WO2024116622A1 - Determination device, determination method, program, and plant system - Google Patents

Determination device, determination method, program, and plant system Download PDF

Info

Publication number
WO2024116622A1
WO2024116622A1 PCT/JP2023/037446 JP2023037446W WO2024116622A1 WO 2024116622 A1 WO2024116622 A1 WO 2024116622A1 JP 2023037446 W JP2023037446 W JP 2023037446W WO 2024116622 A1 WO2024116622 A1 WO 2024116622A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
sensor
prediction target
plant
target sensor
Prior art date
Application number
PCT/JP2023/037446
Other languages
French (fr)
Japanese (ja)
Inventor
英里子 ▲高▼▲崎▼
正法 門脇
Original Assignee
住友重機械工業株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 住友重機械工業株式会社 filed Critical 住友重機械工業株式会社
Publication of WO2024116622A1 publication Critical patent/WO2024116622A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring

Definitions

  • the present invention relates to a determination device for determining abnormalities in a plant, a determination method, a program for the determination device, and a plant system.
  • sensors are used to collect data such as temperature and pressure from various parts during operation, and the data is constantly monitored to ensure that it is within normal values. If some of the data becomes abnormal, the cause of the abnormality is inferred based on the position of the corresponding sensor, and appropriate measures are taken to return the plant to a normal state.
  • a predicted value for the data is calculated depending on the situation, and it is determined how much the actual measured value of the data deviates from the calculated predicted value. Note that the "predicted value” here refers to the value predicted to be indicated by the specific data when it is assumed that the entire plant is operating normally.
  • the predicted value for each data item can be calculated, for example, using the actual measured values of multiple other data items that affect the data item in question and a statistical model.
  • the statistical model is a model (e.g., a mathematical formula) that indicates the correlation between multiple data items, and is created based on each data item that has been acquired in advance under normal conditions.
  • the present invention aims to provide a determination device, a determination method, a program for the determination device, and a plant system that can accurately determine abnormalities in a plant.
  • the determination device is a determination device that determines the presence or absence of an abnormality in a plant, and includes a data acquisition unit that acquires data from multiple sensors installed in the plant, and a determination unit that determines the presence or absence of an abnormality based on the data acquired by the data acquisition unit and a statistical model that indicates the correlation between the multiple data.
  • the multiple sensors installed in the plant include a prediction target sensor.
  • the determination unit of the determination device calculates a predicted value of the data acquired from the prediction target sensor using data acquired from a sensor among the multiple sensors that is installed in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor, and determines the presence or absence of an abnormality based on the predicted value and the actual measured value of the data acquired from the prediction target sensor.
  • a predicted value of data acquired by a sensor to be predicted is calculated using data acquired from a sensor that is provided in the same system as the sensor to be predicted and that measures the same type of physical quantity as the sensor to be predicted.
  • the statistical model used to calculate the predicted value indicates the relationship between data consisting of the same system and type of physical quantity, and is therefore relatively simple and easier to express in mathematical expressions than when multiple types of physical quantities are included. For this reason, even if the amount of data used to train the statistical model is small, it is possible to calculate a predicted value with higher accuracy than in the past.
  • the difference between the actual measured value acquired by the sensor to be predicted and the predicted value calculated as a normal value becomes significant, making it possible to accurately determine the presence or absence of an abnormality based on both.
  • the present invention provides a determination device, a determination method, a program for the determination device, and a plant system that can accurately determine abnormalities in a plant.
  • FIG. 1 is a diagram illustrating a schematic diagram of an overall configuration of a plant system according to an embodiment.
  • FIG. 2 is a diagram showing a plurality of sensors provided in a plant, grouped by system.
  • FIG. 3 is a diagram for explaining the functions of a determination unit and the like included in the determination device.
  • FIG. 4 is a flowchart showing the flow of the process executed by the determination device.
  • FIG. 5 is a diagram showing an example of information displayed on the screen of the monitoring terminal.
  • the determination device 10 constitutes part of the plant system PS, and is a device for determining abnormalities in the plant 11, which will be described later.
  • An "abnormality" in the plant 11 includes a state in which the plant 11 cannot operate due to a physical failure, such as a pipe burst, or a state in which the plant 11 is likely to shut down.
  • the above-mentioned “abnormality” also includes a deviation from a good operating state without a physical failure, such as a significant decrease in the operating efficiency of the plant 11 for some reason.
  • the plant system PS includes a plant 11, a DCS 12, a determination device 10, and a monitoring terminal 20.
  • the plant 11 in this embodiment is a power generation plant including a boiler (not shown). In the boiler, steam is generated by the heat of high-temperature exhaust gas, and the steam drives the turbine of the generator. When the plant 11 is in operation, data (temperature, pressure, etc.) of each part in the exhaust gas system, water system, etc. is constantly monitored, and control is performed so that each data maintains a normal value.
  • the plant 11 may be a power generation plant as in this embodiment, but it may also be another type of plant. For example, it may be an incineration plant or a chemical plant.
  • DCS 12 is also called a “Distributed Control System” and is a device that controls plant 11.
  • DCS 12 controls a part of plant 11 (e.g., a boiler), but may also control the entire plant 11.
  • DCS 12 transmits control signals to each of the various devices (e.g., blowers) installed in plant 11, and controls the operation of plant 11.
  • DCS 12 also receives signals from each of a number of sensors installed in various parts of plant 11, and acquires the status of plant 11.
  • the table shown in FIG. 2 groups the multiple sensors installed in plant 11 by system.
  • a "system” here refers to a series of paths through which the same type of medium flows, such as an exhaust gas system, a water system, an electric power system, etc.
  • the plant 11 has three systems, system A, system B, and system C, and each system is provided with multiple sensors.
  • a total of eight sensors, S11 to S18, are provided in each section of system A.
  • S11, 12, and 13 are sensors for measuring pressure
  • S14, 15, and 16 are sensors for measuring temperature
  • S17 and 16 are sensors for measuring flow rate.
  • the number and types of sensors (S21, etc.) shown in system B and system C shown in FIG. 2 are also as shown in the same figure. Note that the number of sensors actually provided in each system is greater than the number shown in FIG. 2.
  • the determination device 10 is a device for determining abnormalities in the plant 11.
  • the determination device 10 is configured as a computer system having a CPU, ROM, RAM, etc. (not shown), and is installed, for example, on the same premises as the plant 11.
  • the determination device 10 may be configured, for example, as a cloud server located at a location different from the plant 11.
  • the determination device 10 can communicate bidirectionally with the DCS 12.
  • the determination device 10 acquires the actual measured values of each sensor installed in the plant 11 by communication via the DCS 12.
  • the determination device 10 may directly receive signals from each sensor.
  • the monitoring terminal 20 is a terminal device provided as an interface with the user.
  • the user can input information required for the operational settings of the determination device 10 by operating the monitoring terminal 20.
  • the monitoring terminal 20 can also notify the user of the results of the determination made by the determination device 10 by displaying various information on the screen 200.
  • the monitoring terminal 20 may be a stationary terminal installed in a specific location, or it may be a portable communication terminal that can be carried by the user.
  • the monitoring terminal 20 can also be considered as part of the determination device 10.
  • the determination device 10 includes, as elements that represent its functions, a data acquisition unit 110, a selection unit 120, a determination unit 130, a notification unit 140, and a storage unit 150.
  • the data acquisition unit 110 is a part that performs processing to acquire data from multiple sensors installed in the plant 11.
  • the data acquisition unit 110 repeatedly acquires data from all sensors installed in the plant 11 through communication with the DCS 12.
  • the selection unit 120 is a part that performs a process of selecting data to be used in the process of the determination unit 130 from among the multiple pieces of data acquired by the data acquisition unit 110. The specific content of the process performed by the selection unit 120 will be explained later.
  • the determination unit 130 is a part that performs processing to determine the presence or absence of an abnormality in the plant 11 based on the data acquired by the data acquisition unit 110 (specifically, the data selected by the selection unit 120) and a statistical model.
  • a "statistical model” is a model that indicates the correlation between multiple pieces of data acquired by each sensor, and is expressed, for example, as a mathematical formula.
  • the statistical model is created in advance based on the data acquired from each sensor when the entire plant 11 is operating normally, and is stored in the storage unit 150 described below. The statistical model may be learned and updated each time the plant 11 is operating normally.
  • the judgment unit 130 can, for example, input data measured by multiple sensors into a statistical model and calculate, as its output, a predicted value of the data acquired by a specific sensor.
  • the "predicted value” calculated in this way is the value predicted to be indicated by the specific sensor, assuming that the entire plant 11 is operating normally. Therefore, if the actual measurement value actually acquired by the sensor significantly deviates from the calculated predicted value, the judgment unit 130 can determine that an abnormality has occurred in the plant 11.
  • the notification unit 140 is a part that performs the process of notifying the user of the judgment result by the judgment unit 130.
  • the notification unit 140 notifies the user of the judgment result by displaying it on the screen 200 of the monitoring terminal 20. The specific manner in which the notification unit 140 notifies the user will be described later.
  • the storage unit 150 is a non-volatile storage device provided in the determination device 10, such as an HDD or SSD.
  • the storage unit 150 may be a file server installed in a location different from the determination device 10.
  • the storage unit 150 stores various information necessary for the processing performed by the determination device 10, including the statistical model described above.
  • a statistical model is created individually for each sensor installed in the plant 11, and each statistical model is stored in the memory unit 150.
  • Each statistical model stored in the memory unit 150 will be referred to below as a "statistical model 151.”
  • FIG. 3 shows a schematic diagram of the flow of information in each part of the determination device 10.
  • the determination unit 130 selects any one sensor from the multiple sensors as the prediction target sensor, and calculates the above-mentioned "prediction value" for the prediction target sensor. Thereafter, the actual measurement value for the prediction target sensor is compared with the predicted value to determine whether or not there is an abnormality.
  • the data acquisition unit 110 acquires data from all sensors, including the prediction target sensor, and sends this to the selection unit 120. From this data, the selection unit 120 selects data acquired from sensors that are installed in the same system as the prediction target sensor and measure the same type of physical quantity as the prediction target sensor. For example, when S14 in FIG. 2 is the prediction target sensor, the selection unit 120 selects S15 and S16, which are installed in the same system A and measure the same "temperature" as S14. The selection unit 120 sends the data selected in this way to the prediction value calculation unit 131, which is part of the determination unit 130. In FIG. 3, the data sent in this way is shown as "d2".
  • the selection unit 120 sends the actual measurement data acquired by the prediction target sensor, among the data sent from the data acquisition unit 110, to the anomaly degree calculation unit 132, which is part of the judgment unit 130.
  • the data sent in this way is shown as "d1".
  • the prediction value calculation unit 131 reads out a statistical model 151 corresponding to the sensor to be predicted from among the multiple statistical models 151 stored in the storage unit 150.
  • This statistical model 151 is provided in the same system as the sensor to be predicted and was created in advance as a model that indicates the correlation between data acquired from a sensor that measures the same type of physical quantity as the sensor to be predicted (i.e., d2 in FIG. 3) and data measured by the sensor to be predicted.
  • the predicted value calculation unit 131 calculates a predicted value of the data measured by the prediction target sensor based on the statistical model 151 and the data (d2) selected by the selection unit 120, and sends this to the anomaly calculation unit 132. In FIG. 3, the predicted value sent in this manner is shown as "d1'".
  • both the actual measured value d1 of the data acquired by the sensor to be predicted and the predicted value d1' of the data to be measured by the sensor to be predicted are sent to the anomaly degree calculation unit 132.
  • the predicted value d1' is calculated under the assumption that the entire plant 11 is operating normally. Therefore, if no abnormality occurs in the plant 11, the actual measurement value d1 will be roughly equal to the predicted value d1', and the difference between the two will be small. On the other hand, if an abnormality occurs in the plant 11, the actual measurement value d1 will deviate from the predicted value d1', and the difference between the two will be large. Therefore, the difference between the actual measurement value d1 and the predicted value d1' indicates the degree of abnormality occurring in the plant 11.
