A power communication network and power grid combined risk assessment and early warning method and system

Abstract

A method and system for joint risk assessment and early warning of power communication networks and power grids, belonging to the field of power system automation and smart grid technology, is disclosed. The method includes: constructing a bidirectional risk transmission path matrix based on the positive and negative bidirectional transmission relationship between power communication network faults and power grid faults; solving the scenario coupling coefficient using grey relational analysis combined with entropy weight correction; quantifying implicit coupling relationships and obtaining coupling probabilities using a hidden Markov model; building a graph neural network model, calibrating the coupling coefficients using historical fault data, and updating the model's observation sequence, attention weights, and key node features; and completing cross-network feature fusion through graph convolution and attention mechanisms, normalizing to obtain a dynamic coupling risk value. An adaptive adjustment mechanism is set based on the dynamic coupling risk value, a reward function is constructed to dynamically correct the assessment index weights, and early warning levels are divided according to the risk value to carry out risk early warning and decision-making, achieving accurate assessment and early warning of joint risks of power communication networks and power grids.

Description

Technical Field

[0001] This invention belongs to the field of power system automation and smart grid technology, specifically relating to a method and system for joint risk assessment and early warning of power communication networks and power grids. Background Technology

[0002] As power systems rapidly evolve towards intelligence and inter-regional interconnection, power communication networks have upgraded from traditional dispatch support networks to critical infrastructure carrying core services such as relay protection and automation control, forming a deep physical coupling and business dependency relationship with the power grid. Outages in the power communication network can lead to grid dispatch failures, while short circuits and power outages in the grid can directly affect the power supply stability of communication equipment; the propagation of these faults has a significant cascading effect.

[0003] Currently, risk assessments of power communication networks and power grids are independent and lack coordination, failing to fully consider their coupling characteristics. They are mainly divided into two categories: power communication network risk assessment technology and power grid risk assessment technology.

[0004] The existing power communication network risk assessment technology is mainly implemented through three methods: (1) Relying on SNMP (Simple Network Management Protocol) to collect the operating status data of communication equipment such as routers and switches, and to count indicators such as link delay, bandwidth utilization, and packet loss rate. When the indicators exceed the preset fixed threshold, a high-risk alarm is triggered. At the same time, combined with GIS (Geographic Information System), an electronic map of optical cable routes is constructed to mark the pipeline laying method, laying path and surrounding environment information of optical cables. Single-point failure risk is identified through topology connectivity analysis. For example, the only communication optical cable of a substation is marked as a vulnerable link when it passes through a demolition construction area. (2) Using FTA (Fault Tree Analysis) or Markov model, the failure probability of equipment is calculated based on the operating years of communication equipment and historical fault records. For example, the annual average failure rate of the communication screen of a 220kV substation is counted as 0.02 times / unit, and then node-level risk assessment is completed. Both of the above technologies only focus on the communication network's own indicators, such as link connectivity, without considering the impact on the power grid services it carries, such as whether a fiber optic cable break will cause relay protection signal transmission failure. They cannot assess the actual threat of communication network failures to the safe operation of the power grid. (3) They rely on fixed alarm thresholds, such as alarms with a delay > 50ms. They do not dynamically adjust the threshold parameters in conjunction with the real-time operating conditions of the power grid. In scenarios where the amount of communication data surges, such as large-scale grid connection of new energy sources, they are prone to false alarms or missed alarms. In addition, the above existing technologies do not adequately analyze the risks of external environments such as extreme weather and external construction, as well as cross-system coupling risks such as power grid outages leading to communication power interruptions. They are unable to predict communication cascading failures caused by fiber optic cables being washed away by floods or power outages at substations. These limitations make it difficult for existing technologies to support the prevention and control requirements of "communication-power grid" collaborative security in new power systems.

[0005] Existing power grid risk assessment technologies are also mainly implemented through three methods: (1) Relying on SCADA (Supervisory Control And Data Acquisition) systems to collect real-time power grid operation data, including electrical quantities such as voltage, current, and power, and combining power flow calculations to analyze the load status of transmission lines, transformers, and other equipment, and issuing overload risk warnings when the equipment load exceeds the rated capacity. (2) Based on the operating years, maintenance records, and insulation aging curves of power grid equipment, a reliability assessment model is used to calculate the probability of equipment failure, such as the forced shutdown rate of generators, in order to complete the equipment-level risk assessment. (3) Using transient stability analysis software to simulate the dynamic response of the power grid when subjected to large disturbances such as short-circuit faults and large load fluctuations, to determine whether the system will become unstable, and then assess the system-level safety risks. The above technologies only focus on the electrical parameters and equipment status of the power grid itself, without considering the impact of power communication network faults on the transmission of power grid dispatch instructions and relay protection signals, and cannot comprehensively assess cross-system cascading risks. Meanwhile, existing power grid risk assessment models are mostly based on historical data and fixed parameters, making it difficult to adapt to changes in power grid operation characteristics brought about by large-scale grid integration of new energy sources. In scenarios with a high proportion of new energy access, the accuracy of assessment results decreases significantly. In addition, the risk analysis of concentrated power grid equipment failures caused by extreme weather conditions such as typhoons and icing is not in-depth enough, and there is a lack of effective early warning mechanisms, making it difficult to cope with power grid security challenges under such circumstances. These limitations make it impossible for existing technologies to meet the assessment requirements for the safe and stable operation of new power systems. Summary of the Invention

