Real-time road heat map-based urban heat island mitigation system, real-time heat risk prediction method for urban heat island mitigation, and heat island mitigation response method using real-time road heat map

Abstract

One aspect of the present invention discloses an urban heat island mitigation system based on a real-time road heat map. The system comprises: a data collection unit that collects in real time multi-type data including at least two of road-specific temperature, traffic volume, weather information, and population density; a data integration unit that processes the collected data in a time and spatial manner; an AI prediction engine that predicts heat risk for each road section based on the integrated processed data; a visualization unit that visualizes the predicted heat risk as a real-time heat map based on a Geographic Information System (GIS); and an alert unit that provides a heatwave alert to at least one mobile terminal based on the heat map information and provides heat island mitigation infrastructure operation information to an administrative agency server.

Description

Technology Field

[0001] The present invention relates to a heat island mitigation system, and more specifically, to a system for mitigating urban heat islands using real-time heat maps.

[0002] This invention was derived as a result of the AI+X Regional Industrial Technology Development Support Project of the Artificial Intelligence Industry Convergence Group in 2025. Background Technology

[0003] Due to the rapid urbanization of modern society, the urban heat island phenomenon is emerging as a serious environmental issue worldwide. The urban heat island is a phenomenon in which temperatures in urban areas are higher than in surrounding suburban areas, caused by concrete and asphalt-covered roads and buildings absorbing and storing solar heat, which is then released even at night. This phenomenon is particularly intensified during the summer, causing a significant rise in perceived temperatures. This poses a direct threat to citizens' health and is a major cause of the rapid increase in cooling energy consumption.

[0004] Existing urban heat island mitigation technologies have primarily relied on physical methods. Representative examples include vertical greening systems that cultivate vegetation on building exteriors or rooftops, cool roof technology that replaces road paving materials with those having high heat reflectivity, and methods to expand green spaces within the city. Additionally, some regions are implementing response measures during summer heatwaves, such as using water trucks to spray water on roads or installing temporary shade structures.

[0005] However, these existing methods have several limitations. First, due to a lack of spatial precision, they often treat the entire city uniformly or are limited to specific areas, making it difficult to accurately identify sections where the urban heat island effect is actually severe and to respond intensively. Second, due to low temporal precision, they fail to reflect real-time changing factors such as weather changes or traffic conditions. Third, because of a reactive approach, countermeasures are mostly devised only after the urban heat island effect has already occurred.

[0006] In particular, existing weather forecasting systems are limited to providing average temperature information across broad regions, making it difficult to accurately predict or reflect microscopic changes in the thermal environment that vary by road or section. Furthermore, there is still a lack of integrated analysis and prediction systems that comprehensively consider various factors influencing the urban heat island effect, such as traffic volume, road pavement conditions, building density, and population distribution. The problem to be solved

[0007] The objective of one embodiment of the present invention for solving the aforementioned problems is to construct a system capable of precise thermal environment analysis by collecting and integrating various types of public data, such as road-specific temperatures, traffic volume, weather information, and population density, in real time.

[0008] The objective of another aspect of the present invention is to provide an advanced prediction engine capable of predicting heat risk for each road section in real time by utilizing an artificial intelligence-based spatiotemporal deep learning model. Beyond simple temperature measurement or weather forecasting, the aim is to enable proactive response through precise heat risk prediction that comprehensively considers the physical characteristics of the road, traffic patterns, and changes over time.

[0009] The objective of another aspect of the present invention is to provide an interface that can be easily utilized by both citizens and administrative agencies by visualizing predicted heat risk information into an intuitive, real-time heat map based on a Geographic Information System. Furthermore, based on the generated heat map information, the invention aims to establish an integrated response system that provides real-time heatwave alerts to citizens via mobile devices and offers administrative agencies information to efficiently operate heat island mitigation infrastructure, such as water trucks, shade structures, and electronic display boards. Through this, the invention seeks to simultaneously support voluntary responses at the individual level and systematic responses at the public level.

[0010] Finally, the goal is to design a scalable platform structure capable of integrally handling various climate disaster elements, such as not only heatwaves but also heavy rain, cold waves, and fine dust, thereby laying the foundation for developing into a comprehensive urban climate disaster response system. means of solving the problem

[0011] A real-time road heat map-based urban heat island mitigation system according to one embodiment of the present invention for achieving the above-mentioned purpose may include: a data collection unit that collects in real time multi-type data including at least two of road-specific temperature, traffic volume, weather information, and population density; a data integration unit that processes the collected data in an integrated manner in time and space units; an AI prediction engine that predicts heat risk for each road section based on the integrated processed data; a visualization unit that visualizes the predicted heat risk as a real-time heat map based on a Geographic Information System (GIS); and an alert unit that provides a heat wave alert to at least one mobile terminal based on the heat map information and provides heat island mitigation infrastructure operation information to an administrative agency server.

