Customized flood prediction and warning system using smart pole and virtual sensor data

Abstract

A customized flood warning system using smart pole and virtual sensor data according to an embodiment of the present invention is characterized by comprising: a data collection unit that receives measurement data including water level and rainfall from a plurality of smart poles; a data processing unit that corrects missing values ​​or outliers in the received measurement data; a water level prediction unit that predicts a river water level by combining the actual measured water level of the smart pole and the predicted water level of the virtual sensor, wherein the virtual sensor generates a predicted water level for one or more prediction points where no physical sensor is installed using the corrected measurement data; and a warning unit that generates a warning when it is determined that the predicted river water level will exceed a preset reference water level.

Description

Technology Field

[0001] The present invention relates to a technology for predicting and warning of flood risks in rivers, and more specifically, to a system that integrates and analyzes real-time data collected from Internet of Things (IoT) sensors such as Smart Poles and AI-based virtual sensor data, and additionally utilizes upstream dam discharge information as needed to provide flood warnings by reflecting the characteristics of each river point. Background Technology

[0003] Recently, due to the impact of climate change, abnormal weather phenomena such as localized torrential rains are occurring frequently, and the resulting sudden rise in river levels and flood damage are emerging as significant social issues. To address this, flood control centers have long operated a system to install observation stations at key points along rivers, monitor water level data, and issue flood advisories.

[0004] Recently, with the advancement of Internet of Things (IoT) technology, intelligent monitoring equipment such as 'Smart Poles' that collect various environmental data, including water levels, rainfall, temperature, and humidity in remote areas in real time at one-minute intervals, is being introduced. Smart Poles transmit the collected data to a central IoT platform via a communication network, and this data is utilized as important basic data for flood prediction and decision support systems.

[0005] However, these conventional smart pole-based monitoring systems have the following clear limitations.

[0006] First is the issue of data reliability. Since smart poles are installed and operated in remote locations, they are susceptible to communication failures or defects in the equipment itself. This can lead to "missing data" situations where no data is received, or "outliers" where abnormal values ​​deviate from surrounding data are measured due to sensor errors. If such low-quality data is used directly in flood prediction models, the accuracy of the predictions can be severely degraded.

[0007] Second, there is a limitation in the monitoring range. Realistically, installing expensive smart poles densely along every section of a river entails a significant financial and management burden. Consequently, numerous unmeasured points without sensors remain blind spots in monitoring, making it difficult to identify localized water level changes or dangerous situations in those sections.

[0008] Third, there is the issue of the uniformity of prediction methods. Rivers have different characteristics in their upstream and downstream sections, and their topographical features are very complex, including confluences where tributaries merge, bends where the river curves, and straight sections with high flow velocities. Nevertheless, existing early warning systems often fail to adequately reflect these specific characteristics and frequently apply uniform standards, resulting in limitations in prediction accuracy. The problem to be solved

[0010] The present invention was devised to solve the problems of the prior art described above, and its first objective is to provide a flood early warning system capable of securing high-quality data by effectively processing missing values ​​and outliers that impair the reliability of data received from smart poles.

[0011] The second objective of the present invention is to provide a flood early warning system that eliminates blind spots in monitoring by predicting water levels at unmeasured points where physical sensors are not installed using an AI-based virtual sensor, and further improves the reliability of the entire system by providing a mechanism to independently verify and correct the prediction accuracy of the virtual sensor.

[0012] The third objective of the present invention is to provide a flood warning system that maximizes the accuracy and speed of flood prediction by applying a customized warning algorithm that comprehensively considers the topographical characteristics of a river (confluences, bends, etc.) and water level conditions (high water level, low water level).

[0013] Meanwhile, other unspecified objectives of the present invention will be further considered to the extent that they can be easily inferred from the following detailed description and effects. means of solving the problem

[0015] To solve the problem described above, the following solution is proposed.

