Real-time water level prediction device, real-time water level prediction method, and computer program

Abstract

To provide a water level prediction device that performs highly accurate predictions of water levels inside sewer pipes. [Solution] The water level prediction device 1 according to the embodiment acquires time-series data of water level, time-series data of rainfall information, structural information of sewer pipes, and topographic information, calculates water level prediction values ​​at multiple water level gauge installation points using a water level prediction model, selects one of the multiple water level gauge installation points to be updated, calculates updated data for the water level prediction value of the water level gauge installation point to be updated by runoff analysis using the time-series data of water level prediction values ​​excluding the water level prediction value of the water level gauge installation point to be updated, updates the time-series data of the water level prediction value of the water level gauge installation point to be updated using the updated data, and calculates time-series data of water level prediction values ​​at points on the pipeline without water level gauges by runoff analysis using the time-series data of water level prediction values ​​at multiple water level gauge installation points.

Description

Technical Field

[0001] Embodiments of the present invention relate to a real-time water level prediction device, a real-time water level prediction method, and a computer program.

Background Art

[0002] In recent years, heavy rainfall that occurs locally and in a short period of time has been increasing, and urban internal water flooding, where water overflows within the city, has been occurring frequently. In fact, when looking at flood damage (external water flooding, internal water flooding) in terms of damage amount, the damage amount of internal water flooding accounts for about half of the total national damage amount. Also, in Tokyo, the damage amount of internal water flooding exceeds 90% of the total damage amount of the city, and in urban areas where the construction of levees has progressed relatively, internal water flooding has become a new issue.

[0003] As a countermeasure against internal water flooding, research on real-time prediction of water levels based on meteorological data has been underway. For example, an inundation prediction system has been proposed that predicts the possibility of flooding and the depth of inundation within a region based on surface data such as altitude within the predicted region stored in advance, data on the amount of rainfall that actually fell in the past few hours from the current time, data on the amount of rainfall predicted to fall in the next few hours, etc. However, the phenomenon until inundation occurs is complex, and further improvement in prediction accuracy is necessary.

[0004] For example, a technique has been proposed that uses machine learning (multiple regression analysis, Kalman filter, neural network, random forest, support vector machine, etc.) to predict the water level rise rate from the rainfall intensity and calculate the prediction time until internal water flooding occurs. According to this technique, based on the predicted value of the water level rise rate, the time until the water level at which internal water flooding occurs is predicted, but it is difficult for the user to determine how reliable the predicted value is.

[0005] For example, a technique has been proposed to select one of several pre-calculated inundation depth distribution maps as the future inundation depth distribution map, based on observed and predicted rainfall data. However, if the model used to select the future inundation depth distribution is not updated, it may not be able to adapt to changes in precipitation phenomena due to recent climate change.

[0006] Furthermore, if a simple water level prediction is performed using deep learning, and this predicted value is then used to further predict water levels between multiple locations, the relationship between water levels at each location (e.g., relative water levels, which location has a higher water level) may not be learned, potentially making it impossible to correctly interpolate water levels between multiple locations. [Prior art documents] [Patent Documents]

[0007] [Patent Document 1] Japanese Patent Publication No. 2009-008651 [Patent Document 2] Japanese Patent Publication No. 2017-227582 [Patent Document 3] Japanese Patent Publication No. 2021-182408 [Overview of the Initiative] [Problems that the invention aims to solve]

[0008] To mitigate damage from inland flooding, it is desirable to improve the accuracy of water level predictions within sewer pipes. In particular, when predicting water levels at multiple locations and attempting to interpolate water levels between prediction points using these predicted values ​​for purposes such as creating hazard maps, it is crucial to accurately predict the relative levels of water levels at each prediction point.

[0009] The results of predicting individual water levels at multiple locations using AI (artificial intelligence) include errors and do not guarantee the relative levels of water levels because the physical connections between locations are not taken into account. Therefore, inputting the AI ​​prediction results into an outflow analysis model to calculate water levels at unmeasured locations sometimes resulted in poor accuracy.

