] The selection unit 405 is also equipped with a model (in this embodiment, as an example, the model selection unit 4052) that uses learning data including status data at each past point in time, KPI, and the model ID of the target learning model 420(T), and outputs the model ID of the recommended learning model 420(S) in response to being supplied with the history of status data acquired by the acquisition unit 401 and the model ID of the target learning model 420(T) being used. Therefore, it is possible to select with high accuracy the recommended learning model 420(S) that adapts the KPI to the reference conditions.

[0064] Furthermore, the model of the model selection unit 4052 performs a learning process using learning data that further includes control parameters at each past point in time, and outputs the model ID of the recommended learning model 420(S) in response to the supply of the history of acquired state data and the history of control parameters, and the model ID of the target learning model 420(T). Therefore, compared to when the recommended learning model 420(S) is selected without using control parameters, it is possible to select with high accuracy the recommended learning model 420(S) that adapts the KPI to the reference conditions.

[0065] In addition, a report including the model ID of the recommended learning model 420(S) and the reason for selecting the recommended learning model 420(S) is output. Therefore, it is possible to determine whether or not to use the recommended learning model 420(S) based on the reason for selection.

[0066] In addition, the reason for selection includes information indicating the predicted change in KPI when using the recommended learning model 420(S), so it is possible to determine whether or not to use the recommended learning model 420(S) based on the predicted change in KPI.

[0067] Furthermore, since any of a plurality of KPIs related to the equipment 2 is set switchably as the target KPI used to select the recommended learning model 420(S), each of the set target KPIs can be adapted to the reference conditions.

[0068] Furthermore, a KPI designated by the user from among the multiple KPIs is set as the target KPI, so that the user can adapt any KPI to the reference conditions.

[0069] In addition, since the learning model 420 used to control the controlled device 20(T) can be switched to the recommended learning model 420(S), the control parameters output from the recommended learning model 420(S) can be used to control the controlled device 20(T) to adapt the KPI to the reference conditions.

[0070] (operation) 2 shows the operation of the control assistance device 4. The control assistance device 4 switches the target learning model 420(T) by performing the processing of steps S11 to S27. In parallel with the operation shown in this figure, the acquisition unit 401 may acquire status data from each sensor 21, and the AI ​​control unit 402 may supply the status data to the target learning model 420(T), and supply the control parameters output from the target learning model 420(T) to the controlled device 20(T) to control the equipment 2.

[0071] In step S11, the selection unit 405 acquires a query signal requesting selection of a recommended learning model 420(S) from the input unit 403. The query signal may include the KPI ID of the KPI designated by the user.

[0072] In step S13 , the prediction unit 404 predicts state data at each of a plurality of future time points based on the history of state data acquired by the acquisition unit 401 .

[0073] In step S15, the setting unit 4051 sets the KPI indicated by the KPI ID included in the query signal as the target KPI. The setting unit 4051 may further set a reference condition for the target KPI included in the query signal.

[0074] In step S17, the model selection unit 4052 selects a recommended learning model 420(S) that is recommended to be used to control the control-target device 20(T) to make the target KPI conform to the reference conditions, based on at least the current state data acquired by the acquisition unit 401. The selection unit 405 may select a single recommended learning model 420(S), or may select multiple recommended learning models 420(S). The selection unit 405 may output the model ID of the recommended learning model 420(S) and the reason for selecting the recommended learning model 420(S). If the selection unit 405 selects multiple recommended learning models 420(S), it may output the model ID of each recommended learning model 420(S) and the reason for selection.

[0075] In step S19, the determination unit 406 determines the start timing for starting use of the recommended learning model 420(S) based on the state data at each time point predicted in step S13. If multiple recommended learning models 420(S) are selected in step S17, the determination unit 406 may determine a common start timing for each recommended learning model 420(S), or may determine separate start timings for each recommended learning model 420(S).

[0076] In step S21, the generation unit 407 generates a report including the model ID of the recommended learning model 420(S), the reason for selecting the recommended learning model 420(S), and the start timing determined for the recommended learning model 420(S). If multiple recommended learning models 420(S) are selected in step S17, the generation unit 407 may generate a report for each recommended learning model 420(S) including the model ID of the recommended learning model 420(S) and the reason for selecting the recommended learning model 420(S). The report may further include the start timing of each recommended learning model 420(S).