[0016] A customized flood warning system using smart pole and virtual sensor data according to an embodiment of the present invention is characterized by comprising: a data collection unit that receives measurement data including water level and rainfall from a plurality of smart poles; a data processing unit that corrects missing values ​​or outliers in the received measurement data; a water level prediction unit that predicts a river water level by combining the actual measured water level of the smart pole and the predicted water level of the virtual sensor, wherein the virtual sensor generates a predicted water level for one or more prediction points where no physical sensor is installed using the corrected measurement data; and a warning unit that generates a warning when it is determined that the predicted river water level will exceed a preset reference water level.

[0017] In one embodiment, the data processing unit may be characterized by requesting the data of the missing section from the smart pole again to update it when communication is normalized, if missing data occurs in the measurement data due to a communication failure, and correcting the outlier using linear interpolation when the measurement data deviates from a preset range and is determined to be an outlier.

[0018] In one embodiment, the water level prediction unit may be characterized by generating the predicted water level by inputting time series data collected from a plurality of smart poles located around the prediction point into an AI model.

[0019] In one embodiment, the water level prediction unit may be characterized by varying the combination of surrounding smart poles input to the AI ​​model depending on whether the topographical characteristics of the prediction point correspond to a confluence, a curve, or a straight section.

[0020] In one embodiment, the water level prediction unit may be characterized by using a high-level prediction AI model when the current water level of the smart pole around the prediction point is above a preset threshold water level, and using a low-level prediction AI model when the current water level is below the threshold water level to generate the predicted water level.

[0021] In one embodiment, the system may further include: a verification virtual sensor created separately from the virtual sensor at the prediction point at a verification point where an actual sensor of any one of the smart poles is installed; and a virtual sensor verification unit that generates a verification prediction level, which is a predicted value of the verification virtual sensor, using measurement data from other smart poles around the verification point, and calculates a verification error by comparing the verification prediction level with the actual measured level measured by the actual sensor.

[0022] In one embodiment, the water level prediction unit may be characterized by correcting the predicted water level generated by the virtual sensor at the prediction point using a verification error calculated by the virtual sensor verification unit.

[0023] In one embodiment, the virtual sensor verification unit determines the similarity of hydraulic or topographical characteristics between the prediction point and the verification point, generates a correction coefficient by applying a weight to the verification error based on the similarity, and the water level prediction unit corrects the predicted water level using the correction coefficient.

[0024] In one embodiment, the early warning unit may be characterized by setting 'advisory' and 'warning' standard water levels for each point based on past measurement data and planned flood level information of each smart pole point.

[0025] In one embodiment, the early warning unit may be characterized by calculating a future expected water level based on the current water level and the slope of a water level curve representing the amount of change in water level over a past certain period of time (60 minutes), and generating an alarm by comparing the expected water level with the reference water level.

[0026] In one embodiment, the early warning unit may be characterized by transmitting a notification message to a pre-registered administrator's terminal, including the location of occurrence, time of occurrence, current water level, expected time of arrival, expected water level of arrival, and alarm grade, when an alarm occurs. Effects of the invention

[0028] According to the present invention, the following effects are achieved.

[0029] First, by effectively handling missing data and outliers and applying customized water level prediction algorithms that reflect the characteristics of each river point, it is possible to predict flood risks and issue warnings more quickly and accurately than the forecasts of existing flood control centers. This secures sufficient time for disaster response, thereby minimizing casualties and property damage.

[0030] Second, AI-based virtual sensor technology enables dense monitoring at low cost, even in areas where it is difficult to install expensive physical sensors, thereby maximizing the efficiency of river management and eliminating blind spots in monitoring.

[0031] Third, through a unique mechanism that dynamically verifies and corrects the AI ​​prediction model by comparing the values ​​of actual sensors and virtual sensors, the prediction reliability of the entire system can be dramatically improved.