[0006] The purpose of this invention is to address the problems in the prior art by providing a method and system for joint risk assessment and early warning of power communication networks and power grids. This method breaks down the information silos between power communication networks and power grids in risk assessment, constructs a two-way coupled risk matrix between the communication network and the power grid, quantifies the complex risk transmission paths between the communication network and the power grid, and introduces a dynamic adjustment mechanism to intelligently optimize assessment parameters based on the real-time operating conditions of the power grid, thereby achieving accurate and real-time assessment and early warning of joint risks of power communication networks and power grids.

[0007] To achieve the above objectives, the present invention provides the following technical solution: Firstly, a joint risk assessment and early warning method for power communication networks and power grids is provided, including: Based on the bidirectional transmission relationship between power communication network faults and power grid faults, a bidirectional risk transmission path matrix is ​​constructed; The scene coupling coefficient is calculated using grey relational analysis and entropy weight correction method; The implicit coupling relationship is quantified by observing the sequence using a hidden Markov model, and the implicit coupling probability is obtained. By using the scenario coupling coefficient and implicit coupling probability as constraints, and integrating the bidirectional risk transmission path matrix, a graph neural network model is constructed. The scenario coupling coefficient is calibrated using historical fault data, the implicit coupling probability is updated using real-time communication-power grid data as the observation sequence of the hidden Markov model, and the node features and edge weights of the graph neural network model are generated using topology-load data. The hidden Markov model observation sequence is updated using power grid short-circuit data to adjust the scene coupling coefficient; the attention weights of the graph neural network model are updated using communication node load data. Based on the graph neural network model after adjusting the scene coupling coefficient and updating the attention weight, graph convolution and attention mechanisms are used to complete the cross-network feature fusion between the power communication network and the power grid. The coupling features obtained by cross-network feature fusion are transformed into dynamic coupling risk values ​​through normalization calculation. Early warning levels are determined based on the magnitude of the dynamically coupled risk value, enabling risk warning and decision-making.

[0008] As a preferred embodiment, the joint risk assessment and early warning method for the power communication network and the power grid further includes a step of acquiring and preprocessing relevant data from the power communication network and the power grid. The relevant data from the power communication network includes real-time performance data of the power communication network, such as the bit error rate of the optical transmission link, the forwarding delay of routing equipment, and the service bandwidth utilization rate. The relevant data from the power grid includes synchronization phasor measurement data, switch status change information, power flow distribution data, and environmental parameters pushed by meteorological departments. The step of preprocessing the relevant data from the power communication network and the power grid includes: A density-based local outlier factor algorithm is used to identify outliers, and a sliding window mean filter is used to eliminate noise. For missing data caused by transmission interruption, a long short-term memory network is used for time-series prediction and completion. Based on the unified GPS timestamp and power grid topology node coding, communication links are spatially associated and mapped with corresponding transmission lines and substations to form a standardized dataset containing spatiotemporal labels.

[0009] As a preferred embodiment, in the step of constructing a bidirectional risk transmission path matrix based on the positive and negative bidirectional transmission relationship between power communication network faults and power grid faults, the expression of the bidirectional risk transmission path matrix is ​​as follows:

[0010] In the formula, For a forward path, it represents the first... Class of communication failure triggers the first The transmission relationship of power grid-like faults; For the reverse path, it represents the first... Grid-like faults trigger the first The transmission relationship of communication failures.

[0011] As a preferred embodiment, in the step of calculating the scene coupling coefficient using grey relational analysis and entropy weight correction method, the path scene attributes are marked according to the laying method. This provides a basis for assigning values ​​to the scene coupling coefficient; Calculate the grey relational coefficient using the following expression:

[0012] In the formula, Represents the communication network fault type number. Represents the type number of the power grid fault; , This indicates iterating through all communication fault numbers. With power grid fault number Find the global minimum and global maximum values ​​of the fault characteristic differences; For the corresponding scenario The difference in fault characteristics between communication and power grid. The resolution coefficient; Indicates the first In the first type of laying scenario, Class of communication failures and the first Grey relational coefficients directly related to grid-like faults; Scene coupling coefficient Calculate according to the following expression:

[0013] In the formula, The scene weights obtained by the entropy weight method are... This represents the total number of fault combinations. It is the sum of the grey relational coefficients for all fault combinations.

