Power grid icing disaster risk assessment method and system

Abstract

The invention provides a power grid icing disaster risk assessment method and system, and the method comprises the steps: constructing a multi-source data system through the deployment of a customized meteorological terminal, the combination of FNL reanalysis data, power grid disaster situation and operation and maintenance data, improving the data precision through a Kalman filtering algorithm and a dynamic downscaling method, screening icing key disaster-causing factors through partial correlation analysis, and achieving the risk assessment of the icing disaster. According to the method, disaster levels are divided, risks are quantified, a visual platform is built based on the WebGIS technology, multiple types of data are integrated to achieve multi-layer display and interaction, the problem that the spatial-temporal scale of traditional meteorological data is not matched with the power grid requirement is solved, the method better meets the special requirement of the power grid, differentiated emergency measures are made in an assisting mode, and the disaster prevention efficiency is improved; the method comprises the following steps: step 1, collecting data, and constructing a multi-source data system; 2, processing data, and improving forecasting precision; 3, screening icing key disaster-inducing factors by adopting partial correlation analysis; 4, dividing disaster grades, quantifying risks, and drawing a risk zoning map of the disasters; and 5, carrying out visual display and early warning.

Description

Technical Field

[0001] This invention relates to the field of risk assessment technology, specifically to a method and system for assessing the risk of power grid icing disasters. Background Technology

[0002] Icing of transmission lines due to factors such as low temperature, rain and snow is called icing. It can cause insulator flashover, line galloping and tower tilting. As the amount of icing increases, it can easily cause serious accidents such as line breakage and tower collapse, which may lead to large-scale power outages and even grid collapse.

[0003] Currently, the power industry's icing disaster early warning system mainly relies on public warning information from meteorological departments. However, the spatiotemporal accuracy of icing weather warnings differs significantly from the actual needs of the power system, making it difficult to meet the refined management requirements of power grid operation. The main gaps between the meteorological information currently available and actual needs are: The acquired meteorological forecast and early warning information cannot match the short-term flow and behavior of electrical energy on the millisecond to minute scale in the time dimension of power grid operation. In the spatial dimension, power grid operation focuses more on small-scale energy network components, such as substations, within a range of tens to hundreds of meters. Therefore, it is necessary to adopt appropriate downscaling methods to transform long-term, large-scale meteorological data into short-term, fine-range data required for power grid operation. Meteorological departments focus more on the extreme nature, intensity, duration, and scope of icing disasters, while power grid operators are more concerned with the direct impact of icing disasters on the stability and reliability of the power system, and how to ensure the safety and continuity of power supply. Therefore, they may have different emphases in their assessment and response strategies for icing disasters. The emergency response plans for extreme weather that affect power grid operation rely too heavily on public forecasts and warnings issued by local meteorological stations, which falls short of the power company's requirements for the targetedness and precision of emergency management.

[0004] Meanwhile, existing emergency response mechanisms are mostly focused on post-disaster emergency response, lacking technological support for pre-disaster preventative deployment, resulting in low efficiency in disaster prevention and mitigation. Furthermore, meteorological warning standards are primarily geared towards public service, failing to fully consider the specific needs of the power system, such as equipment vulnerability and grid stability, leading to a disconnect between warning information and the actual response needs of the power industry.

[0005] Therefore, matching icing weather forecasts and early warnings with the power grid on a spatiotemporal scale, clarifying the hazard level of icing disasters to power grid operation, and conducting risk assessment research on power grid icing disasters using refined meteorological data have become crucial technical guarantees for power companies' emergency management to better serve the entire company. There is an urgent need to construct numerical forecasting models based on climate microenvironments and "climate-power" joint simulation technology to achieve refined early warning and risk assessment of power grid icing meteorological disasters, providing the power industry with more targeted disaster prevention and mitigation solutions, improving the intelligent management level of the power system, and ensuring social stability and people's livelihood needs. Summary of the Invention

[0006] This invention aims to address the technical deficiencies of existing technologies by providing a method and system for assessing the risk of power grid icing disasters. It constructs a multi-source data system by deploying customized meteorological terminals and combining FNL reanalysis data with power grid disaster and operation and maintenance data. The system utilizes Kalman filtering and dynamic downscaling algorithms to improve data accuracy, while partial correlation analysis is used to screen key icing-causing factors, classify disaster levels, and quantify risks. A visualization platform is built using WebGIS technology, integrating multiple data types to achieve multi-layered display and interaction. This effectively solves the problem of the mismatch between the spatiotemporal scale of traditional meteorological data and the needs of the power grid. The classified disaster levels are more aligned with the specific needs of the power grid, helping to formulate differentiated emergency measures and improve disaster prevention efficiency.

[0007] The first aspect of this invention discloses a method for assessing the risk of power grid icing disasters, comprising the following steps: Step 1: Collect data and build a multi-source data system; by deploying customized meteorological terminals and combining FNL reanalysis data, power grid disaster and operation and maintenance data, establish a high-resolution grid numerical weather prediction model covering the power grid equipment area, and run the numerical weather prediction model to output basic meteorological data including temperature, humidity and wind speed. Step 2: Process the data to improve forecast accuracy; use the Kalman filter algorithm to assimilate the data and then use the dynamic downscaling method to improve the resolution of the meteorological data grid. Step 3: Use partial correlation analysis to screen key factors causing ice accumulation disasters; Step 4: Classify disaster levels and quantify risks, and draw a disaster risk zoning map; Step 5: Visualization and Early Warning. A visualization platform is built based on WebGIS technology to integrate and display meteorological data, power grid data, geospatial information data, and early warning data.

[0008] Furthermore, The customized meteorological terminal in step 1 includes a micro-intelligent meteorological instrument and an intelligent terminal system. The micro-intelligent meteorological instrument is deployed around the power transmission lines of meteorological stations in the power grid area to acquire meteorological data and wirelessly transmit it to the intelligent terminal system. The intelligent terminal integrates a global geographic location information system that supports GNSS and wirelessly transmits meteorological data and location data to a visualization platform. The intelligent weather instrument includes an integrated meteorological element sensor and a tipping bucket rain gauge.

[0009] Furthermore, The numerical weather prediction model in step 1 uses the WRF4.1 model for numerical simulation. FNL reanalysis data with a resolution of 1°×1° and a frequency of once every 6 hours is selected as the initial field conditions. The numerical simulation adopts a double-layer nested structure, including a coarse grid and a fine grid, with horizontal resolutions of 9km and 3km, respectively, and corresponding grid numbers of 649×541 and 673×676, respectively. The top of the model is set to 50hPa, and the vertical direction is divided into 51 layers.

[0010] Furthermore, The Kalman filter algorithm in step 2 assimilates the meteorological data by combining the measured data from the customized meteorological terminal with the meteorological data from the numerical weather prediction model. Using each icing meteorological element in the power grid as a state variable, a linear equation is constructed: , in X k Let k be the state vector at time k. A Here is the state transition matrix. B To control variables, u k For control items such as topography and radiation, w k-1 The state noise is Gaussian distributed; Establish a mapping between observation data and state variables: , In the formula Z k These are the preprocessed observations. H For sparse observation matrices, v k The observation noise is Gaussian distributed. Use the optimal state from the previous moment to deduce the current predicted state: , In the formula Predict the state vector at time k. The optimal state vector at time k-1; Simultaneous calculation of prediction error covariance: , In the formula Let k be the prediction error covariance matrix. The optimal error covariance matrix at time k-1 is... Let A be the state noise covariance matrix. T Transpose of the state transition matrix; Calculate the Kalman gain and adjust the weights according to the accuracy of the observations and the model: , In the formula The Kalman gain at time k, To observe the noise covariance matrix, H T Observation matrix transpose; The forecast is then corrected using measured data from a customized meteorological terminal to obtain the optimal state: , In the formula Let k be the optimal state vector at time k; Last updated error covariance: , In the formula Let I be the optimal error covariance matrix at time k, where I is the identity matrix; Repeat the iteration in 1-hour increments, adjusting every 24 hours based on assimilation error. Q, R To ensure the accuracy of short-term 0-72 hour forecasts, the assimilated 1-20km grid data is converted into NetCDF format, high-risk area data corresponding to key icing thresholds are extracted, and data is pushed to the visualization platform in real time via wireless transmission, with a latency of ≤5 minutes.