[0032] Meanwhile, it should be added that even if an effect is not explicitly mentioned here, the effects described in the following specification and the provisional effects expected by the technical features of the present invention are treated as described in the specification of the present invention. Brief explanation of the drawing

[0034] FIG. 1 is a block diagram showing the overall configuration of a flood early warning system according to one embodiment of the present invention. Figure 2 is a flowchart illustrating the process of processing missing values ​​and outliers in the data processing unit of the present invention. Figure 3 is a conceptual diagram illustrating how the virtual sensor of the present invention applies a different prediction model depending on the topography of the river (confluence, bend, straight section). Figure 4 is a flowchart illustrating the process of verifying and correcting prediction accuracy using a virtual sensor for verification in the virtual sensor verification unit of the present invention. Figure 5 is a flowchart illustrating the process of predicting future water levels and generating alarms in the early warning unit of the present invention. Figure 6 is a graph showing a comparison of the timing of early warning issuance by the system of the present invention and the existing flood control center during an actual heavy rain situation. Figure 7 is an example of a flood warning notification message transmitted to a user terminal according to the present invention. It should be noted that the attached drawings are provided as examples for reference to help understand the technical concept of the present invention, and the scope of the rights of the present invention is not limited by them. Specific details for implementing the invention

[0035] Hereinafter, with reference to the drawings, we will examine the configuration of the present invention as guided by various embodiments thereof and the effects derived therefrom. In describing the present invention, detailed descriptions of related known functions are omitted if they are deemed obvious to a person skilled in the art and could unnecessarily obscure the essence of the invention.

[0036] As used in this document, the terms “part” or “module” may include a unit implemented in hardware, software, or firmware, and may be used interchangeably with terms such as logic, logic block, component, or circuit. A module may be a component formed as a whole, or a minimum unit of said component or a part thereof that performs one or more functions.

[0037] In this document, "~part," "module," or "node" performs tasks such as moving, storing, or converting data using computing devices such as CPUs and APs. For example, "module" or "node" can be implemented as devices such as servers, PCs, tablet PCs, and smartphones.

[0038] The present invention aims to overcome the limitations of existing river monitoring systems and provide faster and more accurate flood warning services by organically combining data processing, AI-based virtual sensors, and location-specific early warning algorithms based on IoT sensor networks such as smart poles.

[0039] FIG. 1 is a block diagram showing the overall configuration of a flood early warning system (100) according to one embodiment of the present invention.

[0040] Referring to FIG. 1, the flood warning system (100) according to the present invention may be configured to include a data collection unit (110), a data processing unit (120), a water level prediction unit (130), a virtual sensor verification unit (140), and a warning unit (150).

[0041] The data collection unit (110) receives raw measurement data, including water level, rainfall, temperature, humidity, etc., from multiple smart poles (10) installed in the river basin at short intervals such as 1 minute. Additionally, if necessary, data from external observation stations such as the National Water Resources Management Information System (WAMIS) and discharge information from upstream dams may be additionally received via API.

[0042] The collected raw data is transmitted to the data processing unit (120). The data processing unit (120) plays a role in preventing prediction errors that may occur if raw data containing missing values ​​due to communication failures or outliers due to sensor errors is used as is. To this end, it supplements missing data and corrects outliers to generate high-quality refined data.

[0043] The refined data is input into the water level prediction unit (130). The water level prediction unit (130) includes not only water level information at the point where the actual smart pole (10) is installed, but also a virtual sensor (132) that predicts the water level at an unmeasured point where there is no physical sensor. The virtual sensor (132) receives data from surrounding actual sensors and predicts the water level at the unmeasured point through an AI model.

[0044] The virtual sensor verification unit (140) is configured to ensure the reliability of the predicted value generated by the virtual sensor (132) of the water level prediction unit (130). The virtual sensor verification unit (140) creates a separate virtual sensor for verification at the location where the actual sensor is located and calculates the difference (verification error) between the actual value and the predicted value. Then, by generating a correction coefficient based on this verification error and feeding it back to the water level prediction unit (130), the prediction accuracy of the virtual sensor (132) is dynamically corrected and improved.

[0045] The early warning unit (150) determines the flood risk based on the final predicted water level (actual sensor water level and corrected virtual sensor water level) received from the water level prediction unit (130). It compares the future predicted water level with the pre-set warning and alert criteria water level for each point, and generates an alert event if it is expected to reach the risk criteria. When an alert event occurs, it quickly sends a notification message via KakaoTalk, email, etc., to the registered administrator's terminal (20).