[0014] As a preferred embodiment, in the step of quantifying the implicit coupling relationship and obtaining the implicit coupling probability through the observation sequence of the Hidden Markov Model, the expression of the Hidden Markov Model observation sequence is: In the formula, It is a set of hidden states; For observation sequence; Here is the state transition matrix. , In the formula , They are respectively , The implicit coupling state at time t, , This corresponds to the number of the hidden state; For the observation probability matrix, , In the formula for Observation sequence samples collected in real time. For the specific data corresponding to the observed samples; Let be the initial state probability vector of the Hidden Markov Model, used to characterize the initial occurrence probability of each implicitly coupled condition at the initial moment of the system; iterative optimization is performed using the Expectation-Maximization (EM) algorithm. The objective function is In the formula, This represents the optimal model parameters that maximize the objective function within the range of all feasible model parameters. To determine the independent variable that makes the function reach its maximum value; Represents all possible model parameters Take the maximum value from the middle. This indicates that the EM algorithm is used for iterative optimization to select the optimal model parameters that best match the actual monitoring data, and to fit the implicit coupling correlation characteristics between the power communication network and the power grid, so that the implicit coupling identification accuracy is ≥90%.

[0015] As a preferred embodiment, the graph neural network model is constructed as follows: Establish node characteristics: Topological betweenness centrality of communication networks Importance of power grid load To perform fusion, the expression is: ; Establish edge weights: Using bidirectional coupling strength as edge weights The calculation expression is: In the formula, The scene coupling coefficient, The implicit coupling probability is output by the Markov model; To extract associated features using a graph convolutional layer, the calculation expression is as follows:

[0016] In the formula, This indicates the target node after graph convolutional layer aggregation, transformation, and normalization. Update the feature vector. () represents a nonlinear activation function. For the target node The adjacent nodes, For the target node The set of all adjacent nodes, Represents the target node The degree of the node is equal to that of the target node. The total number of connected adjacent nodes; Indicates adjacent nodes The degree of a node is equal to that of its neighboring nodes. The total number of connected adjacent nodes; Indicates adjacent nodes The original node feature vector, The graph neural network can learn the weight parameter matrix; An attention mechanism is established to strengthen key nodes, and the calculation expression is as follows:

[0017] In the formula, Indicates adjacent nodes Relative to the target node The original attention association score, It is a nonlinear activation function with leakage, used for nonlinear fitting of node association features. For the transpose of the learnable attention weight vector, This represents the adjacent nodes after the graph convolutional layer output. Update the feature vector; Indicated by the natural index After summing and normalizing the neighboring nodes, the final attention weight coefficients are obtained; Based on the deep extraction of node-coupled features using graph convolution and attention mechanisms, the irregular abstract feature values ​​are linearly mapped to standard risk scores of 0-100 through max-min normalization. The calculation expression is as follows:

[0018] In the formula, For the target node The quantified risk value obtained after normalization mapping is used to intuitively represent the magnitude of risk at a single node. After graph convolution feature extraction and attention mechanism weighting enhancement, the target node The final high-dimensional fused feature values ​​are output. For all target nodes in all power communication-grid coupling nodes The maximum value of the fused feature value For all target nodes in all communication-grid coupled nodes The minimum value of the fused eigenvalues; Using the above maximum-minimum normalization expressions, the node risk scores resulting from the transmission of faults from the power communication network to the power grid are calculated respectively. The node risk score generated by the propagation of grid faults to the power communication network The combined risk value of the power communication network and the power grid is calculated by linear superposition, and the expression is as follows: .

[0019] As a preferred embodiment, the expression for calibrating the scenario coupling coefficient using historical fault data is as follows:

[0020] In the formula, This is historical fault data. This represents the mapping calculation function based on improved grey relational analysis combined with the entropy weight method; The expression for updating the implicit coupling probability using real-time communication-power grid data as an observation sequence in a Hidden Markov Model is as follows:

[0021] In the formula, For real-time communication - power grid data, This represents the probabilistic inference operation function of a hidden Markov model. The steps for generating node features and edge weights of the graph neural network model using topology-load data are as follows: Utilizing topology-load data Calculate the topological betweenness centrality of communication nodes Importance of power grid node loads The node features of the graph neural network model are formed by concatenating vectors. Then, the explicit scenario coupling coefficient is determined based on the topological relationship between communication and power grid line laying. By combining the topological graph association data with the implicit coupling probability obtained through a Hidden Markov Model, the bidirectional coupling strength between nodes is obtained by multiplying the two probabilities, which is then used as the edge weights of the graph neural network. This completes the mapping of topology-load data to the node features and edge weights of the graph neural network model.