[0010] Embodiments of the present invention have been made in view of the above circumstances, and aim to provide a real-time water level prediction device, a real-time water level prediction method, and a computer program that perform highly accurate prediction of water levels inside sewer pipes. [Means for solving the problem]

[0011] The real-time water level prediction device according to an embodiment of the present invention includes a water level prediction unit that acquires time-series data of water levels measured by sensors installed in sewer pipes, time-series data of rainfall information for an area including the water level gauge installation point where the sensor is installed, structural information of the sewer pipe, and topographic information for an area including the water level gauge installation point, and calculates water level prediction values ​​for a plurality of water level gauge installation points using a water level prediction model, and acquires time-series data of the water level prediction values ​​calculated by the water level prediction unit, selects one of the plurality of water level gauge installation points as the update target, and calculates the water level prediction for the water level gauge installation point that is the update target. The system includes: an update unit that outputs time-series data of water level prediction values ​​excluding the actual values ​​and updates the time-series data of water level prediction values ​​at the water level gauge installation locations to be updated using the update data; a second runoff analysis unit that uses the time-series data of water level prediction values ​​output from the update unit to calculate the update data of the water level prediction values ​​at the water level gauge installation locations to be updated by runoff analysis and supplies it to the update unit; and a runoff analysis unit that obtains time-series data of water level prediction values ​​at multiple water level gauge installation locations from the update unit and calculates time-series data of water level prediction values ​​at locations on the pipeline where water level gauges are not installed by runoff analysis. [Brief explanation of the drawing]

[0012] [Figure 1] Figure 1 is a schematic block diagram showing one example configuration of a real-time water level prediction device according to one embodiment. [Figure 2] Figure 2 is a schematic diagram showing an example of time-series data for rainfall information. [Figure 3] Figure 3 is a schematic diagram showing an example of predicted values ​​for time-series water level data calculated by the AI ​​water level prediction unit. [Figure 4] Figure 4 is a flowchart illustrating an example of the operation of a real-time water level prediction device according to one embodiment. [Figure 5] Figure 5 is a schematic diagram showing an example of the connection relationships between points used to calculate water level predictions. [Modes for carrying out the invention]

[0013] The real-time water level prediction device, real-time water level prediction method, and computer program of the embodiment will be described below with reference to the drawings. Figure 1 is a schematic block diagram showing one example configuration of a real-time water level prediction device according to one embodiment.

[0014] The real-time water level prediction device 1 of this embodiment is a computing device comprising, for example, at least one processor and a memory storing a program executed by the processor, and can realize various functions described below by software or by a combination of software and hardware. The real-time water level prediction device 1 can access external databases DB1-DB4 to acquire data.

[0015] The first database, DB1, stores time-series data (water level data) of water levels measured by water level sensors (water level gauges) at predetermined measurement points (water level gauge installation locations) within the target watershed (drainage area). The water level data is time-series data (discrete-time data) measured in real time at predetermined intervals, and may include, for example, past water level values ​​that were actually measured, information identifying the measurement location, and information on the date and time of measurement.

[0016] The second database DB2 stores the structural information of the sewer network. The sewer network structure includes information such as pipe diameter, gradient, and branch at each of multiple locations in the sewer network. The third database DB3 stores the terrain information. The terrain information includes information such as elevation and gradient at multiple locations.

[0017] The fourth database DB4 stores rainfall information data. The rainfall information may include time-series information (discrete-time data) of the distribution of rainfall measurement values and snowfall measurement values at rainfall measurement locations in the area observed by the weather radar. Also, the rainfall information may include information on rainfall intensity. Further, the rainfall information stores time-series information (discrete-time data) of future rainfall measurement values and snowfall measurement values at the rainfall measurement locations based on weather forecasts.

[0018] Also, the real-time water level prediction device 1 is connected to the display unit 2. The display unit 2 includes, for example, a liquid crystal display panel or an organic EL display panel, and can visually present numerical values, images, etc. output from the real-time water level prediction device 1 to the user. The display unit 2 can be, for example, a monitor equipped with a display screen, or a portable terminal such as a tablet terminal or a smartphone. The real-time water level prediction device 1 may be connected to a plurality of display units 2.

