A power supply risk communication early warning method, system and medium

Abstract

This invention relates to a method, system, and medium for power supply risk communication early warning, belonging to the field of power supply early warning technology. The method includes acquiring multi-source heterogeneous power supply risk communication data to form an initial power supply risk communication dataset; inputting a risk distribution matrix, an emergency resource accessibility matrix, and a communication feature vector into a cross-domain correlation mapping engine to perform coupling strength calculation, generating a basic risk probability distribution matrix and a communication quality confidence vector; performing bidirectional feedback iterative correction on the quantitative values ​​of power supply risk status and communication network health, outputting a power supply risk status index and a communication network health index; performing hierarchical boundary optimization and stability verification on the weighted and corrected threshold sequence; and receiving a set of risk early warning hierarchical thresholds to generate a differentiated power supply risk communication early warning instruction set. This invention overcomes the limitations of traditional technologies that separate risk assessment and communication early warning, achieving closed-loop linkage between risk status and communication capabilities, and precise adaptation of early warning strategies.

Description

Technical Field

[0001] This invention relates to the field of power supply early warning technology, and more specifically, to a power supply risk communication early warning method, system, and medium. Background Technology

[0002] Currently, power supply risk early warning technology for distribution networks mainly relies on the independent operation of power grid monitoring systems and manual experience-based judgment. Data silos between various business systems are severe, with key information such as power grid operation monitoring, weather forecasting, emergency resource dispatching, and communication network status scattered across different platforms, lacking a unified spatiotemporal benchmark alignment and structured fusion mechanism. Risk assessments often employ static threshold judgments or single-indicator weighting models, failing to depict the cascading effects of power grid faults and communication interruptions. Fixed early warning thresholds are ill-suited to adapting to the dynamic risk evolution triggered by extreme events. Early warning communication methods are limited to single channels such as mass SMS or telephone notifications, lacking real-time assessment of channel bandwidth, latency, and reliability, as well as differentiated coding strategies. This results in high packet loss rates and poor timeliness of early warning information, preventing emergency command personnel from obtaining accurate information matching the risk level in a timely manner, severely impacting the efficiency of power supply emergency response.

[0003] While existing technologies attempt to integrate multi-source information through data interfaces, they remain at the level of simple data aggregation and fail to achieve two-way extrapolation and feedback correction of power supply risk propagation and communication network health at the model level. This fails to fundamentally solve the core problem of the disconnect between risk assessment and early warning communication. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system, device, and readable storage medium for early warning communication regarding power supply risks, in order to improve the aforementioned problems. To achieve the above objective, the technical solution adopted by this invention is as follows: Firstly, this application provides a method for communication early warning of power supply risks, including: Acquire multi-source heterogeneous power supply risk communication data, including power grid operation monitoring data, meteorological environment data, emergency resource distribution data, and communication network quality data. Perform outlier removal, missing value imputation, and unification of multiple time granularities and spatial coordinate system on the multi-source heterogeneous power supply risk communication data to form an initial power supply risk communication dataset. The initial power supply risk communication dataset is called, and a power grid feature matrix is ​​formed by hierarchical extraction of power grid state features. A meteorological feature vector is formed by encoding meteorological influencing factors. A preliminary risk assessment is performed based on the power grid feature matrix and the meteorological feature vector to obtain a risk distribution matrix. An emergency resource accessibility matrix is ​​calculated based on the risk distribution matrix and emergency resource distribution data. A communication feature vector is formed by extracting communication quality parameter features. The risk distribution matrix, emergency resource accessibility matrix and communication feature vector are input into a cross-domain association mapping engine to perform coupling strength calculation, generating a basic risk probability distribution matrix and a communication quality confidence vector. The system receives the basic risk probability distribution matrix and the communication quality confidence vector. It inputs the basic risk probability distribution matrix into the power supply risk propagation subnetwork to perform forward extrapolation calculation to obtain the power supply risk status quantification value. It inputs the communication quality confidence vector into the communication network health assessment subnetwork to perform reverse impact calculation to obtain the communication network health quantification value. It performs bidirectional feedback iterative correction on the power supply risk status quantification value and the communication network health quantification value, and outputs the power supply risk status index and the communication network health index. The threshold floating benchmark is calculated based on the power supply risk status index and the communication network health index. The real-time external event intensity factor is introduced to weight and correct the threshold floating benchmark. The weighted and corrected threshold sequence is then subjected to hierarchical boundary optimization and stability verification to establish a risk warning hierarchical threshold set. The system receives a set of risk warning classification thresholds, maps these thresholds to SMS channel coding rules, voice channel coding rules, and private network data channel coding rules, prioritizes and couples the coding of each channel with time synchronization, and generates a differentiated power supply risk communication warning instruction set.

[0005] Preferably, the acquisition of multi-source heterogeneous power supply risk communication data includes power grid operation monitoring data, meteorological environment data, emergency resource distribution data, and communication network quality data. Outlier removal, missing value imputation, and unification of multiple time granularities and spatial coordinate systems are performed on the multi-source heterogeneous power supply risk communication data to form an initial power supply risk communication dataset, which includes: Real-time measurements are collected from the power grid SCADA system, synchronous phasor time-series data are collected from the PMU device, gridded numerical forecast data are collected from the meteorological Doppler radar, GIS coordinate point data are collected from the emergency resource management platform, and performance monitoring byte streams are collected from the communication OTN optical network. A five-dimensional data cube structure with time dimension, space dimension, equipment dimension, index dimension and source dimension is constructed. Sliding window anomaly detection and spatiotemporal kriging interpolation are performed on the five-dimensional data cube. Three-point median filtering is used to remove abrupt outliers, and spatiotemporal joint kriging interpolation is used to complete missing data points. The time dimension is uniformly aligned to the UTC standard timestamp, and the spatial dimension is uniformly transformed to the Gauss-Kruger projection coordinate system to obtain the data tensor after the spatiotemporal reference is unified. The data tensor after unifying the spatiotemporal reference is transformed into a dynamically weighted graph structure. Substation nodes, communication base station nodes, and emergency warehouse nodes are mapped as graph vertices, and power transmission lines, communication fiber optic links, and road traffic networks are mapped as graph edges. Edge weights are dynamically calculated based on line load rate and fiber error rate. The features of neighboring nodes are aggregated through the graph neural network message passing mechanism to form an initial power supply risk communication dataset.

[0006] Preferably, the initial power supply risk communication dataset is invoked, and a power grid feature matrix is ​​formed by hierarchical extraction of power grid state features. A meteorological feature vector is formed by encoding meteorological influencing factors. A preliminary risk assessment is performed based on the power grid feature matrix and the meteorological feature vector to obtain a risk distribution matrix. An emergency resource accessibility matrix is ​​calculated based on the risk distribution matrix and emergency resource distribution data. A communication feature vector is formed by extracting communication quality parameter features. The risk distribution matrix, emergency resource accessibility matrix, and communication feature vector are input into a cross-domain association mapping engine to perform coupling strength calculation, generating a basic risk probability distribution matrix and a communication quality confidence vector, including: Voltage hierarchy decomposition and load density clustering are performed on the power grid measurement data in the initial power supply risk communication dataset. The voltage phasors of 220kV and above nodes are used to form a high voltage feature layer, the voltage and power of 110kV nodes are used to form a medium voltage feature layer, and the 10kV feeder loads are used to form a low voltage feature layer. The density peak clustering algorithm is used to dynamically group the load nodes to form a three-level power grid state feature tensor. The typhoon wind field model was fitted and the icing growth differential equation was solved on the meteorological grid data in the initial power supply risk communication dataset. The wind speed and direction of each node were calculated using the empirical formula of the Miyazaki-Takahashi typhoon wind field. The icing thickness of the conductor was calculated using the Jones icing growth model. The calculation results were mapped to the power grid topology nodes to generate the meteorological disaster risk field matrix. The three-level power grid state characteristic tensor and the meteorological disaster risk field matrix are input into the emergency resource accessibility calculation unit. Combined with the emergency resource GIS coordinates and real-time road traffic data, the Dijkstra shortest path algorithm is used to calculate the shortest time for emergency power generation vehicles and maintenance teams to reach each risk node, forming an emergency resource accessibility matrix. Simultaneously, wavelet packet decomposition and modulation domain feature extraction are performed on the communication quality parameters to form a communication quality feature vector. The risk distribution matrix, emergency resource accessibility matrix, and communication quality feature vector are input into the cross-domain association mapping engine. The coupling strength coefficient between features of different domains is calculated through the heterogeneous graph attention mechanism. Bilinear pooling is performed on the coupled features to generate the basic risk probability distribution matrix and communication quality confidence vector.

