Power grid meteorological early warning methods, devices, computer equipment, readable storage media and program products

Abstract

This application relates to a method, device, computer equipment, computer-readable storage medium, and computer program product for power grid meteorological early warning. The method includes: acquiring multiple heterogeneous meteorological probability field data within a target power grid area and performing spatiotemporal dimension concatenation and alignment processing to obtain aligned meteorological probability field data; using a nonlinear cross-validation operator to achieve fusion processing of the aligned meteorological probability field data to obtain a joint comprehensive meteorological disaster probability calculation equation; performing stepped risk level discretization mapping processing to determine the stepped risk level of the virtual spatial grid within the area; further determining the target topology state and state transition information corresponding to the target power grid area; and generating meteorological disaster prevention control signals. This application effectively eliminates the problem of conflict between multiple model forecast results by fusing multiple meteorological data through a nonlinear cross-validation operator, thereby effectively improving the effectiveness and accuracy of meteorological early warning under severe convective weather.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a power grid weather early warning method, device, computer equipment, computer-readable storage medium, and computer program product. Background Technology

[0002] With the evolution of new power systems and the increasing frequency of global climate anomalies, extreme convective weather such as thunderstorms and short-duration heavy rainfall poses a serious threat to the safe and stable operation of provincial power grids. These types of hazardous weather are characterized by their localized spatial scale, rapid formation and dissipation, and significant nonlinearity, easily triggering sudden lightning flashovers at high-voltage transmission sections, mechanical icing damage to long-span towers, damage from instantaneous strong winds, and localized substation flooding—all of which can lead to a surge of power grid emergencies. In the current digital construction of large power grids, the introduction of artificial intelligence algorithms for meteorological disaster prevention and mitigation has become a mainstream trend.

[0003] Currently, the industry is beginning to explore the introduction of large-scale deep learning network models to extrapolate macroscopic multi-physical background fields, and to combine them with high-resolution spatiotemporal extrapolation networks to extrapolate and predict near-end radar echoes and high-resolution satellite data. This primarily employs deterministic location forecasting methods based on a single scale (such as relying solely on conventional numerical weather prediction or single Doppler radar echo extrapolation) to handle power grid-related meteorological forecasts. However, this forecasting method often encounters the dilemma of conflicting and non-convergent forecasts from multiple models during the incubation and development stages of severe convective weather, resulting in persistently high rates of false alarms and missed alarms for the power grid. Summary of the Invention

[0004] Therefore, it is necessary to provide a power grid meteorological early warning method, device, computer equipment, computer-readable storage medium, and computer program product that can improve the accuracy of meteorological early warning under severe convective weather, in response to the above-mentioned technical problems.

[0005] Firstly, this application provides a power grid meteorological early warning method, including:

[0006] Acquire multiple heterogeneous meteorological probability field data within the target power grid area, and perform spatiotemporal dimension concatenation and alignment processing on the multiple heterogeneous meteorological probability field data to obtain the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data.

[0007] By using a nonlinear cross-validation operator, the aligned meteorological probability field data corresponding to each of the multi-path heterogeneous meteorological probability field data are fused to obtain a joint comprehensive meteorological disaster probability calculation equation.

[0008] Based on the meteorological disaster probability value determined by the joint comprehensive meteorological disaster probability calculation equation, the virtual spatial grid in the target power grid area is subjected to a step risk level discretization mapping process to determine the step risk level corresponding to each of the virtual spatial grids;

[0009] Based on the tiered risk levels corresponding to each of the virtual space grids, the target topology state and state transition information corresponding to the target power grid area are determined.

[0010] Based on the target topology state and state transition information, a meteorological disaster prevention and control signal for the target power grid area is generated.

[0011] In one embodiment, the spatiotemporal dimensional concatenation and alignment processing of the multi-channel heterogeneous meteorological probability field data to obtain aligned meteorological probability field data corresponding to each of the multi-channel heterogeneous meteorological probability field data includes:

[0012] Based on the scale information corresponding to the multi-path heterogeneous meteorological probability field data, a unified spatial reference scale is determined;

[0013] The multi-channel heterogeneous meteorological probability field data is dynamically resampled by interpolation operators, and the multi-channel heterogeneous meteorological probability field data is projected and transformed onto a virtual spatial grid under the unified spatial reference scale to obtain the spatially aligned meteorological probability field data corresponding to each of the multi-channel heterogeneous meteorological probability field data.

[0014] By using a dynamic time sliding window, the spatially aligned meteorological probability field data is processed to perform time-series frame alignment, thereby obtaining the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data.

[0015] In one embodiment, the multi-path heterogeneous meteorological probability field data includes refined radar spatiotemporal short-term extrapolation probability field data, extreme severe convective cell identification probability field data, and medium- and long-term large model ensemble forecast precipitation probability field data. The nonlinear cross-validation operator includes a short-term extrapolation time series weight function, an extreme convection time series weight function, and a medium- and long-term time series compensation weight function.

[0016] The process of fusing the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data using a nonlinear cross-validation operator to obtain the joint comprehensive meteorological disaster probability calculation equation includes:

[0017] The following methods are employed: constructing the refined radar spatiotemporal short-term extrapolation probability equation and short-term extrapolation time series weight function corresponding to the refined radar spatiotemporal short-term extrapolation probability field data; constructing the extreme severe convective cell identification probability equation and extreme convection time series weight function corresponding to the extreme severe convective cell identification probability field data; constructing the medium- and long-term large model ensemble forecast precipitation probability equation and medium- and long-term time series compensation weight function corresponding to the medium- and long-term large model ensemble forecast precipitation probability field data; wherein the short-term extrapolation time series weight function is a nonlinear decay function that dynamically evolves with the forecast lead time; the extreme convection time series weight function is a weight control function that exhibits linear dissipation as it dynamically evolves with the forecast lead time; and the medium- and long-term time series compensation weight function is a weight control function that exhibits logarithmic nonlinear gradual increase as it dynamically evolves with the forecast lead time.

[0018] Based on the refined radar spatiotemporal short-term extrapolation probability equation and the short-term extrapolation time series weight function, the first meteorological disaster probability calculation equation is determined; based on the extreme severe convective cell identification probability equation and the extreme convective time series weight function, the second meteorological disaster probability calculation equation is determined; based on the medium- and long-term large model ensemble forecast precipitation probability equation and the medium- and long-term time series compensation weight function, the third meteorological disaster probability calculation equation is determined.

[0019] Based on the first meteorological disaster probability calculation equation, the second meteorological disaster probability calculation equation, and the third meteorological disaster probability calculation equation, the aligned meteorological probability field data corresponding to each of the multi-path heterogeneous meteorological probability field data are fused to obtain the joint comprehensive meteorological disaster probability calculation equation.

[0020] In one embodiment, the step of performing a tiered risk level discretization mapping process on the virtual spatial grid within the target power grid area based on the meteorological disaster probability value determined by the joint comprehensive meteorological disaster probability calculation equation, and determining the tiered risk level corresponding to each virtual spatial grid, includes:

[0021] Based on the joint comprehensive meteorological disaster probability calculation equation, the meteorological disaster probability value corresponding to each virtual spatial grid within the target power grid area is determined;

[0022] When the meteorological disaster probability value is lower than the preset disaster threshold, the virtual space grid that is lower than the preset disaster threshold is pruned to obtain the pruned target virtual space grid;

[0023] Based on the meteorological disaster probability value corresponding to the target virtual space grid, the target virtual space grid is discretized and mapped to various risk levels.

[0024] In one embodiment, generating the meteorological disaster prevention and control signal for the target power grid area based on the target topology state and state transition information includes:

[0025] Determine the virtual space grid of the transmission channels within the target power grid area, and determine the tiered risk level of the virtual space grid of the transmission channels;

[0026] The tiered risk levels of the channel virtual space grid are mapped to the target topology state corresponding to the target power grid area;

[0027] Obtain the current topology state corresponding to the target power grid area;

[0028] By comparing the current topology state with the target topology state, the state transition information corresponding to the target power grid area is obtained.

[0029] In one embodiment, the target topology state includes normal attention state, situational awareness state, defense zone isolation state, emergency pre-positioning state, and recovery and alarm cancellation state, and the meteorological disaster prevention control signal includes standard situational awareness inspection and maintenance signal, digital meteorological defense pre-positioning prompt signal, digital isolation command, highest level disaster prevention linkage defense control signal, and buffer alarm cancellation signal;

[0030] The step of generating the meteorological disaster prevention and control signal for the target power grid area based on the target topology state and state transition information includes:

[0031] When the target topology is in a state of normal interest and the state transition information indicates that the state of normal interest is maintained, a standard situational awareness inspection maintenance signal is generated.

[0032] When the target topology is in a situational awareness state, and the state transition information represents a one-way transition from a normal state of concern to a situational awareness state, a digital meteorological defense advance warning signal is generated.

[0033] When the target topology is in zone isolation state and the state transition information indicates a state transition to zone isolation state, a digital isolation command is generated.

[0034] When the target topology is in an emergency pre-state and the state transition information indicates a strong state transition to the emergency pre-state, the highest level disaster prevention and control linkage signal is generated.

[0035] When the target topology is in the recovery and alarm cancellation state, and the state transition information indicates a one-way state transition to the recovery and alarm cancellation state, a buffered alarm cancellation signal is generated.