[0012] The above AI prediction engine can predict thermal risk by road using a spatiotemporal graph convolution network (ST-GCN) or a convolutional long short-term memory (ConvLSTM) model.

[0013] The above data collection unit can collect data from at least one of the Public Data Portal, the Korea Meteorological Administration, and local government big data platforms through API calls or FTP integration and schedule it regularly.

[0014] The above data integration unit can integrate data by performing location-based integration and time-series synchronization using time windowing.

[0015] The above system may have a modular structure that can be expanded into a climate disaster response platform by integrating multiple disaster elements including at least two of heatwaves, heavy rain, cold waves, and fine dust.

[0016] The above AI prediction engine can calculate a final risk score by integrating at least two of the predicted heat index, perceived temperature, and traffic density.

[0017] The above data integration unit can grid and integrate road section data in units of 50 to 100 meters based on GIS standards.

[0018] The above AI prediction engine includes a ConvLSTM model that takes time-series data in spatiotemporal grid units as input, and an ST-GCN model that combines a normalized adjacency matrix and a Temporal CNN by modeling the road network as a node-edge graph, and can calculate a final heat risk score by integrating the outputs of the two models through a Multi-Head MLP.

[0019] A method for predicting real-time heat risk for mitigating urban heat islands according to another aspect of the present invention for achieving the above-mentioned purpose may include the steps of: collecting characteristic data including at least two of road pavement material, cumulative solar radiation, reflectance, temperature fluctuation coefficient, and intersection congestion; converting the characteristic data into a heat generation index; predicting heat risk for each road section using a spatiotemporal deep learning model; and visualizing the predicted heat risk as a real-time heat map.

[0020] The step of converting to the above heat generation indicator may include a step of converting characteristics using Principal Component Analysis (PCA).

[0021] The above method may further include a step of automatically detecting and correcting abnormal patterns, including sudden surges in vehicle traffic volume or sudden changes in temperature, using an Isolation Forest or IQR (Interquartile Range) rule.

[0022] A method for responding to heat island mitigation using a real-time road heat map according to another aspect of the present invention for achieving the above-mentioned purpose may include the steps of: identifying a high-temperature risk section from a real-time heat map; calculating a heat risk level for the identified high-temperature risk section; transmitting a heat wave warning to a mobile terminal according to the heat risk level; and controlling to support the operation of heat island mitigation infrastructure, including at least one of a water truck, a sunshade, and an electronic display board of an administrative agency, according to the heat risk level.

[0023] The above method may further include the step of visualizing the heat risk level based on a web-based interface.

[0024] According to another aspect of the present invention for achieving the above-mentioned purpose, an urban heat island mitigation program executed by a computer may perform the steps of: collecting various types of public data in real time, including at least two of road-specific temperatures, traffic volume, weather information, and population density; processing the collected data in an integrated manner in time and space units; predicting heat risk for each road section using a ConvLSTM model and an ST-GCN model; visualizing the predicted heat risk as a real-time heat map based on a geographic information system; and providing heatwave alerts to mobile terminals based on the heat map information and providing heat island mitigation infrastructure operation information to administrative agency servers.

[0025] By integrating and analyzing the sprinkler area per operation, the quantity and deployment status of each sprinkler, and the location information of major water sources, the sprinkler routes within the city are optimized, thereby enabling efficient heat island mitigation.

[0026] Furthermore, the present invention is designed to be applicable not only to heatwaves but also to snow removal support during heavy snowfall, road cleaning during yellow dust, and air quality improvement, and can be expanded into an integrated platform capable of actively responding to various climate disaster situations. Effects of the invention

[0027] The real-time road heat map-based urban heat island mitigation system according to the present invention can prevent and minimize damage caused by the heat island phenomenon by implementing a system capable of prior prediction and response, moving away from the existing reactive approach. Through precise AI-based prediction, it is possible to identify in advance the times and regions where heat risk is expected to increase, thereby enabling a preemptive response.

[0028] In addition, spatial precision can be significantly improved by providing detailed heat risk information for each road section. This allows limited resources to be prioritized where they are most needed, thereby increasing administrative efficiency and saving the budget. For example, instead of deploying water trucks evenly across all roads, they can be concentrated in sections with high heat risk for more effective operation.

