A computer system for implementing a plant operation support system and a plant operation support method, and a computer program recording medium.

Abstract

To provide a plant operation assistance system that selects by itself what operation know-how should be learned and requests a person to proceed with learning.SOLUTION: A plant operation assistance system 30 includes: a real-time inference system 310 that inputs plant data generated in a plant system 20 and performs inference; and an AI training system 300 that inputs teacher data created by using the plant data and trains the real-time inference system 310. The AI training system 300 has a know-how evaluation unit 303 that evaluates the reliability of an inference result and, according to the evaluation result of the reliability, selectively requests an operator 40 to participate in teacher data creation.SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The present invention relates to a plant operation support system and a plant operation support method.

Background Art

[0002] Preferably, the present invention relates to a system that uses artificial intelligence to automate at least a part of the operation of a plant and support the plant operator.

[0003] As the scenarios where artificial intelligence automatically performs the operation of a plant increase, the burden on the operator is reduced.

[0004] As a related prior art, there is an AI process control monitoring device described in Patent Document 1.

[0005] This device inputs, in real time from the operator during the operation of the plant, the control information related to process control that is latent in the operator's mind, and discriminates and learns patterns having statistical regularity from the input control information, the operation information by which the operator operates the plant, and the observation data of the plant, and constructs a qualitative model related to the operation of the plant based on the learned patterns. As learning progresses, monitoring control of the plant process is performed based on the constructed qualitative model.

Prior Art Documents

Patent Documents

[0006]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0007] In the minds of plant operators, there is a vast amount of tacit knowledge, encompassing numerous techniques for dealing with the diverse situations that can occur in a plant. From this enormous volume of tacit knowledge, it is difficult for operators to determine which information to select and teach to artificial intelligence (AI) in a way that will enable efficient learning and, furthermore, result in meaningful learning for plant operation. Therefore, if operators were to select the know-how to teach to AI based on their own judgment, the process of verbalizing and defining this tacit knowledge, creating and linking training data to it, and linking it to AI could be inefficient, potentially wasting excessive time and effort before the AI ​​reaches a practical level of development.

[0008] One objective of the present invention is to provide an artificial intelligence-based plant operation support system in which the system itself selects specific operational data that would be improved by registering know-how, and effectively requests the registration of that know-how from a human. [Means for solving the problem]

[0009] A plant operation support system according to one embodiment comprises an inference system that accepts plant data generated at the plant, performs inference using the accepted plant data, and outputs inference results, and a training system that inputs training data created with the involvement of plant operators using plant data, and trains the inference system by incorporating plant operation know-how into the inference system using the input training data. The training system has operator utilization means that selects specific plant data related to specific know-how that is desirable to be incorporated into the inference system from among the given plant data, and requests plant operators to participate in creating new training data based on the selected specific plant data.

[0010] In one embodiment of the plant operation support system, the system itself selects what know-how to incorporate, and develops its reasoning system while requiring human involvement in creating educational data related to the selected know-how. [Brief explanation of the drawing]

[0011] [Figure 1] This figure illustrates the hardware configuration of a plant operation support system according to one embodiment. [Figure 2] This diagram shows the basic functions of the plant operation support system according to this embodiment. [Figure 3] This figure shows a more specific functional configuration of a system in which the plant operation support system according to this embodiment is applied to support plant operations. [Figure 4] This diagram shows the operation of the plant operation support system according to this embodiment. [Figure 5] This diagram illustrates the know-how extraction process in the plant operation support system according to this embodiment. [Figure 6] This figure illustrates the user interface for registering know-how and training data in the plant operation support system according to this embodiment. [Figure 7] This figure illustrates a user interface for suggesting and evaluating know-how in a plant operation support system according to this embodiment. [Figure 8] This figure shows an example of the reliability evaluation flow for real-time inference in the plant operation support system according to this embodiment. [Figure 9] This figure shows a modified configuration for registering know-how in step S5 shown in Figure 4 in the plant operation support system according to this embodiment. [Figure 10] This diagram illustrates the flow of the process for learning driving evaluation methods. [Figure 11] This diagram illustrates the flow of communication between the operator and the system in the user interface during the training of the driving evaluation method. [Figure 12] This diagram illustrates the flow of the operational evaluation process. [Figure 13] This diagram illustrates the flow of communication between the operator and the system in the user interface during operational evaluation. [Figure 14]This is a diagram illustrating a proposed message of a real-time inference system.

Embodiments for Carrying Out the Invention

[0012] Hereinafter, embodiments of the present invention will be described with reference to the drawings. Note that the embodiments described below do not limit the invention according to the claims, and not all of the elements and combinations thereof described in the embodiments are essential for the solution means of the invention.

[0013] In the drawings for explaining the embodiments, the same reference numerals are given to portions having the same function, and repeated explanations thereof are omitted.

[0014] In addition, in the following description, the expression "xxx data" may be used as an example of information, but the data structure of the information may be any type. That is, in order to indicate that the information does not depend on the data structure, "xxx data" can be referred to as "xxx table". Furthermore, "xxx data" may simply be referred to as "xxx". And in the following description, the configuration of each information is an example, and the information may be divided and held, or combined and held.

[0015] In the following description, the "program" may be used as the subject to explain the processing. However, the program is executed by a processor (for example, a CPU (Central Processing Unit)) to perform the defined processing while appropriately using a storage resource (for example, a memory) and / or a communication interface device (for example, a port). Therefore, the subject of the processing may be the program. The processing described with the program as the subject may also be the processing performed by a processor or a computer having the processor.

[0016] FIG. 1 shows the physical configuration of a computer system to which a plant operation support system according to one embodiment is applied.

[0017] The computer system 50 consists of multiple (or one) physical computers 201 connected to the network 240.

