KR102619596B1 - A flood risk warning system that can secure the time nessary for proactive response in case of flooding and a flood risk warning method using the same - Google Patents

Abstract

The flood risk warning method according to the present invention includes the steps of receiving expected rainfall information per first unit time in a specific area, uniformly distributing the received expected rainfall per second unit time shorter than the first unit time, Generating information about the expected rainfall amount, calculating a predicted flood depth per second unit time for a specific area by using the expected rainfall amount per second unit time as an input value of a multi-level feedback model to perform short- and long-term memory type recurrent neural network learning. , receiving actual rainfall information after the elapse of a second unit of time, subtracting the actual rainfall after the elapse of the second unit of time received from the received expected rainfall per unit of time, and evenly redistributing it to the remaining time of the first unit of time. A step of regenerating the expected rainfall information per second unit time, and performing short- and long-term memory type recurrent neural network learning by using the regenerated expected rainfall amount per second unit time as an input value of a multi-level feedback model to obtain a second unit time for a specific area. It includes the step of recalculating the predicted flood depth.

Description

A flood risk warning system that can secure the time necessary for proactive response in the event of flooding and a flood risk warning method using the same SAME}

The present invention relates to a flood risk warning system that can secure the time necessary for proactive response in case of flooding, and a flood risk warning method using the same. In more detail, the present invention uses artificial intelligence to provide highly accurate prediction values for the depth of flooding of facilities, thereby preventing accidents by securing the time necessary for human evacuation or proactive response, and providing objective basis data for decision-making by relevant organizations. It relates to a flood risk warning system that can be used as a flood risk warning system and a flood risk warning method using the same.

Due to localized heavy rain, typhoons, and problems with urban drainage facilities, rivers and streams overflow, causing flooding in nearby underground facilities, resulting in damage to life and property. In particular, in the case of underground driveways or underground parking lots, flooding often occurs within a short period of time, and warnings about the risk of flooding are not properly provided, so there is a high possibility that it will lead to a fatal accident. To solve this problem, various flood risk prediction systems have been proposed, but the accuracy of flood depth predictions is not high and the prediction period is often long, making it difficult to use them as basis data for sudden heavy rain.

Korean Patent No. 10-2277753 (registered on July 9, 2021) Korean Patent No. 10-1379039 (registered on March 21, 2014)

The task of the present invention is to provide highly accurate flood depth prediction data to enable recognition in advance when there is a risk of flooding, such as sudden heavy rain, and to provide a flood risk warning that can secure sufficient time for proactive response such as evacuation. The goal is to provide a method.

Another task of the present invention is to provide a flood risk warning method that provides highly reliable flood depth prediction values using artificial intelligence learning. In particular, the accuracy of the input data used for artificial intelligence learning is improved by using the rainfall data announced by the Korea Meteorological Administration but reflecting the actual rainfall data in real time, and by calculating the predicted flood depth at intervals shorter than the weather forecast period of the Korea Meteorological Administration. , The purpose is to provide a highly reliable flood risk warning method in ever-changing weather situations.

Another task of the present invention is to provide a flood risk warning method that can present highly reliable flood depth forecasts using relationships with surrounding areas even when existing measurement data does not exist.

However, the problem to be solved by the present invention is not limited to the above-described problem, and may be expanded in various ways without departing from the spirit and scope of the present invention.

In order to achieve the above-described object of the present invention, the flood risk warning method according to exemplary embodiments can predict the depth of flooding by recombining rainfall information and applying it to artificial intelligence.

Specifically, the flood risk warning method includes: receiving expected rainfall information per first unit time in a specific area; generating information about the expected rainfall amount per second unit time by uniformly distributing the received expected rainfall amount to each second unit time that is shorter than the first unit time; calculating a second predicted flood depth per unit time for the specific area by performing short- and long-term memory-type recurrent neural network learning using the expected rainfall amount per second unit time as an input value of a multi-level feedback model; Receiving actual rainfall information after the second unit of time has elapsed; Regenerating the expected rainfall information per second unit time by uniformly redistributing the value obtained by subtracting the actual rainfall after the elapse of the received second unit time from the received expected rainfall per unit time to the remaining time of the first unit time. ; and performing short- and long-term memory-type recurrent neural network learning using the regenerated expected rainfall per second unit time as an input value of a multi-level feedback model, thereby recalculating the predicted flood depth per second unit time for the specific area. can do.

Meanwhile, the flood risk warning method may further include issuing a flood risk warning when the second predicted flood depth per unit time is greater than or equal to a preset threshold. It can be applied to warn of future flooding risks.