[0077] In step S23, the output unit 408 outputs the generated report. The output unit 408 may output the report to a display device (not shown).

[0078] In step S25, the switching unit 409 determines whether an approval signal is supplied from the input unit 403. If multiple recommended learning models 420(S) are selected in step S17, the switching unit 409 may determine for which recommended learning model 420(S) an approval signal is supplied. If an approval signal is not supplied (step S25; No), the operation related to switching the target learning model 420(T) is terminated. If an approval signal is supplied (step S25; Yes), the process proceeds to step S27.

[0079] In step S27, the switching unit 409 switches the target learning model 420(T) by setting the recommended learning model 420(S) as the new target learning model 420(T). If multiple recommended learning models 420(S) are selected in step S17, the switching unit 409 may set the approved recommended learning model 420(S) as the new target learning model 420(T). The switching unit 409 may perform switching at the start timing determined for the recommended learning model 420(S) in step S19. This causes the control assistance device 4 to terminate the operation related to switching the target learning model 420(T).

[0080] According to the above operation, for each of the multiple recommended learning models 420(S), a report is output that includes the model ID of the recommended learning model 420(S) and the reason for selecting the recommended learning model 420(S), so that it is possible to determine which recommended learning model 420(S) to use based on the reason for selection.

[0081] (Example of operation) 3 shows changes in the status data acquired by the acquisition unit 401. The acquisition unit 401 may acquire status data for each of temperature, pressure, and flow rate. Each status data may be associated with identification information such as "TIC007," "PIC001," and "FIC001."

[0082] Figure 4 shows the difference in profit depending on the timing of switching the target learning model 420(T). Note that in this figure, the target KPI is set to the quality of the product, so the learning model 420 of "Model A" is used in advance as the target learning model 420(T), and the state data changes as shown in Figure 3.

[0083] In this case, when the user requests selection of a recommended learning model 420(S) with the target KPI as profit, as shown in the column "AI-selected learning model", the selection unit 405 selects the "Model B" learning model 420 as the recommended learning model 420(S), and the determination unit 406 determines timing "T1" as the start timing. When use of the "Model B" learning model 420 starts at timing "T1", the profit per unit time at each point in time will be as shown in the column "Profit by AI selection".

[0084] On the other hand, when a user starts using the "Model B" learning model 420 at their own discretion, the start timing is, for example, "T2" when the quality value, which is the target KPI, exceeds the reference value "10," as shown in the "AI-selected learning model" column. When the use of the "Model B" learning model 420 starts at this timing "T2," the profit per unit time at each point in time is as shown in the "Profit by user selection" column.

[0085] Figure 5 shows the trends in "profits from AI selection" and "profits from user selection." The vertical axis in the figure represents profit per unit time (million USD / h), and the horizontal axis represents time. As shown in this figure, when the learning model 420 of "Model B" selected by the selection unit 405 as the recommended learning model 420(S) is used from the start timing "T1" determined by the determination unit 406 (i.e., the case of "AI selection" in the figure), a decline in the KPI that the user values ​​(profit in the example in this figure) can be prevented compared to when the learning model 420 of "Model B" selected by the user is used from the start timing "T2" determined by the user (i.e., the case of "user selection" in the figure).

[0086] (Example of a dialogue between a user and the generation AI unit 430) When a user inputs an input statement such as "Based on past input values ​​and quality values, please tell me which is more profitable, Model A or Model B. Also, please tell me the reason why," the generation AI unit 430 may generate a first output statement such as "You will increase your profits by switching to Model B. TIC007 has increased by 5 degrees and FIC001 by 1 ton / h in the past three hours, so if you use Model B, you may be able to increase your profits by 10 to 20 million USD / h compared to continuing to use Model A."

[0087] Furthermore, if the user further inputs an input sentence such as "Switching to model B" in response to the first output sentence, the generation AI unit 430 may generate a second output sentence such as "Instructing to switch to model B," and the input sentence may be used as an approval signal to set model B as the target learning model 420(T).

[0088] Furthermore, if the user further inputs an input sentence such as "I will continue to use model A" in response to the first output sentence, the generation AI unit 430 may generate an output sentence such as "Please tell me the reason why you have decided to continue using model A. If you tell me the reason, it will help me decide whether to switch models in the future," and the model selection unit 4052 may perform a learning process using the input reason.