[0018] Network 240 is one or more communication networks, and may include, for example, at least one of an FC (Fibre Channel) network and an IP (Internet Protocol) network. Network 240 may be located outside the computer system 50.

[0019] Each physical computer 201 is, for example, a general-purpose computer and has physical computer resources 330. The physical computer resources 330 include an interface unit 251, a storage unit 252, and a processor unit 253 connected thereto.

[0020] The computing system 50 may be, for example, a cloud computing system that provides XaaS (X as a Service). Generally, "XaaS" refers to a service that makes some of the resources necessary for building or operating a system (e.g., hardware, network lines, software execution environment, application programs, development environment, etc.) available through a network such as the Internet. The letter (or word) used as "X" in XaaS varies depending on the type of XaaS (service model). For example, examples of XaaS include PaaS (Platform as a Service), SaaS (Software as a Service), IaaS (Infrastructure as a Service), or HaaS (Hardware as a Service).

[0021] Figure 2 is a diagram illustrating the basic functions of the plant operation support system according to the embodiment.

[0022] The plant system 20 shown in Figure 2 may include not only the plant itself, but also the people involved with the plant (hereinafter referred to as "operators," but operators may include not only those involved in the operation of the plant, but also those involved in aspects other than operation) 40, as well as the environment and external systems that have a relationship with the plant. The data that enters the plant operation support system 30 from the plant system 20 (hereinafter referred to as "plant data") may include data and signals that are input to and output from the plant system 20, and data and signals that pass through the plant system 20. For example, this may include various operation commands to the plant system 20, sensor data output by various sensors installed in the plant system 20, various DCS data that is input to and output to the DCS (Distributed Control System) that controls the plant system 20, and analysis data obtained by analyzing these various data and analyzing the products and state of the plant.

[0023] The plant operation support system 30 receives plant data from the plant system 20 and performs inferences regarding the input plant data. Preferably, the plant operation support system 30 receives plant data in real time and performs inference work in real time. The reliability of the inferences performed by the plant operation support system 30 (i.e., the inference capability of the plant operation support system 30, i.e., the learning level, or the inference results obtained from that inference) is evaluated. Preferably, the reliability is evaluated for each plant data input at each point in time, or for each classification group of plant data. Therefore, at a given point in time, the reliability of the inference may be evaluated as high for some plant data, but as low for other plant data.

[0024] If, at a certain point in time, the reliability of the inference regarding certain plant data is evaluated as low, it means that at that point, the plant operation support system 30 has not yet acquired sufficient knowledge, such as know-how, regarding that environmental data, and therefore it is desirable for the system to learn about that plant data. Therefore, the plant operation support system 30 presents the plant data to the operator 40 and requests the operator 40 to input the operational know-how related to that plant data as training data into the plant operation support system 30. Based on this training data, the plant operation support system 30 then learns the know-how. In other words, the plant operation support system 30 proactively engages with the operator 40, utilizing human intelligence (HI) to acquire and learn new know-how. On the other hand, if, at a certain point in time, the reliability of the inference regarding certain plant data is evaluated as high, the plant operation support system 30 provides the operator 40 with the inference result (suggested operation).

[0025] In this way, the plant operation support system 30 proactively identifies opportunities, situations, or events in which it is preferable to use HI, and proactively engages the operator 40 to use HI for learning (receives instruction from HI), thereby increasing the reliability of its reasoning.

[0026] As a non-limiting specific example of the process from (2) "Evaluate the reliability of the inference" to (3) "Notify a human if the reliability is low" in Figure 2, the plant operation support system 30 has, for example, the following three processes. 1) From the plant data previously entered and stored in the plant operation support system 30, plant data that has not yet been used to create training data is selected (preferably, plant data with a higher frequency of occurrence is selected first) to be used as material for new training data. 2) When plant data is input in real time at each point in time and inference is performed in real time, the reliability of the inference results that can be obtained from the input plant data is evaluated based on the characteristics of the input plant data at each point in time (for example, based on the similarity of the characteristics of the input plant data), and plant data that is evaluated as having low reliability of the inference results is selected as material for new training data. 3) When plant data is input in real time at each point in time and inference is performed in real time, the inference results for the input plant data at each point in time are provided to a human, and the human is asked to evaluate the validity of the inference results. Based on the validity evaluation results fed back by the human, input plant data that were evaluated as having low validity of the inference results are selected as material for new training data.

[0027] Figure 3 shows a more specific functional configuration of the plant operation support system according to the embodiment.

[0028] The plant operation support system 30 of this embodiment, shown in Figure 3, is capable of communicating with the plant system 20 and the plant operator 40. The plant operation support system 30 includes an AI training system 300 and a real-time inference system 310. The real-time inference system 310 has a configuration such as that disclosed in International Publication 2019 / 003485.

[0029] The plant system 20 comprises an analysis system 21, a control panel 22, a DCS 23, a sensor group 24, an equipment group 25, and a group of materials to be processed 26. The equipment group 25 consists of various pieces of equipment that make up the production facilities of the plant system, and the group of materials to be processed 26 consists of raw materials, intermediate products, products, waste, etc., that are processed on the production facilities. The sensor group 24 senses the state and operation of the equipment group 25 and the group of materials to be processed 26. The DCS 23 controls the process of the equipment group 25. The control panel 22 is operated by the operator 40 to operate the equipment group 25. The analysis system 21 receives data output from the sensor group 24 and performs analysis on various states, operations, performance, and productivity of the plant system 20. As used herein, "plant data" refers to a bundle of various signals and data generated from these components of the plant system 20.