[0019] The real-time water level prediction device 1 of the present embodiment calculates the water level prediction values at a plurality of locations where no water level gauges are installed, for example, for a basin in which water level measurement values at each of a plurality of water level gauge installation locations in a sewer network including a plurality of conduits and rainfall measurement values at each of a plurality of rainfall measurement locations are measured at a predetermined cycle. The real-time water level prediction device 1 has functions of an AI water level prediction unit (water level prediction unit) 10, an AI water level prediction update unit (update unit) 20, an outflow analysis unit 30, and a second outflow analysis unit 40.

[0020] The AI water level prediction unit 10 acquires past water level data in the sewer conduit, which is time-series data (discrete time data) synchronized with a time stamp from the first database DB1, for a plurality of water level gauge installation locations (for example, location 1 to location N) where future water level prediction values are calculated. Also, the AI water level prediction unit 10 acquires sewer conduit structure information from the second database DB2 for an area including a plurality of water level gauge installation locations (for example, location 1 to location R). Also, the AI water level prediction unit 10 acquires terrain information from the third database DB3 for an area including a plurality of water level gauge installation locations (location 1 to location N). Also, the AI water level prediction unit 10 acquires rainfall information data, which is time-series data (discrete time data) synchronized with a time stamp from the fourth database DB4, for an area including a plurality of water level gauge installation locations (location 1 to location N). At this time, when the sewer conduit structure information and terrain information are updated during the period corresponding to the time-series data, the AI water level prediction unit 10 may acquire the sewer conduit structure information and terrain information for the corresponding period.

[0021] FIG. 2 is a diagram schematically showing an example of time-series data of rainfall information. In this example, the time t corresponds to the time when the water level prediction value is calculated, and the time after time t is a future time. The time-series data of rainfall information includes past data for the period (t to t + T) of the water level time-series data calculated by the AI water level prediction unit 10. The time from time t - x to time t - 1 is a time past the time t when the water level prediction value is calculated, and the time from time t + 1 to time t + T is a future time of time t. The time-series data of rainfall information at past times is the measured value of the rainfall amount at the rainfall measurement location. The time-series data of rainfall information at future times can be stored in the fourth database DB4, for example.

[0022] The AI water level prediction unit 10 acquires the time-series data of rainfall information for each of the R rainfall measurement locations (location 1 to location R) in the period from time t - x to time t + T shown in FIG. 2, for example.

[0023] The AI ​​water level prediction unit 10 uses acquired rainfall information data, water level data within sewer pipes, sewer pipe structure, and topographic information to calculate predicted values ​​for time-series water level data at N (N≧1) water level gauge installation locations and outputs them to the AI ​​water level prediction update unit 20. The water level prediction model used by the AI ​​water level prediction unit 10 to calculate predicted values ​​for time-series water level data may be a physical model, a neural network, machine learning, or the like.

[0024] The AI ​​water level prediction unit 10 may, for example, divide the target watershed into multiple sub-watersheds based on the connection relationships of multiple sewer pipes in the sewer network, using a pre-trained machine learning model that has been trained to calculate predicted water levels at water level gauge installation locations using time-series data of water level measurements at multiple water level gauge installation locations and time-series data of rainfall measurements at multiple rainfall measurement locations.

[0025] The AI ​​water level prediction unit 10 may calculate the confidence level of the predicted water level when calculating time-series data of predicted water level values ​​using a water level prediction model. The AI ​​water level prediction unit 10 may calculate the confidence level for each predicted water level value, or it may calculate the confidence level for the time-series data of predicted water level values ​​at water level gauge installation locations. The AI ​​water level prediction unit 10 associates the calculated confidence level values ​​with the data of predicted water level values ​​or identification information of water level gauge installation locations and supplies them to the AI ​​water level prediction update unit 20.

