JP2021033915A - Computer program, dam inflow prediction program, and dam inflow prediction system - Google Patents

Abstract

To solve the problem in which, for example, when a phenomenon to be predicted such as dam inflow is predicted, an induction model has difficulty in prediction at the time of floods, and a deduction model has difficulty in prediction at the time of flat water.SOLUTION: A computer program for predicting a phenomenon to be predicted includes: a first step of acquiring input data related to the phenomenon; a second step of determining whether an induction model will be used and whether a deduction model will be used by comparing the input data with a predetermined threshold value; a third step of outputting first prediction data related to the phenomenon using the deduction model when it is determined that the deduction model will be used; a fourth step of outputting second prediction data related to the phenomenon using the induction model when it is determined that the induction model will be used; and a fifth step of displaying and outputting at least one of the first prediction data and the second prediction data.SELECTED DRAWING: Figure 5

Description

The present invention relates to a computer program, and more particularly to a program for predicting a dam inflow and a dam inflow prediction system.

In dam management, determining the discharge from the dam is an important task.
For example, during floods, when the water is discharged from the dam, it should not cause a sudden rise in the water level of the downstream river, should not exceed the planned flood flow of the downstream river, and should not exceed the maximum flood level of the dam reservoir. Determine the series.

On the other hand, during normal water (when not flooding), it is necessary to perform operations for a different purpose than during flooding. For example, in the case of a power generation dam, it is necessary to keep the water level of the dam reservoir as high as possible to improve the power generation efficiency and to produce a specified amount of power during the time when the power demand is high.
For these flood and flat water operations, it is important to predict the inflow of the dam up to several hours ahead.

For such applications, techniques have been developed to predict dam inflow.
First, the "water level management system of a hydroelectric power plant" disclosed in Patent Document 1 obtains weather information and upstream information of a river required at unsteady state by a data collection means and performs a simulation, and the operator can understand the result. It is a system that provides easy guidance, reduces the burden on the operator, and enables operation without relying on a skilled operator during non-steady conditions. In this system, the inflow amount of the dam is calculated by processing the upstream water level data of the river for a predetermined time of landing delay.

Further, the "water level prediction method, water level prediction program and water level prediction device" disclosed in Patent Document 2 is a method using a neural network for predicting the water level of a river, dam or sewage, and appropriately inputs input data. This is a method aimed at predicting the water level with sufficiently high accuracy and reducing the amount of calculation by extracting the water level.

This method uses a training data extraction step and a neural network to extract training data of the water level and the rainfall and water level that are highly correlated with the water level prediction point and the water level at that time, which are necessary for learning the water level prediction point and the water level at a certain time. It is characterized by sequentially providing a water level prediction point, an extracted rainfall amount having a high correlation with the water level at the relevant time, and a training data learning step for learning the training data of the water level.

Japanese Unexamined Patent Publication No. 2012-244884 JP-A-2019-095240

In the invention disclosed in Patent Document 1, a model is used in which a time series shifted by a predetermined delay time is created based on the flow time series observed in the river upstream of the dam at the time of flood, and the time series is used as the inflow amount of the dam. .. This model can be said to be one of the mathematical models based on the laws of physics. Hereinafter, the mathematical model based on the laws of physics will be referred to as a "deductive model".

In general, this deductive model builds a time-evolving model by expressing the laws of physics with differential equations and then describing them in a computer program. When predicting, the data of the phenomenon that causes the phenomenon of the prediction target is input to the constructed deduction model, and the prediction data of the phenomenon of the prediction target is output.

On the other hand, in the invention disclosed in Patent Document 2, a neural network model is used. This model can be said to be one of the mathematical models based on past case data and statistics. Artificial intelligence (AI) for supervised learning is one of these mathematical models. Hereinafter, these mathematical models will be referred to as "induction models".

This induction model builds a model by machine learning the data of the phenomenon to be predicted and the data that correlates with the phenomenon. When predicting, the data of the phenomenon that correlates with the phenomenon of the prediction target is input to the constructed induction model, and the prediction data of the phenomenon of the prediction target is output.

As described above, there are two methods for predicting the inflow of dams: the deduction model and the induction model.
Here, the advantage of the induction model is that even if the cause of the phenomenon to be predicted is unknown, this model can be applied if there is data having a high correlation with the phenomenon to be predicted. However, a sufficient amount of historical data is needed to know the correlation. The disadvantage is that it is difficult to predict when the phenomenon to be predicted occurs very rarely or has never been observed in the past. For example, it is difficult to make reliable predictions for floods that exceed the maximum history.