[0046] FIG. 2 is a flowchart illustrating the process of processing (a) missing values ​​and (b) outliers in the data processing unit (120) of the present invention.

[0047] The data processing unit (120) of the present invention receives raw measurement data from the data collection unit (110) and performs the role of ensuring the quality and reliability of the data before it is used as an input value for a prediction model. To this end, it includes missing value processing logic and outlier processing logic.

[0048] First, the handling of missing values ​​will be explained with reference to Fig. 2(a).

[0049] Missing values ​​are mainly caused by communication failures between the smart pole (10) and the system. The data processing unit (120) processes the missing values ​​in the following order.

[0050] First, the system attempts to periodically connect to a socket for each smart pole (10). If the connection fails, it attempts to reconnect up to a preset number of times (e.g., up to 3 times). If all attempts to reconnect fail, the data processing unit (120) determines that a communication failure has occurred in the smart pole (10), records the failure information in the database, and can send a message to the administrator notifying them that on-site response is required.

[0051] Subsequently, when the communication connection is successfully restored, the data processing unit (120) checks the last normal measurement data time of the corresponding smart pole (10) stored in the database. Then, it identifies the missing period between the current time and the last measurement time and sends a command to request the data for that period again from the data logger of the smart pole (10). When the data for the missing period is received from the smart pole (10), it updates the database to restore data continuity.

[0052] Next, outlier handling will be explained with reference to Fig. 2(b).

[0053] An outlier is an abnormal measurement value caused by a temporary error in the sensor or sudden noise in the external environment. The data processing unit (120) processes the outlier in the following order.

[0054] When the data processing unit (120) receives new measurement data from the smart pole (10), it checks whether the data is within the normal range. The determination of the normal range can be made by comparing the new measurement value with the previous measurement value and determining whether the amount of change of the new measurement value exceeds a preset threshold (e.g., ±50% of the previous value).

[0055] If a measured value is determined to be an outlier outside the normal range, the data processing unit (120) discards the outlier and generates a corrected value through linear interpolation, which uses the normal measured value before the occurrence of the outlier and the normal measured value afterward. Subsequently, the raw data (outlier) and the corrected data are stored separately in a database to enable tracking and management of the data.

[0056] If an anomaly occurs three or more times consecutively in a specific sensor, it is determined that there is a high probability of a defect in the sensor itself, and a sensor failure notification is sent to the administrator to induce an inspection.

[0057] FIG. 3 is a conceptual diagram illustrating how the virtual sensor (132) of the present invention applies a different prediction model depending on the topography of the river (confluence, bend, straight section).

[0058] The water level prediction unit (130) of the present invention uses refined measurement data received from the data processing unit (120) to predict not only the water level at the point where the smart pole (10) is installed but also the water level at the unmeasured point where there is no physical sensor. To this end, the water level prediction unit (130) includes an AI-based virtual sensor (132).

[0059] The virtual sensor (132) is a software-based prediction model without a physical entity, and receives time-series data (water level, rainfall, etc.) from multiple real sensors (smart poles, WAMIS observation stations, etc.) located around the point to be predicted, and estimates the water level of an unmeasured point in real time through an AI algorithm that learns the hydraulic correlation between the data.

[0060] The AI ​​model used in the present invention may be a Recurrent Neural Network (RNN)-based model that has strengths in time-series data processing, and specifically, a Long Short-Term Memory (LSTM) model may be used. The LSTM model excels at learning the long-term dependencies of past data, and can effectively model the complex influence of water level and rainfall data from previous time periods on the current water level. However, this is exemplary, and the present invention is not limited to a specific AI model.

