Driving assistance system, driving assistance method, and driving assistance program

Abstract

To provide a driving assistance system which can present information that makes it easy to judge a validity of a predicted value.SOLUTION: A driving assistance system 1 comprises: a predicted value calculation unit 11 that applies a prediction technique to a measured value of a plant 4 and a related signal of a time evolution mechanism to predict prediction time-series data of the measured value for each prediction technique; a control input value calculation unit 13 that estimates control time-series data of a control input value corresponding to the prediction time-series data; a scenario generation unit 15 that generates a predicted scenario for the prediction time-series data when controlled by a control input value; an evaluation value calculation unit 16 that calculates a control evaluation value when the predicted scenario is controlled by the control input value; and a driving assistance information generating unit 17 that generates driving assistance information for the plant 4 based on the control evaluation value. The driving assistance information generating unit 17 generates the driving assistance information in which the predicted scenario, the control input value, and prediction basis data are associated with each other, and the prediction basis data includes the measurement value and the related signal applied to the prediction.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to an operation assistance system, an operation assistance method, and an operation assistance program that assist the operation of a plant. [Background technology]

[0002] Operation support systems that support the operation of plants such as water supply and sewerage, electricity, buildings, air conditioning, and factory automation (FA) find operating methods that improve the plant's operational efficiency and safety based on plant measurement values, and feed the results back to the monitoring and control device that monitors the plant's status. This allows the monitoring and control device to control the plant using operating methods that improve the plant's operational efficiency and safety.

[0003] The state of the plant whose operation is being assisted by the operation assistance system changes over time, so it is desirable for the operation assistance system to provide the user with information that allows them to understand the state of the plant after a specific time has passed.

[0004] The girder-below water level prediction device described in Patent Document 1 uses machine learning to generate a trained model for predicting the girder-below water level after a specific time period by using data on multiple sets of watershed water levels and bridge girder-below water levels measured at different times, and predicts the girder-below water level after a specific time period. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Publication No. 2022-77704 Summary of the Invention [Problem to be solved by the invention]

[0006] However, the technology of Patent Document 1 does not manage the predicted value when a specific control is performed in association with the basis for the prediction, and therefore has the problem of not being able to present information that allows easy determination of the validity of the predicted value when a specific control is performed.

[0007] The present disclosure has been made in view of the above, and aims to provide a driving assistance system that can present information that allows the validity of a predicted value to be easily determined. [Means for solving the problem]

[0008] To solve the above-mentioned problems and achieve the object, a driving assistance system disclosed herein includes a prediction value calculation unit that predicts, for each prediction method, time series data of measurement values ​​measured in a plant and time series data of related signals corresponding to the time evolution mechanism of the measurement values ​​by applying, to each of multiple prediction methods, time series data of first future prediction values ​​of the measurement values. The driving assistance system disclosed herein also includes a control input value calculation unit that estimates, for each first prediction value, control time series data that is time series data of control input values ​​corresponding to the prediction time series data, and a scenario generation unit that generates, for a combination of the prediction time series data and the control time series data, a prediction scenario that is a scenario of a second prediction value when control is performed with the control input value for the prediction time series data. The driving assistance system disclosed herein also includes an evaluation value calculation unit that calculates, for each control input value, a control evaluation value that is an evaluation value when control is performed with the control input value for the prediction scenario, and a driving assistance information generation unit that generates driving assistance information to be presented to a user as information for assisting plant operation, based on the control evaluation value. The driving assistance information generation unit generates driving assistance information that associates a prediction scenario, a control input value for the prediction scenario, and prediction basis data that indicates the basis for the prediction of the prediction scenario, and the prediction basis data includes time series data of measurement values ​​and time series data of related signals that were applied when predicting the predicted time series data. [Effects of the Invention]

[0009] The driving assistance system according to the present disclosure has an effect of being able to present information that allows the validity of a predicted value to be easily determined. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 is a diagram showing a configuration of a driving assistance system according to an embodiment. [Figure 2] FIG. 1 is a diagram for explaining physical phenomena, measurement values, and control input values ​​in a plant whose operation is supported by an operation support system according to an embodiment. [Figure 3] FIG. 3 is a diagram for explaining an example of the physical phenomenon, the measurement values, and the control input values ​​shown in FIG. [Figure 4] FIG. 10 is a diagram for explaining an example of a process in which the driving assistance system according to the embodiment predicts a predicted value from a measured value. [Figure 5] FIG. 5 is a diagram illustrating an example of the measured values ​​and predicted values ​​shown in FIG. 4 . [Figure 6] FIG. 1 is a diagram illustrating an example of data recorded by a prediction basis data recording unit of a driving assistance system according to an embodiment. [Figure 7] FIG. 1 is a diagram showing an example of data recorded by a control basis data recording unit of a driving assistance system according to an embodiment; [Figure 8] FIG. 1 is a diagram for explaining an example of predicted time series data predicted by the driving assistance system according to the embodiment; [Figure 9] FIG. 1 is a diagram for explaining an example of a prediction accuracy and a prediction lead time calculated by a driving assistance system according to an embodiment. [Figure 10] FIG. 10 is a diagram for explaining an example of transition of a first control input value calculated by the driving assistance system according to the embodiment. [Figure 11] FIG. 10 is a diagram for explaining an example of transition of a second control input value calculated by the driving assistance system according to the embodiment. [Figure 12] FIG. 10 is a diagram for explaining an example of transition of a third control input value calculated by the driving assistance system according to the embodiment. [Figure 13] FIG. 1 is a diagram for explaining an example of a prediction scenario generated by a driving assistance system according to an embodiment. [Figure 14] FIG. 1 is a diagram for explaining an example of a control evaluation value and a control lead time calculated by the driving assistance system according to the embodiment. [Figure 15] FIG. 1 is a diagram showing a first example of a display screen displayed by the driving assistance system according to the embodiment; [Figure 16] FIG. 10 is a diagram showing a second example of a display screen displayed by the driving assistance system according to the embodiment; [Figure 17] FIG. 1 is a diagram for explaining a control evaluation value calculated by a driving assistance system according to an embodiment. [Figure 18] 1 is a flowchart showing a processing procedure of a process executed by a driving assistance system according to an embodiment; [Figure 19] FIG. 1 is a diagram illustrating a configuration example of a processing circuit when the processing circuit included in the driving assistance system according to the embodiment is realized by a processor and a memory. [Figure 20] FIG. 1 is a diagram illustrating an example of a processing circuit when the processing circuit included in the driving assistance system according to the embodiment is configured with dedicated hardware. DETAILED DESCRIPTION OF THE INVENTION

[0011] A driving assistance system, a driving assistance method, and a driving assistance program according to embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0012] Embodiment FIG. 1 is a diagram illustrating a configuration of an operation assistance system according to an embodiment. The operation assistance system 1 is a system that assists in the operation of a plant 4. The operation assistance system 1 predicts a prediction scenario (fluctuation scenario) of a measurement value based on time series data of a measurement value for a time-evolving mechanism (a physical phenomenon in the embodiment) in the plant 4 and time series data of a related signal corresponding to the time-evolving mechanism (a physical phenomenon in the embodiment). For example, the operation assistance system 1 predicts a prediction scenario (fluctuation scenario) of a measurement value based on time series data of a measurement value (e.g., turbidity) for a physical phenomenon (e.g., water turbidity) in the plant 4 and time series data of a related signal (e.g., water temperature) corresponding to the physical phenomenon. Note that, in the embodiment, a case will be described in which the time-evolving mechanism is a physical phenomenon; however, the time-evolving mechanism is not limited to a physical phenomenon, and may be a chemical phenomenon, a biological phenomenon, or the like. Details of the related signal will be described later.

[0013] In the embodiment, the value of data measured for past physical phenomena in the plant 4 is referred to as a measured value. Also, a measured value predicted by the operation assistance system 1 based on past measured values ​​and past related signals is referred to as a predicted value. Also, a predicted value of a measured value (observed value) when the operation assistance system 1 controls with a specific control input value assuming that the predicted value will be realized is referred to as an observed predicted value.

[0014] The driving assistance system 1 predicts time series data of predicted values ​​(hereinafter sometimes referred to as predicted time series data) for each prediction method using multiple prediction methods, and predicts multiple prediction scenarios using one or multiple methods (control prediction methods described later). A prediction scenario is time series data of observed predicted values ​​and is a scenario of future fluctuations of the predicted values.

[0015] The prediction method is a method created based on time series data of measurement values ​​for a physical phenomenon and time series data of related signals corresponding to this physical phenomenon. The prediction method predicts predicted time series data based on the time series data of measurement values ​​and the time series data of related signals. The driving assistance system 1 predicts future predicted time series data by applying the time series data of measurement values ​​up to the present and the time series data of related signals to the prediction method.

[0016] The operation assistance system 1 estimates the best control input value for each predicted value. The control input value is a value input to the plant 4 to control the plant 4. The operation assistance system 1 generates a prediction scenario, which is time series data of observed predicted values, by applying a control prediction method to a combination of the predicted value and the control input value. The control prediction method is a method for predicting the prediction scenario of the plant 4.

[0017] The driving assistance system 1 calculates a control evaluation value for each control input value when control is performed with the control input value for the generated prediction scenario. That is, the driving assistance system 1 calculates a control evaluation value for evaluating a combination of a prediction value and a control input value. The control evaluation value is an evaluation value of the control input value.

[0018] The driving assistance system 1 predicts a predicted value using a plurality of types of prediction methods, and therefore predicts a plurality of types of predicted values. The driving assistance system 1 applies one control input value to the plurality of types of predicted values, thereby calculating a control evaluation value for the control input value. The driving assistance system 1 applies various control input values ​​to the plurality of types of predicted values, thereby calculating a control evaluation value for each control input value.

[0019] The operation assistance system 1 presents to the user operation assistance information including, for example, a prediction scenario predicted to have the highest control evaluation value at each time, the best control input value for this prediction scenario, and data indicating the basis for the prediction when predicting the prediction scenario (hereinafter referred to as prediction basis data). In the operation assistance information, the prediction scenario, the best control input value, and the prediction basis data are associated with each other. The operation assistance information is information for operation assistance of the plant 4 that is presented to the user.

[0020] The prediction basis data is time-series data used to generate the prediction method. Specifically, the prediction basis data is time-series data of measurement values ​​and related signals for the same physical phenomenon. In other words, the physical phenomenon corresponding to the time-series data of measurement values ​​of the plant 4 and the physical phenomenon corresponding to the time-series data of related signals that serve as the prediction basis data are the same physical phenomenon. The operation assistance system 1 records the time-series data of related signals and the time-series data of measurement values ​​that correspond to the same physical phenomenon in association with each other.

[0021] The prediction method is generated based on measurement values ​​and related signals. The driving assistance system 1 presents the measurement values ​​and related signals used to generate the prediction method to the user as prediction basis data when a predicted value is predicted using the prediction method.

[0022] Here, physical phenomena in the plant 4 will be described. For example, if the plant 4 is located at a dam into which water flows in from a river and the operation assistance system 1 predicts the turbidity of water stored in the dam, the physical phenomenon may include turbidity propagation. For example, turbidity propagation methods include water propagation by advection and water propagation by diffusion. Whether water propagation by advection or water propagation by diffusion is dominant varies depending on the water temperature at each depth of the dam.

