Water level prediction system

Abstract

To provide a prediction system for easily predicting river water levels. [Solution] The water level prediction system, which predicts future water levels at a prediction point on a river, comprises a rainfall acquisition unit 1 that acquires future rainfall in the catchment basin for the prediction point, a memory unit 2 that stores parameters of a first-order lag system that inputs rainfall and outputs the water level at the prediction point on the river, and a water level prediction unit 3 that inputs rainfall acquired by the rainfall acquisition unit 1 and calculates the predicted water level at the prediction point on the river from the first-order lag system using the parameters of the first-order lag system stored in the memory unit 2, and the time constant of the first-order lag system is set according to the water level at the prediction point on the river.

Description

[Technical field]

[0001] The present disclosure relates to a water level prediction system. [Background technology]

[0002] There are various water level prediction systems that use AI, neural networks, probability, and statistical methods to predict the water level of rivers (for example, Non-Patent Document 1). These water level prediction systems target the midstream and downstream areas of large rivers.

[0003] Patent Document 1 describes a water level prediction system for small rivers. Patent Document 1 describes a method of estimating the amount of precipitation flowing into a prediction point and predicting the water level based on the estimated amount of precipitation and a river model. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2024-080389 A [Non-patent literature]

[0005] [Non-Patent Document 1] Fujitsu Ltd., AI Water Management Prediction System, <URL:https: / / www.fujitsu.com / jp / products / network / managed-services-network / resilience / river-prediction-ai / > Summary of the Invention [Problem to be solved by the invention]

[0006] Small and medium-sized rivers in mountainous regions or adjacent areas often do not have full-fledged levees like those found in the middle and lower reaches of large rivers. This is because the rivers have steep slopes and flow at high speeds even during heavy rainfall, so they do not flood even without levees.

[0007] Conventional water level forecasts have generally been aimed at first-class rivers managed by the nation, and water level forecasts have not been made for small and medium-sized rivers managed by prefectures. This is because there are a large number of small and medium-sized rivers and, as mentioned above, it has been thought that there is no possibility of flooding.

[0008] However, due to the effects of recent global warming, small and medium-sized rivers such as the Kinugawa River and the Kuma River are flooding, causing unprecedented flood damage. It is expected that in the future, large-scale flood damage that could result in the loss of life will occur frequently in the upper reaches of small and medium-sized rivers. Therefore, there was a demand for an inexpensive and simple water level prediction system that would enable water level prediction in small and medium-sized rivers as well.

[0009] Furthermore, although the water level prediction system disclosed in Patent Document 1 is applicable to small rivers, it is difficult to create a river model with high prediction accuracy.

[0010] The present disclosure has been made in consideration of the above background, and aims to provide a water level prediction system that can easily predict the water level of a river. [Means for solving the problem]

[0011] One aspect of the present disclosure is A water level prediction system for predicting a future water level at a prediction point of a river, comprising: A rainfall acquisition unit that acquires future rainfall in a catchment area for the prediction point; a storage unit that stores parameters of a first-order lag system in which the rainfall amount is input and the water level at the prediction point of the river is output; a water level prediction unit that receives the rainfall acquired by the rainfall acquisition unit as an input and calculates a predicted water level at the prediction point of the river from the first-order lag system using parameters of the first-order lag system stored in the storage unit, The time constant of the first-order lag system is set according to the water level at the prediction point of the river in a water level prediction system. Effect of the Invention

[0012] In the above embodiment, the predicted water level is calculated by a first-order lag system in which the amount of rainfall is input and the water level at the river's prediction point is output, and the time constant of the first-order lag system is set according to the water level at the river's prediction point. Therefore, the predicted water level can be calculated easily.

[0013] As described above, according to the above aspect, a water level prediction system can be provided that can easily predict the water level of a river. [Brief description of the drawings]

[0014] [Figure 1] FIG. 1 is a block diagram showing a configuration of a water level prediction system in a first embodiment. [Diagram 2] FIG. 2 is a diagram showing a schematic diagram of a catchment area of ​​a river for which water levels are to be predicted. [Diagram 3] FIG. 1 is a diagram illustrating a schematic configuration of a satellite radio wave intensity meter. [Figure 4] This is a diagram showing the amount of rainfall near the predicted point and the water level at the predicted point. [Diagram 5] A schematic diagram showing the relationship between water level, river width, and water level decline. [Figure 6] A schematic diagram showing the relationship between river gradient and the decrease in water level. [Figure 7] A schematic diagram showing the relationship between differences in river size and the decline in water levels. [Figure 8] A schematic diagram showing rivers and predicted water level locations. [Figure 9] A schematic diagram showing the delay of rain that falls at point A. [Figure 10]A schematic diagram showing the delay of rain that falls at point B. [Figure 11] A schematic diagram showing the amount of rainfall at the forecast location. [Figure 12] FIG. 2 is a flowchart illustrating the operation of the water level prediction system in the first embodiment. [Figure 13] 1 is a graph showing curves of actual water level and predicted water level. [Figure 14] 1 is a graph showing curves of actual water level and predicted water level. [Figure 15] 1 is a graph showing curves of actual water level and predicted water level. [Figure 16] 13 is a graph showing the relationship between the water level at the prediction point and the value of the time constant. [Figure 17] This is a graph showing rainfall intensity, predicted water level, and actual water level. [Figure 18] 1 is a graph showing predicted water levels and actual water levels. [Figure 19] 1 is a graph showing predicted water levels and actual water levels. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] The water level prediction system predicts the future water level at a prediction point on a river and has a rainfall acquisition unit that acquires future rainfall in the catchment basin for the prediction point, a memory unit that stores parameters of a first-order lag system that inputs rainfall and outputs the water level at the prediction point on the river, and a water level prediction unit that inputs the rainfall acquired by the rainfall acquisition unit and calculates the predicted water level at the prediction point on the river from the first-order lag system using the parameters of the first-order lag system stored in the memory unit, and the time constant of the first-order lag system is set according to the water level at the prediction point on the river.