[0011] Furthermore, The dynamic downscaling method in step 2 runs the numerical weather model on a defined grid, which includes an inner grid and an outer grid. First, the numerical weather prediction model is parameterized and adapted. Then, the initial field and boundary conditions are interpolated. The FNL low-resolution data is mapped to the outer grid of the downscaled model through bilinear interpolation as the initial field and boundary conditions of the model. During the interpolation process, local topographic data needs to be combined to correct the temperature and air pressure in areas with large altitude differences. Numerical weather forecast model operation settings: The time step is determined according to the grid resolution. The inner grid (1-5km) is set to 1-3 minutes, and the outer grid (10-20km) is set to 5-10 minutes to meet the CFL condition. The simulation duration outputs short-term forecast data of 0-72 hours, with a time resolution of 1 hour. The output elements focus on key meteorological elements of icing, including temperature at 2m height, relative humidity at 2m height, wind speed / direction at 10m height, precipitation, precipitation phase, and supercooled water content in clouds. Post-processing of results: Outliers in the numerical weather prediction model were removed, and missing values ​​were filled in by replacing them with the mean of adjacent grids or correcting them with historical data from the same period. The original data format output by the numerical weather prediction model was converted to a format compatible with the existing power grid platform, and meteorological data of the grid where the power grid equipment is located were extracted to generate a "equipment-meteorological element" correspondence table. The measured data from the customized meteorological terminal was compared with the downscaled output data to calculate the root mean square error (RMSE). The requirements were that the root mean square error of temperature ≤ 0.8℃ and the root mean square error of wind speed ≤ 1.0m / s. If these requirements were not met, the model parameters were optimized by adjusting the boundary layer.

[0012] Furthermore, In step 3, partial correlation analysis is used to screen key icing-causing factors. The t-test or F-test is used to screen key icing-causing factors. A meteorological factor model is constructed based on the key icing-causing factors to reclassify the meteorological disaster level of icing.

[0013] Furthermore, In step 4, disaster levels are classified and risks are quantified. The information diffusion theory assessment model is used to quantify the risks, and the information diffusion theory assessment model is as follows: Let the domain of disaster frequency be: , In the formula u i Let i be the number of possible disasters; Let the sample set of disaster frequency be: , In the formula y j This refers to the j-th sample in the disaster frequency sample set; right Y The j-th sample y j Diffusion is performed, and the calculation formula is as follows: , In the formula h The diffusion coefficient, used to control the breadth of information diffusion, can be calculated using the maximum and minimum values ​​in the sample set and the number of samples.f j ( u i ) represents the j-th sample y j For the i-th value in the universe of discourse u i The diffusion membership degree; make C j The diffusion normalization coefficient for the j-th sample is: , The membership function of its corresponding fuzzy subset is: , right Y m After processing, the risk assessment result can be obtained.

[0014] Furthermore, In step 4, a disaster risk zoning map is drawn using a disaster risk index assessment model. The disaster risk index assessment model is as follows: , In the formula, M is the disaster risk index, n is the number of disaster levels, and p i Let q be the probability of the i-th level disaster occurring. i Let be the intensity of the i-th level of disaster.

[0015] The second aspect of this invention discloses a power grid icing disaster risk assessment system, comprising: The customized numerical weather prediction module collects global / regional climate model data, FNL reanalysis data, and meteorological station observation data in the power grid area to establish a high-resolution grid numerical weather prediction model covering the power grid equipment area, and outputs meteorological data. The meteorological disaster classification and risk assessment module, based on power grid icing disaster records and historical icing meteorological data, classifies icing disasters into levels I to IV according to their impact on the power grid, scope, and severity. Based on operational meteorological data and historical disaster information, it uses partial correlation analysis to control for the influence of other variables, screens key icing meteorological disaster-causing factors, establishes the correlation between disaster levels and meteorological conditions, and ultimately forms a classification standard for icing disasters and a disaster-causing factor index library. Based on the information diffusion theory assessment model and the disaster risk index assessment model, combined with disaster intensity, occurrence probability, and power grid vulnerability, it calculates the disaster risk level of different regions. Through statistical analysis of historical icing disaster records and meteorological data, it generates a spatial distribution map of icing risk probability, and overlays power grid facility distribution information to quantify regional risk levels, outputting a risk zoning map of icing disasters. The power grid meteorological early warning visualization platform module, relying on WebGIS technology, builds a power grid icing meteorological disaster early warning visualization platform, integrating real-time meteorological data, power grid operation status information and risk assessment results, and realizes multi-layer overlay display and interactive operation.

[0016] Furthermore, The numerical weather prediction model is optimized by combining observation data from a customized meteorological terminal with the initial field and boundary conditions of the numerical weather prediction model, and the parameterization scheme is adjusted to simulate icing-related meteorological elements. Compared with the prior art, the beneficial effects of the present invention are: This invention provides a method and system for assessing the risk of power grid icing disasters. By deploying customized meteorological terminals and combining FNL reanalysis data with power grid disaster and operation and maintenance data, a multi-source data system is constructed. The Kalman filter algorithm is used to assimilate data and improve forecast accuracy, while dynamic downscaling is employed to enhance the resolution of the meteorological data grid. Partial correlation analysis is used to screen key icing-causing factors, classify disaster levels, and quantify risks. A visualization platform is built using WebGIS technology, integrating multiple data types to achieve multi-layer display and interaction. This effectively solves the problem of mismatch between the spatiotemporal scale of traditional meteorological data and power grid needs. The classified disaster levels are more aligned with the specific needs of the power grid, assisting in the formulation of differentiated emergency measures, improving disaster prevention efficiency, and enabling more scientific decision-making. This promotes the transformation of power grid management towards data-driven approaches, significantly reducing the risk of equipment failures and large-scale power outages caused by icing, and ensuring the stable operation of the power system and electricity supply for residential use. Attached Figure Description

[0017] Figure 1 This is a flowchart of the steps of the method of the present invention; Figure 2 This is a schematic diagram of the customized numerical weather prediction module structure of the present invention; Figure 3 This is a Kalman filter data diagram of the present invention. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] like Figures 1-3 As shown, it illustrates a specific embodiment of the present invention: The first aspect of this invention discloses a method for assessing the risk of power grid icing disasters, comprising the following steps: Step 1: Collect data and build a multi-source data system; by deploying customized meteorological terminals and combining FNL reanalysis data, power grid disaster and operation and maintenance data, establish a high-resolution grid numerical weather prediction model covering the power grid equipment area, and run the numerical weather prediction model to output basic meteorological data including temperature, humidity and wind speed. Step 2: Process the data to improve forecast accuracy; use the Kalman filter algorithm to assimilate the data and then use the dynamic downscaling method to improve the resolution of the meteorological data grid. Step 3: Use partial correlation analysis to screen key factors causing ice accumulation disasters; Step 4: Classify disaster levels and quantify risks, and draw a disaster risk zoning map; Step 5: Visualization and Early Warning. A visualization platform is built based on WebGIS technology to integrate and display meteorological data, power grid data, geospatial information data, and early warning data.