[0089] (Variation) ((First Modification)) In this modification, the prediction unit 404 may further predict, based on the history of status data acquired by the acquisition unit 401, status data at each of multiple time points after the start timing (also referred to as the first start timing) determined by the determination unit 406 when use of the recommended learning model 420(S) (also referred to as the first recommended learning model 420(S)) begins at the first start timing. The prediction unit 404 may be trained using learning data including the history of status data and the history of the target learning model 420(T), and may predict status data at each future time point after the first recommended learning model 420(S) is used based on the history of status data up to the present time and the history of the target learning model 420(T) up to the first start timing. The history of the target learning model 420(T) up to the first start timing may be the history of the target learning model 420(T) when the first recommended learning model 420(S) is set as the target learning model 420(T) at the first start timing.

[0090] In addition to selecting a first recommended learning model 420(S) based on state data up to the present time, the model selection unit 4052 may further select another recommended learning model 420(S) (also referred to as a second recommended learning model 420(S)) from the multiple learning models 420 that is different from the first recommended learning model 420(S) based on future state data predicted if the first recommended learning model 420(S) is used as the target learning model 420(T).

[0091] The model selection unit 4052 may select, as the second recommended learning model 420(S), a learning model 420 that is recommended for use in controlling the control-target device 20(T) to make the target KPI conform to the reference condition in a state indicated by the state data predicted for a point in time after the first start timing. The model selection unit 4052 may select the second recommended learning model 420(S) in the same manner as the first recommended learning model 420(S). The model selection unit 4052 may output the reason for selecting the second recommended learning model together with the model ID of the second recommended learning model 420(S).

[0092] The determination unit 406 may further determine another start timing (also referred to as a second start timing) at which to start using the second recommended learning model 420(S) based on the state data predicted for each time point after the first start timing. The determination unit 406 may determine the second start timing in the same manner as the first start timing.

[0093] The generation unit 407 may generate a report that further includes the model ID of the second recommended learning model 420(S) and the second start timing. The report may further include the reason for selecting the second recommended learning model 420(S).

[0094] In the first variant described above, when the use of the first recommended learning model 420(S) is started at the first start timing, the status data for each of a plurality of time points after the first start timing is further predicted, and a second recommended learning model 420(S) is further selected that is recommended for use in controlling the controlled device 20(T) to make the KPI conform to the reference conditions in the state indicated by the status data predicted for the time points after the first start timing. Furthermore, based on the status data predicted for each time point after the first start timing, a second start timing at which the use of the second recommended learning model 420(S) should be started is further determined. Therefore, by obtaining control parameters using the second recommended learning model 420(S) from the second start timing, it is possible to reliably make the KPI conform to the reference conditions even at times after the first start timing.

[0095] ((Second Modification)) In this modification, the storage unit 410 may store history data that associates the model ID of the target learning model 420(T) that has started to be used, the history of at least one type of state data before the switching, and the value of the KPI after the switching, each time the target learning model 420(T) is switched. The KPI after the switching may be the KPI immediately after the target learning model 420(T) is switched, or may be the KPI at the point in time when a reference period has elapsed since the switching. The history data may include control parameters at each point in time in the past.

[0096] The model selection unit 4052 may select one history of state data included in the history data based on the similarity between the history of state data acquired by the acquisition unit 401 and each history of state data included in the history data. The model selection unit 4052 may calculate the similarity between the acquired history of state data and each history of state data in the history data. The similarity between the histories of state data may be the similarity between data sets including values ​​of state data at a common number of time points for at least one type of common state data, or may be the distance between the data sets (for example, Euclidean distance or DTW distance). The model selection unit 4052 may select one history data set that is most similar to the acquired history data set from the history data. As a result, one history of state data that is most similar to the acquired history of state data is selected.