  • the anomaly degree calculation unit 132 calculates the "anomaly degree", which is an index showing the degree of anomaly occurring in the plant 11.
  • the anomaly degree is calculated as a normalized value of the difference between the actual measured value d1 and the predicted value d1'.
  • the anomaly degree is calculated as a value in the range from 0 to 1. In this case, if the actual measured value d1 and the predicted value d1' match, the anomaly degree is 0.
  • the anomaly degree increases as the difference between the actual measured value d1 and the predicted value d1' increases, and once the difference reaches a certain level, the anomaly degree is calculated as 1.
  • the degree of abnormality calculated by the abnormality degree calculation unit 132 is sent to the notification unit 140.
  • the notification unit 140 notifies the user by displaying the degree of abnormality on the screen 200 of the monitoring terminal 20.
  • the judgment unit 130 of the judgment device 10 is configured to calculate a predicted value (d1') of the data acquired from the prediction target sensor using data acquired from a sensor, among multiple sensors installed in the plant 11, that is installed in the same system as the prediction target sensor and that measures the same type of physical quantity as the prediction target sensor, and to judge the presence or absence of an abnormality based on the predicted value (d1') and the actual measured value (d1) of the data acquired from the prediction target sensor.
  • the statistical model 151 used to calculate the predicted value is created to show the relationship between data consisting of the same system and type of physical quantity as the sensor to be predicted.
  • the statistical model 151 is created as a model that does not include data measured by a sensor installed in a system different from the sensor to be predicted, or data of physical quantities of a different type than those measured by the sensor to be predicted.
  • the statistical model 151 of this embodiment is relatively simple, since it is easier to express it in a mathematical formula, etc., compared to when multiple types of physical quantities are included.
  • the data d2 selected by the selection unit 120 and sent to the prediction value calculation unit 131 may include data on actual measurements obtained from the sensor to be predicted.
  • the statistical model 151 is updated each time by deep learning, it is preferable to create a statistical model that includes data on actual measurements from the sensor to be predicted, and then use this to calculate the prediction value.
  • the judgment unit 130 of this embodiment can be said to calculate a predicted value of the data acquired from the prediction target sensor using only data acquired from a sensor among multiple sensors that is installed in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor, or using the data acquired from the prediction target sensor plus data measured by the prediction target sensor, and to judge the presence or absence of an abnormality based on the predicted value and the actual measured value of the data acquired from the prediction target sensor.
  • the specific flow of the processing executed by the determination device 10 will be described with reference to the flowchart in FIG. 4.
  • the series of processing shown in the figure is repeatedly executed by the determination device 10 every time a predetermined period elapses during the period during which the plant 11 is in operation.
  • step ST1 the data acquisition unit 110 performs a process of acquiring data from all sensors installed in the plant 11.
  • the process performed in step ST1 corresponds to the "acquisition process” and "data acquisition process” in this embodiment.
  • step ST2 a process of determining one sensor to be predicted from among the sensors installed in the plant 11 is performed by, for example, the determination unit 130.
  • each of the multiple sensors installed in the plant 11 is determined one by one in order as the sensor to be predicted, and the presence or absence of an abnormality is determined based on a comparison between the actual measurement value and the predicted value at each sensor to be predicted.
  • the order in which each sensor is determined as the sensor to be predicted can be set arbitrarily.
  • step ST4 following step ST3, the prediction value calculation unit 131 in FIG. 3 reads out from the storage unit 150 the statistical model 151 corresponding to the sensor to be predicted.
  • step ST5 the prediction value calculation unit 131 calculates a prediction value for the data measured by the sensor to be predicted based on the statistical model 151 read out in step ST4 and the data for calculating the prediction value selected in step ST3 (d2 in FIG. 3).
  • step ST6 the process of calculating the degree of abnormality is performed by the degree of abnormality calculation unit 132 in FIG. 3.
  • the degree of abnormality calculation unit 132 calculates the degree of abnormality based on a comparison between the actual measurement value data (d1 in FIG. 3) acquired by the sensor to be predicted and the predicted value (d1' in FIG. 3) calculated in step ST5.
  • the process performed in step ST6 corresponds to the "determination step” and "determination process” in this embodiment.
  • step ST7 following step ST6 the notification unit 140 performs a process of displaying the degree of abnormality on the screen 200 of the monitoring terminal 20.
  • step ST8 following step ST7, it is determined whether the calculation and reporting of the degree of anomaly has been completed for all sensors installed in the plant 11. If the calculation and reporting has been completed for all sensors, the series of processes shown in FIG. 4 ends. If there is a sensor that has not yet been selected as the sensor to be predicted, the process proceeds to step ST9. In step ST9, one of the sensors that has not yet been selected as the sensor to be predicted is selected and determined to be the next sensor to be predicted. Thereafter, the processes from step ST3 onwards are executed again.
  • the judgment unit 130 will individually judge the presence or absence of an abnormality for each of the multiple sensors installed in the plant 11 when that sensor is set as the sensor to be predicted. Furthermore, the notification unit 140 will individually notify the judgment result corresponding to each of the multiple sensors. In other words, for each of the multiple sensors, the degree of abnormality when that sensor is set as the sensor to be predicted is calculated and notified. Note that all sensors may be set as the sensors to be predicted as in this embodiment, but only some of the sensors installed in the plant 11 may be set as the sensors to be predicted.
  • time series graphs 201, 202, and 203 and an anomaly distribution diagram 210 are displayed on the screen 200 of the monitoring terminal 20.
  • Time series graphs 202 and 203 are graphs that display the change in the degree of anomaly calculated for sensors S14 and S15 in FIG. 2 over time.
  • Time series graph 201 is a graph that displays the difference (temperature difference) between the data measured by sensors S14 and S15 over time.
  • the sensors to be displayed in the time series graph may be sensors other than those mentioned above.
  • the data to be displayed in the time series graph may be arbitrarily set by the user through operations performed on the monitoring terminal 20.
  • the anomaly distribution diagram 210 displays a number of icons 211 representing sensors provided in each part on a configuration diagram showing the entire plant 11.
  • the positions at which the icons 211 are displayed correspond to the positions at which the sensors are provided in the plant 11.
  • the color of the icon 211 represents the value of the anomaly level calculated by the judgment unit 130 for the corresponding sensor.
  • the icon 211 is displayed in a color selected from a gradation in which the closer the calculated anomaly level for the corresponding sensor is to 0, the closer it is to blue, and the closer the anomaly level is to 1 or -1, the closer it is to red.
  • the color of each icon 211 is updated during the processing of the subsequent step ST7 according to the anomaly level calculated in step ST6 of FIG. 4.
  • the notification unit 140 of this embodiment individually notifies the judgment result corresponding to each sensor by changing the color of each icon 211 displayed in the anomaly distribution diagram 210.
  • the icons 211 displayed on the anomaly distribution diagram 210 may be displayed only for the sensors corresponding to the data selected by the selection unit 120 (i.e., the data used to calculate the predicted value), or may be displayed for all sensors installed in the plant 11.
  • the icon 211 may represent the degree of abnormality with a color as described above, but may also display the result of the determination of whether the degree of abnormality has exceeded a predetermined threshold using, for example, two different colors.
  • the notification unit 140 may notify the presence or absence of an abnormality in each part of the plant 11, including the degree of the abnormality (i.e., the degree of abnormality), or may simply notify the presence or absence of an abnormality.
  • the above-mentioned determination method performed by the determination device 10 is realized, for example, by a program stored in a non-volatile storage device (not shown) provided in the determination device 10.
  • the aforementioned "data acquisition process” and “determination process” are executed by operating the determination unit 130 of the determination device 10 in accordance with the program.
  • a part or all of the program may be transmitted to the determination device 10 from an external device and temporarily written to the storage device of the determination device 10.
  • a determination device for determining whether or not an abnormality exists in a plant comprising: a data acquisition unit that acquires data from a plurality of sensors provided in the plant; and a determination unit that determines whether or not an abnormality exists based on the data acquired by the data acquisition unit and a statistical model that indicates a correlation between the plurality of data, wherein the plurality of sensors includes a prediction target sensor, and the determination unit calculates a predicted value of the data acquired from the prediction target sensor using data acquired from a sensor among the plurality of sensors that is provided in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor, and determines whether or not an abnormality exists based on the predicted value and an actual measurement value of the data acquired from the prediction target sensor.
  • (Appendix 2) 2. The determination device according to claim 1, further comprising a selection unit that selects, from the plurality of data acquired by the data acquisition unit, data acquired from a sensor that is provided in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor. (Appendix 3) 3. The determination device according to claim 1, wherein the data used to calculate the predicted value includes data acquired from a sensor to be predicted. (Appendix 4) 4. The determination device according to claim 1, further comprising a notification unit that notifies a result of the determination by the determination unit.
  • a method for determining whether or not an abnormality exists in a plant comprising: an acquisition step of acquiring data from a plurality of sensors provided in the plant; and a determination step of determining whether or not an abnormality exists based on the data acquired in the acquisition step and a statistical model indicating a correlation between the plurality of data, wherein the plurality of sensors includes a prediction target sensor; and in the determination step, a prediction value of the data acquired from the prediction target sensor is calculated using data acquired from a sensor among the plurality of sensors that is provided in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor, and the presence or absence of an abnormality is determined based on the prediction value and an actual measurement value of the data acquired from the prediction target sensor.
  • a program for a determination device that determines whether or not an abnormality exists in a plant the program causing the determination device to perform a data acquisition process that acquires data from a plurality of sensors provided in the plant, and a determination process that determines whether or not an abnormality exists based on the data acquired in the data acquisition process and a statistical model that indicates a correlation between the plurality of data, wherein the plurality of sensors includes a prediction target sensor, and the program causes the determination device to perform a determination process in which a predicted value of data acquired from the prediction target sensor is calculated using data acquired from a sensor among the plurality of sensors that is provided in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor, and determines whether or not an abnormality exists based on the predicted value and an actual measurement value of the data acquired from the prediction target sensor.
  • a plant system comprising a plant and a determination device that determines whether or not an abnormality occurs in the plant, the determination device having a data acquisition unit that acquires data from a plurality of sensors provided in the plant, and a determination unit that determines whether or not an abnormality exists based on the data acquired by the data acquisition unit and a statistical model that indicates a correlation between the plurality of data, wherein the plurality of sensors includes a prediction target sensor, and the determination unit calculates a predicted value of the data acquired from the prediction target sensor using data acquired from a sensor among the plurality of sensors that is provided in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor, and determines whether or not an abnormality exists based on the predicted value and an actual measurement value of the data acquired from the prediction target sensor.
  • Determination device 11 Plant 110: Data acquisition unit 120: Selection unit 130: Determination unit 140: Notification unit PS: Plant system

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Automation & Control Theory (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

This determination device (10) comprises: a data acquisition unit 110 that acquires data from a plurality of sensors provided to a plant 11; and a determination unit 130 that, on the basis of the data acquired by the data acquisition unit 110 and a statistical model 151 indicating correlations between a plurality of items of data, determines whether an abnormality is present. The plurality of sensors include a prediction target sensor. The determination unit 130 calculates, using data acquired from a sensor among the plurality of sensors that is provided to the same group as the prediction target sensor and measures the same physical quantity as the prediction target sensor, a prediction value pertaining to the data acquired from the prediction target sensor, and determines whether an abnormality is present on the basis of the prediction value and the actual value of the data acquired from the prediction target sensor.