[0020] Specifically, it provides comprehensive, accurate, and timely information support to power grid management departments at the data, risk assessment, and decision-making and management levels, making the decision-making process more scientific and rational. By formulating effective disaster prevention measures and emergency plans, it significantly reduces the risk of equipment failures and large-scale power outages caused by icing, and improves the reliability and stability of the power grid.

[0021] (1) Data level A multi-source data system was constructed, providing comprehensive and highly targeted data. By deploying customized meteorological terminals and combining FNL reanalysis data, power grid disaster and operation and maintenance data, a multi-source data system covering the power grid equipment area was built. This integration of multi-source data not only includes traditional observational meteorological data, but also incorporates disaster and operation and maintenance information closely related to power grid operation, making the data more comprehensive and targeted. It can more accurately reflect the relevant situation of power grid icing disasters and provide a solid and reliable data foundation for subsequent risk assessment. To improve forecast accuracy and meet power grid demands: The Kalman filter algorithm is used to assimilate data, effectively integrating data from different sources, eliminating errors and inconsistencies, and improving data consistency and accuracy. Based on this, a dynamic downscaling method is employed to further enhance the resolution of the meteorological data grid, making weather forecasts more refined and able to more accurately capture local weather changes. This series of data processing methods effectively solves the problem of the mismatch between the spatiotemporal scale of traditional meteorological data and power grid needs, providing strong support for the accurate prediction of power grid icing disasters.

[0022] (2) Risk assessment level Screening key disaster-causing factors to improve assessment accuracy: Using partial correlation analysis to screen key disaster-causing factors of power grid icing can identify the factors most closely related to power grid icing disasters from numerous meteorological factors, eliminating interference from irrelevant factors. This helps to more accurately grasp the formation mechanism and development law of power grid icing disasters, improve the scientificity and accuracy of risk assessment, and provide a basis for formulating effective disaster prevention measures.

[0023] Disaster risk classification aligns with power grid needs and facilitates differentiated emergency response: Based on the identified key disaster-causing factors, disaster levels are classified and risks are quantified. These classifications fully consider the specific needs of the power grid and can more accurately reflect the impact of power grid icing disasters on power grid equipment and operation. Based on different disaster risk levels, differentiated emergency measures can be developed, disaster prevention resources can be rationally allocated, disaster prevention efficiency can be improved, and disaster losses can be reduced.

[0024] (3) Decision-making and management level A visualization platform was built to achieve the integration and interaction of multiple data types: Based on WebGIS technology, a visualization platform was built to integrate and display various data types, including meteorological data, power grid data, geospatial information data, and early warning data. Through multi-layer display and interactive functions, decision-makers can intuitively understand the distribution, development trend, and impact on the power grid caused by power grid icing disasters, providing a convenient and efficient tool for scientific decision-making.

[0025] Meteorological data acquisition and construction of numerical weather prediction models Global / regional climate model data is obtained through FNL reanalysis data, and meteorological element data is obtained through customized meteorological terminals at meteorological stations in the power grid area. The customized meteorological terminal consists of a micro-intelligent weather instrument and an intelligent terminal system. The micro-intelligent weather instrument is deployed around the transmission lines of the meteorological station in the power grid area to acquire meteorological data and wirelessly transmit it to the intelligent terminal system. It is generally installed by pre-burying or using clamps to install it on the column. The micro-intelligent weather instrument includes an integrated meteorological element sensor and a tipping bucket rain gauge. The integrated meteorological element sensor can acquire wind speed, wind direction, and temperature and humidity, while the tipping bucket rain gauge can acquire rainfall.

[0026] Data transmission is performed using a smart terminal, which integrates a Global Geospatial Information System (GPS) supporting GNSS, enabling easy location of geographic information at weather stations. It also integrates Wi-Fi communication functionality, allowing convenient network access in areas with comprehensive Wi-Fi coverage. Furthermore, it integrates full-band 4G communication functionality, facilitating convenient network setup in areas with 4G signal coverage. Local communication: RJ45; Wireless communication: Supports 4G and Wi-Fi communication, and is compatible with 2G / 3G, with automatic network identification and switching; Geographic information: Supports GPS, GNSS, and BeiDou positioning systems.

[0027] In response to the meteorological disaster of power grid icing, this study analyzes the types of sensitive meteorological elements in different stages of power grid operation by combining the characteristics of meteorological elements such as temperature, humidity, precipitation and atmospheric stratification, and analyzes the spatiotemporal characteristics of various sensitive meteorological elements.

[0028] A WRF-based numerical weather prediction model was constructed, and the WRF4.1 model was used for numerical simulation. Initial field conditions were selected from 1°×1° resolution FNL reanalysis data updated every 6 hours. The simulation employed a two-layer nested structure, with horizontal resolutions of 9 km for the coarse grid and 3 km for the fine grid, corresponding to 649×541 and 673×676 grid points, respectively. The top of the model was set to 50 hPa, and the vertical direction was divided into 51 layers. Specific physical parameterization schemes included: RRTM longwave radiation scheme, Dudhia shortwave radiation scheme, Monin-Obukhov similarity theory for the surface layer, Noah land surface process model, YSU boundary layer parameterization scheme, and WSM6 microphysical process scheme, ultimately generating hourly 3km×3km meteorological element grid point forecasts.

[0029] Meteorological data assimilation is a technical process that combines observational data with numerical weather prediction models to generate a more accurate initial field. The core objective of data assimilation is to minimize the differences between the analytical field, the observational field, and the background field. In meteorology, because the atmosphere is a complex dynamic system, direct observation of atmospheric states has limitations. Therefore, numerical models are needed to simulate atmospheric evolution, but these models suffer from uncertainties in initial and boundary conditions. By introducing the Kalman filter algorithm for data assimilation and integrating local meteorological observation data, the forecast accuracy of short-term meteorological elements can be improved.

[0030] Kalman filtering provides a method to combine observational data and model predictions to obtain the optimal estimate of atmospheric state. Although atmospheric systems are nonlinear, Kalman filtering typically linearizes the system and assumes the error follows a Gaussian distribution. This assumption allows Kalman filtering to provide the optimal solution within a linear Gaussian framework. The goal of Kalman filtering is to find the state estimate that maximizes the likelihood function of the observed data, which is equivalent to minimizing the covariance of the error between the predicted and observed values. The system state is assumed to depend only on the state at the previous time step and not on earlier states; this is the Markov property. This assumption simplifies the state estimation problem, allowing Kalman filtering to be computed recursively. Under the linear Gaussian assumption, Kalman filtering provides the optimal solution for state estimation, i.e., it minimizes the covariance of the estimation error. The mathematical foundation of Kalman filtering is Bayes' theorem, which allows updating the posterior probability of the state estimate using known prior probabilities and new observational data. Kalman filtering requires defining a dynamic system model to describe the change of state over time, and an observational model to describe the relationship between the observed values ​​and the state variables. Kalman filtering can continuously fuse new observational data to provide the latest estimates of atmospheric conditions, thereby improving the accuracy of weather forecasts.

[0031] Two types of core data need to be matched with the power grid icing requirements: one is background field data (FNL reanalysis data, WRF / MM5 / GRAPS numerical model prediction results, including key icing elements such as temperature, humidity, and wind speed); the other is observation data (minute-level temperature, humidity, and wind speed collected by customized meteorological terminals, hourly rainfall, and data from conventional meteorological stations).

[0032] First, perform spatiotemporal matching (interpolate the observed data to the model time step, and interpolate the FNL data to the 1-20km power grid grid); then remove outliers (identify fault data using the 3σ principle and complete the data with the mean or data from adjacent stations); finally, standardize to eliminate dimensional differences and ensure data availability.