[0007] Preferably, the method involves receiving the basic risk probability distribution matrix and the communication quality confidence vector, inputting the basic risk probability distribution matrix into the power supply risk propagation subnetwork to perform forward extrapolation calculations to obtain a quantitative value of the power supply risk situation, and inputting the communication quality confidence vector into the communication network health assessment subnetwork to perform reverse impact calculations to obtain a quantitative value of the communication network health. The quantitative values ​​of the power supply risk situation and the quantitative values ​​of the communication network health are then subjected to bidirectional feedback iterative correction to output a power supply risk situation index and a communication network health index, including: The basic risk probability distribution matrix is ​​input into the power supply risk propagation subnetwork, and forward extrapolation is performed through the graph attention propagation layer. Each layer of propagation aggregates the risk probabilities of neighboring nodes according to the edge weights and introduces the meteorological risk field intensity of the node itself. The output is obtained by calculating the quantitative value of the power supply risk situation through the ReLU activation function. The communication quality confidence vector is input into the communication network health assessment subnetwork and calculated through the health status back propagation mechanism. The health impact factor is passed layer by layer from the core backbone network node to the edge access network node. The fiber optic link attenuation and wireless channel interference coefficient are introduced at each layer to calculate the quantitative value of the communication network health. A two-way coupled feedback loop is constructed between the quantitative value of power supply risk status and the quantitative value of communication network health. The quantitative value of power supply risk status is used as the input of communication network load stress, and the quantitative value of communication network health is used as the risk propagation efficiency correction factor. Iterative correction is performed until the difference between two consecutive outputs is less than the set convergence tolerance. The power supply risk status index and communication network health index are output.

[0008] Secondly, this application also provides a power supply risk communication early warning system, including: Acquisition Module: Used to acquire multi-source heterogeneous power supply risk communication data, including power grid operation monitoring data, meteorological environment data, emergency resource distribution data, and communication network quality data. It performs outlier removal, missing value imputation, and unification of multiple time granularities and spatial coordinate system on the multi-source heterogeneous power supply risk communication data to form an initial power supply risk communication dataset. The first calculation module is used to call the initial power supply risk communication dataset, extract power grid feature matrix through hierarchical extraction of power grid state features, generate meteorological feature vector through feature encoding of meteorological influencing factors, and perform preliminary risk assessment based on the power grid feature matrix and meteorological feature vector to obtain risk distribution matrix; calculate emergency resource accessibility matrix based on risk distribution matrix and emergency resource distribution data, generate communication feature vector through communication quality parameter feature extraction, and input the risk distribution matrix, emergency resource accessibility matrix and communication feature vector into cross-domain association mapping engine to perform coupling strength calculation, generating basic risk probability distribution matrix and communication quality confidence vector; The second calculation module receives the basic risk probability distribution matrix and the communication quality confidence vector. It inputs the basic risk probability distribution matrix into the power supply risk propagation subnetwork to perform forward extrapolation calculation to obtain the power supply risk status quantification value. It inputs the communication quality confidence vector into the communication network health assessment subnetwork to perform reverse impact calculation to obtain the communication network health quantification value. It performs bidirectional feedback iterative correction on the power supply risk status quantification value and the communication network health quantification value, and outputs the power supply risk status index and the communication network health index. Correction module: It is used to calculate the threshold floating benchmark based on the power supply risk status index and the communication network health index, introduce the real-time external event intensity factor to weight and correct the threshold floating benchmark, perform hierarchical boundary optimization and stability verification on the weighted and corrected threshold sequence, and establish a risk warning hierarchical threshold set. Coupling module: Used to receive the risk warning classification threshold set, map the risk warning classification threshold set to SMS channel encoding rules, voice channel encoding rules and private network data channel encoding rules, prioritize and time-synchronize the encoding of each channel, and generate a differentiated power supply risk communication warning instruction set.

[0009] Thirdly, this application also provides a power supply risk communication early warning device, including: Memory, used to store computer programs; A processor is used to implement the steps of the power supply risk communication early warning method when executing the computer program.

[0010] Fourthly, this application also provides a readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described power supply risk communication early warning method.

[0011] The beneficial effects of this invention are as follows: This invention constructs a five-dimensional data cube encompassing time, space, equipment, indicators, and sources, and transforms it into a dynamically weighted graph model. It integrates the power grid transmission load rate and communication link error rate to calculate edge weights, breaking down data barriers between the power grid, meteorology, emergency resources, and communication networks. This achieves the unification and structured fusion of spatiotemporal benchmarks for multi-source heterogeneous data, significantly improving the data integrity and timeliness of risk assessment.

[0012] This invention employs a three-level power grid feature layer extraction and meteorological disaster risk field matrix modeling, combined with a heterogeneous graph attention mechanism to calculate the cross-domain coupling strength coefficient between the risk domain and the communication domain. This can accurately characterize the bidirectional impact relationship between power supply risk propagation and communication network health, solving the defect of risk assessment and communication status being disconnected in traditional technologies.

[0013] This invention employs a bidirectional coupled simulation engine to perform iterative feedback corrections for forward risk propagation and reverse health assessment. It achieves closed-loop linkage correction between the quantified power supply risk status and the quantified communication network health, enhancing the system's dynamic adaptability and simulation accuracy to extreme events. By introducing intensity factors from external events such as typhoon intensity and earthquake intensity to weighted correction thresholds, and using K-means clustering and Monte Carlo simulation to optimize the early warning classification boundaries, an adaptively evolving set of risk early warning classification thresholds is formed, overcoming the limitation of fixed thresholds failing to match changes in extreme scenarios.

[0014] This invention constructs a channel priority evaluation function based on channel bandwidth, latency, reliability, and load rate. It performs dynamic sorting and collaborative coding on the three channels of SMS, voice, and private network data to generate a differentiated power supply risk communication early warning instruction set. This achieves precise matching between early warning information and channel capabilities, significantly reduces the packet loss rate and transmission latency of early warning information, and improves the response speed and decision-making accuracy of emergency command.

[0015] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the power supply risk communication early warning method described in the embodiments of the present invention; Figure 2 This is a schematic diagram of the power supply risk communication early warning system described in this embodiment of the invention; Figure 3 This is a schematic diagram of the power supply risk communication early warning device described in an embodiment of the present invention.

[0018] In the diagram: 701, Acquisition module; 702, First calculation module; 703, Second calculation module; 704, Correction module; 705, Coupling module; 800, Power supply risk communication early warning device; 801, Processor; 802, Memory; 803, Multimedia component; 804, I / O interface; 805, Communication component. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0020] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance. Example 1:

[0021] This embodiment provides a method for communication and early warning of power supply risks.

[0022] See Figure 1 The figure shows that the method includes steps S100, S200, S300, S400 and S500.