[0026] Figure 3 is a schematic diagram showing an example of predicted values ​​for time-series water level data calculated by the AI ​​water level prediction unit. The AI ​​water level prediction unit 10 calculates predicted values ​​for the time-series water level data (discrete time data) of each of the N water level gauge installation locations (location 1 to location N) for the period from time t to time t+T, and supplies these to the AI ​​water level prediction update unit 20. In this example, time t corresponds to the time of calculation, and time t and beyond are future times.

[0027] The AI ​​water level prediction update unit 20 determines one water level gauge installation location to be updated from among the N water level gauge installation locations corresponding to the time-series data of water level prediction values ​​input from the AI ​​water level prediction unit 10, and outputs the time-series data of water level prediction values ​​corresponding to the remaining N-1 water level gauge installation locations to the second runoff analysis unit 40. The AI ​​water level prediction update unit 20 outputs the time-series data of water level prediction values ​​corresponding to the N-1 water level gauge installation locations to the second runoff analysis unit 40 while sequentially changing the water level gauge installation location to be updated.

[0028] The AI ​​water level prediction update unit 20 can determine the location of a water level gauge to be updated by any method. For example, if numbers, symbols, etc. are assigned sequentially from 1 to N water level gauge locations, the AI ​​water level prediction update unit 20 may process any of the water level gauge locations as update targets in any order (for example, ascending order from 1, descending order from N, random, etc., based on the assigned numbers or symbols). Alternatively, the maximum value of the predicted water level for each of the N water level gauge locations may be calculated, and numbers, symbols, etc. are assigned sequentially from 1 to the N water level gauge locations with the largest predicted water level values ​​from the maximum value, and the AI ​​water level prediction update unit 20 may process them as update targets in any order (for example, ascending order from 1, descending order from N, etc., based on the assigned numbers or symbols).

[0029] The AI ​​water level prediction update unit 20 may, for example, obtain the confidence value when the water level prediction value is calculated by the AI ​​water level prediction unit 10 along with the water level prediction value, assign numbers, symbols, etc., starting from 1 to the water level gauge installation locations in ascending or descending order of confidence value, and process them as update targets in any order (for example, ascending from 1, descending from N, etc., based on the assigned numbers or symbols). In this case, if the confidence value has been calculated for each time series data of the water level prediction value, the AI ​​water level prediction update unit 20 may, for example, adopt at least one of the maximum, minimum, average, or median confidence values ​​included in the time series data, or other representative values, as the confidence value for the water level gauge installation location.

[0030] Furthermore, the AI ​​water level prediction update unit 20 may, for example, obtain the latest data from past measured water levels, assign numbers starting from 1 to the water level gauge installation locations in descending order of the difference between the latest measured value and the predicted water level, and process them as update targets in any order (for example, ascending from 1).

[0031] The AI ​​water level prediction update unit 20 outputs predicted water level data for N-1 locations to the second runoff analysis unit 40, and then obtains time-series data of the calculated predicted water level values ​​for the water level gauge installation locations to be updated from the second runoff analysis unit 40 as analysis results.

[0032] The AI ​​water level prediction update unit 20 updates the time-series data of water level prediction values ​​supplied from the AI ​​water level prediction unit 10 using time-series data of water level prediction values ​​for water level gauge installation locations to be updated, which is obtained from the second runoff analysis unit 40. The AI ​​water level prediction update unit 20 may, for example, update the water level prediction result by the AI ​​water level prediction unit 10 using a weighted average value, which is obtained by using the time-series data of water level prediction values ​​supplied from the AI ​​water level prediction unit 10 but not output to the second runoff analysis unit 40 (for water level gauge installation locations that are not subject to update) and the time-series data of water level prediction values ​​for water level gauge installation locations to be updated (update data) supplied from the second runoff analysis unit 40. The weights between the update data and the time-series data of the water level prediction value to be updated may be set by a predetermined value, set by the AI ​​water level prediction update unit 20 according to the confidence value of the water level prediction value to be updated, set by comparing the update data with the time-series data of the water level prediction value at the water level gauge installation point to be updated, or adjusted by the AI ​​water level prediction update unit 20 so that the amount of water level prediction value update is within a predetermined range.

