Power grid emergency management system and method

By constructing an intelligent perception layer, a dynamic simulation layer, and a decision-making execution layer for the power grid emergency management system, the problem of cross-domain data fragmentation has been solved, enabling proactive control and efficient response of power grid emergency resource scheduling, and adapting to the rapid evolution of extreme disasters.

CN121390897APending Publication Date: 2026-01-23GUANGZHOU JINGKAI TECH CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202511581126.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-10-31
Publication Date
2026-01-23

AI Technical Summary

Technical Problem

In existing power grid emergency command systems, cross-domain data, dynamic simulation, and decision-making execution are disconnected, failing to form a real-time intelligent decision-making closed loop. This results in the system being unable to adapt to the dynamic scenario requirements of rapidly evolving extreme disasters.

Method used

A power grid emergency management system is constructed, comprising an intelligent sensing layer, a dynamic simulation layer, and a decision execution layer. Multi-source raw data is collected through an air-space-ground monitoring network, and spatiotemporal calibration and feature extraction are performed to generate disaster intensity fields and equipment status data. Power flow prediction is performed using physical behavior models to generate equipment failure risk time series maps and power grid status prediction migration paths. Finally, resource scheduling schemes are generated through the decision execution layer.

Benefits of technology

It has achieved a shift from reactive response to proactive prevention and control, improved the efficiency and accuracy of power grid emergency resource dispatch, and formed a closed-loop system of perception-deduction-decision-evolution, which can respond to extreme disasters more efficiently and accurately.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121390897A_ABST
    Figure CN121390897A_ABST
Patent Text Reader

Abstract

The invention discloses a power grid emergency management system and method, and the system comprises an intelligent sensing layer which is used for collecting multi-source original data through a space-air-ground monitoring network when a disaster occurs in an environment where power grid equipment is located, carrying out the time-space calibration, extracting key features, and carrying out the risk initial judgment, generating disaster intensity field data and power grid equipment state data; the dynamic deduction layer is used for matching a target physical model from a physical behavior model library according to the disaster intensity field data and the power grid equipment state data, carrying out power flow prediction, and generating an equipment failure risk time sequence map and a power grid state prediction migration path; and the decision execution layer constructs and solves a target decision function, generates a resource scheduling scheme and outputs the resource scheduling scheme to a corresponding target terminal. A sensing-deduction-decision-evolution closed-loop system is constructed through the layers, so that a response mechanism is changed from lagging disposal to advanced pre-control, and power grid emergency resource scheduling is realized more efficiently and accurately.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of emergency management technology, and in particular to a power grid emergency management system and method. Background Technology

[0002] Global climate change has led to a surge in the frequency and intensity of extreme weather events, posing a systemic threat to power infrastructure. Traditional power grid emergency command, relying heavily on human experience, may suffer from issues such as delayed decision-making and resource misallocation due to factors such as scenario-specific conditions.

[0003] To address this, traditional technologies rely on Geographic Information Systems (GIS) as their core, integrating macro-level early warning information from meteorological departments with power grid equipment operating parameters. This allows for the visualization and labeling of faulty equipment on electronic maps, creating a static overlay view of "meteorology and power grid." Command personnel must manually delineate the disaster area and plan repair routes before sending text instructions containing geographic coordinates to the site, resulting in a unidirectional linear workflow of "data push - manual assessment - instruction issuance."

[0004] However, in the above-mentioned scheme, macro-meteorological data and power grid equipment parameters only achieve coarse-grained superposition of spatial coordinates, without constructing a dynamic correlation model between the intensity of the disaster physical field and the evolution of equipment mechanical performance; the lack of a coupled inference engine for "external disaster excitation - equipment physical failure - power grid system reconfiguration" makes it impossible to simulate disaster paths in advance; manual judgment becomes a bottleneck in information flow, and the delay from data collection to instruction generation far exceeds the golden window for disaster response. Furthermore, cross-domain data, dynamic inference, and decision execution are disconnected, failing to form a real-time response intelligent decision-making closed loop, resulting in the system being unable to adapt to the dynamic scenario requirements of rapidly evolving extreme disasters. Summary of the Invention

[0005] This invention provides a power grid emergency management system and method, which solves the technical problem in existing technical solutions where cross-domain data, dynamic simulation and decision execution are separated, and a real-time response intelligent decision-making closed loop is not formed, resulting in the system being unable to adapt to the dynamic scenario requirements of rapidly evolving extreme disasters.

[0006] The present invention provides a power grid emergency management system, comprising a smart sensing layer, a dynamic simulation layer, and a decision execution layer connected by communication;

[0007] The intelligent sensing layer is used to collect multi-source raw data through the air-space-ground monitoring network and perform spatiotemporal calibration to generate standardized data when a disaster occurs in the environment where the power grid equipment is located; extract key features in the standardized data and make a preliminary risk assessment to generate disaster intensity field data and power grid equipment status data.

[0008] The dynamic simulation layer is used to match the target physical model from the physical behavior model library and perform power flow prediction based on the disaster intensity field data and the power grid equipment status data, and generate equipment failure risk time series map and power grid status prediction migration path.

[0009] The decision execution layer is used to construct and solve the target decision function based on the equipment failure risk time series map and the power grid state prediction migration path, generate a resource scheduling scheme, and output it to the corresponding target terminal.

[0010] Optionally, it also includes a closed-loop evolution layer, specifically used for:

[0011] Calculate the numerical deviation between the predicted target result and the actual target result corresponding to the resource scheduling scheme;

[0012] The distribution of the contribution of positioning strategy deviation is determined by multi-scale evaluation based on the numerical deviation.

[0013] Based on the distribution of the policy deviation contribution and historical data, feedback adjustment instructions are generated for the intelligent perception layer, the dynamic inference layer, and the decision execution layer, and then transmitted.

[0014] Optionally, the air-space-ground monitoring network includes monitoring satellites with multimodal hybrid networking communication, UAV swarms, and an Internet of Things sensor network; the multi-source raw data includes macro-meteorological data, regional micro-data, and second-level device data; the intelligent sensing layer is specifically used for:

[0015] When a disaster occurs in the environment where the power grid equipment is located, the monitoring satellite periodically scans the macro-meteorological data of the area where the power grid equipment is located to locate risk hotspots.

[0016] The drone formation is invoked to move to the risk hotspot area and periodically collect regional micro data;

[0017] The system utilizes an IoT sensor network within the aforementioned risk hotspot area to collect device data at the second-level.

[0018] After performing spatiotemporal calibration and standardization on the macro-meteorological data, the regional micro-data, and the second-level equipment data, standardized data is generated.

[0019] Principal component analysis was performed on the standardized data to extract key features; the key features included at least power grid equipment status data.

[0020] Based on the aforementioned key features, Bayesian inference is performed to generate disaster intensity field data.

[0021] Optionally, the intelligent sensing layer is further provided with a self-maintenance device, specifically used for:

[0022] Periodically diagnose whether any abnormalities occur in the various IoT sensors within the IoT sensor network;

[0023] If so, the calibration model is invoked to perform online correction on the IoT sensor.

[0024] Optionally, the dynamic inference layer is specifically used for:

[0025] The power grid equipment status data is classified, and the target physical model is matched from the physical behavior model library according to the classification results.

[0026] Spatial interpolation is performed based on the disaster intensity field data, and the data is mapped to the location of each target physical model as input load;

[0027] The target physical model is used to predict the power flow of the input load. Based on the power flow prediction results, the location and probability of failure of the equipment in different time periods are determined, and a time series map of equipment failure risk is constructed.

[0028] Based on the location of each failed device and the device change strategy, a time-series evolution is performed to generate a power grid state prediction migration path.