[0036] Secondly, this application also provides a power grid weather early warning device, comprising:

[0037] The data acquisition module is used to acquire multiple heterogeneous meteorological probability field data within the target power grid area, and to perform spatiotemporal dimension concatenation and alignment processing on the multiple heterogeneous meteorological probability field data to obtain the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data.

[0038] The probability field fusion module is used to fuse the aligned meteorological probability field data corresponding to each of the multi-path heterogeneous meteorological probability field data through a nonlinear cross-validation operator to obtain a joint comprehensive meteorological disaster probability calculation equation.

[0039] The risk ladder mapping module is used to perform ladder risk level discretization mapping processing on the virtual spatial grid in the target power grid area based on the meteorological disaster probability value determined by the joint comprehensive meteorological disaster probability calculation equation, and to determine the ladder risk level corresponding to each virtual spatial grid.

[0040] The state machine processing module is used to determine the target topology state and state transition information corresponding to the target power grid area based on the tiered risk level corresponding to each of the virtual space grids.

[0041] The control signal generation module is used to generate meteorological disaster prevention control signals for the target power grid area based on the target topology state and state transition information.

[0042] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0043] Acquire multiple heterogeneous meteorological probability field data within the target power grid area, and perform spatiotemporal dimension concatenation and alignment processing on the multiple heterogeneous meteorological probability field data to obtain the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data.

[0044] By using a nonlinear cross-validation operator, the aligned meteorological probability field data corresponding to each of the multi-path heterogeneous meteorological probability field data are fused to obtain a joint comprehensive meteorological disaster probability calculation equation.

[0045] Based on the meteorological disaster probability value determined by the joint comprehensive meteorological disaster probability calculation equation, the virtual spatial grid in the target power grid area is subjected to a step risk level discretization mapping process to determine the step risk level corresponding to each of the virtual spatial grids;

[0046] Based on the tiered risk levels corresponding to each of the virtual space grids, the target topology state and state transition information corresponding to the target power grid area are determined.

[0047] Based on the target topology state and state transition information, a meteorological disaster prevention and control signal for the target power grid area is generated.

[0048] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0049] Acquire multiple heterogeneous meteorological probability field data within the target power grid area, and perform spatiotemporal dimension concatenation and alignment processing on the multiple heterogeneous meteorological probability field data to obtain the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data.

[0050] By using a nonlinear cross-validation operator, the aligned meteorological probability field data corresponding to each of the multi-path heterogeneous meteorological probability field data are fused to obtain a joint comprehensive meteorological disaster probability calculation equation.

[0051] Based on the meteorological disaster probability value determined by the joint comprehensive meteorological disaster probability calculation equation, the virtual spatial grid in the target power grid area is subjected to a step risk level discretization mapping process to determine the step risk level corresponding to each of the virtual spatial grids;

[0052] Based on the tiered risk levels corresponding to each of the virtual space grids, the target topology state and state transition information corresponding to the target power grid area are determined.

[0053] Based on the target topology state and state transition information, a meteorological disaster prevention and control signal for the target power grid area is generated.

[0054] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0055] Acquire multiple heterogeneous meteorological probability field data within the target power grid area, and perform spatiotemporal dimension concatenation and alignment processing on the multiple heterogeneous meteorological probability field data to obtain the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data.

[0056] By using a nonlinear cross-validation operator, the aligned meteorological probability field data corresponding to each of the multi-path heterogeneous meteorological probability field data are fused to obtain a joint comprehensive meteorological disaster probability calculation equation.

[0057] Based on the meteorological disaster probability value determined by the joint comprehensive meteorological disaster probability calculation equation, the virtual spatial grid in the target power grid area is subjected to a step risk level discretization mapping process to determine the step risk level corresponding to each of the virtual spatial grids;

[0058] Based on the tiered risk levels corresponding to each of the virtual space grids, the target topology state and state transition information corresponding to the target power grid area are determined.

[0059] Based on the target topology state and state transition information, a meteorological disaster prevention and control signal for the target power grid area is generated.

[0060] The aforementioned power grid meteorological early warning method, device, computer equipment, computer-readable storage medium, and computer program product acquire multiple heterogeneous meteorological probability field data within the target power grid area, and perform spatiotemporal dimension concatenation and alignment processing on the multiple heterogeneous meteorological probability field data to obtain the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data; through a nonlinear cross-validation operator, the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data are fused to obtain a joint comprehensive meteorological disaster probability calculation equation; based on the meteorological disaster probability value determined by the joint comprehensive meteorological disaster probability calculation equation, a tiered risk level discretization mapping processing is performed on the virtual spatial grid within the target power grid area to determine the tiered risk level corresponding to each virtual spatial grid; based on the tiered risk level corresponding to each virtual spatial grid, the target topology state and state transition information corresponding to the target power grid area are determined; and based on the target topology state and state transition information, a meteorological disaster prevention and control signal for the target power grid area is generated. After aligning multiple heterogeneous meteorological probability field data, this application uses a nonlinear cross-validation operator to fuse these data to obtain a joint comprehensive meteorological disaster probability calculation equation. This effectively eliminates the problem of conflict between forecast results from multiple models. Then, based on the meteorological disaster probability determined by the equation, the virtual spatial grid of the region is mapped to the corresponding tiered risk level. Through state division and state switching identification processing, meteorological disaster prevention and control signals are generated, which can effectively improve the effectiveness and accuracy of meteorological warnings under severe convective weather. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 This is an application environment diagram of the power grid meteorological early warning method in one embodiment;

[0063] Figure 2 This is a flowchart illustrating a power grid weather warning method in one embodiment;

[0064] Figure 3This is a topology diagram of a multi-scale meteorological joint early warning process for a power grid in one embodiment;

[0065] Figure 4 This is a flowchart illustrating the power grid weather warning method in another embodiment;

[0066] Figure 5 This is a structural block diagram of a power grid weather warning device in one embodiment;

[0067] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0069] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0070] The power grid weather early warning method provided in this application embodiment can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or placed on a cloud or other network server. Server 104 can collect multiple heterogeneous meteorological probability field data within the target power grid area and perform spatiotemporal dimension concatenation and alignment processing on the multiple heterogeneous meteorological probability field data to obtain the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data. Using a nonlinear cross-validation operator, the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data are fused to obtain a joint comprehensive meteorological disaster probability calculation equation. Based on the meteorological disaster probability value determined by the joint comprehensive meteorological disaster probability calculation equation, a tiered risk level discretization mapping process is performed on the virtual spatial grid within the target power grid area to determine the tiered risk level corresponding to each virtual spatial grid. Based on the tiered risk level corresponding to each virtual spatial grid, the target topology state and state transition information corresponding to the target power grid area are determined. Based on the target topology state and state transition information, a meteorological disaster prevention and control signal for the target power grid area is generated. The meteorological disaster prevention and control signals are then transmitted to various terminals 102 to guide the management of meteorological disaster monitoring and prevention in the target power grid area. Terminals 102 can be, but are not limited to, various personal computers, laptops, smartphones, tablets, drones, low-altitude aircraft, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart vehicle devices, and projection equipment. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted displays. Head-mounted displays can be virtual reality (VR) devices, augmented reality (AR) devices, and smart glasses. Server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0071] In one exemplary embodiment, such as Figure 2 As shown, a power grid meteorological early warning method is provided, which can be applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps 201 to 209. Wherein:

[0072] Step 201: Obtain multiple heterogeneous meteorological probability field data within the target power grid area, and perform spatiotemporal dimension concatenation and alignment processing on the multiple heterogeneous meteorological probability field data to obtain the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data.

[0073] The target power grid area refers to the area requiring disaster prevention monitoring for the power grid meteorological early warning in this application. Specifically, the target power grid area can be the provincial power grid main grid disaster prevention responsibility area, centered on the corridor grid along assets such as 500kV and above main grid transmission channels, important transmission sections, long-span tower groups, and local substations. It also includes areas such as 220kV and below distribution networks and distributed new energy power stations, and is superimposed with provincial administrative boundaries. Multi-channel heterogeneous meteorological probability field data refers to at least three meteorological probability fields that are synchronously aggregated through a digital disaster prevention interface, describing the same target power grid area but being "heterogeneous" to each other. Heterogeneity is reflected in the differences in four dimensions: data source, physical quantity, temporal attribute, and spatial attribute. In a specific embodiment, multi-channel heterogeneous meteorological probability field data may include refined radar spatiotemporal short-term extrapolation probability field data, extreme strong convective cell identification probability field data, and medium- and long-term large model ensemble forecast precipitation probability field data, etc. Spatiotemporal dimensional cascade alignment processing refers to the process of aligning multiple heterogeneous meteorological probability field data along the time and spatial dimensions, so that the data from each channel are aligned on the time axis and spatial grid. Aligning meteorological probability field data means that after the original probability fields from each channel have undergone spatiotemporal dimensional cascade alignment processing, the resulting probability field data has a unified timestamp and a unified nested grid.

[0074] For example, server 104 can periodically acquire multiple heterogeneous meteorological probability field data within the target power grid area through a digital disaster prevention interface, thereby completing the basic data collection and processing for the power grid meteorological early warning process. Then, for the collected multiple heterogeneous meteorological probability field data, spatial dimension concatenation alignment and temporal dimension concatenation alignment processing are performed to obtain aligned meteorological probability field data that coexist point by point on a unified timestamp and a unified nested grid, which serves as the standard input data for the subsequent fusion process.