Differently, the flood risk warning method includes receiving actual flood depth measurement data for the specific area; And it may further include issuing a flood risk warning when the maximum value of the measured flood depth value and the second predicted flood depth per unit time is greater than or equal to a preset threshold. This case can be applied when you want to issue an alarm based on the current state.

In one embodiment, the first unit of time may be 1 hour, and the second unit of time may be 10 minutes.

In order to achieve the above-described object of the present invention, the flood risk warning method according to exemplary embodiments can calculate a predicted flood depth of a target area from flood depth data of other adjacent areas through regression analysis.

Specifically, the flood risk warning method includes receiving flood depth data of a first point and a second point spaced apart from the first point, respectively; calculating the elevation water level of the first point over time and the elevation water level of the second point over time using the received flood depth data; Calculating a delay time, which is a time interval between times when the water levels at the first point and the second point reach their maximum; By performing a regression analysis between the water levels at the first point and the second point, a relational expression is derived between the water level at the first point and the water level at the second point at a time with a time difference equal to the delay time. steps; And inputting the flood depth data measured at the first point into the relational equation to calculate a predicted value of the flood depth at the second point after the delay time.

In this case, the step of issuing a flood risk warning may be further included when the predicted flood depth of the second point is greater than or equal to a preset threshold.

In addition, the step of issuing the flood risk warning may include issuing a warning in step 1 when the predicted flood depth of the second point is greater than a preset first threshold value and less than a preset second threshold value; and issuing a second-stage warning when the predicted flood depth of the second point is greater than or equal to the second threshold.

In order to achieve the object of the present invention described above, the flood risk warning method according to exemplary embodiments includes a method of calculating a predicted flood depth by generating expected rainfall information through artificial intelligence learning, and a method of calculating a predicted flood depth through regression analysis. A method of calculating the predicted flood depth of the target area from local flood depth data can be used simultaneously.

Specifically, the flood risk warning method includes receiving flood depth data of a first point and a second point spaced apart from the first point, respectively; Regression analysis between the received flood depth data is performed to derive a relational expression between the elevation water level of the first point and the elevation water level of the second point, and the flood depth data of the first point is input into the relational expression to obtain the first point. Calculating a first flood depth prediction value expected at point 2; By generating expected rainfall information per unit time at the second point and performing short- and long-term memory type recurrent neural network learning by using the generated expected rainfall as an input value of a multi-level feedback model, the expected rainfall per unit time at the second point is performed. 2 Calculating a predicted flood depth; and issuing a flood risk warning when the maximum value among the first and second predicted flood depth values is greater than or equal to a preset threshold.

In this case, the step of calculating the first flood depth predicted value includes the elevation water level of the first point over time using the received flood depth data of the first and second points, and the water level over time. Calculating the elevation water level of each of the second points; Calculating a delay time, which is a time interval between times when the water levels at the first point and the second point reach their maximum; Regression analysis is performed between the elevation water levels of the first point and the second point, and the relationship between the elevation water level of the first point and the elevation water level appearing at the second point after a time difference equal to the delay time has elapsed. A step of deriving; And inputting the flood depth data measured at the first point into the relational equation to calculate a first flood depth predicted value expected at the second point when the delay time has elapsed.

In addition, calculating the second predicted depth of flooding may include receiving expected rainfall information per first unit time at the second point; generating information about the expected rainfall amount per second unit time by uniformly distributing the received expected rainfall amount to each second unit time that is shorter than the first unit time; calculating a predicted flood depth per second unit time for the second point by performing short- and long-term memory-type recurrent neural network learning using the expected rainfall amount per second unit time as an input value of a multi-level feedback model; Receiving actual rainfall information after the second unit of time has elapsed; A value obtained by subtracting the actual rainfall amount after the elapse of the received second unit time from the received expected rainfall amount per unit time is uniformly redistributed to the remaining time of the first unit time, so that the second stage regenerates the expected rainfall information per hour. step; and performing short- and long-term memory-type recurrent neural network learning using the regenerated expected rainfall per second unit time as an input value of a multi-level feedback model, thereby recalculating the predicted flood depth per second unit time for the second point. It can be included.

The flood risk warning method according to exemplary embodiments of the present invention provides highly accurate flood depth prediction data to enable recognition in advance when there is a risk of flooding, such as sudden heavy rain, for proactive response such as evacuation. You can secure enough time as needed. In addition, by providing objective flood depth prediction data, personnel from related organizations can use it as useful data to determine administrative policies.

In addition, the flooding risk warning method according to the present invention can greatly improve the accuracy of prediction because it calculates the flood depth prediction value using artificial intelligence learning. In particular, the accuracy of the input data used for artificial intelligence learning is improved by using the rainfall data announced by the Korea Meteorological Administration but reflecting the actual rainfall data in real time, and by calculating the predicted flood depth at intervals shorter than the weather forecast period of the Korea Meteorological Administration. , can respond quickly and appropriately to ever-changing weather situations.