[0023] S100. Acquire multi-source heterogeneous power supply risk communication data, including power grid operation monitoring data, meteorological environment data, emergency resource distribution data and communication network quality data. Perform outlier removal, missing value imputation, and unification of multiple time granularities and spatial coordinate system on the multi-source heterogeneous power supply risk communication data to form an initial power supply risk communication dataset. It is understood that step S100 includes S101, S102, and S103, wherein: S101. Collect real-time measurement values ​​from the power grid SCADA system, collect synchronous phasor time series data from the PMU device, collect gridded numerical forecast data from the meteorological Doppler radar, collect GIS coordinate point data from the emergency resource management platform, and collect performance monitoring byte streams from the communication OTN optical network to construct a five-dimensional data cube structure with time dimension, space dimension, equipment dimension, index dimension and source dimension. It should be noted that the time dimension uses a 64-bit UTC nanosecond-level timestamp, the spatial dimension uses a dual representation of Gauss-Kruger projection plane coordinates and WGS84 latitude and longitude, the equipment dimension establishes a unified coding system to map power grid equipment IDs, communication network element IDs, and emergency resource numbers, the index dimension defines a standardized dimensional library to normalize voltage, wind speed, bandwidth, and distance, and the source dimension marks data credibility weights. This five-dimensional structure breaks through the paradigm constraints of traditional two-dimensional relational databases and supports parallel operations at the tensor level. In actual deployment, a multi-protocol adapter needs to be deployed on the front-end machine of the dispatch master station to capture SCADA data via Modbus-TCP, parse PMU data via the IEEE C37.118 protocol, download meteorological grid data via FTP, call emergency resource services via RESTful APIs, and receive OTN alarms via SNMP Trap. The adapter parses the raw messages and injects them into the Apache Kafka message queue, buffering and aligning them according to time windows. The innovation here lies in not relying on a centralized data warehouse, but building a dynamic data cube directly in memory through the streaming computing engine Flink, compressing the data access latency from minutes to seconds, and realizing semantic-level fusion of data from four domains: power grid, meteorology, emergency response, and communications. This provides a unified spatiotemporal benchmark and data structure support for subsequent cross-domain coupled analysis.

[0024] S102. Perform sliding window anomaly detection and spatiotemporal kriging interpolation on the five-dimensional data cube. Use three-point median filtering to remove abrupt outliers and use spatiotemporal joint kriging interpolation to fill in missing data points. Unify the time dimension to the UTC standard timestamp and unify the spatial dimension to the Gauss-Kruger projection coordinate system to obtain the data tensor after the spatiotemporal reference is unified. It should be noted that the sliding window anomaly detection employs a variable-length window strategy, with the window length dynamically adjusted according to the data type: a 100ms window for PMU phasor data to capture low-frequency oscillations, a 30-minute window for meteorological data to filter out random disturbances from convective cells, and a 2-hour window for emergency resource data to smooth out instantaneous fluctuations in traffic conditions. The three-point median filtering does not simply take the median; instead, it constructs a weighted median filter, assigning a weight of 0.5 to the center point and 0.25 to the two points before and after it. This design suppresses impulse noise while preserving data trend characteristics, avoiding excessive smoothing of effective transition signals by traditional median filtering. For missing value imputation, the spatiotemporal joint Kriging interpolation method creatively introduces the physical coupling relationship between power, meteorology, and transportation as an anisotropy constraint. The semi-variogram no longer relies solely on Euclidean distance but constructs a composite distance metric of "electrical distance + meteorological similarity + traffic accessibility." For example, although two devices may be geographically close, if they are located in different power supply zones and have vastly different meteorological conditions, their spatial correlation weight will be suppressed. Aligning the time dimension to UTC timestamps requires handling complex scenarios such as daylight saving time switching and leap second insertion. This application adopts a linear interpolation alignment strategy, scaling the time period spanning leap seconds according to the actual physical time ratio, thus avoiding the timing distortion caused by simple frame duplication or deletion.

[0025] S103. The data tensor after unifying the spatiotemporal reference is transformed into a dynamically weighted graph structure. Substation nodes, communication base station nodes, and emergency warehouse nodes are mapped as graph vertices, and power transmission lines, communication fiber optic links, and road traffic networks are mapped as graph edges. Edge weights are dynamically calculated based on line load rate and fiber optic bit error rate. Neighbor node features are aggregated through a graph neural network message passing mechanism to form an initial power supply risk communication dataset. The edge weight calculation formula is as follows: In the formula, For nodes With nodes The weight of the connecting edges between them, The power transmission weighting coefficient ranges from 0.6 to 0.8, indicating the line... The current load power, Indicates the line Rated capacity, The communication quality weighting coefficient ranges from 0.2 to 0.4, representing the communication link. The bit error rate.

[0026] It should be noted that transforming the unified spatiotemporal data tensor into a graph structure is a crucial bridge connecting the data layer and the model layer. Graph vertex mapping is not limited to node enumeration, but rather injects a multi-domain attribute vector into each vertex: substation vertices include voltage level, load rate, and important user correlation; communication base station vertices include channel capacity, interference level, and backup power supply duration; emergency warehouse vertices include generator power, fuel reserves, and team size. Graph edge mapping needs to handle the issues of directionality and multiplicity. Power lines establish directed edges according to power flow direction, while parallel communication optical fibers and road traffic constitute multi-edges. This application uses a hypergraph representation to aggregate multi-edges into hyperedges, where the weight of the hyperedge is the weighted sum of the weights of its sub-edges. The graph neural network message passing mechanism adopts the sampling aggregation strategy of GraphSAGE, sampling the importance of higher-order neighbors instead of full aggregation. The sampling probability is proportional to the edge weight. The aggregation function uses LSTM to capture the temporal dependencies between neighboring nodes, rather than simple mean or maximum pooling. In practical applications, the graph structure is dynamically updated every 5 minutes, and the addition and deletion of vertices and edges are achieved through incremental calculation, avoiding the computational overhead of full graph reconstruction.

[0027] S200: Call the initial power supply risk communication dataset, extract power grid feature matrix through hierarchical extraction of power grid state features, and form meteorological feature vector through feature encoding of meteorological influencing factors. Based on the power grid feature matrix and meteorological feature vector, perform preliminary risk assessment to obtain risk distribution matrix; calculate emergency resource accessibility matrix based on risk distribution matrix and emergency resource distribution data, and form communication feature vector through communication quality parameter feature extraction. Input the risk distribution matrix, emergency resource accessibility matrix and communication feature vector into cross-domain association mapping engine to perform coupling strength calculation, and generate basic risk probability distribution matrix and communication quality confidence vector. It is understood that step S200 includes S201, S202, S203, and S204, wherein: S201. Perform voltage hierarchy decomposition and load density clustering on the power grid measurement data in the initial power supply risk communication dataset. Construct a high-voltage feature layer from the voltage phasors of 220kV and above nodes, construct a medium-voltage feature layer from the voltage and power of 110kV nodes, construct a low-voltage feature layer from the 10kV feeder loads, and use the density peak clustering algorithm to dynamically group the load nodes to form a three-level power grid state feature tensor. It should be noted that this step breaks away from the simplistic approach of treating the power grid as a homogeneous node in traditional risk assessment. Instead, it constructs a voltage-level sensitive three-level feature tensor to achieve differentiated modeling of fault propagation characteristics at different voltage levels. The characteristics of the 220kV and above high-voltage layer not only include node voltage phasors (amplitude and phase angle), but also creatively introduce a coupling strength index with the lower-level power grid. The cascading impact effect of high-voltage faults on the medium-voltage distribution network is quantified through the tie-line power fluctuation covariance matrix. That is, through three-level decoupling, the global stability of the high-voltage layer, the regional balance of the medium-voltage layer, and the local vulnerability of the low-voltage layer are separated. In subsequent risk propagation calculations, high-voltage faults can be quickly mapped to the medium-voltage influence domain, and medium-voltage disturbances can be accurately located to low-voltage vulnerable clusters. This achieves a dual improvement in computational efficiency and assessment accuracy, avoiding the false alarm problem of "small high-voltage disturbances triggering large low-voltage alarms" caused by feature confusion in traditional methods.

[0028] S202. Perform typhoon wind field model fitting and icing growth differential equation solution on the meteorological grid data in the initial power supply risk communication dataset. Use the Miyazaki-Takahashi typhoon wind field empirical formula to calculate the wind speed and direction of each node. Use the Jones icing growth model to calculate the conductor icing thickness. Map the calculation results to the power grid topology nodes to generate the meteorological disaster risk field matrix. It should be noted that the wind field model fitting adopts the Miyazaki-Takahashi empirical formula. Its innovation lies in not directly using the wind speeds observed at weather stations, but instead substituting parameters such as the typhoon's central pressure, movement speed, and maximum wind speed radius into the formula to calculate the gradient wind field at each power grid node location, and then superimposing a terrain friction correction coefficient. The calculated wind speed and icing thickness are not stored in isolation, but are embedded into the power grid node attributes through "topology mapping". Specifically, for each transmission tower node, the wind speed vector is decomposed into the vertical span direction and the longitudinal direction. The vertical component is used to calculate the conductor wind deflection angle, and the longitudinal component is used to calculate the span change. The icing thickness is calculated by combining the conductor type, span length, and sag height to calculate the equivalent mechanical load increment. This mapping transforms meteorological risk from the "meteorological station space" to the "power grid topology space", achieving equipment-level accuracy in the risk field. In other words, through topology mapping, the risk field and the power grid structure are deeply integrated, enabling subsequent accessibility calculations to accurately assess real-world scenarios such as "repair vehicles needing to cross icy road sections." This avoids the drawbacks of traditional methods that separate the assessment of meteorological risk and power grid risk, and provides an input field that conforms to the laws of physics for coupling strength calculation.