[0023] For example, if the water below a dam is cold and the water above is warmer than the water below, and the water flowing into the dam is warmer, advection, in which water flows only in the upper layer, will be dominant. Also, if the temperature difference between the water below and the water above the dam is small and the water flowing into the dam is close in temperature to the water above and below, diffusion will be dominant. Also, if the water below a dam is cold and the water above is warmer than the water below, and the water flowing into the dam is much colder than the water below, total circulation will occur, with the water sinking into the lower layer of the dam, stirring up sediments at the bottom of the dam and causing extremely high turbidity.

[0024] The prediction scenario used by the driving assistance system 1 is a scenario generated based on time-series data of measurement values ​​and time-series data of related signals corresponding to the measurement values. Therefore, by checking the contents of the measurement values ​​and related signals included in the prediction basis data, the user can understand what physical phenomenon the prediction value is based on.

[0025] For example, when the driving assistance system 1 predicts the turbidity of water in a dam, the measured value is turbidity, etc., and the related signal is water temperature, etc. In this case, the driving assistance system 1 presents to the user driving assistance information including predicted time series data of turbidity, a turbidity prediction scenario, a control input value for turbidity, and time series data of past turbidity and water temperature, which are prediction basis data.

[0026] This allows the user to refer to the presented prediction scenario and the past time series data of turbidity and water temperature corresponding to this prediction scenario. Therefore, the user can understand the association between the prediction scenario and the time series data of turbidity and water temperature, which are examples of physical phenomena. As a result, the user can determine the validity of the prediction time series data while referring to the physical phenomena, and can control the operation of the plant 4 based on the validity of the prediction scenario. In addition, the user can understand the association between the prediction scenario and the control input value. In addition, the user can understand the association between the prediction time series data and the physical phenomena.

[0027] The driving assistance information includes multiple combinations of predicted time-series data, predicted scenarios, control input values, and prediction basis data, which allows the user to consider control based on the driving assistance information even when the prediction of the control input value that is predicted to have the highest control evaluation value turns out to be incorrect.

[0028] Whether the prediction scenario is wrong may be subjectively determined by the user, or may be determined and notified to the user by the driving assistance system 1. The driving assistance system 1 determines that the prediction scenario is wrong, for example, when the difference (deviation range) between the latest predicted value predicted based on the current value or information up to the current value and the past predicted value predicted at an earlier timing based on the measured value at that time is larger than a specific value.

[0029] The driving assistance information may include at least one of the following information (1) to (7). (1) Forecasting method used to forecast the forecast scenario (2) Forecast lead time, which is the time from the current time to the time when the forecast value becomes characteristic. (3) Prediction accuracy of predicted values (4) Control lead time, which is the time from the current time until the control input value reaches a characteristic time. (5) Control evaluation value, which is the evaluation value of the control input value (6) Data that served as the basis for calculating the control input value (hereinafter referred to as control basis data) (7) Transition of control input values ​​to be presented

[0030] The multiple prediction scenarios may be prediction scenarios for each time (prediction time) predicted using the same prediction method, or may be prediction scenarios at the same time predicted using different prediction methods. The driving assistance system 1 displays the prediction scenarios in a graph, for example. In addition, the driving assistance system 1 displays recommended control input values ​​in a message, for example.

[0031] The operation assistance system 1 is connected to a monitoring and control device 3 and an input / output device 5. The monitoring and control device 3 is a device that monitors and controls the plant 4. The monitoring and control device 3 collects monitoring and control data used for monitoring and controlling the plant 4. The monitoring and control device 3 extracts measurement values ​​and related signals, which are data measured in the plant 4, from the data included in the monitoring and control data. The plant 4 monitored by the monitoring and control device 3 may be any plant, such as a water supply and sewage plant, an electric power plant, a building plant, an air conditioning plant, or an FA plant. The monitoring and control device 3 transmits the measurement values ​​and related signals acquired from the plant 4 to the operation assistance system 1.

[0032] The operation assistance system 1 uses measurement values ​​received from the monitoring and control device 3 to assist the operation of the plant 4. The operation assistance system 1 may be applied to assist the operation of any plant 4, such as water supply and sewerage, electricity, buildings, air conditioning, and factory automation. Note that, although the embodiment will be described assuming that the plant 4 is a plant that takes in water from a dam, the plant 4 may be any type of plant.

[0033] The driving assistance system 1 includes a time-series data recording unit 21, a prediction method recording unit 22, a prediction basis data recording unit 23, a control estimation method recording unit 24, and a control basis data recording unit 25. The driving assistance system 1 also includes a prediction value calculation unit 11, a prediction lead time calculation unit 12, a control input value calculation unit 13, a control lead time calculation unit 14, a scenario generation unit 15, an evaluation value calculation unit 16, and a driving assistance information generation unit 17.

[0034] The time-series data recording unit 21 records time-series data of measurement values ​​and time-series data of related signals of the plant 4 received from the monitoring and control device 3. The prediction method recording unit 22 records multiple types of prediction methods (prediction models). The prediction methods recorded by the prediction method recording unit 22 include multiple types of prediction methods for calculating predicted values ​​of measurement values ​​from time-series data of measurement values ​​and related signals, and observation prediction methods for calculating observed predicted values ​​when the predicted values ​​are controlled using control input values. The multiple types of prediction methods are used by the predicted value calculation unit 11, and the observation prediction methods are used by the evaluation value calculation unit 16. The predicted value calculated by the predicted value calculation unit 11 is the first predicted value, and the observed predicted value calculated by the evaluation value calculation unit 16 is the second predicted value.

[0035] The prediction basis data recording unit 23 records, via tags, the correspondence between the prediction method and the time-series data used to predict the measurement value. That is, the prediction basis data recording unit 23 records information indicating the type of prediction method and tags corresponding to the time-series data used for the prediction in association with each other.

[0036] The prediction basis data recording unit 23 also stores information about the time-series data used in the prediction, such as the measurement period of the measurement values ​​and related signals used in the prediction, the type of related signals, and tags corresponding to the time-series data. The tags are information that identify the time-series data and are expressed using any symbol, number, natural language, etc.

[0037] The control estimation method recording unit 24 records a plurality of types of control estimation methods (control estimation models). The control basis data recording unit 25 records, via tags, the correspondence between the control estimation method and the time-series data used to estimate the control input value. That is, the control basis data recording unit 25 records information indicating the type of control estimation method and tags corresponding to the time-series data used to estimate the control input value in association with each other.

[0038] In addition, the control estimation method recording unit 24 includes, as information on the time series data used to estimate the control input value, the measurement period of the measurement values ​​and related signals used to estimate the control input value, the type of related signal, and a tag corresponding to the time series data.

[0039] The predicted value calculation unit 11 reads time series data of the measurement values ​​and related signals from the time series data recording unit 21. The predicted value calculation unit 11 also reads a prediction method from the prediction method recording unit 22 and reads prediction basis data from the prediction basis data recording unit 23. The predicted value calculation unit 11 calculates a future predicted value of the measurement value by applying the prediction method and the prediction basis data to the time series data of the measurement values ​​and related signals. The predicted value calculation unit 11 also calculates the prediction accuracy of the predicted value using the time series data of the measurement values ​​and the prediction basis data when executing the prediction method. The method of calculating the prediction accuracy will be described later.

[0040] The predicted value calculation unit 11 sequentially reads out the multiple types of prediction methods recorded in the prediction method recording unit 22, and also sequentially reads out the prediction basis data corresponding to the read prediction methods from the prediction basis data recording unit 23. The predicted value calculation unit 11 applies the prediction methods and the prediction basis data sequentially to the time-series data of the measurement values, thereby calculating predicted values ​​and prediction accuracy using the multiple types of prediction methods.

[0041] Here, an example of a method for calculating a predicted value by the predicted value calculation unit 11 will be described. For example, if time series data of a plurality of measurement values ​​(signals) from time tn to time t is expressed as x t,t-n Let x t,t-n is the time series data of the measurement values ​​and related signals. t,t-n In this case, time series data including x t,t-n Based on this, the predicted value y from time t+1 to time t+m in the future is calculated. t+m,t+1 Forecasting method to predict f(x t,t-n ), the predicted value y t+m,t+1 is y t+m =f(x t,t-n ) can be calculated using the prediction method f(x t,t-n) is recorded in the prediction method recording unit 22.

[0042] Forecasting method f(x t,t-n ) may be a prediction method based on equations of physical, chemical, or biological phenomena, or a prediction method based on statistics. t,t-n ) may be a prediction method based on machine learning or a prediction method based on matching similar past data with current phenomena.

[0043] The prediction basis data recording unit 23 records prediction basis data in association with a prediction method generated mainly based on statistical processing or machine learning processing for creating a prediction method from data. That is, the prediction basis data recording unit 23 records information in which a prediction method is associated with prediction basis data. The prediction basis data recorded by the prediction basis data recording unit 23 includes time series data x of measurement values ​​and related signals used when learning a prediction method for a predicted value. s (0≦s≦N). The time series data x used to train the prediction method s The prediction basis data includes time series data of measurement values ​​and related signals.

[0044] The predicted value calculation unit 11 calculates the current measured value x t,t-n and the predicted value y t+m,t+1 and match it with the prediction basis data for each prediction method, and find similar data x s+m,s-n Extract the data as the basis for prediction. Similar data x s+m,s-n is the time series data of the forecast basis data used to generate the forecast method, corresponding to the period used to predict the forecast value and the predicted period. The time series data of the forecast basis data extracted by the forecast value calculation unit 11 is the forecast basis data presented to the user.

[0045] Here, an example of a method for calculating the prediction accuracy by the prediction value calculation unit 11 will be described. The prediction value calculation unit 11 calculates the prediction accuracy by using time series data x s Based on (0≦s≦N), the time series data of past forecast values ​​is time series data x s,s-nIf the future time series data is time series data x s+m,s+1 The distribution P(x s+m,s+1 |x s,s-n ) where the time series data x s is the forecasting method f(x t,t-n ) is the time series data x of the measurement values ​​used to learn s In addition, the time series data x s,s-n is the time series data of the measurement value from time sn to time s, and the time series data x s+m,s+1 is the predicted time series data from the present to the future time s+m.

[0046] When calculating the prediction accuracy, the prediction value calculation unit 11 calculates the prediction accuracy by using x s,s-n and the realization of x s+m,s+1 Based on the estimated value of p(x s+m,s+1 |x s,s-n The predicted value calculation unit 11 calculates x s+m,s+1 To estimate this, we use estimation methods such as the maximum likelihood method.

[0047] The forecast value calculation unit 11 transmits the calculated forecast time series data to the forecast lead time calculation unit 12, the control input value calculation unit 13, and the scenario generation unit 15. The forecast value calculation unit 11 also transmits the calculated forecast accuracy to the evaluation value calculation unit 16. The forecast value calculation unit 11 also transmits forecast basis data extracted from the forecast time series data to the driving support information generation unit 17.

[0048] The predicted lead time calculation unit 12 calculates a predicted lead time, which is the time from the current time (prediction start time), at which the predicted value is predicted, to the time at which the predicted value is characteristic (for example, the time at which the predicted value peaks), based on the predicted time series data. That is, the predicted lead time calculation unit 12 calculates the difference between the time at which the predicted value is characteristic and the current time as the predicted lead time.