[0097] The model selection unit 4052 may select the target learning model 420(T) associated with the history of the selected piece of status data as the recommended learning model 420(S) on condition that the value of the KPI associated in the history data for the selected piece of status data satisfies the standard condition. If the value of the KPI associated with the history of the selected piece of status data does not satisfy the standard condition, the model selection unit 4052 may select the history of the piece of status data from the dataset with the next highest similarity. The model selection unit 4052 may select the target learning model 420(T) associated with the history of the selected piece of status data as the recommended learning model 420(S) on condition that the value of the KPI associated with the history of the selected piece of status data satisfies the standard condition. Thereafter, the model selection unit 4052 may repeat the same process until the value of the KPI associated with the history of the selected piece of status data satisfies the standard condition, and select the recommended learning model 420(S).

[0098] In the second modified example described above, the recommended learning model 420(S) that adapts the KPI to the reference conditions can also be selected with high accuracy.

[0099] ((Other variations)) In the above embodiment and modified examples, the setting unit 4051 sets a user-specified KPI as the target KPI. However, the target KPI may be set by other methods. For example, the setting unit 4051 may set a KPI that can be improved from its current state as the target KPI from among multiple KPIs. In this case, the query signal supplied from the input unit 403 may include a signal instructing the setting of any improvable KPI as the target KPI. The setting unit 4051 may detect a KPI that has previously shown a value better than the current value from the history of each KPI stored in the storage unit 410 and set the KPI as an improvable KPI. The setting unit 4051 may acquire, from the model selection unit 4052, a predicted value of each KPI when each learning model 420 is set as the target learning model 420(T), and set a KPI with a predicted value better than the current value as an improvable KPI. This allows the KPI to be improved to conform to the reference conditions even if the user is not aware of the KPI to be improved.

[0100] Furthermore, although the model selection unit 4052 and the generation AI unit 430 have been described as having been trained, they may be further trained using the contents stored in the storage unit 410 as training data. In this case, the control assistance device 4 may further include a learning processing unit that performs learning processing on the model selection unit 4052 and the generation AI unit 430.

[0101] Furthermore, although the control assistance device 4 has been described as including an AI control unit 402, an input unit 403, a prediction unit 404, a determination unit 406, a generation unit 407, a switching unit 409, and a storage unit 410, any of these may be omitted. If the control assistance device 4 does not include the AI ​​control unit 402, a user may manually control the control target device 20(T) using control parameters output from the target learning model 420(T). If the control assistance device 4 does not include the input unit 403, the setting unit 4051 may set an improvable KPI as the target KPI in response to detecting the KPI, regardless of a query signal, and the model selection unit 4052 may select a recommended learning model 420(S) each time the target KPI is changed. Furthermore, the switching unit 409 may automatically switch the recommended learning model 420(S) to the target learning model 420(T). If the control assistance device 4 does not have the prediction unit 404 and the determination unit 406, the switching unit 409 may switch the recommended learning model 420(S) to the target learning model 420(T) at the timing when an approval signal for switching is received. If the control assistance device 4 does not have the generation unit 407, the model ID of the recommended learning model 420(S) may be output to a display device. If the control assistance device 4 does not have the switching unit 409, the control assistance device 4 may be used to instruct a new user on the switching content of the target learning model 420(T).

[0102] Furthermore, although the selecting unit 405 has been described as including the setting unit 4051, it may not necessarily include the setting unit 4051. In this case, one of the multiple KPIs may be fixedly set in advance as the target KPI.

[0103] Various embodiments of the present invention may be described with reference to flowcharts and block diagrams, where the blocks may represent (1) stages of a process in which operations are performed or (2) sections of an apparatus responsible for performing the operations. Particular stages and sections may be implemented by dedicated circuitry, programmable circuitry provided with computer-readable instructions stored on a computer-readable medium, and / or a processor provided with computer-readable instructions stored on a computer-readable medium. Dedicated circuitry may include digital and / or analog hardware circuitry, and may include integrated circuits (ICs) and / or discrete circuits. Programmable circuitry may include reconfigurable hardware circuitry, including logical AND, OR, XOR, NAND, NOR, and other logical operations, flip-flops, registers, memory elements such as field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), and the like.

[0104] A computer-readable medium may include any tangible device capable of storing instructions that are executed by an appropriate device, such that the computer-readable medium having instructions stored thereon comprises an article of manufacture containing instructions that can be executed to create means for performing the operations specified in the flowcharts or block diagrams. Examples of computer-readable media may include electronic, magnetic, optical, electromagnetic, and semiconductor storage media. More specific examples of computer-readable media may include floppy disks, diskettes, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), electrically erasable programmable read-only memory (EEPROM), static random access memory (SRAM), compact disc read-only memory (CD-ROM), digital versatile disc (DVD), Blu-ray disc, memory stick, integrated circuit card, and the like.