[0061] The training of an AI model can be carried out in the following manner. First, a training dataset is constructed using historically collected data. For example, to predict the water level at a specific virtual sensor point, 10-minute interval water level and rainfall data from surrounding real sensors over the past 24 hours are used as input features, and the actual water level measured at that time (such as the value from a temporarily installed sensor during the initial model construction) is set as the target value. Meanwhile, if a dam is installed upstream of the prediction point, real-time dam discharge data can be additionally included as input variables. Through this, the model can learn the impact of rapid flow changes caused by artificial dam discharges on downstream water levels, in addition to natural inflow, thereby improving prediction accuracy. The AI ​​model is trained using multiple data pairs constructed in this way to find optimal weights that represent the non-linear relationship between the input data and the target value.

[0062] In particular, the present invention applies a virtual sensor model that uses a combination of input data optimized for each terrain type to improve prediction accuracy by considering the complex and diverse topographical characteristics of rivers.

[0063] The case of the straight section model is shown in Fig. 3(a).

[0064] As illustrated in FIG. 3(a), sections of a river that are relatively straight exhibit the characteristic that changes in flow rate upstream reach downstream with a time lag without significant variables. A virtual sensor (132a) located in such a section uses time-series water level and rainfall data from a real sensor (10a) located upstream of the prediction point as key inputs to increase prediction accuracy. Through this input data, the AI ​​model learns the time it takes for the upstream water level wave to propagate to the downstream virtual sensor point and the water level change pattern to predict the water level at the downstream point.

[0065] The case of the meandering section model is shown in Fig. 3(b).

[0066] As illustrated in FIG. 3(b), in the bending section of the river, the flow velocity differs between the inside and outside, and vortices are generated. Consequently, the influence of the upstream water level is not transmitted simply as in the straight section, and a complex backwater effect may occur. To reflect this complexity in the model, a virtual sensor (132b) located in the bend receives data from the upstream sensor (10b-1) at the prediction point as well as data from the downstream sensor (10b-2). By simultaneously learning the backwater effect resulting from changes in upstream inflow and downstream water level fluctuations, the AI ​​model predicts non-linear water level changes at the midpoint of the bend more precisely.

[0067] The case of the Confluence Section model is shown in Fig. 3(c).

[0068] As illustrated in FIG. 3(c), the water level at the point immediately downstream of the confluence where two tributaries merge into one is determined by the total sum of the inflow from each tributary. Accordingly, a virtual sensor (132c) located immediately downstream of the confluence receives both water level and rainfall data from all actual sensors (10c-1, 10c-2) located in each upstream tributary. The AI ​​model predicts the complex water level change after the two flows are combined by comprehensively considering the difference in inflow and arrival time of each tributary.

[0069] Meanwhile, the water level prediction unit (130) can selectively apply different AI models depending on the current water level of the river to further improve prediction accuracy. For example, when the water level of the upstream sensor is below a pre-set critical water level (e.g., 73.56 m, which is close to the flood warning level), a 'low water level prediction AI model' trained to accurately reflect the normal flow rate change pattern is used. On the other hand, when the upstream water level exceeds the critical water level and the possibility of a flood increases, the prediction is performed by switching to a 'high water level prediction AI model' trained to be specialized for extreme situation data, such as rapid flow rate changes and the possibility of flooding. By separately applying optimized models suited to the situation in this way, high prediction accuracy can be maintained across the entire range, from subtle changes during normal times to rapid changes during floods.

[0070] FIG. 4 is a flowchart illustrating the process of verifying and correcting prediction accuracy using a virtual sensor for verification in the virtual sensor verification unit (140) of the present invention.

[0071] A virtual sensor (132) based on an AI model may inevitably contain a certain level of prediction error. The virtual sensor verification unit (140) is a key component for minimizing this uncertainty and dynamically improving the prediction reliability of the entire system. It goes beyond simply evaluating the performance of the model and performs the role of correcting the predicted value by feeding the results back to the actual prediction process in real time.

[0072] The operation of the virtual sensor verification unit (140) includes the following steps.

[0073] First, the system selects an arbitrary point where actual smart poles (10) are already installed and actual water level data can be obtained as a 'verification point'. Then, a 'virtual sensor for verification' is created in software at the verification point. This virtual sensor for verification uses the same AI model and algorithm as the virtual sensor (132) at another unmeasured point to be predicted.