[0022] As a preferred embodiment, the joint risk assessment and early warning method for power communication network and power grid further includes implementing a dynamic parameter adaptive adjustment mechanism based on dynamically coupled risk value; and constructing a reward function based on minimizing risk prediction deviation, and dynamically correcting the weight coefficients of each assessment indicator based on the weight of each assessment indicator. In the step of implementing a dynamic parameter adaptive adjustment mechanism based on dynamic coupling risk value, key power grid operation indicators associated with the dynamic coupling risk value are extracted. When it is detected that the output of new energy sources in the coupled risk associated area fluctuates by more than 20% within a set time range, or the load mutation rate reaches 5% per minute, the current power grid operation condition is determined to be abnormal, and the parameter adjustment process is triggered. The key power grid operation indicators include the fluctuation range of new energy output, the load mutation rate, and the system frequency deviation.

[0023] As a preferred approach, in the step of classifying early warning levels based on the magnitude of dynamic coupling risk values ​​and conducting risk warning and decision-making, a dynamic coupling risk value greater than or equal to 80 and less than or equal to 100 is classified as a Level 1 warning, indicating the existence of system-level cascading failure risk; a dynamic coupling risk value greater than or equal to 60 and less than 80 is classified as a Level 2 warning, indicating the existence of regional risk; a dynamic coupling risk value greater than or equal to 30 and less than 60 is classified as a Level 3 warning, representing the existence of local equipment risk; a dynamic coupling risk value less than 30 indicates a normal state. For Level 1 and Level 2 warnings, early warning information is pushed to the scheduling terminal and operation and maintenance platform through a distributed message queue. The early warning information includes the risk location, impact range, and development trend. Simultaneously, a digital twin simulation engine is invoked to simulate the risk diffusion path based on real-time power grid topology and communication network status. While generating early warning information, based on a genetic algorithm, under the premise of satisfying power grid safety constraints, the optimal risk control strategy is searched, and the implementation cost and risk reduction rate of each risk control strategy are calculated to provide quantitative decision-making reference.

[0024] Secondly, a joint risk assessment and early warning system for power communication networks and power grids is provided, including: The two-way risk transmission path matrix construction module is used to construct a two-way risk transmission path matrix based on the positive and negative two-way transmission relationship between power communication network faults and power grid faults. The scene coupling coefficient calculation module is used to calculate the scene coupling coefficient using grey relational analysis and entropy weight correction methods. The implicit coupling probability acquisition module is used to quantify implicit coupling relationships through the observation sequence of the Hidden Markov Model and obtain the implicit coupling probability. The graph neural network model building module is used to construct a graph neural network model by using the scene coupling coefficient and implicit coupling probability as constraints and fusing the bidirectional risk transmission path matrix. The data layering and mapping module is used to calibrate the scenario coupling coefficient through historical fault data, update the implicit coupling probability by using real-time communication-power grid data as the observation sequence of the hidden Markov model, and generate the node features and edge weights of the graph neural network model using topology-load data. The data-driven parameter update module is used to update the observation sequence of the hidden Markov model using power grid short-circuit data and adjust the scene coupling coefficient; and to update the attention weights of the graph neural network model using communication node load data to strengthen the features of key nodes. The feature fusion and risk quantification module is used to complete the cross-network feature fusion between the power communication network and the power grid based on the graph neural network model after the scene coupling coefficient is adjusted and the attention weight is updated. It uses graph convolution and attention mechanism to convert the coupling features obtained by cross-network feature fusion into dynamic coupling risk values ​​through normalization calculation. The early warning and decision-making module is used to classify early warning levels based on the magnitude of dynamically coupled risk values, and to conduct risk early warning and decision-making.

[0025] As a preferred embodiment, the power communication network and power grid joint risk assessment and early warning system further includes an assessment model dynamic adjustment module, which is used to execute a dynamic parameter adaptive adjustment mechanism based on the dynamically coupled risk value; and to construct a reward function based on minimizing risk prediction deviation, and dynamically correct the weight coefficients of each assessment indicator based on the weight of each assessment indicator. When implementing a dynamic parameter adaptive adjustment mechanism based on dynamic coupling risk values, key power grid operation indicators associated with the dynamic coupling risk values ​​are extracted. When it is detected that the output of new energy sources in the coupled risk-associated area fluctuates by more than 20% within a set time range, or the load mutation rate reaches 5% per minute, the current power grid operation condition is determined to be abnormal, and the parameter adjustment process is triggered. The key power grid operation indicators include the fluctuation range of new energy output, the load mutation rate, and the system frequency deviation.

[0026] Thirdly, an electronic device is provided, including a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement the joint risk assessment and early warning method of the power communication network and the power grid.

[0027] Fourthly, a computer-readable storage medium is provided, the computer-readable storage medium storing at least one instruction, which, when executed by a processor, implements the joint risk assessment and early warning method of the power communication network and the power grid.

