Beidou fusion multi-source data power weather disaster early warning device

Abstract

The application discloses a Beidou fusion multi-source data power weather disaster early warning device, and belongs to the technical field of power weather disaster early warning, which comprises a multi-source data sensing and fusion layer, a Beidou and AI disaster early warning analysis layer, an emergency dispatch decision and optimization layer and a visual display and operation guarantee layer. The Beidou fusion multi-source data power weather disaster early warning device realizes real-time collection of multi-source data such as Beidou positioning, micro-meteorological sensors and meteorological satellites, input of the multi-source data into an AI early warning model after space-time unification, calculation of disaster probability and risk level such as icing and wind deflection for each tower, generation of refined early warning and automatic matching of repair schemes, realization of accurate tower-level disaster early warning and a significant reduction of repair response time. The device relies on Beidou short message to guarantee smooth communication under extreme weather and avoid blind line patrol, optimizes resource dispatch paths, effectively reduces the range and duration of power failure, reduces economic losses and improves the disaster resistance of the power grid.

Description

Technical Field

[0001] This invention relates to the field of power meteorological disaster early warning technology, specifically a power meteorological disaster early warning device that integrates BeiDou multi-source data. Background Technology

[0002] Power meteorological disasters can cause transmission line breaks, tower collapses, and damage to transformer equipment, leading to widespread power outages. This disrupts residents' lives, industrial production, and the operation of core urban functions such as transportation and communication. Furthermore, torrential rains and floods can block repair roads, prolonging power restoration time and causing direct and indirect economic losses, including industrial shutdowns, equipment replacement, and emergency repair costs. Therefore, early warning systems for power meteorological disasters are necessary. These warnings can identify risks in advance, allowing time for equipment reinforcement and personnel evacuation, reducing equipment damage and casualties. They also assist dispatching departments in deploying repair forces in advance, optimizing grid operation, shortening outage time, and minimizing economic losses. This necessitates the use of power meteorological disaster early warning systems, such as the Chinese invention patent application No. 202211739217.5, filed on December 30, 2022, which proposes a power meteorological early warning method based on refined forecasting. This method comprehensively considers the probability of power meteorological disasters and their risk severity. Finally, it establishes a power meteorological disaster early warning system based on refined meteorological forecast data, enabling regional monitoring of the power grid. The system provides early warnings of power meteorological disaster risk levels through progressive updates, laying a technical foundation for scientifically assessing the impact of power meteorological disasters. This is significant for rapidly initiating emergency responses to power meteorological disasters, improving power grid reliability, and enhancing the grid's disaster prevention and mitigation capabilities. Furthermore, Chinese invention patent application number 202411135143.3, filed on August 19, 2024, describes an intelligent power facility meteorological disaster risk early warning system. This system utilizes sensor networks, automatic weather stations, and remote sensing technology to monitor and collect various meteorological data related to the area where power facilities are located in real time. It employs an improved Transformer generative model to comprehensively analyze the meteorological data and generate a disaster prediction model, improving the real-time performance and accuracy of meteorological disaster prediction. Simultaneously, a multi-scale local attention mechanism is added to the improved Transformer generative model to reduce the influence of irrelevant information and capture the hierarchical dependencies between different factors, ensuring timely and accurate early warnings before meteorological disasters occur, protecting power infrastructure from damage.

[0003] During system operation, due to significant differences in terrain, microclimate, and structural strength among different towers, the risk of damage varies under the same meteorological conditions. Therefore, it is necessary to refine the early warning granularity to the level of a single tower and link it with its design disaster resistance parameters and real-time micro-meteorological data. Without the ability to conduct differentiated risk assessments at the tower level, high-risk towers may be missed while low-risk towers may over-respond, reducing the accuracy of emergency repair resource allocation, preventing priority treatment of truly severe disaster locations, expanding the power outage area, and significantly extending the recovery time. Summary of the Invention

[0004] The purpose of this invention is to provide a power meteorological disaster early warning device that integrates BeiDou multi-source data to solve the problems mentioned in the background art, such as the lack of pole-level differentiated risk assessment capabilities, which leads to high-risk poles being missed while low-risk poles over-respond, reduced accuracy of emergency repair resource allocation, failure to prioritize the truly severe disaster locations, expanded power outage areas, and significantly prolonged recovery time.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a power meteorological disaster early warning device integrating BeiDou multi-source data, comprising a multi-source data perception and fusion layer, a BeiDou and AI disaster early warning analysis layer, an emergency dispatch decision and optimization layer, and a visualization and operation and maintenance support layer. When the multi-source data perception and fusion layer is working, it solves the problem of data silos and achieves integrated data acquisition from space, ground, and air. The BeiDou and AI disaster early warning analysis layer performs short-term early warning, trend prediction, and risk quantification based on the fused data. The emergency dispatch decision and optimization layer serves as the action center, transforming early warnings into executable and traceable dispatch plans, particularly suitable for large-scale disaster scenarios. The visualization and operation and maintenance support layer provides a visual interface for disaster situation and dispatch status, and monitors the system's own operational health.

[0006] Preferably, the multi-source data perception and fusion layer includes a BeiDou satellite real-time observation submodule and a meteorological multi-element perception submodule. The BeiDou satellite real-time observation submodule realizes the BeiDou high-precision positioning function and also has the BeiDou short message communication function. When 4G and 5G on the ground are interrupted, it serves as an emergency data backhaul channel. It also has the BeiDou time synchronization function to provide a unified time reference for all network devices and ensure the consistency of time-series data.

[0007] Preferably, the multi-source data sensing and fusion layer further includes a power private network and Internet of Things submodule and an external data access submodule;

[0008] The power grid and Internet of Things submodule monitors conductor galloping and ice thickness through micro-meteorological sensors on transmission lines, while the external data access submodule receives numerical weather forecasts from the meteorological bureau and accesses the database of geological disaster hazard points and historical disaster case databases.