[0074] To predict the water level at a verification point, the virtual verification sensor receives measurement data from other surrounding real sensors, excluding the actual measurement data from the verification point itself. For example, if there are real sensors located both upstream and downstream of the verification point, the water level at the verification point is predicted using only the data from these two sensors. The value calculated in this way is called the 'verification predicted water level'.

[0075] At the same time, the 'actual water level' is measured at the actual smart pole (10) at the verification point. The virtual sensor verification unit (140) compares the verification predicted water level with the actual water level and calculates the difference value, i.e., the 'verification error'. This verification error is a quantitative indicator that shows how much the current AI model over- or under-predicts under specific conditions.

[0076] The calculated verification error is fed back to the water level prediction unit (130) and used to correct the initial predicted water level generated by the virtual sensor (132) at the actual unmeasured point.

[0077] At this time, to resolve the problem that the physical locations of the verification point and the prediction point are different, the virtual sensor verification unit (140) determines the 'similarity of hydraulic and topographical characteristics' between the two points. For example, the similarity is evaluated as higher the more similar the river width, slope, and degree of curvature of the two points are.

[0078] The virtual sensor verification unit (140) generates a 'correction coefficient' that applies weights to the verification error based on this similarity. If the characteristics of the two points are very similar, the correction coefficient is set close to 1 so that most of the verification error is reflected, and if the characteristics are different, it is set close to 0 so that little is reflected or not reflected at all.

[0079] Finally, the water level prediction unit (130) corrects the predicted water level of another unmeasured point through an operation such as ‘final predicted water level = initial predicted water level + (verification error Х correction coefficient)’. Through this process, the system of the present invention is equipped with a dynamic correction capability that diagnoses its own prediction performance in real time and reflects the results in other predictions, thereby securing even higher reliability.

[0080] FIG. 5 is a flowchart illustrating the process of predicting future water levels and generating an alarm in the early warning unit (150) of the present invention.

[0081] The early warning unit (150) comprehensively determines the flood risk based on the final predicted water level (water level of the actual sensor and the corrected virtual sensor) received from the water level prediction unit (130), and performs the role of quickly delivering a warning to the user when a dangerous situation occurs.

[0082] First, a customized standard water level is set for each point. The present invention does not apply a uniform standard but sets a customized standard water level that reflects the unique characteristics of each monitoring point.

[0083] The early warning unit (150) analyzes the river basic plan information (e.g., planned flood level, riverbed elevation) of each smart pole installation point and actual measurement data accumulated in the past. For example, the 'warning level' can be set to a water level corresponding to 60% of the height from the average low water level over the last 5 years to the planned flood level, and the 'warning level' can be set to a water level corresponding to 70% of the planned flood volume. The warning and warning standard water levels for each point set in this way are stored in a database and used as basic data for early warning judgment.

[0084] Next, future water level prediction and early warning determination are performed.

[0085] In the present invention, the early warning unit (150) does not provide a warning in a passive manner when the current water level reaches the reference water level, but rather provides a warning by predicting future risks and providing a preemptive warning.

[0086] The early warning unit (150) calculates the 'water level curve slope' by analyzing the current water level and the amount of change in water level over a past period of time (e.g., 60 minutes). This slope is an indicator of how quickly the current water level is rising or falling. Subsequently, based on the calculated slope, the future water level for up to 60 minutes is predicted in 10-minute intervals using linear extrapolation or similar methods. Then, it determines whether the predicted future water level exceeds the warning or alert threshold water level at the point. If it is determined that the predicted future water level exceeds the threshold water level, the early warning unit (150) immediately triggers an early warning event. This method allows for issuing an alert in advance before the actual water level reaches a dangerous level, thereby securing valuable time for evacuation and disaster prevention measures.

[0087] Finally, the alarm notification transmission step is performed.

[0088] When an early warning event occurs, the early warning unit (150) quickly sends a notification message to administrators and relevant personnel who are registered in the system in advance. The notification can be sent through multiple channels, such as KakaoTalk notifications and email.