[0029] Optionally, it also includes a prediction correction system, specifically used for:

[0030] Based on the difference between the predicted power flow results and the actual observation results, the model parameters of the target physical model are iteratively optimized in reverse.

[0031] Optionally, the decision execution layer is specifically used for:

[0032] Security constraints are constructed based on the predicted migration path of the power grid state, and constraint conditions are created by combining resource constraints and time constraints.

[0033] Based on the aforementioned equipment failure risk time series map, risk terms of the target decision function are constructed, and cost and time terms are superimposed to create the target decision function;

[0034] With the objectives of minimizing risk, cost, and time, a resource scheduling scheme is generated by using a genetic algorithm or particle swarm optimization algorithm to find the approximate optimal solution of the objective decision function under the constraints within a preset solution period.

[0035] According to the resource scheduling scheme, scheduling instructions are issued to the corresponding target terminals; the scheduling instructions include a resource scheduling spatiotemporal trajectory diagram, a state operation sequence, and a cross-agency collaboration protocol.

[0036] Optionally, the objective decision function is:

[0037]

[0038] in, This is a risk item. For cost items, For time terms, As the weight of risk items, As the weight of cost items, Let X be the weight of the time term, X be the resource scheduling scheme, and t be the time point.

[0039] Optionally, the intelligent sensing layer is also used to generate a panoramic view of disaster risk based on the standardized data.

[0040] The present invention also provides a power grid emergency management method, applied to the power grid emergency management system described in any of the above claims, the method comprising:

[0041] When a disaster occurs in the environment where the power grid equipment is located, the intelligent sensing layer calls the air-space-ground monitoring network to collect multi-source raw data and performs spatiotemporal calibration to generate standardized data.

[0042] The intelligent sensing layer extracts key features from the standardized data and performs preliminary risk assessment to generate disaster intensity field data and power grid equipment status data.

[0043] Based on the disaster intensity field data and the power grid equipment status data, the dynamic inference layer matches the target physical model from the physical behavior model library and performs power flow prediction to generate a time series map of equipment failure risk and a power grid status prediction migration path.

[0044] The decision execution layer constructs and solves the target decision function based on the equipment failure risk time series map and the power grid state prediction migration path, generates a resource scheduling scheme, and outputs it to the corresponding target terminal.

[0045] As can be seen from the above technical solutions, the present invention has the following advantages:

[0046] This invention provides a power grid emergency management system, comprising a smart sensing layer, a dynamic inference layer, and a decision execution layer connected by communication. The smart sensing layer collects multi-source raw data through a space-air-ground monitoring network and performs spatiotemporal calibration to generate standardized data when a disaster occurs in the environment where power grid equipment is located. It then extracts key features from the standardized data and performs preliminary risk assessment, generating disaster intensity field data and power grid equipment status data. The dynamic inference layer matches a target physical model from a physical behavior model library and performs power flow prediction based on the disaster intensity field data and power grid equipment status data, generating a time-series map of equipment failure risk and a predicted migration path for power grid status. The decision execution layer constructs and solves a target decision function based on the time-series map of equipment failure risk and the predicted migration path for power grid status, generating a resource scheduling scheme and outputting it to the corresponding target terminal. By constructing a closed-loop system of perception-inference-decision-evolution through these layers, the response mechanism shifts from reactive handling to proactive control, enabling more efficient and accurate power grid emergency resource scheduling. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 A structural block diagram of a power grid emergency management system provided in an embodiment of the present invention;

[0049] Figure 2 A data processing flowchart of a closed-loop evolution layer is provided in an embodiment of the present invention;

[0050] Figure 3 A data processing flowchart for a dynamic inference layer provided in an embodiment of the present invention;

[0051] Figure 4 A data processing flowchart for the decision execution layer provided in this embodiment of the invention;

[0052] Figure 5 A flowchart illustrating the steps of a power grid emergency management method provided in an embodiment of the present invention. Detailed Implementation

[0053] This invention provides a power grid emergency management system and method to address the technical problem in existing technical solutions where cross-domain data, dynamic simulation, and decision execution are disconnected, failing to form a real-time intelligent decision-making closed loop, thus making the system unable to adapt to the dynamic scenario requirements of rapidly evolving extreme disasters.

[0054] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0055] Please see Figure 1 , Figure 1 This is a structural block diagram of a power grid emergency management system provided in an embodiment of the present invention.

[0056] The present invention provides a power grid emergency management system, comprising an intelligent sensing layer 101, a dynamic simulation layer 102, and a decision execution layer 103 connected by communication;

[0057] The intelligent sensing layer 101 is used to collect multi-source raw data through the air-space-ground monitoring network and perform spatiotemporal calibration to generate standardized data when a disaster occurs in the environment where the power grid equipment is located; extract key features in the standardized data and make a preliminary risk assessment to generate disaster intensity field data and power grid equipment status data.

[0058] Dynamic simulation layer 102 is used to match the target physical model from the physical behavior model library and perform power flow prediction based on disaster intensity field data and power grid equipment status data, and generate equipment failure risk time series map and power grid status prediction migration path;

[0059] The decision execution layer 103 is used to construct and solve the target decision function based on the equipment failure risk time series map and the power grid state prediction migration path, generate a resource scheduling scheme and output it to the corresponding target terminal.

[0060] The air-space-ground monitoring network refers to a three-level collaborative monitoring system consisting of outer space satellites, mid-air unmanned aerial vehicle (UAV) formations, and ground-based Internet of Things (IoT) sensor networks.

[0061] Disaster intensity field data refers to quantitative data that characterizes the evolutionary features of disasters (such as development path, intensity gradient, impact range, and derivative correlation characteristics), and is used to reflect the spatiotemporal variation patterns of natural external forces.

[0062] Power grid equipment status data refers to a dataset formed by integrating the mechanical structure performance evolution data of power grid equipment with real-time electrical quantity monitoring values ​​(such as vibration, displacement, icing thickness, and circuit breaker status), reflecting the dynamic response characteristics of power facilities in disaster environments.

[0063] Equipment failure risk time series map refers to a structured map that identifies the location of high-risk facilities and their failure probability at different times.

[0064] The power grid state prediction migration path refers to the path that describes the sequence of changes in power grid operation mode that may result from the evolution of faults. It is obtained through the system-level cascading fault deduction process, such as simulating protection actions, analyzing power flow transfer, and assessing system stability.

[0065] The resource scheduling scheme refers to the structured scheme generated by the decision execution layer 103, which includes a resource scheduling spatiotemporal trajectory diagram (describing the movement paths and time nodes of personnel and equipment), a state change instruction set (specifying the operation sequence of power grid facilities), and a collaborative action protocol (clarifying the cross-agency task handover), used to guide physical disaster response actions.

[0066] The target terminals include, but are not limited to, mobile terminals of power grid emergency repair teams, terminals of power grid control centers, and cross-agency collaborative terminals such as fire protection and medical services. They obtain and execute resource scheduling plans through encrypted communication channels.

[0067] In this embodiment of the invention, the power grid emergency management system is based on a closed-loop design encompassing perception, deduction, decision-making, and evolution, constructing a digital technology system characterized by interconnected communication, data integration, and synergistic efficiency. The system uses an intelligent perception layer 101 as its data foundation, a dynamic deduction layer 102 as its cognitive core, and a decision-making execution layer 103 as its action transformation center, supplemented by a closed-loop evolution layer to achieve autonomous evolution. Each layer achieves real-time data interaction and functional collaboration through a robust communication network, jointly achieving the core objectives of proactive prevention and control, precise handling, and continuous optimization of power grid disasters.