[0029] S203. Input the three-level power grid state characteristic tensor and the meteorological disaster risk field matrix into the emergency resource accessibility calculation unit. Combine the emergency resource GIS coordinates and real-time road traffic data. Use Dijkstra's shortest path algorithm to calculate the shortest time for emergency power generation vehicles and maintenance teams to reach each risk node, forming an emergency resource accessibility matrix. Simultaneously, perform wavelet packet decomposition and modulation domain feature extraction on the communication quality parameters to form a communication quality feature vector. It should be noted that traditional methods often use static Euclidean distance or a single fastest path, neglecting the impact of real-time traffic conditions and communication quality on the transmission of scheduling instructions. When using Dijkstra's algorithm to calculate the shortest time path, the edge weights are no longer road lengths, but rather a composite cost that integrates "travel time + road capacity + traffic control status." Communication quality feature extraction uses wavelet packet decomposition instead of Fourier transform, which has the advantage of simultaneously analyzing time-frequency localization features. The performance monitoring byte stream of the OTN optical network is decomposed into eight sub-bands. For each sub-band, three features are extracted: energy entropy, kurtosis, and impulse factor. Energy entropy reflects the dispersion of link noise, kurtosis captures the spike characteristics of sudden interference, and impulse factor identifies periodic bit errors. The vector formed by these three features can distinguish different fault modes such as fiber aging, bending loss, and external construction vibration.

[0030] S204. Input the risk distribution matrix, emergency resource accessibility matrix, and communication quality feature vector into the cross-domain association mapping engine. Calculate the coupling strength coefficient between features from different domains using a heterogeneous graph attention mechanism. Perform bilinear pooling on the coupled features to generate the basic risk probability distribution matrix and communication quality confidence vector. The formula for calculating the cross-domain coupling strength coefficient is as follows: In the formula, Represents risk domain nodes With communication domain nodes The cross-domain coupling strength coefficient between them Represents the risk domain query vector. Represents the communication domain key vector. This represents the dimension of the feature vector, where N represents the total number of neighboring nodes. This represents the normalized denominator over all neighboring nodes.

[0031] It should be noted that the matrix generated by the outer product in this step is sparse. The combination of high load and low communication quality will produce a maximum value, becoming a key triggering factor for subsequent risk assessment. The technical effect is reflected in the fact that the generated coupling strength coefficient matrix can be used to identify "critical coupling edges," that is, those weak links with high power load and poor communication quality. These edges will exhibit a nonlinear amplification effect in subsequent risk propagation. The basic risk probability distribution matrix is ​​no longer an independent grid risk, but a conditional probability distribution under communication constraints. The communication quality confidence vector is also no longer independent, but a dynamic assessment driven by risk. The two together constitute the input state for subsequent bidirectional coupling inference, realizing a paradigm leap from "data fusion" to "causal coupling." This is the essential innovation in this step that distinguishes it from the "data-stacking fusion" of existing technologies.

[0032] S300 receives the basic risk probability distribution matrix and the communication quality confidence vector. It inputs the basic risk probability distribution matrix into the power supply risk propagation subnetwork to perform forward extrapolation calculation to obtain the power supply risk status quantification value. It inputs the communication quality confidence vector into the communication network health assessment subnetwork to perform reverse impact calculation to obtain the communication network health quantification value. It performs bidirectional feedback iterative correction on the power supply risk status quantification value and the communication network health quantification value, and outputs the power supply risk status index and the communication network health index. It is understood that step S300 includes S301, S302, and S303, wherein: S301. Input the basic risk probability distribution matrix into the power supply risk propagation subnetwork, and perform forward inference through the graph attention propagation layer. Each layer of propagation aggregates the risk probabilities of neighboring nodes according to the edge weights, and introduces the meteorological risk field intensity of the node itself. The output is obtained by calculating the quantitative value of the power supply risk situation through the ReLU activation function. S302. Input the communication quality confidence vector into the communication network health assessment sub-network, and calculate it through the health status back propagation mechanism. The health impact factor is passed from the core backbone network node to the edge access network node layer by layer. The fiber optic link attenuation and wireless channel interference coefficient are introduced at each layer to calculate the quantitative value of the communication network health. It should be noted that when the gating factor exceeds the design threshold, node i's "immunity" to neighbor risk propagation decreases, and the aggregation weight automatically increases. This mechanism accurately simulates the actual operating conditions where equipment insulation margin is reduced and it is more susceptible to the impact of neighboring faults under extreme weather conditions. Finally, nonlinear mapping is performed through the ReLU activation function. The introduction of ReLU is not a simple numerical pruning; its essence is to simulate the operating characteristics of the protection device—the output is zero when the accumulated risk is below the activation threshold (protection blocking), and the output is linearly proportional when it is above the threshold (protection action). This biomimetic design gives risk propagation a clear physical triggering mechanism, avoiding the distortion problem of "chronic penetration" of risk caused by the traditional Sigmoid function. In practical applications, forward extrapolation is usually set to 3-4 layers of propagation, corresponding to the 3-4 order of the fault's impact range in the power grid. Taking a typhoon scenario in a coastal power grid as an example, the simulation process involves four layers: the first layer focuses on capturing the risk of tower failures directly swept by the typhoon's eyewall; the second layer simulates the risk of 10kV bus voltage collapse caused by these tower failures; the third layer simulates the risk of regional low-frequency load shedding triggered by voltage collapse; and the fourth layer simulates the economic losses to industrial users due to power outages caused by load shedding. The output of each layer serves as the input for the next layer, forming a cascaded computational chain for risk propagation.

[0033] Understandably, the design of the backpropagation mechanism in this step to assess the health status of the communication network stems from the fault propagation characteristics of the "core-edge" architecture of the communication network. Traditional methods for assessing communication quality often adopt a "sensing-reporting" model, where edge nodes report anomalies to the core network management system. This model suffers from high latency and cannot predict the secondary impact of core failures on the edge. In the backpropagation process, health impact factors are not simply added together, but rather a "barrel effect" mechanism is used—the final health of a node is determined by the worst health value among all its upstream paths. This aligns with the actual characteristic of communication networks where "a single point of failure leads to a complete service interruption." The calculated quantitative value of the communication network health is an end-to-end assessment of the quality of the early warning information transmission channel, directly determining whether risk alarms can be reliably delivered. In contrast to the quantitative value of the risk situation derived from the forward extrapolation, these two "positive" and "negative" quantitative values ​​will form a closed-loop feedback in the next step—low health will reduce the success rate of risk information transmission, while high-risk areas will exacerbate the communication network load and further lower the health. This two-way feedback mechanism is the key to this application's breakthrough in traditional one-way assessment.

[0034] S303. Construct a bidirectional coupled feedback loop between the quantitative value of power supply risk status and the quantitative value of communication network health. Use the quantitative value of power supply risk status as the input of communication network load stress, and the quantitative value of communication network health as a risk propagation efficiency correction factor. Perform iterative correction until the difference between two consecutive outputs is less than a set convergence tolerance, and output the power supply risk status index and the communication network health index. The bidirectional coupled feedback correction formula is: In the formula, This represents the power supply risk situation index vector for the (t+1)th iteration. This represents the vector of quantitative values ​​of the power supply risk situation in the t-th iteration. This represents the vector of communication network health quantization values ​​in the t-th iteration. This represents the communication network health index vector for the (t+1)th iteration. Represents the risk state weight matrix. This represents the weight matrix indicating the impact of communication. Represents the weight matrix of health status. This represents the communication quality confidence vector. Let represent the meteorological stress matrix, σ represent the Sigmoid activation function, and ⊙ represent the Hadamard product.