[0075] Step 203: Using a nonlinear cross-validation operator, the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data are fused to obtain the joint comprehensive meteorological disaster probability calculation equation.

[0076] Among them, the nonlinear cross-validation operator refers to the operator applied to aligned meteorological probability field data, which integrates multiple heterogeneous probabilities onto a unified grid. With timeliness The mapping operator performs point-by-point fusion and mutual verification. Its cross-validation is reflected in the fact that the probability of any path is not determined independently, but rather referenced and weighted against other paths under a unified spatiotemporal benchmark. In one embodiment, the nonlinear cross-validation operator includes a short-term extrapolation time-series weight function and an extreme convection time-series weight function. The short-term extrapolation time-series weight function is a nonlinear decay function that dynamically evolves with the prediction time, while the extreme convection time-series weight function is a weight control function that exhibits linear dissipation as it dynamically evolves with the prediction time. The joint comprehensive meteorological disaster probability calculation equation refers to the equation determined by the nonlinear cross-validation operator, used to calculate the unified virtual space grid. Timeliness The Standalone cascade mathematical formula for the final convergence probability of severe convective disasters (thunderstorms, short-duration heavy rainfall, etc.) is presented. The joint comprehensive meteorological disaster probability refers to the final comprehensive severe convective disaster occurrence confidence level after deep cascade fusion and cross-validation using the nonlinear time decay weighting operator of this application, and convergence on a unified digital virtual grid.

[0077] For example, in a unified virtual space grid Within, for any given prediction lead time The system reads the corresponding probability values ​​point by point from the multi-channel aligned meteorological probability field data. Then, a nonlinear cross-validation operator is used to determine the time-varying weights of the aligned meteorological probability field data corresponding to each multi-channel heterogeneous meteorological probability field data. Based on the time-varying weights, the aligned meteorological probability field data corresponding to each multi-channel heterogeneous meteorological probability field data are weighted and fused to obtain the joint comprehensive meteorological disaster probability calculation equation. Using this equation, all effective grids and all time-dependent points within the target power grid area can be solved point by point, outputting a joint comprehensive meteorological disaster probability field containing the meteorological disaster probability values ​​of each grid, which serves as the basis for subsequent processing.

[0078] Step 205: Based on the meteorological disaster probability value determined by the joint comprehensive meteorological disaster probability calculation equation, perform step risk level discretization mapping processing on the virtual space grid in the target power grid area to determine the step risk level corresponding to each virtual space grid.

[0079] Among them, the meteorological disaster probability value refers to the unified virtual space grid. With forecast timeliness The specific numerical value obtained by substituting it into the joint comprehensive meteorological disaster probability calculation equation represents the final convergence probability of severe convective disasters (thunderstorms, short-duration heavy precipitation, etc.) occurring in that grid within that time period, and serves as the input for the discretization mapping. The tiered risk level discretization mapping refers to the quantization process of converting continuous meteorological disaster probability values ​​grid-by-grid into finite integer tiers according to a preset monotonic threshold sequence. The tiered risk level is the output of the discretization mapping, i.e., the discrete integer level corresponding to each virtual spatial grid.

[0080] For example, by combining the comprehensive meteorological disaster probability calculation equation, all virtual spatial grids in the target power grid area , all prediction time horizons are processed to obtain meteorological disaster probability value results, so as to form a continuous probability field as the mapping input. Then, according to the meteorological disaster probability value results, all virtual spatial grids in the target power grid area are processed to eliminate the disturbance of high-frequency fluctuations of continuous floating-point probability values to the control system through a monotonic threshold sequence, and a quantization operator is used to discretize and map the continuous probability values into respective fixed integer stepped risk levels. In one embodiment, virtual spatial grids with relatively low meteorological disaster probability values can be directly pruned, it is determined that the grids belong to absolute safety or meteorological spatial noise within the current sliding window, and "feature space pruning and hiding" is directly performed. The pruned grids will no longer be allocated memory space, nor will they participate in the calculation of subsequent steps. Then, according to the continuous probability values, the values are divided into integer steps of levels 1 to 4 according to the monotonic threshold sequence T1<T2<T3 , that is:

[0081] = 0 (pruned), if < 0.15;

[0082] = 1, if 0.15 ≤ < T1, which is quantized and mapped to [Risk Level 1 (Mild Risk)];

[0083] = 2, if T1 ≤ < T2, which is quantized and mapped to [Risk Level 2 (Moderate Risk)];

[0084] = 3, if T2 ≤ < T3, which is quantized and mapped to [Risk Level 3 (Severe Risk)];

[0085] = 4, if ≥ T3, which is quantized and mapped to [Risk Level 4 (Extreme Risk)].

[0086] Step 207: Determine the target topology state and state transition information corresponding to the target power grid area based on the respective stepped risk levels of the virtual spatial grids.

[0087] The target topology state refers to the state that the state machine should or has already entered, determined by mapping and arbitration based on the current cycle's tiered risk level among the various topologies of the early warning state machine. It has two meanings: first, the mapping layer target, which is the "state to be entered" read from the static mapping table based on the current cycle's tiered aggregation result; second, the arbitration layer landing point, which is the state actually written into the state register in the current cycle after adjudication based on guard conditions and step-by-step constraints. When the mapping layer target is inconsistent with the current state and the guard conditions are not met, the target topology state may still be the current state (maintained) or an intermediate state on the path (the next station in the step-by-step descent). The target topology state is the direct driving source of the output control signal. State transition information is a structured edge description of whether and how the state machine shifts gears in the current cycle. It includes at least the following fields: transition occurrence flag, source state, destination state, transition direction (upgrade / downgrade / direct downgrade / hold), transition edge (upgrade edge / backgrade edge / close edge / direct downgrade channel), trigger event type (far-end trend activation / near-end dual activation strong transition / probability fallback / sliding window safety closure / secondary surge interruption / full pruning and repositioning), and accompanying pulse action commands. State transition information is the data source for pulse-type output signals and event logs. Specifically, the joint early warning spatiotemporal state machine controlling state transitions is a spatiotemporal topology state transition control architecture deployed at the digital decision-making front end of the large power grid, driven by a deterministic meteorological risk ladder. It is used to convert continuous meteorological field probabilistically into discrete, dimensionless disaster prevention control in-situ action signals.

[0088] For example, for all virtual spatial grids within the target power grid area Aggregation is performed to extract the driving events required by the state machine. The aggregated events are substituted into the static mapping table to read the corresponding target topology state, and compared with the current state to determine the transition direction and candidate edges. Upgrading edges are event-driven and executed immediately, while downgrading edges are condition-driven and confirmed with a delay. Finally, state transition information is generated and the output is driven.

[0089] Step 209: Generate meteorological disaster prevention and control signals for the target power grid area based on the target topology status and status transition information.

[0090] Among them, the meteorological disaster prevention control signal refers to the dimensionless control action signal generated by the early warning state machine in each scanning cycle, based on the target topology state and state transition information, and output to the application layer (unified meteorological emergency front-end system).

[0091] For example, server 104 receives the target topology state and state transition information determined in the previous step. It then generates a level signal based on the target topology state and a pulse signal based on the state transition information. Finally, it extracts the grid set representing the driving state from the grid aggregation result, converts it into a GPS latitude and longitude grid set, and attaches it to the signal, enabling downstream operation and maintenance units to deploy defenses precisely according to coordinates. Finally, the unified meteorological emergency front-end system for the target power grid area maps the dimensionless code to contingency plan actions, executing macro-level disaster prevention preparation, enhanced inspection, resource pre-positioning and standby, highest-level linkage control, buffer holding / prohibition of release, and complete release, ensuring power grid safety under extreme weather conditions.

[0092] The aforementioned power grid meteorological early warning method acquires multiple heterogeneous meteorological probability field data within the target power grid area and performs spatiotemporal dimension concatenation and alignment processing on these data to obtain aligned meteorological probability field data corresponding to each heterogeneous meteorological probability field data. Using a nonlinear cross-validation operator, the aligned meteorological probability field data corresponding to each heterogeneous meteorological probability field data are fused to obtain a joint comprehensive meteorological disaster probability calculation equation. Based on the meteorological disaster probability values ​​determined by the joint comprehensive meteorological disaster probability calculation equation, a tiered risk level discretization mapping process is performed on the virtual spatial grid within the target power grid area to determine the tiered risk level corresponding to each virtual spatial grid. Based on the tiered risk level corresponding to each virtual spatial grid, the target topology state and state transition information corresponding to the target power grid area are determined. Based on the target topology state and state transition information, a meteorological disaster prevention and control signal for the target power grid area is generated. After aligning multiple heterogeneous meteorological probability field data, this application uses a nonlinear cross-validation operator to fuse these data to obtain a joint comprehensive meteorological disaster probability calculation equation. This effectively eliminates the problem of conflict between forecast results from multiple models. Then, based on the meteorological disaster probability determined by the equation, the virtual spatial grid of the region is mapped to the corresponding tiered risk level. Through state division and state switching identification processing, meteorological disaster prevention and control signals are generated, which can effectively improve the effectiveness and accuracy of meteorological warnings under severe convective weather.

[0093] In an exemplary embodiment, spatiotemporal dimensional concatenation and alignment processing is performed on multiple heterogeneous meteorological probability field data to obtain aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data, including:

[0094] Based on the scale information corresponding to the multiple heterogeneous meteorological probability field data, a unified spatial reference scale is determined. The multiple heterogeneous meteorological probability field data are dynamically resampled using interpolation operators, and the multiple heterogeneous meteorological probability field data are projected and transformed onto a virtual spatial grid under the unified spatial reference scale to obtain the spatially aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data. The spatially aligned meteorological probability field data are then subjected to temporal frame alignment processing using dynamic time sliding windows to obtain the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data.