[0033] Using each icing meteorological element in the power grid as a state variable, a linear equation is constructed: , in X k Let k be the state vector at time k. A Here is the state transition matrix. B To control variables, u k For control items such as topography and radiation, w k-1 The state noise is Gaussian distributed.

[0034] Establish a mapping between observation data and state variables: , In the formula Z k These are the preprocessed observations. H This is a sparse observation matrix (matching observation stations with grids). v k The observation noise is Gaussian distributed.

[0035] Use the optimal state from the previous moment to deduce the current predicted state: , In the formula Predict the state vector at time k. This is the optimal state vector at time k-1.

[0036] Simultaneous calculation of prediction error covariance: , In the formula Let k be the prediction error covariance matrix. The optimal error covariance matrix at time k-1 is... Let A be the state noise covariance matrix. T Transpose of the state transition matrix.

[0037] Calculate the Kalman gain (adjusting weights based on observation and model accuracy): , In the formula The Kalman gain at time k, To observe the noise covariance matrix, H T Transpose the observation matrix.

[0038] Then, the prediction is corrected using the observed values ​​to obtain the optimal state: , In the formula Let be the optimal state vector at time k.

[0039] Last updated error covariance: , In the formula Let I be the optimal error covariance matrix at time k, and let I be the identity matrix.

[0040] Repeat the iteration in 1-hour increments, adjusting every 24 hours based on assimilation error. Q, R To ensure the accuracy of short-term (0-72 hours) forecasts.

[0041] The assimilated 1-20km grid data is converted into NetCDF format, and high-risk area data corresponding to key icing thresholds (such as temperature ≤0℃ and humidity ≥85%) are extracted and pushed to the power grid dispatch center in real time via 4G / WiFi with a latency of ≤5 minutes, directly supporting icing forecasting and risk assessment.

[0042] Then, a dynamic downscaling method is used to achieve numerical modeling based on the climate microenvironment. A high-resolution grid (1-20km) covering the power grid equipment area is established, and the numerical weather model is run on this high-resolution grid to output high-resolution meteorological data, including temperature, humidity, wind speed, wind direction, etc.

[0043] Numerical weather forecasting (NWFR) dynamic downscaling is a technique that uses numerical weather models to simulate small-scale atmospheric processes to obtain climate data with higher spatial resolution. This method simulates local weather systems based on physical equations, capturing airflow changes caused by local features such as topography, ocean, and vegetation. The project will utilize currently mainstream numerical weather forecasting models, such as WRF, MM5, and GRAPS, to obtain data from global or large-scale climate models as initial conditions. This data typically has low spatial resolution (e.g., approximately 100 km). A high-resolution grid (e.g., 1-20 km) will be defined for the power grid area within the numerical weather model to better simulate local meteorological characteristics. The numerical weather model will be run on this defined grid, simulating physical processes such as convection, surface exchange, radiation balance, momentum transport, cloud physics, and microphysical processes. The ultimate goal is to output high-resolution meteorological data, including temperature, humidity, wind speed, and wind direction.

[0044] To address the need for power grid icing monitoring, a scaled-down grid range was determined to ensure that the output data accurately reflects the distribution of power grid equipment. A "nested grid" technique was used to divide the spatial grid, centered on the core areas of power grid operation and maintenance (such as along transmission lines and areas with concentrated substations). Outer mesh: Covers the province / region where the power grid is located, with a resolution of 10-20km, used to receive large-scale signals from FNL low-resolution data; Inner grid: Focus on high-risk areas of icing (such as power transmission lines in mountainous areas), with a resolution of 1-5km, ensuring that each grid can accurately correspond to 1-2 key power grid devices (such as poles and substations). The grid range must completely include the coordinates of all power grid devices to avoid data omission.

[0045] For power grid icing scenarios, the model parameterization scheme was adjusted to improve the accuracy of local microclimate simulation.

[0046] First, parameterization schemes were adapted. Schemes capable of accurately simulating "supercooled water droplets" (the core source of icing material) were selected, such as the WSM6 (WRF Single-Moment 6-class) scheme, which accurately outputs the supercooled water content in clouds, the size and concentration of raindrops / snow particles, supporting indirect estimation of icing amount. The YSU (Yonsei University) scheme was adopted to optimize the turbulent mixing simulation in complex terrain areas (such as mountainous regions), improving the simulation accuracy of near-surface (10m height) wind speed and temperature (near-surface meteorological elements directly affect the formation of icing on transmission lines). The RRTMG (Rapid Radiative Transfer Model for GCMs) scheme was selected to accurately calculate solar shortwave radiation and atmospheric longwave radiation, improving the diurnal variation characteristics of temperature simulation (diurnal temperature variation affects the freezing / melting process of icing).

[0047] Next, initial field and boundary condition interpolation is performed. The FNL low-resolution data (100km) is mapped to the outer grid (10-20km) of the downscaled model using the "bilinear interpolation method" as the initial field (atmospheric state at the initial moment) and boundary conditions (meteorological elements input at the grid boundary during the simulation). During the interpolation process, local topographic data needs to be combined to perform "topographic correction" on temperature and air pressure in areas with large altitude differences (such as the transition zone between mountains and plains) (e.g., temperature decreases by 0.6℃ for every 100m increase in altitude) to reduce interpolation errors.

[0048] High-resolution data is generated by running numerical weather prediction models, and quality control and format conversion are performed to ensure that the data can be directly used for power grid icing forecasting.

[0049] First, the numerical weather prediction model is set up. The time step is determined based on the grid resolution: 1-3 minutes for the inner grid (1-5km) and 5-10 minutes for the outer grid (10-20km), satisfying the "CFL condition" (ensuring numerical calculation stability). The simulation duration outputs short-term forecast data of 0-72 hours (covering the critical response period for power grid icing disasters), with a time resolution of 1 hour (matching the monitoring frequency of power grid operation and maintenance). The output elements focus on key meteorological elements related to icing, including temperature at 2m altitude, relative humidity at 2m altitude, wind speed / direction at 10m altitude, precipitation, precipitation phase (rain / snow / freezing rain), and supercooled water content in clouds.

[0050] The results are then post-processed. Outliers (such as wind speed > 50 m / s and temperature > 20℃) are removed from the model. Missing values ​​are filled in using either "mean replacement of adjacent grids" or "correction of historical data". The original data output by the model (such as WRF's NetCDF format) is converted to a format compatible with the existing power grid platform (such as ASCII and JSON). Meteorological data of the grid where the power grid equipment is located is extracted to generate a "equipment-meteorological element" correspondence table (such as hourly temperature and wind speed data of the grid where a certain tower is located). The measured data (minute-level temperature, humidity, and wind speed) from the power grid customized meteorological terminal (including micro-intelligent meteorological instrument) are compared with the downscaled output data to calculate the root mean square error (RMSE). The requirements are that the temperature RMSE ≤ 0.8℃ and the wind speed RMSE ≤ 1.0 m / s. If these requirements are not met, the model parameters are re-optimized (such as adjusting the boundary layer scheme).

[0051] The processed high-resolution data is combined with power grid operations to directly support icing forecasting and risk assessment.

[0052] Based on data on transmission line accidents and disasters, and according to the descriptions in power grid disaster records, the severity of the damage caused by icing weather to power grid operation is classified. This classification is determined by the degree, scope, and severity of the impact of icing disasters on the power grid. Disaster weather is categorized into different levels: ① Icing weather occurs, causing some impact on power supply and damaging power infrastructure; ② Large-scale damage to transmission and transformation facilities, power outages in some areas, and the collapse of some power lines and poles; ③ The duration of communication outages, the magnitude of economic losses, and the amount of manpower and material resources consumed are unprecedented in the province, reaching the highest recorded values, and many power lines and poles have been collapsed. Based on the above disaster descriptions, meteorological data, historical disaster data, and power grid operation data are collected. The data is cleaned and preprocessed, and data samples exhibiting one of the descriptions within a certain level are classified as belonging to that level.