[0016] In the water level prediction system, the time constant in the first-order lag system may be set based on the drop characteristic of the measured water level at the prediction point of the river after the rain has stopped. The time constant can be set with high accuracy, and the water level prediction accuracy can be improved.

[0017] In the water level prediction system, the rainfall acquisition unit may acquire rainfall forecast data from the Japan Meteorological Agency as the rainfall amount.

[0018] In the water level prediction system, the water level prediction unit may convert rainfall forecast data from the Japan Meteorological Agency into data with equal time intervals and smoothed values, and use the data as the amount of rainfall per unit time. This can improve the prediction accuracy.

[0019] In the above water level prediction system, the rainfall acquisition unit may acquire data on precipitation nowcast (5 minutes) from the Japan Meteorological Agency as the amount of rainfall from the current time to one hour ahead, and may acquire data on the Japan Meteorological Agency's quick analysis rainfall and quick short-term precipitation forecast as the amount of rainfall more than one hour ahead.

[0020] The water level prediction system may further include a water level measurement unit that measures the actual water level at the prediction point of the river, and the water level prediction unit may perform a first correction to the predicted water level, so that the value of the predicted water level at the current time coincides with the value of the actual water level at the current time measured by the water level measurement unit. This can improve the accuracy of the water level prediction.

[0021] In the above water level prediction system, the water level prediction unit may perform a second correction on the predicted water level after the first correction, and the second correction may be performed so that the value of the time change of the predicted water level after the first correction at the current time coincides with the value of the time change of the actual water level at the current time measured by the water level measurement unit. This can further improve the accuracy of the water level prediction.

[0022] In the above water level prediction system, the water level prediction unit may perform both the first correction and the second correction for the water level from the current time to a predetermined time ahead, and may perform the first correction but not the second correction for the water level beyond the predetermined time ahead. This can improve the prediction accuracy of the water level beyond the predetermined time ahead.

[0023] In the above water level prediction system, the water level prediction unit may perform both the first correction and the second correction, and if there is a section in which the predicted water level after the correction exceeds a predetermined upper limit, the first correction may be performed for that section, but the second correction may not be performed to return the value to the original value. This can further improve the accuracy of the water level prediction.

[0024] The water level prediction system may further include a water level measurement unit that measures the actual water level at the prediction point of the river, and the water level prediction unit may correct the predicted water level by multiplying the predicted water level by a coefficient, and the coefficient may be determined so that the sum of the differences between the actual water level measured by the water level measurement unit at a certain time and the predicted water level becomes 0. This can further improve the accuracy of water level prediction.

[0025] The water level prediction system may further include a water level measurement unit that measures the actual water level at the prediction point of the river, and the water level prediction unit may correct the predicted water level by multiplying the predicted water level by a coefficient, and the coefficient may be determined so as to minimize the sum of squares of the difference between the actual water level measured by the water level measurement unit at a certain time and the predicted water level. This can further improve the accuracy of water level prediction.

[0026] The water level prediction system may further include a water level measurement unit that measures the actual water level at the prediction point of the river, and the water level prediction unit may correct the predicted water level, calculate an inflection point of the actual water level, and make the absolute value of the correction amount after the time of the inflection point smaller than the absolute value of the correction amount before the time of the inflection point. This can further improve the accuracy of the water level prediction.

[0027] The water level prediction system may further include a satellite radio wave intensity meter arranged in the water catchment area and receiving radio waves from satellite broadcasting, and the water level prediction unit may calculate the amount of rainfall from the intensity of the radio waves received by the satellite radio wave intensity meter and correct the amount of rainfall acquired by the rainfall acquisition unit using the calculated amount of rainfall. This can improve the accuracy of water level prediction. In particular, in mountainous areas, the radio waves of the weather radar see positions higher than the mountains, and rain clouds at altitudes lower than the mountains are not visible. Therefore, if the amount of rainfall is corrected using the satellite radio wave intensity meter, the accuracy of water level prediction can be improved.

[0028] In the water level prediction system, the water level prediction unit may calculate a delay time for water to reach the prediction point from the rainfall point based on the river flow length from the rainfall point to the prediction point and the flow rate of water flowing through the river, and may correct the rainfall amount for the delay time. This can improve the accuracy of water level prediction.