Description

判定装置、判定方法、プログラム、及びプラントシステムDetermination device, determination method, program, and plant system
　本発明は、プラントにおける異常を判定する判定装置、判定方法、判定装置用のプログラム、及びプラントシステムに関する。 The present invention relates to a determination device for determining abnormalities in a plant, a determination method, a program for the determination device, and a plant system.
　例えば発電プラントや焼却プラントのようなプラントにおいては、運転中における各部の温度や圧力等のデータをセンサで取得しながら、各データが正常値となっているか否かを常に監視している。一部のデータが異常値となった場合には、対応するセンサの位置等に基づいて異常の原因が推定され、正常な状態に戻すための適切な対応がとられる。 For example, in plants such as power plants and incineration plants, sensors are used to collect data such as temperature and pressure from various parts during operation, and the data is constantly monitored to ensure that it is within normal values. If some of the data becomes abnormal, the cause of the abnormality is inferred based on the position of the corresponding sensor, and appropriate measures are taken to return the plant to a normal state.
　取得した特定のデータが正常値か否かを判定するにあたっては、例えば下記特許文献１に記載されているように、当該データについての予測値を状況に応じて算出し、当該データの実測値が、算出された予測値からどの程度乖離しているのかを判定すればよい。尚、ここでいう「予測値」とは、プラントの全体が正常に動作していると仮定した場合において、上記特定のデータが示すと予測される値のことである。 To determine whether the acquired specific data is a normal value, for example, as described in Patent Document 1 below, a predicted value for the data is calculated depending on the situation, and it is determined how much the actual measured value of the data deviates from the calculated predicted value. Note that the "predicted value" here refers to the value predicted to be indicated by the specific data when it is assumed that the entire plant is operating normally.
特開昭６４－２１５０９号公報Japanese Patent Application Laid-Open No. 64-21509
　各データについての上記予測値は、例えば、当該データに影響を与える他の複数のデータの実測値と、統計モデルと、を用いて算出することができる。統計モデルは、複数のデータ間における相関を示すモデル（例えば数式）であって、正常時に予め取得された各データに基づいて作成されるものである。 The predicted value for each data item can be calculated, for example, using the actual measured values of multiple other data items that affect the data item in question and a statistical model. The statistical model is a model (e.g., a mathematical formula) that indicates the correlation between multiple data items, and is created based on each data item that has been acquired in advance under normal conditions.
　ただし、上記における「当該データに影響を与える他の複数のデータ」の種類や数は膨大であり、プラントの規模が大きくなると、そのようなデータは更に膨大なものとなる。従って、予測値の算出に必要な統計モデルは非常に複雑なものとなることが多く、正確な統計モデルを得るためには、予め膨大なサンプル数のデータを用いて学習を行っておくことが必要となる。 However, the types and numbers of the "multiple other data that affect the data in question" mentioned above are enormous, and as the scale of the plant increases, the amount of such data becomes even more enormous. Therefore, the statistical models required to calculate predicted values are often very complex, and in order to obtain an accurate statistical model, it is necessary to train it in advance using a huge number of sample data.
　しかしながら、学習に費やすことのできる時間は限られており、十分なサンプル数のデータを用いて学習を行うのは難しい場合が多い。少ないサンプル数のデータで統計モデルの学習を完了させると、予測値の算出精度が低くなるので、実測値と比較して異常の有無を判定することが難しくなってしまう。つまり、測定される各データについて、正常時と異常時との違いが見えにくくなってしまう。 However, the time available for learning is limited, and it is often difficult to learn using a sufficient number of sample data. If a statistical model is trained using a small number of sample data, the accuracy of the calculation of predicted values will be low, making it difficult to determine whether or not there is an abnormality by comparing the values with actual measured values. In other words, it becomes difficult to see the difference between normal and abnormal cases for each piece of measured data.
　本発明は、プラントにおける異常を正確に判定することのできる判定装置、判定方法、判定装置用のプログラム、及びプラントシステムを提供することを目的とする。 The present invention aims to provide a determination device, a determination method, a program for the determination device, and a plant system that can accurately determine abnormalities in a plant.
　本発明に係る判定装置は、プラントにおける異常を判定する判定装置であって、プラントに設けられた複数のセンサからデータを取得するデータ取得部と、データ取得部により取得されたデータと、複数のデータ間における相関を示す統計モデルと、に基づいて、異常の有無を判定する判定部と、を備える。プラントに設けられた複数のセンサには、予測対象センサが含まれる。上記判定装置の判定部は、複数のセンサのうち、予測対象センサと同じ系統に設けられ、かつ、予測対象センサと同種の物理量を測定するセンサから取得されたデータを利用して、予測対象センサから取得されるデータの予測値を算出し、当該予測値と、予測対象センサから取得されたデータの実測値とに基づいて、異常の有無を判定する。 The determination device according to the present invention is a determination device that determines the presence or absence of an abnormality in a plant, and includes a data acquisition unit that acquires data from multiple sensors installed in the plant, and a determination unit that determines the presence or absence of an abnormality based on the data acquired by the data acquisition unit and a statistical model that indicates the correlation between the multiple data. The multiple sensors installed in the plant include a prediction target sensor. The determination unit of the determination device calculates a predicted value of the data acquired from the prediction target sensor using data acquired from a sensor among the multiple sensors that is installed in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor, and determines the presence or absence of an abnormality based on the predicted value and the actual measured value of the data acquired from the prediction target sensor.
　このような構成の判定装置では、予測対象センサで取得されるデータの予測値を、予測対象センサと同じ系統に設けられ、かつ、予測対象センサと同種の物理量を測定するセンサから取得されたデータを利用して算出する。この場合、予測値の算出に用いられる統計モデルは、同系統且つ同種の物理量からなるデータ間の関係を示すものとのなるので、複数種類の物理量を含む場合に比べて数式等の表現が行いやすく、比較的シンプルなものとなる。このため、統計モデルの学習に用いるデータ数が少ない場合であっても、従来に比べて高い精度で予測値を算出することが可能となる。異常が生じた場合には、予測対象センサで実際に取得された実測値と、正常時の値として算出された予測値と、の差が顕著なものとなるので、両者に基づいて異常の有無を正確に判定することが可能となる。 In a determination device configured in this manner, a predicted value of data acquired by a sensor to be predicted is calculated using data acquired from a sensor that is provided in the same system as the sensor to be predicted and that measures the same type of physical quantity as the sensor to be predicted. In this case, the statistical model used to calculate the predicted value indicates the relationship between data consisting of the same system and type of physical quantity, and is therefore relatively simple and easier to express in mathematical expressions than when multiple types of physical quantities are included. For this reason, even if the amount of data used to train the statistical model is small, it is possible to calculate a predicted value with higher accuracy than in the past. When an abnormality occurs, the difference between the actual measured value acquired by the sensor to be predicted and the predicted value calculated as a normal value becomes significant, making it possible to accurately determine the presence or absence of an abnormality based on both.
　本発明によれば、プラントにおける異常を正確に判定することのできる判定装置、判定方法、判定装置用のプログラム、及びプラントシステムが提供される。 The present invention provides a determination device, a determination method, a program for the determination device, and a plant system that can accurately determine abnormalities in a plant.
図１は、実施形態に係るプラントシステムの全体構成を模式的に示す図である。FIG. 1 is a diagram illustrating a schematic diagram of an overall configuration of a plant system according to an embodiment. 図２は、プラントに設けられた複数のセンサを、系統ごとに纏めて表した図である。FIG. 2 is a diagram showing a plurality of sensors provided in a plant, grouped by system. 図３は、判定装置が備える判定部等の機能について説明するための図である。FIG. 3 is a diagram for explaining the functions of a determination unit and the like included in the determination device. 図４は、判定装置によって実行される処理の流れを示すフローチャートである。FIG. 4 is a flowchart showing the flow of the process executed by the determination device. 図５は、監視端末の画面に表示される情報の一例を示す図である。FIG. 5 is a diagram showing an example of information displayed on the screen of the monitoring terminal.
　以下、添付図面を参照しながら本実施形態について説明する。説明の理解を容易にするため、各図面において同一の構成要素に対しては可能な限り同一の符号を付して、重複する説明は省略する。 The present embodiment will now be described with reference to the accompanying drawings. To facilitate understanding, the same components in each drawing will be given the same reference numerals as much as possible, and duplicate descriptions will be omitted.
　本実施形態に係る判定装置１０は、プラントシステムＰＳの一部を構成するものであって、後述のプラント１１における異常を判定するための装置である。プラント１１における「異常」には、例えば配管の噴破のように、物理的な故障によってプラント１１が動作し得ない状態となることや、プラント１１が停止する可能性が高い状態となること等が含まれる。また、例えば何らかの原因でプラント１１の運転効率が著しく低下する等、物理的な故障を伴うことなく良好な運転状態から外れてしまうことも、上記の「異常」に含まれる。 The determination device 10 according to this embodiment constitutes part of the plant system PS, and is a device for determining abnormalities in the plant 11, which will be described later. An "abnormality" in the plant 11 includes a state in which the plant 11 cannot operate due to a physical failure, such as a pipe burst, or a state in which the plant 11 is likely to shut down. In addition, the above-mentioned "abnormality" also includes a deviation from a good operating state without a physical failure, such as a significant decrease in the operating efficiency of the plant 11 for some reason.
　判定装置１０の説明に先立ち、プラントシステムＰＳの構成について、主に図１を参照しながら説明する。図１に示されるように、プラントシステムＰＳは、プラント１１と、ＤＣＳ１２と、判定装置１０と、監視端末２０と、を備える。 Before describing the determination device 10, the configuration of the plant system PS will be described mainly with reference to FIG. 1. As shown in FIG. 1, the plant system PS includes a plant 11, a DCS 12, a determination device 10, and a monitoring terminal 20.