[0053] By combining power grid operation and maintenance data, partial correlation analysis was conducted on meteorological factors during power grid icing weather to identify the main meteorological disaster-causing factors. Based on these main disaster-causing factors, a meteorological factor model was constructed to reclassify the level of icing meteorological disasters.

[0054] Based on high-resolution meteorological data that has undergone data assimilation (Kalman filtering) and dynamic downscaling, and combined with key icing-causing factors selected through partial correlation analysis (such as temperature T at 2m altitude, relative humidity RH at 2m altitude, wind speed V at 10m altitude, precipitation phase P, and supercooled water content S in clouds, which need to be verified by t-test / F-test with a significance level α≤0.05), a meteorological factor model is constructed according to the following steps: 1. Determine the model's input and output variables. Input variables: Select key hazard factors with a partial correlation coefficient absolute value ≥ 0.6 (or determined to be "highly significant correlation" by t-test / F-test), specifically including: (1) Temperature T at a height of 2m (unit: °C, core influencing factor of icing, usually T≤0℃ is the premise of icing); (2) Relative humidity RH at a height of 2m (unit: %, when RH≥85%, there is sufficient water vapor and ice formation is easy); (3) Wind speed V at 10m height (unit: m / s, wind speed affects ice thickness, usually ice growth is fastest when V∈[2,15]m / s). (4) Precipitation phase P (discrete variable, assigned values: 0 = rain, 1 = snow, 2 = freezing rain, freezing rain is a high-risk phase for icing); (5) Cloud supercooled water content S (unit: g / m³, which directly determines the source of icing material; when S≥0.5g / m³, the risk of icing increases significantly).

[0055] Output variable: Define “Icing Impact Index Y” (dimensionless, value range 0-100) to quantify the comprehensive impact of meteorological conditions on power grid icing. The larger the Y value, the higher the icing risk.

[0056] 2. Data Preprocessing and Sample Construction (1) Data sources: measured data collected by customized meteorological terminals, data assimilated from FNL reanalysis data, and 3km grid meteorological data output by dynamic downscaling, with a time resolution of 1 hour and spatial coverage of all grids where the power grid equipment is located; (2) Outlier handling: Outliers (such as unreasonable data such as wind speed > 30m / s and temperature > 10℃) are removed using the “3σ principle”. Missing values ​​are filled by “replacing with the mean of adjacent grids + correcting with historical data of the same period”. (3) Data normalization: To eliminate differences in dimensions, the input variables are normalized using the Min-Max method, as shown in the following formula: , in, X The original variable value, X min / X max This represents the minimum / maximum value of the variable in the historical samples. X norm These are the normalized variable values; (4) Sample set division: Select historical data of power grid icing disasters (last 5-10 years) as samples, of which 70% is the training set (used to fit model parameters) and 30% is the validation set (used to verify model accuracy). The samples must include "meteorological factor combination + corresponding actual icing disaster level" (e.g., a sample has T=-3℃, RH=90%, V=8m / s, P=2, S=0.8g / m³, corresponding to the actual disaster level "Level II icing").

[0057] 3. Model selection and parameter fitting Considering the practicality and computational efficiency of power grid operation and maintenance, a multiple linear regression model (or a simplified random forest model) is selected to construct the meteorological factor model, as follows: (1) Formula for multiple linear regression model: , Where: Y is the icing impact index (output variable); T norm , RH norm , V norm , S norm The values ​​are the normalized temperature at 2m altitude, relative humidity at 2m altitude, wind speed at 10m altitude, and supercooled water content in the clouds; P is the precipitation phase (discrete variable, directly involved in the calculation); a, b, c, d, and e are factor weight coefficients (which need to be fitted through the training set); and f is a constant term. (2) Parameter fitting method: The training set samples are fitted by the "least square method" to minimize the error between the Y value predicted by the model and the actual degree of ice impact of the samples (quantified by the historical disaster level, such as Level I = 10, Level II = 30, Level III = 50, Level IV = 80). (3) Weighting coefficient constraint: Adjust the sign of the coefficient according to the physical mechanism of icing, such as the coefficient of T (a < 0) (the lower the temperature, the larger Y), the coefficient of RH (b > 0) (the higher the humidity, the larger Y), and the coefficient of P (d > 0) (P = 2 during freezing rain, with the largest contribution).

[0058] 4. Model accuracy verification Using "accuracy of grade determination" and "root mean square error (RMSE) of icing impact index" as verification indicators, the requirements are as follows: (1) The model's accuracy in determining the severity of icing disasters on the validation set is ≥85%; (2) The RMSE of the icing impact index Y is ≤5 (i.e., the average deviation between the model prediction and the actual quantified value is ≤5). (3) If the accuracy requirements are not met, key disaster-causing factors need to be re-screened (such as adding the "ice-covering duration" factor) or the model needs to be optimized (such as using a random forest model to improve nonlinear fitting ability) until the accuracy standards are met.

[0059] A Method for Reclassifying Icing Meteorological Disaster Levels Based on Meteorological Factor Models Based on the "icing impact index Y" output by the meteorological factor model, and combined with the actual operation and maintenance needs of the power grid (such as equipment tolerance to icing thickness and power supply guarantee priority), the icing meteorological disaster level is reclassified. The specific steps are as follows: 1. Determining the core basis for grade classification Abandoning the traditional "single meteorological indicator judgment" (such as relying solely on temperature), this paper takes the icing impact index Y output by the model as the core, while also taking into account the vulnerability of power grid equipment (such as the lower icing tolerance threshold of old towers and high-altitude lines), forming a two-dimensional classification logic of "Y value + equipment vulnerability correction".

[0060] 2. Determination of the threshold for classifying levels Based on the statistical relationship between historical icing disasters and corresponding Y values, and combined with power grid operation and maintenance experience, the following level thresholds are determined (which can be fine-tuned according to the characteristics of different regional power grids), as shown in Table 1: Icing Disaster Level Judgment Table: Table 1. Icing Disaster Level Determination Table Ice cover disaster level Icing Impact Index Y Range Core judgment criteria (combining meteorological factors and actual impacts) Level I (Minor Risk) [0,20) The Y value is low, corresponding to the following meteorological conditions: T∈[-5,0)℃, RH∈[85,90)%, V∈[2,5)m / s, P=1 (snow), S∈[0.2,0.5)g / m³; Actual impact: slight icing on the insulators, no power outage, and no special emergency measures required. Level II (General Risk) [20,40) The Y value is moderate, corresponding to the following meteorological conditions: T∈[-10,-5)℃, RH∈[90,95)%, V∈[5,10)m / s, P=1 or 2 (snow / freezing rain), S∈[0.5,0.8) g / m³; Actual impact: Some towers are slightly iced (ice thickness 5-10mm), and a few remote areas experience short-term power outages, requiring the activation of "routine inspection + early warning monitoring". Level III (Severe Risk) [40,60) The Y value is relatively high, corresponding to the following meteorological conditions: T∈[-15,-10)℃, RH∈[95,100]%, V∈[10,15)m / s, P=2 (freezing rain), S∈[0.8,1.2)g / m³; Actual impact: less than 30% of the towers are moderately iced (ice thickness 10-20mm), power outages occur in multiple areas, and "emergency teams on standby + de-icing equipment deployment" needs to be activated. Level IV (Extreme Risk) [60,100] The Y value is extremely high, corresponding to the following meteorological conditions: T≤-15℃, RH∈[95,100]%, V≥15m / s, P=2 (freezing rain), S≥1.2g / m³; Actual impact: More than 30% of the towers are heavily iced (ice thickness ≥20mm), posing a risk of line breakage and tower collapse, and communication may be interrupted, requiring the activation of the "network-wide emergency response + personnel evacuation plan". .