[0105] The computer readable instructions may include either assembler instructions, Instruction Set Architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state-setting data, or source or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk®, JAVA®, C++, etc., and conventional procedural programming languages ​​such as the “C” programming language or similar programming languages.

[0106] The computer-readable instructions may be provided to a processor or programmable circuit of a programmable data processing device, such as a computer, locally or over a wide area network (WAN) such as a local area network (LAN) or the Internet, and the computer-readable instructions may be executed to create means for performing the operations specified in the flowcharts or block diagrams. Here, the computer may be a personal computer (PC), a tablet computer, a smartphone, a workstation, a server computer, a general-purpose computer, a special-purpose computer, or the like, or may be a computer system in which multiple computers are connected. Such a computer system in which multiple computers are connected is also called a distributed computing system, and is a broad definition of computer. In a distributed computing system, the multiple computers collectively execute a program by each executing a portion of the program and passing data between the computers as needed during program execution.

[0107] Examples of processors include a computer processor, a central processing unit (CPU), a processing unit, a microprocessor, a digital signal processor, a controller, a microcontroller, etc. A computer may have one or more processors. In a multiprocessor system with multiple processors, each processor executes a portion of a program and passes data between processors as needed during program execution, allowing the multiple processors to collectively execute a program. For example, in multitasking, each of the multiple processors may execute a portion of each task in small chunks by switching tasks at time slice intervals. In this case, which portion of a program each processor executes changes dynamically. Which portion of a program each of the multiple processors executes may also be statically determined by multiprocessor-aware programming.

[0108] 6 illustrates an example of a computer 1200 in which aspects of the present invention may be embodied, in whole or in part. Programs installed on the computer 1200 may cause the computer 1200 to function as or perform operations associated with an apparatus or one or more sections of the apparatus according to embodiments of the present invention, and / or to perform a process or steps of a process according to embodiments of the present invention. Such programs may be executed by the CPU 1212 to cause the computer 1200 to perform specific operations associated with some or all of the blocks in the flowcharts and block diagrams described herein.

[0109] A computer 1200 according to this embodiment includes a CPU 1212, a RAM 1214, a graphics controller 1216, and a display device 1218, which are interconnected by a host controller 1210. The computer 1200 also includes input / output units such as a communication interface 1222, a storage device 1224 such as a hard disk drive, a DVD-ROM drive 1226, and an IC card drive, which are connected to the host controller 1210 via an input / output controller 1220. The computer also includes legacy input / output units such as a ROM 1230 and a keyboard 1242, which are connected to the input / output controller 1220 via an input / output chip 1240.

[0110] The CPU 1212 operates according to programs stored in the ROM 1230 and the RAM 1214, thereby controlling each unit. The graphics controller 1216 acquires image data generated by the CPU 1212 into a frame buffer or the like provided in the RAM 1214 or into the graphics controller 1216 itself, and causes the image data to be displayed on the display device 1218.

[0111] The communication interface 1222 communicates with other electronic devices via a network. The storage device 1224 stores programs and data used by the CPU 1212 in the computer 1200. The DVD-ROM drive 1226 reads programs or data from a DVD-ROM 1227 and provides the programs or data to the storage device 1224 via the RAM 1214. The IC card drive reads programs and data from an IC card and / or writes programs and data to an IC card.

[0112] The ROM 1230 stores therein a boot program or the like that is executed by the computer 1200 upon activation, and / or programs that depend on the hardware of the computer 1200. The input / output chip 1240 may also connect various input / output units to the input / output controller 1220 via a parallel port, a serial port, a keyboard port, a mouse port, etc.

[0113] The programs are provided by a computer-readable medium such as a DVD-ROM 1227 or an IC card. The programs are read from the computer-readable medium, installed in the storage device 1224, RAM 1214, or ROM 1230, which are also examples of computer-readable media, and executed by the CPU 1212. Information processing described in these programs is read by the computer 1200, and causes cooperation between the programs and the various types of hardware resources described above. An apparatus or a method may be configured by implementing information manipulation or processing in accordance with the use of the computer 1200.