In addition, the flood risk warning method according to the present invention can calculate a highly reliable flood depth forecast using the relationship with the surrounding area even when existing measurement data does not exist. Therefore, it has the advantage of being applicable under various environments and conditions.

In addition, the flood risk warning method according to the present invention can provide a full-cycle risk warning solution in the initial stage of introduction when there is no data at all, the early stage of operation when some data is secured, and the mature stage of operation when sufficient data is accumulated.

In addition, the flood risk warning method according to the present invention is a field trip to prevent the occurrence of an unexpected major accident in cases where the site environment changes due to facility and river construction or when the rainwater drain is blocked by foreign substances, etc. This can be forwarded to the relevant organizations for immediate action.

Figure 1 is a block diagram showing a flooding risk warning system according to an embodiment of the present invention.
Figure 2 is a flowchart for explaining the steps of a flood risk warning method according to an embodiment of the present invention.
FIG. 3 is a diagram illustrating short- and long-term memory-type recurrent neural network learning to which the multi-level feedback model of FIG. 2 is applied.
FIG. 4 is a diagram illustrating an example of generating the expected rainfall information per unit time of FIG. 2 .
Figure 5 is a flowchart for explaining the steps of a flood risk warning method according to another embodiment of the present invention.
FIG. 6 is a diagram for explaining the derivation of a relational expression through the regression analysis of FIG. 5.
Figure 7 is a flowchart for explaining the steps of a flood risk warning method according to another embodiment of the present invention.
Figure 8 is a flowchart explaining the flood warning cancellation steps according to the present invention.

Regarding the embodiments of the present invention disclosed in the text, specific structural and functional descriptions are merely illustrative for the purpose of explaining the embodiments of the present invention, and the embodiments of the present invention may be implemented in various forms. It should not be construed as limited to the embodiments described in.

Since the present invention can be subject to various changes and have various forms, specific embodiments will be illustrated in the drawings and described in detail in the text. However, this is not intended to limit the present invention to a specific disclosed form, and should be understood to include all changes, equivalents, and substitutes included in the spirit and technical scope of the present invention.

Terms such as first, second, etc. may be used to describe various components, but the components should not be limited by the terms. The above terms may be used for the purpose of distinguishing one component from another component. For example, a first component may be referred to as a second component, and similarly, the second component may be referred to as a first component without departing from the scope of the present invention.

When a component is said to be "connected" or "connected" to another component, it is understood that it may be directly connected to or connected to the other component, but that other components may exist in between. It should be. On the other hand, when it is mentioned that a component is “directly connected” or “directly connected” to another component, it should be understood that there are no other components in between. Other expressions that describe the relationship between components, such as "between" and "immediately between" or "neighboring" and "directly adjacent to" should be interpreted similarly.

The terms used in this application are only used to describe specific embodiments and are not intended to limit the invention. Singular expressions include plural expressions unless the context clearly dictates otherwise. In this application, terms such as “comprise” or “have” are intended to designate the presence of a described feature, number, step, operation, component, part, or combination thereof, but are not intended to indicate the presence of one or more other features or numbers. It should be understood that this does not preclude the existence or addition of steps, operations, components, parts, or combinations thereof.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person of ordinary skill in the technical field to which the present invention pertains. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related technology, and unless clearly defined in the present application, should not be interpreted as having an ideal or excessively formal meaning. .

Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the attached drawings. The same reference numerals are used for the same components in the drawings, and duplicate descriptions for the same components are omitted.

First, the flooding risk warning system according to the present invention will be described with reference to FIG. 1. Figure 1 is a block diagram showing a flooding risk warning system according to an embodiment of the present invention.

Referring to FIG. 1, the flooding risk warning system 10 according to an exemplary embodiment of the present invention includes a sensing unit 100 that measures water level changes in real time, data from the external server 20, and the sensing unit 100. It includes a flood risk detection server 200 that determines whether there is a risk of flooding using measured measurement data, and an alarm unit 300 that notifies the user if it is determined that there is a risk of flooding.

The sensing unit 100 includes a river water level meter 110 that monitors changes in the water level of the river, a riverside flood depth meter 120 that monitors changes in the flood depth of riverside areas adjacent to the river, and a riverside flood depth meter 120 that monitors changes in the flood depth of a specific facility. The facility may include a submersion depth meter (130). The river water level gauge 110 may be equipment that measures the water level of the river using electromagnetic waves, etc., and the flood depth gauges 120 and 130 may be equipment that measures the flood depth using pressure changes. However, the present invention is not limited to this, and it is sufficient as long as changes in water level can be measured precisely in real time. Meanwhile, the above facilities refer to facilities that require flood depth monitoring, and may include tunnels, bridges, underground parking lots, etc.