[0061] 3. Dynamic calibration of grade thresholds To ensure that the classification aligns with the actual needs of the power grid, threshold calibration must be performed every six months based on the latest data. (1) Data update: Incorporate the customized meteorological terminal measured data of the past six months and the power grid icing disaster record (such as the newly added Level IV icing event). (2) Threshold adjustment: If a region has more than 3 instances of deviation where “the model predicts it to be level II but it is actually level III”, the model weight coefficients need to be refitted (e.g., increase the weight d of P in that region), or the threshold of Y value needs to be fine-tuned (e.g., reduce the lower limit of Y for level III in that region from 40 to 35). (3) Regional differentiation adjustment: For high-altitude (e.g., altitude > 1500m) and mountainous areas with high risk of icing, a separate “region-specific threshold” can be formulated (e.g., the lower limit of Y for Level IV in high-altitude areas is reduced to 55) to match the vulnerability of equipment in the region.

[0062] 4. Application of the classification results: The reclassified icing meteorological hazard level needs to be aligned with the risk assessment process. The classification results are used as input to the "Disaster Risk Index Assessment Model", where q iThe range of Y values ​​directly corresponding to the (disaster intensity) level (e.g., Level I q1=10, Level II q2=30); In the WebGIS visualization platform, the classification results are overlaid with the distribution of power grid equipment to generate an "equipment-level-risk" associated layer. For example, blue is used to mark level I risk towers and red is used to mark level IV risk towers, which supports operation and maintenance personnel in accurately formulating differentiated emergency measures.

[0063] Data on transmission line accidents during power grid operation and interpolated meteorological data were collected. Using a correlation coefficient matrix, matrix algebra was employed to calculate the partial correlation coefficients between the target variables, controlling for the influence of other variables. Statistical significance tests were then performed on the obtained partial correlation coefficients. This typically involves hypothesis testing, such as t-tests or F-tests, to determine whether the observed relationship is unlikely to be caused by random variation. Based on the magnitude and statistical significance of the partial correlation coefficients, the relationship between the explanatory variables was determined. If the partial correlation coefficient is significantly non-zero, it indicates that a correlation still exists between the two target variables after controlling for other variables.

[0064] Estimation of the probability of meteorological disasters caused by power grid icing based on an information diffusion theory assessment model: Based on the constructed disaster level index, the probability of power grid icing meteorological disaster risk is estimated using an information diffusion theory assessment model, and the probability of occurrence of different levels of icing meteorological disasters in the power grid operating area is calculated.

[0065] The core idea of ​​power grid meteorological disaster risk probability estimation based on the information diffusion theory assessment model is to treat the risk of power grid icing disasters as a kind of "information" and estimate the probability of the risk by simulating its diffusion process in the power grid system. First, the power grid system needs to be modeled as a network graph, where nodes represent various parts of the power grid (such as substations, transmission lines, etc.) and edges represent the connections between them. The information diffusion theory assessment model is as follows: Let the domain of disaster frequency be: , In the formula u i Let be the number of possible disasters of type i.

[0066] Let the sample set of disaster frequency be: , In the formula y j It is the j-th sample in the disaster frequency sample set.

[0067] right Y The j-th sample y j Diffusion is performed, and the calculation formula is as follows: , In the formula h The diffusion coefficient is used to control the width of information diffusion and can be calculated from the maximum and minimum values ​​in the sample set and the number of samples. f j ( u i ) represents the j-th sample y j For the i-th value in the universe of discourse u i The diffusion membership degree.

[0068] make C j The diffusion normalization coefficient for the j-th sample is: , The membership function of its corresponding fuzzy subset is: , right Y m After processing, the risk assessment result can be obtained.

[0069] That is, the above m data points (a total of m*n data points, corresponding to m samples and n types of disaster frequencies) need to be transformed into "quantitative risk results that can directly support the drawing of risk zoning maps" through subsequent processing. The specific steps are as follows: 1. Consistency verification of single-sample diffusion results (removal of outliers) Because actual observations may contain samples y j Outliers (such as records that are incorrect) j (far greater than the historical maximum), it is necessary to first analyze each Perform consistency checks to prevent the spread of abnormal results from affecting the accuracy of risk estimation. (1) Criteria for identifying outliers: If a certain sample y j Corresponding membership function An error is considered to occur if any of the following conditions are met: exist >1.05 or <-0.05 (outside the reasonable range of membership degree); Normalization coefficient C j <0.1 (excessive diffusion leads to distorted membership). The sample corresponds to <0.2 (for all disaster occurrences u) i The membership degrees are too low to be of any practical reference value.

[0070] (2) Outlier handling methods: If the percentage of abnormal samples is ≤5%, the corresponding sample will be removed directly. ; If the proportion of outlier samples is greater than 5%, the "neighboring sample correction method" is used: This method is applied to outlier sample y. j The two normal samples y that are closest in time / space j-1 y j+1 The membership function mean substitution, i.e.: , Objective: To ensure that everyone involved in the subsequent integration... All data are valid, avoiding outliers that could lead to biased risk assessments.

[0071] 2. Fusion calculation of multi-sample diffusion results (to obtain the comprehensive membership degree of the universe of discourse U) Since the m samples correspond to icing disaster records from different times / spaces, it is necessary to sort the m samples... The integration is expressed as "the comprehensive membership degree of the disaster frequency domain U", quantifying "the frequency of each disaster u". i "Overall correlation": (1) Selection of fusion algorithm: The weighted average method is adopted, with weight ω j Determine based on the sample's "confidence level"—sample y j The greater the actual impact of the corresponding icing disaster (e.g., the longer the power outage lasts and the more users affected), the higher the confidence level and the greater the weight, ensuring that high-impact samples contribute more strongly to risk estimation. (2) Weight ω j Calculation method: For each sample y j The basic weights are assigned according to the corresponding ice accretion disaster level (Level I-IV): Level I (minor) = 1, Level II (moderate) = 2, Level III (severe) = 3, Level IV (extreme) = 4; Combining this with the observation accuracy of the sample (e.g., adding 0.5 to the sample weight of measured data from customized meteorological terminals, and indirectly subtracting 0.3 from the sample weight), the final weight ω j Normalization is required The formula is: (ω) base,j ω represents the basic weight corresponding to the j-th sample. acc,j ω represents the precision weight corresponding to the j-th sample. base,k ω represents the basic weight corresponding to the k-th sample. acc,k (The precision weight corresponding to the k-th sample). (3) Calculation of overall membership degree: For each type of disaster number u i Calculate the sum of weighted membership degrees for all samples to obtain the overall membership degree R(u). i ): (R(u) i (∈[0,1]), the larger the value, the more likely "u" occurs in this region. i The higher the correlation with "secondary icing disasters".

[0072] Objective: To integrate the diffusion results of multiple discrete samples into a unified "universe of discourse-integrated membership" relationship, laying the foundation for subsequent quantification of risk probability.