[0114] For example, when communication is performed between the computer 1200 and an external device, the CPU 1212 may execute a communication program loaded into the RAM 1214 and instruct the communication interface 1222 to perform communication processing based on the processing described in the communication program. Under the control of the CPU 1212, the communication interface 1222 reads transmission data stored in a transmission buffer processing area provided in the RAM 1214, the storage device 1224, the DVD-ROM 1227, or a recording medium such as an IC card, and transmits the read transmission data to the network, or writes reception data received from the network to a reception buffer processing area or the like provided on the recording medium.

[0115] Furthermore, the CPU 1212 may cause all or a necessary portion of a file or database stored in an external recording medium such as the storage device 1224, the DVD-ROM drive 1226 (DVD-ROM 1227), an IC card, etc. to be read into the RAM 1214, and perform various types of processing on the data in the RAM 1214. The CPU 1212 then writes back the processed data to the external recording medium.

[0116] Various types of information, such as various types of programs, data, tables, and databases, may be stored on the recording medium and may undergo information processing. CPU 1212 may perform various types of processing on data read from RAM 1214, including various types of operations, information processing, conditional judgment, conditional branching, unconditional branching, information search / replacement, etc., as described throughout this disclosure and specified by the instruction sequences of the programs, and write the results back to RAM 1214. CPU 1212 may also search for information in a file, database, etc. on the recording medium. For example, if multiple entries each having an attribute value of a first attribute associated with an attribute value of a second attribute are stored on the recording medium, CPU 1212 may search for an entry that matches a condition specified by the attribute value of the first attribute from among the multiple entries, read the attribute value of the second attribute stored in the entry, and thereby obtain the attribute value of the second attribute associated with the first attribute that satisfies a predetermined condition.

[0117] The above-described programs or software modules may be stored in a computer-readable medium on or near the computer 1200. A recording medium such as a hard disk or RAM provided in a server system connected to a dedicated communication network or the Internet can also be used as a computer-readable medium, thereby providing the programs to the computer 1200 via the network.

[0118] Although the present invention has been described above using embodiments, the technical scope of the present invention is not limited to the scope described in the above embodiments. It will be apparent to those skilled in the art that various modifications and improvements can be made to the above embodiments. It is clear from the claims that such modifications and improvements can also be included within the technical scope of the present invention.

[0119] It should be noted that the execution order of each process, such as operations, procedures, steps, and stages, in the devices, systems, programs, and methods shown in the claims, specifications, and drawings is not specifically stated as "before," "prior to," etc., and that the processes can be performed in any order unless the output of a previous process is used in a subsequent process. Even if the operational flow in the claims, specifications, and drawings is described using "first," "next," etc. for convenience, this does not mean that the processes must be performed in this order. [Explanation of symbols]

[0120] 1 System 2 Equipment 4 Control support device 20 equipment 20(T) Controlled equipment 21 Sensors 401 Acquisition Department 402 AI control section 403 Input section 404 Prediction Department 405 Selection Section 406 Decision Section 407 Generation part 408 Output Section 409 Switching section 410 Storage section 420 Learning Model 420(T) Target Learning Model 420(S) Recommended Study Model 430 Generation AI Department 1200 Computer 1210 host controller 1212 CPU 1214 RAM 1216 Graphics Controller 1218 Display Devices 1220 Input / Output Controller 1222 communication interface 1224 Storage device 1226 DVD-ROM drive 1227 DVD-ROM 1230 ROM 1240 Input / Output Chip 1242 keyboard 4051 Settings 4052 Model Selection Section

Claims

1. an acquisition unit that acquires status data related to the equipment; a selection unit that, in response to the supply of status data related to the equipment, selects, from a plurality of learning models that output control parameters to be applied to a control object in the equipment, a recommended learning model that is recommended to be used to control the control object in order to make a KPI related to the equipment conform to a reference condition, based on the status data acquired by the acquisition unit; an output unit that outputs identification information of the recommended learning model; An apparatus comprising:

2. The device according to claim 1 , wherein the selection unit selects the recommended learning model based on a history of the state data acquired by the acquisition unit.