[0068] Specifically, the intelligent sensing layer 101 mainly consists of an air-space-ground monitoring network, a multi-source data processing engine, a risk feature identification module, and a self-maintenance device. It collects multi-source raw data on the disaster environment and power grid equipment through the air-space-ground monitoring network, eliminates data heterogeneity through spatiotemporal calibration and standardization, and completes initial risk assessment through feature extraction and decision fusion. Finally, it outputs disaster intensity field data depicting the evolution of the disaster, and power grid equipment status data reflecting the dynamic response of the equipment, providing a high-fidelity data foundation for subsequent simulations and decisions. The dynamic simulation layer 102 consists of a physical behavior model library, a multi-physics coupling engine, a multi-scale correlation framework, and a prediction and correction module. It receives data output from the intelligent sensing layer 101, matches and adapts the target physical model through an adaptive modeling mechanism, and completes power flow prediction by combining multi-physics coupling calculations and system-level cascading fault simulation. This generates a time-series map of equipment failure risks that identifies high-risk equipment, and a power grid status prediction migration path describing the evolution of the power grid operation mode. The decision execution layer 103 uses a multi-dimensional value platform engine as its core to jointly optimize and solve problems, issuing instructions to various target terminals. The multi-dimensional value balancing engine constructs a target decision function with resource, timeliness, and security constraints. The optimization and solution module uses metaheuristic algorithms (such as genetic algorithms) to formalize the function into a mixed integer programming problem. The instruction generation and distribution module transforms the optimal solution into a resource scheduling scheme and outputs it to the target terminal through an encrypted channel. The dynamic adaptation mechanism receives on-site feedback data, and if the environment or resource status changes, it re-triggers the solution process to update the scheme, ensuring that actions adapt to the evolution of the disaster.

[0069] In this embodiment of the invention, when a disaster occurs in the environment where the power grid equipment is located, the intelligent sensing layer 101 first acquires multi-source raw data through the air-space-ground monitoring network. The multi-source data processing engine uses a spatiotemporal registration algorithm and a sensor error compensation algorithm to perform spatiotemporal calibration on the raw data, converting it into standardized data with a unified format. The risk feature identification module then extracts key features of disaster evolution and equipment status from the standardized data, completes the initial risk assessment through decision-layer fusion technology, and generates disaster intensity field data and power grid equipment status data. After the data quality is verified by the self-maintenance system, the data is transmitted to the dynamic inference layer 102 through a robust communication network. After receiving the data, the dynamic inference layer 102 uses an adaptive modeling mechanism to identify the current disaster type through a lightweight machine learning classification algorithm, matches the appropriate target physical model from the physical behavior model library, and the multi-physics coupling engine realizes the interactive calculation of mechanical-electrical-thermodynamic fields based on the target physical model. At the same time, the system-level cascading fault inference process is initiated (setting the initial fault set, simulating the protection action sequence, analyzing the power flow transfer path, and evaluating the system stability characteristics) to complete the power flow prediction. The multi-scale correlation framework maps the disaster intensity field data to the equipment coordinates. The system calculates failure probability and timing characteristics, analyzes changes in power grid operation mode, generates a time-series map of equipment failure risk and a predicted migration path for power grid status. The prediction and correction module compares real-time data with predicted values ​​using a Bayesian update method, adjusts key model parameters to improve inference accuracy, and then transmits the predicted data to the decision execution layer 103. After receiving the predicted data, the decision execution layer 103 uses a multi-dimensional value balance engine to construct a target decision function (aiming to minimize comprehensive loss) containing resource constraints (action resource consumption does not exceed the total amount), time pressure constraints (key actions must be completed within a time limit), and safety boundary constraints (avoiding dangerous power grid states). The optimization and solution module formalizes the function into a constraint satisfaction problem and uses a metaheuristic algorithm to solve for the Pareto optimal solution. The instruction generation and distribution module transforms the optimal solution into a resource scheduling scheme that includes a resource scheduling spatiotemporal trajectory map, a state change instruction set, and a collaborative action protocol. This scheme is then output to the corresponding target terminal through an encrypted channel. The dynamic adaptation mechanism receives real-time feedback on environmental mutations or resource status changes from the field. If adjustments are needed, the solution process is re-triggered to update the scheme, achieving closed-loop interaction of data flow between layers and precise transformation from digital cognition to physical action.

[0070] Optionally, it also includes a closed-loop evolution layer, specifically used for:

[0071] The numerical deviation between the predicted target result and the actual target result corresponding to the resource scheduling scheme is calculated.

[0072] Multi-scale evaluation based on numerical deviation, distribution of the contribution of positioning strategy deviation;

[0073] Based on the distribution of the contribution of strategy deviation and historical data, feedback adjustment instructions are generated and transmitted for the intelligent perception layer 101, the dynamic inference layer 102, and the decision execution layer 103, respectively.

[0074] The predictive target structure refers to the data on the handling effect predicted by the dynamic inference layer 102 model based on the resource scheduling scheme output by the decision execution layer 103, including the expected degree of equipment risk reduction, task completion time, and resource utilization efficiency.

[0075] Actual target results refer to the real handling effect data obtained through on-site feedback such as emergency repair terminal data and power grid operation monitoring data after the resource scheduling plan is implemented. This includes the actual level of risk control achieved, the actual task time, and the actual resource consumption, reflecting the actual implementation effectiveness of the plan.

[0076] The strategy deviation contribution distribution refers to a graph generated by contribution decomposition technology, which depicts the contribution ratio of each link to the total value deviation, such as "data error in the intelligent perception layer 101, model deviation in the dynamic inference layer 102, and insufficient adaptability of the decision execution layer 103".

[0077] Feedback adjustment instructions refer to a set of instructions generated based on the distribution of policy deviation contribution and historical data, used to optimize the operation logic of the other three layers. Instructions for the intelligent perception layer 101 include monitoring resource reconfiguration and sampling frequency adjustment; instructions for the dynamic inference layer 102 include model parameter correction and confidence coefficient update; instructions for the decision execution layer 103 include objective decision function weight optimization and constraint threshold adjustment.

[0078] In this embodiment of the invention, after the decision execution layer 103 issues the resource scheduling plan to the target terminal, it simultaneously collects the actual target results after execution, such as the actual risk control situation and task time, to form execution feedback data, which is then transmitted to the closed-loop evolution layer. The closed-loop evolution layer receives the predicted target results and the actual target results, calculates the numerical deviation, and conducts multi-scale evaluations, such as attribution analysis and contribution decomposition, through full-element data synchronization and digital twin trajectory reconstruction, to determine the distribution of the contribution of strategy deviations, such as the proportion of data error in the perception layer, the proportion of model deviation in the inference layer, and the proportion of insufficient adaptability in the decision layer. Specifically, through full-element data synchronization and digital twin trajectory reconstruction, the deviation between the entity action trajectory fed back by the execution layer and the target achievement is continuously analyzed from multiple dimensions. The deviation is quantified and features are extracted, and root cause tracing and cognitive defect localization are performed on environmental response and resource scheduling, ultimately achieving knowledge system evolution and strategy generation optimization. In the time dimension, it connects the dual paths of short-term correction and long-term evolution, achieving synergy between minute-level parameter fine-tuning and annual-level rule reconstruction; in the spatial dimension, it establishes a correlation mapping between individual events and system laws, extracting a universal cognitive framework from discrete cases; in the efficiency dimension, it integrates the contradictory needs of accuracy and robustness, forming a flexible knowledge body that balances optimal solutions and disturbance resistance. The synergy between this layer and the perception / inference / decision layer is manifested in a holistic empowerment relationship: evolutionary output dynamically refreshes the core parameters and logical structures of each layer, while the operational data of each layer feeds back into the raw materials for evolutionary analysis.