　本実施形態のプラント１１は、不図示のボイラを含む発電プラントである。ボイラでは高温の排ガスの熱によって水蒸気が生成され、当該水蒸気によって発電機のタービンが駆動される。プラント１１の稼働時においては、排ガス系統や水系統等における各部のデータ（温度や圧力等）が常に監視され、それぞれのデータが正常値を維持するように制御が行われる。尚、プラント１１は、本実施形態のように発電プラントであってもよいが、他の種類のプラントであってもよい。例えば、焼却プラントや化学プラントであってもよい。 The plant 11 in this embodiment is a power generation plant including a boiler (not shown). In the boiler, steam is generated by the heat of high-temperature exhaust gas, and the steam drives the turbine of the generator. When the plant 11 is in operation, data (temperature, pressure, etc.) of each part in the exhaust gas system, water system, etc. is constantly monitored, and control is performed so that each data maintains a normal value. Note that the plant 11 may be a power generation plant as in this embodiment, but it may also be another type of plant. For example, it may be an incineration plant or a chemical plant.
　ＤＣＳ１２は、「分散制御システム（Distributed Control System）」とも称されるものであり、プラント１１の制御を行う装置である。ＤＣＳ１２は、プラント１１の一部（例えばボイラ）の制御を行うものであるが、プラント１１の全体の制御を行うものであってもよい。 DCS 12 is also called a "Distributed Control System" and is a device that controls plant 11. DCS 12 controls a part of plant 11 (e.g., a boiler), but may also control the entire plant 11.
　ＤＣＳ１２は、プラント１１に設けられた各種機器（例えばブロア）のそれぞれに制御信号を送信し、プラント１１の動作を制御する。また、ＤＣＳ１２は、プラント１１の各部に設けられた複数のセンサのそれぞれから信号を受信し、プラント１１の状態を取得する。図２に示される表は、プラント１１に設けられた複数のセンサを、系統ごとに纏めたものである。ここでいう「系統」とは、例えば排ガス系統や水系統、電力系統等のように、同種の媒体が流れる一連の経路を構成するものである。 DCS 12 transmits control signals to each of the various devices (e.g., blowers) installed in plant 11, and controls the operation of plant 11. DCS 12 also receives signals from each of a number of sensors installed in various parts of plant 11, and acquires the status of plant 11. The table shown in FIG. 2 groups the multiple sensors installed in plant 11 by system. A "system" here refers to a series of paths through which the same type of medium flows, such as an exhaust gas system, a water system, an electric power system, etc.
　本実施形態では、プラント１１に、系統Ａ、系統Ｂ、及び系統Ｃからなる３つの系統が存在しており、それぞれの系統ごとに複数のセンサが設けられている。図２の例では、系統Ａの各部に、Ｓ１１からＳ１８までの計８個のセンサが設けられている。このうち、Ｓ１１、１２、１３は圧力を測定するためのセンサであり、Ｓ１４、１５、１６は温度を測定するためのセンサであり、Ｓ１７、１６は流量を測定するためのセンサである。図２に示される系統Ｂ、系統Ｃに示されるセンサ（Ｓ２１等）の数や種類も、同図に示される通りである。尚、各系統に実際に設けられているセンサの個数は、図２に示される数よりも多い。 In this embodiment, the plant 11 has three systems, system A, system B, and system C, and each system is provided with multiple sensors. In the example of FIG. 2, a total of eight sensors, S11 to S18, are provided in each section of system A. Of these, S11, 12, and 13 are sensors for measuring pressure, S14, 15, and 16 are sensors for measuring temperature, and S17 and 16 are sensors for measuring flow rate. The number and types of sensors (S21, etc.) shown in system B and system C shown in FIG. 2 are also as shown in the same figure. Note that the number of sensors actually provided in each system is greater than the number shown in FIG. 2.
　判定装置１０は、先に述べた通り、プラント１１における異常を判定するための装置である。判定装置１０は、不図示のＣＰＵ、ＲＯＭ、ＲＡＭ等を有するコンピュータシステムとして構成されており、例えばプラント１１と同じ敷地内に設置されている。判定装置１０は、例えば、プラント１１とは異なる場所にあるクラウドサーバーとして構成されていてもよい。 As described above, the determination device 10 is a device for determining abnormalities in the plant 11. The determination device 10 is configured as a computer system having a CPU, ROM, RAM, etc. (not shown), and is installed, for example, on the same premises as the plant 11. The determination device 10 may be configured, for example, as a cloud server located at a location different from the plant 11.
　判定装置１０は、ＤＣＳ１２との間で双方向の通信を行うことができる。判定装置１０は、プラント１１に設けられた各センサのそれぞれの実測値を、ＤＣＳ１２を介した通信によって取得する。このような態様に換えて、判定装置１０が、各センサからの信号を直接受信することとしてもよい。 The determination device 10 can communicate bidirectionally with the DCS 12. The determination device 10 acquires the actual measured values of each sensor installed in the plant 11 by communication via the DCS 12. Alternatively, the determination device 10 may directly receive signals from each sensor.
　監視端末２０は、使用者との間のインターフェイスとして設けられた端末装置である。使用者は、監視端末２０を操作することにより、判定装置１０の動作設定等に必要な情報を入力することができる。また、監視端末２０は、画面２００に各種の情報を表示させることにより、判定装置１０による判定の結果を使用者に報知することができる。監視端末２０は、特定の場所に設置された据え置き型の端末であってもよいが、使用者が携帯することのできる携帯通信端末であってもよい。監視端末２０は、判定装置１０の一部とみなすこともできる。 The monitoring terminal 20 is a terminal device provided as an interface with the user. The user can input information required for the operational settings of the determination device 10 by operating the monitoring terminal 20. The monitoring terminal 20 can also notify the user of the results of the determination made by the determination device 10 by displaying various information on the screen 200. The monitoring terminal 20 may be a stationary terminal installed in a specific location, or it may be a portable communication terminal that can be carried by the user. The monitoring terminal 20 can also be considered as part of the determination device 10.
　引き続き図１を参照しながら、判定装置１０の構成について説明する。判定装置１０は、その機能を表す要素として、データ取得部１１０と、選定部１２０と、判定部１３０と、報知部１４０と、記憶部１５０と、を備えている。 Continuing to refer to FIG. 1, the configuration of the determination device 10 will be described. The determination device 10 includes, as elements that represent its functions, a data acquisition unit 110, a selection unit 120, a determination unit 130, a notification unit 140, and a storage unit 150.
　データ取得部１１０は、プラント１１に設けられた複数のセンサからデータを取得する処理を行う部分である。データ取得部１１０は、プラント１１に設けられた全てのセンサからのデータを、ＤＣＳ１２との通信を介して繰り返し取得する。 The data acquisition unit 110 is a part that performs processing to acquire data from multiple sensors installed in the plant 11. The data acquisition unit 110 repeatedly acquires data from all sensors installed in the plant 11 through communication with the DCS 12.
　選定部１２０は、データ取得部１１０で取得された複数のデータの中から、判定部１３０の処理で用いるためのデータを選定する処理、を行う部分である。選定部１２０が行う処理の具体的な内容については後に説明する。 The selection unit 120 is a part that performs a process of selecting data to be used in the process of the determination unit 130 from among the multiple pieces of data acquired by the data acquisition unit 110. The specific content of the process performed by the selection unit 120 will be explained later.
　判定部１３０は、データ取得部１１０により取得されたデータ（具体的には、選定部１２０によって選定されたデータ）と、統計モデルと、に基づいて、プラント１１における異常の有無を判定する処理を行う部分である。「統計モデル」とは、各センサで取得される複数のデータ間における相関を示すモデルであって、例えば数式として表現されるものである。統計モデルは、プラント１１の全体が正常に動作しているときに各センサから取得されたデータに基づいて予め作成され、後述の記憶部１５０に記憶されている。統計モデルは、プラント１１が正常に動作している期間において都度学習され更新されることとしてもよい。 The determination unit 130 is a part that performs processing to determine the presence or absence of an abnormality in the plant 11 based on the data acquired by the data acquisition unit 110 (specifically, the data selected by the selection unit 120) and a statistical model. A "statistical model" is a model that indicates the correlation between multiple pieces of data acquired by each sensor, and is expressed, for example, as a mathematical formula. The statistical model is created in advance based on the data acquired from each sensor when the entire plant 11 is operating normally, and is stored in the storage unit 150 described below. The statistical model may be learned and updated each time the plant 11 is operating normally.
　判定部１３０は、例えば複数のセンサで測定されたデータを統計モデルに入力し、そのアウトプットとして、特定のセンサで取得されるデータの予測値を算出することができる。このように算出される「予測値」は、プラント１１の全体が正常に動作していると仮定した場合において、上記特定のセンサが示すと予測される値ということになる。従って、当該センサで実際に取得された実測値が、算出された予測値から大きく乖離した場合には、判定部１３０は、プラント１１において異常が生じていると判定することができる。 The judgment unit 130 can, for example, input data measured by multiple sensors into a statistical model and calculate, as its output, a predicted value of the data acquired by a specific sensor. The "predicted value" calculated in this way is the value predicted to be indicated by the specific sensor, assuming that the entire plant 11 is operating normally. Therefore, if the actual measurement value actually acquired by the sensor significantly deviates from the calculated predicted value, the judgment unit 130 can determine that an abnormality has occurred in the plant 11.
　報知部１４０は、判定部１３０による判定結果を報知する処理、を行う部分である。報知部１４０は、判定結果を監視端末２０の画面２００に表示させることにより、使用者への報知を行う。報知部１４０による報知の具体的な態様については後に説明する。 The notification unit 140 is a part that performs the process of notifying the user of the judgment result by the judgment unit 130. The notification unit 140 notifies the user of the judgment result by displaying it on the screen 200 of the monitoring terminal 20. The specific manner in which the notification unit 140 notifies the user will be described later.
　記憶部１５０は、判定装置１０に設けられた不揮発性の記憶装置であって、例えばＨＤＤやＳＳＤである。記憶部１５０は、判定装置１０とは異なる場所に設置されたファイルサーバーであってもよい。記憶部１５０には、先に述べた統計モデルを含む、判定装置１０が行う処理に必要な種々の情報が記憶されている。 The storage unit 150 is a non-volatile storage device provided in the determination device 10, such as an HDD or SSD. The storage unit 150 may be a file server installed in a location different from the determination device 10. The storage unit 150 stores various information necessary for the processing performed by the determination device 10, including the statistical model described above.
　後に説明するように、統計モデルは、プラント１１に設けられたそれぞれのセンサに対応して個別に作成され、それぞれの統計モデルが記憶部１５０に記憶されている。記憶部１５０に記憶されたそれぞれの統計モデルのことを、以下では「統計モデル１５１」とも称する。 As will be described later, a statistical model is created individually for each sensor installed in the plant 11, and each statistical model is stored in the memory unit 150. Each statistical model stored in the memory unit 150 will be referred to below as a "statistical model 151."
　判定装置１０で実行される処理の概要について、図３を参照しながら説明する。図３には、判定装置１０の各部における情報の流れが模式的に描かれている。 The process executed by the determination device 10 will be outlined with reference to FIG. 3. FIG. 3 shows a schematic diagram of the flow of information in each part of the determination device 10.