[0029] The water level prediction system may further include a display unit that displays the predicted water level, the precipitation nowcast (5 minutes), and the precipitation amount based on the quick analysis precipitation amount and the quick short-term precipitation forecast, and the update of the predicted water level may be synchronized with the update of the precipitation nowcast (5 minutes), the quick analysis precipitation amount and the quick short-term precipitation forecast. The predicted water level can be displayed to consumers in an easy-to-understand manner.

[0030] (First embodiment) 1. Water level prediction system configuration Fig. 1 is a block diagram showing the configuration of a water level prediction system in the first embodiment. The water level prediction system in the first embodiment is a system that predicts the water level at a prediction point in a river whose water level is to be predicted. As shown in Fig. 1, the water level prediction system in the first embodiment has a rainfall acquisition unit 1, a memory unit 2, a water level prediction unit 3, a display unit 4, a current meter 5, a water level gauge 6, a satellite radio wave intensity meter 7, and a rain gauge 8.

[0031] The rainfall acquisition unit 1 acquires rainfall data within a watershed 20 from a Japan Meteorological Agency server 10 via a network NW. The rainfall acquisition unit 1 also acquires various data from a current meter 5, a water level gauge 6, a satellite radio wave intensity meter 7, and a rain gauge 8 from a cloud server 11 via the network NW.

[0032] 2 is a schematic diagram showing a river for which water level is to be predicted and its catchment area 20. The catchment area 20 is an area upstream of the prediction point of the river, and is an area into which rain flows during rainfall.

[0033] The precipitation data is data predicting the amount of precipitation at each point in the watershed 20. For the precipitation data, precipitation nowcast (5 minutes) data is used for the period up to one hour from the current time, and data from the quick analysis precipitation and quick short-term precipitation forecast is used for the period beyond one hour. In addition, actual precipitation data prior to the current time may be used.

[0034] Precipitation Nowcast (5 min) data is rainfall intensity forecast data for every 1 km square mesh every 5 minutes up to 1 hour ahead. Rapid Analysis Rainfall and Rapid Short-Term Precipitation Forecast data is rainfall intensity forecast data for every 1 km square mesh every 10 minutes up to 6 hours ahead, and rainfall amount forecast data for every 5 km square mesh every hour up to 15 hours ahead.

[0035] For the range from the current time to 30 minutes ahead or from the current time to 1 hour ahead, high-resolution precipitation nowcast data may be used instead of precipitation nowcast (5 minutes). This will improve the accuracy of water level predictions. Rainfall data provided by other meteorological agencies may also be used. Specifically, nationwide precipitation nowcast GPV, analyzed precipitation, short-term precipitation forecast, and 15-hour precipitation forecast may be used.

[0036] The rainfall data may be obtained from sources other than the Japan Meteorological Agency, or may be obtained from both the Japan Meteorological Agency and sources other than the Japan Meteorological Agency.

[0037] The memory unit 2 stores various parameters of a first-order lag system that inputs rainfall and outputs the water level at the river's water level prediction point, as well as various data acquired from the current meter 5, water level gauge 6, satellite radio wave intensity meter 7, and rain gauge 8. The various parameters of the first-order lag system include a time constant. The memory unit 2 also stores map data of the watershed 20. The memory unit 2 also stores the flow path length from each point in the watershed 20 to the prediction point.

[0038] The time constant is set according to the water level at the prediction point of the river. This allows the shape of the river and other factors to be reflected in the water level prediction, improving the accuracy of the water level prediction. The time constant of the first-order lag system is set, for example, by measuring the water level at the prediction point of the river after the rain has stopped and based on the characteristics of the measured water level drop. Once the rain has stopped, the water level is more stable than during rainfall, so the time constant can be determined with high accuracy. Also, while the water level is rising, the water level fluctuates due to rainfall, making it difficult to measure the time constant. By using the characteristics during the water level drop, the time constant can be easily measured.

[0039] The water level prediction unit 3 calculates the predicted water level at the prediction point as the response characteristic of a first-order lag system using the rainfall in the catchment area 20 as an input from the rainfall data acquired by the rainfall acquisition unit 1. The time constant of the first-order lag system stored in the memory unit 2 is used for the calculation of the first-order lag system. As described above, the time constant is not a constant used in a general first-order lag system, but is set according to the water level at the prediction point of the river. Therefore, in the following, it may be called a nonlinear first-order lag system to distinguish it from a general first-order lag system. In addition, the rainfall data is corrected using various data acquired from the current meter 5, water level gauge 6, satellite radio wave intensity meter 7, and rain gauge 8. A specific method for calculating the predicted water level will be described later.

[0040] The display unit 4 is a device that displays the water level predicted by the water level prediction unit 3. For example, the predicted water level is displayed on a display of a smartphone, a tablet terminal, a PC, or the like.