[0035] It should be noted that the core idea of ​​bidirectional coupling is rooted in the real mutual feedback in the operation of the power system: On the one hand, an increase in grid risk will trigger a large number of sudden uplink traffic such as protection action signals, PMU high-speed data, and video monitoring streams, causing a surge in communication network load, resulting in switch buffer overflow, tight optical link power budget, and intensified wireless channel competition, thereby deteriorating communication quality; on the other hand, the deterioration of communication quality will delay the uploading of protection signals, reduce the success rate of dispatching instructions, and interrupt emergency resource location information, making it impossible for dispatchers to grasp the risk evolution in a timely manner, delaying control measures, and thus exacerbating the spread of risks.

[0036] Furthermore, the iterative correction process employs a dual-timescale advancement strategy: a forward risk projection is performed every second on the fast timescale to capture the millisecond-second dynamics of the fault; a backward health propagation is performed every 10 seconds on the slow timescale to reflect the second-minute evolution of the communication fault. The two quantified values ​​are interactively updated after each iteration, forming a computational flow.

[0037] Understandably, the output power supply risk status index and communication network health index are no longer open-loop calculation results, but closed-loop equilibrium solutions after multiple rounds of mutual feedback correction. They can self-consistently reflect the vicious cycle of risk deterioration → communication pressure → further risk deterioration, and their numerical accuracy and scenario realism far exceed traditional one-way assessments. This mechanism enables the system to predict the emergent behavior of complex "risk-communication" systems, providing early warning signals for preventing cascading collapses in extreme events.

[0038] S400: Calculate the threshold floating benchmark based on the power supply risk status index and the communication network health index, introduce the real-time external event intensity factor to weight and correct the threshold floating benchmark, perform hierarchical boundary optimization and stability verification on the weighted and corrected threshold sequence, and establish a risk warning hierarchical threshold set. Understandably, in this step, the historical time series of the power supply risk status index and the communication network health index are read, the trend component is extracted by using the moving average filter, the comprehensive weighted value of risk and health is calculated, and the dynamic standard deviation of the comprehensive weighted value is used as the threshold floating benchmark. By incorporating external event intensity factors such as typhoon intensity level, earthquake intensity, wildfire spread rate, and peak power grid load, and normalizing each factor to construct a weight vector, a vector dot product weighted correction is performed on the threshold floating benchmark to obtain an event-driven threshold sequence. Hierarchical boundary optimization is performed on event-driven threshold sequences. The threshold space is divided into four warning intervals using the K-means clustering algorithm. Cubic spline smoothing is performed on the interval boundaries. The stability of the threshold set is verified by Monte Carlo simulation, and the risk warning hierarchical threshold set is established. The external event-driven weighted correction formula is as follows: In the formula, This represents an event-driven threshold sequence. Indicates the threshold floating benchmark. This represents the weighting coefficient of the intensity factor of the k-th external event. This represents the normalized value of the typhoon intensity level. This represents the normalized value of earthquake intensity. This represents the normalized value of the wildfire spread rate. This represents the normalized value of the peak load of the power grid. This represents the weighted sum of four external event factors.

[0039] S500 receives a set of risk warning classification thresholds, maps the risk warning classification thresholds to SMS channel coding rules, voice channel coding rules, and private network data channel coding rules, prioritizes and couples the coding of each channel with time sequence, and generates a differentiated power supply risk communication warning instruction set.

[0040] It is understandable that the risk warning classification threshold set is mapped to SMS text encoding rules, voice waveform encoding rules and private network data frame encoding rules, and the upper limit of SMS character length, speech synthesis speed parameters and data frame compression ratio parameters are determined according to the threshold level to form a three-channel encoding parameter vector; Construct a channel priority evaluation function. The function input includes channel bandwidth, transmission delay, reliability index and current load rate. Use the analytic hierarchy process to calculate the dynamic priority score of each channel. Based on the score ranking result, perform time synchronization marking on the three-channel coding parameter vector. The marking information includes the transmission time offset and the upper limit of retransmission times. The three-channel encoding parameter vector after synchronization is input into the multi-channel collaborative encoder. Interleaving encoding and error correction code addition are performed according to the timing mark to generate short messages for the SMS channel, TTS instructions for the voice channel, and binary streams for the private network data channel. These are combined to form a differentiated power supply risk communication early warning instruction set. Example 2:

[0041] like Figure 2 As shown, this embodiment provides a power supply risk communication early warning system. See [link / reference]. Figure 2 The system includes: Acquisition Module 701: Used to acquire multi-source heterogeneous power supply risk communication data, including power grid operation monitoring data, meteorological environment data, emergency resource distribution data and communication network quality data. It performs outlier removal, missing value imputation, and unification of multiple time granularities and spatial coordinate system on the multi-source heterogeneous power supply risk communication data to form an initial power supply risk communication dataset. The first calculation module 702 is used to call the initial power supply risk communication dataset, extract power grid feature matrix through hierarchical extraction of power grid state features, and form meteorological feature vector through feature encoding of meteorological influencing factors. Based on the power grid feature matrix and meteorological feature vector, a preliminary risk assessment is performed to obtain a risk distribution matrix. Based on the risk distribution matrix and emergency resource distribution data, an emergency resource accessibility matrix is ​​calculated, and a communication feature vector is formed through feature extraction of communication quality parameters. The risk distribution matrix, emergency resource accessibility matrix and communication feature vector are input into the cross-domain association mapping engine to perform coupling strength calculation, generating a basic risk probability distribution matrix and a communication quality confidence vector. The second calculation module 703 is used to receive the basic risk probability distribution matrix and the communication quality confidence vector, input the basic risk probability distribution matrix into the power supply risk propagation subnetwork to perform forward extrapolation calculation to obtain the power supply risk situation quantification value, input the communication quality confidence vector into the communication network health assessment subnetwork to perform reverse impact calculation to obtain the communication network health quantification value, perform bidirectional feedback iterative correction on the power supply risk situation quantification value and the communication network health quantification value, and output the power supply risk situation index and the communication network health index. Correction module 704: is used to calculate the threshold floating benchmark based on the power supply risk status index and the communication network health index, introduce the real-time external event intensity factor to weight and correct the threshold floating benchmark, perform hierarchical boundary optimization and stability verification on the weighted and corrected threshold sequence, and establish a risk warning hierarchical threshold set. Coupling module 705: Used to receive the risk warning classification threshold set, map the risk warning classification threshold set to SMS channel coding rules, voice channel coding rules and private network data channel coding rules, prioritize and time-synchronize the coding of each channel, and generate a differentiated power supply risk communication warning instruction set.

[0042] Specifically, the acquisition module 701 includes: Construction Unit: Used to collect real-time measurement values ​​from the power grid SCADA system, synchronous phasor time series data from the PMU device, gridded numerical forecast data from the meteorological Doppler radar, GIS coordinate point data from the emergency resource management platform, and performance monitoring byte streams from the communication OTN optical network, constructing a five-dimensional data cube structure with time dimension, spatial dimension, equipment dimension, index dimension and source dimension; Unified Unit: Used to perform sliding window anomaly detection and spatiotemporal kriging interpolation on a five-dimensional data cube. It uses three-point median filtering to remove abrupt outliers, spatiotemporal joint kriging interpolation to fill in missing data points, and unifies the time dimension to the UTC standard timestamp and the spatial dimension to the Gauss-Kruger projection coordinate system to obtain a data tensor with unified spatiotemporal reference. Mapping Calculation Unit: This unit transforms the data tensor after unifying the spatiotemporal reference into a dynamically weighted graph structure. It maps substation nodes, communication base station nodes, and emergency warehouse nodes as graph vertices, and power transmission lines, communication fiber optic links, and road networks as graph edges. It dynamically calculates edge weights based on line load rate and fiber optic bit error rate, and aggregates neighbor node features via a graph neural network message passing mechanism to form an initial power supply risk communication dataset. The edge weight calculation formula is as follows: In the formula, For nodes With nodes The weight of the connecting edges between them, The power transmission weighting coefficient ranges from 0.6 to 0.8, indicating the line... The current load power, Indicates the line Rated capacity, The communication quality weighting coefficient ranges from 0.2 to 0.4, representing the communication link. The bit error rate.