[0049] The current time in the embodiment is the time when prediction starts, and is not necessarily the current time itself. In other words, the current time is the first prediction time in the prediction period, and is not necessarily the time when the prediction process is executed. Therefore, when predicting a predicted value after time Tx, which is a future time, time Tx becomes the current time. For example, when the driving assistance system 1 predicts a predicted value after 1:00 on January 1st, 1:00 on January 1st is the current time. The predicted lead time calculation unit 12 transmits the calculated predicted lead time to the driving assistance information generation unit 17.

[0050] The control input value calculation unit 13 receives the predicted time series data from the predicted value calculation unit 11. The control input value calculation unit 13 also reads out the control estimation method from the control estimation method recording unit 24 and reads out the control basis data from the control basis data recording unit 25.

[0051] The control input value calculation unit 13 applies a control estimation method and control basis data to the predicted time series data to calculate (estimate) time series data of control input values ​​(hereinafter, sometimes referred to as control time series data) for making the predicted value fall within an allowable range. Furthermore, when calculating the control input value, the control input value calculation unit 13 calculates the estimation accuracy of the control input value using the predicted time series data and the control basis data.

[0052] The control input value calculation unit 13 sequentially reads out the multiple types of control estimation methods recorded in the control estimation method recording unit 24, and also sequentially reads out the control basis data corresponding to the read control estimation methods from the control basis data recording unit 25. The control input value calculation unit 13 sequentially applies the control estimation methods and the control basis data to the time series data of the measurement values ​​and related signals, thereby calculating the control time series data and the estimation accuracy of the control input value using the multiple types of control estimation methods.

[0053] The control input value calculation unit 13 transmits the calculated control time series data to the control lead time calculation unit 14 and the scenario generation unit 15. In addition, the control input value calculation unit 13 transmits the calculated estimation accuracy to the evaluation value calculation unit 16. In addition, the control input value calculation unit 13 extracts a portion of the control basis data used to calculate the control input value that corresponds to the control time series data, and transmits the extracted control basis data to the driving support information generation unit 17.

[0054] Here, an example of a method for calculating the control input value by the control input value calculation unit 13 will be described. For example, if the time series data of the signal of the predicted value (signal) from time t to time t+m is y t+m,t The control input value calculation unit 13 receives y t+m,t In this case, the time series data including the predicted value (signal) is received. t+m,t Based on this, the optimal control input value u from time t to future time t+m is calculated. t+m,t The control estimation method used to estimate is called the control estimation method g(y t+m,t ), the control input value is u t+m,t =g(y t+m,t ) can be calculated by the control estimation method g(y t+m,t ) is recorded in the control estimation method recording unit 24.

[0055] Control estimation method g(y t+m,t ) may be a rule-based control estimation method, or a control estimation method that uses a simulator to optimize the control input value for the predicted time series data. t+m,t ) may be a control estimation method based on statistics or a control estimation method based on machine learning. t+m,t ) may be a control estimation method based on matching similar past data with current predicted time series data.

[0056] The control basis data recording unit 25 records the control basis data in association with the control estimation method generated mainly based on statistical processing or machine learning for creating the control estimation method from data. The control basis data recorded by the control basis data recording unit 25 is the time series data x of the control input value used when learning the control estimation method. s (0≦s≦N). The control basis data includes time series data of control input values ​​and related signals. Hereinafter, the time series data x used in learning the control estimation method will be referred to as s is sometimes called control learning time series data.

[0057] The control input value calculation unit 13 calculates the time series data y t+m,t and the estimated control input value u t+m,t and the control basis data for each control estimation method are matched, and similar data x s+m,s Extract the data as the basis for prediction. Similar data x s+m,s is the time series data of the control basis data used to generate the control prediction method, corresponding to the period used to estimate the control input value (the period during which the control input value was estimated). The time series data of the control basis data extracted by the control input value calculation unit 13 is the control basis data presented to the user.

[0058] Here, an example of a method for calculating the estimation accuracy by the control input value calculation unit 13 will be described. The control input value calculation unit 13 calculates the estimation accuracy by using the time series data x s Based on (0≦s≦N), the past forecast value is the forecast value y s+m,s When the control input value is the control input value u s+m,s The distribution of the conditional probability P(u s+m,s |y s+m,s ) is learned. Here, the time series data y s is the control estimation method g(y t+m,t ) is the time series data of the predicted values ​​used in learning the time series data y s+m,s is the predicted time series data from time s to time s+m, and the control input value u s+m,s is the control time series data from time s to time s+m.

[0059] When calculating the estimation accuracy, the control input value calculation unit 13 calculates y s+m,s The realization of u s+m,s Based on the estimated value and p(u s+m,s |y s+m,s The control input value calculation unit 13 calculates u s+m,s The estimation accuracy estimated by the control input value calculation unit 13 is used by the evaluation value calculation unit 16 when calculating the control evaluation value.

[0060] The control lead time calculation unit 14 calculates a control lead time, which is the time from the current time to a time (a time of interest) when the control input value is characteristic, based on the control time-series data. That is, the control lead time calculation unit 14 calculates the difference between the current time and the time when the control input value is characteristic, as the control lead time. Examples of the characteristic time when the control input value is characteristic include the time when the control input value is next changed, the time when the control input value is maximized, and the time when the control input value is minimized. The control lead time calculation unit 14 transmits the estimated control lead time to the driving assistance information generation unit 17.

[0061] The scenario generator 15 generates a prediction scenario, which is time series data of observed predicted values ​​when the prediction time series data is controlled by the control time series data, based on the prediction time series data and the control time series data. That is, the scenario generator 15 generates a prediction scenario when the prediction value is controlled by the control input value.

[0062] The scenario generator 15 generates a forecast scenario for each combination of a forecast value and a control input value. That is, the scenario generator 15 generates a forecast scenario for each combination of a characteristic forecast value in each forecasting method and an optimal control input value for each forecast value.

[0063] For example, if there are M types of forecasting methods (M is a natural number), each forecasting method has one characteristic forecasted value, so there are M characteristic forecasted values ​​in total. Also, there is one optimal control input value for each forecasted value, so there are M control input values ​​in total. Therefore, the scenario generation unit 15 generates M x M forecasting scenarios, which are combinations of M forecasted values ​​and M control input values.

[0064] Since there is one optimal control input value for one predicted value, M of the M × M prediction scenarios are optimal prediction scenarios. The optimal prediction scenario is a prediction scenario in which a predicted value is correct and an optimal control input value is input for this correct predicted value. Prediction scenarios other than the optimal prediction scenario are prediction scenarios in which a predicted value is not correct but a control input value that would be used if the predicted value were correct is input. In the plant 4, if the predicted value is correct, the plant will be in a state similar to the optimal prediction scenario, but the predicted value will not necessarily be correct. In the embodiment, by providing the user with prediction scenarios in which the predicted value is not correct, the user can control the operation of the plant 4 while also considering the case in which the predicted value is not correct.

[0065] The scenario generator 15 receives control input values ​​corresponding to various predicted values ​​calculated by the control input value calculator 13. For example, the scenario generator 15 receives the control input value u corresponding to the predicted value A1 calculated by the control input value calculator 13. A1 , the control input value u for the predicted value B1 B1 , and the control input value u for the predicted value C1 C1 Receive.

[0066] The driving assistance system 1 predicts that the predicted value A1 will be realized, and sets the control input value u A1 In this case, the prediction may be wrong and the future measurement value may be the measurement value corresponding to the predicted value B1. That is, the driving assistance system 1 may calculate the optimal control input value u A1 However, in reality, the control input value uA1 In addition, the prediction may be wrong and the future measured value may be the measured value corresponding to the predicted value C1.

[0067] In these cases, the scenario generation unit 15 generates various prediction scenarios consisting of combinations of predicted values ​​and control input values, with the aim of having the evaluation value calculation unit 16 evaluate which control evaluation value will be the observed predicted value of the measurement value, which is used as an indicator for operational management.

[0068] The scenario generation unit 15 generates prediction scenarios for all possible combinations of, for example, the multiple pieces of prediction time series data calculated by the prediction value calculation unit 11 and the control time series data calculated by the control input value calculation unit 13. The scenario generation unit 15 transmits the generated prediction scenarios to the evaluation value calculation unit 16.

[0069] The evaluation value calculation unit 16 calculates a control evaluation value when control is performed with a control input value for the generated prediction scenario. The evaluation value calculation unit 16 calculates the control evaluation value based on the prediction scenario, the prediction accuracy of the prediction value, the estimation accuracy of the control input value, and the control evaluation value for each observed prediction value. Details of the control evaluation value for each observed prediction value will be described later. The evaluation value calculation unit 16 transmits the control evaluation value and the prediction accuracy to the driving assistance information generation unit 17.

[0070] The driving assistance information generator 17 determines driving assistance information to be presented to the user based on the estimated control evaluation value. For example, the driving assistance information generator 17 selects the predicted scenario with the highest control evaluation value as the best scenario, and generates driving assistance information including the selected best scenario as driving assistance information to be presented to the user.

[0071] Furthermore, the driving assistance information generation unit 17 determines a main scenario based on prediction accuracy. For example, the driving assistance information generation unit 17 determines the predicted scenario with the highest prediction accuracy as the main scenario. The driving assistance information generation unit 17 selects the predicted scenario with the highest prediction accuracy as the main scenario, and generates driving assistance information including the selected main scenario as driving assistance information to be presented to the user.

[0072] Furthermore, the driving assistance information generator 17 selects, for example, the predicted scenario with the lowest control evaluation value as the worst scenario, and generates driving assistance information including the selected worst scenario as driving assistance information to be presented to the user. The driving assistance information generator 17 transmits the driving assistance information including at least one of the best scenario, the main scenario, and the worst scenario to the input / output device 5.

[0073] In the embodiment, the driving assistance system 1 generates various prediction scenarios using multiple types of prediction methods. Therefore, the prediction scenarios generated by the driving assistance system 1 include a best scenario, a main scenario, and a worst scenario.

[0074] The best scenario is a scenario in which external factors that have a negative effect on the measurement values, which are operational control items of plant 4, have the least impact on the measurement values. For example, if the measurement value, which is an operational control item, is turbidity, the best scenario is, among the predicted scenarios, the scenario with the lowest peak value of turbidity, the scenario with the shortest time for turbidity to exceed the threshold, the scenario with the smallest accumulated value of rainfall during a specific period, etc. Note that the operational control items can be set arbitrarily by the user for the time series data.

[0075] Furthermore, if the measurement value that is an operational management item is rainfall, among the forecast scenarios, the scenario with the lowest peak rainfall value, the scenario with the shortest time for rainfall to exceed the threshold, or the scenario with the smallest cumulative rainfall value during a specific period may be used.

[0076] The worst-case scenario is the scenario in which external factors that adversely affect the measured values, which are the operational management items of the plant, have the greatest impact on the measured values. The main scenario is the scenario with the highest probability of occurrence (prediction accuracy) among multiple fluctuation scenarios predicted based on information up to the current time.

[0077] In this way, the predicted scenarios generated by the driving assistance system 1 include a best scenario, a main scenario, and a worst scenario, and the user can evaluate the best scenario, the main scenario, and the worst scenario.