[0029] Next, the integrated utilization of diverse public data can present best practices for data use and increase the potential for expansion to other cities or regions. Through standardized API integration methods and a modular system structure, universality can be secured, enabling easy adoption and operation in cities of various sizes and characteristics.

[0030] Finally, in preparation for the future increase in extreme weather events caused by climate change, it can be expanded into a platform capable of integrally managing various disaster elements such as heavy rain, cold waves, and fine dust. Through this, it can contribute to establishing a comprehensive urban climate safety network beyond single disaster responses and serve as a core foundational technology for the realization of smart cities. Brief explanation of the drawing

[0031] FIG. 1 is a block diagram showing the overall configuration of a real-time road heat map-based urban heat island mitigation system according to one embodiment of the present invention. FIG. 2 is a flowchart showing the entire process from the collection to integrated processing of various types of public data according to one embodiment of the present invention. FIG. 3 is a block diagram showing in detail the internal configuration of an AI prediction engine according to one embodiment of the present invention. FIG. 4 is a user interface screen showing a real-time heat map visualization result based on a geographic information system according to an embodiment of the present invention. FIG. 5 is a diagram showing the configuration of a mobile-based heatwave notification service provided to citizens according to an embodiment of the present invention. FIG. 6 is a diagram showing the configuration of a heat island mitigation infrastructure operation support system provided to an administrative agency according to one embodiment of the present invention. FIG. 7 is a flowchart showing the process of the entire system operating in real time according to one embodiment of the present invention. Specific details for implementing the invention

[0032] The present invention is capable of various modifications and may have various embodiments, and specific embodiments are illustrated in the drawings and described in detail.

[0033] However, this is not intended to limit the invention to specific embodiments, and it should be understood that it includes all modifications, equivalents, and substitutions that fall within the spirit and scope of the invention.

[0034] Terms such as "first," "second," etc., may be used to describe various components, but said components should not be limited by said terms. Such terms are used solely for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be named the second component, and similarly, the second component may be named the first component. The term "and / or" includes a combination of a plurality of related described items or any of a plurality of related described items.

[0035] When it is stated that one component is "connected" or "connected" to another component, it should be understood that while it may be directly connected or connected to that other component, there may also be other components in between. On the other hand, when it is stated that one component is "directly connected" or "directly connected" to another component, it should be understood that there are no other components in between.

[0036] The terms used in this application are used merely to describe specific embodiments and are not intended to limit the invention. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this application, terms such as "comprising" or "having" are intended to specify the presence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0037] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the present invention pertains. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this application.

[0038] Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the attached drawings. In order to facilitate an overall understanding of the present invention, the same reference numerals are used for identical components in the drawings, and redundant descriptions of identical components are omitted.

[0040] FIG. 1 is a block diagram showing the overall configuration of a real-time road heat map-based urban heat island mitigation system according to an embodiment of the present invention. As shown in FIG. 1, the real-time road heat map-based urban heat island mitigation system (100) according to an embodiment of the present invention may include a data collection unit (110), a data integration unit (120), an AI prediction engine (130), a visualization unit (140), and a notification unit (150). The real-time road heat map-based urban heat island mitigation system (100) may be implemented as a computing device. Hereinafter, this may be referred to as a server, a device, etc.

[0041] Referring to FIG. 1, the data collection unit (110) is responsible for collecting various types of data required for urban thermal environment analysis in real time. Specifically, it collects weather information such as real-time temperature, humidity, UV index, and wind speed at 5-minute intervals via the Korea Meteorological Administration API, and obtains air quality information such as PM10, PM2.5, and ozone concentration in real time via the Ministry of Environment Air Quality API. In addition, it collects traffic volume, congestion, and speed information by road from the traffic DB, and collects road surface temperature and real-time wind chill temperature in real time via IoT temperature sensors using MQTT or LoRa communication protocols. Video data from CCTV and traffic cameras is streamed frame by frame and summarized. These data are collected from the public data server (10) via API calls, and during the data collection process, data collection logs are stored and API keys are systematically managed for security and access control. In particular, for the protection of personal information, GPS coordinates are collected after being de-identified through grid processing.