[0079] When the dual-mode driven mechanism detects a deviation exceeding a threshold, it generates a learning event alert. The system prioritizes allocating significant computing resources (such as activating high-performance computing nodes) to this event, generating a detailed incident analysis report through deep analysis to identify the root cause and directly drive major updates to the rule base and model base. An automated data pipeline is established, continuously injecting execution deviation data, environmental data, and operation records. Online machine learning algorithms and incremental learning models are used to continuously learn from the data, with the learning results being fine-tuned in small, rapid steps. By filtering abnormal data, using incremental and robust learning algorithms to handle abnormal data, verifying the consistency of results generated from new and old data, and implementing version control and snapshot rollback, the knowledge update process is protected from interference by abnormal data, maintaining system stability. The three mechanisms of the core evolutionary architecture do not work independently, but rather form a rigorous collaborative workflow: Regular, cumulative learning acts like the system's metabolism, continuously fine-tuning; when deviations become too large, event-triggered learning is activated as an immune response, performing deep cleansing and repair; the multi-scale evaluation framework acts as a precise diagnostic tool for the system, pinpointing the root cause of problems; and the perturbation-resistant evolutionary design acts like an immune system monitoring and version control, ensuring that any updates do not cause system crashes, maintaining overall stability and reliability. Through this design, the system's evolution is no longer blind trial and error, but a precise, robust, and highly automated cognitive sublimation process. For the intelligent perception layer 101, it outputs optimized vectors from the perception network to guide the dynamic tilting of monitoring resources towards high-value target areas; for the dynamic inference layer 102, it outputs inference confidence correction coefficients to improve the prediction accuracy of specific disaster scenarios; and for the decision execution layer 103, it outputs a set of decision rule patches to repair potential vulnerabilities in the strategy generation logic.

[0080] Specifically, such as Figure 2As shown, the execution layer feedback data pool inputs data into the dual-mode drive engine. This engine works in tandem through event-triggered learning (when deviations exceed limits, high computing power is scheduled to conduct deep root cause tracing, outputting an incident analysis report to drive major updates to the rule base / model base) and routine accumulation learning (relying on an automated data pipeline to continuously inject execution deviations, environmental and operational records, and using online machine learning for incremental fine-tuning). After updating the cognitive gene library through the rule reconstruction matrix, it outputs the perception layer optimization vector (guiding the reconfiguration of perception resources), the inference layer confidence correction coefficient (driving model parameter updates), and the decision layer rule patch set (achieving strategy logic repair) through the global empowerment radiation network. Ultimately, the new data is fed back to the execution layer's feedback data pool. Simultaneously, the system ensures stability through anti-disturbance evolution design, including abnormal data filtering, incremental robust algorithms, result consistency verification, and version control. These three mechanisms form a collaborative workflow of "normal accumulation (metabolic fine-tuning) + event triggering (immune response repair) + multi-scale evaluation (precise diagnosis) + anti-disturbance guarantee," enabling the system to evolve into a precise and robust automated cognitive sublimation process. This achieves closed-loop optimization, including the intelligent perception layer 101 dynamically tilting monitoring resources towards high-value areas, the dynamic inference layer 102 improving the prediction accuracy of specific disaster scenarios, and the decision execution layer 103 patching loopholes in the strategy generation logic.

[0081] Optionally, the air-space-ground monitoring network includes monitoring satellites with multimodal hybrid networking communication, UAV swarms, and IoT sensor networks; the multi-source raw data includes macro-meteorological data, regional micro-data, and second-level device data; the intelligent sensing layer 101 is specifically used for:

[0082] When a disaster occurs in the environment where the power grid equipment is located, risk hotspots can be located by periodically scanning the macro-meteorological data of the area where the power grid equipment is located by detection satellites.

[0083] The drone formation was deployed to move to high-risk hotspots and periodically collect micro-data of the area.

[0084] It utilizes IoT sensor networks within high-risk hotspot areas to collect device data at the second-level.

[0085] After spatiotemporal calibration and standardization of macro-meteorological data, regional micro-data, and second-level equipment data, standardized data is generated.

[0086] Principal component analysis was performed on the standardized data to extract key features; the key features included at least the power grid equipment status data.

[0087] Based on key features, Bayesian inference is performed to generate disaster intensity field data.

[0088] Multimodal hybrid networking refers to the communication technology architecture adopted by the air-space-ground monitoring network, which integrates multiple communication methods to form a heterogeneous network without relying on any single communication method.

[0089] Monitoring satellites refer to the outer space acquisition nodes of a space-air-ground monitoring network. They are used to periodically scan the areas where power grid equipment is located, acquiring macro-meteorological data reflecting the macro-evolution patterns of disasters. This enables preliminary disaster monitoring and risk hotspot location with full coverage. They continuously acquire large-scale meteorological evolution data (typhoon paths, precipitation cloud systems), geological deformation information (surface displacement, landslide trends), and ecological change indicators (vegetation cover, water area); providing periodic scanning covering the entire region, enabling continuous tracking of the macro-evolution patterns of disaster systems.

[0090] Unmanned aerial vehicle (UAV) swarms refer to mid-altitude data acquisition nodes in an air-space-ground monitoring network. These nodes can move to high-risk hotspots upon command, periodically collecting regional micro-data reflecting local micro-environmental characteristics to achieve refined disaster monitoring of high-risk areas. Enhanced scanning is implemented on high-risk areas identified by satellites to achieve centimeter-level accuracy in terrain modeling and facility monitoring, enabling emergency mobile monitoring and rapid response and refined reconnaissance in the event of sudden disasters.

[0091] The Internet of Things (IoT) sensor network refers to the ground-level acquisition nodes of a space-air-ground monitoring network. These nodes are deployed around power grid equipment and in high-risk areas to collect second-level data reflecting the real-time status of the equipment. Sensor types include, but are not limited to, environmental parameter sensors (temperature, humidity, wind speed and direction, rainfall monitoring), equipment status sensors (vibration, tilt, displacement, icing thickness detection), and image acquisition devices (high-definition cameras, infrared thermal imagers). This enables second-level data acquisition and builds a holographic perception capability for the micro-state and micro-environmental changes at the equipment level.

[0092] Macro-meteorological data refers to data that reflects the macro-evolutionary characteristics of disasters in the area where power grid equipment is located.

[0093] Regional micro-data refers to data that reflects the characteristics of local micro-environmental disasters and can supplement the deficiencies of macro-meteorological data at the micro-scale.

[0094] Second-level equipment data refers to real-time status data of power grid equipment, which can accurately reflect the dynamic response of equipment in disaster environments.

[0095] In this embodiment of the invention, the intelligent sensing layer 101 is the central hub for the system's global data acquisition and preprocessing. It uses a multimodal hybrid networking communication network (including monitoring satellites, UAV formations, and IoT sensor networks) as its core acquisition carrier, coupled with a spatiotemporal calibration module, a standardization processing module, a principal component analysis module, and a Bayesian inference module to form a complete functional chain of "data acquisition - preprocessing - feature extraction - risk quantification." Specifically, monitoring satellites are responsible for macro-meteorological data acquisition and risk hotspot location; UAV formations are responsible for regional micro-data acquisition in risk hotspot areas; and the IoT sensor network is responsible for second-level equipment data acquisition in risk hotspot areas. Multimodal hybrid networking communication ensures stable data interaction between the three types of acquisition nodes. The spatiotemporal calibration module and the standardization processing module are used to eliminate the heterogeneity of multi-source raw data; the principal component analysis module is used to extract key features including power grid equipment status data; and the Bayesian inference module is used to generate disaster intensity field data, ultimately providing high-fidelity disaster environment and equipment status data support for the dynamic simulation layer 102.