[0030] The plant operation support system 30 inputs plant data generated moment by moment in the plant system 20 in real time and performs real-time inference using this real-time input plant data ("(1) Real-time data" in Figure 3). The inference results (typically, notifications indicating which know-how to apply to which plant data and what operation to propose) are output to the operator 40. The plant operation support system 30 also inputs plant data generated in the past in the plant system 20 ("(1) Time-series data" in Figure 3) (time-series data may be past plant data accumulated externally over a certain period in advance, or it may be real-time input data sequentially saved and accumulated within the plant operation support system 30) and uses this time-series data in the learning cycle.

[0031] The AI ​​training system 300 includes a data classification unit 301, a know-how registration unit 302, and a know-how evaluation unit 303.

[0032] The data classification unit 301 classifies the input plant data (real-time data and time-series data) into multiple groups according to classification rules (i.e., labels the input plant data). The classification rules may be set in advance by the system designer or operator. The groups (i.e., labels) may be arbitrarily set, for example, groups based on various operating operations, groups based on various alarms, or groups based on various operating scenarios such as equipment startup or target value changes. With regard to real-time data, after classifying (labeling) the real-time data, the data classification unit 301 provides that real-time data to the real-time inference system 310.

[0033] On the other hand, with respect to time-series data, the data classification unit 301 classifies the time-series data and, based on the classification results, selects and extracts time-series data (plant data) related to operational know-how that should be prioritized for learning (i.e., operational know-how that the real-time inference system 310 has not yet learned or whose learning level is insufficient). The method of selection and extraction will be described later. The data classification unit 301 then provides the extracted plant data to the know-how registration unit 302.

[0034] The know-how registration unit 302 notifies the operator 40 of the plant data extracted by the data classification unit 301 and requests the operator 40 to register the know-how. The operator 40 registers the know-how in response to the request from the know-how registration unit 302. There are two stages of know-how registration, indicated in the diagram as "Know-how Registration" and "Training Data Registration". "Know-how Registration" is performed by the operator 40 inputting basic information about the know-how that formed the basis of the operating operations shown in the notified plant data (for example, the name of the know-how and the content of the operating operations). "Training Data Registration" is performed by the operator 40 inputting more detailed information about the know-how required for the training data (for example, the conditions that created the need for the operation, such as the temperature and pressure of the equipment being such that the operation was performed).

[0035] Once the operator 40 has registered the know-how and training data, the know-how registration unit 302 creates training data from the registered information and provides the training data to the data classification unit 301. The data classification unit 301 has an inference model in the process of learning (not shown; hereinafter referred to as the learning model), and provides the training data to the learning model to allow it to learn the know-how. This process is repeated for a large amount of plant data, improving the inference performance of the learning model. When the learning model has progressed to a level that meets predetermined conditions, the data classification unit 301 deploys the learning model to the real-time inference system 310 (that is, it updates, or trains, the inference model in the real-time inference system 310 to one with higher inference performance).

[0036] The real-time inference system 310 inputs plant data in real time, performs inference, and passes the inference results for each plant data to the know-how evaluation unit 303. The know-how evaluation unit 303 evaluates the reliability of the inference results for each plant data and determines whether the reliability is lower than a certain level (i.e., the plant data relates to know-how to be learned) or higher than a certain level (i.e., the plant data does not relate to know-how to be learned). low In this case, it means that the input plant data at that time is still unlearned or under-learned, and therefore new know-how should be registered. Accordingly, the real-time inference system 310 provides the inference result (which includes at least the input plant data used for the inference at that time) to the data classification unit 301.

[0037] The data classification unit 301 then passes the input plant data to the know-how registration unit 302 in the manner already described. The know-how registration unit 302 then requests the operator 40 to register the know-how and training data for the input plant data in the manner already described. This know-how registration and training data registration may be performed immediately after the inference results are obtained, or it may be performed after a certain amount of unreliable inference results have accumulated. However, performing it immediately is convenient for the operator 40 because they are requested to register the know-how and training data for an operation immediately after performing that operation. In any case, once the training data is newly registered in this way, the learning model is then trained using the newly registered training data in the manner already described, and eventually, the learning model with an even higher learning level is deployed to the inference system 310. This further enhances the inference capabilities of the real-time inference system 310.

[0038] On the other hand, if the inference result regarding the input plant data at that time is deemed to be more reliable than a certain level, the know-how evaluation unit 303 displays the inference result (which includes the input plant data, the learned know-how used in the inference, and the proposed operation derived from the inference) to the operator 40 in the form of a "know-how proposal" (for example, "Since event XXXX has occurred, we propose operation XXX according to know-how XXXX"). The operator 40 evaluates whether the know-how proposal is appropriate or inappropriate and inputs the evaluation result as feedback to the know-how evaluation unit 303.

[0039] The know-how evaluation unit 303 notifies the data classification unit 301 whether the know-how evaluation result feedback from the operator 40 indicates validity or invalidity. If the operator 40 determines that the know-how proposal is invalid, it means that new know-how should be registered for the input plant data used in the inference at that time. Therefore, in this case, as with the case where the reliability of the inference result described above was low, the data classification unit 301 passes the input plant data to the know-how registration unit 302, and the know-how registration unit 302 registers the know-how and training data for the input plant data in the manner already described. The know-how registration and training data registration in this case may also be performed immediately after the feedback from the operator 40, or after a certain amount of feedback from the operator 40 has accumulated, but the former would be more convenient for the operator 40. In any case, the registration of know-how and training data based on this feedback further enhances the inference capabilities of the real-time inference system 310.

[0040] Figure 4 shows the operation of the plant operation support system according to this embodiment.

[0041] In this embodiment, the AI ​​system 30 acquires plant data (real-time data and / or time-series data) from the plant system 20 (step S1).

[0042] The real-time inference system 310 acquires real-time data from the plant system 20 (S2), performs inference operations based on this real-time data, and outputs the inference results to the AI ​​training system 300 (S3).