The water level and flood depth data measured by the sensing unit 100 may be provided to the flood risk detection server 200 by wire or wirelessly. The flooding risk detection server 200 may be, for example, a central management server or processor.

The flooding risk detection server 200 can receive measured water level data from the sensing unit 100 and can receive weather forecasts, topographic information, flooding history information, etc. from the external server 20. The flooding risk detection server 200 can learn an artificial intelligence model using the received information (learning unit, 210) and, based on this, predict the depth of flooding of the facility over time (predicting unit, 220).

In particular, the prediction unit 220 can not only calculate the predicted flood depth of the facility over time using the learned artificial intelligence model, but also determine whether the calculated predicted flood depth is at a level that requires a warning. . This will be described in more detail later with reference to FIGS. 2 to 8.

The predicted flood depth calculated by the flood risk detection server 200, the actual measured flood depth, and the need to issue an alarm may be provided to the user through the user terminal 310. For example, the user terminal 310 may be a smartphone, a computer of a related organization, etc. Individual users can use the information provided on their smartphones to immediately check the risk of flooding or whether an alert has been issued, and public officials from related organizations can use it as data to determine the possibility of flooding.

If the flooding risk detection server 200 determines that an alert is necessary, an alert can be issued through the alert unit 300. For example, if it is determined that there is a risk of flooding in the riverside, the riverside alarm 320 can be activated to warn people around, and if it is determined that there is a risk of flooding in the facility, the facility alarm 330 can be activated to alert people around. You can urge people to evacuate. At this time, the alarms 320 and 330 may display information related to flooding, operate warning lights of different colors depending on the risk level, or emit warning sounds.

In one embodiment, the alarms provided by the alarms 320 and 330 may operate in different ways depending on the level. For example, if a total of two levels of alarm are set, an orange warning light can be turned on and a short warning sound can be transmitted during the first level warning, and a red warning light can be turned on and a long warning sound can be transmitted during the second level warning. .

Meanwhile, the alarm unit 300 may be connected to the flooding risk detection server 200 by wire or wirelessly. Differently, the riverside alarm 320 may be connected to the riverside flood depth meter 120 by wire, and the facility alarm 330 may be connected to the facility flood depth meter 130 by wire. This is to receive flooding information directly from the measuring devices 120 and 130 and warn the user even when wireless communication is not possible due to a flood or the like.

Hereinafter, a method for warning of a flood risk using the flood risk warning system 10 described in FIG. 1 will be described in detail. FIG. 2 is a flowchart for explaining the steps of a flood risk warning method according to an embodiment of the present invention, FIG. 3 is a diagram for explaining long-term memory type recurrent neural network learning to which the multi-level feedback model of FIG. 2 is applied, and FIG. 4 is a diagram for explaining an example of generating expected rainfall information per unit time in FIG. 2. Figure 5 is a flowchart for explaining the steps of a flood risk warning method according to another embodiment of the present invention, and Figure 6 is a diagram for explaining the derivation of a relational expression through the regression analysis of Figure 5. Figure 7 is a flowchart for explaining the steps of a flood risk warning method according to another embodiment of the present invention. Additionally, Figure 8 is a flowchart for explaining the flood warning cancellation steps according to the present invention.

First, Figures 2 to 4 present a method of determining the risk of flooding by calculating a predicted flood depth using expected rainfall information. Flood depth information measured in real time has the advantage of being an accurate value, but it is insufficient to be used as data to determine the possibility of flooding of underground facilities. Therefore, there is a need to predict and provide future flood depth, and it is necessary to issue a warning in advance using the predicted flood depth value or to take appropriate response from related organizations. The embodiment of FIGS. 2 to 4 predicts future flooding depth by reflecting rainfall forecasts in artificial intelligence learning. In particular, it has the advantage of being able to predict changes in inundation depth at shorter intervals than the general rainfall forecast cycle, and greatly improving the accuracy of prediction by continuously reflecting actual rainfall information. This will be explained in detail.

First, the flood depth data of the facility requiring monitoring is continuously received (S110). For example, a facility flood depth measuring device 130 can be installed in a facility such as an underground roadway, and the actual flood depth measurement value of the facility can be continuously monitored using this. At the same time, the expected rainfall information provided by the Korea Meteorological Administration is used, but the expected rainfall information with a shorter cycle is generated by reflecting the actual rainfall data (S120). This is to provide a flood risk warning to ensure sufficient time for evacuation or response even when the depth of flooding changes rapidly, such as when heavy rain falls in a short period of time.

At this time, the present invention can generate a flood depth prediction value using a Long Short-Term Memory model (LSTM) Recurrent Neural Network (RNN) applying a multi-step feedback model. there is. This is shown in Figure 3.