[0033] The AI ​​water level prediction update unit 20 updates the time-series data of the predicted water level and then determines whether to continue processing. The AI ​​water level prediction update unit 20 can determine whether to continue processing by any method. For example, the AI ​​water level prediction update unit 20 may terminate the iterative processing when it has finished updating all N water level gauge installation locations. Alternatively, the AI ​​water level prediction update unit 20 may continue the iterative processing after it has finished updating all N water level gauge installation locations, and terminate the iterative processing when the amount of update to the predicted water level (the difference between the predicted water level before the update and the predicted water level after the update) falls below a threshold. Or, the AI ​​water level prediction update unit 20 may terminate the iterative processing based on the amount of update, etc., even before it has finished updating all N water level gauge installation locations.

[0034] If the AI ​​water level prediction update unit 20 determines that it is necessary to continue the update process, it selects another water level gauge installation location to be updated, outputs the predicted water level data for N-1 locations to the second outflow analysis unit 40, and repeats the same update process. When the AI ​​water level prediction update unit 20 determines that the update process is complete, it outputs time-series data of predicted water levels at all N water level gauge locations to the runoff analysis unit 30.

[0035] The runoff analysis unit 30 uses time-series data of rainfall information obtained from the fourth database DB4 and time-series data of predicted water levels at all N water level gauge locations input from the AI ​​water level prediction update unit 20 to calculate the water level at any point on the pipeline (water level at locations without water level gauges), and outputs the time-series data of predicted water levels at both locations with and without water level gauges to an external device such as the display unit 2.

[0036] The runoff analysis model used in the runoff analysis unit 30 may be a model that performs a simplified runoff analysis and flood prediction using a model that closely resembles an actual sewer pipeline, based on physical parameters of the sewer pipeline network, such as "flow rate calculation tables" and "section division plans," which are used by local governments and other entities involved in sewerage projects for planning and managing sewerage systems. Alternatively, it may be a model that tracks rainfall flow using a runoff analysis model that appropriately combines hydrological and hydraulic models (such as the Saint-Venin equation) with civil engineering information such as land use patterns, topographic information such as elevation, and sewer pipeline laying conditions in a given area. The runoff analysis unit 30 may also use general-purpose analysis software for representative models such as MOUSE and MIKE from Denmark, Infoworks from the UK, and SWMM from the US.

[0037] The runoff analysis unit 30 may be configured to calculate time-series data of predicted water levels at multiple locations without water level gauges in parallel by dividing the runoff analysis model, which performs runoff analysis of the target watershed to calculate predicted water levels at locations without water level gauges included in each of the multiple sub-watersheds, using time-series data of predicted water levels at locations with water level gauges on the pipeline as boundary conditions for water levels in multiple sub-watersheds, and inputting time-series data of measured rainfall and predicted rainfall at multiple rainfall measurement locations into multiple sub-runoff analysis models.

[0038] The second runoff analysis unit 40 calculates time-series data of the water level prediction value for the remaining 1 water level gauge installation location (to be updated) from the time-series data of N-1 water level prediction values ​​input from the AI ​​water level prediction update unit 20 and the time-series data of rainfall information, and outputs the calculation result to the AI ​​water level prediction update unit 20. The runoff analysis model used in the second runoff analysis unit 40 may be the exact same model as that used in the runoff analysis unit 30, or it may be a model that has the minimum subset necessary for the above calculation. When the second runoff analysis unit 40 determines that the time-series data of the water level prediction values ​​for N-1 locations is valid, it calculates the time-series data of the remaining 1 water level prediction value that is most appropriate to that (time-series data of the water level prediction values ​​for N-1 locations) while considering the physical structure. By controlling the AI ​​water level prediction update unit 20 to change the water level gauge installation location to be updated, and by the second runoff analysis unit 40 repeating the process sequentially, the time-series data of all N water level prediction values ​​will be corrected to a water level relationship that considers the physical structure.

[0039] Figure 4 is a flowchart illustrating an example of the operation of a real-time water level prediction device according to one embodiment. Figure 5 is a schematic diagram showing an example of the connection relationships between points used to calculate water level predictions.