[0009] Preferably, the BeiDou and AI disaster early warning analysis layer includes a disaster correlation submodule based on the BeiDou spatiotemporal reference, a multi-hazard intelligent identification submodule, a disaster impact prediction and risk thermodynamics submodule, and an early warning credibility assessment submodule;

[0010] The disaster correlation submodule based on the BeiDou spatiotemporal reference incorporates a spatiotemporal alignment engine, unifying meteorological data, geographic data, and power grid equipment coordinates to the BeiDou coordinate system. It also performs dynamic and static overlay analysis of disasters, combining dynamic meteorological disasters with static equipment vulnerability analysis, involving formulas... , Indicates the target coordinate system CGCS2000 coordinate, Represents the original coordinate system of BeiDou coordinate, Indicates that the origin of the two coordinate systems is at... Translation amount in direction, This represents the rotation angle between the two coordinate systems. Represents the original coordinate system of BeiDou coordinate, The origins of the two coordinate systems are at The amount of translation in the direction;

[0011] The multi-hazard intelligent identification submodule includes a rainstorm and flood early warning subunit for establishing a correlation model between water depth and substation terrain. It also includes a gale and typhoon early warning subunit, which matches wind speed with tower wind resistance levels during operation. Additionally, it includes a lightning early warning subunit, which integrates atmospheric electric field analysis and BeiDou ionospheric disturbance data. Finally, it incorporates a formula for estimating conductor icing thickness. In the formula Indicates equivalent icing thickness, in millimeters. This indicates the current measured tension of the conductor, in Newtons. This represents the reference tension of the conductor in the absence of ice and wind, measured in Newtons. This represents the coefficient indicating the effect of wind speed on tension. This indicates the current wind speed, in m / s. Indicates the density of ice. It is the acceleration due to gravity. Indicates the conductor span;

[0012] It also includes a built-in formula for calculating tower wind load. , This represents the wind load on the tower, measured in Newtons. air density, Wind speed (m / s) The windward area of ​​the tower (m) 2 , This is the shape drag coefficient.

[0013] Preferably, when the disaster impact prediction and risk thermodynamics submodule is working, it calculates the probability of damage to power grid equipment based on fault tree and Bayesian network using the equipment failure probability formula. The formula is as follows: ,in, Let be the probability of failure within time t. The duration of exposure to the disaster, in hours. For characteristic lifetime parameters, in hours. For shape parameters, Using a natural exponential function and combining geographic information, the system predicts potential power outage areas. Finally, it automatically generates a visual risk level using a risk level formula—red, orange, yellow, and blue. The risk level formula is as follows: ,in, Overall risk value, Equipment failure probability, Severity of failure consequences;

[0014] The early warning reliability assessment submodule performs multi-model consistency verification during operation, and continuously optimizes threshold judgment based on historical feedback, performing self-learning correction for false alarms and false negatives. This involves multi-model consistency verification formulas. , The Pearson correlation coefficient between the outputs of the two models. For the model In the The output value of each sample point is the model. The arithmetic mean of all output values ​​is the model's... In the The output value of each sample point is the model. The arithmetic mean of all output values.

[0015] Preferably, the emergency dispatch decision and optimization layer includes an optimal dispatch submodule for repair resources, a load transfer and islanded operation submodule, a multi-department collaborative command submodule, and a dynamic reassessment submodule during the disaster;

[0016] The optimal scheduling submodule for emergency repair resources operates by managing the real-time location of repair teams, vehicles, and drones based on BeiDou positioning. It also incorporates a path planning algorithm for dynamic obstacle avoidance in cases of road flooding and landslides. Following this, it performs material matching—automatically recommending a list of necessary materials based on the type of disaster damage—which involves Dijkstra's algorithm formula. ,in From the starting point to the node The current shortest distance, in kilometers. From the starting point to the node The known shortest distance, in kilometers. For nodes To the node The actual travel distance between the roads, The node number is the one for which the shortest path has been determined. Number the adjacent nodes to be updated;

[0017] When the load transfer and islanding operation submodule is working, it generates priority supply plans for important loads such as hospitals, communications, and flood control facilities based on the microgrid and distributed power source rapid disconnection and grid connection strategies.

[0018] Preferably, the multi-department collaborative command submodule has a one-click emergency command issuance function, which can be implemented via Beidou short message, SMS, and voice broadcast. It also has a function for visual management of task execution status and is equipped with reserved interfaces for government emergency offices, meteorological bureaus, and transportation departments.

[0019] The disaster dynamic reassessment submodule has the function of refreshing the early warning level every 15 minutes during the disaster development process, and it performs rolling corrections of the scheduling plan based on the latest real-time data during operation, involving a smooth update formula for the dynamic early warning index. ,in For the current moment Smoothed warning level values For smoothing coefficients, For the current moment The instantaneous warning level value calculated from measured data. This is the smoothed warning level value from the previous moment. Current time step, every 15 minutes.

[0020] Preferably, the visualization and operation and maintenance support layer belongs to the human-computer interaction layer, including a three-dimensional digital sandbox module, a mobile terminal command sentinel sub-module, a system health self-check sub-module, and a post-disaster review and knowledge base sub-module;

[0021] When the three-dimensional digital sand table module is working, it displays an overlay map of power grid, meteorology, topography, and densely populated areas based on GIS and Beidou high-precision maps, while simultaneously simulating disaster evolution animations and supporting timeline drag-and-drop playback.

[0022] The mobile command sentry submodule is a lightweight APP for on-site repair personnel. It receives early warnings, reports on-site photos and videos, and supports offline mode, sending and receiving short messages via Beidou short message service when there is no signal.