[0078] The driving assistance system 1 may be realized by a single PC (Personal Computer), or at least some of the components included in the driving assistance system 1 may be connected via a network.

[0079] The input / output device 5 receives and outputs driving assistance information from the driving assistance information generation unit 17 of the driving assistance system 1. The input / output device 5 is, for example, a display device such as a liquid crystal monitor. In this case, the input / output device 5 provides the driving assistance information to the user by displaying the driving assistance information. The input / output device 5 may also receive instructions from the user. The input / output device 5 displays information in accordance with the instructions received from the user. For example, when the input / output device 5 receives an instruction from the user to display predicted value basis data, it displays the predicted value basis data. When the input / output device 5 receives an instruction from the user to display control value basis data, it displays the control value basis data.

[0080] FIG. 2 is a diagram for explaining physical phenomena, measurement values, and control input values ​​in a plant whose operation is assisted by an operation assistance system according to an embodiment. In the plant 4, various physical phenomena occur, and measurement values ​​corresponding to the physical phenomena are measured and control input values ​​are input. In FIG. 2, physical phenomena P1 to P6 occur, measurement values ​​M1 to M5 are measured, and a control input value U i ~U iv The figure shows a case where the physical phenomena P1 to P6, the measurement values ​​M1 to M5, and the control input value U i ~U iv is an example, and various physical phenomena occur in each plant 4, various measurement values ​​are taken, and various control input values ​​are input.

[0081] For example, when a physical phenomenon P1 occurs, a measurement value M1 corresponding to the physical phenomenon P1 is measured. Then, when the physical phenomenon P1 occurs, a physical phenomenon P2 occurs, and a measurement value M2 corresponding to the physical phenomenon P2 is measured.

[0082] When the measured values ​​M1 and M2 are measured, the appropriate control input value U i ,U ii is input to the monitoring control device 3, and an appropriate control input value U i ,U ii The plant 4 is controlled by the following. In the operation assistance system 1, the user can refer to the measured values ​​M1, M2 and the past measured values ​​M3 to M5 to calculate the control input value U i ,U ii is input to the monitoring control device 3.

[0083] This allows the physical phenomenon P2 and the control input value U i ,U ii A physical phenomenon P3 occurs according to the above, and a measurement value M3 corresponding to the physical phenomenon P3 is measured. When the physical phenomenon P3 occurs, a physical phenomenon P4 may occur. When the physical phenomenon P4 occurs, a measurement value M3 corresponding to the physical phenomenon P4 is measured.

[0084] When the measured value M3 is measured, the appropriate control input value U iii is input to the monitoring control device 3, and an appropriate control input value U iii The plant 4 is controlled by the following. In the operation assistance system 1, the user can refer to the measured values ​​M1 to M3 and the past measured values ​​M4 and M5 to calculate the control input value U iii is input to the monitoring control device 3.

[0085] This allows the physical phenomenon P3 or the physical phenomenon P4 and the control input value U iii A physical phenomenon P5 occurs according to the above, and a measurement value M4 corresponding to the physical phenomenon P5 is measured. When the measurement value M4 is measured, an appropriate control input value U iv is input to the monitoring control device 3, and an appropriate control input value U ivThe plant 4 is controlled by the following. In the operation assistance system 1, the user can refer to the measured values ​​M1 to M4 and the past measured value M5 to calculate the control input value U iv is input to the monitoring control device 3. As a result, the physical phenomenon P5 and the control input value U iv A physical phenomenon P6 occurs in response to the above, and a measurement value M5 corresponding to the physical phenomenon P6 is measured. After this, a physical phenomenon (not shown) occurs in response to the physical phenomenon P6.

[0086] The measured values ​​M1 to M5 may be predicted values. For example, if the physical phenomenon is rainfall and the measured value M1 is a rainfall measurement value M1x, the physical phenomenon of rainfall and the rainfall measurement value M1x may be predicted.

[0087] The driving assistance system 1 may, for example, iii When generating driving assistance information for the control input value U, the measurement values ​​M1 to M3 may be predicted values ​​or actual measurement values. iii are predicted values, so the measured values ​​M4 and M5 are predicted values. In this way, the measured values ​​before the stage at which the control input values ​​are input may be predicted values ​​or actual measured values, but the measured values ​​after the stage at which the control input values ​​are input are predicted values.

[0088] Here, the physical phenomena P1 to P6, the measured values ​​M1 to M5, and the control input value U i ~U iv 3 is a diagram for explaining an example of the physical phenomena, measurement values, and control input values ​​shown in FIG. 2. In FIG. 3, physical phenomena P1 to P6, measurement values ​​M1 to M5, and control input value U are shown in a case where the plant 4 is a plant that takes water from a dam. i ~U iv An example of this will be described.

[0089] For example, physical phenomenon P1 is rainfall, and physical phenomenon P2 is river flow (advection). River flow is the flow of water downstream from a river to a dam. Furthermore, physical phenomenon P3 is reservoir inflow (advection), and physical phenomenon P4 is reservoir inflow (diffusion). Reservoir inflow (advection) is the inflow of suspended matter (turbidity) from the river to the dam due to advection, and reservoir inflow (diffusion) is the inflow of suspended matter from the river to the dam due to diffusion. Physical phenomenon P5 is coagulation, and physical phenomenon P6 is filtration. Coagulation is the coagulation of suspended matter at the dam's intake. Filtration is the filtering of water at the intake.

[0090] For example, the control input value U i is the control value that controls the depth of the intake (Intake depth control value UX i ) and the control input value U ii is the control value that controls the amount of sprayed pesticide (pesticide spray control value UX ii ) and the control input value U iii is the control value that controls the amount of medicine injected (medicine injection control value UX iii ) and the control input value U iv is the control value that controls the amount of filtered water (filtered water amount control value UX iv )

[0091] Measurement value M1 is the measurement of rainfall (rainfall measurement value M1x), measurement value M2 is the measurement of the amount of water inflow from the river to the dam (inflow measurement value M2x), measurement value M3 is the measurement of turbidity at the intake (intake turbidity measurement value M3x), measurement value M4 is the measurement of turbidity at the sedimentation basin (sedimentation basin turbidity measurement value M4x), and measurement value M5 is the measurement of turbidity at the filter basin (filter basin turbidity measurement value M5x).

[0092] Next, an example of a process for predicting a predicted value by applying a prediction method to a measured value will be described. Fig. 4 is a diagram for explaining an example of a process for predicting a predicted value from a measured value by the driving assistance system according to the embodiment.

[0093] Here, a case will be described in which the driving assistance system 1 predicts three predicted values ​​Sa3, Sb3, and Sc3 using three types of prediction methods Aa, Bb, and Cc. Prediction method Aa is a prediction method that predicts the predicted value Sa3, which is a predicted value of the measured value M3, from the measured value M2 and the related signal C2. Furthermore, prediction method Bb and prediction method Cc are prediction methods that predict the predicted values ​​Sb3 and Sc3, which are predicted values ​​of the measured value M3, from the measured value M3 and the related signal C3. The measured value M2 and the related signal C2 are information acquired for the physical phenomenon P2. Furthermore, the measured value M3 and the related signal C3 are information acquired for the physical phenomena P3 and P4.

[0094] The predicted value calculation unit 11 of the driving assistance system 1 predicts the predicted value Sa3 by applying the measurement value M2 and the related signal C2 to the prediction method Aa. The predicted value calculation unit 11 also predicts the predicted value Sb3 by applying the measurement value M3 and the related signal C3 to the prediction method Bb, and predicts the predicted value Sc3 by applying the measurement value M3 and the related signal C2 to the prediction method Cc. In this way, the predicted value calculation unit 11 predicts various predicted values ​​by applying various prediction methods to various measurement values.

[0095] Since prediction method Aa and prediction methods Bb and Cc use different measurement values ​​and related signals to predict predicted values, the predicted value Sa3 and the predicted values ​​Sb3 and Sc3 are different values. Also, since prediction method Bb and prediction method Cc are different prediction methods, the predicted value Sb3 and the predicted value Sc3 are different values.

[0096] The predicted value calculation unit 11 may use a prediction method for predicting the measured value M3 from the measured value M1. That is, the predicted value calculation unit 11 is not limited to using a prediction method for predicting, from a measured value, a measured value at the same stage as the measured value or one stage later than the measured value. The predicted value calculation unit 11 may use a prediction method for predicting, from a measured value, a measured value two or more stages later than the measured value.

[0097] Here, an example will be described of the measured values ​​M2 and M3 and the predicted values ​​Sa3, Sb3, and Sc3 described in Fig. 4. Fig. 5 is a diagram for explaining an example of the measured values ​​and predicted values ​​shown in Fig. 4.

[0098] The predicted value calculation unit 11 predicts the intake turbidity predicted value Sa3x, which is an example of the predicted value Sa3, by applying the inflow measurement value M2x, which is an example of the measurement value M2, and the valve opening / closing signal C2x, which is an example of the related signal C2, to the prediction method Aa. The valve opening / closing signal C2x is a signal that indicates the opening / closing amount of the valve.

[0099] The predicted value calculation unit 11 also predicts the intake turbidity predicted value Sb3x, which is an example of the predicted value Sb3, by applying the intake turbidity measured value M3x, which is an example of the measured value M3, and the water temperature signal C3x, which is an example of the related signal C3, to the prediction method Bb. The water temperature signal C3x is a signal indicating the water temperature.

[0100] In addition, the prediction value calculation unit 11 predicts the intake turbidity prediction value Sc3x, which is an example of the prediction value Sc3, by applying the intake turbidity measurement value M3x, which is an example of the measurement value M3, and the water temperature signal C3x, which is an example of the related signal C3, to the prediction method Cc.

[0101] Here, there will be explained data recorded by the prediction basis data recording unit 23 and the control basis data recording unit 25. Fig. 6 is a diagram showing an example of data recorded by the prediction basis data recording unit of the driving assistance system according to the embodiment.

[0102] The prediction basis data recording unit 23 records prediction method information and time series data information. Fig. 6 shows a case where the prediction basis data recording unit 23 records prediction method information 51 and 52 and time series data information 5A1, 5A2, and 5B1.

[0103] Prediction method information 51 is information on prediction method Aa, and prediction method information 52 is information on prediction method Bb. Prediction method information 51 includes information for identifying prediction method Aa and tags (information for identifying time series data) of time series data information used when prediction method Aa was generated. Here, a case is shown in which prediction method information 51 includes tags T1 to T3.

[0104] The prediction method information 52 includes information for identifying the prediction method Bb and the tag of the time-series data information used when the prediction method Bb was generated. Here, the prediction method information 52 includes tags T4 to T6.

[0105] The time series data information 5A1, 5A2, and 5B1 are information on time series data of measurement values ​​and related signals. Note that the time series data information 5A1, 5A2, and 5B1 have the same configuration, so here, the configuration of the time series data information 5A1 will be described.

[0106] The time-series data information 5A1 includes information on the period during which the time-series data of the measurement values ​​and related signals indicated in the time-series data information 5A1 were measured, information identifying the related signals corresponding to the measurement values, and a tag corresponding to the time-series data information 5A1. The period during which the time-series data of the related signals was measured is the same as the period during which the time-series data of the measurement values ​​was measured.