On the other hand, the strengths and weaknesses of the deduction model are the opposite of the strengths and weaknesses of the induction model above. For example, in a non-flood state (flat water state), the flow rate fluctuates due to the influence of inhuman factors (snow melting, spring water, etc.) and human factors (power generation water, agricultural water, etc.). These effects are negligible during floods with high flow rates, but cannot be ignored during flat water with low flow rates. Therefore, in the deduction model, it is difficult to make a prediction with a small error in normal water for a phenomenon caused by an unidentifiable cause.

As described above, the induction model had difficulty in predicting the flood time, and the deduction model had difficulty in predicting the time of flat water.
The present invention has been made in view of the above-mentioned problems, and in order to solve the above-mentioned problems, it is possible to output a prediction with a small error regardless of whether the water is flat or flood. Provides a dam inflow forecasting program.

As a computer program for predicting a phenomenon to be predicted, such as a dam inflow, according to the present invention, the first step of acquiring input data related to the phenomenon is compared with the input data and a predetermined threshold. The second step of determining whether to use the deduction model and the induction model, and the first prediction related to the phenomenon using the deduction model when it is determined to use the deduction model. The third step of outputting data, the fourth step of outputting the second prediction data related to the phenomenon using the induction model when it is determined to use the induction model, the first prediction data and the first It is characterized by having a fifth step of displaying and outputting at least one of the prediction data of 2.

According to the present invention, the distinction between flat water and flood is automatically determined, and in the case of flat water, the induction model is used, and in the case of flood, the deduction model is used. Also, it is possible to provide a dam inflow prediction program that makes it possible to output a prediction with a small error.

It is a schematic block diagram which shows the structure of the computer system for realizing the Example which concerns on this invention. It is a sequence diagram which shows the process executed by the computer system for realizing the Example which concerns on this invention. It is a schematic block diagram which shows the software structure of the dam inflow prediction program. It is a sequence diagram which shows the processing of the dam inflow amount prediction program. It is a figure which shows the 1st GUI by the dam inflow prediction program. It is a figure which shows the 2nd GUI by the dam inflow prediction program.

Hereinafter, examples for carrying out the present invention will be described with reference to the accompanying drawings.

FIG. 1 is a schematic block diagram showing a configuration of a computer system for realizing an embodiment of the present invention.
The computer system of this embodiment is composed of a prediction server 110, a data server 180, and a client terminal 170 connected to the Internet 160.

The prediction server 110 includes a memory 120, a CPU (Central Processing Unit) 130, and an HDD (Hard Disk Drive) 140.

A dam inflow prediction program 121 is deployed in the memory 120. The dam inflow prediction program 121 is configured by an instruction to the CPU 130.

The CPU 130 performs processing such as calculation according to the instruction of the dam inflow amount prediction program 121, access to the data server 180, and response to a request from the client terminal 170. The process executed by the dam inflow amount prediction program 121 is actually executed by the CPU 130 according to the instruction described in the dam inflow amount prediction program 121.
The HDD 140 stores the processing result of the dam inflow amount prediction program 121.

The client terminal 170 is a computer having a function of operating a web browser, such as a PC (Personal Computer).

The data server 180 stores weather data 181 and river observation data 182 and map data 183, and distributes the respective data in response to requests from the prediction server 110 and the client terminal 170.

Next, various types of data stored in the data server 180 will be described.
The meteorological data 181 is data of a past meteorological observation time series or a future meteorological prediction time series, and is stored as time series mesh data. That is, space-time is defined by four axes of time, altitude, latitude and longitude, and the start point, end point and resolution are defined for each axis, and the space-time is divided into cells based on the definition, and each cell is divided into cells. The value is stored.

The user can specify the time, altitude, latitude and longitude and get the corresponding values. However, the defined axes are not limited to four. For example, if the data is defined only on the ground surface, it is composed of three axes of time, latitude, and longitude. Alternatively, if it is a forecast, the initial time axis is additionally used.

The river observation data 182 is data of a past river observation time series or a future river prediction time series. This includes river water level, river flow rate, dam inflow, dam discharge and dam storage position water level at the observation point. In addition, the river observation data 182 is stored in a tabular format in which the observation elements are columns, dates, and times. Here, the future river prediction time series is, for example, the planned release time series of the dam located upstream.

The map data 183 is map image data, and may be implemented in accordance with, for example, WMTS (Web Map Tile Service). Further, as the map data 183, data such as a river channel crossing map and a river channel longitudinal map are included.

FIG. 2 is a sequence diagram showing a process executed by a computer system for realizing the embodiment according to the present invention.
In step 201 (S201), the prediction server 110 sends a request to the data server 180 to acquire meteorological data 181 and river observation data 182.