[0095] The unified spatial reference scale refers to the unique spatial resolution and grid partitioning reference determined based on the scale information of each heterogeneous meteorological probability field data, serving as the common target for projection transformation of all data streams. This includes grid size, coordinate reference, grid origin, and alignment method. The interpolation operator, also known as the numerical remapping operator, is a spatial interpolation rule that calculates the values ​​of a given probability field on its original grid to the values ​​on each virtual grid at the unified spatial reference scale, achieving scale transformation from the original scale to the unified spatial reference scale. Dynamic resampling processing refers to the process of dynamically configuring interpolation operator parameters (interpolation kernel, weighting scheme, footprint geometry) for each heterogeneous meteorological probability field data stream based on its scale information and grid geometry, and updating the projection transformation frame-by-frame as each data stream is processed to obtain spatially aligned meteorological probability field data. Dynamic time sliding windows refer to a time window mechanism that slides along a unified scheduling time axis, with the window length, step size, and anchor point dynamically configured according to the time of each data stream. It is used for temporal frame alignment of spatially aligned meteorological probability field data. Using the current anchor point time of the sliding window as a unified time reference, it matches, maintains, or interpolates data frames with different update cycles and delays within the window, eliminating timescale and cycle differences between streams and ensuring that all probability values ​​are grouped into the same timestamp sequence. Its dynamic nature is reflected in the fact that the window parameters adaptively adjust according to the actual update rhythm and delay of each stream, rather than being a rigid window with a fixed window length and step size.

[0096] For example, in the process of spatiotemporal dimensional cascade alignment, the scale information corresponding to multiple heterogeneous meteorological probability field data can be determined first, and then the finest common scale can be selected as a unified spatial reference scale. Spatial alignment is then performed under the reference of this unified spatial reference scale. Dynamic resampling of the multiple heterogeneous meteorological probability field data is performed using interpolation operators, projecting and transforming the data onto a virtual spatial grid under the unified spatial reference scale. Then, temporal frame alignment is performed on the spatially aligned meteorological probability field data using dynamic temporal sliding windows, obtaining the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data. This bridges the multi-scale spatial and temporal discontinuities between the macroscopic background field and the microscopic short-term characteristic flow in the multiple heterogeneous meteorological probability field data.

[0097] In this embodiment, spatial alignment of multiple heterogeneous meteorological probability field data is achieved by determining a unified spatial reference scale, and temporal frame alignment is achieved by using a dynamic time sliding window. This effectively completes the spatiotemporal dimension cascade alignment process, thereby providing effective aligned meteorological probability field data for subsequent processing.

[0098] In one specific embodiment, the multi-path heterogeneous meteorological probability field data includes refined radar spatiotemporal short-term extrapolation probability field data, extreme severe convective cell identification probability field data, and medium- and long-term large model ensemble forecast precipitation probability field data, which are output by the medium- and long-term precipitation forecast network model, the short-term precipitation extrapolation network model, and the extreme severe convective identification and early warning model, respectively. Wherein:

[0099] The medium- and long-term precipitation forecasting network model outputs the macro-precipitation and severe convection probability matrix for the target area over the next 3 days (72 hours). Its spatial resolution is the first scale. (In this embodiment, the degree is set to 0.25 degrees, which is equivalent to a spatial grid of 25km~28km). This scale naturally covers cross-regional, large-area 500kV and above main grid transmission channels and long-span tower groups.

[0100] The short-term precipitation extrapolation network model, based on Doppler radar vertically integrated liquid water content (VIL) data with a 6-minute update interval, outputs a deterministic spatiotemporal extrapolated precipitation probability field for the next 0-2 hours. Its spatial resolution is a refined second scale. (In this embodiment, a 1km × 1km pixel grid is used. This scale is used for high-precision defense of dense and variable 220kV and below distribution networks, local substations, and distributed new energy power stations.)

[0101] The extreme severe convection identification and early warning model outputs multi-channel high-resolution satellite cloud images updated every 15 minutes. Through a multi-scale spatial geometry preservation operator, it identifies instantaneous, localized, and sudden extreme severe convective cells and outputs the corresponding morphological convection probability field. (Spatial resolution is 4km grid).

[0102] Since the spatial physical resolution and temporal update frequency of various data sources are completely inconsistent, this embodiment uses bilinear interpolation or inverse distance weighted interpolation (IDW) as interpolation operators to refine the second scale. To unify spatial benchmarks, the first scale will be... The coarse-grid probability field and the 4km probability field identified by satellite are dynamically resampled, and all of them are projected, transformed, and aligned to a unified 1km×1km digital power grid virtual space grid. The above describes a process where a dynamic time window with continuous overlap and a step size of 6 minutes is established on the timeline. This window rigorously aligns the medium- to long-term background data for the next 72 hours, the short-term extrapolation data for the next 2 hours, and the real-time satellite identification data along the timestamp dimension. This mapping step eliminates coordinate misalignment noise between multi-source heterogeneous meteorological data and the physical spatial assets of the power grid.

[0103] In another embodiment, the bilinear interpolation operator can be replaced with the Kriging interpolation operator, the cubic spline interpolation operator, or the nearest neighbor interpolation operator. The set 6-minute dynamic time window update step size is a preferred setting based on the physical scanning cycle of the basic reflectivity data of the standard Doppler radar. In actual deployment, it can be adaptively replaced with an isochronous update window step size of 3 minutes, 5 minutes, or 10 minutes depending on the upgrade of the weather radar hardware or the timeliness of the high-resolution satellite.

[0104] In an exemplary embodiment, the multi-path heterogeneous meteorological probability field data includes refined radar spatiotemporal short-term extrapolation probability field data, extreme severe convective cell identification probability field data, and medium- and long-term large model ensemble forecast precipitation probability field data. The nonlinear cross-validation operator includes a short-term extrapolation time series weight function, an extreme convection time series weight function, and a medium- and long-term time series compensation weight function.

[0105] Step 203 includes:

[0106] We construct refined radar spatiotemporal short-term extrapolation probability equations and short-term extrapolation time-series weight functions corresponding to refined radar spatiotemporal short-term extrapolation probability field data; we construct extreme severe convective cell identification probability equations and extreme convection time-series weight functions corresponding to extreme severe convective cell identification probability field data; and we construct medium- and long-term large model ensemble forecast precipitation probability equations and medium- and long-term time-series compensation weight functions corresponding to medium- and long-term large model ensemble forecast precipitation probability field data. The short-term extrapolation time-series weight function is a nonlinear decay function that dynamically evolves with the forecast lead time; the extreme convection time-series weight function is a weight control function that exhibits linear dissipation as it dynamically evolves with the forecast lead time; and the medium- and long-term time-series compensation weight function is a weight control function that exhibits logarithmic nonlinear gradual increase as it dynamically evolves with the forecast lead time.

[0107] Based on the refined radar spatiotemporal short-term extrapolation probability equation and the short-term extrapolation time series weight function, the first meteorological disaster probability calculation equation is determined. Based on the extreme severe convective cell identification probability equation and the extreme convection time series weight function, the second meteorological disaster probability calculation equation is determined. Based on the medium- and long-term large model ensemble forecast precipitation probability equation and the medium- and long-term time series compensation weight function, the third meteorological disaster probability calculation equation is determined.

[0108] Based on the first, second, and third meteorological disaster probability calculation equations, the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data are fused to obtain the joint comprehensive meteorological disaster probability calculation equation.

[0109] The refined short-term radar extrapolation probability refers to a feature field characterized by the probability of localized heavy precipitation occurring within the near-field within two hours, derived from high-frequency Doppler radar image sequence extrapolation operators with a spatial resolution of 1km-level pixel alignment. It is primarily based on Doppler radar vertically integrated liquid water content (VIL) data updated every 6 minutes. The Doppler radar vertically integrated liquid water content refers to the total amount of liquid water per unit area (in units of...) obtained by integrating the reflectivity factor detected by Doppler weather radar along the vertical direction. ), is a core meteorological physical quantity used to characterize the intensity of radar echoes from localized heavy rain, hail, and other severe convective weather.

[0110] Medium-range ensemble precipitation probability refers to the joint strong convective precipitation probability feature field output by a large-scale multi-physical quantity spatiotemporal numerical forecasting deep learning network model, which has a relatively coarse spatial resolution and represents the medium- to long-term macro-meteorological evolution trend.

[0111] The Extreme Convective Monomer Recognition Probability is based on multi-channel high-resolution satellite cloud images with a 15-minute update interval. It uses a multi-scale spatial geometry preservation operator to identify instantaneous local sudden extreme convective monomers and outputs the corresponding morphological convection probability field.

[0112] For example, the multi-channel heterogeneous meteorological probability field data specifically includes three types of data: refined radar spatiotemporal short-term extrapolation probability field data, extreme severe convective cell identification probability field data, and medium- and long-term large model ensemble forecast precipitation probability field data. The joint comprehensive meteorological disaster probability calculation equation requires the fusion of these three types of probability field data. Therefore, in a unified virtual space grid... Within, for any given prediction lead time The probability equations and weighting functions corresponding to the three probability fields can be pre-constructed, and then the three probability equations can be fused through the weighting function to obtain the joint comprehensive meteorological disaster probability calculation equation.