[0040] The following describes an example of calculating time-series data of predicted water levels at water level gauge installation points A, B, and C (N=3) using an AI model, and then calculating time-series data of predicted water levels at point X on the pipeline where no water level gauge is installed using an outflow analysis model. Point X is a point on the pipeline where water flowing from upstream water level gauge installation point A branches off to downstream water level gauge installation point B and water level gauge installation point C.

[0041] The location information of water level gauge installation points A, B, and C, along with past measured water levels, is stored in the first database DB1. The pipe structure, such as pipe diameter and gradient, of the sewer pipes at each of the water level gauge installation points A, B, C, and the uninstalled water level gauge location X is stored in the second database DB2. Topographic information, such as elevation and gradient, for the area including water level gauge installation points A, B, C, and the uninstalled water level gauge location X is stored in the third database DB3. Rainfall information, such as rainfall amount and rainfall intensity, for the area including water level gauge installation points A, B, C, and the uninstalled water level gauge location X is stored in the fourth database DB4.

[0042] First, the AI ​​water level prediction unit 10 obtains, for example, past water level data for a predetermined period measured at water level gauge installation points A, B, and C from the first database DB1, structural information of sewer pipes in the area including water level gauge installation points A, B, and C from the second database DB2, topographic information of the area including water level gauge installation points A, B, and C from the third database DB3, and past and future rainfall information for the area including water level gauge installation points A, B, and C from the fourth database DB4.

[0043] The AI ​​water level prediction unit 10 uses the acquired data to calculate time-series data of predicted water levels at water level gauge installation points A, B, and C using a water level prediction model, and outputs the calculation results to the AI ​​water level prediction update unit 20 (step S1). At this time, the AI ​​water level prediction unit 10 may also output the confidence level of the calculated predicted water levels to the AI ​​water level prediction update unit 20.

[0044] Next, the AI ​​water level prediction update unit 20 outputs the water level prediction results for two water level gauge installation locations (N-1) excluding the water level gauge installation location to be updated (for example, water level gauge installation location C) to the second runoff analysis unit 40 (step S2).

[0045] The second runoff analysis unit 40 uses the time-series data of predicted water levels at the input water level gauge installation locations (water level gauge installation locations A and B) to calculate time-series data of predicted water levels at the water level gauge installation location to be updated (water level gauge installation location C) using a runoff analysis model (or a subset of the runoff analysis model), and outputs the calculation results to the AI ​​water level prediction update unit 20 (step S3).

[0046] The AI ​​water level prediction update unit 20 uses the time-series data of the predicted water level at the water level gauge installation location (water level gauge installation location C) to be updated, input from the AI ​​water level prediction unit 10, and the time-series data of the predicted water level at the water level gauge installation location (water level gauge installation location C) to be updated, input from the second runoff analysis unit 40, to calculate, for example, a weighted average value. The AI ​​water level prediction update unit 20 overwrites (updates) the time-series data of the predicted water level at the water level gauge installation location (water level gauge installation location C) to be updated, input from the AI ​​water level prediction unit 10, with the calculated weighted average value (step S4).

[0047] Here, for example, if the water level gauge installation locations to be updated are locations that include water level prediction values ​​with low confidence, it becomes possible to update the water level values ​​with high uncertainty among the water level prediction values ​​calculated by the AI ​​water level prediction unit 10 to water level values ​​that take into account the structural information and topographic information of the sewer pipes. Furthermore, in subsequent update processes (update processes for other water level gauge installation locations), the second runoff analysis unit 40 will perform runoff analysis using the water level prediction values ​​that have already been updated, thereby improving the prediction accuracy of the update data calculated by the second runoff analysis unit 40.

[0048] The AI ​​water level prediction update unit 20 determines whether the conditions for terminating the data update process (repeated process) input from the AI ​​water level prediction unit 10 have been met (step S5).

[0049] Conditions for terminating the loop include, for example, the completion of updating all N water level gauge locations, or the change in the amount of water level prediction (the difference between the water level prediction before and after the update) falling below a threshold.