[0096] When a disaster occurs in the environment where the power grid equipment is located, the intelligent sensing layer 101 relies on an air-ground monitoring network with multimodal hybrid networking communication to carry out data collection and processing: First, monitoring satellites periodically scan the area where the power grid equipment is located to obtain macro-meteorological data and locate risk hotspots based on this; then, through multimodal hybrid networking communication, a UAV formation is called to move to the risk hotspot area, and the UAV formation periodically collects regional micro-data; at the same time, the Internet of Things sensor network in the risk hotspot area is called to collect second-level equipment data reflecting the real-time status of the equipment; then, the collected macro-meteorological data, regional micro-data, and second-level equipment data are transmitted to the spatiotemporal calibration module for spatiotemporal consistency correction, and then converted into standardized data with a unified format by the standardization processing module. The sensor error model and data compensation algorithm are used to automatically correct equipment deviations and environmental interference; then, the principal component analysis module processes the standardized data to extract key features containing power grid equipment status data; finally, the Bayesian inference module performs inference calculations based on the extracted key features to generate disaster intensity field data that quantitatively reflects the spatiotemporal distribution of disaster intensity.

[0097] In its implementation, the system identifies potential risk hotspots using initial wide-area scanning data and then activates an enhanced observation mode: spatially, it commands nearby drone swarms to converge on the hotspot area for intensive, three-dimensional surveillance; simultaneously, it schedules satellite remote sensing resources to increase the frequency of overflight and imaging of the area; at the ground level, it adaptively activates surrounding backup mobile sensor nodes and adjusts the sampling frequency of fixed nodes to the highest level, forming a collaborative "satellite-drone-ground" enhanced perception network. Once the risk diminishes, the system automatically deactivates the enhanced mode, and resources revert to wide-area patrol mode, thereby maximizing observation efficiency. Spatiotemporal reference registration algorithms and sensor data standardization are employed to align and preprocess multiple raw sensor data from the same monitoring target. Principal component analysis, feature stitching, and deep learning feature extraction techniques are used to extract representative features from various data types, and these feature vectors are then fused into a unified feature set for subsequent analysis and identification. Bayesian inference and DS evidence theory are used to independently make preliminary judgments on each data source or algorithm, and then the independent decisions are synthesized to arrive at a final conclusion.

[0098] The disaster intensity field data can be obtained through feature simulation using a dynamic field model based on the fusion of physical mechanisms and data-driven approaches. Specifically, it can include spatiotemporal dynamic features: disaster development path, spread speed, and evolution of the scope of influence; intensity evolution features: disaster intensity change gradient and extreme value distribution patterns; and derivative correlation features: secondary disaster triggering mechanisms and chain reaction paths.

[0099] Furthermore, to ensure the continued transmission of critical alarm information from the sensing network to the command center under extreme disaster conditions (such as strong winds, heavy rain, or earthquakes causing damage to some infrastructure), guaranteeing the survivability of the information link, the intelligent sensing layer 101 in this embodiment employs a multimodal hybrid networking approach for monitoring satellites, drone swarms, and IoT sensor networks. Simultaneously, the network possesses intelligent management capabilities, enabling real-time monitoring of the communication quality of each link. When an interruption is detected in a primary path, the data stream automatically and seamlessly switches to a backup path to ensure uninterrupted communication. When bandwidth is limited, the system automatically activates a data priority and hierarchical transmission mechanism.

[0100] In one example of the present invention, the intelligent sensing layer 101 is further provided with a self-maintenance device, specifically used for:

[0101] Periodically diagnose whether any abnormalities occur in the various IoT sensors within the IoT sensor network;

[0102] If so, the calibration model is invoked to perform online correction of the IoT sensor.

[0103] The calibration model refers to the algorithm model pre-set in the self-maintenance device for correcting abnormal data from IoT sensors. It can adjust the collected data of abnormal sensors online based on historical normal data of the sensors, cross-validation data of similar sensors, and physical law verification logic, so as to restore the accuracy of the data.

[0104] In this embodiment of the invention, the self-maintenance device is directly associated with the Internet of Things (IoT) sensor network. It scans the operating status of each IoT sensor through a periodic diagnostic procedure and identifies data anomalies in real time. It has a built-in calibration model library. When a sensor anomaly is detected, it automatically calls the matching calibration model to correct the second-level device data collected by the sensor online, ensuring the accuracy of device data in multi-source raw data.

[0105] Specifically, spatial consistency can be verified by comparing readings from adjacent or similar sensors; physical logic consistency verification can be performed by using known physical laws to check data; a dynamically changing confidence score can be assigned to each data point or each sensor; and for data identified as abnormal, the system can call a calibration model to correct it online.

[0106] In one example of the invention, the intelligent sensing layer 101 is also used to generate a panoramic view of disaster risk based on standardized data.

[0107] A disaster risk panorama view refers to a dynamic, visualized sand table that integrates multi-dimensional risk information based on standardized data. It combines real-time monitoring data (macro-meteorological, regional micro-level, and second-level equipment data) and preliminary risk assessment results (disaster intensity field data and power grid equipment status data). It can perform spatial zooming and time-lapse operations, intuitively presenting the disaster distribution, risk level, and equipment status of the area where the power grid equipment is located, providing support for commanders to understand the overall risk.

[0108] In this embodiment, a correlation model between environmental parameters and disaster characteristics is established based on machine learning algorithms. This enables early identification of disaster risks and dynamic extrapolation of their evolution trends, generating a disaster development probability map and a spatiotemporal evolution sequence. Real-time monitoring data, predicted trends, and historical cases are integrated to construct a multi-dimensional risk profile, forming a spatially scalable and temporally shiftable dynamic risk sandbox. This panoramic view of disaster risk can provide disaster development scenario extrapolation based on real data; support consequence simulation and benefit assessment under different response strategies; provide quantitative basis for power facility risk pre-control; provide optimization suggestions for disaster prevention and mitigation resource allocation; and provide contextualized support for emergency command and decision-making.

[0109] It should be noted that the Intelligent Sensing Layer 101 reconstructs the traditional monitoring paradigm through three innovations: Spatial dimension innovation: breaking down the cognitive barriers between macro and micro levels, achieving a collaborative closed loop of "satellite survey - UAV detailed survey - sensor verification"; establishing an observation system that unifies full coverage and key enhancement. Physical dimension innovation: bridging the observation gap between environmental parameters and equipment status; establishing a correlation mapping model of "natural external forces - equipment response - system impact". Temporal dimension innovation: overcoming the information lag defects of discrete sampling; forming a continuously evolving dynamic field description and prediction capability. The Intelligent Sensing Layer 101 establishes a complete technological chain from data acquisition to cognitive enhancement by constructing an integrated air-space-ground intelligent observation network. Its innovative value is not only reflected in technological breakthroughs, but also in the fundamental reconstruction of the traditional disaster monitoring paradigm, providing a completely new level of perception capabilities and decision support for modern emergency command systems. The perception layer outputs a structured environmental description system: a spatial distribution model of disaster intensity, quantitatively expressing the spatiotemporal variation patterns of natural external forces; a set of equipment operating status parameters reflecting the dynamic response characteristics of power facilities in a disaster environment; and risk feature identification results, pointing to potential threat locations requiring key attention. The capabilities of this layer have been validated through system testing: spatial observation resolution has been improved by orders of magnitude, the timeliness of key parameter acquisition has been compressed to the second level, and information fidelity meets the input requirements of subsequent modules. These characteristics lay the cognitive foundation for disaster transmission simulation, enabling the technical system to possess the ability to predict environmental responses in advance.

[0110] In one example of the present invention, the dynamic inference layer 102 is specifically used for:

[0111] Classify the power grid equipment status data, and match the target physical model from the physical behavior model library according to the classification results;

[0112] Spatial interpolation is performed based on the disaster intensity field data, and the data is mapped to the location of each target physical model as the input load;

[0113] The target physical model is used to predict the power flow of the input load. Based on the power flow prediction results, the location and failure probability of the failed equipment at different time periods are determined, and a time series map of equipment failure risk is constructed.