[0028] Compared with the prior art, the first aspect of the present invention has at least the following beneficial effects: This invention constructs a bidirectional risk transmission path matrix based on the positive and negative bidirectional transmission relationship between power communication network faults and power grid faults, thereby quantifying the correlation strength of fault transmission between the two and making up for the shortcomings of traditional methods that ignore cross-system bidirectional risk transmission. It adopts a Hidden Markov Model (HMM) and uses the operating state parameters of the power communication network and the power grid as the HMM observation sequence to achieve accurate quantification of implicit coupling relationships such as electromagnetic interference and ground potential difference. This solves the problem that traditional methods cannot capture such implicit correlations. The measured results show that it can improve the coupling risk identification rate by 60%, and solve the problem that traditional assessment methods are not comprehensive and accurate in identifying cross-system coupling risks. The initial scenario coupling coefficient is calculated using grey relational analysis and then corrected using entropy weight correction, achieving preliminary fusion and calibration of multi-source data. Historical fault data is used to calibrate the scenario coupling coefficient. Real-time communication-power grid data is used as the observation sequence of the Hidden Markov Model to update the implicit coupling probability. Topology-load data is used to generate node features and edge weights for the graph neural network model, completing hierarchical mapping and deep integration of multi-source data. Based on this, communication network performance data, power grid electrical quantities, and environmental parameters are transformed into a unified risk assessment dimension, effectively capturing faults such as latent damage to optical cables that are difficult to identify using traditional methods. Simultaneously, the multi-level algorithmic collaborative architecture of the Hidden Markov Model, graph neural network model, grey relational analysis, and entropy weight correction method ensures an outlier identification accuracy of ≥95% and a missing data completion error of ≤3%, providing high-quality and highly reliable data support for risk assessment and improving the credibility of the assessment results. Early warning levels are divided according to the magnitude of the dynamic coupling risk value for risk warning and decision-making, achieving the integration of fault propagation simulation and emergency repair decision optimization. This invention enables comprehensive quantitative assessment of cross-system risks, supports real-time calculations at a scale of at least 1,000 nodes and flexible configuration of at least 100 early warning scenarios, adapts to the development needs of intelligent and cross-regional interconnection of new power systems, and improves the overall level of intelligent risk prevention and control and safety redundancy capabilities of power systems, providing strong support for the stable operation of new power systems.

[0029] Furthermore, this invention uses a dynamic coupling risk value as a benchmark, adopts a reinforcement learning-driven dynamic parameter adaptive adjustment mechanism, constructs a reward function by minimizing risk prediction deviation, and dynamically corrects the weight coefficients of each evaluation indicator by adjusting the weight of each evaluation indicator. It can adapt to operating conditions such as new energy fluctuations and load mutations in real time, reducing the false alarm rate in complex scenarios from 30% to 8%, and significantly improving the environmental adaptability of the evaluation model.

[0030] Furthermore, in this invention, when conducting risk warning and decision-making, a dynamic coupling risk value greater than or equal to 80 and less than or equal to 100 is designated as a Level 1 warning, indicating the existence of system-level cascading failure risk; a dynamic coupling risk value greater than or equal to 60 and less than 80 is designated as a Level 2 warning, indicating the existence of regional risk; a dynamic coupling risk value greater than or equal to 30 and less than 60 is designated as a Level 3 warning, representing the existence of local equipment risk; and a dynamic coupling risk value less than 30 indicates a normal state. For Level 1 and Level 2 warnings, complete warning information is pushed through a distributed message queue, and a digital twin simulation engine is invoked to simulate the risk diffusion path based on real-time power grid topology and communication network status (such as the line tripping sequence that may be caused by the loss of protection signals after a communication optical cable interruption). Simultaneously, while generating warning information, based on a genetic algorithm, under the premise of satisfying power grid safety constraints such as power flow not exceeding limits and voltage qualification, the optimal risk control strategy (including communication network backup link switching, power grid load transfer, emergency power supply activation sequence, etc.) is automatically searched, and the implementation cost and risk reduction rate of each scheme are calculated, providing quantitative decision-making references. By combining digital twins with genetic algorithms, we can not only accurately predict fault propagation paths through Monte Carlo simulation with an error of ≤5%, but also generate optimized emergency repair decision schemes, improving fault response efficiency by 50%, significantly shortening fault handling time, and significantly enhancing the initiative and accuracy of power system risk prevention and control.

[0031] It is understood that the beneficial effects of the second to fourth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 Flowchart of the joint risk assessment and early warning method for power communication network and power grid according to an embodiment of the present invention; Figure 2 A schematic diagram of the structure of the joint risk assessment and early warning system of the power communication network and the power grid in an embodiment of the present invention. Detailed Implementation