[0041] The predicted water level may be displayed as follows: A map of the vicinity of the prediction point is displayed, the predicted water level is divided into different colors according to its height, and the color of the prediction point on the map is changed to a color according to the predicted water level. The predicted water level may be updated in synchronization with updates to the Japan Meteorological Agency's precipitation nowcast (5 minutes) and the quick analysis rainfall and quick short-term precipitation forecast. This allows the predicted water level to be displayed in an easy-to-understand manner for consumers.

[0042] Also, if the predicted water level exceeds a specified water level, a warning may be displayed. The warning may be made by voice. Also, if the predicted water level exceeds a specified water level, information on the predicted water level may be sent to a smartphone or tablet device owned by the head or manager of the local government to inform them.

[0043] The current meter 5 is a device that measures the flow speed of water flowing in a river. The current meter 5 also has a communication device, and transmits the measured flow speed data to the cloud server 11. As shown in Fig. 2, a plurality of current meters 5 are placed in the river that is the target of prediction, upstream of the prediction point. The placement interval of the current meters 5 is, for example, every 1 to 5 km of the river channel length.

[0044] The water level gauge 6 is a device that measures the water level of the river. The water level gauge 6 also has a communication device, and transmits the measured water level data to the cloud server 11. As shown in Fig. 2, the water level gauge 6 is placed near the prediction point on the river that is the prediction target, and a plurality of water level gauges 6 are placed at positions upstream of the prediction point. The placement interval of the water level gauges 6 is, for example, every 1 to 5 km of the river channel length.

[0045] The satellite signal strength meter 7 is a device that receives radio waves from satellite broadcasting and measures the reception strength. The satellite signal strength meter 7 transmits the satellite signal strength data to the cloud server 11. As shown in Fig. 2, the satellite signal strength meters 7 are installed at multiple locations within the watershed 20. For example, they are placed at intervals of about 5 to 10 km square.

[0046] Fig. 3 is a diagram showing a schematic configuration of the satellite signal strength meter 7. As shown in Fig. 3, the satellite signal strength meter 7 has an antenna 70, a data generating unit 71, a solar panel 72, a power supply device 73, and a wireless device 74.

[0047] The antenna 70 is a device that receives radio waves from satellite broadcasting. The frequency of the radio waves received is, for example, the 12 GHz band. The data generation unit 71 is a device that generates data on the reception strength of the radio waves received by the antenna 70. The solar panel 72 is a device that converts solar energy into electricity and outputs it. The power supply unit 73 is a device that drives the data generation unit 71 and the wireless device 74. The power supply unit 73 also has a storage battery. The storage battery is charged by the power from the solar panel 72. The wireless device 74 transmits the reception strength data generated by the data generation unit 71 to the cloud server 11.

[0048] The satellite radio wave strength data measured by the satellite radio wave strength meter 7 is used to measure the amount of rainfall in the area where the satellite radio wave strength meter 7 is installed. There is a correlation between the level of the received satellite radio wave and the amount of rainfall. The satellite radio wave is attenuated by rainfall, and the amount of attenuation is related to the rainfall intensity. Therefore, the amount of rainfall can be calculated from the satellite radio wave strength data.

[0049] The rain gauge 8 is a device that measures the amount of rainfall. The rain gauge 8 also has a communication device and transmits rainfall data to the cloud server 11. As shown in Fig. 2, the rain gauges 8 are installed at multiple locations within the watershed 20. The rain gauge 8 measures the amount of rainfall, for example, every 1 to 5 minutes.

[0050] 2. Overview of water level forecast Next, an overview of the water level prediction in the first embodiment will be described.

[0051] To predict the water level of a river at a certain prediction point, the relationship between rainfall and water level at the prediction point is considered. A basic model for water level prediction is the tank model. This approximates the relationship between rainfall and water level to the relationship between the amount of water flowing into a tank and the amount of water flowing out of the tank. In other words, the water level is predicted using a first-order lag system in which the rainfall at the prediction point is input and the water level at the prediction point is output.

[0052] Figure 4 shows the amount of rainfall at a prediction point and the water level at the prediction point. When there is pulsed rainfall at a certain time t0 at a prediction point, the water level instantaneously increases to a certain value at time t0 and then gradually decreases. Here, the time when the water level becomes 0 with respect to the tangent of the water level decrease curve at time t0 is defined as t1, and the time from t0 to t1 is defined as the time constant T.

[0053] Here, the actual water level is affected by factors such as the shape of the river, its gradient, and the time it takes for rain to reach the river. For example, as shown in Figure 5, if the river bank is sloping, the higher the water level, the wider the river will be. Therefore, when the water level is low, the decrease in water level will be gradual, and the time constant T is likely to be large. On the other hand, when the water level is high, the decrease in water level will be rapid, and the time constant T is likely to be small.

[0054] Also, as shown in Figure 6, when the gradient of a river is gentle, the water flow speed is slow, so the decrease in water level is gradual, and the time constant T is expected to be large. On the other hand, when the gradient of a river is steep, the water flow speed is fast, so the decrease in water level is rapid, and the time constant T is expected to be small.