[0042] The data integration unit (120) is responsible for the function of integrating and processing various types of heterogeneous data collected from the data collection unit (110) in time and space units. For processing missing values, Kriging and IDW (Inverse Distance Weighting) methods, which are spatiotemporal interpolation methods, are applied in addition to linear interpolation, and in cases of severe missing values, they are replaced with the recent average value. For outlier detection, an Autoencoder-based deep learning outlier detection method is applied in addition to the existing Z-Score and IQR-based outlier removal methods. In the data standardization process, Feature Scaling is performed through Min-Max normalization, and data is gridded and grouped into units of 50 to 100 meters per road section based on GIS. Overall data pipeline management manages the schedule of the ETL process through workflow management tools such as Airflow or Prefect.

[0043] The AI ​​prediction engine (130) is a core component that predicts the heat risk for each road section based on the integrated data provided by the data integration unit (120). The AI ​​prediction engine (130) consists of five core components: a feature engineering module (132), a spatiotemporal deep learning model (134), an outlier correction module (136), a heat risk calculation module (138), and a Multi-Head MLP module (139). The input data is time-series data in spatiotemporal grid units, configured in the form of time T, W×H grid, and Multi-Feature. The spatiotemporal deep learning model (134) includes two core models. The first is a ConvLSTM model that combines a 2D Convolution Layer and an LSTM Layer to simultaneously process spatiotemporal patterns of image data. The second is an ST-GCN model that models the road network as a node-edge graph, learns the spatial structure with GCN, and learns the time flow with Temporal CNN. The Multi-Head MLP module (139) is responsible for calculating the final risk score by integrating the predicted heat index, perceived temperature, traffic density, etc. The final output generates a heat risk heatmap for each road with values ​​between 0 and 1, and provides visualization linked with GIS.

[0044] The visualization unit (140) performs the function of converting and displaying heat risk information calculated by the AI ​​prediction engine (130) into an intuitive and easy-to-understand real-time heat map based on a geographic information system. It outputs a real-time heat map using GIS APIs such as Mapbox GL JS and Leaflet.js, and displays the heat risk for each road in the form of Low, Medium, High classes or a Score between 0 and 1. It provides an interactive interface that allows the user to check detailed information by selecting a desired time period and region, and enables intuitive situational awareness through heat risk information distinguished by color.

[0045] The notification unit (150) is responsible for providing appropriate notifications and response information to citizens and administrative agencies based on heat map information generated by the visualization unit (140). It provides personalized heat wave warnings and safety information to citizens via mobile terminals (20), and provides decision support information for the efficient operation of heat island mitigation infrastructure by linking with water truck and shelter placement algorithms via the administrative agency server (30). The notification unit (150) implements an intelligent notification system that generates customized notifications considering the user's location, time, personal characteristics, etc., and adjusts the form and frequency of notifications according to the level of urgency.

[0046] FIG. 2 is a flowchart showing the entire process from the collection to integrated processing of various types of public data according to one embodiment of the present invention.

[0047] Referring to Fig. 2, first, the device can collect five major data types in parallel. Road-specific temperature data is collected at 5-minute intervals via the Korea Meteorological Administration API, providing information such as real-time temperature, humidity, UV index, and wind speed, while air quality information such as PM10, PM2.5, and ozone concentration is obtained in real time via the Ministry of Environment Air Quality API. Traffic volume data is collected from the traffic database, providing information on traffic volume, congestion, and speed for each road, and road surface temperature and real-time wind chill temperature are collected in real time via IoT temperature sensors using MQTT or LoRa communication protocols. CCTV and traffic camera footage is streamed frame by frame and summarized.

[0048] The collected raw data undergoes advanced data preprocessing to be transformed into a format suitable for analysis. For missing value handling, spatiotemporal interpolation methods such as Kriging and IDW are applied in addition to linear interpolation to perform precise missing value estimation considering spatial continuity. In cases of severe missing values, they are replaced with the most recent average value to ensure data continuity. For outlier detection, complex patterns of anomalies can be effectively detected by applying an Autoencoder-based deep learning outlier detection method in addition to existing Z-Score and IQR criteria. In the data cleaning stage, a reliable dataset is constructed through noise removal, duplicate data elimination, and data quality verification.

[0049] Next, the device performs feature scaling through Min-Max normalization to transform data of different scales into a unified range. During the gridding process, spatial consistency is ensured by grouping data into units of 50 to 100 meters per road section based on GIS standards. Data pipeline management automates and monitors the schedule of the ETL process through workflow management tools such as Airflow or Prefect. In the spatiotemporal synchronization stage, data with different time periods and spatial resolutions are normalized into a unified spatiotemporal grid system to generate an integrated dataset optimized for AI models.