　プラント１１に設けられた複数のセンサのうち、異常の判定に用いられる特定のセンサのことを、以下では「予測対象センサ」とも称する。判定部１３０は、複数のセンサの中から、任意の１つのセンサを予測対象センサとして選定し、予測対象センサにおける上記「予測値」を算出する。その後、予測対象センサにおける実測値と、予測値とを比較することで、異常の有無を判定する。 Among the multiple sensors installed in the plant 11, a specific sensor used to determine whether or not there is an abnormality is hereinafter also referred to as the "prediction target sensor." The determination unit 130 selects any one sensor from the multiple sensors as the prediction target sensor, and calculates the above-mentioned "prediction value" for the prediction target sensor. Thereafter, the actual measurement value for the prediction target sensor is compared with the predicted value to determine whether or not there is an abnormality.
　データ取得部１１０は、予測対象センサを含む全てのセンサからのデータを取得し、これを選定部１２０に送る。選定部１２０は、これらのデータの中から、予測対象センサと同じ系統に設けられ、且つ、予測対象センサと同種の物理量を測定するセンサから取得されたデータを選定する。例えば、図２のＳ１４が予測対象センサとなっている場合には、選定部１２０は、同じ系統Ａに設けられたセンサであって、Ｓ１４と同じ「温度」を測定するセンサであるＳ１５及びＳ１６を選定する。選定部１２０は、このように選定したデータを、判定部１３０の一部である予測値算出部１３１に送る。図３では、このように送られるデータが「ｄ２」として示されている。 The data acquisition unit 110 acquires data from all sensors, including the prediction target sensor, and sends this to the selection unit 120. From this data, the selection unit 120 selects data acquired from sensors that are installed in the same system as the prediction target sensor and measure the same type of physical quantity as the prediction target sensor. For example, when S14 in FIG. 2 is the prediction target sensor, the selection unit 120 selects S15 and S16, which are installed in the same system A and measure the same "temperature" as S14. The selection unit 120 sends the data selected in this way to the prediction value calculation unit 131, which is part of the determination unit 130. In FIG. 3, the data sent in this way is shown as "d2".
　選定部１２０は、データ取得部１１０から送られたデータのうち、予測対象センサで取得された実測値のデータを、判定部１３０の一部である異常度算出部１３２に送る。図３では、このように送られるデータが「ｄ１」として示されている。 The selection unit 120 sends the actual measurement data acquired by the prediction target sensor, among the data sent from the data acquisition unit 110, to the anomaly degree calculation unit 132, which is part of the judgment unit 130. In FIG. 3, the data sent in this way is shown as "d1".
　予測値算出部１３１は、記憶部１５０に記憶されている複数の統計モデル１５１の中から、予測対象センサに対応する統計モデル１５１を読み出す。この統計モデル１５１は、予測対象センサと同じ系統に設けられ、且つ、予測対象センサと同種の物理量を測定するセンサから取得されるデータ（つまり、図３のｄ２）と、予測対象センサで測定されるデータとの相関を示すものとして、予め作成されていたものである。 The prediction value calculation unit 131 reads out a statistical model 151 corresponding to the sensor to be predicted from among the multiple statistical models 151 stored in the storage unit 150. This statistical model 151 is provided in the same system as the sensor to be predicted and was created in advance as a model that indicates the correlation between data acquired from a sensor that measures the same type of physical quantity as the sensor to be predicted (i.e., d2 in FIG. 3) and data measured by the sensor to be predicted.
　予測値算出部１３１は、上記の統計モデル１５１と、選定部１２０で選定されたデータ（ｄ２）とに基づいて、予測対象センサで測定されるデータの予測値を算出し、これを異常度算出部１３２に送る。図３では、このように送られる予測値が「ｄ１’」として示されている。 The predicted value calculation unit 131 calculates a predicted value of the data measured by the prediction target sensor based on the statistical model 151 and the data (d2) selected by the selection unit 120, and sends this to the anomaly calculation unit 132. In FIG. 3, the predicted value sent in this manner is shown as "d1'".
　以上のように、異常度算出部１３２には、予測対象センサで取得されたデータの実測値ｄ１と、予測対象センサで測定されるデータの予測値ｄ１’と、の両方が送られる。 As described above, both the actual measured value d1 of the data acquired by the sensor to be predicted and the predicted value d1' of the data to be measured by the sensor to be predicted are sent to the anomaly degree calculation unit 132.
　予測値ｄ１’は、プラント１１の全体が正常に動作しているという前提の下で算出されたものである。このため、プラント１１において異常が生じていない場合には、実測値ｄ１は、予測値ｄ１’に概ね等しい値となり、両者の差は小さくなる。一方、プラント１１において異常が生じている場合には、実測値ｄ１は予測値ｄ１’から乖離した値となり、両者の差は大きくなる。従って、実測値ｄ１と予測値ｄ１’との差は、プラント１１において生じている異常の度合いを示すものとなる。 The predicted value d1' is calculated under the assumption that the entire plant 11 is operating normally. Therefore, if no abnormality occurs in the plant 11, the actual measurement value d1 will be roughly equal to the predicted value d1', and the difference between the two will be small. On the other hand, if an abnormality occurs in the plant 11, the actual measurement value d1 will deviate from the predicted value d1', and the difference between the two will be large. Therefore, the difference between the actual measurement value d1 and the predicted value d1' indicates the degree of abnormality occurring in the plant 11.
　異常度算出部１３２は、プラント１１において生じている異常の度合いを示す指標、である「異常度」を算出する。本実施形態では、異常度は、実測値ｄ１と予測値ｄ１’との差を正規化した値として算出される。例えば、異常度は、０から１までの範囲の値として算出される。この場合、実測値ｄ１と予測値ｄ１’とが一致している場合には異常度は０となる。また、実測値ｄ１と予測値ｄ１’の差が大きくなるにしたがって異常度は大きくなり、当該差がある程度大きくなった以降は、異常度は１として算出される。 The anomaly degree calculation unit 132 calculates the "anomaly degree", which is an index showing the degree of anomaly occurring in the plant 11. In this embodiment, the anomaly degree is calculated as a normalized value of the difference between the actual measured value d1 and the predicted value d1'. For example, the anomaly degree is calculated as a value in the range from 0 to 1. In this case, if the actual measured value d1 and the predicted value d1' match, the anomaly degree is 0. Furthermore, the anomaly degree increases as the difference between the actual measured value d1 and the predicted value d1' increases, and once the difference reaches a certain level, the anomaly degree is calculated as 1.
　実測値ｄ１と予測値ｄ１’との差の符号を考慮して、－１から１までの値として異常度が算出されてもよい。この場合、実測値ｄ１が予測値ｄ１’よりも著しく大きい場合には異常度は１となり、実測値ｄ１が予測値ｄ１’よりも著しく小さい場合には異常度は－１となる。 The degree of anomaly may be calculated as a value between -1 and 1, taking into account the sign of the difference between the actual measured value d1 and the predicted value d1'. In this case, if the actual measured value d1 is significantly larger than the predicted value d1', the degree of anomaly will be 1, and if the actual measured value d1 is significantly smaller than the predicted value d1', the degree of anomaly will be -1.
　尚、異常度算出部１３２によって算出される異常度は、プラント１１において生じている異常の度合いを示す指標、として用いることができるのであれば、上記とは異なる方法で算出される値であってもよい。例えば、実測値ｄ１と予測値ｄ１’との差を、そのまま異常度として用いることとしてもよい。 The degree of anomaly calculated by the anomaly degree calculation unit 132 may be a value calculated by a method different from the above, as long as it can be used as an index showing the degree of anomaly occurring in the plant 11. For example, the difference between the actual measurement value d1 and the predicted value d1' may be used as the degree of anomaly.
　異常度算出部１３２で算出された異常度は、報知部１４０に送られる。報知部１４０は、監視端末２０の画面２００に異常度を表示させることで、使用者への報知を行う。 The degree of abnormality calculated by the abnormality degree calculation unit 132 is sent to the notification unit 140. The notification unit 140 notifies the user by displaying the degree of abnormality on the screen 200 of the monitoring terminal 20.
　以上のように、本実施形態に係る判定装置１０の判定部１３０は、プラント１１に設けられた複数のセンサのうち、予測対象センサと同じ系統に設けられ、且つ、予測対象センサと同種の物理量を測定するセンサから取得されたデータを利用して、予測対象センサから取得されるデータの予測値（ｄ１’）を算出し、当該予測値（ｄ１’）と、予測対象センサから取得されたデータの実測値（ｄ１）とに基づいて、異常の有無を判定するように構成されている。 As described above, the judgment unit 130 of the judgment device 10 according to this embodiment is configured to calculate a predicted value (d1') of the data acquired from the prediction target sensor using data acquired from a sensor, among multiple sensors installed in the plant 11, that is installed in the same system as the prediction target sensor and that measures the same type of physical quantity as the prediction target sensor, and to judge the presence or absence of an abnormality based on the predicted value (d1') and the actual measured value (d1) of the data acquired from the prediction target sensor.
　予測値の算出に用いられる統計モデル１５１は、予測対象センサと同系統且つ同種の物理量からなるデータ間の関係を示すものとして作成されている。すなわち、予測対象センサとは異なる系統に設けられたセンサで測定されるデータや、予測対象センサで測定されるものとは異なる種類の物理量のデータを含まないモデルとして、統計モデル１５１が作成されている。その結果、本実施形態の統計モデル１５１は、複数種類の物理量を含む場合に比べて数式等の表現が行いやすいため、比較的シンプルなものとなっている。 The statistical model 151 used to calculate the predicted value is created to show the relationship between data consisting of the same system and type of physical quantity as the sensor to be predicted. In other words, the statistical model 151 is created as a model that does not include data measured by a sensor installed in a system different from the sensor to be predicted, or data of physical quantities of a different type than those measured by the sensor to be predicted. As a result, the statistical model 151 of this embodiment is relatively simple, since it is easier to express it in a mathematical formula, etc., compared to when multiple types of physical quantities are included.
　その結果、少ないデータ数で統計モデルの学習が行われた場合であっても、予測値算出部１３１では、従来に比べて高い精度で予測値を算出することができる。異常が生じた場合には、予測値ｄ１’と実測値ｄ１との差が顕著なものとなるので、判定部１３０は、両者の比較に基づいて異常の有無を正確に判定することができる。 As a result, even if a statistical model is trained using a small amount of data, the predicted value calculation unit 131 can calculate a predicted value with higher accuracy than in the past. When an abnormality occurs, the difference between the predicted value d1' and the actual measured value d1 becomes significant, and the determination unit 130 can accurately determine the presence or absence of an abnormality based on a comparison between the two.