[0089] The notification message includes not only the fact of a risk occurrence but also specific and intuitive information such as the observation point name, alert level (advisory / alert), time of the event, water level at that time, estimated time of reaching the reference level, and estimated water level. Additionally, it includes a web link within the message to access the monitoring system directly, allowing users to immediately check the detailed situation.

[0090] Hereinafter, specific embodiments are presented to further clarify the effects of the present invention.

[0091] Figure 6 is a graph comparing the timing of early warning issuance by the system of the present invention and the existing flood control center during an actual heavy rain situation, and the table in Figure 6 summarizes the data for the situation.

[0092] The top graph of Fig. 6 shows the monitoring status of the Flood Control Center, and the bottom graph shows the monitoring status of the smart pole system according to the present invention. Taking the actual heavy rain situation that occurred at the 'Seomjin River Geumgok Bridge' as an example, the system of the present invention analyzed the rate of water level rise at 16:20, predicted that it would reach the warning level (52.65m) 50 minutes later at 17:10, and preemptively issued a 'warning'. On the other hand, the Flood Control Center issued a warning at 17:20, when the actual water level was close to the warning criteria, indicating that the system of the present invention provided prediction and warning about one hour faster.

[0093] The effect is even more evident in the case of an 'alarm' situation. At 16:50, the system of the present invention predicted that the alarm level (53.37m) would be reached at 17:40, 50 minutes later, and issued an 'alarm'. In contrast, the Flood Control Center issued the alarm at 19:40, confirming that the present invention secured an overwhelming lead time of approximately 2 hours and 50 minutes. This is an example that clearly demonstrates how effective the present invention is in securing the golden time.

[0094] Figure 7 is an example of a flood warning notification message transmitted to a user terminal according to the present invention.

[0095] As illustrated in FIG. 7, when an alarm event occurs by the early warning unit (150), a notification message is immediately sent to the KakaoTalk, etc., of a manager registered in the system. The message clearly displays the key content of the alarm, such as "Gasan Bridge Flood Advisory Notification."

[0096] In addition, the message body specifically includes information essential for decision-making, such as ▲_Observation Point: Gasan Bridge (SP007), ▲_Date and Time of Occurrence: 2024-10-10 11:53 AM, ▲_Water Level: -0.38m, ▲_Estimated Time: 2024-10-10 12:13 PM, ▲_Estimated Water Level: -0.30m, and ▲_Flood Grade: Advisory. This allows administrators to immediately grasp the severity and urgency of the situation without accessing a separate system. A 'Go to Monitoring Page' link is included at the bottom, enabling rapid follow-up actions by allowing direct navigation to a web page where detailed status information can be viewed with a single click.

[0097] The system and method of the present invention as described above may be implemented as a program (or application) comprising an executable algorithm that can be executed on a computer. The program may be provided by being stored on a non-transitory computer-readable medium. Here, a non-transitory computer-readable medium refers to a medium that stores data semi-permanently and is readable by a device, rather than a medium that stores data for a short period of time, such as a register, cache, or memory. Specifically, the various applications or programs described above may be provided by being stored on a non-transitory computer-readable medium such as a CD, DVD, hard disk, Blu-ray disc, USB, memory card, ROM, etc.

[0098] The scope of protection of the present invention is not limited to the description and expression of the embodiments explicitly described above. Furthermore, it is added once again that the scope of protection of the present invention cannot be limited by obvious changes or substitutions in the technical field to which the present invention belongs.