[0043] Meanwhile, the AI ​​training system 300 acquires and stores time-series data from the plant system 20 and classifies this time-series data into multiple groups according to predetermined classification rules (S4). Next, the AI ​​training system 300 prioritizes selecting groups with a high frequency of occurrence (i.e., a relatively large amount of plant data classified into those groups) from among these groups (i.e., various know-how domains), extracts data that has not yet been used to create training data (i.e., plant data belonging to a new know-how domain) from within the selected groups, presents the extracted plant data to the operator 40, and requests the operator 40 to register the know-how and training data (S5). The AI ​​training system 300 then receives the know-how registered by the operator 40 (S6) and the training data related to that know-how (S5). Furthermore, the AI ​​training system 300 uses the training data registered by the operator 40 to train its own learning model with the new (i.e., registered by the operator 40) know-how (S7). Then, the AI ​​training system 300 registers (deploys) the newly learned model to the real-time inference system 310 (S8).

[0044] Furthermore, the AI ​​training system 300 determines the reliability of the inference results input from the real-time inference system 310, and if it determines that the reliability is high, it proposes know-how based on the inference results to the operator 40 (S9). If it determines that the reliability of the inference results is low, it proceeds to step S5 using the input plant data used for the inference at that time (S9). In addition, it receives the operator 40's evaluation of the proposed know-how, and if the know-how proposal is evaluated as inappropriate, it proceeds to step S5 using the input plant data used for the inference at that time (S9).

[0045] Figure 5 is a diagram illustrating the know-how extraction process in the plant operation support system according to this embodiment. Figure 6 is a diagram illustrating the user interface for know-how registration and training data registration in the plant operation support system according to this embodiment.

[0046] The process shown in Figure 5 corresponds to the process performed by the data classification unit 301 and the know-how registration unit 302 in Figure 3, that is, the process from steps S4 to S5 in Figure 4.

[0047] In Figure 5, first, in step S11, data classification, graphing, and frequency calculation, time-series data is classified into multiple groups based on predetermined classification rules (for example, groups corresponding to multiple types of operation, or groups corresponding to multiple types of alarms, etc., groups corresponding to multiple types of events that can occur, i.e., multiple areas of know-how). Plant data at each point in time is then saved for each group (i.e., event type, in other words, area of ​​know-how). The amount of saved data for each group, in other words, the occurrence frequency of each group, is then calculated and graphed. In the example shown, the data is classified into three groups A, B, and C, with group A having the highest occurrence frequency.

[0048] Such histograms for each group may be presented to the operator 40. Details of the plant data sets belonging to each group may also be presented to the operator 40. This makes it easier for the operator to determine what know-how should be prioritized for training the model.

[0049] Next, in step S12, during know-how registration, a group with a larger amount of plant data that has not yet been used to create training data (i.e., including unlearned events with a higher frequency of occurrence) is preferentially selected from among multiple groups (typically the group with the highest frequency of occurrence is selected), and plant data that has not been used to create training data is extracted from that group. For example, in Figure 5, some unused plant data is extracted from group A, which has the highest frequency of occurrence. Then, along with the extracted plant data, a know-how registration form, as exemplified in Figure 6(A), is displayed to the operator, and the operator is requested to register the know-how. The operator registers the know-how by entering the necessary information into the form.

[0050] As illustrated in Figure 6(A), when registering know-how, basic information about the operational know-how related to the extracted plant data, such as the know-how name, the equipment operated, the content of the operation, and the time of the operation, is entered by the operator 40.

[0051] In the subsequent step S13, during the registration of training data, a training data registration form, as illustrated in Figure 6(B), is displayed to the operator 40. The operator 40 enters more detailed information into this form, which is necessary for creating the training data and should be registered in addition to the know-how registration. In the example shown in Figure 6(B), one or more conditions that must be checked (i.e., must be met in order to perform the registered operation), such as "liquid temperature of equipment A" and "vaporization rate of equipment A," are registered. Based on these registered items, training data is created.

[0052] Through the registration of know-how and training data as described above, once training data for a particular operational know-how is created and registered, then, as shown in step S13 in Figure 5, plant data that is deemed applicable to the registered know-how (for example, plant data with a high degree of similarity to the plant data used for know-how registration) is automatically searched and extracted from the already accumulated set of past plant data (time-series data) based on that registered training data. The extracted plant data is then automatically labeled with the same label as the registered know-how. This creates a new group (label) that contains only the plant data corresponding to that registered know-how. In other words, as illustrated in the graph of Group A in step S12 in Figure 5, within Group A (which is a relatively broad classification) classified according to the predetermined classification rules in step S11, more detailed groups 1, 2, and 3 are created for each registered know-how, and the data in each group is labeled with the label of that group. Then, based on the extracted data classified for each registered know-how, additional training data related to that registered know-how is created, and additional training of the learning model is performed using this additional training data, thereby further improving the learning level.

[0053] As described above, once detailed groups are created for each registered know-how, these detailed groups (labels) are applied when the data classification unit 301, shown in Figure 3, classifies (labels) time-series data and real-time data (S11 in Figure 5). Therefore, as the amount of registered know-how increases, the data classification becomes more refined, which contributes to improving the accuracy of inference. The creation of detailed groups can be done completely automatically, but the assistance of the operator 40 may also be used. For example, when data corresponding to registered know-how is extracted from past accumulated datasets, the extracted data can be presented to the operator 40, who can judge whether the extraction date is valid or not, and only if it is valid the extracted data can be added to that group. Alternatively, training data created based on that extracted data can be presented to the operator 40, who can judge whether the training date is valid or not, and only if it is valid the training data can be used for learning.

[0054] Figure 7 illustrates the user interface for know-how suggestion and evaluation in the plant operation support system according to this embodiment.

