Search

KR-20260065995-A - Climate Prediction-Integrated Phase-Critical Point Based Disaster and Industrial Safety Resource Scheduling Operating System, Resource Allocation Method Using the Same, and Computer Program Implementing the Same

KR20260065995AKR 20260065995 AKR20260065995 AKR 20260065995AKR-20260065995-A

Abstract

The present invention relates to a disaster and industrial safety integrated operating system (Guardian OS) applicable to various disaster domains, such as natural disasters, industrial accidents, and complex disasters. The invention receives multi-time series (short-term, medium-term, and long-term) weather parameters and uncertainty information from an external climate forecasting system, combines this with real-time field sensor data to detect the point in time when the time derivative of a grid-unit risk index exceeds a critical slope as a phase threshold. For the detected phase threshold, it executes a partial differential equation (PDE)-based digital twin simulation to autonomously generate an optimal schedule that minimizes the resource allocation objective function, and continuously updates the execution results through closed-loop meta-learning. It includes a fail-safe distributed architecture in which edge nodes operate autonomously in the event of a central node failure. [Representative Claim] Claim 1

Inventors

  • 김성금

Assignees

  • 김성금

Dates

Publication Date
20260512
Application Date
20260225

Claims (13)

  1. For one or more disaster domains, A climate interface receiver that receives climate prediction data including one or more weather parameters from an external climate prediction system; A phase threshold detection engine that combines the above climate prediction data with one or more real-time field sensor data to calculate a risk index R(x,t) in spatial grid units, and detects a grid area where the temporal rate of change of the risk index exceeds a predefined critical slope θ_c as a phase threshold; A resource scheduling kernel that simulates disaster spread in response to detected phase thresholds and autonomously generates a resource allocation schedule that minimizes a resource allocation objective function; and A closed-loop meta-learning module that collects result data of executed resource placement as feedback and automatically updates the risk index calculation parameters of the phase threshold detection engine and the objective function weights of the resource scheduling kernel; A phase threshold-based disaster and industrial safety resource scheduling operating system linked with an external climate prediction system including
  2. In claim 1, The above climate interface receiver is, A first channel receiving short-term prediction values with a prediction time horizon of 6 hours or less; A second channel receiving a medium-term forecast value with a forecast time horizon of more than 1 day and less than or equal to 7 days; and It includes a third channel that receives long-term forecast values with a forecast time horizon of more than 7 days and less than or equal to 90 days, and Each channel receives the uncertainty (σ_i) of the corresponding prediction along with the predicted value, and A phase threshold-based disaster and industrial safety resource scheduling operating system linked with an external weather forecasting system, characterized in that the resource scheduling kernel above calculates a buffer resource amount B(x,t) = α_buf · σ_mean(x,t) · R(x,t) in proportion to the uncertainty (σ_i), and includes a predictive pre-placement function that pre-places a resource corresponding to the buffer resource amount at a base near the predicted ignition location prior to phase threshold detection.
  3. In claim 1, The above operating system includes a distributed architecture comprising a central cloud node and a plurality of edge nodes, and includes a fail-safe autonomous mode in which, when communication with the central cloud node is disconnected, each edge node independently performs phase threshold detection and resource allocation schedule generation using only sensor data within its jurisdiction by utilizing the parameters of the phase threshold detection engine and the weights of the resource scheduling kernel that were synchronized immediately prior. A phase threshold-based disaster and industrial safety resource scheduling operating system linked with an external weather prediction system, characterized in that when the above communication is restored, feedback data from each edge node is synchronized with a closed-loop meta-learning module of a central cloud node.
  4. In claim 1, The above resource scheduling kernel calls the digital twin simulation engine immediately upon detecting a phase threshold, and A phase threshold-based disaster and industrial safety resource scheduling operating system linked with an external climate prediction system, characterized in that the digital twin simulation engine expresses the spatial spread of a disaster using a partial differential equation (PDE) model ∂u(x,t)/∂t = ∇·[D(x,t)∇u(x,t)] + F(x,t) - k(x,t)·u(x,t), executes parallel simulations for multiple proposed resource placement scenarios, selects an optimal scenario based on damage minimization criteria, and returns it to the resource scheduling kernel.