[0050] FIG. 3 is a block diagram showing in detail the internal configuration of an AI prediction engine according to an embodiment of the present invention. As shown in FIG. 3, an AI prediction engine (130) according to an embodiment of the present invention may include a feature engineering module (132), a spatiotemporal deep learning model (134), an outlier correction module (136), and a heat risk calculation module (138).

[0051] The feature engineering module (132) is responsible for generating derived features useful for predicting thermal risk from integrated raw data. It extracts and transforms features such as road-specific pavement material information, cumulative solar radiation data, reflectance, temperature variation coefficient, and intersection congestion. These features can be optimized through dimensionality reduction techniques, such as principal component analysis or independent component analysis, to increase the learning efficiency of the model.

[0052] The spatiotemporal deep learning model (134) is a core algorithm that predicts thermal risk for each road section by receiving features generated in the feature engineering module (132) as input. It includes two main model structures. The ConvLSTM model combines a 2D Convolution Layer and an LSTM Layer to simultaneously process spatiotemporal patterns of image data. The core formula of the ConvLSTM is as follows.

[0053] i in the Input Gate t = σ(W xi * X t + W hi * H t- ₁ + b i ), at the Forget Gate, f t = σ(W xf * X t + W hf * H t- ₁ + b f ), at the Output Gate o t = σ(W xo * X t + W ho * H t- ₁ + b o ), in Cell Candidate ￠ t = tanh(W x c * X t + W hc * H t- ₁ + b c ), in Cell State C t = f t ˚ C t- ₁ + i t˚ ￠ t , H in the Hidden State t = o t ˚ tanh(C t ) is calculated. Here, * represents the Convolution operation, and ˚ represents the element-wise product.

[0054] The ST-GCN model models road networks as node-edge graphs, learns spatial structures using GCN, and learns temporal flow using Temporal CNN. The core formula of ST-GCN is H^(l+1) = σ(Σ k=0 KA k H^(l) W k It is expressed as ), where A k represents the road network graph as a normalized adjacency matrix, H^(l) is the l-th layer feature, and W k represents the learning parameter.

[0055] The outlier correction module (136) performs the function of detecting and correcting abnormal values ​​that may appear in the prediction results. It identifies prediction errors that may occur due to exceptional situations, such as a sudden increase in vehicle traffic volume or a sudden change in temperature, and detects outliers by utilizing statistical methods such as isolated forests or IQR rules. For the detected outliers, it adjusts the traffic volume coefficient using real-time traffic sensor data, reflects actual road surface temperature measurements using CCTV thermal image data, and readjusts the model weights to update the prediction values ​​in real time.

[0056] The heat risk calculation module (138) is responsible for calculating the final heat risk for each road section and assigning a grade by combining the results of the spatiotemporal deep learning model (134) and the outlier correction module (136). It calculates the perceived temperature by combining the predicted temperature value and various environmental factors, and determines the heat risk grade based on this.

[0057] Additionally, the Multi-Head MLP module (not shown) calculates a final risk score by integrating various indicators such as the predicted heat index, perceived temperature, and traffic density. This module calculates risk levels from different aspects in parallel through multiple MLP heads and generates a comprehensive risk index by performing a weighted sum of these values. The final output creates a heat risk heatmap for each road with values ​​ranging from 0 to 1, providing visualization linked with GIS.

[0058] FIG. 4 is a user interface screen showing a real-time heat map visualization result based on a geographic information system according to an embodiment of the present invention.

[0059] Referring to Fig. 4, a time zone selection menu is located at the top of the user interface screen along with the title of the heat map visualization system, so that the user can check heat risk information at a desired time. The time zone selection function provides forecast information from the current time to the next 24 hours by time zone, and can also refer to past data, which can be used to analyze patterns of heat environment change.

[0060] Located in the central area of ​​the user interface screen is a main heat map where heat risk information is displayed in color over a map showing the actual road network. Each road section is marked in a different color according to the heat risk level of the area, and the heat map is displayed through GIS APIs such as Mapbox GL JS or Leaflet.js that are updated in real time. Heat risk is displayed in the form of Low, Medium, High classes or a score between 0 and 1, allowing for an intuitive assessment of the risk level. When a user clicks on a specific section or hovers the mouse over it, a detailed information pop-up appears, allowing them to check specific details such as the exact temperature, wind chill, risk grade, and estimated duration for that section.