[0113] This involves constructing refined radar spatiotemporal short-term extrapolation probability equations and short-term extrapolation time-series weighting functions corresponding to refined radar spatiotemporal short-term extrapolation probability field data. Furthermore, based on these equations and weighting functions, the probability calculation equation for the first meteorological disaster can be determined, specifically expressed as follows:

[0114]

[0115]

[0116] in, This is the equation for calculating the probability of the first meteorological disaster. To refine the radar spatiotemporal short-term extrapolation probability equations, a short-term extrapolation network model is used in a grid. and prediction timeliness The original probability of the deduced output. This is the Short-term Extrapolation Temporal Weight Function, which varies with the prediction time. A dynamically evolving nonlinear decay function. The initial confidence baseline coefficient can be set to 0.75 to characterize the high spatial accuracy confidence level of the self-developed short-term extrapolation network and satellite single-unit capture under near-term time-sensitivity conditions. It is the Temporal Dissipation Penalty Coefficient, a positive characteristic constant used to characterize the logarithmic decay of the refined pixel prediction capability of the short-range radar extrapolation model as it approaches the far end over time due to the strong nonlinear chaotic evolution of the atmosphere.

[0117] Construct the extreme severe convective cell identification probability equation and extreme convection time-series weighting function corresponding to the extreme severe convective cell identification probability field data, and determine the second meteorological disaster probability calculation equation based on the extreme severe convective cell identification probability equation and extreme convection time-series weighting function. This process can be expressed as:

[0118]

[0119]

[0120] in, This is the equation for calculating the probability of the first meteorological disaster. The probability equation for identifying extremely strong convective cells is given by the morphological operator of a multi-channel high-resolution satellite, which represents the probability of sudden convection occurrence. The extreme convection time series weighting function varies with the prediction lead time. A weighted control function exhibiting linear dissipation. This is also the initial confidence baseline coefficient, in this embodiment It can be set to 0.65 to characterize the high spatial accuracy trust level of the self-developed short-term extrapolation network and satellite single-unit capture under near-term timeliness. In other embodiments, it can be adjusted according to the actual situation. This is the Maximum Short-term Prediction Time Window. In this embodiment, it is set to 120 minutes, but in other embodiments, it can be adjusted according to the actual situation.

[0121] We construct the medium- and long-term large-scale model ensemble forecast precipitation probability equation and the medium- and long-term time-series compensation weight function corresponding to the medium- and long-term large-scale model ensemble forecast precipitation probability field data. Based on the medium- and long-term large-scale model ensemble forecast precipitation probability equation and the medium- and long-term time-series compensation weight function, we determine the probability calculation equation for the third meteorological disaster, which can be specifically expressed as follows:

[0122]

[0123]

[0124] in, This represents the probability of precipitation forecasts from a medium- to long-term large model ensemble. It is the result of resampling the large model base onto the grid after coarse-grid resampling. and timeliness The probability of the macroscopic background field at that location. This is the medium-to-long-term time-series compensation weight function, which represents the weighting function as time progresses. It continuously extends to distant points (e.g., approaching 1 or 2 hours), automatically amplifies, and takes absolute control.

[0125] Finally, based on the first, second, and third meteorological disaster probability calculation equations, the aligned meteorological probability field data corresponding to the various heterogeneous meteorological probability field data are fused to obtain the joint comprehensive meteorological disaster probability calculation equation. This process can be expressed as:

[0126]

[0127] This refers to a specific digital power grid virtual space grid. Forecasting timeframe for the present and future is The final convergence probability of severe convective disasters (thunderstorms, short-duration heavy rainfall, etc.) is calculated. The formula uses real-time summation and normalization of the denominator to ensure consistency across all prediction timeframes. Under these conditions, the joint weights of the three probabilities converge tightly within the closed interval [0,1], completely eliminating the overflow and collapse problem caused by traditional linear weighting.

[0128] Based on the mathematical constraints of the above equations, when the prediction time... When at the very proximal end (e.g.) ), due to the attenuation term Close to 1 and The weights are extremely large, resulting in the final joint probability. It is dominated by a spatiotemporal extrapolation network with high-fidelity pixel precision alignment and satellite extreme strong convective cell identification probability, making full use of the spatial positioning advantages of the self-developed short-term model for local sudden disasters.

[0129] With timeliness It continues to extend further away (for example, approaching 1 or 2 hours). It decays exponentially. It also dissipates linearly to the critical edge, at which point the temporal weights of both rapidly compress towards zero. It automatically amplifies and assumes absolute control. This means that the macroeconomic trend probability predicted by the large ensemble of medium- and long-term models will safely and smoothly take over the dominant position. Through this dynamic feedback verification mechanism, automatic probability convergence of multi-source models is achieved, effectively avoiding scheduling decision chaos caused by contradictions in multi-source model forecasts.

[0130] In this embodiment, a nonlinear cascaded fusion function that dissipates with the depth of prediction time is used. A nonlinear cross-validation operator introduces a time dissipation penalty term coefficient and a maximum time window, assigning extremely high temporal weights to a high-frequency, fine-grained spatiotemporal extrapolation network and an extremely strong convection identification network at the very near-term timeframe. As the prediction time extends further into the future, exponential decay smoothly and safely transfers dominance to the medium- to long-term large model base. This multi-path probability feedback verification loop achieves dynamic convergence of multiple probabilities, fundamentally reducing the false alarm and missed alarm rates.

[0131] In an exemplary embodiment, step 205 includes: determining the meteorological disaster probability value corresponding to each virtual space grid within the target power grid area based on the joint comprehensive meteorological disaster probability calculation equation; pruning the virtual space grids below the preset disaster threshold when the meteorological disaster probability value is lower than the preset disaster threshold to obtain the pruned target virtual space grid; and discretizing and mapping the target virtual space grid to each tiered risk level according to the meteorological disaster probability value corresponding to the target virtual space grid.

[0132] For example, to significantly reduce the distributed computing load of the provincial power grid dispatching core host when facing massive virtual spatial grids (typically hundreds of thousands of grids), this application introduces a spatial noise filtering and pruning operator and a discrete ladder mapping mechanism. Therefore, for the processing of ladder risk level mapping, a pruning-then-grading strategy can be implemented to reduce the amount of data in subsequent processing and ensure processing efficiency.

[0133] Preset a low significance confidence threshold (In this embodiment, it is set to 0.15). A full traversal scan of all virtual space grids is performed at the decision layer. If a certain virtual space grid... In the future The joint synthesis probability within satisfies If the grid is determined to be either absolutely safe or meteorological spatial noise within the current sliding window, "feature space pruning and hiding" will be performed directly. The pruned grid will no longer be allocated memory space, nor will it participate in the state machine transition calculations of subsequent processes.

[0134] Regarding the process of discretizing and quantizing the probability of high-risk grids, for the set of high-risk grids that have not been pruned and hidden, in order to eliminate the disturbance of the control system caused by the high-frequency fluctuations of continuous floating-point probability values, the continuous probability values ​​can be discretized and mapped to four fixed integer step-risk levels using quantization operators, namely:

[0135] If 0.15 ≤ < 0.45, quantitatively mapped to [Risk Level 1 (Mild Risk)];

[0136] If 0.45 ≤ < 0.65, quantitatively mapped to [Risk Level 2 (Medium Risk)];

[0137] If 0.65 ≤ < 0.75, quantitatively mapped to [Risk Level 3 (Severe Risk)];

[0138] ≥0.75, quantitatively mapped to [Risk Level 4 (Extremely High Risk)].

[0139] In this embodiment, a virtual space grid below a preset disaster threshold is pruned, and then the pruned target virtual space grid is discretized and mapped to each risk level. This can effectively reduce the computational load of subsequent processing and improve processing efficiency.

[0140] In an exemplary embodiment, step 207 includes: determining the channel virtual space grid corresponding to the transmission channel within the target power grid area, and determining the tiered risk level of the channel virtual space grid; mapping the tiered risk level of the channel virtual space grid to the target topology state corresponding to the target power grid area; obtaining the current topology state corresponding to the target power grid area; comparing the current topology state and the target topology state to obtain the state transition information corresponding to the target power grid area.

[0141] For example, the process of generating state transition information is mainly driven by the data logic layer, achieving soft decoupling from the high-dynamic transient parameters of the large power grid. Therefore, the process of determining the target topology state and state transition information requires first determining the virtual space grid of the transmission channels within the target power grid area and determining the tiered risk level of the virtual space grid. That is, identification is performed based on the virtual space grid of the transmission channels within the target power grid area to determine whether the transmission channels in the area face the threat of extreme weather. Then, the tiered risk level of the virtual space grid is mapped to the target topology state corresponding to the target power grid area, that is, determining the target topology state that these virtual space grids may reach. Then, the current topology state corresponding to the target power grid area is obtained. Based on the current topology state and the target topology state, it is determined whether a state transition has occurred, and if so, what kind of state transition has occurred, thus obtaining the state transition information corresponding to the target power grid area. In this embodiment, by determining the virtual space grid of the transmission channels within the target power grid area and then performing topology state identification and state transition identification processing for power grid weather warning on these virtual space grids, the accuracy of the processing can be effectively guaranteed.