[0023] Preferably, the system health self-check submodule performs data source quality monitoring during operation, including outlier detection, latency and packet loss rate detection, and monitors the model running status, including CPU and GPU load and inference time.

[0024] Preferably, the post-disaster review and knowledge base submodule generates a full-chain time-series report of the disaster situation, covering early warning, response, and recovery, while automatically extracting typical cases and updating the emergency plan library.

[0025] Compared with existing technologies, the beneficial effects of this invention are as follows: It adopts a novel structural design, collects multi-source data in real time from BeiDou positioning, micro-meteorological sensors, and meteorological satellites, and inputs this data into an AI early warning model after spatiotemporal unification. It calculates the probability and risk level of disasters such as icing and wind deflection on a tower-by-tower basis, generates refined early warnings, and automatically matches repair plans, achieving precise early warning of tower-level disasters and significantly shortening repair response time. It relies on BeiDou short message service to ensure uninterrupted communication under extreme weather conditions, avoiding blind line inspections. It optimizes resource scheduling paths, effectively reducing the scope and duration of power outages, minimizing economic losses, and enhancing the disaster resilience of the power grid. The specific details are as follows:

[0026] (1) This BeiDou-integrated multi-source data power meteorological disaster early warning device, when the multi-source data sensing and fusion layer is working, achieves wide coverage and high-density real-time collection of power grid equipment status and surrounding meteorological environment by integrating BeiDou high-precision positioning, micro-meteorological sensors, radar satellite cloud images and Internet of Things monitoring equipment. The beneficial effects of this module are: it breaks the limitations of traditional single data source, provides a complete and reliable data foundation for the upper-level early warning model, and solves the problem of data back transmission when ground communication is interrupted by BeiDou short message service, ensuring that information is not interrupted under extreme weather conditions, and laying a solid foundation for subsequent accurate early warning and scientific scheduling.

[0027] (2) The power meteorological disaster early warning device that integrates Beidou multi-source data, when working with the Beidou and AI disaster early warning analysis layer, unifies multi-source data through the Beidou high-precision spatiotemporal reference, and calculates the probability and risk level of disasters such as icing, wind deflection and flooding on a pole-by-pole basis by AI model, thereby raising the early warning from the regional level to the pole-tower level. This effectively solves the problems of ambiguous location and delayed timeliness in traditional forecasts, significantly reduces the false alarm and missed alarm rates, provides an accurate and reliable risk list and priority ranking for emergency repair decisions, reduces blind line inspection and ineffective dispatch from the source, and compresses the disaster response window period to the greatest extent.

[0028] Furthermore, during emergency dispatch decision-making and optimization, based on BeiDou real-time positioning and dynamic road network data, the system automatically calculates the optimal route for repair teams and intelligently matches the required materials and equipment, avoiding detours and delays caused by road flooding or landslides. Simultaneously, it rapidly generates load transfer and supply guarantee plans based on disaster damage assessment results, ensuring priority power restoration for critical users such as hospitals and telecommunications companies. A rolling reassessment mechanism continuously tracks the disaster's development, dynamically corrects dispatch instructions, significantly shortens power outage restoration time, reduces secondary losses, and improves overall repair efficiency.

[0029] (3) The Beidou-integrated multi-source data power meteorological disaster early warning device, when the operation and maintenance support layer is working, will present information such as disaster early warning, equipment status, repair progress and resource location in a three-dimensional digital sand table, which will help decision-makers quickly grasp the overall situation and reduce the time spent on information reading and analysis; at the same time, it will provide mobile terminal on-site operation support to realize closed-loop management of instruction issuance and on-site feedback; the system health self-check function will continuously monitor data quality and model operation status to ensure the reliability of early warning results; after the disaster, it will automatically generate a full-link review report to provide data support for plan optimization and experience accumulation. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the module structure of the present invention. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] Please see Figure 1 The present invention provides the following technical solution: a power meteorological disaster early warning device that integrates Beidou multi-source data.

[0033] It includes a multi-source data perception and fusion layer, a BeiDou and AI disaster early warning and analysis layer, an emergency dispatch decision-making and optimization layer, and a visualization and operation and maintenance support layer:

[0034] When the multi-source data sensing and fusion layer is working, it solves the problem of data silos and realizes integrated data acquisition from space, ground, and air. The multi-source data sensing and fusion layer includes a BeiDou satellite real-time observation submodule and a meteorological multi-element sensing submodule. The BeiDou satellite real-time observation submodule realizes the BeiDou high-precision positioning function and also has the BeiDou short message communication function. When terrestrial 4G and 5G are interrupted, it serves as an emergency data backhaul channel. It also has the BeiDou time synchronization function, providing a unified time reference for all network devices and ensuring the consistency of time-series data.

[0035] The multi-source data sensing and fusion layer also includes a power private network and Internet of Things sub-module and an external data access sub-module;

[0036] The power grid and Internet of Things submodule monitors conductor galloping and ice thickness through micro-meteorological sensors on transmission lines, while the external data access submodule receives numerical weather forecasts from the meteorological bureau and accesses the database of geological disaster hazard points and historical disaster case databases.