As shown in Figure 3, the multi-level model predicts future m data values using n past data, and the accuracy of the predicted values can be improved by using the predicted values as input (feedback). However, this multi-level feedback model has the problem that if the accuracy of the predicted value fed back is low, the reliability of subsequent predictions may be lowered, and the accuracy gradually decreases as the prediction is repeated. In order to complement this, the present invention does not simply utilize the expected rainfall information announced by the Korea Meteorological Administration, but continuously reflects the actual rainfall information to improve the accuracy of the expected rainfall, thereby greatly improving the accuracy of predicting the depth of flooding. In addition, by generating the expected rainfall per unit time that is shorter than the forecast unit time and using it to predict the depth of flooding, it is possible to secure sufficient response time in urgent flooding situations.

Specifically, as shown in FIGS. 2 and 4, expected rainfall information per first unit time is received (S121). The Korea Meteorological Administration announces expected rainfall information at hourly intervals, and the first unit time may be, for example, 1 hour. In the present invention, expected rainfall information is generated for a second unit time that is shorter than the first unit time, and the second unit time may be, for example, 10 minutes. The first and second unit times may be changed in various ways depending on local weather characteristics, weather forecast cycles, policy necessities, etc.

In other words, the present invention generates expected rainfall information in 10-minute increments using the Korea Meteorological Administration's expected rainfall information announced at hourly intervals. At this time, actual rainfall information is reflected in order to improve the accuracy of the expected rainfall information. For example, as shown in FIG. 4, assume that there was a forecast from the Korea Meteorological Administration that 60 mm of rain was expected for 1 hour from 1 o'clock. In this case, the total rainfall for 1 hour (60mm) can be divided into 10-minute increments (10mm) and distributed as 10mm of rainfall is expected for each 10-minute interval from 1 o'clock to 2 o'clock. At 1 o'clock sharp, the predicted flood depth can be calculated by applying the six data set in 10-minute increments to a multi-level feedback model. Afterwards, actual rainfall information is obtained every 10 minutes and the predicted rainfall information is continuously modified. For example, if the actual rainfall from 1:00 to 1:10 is 0 mm, the Korea Meteorological Administration's forecast rainfall of 60 mm can be redistributed to the remaining 50 minutes. In this case, the expected rainfall per 10 minutes may be revised from 10 mm to 12 mm, and the predicted flood depth may also change accordingly. Additionally, if the actual rainfall between 1:10 and 1:20 is 20mm, the actual rainfall (20mm) minus the Korea Meteorological Administration's predicted rainfall (60mm) can be redistributed to the remaining 40 minutes. In this case, the expected rainfall after 1:20 is revised to 10 mm every 10 minutes, and the predicted flood depth may also be changed again. By continuously repeating this process, the accuracy of expected rainfall and inundation depth information can be greatly improved.

In this way, the present invention receives expected rainfall information per first unit time (S121), receives actual rainfall information per second unit time that is shorter than the first unit time (S123), and receives the expected rainfall amount per first unit time (S123). The value obtained by subtracting the actual rainfall amount can be distributed to the remaining time of the first unit time to generate expected rainfall information per second unit time (S125). By applying the predicted rainfall information per second unit time generated in this way to learning a short- and long-term memory type recurrent neural network, it is possible to calculate a predicted flood depth per second unit time with dramatically improved accuracy (S130).

The predicted flood depth generated in this way can be provided to the user terminal 310 or related organizations. Users can prepare for the risk of flooding in advance by looking at the forecast information displayed on the terminal 310, and related organizations can secure sufficient time for evacuation or response by using this for decision-making.

Additionally, the generated flood depth predictions can issue immediate alerts to underground facilities. For example, if the actual measured flood depth value or the predicted flood depth value within a certain time is greater than a preset threshold (S140), an immediate alert can be issued to the alarms 320, 330 and the user terminal 310 at the corresponding location. (S150).

For example, as shown in FIG. 2, when the maximum value of the measured flood depth value and the predicted flood depth value is greater than or equal to the preset first threshold value or less than the second threshold value, a first-stage warning may be issued (S151). . The first level alarm may, for example, flash an orange warning light or sound a short warning sound. Differently, if the maximum value of the actual measured flood depth value and the predicted flood depth value is greater than or equal to a preset second threshold value, a second-level warning can be issued to induce immediate evacuation (S153). The second level warning may be, for example, flashing a red warning light or sounding a continuous warning sound. In this case, the predicted flood depth value used to determine whether to issue a warning may vary depending on the climate characteristics, geographical characteristics, policy decisions, etc. of the area. For example, the predicted flood depth value within 20 minutes can be used to issue a warning.