[0114] Based on the location of each failed device and the device change strategy, a time-series evolution is performed to generate a power grid state prediction migration path.

[0115] The physical behavior model library refers to a collection of equipment-level models, including but not limited to structural mechanics models (simulating the stress and strain of towers and conductors under external forces such as wind load, ice load, and geological displacement), electrical characteristic models (simulating power flow transfer, voltage fluctuation, and protection action sequence under faults), thermodynamic models (calculating the evolution of thermal processes such as equipment overheating and sag increase), and a multi-field coupling engine (realizing real-time interactive calculation of mechanical, electrical, and thermodynamic fields). Each model corresponds to a specific type / state of power grid equipment.

[0116] Input loads refer to quantified disaster intensity data (such as wind load, thermal radiation flux, and ice load) mapped to the location of equipment in each target physical model. They are the core input parameters for the target physical model to carry out power flow prediction.

[0117] Equipment change strategies refer to pre-defined adjustment schemes used to deal with equipment failures, including strategies such as grid topology reconfiguration, standby equipment switching, and load transfer. These strategies are formulated based on grid operation rules and disaster response experience, and provide operational basis for the evolution of grid status.

[0118] In this embodiment of the invention, after the dynamic simulation layer 102 receives the power grid equipment status data and disaster intensity field data, the power grid equipment status data classification module first classifies the power grid equipment status data according to equipment attributes and status characteristics. Based on the classification results, the target physical model of the corresponding equipment is matched from the physical behavior model library. Then, the disaster intensity field data is processed by the spatial interpolation module using an interpolation algorithm. The disaster intensity data of the entire domain is quantified and mapped to the geographical coordinates of the power grid equipment where each target physical model is located, forming the input load required for the target physical model to carry out calculations. Specifically, the environmental field data obtained by the perception layer can be transformed into equipment risk quantification indicators through spatial gradient mapping, dynamic response calculation, and temporal evolution prediction. Subsequently, the power flow prediction calculation module calls the target physical model and combines it with the multiphysics coupling engine to realize real-time interactive calculation of mechanical, electrical, and thermodynamic fields. It also links with the system-level cascading fault simulation component to simulate relay protection actions and power redistribution. Through power flow prediction, it determines the location of failed equipment and the corresponding failure probability in different time periods. The equipment failure risk time series map construction module integrates this information to construct the equipment failure risk time series map. Finally, the power grid state prediction migration path generation module, based on the location of each failed device in the time series map and combined with preset equipment change strategies (such as topology reconfiguration and load transfer), conducts power grid state time series evolution simulation to generate power grid state prediction migration paths that describe the sequence of power grid operation mode changes. This constructs an accurate digital mirror system, realizing the full-process advanced simulation of disaster evolution and power grid response, and providing a forward-looking and quantitative scientific basis for disaster prevention decision-making.

[0119] It should be noted that the dynamic simulation layer 102 has a complete ability to simulate the power grid's operating state, such as... Figure 3As shown, the specific implementation is as follows:

[0120] The real-time data stream input from the perception layer uses a lightweight machine learning classification algorithm to quickly identify and classify the current disaster. Based on the classification results, the system matches a set of optimal models from a pre-set physical behavior model library, enabling it to autonomously select the optimal physical law expression form according to the disaster type characteristics.

[0121] The simulation engine first calculates the global disaster intensity field of the environmental field data. Using spatial interpolation algorithms (such as Kriging interpolation), the field strength is mapped to the geographic coordinates of each specific device, serving as the input load for the device-level model. The device-level model then calculates the failure probability or state. Once a device "fails," the event is converted into an event command by the power grid analysis software. Upon receiving this command, the software performs power flow calculations, dynamic security analysis, and transient stability analysis, outputting the operating status, stability indicators, and potential cascading failure paths. Finally, through a collaborative simulation platform (such as the FMU / FMI standard), data exchange and collaborative calculation between the device's micro-response and the power grid's macro-state are achieved, resulting in a time-series map of device failure risk, power grid migration paths, and system fault early warning information.

[0122] Model downscaling involves pre-training a downscaled model for the high-fidelity complex model. When the system detects insufficient computing resources, it automatically switches from the high-fidelity model to a downscaled proxy model that is 1-2 orders of magnitude faster, trading speed for necessary accuracy loss. Dynamic computing resource scheduling monitors CPU / memory / network load, automatically requesting more computing nodes from the resource management platform in case of overload. Redundancy and backup utilize hot or cold backup strategies for the primary and backup nodes of the inference engine. If the primary node fails, the backup node can take over the task within seconds and restore the inference state from the latest checkpoint, ensuring the continuity of the inference process. These three methods guarantee the sustainability of core inference functions under extreme scenarios.

[0123] The extrapolation layer outputs a structured predictive system, including a time-series map of equipment failure risks, identifying the locations and failure probabilities of high-risk facilities at different times; a set of power grid state transition paths, describing the sequence of operational mode changes that may result from fault evolution; and a system collapse critical point early warning system, indicating key weak points that may trigger cascading failures. The capability boundaries of this layer have been system-verified, with the fault chain prediction time span extended to a significant period before the disaster occurs, the accuracy of critical equipment risk identification meeting proactive defense requirements, and the calculation timeliness controlled within the decision-making window. These characteristics provide predictive support for the formulation of response strategies, enabling the technical system to transition from passive response to proactive intervention.

[0124] In one example of the invention, a prediction correction system is also included, specifically for:

[0125] Based on the difference between the predicted tidal current and the actual observation results, the model parameters of the target physical model are optimized iteratively in reverse.

[0126] In this embodiment of the invention, data assimilation techniques (such as Kalman filtering) or Bayesian update methods are used to calculate the difference between the predicted value and the observed value, reverse the calculation and adjust the key uncertain parameters in the model, and then continuously optimize the confidence of the model parameters.

[0127] In one example of the present invention, the decision execution layer 103 is specifically used for:

[0128] Security constraints are constructed based on the predicted migration path of the power grid state, and constraint conditions are created by combining resource constraints and time constraints.

[0129] Based on the time series diagram of equipment failure risk, risk items are constructed for the objective decision function, and cost and time items are superimposed to create the objective decision function;

[0130] With the goals of minimizing risk, cost, and time, a resource scheduling scheme is generated by using a genetic algorithm or particle swarm optimization algorithm to find the approximate optimal solution of the objective decision function under constraints within a preset solution period.

[0131] Dispatch instructions are issued to the corresponding target terminals in accordance with the resource scheduling plan; the dispatch instructions include the resource scheduling spatiotemporal trajectory diagram, the status operation sequence and the cross-organizational collaboration protocol.

[0132] Genetic algorithms, belonging to the metaheuristic category, search for an approximate optimal solution to the objective decision function under constraints within a preset solution period by simulating the biological process of genetic evolution (selection, crossover, and mutation), thus adapting to multi-objective optimization needs. Particle swarm optimization (PSO) is also a metaheuristic algorithm. By simulating the flight and cooperation of particles in space, it dynamically adjusts the search direction to find an approximate optimal solution to the objective decision function under constraints within a preset solution period, exhibiting fast convergence speed.

[0133] Resource scheduling spatiotemporal trajectory diagrams are used to describe the movement paths and corresponding time nodes of resources such as personnel and equipment in physical space, clarifying the spatial and temporal requirements of resource scheduling. Status operation sequences refer to sequences of operation instructions used to specify the operation steps of power grid equipment (such as switch opening and closing, load transfer sequence), ensuring that power grid status adjustments comply with safe operation standards. Cross-agency collaboration protocols are used to clarify the division of tasks, handover nodes, and cooperation rules between power grid companies and external agencies such as fire departments and medical institutions, enabling collaborative scheduling of resources across different domains.