[0037] Specifically, during the operation of the multi-source data sensing and fusion layer, the status of power grid equipment and the surrounding meteorological environment are collected in real time through an integrated air-space-ground sensing network. Leveraging the spatiotemporal reference and short message capability of the BeiDou system, all heterogeneous data is cleaned and fused under a unified coordinate system and time label. First, relying on BeiDou's high-precision positioning and timing functions, each tower, substation, and mobile terminal is assigned a unique spatiotemporal identifier to ensure that all monitoring data have a consistent spatial location and absolute time reference. At the sensing end, a multi-layer sensor network is deployed: micro-meteorological sensors are installed on the lines to measure wind speed, wind direction, temperature, humidity, and air pressure; tension sensors measure icing and galloping; and tilt sensors measure tower deformation. Water level gauges and partial discharge monitoring equipment are deployed in substations. Simultaneously, data from meteorological radar, satellite cloud images, and regional automatic weather stations are accessed, taking into account both macro-weather evolution and micro-local abrupt changes. After collection, the data enters the fusion stage, passing through a unified access gateway. The system performs protocol conversion and preliminary quality control to eliminate obvious outliers. It uses BeiDou time synchronization to interpolate asynchronous observation data, unifying them to the same time segment. Spatially, it uses a seven-parameter coordinate transformation model to convert WGS84 or latitude and longitude coordinates to the CGCS2000 coordinate system and precisely matches them with the tower coordinates in the power grid GIS database, linking each meteorological data point to a specific device. For extreme cases of ground communication interruption, this layer integrates BeiDou short message bidirectional communication as a backup channel. If 4G or fiber optic fails, critical on-site status data automatically switches to short message transmission back to the command center, ensuring the system remains operational. Processed data is cached at high speed in an in-memory database and pushed to the upper-level early warning module on demand via a publish-subscribe model. It is also persisted to a historical database for subsequent training and traceability. Data quality monitoring is implemented throughout the entire process, with real-time statistics on latency, packet loss rate, and outlier ratios. Inferior data is automatically marked for reliability and triggers alarms. This layer transforms discrete and heterogeneous raw observations into an ordered and reliable standardized data stream, providing a real-time and high-precision data foundation for disaster identification and scheduling at the upper level. It is the fundamental guarantee for the effective operation of the entire system.

[0038] The BeiDou-AI disaster early warning and analysis layer performs short-term early warning, trend prediction, and risk quantification based on fused data;

[0039] The BeiDou and AI disaster early warning analysis layer includes a disaster correlation submodule based on the BeiDou spatiotemporal reference, a multi-hazard intelligent identification submodule, a disaster impact prediction and risk thermodynamics submodule, and an early warning credibility assessment submodule.

[0040] The disaster correlation submodule based on the BeiDou spatiotemporal reference incorporates a spatiotemporal alignment engine, unifying meteorological data, geographic data, and power grid equipment coordinates to the BeiDou coordinate system. It also performs dynamic and static overlay analysis of disasters, combining dynamic meteorological disasters with static equipment vulnerability analysis, involving formulas... , Indicates the target coordinate system CGCS2000 coordinate, Represents the original coordinate system of BeiDou coordinate, Indicates that the origin of the two coordinate systems is at... Translation amount in direction, This represents the rotation angle between the two coordinate systems. Represents the original coordinate system of BeiDou coordinate, The origins of the two coordinate systems are at The amount of translation in the direction;

[0041] The aforementioned disaster correlation submodule based on the BeiDou spatiotemporal reference uses the high-precision spatial coordinates and unified time labels provided by the BeiDou system to accurately correlate meteorological data, power grid equipment data, and environmental monitoring data in the spatiotemporal dimension. This ensures that each early warning message can be assigned to a specific tower or line section. In the spatial dimension, this submodule first establishes a unified benchmark spatial database for power grid equipment. Using carrier phase observations collected by BeiDou receivers, it obtains the millimeter-level latitude, longitude, and altitude coordinates of each tower and substation after differential positioning calculation, and then uniformly converts them to the CGCS2000 national geodetic coordinate system. For external meteorological data, such as meteorological radar reflectivity, satellite cloud images, and numerical forecast products, coordinate transformation and grid resampling are also performed to ensure strict alignment with the coordinates of power grid equipment. Then, through nearest neighbor or inverse distance weighted interpolation, meteorological grid data is mapped to the location of each tower, achieving a one-to-one correspondence between meteorological elements and equipment, thereby eliminating the location ambiguity problem in traditional regional forecasts.

[0042] In terms of time dimension, this submodule utilizes nanosecond-level pulse signals output by the BeiDou time synchronization module to perform time synchronization calibration on various data sources, unifying the acquisition times of field equipment such as micro-meteorological sensors, tension gauges, and water level gauges to the BeiDou standard time. It also performs linear interpolation on data misaligned due to network latency or different sampling periods, generating aligned data snapshots of the same time segment to ensure that subsequent analysis is not misjudged due to time inconsistencies. After completing spatiotemporal correlation, this submodule outputs structured data, namely standardized records with tower number as the primary key, accompanied by precise coordinates, real-time meteorological elements (wind speed, temperature, rainfall, icing thickness, etc.), and corresponding timestamps. It also establishes a vulnerability attribute table for equipment (such as wind resistance level and design icing thickness), providing directly input feature vectors for upper-level AI early warning models. This module also dynamically responds to the arrival of new data, refreshing the correlation results in real time, supporting the overlay analysis of disaster movement paths and equipment locations, and rapidly triggering alerts when extreme weather systems approach specific tower groups.

[0043] Through the above mechanism, this submodule weaves the originally scattered and isolated multi-source data into a spatiotemporal correlation network with power grid equipment as the core, providing an accurate and traceable data foundation for subsequent multi-hazard identification, risk quantification and resource scheduling.

[0044] The multi-hazard intelligent identification submodule includes a rainstorm and flood early warning subunit for establishing a correlation model between water depth and substation terrain. It also includes a gale and typhoon early warning subunit, which matches wind speed with tower wind resistance levels during operation. Additionally, it includes a lightning early warning subunit, which integrates atmospheric electric field analysis and BeiDou ionospheric disturbance data. Finally, it incorporates a formula for estimating conductor icing thickness. In the formula Indicates equivalent icing thickness, in millimeters. This indicates the current measured tension of the conductor, in Newtons. This represents the reference tension of the conductor in the absence of ice and wind, measured in Newtons. This represents the coefficient indicating the effect of wind speed on tension. This indicates the current wind speed, in m / s. Indicates the density of ice. It is the acceleration due to gravity. Indicates the conductor span;

[0045] It also includes a built-in formula for calculating tower wind load. , This represents the wind load on the tower, measured in Newtons. air density, Wind speed (m / s) The windward area of ​​the tower (m) 2 , This is the shape drag coefficient.