[0055] As already explained with reference to Figure 3, the know-how evaluation unit 303 evaluates the reliability of the inference results output by the real-time inference system 310. If the reliability of the inference is above a predetermined level, the inference result (for example, what operation should be performed) is presented to the operator 40 in the form of a know-how suggestion, along with the plant data used in the inference and the know-how (label) applied to it. For example, as illustrated in Figure 7, the name of the applicable know-how, the proposed operation (inference result), the date and time of the target plant data, and the registrant of the know-how are presented. The operator 40 evaluates the suggestion to determine whether it is appropriate or not, and enters the result into the know-how suggestion form.

[0056] If a proposed know-how is deemed invalid, as explained earlier, new know-how registration and training data registration will be performed using the plant data used in that reasoning process.

[0057] Figure 8 shows an example of the reliability assessment process performed by the know-how evaluation unit 303 on the inference results from the real-time inference system 310.

[0058] As shown in Figure 8, the real-time inference system 310 extracts plant data with the same labels (i.e., belonging to the same group) as the input real-time data used for inference from the set of time-series data (step S21). Then, it calculates the similarity between a predetermined feature of the extracted time-series data with the same labels and the feature of the input real-time data (S22). In this case, if training data has already been created for the time-series data, the operating conditions included in that training data (for example, "temperature sensor 1 is XX", "pressure sensor 1 is XXX", etc., as exemplified in Figures 5 and 6(B)) can be treated as components with a greater weight than other components among the features.

[0059] Subsequently, if the calculated similarity is higher than a predetermined level, the inference result is judged to be highly reliable; otherwise, it is judged to be unreliable (S23).

[0060] Figure 9 shows a modified configuration for performing step S5, "presentation of new know-how areas and registration of know-how," as shown in Figure 4.

[0061] As shown in Figure 9, the know-how registration unit 302 (which is part of the AI ​​training system 300 as shown in Figure 3) includes an operation evaluation method learning unit 322, an operation evaluation unit 324, and a know-how / teaching database 326. The operation evaluation method learning unit 322 is a system that learns operation evaluation methods, that is, methods for evaluating the quality of operations performed by the operator 40 on the plant system 20, i.e., operation actions (for example, scoring them as 60 out of 100 points), and creates an operation evaluation model that evaluates operation actions using that method. The operation evaluation unit 324 uses the operation evaluation model created by the operation evaluation method learning unit 322 to evaluate the operation actions performed by the operator 40, communicates the evaluation result (for example, a score of 60 out of 100 points) to the operator 40, and requests the operator 40 to register the know-how related to that operation action (especially the know-how related to operation actions with low evaluation results). The know-how / training database 326 is a system that stores know-how and training data registered by the operators 40.

[0062] As shown in Figure 9, the operation evaluation method learning unit 322 retrieves a dataset related to a specific type of past operation action (for example, operating a specific piece of equipment for a specific purpose) from a time-series database 320 (which may be located within or outside the system according to this embodiment) that stores operation data (time-series data) previously output from the plant system 20, and presents this dataset to the operator 40, requesting the operator 40 to evaluate the quality of the operation action shown in the dataset (a preliminary evaluation performed for learning the evaluation method). In response to this request, the operator 40 judges the quality of the operation action shown in the presented dataset and inputs (registers) the evaluation result into the operation evaluation method learning unit 322. For one type of operation action, the operation evaluation method learning unit 322 retrieves a large number of datasets from the time-series database 320 and learns how to evaluate that type of operation action to the required level by repeating the above-mentioned preliminary operation evaluation request and registration of operation evaluation results.

[0063] Furthermore, a plant contains numerous pieces of equipment, and there are many types of operating actions required for plant operation. For example, each piece of equipment may have operations such as startup, shutdown, and changes in operating conditions. An operating evaluation method learning unit 322 and an operating evaluation unit 324 may be provided for each of these types of operating actions. However, in the following explanation, we will focus on and describe the operating evaluation method learning unit 322 and the operating evaluation unit 324 related to one type of operating action.

[0064] The trained operation evaluation model created by the operation evaluation method learning unit 322 is deployed to the operation evaluation unit 324, thereby enabling the operation evaluation unit 324 to operate. The operation evaluation unit 324 receives operation data output from the plant system 20 (which may be past operation data stored in the time-series database 320, or current operation data (real-time data) obtained from the plant system 20), extracts a dataset from that operation data that shows a predetermined type of operation action for a predetermined piece of equipment, and evaluates that dataset (for example, assigning a score of 90 or 65). The operation evaluation unit 324 then presents the dataset and the evaluation result of that dataset (operation action) (for example, a score of 90 or 65) to the operator. In this case, the dataset is displayed in the form of a graph showing the change of each data value along the time axis, so that the operator 40 can easily grasp it visually. The timing of the operation evaluation unit 324's evaluation and the timing of presenting the evaluation results to the operator 40 may be a predetermined time or a selected time after the operation action being evaluated has been performed, or it may be almost simultaneous with the operation action (i.e., in real time). For example, both the evaluation of the operation action and the presentation of the results to the operator may be at a time convenient for the operator 40, such as a time selected by the operator 40 after the operation action has been completed. Alternatively, for example, the evaluation of the operation action may be performed in real time, almost simultaneously with the operation action, and the evaluation results may be temporarily stored in this system, and then presented to the operator 40 at a time convenient for the operator 40, such as a time selected by the operator 40.

[0065] When presenting the evaluation results to the driver 40, the driver 40 may be presented with all evaluated driving actions (datasets), regardless of whether the evaluation results are good or bad. Alternatively, only driving actions with evaluation results worse than a predetermined standard (for example, scores below 80 points) may be selected and presented to the driver 40. In any case, at the very least, driving actions with evaluation results worse than a predetermined standard are actions that the driver 40 is not good at, so it is desirable to register them as know-how to help the driver 40. Therefore, when the driving evaluation unit 324 notifies the driver 40 of driving actions with poor evaluation results, it requests the driver 40 to register the know-how for those actions.