　尚、選定部１２０によって選定され予測値算出部１３１に送られるデータｄ２には、予測対象センサから取得された実測値のデータが含まれていてもよい。例えば、深層学習によって統計モデル１５１を都度更新して行くような構成においては、予測対象センサによる実測値のデータを含めた形の統計モデルとした上で、これを用いて予測値の算出を行うことが好ましい。 The data d2 selected by the selection unit 120 and sent to the prediction value calculation unit 131 may include data on actual measurements obtained from the sensor to be predicted. For example, in a configuration in which the statistical model 151 is updated each time by deep learning, it is preferable to create a statistical model that includes data on actual measurements from the sensor to be predicted, and then use this to calculate the prediction value.
　以上のことから、本実施形態の判定部１３０は、複数のセンサのうち、予測対象センサと同じ系統に設けられ、且つ、予測対象センサと同種の物理量を測定するセンサから取得されたデータのみを利用して、もしくは、当該データに対し予測対象センサで測定されたデータを加えたものを利用して、予測対象センサから取得されるデータの予測値を算出し、当該予測値と、予測対象センサから取得されたデータの実測値とに基づいて、異常の有無を判定するもの、ということができる。 From the above, the judgment unit 130 of this embodiment can be said to calculate a predicted value of the data acquired from the prediction target sensor using only data acquired from a sensor among multiple sensors that is installed in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor, or using the data acquired from the prediction target sensor plus data measured by the prediction target sensor, and to judge the presence or absence of an abnormality based on the predicted value and the actual measured value of the data acquired from the prediction target sensor.
　判定装置１０によって実行される処理の具体的な流れについて、図４のフローチャートを参照しながら説明する。同図に示される一連の処理は、プラント１１の運転が行われている期間において、所定の周期が経過する毎に、判定装置１０によって繰り返し実行されるものである。 The specific flow of the processing executed by the determination device 10 will be described with reference to the flowchart in FIG. 4. The series of processing shown in the figure is repeatedly executed by the determination device 10 every time a predetermined period elapses during the period during which the plant 11 is in operation.
　最初のステップＳＴ１では、データ取得部１１０によって、プラント１１に設けられた全てのセンサからのデータを取得する処理が行われる。ステップＳＴ１で行われる処理は、本実施形態における「取得工程」及び「データ取得処理」に該当する。 In the first step ST1, the data acquisition unit 110 performs a process of acquiring data from all sensors installed in the plant 11. The process performed in step ST1 corresponds to the "acquisition process" and "data acquisition process" in this embodiment.
　ステップＳＴ１に続くステップＳＴ２では、プラント１１に設けられたセンサの中から、１つの予測対象センサを決定する処理が、例えば判定部１３０によって行われる。本実施形態では、プラント１１に設けられた複数のセンサのそれぞれが、１つずつ順に予測対象センサとして決定され、それぞれの予測対象センサにおける実測値と予測値との比較に基づいて異常の有無が判定される。それぞれのセンサが予測対象センサとして決定される順序は、任意に設定することができる。 In step ST2 following step ST1, a process of determining one sensor to be predicted from among the sensors installed in the plant 11 is performed by, for example, the determination unit 130. In this embodiment, each of the multiple sensors installed in the plant 11 is determined one by one in order as the sensor to be predicted, and the presence or absence of an abnormality is determined based on a comparison between the actual measurement value and the predicted value at each sensor to be predicted. The order in which each sensor is determined as the sensor to be predicted can be set arbitrarily.
　ステップＳＴ２に続くステップＳＴ３では、予測対象センサと同じ系統に設けられ、且つ、予測対象センサと同種の物理量を測定するセンサから取得されたデータ、すなわち、予測値の算出に用いられるデータを選定する処理が、選定部１２０によって行われる。 In step ST3 following step ST2, the selection unit 120 selects data acquired from a sensor that is provided in the same system as the sensor to be predicted and that measures the same type of physical quantity as the sensor to be predicted, i.e., data to be used in calculating the predicted value.
　ステップＳＴ３に続くステップＳＴ４では、予測対象センサに対応する統計モデル１５１を記憶部１５０から読み出す処理が、図３の予測値算出部１３１によって行われる。続くステップＳＴ５において、予測値算出部１３１は、ステップＳＴ４で読み出された統計モデル１５１と、ステップＳＴ３で選定された予測値算出用のデータ（図３のｄ２）とに基づいて、予測対象センサで測定されるデータの予測値を算出する。 In step ST4 following step ST3, the prediction value calculation unit 131 in FIG. 3 reads out from the storage unit 150 the statistical model 151 corresponding to the sensor to be predicted. In the following step ST5, the prediction value calculation unit 131 calculates a prediction value for the data measured by the sensor to be predicted based on the statistical model 151 read out in step ST4 and the data for calculating the prediction value selected in step ST3 (d2 in FIG. 3).
　ステップＳＴ５に続くステップＳＴ６では、異常度を算出する処理が、図３の異常度算出部１３２によって行われる。先に述べたように、異常度算出部１３２は、予測対象センサで取得された実測値のデータ（図３のｄ１）と、ステップＳＴ５で算出された予測値（図３のｄ１’）との比較に基づいて、異常度を算出する。ステップＳＴ６で行われる処理は、本実施形態における「判定工程」及び「判定処理」に該当する。ステップＳＴ６に続くステップＳＴ７では、監視端末２０の画面２００に異常度を表示させる処理が、報知部１４０によって行われる。 In step ST6 following step ST5, the process of calculating the degree of abnormality is performed by the degree of abnormality calculation unit 132 in FIG. 3. As described above, the degree of abnormality calculation unit 132 calculates the degree of abnormality based on a comparison between the actual measurement value data (d1 in FIG. 3) acquired by the sensor to be predicted and the predicted value (d1' in FIG. 3) calculated in step ST5. The process performed in step ST6 corresponds to the "determination step" and "determination process" in this embodiment. In step ST7 following step ST6, the notification unit 140 performs a process of displaying the degree of abnormality on the screen 200 of the monitoring terminal 20.
　ステップＳＴ７に続くステップＳＴ８では、異常度の算出及び報知が、プラント１１に設けられた全てのセンサについて完了したか否かが判定される。全てのセンサについて完了した場合には、図４に示される一連の処理を終了する。予測対象センサとして未だ選定されていないセンサが存在する場合には、ステップＳＴ９に移行する。ステップＳＴ９では、予測対象センサとして未だ選定されていないセンサの中の一つを選定し、これを次の予測対象センサとして決定する処理が行われる。以降は、ステップＳＴ３以降の処理が再度実行される。 In step ST8 following step ST7, it is determined whether the calculation and reporting of the degree of anomaly has been completed for all sensors installed in the plant 11. If the calculation and reporting has been completed for all sensors, the series of processes shown in FIG. 4 ends. If there is a sensor that has not yet been selected as the sensor to be predicted, the process proceeds to step ST9. In step ST9, one of the sensors that has not yet been selected as the sensor to be predicted is selected and determined to be the next sensor to be predicted. Thereafter, the processes from step ST3 onwards are executed again.
　このような処理が繰り返される結果として、判定部１３０は、プラント１１に設けられた複数のセンサのそれぞれについて、当該センサを予測対象センサとした場合における異常の有無を個別に判定することとなる。また、報知部１４０は、複数のセンサのそれぞれに対応する判定結果を個別に報知することとなる。つまり、複数のセンサのそれぞれについて、当該センサを予測対象センサとした場合における異常度が算出され報知されることとなる。尚、予測対象センサとして設定されるのは、本実施形態のように全てのセンサであってもよいが、プラント１１に設けられた一部のセンサのみであってもよい。 As a result of repeating such processing, the judgment unit 130 will individually judge the presence or absence of an abnormality for each of the multiple sensors installed in the plant 11 when that sensor is set as the sensor to be predicted. Furthermore, the notification unit 140 will individually notify the judgment result corresponding to each of the multiple sensors. In other words, for each of the multiple sensors, the degree of abnormality when that sensor is set as the sensor to be predicted is calculated and notified. Note that all sensors may be set as the sensors to be predicted as in this embodiment, but only some of the sensors installed in the plant 11 may be set as the sensors to be predicted.
　報知部１４０による報知の態様の一例について、図５を参照しながら説明する。報知が行われる際において、監視端末２０の画面２００には、時系列グラフ２０１、２０２、２０３と、異常度分布図２１０と、が表示される。 An example of a mode of notification by the notification unit 140 will be described with reference to FIG. 5. When a notification is made, time series graphs 201, 202, and 203 and an anomaly distribution diagram 210 are displayed on the screen 200 of the monitoring terminal 20.
　時系列グラフ２０２、２０３は、それぞれ、図２のセンサＳ１４、Ｓ１５について算出された異常度の変化を時系列で表示したグラフである。時系列グラフ２０１は、センサＳ１４、Ｓ１５で測定されたデータの差（温度差）を時系列で表示したグラフである。 Time series graphs 202 and 203 are graphs that display the change in the degree of anomaly calculated for sensors S14 and S15 in FIG. 2 over time. Time series graph 201 is a graph that displays the difference (temperature difference) between the data measured by sensors S14 and S15 over time.
　尚、時系列グラフで表示する対象となるセンサは、上記以外のセンサであってもよい。時系列グラフとして表示の対象となるデータは、監視端末２０に対し使用者が行う操作によって任意に設定できるようにすればよい。 The sensors to be displayed in the time series graph may be sensors other than those mentioned above. The data to be displayed in the time series graph may be arbitrarily set by the user through operations performed on the monitoring terminal 20.
　異常度分布図２１０は、プラント１１の全体を示す構成図の上に、各部に設けられたセンサを表すアイコン２１１を複数表示したものである。アイコン２１１が表示されている位置は、プラント１１においてセンサが設けられている位置に対応する。また、アイコン２１１の色は、対応するセンサについて判定部１３０により算出された異常度の値を表している。例えば、アイコン２１１は、対応センサについて算出された異常度が０に近いほど青色に近づき、異常度が１又は－１に近いほど赤色に近づくようなグラデーション、から選択された色で表示される。それぞれのアイコン２１１の色は、図４のステップＳＴ６で算出された異常度に応じて、続くステップＳＴ７の処理時において更新される。つまり、本実施形態の報知部１４０は、異常度分布図２１０に表示された各アイコン２１１の色を変化させることで、各センサのそれぞれに対応する判定結果を個別に報知する。 The anomaly distribution diagram 210 displays a number of icons 211 representing sensors provided in each part on a configuration diagram showing the entire plant 11. The positions at which the icons 211 are displayed correspond to the positions at which the sensors are provided in the plant 11. The color of the icon 211 represents the value of the anomaly level calculated by the judgment unit 130 for the corresponding sensor. For example, the icon 211 is displayed in a color selected from a gradation in which the closer the calculated anomaly level for the corresponding sensor is to 0, the closer it is to blue, and the closer the anomaly level is to 1 or -1, the closer it is to red. The color of each icon 211 is updated during the processing of the subsequent step ST7 according to the anomaly level calculated in step ST6 of FIG. 4. In other words, the notification unit 140 of this embodiment individually notifies the judgment result corresponding to each sensor by changing the color of each icon 211 displayed in the anomaly distribution diagram 210.