  5. As an integrated disaster and industrial safety operation method for one or more disaster domains, (a) receiving weather parameters and prediction uncertainties corresponding to multiple prediction time horizons from an external climate prediction system; (b) a step of calculating a risk index R(x,t) in spatial grid units by combining the above weather parameters and real-time field sensor data; (c) A step of detecting a grid region where the temporal rate of change of the above risk index exceeds a critical slope θ_c as a phase threshold; (d) a step of running a digital twin simulation for the detected phase threshold and generating a resource placement schedule that minimizes the resource placement objective function; (e) A step of executing resource placement commands according to the generated resource placement schedule; (f) a step of collecting the results of resource placement execution as feedback data; and (g) automatically updating the risk index calculation parameters and resource allocation objective function weights using the collected feedback data, and returning to step (a); A disaster and industrial safety integrated operation method using a phase threshold-based disaster and industrial safety resource scheduling operating system linked with an external climate prediction system, wherein steps (a) to (g) are continuously repeated in a closed-loop cycle.
  6. In claim 1, The above closed-loop meta-learning module is, It includes a dual-loop meta-learning structure that uses synthetic disaster scenario data generated through digital twin simulation for inner loop learning and actual disaster response result data for outer loop updating, A phase threshold-based disaster and industrial safety resource scheduling operating system linked with an external climate prediction system, characterized in that the gradient update of the inner loop is performed independently by domain, and the meta-parameters of the outer loop are shared across all domains to perform knowledge transfer between domains.
  7. In claim 1, If the above disaster domain is a wildfire, The above climate interface receiver receives weather parameters including temperature, relative humidity, and wind speed, and The above phase threshold detection engine prioritizes classifying grid regions where predefined threshold conditions, such as a temperature of 38°C or higher, a relative humidity of 15% or lower, and a wind speed of 20 m/s or higher, are simultaneously satisfied as phase threshold candidates, and The above digital twin simulation engine is a phase threshold-based disaster and industrial safety resource scheduling operating system linked with an external climate prediction system, characterized by executing a wildfire spread PDE model that reflects terrain data, vegetation fuel distribution, and real-time weather data.
  8. In claim 1, In the case where the above disaster domain is a complex industrial facility disaster, The above phase criticality detection engine combines IoT temperature sensors, gas detection sensors, SCADA system data, and weather parameters to detect complex phase critical points of fire ignition and hazardous substance diffusion, and A phase threshold-based disaster and industrial safety resource scheduling operating system linked with an external climate forecasting system, characterized in that the above resource scheduling kernel performs multi-objective optimization for fire suppression resources and hazardous substance response resources.
  9. A computer program stored on a computer-readable recording medium comprising instructions that cause a computer to execute the method of claim 5.
  10. For one or more disaster domains, Calculate a risk index R(x,t) at the spatial grid level by combining weather parameters received from an external climate prediction system and real-time field sensor data, and Detecting a grid region where the temporal rate of change ∂R(x,t)/∂t of the above risk index exceeds a predefined critical slope θ_c as a phase threshold Disaster risk early detection system including a phase criticality detection engine.
  11. As a distributed disaster response system including multiple edge nodes and a central cloud node, In the event that communication with the central cloud node is disconnected, it includes a fail-safe autonomous execution mode in which each edge node independently performs risk detection and generates resource placement schedules using only sensor data within its jurisdiction, utilizing risk detection parameters synchronized immediately prior to the event. A distributed disaster response system characterized by synchronizing feedback data with a central node upon communication restoration.
  12. In claim 5, (h) A step of calculating a buffer resource amount B(x,t) = α_buf · σ_mean(x,t) · R(x,t) based on the medium-to-long-term prediction uncertainty received in step (a) above, and pre-deploying it at a base near the predicted ignition location prior to phase threshold detection; A disaster and industrial safety integrated operation method using a phase threshold-based disaster and industrial safety resource scheduling operating system linked with an external climate prediction system, characterized by further including
  13. In claim 5, (i) A fail-safe autonomous execution step in which each edge node independently performs steps (b) through (d) using synchronized parameters when communication with the central node is disconnected; A disaster and industrial safety integrated operation method using a phase threshold-based disaster and industrial safety resource scheduling operating system linked with an external climate prediction system, characterized by further including