[0107] 6 shows a case where the measurement period of the time series data 1001 of measurement values ​​is from 10:00 on May 1, 2019 to 17:00 on May 3, 2019. Also, the related signals corresponding to the time series data 1001 are signals s1 to s3, and the tag of the time series data 1001 is tag T1.

[0108] When the predicted value calculation unit 11 predicts a predicted value using the prediction method Aa, it extracts the time-series data of the tags T1 to T3 associated with the prediction method Aa as prediction basis data.

[0109] 7 is a diagram illustrating an example of data recorded by a control basis data recording unit of a driving assistance system according to an embodiment. The control basis data recording unit 25 records control estimation method information and time series data information. FIG. 7 illustrates a case in which the control basis data recording unit 25 records control estimation method information 61 and 62 and time series data information 6A1, 6A2, and 6B1.

[0110] The control and estimation method information 61 is information on the control and estimation method X, and the control and estimation method information 62 is information on the control and estimation method Y. The control and estimation method information 61 includes information for identifying the control and estimation method X and a tag of the time-series data information used when the control and estimation method X was generated. Here, the case is shown where the control and estimation method information 61 includes tags T11 to T13.

[0111] The control and estimation method information 62 includes information for identifying the control and estimation method Y and a tag of the time-series data information used when the control and estimation method Y was generated. Here, the case is shown where the control and estimation method information 62 includes tags T14 to T16.

[0112] The time-series data information 6A1, 6A2, and 6B1 is information on time-series data of control input values ​​and related signals. The related signals of the control input values ​​are signals related to the control input values. The related signals of the control input values ​​are, for example, signals that are referenced or input when the control input values ​​are input. For example, when the control input value is the chemical injection control value UX iii In this case, the related signal is the pH of the water that the medicine has been poured into. Since the time series data information 6A1, 6A2, and 6B1 have the same configuration, only the configuration of the time series data information 6A1 will be described here.

[0113] The time series data information 6A1 includes information on the period during which the time series data of the control input value and related signal of the time series data information 6A1 was input to the plant 4, information for identifying the related signal corresponding to the control input value, and a tag corresponding to the time series data information 6A1. The period during which the time series data of the related signal was referenced or input is the same as the period during which the time series data of the control input value was input.

[0114] 7 shows a case where the input period of the time-series data 2001 of the control value is from 10:00 on May 1, 2019 to 17:00 on May 3, 2019. Also, the related signals corresponding to the time-series data 2001 are signals s3, s8 to s10, and the tag of the time-series data 2001 is tag T11.

[0115] When the control input value calculation unit 13 predicts the control input value using the control estimation method X, it extracts the time-series data of the tags T11 to T13 associated with the control estimation method X as control basis data.

[0116] Next, an example of predicted time series data predicted by the driving assistance system 1 will be described. Fig. 8 is a diagram for explaining an example of predicted time series data predicted by the driving assistance system according to the embodiment. Here, a case will be described in which the driving assistance system 1 generates predicted time series data from the measurement value M3. The horizontal axis of the graph shown in Fig. 8 represents time, and the vertical axis represents the predicted value or the measurement value.

[0117] The scenario generation unit 15 generates predicted time series data At1, Bt1, and Ct1 that indicate predicted trends in predicted values ​​from the current time t1 onward, based on measured values ​​up to the current time t1. The predicted time series data At1 is predicted time series data generated using prediction method Aa. The predicted time series data Bt1 is predicted time series data generated using prediction method Bb. The predicted time series data Ct1 is predicted time series data generated using prediction method Cc.

[0118] The predicted value calculation unit 11 calculates a predicted lead time, which is the time from the current time t1 to the time when the predicted values ​​of the predicted time series data At1, Bt1, and Ct1 are characteristic (here, the time when the predicted values ​​reach their peaks).

[0119] Here, the time when the predicted value of the predicted time series data At1 reaches its peak is time ta, the time when the predicted value of the predicted time series data Bt1 reaches its peak is time tb, and the time when the predicted value of the predicted time series data Ct1 reaches its peak is time tc.

[0120] The predicted value calculation unit 11 calculates ta-t1 as the predicted lead time for the predicted time series data At1. Similarly, the predicted value calculation unit 11 calculates tb-t1 as the predicted lead time for the predicted time series data Bt1, and calculates tc-t1 as the predicted lead time for the predicted time series data Ct1.

[0121] The predicted lead time is displayed to the user as driving assistance information. This allows the user to refer to the predicted lead time when the prediction is correct and the predicted lead time when the prediction is incorrect. By the driving assistance system 1 presenting the predicted lead time to the user, the user can make preparations according to the length of the predicted lead time.

[0122] 9 is a diagram illustrating an example of the prediction accuracy and the prediction lead time calculated by the driving assistance system according to the embodiment. The horizontal axis of the graph shown in FIG. 9 represents the prediction lead time, and the vertical axis represents the prediction accuracy.

[0123] In the embodiment, the forecast value calculation unit 11 calculates the forecast accuracy of the forecast value using the time-series data of the measurement value and the forecast basis data. The forecast value calculation unit 11 calculates the correspondence relationship between the forecast accuracy and the forecast lead time.

[0124] 9 shows the prediction accuracy and prediction lead time of predicted time series data At1-At3, Bt1-Bt3, and Ct1-Ct3. The predicted time series data At1-At3 are predicted using prediction method Aa. The predicted time series data Bt1-Bt3 are predicted using prediction method Bb, and the predicted time series data Ct1-Ct3 are predicted using prediction method Cc.

[0125] The predicted time series data At1, Bt1, and Ct1 are predicted time series data when predicted values ​​from the current time t1 onwards. The predicted time series data At2, Bt2, and Ct2 are predicted time series data when predicted values ​​from the current time t2 onwards. The predicted time series data At3, Bt3, and Ct3 are predicted time series data when predicted values ​​from the current time t3 onwards. The current time t2 is a time later than the current time t1, and the current time t3 is a time later than the current time t2.

[0126] Since the predicted time series data Ct3 is predicted at a later time than the predicted time series data Ct1, the predicted time series data Ct3 has a shorter prediction lead time than the predicted time series data Ct1, as shown in Fig. 9. For example, the driving assistance system 1 may predict the intake turbidity measurement value M3x using the inflow measurement value M2x. In this case, the closer the inflow measurement value M2x is predicted to the time when the predicted value peaks, the higher the prediction accuracy will be, but the shorter the lead time will be.

[0127] Meanwhile, the driving assistance system 1 may predict the intake turbidity measurement value M3x using the rainfall measurement value M1x at an upstream point in the river. Because the rainfall measurement value M1x measured upstream affects the turbidity at the intake after a certain time has passed, the prediction accuracy may be reduced if the rainfall measurement value M1x is predicted close to the time when the predicted value peaks. Furthermore, the lead time becomes shorter as the rainfall measurement value M1x is predicted close to the time when the predicted value peaks.

[0128] 10 is a diagram for explaining an example of transition of the first control input value calculated by the driving assistance system according to the embodiment. The horizontal axis of the graph shown in FIG. 10 represents time, and the vertical axis represents the control input value. In FIG. 10, the control input value UAt1 is calculated when the driving assistance system 1 performs control with the optimal control input value UAt1 from the current time t1 for the predicted time series data At1. i ,U ii ,U iv The control input value UAt1 shows the transition of the control input value U i ,U ii ,U iv Contains:

[0129] In FIG. 10, the control input value calculation unit 13 calculates the control input value U i ,U ii Without changing the control input value U iv The control input value U i ,U ii ,U iv The control input value U iv The predicted lead time is ta1-t1.

[0130] In the driving assistance system 1, there are cases where the measurement values ​​change according to the predicted time series data At1, but also cases where the measurement values ​​change according to the predicted time series data Bt1 or Ct1. For this reason, the control input value calculation unit 13 also calculates the control input values ​​when the measurement values ​​change according to the predicted time series data Bt1 or Ct1.

[0131] 11 is a diagram for explaining an example of transition of the second control input value calculated by the driving assistance system according to the embodiment. The horizontal axis of the graph shown in FIG. 11 represents time, and the vertical axis represents the control input value. In FIG. 11, the control input value U when the driving assistance system 1 controls the predicted time series data Bt1 with the optimal control input value UBt1 from the current time t1 is shown. i ,U ii ,U iv The control input value UBt1 shows the transition of the control input value U i ,Uii ,U iv Contains:

[0132] In FIG. 11, the control input value calculation unit 13 calculates the control input value U i Without changing the control input value U ii starts to change, and at time tb2, the control input value U iv The control input value U i ,U ii ,U iv The control input value U ii The predicted lead time of is tb1-t1, and the control input value U iv The forecast lead time is tb2-t1.

[0133] For example, when the driving assistance system 1 wants to control the amount of filtered water, the driving assistance system 1 sets a filtered water amount control value UX iv Controlling the filtered water volume by itself is not enough. iv and pesticide spray control value UX ii In some cases, the amount of filtered water can be controlled to an optimum amount by controlling the control input value U i ,U ii ,U iv Calculate the following.

[0134] 12 is a diagram for explaining an example of transition of the third control input value calculated by the driving assistance system according to the embodiment. The horizontal axis of the graph shown in FIG. 12 represents time, and the vertical axis represents the control input value. In FIG. 12, the control input value U i ,U ii ,U iv The control input value UCt1 shows the transition of the control input value U i ,U ii ,U iv Contains:

[0135] In FIG. 12, the control input value calculation unit 13 calculates the control input value U i starts to change, and at time tc2, the control input value U ii starts to change, and at time tc3, the control input value U iv The control input value U i ,U ii ,U iv The control input value U i The predicted lead time of is tc1-t1. Also, the control input value U ii The predicted lead time of is tc2-t1, and the control input value U iv The predicted lead time is tc3-t1.

[0136] For example, when the driving assistance system 1 wants to control the amount of filtered water, the driving assistance system 1 sets a filtered water amount control value UX iv Controlling the filtered water volume by itself is not enough. iv , Intake depth control value UX i , and pesticide spray control value UX ii In some cases, the amount of filtered water can be controlled to an optimum amount by controlling the control input value U . Taking such cases into consideration, the driving assistance system 1 adjusts the control input value U so that the transition shown in FIG. i ,U ii ,U iv Calculate the following.

[0137] Fig. 13 is a diagram for explaining examples of prediction scenarios generated by a driving assistance system according to an embodiment. In the graph of prediction scenarios from the current time t1 shown in Fig. 13, the horizontal axis represents time and the vertical axis represents the observed predicted value of the measurement value M5. The nine prediction scenarios shown in Fig. 13 are prediction scenarios that indicate the observed predicted value of the measurement value M5 when various control input values ​​are input to the predicted time-series data of the measurement value M3.

[0138] FIG. 13 shows prediction scenarios 30aa, 30ab, 30ac, 30ba, 30bb, 30bc, 30ca, 30cb, and 30cc as examples of prediction scenarios generated by the scenario generation unit 15 of the driving assistance system 1.

[0139] The three forecast scenarios 30aa, 30ab, and 30ac on the left are forecast scenarios in which control input values ​​are input to forecast values ​​calculated using forecasting method Aa, while the three forecast scenarios 30ba, 30bb, and 30bc in the center are forecast scenarios in which control input values ​​are input to forecast values ​​calculated using forecasting method Bb, and the three forecast scenarios 30ca, 30cb, and 30cc on the right are forecast scenarios in which control input values ​​are input to forecast values ​​calculated using forecasting method Cc.