[0046] The aforementioned multi-hazard intelligent identification submodule, the core of the entire system's disaster identification, works on the principle of constructing identification algorithms that combine physical and statistical driving methods for four major meteorological disasters: icing, strong winds, rainstorms and floods, and lightning, based on spatiotemporal correlation data provided by the upper layer. It calculates the probability and intensity level of each disaster on a tower-by-tower basis and comprehensively assesses the concurrent risk of multiple disasters. In terms of icing identification, this submodule integrates measured data from conductor tension sensors with temperature, humidity, and wind speed information from micro-weather stations. It calculates the equivalent icing thickness using mechanical equilibrium equations and overlays the duration of freezing rain forecasts to predict the icing growth trend over the next 6 hours. When the equivalent thickness exceeds 70% of the tower's designed icing threshold, it is automatically marked as a concern; if it exceeds 90%, it is automatically flagged as a concern. A 0% threshold triggers an early warning. For strong wind identification, the system integrates measured values ​​from tower anemometers with ground wind field data retrieved from meteorological radar, combined with tower wind resistance design parameters, to calculate wind loads and compare them with the structural bearing capacity limit. During typhoons or squall lines, real-time displacement monitoring data from the BeiDou Navigation Satellite System is used to identify tower tilting or swaying anomalies, providing early warnings of tower collapse risks. For rainstorm and flood identification, the system incorporates radar quantitative precipitation estimation and regional automatic weather station rainfall data, combined with the elevation of substations and distribution rooms and the distribution of surrounding water systems, to calculate water depth using water balance equations, focusing on low-lying sites and underground power distribution facilities. For lightning identification, the system assesses the insulator flashover probability based on atmospheric electric field meter monitoring data and lightning density and current amplitude provided by the lightning location system. When multiple disasters coexist, this submodule uses superimposed discrimination logic to calculate a weighted comprehensive risk coefficient, ultimately outputting the main disaster-causing factors, risk level (red, orange, yellow, blue), and future 1-6 hour trend for each tower, for further quantification by the subsequent risk thermodynamics module.

[0047] When the disaster impact prediction and risk thermodynamics submodule is working, it calculates the probability of damage to power grid equipment based on fault tree and Bayesian network using the equipment failure probability formula. The formula is as follows: ,in, Let be the probability of failure within time t. The duration of exposure to the disaster, in hours. For characteristic lifetime parameters, in hours. For shape parameters, Using a natural exponential function and combining geographic information, the system predicts potential power outage areas. Finally, it automatically generates a visual risk level using a risk level formula—red, orange, yellow, and blue. The risk level formula is as follows: ,in, Overall risk value, Equipment failure probability, Severity of failure consequences;

[0048] The core task of the aforementioned disaster impact prediction and risk thermodynamics submodule is to transform disaster identification results into quantifiable equipment damage probabilities and power grid operation risks. At the equipment level, for disaster types such as icing, strong winds, heavy rain, and lightning strikes, disaster intensity indicators (equivalent icing thickness, wind load value, water depth, lightning density, etc.) are input into the Weibull distribution failure probability model. Combined with the equipment's historical failure data and design life parameters, the conditional probability of damage under the current disaster conditions is calculated. Simultaneously, the cumulative effect of disaster duration is considered; when icing or heavy rain continues to exceed a set threshold, the failure probability increases over time. At the power grid level, the failure probability of a single device is multiplied by weighted factors such as the device's voltage level, the importance of the supplied load, and the number of power users to calculate a comprehensive risk value, which is then automatically classified into four risk levels: red, orange, yellow, and blue. This submodule also introduces a thermodynamic equation of state, treating the disaster system as a dynamic evolution process. Based on the rate of change of disaster intensity across multiple consecutive time sections, risk trends are extrapolated: if the derivative of disaster intensity remains positive, the risk increases; conversely, it decreases. Furthermore, this module employs a risk superposition and discrimination logic. When the same tower is threatened by multiple types of disasters simultaneously, it generates a comprehensive risk coefficient through weighted summation or taking the maximum value. The final output is a risk list for each tower, including major disaster-causing factors, equipment failure probability, comprehensive risk level, and risk change curves for the next 1-6 hours, providing a quantitative basis for subsequent emergency repair priority ranking.

[0049] The early warning credibility assessment submodule performs multi-model consistency verification during operation, and continuously optimizes threshold judgment based on historical feedback, performing self-learning correction for false alarms and false negatives. This involves multi-model consistency verification formulas. , The Pearson correlation coefficient between the outputs of the two models. For the model In the The output value of each sample point is the model. The arithmetic mean of all output values ​​is the model's... In the The output value of each sample point is the model. The arithmetic mean of all output values.