[0066] In response to the registration request, the driver 40 can register know-how and training data related to their driving actions with the driving evaluation unit 324. The driving evaluation unit 324 stores the registered know-how and training data in the know-how / training database 326. The training data stored here is used to train the inference model or design the inference rules that make up the real-time inference system 310, as has already been explained with reference to Figures 3 and 4 (steps S7-S8 in Figure 4).

[0067] Figure 10 illustrates the flow of the process by which the aforementioned driving evaluation learning unit 322 learns the driving evaluation method. Figure 11 illustrates the flow of communication between the driving evaluation learning unit 22 and the driver 40 in the user interface during the learning of the driving evaluation method.

[0068] As shown in Figure 10, the operation evaluation learning unit 322 prompts the operator 40 to select the equipment and type of operation action within the plant system 20 that they wish to support (S31). For example, as shown in Figure 11, in S311, the operation evaluation learning unit 322 presents the operator 40 with a request message such as "Please select equipment and operation," and in response, in S312, the operator 40 specifies the equipment and type of operation action they wish to support, such as "Startup operation of the XXX reaction system."

[0069] Subsequently, as shown in Figure 10, the operation evaluation learning unit 322 prompts the operator 40 to select a set of data items that they want to display regarding the target device and the type of operation action (i.e., items that should be viewed to evaluate the quality of the operation action) (S32). For example, as shown in Figure 11, in S321, the operation evaluation learning unit 322 presents the operator 40 with a request message such as "Please select the items you want to display," and in response, in S322, the operator 40 specifies a set of data items that should be viewed for evaluation, such as "XX heater outlet gas temperature, XX catalyst first layer bottom temperature, XXX heater outlet gas temperature, XXX outlet gas pressure."

[0070] Subsequently, as shown in Figure 10, the operation evaluation learning unit 322 extracts one or more sets of data items specified by the operator 40 from the time-series database 320 with respect to the target equipment and operation action type specified by the operator 40, and presents these data sets to the operator 40 in graph format (S33). For example, as shown in Figure 11, in S331, the operation evaluation learning unit 322 creates a graph representing the change in data values ​​along the time axis of each extracted data set, and displays the graph on the user interface screen along with a request message such as "Please check if this graph is correct."

[0071] Subsequently, as shown in Figure 10, the driving evaluation learning unit 322 prompts the driver 40 to evaluate (score) each set of driving data displayed in graph format (i.e., each past driving action) (S34). For example, as shown in Figure 11, in S341, the driving evaluation learning unit 322 displays a request message such as "Please select past driving data and assign a score," and in response, in S342, the driver 40 inputs an evaluation score, such as 70 points or 20 points, for each graph (driving action).

[0072] Subsequently, as shown in Figure 10, the operation evaluation learning unit 322 uses the data sets (operation actions) presented to the operator 40 and the evaluation results (scores) of each data set (operation action) input by the operator 40 to create an operation evaluation model that has learned the operation evaluation method for the corresponding operation action type of the device (S35). Note that the evaluation method learned here does not necessarily have to be a detailed evaluation method such as assigning a two-digit score as exemplified, but a simpler evaluation method may also be used. As a simpler evaluation method, for example, there may be a two-stage evaluation method of good or bad. In the case of this two-stage evaluation method, it is also possible to learn using only the data sets that have obtained one of the evaluation results, good (or bad). For example, if learning is done using only the data sets that have received a good evaluation, it is possible to create an operation evaluation model that calculates the distance in the data space from the learned good data sets and outputs a score corresponding to that distance or simply a judgment of good or bad as the evaluation result.

[0073] Using the driving evaluation model created through the learning process described above, the driving evaluation unit 324 evaluates the driving actions performed by the driver 40 in real time. Figure 12 illustrates the flow of the driving evaluation process performed by the driving evaluation unit 324. Figure 13 illustrates the flow of communication between the driving evaluation unit 324 and the driver 40 in the user interface at that time.

[0074] As shown in Figure 12, the operation evaluation unit 324 receives real-time operation data (real-time data) from the plant system 20 and extracts a dataset of selected data items from the real-time data relating to the target equipment and operation action type (S41). Then, the operation evaluation unit 324 evaluates (scores) the extracted dataset (i.e., the operation action just performed) using a learned evaluation method (S42).

[0075] The driving evaluation unit 324 then determines whether the evaluation result meets predetermined selection conditions (for example, the scored score is lower than a certain standard value, meaning the driving action performed is worse than a certain standard), and if the condition is met, it presents the evaluated dataset and its evaluation result to the driver 40 (S43). The dataset is displayed in a format that is easy for the driver 40 to understand visually, for example, by showing the changes in data values ​​along a time axis. The driving evaluation unit 324 then asks the driver 40 to specify what the data state that serves as the indicator (basis) for the evaluation result is on the displayed dataset graph (S44).

[0076] In S43 above, only low evaluation results are selectively presented to the operator 40, but this is not necessarily required. Alternatively, all evaluation results, regardless of whether they are high or low, may be presented to the operator 40, or only high evaluation results may be selectively presented to the operator 40.