[0061] A heat risk color legend is located at the bottom left of the user interface screen, clearly conveying the meaning of the colors displayed on the map. The legend is divided into four levels: Low, Medium, High, and Danger, with the corresponding color and temperature range displayed for each level. Additionally, recommended action guidelines for each risk grade are provided to help users take appropriate action. The legend area also displays the update time of the currently shown data and prediction reliability information, allowing users to verify the reliability of the data.

[0062] FIG. 5 is a diagram showing the configuration of a mobile-based heatwave notification service provided to citizens according to an embodiment of the present invention.

[0063] Referring to Fig. 5, the mobile notification service for citizens can be configured and visualized in the form of a smartphone screen. At the top of the smartphone screen, the app title "Heat Island Notification Service" is displayed, along with basic smartphone information such as the current time and battery status.

[0064] A push notification area is located at the very top of the smartphone screen to immediately deliver urgent heatwave alerts or important safety information. This area concisely displays key information, such as the type of alert currently issued, the heat risk level at the user's location, and the estimated wind chill temperature. Notification messages are generated based on risk scores calculated by an AI prediction engine, and are designed to allow users to intuitively grasp the severity of the situation by displaying different colors and icons according to the risk level.

[0065] The current location-based heat risk display area shows real-time thermal environment information for the user's current location, identified via GPS. It includes precise location coordinates, the heat risk grade of the point, the current temperature and wind chill, and a comprehensive safety index calculated by the Multi-Head MLP module. Additionally, forecasts of heat risk changes over the next few hours are provided to help users plan their outings.

[0066] The safe route guidance area displays information recommending safe routes with low heat risk from among various paths to the user's destination. It suggests detours that prioritize shaded paths, underpasses, and indoor passages, and displays the additional travel time and the expected reduction in perceived temperature. Route recommendations are dynamically updated by comprehensively considering real-time heat map data and traffic conditions.

[0067] The nearby cooling center information area displays the locations and availability of free cooling facilities, shade structures, and rest areas near the user's location. It includes information such as the distance to each facility, walking time, type of facility, operating hours, and current congestion levels. In particular, when a heat wave advisory is issued, location information for temporary cooling centers or free water fountains operated by public institutions is also provided to help citizens easily access them.

[0068] FIG. 6 is a diagram showing the configuration of a heat island mitigation infrastructure operation support system provided to an administrative agency according to one embodiment of the present invention.

[0069] Referring to Fig. 6, the dashboard and infrastructure linkage configuration for administrative agencies can be configured in the form of a comprehensive control screen. At the top of the administrative agency dashboard screen, a real-time monitoring dashboard is located to allow for an overview of the thermal environment conditions of the entire city. In this area, key indicators such as the current average temperature of the entire city, the number of high-risk zones identified by the AI ​​prediction engine, and the number of active alerts currently in effect are displayed in the form of numbers and graphs. Additionally, the change from the previous day and the predicted trend are also displayed, making it easy to identify the direction of change in the situation.

[0070] Located on the left side of the administrative agency dashboard is a water tank deployment status management area, allowing for the real-time monitoring of the locations and operational status of water tankers operating within the city. Information such as the number of currently active tankers, the number of high-risk sections requiring priority deployment, and the scheduled time for the next patrol are displayed. The tanker deployment algorithm automatically calculates and proposes optimized routes based on thermal risk information derived from an AI prediction engine. Detailed information for each tanker, including current location, fuel status, water tank capacity, and work progress, can also be checked, and emergency deployment orders to specific sections can be issued if necessary.

[0071] Located on the right side of the center of the administrative agency dashboard is a shade installation priority management area, allowing for the management of the installation status and plans for temporary or sunshades. The number and locations of already installed shades, the number of sites awaiting installation, and the number of high-risk sites requiring urgent installation are displayed. Based on the results of heat risk analysis, the priority of shade installation is automatically calculated and presented, and the current location and progress of the installation team can also be checked in real time.

[0072] At the bottom of the administrative agency dashboard screen, an area for managing electronic display alert messages is located, allowing for the integrated management of public announcements broadcast via electronic displays installed within the city. The content of currently transmitted alert messages is displayed; messages can be automatically generated based on the current heat risk level or edited directly by the administrator. Status information is displayed, including the total number of connected electronic displays within the city, the number of displays currently in normal operation, and the operation rate. In the event of a risk situation confined to a specific area, customized messages can be transmitted only to the displays in that region, and the display method and duration can be adjusted according to the urgency of the message.

[0073] FIG. 7 is a flowchart showing the process of the entire system operating in real time according to one embodiment of the present invention.