[0142] In an exemplary embodiment, the target topology states include normal concern state, situational awareness state, defense zone isolation state, emergency pre-positioning state, and recovery and alarm cancellation state. The meteorological disaster prevention control signals include standard situational awareness inspection and maintenance signals, digital meteorological defense pre-positioning prompt signals, digital isolation commands, the highest level disaster prevention linkage defense control signals, and buffered alarm cancellation signals. Based on the target topology states and state transition information, the generation of meteorological disaster prevention control signals for the target power grid area includes: generating a standard situational awareness inspection and maintenance signal when the target topology state is in the normal concern state and the state transition information indicates maintenance of the normal concern state; generating a digital meteorological defense pre-positioning prompt signal when the target topology state is in the situational awareness state and the state transition information indicates a unidirectional transition from the normal concern state to the situational awareness state; generating the highest level disaster prevention linkage defense control signal when the target topology state is in the emergency pre-positioning state and the state transition information indicates a strong transition to the emergency pre-positioning state; and generating a buffered alarm cancellation signal when the target topology state is in the recovery and alarm cancellation state and the state transition information indicates a unidirectional transition to the recovery and alarm cancellation state.

[0143] For example, based on the processing procedure and state transition logic determined above, the four standard topology states and core transition rules of the early warning state machine of this application are set as follows:

[0144] For the target topology state being in a normal state of interest, when the virtual space grids of each transmission channel along the main grid line are located... When all elements are in a pruned state (i.e., risk level is zero), the state machine remains in the normal state of interest. At this time, if the target topology is in the normal state of interest and the state transition information represents the normal state of interest, a standard situational awareness inspection maintenance signal is generated, and the system maintains standard situational awareness inspection and regular low-frequency data polling.

[0145] When the target topology is in a situational awareness state, for example, if a medium-to-long-term precipitation forecast model reports a strong convective evolution trend in the coarse grid of a certain important 500kV transmission section across a region within a future long-term lead time, while the near-end 1km-level short-term radar extrapolation model has not yet predicted the entity echo due to time constraints, but after resampling and overlapping calculations, the joint probability of the grid in this area enters the risk level range of risk level 1 or risk level 2. The state machine unidirectionally jumps from the normal state of concern to the situational awareness state. At this time, a digital meteorological defense advance warning signal is generated, and the "digital meteorological defense advance warning signal" containing the geospatial coordinates (GPS latitude and longitude grid set) of the section is output to the unified meteorological emergency front-end system of the application layer, guiding the downstream operation and maintenance units to enter the macro-disaster prevention preparation stage.

[0146] When the target topology is in a zone isolation state, for example, when the fusion result of the medium- and long-term precipitation forecast large model and the near-end short-term radar extrapolation model continues to rise, and the grid of a certain cross-regional 500kV important transmission section is resampled and overlapped, and its joint probability exceeds the upper threshold of risk level 2 and enters the interval of risk level 3 in the step risk level, but has not yet exceeded the high threshold of risk level 4, the state machine unidirectionally jumps from the situational awareness state to the zone isolation state. At this time, a digital isolation command is generated, and a digital isolation command containing the geospatial coordinates (GPS latitude and longitude grid set) of the transmission section and related stations is output to the unified meteorological emergency front-end system of the application layer. The planned maintenance work of the transmission section and related stations is suspended in advance, and the remote operation of the protection device is locked, guiding the downstream operation and maintenance unit to enter the zone isolation standby stage.

[0147] For cases where the target topology is in an emergency pre-emptive state, such as when severe weather enters a near-term sliding window as it dynamically evolves over time, the short-term precipitation extrapolation network model and the extreme severe convection identification model are simultaneously activated within the same spatial grid. Doppler radar detects strong VIL echo cells and triggers the multi-scale spatial geometry preservation operator, leading to a joint comprehensive disaster probability in the grid containing the transmission section. A strong transition occurs, directly exceeding the high-risk confidence threshold, i.e., entering risk level 4. At this point, the state machine instantly transitions to the highest-level emergency pre-state. It is then necessary to urgently generate the highest-level disaster prevention and control linkage signal to bypass the time-consuming power flow Jacobian matrix calculation of the large power grid and directly output the dimensionless highest-level disaster prevention and control linkage signal (such as the Level I emergency warning action command). This drives the automated emergency front-end to execute proactive defense management actions such as the pre-positioning of the transmission line repair team, the active ratio regulation and isolation of distributed new energy, and the control interlocking of the dispatch backup automatic transfer system.

[0148] When the target topology is in a restored state, and the strong convective echo system and thunderstorm cells have completely moved out of the spatial grid coordinates where the asset is located, the heterogeneity probabilities of all paths fall back to a significant low threshold. The grid is then reclassified as being in a pruned state. At this point, the state machine unidirectionally transitions to the recovery and de-alarm state, generating a buffered de-alarm signal. Specifically, a 15-minute delayed confirmation and reset window is initiated within the recovery and de-alarm state. During this window's closure, if no secondary surge of localized convection occurs in the meteorological field, the state machine automatically and safely returns to the initial normal concern state after the 15-minute confirmation period. If the probability surges again within the window, the process is interrupted and switches back to the corresponding high-risk state. This mechanism effectively prevents the control dead zone problem caused by frequent state machine oscillations and switching due to jagged fluctuations at the edge of severe convective weather.

[0149] In this embodiment, by setting a unified digital power grid meteorological spatial virtual grid at the application layer, the main power grid structure is reduced in dimension and abstracted into a static asset point / line topology within the grid. Based purely on the confidence interval after joint probability cross-validation at the decision layer, the early warning state machine is driven to smoothly transition between five standard topological states: "normal attention state, situational awareness state, defense zone isolation state, emergency pre-emptive state, and recovery and alarm cancellation state." Externally, it directly outputs dimensionless, highest-level disaster prevention and control linkage signals, driving downstream systems to execute fine-grained scheduling with a single click, achieving millisecond-level rapid automated emergency deployment and response of the large power grid in the face of extreme weather.

[0150] In a specific embodiment, the topology diagram of the power grid multi-scale meteorological joint early warning system for severe convective weather based on the power grid meteorological early warning method of this application can be referred to... Figure 3 As shown.

[0151] This topology comprises four layers: a data layer (301), a model layer (302), a decision layer (303), and an application layer (304). The data layer (301) controls the high-frequency input of multi-source heterogeneous data. The model layer (302) implements a multi-scale weather forecasting model matrix; the three models, arranged according to their respective scales and time-series requirements, form the model matrix, serving as the physical carrier for the multi-scale process. The decision layer (303) acts as the joint decision-making center for the model matrix, and the application layer (304) is the mapping for the power system's emergency front-end. The main link direction is data layer (301) → model layer (302) → decision layer (303) → application layer (304), from top to bottom. Simultaneously, there exists a bottom-up multi-path probabilistic dynamic feedback verification loop: the execution and verification results of the application layer (304) and decision layer (303) are fed back to the model layer for dynamic adjustment of model weights, forming a closed loop.

[0152] and Figure 3 The lower half corresponds to the three improvements of this application scheme compared to the existing scheme, including 305 multi-scale spatiotemporal grid alignment and cascading, 306 nonlinear time decay weighted cross-validation operator, and 307 dimensionless joint early warning spatiotemporal state machine control.

[0153] Multi-scale spatiotemporal grid alignment and cascading refers to the process of dynamically resampling (overlap area weighting) three probability fields using interpolation operators at a unified spatial reference scale, followed by temporal frame alignment via dynamic time sliding windows to obtain point-to-point coexisting aligned meteorological probability field data. It addresses the problem of how heterogeneous data can be compared and is a prerequisite for data fusion.

[0154] The nonlinear time decay weighted cross-validation operator refers to the process of fusing the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data using a nonlinear cross-validation operator. It solves the problem of how to converge multi-source conflicting forecasts into a single reliable probability, i.e. the time-series loss probability stitching process in the figure, which can smoothly connect the advantageous models of different time windows one by one.

[0155] The dimensionless joint early warning spatiotemporal state machine control is the process of driving the state machine to smoothly transition and output dimensionless control signals. The state machine only consumes integer steps and only outputs dimensionless signals. No physical dimension parameters are transmitted between the meteorological sensing layer and the power control layer. The state transition is completed entirely within the topology graph.

[0156] for Figure 3In the data flow process, the state machine output is sent back to the application layer (emergency front-end mapping). After the application layer executes the plan, the execution status is sent back to the model layer / decision layer through the feedback verification loop, forming an outer loop of "decision → execution → feedback". The fusion result and state jump information are sent back to the model layer for dynamic verification of weights and models, forming an inner loop of "fusion → verification → correction". The alignment parameters (grid partitioning, sliding window configuration) are driven by the asset domain settings of the decision layer / application layer and can dynamically adjust the monitoring domain and scale chain, forming a configuration loop of "application requirements → perception configuration".

[0157] In summary, this application proposes a full-process control architecture consisting of a "data layer—model layer—decision layer—application layer." The data layer aggregates global / regional forecast initial fields, radar VIL echo data, and multi-channel high-resolution satellite cloud images. The model layer constructs a model matrix that includes extrapolation of medium- and long-term precipitation, short-term heavy precipitation, and identification of extreme severe convection. The decision layer performs multi-scale fusion alignment and cross-validation. Finally, the application layer realizes grid meteorological grid risk mapping and drives a unified meteorological emergency front-end system, forming a standardized data realization control chain.