　尚、異常度分布図２１０に表示されるアイコン２１１は、選定部１２０によって選定されたデータ（つまり、予測値の算出に用いられたデータ）に対応するセンサについてのみ表示されてもよく、プラント１１に設けられた全てのセンサについて表示されてもよい。 In addition, the icons 211 displayed on the anomaly distribution diagram 210 may be displayed only for the sensors corresponding to the data selected by the selection unit 120 (i.e., the data used to calculate the predicted value), or may be displayed for all sensors installed in the plant 11.
　また、アイコン２１１は、上記のように異常度の値を色で表すものであってもよいが、異常度が所定の閾値を超えたか否かの判定結果を、２種類の色等で表示するものであってもよい。換言すれば、報知部１４０は、プラント１１の各部における異常の有無を、その度合い（つまり異常度）を含めて報知するものであってもよく、単に異常の有無のみを報知するものであってもよい。 In addition, the icon 211 may represent the degree of abnormality with a color as described above, but may also display the result of the determination of whether the degree of abnormality has exceeded a predetermined threshold using, for example, two different colors. In other words, the notification unit 140 may notify the presence or absence of an abnormality in each part of the plant 11, including the degree of the abnormality (i.e., the degree of abnormality), or may simply notify the presence or absence of an abnormality.
　判定装置１０によって行われる以上のような判定方法は、例えば、判定装置１０に設けられた不揮発性の記憶装置（不図示）に記憶されたプログラム、によって実現される。当該プログラムに従って判定装置１０の判定部１３０等を動作させることで、先に述べた「データ取得処理」や「判定処理」等が実行される。プログラムの一部又は全部は、判定装置１０に対し外部から送信され、判定装置１０の記憶装置に一時的に書き込まれるようなものであってもよい。 The above-mentioned determination method performed by the determination device 10 is realized, for example, by a program stored in a non-volatile storage device (not shown) provided in the determination device 10. The aforementioned "data acquisition process" and "determination process" are executed by operating the determination unit 130 of the determination device 10 in accordance with the program. A part or all of the program may be transmitted to the determination device 10 from an external device and temporarily written to the storage device of the determination device 10.
　［付記］
上記各実施形態から把握できる技術的思想を以下に記載する。
（付記１）
プラントにおける異常を判定する判定装置であって、プラントに設けられた複数のセンサからデータを取得するデータ取得部と、データ取得部により取得されたデータと、複数のデータ間における相関を示す統計モデルと、に基づいて、異常の有無を判定する判定部と、を備え、複数のセンサには、予測対象センサが含まれ、判定部は、複数のセンサのうち、予測対象センサと同じ系統に設けられ、且つ、予測対象センサと同種の物理量を測定するセンサから取得されたデータを利用して、予測対象センサから取得されるデータの予測値を算出し、当該予測値と、予測対象センサから取得されたデータの実測値とに基づいて、異常の有無を判定する、判定装置。
（付記２）
データ取得部で取得された複数のデータの中から、予測対象センサと同じ系統に設けられ、且つ、予測対象センサと同種の物理量を測定するセンサから取得されたデータを選定する選定部、を更に備える、付記１に記載の判定装置。
（付記３）
予測値の算出に用いられるデータには、予測対象センサから取得されたデータも含まれる、付記１又は２に記載の判定装置。
（付記４）
判定部による判定結果を報知する報知部を更に備える、付記１乃至３のいずれか１つに記載の判定装置。
（付記５）
判定部は、複数のセンサのそれぞれについて、当該センサを予測対象センサとした場合における異常の有無を個別に判定し、報知部は、複数のセンサのそれぞれに対応する判定結果を個別に報知する、付記４に記載の判定装置。
（付記６）
プラントにおける異常を判定する判定方法であって、プラントに設けられた複数のセンサからデータを取得する取得工程と、取得工程において取得されたデータと、複数のデータ間における相関を示す統計モデルと、に基づいて、異常の有無を判定する判定工程と、を有し、複数のセンサには、予測対象センサが含まれ、判定工程においては、複数のセンサのうち、予測対象センサと同じ系統に設けられ、かつ、予測対象センサと同種の物理量を測定するセンサから取得されたデータを利用して、予測対象センサから取得されるデータの予測値を算出し、当該予測値と、予測対象センサから取得されたデータの実測値とに基づいて、異常の有無を判定する、判定方法。
（付記７）
プラントにおける異常を判定する判定装置用のプログラムであって、プラントに設けられた複数のセンサからデータを取得するデータ取得処理と、データ取得処理において取得されたデータと、複数のデータ間における相関を示す統計モデルと、に基づいて、異常の有無を判定する判定処理と、を、判定装置に行わせるものであり、複数のセンサには、予測対象センサが含まれ、複数のセンサのうち、予測対象センサと同じ系統に設けられ、かつ、予測対象センサと同種の物理量を測定するセンサから取得されたデータを利用して、予測対象センサから取得されるデータの予測値を算出し、当該予測値と、予測対象センサから取得されたデータの実測値とに基づいて、異常の有無を判定する処理を、判定処理として判定装置に行わせる、プログラム。
（付記８）
プラントと、プラントにおける異常を判定する判定装置と、を備えるプラントシステムであって、判定装置は、プラントに設けられた複数のセンサからデータを取得するデータ取得部と、データ取得部により取得されたデータと、複数のデータ間における相関を示す統計モデルと、に基づいて、異常の有無を判定する判定部と、を有し、複数のセンサには、予測対象センサが含まれ、判定部は、複数のセンサのうち、予測対象センサと同じ系統に設けられ、かつ、予測対象センサと同種の物理量を測定するセンサから取得されたデータを利用して、予測対象センサから取得されるデータの予測値を算出し、当該予測値と、予測対象センサから取得されたデータの実測値とに基づいて、異常の有無を判定する、プラントシステム。 [Additional Notes]
The technical ideas that can be understood from the above embodiments will be described below.
(Appendix 1)
A determination device for determining whether or not an abnormality exists in a plant, the determination device comprising: a data acquisition unit that acquires data from a plurality of sensors provided in the plant; and a determination unit that determines whether or not an abnormality exists based on the data acquired by the data acquisition unit and a statistical model that indicates a correlation between the plurality of data, wherein the plurality of sensors includes a prediction target sensor, and the determination unit calculates a predicted value of the data acquired from the prediction target sensor using data acquired from a sensor among the plurality of sensors that is provided in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor, and determines whether or not an abnormality exists based on the predicted value and an actual measurement value of the data acquired from the prediction target sensor.
(Appendix 2)
2. The determination device according to claim 1, further comprising a selection unit that selects, from the plurality of data acquired by the data acquisition unit, data acquired from a sensor that is provided in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor.
(Appendix 3)
3. The determination device according to claim 1, wherein the data used to calculate the predicted value includes data acquired from a sensor to be predicted.
(Appendix 4)
4. The determination device according to claim 1, further comprising a notification unit that notifies a result of the determination by the determination unit.
(Appendix 5)
The determination unit individually determines whether or not an abnormality exists for each of the multiple sensors when the sensor is the sensor to be predicted, and the notification unit individually notifies the determination result corresponding to each of the multiple sensors.
(Appendix 6)
A method for determining whether or not an abnormality exists in a plant, the method comprising: an acquisition step of acquiring data from a plurality of sensors provided in the plant; and a determination step of determining whether or not an abnormality exists based on the data acquired in the acquisition step and a statistical model indicating a correlation between the plurality of data, wherein the plurality of sensors includes a prediction target sensor; and in the determination step, a prediction value of the data acquired from the prediction target sensor is calculated using data acquired from a sensor among the plurality of sensors that is provided in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor, and the presence or absence of an abnormality is determined based on the prediction value and an actual measurement value of the data acquired from the prediction target sensor.
(Appendix 7)
A program for a determination device that determines whether or not an abnormality exists in a plant, the program causing the determination device to perform a data acquisition process that acquires data from a plurality of sensors provided in the plant, and a determination process that determines whether or not an abnormality exists based on the data acquired in the data acquisition process and a statistical model that indicates a correlation between the plurality of data, wherein the plurality of sensors includes a prediction target sensor, and the program causes the determination device to perform a determination process in which a predicted value of data acquired from the prediction target sensor is calculated using data acquired from a sensor among the plurality of sensors that is provided in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor, and determines whether or not an abnormality exists based on the predicted value and an actual measurement value of the data acquired from the prediction target sensor.
(Appendix 8)
A plant system comprising a plant and a determination device that determines whether or not an abnormality occurs in the plant, the determination device having a data acquisition unit that acquires data from a plurality of sensors provided in the plant, and a determination unit that determines whether or not an abnormality exists based on the data acquired by the data acquisition unit and a statistical model that indicates a correlation between the plurality of data, wherein the plurality of sensors includes a prediction target sensor, and the determination unit calculates a predicted value of the data acquired from the prediction target sensor using data acquired from a sensor among the plurality of sensors that is provided in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor, and determines whether or not an abnormality exists based on the predicted value and an actual measurement value of the data acquired from the prediction target sensor.
　以上、具体例を参照しつつ本実施形態について説明した。しかし、本開示はこれらの具体例に限定されるものではない。これら具体例に、当業者が適宜設計変更を加えたものも、本開示の特徴を備えている限り、本開示の範囲に包含される。前述した各具体例が備える各要素およびその配置、条件、形状などは、例示したものに限定されるわけではなく適宜変更することができる。前述した各具体例が備える各要素は、技術的な矛盾が生じない限り、適宜組み合わせを変えることができる。 The present embodiment has been described above with reference to specific examples. However, the present disclosure is not limited to these specific examples. Design modifications to these specific examples made by a person skilled in the art are also included within the scope of the present disclosure as long as they have the features of the present disclosure. The elements of each of the above-mentioned specific examples, as well as their arrangement, conditions, shape, etc., are not limited to those exemplified and can be modified as appropriate. The elements of each of the above-mentioned specific examples can be combined in different ways as appropriate, as long as no technical contradictions arise.
　１０：判定装置
　１１：プラント
　１１０：データ取得部
　１２０：選定部
　１３０：判定部
　１４０：報知部
　ＰＳ：プラントシステム
 