[0055] Thus, in an actual river, the time constant T is thought to vary depending on the shape, gradient, etc. of the river. Figure 7 is a schematic diagram showing the fall in water level for large, medium-sized, and small rivers. In small and medium-sized rivers, the river width is narrow, the riverbed is shallow, and the gradient is steep, so the time constant T tends to be small, while in large rivers, the river width is wide, the riverbed is deep, and the gradient is gentle, so the time constant T tends to be large. Also, in large rivers, the rise in water level during rainfall is small, while in small and medium-sized rivers, the rise in water level during rainfall is large.

[0056] Therefore, the inventors thought that if the time constant T of the first-order lag system is considered as a function of the water level, rather than being fixed, it would be possible to take into account that the time constant T varies depending on the shape and gradient of the river. In other words, they came up with the idea of ​​predicting the water level using a nonlinear first-order lag system with the time constant T as a function of the water level, and with the input being the amount of rainfall and the output being the water level. As described below, they found that the water level can be predicted with high accuracy using a nonlinear first-order lag system. The water level prediction system in the first embodiment is a system that predicts the water level using such a nonlinear first-order lag system with the time constant T as a function of the water level.

[0057] In addition, in actual water level prediction, there is a time delay between the rainfall point and the prediction point. Figure 8 is a schematic diagram showing a river and the water level prediction point. As shown in Figure 8, point A is taken to be located upstream of the prediction point in the river, and point B is taken to be located further upstream than point B. It is assumed that there was pulse-like rainfall at points A and B at a certain time t0.

[0058] Rain that falls at point A flows into the river and travels along the river to the prediction point. In other words, as shown in Figure 9, rain that falls at point A reaches the prediction point with a delay that depends on the flow path length from point A to the prediction point. For example, if the flow path length from point A to the prediction point is LA, the flow velocity of the water flowing through the river is v, and the time it reaches the prediction point is tA, then tA = t0 + LA / v.

[0059] Similarly, as shown in Figure 10, rain that falls at point B also arrives at the prediction point with a delay depending on the flow path length from point A to the prediction point. Here, point B is upstream of point A, so it arrives at the prediction point even later than the rain that fell at point A. If the time it arrives at the prediction point is tB, then tB>tA.

[0060] The rain that reaches the forecast point is the sum of the rain that fell at point A and the rain that fell at point B. Therefore, as shown in Figure 11, the rainfall amount at the forecast point is the sum of the rainfall amount at point A delayed by tA and the rainfall amount at point B delayed by tB. This total rainfall amount is used as the input of the nonlinear first-order lag system.

[0061] As described above, the rainfall amount used as the input of the nonlinear first-order lag system is corrected for delay according to the flow path length from the rainfall point to the prediction point, thereby improving the accuracy of water level prediction.

[0062] 3. Operation of the Water Level Prediction System Next, the operation of the water level prediction system in the first embodiment will be described with reference to the flowchart in FIG.

[0063] First, the rainfall acquisition unit 1 acquires rainfall forecast data from the Japan Meteorological Agency server 10 and stores it in the memory unit 2. As rainfall forecast data, precipitation nowcast (5 min) data is acquired for the range of one hour from the current time, and quick analysis rainfall / quick short-term precipitation forecast data is acquired for the range of more than one hour. Then, the water level prediction unit 3 compares the rainfall forecast data with the map data of the catchment basin 20 stored in the memory unit 2, and extracts the rainfall forecast data for the catchment basin 20 (step S1 in FIG. 12).

[0064] The rainfall acquisition unit 1 also acquires various data from a current meter 5, a water level meter 6, a satellite radio wave intensity meter 7, and a rain gauge 8. The memory unit 2 stores the various data acquired by the rainfall acquisition unit 1.

[0065] In parallel with this, the water level prediction unit 3 calculates the amount of rainfall in the area where the satellite signal strength meter 7 is installed from the data of the satellite signal strength. The relationship between the satellite signal strength and the amount of rainfall is stored in advance in the memory unit 2, and the amount of rainfall is calculated using this relationship. The memory unit 2 stores the calculated amount of rainfall.

[0066] Next, the water level prediction unit 3 corrects the rainfall forecast data based on the amount of rainfall in the area where the satellite signal strength meter 7 is installed, obtained from the satellite signal strength meter 7, and the amount of rainfall in the area where the rain gauge 8 is installed, measured by the rain gauge 8 (step S2 in FIG. 12). For example, the rainfall data from the Japan Meteorological Agency is divided into one-minute intervals, and the data for one minute from the current time is replaced with the amount of rainfall obtained from the satellite signal strength of the satellite signal strength meter 7 or the amount of rainfall measured by the rain gauge 8.

[0067] The Japan Meteorological Agency's rainfall data is based on radar reflections of rain clouds at an altitude of around 3000m to avoid reflections from mountains in mountainous areas, and is used to predict rainfall. Also, the effects of mountain shapes are not always fully taken into account. As a result, rainfall from rain clouds lower than the mountains may not be predicted accurately. In contrast, rainfall measurements using satellite radio wave strength can accurately measure rainfall from rain clouds lower than the mountains.

[0068] In this way, the accuracy of rainfall prediction can be improved by correcting the rainfall forecast data of the Japan Meteorological Agency with the amount of rainfall actually measured by the satellite signal intensity meter 7 and the rain gauge 8. As a result, the accuracy of water level prediction can be improved.