[0034] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0035] Please see Figure 1 To address the shortcomings of existing power communication network risk assessment technologies, which are isolated from power grid operations, focusing only on their own links and equipment while ignoring the actual threat of communication failures to power grid security, and existing power grid risk assessment technologies that focus only on their own electrical parameters and equipment status without considering the impact of communication network failures on the transmission of power grid dispatch instructions and relay protection signals, and whose assessment models are difficult to adapt to power grid changes after large-scale grid integration of new energy sources, this invention proposes a joint risk assessment and early warning method for power communication networks and the power grid. This method breaks the current situation where risk assessments of communication networks and the power grid are independent, comprehensively integrating multi-source data, including real-time power grid operation data, communication network status data, and external environmental data, constructing a two-way risk transmission path matrix between the communication network and the power grid, and quantifying the complex risk transmission paths between them. Furthermore, a dynamic adjustment mechanism is introduced to intelligently optimize assessment parameters based on real-time power grid operating conditions, such as load fluctuations and new energy access status, achieving accurate and real-time assessment and early warning of joint risks between the power communication network and the power grid. This allows for early detection of cross-system risks and helps maintenance personnel formulate timely response strategies, comprehensively improving the safety and reliability of the power system and ensuring its stable operation. This embodiment mainly includes the following steps: S1. Construct a two-way risk transmission path matrix based on the positive and negative bidirectional transmission relationship between power communication network faults and power grid faults; S2. The scene coupling coefficient is calculated using grey relational analysis and entropy weight correction method; S3. Quantify the implicit coupling relationship by observing the sequence through a hidden Markov model and obtain the implicit coupling probability; S4. Using the scene coupling coefficient and implicit coupling probability as constraints, and integrating the bidirectional risk transmission path matrix, a graph neural network model is constructed. S5. The scenario coupling coefficient is calibrated by using historical fault data, the implicit coupling probability is updated by using real-time communication-power grid data as the observation sequence of the hidden Markov model, and the node features and edge weights of the graph neural network model are generated by using topology-load data. S6. Update the observation sequence of the Hidden Markov Model using power grid short-circuit data and adjust the scene coupling coefficient; update the attention weight of the graph neural network model using communication node load data to enhance the features of key nodes. S7. Based on the graph neural network model after adjusting the scene coupling coefficient and updating the attention weight, graph convolution and attention mechanisms are used to complete the cross-network feature fusion between the power communication network and the power grid. The coupling features obtained by cross-network feature fusion are transformed into dynamic coupling risk values ​​through normalization calculation. S8. Based on the dynamic coupling risk value, implement a dynamic parameter adaptive adjustment mechanism; construct a reward function by minimizing the risk prediction deviation, and dynamically adjust the weight coefficients of each evaluation indicator in the coupled model by taking the weights of each evaluation indicator as the adjustment object. S9. Recalculate the final dynamic coupling risk value based on the dynamically corrected weight coefficients, classify the warning level according to the magnitude of the final dynamic coupling risk value, and conduct risk warning and decision-making.

[0036] The present invention provides a joint risk assessment and early warning method for power communication networks and power grids. Based on the deep coupling characteristics of "communication-power grid", it achieves quantitative assessment and accurate early warning of cross-system risks through multi-dimensional data fusion and dynamic modeling.

[0037] Because there is a two-way risk transmission path between the power communication network and the power grid, a communication network link interruption may lead to the failure of power grid control signal transmission, and a power grid short circuit may affect the integrity of communication signals through electromagnetic interference. This invention constructs a two-way risk transmission path matrix to transform the causal relationship between communication failure modes and power grid failure consequences into quantifiable risk values ​​ranging from 0 to 100, achieving a mapping from a single failure to system risk. Simultaneously, this invention utilizes a reinforcement learning algorithm (DQN) to construct a parameter adaptive mechanism, enabling the evaluation model to dynamically adjust weight coefficients based on real-time power grid operating conditions such as renewable energy output fluctuations and load changes. For example, when wind power output fluctuations exceed 15%, the weight of the communication network's stability index for renewable energy control signal transmission is automatically increased, solving the adaptability problem of traditional static models in complex scenarios. Furthermore, a spatiotemporal alignment algorithm transforms performance data such as communication network link delay and bit error rate, power grid electrical quantities such as power angle and frequency, and environmental parameters such as wind speed and construction plans into unified-dimensional evaluation features. Combined with blockchain encryption technology, this ensures data credibility and provides comprehensive input for risk assessment.

[0038] In one possible implementation, the joint risk assessment and early warning method for power communication networks and power grids in this embodiment further includes a step of acquiring and preprocessing relevant data from the power communication network and power grid. The relevant data from the power communication network includes real-time performance data of the power communication network, such as the bit error rate of optical transmission links, forwarding delay of routing equipment, and service bandwidth utilization. The relevant data from the power grid includes synchronization phasor measurement data, switch status change information, power flow distribution data, and environmental parameters pushed by meteorological departments. The step of preprocessing the relevant data from the power communication network and power grid includes: A density-based local outlier factor algorithm is used to identify outliers (such as sudden anomalies where link latency suddenly jumps from 20ms to 500ms). High-frequency noise is eliminated by sliding window mean filtering. For missing data caused by transmission interruption, an improved Long Short-Term Memory (LSTM) network is used for time-series prediction and completion, with the completion error controlled within 5%.