[0073] 3. Quantitative conversion of risk probability (converting comprehensive membership degree into occurrence probability) Overall membership degree R(u) i This only indicates the "degree of correlation" and needs to be further converted into the "probability of disaster occurrence" commonly used in power grid risk assessment in order to directly support the "probability classification" of the risk zoning map. (1) Probability transformation logic: Since R(u i This has already reflected the various u i The relative correlation is converted into the probability of occurrence P(u) using the "normalization method". i (i.e., "of all possible disaster occurrences, u...") i The probability of (times) (2) Probability calculation formula: It satisfies the basic axiom of probability, P(u i The larger the value, the more likely u is to occur in that region. i The higher the probability of secondary ice accumulation disasters. (3) Risk probability result output: The final output is a table of correspondence between "number of disasters - probability of occurrence".

[0074] Objective: To transform abstract membership degrees into concrete probability values, providing a quantitative basis for setting the probability threshold of risk levels in risk zoning maps.

[0075] 4. Risk level mapping (corresponding to icing disaster levels, linking zoning maps) Risk probability P(u) i It needs to be further mapped to "power grid icing risk level" (Level I-IV disaster level) before it can be directly used for "grading and coloring" in the risk zoning map: (1) Risk level mapping basis: combining "probability of disaster frequency" and "intensity of single disaster" (q in the instruction manual) i For Level I, q1=10; for Level II, q2=30; for Level III, q3=50; and for Level IV, q4=80, calculate the "expected risk value E" as the core indicator for level mapping—the larger E is, the higher the overall risk of the region. (2) Formula for calculating expected risk value E: , in: This represents the average intensity of historical ice-covering disasters in the region. For example, if a region is primarily affected by Level II disasters, =30; The unit of E is "risk index unit", which is compatible with the disaster risk index M; (3) Risk level mapping standard: Based on the range of values ​​of E, the corresponding icing disaster risk level is determined.

[0076] The probability results are converted into "risk levels" that are directly matched with the risk zoning map, realizing a complete link of "information dissemination results → risk levels → visual colors".

[0077] Drawing of risk zoning maps for ice accumulation disasters Using a disaster risk index assessment model, a risk zoning map of ice accumulation disasters was drawn.

[0078] The disaster risk index refers to the probability of n levels of power grid meteorological disasters occurring within a certain area and time period. Taking into account the intensity of different levels of disasters and the corresponding probability of occurrence, a weighted summation model is used to construct a disaster risk index assessment model for power line icing disasters. , In the formula, M is the disaster risk index, n is the number of disaster levels, and p i Let q be the probability of the i-th level disaster occurring. i Let be the intensity of the i-th level of disaster.

[0079] A risk zoning map is created using a tiered color scheme. Based on the calculated probability of disaster occurrence and disaster risk index, areas are divided into different levels, such as low risk, medium risk, high risk, and extremely high risk, represented by colors such as blue, yellow, orange, and red, respectively. For example, in the color scheme map of icing disaster risk grading for a certain province, areas with darker colors indicate a higher risk of icing.

[0080] Based on observational data and multi-temporal time-series data of meteorological forecasts, and through the analysis of the needs and application scenarios of basic management, meteorological monitoring, meteorological forecasting, and meteorological early warning, this system provides gridded display of forecast and early warning information, query display and statistical evaluation technology for power grid icing meteorological disaster risk forecasts in specific ranges and regions, and is based on a GIS visualization platform and a high-resolution geographic base map.

[0081] GIS (Geographic Information System)-based power grid weather forecasting and early warning technology is a technology that combines meteorological data, power grid data, and geographic information. It aims to improve the power grid's ability to warn of extreme weather events and ensure the safe and stable operation of the power system. It consists of several components: meteorological data (including real-time weather observation and forecast data, such as temperature, humidity, wind speed, and rainfall); power grid data (including infrastructure information such as the location and status of transmission lines, substations, and distribution networks); a GIS platform (providing map display, spatial analysis, and data management functions, serving as a visualization tool integrating meteorological and power grid data); an early warning model (assessing the potential impact of extreme weather on the power grid and determining the early warning level based on meteorological and power grid data); and a communication system (used to transmit real-time meteorological data, power grid status information, and early warning information).

[0082] Power grid icing disaster risk assessment system The customized numerical weather prediction module collects global / regional climate model data (FNL reanalysis data) and meteorological station observation data in the power grid area to establish a high-resolution grid numerical weather prediction model covering the power grid equipment area. The model outputs basic meteorological data such as temperature, humidity, and wind speed. The initial field and boundary conditions of the model are optimized by combining local meteorological observation data, and the parameterization scheme is adjusted to simulate icing-related meteorological elements, forming a foundation for refined icing meteorological numerical weather prediction technology tailored to the needs of power grid operation. Meteorological data assimilation is carried out using mainstream numerical weather prediction models, and dynamic downscaling methods are employed to achieve numerical modeling based on the climate microenvironment.

[0083] The disaster classification and risk assessment module classifies disasters into levels (Level I to Level IV) based on power grid icing disaster records and historical icing meteorological data, according to the degree, scope, and severity of the impact of icing disasters on the power grid. Based on power grid operation and maintenance data and historical disaster information, partial correlation analysis is used to control for the influence of other variables and screen key icing meteorological disaster-causing factors (humidity, temperature, wind speed, etc.). A correlation between disaster levels and meteorological conditions is established, ultimately forming a classification standard for icing disasters and a disaster-causing factor index library, providing a basis for subsequent risk assessment. Then, based on information diffusion theory and a disaster risk index assessment model, combined with disaster intensity, probability of occurrence, and power grid vulnerability, the disaster risk level for different regions is calculated. Through statistical analysis of historical icing disasters and meteorological data, a spatial distribution map of icing risk probability is generated, and power grid facility distribution information is overlaid to quantify the regional risk level. Finally, a risk zoning map of icing disasters is output.

[0084] The power grid meteorological early warning visualization platform module, relying on WebGIS technology, establishes a platform for visualizing early warnings of power grid icing meteorological disasters. It integrates real-time meteorological data, power grid operational status information, and risk assessment results, enabling multi-layer overlay display and interactive operation. Visualization includes meteorological data (such as real-time weather observation and forecast data); power grid data (such as power grid infrastructure information); geospatial information data; and early warning data (based on meteorological and power grid data, assessing the potential impact of extreme weather on the power grid and determining the early warning level). This results in a power grid icing meteorological disaster early warning technology with visualization and early warning information dissemination capabilities, capable of intuitively presenting meteorological change trends and the distribution of power grid icing risks.

[0085] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention. These changes involve related technologies well known to those skilled in the art, and all of them fall within the protection scope of the present invention.

[0086] Many other changes and modifications can be made without departing from the concept and scope of this invention. It should be understood that this invention is not limited to the specific embodiments, and the scope of this invention is defined by the appended claims.

Claims

1. A method for assessing the risk of power grid icing disasters, characterized in that, Includes the following steps: Step 1: Collect data and build a multi-source data system; By deploying customized meteorological terminals and combining FNL reanalysis data, power grid disaster and operation and maintenance data, a high-resolution grid numerical weather prediction model covering the power grid equipment area is established. The numerical weather prediction model outputs basic meteorological data including temperature, humidity and wind speed. Step 2: Process the data to improve forecast accuracy; use the Kalman filter algorithm to assimilate the data and then use the dynamic downscaling method to improve the resolution of the meteorological data grid. Step 3: Use partial correlation analysis to screen key factors causing ice accumulation disasters; Step 4: Classify disaster levels and quantify risks, and draw a disaster risk zoning map; Step 5: Visualization and Early Warning. A visualization platform is built based on WebGIS technology to integrate and display meteorological data, power grid data, geospatial information data, and early warning data.

2. The power grid icing disaster risk assessment method according to claim 1, characterized in that, The customized meteorological terminal in step 1 includes a micro-intelligent meteorological instrument and an intelligent terminal system. The micro-intelligent meteorological instrument is deployed around the power transmission lines of meteorological stations in the power grid area to acquire meteorological data and wirelessly transmit it to the intelligent terminal system. The intelligent terminal integrates a global geographic location information system that supports GNSS and wirelessly transmits meteorological data and location data to a visualization platform. The intelligent weather instrument includes an integrated meteorological element sensor and a tipping bucket rain gauge.