[0043] Specifically, the first computing module 702 includes: Clustering Unit: Used to perform voltage hierarchy decomposition and load density clustering on the power grid measurement data in the initial power supply risk communication dataset. It constructs a high-voltage feature layer by forming the voltage phasors of 220kV and above nodes, a medium-voltage feature layer by forming the voltage and power of 110kV nodes, and a low-voltage feature layer by forming the load of 10kV feeder loads. It uses the density peak clustering algorithm to dynamically cluster the load nodes and form a three-level power grid state feature tensor. Fitting and solving unit: used to perform typhoon wind field model fitting and icing growth differential equation solving on meteorological grid data in the initial power supply risk communication dataset. It uses the Miyazaki-Takahashi typhoon wind field empirical formula to calculate the wind speed and direction of each node, uses the Jones icing growth model to calculate the conductor icing thickness, maps the calculation results to the power grid topology nodes, and generates a meteorological disaster risk field matrix. Input extraction unit: Used to input the three-level power grid state feature tensor and meteorological disaster risk field matrix into the emergency resource accessibility calculation unit. Combined with emergency resource GIS coordinates and real-time road traffic data, the Dijkstra shortest path algorithm is used to calculate the shortest time for emergency power generation vehicles and maintenance teams to reach each risk node, forming an emergency resource accessibility matrix. Simultaneously, wavelet packet decomposition and modulation domain feature extraction are performed on communication quality parameters to form a communication quality feature vector. The operation input unit is used to input the risk distribution matrix, emergency resource accessibility matrix, and communication quality feature vector into the cross-domain correlation mapping engine. It calculates the coupling strength coefficient between features from different domains via a heterogeneous graph attention mechanism, performs bilinear pooling on the coupled features, and generates the basic risk probability distribution matrix and communication quality confidence vector. The formula for calculating the cross-domain coupling strength coefficient is as follows: In the formula, Represents risk domain nodes With communication domain nodes The cross-domain coupling strength coefficient between them Represents the risk domain query vector. Represents the communication domain key vector. This represents the dimension of the feature vector, where N represents the total number of neighboring nodes. This represents the normalized denominator over all neighboring nodes.

[0044] Specifically, the second computing module 703 includes: Aggregation Unit: Used to input the basic risk probability distribution matrix into the power supply risk propagation subnetwork, and perform forward inference through the graph attention propagation layer. Each layer of propagation aggregates the risk probabilities of neighboring nodes according to the edge weights, and introduces the node's own meteorological risk field intensity. The output is obtained by calculating the quantitative value of the power supply risk situation through the ReLU activation function. Calculation Unit: Used to input the communication quality confidence vector into the communication network health assessment sub-network, and calculate it through the health status back propagation mechanism. The health impact factor is passed layer by layer from the core backbone network node to the edge access network node. The fiber optic link attenuation and wireless channel interference coefficient are introduced at each layer to calculate the quantitative value of the communication network health. Feedback Correction Unit: This unit constructs a bidirectional coupled feedback loop between the quantified power supply risk status and the quantified communication network health. The quantified power supply risk status is used as the input to the communication network load stress, and the quantified communication network health is used as a risk propagation efficiency correction factor. It iterative correction is performed until the difference between two consecutive outputs is less than a set convergence tolerance, outputting the power supply risk status index and the communication network health index. The bidirectional coupled feedback correction formula is as follows: In the formula, This represents the power supply risk situation index vector for the (t+1)th iteration. This represents the vector of quantitative values ​​of the power supply risk situation in the t-th iteration. This represents the vector of communication network health quantization values ​​in the t-th iteration. This represents the communication network health index vector for the (t+1)th iteration. Represents the risk state weight matrix. This represents the weight matrix indicating the impact of communication. Represents the weight matrix of health status. This represents the communication quality confidence vector. Let represent the meteorological stress matrix, σ represent the Sigmoid activation function, and ⊙ represent the Hadamard product.

[0045] In summary, this invention achieves spatiotemporal benchmark unification and graph structure transformation of multi-source heterogeneous data, including power grids, meteorological data, emergency resources, and communication networks, by constructing a five-dimensional data cube and a dynamic weighted graph model. A cross-domain heterogeneous graph attention mechanism is employed to calculate the coupling strength between the risk domain and the communication domain, and a bidirectional coupling inference engine is established to iteratively correct the quantitative values ​​of power supply risk status and communication network health, outputting dynamic risk indices and health indicators. External event intensity factors are introduced to weight and correct thresholds, and K-means clustering and Monte Carlo simulation are used to optimize the early warning classification boundary, forming an adaptive threshold set. Finally, based on a channel priority evaluation function, multiple channels such as SMS, voice, and private network data are dynamically sorted and collaboratively encoded to generate a differentiated early warning instruction set. This method overcomes the limitations of traditional technologies that separate risk assessment and communication early warning, achieving closed-loop linkage between risk status and communication capabilities, and precise adaptation of early warning strategies.

[0046] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here. Example 3:

[0047] Corresponding to the above method embodiments, this embodiment also provides a power supply risk communication early warning device. The power supply risk communication early warning device described below and the power supply risk communication early warning method described above can be referred to in correspondence.

[0048] Figure 3 This is a block diagram illustrating a power supply risk communication early warning device 800 according to an exemplary embodiment. For example... Figure 3 As shown, the power supply risk communication early warning device 800 includes a processor 801 and a memory 802. The power supply risk communication early warning device 800 also includes one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0049] The processor 801 controls the overall operation of the power supply risk communication early warning device 800 to complete all or part of the steps in the aforementioned power supply risk communication early warning method. The memory 802 stores various types of data to support the operation of the power supply risk communication early warning device 800. This data may include, for example, instructions for any application or method operating on the power supply risk communication early warning device 800, as well as application-related data such as contact data, sent and received messages, images, audio, video, etc. The memory 802 can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. I / O interface 804 provides an interface between processor 801 and other interface modules, such as keyboards, mice, or buttons. These buttons can be virtual or physical. Communication component 805 is used for wired or wireless communication between the power supply risk communication early warning device 800 and other devices. Wireless communication includes, for example, Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 805 may include a Wi-Fi module, a Bluetooth module, or an NFC module.

[0050] In an exemplary embodiment, the power supply risk communication early warning device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the aforementioned power supply risk communication early warning method.

[0051] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the power supply risk communication early warning method described above. For example, the computer-readable storage medium may be the memory 802 including the program instructions described above, which may be executed by the processor 801 of the power supply risk communication early warning device 800 to complete the power supply risk communication early warning method described above. Example 4:

[0052] Corresponding to the above method embodiments, this embodiment also provides a readable storage medium. The readable storage medium described below can be referred to in conjunction with the power supply risk communication early warning method described above.

[0053] A computer program is stored on a readable storage medium, and when the computer program is executed by a processor, it implements the steps of the power supply risk communication early warning method described in the above method embodiments.

[0054] Specifically, the readable storage medium can be a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, or any other readable storage medium capable of storing program code.