[0134] In this embodiment of the invention, safety constraints are constructed based on the evolution of power grid operation modes in the predicted migration path of power grid status. Simultaneously, resource constraints and time constraints are constructed by combining resource availability and key task completion deadlines, integrating them to form complete constraint conditions. The objective decision function creation module constructs risk terms based on the equipment failure probability and loss coefficient at different time periods in the equipment failure risk time series graph. It then superimposes cost terms corresponding to resource usage costs and time terms corresponding to task execution time to create an objective decision function with the goal of "minimum risk, lowest cost, and shortest time." Subsequently, the optimization and solution module calls the built-in genetic algorithm or particle swarm optimization algorithm, inputting the constraints and objective decision function, and searches for an approximately optimal solution that satisfies the constraints within a preset solution period to generate a resource scheduling scheme. Finally, the instruction generation and issuance module parses the resource scheduling scheme into scheduling instructions containing a resource scheduling spatiotemporal trajectory graph (clarifying resource movement paths and times), a state operation sequence (specifying power grid equipment operation steps), and a cross-agency collaboration protocol (clarifying cross-domain collaboration rules). These instructions are then issued to the corresponding target terminals (such as emergency repair team terminals and cross-agency collaboration terminals) through an encrypted communication network, completing the transformation from predicted data to entity execution instructions.

[0135] The objective decision function is:

[0136]

[0137] in, This is a risk item. For cost items, For time terms, As the weight of risk items, As the weight of cost items, Let X be the weight of the time term, X be the resource scheduling scheme, and t be the time point.

[0138] The risk item measures the remaining risk exposure, as follows:

[0139]

[0140] Let be the failure probability of component b at time t under scheme X (derived from the inference layer risk map).

[0141] This is the loss factor (such as load loss, social impact) caused by the failure of component b.

[0142] This minimization directly corresponds to the "risk minimization" objective.

[0143] The cost item is the total execution cost, as detailed below:

[0144]

[0145] The fixed cost of initiating action i, The variable cost of using one unit of resource k.

[0146] The time field represents the total task completion time, i.e., the total time elapsed from the start of the first action to the end of the last action. Details are as follows:

[0147]

[0148] The specific constraints are expressed as follows:

[0149] Resource constraints ensure that the resources consumed at any given time do not exceed the total available resources.

[0150]

[0151] Let i be the amount of resources k (such as manpower, special vehicles, drones, spare parts) occupied by action i at time t. Let K be the total amount of resource k available at time t.

[0152] Time pressure constraints are used to require a critical action to be completed before a certain absolute time.

[0153]

[0154] Let be the end time of action i (decision variable). The final absolute time that action i must complete. This is the collection of all key actions.

[0155] Safety boundary constraints are used to ensure that no operation puts the system in a dangerous state:

[0156]

[0157] Let X be the state of the system at time t under the influence of sequence X (predicted by the extrapolation layer model). (For thermal stability limits and node voltages within acceptable range).

[0158] In the specific solution process, the aforementioned constraints and objective function are formalized into a mixed-integer programming or constraint satisfaction problem. Due to the real-time requirements of disaster response, metaheuristic algorithms (such as genetic algorithms and particle swarm optimization) or dedicated scheduling algorithms are typically used to find an approximate optimal solution within a finite time. After environmental changes (such as new simulation results or resource depletion), the system will re-trigger the above solution process, updating the action sequence online to ensure it always remains on the optimal or suboptimal trajectory. By generating the optimal action sequence through the above mathematical modeling, a structured and optimal resource scheduling spatiotemporal trajectory diagram and state change instruction set are finally output, realizing closed-loop control from digital cognition to physical action. By establishing a dynamic mapping relationship between virtual strategies and real-world execution, the execution engine transforms abstract decisions into concrete instruction flows, driving the directional flow of physical elements such as personnel, equipment, and materials. The instruction system has built-in adaptive adjustment capabilities to ensure that the action plan continues to maintain its optimal state as the environment evolves, forming a precise projection from digital space to physical space.

[0159] The decision-making execution layer 103 achieves flexible scaling of action sequences in the time dimension, dynamically compressing or extending the operation time window according to the disaster evolution rate; in the spatial dimension, it establishes a dynamic matching model of resource distribution and demand hotspots, eliminating response gaps in cross-regional coordination; in the efficiency dimension, it integrates the dual objectives of efficiency optimization and risk minimization, generating a balanced solution that takes into account both speed and safety. The synergy between this layer and the closed-loop evolution layer is manifested in a knowledge feedback relationship: the full-element tracking data of the execution process drives the continuous optimization of decision rules, and the evolutionary requirements reverse-calibrate the execution accuracy threshold. It autonomously interrupts the original instruction flow and initiates the generation of alternative solutions based on environmental mutation signals; a multi-agency collaborative framework achieves cross-organizational boundary action sequence synchronization and seamless connection of responsibility areas; an efficiency evaluation system monitors the deviation of actions in real time through multi-dimensional indicators and triggers correction processes; and a robust control design ensures the mandatory implementation capability of core instructions under extreme conditions. It outputs a structured intervention system: a resource scheduling spatiotemporal trajectory map, describing the movement paths and time nodes of personnel and equipment in physical space; a state change instruction set, specifying the operation sequence and parameter adjustment values ​​of power grid facilities; and a collaborative action protocol, clarifying the cross-agency task handover points and joint operation specifications. The capability boundaries of this layer have been verified by the system: the timeliness of action plan generation meets the disaster response window requirements, the resource scheduling path optimization rate breaks through the bottleneck of traditional methods, and the command error execution rate is reduced to below an acceptable threshold. These characteristics provide an action blueprint for proactive disaster intervention, enabling the technical system to have a closed-loop capability from risk prediction to effective control, such as... Figure 4 As shown.

[0160] This invention provides a power grid emergency management system, comprising a smart sensing layer, a dynamic inference layer, and a decision execution layer connected by communication. The smart sensing layer collects multi-source raw data through a space-air-ground monitoring network and performs spatiotemporal calibration to generate standardized data when a disaster occurs in the environment where power grid equipment is located. It then extracts key features from the standardized data and performs preliminary risk assessment, generating disaster intensity field data and power grid equipment status data. The dynamic inference layer matches a target physical model from a physical behavior model library and performs power flow prediction based on the disaster intensity field data and power grid equipment status data, generating a time-series map of equipment failure risk and a predicted migration path for power grid status. The decision execution layer constructs and solves a target decision function based on the time-series map of equipment failure risk and the predicted migration path for power grid status, generating a resource scheduling scheme and outputting it to the corresponding target terminal. By constructing a closed-loop system of perception-inference-decision-evolution through these layers, the response mechanism shifts from reactive handling to proactive control, enabling more efficient and accurate power grid emergency resource scheduling.

[0161] Please see Figure 5 , Figure 5 A flowchart illustrating the steps of a power grid emergency management method according to an embodiment of the present invention is shown.

[0162] This embodiment provides a power grid emergency management method, applied to the power grid emergency management system of any of the above embodiments, the method comprising:

[0163] Step 501: When a disaster occurs in the environment where the power grid equipment is located, the intelligent sensing layer calls the air-space-ground monitoring network to collect multi-source raw data and performs spatiotemporal calibration to generate standardized data.

[0164] Step 502: Extract key features from the standardized data through the intelligent sensing layer and make a preliminary risk assessment to generate disaster intensity field data and power grid equipment status data;

[0165] Step 503: Based on the disaster intensity field data and power grid equipment status data, the dynamic inference layer matches the target physical model from the physical behavior model library and performs power flow prediction to generate a time series map of equipment failure risk and a power grid status prediction migration path.