[0069] Next, the water level prediction unit 3 corrects the rainfall forecast data based on the delay time of the river water (step S3 in FIG. 12). Rain that falls at a certain point in the watershed 20 reaches the prediction point through the river. Therefore, there is a delay depending on the river's channel length and the water flow speed. The delay time is calculated by dividing the channel length from the rainfall point to the prediction point by the flow speed. Therefore, the water level prediction unit 3 calculates the delay time from the channel length from each point to the prediction point stored in the memory unit 2 and the flow speed measured by the flow speed meter 5. Then, the rainfall amount at each point is shifted by the delay time. For example, if the rainfall amount at a certain point i is ui(t) and the delay time is ti, it is set to ui(t-ti). In this way, by correcting the rainfall forecast data by the delay time, it is possible to improve the accuracy of the water level prediction.

[0070] In the first embodiment, the flow velocity measured by the flow meter 5 is used to calculate the delay time, but a flow velocity predicted from the amount of rainfall or the like may also be used. The predicted flow velocity may be corrected using the flow velocity measured by the flow meter 5, and the delay time may be calculated using the corrected flow velocity. The delay time may also be corrected using the water level measured by each water level gauge.

[0071] Next, the water level prediction unit 3 divides the rainfall amount at each point of the rainfall forecast data into equal intervals at a predetermined time interval, converts it into rainfall amount per unit time, and corrects and smooths the rainfall amount per unit time so that the temporal change of the rainfall amount becomes smoother (step S4 in FIG. 12). Then, the water level prediction unit 3 adds up the rainfall amount (rainfall amount per unit time) at each point to calculate the total rainfall amount. The rainfall amount at each point may be added up first and then smoothed. For example, the time course of the rainfall amount is divided into predetermined intervals, each interval is approximated by a polynomial, and each polynomial is connected.

[0072] Next, the water level prediction unit 3 inputs the smoothed rainfall amount and calculates the predicted water level, which is the output, from a first-order lag system that outputs the water level at the prediction point (step S5 in FIG. 12). Here, the time constant of the first-order lag system uses a value that changes depending on the water level at the prediction point.

[0073] Next, the water level prediction unit 3 corrects the calculated predicted water level (step S6 in FIG. 12). A specific correction method will be described with reference to FIG. 13. FIG. 13(a) is a graph showing a curve of the actual water level and the predicted water level before correction. As shown in FIG. 13(a), the actual water level value at the current time usually does not match the calculated predicted water level value at the current time.

[0074] First, as shown in FIG. 13(b), a correction is performed by multiplying the curve of the predicted water level by a coefficient so that the value of the actual water level near the prediction point measured by the water level gauge 6 at the current time coincides with the value of the calculated predicted water level at the current time. Hereinafter, this is referred to as the first correction. Next, as shown in FIG. 13(c), the curve of the predicted water level is rotated around the value of the water level at the current time so that the slope of the actual water level at the current time (time change in the actual water level) coincides with the slope of the predicted water level at the current time. Hereinafter, this is referred to as the second correction. By correcting the predicted water level twice as described above, the first correction and the second correction, the actual water level and the predicted water level sufficiently coincide near the current time. In other words, the accuracy of the water level prediction can be improved.

[0075] Furthermore, when the second correction is performed, a restriction may be imposed on the time range. As shown in FIG. 14, the predicted water levels may be the first and second corrections for the range from the current time to a predetermined time ahead, and the predicted water levels may be the first correction only for the range beyond the predetermined time. This can further improve the prediction accuracy. For example, the time range when the second correction is performed may be from the current time to 1 to 2 hours ahead.

[0076] Also, when the second correction is performed, a limit based on the water level may be added. As shown in FIG. 15, for the predicted water level after the second correction, the predicted water level of the first and second corrections may be used for the range below the upper limit water level, and for the range above the upper limit water level, the predicted water level of the first correction may be restored without the second correction. This can further improve the prediction accuracy. The upper limit water level is, for example, in the range of 200 to 300% of the actual water level at the current time.

[0077] The predicted water level may be corrected each time a predicted water level for a predetermined time ahead is calculated. This allows for improved prediction accuracy. For example, the correction may be performed each time a predicted water level for 1 to 6 hours ahead is calculated. In particular, if the correction is performed each time a predicted water level for 1 to 3 hours ahead is calculated, the water level can be predicted with higher accuracy.

[0078] Next, the display unit 4 displays the predicted water level calculated by the water level prediction unit. The display unit 4 may display the predicted water level numerically. Alternatively, a map of the vicinity of the prediction point may be displayed, and the prediction point on the map may be displayed in a color corresponding to the predicted water level. Alternatively, the map may display the precipitation nowcast (5 minutes) and the precipitation amount based on the quick analysis precipitation amount and the quick short-term precipitation forecast, and the update of the predicted water level may be synchronized with the update of the precipitation nowcast (5 minutes), the quick analysis precipitation amount and the quick short-term precipitation forecast. This allows the predicted water level to be displayed in a way that is easier for consumers to understand. This is particularly suitable when there are multiple prediction points.