[0050] The AI ​​water level prediction update unit 20, until the conditions for terminating the repetitive process are met (step S5, No), resets the water level gauge installation locations to be updated by any method (for example, setting water level gauge installation location B as the update target) (step S6), and transmits time-series data of water level prediction values ​​for N-1 locations other than the water level gauge installation locations to the second runoff analysis unit 40 (step S2). At this time, the time-series data of the water level prediction value for the previously updated water level gauge installation location (water level gauge installation location C) becomes the updated value. The AI ​​water level prediction update unit 20 and the second runoff analysis unit 40 repeatedly perform steps S2-S6 until the conditions for terminating the iterative processing are met.

[0051] If the conditions for terminating the iterative process are met (Step S5, Yes), the AI ​​water level prediction update unit 20 terminates the iterative process and outputs time-series data of predicted water levels for all water level gauge installation locations (water level gauge installation locations A, B, C) to the runoff analysis unit 30 (Step S7).

[0052] For example, the AI ​​water level prediction update unit 20 may repeatedly perform the update process until the amount of update to the time series data of the water level prediction value (the difference between the water level prediction value before the update and the water level prediction value after the update) falls below a threshold. The amount of update to the time series data may be, for example, the average, median, maximum, minimum, or other representative value of the amount of update to the water level prediction value data included in the time series data.

[0053] For example, the AI ​​water level prediction update unit 20 may terminate the iterative process based on the amount of update, etc., even before the process of updating all N water level gauge installation points is completed. When the amount of update converges to a certain value, it is considered that the topography of the area (drainage district) including water level gauge installation points A, B, and C and the structure of the sewer pipes are reflected in the water level prediction value. Therefore, there is no need to repeat the update process excessively, and a high level of accuracy in calculating the time-series data of the water level prediction value can be ensured.

[0054] The runoff analysis unit 30 uses time-series data of predicted water levels at water level gauge installation locations input from the AI ​​water level prediction update unit 20 to calculate time-series data of predicted water levels at any point on the pipeline (for example, point X where no water level gauge is installed) using a runoff analysis model (or a subset of the runoff analysis model), and outputs the time-series data of predicted water levels at water level gauge installation locations A, B, C and point X where no water level gauge is installed to an external device such as the display unit 2 (step S8).

[0055] The display unit 2 can visually present time-series data of water level predictions input from the runoff analysis unit 30 to the user, for example, by displaying numerical values ​​and colors on a map indicating the corresponding locations, or by displaying tables showing the time-series data.

[0056] According to the configuration of the real-time water level prediction device and real-time water level prediction method described above, runoff analysis is performed using water level prediction values ​​that take into account the height relationship of water levels due to physical connections, and water level prediction values ​​for locations where water level gauges are not installed are calculated using the runoff analysis model, thereby improving the accuracy of water level prediction calculations.

[0057] In other words, according to this embodiment, it is possible to provide a real-time water level prediction device, a real-time water level prediction method, and a computer program that perform highly accurate prediction of water levels inside sewer pipes.

[0058] In the above embodiment, the AI ​​water level prediction update unit 20 of the real-time water level prediction device repeatedly performs the update process until the termination condition is met, but the number of water level gauge installation locations to be updated may be as few as one. It is sufficient to update the time-series data of the water level prediction values ​​for at least one water level gauge installation location that satisfies the update condition. Even in this case, the calculation accuracy of the time-series data of the water level prediction values ​​for locations without water level gauges in the runoff analysis unit 30 can be improved. For example, the AI ​​water level prediction update unit 20 may update the one location with the lowest confidence level of the water level prediction value as the water level gauge installation location to be updated, and then output the time-series data of the water level prediction values ​​for all water level gauge installation locations to the runoff analysis unit 30.

[0059] Furthermore, while the real-time water level prediction device, real-time water level prediction method, and computer program of the above embodiment are applicable to processes that calculate water level prediction values ​​in real time, they may also be used in processes that calculate water level prediction values ​​offline. In that case as well, the same effects as those of the above embodiment can be obtained.