[0077] An example of the user interface display in the process S43-S44 of Figure 12 described above is shown in S431-S442 of Figure 13. (Note that the operator 40 who responds to the display message shown in Figure 13 may be a different person from the operator 40 who performed the evaluated operation action.) For example, if operator 40 performs a startup operation of the target equipment, and the operation evaluation unit 324 evaluates the startup operation as being below a certain standard, the operation evaluation unit 324 displays a message to operator 40 in S431 of Figure 13 indicating the low evaluation result, such as, "Today's startup operation is worth 65 points. Is this evaluation score correct?" In response to this message, if the driver 40 agrees to the evaluation result, such as by answering "Yes" in S432, the driving evaluation unit 324 displays a graph of the evaluated dataset to the driver 40, as shown in S442, and also displays a message such as "Please tell us where points were deducted," as shown in S441, requesting the driver 40 to specify which part of the graph corresponds to the data state that serves as an indicator of the evaluation result. In response to this request, the driver 40 places an indicator mark 51 on the part of the graph corresponding to the data state of the above indicator (in this example, the part where a certain data value changed abruptly), as shown in S442. The driving evaluation unit 324 stores the data state of the part where this designated mark 51 is placed as an indicator of the evaluation result (low evaluation result).

[0078] Subsequently, as shown in Figure 12, the operation evaluation unit 324 prompts the operator 40 to input detailed information about what the data state means and the correct operation that should have been performed when that data state occurred (S45). For example, as shown in Figure 13, in S451, the operation operation unit 324 presents the operator 40 with a request message such as, "Please tell me what the state is and what operation I should perform." In response to this request, the operator 40 inputs information in S452 such as, "State 'XX heater outlet temperature' is 'dropping rapidly'. State 'there is a possibility of the system stopping.' Operation 'Please reduce the XXXX operation range'." The operation evaluation unit 324 stores this input information as the content of a suggestion message that the real-time inference unit 310 (see Figure 3) will present to the operator 40.

[0079] Subsequently, as shown in Figure 12, the operation evaluation unit 324 extracts one or more (usually many) datasets corresponding to the target equipment and operation action type from the time-series database 320, and presents each dataset to the operator 40 in the same graph format as above (S46). The operation evaluation unit 324 then requests the operator 40 to specify a data state from among the graphs of these datasets that serves as an indicator of an evaluation result (high evaluation result) in contrast to the previous evaluation result (low evaluation result) (S47). For example, as shown in Figure 13, in S471, the operation control unit 324 displays a message to the operator 40 such as "Please tell me what kind of operation you should aim for," and in S472, displays graphs of the one or more (usually many) extracted datasets. In response, as shown in S472, the operator 40 selects a data state from the displayed graph that serves as an indicator of an operation action to aim for (highly evaluated), and specifies the data state corresponding to that indicator by placing an indicator mark 53 at that location. The operation evaluation unit 324 stores the specified data state as an indicator of the contrasting evaluation result (high evaluation result).

[0080] Subsequently, as shown in Figure 12, the operation evaluation unit 324 prompts the operator 40 to input know-how names for the data states that serve as indicators of low and high evaluations and the operation actions to be taken, which the operator 40 has input in S44 to S47 (S48) (for example, S481 to S482 in Figure 13), and stores this input information as training data in the know-how / training database 326, associated with the know-how names (S49).

[0081] As already explained, the training data stored in the know-how / training database 326 is used to create the inference model that constitutes the real-time inference system 310 (see Figure 3). For example, if the inference model is constructed using a machine learning-capable neural network, the training data in the know-how / training database 326 can be used as training data for that machine learning. Alternatively, if the inference model is constructed using a von Neumann-type system that performs inference using pre-programmed inference rules, the training data in the know-how / training database 326 can be used for programming, or designing, those inference rules.

[0082] The real-time inference system 310, thus created, receives real-time operating data (real-time data) from the plant system 20 and extracts from that real-time data a dataset of selected data items related to the type of operating action of the target equipment, which includes data states corresponding to indicators of a low evaluation result. The real-time inference system 310 then identifies the data states corresponding to indicators of a low evaluation from the extracted dataset, identifies what those data states mean and what operations should be performed regarding that dataset, and uses these identified matters to create a suggestion message and present it to the operator 40.

[0083] Figure 14 shows an example of a suggestion message presented by the real-time inference system 310 to the driver 40. As shown in Figure 14, the extracted dataset is displayed in graph format, and an indicator mark 55 is displayed on the graph at the location of data states corresponding to low-rated indicators. For that dataset, the target equipment and the type of operating action are displayed, such as "Equipment: XXX reaction system. Operation: Startup." Furthermore, a description of the indicator's data state and detailed information on the operation to be performed are displayed, such as "Status: XX heater outlet temperature is dropping rapidly. Suggestion: Reduce the XXXX operation range."

[0084] As described in detail above, the plant operation support system 30 of this embodiment makes it possible to realize a plant operation support system and a learning method in a plant operation support system that allows the system to select what knowledge to learn and request it from a human as it progresses through the learning process.

[0085] The above-described embodiments are provided to illustrate the present invention in detail for clarity, and are not necessarily limited to those comprising all the described components. Furthermore, some components of each embodiment can be added, deleted, or replaced with other components.

[0086] Furthermore, each of the above-mentioned configurations, functions, processing units, processing means, etc., may be implemented in hardware, either partially or entirely, by designing them as integrated circuits, for example. The present invention can also be implemented by software program code that realizes the functions of the embodiment. In this case, a storage medium on which the program code is recorded is provided to a computer, and the processor of that computer reads the program code stored in the storage medium. In this case, the program code read from the storage medium itself realizes the functions of the embodiment described above, and the program code itself and the storage medium on which it is stored constitute the present invention. Examples of storage media used to supply such program code include flexible disks, CD-ROMs, DVD-ROMs, hard disks, SSDs (Solid State Drives), optical disks, magneto-optical disks, CD-Rs, magnetic tapes, non-volatile memory cards, ROMs, and the like.

[0087] Furthermore, the program code that implements the functions described in this embodiment can be implemented in a wide range of programming or scripting languages, such as assembler, C / C++, Perl, Shell, PHP, Java®, and Python.