In step 202 (S202), the prediction server 110 executes the dam inflow prediction program 121. The dam inflow amount prediction result data output as the processing result is stored in the HDD 140.

In step 203 (S203), the client terminal 170 sends a request to the prediction server 110 to provide the prediction result at an arbitrary timing. In response to this request, the prediction server 110 returns the latest dam inflow amount prediction result data stored in the HDD 140 to the client terminal 170.

In step 204 (S204), the client terminal 170 sends a request to the data server 180 and acquires the map data 183.

In step 205 (S205), the client terminal 170 draws and displays a GUI (Graphical User Interface) on the web browser based on the prediction result data acquired in step 203 (203) and the map data acquired in step 204 (S204). To do. This GUI will be described later with reference to FIGS. 5 and 6.

In step 206 (S206), when the user operates the selection of the deduction model or the induction model by the radio button (540 shown in FIG. 5) provided in the GUI on the client terminal 170, the operation result is predicted. It is sent to the server 110.

In step 207 (S207), the prediction server 110 executes the dam inflow amount prediction program 121 and updates the dam inflow amount prediction result data.

In step 208 (S208), the client terminal 170 sends a request to the prediction server 110 at an arbitrary timing, and acquires the updated dam inflow amount prediction result data.

In step 209 (S209), the client terminal 170 draws and displays the GUI on the web browser based on the update data acquired in step 209 (S209) and the map data acquired earlier.

FIG. 3 is a schematic block diagram showing a software configuration of the dam inflow prediction program 121.
The dam inflow prediction program 121 is composed of a control module 310, an induction model module 320, a deduction model module 330, a pre-judgment module 340, a post-judgment module 350, and a data access module 360. The exchange between each module is realized by calling a function.

The control module 310 acquires meteorological data 181 and river observation data 182 from the data access module 360. Further, the control module 310 issues instructions to the induction model module 320, the deduction model module 330, the pre-determination module 340, and the post-determination module 350 to create dam inflow prediction result data. The created dam inflow prediction result data will be stored in the HDD 140.

The induction model module 320 inputs the meteorological data 181 and the river observation data 182 and outputs the dam inflow prediction result data. Here, the induction model module 320 creates a dam inflow prediction result based on past observation data (weather data 181 and river observation data 182). As the logic of the induction model module, for example, a machine learning model such as a neural network model, a random forest, or a multiple regression model, or an ensemble model combining these machine learning models may be used.

Further, the past observation data used for machine learning may be limited to the data at the time of normal water. For example, a case where the river flow rate at a specific point is equal to or less than a predetermined threshold value (flood flow rate) may be defined as "flat water", and learning may be performed using only the data at that time. Due to this limitation, it can be expected that the accuracy of the created induction model will be improved.

The deduction model module 330 inputs the meteorological data 181 and the river observation data 182 and outputs the dam inflow prediction result data. Here, as the deduction model module 330, a computer program that describes the physical laws regarding runoff and river flow is used. Specifically, a centralized runoff model such as a storage function, or a combined model of a distributed runoff model and a one-dimensional river channel model can be used.

If the planned discharge of the upstream dam is known in advance, the time series of the planned discharge may be given as the lateral inflow amount or the upstream end flow rate of the one-dimensional river channel model. The parameters of the distributed runoff model (soil layer thickness, roughness, etc.) are adjusted so that the flood period can be accurately reproduced from the past observation data (meteorological data 181 and river observation data 182). However, when the river flow at a specific point exceeds a predetermined threshold (flood flow), it is defined as "flood". In addition, a series of periods from flat water to flood and return to flat water is defined as one "flood event". Adjust the above parameters using the data of multiple flood events.

Each of the pre-determination module 340 and the post-determination module 350 has a threshold value for determination, and if it is below the threshold value, it is determined to be flat water, and if it exceeds the threshold value, it is determined to be flood. Here, the threshold value is, for example, a water level at a certain point, a flow rate at a certain point, or a combination of these conditions.

The data access module 360 sends a request to the data server 180 and acquires meteorological data 181 and river observation data 182 from the data server 180.

FIG. 4 is a sequence diagram showing the processing of the dam inflow prediction program 121. This processing sequence can be operated by executing it at regular intervals.
In step 401 (S401), the control module 310 issues a data request to the data access module 360 and acquires the meteorological data 181 and the river observation data 182 from the data access module 360.

In step 402 (S402), the control module 310 passes the weather data 181 and the river observation data 182 to the pre-determination module 340, and whether or not to use the induction model module 320 to the pre-determination module 340 and the deduction model module 330. To determine whether or not to use.