[0074] Referring to Fig. 7, the system completes preparation for operation through an initialization process during the startup phase and then provides real-time services through continuous cyclic operation. First, the device automatically collects various types of public data according to a pre-set schedule or in real-time (S710). The device collects weather information via a pull method at 5-minute intervals through the Korea Meteorological Administration API and collects air quality information in real-time through the Ministry of Environment Air Quality API. Data is collected from IoT temperature sensors via a push method using MQTT or LoRa protocols, and CCTV and traffic camera footage is streamed frame by frame and summarized. During the data collection process, data collection logs are stored and API keys are managed for security and access control, and GPS coordinates are gridded and anonymized for the protection of personal information.

[0075] Next, the device performs data integration processing, including preprocessing tasks such as format standardization, coordinate system unification, and time synchronization of the collected multi-type data (S720). To handle missing values, linear interpolation and spatiotemporal interpolation methods such as Kriging and IDW are applied, and in cases of severe missing values, they are replaced with the recent average value. For outlier detection, an Autoencoder-based deep learning outlier detection method is applied along with Z-Score and IQR criteria methods. Min-Max normalization is performed for feature scaling, and the data is gridded into units of 50 to 100 meters per road section based on GIS. The entire data pipeline manages the ETL schedule through Airflow or Prefect.

[0076] Next, the device calculates the heat risk for each road section using an AI model based on integrated data (S730). Derived features are generated in the feature engineering module, and predictions are performed using ConvLSTM and ST-GCN models in the spatiotemporal deep learning model. Abnormal patterns, such as sudden increases in vehicle traffic volume or sudden changes in temperature, are detected and corrected in the outlier correction module. The Multi-Head MLP module calculates the final risk score by integrating the predicted heat index, wind chill temperature, and traffic density. A confidence index is also calculated for the prediction results to quantitatively present the degree of uncertainty of the prediction.

[0077] The system analyzes the predicted heat risk information to determine whether a high-risk zone exists (S740). In this determination process, the relative risk is evaluated by considering the average risk level of the same period in the past along with a pre-set threshold. If no high-risk zone exists, the system returns to a standby state and waits until the next collection cycle. On the other hand, if a high-risk zone is detected, it proceeds immediately to the subsequent response step.

[0078] If the device determines that a high-risk section exists, it generates map data to visually represent the predicted heat risk information (S750). In this determination process, the relative risk is evaluated by considering the average risk level of the same period in the past along with a pre-set threshold. If no high-risk section exists, the system returns to a standby state and waits until the next collection cycle. On the other hand, if a high-risk section is detected, it proceeds immediately to a follow-up response step. Meanwhile, in one example, the device can generate map data to visually represent the predicted heat risk information. It outputs a real-time heat map using GIS APIs such as Mapbox GL JS or Leaflet.js, and displays the heat risk for each road in the form of low, medium, high classes or a score between 0 and 1. The generated heat map is provided in real-time through a web service and a mobile app, and is updated immediately upon a user's request.

[0079] Finally, the device transmits customized heatwave alerts and safety information to citizens' mobile devices (S760). In this process, personalized alerts are generated by considering the user's current location, movement patterns, and personal settings. The content and format of the alerts are differentiated according to the level of risk, and in urgent situations, mandatory alerts are used to induce an immediate response from the user. Additionally, the results of the alert transmission are tracked to analyze the delivery rate and user response, and based on this, the effectiveness of the alert system is continuously improved.

[0080] In addition, along with transmission to mobile devices, the device provides decision support information to administrative agency servers, etc., to enable the efficient operation of the administrative agency's heat island mitigation infrastructure (S770). At this time, the device calculates the optimal deployment route for water trucks, determines the priority for installing shade canopies, and automatically generates citizen guidance messages via electronic display boards. This information is updated in real time to support immediate response to changes in on-site conditions. Furthermore, it monitors the operational effectiveness of each infrastructure in real time to evaluate the effectiveness relative to input resources and suggests more efficient operational plans.

[0081] Once all steps are completed, the system returns to a standby state and waits for the next data collection cycle. Through this cyclical operation, the system continuously monitors the city's thermal environment 24 hours a day and provides necessary information and services to citizens and administrative agencies in real time. The overall performance and stability of the system are continuously monitored through a separate monitoring module, and in the event of an abnormal situation, a recovery procedure is automatically executed to minimize service interruption.