[0079] If the predicted water level exceeds a predetermined threshold, an alarm may be sounded. Also, the predicted water level may be notified to a smartphone or tablet carried by a local government manager or chief.

[0080] As described above, in the water level prediction system in the first embodiment, the water level of a river is predicted as the output of a nonlinear first-order lag system whose input is the rainfall in the catchment area. The time constant of the nonlinear first-order lag system is set according to the water level at the prediction point of the river. Therefore, the water level at the prediction point of the river can be predicted easily and accurately. Furthermore, the water level prediction system in the first embodiment is suitable for predicting the water level of small and medium-sized rivers, and is particularly suitable for small and medium-sized rivers in mountainous areas.

[0081] Next, a simulation result for the water level prediction system in the first embodiment will be described.

[0082] Using the water level prediction system of the first embodiment, the water level of the Tsubo River in October 2019 was predicted. The prediction point was a certain point in Kaminoho, Seki City. The time constant was obtained by measuring the drop in water level after rainfall from April 2018 to March 2021. FIG. 16 is a graph showing the relationship between the water level at the prediction point and the value of the time constant.

[0083] The predicted water level was calculated using a nonlinear first-order lag system using the time constant shown in Figure 16. The input rainfall amount was measured by the Japan Meteorological Agency in catchment area 20. The peak of the predicted water level was made to coincide with the peak of the actual water level. Figure 17 is a graph showing the rainfall intensity, predicted water level, and actual water level. It was confirmed that the curve of the predicted water level roughly coincides with the curve of the actual water level.

[0084] In addition, after calculating the predicted water level for the next hour, the predicted water level was adjusted to match the actual water level at the current time with the predicted water level, and the slope of the actual water level at the current time was adjusted to match the slope of the predicted water level. This adjustment was repeated each time the predicted water level for the next hour was calculated.

[0085] Figure 18 is a graph showing the actual water level and the predicted water level. As shown in Figure 18, the curves of the actual water level and the predicted water level are almost identical, and it was found that the water level at the prediction point can be predicted with extremely high accuracy by correcting the predicted water level successively.

[0086] We also changed the time from one hour to three hours ahead, and repeated the same sequential correction as in Figure 18 each time we calculated the predicted water level up to three hours ahead. Figure 19 is a graph showing the actual water level and the predicted water level. As shown in Figure 19, even when correcting every three hours, the curves of the actual water level and the predicted water level roughly match, indicating that the water level at the prediction point could be predicted with high accuracy.

[0087] As shown in Figures 18 and 19, by successively correcting the predicted water level each time a water level forecast for one hour or three hours ahead was calculated, it was possible to make the predicted water level almost match the actual water level.

[0088] (Modification of the first embodiment) In the first embodiment, a correction is performed to match the actual water level at the current time with the predicted water level, and to match the slope of the actual water level at the current time with the slope of the predicted water level, but other correction methods may also be used.

[0089] For example, the coefficient A to be multiplied with the predicted water level may be determined so that the sum of the differences between the actual water level and the predicted water level is 0. In other words, the actual water level at a certain time t is W(t), the predicted water level is R(t), and the sum of the differences between the actual water level W(t) and the predicted water level R(t) from a certain time t0 to a certain time ti is S, and A is determined so that S = Σ(W(t)-A×R(t)) = 0 is satisfied. For example, A is determined so that S = 0 is satisfied from three hours before the current time to the current time. By determining A in this way and correcting the predicted water level R(t) to A×R(t), the predicted water level can be brought closer to the actual water level.

[0090] Also, for example, the coefficient B multiplied by the predicted water level may be determined by the least squares method. In other words, the actual water level at a certain time t is W(t), the predicted water level is R(t), and the sum of squares of the difference between the actual water level W(t) and the predicted water level R(t) from a certain time t0 to a certain time ti is RSS, where RSS=Σ(W(t)-B×R(t)). 2 The coefficient B is determined so that is minimized. By finding B in this way and correcting the predicted water level R(t) to B × R(t), the predicted water level can be brought closer to the actual water level.

[0091] Also, an inflection point of the actual water level may be found, and the absolute value of the correction amount for the predicted water level after the time of the inflection point may be made smaller than the correction amount before the time of the inflection point. For example, the above-mentioned coefficients A and B are used before the time of the inflection point, and a coefficient smaller than coefficients A and B is used after the time of the inflection point. When the water level approaches its peak, an inflection point appears where the first-order time derivative of the water level is maximized and the second-order time derivative of the water level is zero. After the time of this inflection point, the increase in the water level slows down. Therefore, by changing the correction amount of the predicted water level before and after the time of the inflection point as described above, the accuracy of the water level prediction can be improved.