[0140] The three forecast scenarios 30aa, 30ba, and 30ca in the top row are forecast scenarios in which control is performed using a control input value UAt1 that is optimal for the forecast time series data At1. The three forecast scenarios 30ab, 30bb, and 30cb in the middle row are forecast scenarios in which control is performed using a control input value UBt1 that is optimal for the forecast time series data Bt1. The three forecast scenarios 30ac, 30bc, and 30cc in the bottom row are forecast scenarios in which control is performed using a control input value UCt1 that is optimal for the forecast time series data Ct1.

[0141] For example, when controlled by the control input value UAt1, the control input value U iv When the control input value UBt1 is used, the control input value U ii ,U iv When the control input value UCt1 is used, the control input value U i ,U ii ,U iv is controlled.

[0142] Of the prediction scenarios shown in Figure 13, prediction scenarios 30aa, 30bb, and 30cc indicate prediction scenarios for observed predicted values ​​when predicted time series data for measurement value M3 occurs as predicted and optimal control is performed on the predicted time series data for this generated measurement value M3.

[0143] In the plant 4, there are cases where predicted time series data of the measured value M3 does not occur as predicted. Therefore, the scenario generation unit 15 of the embodiment also generates a prediction scenario of the observed predicted value when control is performed on the predicted time series data of the measured value M3 that is not as predicted (predicted time series data when the prediction is incorrect). This allows the user to refer to both the prediction scenario of the measured value M5 when the prediction is correct and the prediction scenario of the measured value M5 when the prediction is incorrect. In addition, the scenario generation unit 15 calculates the characteristic time of the generated prediction scenario and the control lead time.

[0144] The control lead time is displayed to the user as driving support information. This allows the user to refer to the control lead time when the prediction is correct and the control lead time when the prediction is incorrect. For example, when the control input value U i is the control input value U ii ,U iv However, the driving assistance system 1 presents the control lead time to the user, allowing the user to make preparations according to the length of the control lead time. Also, by referring to the control lead time, the user can consider control input values ​​that have a sufficient lead time in case the prediction turns out to be incorrect.

[0145] 14 is a diagram illustrating an example of a control evaluation value and a control lead time calculated by the driving assistance system according to the embodiment. The horizontal axis of the graph shown in FIG. 14 represents the control lead time, and the vertical axis represents the control evaluation value.

[0146] Fig. 14 shows the control evaluation value and control lead time when predicted time series data At1 is controlled using control input values ​​UAt1 to UAt3. Fig. 14 also shows the control evaluation value and control lead time when predicted time series data Bt1 is controlled using control input values ​​UBt1 to UBt3, and the control evaluation value and control lead time when predicted time series data Ct1 is controlled using control input values ​​UCt1 to UCt3.

[0147] The control input value UAt1 is the optimal control performed on the predicted time series data At1 from the current time t1, the control input value UAt2 is the optimal control performed on the predicted time series data At2 from the current time t2, and the control input value UAt3 is the optimal control performed on the predicted time series data At3 from the current time t3.

[0148] The control input value UBt1 is the optimal control performed on the predicted time series data Bt1 from the current time t1, the control input value UBt2 is the optimal control performed on the predicted time series data Bt2 from the current time t2, and the control input value UBt3 is the optimal control performed on the predicted time series data Bt3 from the current time t3.

[0149] The control input value UCt1 is the optimal control performed from the current time t1 on the predicted time series data Ct1, the control input value UCt2 is the optimal control performed from the current time t2 on the predicted time series data Ct2, and the control input value UCt3 is the optimal control performed from the current time t3 on the predicted time series data Ct3.

[0150] For example, the driving assistance system 1 presents the control input value UCt3 to the user because the control evaluation value of the control input value UCt3 is higher than that of the control input values ​​UCt1 to UCt3. Note that the driving assistance system 1 also presents the control input values ​​UCt1 and UCt2 having lower control evaluation values ​​than the control input value UCt3 in response to a request from the user.

[0151] Furthermore, for example, the driving assistance system 1 presents the control input value UAt1 to the user because the control evaluation value of the control input value UAt1 is higher than that of the control input values ​​UAt1 to UAt3. Note that the driving assistance system 1 also presents the control input values ​​UAt2 and UAt3, which have lower control evaluation values ​​than the control input value UAt1, in response to a request from the user.

[0152] Furthermore, for example, the control input value UBt2 has a higher control evaluation value than the control input values ​​UBt1 to UBt3, so the driving assistance system 1 presents the control input value UBt2 to the user. Note that the driving assistance system 1 also presents the control input values ​​UBt1 and UBt3, which have lower control evaluation values ​​than the control input value UBt2, in response to a request from the user.

[0153] This allows the driving assistance system 1 to present the user with the control evaluation value and control lead time for each current time. For example, the driving assistance system 1 may present the user with the control input value with the highest control evaluation value at each time, or may present the user with the control input value with the longest control lead time. This allows the user to know the control input value with the highest evaluation value at each time, the control input value with the longest control lead time, etc.

[0154] Next, a description will be given of an example of a display screen that the driving assistance system 1 causes to be displayed on the entrance / exit device. Fig. 15 is a diagram showing a first example of a display screen that the driving assistance system according to the embodiment causes to be displayed.

[0155] The driving assistance information generator 17 of the driving assistance system 1 determines driving assistance information based on the estimated control evaluation value and displays the driving assistance information on a display screen 40X. A prediction display button 41, which is a button for displaying prediction basis data, and a control display button 42, which is a button for displaying control basis data, are displayed on the display screen 40X. In addition, a message 43 for the user, predicted time-series data 44, which is a graph of predicted values, and control input value guidance 45 for inputting control input values ​​are displayed on the display screen 40X.

[0156] The predicted time series data 44 is, for example, a graph of predicted time series data as shown in Fig. 8. Fig. 15 shows a case where a graph of predicted time series data for the current time t3 is displayed as the predicted time series data 44. Note that the predicted time series data 44 may be displayed as a graph as shown in Fig. 9.

[0157] The message 43 displays a message for the predicted time series data to be presented to the user from the predicted time series data 44 (here, the predicted time series data Ct3 with the highest prediction accuracy). The message 43 includes, for example, a message indicating a predicted value of the predicted time series data and a message indicating a control input value to be presented for this predicted value. The message indicating the predicted value displays, for example, a message indicating a characteristic time in the predicted time series data and the predicted value at this time. Furthermore, the message indicating the control input value displays a message indicating the time at which the control input value is input, the control input value before change, and the control input value after change.

[0158] The prediction display button 41 is a button that, when pressed, displays prediction basis data. When the prediction display button 41 is pressed, the driving support information generation unit 17 displays prediction basis data for the predicted value (here, predicted time series data Ct3) displayed in the message 43 on the display screen 40X. The driving support information generation unit 17 displays the prediction basis data received from the prediction value calculation unit 11 on the display screen 40X. The driving support information generation unit 17 displays, for example, graphs of time series data of measurement values ​​and related signals as prediction basis data on the display screen 40X.

[0159] The control display button 42 is a button that displays control basis data when pressed. When the control display button 42 is pressed, the driving support information generation unit 17 displays the control basis data for the predicted value (here, predicted time series data Ct3) displayed in the message 43 on the display screen 40X. The driving support information generation unit 17 displays the control basis data received from the control input value calculation unit 13 on the display screen 40X. The driving support information generation unit 17 displays, for example, graphs of time series data of the control input value and related signals as the control basis data on the display screen 40X.

[0160] The control input value guidance 45 is, for example, a graph of the first to third control time-series data as shown in Figs. 10 to 12. In Fig. 15, the control input value Ui ,U ii ,U iv 14. In addition, the control input value guidance 45 may display a graph such as that shown in FIG.

[0161] When a specific predicted time series data is selected by the user from the predicted time series data 44, the driving support information generation unit 17 changes the message 43 to be displayed to a message 43 corresponding to the selected predicted time series data. The driving support information generation unit 17 also changes the control input value guidance 45 to be displayed to the control input value guidance 45 corresponding to the selected predicted time series data. The driving support information generation unit 17 also changes the prediction basis data to be displayed when the prediction display button 41 is pressed to the prediction basis data corresponding to the selected predicted time series data. The driving support information generation unit 17 also changes the control basis data to be displayed when the control display button 42 is pressed to the control basis data corresponding to the selected predicted time series data.

[0162] 16 is a diagram showing a second example of a display screen displayed by the driving assistance system according to the embodiment. When the user presses the prediction display button 41, the driving assistance information generator 17 of the driving assistance system 1 displays a predicted scenario 47 and prediction basis data 48 on a display screen 40Y, which is the second example of the display screen. For example, on the display screen 40Y, a control display button 42, a message 43, predicted time series data 46, which is predicted time series data selected by the user, the predicted scenario 47, and the prediction basis data 48 are displayed.

[0163] The predicted time series data 46 displays the predicted time series data selected by the user (the predicted time series data corresponding to the predicted value displayed in the message 43 on the display screen 40X). For example, if the predicted time series data selected by the user is the predicted time series data Ct3 of the measured value M3, the predicted time series data Ct3 of the measured value M3 is displayed as the predicted time series data 46.

[0164] The forecast scenario 47 is a graph of the forecast scenario corresponding to the forecast time series data selected by the user. The forecast method used to forecast the forecast time series data Ct3 selected by the user is forecast method Cc. Therefore, the forecast scenario when a control input value is input to the forecast time series data Ct3 calculated using forecast method Cc is displayed as the forecast scenario 47.

[0165] The forecast basis data 48 is forecast basis data for the forecast time series data selected by the user. When the user selects the forecast time series data Ct3, data on the physical phenomenon corresponding to the forecast time series data Ct3 is displayed as the forecast basis data 48. That is, when the user selects the forecast time series data Ct3, time series data on the control input value (e.g., intake depth) corresponding to the forecast time series data Ct3 and time series data on the related signal (e.g., water temperature) corresponding to this physical phenomenon are displayed as the forecast basis data 48. This allows the user to easily link the forecast scenario with the physical phenomenon (forecast basis data) that is predicted to develop or occur, and understand the forecast scenario.

[0166] 17 is a diagram for explaining a control evaluation value calculated by the driving assistance system according to the embodiment. The evaluation value calculation unit 16 of the driving assistance system 1 calculates a control evaluation value, which is an evaluation value for a control input value, based on a prediction scenario, the prediction accuracy of the predicted time-series data, the estimation accuracy of the control input value, and the control evaluation value for each observed predicted value.

[0167] 17 shows a formula for calculating the control evaluation value for the control input values ​​UAt1, UBt1, and UCt1. Note that the calculation method for the control evaluation value for the control input values ​​UAt1, UBt1, and UCt1 is similar, so here we will explain the calculation method for the control evaluation value for the control input value UAt1.

[0168] The evaluation value calculation unit 16 calculates the control evaluation value of the control input value UAt1 by summing the control evaluation value when the predicted time series data At1 is realized when controlled with the control input value UAt1, the control evaluation value when the predicted time series data Bt1 is realized when controlled with the control input value UAt1, and the control evaluation value when the predicted time series data Ct1 is realized when controlled with the control input value UAt1.