[0060] The program according to this embodiment may be transferred while stored on an electronic device, or it may be transferred while not stored on an electronic device. In the latter case, the program may be transferred via a network, or it may be transferred while stored on a storage medium. The storage medium is a non-temporary tangible medium. The storage medium is a computer-readable medium. The storage medium can be any medium that is capable of storing a program and is readable by a computer, such as a CD-ROM or memory card, and its form is not limited.

[0061] While several embodiments of the present invention have been described, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These novel embodiments can be carried out in a variety of other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims of the invention and its equivalents. [Explanation of Symbols]

[0062] 1...Real-time water level prediction device, 2...Display unit, 10...AI water level prediction unit (water level prediction unit), 20...AI water level prediction update unit (update unit), 30...Outflow analysis unit, 40...Second outflow analysis unit, DB1-DB4...Database

Claims

1. A water level prediction unit acquires time-series data of water levels measured by sensors installed in sewer pipes, time-series data of rainfall information for the area including the water level gauge installation points where the sensors are installed, structural information of the sewer pipes, and topographic information for the area including the water level gauge installation points, and calculates predicted water levels for multiple water level gauge installation points using a water level prediction model. An update unit acquires time-series data of water level prediction values ​​calculated by the water level prediction unit, outputs time-series data of water level prediction values ​​excluding the water level prediction values ​​of one of the multiple water level gauge installation locations to be updated, and updates the time-series data of the water level prediction values ​​of the water level gauge installation location to be updated using update data for updating the water level prediction values ​​of the water level gauge installation location to be updated. A second runoff analysis unit calculates the update data by runoff analysis using the time-series data of water level prediction values ​​output from the update unit and supplies it to the update unit. A real-time water level prediction device comprising: an update unit that acquires time-series data of water level prediction values ​​at multiple water level gauge installation locations from the update unit, and an outflow analysis unit that calculates time-series data of water level prediction values ​​at locations on the pipeline where water level gauges are not installed by outflow analysis.

2. The real-time water level prediction device according to claim 1, wherein the update unit repeatedly updates the time-series data of the predicted water level values ​​at the water level gauge installation locations by changing the water level gauge installation locations to be updated.

3. The real-time water level prediction device according to claim 2, wherein the update unit terminates the repetition of the update in accordance with the fact that it has updated the time-series data of water level prediction values ​​at all of the plurality of water level gauge installation locations.

4. The real-time water level prediction device according to claim 2, wherein the update unit terminates the repetition of the update when the amount of update of the predicted water level value at the water level gauge installation point falls below a threshold.

5. The real-time water level prediction device according to claim 2, wherein the update unit targets the water level gauge installation locations in an order based on the numbers assigned to each location.

6. The water level prediction unit supplies the confidence level of the time-series data of the water level prediction values ​​calculated by the water level prediction model to the update unit. The real-time water level prediction device according to claim 2, wherein the update unit targets a plurality of water level gauge installation locations in order of decreasing confidence level.

7. The real-time water level prediction device according to claim 1, wherein the update unit performs the update using a weighted average value of the update data obtained from the second outflow analysis unit and the time-series data of the predicted water level at the water level gauge installation point to be updated.

8. Time-series data of water levels measured by sensors installed in sewer pipes, time-series data of rainfall information for the area including the water level gauge installation points where the sensors are installed, structural information of the sewer pipes, and topographic information for the area including the water level gauge installation points are acquired, and water level prediction values ​​for multiple water level gauge installation points are calculated using a water level prediction model. Selecting one of the multiple water level gauge installation locations as the target for updating, the update data for the water level prediction value of the water level gauge installation location that is the target for updating is calculated by runoff analysis using time-series data of water level prediction values ​​excluding the water level prediction value of the water level gauge installation location that is the target for updating. The time-series data of the predicted water level at the water level gauge installation location to be updated is updated using the aforementioned updated data. A real-time water level prediction method that uses time-series data of predicted water levels at multiple water level gauge installation locations to calculate time-series data of predicted water levels at locations on the pipeline where water level gauges are not installed, through runoff analysis.

9. A computer program that causes a computer to execute the real-time water level prediction method described in claim 8.

Images