[0088] Furthermore, all or part of the program code of the software that implements the functions of each embodiment may be stored in the computer's memory resources in advance, or, if necessary, may be stored in the computer's memory resources from a non-temporary storage device of another device connected to the network, or from a non-temporary storage medium via an external I / F (not shown) provided by the computer.

[0089] Furthermore, the program code for the software that implements the functions of the embodiment may be distributed via a network and stored in a storage means such as a computer's hard disk or memory, or in a storage medium such as a CD-RW or CD-R, and the computer's processor may read and execute the program code stored in the storage means or storage medium.

[0090] In the above-described embodiment, the control lines and information lines shown are those deemed necessary for explanation and do not necessarily represent all control lines and information lines in the actual product. All components may be interconnected. [Explanation of symbols]

[0091] 20...Plant System 30...Plant Operation Support System 40...Operator 300...AI Training System 301...Data Classification Unit 302...Knowledge Registration Unit 303...Knowledge Evaluation Unit 310...Real-time Inference System 320...Time Series Database 322...Operation Evaluation Method Learning Unit 324...Operation Evaluation Unit 326...Knowledge / Trainer Database

Claims

1. An inference system that accepts plant data generated at a plant, performs inference using the accepted plant data, and outputs suggestions for operational procedures as an inference result, A training system that uses the aforementioned plant data to input training data created with the involvement of the plant operators, and uses the input training data to incorporate plant operation know-how into the inference system, thereby training the inference system. Equipped with, The training system has operator utilization means that selects specific plant data from given plant data that relates to specific know-how that is desirable to be incorporated into the inference system, and requests the operator to participate in creating new training data based on the selected specific plant data. The aforementioned means of using the operator An evaluation method learning means that receives evaluation information regarding an evaluation method for evaluating the plant data from the operator and learns the evaluation method using the evaluation information, An evaluation means that evaluates the given plant data using the learned evaluation method, selects the plant data from which evaluation results satisfying predetermined conditions have been obtained as the specific plant data, presents the selected specific plant data to the operator, and requests the operator to participate in creating training data using the specific plant data. has Plant operation support system.

2. In the plant operation support system according to claim 1, The operator access means is configured to determine the reliability of the inference result, and if the reliability is higher than a predetermined level, to propose the inference result to the operator. If the operator evaluates the proposed inference result as unreasonable, the operator is required to use the plant data used in the unreasonable inference result to request the operator to participate in creating training data. Plant operation support system.

3. In the plant operation support system according to either claim 1 or 2, The operator access means is configured to determine the reliability of the inference result, and if the reliability is lower than a predetermined level, to request the operator to participate in creating training data using the plant data used in the inference result that was evaluated as having low reliability. Plant operation support system.

4. In the plant operation support system according to any one of claims 1 to 3, The operator access means is configured to collect historical environmental data that is highly similar to the environmental data used to create the training data, and to create additional training data using the collected historical environmental data. Plant operation support system.

5. In the plant operation support system according to any one of claims 1 to 4, The operator utilization means is configured to evaluate the similarity of the plant data input to the inference system and, depending on the similarity evaluation result, to selectively request the operator to participate in the creation of training data. Plant operation support system.

6. In the plant operation support system according to any one of claims 1 to 5, The operator utilization means is configured to provide the operator with the inference results output from the inference system, receive feedback from the operator regarding the inference results, and selectively request the operator to participate in the creation of training data in response to the feedback. Plant operation support system.

7. In the plant operation support system according to claims 1 to 6, The operator access means is configured to have a set of plant data generated in the past, to select a type of plant data specified by the operator from the set, and to receive the evaluation information related to the selected plant data from the operator. Plant operation support system.

8. A computer system comprising a storage device that stores a computer program and a CPU that mechanically reads and executes the computer program, and capable of communicating with a plant and its operators, The CPU implements the plant operation support method by executing the computer program, and the plant operation support method is An inference process that receives plant data generated at the plant, performs inference using the received plant data, and outputs a suggestion for operation as an inference result, A training process is performed to train the inference process by inputting training data created with the involvement of operators using the aforementioned plant data, and incorporating plant operation know-how into the inference process using the input training data. Equipped with, The aforementioned training process includes an operator utilization process in which the operator selects specific plant data from given plant data that relates to specific know-how that is desirable to be incorporated into the inference process, and requests the operator to participate in creating new training data based on the selected plant data. The aforementioned operator utilization process is, A step of inputting evaluation information regarding an evaluation method for evaluating the plant data from the operator, and learning the evaluation method using the evaluation information, The process includes evaluating the given plant data using the learned evaluation method, selecting the plant data from which evaluation results satisfying predetermined conditions are obtained as the specific plant data, presenting the selected specific plant data to the operator, and requesting the operator to participate in creating training data using the specific plant data; including A computer system that implements plant operation support methods.

9. In a program storage medium containing machine-readable computer programs, The program includes instruction code that causes the processor to perform a plant operation support method, The aforementioned plant operation support method An inference process that receives plant data generated at the plant, performs inference using the received plant data, and outputs suggestions for operation as an inference result, A training process is performed to train the inference process by inputting training data created with the involvement of operators using the aforementioned plant data, and incorporating plant operation know-how into the inference process using the input training data. Equipped with, The aforementioned training process includes an operator utilization process that selects specific plant data from given plant data that relates to specific know-how that is desirable to be incorporated into the inference process, and requests a human to participate in creating new training data based on the selected plant data. The aforementioned operator utilization process is, A step of inputting evaluation information regarding an evaluation method for evaluating the plant data from the operator, and learning the evaluation method using the evaluation information, The process includes evaluating the given plant data using the learned evaluation method, selecting the plant data from which evaluation results satisfying predetermined conditions are obtained as the specific plant data, presenting the selected specific plant data to the operator, and requesting the operator to participate in creating training data using the specific plant data; including Program recording medium.

Images