Description

Climate Prediction-Integrated Phase-Critical Point Based Disaster and Industrial Safety Resource Scheduling Operating System, Resource Allocation Method Using the Same, and Computer Program Implementing the Same The present invention relates to an integrated disaster and industrial safety operating system (hereinafter 'Guardian OS') that automatically detects a risk critical state (phase critical point) by combining real-time sensor data with multiple time-series weather parameters received from an external climate prediction system across one or more disaster domains, such as natural disasters, industrial accidents, and complex disasters, autonomously schedules disaster response resources through digital twin-based simulation, and continuously updates execution results through closed-loop meta-learning. In this specification, the term 'Operating System (OS)' refers to a kernel-based control layer comprising at least one of an event bus, a policy engine, a resource scheduler, and a state store, and includes a control function that integrates and manages multiple heterogeneous systems and generates and executes real-time resource allocation commands. The Guardian OS of the present invention is responsible for the upper kernel layer (Layer 0~1) that converts climate prediction data into real-time resource placement commands, and operates in conjunction with a field multi-sensor-based disaster detection and response operating system (e.g., a domain-specific sensor OS). The Guardian OS is specialized in phase threshold-based resource scheduling and climate-disaster interfaces, and delegates detailed control at the field sensor level to the lower domain OS. 1. Structural Segmentation Problem of the National Disaster and Safety Information System South Korea's disaster safety information system is structured such that a total of 34 individual information systems under the Ministry of the Interior and Safety operate independently. As the current system involves individual agencies operating separate systems for each stage of prevention, preparedness, response, and recovery, there is no real-time data sharing or integrated decision-making function between systems. The Ministry of the Interior and Safety is pursuing an information system overhaul project spanning 2025 to 2027, but this is limited to the integration of user interfaces for public portals and does not include the integration of backend operational kernels, phase criticality detection engines, or autonomous resource scheduling functions in the overhaul plan. 2. Limitations of the Open-Loop Structure of Existing Disaster Detection and Warning Systems Existing disaster detection and warning systems adopt an open-loop structure that relies entirely on the manual judgment of human commanders for resource deployment decisions following event detection. This structure entails the following technical limitations. First, human judgment delays occur between the detection signal and the resource deployment command, making it structurally difficult to reach resources within the golden time. Second, since the execution results of the detection system are not used to update model parameters, the same error may be repeated under the same environmental conditions. Third, since data generated by multiple heterogeneous sensors is not integrated into a single risk index, it is impossible to determine priorities in complex risk situations. Fourth, since the forecast data from the Numerical Weather Prediction (NWP) model provided by the Korea Meteorological Administration is not linked to the disaster response resource deployment system, forecast information from hours to days prior to a disaster cannot be converted into actual response actions. In fact, the 398 wildfire monitoring drones and AI/CCTV detection systems currently in operation in Korea only managed to detect 16 cases (0.67% of the total) out of 2,376 wildfires that occurred over the past five years (data from the 2024 National Assembly audit). Despite the fact that the large-scale Yeongnam wildfire that occurred in March 2025 destroyed 103,876 hectares, resulted in 31 deaths, and caused approximately 1.818 trillion won in property damage, the meteorological critical conditions immediately preceding ignition (simultaneous occurrence of extreme dryness, high temperature, and strong winds) were not automatically converted into a command for the pre-deployment of disaster response resources. 3. The Problem of Practical Underutilization of the National Disaster and Safety Communication Network The Public Safety LTE network is a disaster communication infrastructure into which approximately 1.5 trillion won of the national budget has been invested. However, limitations have been consistently confirmed in that it cannot be practically utilized as an integrated command and communication tool in large-scale disaster situations due to the coexistence of heterogeneous communication channels between agencies, the absence