Meanwhile, in the warning necessity determination step (S140) shown in FIG. 2, it is determined whether the current state requires a flood warning. For this reason, the need for a warning is determined by simultaneously utilizing the predicted value and the current actual measurement value. Differently, in the case of determining the need for a warning at a specific point in the future, since there are no actual measured values, the need for a warning can be determined simply using the predicted value.

As described above, the flood risk warning method according to an embodiment of the present invention disclosed in FIGS. 2 to 4 utilizes expected rainfall data for artificial intelligence learning to provide a predicted flood depth value, thereby providing sufficient time for proactive response in case of flooding. It can be secured. In particular, rather than using the Korea Meteorological Administration's rainfall forecast data as is, it has the advantage of being able to generate flood depth forecasts at intervals shorter than the Korea Meteorological Administration's announcement cycle, and greatly improving the accuracy of predictions by continuously reflecting actual rainfall data.

However, since this embodiment is a method of predicting future data using past data, sufficient reliability can be expected when measurement data is accumulated to a certain extent. Accordingly, in Figures 5 and 6, an embodiment that can provide reliable flood depth prediction values even when existing data does not exist will be described.

Figures 5 and 6 present a method of predicting the flood depth of an area where a facility is located by using flood depth data from another point distant from it when there is no measured data for the point requiring prediction. In this case, it is possible to provide predicted values for the depth of flooding even when measurement data does not exist, and has the advantage of being able to further improve accuracy as measured values are accumulated. This will be explained in detail.

First, it continuously receives flood depth data from surrounding areas and facilities requiring monitoring (S210). Here, the surrounding area may be the water level of a river adjacent to the facility, the water level of a riverside, the water level of another facility, etc., and can be obtained through the sensing unit 100. In other words, the flood depth information of the target point is predicted by using the water level change in the adjacent river or the flood depth information of other areas where flooding occurred first. In this way, using the flood depth data of adjacent areas and the flood depth data of the area where the facility is located, a relationship between them is derived (S220), and the predicted flood depth of the area where the facility is located is calculated using the relationship equation (S230). For convenience of explanation, hereinafter, the point where the underground facility requiring flooding monitoring is located will be referred to as the comparison point, and the surrounding area that serves as the standard will be referred to as the reference point.

In the present invention, a correlation equation is derived between the altitude of the reference point and the comparison point using regression analysis (S220). Specifically, the elevation water levels of the reference point and comparison point over time are calculated respectively (S221). As shown in FIG. 6(a), the elevation water level of the reference point may be displayed as an A line, and the elevation water level of the comparison point may be displayed as a B line. Next, the delay time (Δt) between the reference point and the comparison point is calculated (S223). Here, the delay time (Δt) may be the time between the time when the water level of the reference point is maximum and the time when the water level of the comparison point is maximum. For example, if the reference point is a river in a nearby area, it can be understood that changes in the water level of the river appear at the comparison point with a time difference equal to the delay time (Δt).

Afterwards, the time when the elevation water level of the reference point is maximum and the time when the elevation water level of the comparison point is maximum are matched (Figure 6(b)), and the elevation water level at this time is plotted (Figure 6(c)) to perform regression analysis. By performing , the relationship between the water level at the reference point and the water level at the comparison point is derived (S225). The above relational expression may be a relational expression that outputs the elevation water level of the comparison point at a specific time (t ₁ +Δt) corresponding to the input of the elevation water level of the reference point at a specific time (t ₁ ).

Afterwards, when rainfall occurs, the flood depth data at the reference point measured by the sensing unit 100 is input into the above-derived relational equation, thereby calculating the predicted flood depth at the comparison point after the delay time (Δt) (S230). .

Meanwhile, the above-described flood depth prediction method can be used when existing data does not exist, but can also be used when data is secured through continuous measurement. For example, after sufficient inundation depth data is secured using the above-described method, the short- and long-term memory type recurrent neural network applying the multi-level feedback model described in Figure 2 can be used as is, and the actual elevation water level value is used as the feedback value. The accuracy of predictions can be greatly improved.

The predicted flood depth generated thereafter may be provided to the user terminal 310 or related organizations. Users can prepare for the risk of flooding in advance by looking at the forecast information displayed on the terminal 310, and related organizations can secure sufficient time for evacuation or response by using this for decision-making.

Additionally, the generated flood depth predictions can issue immediate alerts to underground facilities. For example, if the actual measured flood depth value or the predicted flood depth value within a certain time is greater than a preset threshold (S240), an immediate alert can be issued to the alarms 320, 330 and the user terminal 310 at the corresponding location. (S250). More specifically, when the maximum value of the actual measured flood depth value and the predicted flood depth value is greater than or equal to a preset first threshold value or less than a second threshold value, a first-stage warning may be issued (S251). Differently, if the maximum value of the actual measured flood depth value and the predicted flood depth value is greater than or equal to a preset second threshold, a second-level warning can be issued to induce immediate evacuation (S253).