3. The power grid icing disaster risk assessment method according to claim 2, characterized in that, The numerical weather prediction model in step 1 uses the WRF4.1 model for numerical simulation. FNL reanalysis data with a resolution of 1°×1° and a frequency of once every 6 hours is selected as the initial field conditions. The numerical simulation adopts a double-layer nested structure, including a coarse grid and a fine grid, with horizontal resolutions of 9km and 3km, respectively, and corresponding grid numbers of 649×541 and 673×676, respectively. The top of the model is set to 50hPa, and the vertical direction is divided into 51 layers.

4. The power grid icing disaster risk assessment method according to claim 3, characterized in that, The Kalman filter algorithm in step 2 assimilates the meteorological data by combining the measured data from the customized meteorological terminal with the meteorological data from the numerical weather prediction model. Using each icing meteorological element in the power grid as a state variable, a linear equation is constructed: ， in X k Let k be the state vector at time k. A Here is the state transition matrix. B To control variables, u k For control items such as topography and radiation, w k-1 The state noise is Gaussian distributed; Establish a mapping between observation data and state variables: ， In the formula Z k These are the preprocessed observations. H For sparse observation matrices, v k The observation noise is Gaussian distributed. Use the optimal state from the previous moment to deduce the current predicted state: ， In the formula Predict the state vector at time k. The optimal state vector at time k-1; Simultaneous calculation of prediction error covariance: ， In the formula Let k be the prediction error covariance matrix. The optimal error covariance matrix at time k-1 is... Let A be the state noise covariance matrix. T Transpose of the state transition matrix; Let A be the state noise covariance matrix. T Transpose of the state transition matrix; ， In the formula The Kalman gain at time k, To observe the noise covariance matrix, H T Observation matrix transpose; The forecast is then corrected using measured data from a customized meteorological terminal to obtain the optimal state: ， In the formula Let k be the optimal state vector at time k; Last updated error covariance: ， In the formula Let I be the optimal error covariance matrix at time k, where I is the identity matrix; Repeat the iteration in 1-hour increments, adjusting every 24 hours based on assimilation error. Q, R To ensure the accuracy of short-term 0-72 hour forecasts, the assimilated 1-20km grid data is converted into NetCDF format, high-risk area data corresponding to key icing thresholds are extracted, and data is pushed to the visualization platform in real time via wireless transmission, with a latency of ≤5 minutes.

5. The power grid icing disaster risk assessment method according to claim 4, characterized in that, The dynamic downscaling method in step 2 runs the numerical weather model on a defined grid, which includes an inner grid and an outer grid. First, the numerical weather prediction model is parameterized and adapted. Then, the initial field and boundary conditions are interpolated. The FNL low-resolution data is mapped to the outer grid of the downscaled model through bilinear interpolation as the initial field and boundary conditions of the model. During the interpolation process, local topographic data needs to be combined to correct the temperature and air pressure in areas with large altitude differences. Numerical weather forecast model operation settings: The time step is determined according to the grid resolution. The inner grid (1-5km) is set to 1-3 minutes, and the outer grid (10-20km) is set to 5-10 minutes to meet the CFL condition. The simulation duration outputs short-term forecast data of 0-72 hours, with a time resolution of 1 hour. The output elements focus on key meteorological elements of icing, including temperature at 2m height, relative humidity at 2m height, wind speed / direction at 10m height, precipitation, precipitation phase, and supercooled water content in clouds. Post-processing of results; Outliers in the numerical weather prediction model were removed, and missing values ​​were filled in by replacing them with the mean of adjacent grids or correcting them with historical data from the same period. The original data format output by the numerical weather prediction model was converted into a format compatible with the existing power grid platform, and meteorological data of the grid where the power grid equipment is located were extracted to generate a "equipment-meteorological element" correspondence table. The measured data from customized meteorological terminals were compared with the downscaled output data to calculate the root mean square error (RMSE). The requirements were that the root mean square error of temperature ≤ 0.8℃ and the root mean square error of wind speed ≤ 1.0m / s. If these requirements were not met, the model parameters were optimized by adjusting the boundary layer.

6. The power grid icing disaster risk assessment method according to claim 1 or 5, characterized in that, In step 3, partial correlation analysis is used to screen key icing-causing factors. The t-test or F-test is used to screen key icing-causing factors. A meteorological factor model is constructed based on the key icing-causing factors to reclassify the meteorological disaster level of icing.

7. The power grid icing disaster risk assessment method according to claim 1, characterized in that, In step 4, disaster levels are classified and risks are quantified. The information diffusion theory assessment model is used to quantify the risks, and the information diffusion theory assessment model is as follows: Let the domain of disaster frequency be: ， In the formula u i Let i be the number of possible disasters; Let the sample set of disaster frequency be: ， In the formula y j This refers to the j-th sample in the disaster frequency sample set; right Y The j-th sample y j For diffusion, the calculation formula is: ， In the formula h The diffusion coefficient, used to control the breadth of information diffusion, can be calculated using the maximum and minimum values ​​in the sample set and the number of samples. f j ( u i ) represents the j-th sample y j For the i-th value in the universe of discourse u i The diffusion membership degree; make C j The diffusion normalization coefficient for the j-th sample is: ， The membership function of its corresponding fuzzy subset is: ， right Y m After processing, the risk assessment result can be obtained.

8. The power grid icing disaster risk assessment method according to claim 1, characterized in that, In step 4, a disaster risk zoning map is drawn using a disaster risk index assessment model. The disaster risk index assessment model is as follows: ， In the formula, M is the disaster risk index, n is the number of disaster levels, and p i Let q be the probability of the i-th level disaster occurring. i Let be the intensity of the i-th level of disaster.

9. A power grid icing disaster risk assessment system, characterized in that, The method for assessing the risk of power grid icing disasters as described in any one of claims 1-8 includes: The customized numerical weather prediction module collects global / regional climate model data, FNL reanalysis data, and meteorological station observation data in the power grid area to establish a high-resolution grid numerical weather prediction model covering the power grid equipment area, and outputs meteorological data. The meteorological disaster classification and risk assessment module, based on power grid icing disaster records and historical icing meteorological data, classifies icing disasters into levels I to IV according to their impact on the power grid, scope, and severity. Based on operational meteorological data and historical disaster information, it uses partial correlation analysis to control for the influence of other variables, screens key icing meteorological disaster-causing factors, establishes the correlation between disaster levels and meteorological conditions, and ultimately forms a classification standard for icing disasters and a disaster-causing factor index library. Based on the information diffusion theory assessment model and the disaster risk index assessment model, combined with disaster intensity, occurrence probability, and power grid vulnerability, it calculates the disaster risk level of different regions. Through statistical analysis of historical icing disaster records and meteorological data, it generates a spatial distribution map of icing risk probability, and overlays power grid facility distribution information to quantify regional risk levels, outputting a risk zoning map of icing disasters. The power grid meteorological early warning visualization platform module, relying on WebGIS technology, builds a power grid icing meteorological disaster early warning visualization platform, integrating real-time meteorological data, power grid operation status information and risk assessment results, and realizes multi-layer overlay display and interactive operation.

10. The power grid icing disaster risk assessment system according to claim 9, characterized in that, The numerical weather prediction model is optimized by combining observation data from customized meteorological terminals with the initial field and boundary conditions of the numerical weather prediction model, and the parameterization scheme is adjusted to simulate icing-related meteorological elements.

Images