[0055] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A power supply risk communication early warning method, characterized in that, include: Acquire multi-source heterogeneous power supply risk communication data, including power grid operation monitoring data, meteorological environment data, emergency resource distribution data, and communication network quality data. Perform outlier removal, missing value imputation, and unification of multiple time granularities and spatial coordinate system on the multi-source heterogeneous power supply risk communication data to form an initial power supply risk communication dataset. The initial power supply risk communication dataset is called, and a power grid feature matrix is ​​formed by hierarchical extraction of power grid state features. A meteorological feature vector is formed by encoding meteorological influencing factors. A preliminary risk assessment is performed based on the power grid feature matrix and the meteorological feature vector to obtain a risk distribution matrix. An emergency resource accessibility matrix is ​​calculated based on the risk distribution matrix and emergency resource distribution data. A communication feature vector is formed by extracting communication quality parameter features. The risk distribution matrix, emergency resource accessibility matrix and communication feature vector are input into a cross-domain association mapping engine to perform coupling strength calculation, generating a basic risk probability distribution matrix and a communication quality confidence vector. The system receives the basic risk probability distribution matrix and the communication quality confidence vector. It inputs the basic risk probability distribution matrix into the power supply risk propagation subnetwork to perform forward extrapolation calculation to obtain the power supply risk status quantification value. It inputs the communication quality confidence vector into the communication network health assessment subnetwork to perform reverse impact calculation to obtain the communication network health quantification value. It performs bidirectional feedback iterative correction on the power supply risk status quantification value and the communication network health quantification value, and outputs the power supply risk status index and the communication network health index. The threshold floating benchmark is calculated based on the power supply risk status index and the communication network health index. The real-time external event intensity factor is introduced to weight and correct the threshold floating benchmark. The weighted and corrected threshold sequence is then subjected to hierarchical boundary optimization and stability verification to establish a risk warning hierarchical threshold set. The system receives a set of risk warning classification thresholds, maps these thresholds to SMS channel coding rules, voice channel coding rules, and private network data channel coding rules, prioritizes and couples the coding of each channel with time synchronization, and generates a differentiated power supply risk communication warning instruction set.

2. The power supply risk communication early warning method according to claim 1, characterized in that, The acquisition of multi-source heterogeneous power supply risk communication data includes power grid operation monitoring data, meteorological environment data, emergency resource distribution data, and communication network quality data. Outlier removal, missing value imputation, and unification of multiple time granularities and spatial coordinate systems are performed on the multi-source heterogeneous power supply risk communication data to form an initial power supply risk communication dataset, which includes: Real-time measurements are collected from the power grid SCADA system, synchronous phasor time-series data are collected from the PMU device, gridded numerical forecast data are collected from the meteorological Doppler radar, GIS coordinate point data are collected from the emergency resource management platform, and performance monitoring byte streams are collected from the communication OTN optical network. A five-dimensional data cube structure with time dimension, space dimension, equipment dimension, index dimension and source dimension is constructed. Sliding window anomaly detection and spatiotemporal kriging interpolation are performed on the five-dimensional data cube. Three-point median filtering is used to remove abrupt outliers, and spatiotemporal joint kriging interpolation is used to complete missing data points. The time dimension is uniformly aligned to the UTC standard timestamp, and the spatial dimension is uniformly transformed to the Gauss-Kruger projection coordinate system to obtain the data tensor after the spatiotemporal reference is unified. The data tensor after unifying the spatiotemporal reference is transformed into a dynamically weighted graph structure. Substation nodes, communication base station nodes, and emergency warehouse nodes are mapped as graph vertices, and power transmission lines, communication fiber optic links, and road traffic networks are mapped as graph edges. Edge weights are dynamically calculated based on line load rate and fiber optic bit error rate. Neighbor node features are aggregated through a graph neural network message passing mechanism to form an initial power supply risk communication dataset. The edge weight calculation formula is as follows: In the formula, For nodes With nodes The weight of the connecting edges between them, The power transmission weighting coefficient ranges from 0.6 to 0.8, indicating the line... The current load power, Indicates the line Rated capacity, The communication quality weighting coefficient ranges from 0.2 to 0.4, representing the communication link. The bit error rate.

3. The power supply risk communication early warning method according to claim 1, characterized in that, The process involves calling the initial power supply risk communication dataset, extracting power grid state features to form a power grid feature matrix, encoding meteorological influencing factors to form a meteorological feature vector, and performing a preliminary risk assessment based on the power grid feature matrix and meteorological feature vector to obtain a risk distribution matrix. Based on the risk distribution matrix and emergency resource distribution data, an emergency resource accessibility matrix is ​​calculated, and a communication feature vector is formed through communication quality parameter feature extraction. The risk distribution matrix, emergency resource accessibility matrix, and communication feature vector are then input into a cross-domain association mapping engine to perform coupling strength calculation, generating a basic risk probability distribution matrix and a communication quality confidence vector, including: Voltage hierarchy decomposition and load density clustering are performed on the power grid measurement data in the initial power supply risk communication dataset. The voltage phasors of 220kV and above nodes are used to form a high voltage feature layer, the voltage and power of 110kV nodes are used to form a medium voltage feature layer, and the 10kV feeder loads are used to form a low voltage feature layer. The density peak clustering algorithm is used to dynamically group the load nodes to form a three-level power grid state feature tensor. The typhoon wind field model was fitted and the icing growth differential equation was solved on the meteorological grid data in the initial power supply risk communication dataset. The wind speed and direction of each node were calculated using the empirical formula of the Miyazaki-Takahashi typhoon wind field. The icing thickness of the conductor was calculated using the Jones icing growth model. The calculation results were mapped to the power grid topology nodes to generate the meteorological disaster risk field matrix. The three-level power grid state characteristic tensor and the meteorological disaster risk field matrix are input into the emergency resource accessibility calculation unit. Combined with the emergency resource GIS coordinates and real-time road traffic data, the Dijkstra shortest path algorithm is used to calculate the shortest time for emergency power generation vehicles and maintenance teams to reach each risk node, forming an emergency resource accessibility matrix. Simultaneously, wavelet packet decomposition and modulation domain feature extraction are performed on the communication quality parameters to form a communication quality feature vector. The risk distribution matrix, emergency resource accessibility matrix, and communication quality feature vector are input into the cross-domain association mapping engine. A heterogeneous graph attention mechanism is used to calculate the coupling strength coefficient between features from different domains. Bilinear pooling is then performed on the coupled features to generate the basic risk probability distribution matrix and the communication quality confidence vector. The formula for calculating the cross-domain coupling strength coefficient is as follows: In the formula, Represents risk domain nodes With communication domain nodes The cross-domain coupling strength coefficient between them Represents the risk domain query vector. Represents the communication domain key vector. This represents the dimension of the feature vector, where N represents the total number of neighboring nodes. This represents the normalized denominator over all neighboring nodes.

4. The power supply risk communication early warning method according to claim 1, characterized in that, The system receives the basic risk probability distribution matrix and the communication quality confidence vector. The basic risk probability distribution matrix is ​​input into the power supply risk propagation subnetwork to perform forward inference calculations to obtain a quantitative value of the power supply risk situation. The communication quality confidence vector is input into the communication network health assessment subnetwork to perform reverse impact calculations to obtain a quantitative value of the communication network health. The quantitative values ​​of the power supply risk situation and the quantitative values ​​of the communication network health are then iteratively corrected through bidirectional feedback to output a power supply risk situation index and a communication network health index, including: The basic risk probability distribution matrix is ​​input into the power supply risk propagation subnetwork, and forward extrapolation is performed through the graph attention propagation layer. Each layer of propagation aggregates the risk probabilities of neighboring nodes according to the edge weights and introduces the meteorological risk field intensity of the node itself. The output is obtained by calculating the quantitative value of the power supply risk situation through the ReLU activation function. The communication quality confidence vector is input into the communication network health assessment subnetwork and calculated through the health status back propagation mechanism. The health impact factor is passed layer by layer from the core backbone network node to the edge access network node. The fiber optic link attenuation and wireless channel interference coefficient are introduced at each layer to calculate the quantitative value of the communication network health. A bidirectional coupled feedback loop is constructed, integrating the quantitative values ​​of power supply risk status and communication network health. The quantitative value of power supply risk status is used as the input to the load stress of the communication network, and the quantitative value of communication network health is used as a risk propagation efficiency correction factor. Iterative correction is performed until the difference between two consecutive outputs is less than a set convergence tolerance, outputting the power supply risk status index and the communication network health index. The bidirectional coupled feedback correction formula is as follows: In the formula, This represents the power supply risk situation index vector for the (t+1)th iteration. This represents the vector of quantitative values ​​of the power supply risk situation in the t-th iteration. This represents the vector of communication network health quantization values ​​in the t-th iteration. This represents the communication network health index vector for the (t+1)th iteration. Represents the risk state weight matrix. This represents the weight matrix indicating the impact of communication. Represents the weight matrix of health status. This represents the communication quality confidence vector. Let represent the meteorological stress matrix, σ represent the Sigmoid activation function, and ⊙ represent the Hadamard product.