Meanwhile, in the warning necessity determination step (S240) shown in FIG. 5, it is a step to determine whether the current state requires a flood warning. For this reason, the need for a warning is determined by simultaneously utilizing the predicted value and the current actual measurement value. Differently, in the case of determining the need for a warning at a specific point in the future, since there are no actual measured values, the need for a warning can be determined simply using the predicted value.

As described above, the flood risk warning method according to an embodiment of the present invention disclosed in FIGS. 5 and 6 can predict the flood depth of an area requiring flood monitoring using flood depth data of other adjacent areas, and use this to Sufficient time can be secured for proactive response in the event of flooding. In particular, this embodiment can be used even when there is no existing measurement data at all, and has the advantage that the accuracy gradually improves as measurement data is accumulated.

In Figure 7, a flood risk warning method using the two methods described above is disclosed. In other words, the first flood depth predicted value is obtained by applying the expected rainfall information to artificial intelligence learning, and at the same time, the second flood depth predicted value is obtained by deriving a relationship with the adjacent area through regression analysis, and the first and This is a method that utilizes all the second flood depth prediction values. Although this embodiment is applicable only when a certain amount of measured data exists, it has the advantage of being able to predict the depth of flooding more precisely and improving the reliability of flood risk warnings.

Specifically, the sensing unit 100 is used to continuously receive data on the water level of a river or riverside and the depth of flooding of a facility requiring monitoring (S310).

At the same time, expected rainfall information is generated (S320). Specifically, as shown in FIG. 2, expected rainfall information per first unit time is received, actual rainfall information per second unit time that is shorter than the first unit time is received, and actual rainfall information is received in the expected rainfall amount per first unit time. The value obtained by subtracting the rainfall amount may be distributed to the remaining time of the first unit time to generate expected rainfall information per second unit time. By applying the predicted rainfall information generated in this way to learning a short- and long-term memory type recurrent neural network, the second predicted flood depth per unit time (hereinafter referred to as the first predicted flood depth) can be calculated (S330).

In addition, regression analysis is used to derive the relationship between the elevation levels of the reference point and the comparison point (S325). Specifically, as shown in Figure 5, the elevation water levels of the reference point and comparison point are calculated over time, and the difference between the time when the elevation water level of the reference point is maximum and the time when the elevation water level of the comparison point is maximum is calculated. Calculate the delay time (Δt), which is the time. Afterwards, the time when the water level of the reference point is the maximum and the time when the water level of the comparison point is the maximum are matched, the water levels are plotted, and regression analysis is performed to form the relationship between the water level of the reference point and the water level of the comparison point. can be derived. By substituting the actual measured flood depth value of the reference point or the predicted flood depth predicted by artificial intelligence learning into the relational expression derived in this way, the predicted flood depth value of the comparison point (hereinafter referred to as the first predicted flood depth value) can be calculated (S330 ).

Through the above process, the first and second flood depth predicted values can be obtained (S330). In the present invention, only an example of obtaining a total of two predicted flood depth values has been described, but if necessary, a larger number of predicted flood depth values can be obtained using various types of prediction methods or artificial intelligence learning models.

Using the actual measured and predicted flood depth values obtained in this way, the need for issuing an alarm is determined (S340), and if necessary, an alarm is issued (S350). For example, when the maximum value of the actual measured flood depth value and the predicted flood depth value is greater than or equal to a preset first threshold value or less than a second threshold value, a first-level warning may be issued (S351). Differently, if the maximum value of the actual measured flood depth value and the predicted flood depth value is greater than or equal to a preset second threshold, a second-stage warning can be issued to induce immediate evacuation (S353).

Meanwhile, in the warning necessity determination step (S340) shown in FIG. 7, it is determined whether the current state requires a flood warning. For this reason, the need for warning is determined by simultaneously utilizing predicted values and current actual measurements. Unlike this, in the case of determining the need for a warning at a specific point in the future, since there is no actual measured value, the need for a warning can be determined simply by using the largest value among the predicted values.

As described above, the flooding risk warning method according to an embodiment of the present invention disclosed in FIG. 7 can calculate a plurality of flood depth prediction values by simultaneously utilizing various highly reliable methods, and at least one of the calculated prediction values is a threshold value. If it exceeds , an alert is issued, allowing the risk of flooding to be predicted more reliably. In addition, various forecast values can be provided to the user at the same time, so the person in charge of the relevant organization can use it as useful evidence to make policy decisions such as evacuation.