[0169] Here, the prediction accuracy of the predicted time series data At1 is defined as P(A1), the prediction accuracy of the predicted time series data Bt1 as P(B1), and the prediction accuracy of the predicted time series data Ct1 as P(C1). The estimation accuracy of the control input value UAt1 for the predicted time series data At1 is defined as P(uA1|A1), the estimation accuracy of the control input value UAt1 for the predicted time series data Bt1 as P(uA1|B1), and the estimation accuracy of the control input value UAt1 for the predicted time series data Ct1 as P(uA1|C1). For example, P(uA1|A1) is the distribution of the conditional probability that the control time series data will be the control input value UAt1 when the predicted time series data is the predicted time series data At1.

[0170] The evaluation value calculation unit 16 calculates P(uA1|A1)×P(A1) as the occurrence probability when the predicted time series data At1 is realized when controlled using the control input value UAt1. P(uA1|A1)×P(A1) is the occurrence probability of a prediction scenario in which the predicted time series data At1 is realized when controlled using the control input value UAt1. In other words, P(uA1|A1)×P(A1) is the probability that the observed predicted value becomes the predicted time series data At1 when controlled using the control input value UAt1.

[0171] Similarly, the evaluation value calculation unit 16 calculates P(uA1|B1)×P(B1) as the probability of occurrence when the predicted time series data Bt1 is realized when controlled with the control input value UAt1, and calculates P(uA1|C1)×P(C1) as the probability of occurrence when the predicted time series data Ct1 is realized when controlled with the control input value UAt1.

[0172] Furthermore, the evaluation value calculation unit 16 calculates the calculation result of the predicted value (prediction scenario) as yt+m,t The control input value is u t+m,t Then, the observation prediction method h(y t+m,t ,u t+m,t ) to obtain the observed predicted value v t+m,t v t+m,t =h(y t+m,t ,u t+m,t ) is calculated.

[0173] Observed predicted value v t+m,t is the control input value u t+m,t This is the future observation prediction value when controlled by the observation prediction method h(y t+m,t ,u t+m,t ) is the predicted value y t+m,t and the control input value u t+m,t From the observed predicted value v t+m,t This is an observational prediction method (observational prediction model) that calculates the following.

[0174] This observation prediction method h(y t+m,t ,u t+m,t ) is recorded in the prediction method recording unit 22, and the evaluation value calculation unit 16 reads out and uses the observation prediction method from the prediction method recording unit 22. t+m,t ,u t+m,t ) is the prediction method f(x t,t-n ), any prediction method may be used. That is, the observation prediction method h(y t+m,t ,u t+m,t ) may be a prediction method based on equations of physical, chemical, or biological phenomena, or a prediction method based on statistics. t+m,t ,u t+m,t ) may be a prediction method based on machine learning or a prediction method based on matching similar past data with current phenomena.

[0175] The evaluation value calculation unit 16 calculates the predicted value as y t+m,t The control input value is u t+m,t Let the observed predicted value be v t+m,t In this case, the observation prediction method h(y t+m,t ,u t+m,t ) to obtain the control evaluation value z for each observed and predicted value.t+m,t , z t+m,t =I(v t+m,t ) is calculated as

[0176] 17, z1 indicates the control evaluation value for each observed predicted value when predicted time series data At1 is realized when control is performed with the control input value UAt1, z2 indicates the control evaluation value for each observed predicted value when predicted time series data Bt1 is realized when control is performed with the control input value UAt1, and z3 indicates the control evaluation value for each observed predicted value when predicted time series data Ct1 is realized when control is performed with the control input value UAt1.

[0177] For example, in a case where there are three prediction scenarios corresponding to the predicted time series data At1, Bt1, and Ct1, the evaluation value calculation unit 16 calculates the evaluation value of the control input value UAt1 when the control input value estimated for the prediction scenario of the predicted time series data At1 is the control input value UAt1 from the control evaluation value of the observed prediction value when the predicted prediction scenario occurs and the control evaluation value of the observed prediction value when the unexpected prediction scenario occurs.

[0178] Specifically, the evaluation value calculation unit 16 calculates a control evaluation value z1 of the expected observation predicted value when the control input value UAt1 is set when the prediction scenario of the predicted time series data At1 is realized as predicted. The evaluation value calculation unit 16 also calculates a control evaluation value z2 of the expected observation predicted value when the control input value UAt1 is set when the prediction is wrong and the prediction scenario of the predicted time series data Bt1 is realized. The evaluation value calculation unit 16 also calculates a control evaluation value z3 of the expected observation predicted value when the control input value UAt1 is set when the prediction is wrong and the prediction scenario of the predicted time series data Ct1 is realized. The evaluation value calculation unit 16 then uses the control evaluation values ​​z1 to z3 of the observation predicted value to calculate a control evaluation value E(uA1) when the control input value UAt1 is set.

[0179] Specifically, the evaluation value calculation unit 16 calculates the control evaluation value E(uA1) when the control input value UAt1 is set by E(uA1) ≒ z1 × P(uA1|A1) × P(A1) + z2 × P(uA1|B1) × P(B1) + z3 × P(uA1|C1) × P(C1).

[0180] 18 is a flowchart showing the processing procedure of processing executed by the operation assistance system according to the embodiment. The monitoring and control device 3 collects monitoring and control data from the plant 4, and extracts measurement values ​​and related signals, which are data measured in the plant 4, from the data included in the monitoring and control data. The monitoring and control device 3 transmits the measurement values ​​and related signals to the operation assistance system 1.

[0181] The operation assistance system 1 records the measurement values ​​and related signals collected by the monitoring and control device 3 from the plant 4 (step S10). The predicted value calculation unit 11 calculates a predicted value and prediction accuracy from the measurement values ​​and related signals based on the prediction method and prediction basis data (step S20). The predicted lead time calculation unit 12 calculates a predicted lead time from predicted time series data, which is time series data of the predicted value, and the current time (the start time of the prediction) (step S30).

[0182] The control input value calculation unit 13 calculates an appropriate control input value and the estimation accuracy of the control input value for the predicted time series data based on the control estimation method and the control basis data (step S40). The control lead time calculation unit 14 calculates a control lead time from the control time series data and the current time (step S50).

[0183] The scenario generator 15 generates a forecast scenario from the forecast time series data and the control time series data (step S60). The evaluation value calculator 16 calculates a control evaluation value for each control input value based on the forecast scenario, the forecast accuracy of the forecast value, the estimation accuracy of the control input value, and the control evaluation value for each observed forecast value (step S70).

[0184] The driving support information generator 17 generates driving support information based on the predicted time series data, the prediction accuracy of the predicted value, the predicted lead time, the control time series data, the control lead time, the control evaluation value, the prediction basis data, and the control basis data (step S80).The driving support information generator 17 causes the input / output device 5 to display the driving support information (step S90).

[0185] Next, we will explain the hardware configuration of the driving assistance system 1. The driving assistance system 1 is realized by a processing circuit. The processing circuit may be a processor and memory that executes a program stored in a memory, or may be dedicated hardware.

[0186] FIG. 19 is a diagram illustrating an example of the configuration of a processing circuit included in a driving assistance system according to an embodiment, when the processing circuit is realized by a processor and a memory. The processing circuit 90 illustrated in FIG. 19 includes a processor 91 and a memory 92. When the processing circuit 90 includes the processor 91 and the memory 92, each function of the processing circuit 90 is realized by software, firmware, or a combination of software and firmware. The software or firmware is written as a driving assistance program and stored in the memory 92. In the processing circuit 90, each function is realized by the processor 91 reading and executing the driving assistance program stored in the memory 92. That is, the processing circuit 90 includes the memory 92 for storing the driving assistance program that results in the processing of the driving assistance system 1 being executed. This driving assistance program can also be said to be a program that causes the driving assistance system 1 to execute each function realized by the processing circuit 90. This driving assistance program may be provided by a computer-readable recording medium on which the driving assistance program is recorded, or by other means such as a communication medium.

[0187] The driving assistance program can also be said to be a program that causes the driving assistance system 1 to execute the processes of steps S10 to S90 in Fig. 18. Here, the processor 91 is, for example, a CPU (Central Processing Unit), a processing device, an arithmetic device, a microprocessor, a microcomputer, or a DSP (Digital Signal Processor). Furthermore, the memory 92 is, for example, a non-volatile or volatile semiconductor memory such as a RAM (Random Access Memory), a ROM (Read Only Memory), a flash memory, an EPROM (Erasable Programmable ROM), or an EEPROM (registered trademark) (Electrically EPROM), a magnetic disk, a flexible disk, an optical disk, a compact disk, a minidisk, or a DVD (Digital Versatile Disc).

[0188] FIG. 20 is a diagram illustrating an example of a processing circuit included in the driving assistance system according to the embodiment, configured with dedicated hardware. The processing circuit 93 illustrated in FIG. 20 corresponds to, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or a combination thereof. The processing circuit 93 may be partially implemented with dedicated hardware and partially implemented with software or firmware. In this way, the processing circuit 93 can realize each of the above-described functions by dedicated hardware, software, firmware, or a combination thereof.

[0189] In addition, at least one of the predicted value calculation unit 11, predicted lead time calculation unit 12, control input value calculation unit 13, control lead time calculation unit 14, scenario generation unit 15, evaluation value calculation unit 16, and driving assistance information generation unit 17 of the driving assistance system 1 may be realized by the above-mentioned processing circuit 90 or processing circuit 93.

[0190] In this way, the driving assistance system 1 of the embodiment presents to the user prediction basis data including a prediction scenario, a control input value for the prediction scenario, time series data of measurement values ​​applied when predicting the predicted time series data, and time series data of related signals corresponding to physical phenomena of the measurement values. This allows the driving assistance system 1 to present driving assistance information that allows the user to easily determine the validity of the observed predicted values ​​included in the prediction scenario when control using the control input value is performed.

[0191] Furthermore, because the operation assistance system 1 generates various prediction scenarios using multiple prediction methods, the user can evaluate the best scenario, main scenario, and worst scenario among the prediction scenarios. This allows the user to consider an operation method that will provide high operational efficiency, safety, etc. for the plant 4, and control the monitoring and control device 3. This allows the monitoring and control device 3 to control the plant 4 using an operation method that will provide high operational efficiency, safety, etc. for the plant 4.

[0192] The configurations shown in the above embodiments are merely examples, and may be combined with other known technologies, and parts of the configurations may be omitted or modified without departing from the spirit of the invention.

[0193] Various aspects of the present disclosure are summarized below as appendices.