Claims

Claim 1 A customized flood early warning system using smart pole and virtual sensor data, comprising: a data collection unit that receives measurement data including water level and rainfall from a plurality of smart poles; a data processing unit that corrects missing values ​​or outliers in the received measurement data; a water level prediction unit that predicts a river water level by combining the actual measured water level of the smart pole and the predicted water level of the virtual sensor, wherein the water level prediction unit generates the predicted water level by inputting time-series data collected from a plurality of smart poles located around the prediction point into an AI model, and generates the predicted water level by inputting time-series data collected from a plurality of smart poles located around the prediction point into an AI model, wherein if a dam is located upstream of the prediction point, the discharge data of the dam is additionally included in the time-series data and input into the AI ​​model. Claim 2 A customized flood warning system using smart pole and virtual sensor data, comprising: a data collection unit that receives measurement data including water level and rainfall from a plurality of smart poles; a data processing unit that corrects missing values ​​or outliers in the received measurement data; a water level prediction unit that predicts a river water level by combining the actual measured water level of the smart pole and the predicted water level of the virtual sensor, wherein the water level prediction unit generates the predicted water level by inputting time-series data collected from a plurality of smart poles located around the prediction point into an AI model, and generates the predicted water level by inputting time-series data collected from a plurality of smart poles located around the prediction point into an AI model, wherein the combination of surrounding smart poles input into the AI ​​model is varied depending on whether the topographical characteristics of the prediction point correspond to a confluence, a curved section, or a straight section. Claim 3 A data collection unit that receives measurement data including water level and rainfall from a plurality of smart poles; a data processing unit that corrects missing values ​​or outliers in the received measurement data; and a water level prediction unit that predicts a river water level by combining the actual measured water level of the smart poles and the predicted water level of the virtual sensors, including a virtual sensor that generates a predicted water level for one or more prediction points where physical sensors are not installed, using the corrected measurement data. A customized flood early warning system using smart poles and virtual sensor data, comprising: an early warning unit that generates an alarm when it is determined that the predicted river water level will exceed a preset standard water level; wherein the water level prediction unit generates the predicted water level by inputting time-series data collected from a plurality of smart poles located around the prediction point into an AI model, and generates the predicted water level by inputting time-series data collected from a plurality of smart poles located around the prediction point into an AI model, wherein the prediction level is generated by using a high-level prediction AI model when the current water level of the smart poles around the prediction point is above a preset threshold water level and by using a low-level prediction AI model when it is below the threshold water level. Claim 4 A flood warning system according to any one of claims 1 to 3, wherein the data processing unit, when a missing value occurs in the measurement data due to a communication failure, requests the data of the missing section from the smart pole again at the time of normalization of communication and updates it, and when the measurement data deviates from a preset range and is determined to be an outlier, corrects the outlier using linear interpolation. Claim 5 A flood warning and alert system characterized by further comprising: a verification virtual sensor created separately from the virtual sensor at the prediction point at a verification point where an actual sensor of any one of the smart poles is installed, in any one of claims 1 to 3; and a virtual sensor verification unit that generates a verification prediction water level, which is a predicted value of the verification virtual sensor, using measurement data of other smart poles around the verification point, and calculates a verification error by comparing the verification prediction water level with the actual water level measured by the actual sensor. Claim 6 A flood warning system according to claim 5, wherein the water level prediction unit corrects the predicted water level generated by the virtual sensor at the prediction point using the verification error calculated by the virtual sensor verification unit. Claim 7 A flood warning system according to claim 6, wherein the virtual sensor verification unit determines the similarity of hydraulic or topographical characteristics between the prediction point and the verification point, generates a correction coefficient by applying a weight to the verification error based on the similarity, and the water level prediction unit corrects the predicted water level using the correction coefficient. Claim 8 A flood warning system characterized in that, in any one of paragraphs 1 to 3, the warning unit sets the reference water levels for 'advisory' and 'warning' for each point based on past measurement data and planned flood level information of each smart pole point. Claim 9 A flood warning system according to claim 8, wherein the warning unit calculates a future expected water level based on the slope of a water level curve representing the amount of change in water level over a past certain period of time (60 minutes) and generates a warning by comparing the expected water level with the reference water level. Claim 10 A flood warning system according to any one of claims 1 to 3, wherein the warning unit transmits a notification message including the location of occurrence, time of occurrence, current water level, expected time of arrival, expected water level of arrival, and warning grade to a terminal of a pre-registered manager when a warning occurs. Claim 11 delete Claim 12 delete

Images