[0166] Step 504: The decision execution layer constructs and solves the target decision function based on the equipment failure risk time series map and the power grid state prediction migration path, generates a resource scheduling scheme, and outputs it to the corresponding target terminal.

[0167] In this embodiment of the invention, when a disaster occurs in the environment where the power grid equipment is located, the intelligent sensing layer invokes an air-ground monitoring network composed of monitoring satellites, drone formations, and IoT sensor networks with multimodal hybrid networking communication capabilities. First, the monitoring satellites periodically scan and collect macroscopic meteorological data and locate risk hotspots. Then, the drone formations are invoked to collect regional microscopic data in the risk hotspots, and the IoT sensor network in the area is simultaneously activated to collect second-level device data. During this process, the self-maintenance device periodically diagnoses sensors and uses calibration models to correct abnormal sensors online. Afterwards, the three types of data undergo spatiotemporal calibration and standardization to generate standardized data. The intelligent sensing layer uses principal component analysis to extract key features containing power grid equipment status data from the standardized data, generates disaster intensity field data through Bayesian inference, and simultaneously constructs a panoramic view of disaster risk. The dynamic inference layer first calculates the power grid equipment status data... The classification results are matched with the target physical model from the physical behavior model library. Then, spatial interpolation is performed on the disaster intensity field data to obtain the input load. The target physical model is called in conjunction with the multiphysics coupling engine and system-level cascading failure simulation to carry out power flow prediction. The location and probability of equipment failure at different time periods are determined to construct the equipment failure risk time series map. Combined with equipment change strategy, the time series evolution is simulated to generate the power grid state prediction migration path. The decision execution layer constructs safety constraints based on the power grid state prediction migration path, and forms constraint conditions by combining resource and time constraints. Based on the equipment failure risk time series map, the risk term of the target decision function is constructed and the cost term and time term are superimposed. The genetic algorithm or particle swarm algorithm is used to solve the approximate optimal solution within a preset period to generate a resource scheduling scheme. Finally, the scheme is parsed into a scheduling instruction containing a resource scheduling spatiotemporal trajectory map, state operation sequence and cross-agency coordination protocol, and sent to the corresponding target terminal through an encrypted channel.

[0168] In the embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between systems or modules may be electrical, mechanical, or other forms.

[0169] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0170] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0171] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A power grid emergency management system, characterized in that, It includes an intelligent sensing layer, a dynamic inference layer, and a decision execution layer for communication connectivity; The intelligent sensing layer is used to collect multi-source raw data through the air-space-ground monitoring network and perform spatiotemporal calibration to generate standardized data when a disaster occurs in the environment where the power grid equipment is located; extract key features in the standardized data and make a preliminary risk assessment to generate disaster intensity field data and power grid equipment status data. The dynamic simulation layer is used to match the target physical model from the physical behavior model library and perform power flow prediction based on the disaster intensity field data and the power grid equipment status data, and generate equipment failure risk time series map and power grid status prediction migration path. The decision execution layer is used to construct and solve the target decision function based on the equipment failure risk time series map and the power grid state prediction migration path, generate a resource scheduling scheme, and output it to the corresponding target terminal.

2. The system according to claim 1, characterized in that, It also includes a closed-loop evolution layer, specifically used for: Calculate the numerical deviation between the predicted target result and the actual target result corresponding to the resource scheduling scheme; The distribution of the contribution of positioning strategy deviation is determined by multi-scale evaluation based on the numerical deviation. Based on the distribution of the policy deviation contribution and historical data, feedback adjustment instructions are generated for the intelligent perception layer, the dynamic inference layer, and the decision execution layer, and then transmitted.

3. The system according to claim 1, characterized in that, The air-space-ground monitoring network includes monitoring satellites with multimodal hybrid networking communication, UAV swarms, and an IoT sensor network. The multi-source raw data includes macro-meteorological data, regional micro-data, and second-level device data. The intelligent sensing layer is specifically used for: When a disaster occurs in the environment where the power grid equipment is located, the monitoring satellite periodically scans the macro-meteorological data of the area where the power grid equipment is located to locate risk hotspots. The drone formation is invoked to move to the risk hotspot area and periodically collect regional micro data; The system utilizes an IoT sensor network within the aforementioned risk hotspot area to collect device data at the second-level. After performing spatiotemporal calibration and standardization on the macro-meteorological data, the regional micro-data, and the second-level equipment data, standardized data is generated. Principal component analysis was performed on the standardized data to extract key features; The key features include at least power grid equipment status data; Based on the aforementioned key features, Bayesian inference is performed to generate disaster intensity field data.

4. The system according to claim 3, characterized in that, The intelligent sensing layer is also equipped with a self-maintenance device, specifically used for: Periodically diagnose whether any abnormalities occur in the various IoT sensors within the IoT sensor network; If so, the calibration model is invoked to perform online correction on the IoT sensor.

5. The system according to claim 1, characterized in that, The dynamic deduction layer is specifically used for: The power grid equipment status data is classified, and the target physical model is matched from the physical behavior model library according to the classification results. Spatial interpolation is performed based on the disaster intensity field data, and the data is mapped to the location of each target physical model as input load; The target physical model is used to predict the power flow of the input load. Based on the power flow prediction results, the location and probability of failure of the equipment in different time periods are determined, and a time series map of equipment failure risk is constructed. Based on the location of each failed device and the device change strategy, a time-series evolution is performed to generate a power grid state prediction migration path.

6. The system according to claim 5, characterized in that, It also includes a predictive correction system, specifically used for: Based on the difference between the predicted power flow results and the actual observation results, the model parameters of the target physical model are iteratively optimized in reverse.

7. The system according to claim 1, characterized in that, The decision execution layer is specifically used for: Security constraints are constructed based on the predicted migration path of the power grid state, and constraint conditions are created by combining resource constraints and time constraints. Based on the aforementioned equipment failure risk time series map, risk terms of the target decision function are constructed, and cost and time terms are superimposed to create the target decision function; With the objectives of minimizing risk, cost, and time, a resource scheduling scheme is generated by using a genetic algorithm or particle swarm optimization algorithm to find the approximate optimal solution of the objective decision function under the constraints within a preset solution period. According to the resource scheduling scheme, scheduling instructions are issued to the corresponding target terminals; the scheduling instructions include a resource scheduling spatiotemporal trajectory diagram, a state operation sequence, and a cross-agency collaboration protocol.

8. The system according to claim 7, characterized in that, The objective decision function is: in, This is a risk item. For cost items, For time terms, As the weight of risk items, As the weight of cost items, Let X be the weight of the time term, X be the resource scheduling scheme, and t be the time point.

9. The system according to claim 1, characterized in that, The intelligent sensing layer is also used to generate a panoramic view of disaster risks based on the standardized data.

10. A power grid emergency management method, characterized in that, The method, applied to the power grid emergency management system according to any one of claims 1 to 9, comprises: When a disaster occurs in the environment where the power grid equipment is located, the intelligent sensing layer calls the air-space-ground monitoring network to collect multi-source raw data and performs spatiotemporal calibration to generate standardized data. The intelligent sensing layer extracts key features from the standardized data and performs preliminary risk assessment to generate disaster intensity field data and power grid equipment status data. Based on the disaster intensity field data and the power grid equipment status data, the dynamic inference layer matches the target physical model from the physical behavior model library and performs power flow prediction to generate a time series map of equipment failure risk and a power grid status prediction migration path. The decision execution layer constructs and solves the target decision function based on the equipment failure risk time series map and the power grid state prediction migration path, generates a resource scheduling scheme, and outputs it to the corresponding target terminal.

Citation Information

Cited By

  • Multi-disaster-type disaster chain-oriented sky-ground cooperative monitoring task planning method, apparatus and device, and storage medium

    CN122047944A