5. A power supply risk communication early warning system, based on the power supply risk communication early warning method according to claim 1, characterized in that, include: Acquisition Module: Used to acquire multi-source heterogeneous power supply risk communication data, including power grid operation monitoring data, meteorological environment data, emergency resource distribution data, and communication network quality data. It performs outlier removal, missing value imputation, and unification of multiple time granularities and spatial coordinate system on the multi-source heterogeneous power supply risk communication data to form an initial power supply risk communication dataset. The first calculation module is used to call the initial power supply risk communication dataset, extract power grid feature matrix through hierarchical extraction of power grid state features, generate meteorological feature vector through feature encoding of meteorological influencing factors, and perform preliminary risk assessment based on the power grid feature matrix and meteorological feature vector to obtain risk distribution matrix; calculate emergency resource accessibility matrix based on risk distribution matrix and emergency resource distribution data, generate communication feature vector through communication quality parameter feature extraction, and input the risk distribution matrix, emergency resource accessibility matrix and communication feature vector into cross-domain association mapping engine to perform coupling strength calculation, generating basic risk probability distribution matrix and communication quality confidence vector; The second calculation module receives the basic risk probability distribution matrix and the communication quality confidence vector. It inputs the basic risk probability distribution matrix into the power supply risk propagation subnetwork to perform forward extrapolation calculation to obtain the power supply risk status quantification value. It inputs the communication quality confidence vector into the communication network health assessment subnetwork to perform reverse impact calculation to obtain the communication network health quantification value. It performs bidirectional feedback iterative correction on the power supply risk status quantification value and the communication network health quantification value, and outputs the power supply risk status index and the communication network health index. Correction module: It is used to calculate the threshold floating benchmark based on the power supply risk status index and the communication network health index, introduce the real-time external event intensity factor to weight and correct the threshold floating benchmark, perform hierarchical boundary optimization and stability verification on the weighted and corrected threshold sequence, and establish a risk warning hierarchical threshold set. Coupling module: Used to receive the risk warning classification threshold set, map the risk warning classification threshold set to SMS channel encoding rules, voice channel encoding rules and private network data channel encoding rules, prioritize and time-synchronize the encoding of each channel, and generate a differentiated power supply risk communication warning instruction set.

6. The power supply risk communication early warning system according to claim 5, characterized in that, The acquisition module includes: Construction Unit: Used to collect real-time measurement values ​​from the power grid SCADA system, synchronous phasor time series data from the PMU device, gridded numerical forecast data from the meteorological Doppler radar, GIS coordinate point data from the emergency resource management platform, and performance monitoring byte streams from the communication OTN optical network, constructing a five-dimensional data cube structure with time dimension, spatial dimension, equipment dimension, index dimension and source dimension; Unified Unit: Used to perform sliding window anomaly detection and spatiotemporal kriging interpolation on a five-dimensional data cube. It uses three-point median filtering to remove abrupt outliers, spatiotemporal joint kriging interpolation to fill in missing data points, and unifies the time dimension to the UTC standard timestamp and the spatial dimension to the Gauss-Kruger projection coordinate system to obtain a data tensor with unified spatiotemporal reference. Mapping Calculation Unit: This unit transforms the data tensor after unifying the spatiotemporal reference into a dynamically weighted graph structure. It maps substation nodes, communication base station nodes, and emergency warehouse nodes as graph vertices, and power transmission lines, communication fiber optic links, and road networks as graph edges. It dynamically calculates edge weights based on line load rate and fiber optic bit error rate, and aggregates neighbor node features via a graph neural network message passing mechanism to form an initial power supply risk communication dataset. The edge weight calculation formula is as follows: In the formula, For nodes With nodes The weight of the connecting edges between them, The power transmission weighting coefficient ranges from 0.6 to 0.8, indicating the line... The current load power, Indicates the line Rated capacity, The communication quality weighting coefficient ranges from 0.2 to 0.4, representing the communication link. The bit error rate.

7. The power supply risk communication early warning system according to claim 5, characterized in that, The first computing module includes: Clustering Unit: Used to perform voltage hierarchy decomposition and load density clustering on the power grid measurement data in the initial power supply risk communication dataset. It constructs a high-voltage feature layer by forming the voltage phasors of 220kV and above nodes, a medium-voltage feature layer by forming the voltage and power of 110kV nodes, and a low-voltage feature layer by forming the load of 10kV feeder loads. It uses the density peak clustering algorithm to dynamically cluster the load nodes and form a three-level power grid state feature tensor. Fitting and solving unit: used to perform typhoon wind field model fitting and icing growth differential equation solving on meteorological grid data in the initial power supply risk communication dataset. It uses the Miyazaki-Takahashi typhoon wind field empirical formula to calculate the wind speed and direction of each node, uses the Jones icing growth model to calculate the conductor icing thickness, maps the calculation results to the power grid topology nodes, and generates a meteorological disaster risk field matrix. Input extraction unit: Used to input the three-level power grid state feature tensor and meteorological disaster risk field matrix into the emergency resource accessibility calculation unit. Combined with emergency resource GIS coordinates and real-time road traffic data, the Dijkstra shortest path algorithm is used to calculate the shortest time for emergency power generation vehicles and maintenance teams to reach each risk node, forming an emergency resource accessibility matrix. Simultaneously, wavelet packet decomposition and modulation domain feature extraction are performed on communication quality parameters to form a communication quality feature vector. The operation input unit is used to input the risk distribution matrix, emergency resource accessibility matrix, and communication quality feature vector into the cross-domain correlation mapping engine. It calculates the coupling strength coefficient between features from different domains via a heterogeneous graph attention mechanism, performs bilinear pooling on the coupled features, and generates the basic risk probability distribution matrix and communication quality confidence vector. The formula for calculating the cross-domain coupling strength coefficient is as follows: In the formula, Represents risk domain nodes With communication domain nodes The cross-domain coupling strength coefficient between them Represents the risk domain query vector. Represents the communication domain key vector. This represents the dimension of the feature vector, where N represents the total number of neighboring nodes. This represents the normalized denominator over all neighboring nodes.

8. The power supply risk communication early warning system according to claim 5, characterized in that, The second computing module includes: Aggregation Unit: Used to input the basic risk probability distribution matrix into the power supply risk propagation subnetwork, and perform forward inference through the graph attention propagation layer. Each layer of propagation aggregates the risk probabilities of neighboring nodes according to the edge weights, and introduces the node's own meteorological risk field intensity. The output is obtained by calculating the quantitative value of the power supply risk situation through the ReLU activation function. Calculation Unit: Used to input the communication quality confidence vector into the communication network health assessment sub-network, and calculate it through the health status back propagation mechanism. The health impact factor is passed layer by layer from the core backbone network node to the edge access network node. The fiber optic link attenuation and wireless channel interference coefficient are introduced at each layer to calculate the quantitative value of the communication network health. Feedback Correction Unit: This unit constructs a bidirectional coupled feedback loop between the quantified power supply risk status and the quantified communication network health. The quantified power supply risk status is used as the input to the communication network load stress, and the quantified communication network health is used as a risk propagation efficiency correction factor. It iterative correction is performed until the difference between two consecutive outputs is less than a set convergence tolerance, outputting the power supply risk status index and the communication network health index. The bidirectional coupled feedback correction formula is as follows: In the formula, This represents the power supply risk situation index vector for the (t+1)th iteration. This represents the vector of quantitative values ​​of the power supply risk situation in the t-th iteration. This represents the vector of communication network health quantization values ​​in the t-th iteration. This represents the communication network health index vector for the (t+1)th iteration. Represents the risk state weight matrix. This represents the weight matrix indicating the impact of communication. Represents the weight matrix of health status. This represents the communication quality confidence vector. Let represent the meteorological stress matrix, σ represent the Sigmoid activation function, and ⊙ represent the Hadamard product.

9. A power supply risk communication early warning device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the power supply risk communication early warning method as described in any one of claims 1 to 4 when executing the computer program.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, which, when executed by a processor, implements the power supply risk communication early warning method as described in any one of claims 1 to 4.

Images