[0158] In a specific embodiment, the complete flowchart of the power grid meteorological early warning method of this application can be referred to Figure 4 As shown, it includes:

[0159] Step 401: Input the resampled multi-source probabilities and the current prediction time t. Read the three-way aligned meteorological probability fields P_short, P_extreme, and P_medium after spatiotemporal concatenation alignment (invention point 1) and the current prediction time t as the fusion input.

[0160] Step 402, calculate the time-series dynamic weights (Exponential decay) (Linear dissipation) and (Medium- to long-term time series compensation).

[0161] Step 403: Calculate the combined probability of severe convective weather disasters. The continuous fusion probability field is obtained by solving the problem grid by grid and time by time.

[0162] Step 404, determine P_joint <P_min(0.15)。

[0163] If so, proceed to step 405, perform feature space pruning and hiding, release the computational load, do not trigger the state machine, and terminate the current round of the mesh process.

[0164] If not, proceed to step 406, perform discretization ladder quantization mapping to convert continuous probability values ​​into 1–4 level integer risk ladders, and output grid-level ladders.

[0165] Step 407: Activate the unified early warning spatiotemporal state machine driving engine to aggregate the grid ladder into regional driving events and read the current state.

[0166] Step 408: Based on the risk ladder and the time-limited sliding window trigger state jump branch judgment, determine the landing point and transfer edge of this cycle, and divert to steps 409-413.

[0167] Step 409, [Situational Awareness Status] Trigger Conditions: Long-distance, time-sensitive major fault line excavation trend, and the step ladder is in the risk level 1 or 2 range. Output: Digital meteorological defense advance warning signal.

[0168] Step 410, [Zone Isolation State] Trigger Condition: The ladder is in the risk level 3 range. Output: Digital isolation command.

[0169] Step 411, [Emergency Pre-State] Trigger Condition: Near-end timeliness window / satellite dual activation suddenly exceeds the threshold, stepping into level 4. Output: Strong transition top-level output of the highest-level associated motion control signal.

[0170] Step 412, [Restore Alarm Deactivation State] Trigger Conditions: The echo system completely moves out of the network coordinates, and the probability drops below the low threshold. Action: Enable 15-minute delay confirmation reset sliding window; output buffered alarm deactivation (lock-up) level, and prohibit complete disarming.

[0171] Step 413: Determine if a second sudden surge in probability occurs within the 15-minute sliding window. If yes, jump back to step 411 along the upgrade edge (upgrade and hot recovery); if no, the sliding window closes safely and proceeds to step 414.

[0172] Step 414, [Normal Attention State] Triggering conditions: Sliding window is safely closed, or the entire mesh is in a pruned state (direct drop channel). Output: Conventional low-frequency situational awareness inspection signal; the system returns to the dormant ground state, waiting for the input of the next round of step 401, realizing the full-process cycle loop.

[0173] Based on the above process, this application completely overcomes the dissipation faults and ambiguities in multi-scale weather forecasts along the time axis. Through an innovatively designed nonlinear time decay weighting formula, this application deeply integrates the background field of a large model with medium- to long-term lead time with a high-frequency, refined short-term radar / satellite extrapolation network. Experimental and operational data show that this application successfully maintains the Mean Success Index (Mean CSI) of 1- to 2-hour long-term strong convective weather comprehensive forecasts along large power grids at an industry-leading level above 0.42, breaking the technical bottleneck of the logarithmic collapse in forecast capability of traditional extrapolation models after 30 minutes.

[0174] This application significantly reduces the "double penalty" and false alarm / missed alarm rates in emergency command of provincial power grids. By introducing a fusion verification operator based on depth dissipation with forecast lead time, it achieves dynamic convergence of multi-source models and automatic uncertainty comparison. This application effectively filters out smoothing artifacts caused by spatial displacement noise of weather systems, reducing the "missed alarm rate" of tripping caused by sudden severe convective weather in provincial power grids by more than 28%. At the same time, due to the implementation of accurate cross-verification, it significantly reduces the number of logistical redundancies caused by false weather alerts, such as redundant disaster prevention mobilization and blind pre-positioning of repair vehicles, by more than 35%, significantly improving the digital economic benefits of disaster prevention and mitigation in the new power system.

[0175] This application circumvents the problem of convergence due to physical constraints in a network outage environment, significantly improving computational response time. It abstracts complex transmission sections, towers, and other assets into static asset segment topology within a spatial virtual grid, eliminating over 85% of risk-free grid elements through spatial noise filtering and pruning operators. The state machine transitions are driven by a spatiotemporal geometric topology mapping operator, achieving soft decoupling from the high-dynamic transient parameters of the large power grid, without consuming the conventional power flow computation power of the large power grid dispatch core host. Its state machine transitions and defense level signal outputs achieve millisecond-level computational response speeds, making it highly suitable for rapid deployment at the front end of grassroots distribution network automation master stations and the main grid's active defense.

[0176] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0177] Based on the same inventive concept, this application also provides a power grid weather warning device for implementing the power grid weather warning method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more power grid weather warning device embodiments provided below can be found in the limitations of the power grid weather warning method described above, and will not be repeated here.

[0178] In one exemplary embodiment, such as Figure 5 As shown, a power grid weather early warning device is provided, comprising:

[0179] The data acquisition module 502 is used to acquire multiple heterogeneous meteorological probability field data within the target power grid area, and to perform spatiotemporal dimension concatenation and alignment processing on the multiple heterogeneous meteorological probability field data to obtain the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data.

[0180] The probability field fusion module 504 is used to fuse the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data through a nonlinear cross-validation operator to obtain a joint comprehensive meteorological disaster probability calculation equation.

[0181] The risk ladder mapping module 506 is used to perform ladder risk level discretization mapping on the virtual space grid within the target power grid area based on the meteorological disaster probability value determined by the joint comprehensive meteorological disaster probability calculation equation, and to determine the ladder risk level corresponding to each virtual space grid.

[0182] The state machine processing module 508 is used to determine the target topology state and state transition information of the target power grid area based on the corresponding tiered risk level of each virtual space grid.

[0183] The control signal generation module 510 is used to generate meteorological disaster prevention control signals for the target power grid area based on the target topology status and status transition information.

[0184] In one embodiment, the data acquisition module 502 is specifically used to: determine a unified spatial reference scale based on the scale information corresponding to the multi-channel heterogeneous meteorological probability field data; perform dynamic resampling processing on the multi-channel heterogeneous meteorological probability field data through interpolation operators, project and transform the multi-channel heterogeneous meteorological probability field data onto a virtual spatial grid under the unified spatial reference scale, and obtain spatially aligned meteorological probability field data corresponding to each of the multi-channel heterogeneous meteorological probability field data; and perform temporal frame alignment processing on the spatially aligned meteorological probability field data through dynamic time sliding windows to obtain aligned meteorological probability field data corresponding to each of the multi-channel heterogeneous meteorological probability field data.

[0185] In one embodiment, the multi-path heterogeneous meteorological probability field data includes refined radar spatiotemporal short-term extrapolation probability field data, extreme severe convective cell identification probability field data, and medium- and long-term large model ensemble forecast precipitation probability field data. The nonlinear cross-validation operator includes short-term extrapolation time series weighting function, extreme convection time series weighting function, and medium- and long-term time series compensation weighting function.

[0186] The probability field fusion module 504 is specifically used for: constructing refined radar spatiotemporal short-term extrapolation probability equations and short-term extrapolation time-series weight functions corresponding to refined radar spatiotemporal short-term extrapolation probability field data; constructing extreme severe convective cell identification probability equations and extreme convection time-series weight functions corresponding to extreme severe convective cell identification probability field data; and constructing medium- and long-term large-scale model ensemble forecast precipitation probability equations and medium- and long-term time-series compensation weight functions corresponding to medium- and long-term large-scale model ensemble forecast precipitation probability field data. The short-term extrapolation time-series weight function is a nonlinear decay function that dynamically evolves with the forecast lead time. The extreme convection time-series weight function is a weight control function that exhibits linear dissipation as it dynamically evolves with the forecast lead time. The medium- and long-term time-series compensation weight function is a weight control function that dynamically evolves with the forecast lead time. A logarithmically nonlinear, gradually increasing weight control function is used. Based on the refined radar spatiotemporal short-term extrapolation probability equation and the short-term extrapolation time-series weight function, the first meteorological disaster probability calculation equation is determined. Based on the extreme severe convective cell identification probability equation and the extreme convection time-series weight function, the second meteorological disaster probability calculation equation is determined. Based on the medium- and long-term large model ensemble forecast precipitation probability equation and the medium- and long-term time-series compensation weight function, the third meteorological disaster probability calculation equation is determined. Based on the first, second, and third meteorological disaster probability calculation equations, the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data are fused to obtain the joint comprehensive meteorological disaster probability calculation equation.

[0187] In one embodiment, the risk ladder mapping module 506 is specifically used to: determine the meteorological disaster probability value corresponding to each virtual space grid in the target power grid area based on the joint comprehensive meteorological disaster probability calculation equation; when the meteorological disaster probability value is lower than the preset disaster threshold, prune the virtual space grids that are lower than the preset disaster threshold to obtain the pruned target virtual space grid; and discretize and map the target virtual space grid to each ladder risk level according to the meteorological disaster probability value corresponding to the target virtual space grid.