[0050] The aforementioned early warning credibility assessment submodule uses multi-model cross-validation and historical error statistics to score the confidence level of early warning results in real time, avoiding erroneous decisions due to false alarms from a single model or data anomalies. It horizontally compares the same output item of different algorithms (such as physical models, statistical regression, and lightweight neural networks) used in submodules 2.2 (multi-hazard identification) and 2.3 (risk prediction) to calculate the dispersion between the results of each model. If multiple models consistently determine the risk level of the same tower, the credibility score is higher; otherwise, it is lowered and marked for review. Simultaneously, this submodule performs a similarity search between the current early warning result and actual records under similar meteorological conditions in the historical disaster database. If the current early warning intensity closely matches the actual damage under similar historical conditions, the credibility weight is increased. For measured data sources, this module also monitors the quality of the input data in real time. When sensor data in a certain area exceeds its range, remains unchanged for a long time, or exhibits abnormal jumps, it automatically lowers the credibility of all early warning results for that area and prompts manual intervention for verification. The system outputs a confidence percentage (0-100%) for each warning record, as well as a warning label for low-confidence warnings, for dispatchers to refer to when making decisions. When the confidence level is lower than the set threshold, the system does not automatically trigger dispatch instructions, but only pushes them as reference information. This mechanism effectively reduces the false alarm rate and improves the reliability of the system under complex weather conditions.

[0051] The emergency dispatch decision-making and optimization layer serves as the action center, transforming early warnings into executable and traceable dispatch plans, which are particularly suitable for large-scale disaster scenarios;

[0052] The emergency dispatch decision and optimization layer includes a sub-module for optimal dispatch of repair resources, a sub-module for load transfer and islanded operation, a sub-module for multi-departmental collaborative command, and a sub-module for dynamic reassessment during disasters.

[0053] The optimal scheduling submodule for emergency repair resources operates by managing the real-time location of repair teams, vehicles, and drones based on BeiDou positioning. It also incorporates a path planning algorithm for dynamic obstacle avoidance in cases of road flooding and landslides. Following this, it performs material matching—automatically recommending a list of necessary materials based on the type of disaster damage—which involves Dijkstra's algorithm formula. ,in From the starting point to the node The current shortest distance, in kilometers. From the starting point to the node The known shortest distance, in kilometers. For nodes To the node The actual travel distance between the roads, The node number is the one for which the shortest path has been determined. Number the adjacent nodes to be updated;

[0054] The aforementioned submodule for optimal scheduling of emergency repair resources automatically generates executable repair plans based on early warning results and real-time road conditions during a disaster. Its core mechanism is as follows: First, it uses BeiDou high-precision positioning to obtain the real-time locations of repair teams, vehicles, drones, and emergency supply warehouses, and dynamically calibrates the status of various resources—types of onboard equipment, fuel levels, and personnel skill levels—to form a complete resource situation map. After receiving the upper-level risk list, this module determines the repair priority order based on risk level, prioritizing high-risk poles. For route planning, this module integrates BeiDou road network data with road flooding, landslide, and road closure information provided by meteorological departments, using a dynamic Dijkstra algorithm to calculate the optimal travel path from each repair team to the target pole. The actual distance weight is dynamically adjusted according to real-time road conditions and traffic control to minimize the total travel time. For resource matching, this module uses an integer linear programming model to solve for the optimal material allocation scheme based on the material types required for the task, ensuring equipment matching for high-priority tasks. The module ultimately outputs a repair team assignment plan, optimal route, and estimated arrival time, which is automatically sent to the repair personnel's mobile terminals. This module supports rolling recalculation, automatically adjusting the original plan when a new warning is triggered. Simultaneously, the progress of the repair task is transmitted back to the command center in real time, forming a closed-loop management system. Through this mechanism, the module significantly avoids blind patrols and ineffective deployments, improving repair efficiency.

[0055] When the load transfer and islanding operation submodule is working, it generates priority supply plans for important loads such as hospitals, communications, and flood control facilities based on the microgrid and distributed power generation fast disconnection and grid connection strategies.

[0056] The multi-department collaborative command submodule has the function of one-click issuance of emergency instructions, which can be sent via Beidou short message, SMS and voice broadcast. It also has the function of visual management of task execution status, and has reserved interfaces with government emergency offices, meteorological bureaus and transportation departments.

[0057] The disaster dynamic reassessment submodule has the function of refreshing the early warning level every 15 minutes during the disaster development process. Moreover, it performs rolling corrections of the scheduling plan based on the latest real-time data during operation, involving a smooth update formula for the dynamic early warning index. ,in For the current moment Smoothed warning level values For smoothing coefficients, For the current moment The instantaneous warning level value calculated from measured data. This is the smoothed warning level value from the previous moment. Current time step, every 15 minutes.

[0058] The aforementioned dynamic reassessment submodule during a disaster continuously updates warning levels and dispatch plans based on the influx of new real-time data, ensuring that decisions are always synchronized with the actual situation on the ground. Using a 15-minute rolling cycle, each new data arrival triggers a complete reassessment process. First, it re-reads the latest micro-meteorological sensor data, BeiDou tower deformation monitoring data, and updated weather radar products. Combined with feedback on actual disaster damage that has occurred during that period (e.g., whether a line has lost power), it repeatedly calls the upper-level warning analysis model to calculate the latest risk level for each tower at the current moment. Then, it compares the new risk level with the result from the previous period. If the risk level increases, such as from yellow to orange or the appearance of a new high-risk tower, the system immediately generates an incremental warning and pushes it to the dispatch interface for highlighting. Regarding dispatch plan adjustments, this module uses an exponentially weighted moving average method to smooth risk trends, avoiding drastic fluctuations in the dispatch plan caused by single data spikes. It also sets a trigger threshold; the plan adjustment is only formally triggered when the risk value exceeds the upgrade boundary for two consecutive rolling cycles, maintaining the stability of dispatch instructions. For dispatched repair teams, if the latest risk level of their target towers decreases or has dropped to a safe level, the system automatically issues a reassignment suggestion to avoid resource congestion. Ultimately, this module continuously outputs optimized scheduling plans based on the latest situation until the disaster ends, ensuring the system's adaptability during the disaster's evolution.

[0059] The visualization and operation and maintenance support layer provides a visual interface for disaster situation and scheduling status, and monitors the system's own operational health.