3. a prediction unit that predicts state data at each of a plurality of future time points based on the history of the state data acquired by the acquisition unit; A determination unit that determines a start timing for starting use of the recommended learning model based on the predicted state data at each point in time; a generation unit that generates a report including identification information of the recommended learning model and the start timing; Furthermore, The apparatus according to claim 2 , wherein the output unit outputs the report.

4. The prediction unit further predicts state data at each of a plurality of time points after the start timing when use of the recommended learning model is started at the start timing, The selection unit further selects another recommended learning model from the plurality of learning models, which is different from the recommended learning model and is recommended to be used for controlling the control object in order to make the KPI conform to the reference condition in a state indicated by state data predicted for a point in time after the start timing; The determination unit further determines another start timing at which to start using the other recommended learning model based on state data predicted for each time point after the start timing; The device of claim 3 , wherein the generator generates a report further including identification information of the other recommended learning models and the other start timings.

5. The device according to claim 2 , wherein the selection unit selects the recommended learning model further based on a history of the past status data, a history of the KPI, and a history of the learning model used to control the control object.

6. The device described in claim 5, wherein the selection unit performs a learning process using learning data including status data at each past point in time, the KPI, and identification information of the learning model used to control the control object, and has a model that outputs identification information of the recommended learning model in response to being supplied with the history of the status data acquired by the acquisition unit and the identification information of the learning model used to control the control object.

7. the acquisition unit further acquires a control parameter applied to the control object; The device described in claim 6, wherein the model undergoes a learning process using learning data that further includes control parameters at each past point in time, and outputs identification information of the recommended learning model in response to being supplied with the history of the status data and the history of the control parameters acquired by the acquisition unit, and identification information of the learning model used to control the controlled object.

8. Further provided is a storage unit that stores history data that associates, each time the learning model used for controlling the control object is switched, identification information of the learning model that has started to be used, a history of the status data before the switching, and the value of the KPI after the switching, The selection unit selecting one history of the status data included in the history data based on a similarity between the history of the status data acquired by the acquisition unit and each history of the status data included in the history data; The device described in claim 5 selects a learning model associated with a history of the status data as the recommended learning model, on the condition that the value of the KPI associated with the history of the status data in the historical data satisfies the standard condition.

9. A generating unit generates a report including identification information of the recommended learning model and a reason why the selecting unit selected the recommended learning model, The apparatus according to claim 1 , wherein the output unit outputs the report.

10. The selection unit selects a plurality of the recommended learning models, The apparatus of claim 9 , wherein the generator generates the report including, for each recommended learning model, identification information of the recommended learning model and a reason for selecting the recommended learning model.

11. The apparatus of claim 9 , wherein the reason includes information indicating a predicted change in the KPI when the recommended learning model is used.

12. The device according to claim 1 , wherein the selection unit has a setting unit that switchably sets one of a plurality of KPIs related to the equipment as the KPI to be used in selecting the recommended learning model.

13. The device according to claim 12 , wherein the setting unit sets a KPI designated by a user from among the plurality of KPIs as the KPI to be used for selecting the recommended learning model.

14. The device according to claim 12 , wherein the setting unit sets, among the plurality of KPIs, a KPI that can be improved from a current state as the KPI to be used for selecting the recommended learning model.

15. The device according to claim 1 , further comprising a switching unit that switches a learning model used for controlling the control target to the recommended learning model.

16. an acquisition stage for acquiring status data relating to the equipment; a selection step of selecting, based on the status data acquired in the acquisition step, a recommended learning model to be used to control the control object in the equipment from among a plurality of learning models that output control parameters to be applied to the control object in the equipment in response to the supply of status data related to the equipment, the recommended learning model being recommended to be used to control the control object in order to make the KPI related to the equipment conform to a reference condition; an output stage for outputting identification information of the recommended learning model; A method for providing the above.

17. When executed by a computer, the computer an acquisition unit that acquires status data related to the equipment; a selection unit that, in response to the supply of status data related to the equipment, selects, from a plurality of learning models that output control parameters to be applied to a control object in the equipment, a recommended learning model that is recommended to be used to control the control object in order to make a KPI related to the equipment conform to a reference condition, based on the status data acquired by the acquisition unit; an output unit that outputs identification information of the recommended learning model; A program that functions as a

Images