[0194] (Appendix 1) a prediction value calculation unit that predicts prediction time series data, which is time series data of a first future prediction value of the measurement value, for each of a plurality of prediction methods by applying time series data of measurement values ​​measured in the plant and time series data of related signals corresponding to a time evolution mechanism of the measurement values ​​to each of the plurality of prediction methods; a control input value calculation unit that estimates control time series data, which is time series data of a control input value corresponding to the predicted time series data, for each of the first predicted values; a scenario generation unit that generates, for a combination of the predicted time series data and the control time series data, a prediction scenario that is a scenario of a second predicted value when the predicted time series data is controlled by the control input value; an evaluation value calculation unit that calculates, for each control input value, a control evaluation value that is an evaluation value when the forecast scenario is controlled using the control input value; an operation support information generation unit that generates operation support information, which is information on operation support of the plant to be presented to a user, based on the control evaluation value; Equipped with the driving assistance information generation unit generates the driving assistance information in which the prediction scenario, the control input value for the prediction scenario, and prediction basis data indicating a basis for prediction of the prediction scenario are associated with each other; the prediction basis data includes time series data of the measurement values ​​and time series data of the related signals applied when predicting the predicted time series data; A driving assistance system characterized by: (Appendix 2) the predicted value calculation unit calculates the prediction accuracy of the predicted time series data; The driving assistance information includes the prediction accuracy. 2. A driving assistance system according to claim 1, (Appendix 3) the control input value calculation unit calculates the estimation accuracy of the control time-series data; the evaluation value calculation unit calculates the control evaluation value based on the prediction scenario, the estimation accuracy, the prediction accuracy, and the control evaluation value for each prediction scenario. 3. A driving assistance system according to claim 2, (Appendix 4) a forecast lead time calculation unit that calculates a forecast lead time, which is a time from when the forecast is started until when the first forecast value of the forecast time series data becomes a characteristic time; The driving assistance information includes the predicted lead time. 4. A driving assistance system according to any one of claims 1 to 3. (Appendix 5) The driving assistance information includes the control evaluation value. 5. A driving assistance system according to any one of claims 1 to 4, (Appendix 6) The driving assistance information includes a message indicating the control input value for each time. 6. A driving assistance system according to any one of appendices 1 to 5, (Appendix 7) combinations of the predicted time series data and the control time series data include combinations in which the prediction scenario turns out as predicted and combinations in which the prediction scenario turns out to be different from the prediction, the scenario generation unit generates the prediction scenarios for combinations in which the predictions are met and combinations in which the predictions are not met. 7. A driving assistance system according to any one of claims 1 to 6, (Appendix 8) an operation support system for supporting the operation of a plant, applying time series data of measurement values ​​measured in the plant and time series data of related signals corresponding to time evolution mechanisms of the measurement values ​​to each of a plurality of types of prediction methods, thereby predicting predicted time series data that is time series data of a first future predicted value of the measurement values ​​for each of the prediction methods; a control input value calculation step in which the driving assistance system estimates control time-series data, which is time-series data of a control input value according to the predicted time-series data, for each of the first predicted values; a scenario generation step in which the driving assistance system generates, for a combination of the predicted time series data and the control time series data, a prediction scenario that is a scenario of a second predicted value when the predicted time series data is controlled by the control input value; an evaluation value calculation step of calculating, for each control input value, a control evaluation value that is an evaluation value when the driving assistance system controls the prediction scenario using the control input value; an operation assistance information generation step in which the operation assistance system generates operation assistance information, which is information on operation assistance for the plant to be presented to a user, based on the control evaluation value; Including, In the driving assistance information generating step, the driving assistance system generates the driving assistance information in which the prediction scenario, the control input value for the prediction scenario, and prediction basis data indicating a basis for prediction of the prediction scenario are associated with each other; the prediction basis data includes time series data of the measurement values ​​and time series data of the related signals applied when predicting the predicted time series data; A driving assistance method comprising: (Appendix 9) a prediction value calculation step of applying time series data of measurement values ​​measured in the plant and time series data of related signals corresponding to a time evolution mechanism of the measurement values ​​to each of a plurality of types of prediction methods, thereby predicting prediction time series data that is time series data of a first future prediction value of the measurement values ​​for each of the prediction methods; a control input value calculation step of estimating control time series data, which is time series data of a control input value corresponding to the predicted time series data, for each of the first predicted values; a scenario generation step of generating, for a combination of the predicted time series data and the control time series data, a prediction scenario that is a scenario of a second predicted value when the predicted time series data is controlled by the control input value; an evaluation value calculation step of calculating, for each control input value, a control evaluation value which is an evaluation value when the forecast scenario is controlled by the control input value; an operation support information generating step of generating operation support information, which is information on operation support of the plant to be presented to a user, based on the control evaluation value; on the computer, In the driving assistance information generating step, the driving assistance information is generated by associating the prediction scenario, the control input value for the prediction scenario, and prediction basis data indicating a basis for prediction of the prediction scenario, the prediction basis data includes time series data of the measurement values ​​and time series data of the related signals applied when predicting the predicted time series data; A driving assistance program characterized by: [Explanation of symbols]

[0195] 1 Operation support system, 3 Monitoring control device, 4 Plant, 5 Input / output device, 5A1, 5A2, 5B1, 6A1, 6A2, 6B1 Time series data information, 11 Forecast value calculation unit, 12 Forecast lead time calculation unit, 13 Control input value calculation unit, 14 Control lead time calculation unit, 15 Scenario generation unit, 16 Evaluation value calculation unit, 17 Operation support information generation unit, 21 Time series data recording unit, 22 Forecast method recording unit, 23 Forecast basis data recording unit, 24 Control estimation method recording unit, 25 Control basis data recording unit, 30aa, 30ab, 30ac, 30ba, 30bb, 30bc, 30ca, 30cb, 30cc Forecast scenario, 40X, 40Y Display screen, 41 Forecast display button, 42 Control display button, 43 Messages, 44, 46, At1-At3, Bt1-Bt3, Ct1-Ct3: Forecast time series data, 45: Control input value guidance, 47: Forecast scenario, 48: Forecast basis data, 51, 52: Forecast method information, 61, 62: Control estimation method information, 90, 93: Processing circuit, 91: Processor, 92: Memory, A1, B1, C1, Sa3, Sb3, Sc3: Forecast value, Aa, Bb, Cc: Forecast method, C2, C3: Related signal, C2x: Valve opening / closing signal, C3x: Water temperature signal, M1-M5: Measurement value, M1x: Rainfall measurement value, M2x: Inflow measurement value, M3x: Intake turbidity measurement value, M4x: Settling tank turbidity measurement value, M5x: Filter basin turbidity measurement value, P1-P6: Physical phenomenon, Sa3x, Sb3x, Sc3x Intake turbidity forecast value, T1~T6, T11~T16 tags, Tx, ta, ta1, tb, tb1, tb2, tc, tc1~tc3 times, U i ~U iv ,UAt1~UAt3,UBt1~UBt3,UCt1~UCt3,u A1 ,u B1 ,u C1 Control input value, UX i Intake depth control value, UX iiChemical spray control value, UX iii Chemical injection control value, UX iv Filtration water volume control value, X, Y control estimation method, t1~t3 current time, z1~z3 control evaluation value.

Claims

1. a prediction value calculation unit that predicts prediction time series data, which is time series data of a first future prediction value of the measurement value, for each of a plurality of prediction methods by applying time series data of measurement values ​​measured in the plant and time series data of related signals corresponding to a time evolution mechanism of the measurement values ​​to each of the plurality of prediction methods; a control input value calculation unit that estimates control time-series data, which is time-series data of a control input value corresponding to the predicted time-series data, for each of the first predicted values; a scenario generation unit that generates, for a combination of the predicted time series data and the control time series data, a prediction scenario that is a scenario of a second predicted value when the predicted time series data is controlled by the control input value; an evaluation value calculation unit that calculates, for each control input value, a control evaluation value that is an evaluation value when the forecast scenario is controlled using the control input value; an operation support information generation unit that generates operation support information, which is information on operation support of the plant to be presented to a user, based on the control evaluation value; Equipped with the driving assistance information generation unit generates the driving assistance information in which the prediction scenario, the control input value for the prediction scenario, and prediction basis data indicating a basis for prediction of the prediction scenario are associated with each other; the prediction basis data includes time series data of the measurement values ​​and time series data of the related signals applied when predicting the predicted time series data; A driving assistance system characterized by:

2. the predicted value calculation unit calculates the prediction accuracy of the predicted time series data; The driving assistance information includes the prediction accuracy.

2. The driving assistance system according to claim 1.

3. the control input value calculation unit calculates the estimation accuracy of the control time-series data; the evaluation value calculation unit calculates the control evaluation value based on the prediction scenario, the estimation accuracy, the prediction accuracy, and the control evaluation value for each prediction scenario.

3. The driving assistance system according to claim 2.

4. a forecast lead time calculation unit that calculates a forecast lead time, which is a time from when the forecast is started until when the first forecast value of the forecast time series data becomes a characteristic time; The driving assistance information includes the predicted lead time.

2. The driving assistance system according to claim 1.

5. The driving assistance information includes the control evaluation value.

2. The driving assistance system according to claim 1.

6. The driving assistance information includes a message indicating the control input value for each time.

2. The driving assistance system according to claim 1.

7. combinations of the predicted time series data and the control time series data include combinations in which the prediction scenario turns out as predicted and combinations in which the prediction scenario turns out to be different from the prediction, the scenario generation unit generates the prediction scenarios for combinations in which the predictions are met and combinations in which the predictions are not met.

7. A driving assistance system according to claim 1, wherein the driving assistance system comprises: a driver assistance system;

8. an operation support system for supporting the operation of a plant, applying time series data of measurement values ​​measured in the plant and time series data of related signals corresponding to time evolution mechanisms of the measurement values ​​to each of a plurality of types of prediction methods, thereby predicting prediction time series data that is time series data of a first future prediction value of the measurement values ​​for each of the prediction methods; a control input value calculation step in which the driving assistance system estimates control time-series data, which is time-series data of a control input value corresponding to the predicted time-series data, for each of the first predicted values; a scenario generation step in which the driving assistance system generates, for a combination of the predicted time series data and the control time series data, a prediction scenario that is a scenario of a second predicted value when the predicted time series data is controlled by the control input value; an evaluation value calculation step of calculating, for each control input value, a control evaluation value that is an evaluation value when the driving assistance system controls the prediction scenario using the control input value; an operation assistance information generation step in which the operation assistance system generates operation assistance information, which is information on operation assistance for the plant to be presented to a user, based on the control evaluation value; Including, In the driving assistance information generating step, the driving assistance system generates the driving assistance information in which the prediction scenario, the control input value for the prediction scenario, and prediction basis data indicating a basis for prediction of the prediction scenario are associated with each other; the prediction basis data includes time series data of the measurement values ​​and time series data of the related signals applied when predicting the predicted time series data; A driving assistance method comprising:

9. a prediction value calculation step of applying time series data of measurement values ​​measured in the plant and time series data of related signals corresponding to a time evolution mechanism of the measurement values ​​to each of a plurality of types of prediction methods to predict prediction time series data, which is time series data of a first future prediction value of the measurement values, for each of the prediction methods; a control input value calculation step of estimating control time-series data, which is time-series data of a control input value corresponding to the predicted time-series data, for each of the first predicted values; a scenario generation step of generating, for a combination of the predicted time series data and the control time series data, a prediction scenario which is a scenario of a second predicted value when the predicted time series data is controlled by the control input value; an evaluation value calculation step of calculating, for each control input value, a control evaluation value which is an evaluation value when the forecast scenario is controlled by the control input value; an operation support information generating step of generating operation support information, which is information on operation support of the plant to be presented to a user, based on the control evaluation value; on the computer, In the driving assistance information generating step, the driving assistance information is generated by associating the prediction scenario, the control input value for the prediction scenario, and prediction basis data indicating a basis for prediction of the prediction scenario, the prediction basis data includes time series data of the measurement values ​​and time series data of the related signals applied when predicting the predicted time series data; A driving assistance program characterized by:

Images