[0060] The visualization and operation and maintenance support layer belongs to the human-computer interaction layer, including a 3D digital sandbox module, a mobile command sentinel submodule, a system health self-check submodule, and a post-disaster review and knowledge base submodule. When the 3D digital sandbox module is working, it displays overlay maps of power grids, meteorology, terrain, and densely populated areas based on GIS and BeiDou high-precision maps, while simultaneously performing disaster evolution animation simulations and supporting timeline drag-and-drop playback. When the mobile command sentinel submodule is working, it is a lightweight APP for on-site repair personnel, receiving early warnings, reporting on-site pictures and videos, and supporting offline mode, sending and receiving short messages via BeiDou short messages when there is no signal.

[0061] The system health self-check submodule monitors the quality of the data source, including outlier detection, latency and packet loss rate detection, and monitors the model running status, including CPU and GPU load and inference time. The disaster recovery and knowledge base submodule generates a full-link time-series report of the disaster, from early warning to response to recovery, and automatically extracts typical cases and updates the emergency plan library.

[0062] The aforementioned visualization and operation and maintenance support layer serves as the system's user-facing portal and core for operational support. During operation, it transforms the analysis results from the upper layers into an intuitive graphical interface and continuously monitors the health of the entire system link, ensuring accurate transmission of early warnings and dispatch instructions. In terms of visualization, this layer uses a 3D digital sand table as its core carrier, overlaying a high-precision BeiDou map with power grid equipment distribution, real-time meteorological cloud maps, disaster risk heat maps, and the location of emergency repair resources to form a unified situational awareness view. Decision-makers can view the overall situation on a large screen or zoom in to any tower to view detailed real-time data, risk levels, and trend curves. Simultaneously, the system provides a timeline playback function, allowing users to trace the entire disaster evolution process and assist in post-disaster analysis. On the mobile end, this layer provides a dedicated APP for on-site repair personnel, pushing task instructions, optimal route navigation, and on-site image reporting entry in a concise card format. Even without ground signals, key instructions can be sent and received via BeiDou short messages, achieving a closed-loop information system between the front and rear.

[0063] In terms of operation and maintenance, this layer establishes a system health self-check module to continuously monitor the latency, packet loss rate, and outlier ratio of each data source, as well as the inference time and GPU load of the upper-layer AI model. If a data source quality falls below a threshold or the model response times out, an alarm is immediately triggered in red on the operation and maintenance interface, and an anomaly log is recorded, providing operation and maintenance personnel with clear fault location information. This module also automatically verifies the deviation between the early warning results and the actual disaster damage, and regularly generates false alarm and missed alarm analysis reports to provide a basis for model iterative training. Furthermore, the system has an automatic post-disaster review function, generating a full-link time-series report with one click after the disaster ends, including early warning trigger records, dispatch instructions, repair progress, and recovery time, assisting managers in summarizing experience and updating the emergency plan library. Through these mechanisms, this layer not only makes complex data "understandable" but also ensures the reliability and traceability of system operation.

[0064] The above is the entire working process of the device, and all contents not described in detail in this specification are existing technologies known to those skilled in the art.

[0065] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A power meteorological disaster early warning device integrating BeiDou multi-source data, comprising a multi-source data sensing and fusion layer, a BeiDou and AI disaster early warning analysis layer, an emergency dispatch decision-making and optimization layer, and a visualization and operation and maintenance support layer, characterized in that: When the multi-source data perception and fusion layer is working, it solves the problem of data silos and realizes integrated data collection from the sky, ground and air. The Beidou and AI disaster early warning and analysis layer performs short-term early warning, trend prediction and risk quantification based on the fused data. The emergency dispatch decision-making and optimization layer serves as the action center, transforming early warnings into executable and traceable dispatch plans, which is particularly suitable for large-scale disaster scenarios; the visualization and operation and maintenance support layer provides a visual interface for disaster situation and dispatch status, and monitors the system's own operational health.

2. The power meteorological disaster early warning device based on BeiDou fusion multi-source data according to claim 1, characterized in that: The multi-source data perception and fusion layer includes a BeiDou satellite real-time observation submodule and a meteorological multi-element perception submodule. The BeiDou satellite real-time observation submodule realizes the BeiDou high-precision positioning function and also has the BeiDou short message communication function. When 4G and 5G on the ground are interrupted, it serves as an emergency data backhaul channel. It also has the BeiDou time synchronization function, providing a unified time reference for all network devices and ensuring the consistency of time-series data.

3. A power meteorological disaster early warning device based on BeiDou fusion multi-source data as described in claim 1, characterized in that: The multi-source data sensing and fusion layer also includes a power private network and Internet of Things submodule and an external data access submodule; The power grid and Internet of Things submodule monitors conductor galloping and ice thickness through micro-meteorological sensors on transmission lines, while the external data access submodule receives numerical weather forecasts from the meteorological bureau and accesses the database of geological disaster hazard points and historical disaster case databases.