Figure 8 is a diagram to explain the process of releasing an alarm after it is issued. Specifically, river water level and flood depth data are continuously monitored (S410). Afterwards, predicted flood depth values are calculated using various prediction methods illustrated in FIG. 7. And only when both the actual measured flood depth value and the calculated predicted flood depth value are smaller than the first threshold (S420), the warning is issued (S440). In other words, the alarm is canceled only when the alarm level falls below level 1. For example, if a level 2 warning is currently being issued, the warning is not canceled even if the flood depth level falls to the level 1 warning level, and the warning is canceled only when it falls lower than the level 1 warning level.

10: Flood risk warning system 20: External server
100: sensing unit 110: river water level measuring device
120: Riverside flood depth measuring device 130: Facility flooding depth measuring device
200: Flooding risk detection server 210: Learning department
220: prediction unit 300: warning unit
310: User terminal 320: River change alarm
330: Facility alarm

Claims (10)

Receiving expected rainfall information per first unit of time in a specific area;
generating information about the expected rainfall amount per second unit time by uniformly distributing the received expected rainfall amount to each second unit time that is shorter than the first unit time;
calculating a second predicted flood depth per unit time for the specific area by performing short- and long-term memory-type recurrent neural network learning using the expected rainfall amount per second unit time as an input value of a multi-level feedback model;
Receiving actual rainfall information after the second unit of time has elapsed;
Regenerating the expected rainfall information per second unit time by uniformly redistributing the value obtained by subtracting the actual rainfall after the elapse of the received second unit time from the received expected rainfall per unit time to the remaining time of the first unit time. ; and
Comprising the step of recalculating the predicted flood depth per second unit time for the specific area by performing short- and long-term memory-type recurrent neural network learning using the regenerated expected rainfall amount per second unit time as an input value of a multi-level feedback model. How to warn of flooding hazards. The flood risk warning method according to claim 1, further comprising issuing a flood risk warning when the second predicted flood depth per unit time is greater than or equal to a preset threshold. According to paragraph 1,
Receiving actual flood depth data for the specific area; and
A flood risk warning method further comprising issuing a flood risk warning when the maximum value of the measured flood depth value and the second predicted flood depth per unit time is greater than or equal to a preset threshold. The method according to claim 1, wherein the first unit time is 1 hour and the second unit time is 10 minutes. delete delete delete Receiving flood depth data of a first point and a second point spaced apart from the first point, respectively;
Regression analysis between the received flood depth data is performed to derive a relational expression between the elevation water level of the first point and the elevation water level of the second point, and the flood depth data of the first point is input into the relational expression to obtain the first point. Calculating a first flood depth prediction value expected at point 2;
By generating expected rainfall information per unit time at the second point and performing short- and long-term memory type recurrent neural network learning by using the generated expected rainfall as an input value of a multi-level feedback model, the expected rainfall per unit time at the second point is performed. 2 Calculating a predicted flood depth; and
A flood risk warning method comprising issuing a flood risk warning when the maximum value among the first and second flood depth predicted values is greater than or equal to a preset threshold. The method of claim 8, wherein the step of calculating the first flood depth predicted value is,
calculating the elevation water level of the first point over time and the elevation water level of the second point over time using the received flood depth data of the first and second points;
Calculating a delay time, which is a time interval between times when the water levels at the first point and the second point reach their maximum;
By performing a regression analysis between the elevation water levels of the first point and the second point, the relationship between the elevation water level of the first point and the elevation water level appearing at the second point after a time difference equal to the delay time has elapsed. A step of deriving; and
A flood risk warning, comprising the step of inputting the flood depth data measured at the first point into the relational equation to calculate a first flood depth predicted value expected at the second point when the delay time has elapsed. method. The method of claim 8, wherein the step of calculating the second flood depth predicted value is,
Receiving expected rainfall information per first unit time at the second point;
generating information about the expected rainfall amount per second unit time by uniformly distributing the received expected rainfall amount to each second unit time that is shorter than the first unit time;
calculating a predicted flood depth per second unit time for the second point by performing short- and long-term memory-type recurrent neural network learning using the expected rainfall amount per second unit time as an input value of a multi-level feedback model;
Receiving actual rainfall information after the second unit of time has elapsed;
A value obtained by subtracting the actual rainfall amount after the elapse of the received second unit time from the received expected rainfall amount per unit time is uniformly redistributed to the remaining time of the first unit time, so that the second stage regenerates the expected rainfall information per hour. step; and
Comprising a step of recalculating a predicted flood depth per second unit time for the second point by performing short- and long-term memory-type recurrent neural network learning using the regenerated expected rainfall per second unit time as an input value of a multi-level feedback model. A flood risk warning method characterized in that.

Images