In the following cases a) to c), the induction model module 320 is used.
a) In the latest river observation data 182, when the river flow rate at a specific point is less than or equal to a predetermined threshold (flood flow rate) b) In the previous dam inflow prediction result data, the river flow rate at a specific point is predetermined. When it becomes less than or equal to the threshold (flood flow rate) c) When the user performs the deduction model selection operation from the selection of the deduction model or the deduction model by the radio button (540 shown in FIG. 5) provided in the GUI.

In the following cases d) to f), the deduction model module 320 is used.
d) In the latest river observation data 182, when the river flow rate at a specific point exceeds a predetermined threshold (flood flow rate) e) In the previous dam inflow prediction result data, the river flow rate at a specific point is predetermined. When the threshold (flood flow rate) is exceeded f) When the user selects the deduction model from the selection of the deduction model or the deduction model by using the radio button (540 shown in FIG. 5) provided in the GUI.

In step 403 (S403), the control module 310 causes the induction model module 320 to predict the dam inflow amount. When meteorological data 181 and river observation data 182 are input to the induction model module 320, dam inflow prediction result data according to the induction model is output.

In step 404 (S404), the control module 310 causes the deduction model module 330 to predict the dam inflow amount. When the meteorological data 181 and the river observation data 182 are input to the deduction model module 330, the dam inflow prediction result data according to the deduction model is output.

In step 405 (S405), the control module 310 causes the post-determination module 350 to select either the prediction result output by the induction model module 320 or the prediction result output by the deduction model module 330.
If the pre-determination module 340 determines in step 402 that only one of the model modules will be used, the prediction result of the model module determined to be used is selected.

On the other hand, if the pre-determination module 340 determines in step 402 that both of the two model modules will be used, the post-determination module 350 makes the following determination. That is, the absolute value of the difference between the dam inflow amount at the current time T1 in the prediction made at the previously predicted time T0 and the dam inflow amount at the time T1 stored in the river observation data 182 is obtained, and this value is set to "previous time". The prediction result of the model module with the smaller one is adopted.
The control module 310 stores the adopted prediction result data in the HDD 140.

FIG. 5 is a diagram showing a first GUI by the dam inflow prediction program 121.
The first GUI is composed of 500 web pages. The web page 500 is displayed on the web browser of the client terminal 170. Further, the web page 500 is displayed at the same time as the web page 600 by the second GUI described later, and the user can operate either screen.

The web page 500 displays a time-series graph 510 of actual and predicted precipitation, and a time-series graph 520 of actual and predicted dam inflow, dam discharge and dam reservoir water level. The horizontal axis of these two graphs is common, and the date and time are shown.
Further, on the horizontal axis, an icon 530 (inverted triangle mark) indicating the initial time of the latest forecast is shown. The graph at the time before the time indicated by the icon 530 shows the observation result, and the graph at the time after the icon 530 shows the prediction result.

In the actual and predicted precipitation time series graph 510, the rainfall intensity 511 is shown as a bar graph, and the cumulative precipitation 512 is shown as a line graph. Following convention, the vertical axis of the graph is downward. The vertical axis on the right side of the graph shows the rainfall intensity, and the vertical axis on the left side of the graph shows the accumulated precipitation.

On the other hand, in the time series graph 520 of the actual and predicted dam inflow, dam discharge and dam reservoir water level, the dam inflow 521, the dam discharge 522 and the dam reservoir water level 523 are shown as line graphs, respectively. Following convention, the vertical axis of the graph is upward. The vertical axis on the left side of the graph shows the flow rate, and the vertical axis on the right side of the graph shows the water level.

Further, on the web page 500, a radio button 540 indicating a selection state between the deduction model and the induction model is displayed (in FIG. 5, the upper right of the screen). As a default value (default), select the check box 543 indicating "automatic selection".

In the case of the default value (default), if the model module selected by the pre-judgment module 340 is the induction model module 320, the label "induction model" 541 is automatically displayed on and the label "deduction model" is displayed. 542 is automatically displayed as off. On the other hand, when the model module selected by the pre-judgment module 340 is the deduction model module 330, the label "induction model" 541 is automatically displayed as off, and the label "deduction model" 542 is automatically displayed as on. Become.

The user can switch the check box 543 indicating "automatic selection" on and off at any time. When turned off, the user can also select either the label "Induction Model" 541 or the label "Deduction Model" 542.

FIG. 6 is a diagram showing a second GUI by the dam inflow prediction program 121.
The second GUI consists of a web page 600. Similar to the previous web page 500, the web page 600 is displayed on the web browser of the client terminal 170.