[0188] In one embodiment, the state machine processing module 508 is specifically used to: determine the channel virtual space grid corresponding to the transmission channel within the target power grid area, and determine the step risk level of the channel virtual space grid; map the step risk level of the channel virtual space grid to the target topology state corresponding to the target power grid area; obtain the current topology state corresponding to the target power grid area; compare the current topology state and the target topology state to obtain the state transition information corresponding to the target power grid area.

[0189] In one embodiment, the target topology states include normal attention state, situational awareness state, defense zone isolation state, emergency pre-positioning state, and recovery and alarm cancellation state, and the meteorological disaster prevention control signals include standard situational awareness inspection and maintenance signals, digital meteorological defense pre-positioning prompt signals, digital isolation commands, the highest level disaster prevention linkage defense control signals, and buffer alarm cancellation signals;

[0190] The control signal generation module 510 is specifically used for: generating a standard situational awareness inspection maintenance signal when the target topology is in a normal state of concern and the state transition information indicates that the state remains in a normal state of concern; generating a digital meteorological defense advance warning signal when the target topology is in a situational awareness state and the state transition information indicates that the state has unidirectionally transitioned from the normal state of concern to the situational awareness state; generating a digital isolation command when the target topology is in a defense zone isolation state and the state transition information indicates that the state has transitioned to the defense zone isolation state; generating the highest level disaster prevention and control linkage control signal when the target topology is in an emergency advance state and the state transition information indicates that the state has strongly transitioned to the emergency advance state; and generating a buffered alarm cancellation signal when the target topology is in a recovery and alarm cancellation state and the state transition information indicates that the state has unidirectionally transitioned to the recovery and alarm cancellation state.

[0191] Each module in the aforementioned power grid weather warning device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0192] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores data related to power grid weather warnings. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a power grid weather warning method.

[0193] Those skilled in the art will understand that Figure 6The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0194] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0195] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0196] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and executes the computer instructions, causing the computer device to perform the steps in the above method embodiments.

[0197] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0198] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0199] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0200] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A power grid meteorological early warning method, characterized in that, The method includes: Acquire multiple heterogeneous meteorological probability field data within the target power grid area, and perform spatiotemporal dimension concatenation and alignment processing on the multiple heterogeneous meteorological probability field data to obtain the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data. By using a nonlinear cross-validation operator, the aligned meteorological probability field data corresponding to each of the multi-path heterogeneous meteorological probability field data are fused to obtain a joint comprehensive meteorological disaster probability calculation equation. Based on the meteorological disaster probability value determined by the joint comprehensive meteorological disaster probability calculation equation, the virtual spatial grid in the target power grid area is subjected to a step risk level discretization mapping process to determine the step risk level corresponding to each of the virtual spatial grids; Based on the tiered risk levels corresponding to each of the virtual space grids, the target topology state and state transition information corresponding to the target power grid area are determined. Based on the target topology state and state transition information, a meteorological disaster prevention and control signal for the target power grid area is generated.

2. The method according to claim 1, characterized in that, The spatiotemporal dimensional concatenation and alignment processing of the multi-channel heterogeneous meteorological probability field data to obtain the aligned meteorological probability field data corresponding to each of the multi-channel heterogeneous meteorological probability field data includes: Based on the scale information corresponding to the multi-path heterogeneous meteorological probability field data, a unified spatial reference scale is determined; The multi-channel heterogeneous meteorological probability field data is dynamically resampled by interpolation operators, and the multi-channel heterogeneous meteorological probability field data is projected and transformed onto a virtual spatial grid under the unified spatial reference scale to obtain the spatially aligned meteorological probability field data corresponding to each of the multi-channel heterogeneous meteorological probability field data. By using a dynamic time sliding window, the spatially aligned meteorological probability field data is processed to perform time-series frame alignment, thereby obtaining the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data.

3. The method according to claim 1, characterized in that, The multi-path heterogeneous meteorological probability field data includes refined radar spatiotemporal short-term extrapolation probability field data, extreme severe convective cell identification probability field data, and medium- and long-term large model ensemble forecast precipitation probability field data. The nonlinear cross-validation operator includes a short-term extrapolation time series weight function, an extreme convection time series weight function, and a medium- and long-term time series compensation weight function. The process of fusing the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data using a nonlinear cross-validation operator to obtain the joint comprehensive meteorological disaster probability calculation equation includes: The following methods are employed: constructing the refined radar spatiotemporal short-term extrapolation probability equation and short-term extrapolation time series weight function corresponding to the refined radar spatiotemporal short-term extrapolation probability field data; constructing the extreme severe convective cell identification probability equation and extreme convection time series weight function corresponding to the extreme severe convective cell identification probability field data; constructing the medium- and long-term large model ensemble forecast precipitation probability equation and medium- and long-term time series compensation weight function corresponding to the medium- and long-term large model ensemble forecast precipitation probability field data; wherein the short-term extrapolation time series weight function is a nonlinear decay function that dynamically evolves with the forecast lead time; the extreme convection time series weight function is a weight control function that exhibits linear dissipation as it dynamically evolves with the forecast lead time; and the medium- and long-term time series compensation weight function is a weight control function that exhibits logarithmic nonlinear gradual increase as it dynamically evolves with the forecast lead time. Based on the refined radar spatiotemporal short-term extrapolation probability equation and the short-term extrapolation time series weight function, the first meteorological disaster probability calculation equation is determined; based on the extreme severe convective cell identification probability equation and the extreme convective time series weight function, the second meteorological disaster probability calculation equation is determined; based on the medium- and long-term large model ensemble forecast precipitation probability equation and the medium- and long-term time series compensation weight function, the third meteorological disaster probability calculation equation is determined. Based on the first meteorological disaster probability calculation equation, the second meteorological disaster probability calculation equation, and the third meteorological disaster probability calculation equation, the aligned meteorological probability field data corresponding to each of the multi-path heterogeneous meteorological probability field data are fused to obtain the joint comprehensive meteorological disaster probability calculation equation.

4. The method according to claim 1, characterized in that, The meteorological disaster probability value determined based on the joint comprehensive meteorological disaster probability calculation equation is used to perform a stepped risk level discretization mapping process on the virtual spatial grid within the target power grid area to determine the stepped risk level corresponding to each virtual spatial grid, including: Based on the joint comprehensive meteorological disaster probability calculation equation, the meteorological disaster probability value corresponding to each virtual spatial grid within the target power grid area is determined; When the meteorological disaster probability value is lower than the preset disaster threshold, the virtual space grid that is lower than the preset disaster threshold is pruned to obtain the pruned target virtual space grid; Based on the meteorological disaster probability value corresponding to the target virtual space grid, the target virtual space grid is discretized and mapped to various risk levels.

5. The method according to any one of claims 1 to 4, characterized in that, The step of generating the meteorological disaster prevention and control signal for the target power grid area based on the target topology state and state transition information includes: Determine the virtual space grid of the transmission channels within the target power grid area, and determine the tiered risk level of the virtual space grid of the transmission channels; The tiered risk levels of the channel virtual space grid are mapped to the target topology state corresponding to the target power grid area; Obtain the current topology state corresponding to the target power grid area; By comparing the current topology state with the target topology state, the state transition information corresponding to the target power grid area is obtained.

6. The method according to any one of claims 1 to 4, characterized in that, The target topology states include normal attention state, situational awareness state, defense zone isolation state, emergency pre-positioning state, and recovery and alarm cancellation state. The meteorological disaster prevention control signals include standard situational awareness inspection and maintenance signals, digital meteorological defense pre-positioning prompt signals, digital isolation commands, the highest level disaster prevention linkage defense control signals, and buffer alarm cancellation signals. The step of generating the meteorological disaster prevention and control signal for the target power grid area based on the target topology state and state transition information includes: When the target topology is in a state of normal interest and the state transition information indicates that the state of normal interest is maintained, a standard situational awareness inspection maintenance signal is generated. When the target topology is in a situational awareness state, and the state transition information represents a one-way transition from a normal state of concern to a situational awareness state, a digital meteorological defense advance warning signal is generated. When the target topology is in zone isolation state and the state transition information indicates a state transition to zone isolation state, a digital isolation command is generated. When the target topology is in an emergency pre-state and the state transition information indicates a strong state transition to the emergency pre-state, the highest level disaster prevention and control linkage signal is generated. When the target topology is in the recovery and alarm cancellation state, and the state transition information indicates a one-way state transition to the recovery and alarm cancellation state, a buffered alarm cancellation signal is generated.

7. A power grid weather early warning device, characterized in that, The device includes: The data acquisition module is used to acquire multiple heterogeneous meteorological probability field data within the target power grid area, and to perform spatiotemporal dimension concatenation and alignment processing on the multiple heterogeneous meteorological probability field data to obtain the aligned meteorological probability field data corresponding to each of the multiple heterogeneous meteorological probability field data. The probability field fusion module is used to fuse the aligned meteorological probability field data corresponding to each of the multi-path heterogeneous meteorological probability field data through a nonlinear cross-validation operator to obtain a joint comprehensive meteorological disaster probability calculation equation. The risk ladder mapping module is used to perform ladder risk level discretization mapping processing on the virtual spatial grid in the target power grid area based on the meteorological disaster probability value determined by the joint comprehensive meteorological disaster probability calculation equation, and to determine the ladder risk level corresponding to each virtual spatial grid. The state machine processing module is used to determine the target topology state and state transition information corresponding to the target power grid area based on the tiered risk level corresponding to each of the virtual space grids. The control signal generation module is used to generate meteorological disaster prevention control signals for the target power grid area based on the target topology state and state transition information.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images