[0092] It is preferable to update the time constant of the first-order lag system as needed. For example, the time period after the rain has stopped is detected from the rainfall data, and each time such a time period is detected, the time constant is set and updated based on the characteristics of the water level drop at the prediction point of the river during that time period. At this time, the rainfall data used is data obtained from the satellite radio wave intensity meter 7 and the rain gauge 8, and the water level at the prediction point is data obtained from the water level gauge 6. [Explanation of symbols]

[0093] 1: Rainfall acquisition part 2: Storage part 3: Water level prediction section 4: Display section 5: Velocity meter 6: Water level gauge 7:Satellite radio wave strength meter 8:Rain gauge 10: Japan Meteorological Agency server 11: Cloud Server 20: Catchment Area

Claims

1. A water level prediction system for predicting a future water level at a prediction point of a river, comprising: A rainfall acquisition unit that acquires future rainfall in a catchment area for the prediction point; a storage unit that stores parameters of a first-order lag system in which the rainfall amount is input and the water level at the prediction point of the river is output; a water level prediction unit that receives the rainfall acquired by the rainfall acquisition unit as an input and calculates a predicted water level at the prediction point of the river from the first-order lag system using parameters of the first-order lag system stored in the storage unit, A water level prediction system, wherein the time constant of the first-order lag system is set according to the water level at the prediction point of the river.

2. The water level prediction system according to claim 1 , wherein the time constant in the delay system is set based on the drop characteristics of the measured water level at the prediction point of the river after the rain has stopped.

3. The water level prediction system according to claim 1 , wherein the rainfall amount acquisition unit acquires rainfall forecast data from the Japan Meteorological Agency as the rainfall amount.

4. 4. The water level prediction system according to claim 3, wherein the water level prediction unit converts the rainfall forecast data from the Japan Meteorological Agency into data with equal time intervals and smoothed values, and uses the data as the amount of rainfall per unit time.

5. The water level prediction system of claim 3 or claim 4, wherein the rainfall acquisition unit acquires precipitation nowcast (5 minutes) data from the Japan Meteorological Agency as the rainfall amount from the current time to one hour ahead, and acquires quick analysis rainfall data and quick short-term precipitation forecast data from the Japan Meteorological Agency as the rainfall amount more than one hour ahead.

6. Further, a water level measuring unit is provided for measuring an actual water level at the predicted point of the river, The water level prediction unit performs a first correction on the predicted water level, The water level prediction system according to claim 1 , wherein the first correction is performed so that the value of the predicted water level at the current time coincides with the value of the actual water level at the current time measured by the water level measurement unit.

7. The water level prediction unit performs a second correction on the predicted water level after the first correction, The water level prediction system described in claim 6, wherein the second correction is performed so that the value of the time change of the predicted water level after the first correction at the current time matches the value of the time change of the actual water level at the current time measured by the water level measurement unit.

8. The water level prediction unit is For the water level from the current time to a predetermined time ahead, both the first correction and the second correction are performed; The water level prediction system according to claim 7 , wherein the first correction is performed but the second correction is not performed for a water level beyond a predetermined time.

9. The water level prediction system described in claim 7 or claim 8, wherein the water level prediction unit performs both the first correction and the second correction, and if there is a section in which the predicted water level after the correction exceeds a predetermined upper limit value, the first correction is performed for that section and the value is returned to one in which the second correction is not performed.

10. Further, a water level measuring unit is provided for measuring an actual water level at the predicted point of the river, The water level prediction unit corrects the predicted water level by multiplying the predicted water level by a coefficient, The water level prediction system according to claim 1 , wherein the coefficient is determined so that a sum of a difference between an actual water level measured by the water level measuring unit at a certain time and the predicted water level becomes zero.

11. Further, a water level measuring unit is provided for measuring an actual water level at the predicted point of the river, The water level prediction unit corrects the predicted water level by multiplying the predicted water level by a coefficient, The water level prediction system according to claim 1 , wherein the coefficient is determined so as to minimize a sum of squares of the difference between the actual water level measured by the water level measuring unit at a certain time and the predicted water level.

12. Further, a water level measuring unit is provided for measuring an actual water level at the predicted point of the river, The water level prediction system of claim 1, wherein the water level prediction unit corrects the predicted water level, calculates an inflection point of the actual water level, and makes the absolute value of the correction amount after the time of the inflection point smaller than the absolute value of the correction amount before the time of the inflection point.

13. The water catchment area further includes a satellite signal strength meter that is disposed in the water catchment area and receives satellite broadcasting signals, The water level prediction system according to claim 1 , wherein the water level prediction unit calculates the amount of rainfall from the strength of the radio waves received by the satellite radio wave intensity meter, and corrects the amount of rainfall acquired by the rainfall acquisition unit using the calculated amount of rainfall.

14. The water level prediction system of claim 1, wherein the water level prediction unit calculates a delay time for water to reach the prediction point from the rainfall point based on the flow path length of the river from the rainfall point to the prediction point and the flow velocity of water flowing through the river, and corrects the rainfall amount for the delay time.

15. Further, a display unit is provided for displaying the predicted water level, the precipitation nowcast (5 minutes), and the precipitation amount based on the quick report analysis precipitation amount and the quick report short-term precipitation forecast, The water level prediction system according to claim 5, wherein the update of the predicted water level is synchronized with the update of the precipitation nowcast (5 minutes) and the quick analysis rainfall / quick precipitation short-term forecast.

Images