[0083] Although the invention has been described with reference to the drawings and embodiments, this does not imply that the scope of protection of the present invention is limited by the drawings or embodiments, and those skilled in the art will understand that various modifications and changes can be made to the present invention without departing from the spirit and scope of the invention as described in the following claims.

Claims

Claim 1 A real-time road heat map-based urban heat island mitigation system comprises: a data collection unit that collects in real time multi-type data including at least two of road-specific temperature, traffic volume, weather information, and population density; a data integration unit that processes the collected data in an integrated manner in time and space units; an AI prediction engine that predicts heat risk for each road section based on the integrated processed data; a visualization unit that visualizes the predicted heat risk as a real-time heat map based on a Geographic Information System (GIS); and an alert unit that provides a heat wave alert to at least one mobile terminal based on the heat map information and provides heat island mitigation infrastructure operation information to an administrative agency server, wherein the AI ​​prediction engine predicts heat risk for each road using at least one of a spatiotemporal graph convolution network (ST-GCN) and a convolutional long short-term memory (ConvLSTM) model. Claim 2 A real-time road heat map-based urban heat island mitigation system according to claim 1, wherein the AI ​​prediction engine includes a ConvLSTM model that takes time-series data in spatiotemporal grid units as input, and an ST-GCN model that combines a normalized adjacency matrix and a Temporal CNN by modeling a road network as a node-edge graph, and integrates the outputs of the two models through a Multi-Head MLP to calculate a final risk score. Claim 3 In claim 1, the data collection unit collects data from at least one of the Public Data Portal, the Korea Meteorological Administration, and a local government big data platform through API calls or FTP integration and schedules it regularly, a real-time road heat map-based urban heat island mitigation system. Claim 4 In claim 1, the data integration unit integrates data by performing location-based integration and time-series synchronization using time windowing, thereby forming a real-time road heat map-based urban heat island mitigation system. Claim 5 A real-time road heat map-based urban heat island mitigation system according to claim 1, having a modular structure that can be expanded into a climate disaster response platform by integrating multiple disaster elements including at least two of heatwaves, heavy rain, cold waves, and fine dust. Claim 6 In claim 1, the AI ​​prediction engine calculates a final risk score by integrating at least two of the predicted heat index, the perceived temperature, and the traffic density, in a real-time road heat map-based urban heat island mitigation system. Claim 7 In claim 1, the data integration unit grids and integrates road section data in units of 50 to 100 meters based on GIS standards, a real-time road heat map-based urban heat island mitigation system. Claim 8 A real-time heat risk prediction method for mitigating urban heat islands, comprising: a step of collecting characteristic data including at least two of pavement material, cumulative solar radiation, reflectance, temperature variation coefficient, and intersection congestion by road; a step of converting the characteristic data into a heat generation index; a step of predicting heat risk by road section using a spatiotemporal deep learning model; and a step of visualizing the predicted heat risk as a real-time heat map. Claim 9 A real-time heat risk prediction method according to claim 8, wherein the step of converting into the heat generation indicator includes the step of converting characteristics using Principal Component Analysis (PCA). Claim 10 A real-time heat risk prediction method according to claim 8, further comprising the step of automatically detecting and correcting abnormal patterns, including sudden surges in vehicle traffic volume or sudden changes in temperature, using an Isolation Forest or Interquartile Range (IQR) rule. Claim 11 delete Claim 12 A real-time heat risk prediction method according to claim 8, further comprising: a step of identifying a high-temperature risk section from the real-time heat map; a step of calculating a heat risk level for the identified high-temperature risk section; a step of transmitting a heat wave warning to a mobile terminal according to the heat risk level; and a step of controlling to support the operation of heat island mitigation infrastructure, including at least one of a water truck, a shade canopy, and an electronic display board of an administrative agency, according to the heat risk level. Claim 13 A real-time heat risk prediction method according to claim 12, further comprising the step of visualizing the heat risk level based on a web-based interface. Claim 14 A computer-readable recording medium storing an urban heat island mitigation program, wherein the urban heat island mitigation program is executed on a computer to perform the following steps: collecting various types of public data in real time, including at least two of road-specific temperatures, traffic volume, weather information, and population density; processing the collected data in an integrated manner in time and space units; predicting heat risk for each road section using at least one of a spatiotemporal graph convolution network (ST-GCN) and a convolutional long short-term memory (ConvLSTM) model; visualizing the predicted heat risk as a real-time heat map based on a geographic information system; and providing a heatwave alert to a mobile terminal based on the heat map information and providing heat island mitigation infrastructure operation information to an administrative agency server.

Images