  10: Determination device 11: Plant 110: Data acquisition unit 120: Selection unit 130: Determination unit 140: Notification unit PS: Plant system

Claims (8)

  1. 　プラントにおける異常を判定する判定装置であって、
    　前記プラントに設けられた複数のセンサからデータを取得するデータ取得部と、
    　前記データ取得部により取得されたデータと、複数のデータ間における相関を示す統計モデルと、に基づいて、異常の有無を判定する判定部と、を備え、
    　複数の前記センサには、予測対象センサが含まれ、
    　前記判定部は、
    　複数の前記センサのうち、前記予測対象センサと同じ系統に設けられ、且つ、前記予測対象センサと同種の物理量を測定する前記センサから取得されたデータを利用して、前記予測対象センサから取得されるデータの予測値を算出し、当該予測値と、前記予測対象センサから取得されたデータの実測値とに基づいて、異常の有無を判定する、判定装置。     A determination device for determining an abnormality in a plant,
    A data acquisition unit that acquires data from a plurality of sensors installed in the plant;
    a determination unit that determines the presence or absence of an abnormality based on the data acquired by the data acquisition unit and a statistical model that indicates a correlation between a plurality of data,
    The plurality of sensors includes a prediction target sensor;
    The determination unit is
    a determination device that calculates a predicted value of data obtained from the prediction target sensor using data obtained from a sensor among the multiple sensors that is provided in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor, and determines the presence or absence of an abnormality based on the predicted value and an actual measured value of the data obtained from the prediction target sensor.
  2. 　前記データ取得部で取得された複数のデータの中から、前記予測対象センサと同じ系統に設けられ、且つ、前記予測対象センサと同種の物理量を測定する前記センサから取得されたデータを選定する選定部、を更に備える、請求項１に記載の判定装置。 The determination device according to claim 1, further comprising a selection unit that selects, from among the multiple pieces of data acquired by the data acquisition unit, data acquired from a sensor that is provided in the same system as the prediction target sensor and that measures the same type of physical quantity as the prediction target sensor.
  3. 　前記予測値の算出に用いられるデータには、前記予測対象センサから取得されたデータも含まれる、請求項１に記載の判定装置。 The determination device according to claim 1, wherein the data used to calculate the predicted value also includes data acquired from the sensor to be predicted.
  4. 　前記判定部による判定結果を報知する報知部を更に備える、請求項１に記載の判定装置。 The determination device according to claim 1, further comprising a notification unit that notifies the result of the determination by the determination unit.
  5. 　前記判定部は、
    　複数の前記センサのそれぞれについて、当該センサを前記予測対象センサとした場合における異常の有無を個別に判定し、
    　前記報知部は、複数の前記センサのそれぞれに対応する判定結果を個別に報知する、請求項４に記載の判定装置。     The determination unit is
    For each of the plurality of sensors, individually determining whether or not there is an abnormality when the sensor is the prediction target sensor;
    The determination device according to claim 4 , wherein the notification unit individually notifies the determination results corresponding to the plurality of sensors.
  6. 　プラントにおける異常を判定する判定方法であって、
    　前記プラントに設けられた複数のセンサからデータを取得する取得工程と、
    　前記取得工程において取得されたデータと、複数のデータ間における相関を示す統計モデルと、に基づいて、異常の有無を判定する判定工程と、を有し、
    　複数の前記センサには、予測対象センサが含まれ、
    　前記判定工程においては、
    　複数の前記センサのうち、前記予測対象センサと同じ系統に設けられ、かつ、前記予測対象センサと同種の物理量を測定する前記センサから取得されたデータを利用して、前記予測対象センサから取得されるデータの予測値を算出し、当該予測値と、前記予測対象センサから取得されたデータの実測値とに基づいて、異常の有無を判定する、判定方法。     A method for determining an abnormality in a plant, comprising:
    An acquisition step of acquiring data from a plurality of sensors installed in the plant;
    a determination step of determining the presence or absence of an abnormality based on the data acquired in the acquisition step and a statistical model indicating a correlation between a plurality of data,
    The plurality of sensors includes a prediction target sensor;
    In the determination step,
    A method for determining whether or not an abnormality exists, comprising: calculating a predicted value of data to be acquired from a sensor among the plurality of sensors, the data being acquired from a sensor that is provided in the same system as the sensor to be predicted and that measures the same type of physical quantity as the sensor to be predicted; and determining whether or not an abnormality exists based on the predicted value and an actual measured value of the data acquired from the sensor to be predicted.
  7. 　プラントにおける異常を判定する判定装置用のプログラムであって、
    　前記プラントに設けられた複数のセンサからデータを取得するデータ取得処理と、
    　前記データ取得処理において取得されたデータと、複数のデータ間における相関を示す統計モデルと、に基づいて、異常の有無を判定する判定処理と、
    　を、前記判定装置に行わせるものであり、
    　複数の前記センサには、予測対象センサが含まれ、
    　複数の前記センサのうち、前記予測対象センサと同じ系統に設けられ、かつ、前記予測対象センサと同種の物理量を測定する前記センサから取得されたデータを利用して、前記予測対象センサから取得されるデータの予測値を算出し、当該予測値と、前記予測対象センサから取得されたデータの実測値とに基づいて、異常の有無を判定する処理を、前記判定処理として前記判定装置に行わせる、プログラム。     A program for a determination device for determining an abnormality in a plant,
    A data acquisition process for acquiring data from a plurality of sensors installed in the plant;
    A determination process for determining the presence or absence of an abnormality based on the data acquired in the data acquisition process and a statistical model showing correlations between a plurality of data;
    The determination device is caused to perform the following:
    The plurality of sensors includes a prediction target sensor;
    a program that causes the determination device to perform, as the determination process, a process of calculating a predicted value of data acquired from the prediction target sensor using data acquired from a sensor among the multiple sensors that is provided in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor, and determining whether or not there is an abnormality based on the predicted value and an actual measurement value of the data acquired from the prediction target sensor.
  8. 　プラントと、前記プラントにおける異常を判定する判定装置と、を備えるプラントシステムであって、
    　前記判定装置は、
    　前記プラントに設けられた複数のセンサからデータを取得するデータ取得部と、
    　前記データ取得部により取得されたデータと、複数のデータ間における相関を示す統計モデルと、に基づいて、異常の有無を判定する判定部と、を有し、
    　複数の前記センサには、予測対象センサが含まれ、
    　前記判定部は、
    　複数の前記センサのうち、前記予測対象センサと同じ系統に設けられ、かつ、前記予測対象センサと同種の物理量を測定する前記センサから取得されたデータを利用して、前記予測対象センサから取得されるデータの予測値を算出し、当該予測値と、前記予測対象センサから取得されたデータの実測値とに基づいて、異常の有無を判定する、プラントシステム。
     
          A plant system including a plant and a determination device that determines an abnormality in the plant,
    The determination device includes:
    A data acquisition unit that acquires data from a plurality of sensors installed in the plant;
    a determination unit that determines the presence or absence of an abnormality based on the data acquired by the data acquisition unit and a statistical model that indicates a correlation between a plurality of data,
    The plurality of sensors includes a prediction target sensor;
    The determination unit is
    A plant system that uses data acquired from a sensor among the multiple sensors that is provided in the same system as the prediction target sensor and measures the same type of physical quantity as the prediction target sensor to calculate a predicted value of the data acquired from the prediction target sensor, and determines the presence or absence of an abnormality based on the predicted value and an actual measured value of the data acquired from the prediction target sensor.
PCT/JP2023/037446 2022-11-29 2023-10-16 Determination device, determination method, program, and plant system WO2024116622A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2022-190180 2022-11-29
JP2022190180 2022-11-29

Publications (1)

Publication Number Publication Date
WO2024116622A1 true WO2024116622A1 (en) 2024-06-06

Family

ID=91323658

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2023/037446 WO2024116622A1 (en) 2022-11-29 2023-10-16 Determination device, determination method, program, and plant system

Country Status (1)

Country Link
WO (1) WO2024116622A1 (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018104985A1 (en) * 2016-12-08 2018-06-14 日本電気株式会社 Abnormality analysis method, program, and system
WO2019026193A1 (en) * 2017-08-02 2019-02-07 日本電気株式会社 Information processing device, information processing system, information processing method, and recording medium
JP2021038946A (en) * 2019-08-30 2021-03-11 シンフォニアテクノロジー株式会社 Abnormality detection device and creation method of machine learning input data
KR20220074277A (en) * 2020-11-27 2022-06-03 한국과학기술연구원 Monitoring system and method for sensing failure of power generation facilities early

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018104985A1 (en) * 2016-12-08 2018-06-14 日本電気株式会社 Abnormality analysis method, program, and system
WO2019026193A1 (en) * 2017-08-02 2019-02-07 日本電気株式会社 Information processing device, information processing system, information processing method, and recording medium
JP2021038946A (en) * 2019-08-30 2021-03-11 シンフォニアテクノロジー株式会社 Abnormality detection device and creation method of machine learning input data
KR20220074277A (en) * 2020-11-27 2022-06-03 한국과학기술연구원 Monitoring system and method for sensing failure of power generation facilities early

Similar Documents

Publication Publication Date Title
JP5081998B1 (en) Abnormal sign diagnostic apparatus and abnormal sign diagnostic method
US7729882B2 (en) Method and system for assessing performance of control systems
US7853431B2 (en) On-line monitoring and diagnostics of a process using multivariate statistical analysis
US7200520B2 (en) Device and method for monitoring an electric power station
JP6933899B2 (en) Plant operation support equipment, plant operation support method, and plant operation support program
WO2018123144A1 (en) Diagnostic device, diagnostic method, and program
JP2010506331A (en) Defect detection method and system in manufacturing plant
JP5081999B1 (en) How to display abnormal sign diagnosis results
JP6088131B2 (en) Turbine performance diagnostic system and method
JP2013516674A (en) Method and apparatus for monitoring plant equipment performance and predicting failures
CN101676559B (en) Method of alarm mask generation and condition monitoring of wind turbines
BR112015009029B1 (en) System and process of monitoring a set of components of an equipment
EP3584657B1 (en) Risk assessment device, risk assessment method, and risk assessment program
JP2018190245A (en) Facility equipment abnormality diagnosis system
JP6523815B2 (en) Plant diagnostic device and plant diagnostic method
EP3584656A1 (en) Risk assessment device, risk assessment method, and risk assessment program
WO2018055808A1 (en) Sensor diagnosis device, sensor diagnosis method, and program
US10429828B2 (en) Plant simulation device and plant simulation method with first parameter adjustable at start and second parameter adjustable during operation of the plant
CN110431496B (en) Evaluation device, evaluation system, evaluation method, and computer-readable storage medium
WO2024116622A1 (en) Determination device, determination method, program, and plant system
CN117092437A (en) LED display device for monitoring health state of transformer in power grid
US7610159B2 (en) Multivariable mass flow rate transmitter system and flow rate setting tool
JP6554162B2 (en) Power plant performance evaluation method and power plant performance evaluation program
JP2021111207A (en) Plant performance evaluation system and plant performance evaluation method
CN116384526A (en) Important parameter fault early warning method and system for nuclear power plant system or equipment