4. A power meteorological disaster early warning device based on BeiDou fusion multi-source data according to claim 1, characterized in that: The BeiDou and AI disaster early warning analysis layer includes a disaster correlation submodule based on BeiDou spatiotemporal reference, a multi-hazard intelligent identification submodule, a disaster impact prediction and risk thermodynamics submodule, and an early warning credibility assessment submodule. The disaster correlation submodule based on the BeiDou spatiotemporal reference incorporates a spatiotemporal alignment engine, unifying meteorological data, geographic data, and power grid equipment coordinates to the BeiDou coordinate system. It also performs dynamic and static overlay analysis of disasters, combining dynamic meteorological disasters with static equipment vulnerability analysis, involving formulas... , Indicates the target coordinate system CGCS2000 coordinate, Represents the original coordinate system of BeiDou coordinate, Indicates that the origin of the two coordinate systems is at... Translation amount in direction, This represents the rotation angle between the two coordinate systems. Represents the original coordinate system of BeiDou coordinate, The origins of the two coordinate systems are at The amount of translation in the direction; The multi-hazard intelligent identification submodule includes a rainstorm and flood early warning subunit for establishing a correlation model between water depth and substation terrain. It also includes a gale and typhoon early warning subunit, which matches wind speed with tower wind resistance levels during operation. Additionally, it includes a lightning early warning subunit, which integrates atmospheric electric field analysis and BeiDou ionospheric disturbance data. Finally, it incorporates a formula for estimating conductor icing thickness. In the formula Indicates equivalent icing thickness, in millimeters. This indicates the current measured tension of the conductor, in Newtons. This represents the reference tension of the conductor in the absence of ice and wind, measured in Newtons. This represents the coefficient indicating the effect of wind speed on tension. This indicates the current wind speed, in m / s. Indicates the density of ice. It is the acceleration due to gravity. Indicates the conductor span; It also includes a built-in formula for calculating tower wind load. , This represents the wind load on the tower, measured in Newtons. air density, Wind speed (m / s) The windward area of ​​the tower (m) 2 , This is the shape drag coefficient.

5. A power meteorological disaster early warning device based on BeiDou fusion multi-source data according to claim 4, characterized in that: When the disaster impact prediction and risk thermodynamics submodule is working, it calculates the probability of damage to power grid equipment based on fault trees and Bayesian networks using the equipment failure probability formula. The formula is as follows: ,in, Let be the probability of failure within time t. The duration of exposure to the disaster, in hours. For characteristic lifetime parameters, in hours. For shape parameters, Using a natural exponential function and combining geographic information, the system predicts potential power outage areas. Finally, it automatically generates a visual risk level using a risk level formula—red, orange, yellow, and blue. The risk level formula is as follows: ,in, Overall risk value, Equipment failure probability, Severity of failure consequences; The early warning reliability assessment submodule performs multi-model consistency verification during operation, and continuously optimizes threshold judgment based on historical feedback, performing self-learning correction for false alarms and false negatives. This involves multi-model consistency verification formulas. , The Pearson correlation coefficient between the outputs of the two models. For the model In the The output value of each sample point is the model. The arithmetic mean of all output values ​​is the model's... In the The output value of each sample point is the model. The arithmetic mean of all output values.

6. A power meteorological disaster early warning device based on BeiDou fusion multi-source data according to claim 1, characterized in that: The emergency dispatch decision and optimization layer includes a sub-module for optimal dispatch of repair resources, a sub-module for load transfer and islanded operation, a sub-module for multi-department collaborative command, and a sub-module for dynamic reassessment during disasters. The optimal scheduling submodule for emergency repair resources operates by managing the real-time location of repair teams, vehicles, and drones based on BeiDou positioning. It also incorporates a path planning algorithm for dynamic obstacle avoidance in cases of road flooding and landslides. Following this, it performs material matching—automatically recommending a list of necessary materials based on the type of disaster damage—which involves Dijkstra's algorithm formula. ,in From the starting point to the node The current shortest distance, in kilometers. From the starting point to the node The known shortest distance, in kilometers. For nodes To the node The actual travel distance between the roads, The node number is the one for which the shortest path has been determined. Number the adjacent nodes to be updated; When the load transfer and islanding operation submodule is working, it generates priority supply plans for important loads such as hospitals, communications, and flood control facilities based on the microgrid and distributed power source rapid disconnection and grid connection strategies.

7. A power meteorological disaster early warning device based on BeiDou fusion multi-source data according to claim 6, characterized in that: The multi-department collaborative command submodule has a one-click emergency command issuance function, which can be implemented via Beidou short message, SMS, and voice broadcast. It also has a function for visual management of task execution status and is equipped with reserved interfaces for government emergency offices, meteorological bureaus, and transportation departments. The disaster dynamic reassessment submodule has the function of refreshing the early warning level every 15 minutes during the disaster development process, and it performs rolling corrections of the scheduling plan based on the latest real-time data during operation, involving a smooth update formula for the dynamic early warning index. ,in For the current moment Smoothed warning level values For smoothing coefficients, For the current moment The instantaneous warning level value calculated from measured data. This is the smoothed warning level value from the previous moment. Current time step, every 15 minutes.

8. A power meteorological disaster early warning device based on BeiDou fusion multi-source data according to claim 1, characterized in that: The visualization and operation and maintenance support layer belongs to the human-computer interaction layer, including a 3D digital sandbox module, a mobile command sentinel sub-module, a system health self-check sub-module, and a post-disaster review and knowledge base sub-module; When the three-dimensional digital sand table module is working, it displays an overlay map of power grid, meteorology, topography, and densely populated areas based on GIS and Beidou high-precision maps, while simultaneously simulating disaster evolution animations and supporting timeline drag-and-drop playback. The mobile command sentry submodule is a lightweight APP for on-site repair personnel. It receives early warnings, reports on-site photos and videos, and supports offline mode, sending and receiving short messages via Beidou short message service when there is no signal.

9. A power meteorological disaster early warning device based on BeiDou fusion multi-source data according to claim 8, characterized in that: The system health self-check submodule monitors the quality of the data source, including outlier detection, latency and packet loss rate detection, and monitors the model running status, including CPU and GPU load and inference time.

10. A power meteorological disaster early warning device based on BeiDou fusion multi-source data according to claim 9, characterized in that: When the post-disaster review and knowledge base sub-module is working, it generates a full-chain time-series report on the disaster situation, covering early warning, response, and recovery. At the same time, it automatically extracts typical cases and updates the emergency plan library.

Images