On the web page 600, a map 610, a river crossing diagram 620, a river longitudinal diagram 630, and a time pane 640 are displayed.
First, the time pane 640 displays the current time 641 and the control bar 642. When the user operates the control bar 642, the data can be moved to any time during a certain period.

On the map 610, the icon 611 indicating the dam and the icon 614 indicating the water level gauge are displayed superimposed on the river 612. The map data 183 is acquired from the data server 180.

The river crossing map 620 displays a river channel crossing map at a certain point. In the screen, the left bank is the left bank and the right bank is the right bank, and the river water level 621 is displayed together with the riverbed 622. It also has a user interface for changing the display point.

A river channel profile at a certain point is displayed on the river profile 630. In the screen, the left side is upstream and the right side is downstream, and the river water level 631 is displayed together with the riverbed 632. However, this profile is invalid when displaying the result of the induction model module 320, and is valid when displaying the result of the deduction model module 330. This is because the induction model module 320 predicts only the water level at a predetermined point, whereas the deduction model module 330 obtains the water level and the flow rate of the entire river as a one-dimensional indefinite flow.

As described above, according to the embodiment of the present invention, the distinction between flat water and flood is automatically determined, and in the case of flat water, the induction model is used, and in the case of flood, the deduction model is used. It enables the dam inflow prediction program to output small error predictions, even during floods.

110 Prediction server 120 Memory 121 Dam inflow prediction program 130 CPU
140 HDD
160 Internet 170 Client terminal 180 Data server 181 Meteorological data 182 River observation data 183 Map data 310 Control module 320 Deduction model module 330 Demonstration model module 340 Pre-judgment module 350 Post-judgment module 360 Data access module 500, 600 Web page 510 Actual results and forecasts Precipitation time series graph 511 Rain intensity bar graph 512 Cumulative precipitation line graph 520 Actual and predicted dam inflow, dam discharge and dam reservoir water level time series graph 521 Dam inflow line graph 522 Dam discharge Line graph 523 Dam reservoir water level line graph 530 Icon 540 Radio button 610 Map 620 River crossing map 630 River profile 640 Time pane

Claims (7)

A computer program that predicts phenomena
The first step of acquiring the input data related to the phenomenon, and
A second step of comparing the input data with a predetermined threshold to determine whether to use the deduction model and the induction model.
When it is determined that the deduction model is used, the third step of outputting the first prediction data related to the phenomenon using the deduction model, and
When it is determined that the induction model is used, the fourth step of outputting the second prediction data related to the phenomenon using the induction model, and
A computer program having a fifth step of displaying and outputting at least one of the first prediction data and the second prediction data. An input section that uses observation data related to rivers and weather as input values,
A deduction model unit that predicts the amount of dam inflow with respect to the input value and outputs the first prediction result.
An induction model unit that predicts the amount of dam inflow with respect to the input value and outputs the second prediction result.
A dam inflow prediction system including a display unit that displays at least one of the first prediction result and the second prediction result. The dam inflow prediction system according to claim 2.
A first determination unit for making a comparison determination between the input value and a predetermined threshold value is provided.
The first determination unit is a dam inflow prediction system, characterized in that it determines whether or not to use each of the deduction model unit and the induction model unit according to the result of the comparison determination. The dam inflow prediction system according to claim 3.
It is provided with a second determination unit for comparing and determining the deviation between the first prediction result and the observation data and the deviation between the second prediction result and the observation data.
The second determination unit is a dam inflow prediction system characterized in that the prediction result having the smaller deviation is adopted. The dam inflow prediction system according to any one of claims 2 to 4.
The deduction model part uses a centralized outflow model such as a storage function, or a combined model of a distributed outflow model and a one-dimensional river channel model.
The induction model unit is a dam inflow prediction system characterized by using a machine learning model such as a neural network, a random forest, or a multiple regression model, or an ensemble model in which the machine learning models are combined. A computer-run dam inflow prediction program
Input steps to acquire observational data on rivers and meteorology,
A determination step of comparing the observed data with a predetermined threshold value to determine whether or not to use the deduction model and whether or not to use the induction model.
When it is determined that the deduction model is used, the first prediction step of outputting the first prediction data of the dam inflow amount using the deduction model, and the first prediction step.
When it is determined that the induction model is used, the second prediction step of outputting the second prediction data of the dam inflow amount using the induction model and the second prediction step.
A dam inflow prediction program having a display step for displaying at least one of the first prediction data and the second prediction data. The dam inflow prediction program according to claim 6.
Between the first and second prediction steps and the display step,
A second determination step is performed in which the deviation between the first prediction data and the observation data and the deviation between the second prediction data and the observation data are compared and determined, and the prediction data having the smaller deviation is adopted. Dam inflow prediction program to have.

Images