[0039] Based on unified GPS timestamps and power grid topology node codes, communication links are spatially associated and mapped with corresponding transmission lines and substations (for example, marking the co-pole erection relationship between "220kV East Line 1" and "Communication Optical Cable Section A"), forming a standardized dataset containing spatiotemporal labels, providing a high-quality data foundation for subsequent evaluation.

[0040] In one possible implementation, a set of communication network faults is defined. (8 types of faults including link interruption and node failure) Power grid fault set (Six categories including line tripping and short circuit faults); The expression for constructing the two-way risk transmission path matrix is ​​as follows:

[0041] In the formula, For a forward path, it represents the first... Class of communication failure triggers the first The transmission relationship of power grid-like faults; For the reverse path, it represents the first... Grid-like faults trigger the first The transmission relationship of communication failures.

[0042] Furthermore, the path scene attributes are marked according to the laying method. This provides a basis for assigning values ​​to the scene coupling coefficient; Calculate the grey relational coefficient using the following expression:

[0043] In the formula, Represents the communication network fault type number. Represents the type number of the power grid fault; , This indicates iterating through all communication fault numbers. With power grid fault number Find the global minimum and global maximum values ​​of the fault characteristic differences; For the corresponding scenario The difference in fault characteristics between communication and power grid. In this embodiment, the resolution coefficient is used. ; Indicates the first In the first type of laying scenario, Class of communication failures and the first Grey relational coefficients for grid-like faults.

[0044] Scene coupling coefficient Calculate according to the following expression:

[0045] In the formula, The scene weights obtained by the entropy weight method are... This represents the total number of fault combinations. The sum of the grey relational coefficients for all fault combinations; Substitute the scenario and assign values: , , , This enables dynamic switching.

[0046] Furthermore, the expression for the observation sequence of the Hidden Markov Model is defined as follows: In the formula, This is a set of latent states, including four categories: electromagnetic interference, ground potential difference, etc. ; The observation sequence includes measurable data such as bit error rate and short-circuit current. ; Here is the state transition matrix. , In the formula , They are respectively , The implicit coupling state at time t, , This corresponds to the number of the hidden state; For the observation probability matrix, , In the formula for Observation sequence samples collected in real time. For the specific data corresponding to the observed samples; is the initial state probability vector of the hidden Markov model, used to characterize the initial occurrence probability of each implicit coupling condition at the initial moment of the system; Iterative optimization using the Expectation-Maximization (EM) algorithm The objective function is In the formula, This represents the optimal model parameters that maximize the objective function across all feasible model parameters. To find the independent variable that makes the function reach its maximum value, Represents all possible model parameters Take the maximum value from the middle. This indicates that the EM algorithm is used for iterative optimization to select the optimal model parameters that best match the actual monitoring data, and to fit the implicit coupling correlation characteristics between the power communication network and the power grid, so that the implicit coupling identification accuracy is ≥90%.

[0047] In one possible implementation, the graph neural network model is constructed as follows: Establish node characteristics: Topological betweenness centrality of communication networks Importance of power grid load To perform fusion, the expression is: ; Establish edge weights: Using bidirectional coupling strength as edge weights The calculation expression is: In the formula, The scene coupling coefficient, The implicit coupling probability is output by the Markov model; To extract associated features using a graph convolutional layer, the calculation expression is as follows:

[0048] In the formula, This indicates the target node after graph convolutional layer aggregation, transformation, and normalization. Update the feature vector. () represents a nonlinear activation function. For the target node The adjacent nodes, For the target node The set of all adjacent nodes, Represents the target node The degree of the node is equal to that of the target node. The total number of connected adjacent nodes; Indicates adjacent nodes The degree of a node is equal to that of its neighboring nodes. The total number of connected adjacent nodes; Indicates adjacent nodes The original node feature vector, The graph neural network can learn the weight parameter matrix; An attention mechanism is established to strengthen key nodes, and the calculation expression is as follows:

[0049] In the formula, Indicates adjacent nodes Relative to the target node The original attention association score, It is a nonlinear activation function with leakage, used for nonlinear fitting of node association features. For the transpose of the learnable attention weight vector, This represents the adjacent nodes after the graph convolutional layer output. Update the feature vector; Indicated by the natural index After summing and normalizing the neighboring nodes, the final attention weight coefficients are obtained; Based on the deep extraction of node coupling features using graph convolution and attention mechanisms, the irregular abstract feature values ​​are linearly mapped to standard risk scores of 0-100 through max-min normalization. This enables a comparable, graded, and predictable quantitative representation of the risk of each coupled node in the entire network, providing a unified quantitative basis for subsequent dynamic parameter adaptive adjustment and risk level early warning. The calculation expression is as follows:

[0050] In the formula, For the target node The quantified risk value obtained after normalization mapping is used to intuitively represent the magnitude of risk at a single node. After graph convolution feature extraction and attention mechanism weighting enhancement, the target node The final high-dimensional fused feature values ​​are output. For all target nodes in all power communication-grid coupling nodes The maximum value of the fused feature value For all target nodes in all communication-grid coupled nodes The minimum value of the fused eigenvalues; Using the above maximum-minimum normalization expressions, the node risk scores resulting from the transmission of faults from the power communication network to the power grid are calculated respectively. The node risk score generated by the propagation of grid faults to the power communication network The combined risk value of the power communication network and the power grid is calculated by linear superposition, and the expression is as follows:

[0051] In the formula, This represents the node risk score resulting from the transmission of faults in the power communication network to the power grid. It is the risk score caused by communication anomalies such as communication link interruption, signal distortion, and node overload, which lead to anomalies in power grid protection, dispatching, and business transmission. This represents the node risk score resulting from the transmission of power grid faults to the power communication network. It is the risk score caused by changes in power grid operating conditions such as short circuits, heavy loads, electromagnetic disturbances, and abnormal power supply, which interfere with communication transmission quality and equipment operation.

[0052] In one possible implementation, the expression for calibrating the scenario coupling coefficient using historical fault data is as follows:

[0053] In the formula, This is historical fault data. This represents the mapping calculation function based on improved grey relational analysis combined with the entropy weight method; The expression for updating the implicit coupling probability using real-time communication-power grid data as an observation sequence in a Hidden Markov Model is as follows:

[0054] In the formula, For real-time communication - power grid data, This represents the probabilistic inference operation function of a hidden Markov model. The steps for generating node features and edge weights of a graph neural network model using topology-load data are as follows: Utilizing topology-load data Calculate the topological betweenness centrality of communication nodes Importance of power grid node loads The node features of the graph neural network model are formed by concatenating vectors. Then, the explicit scenario coupling coefficient is determined based on the topological relationship between communication and power grid line laying. By combining the topological graph association data with the implicit coupling probability obtained through a Hidden Markov Model, the bidirectional coupling strength between nodes is obtained by multiplying the two probabilities, which is then used as the edge weights of the graph neural network. This completes the mapping of topology-load data to the node features and edge weights of the graph neural network model.

[0055] Furthermore, data-driven parameter updates are implemented. On one hand, real-time data triggers adaptive correction of model parameters, using grid short-circuit data to update the HMM observation sequence, thereby driving the scenario coupling coefficient. On the one hand, the weights of the GNN are increased by 0.2; on the other hand, the attention weights of the GNN are updated using the load data of the communication nodes, thereby strengthening the features of key nodes.

[0056] Furthermore, feature fusion and risk quantification are performed. Graph convolution and attention mechanisms are used to achieve cross-network feature fusion, and coupled features are transformed into scores of 0-100 using a normalization formula. This achieves closed-loop coupling of data, model, and risk value.

[0057] In one possible implementation, based on the combined risk value of the coupled output. The logic is based on a coupling risk benchmark, operating condition triggering, parameter adaptation, and threshold correction, and is strongly correlated with the results of previous data coupling. In the step of executing a dynamic parameter adaptive adjustment mechanism based on dynamic coupling risk values, key power grid operating indicators associated with the dynamic coupling risk values ​​are extracted. When it is detected that the output of new energy sources in the coupling risk associated area fluctuates by more than 20% within a set time range (10 minutes in this embodiment), or the load mutation rate reaches 5% per minute, the current power grid operating condition is determined to be abnormal, and the parameter adjustment process is triggered. The key power grid operating indicators include the fluctuation range of new energy output, the load mutation rate, and the system frequency deviation.

[0058] At this point, a deep reinforcement learning process is invoked, using the minimization of risk prediction bias as the reward function and the weights of the coupled model's indicators as the adjustment targets. Through continuous interaction between the agent and the environment, the weight coefficients of each indicator in the evaluation model are dynamically adjusted. For example, during periods of high new energy development, the weight of the communication network on the transmission delay of control commands is automatically increased from the usual 0.15 to 0.3. Simultaneously, the system is based on the coupled risk value... Based on the boundary conditions for the safe and stable operation of the power grid, such as the transient stability reserve coefficient and voltage qualification rate, the risk warning threshold is adjusted in real time. When the system is under heavy load, i.e. the load rate exceeds 85%, the warning threshold is automatically lowered by 20%, ensuring that the risk assessment results based on coupled data can sensitively capture potential risks.

[0059] When the risk value calculated by the model exceeds the dynamically adjusted warning threshold, the risk warning and decision-making stage begins. A dynamic coupling risk value greater than or equal to 80 and less than or equal to 100 indicates a Level 1 warning, suggesting a system-level cascading failure risk; a dynamic coupling risk value greater than or equal to 60 and less than 80 indicates a Level 2 warning, suggesting a regional risk; a dynamic coupling risk value greater than or equal to 30 and less than 60 indicates a Level 3 warning, representing a local equipment risk; a dynamic coupling risk value less than 30 indicates a normal state. For Level 1 and Level 2 warnings, warning information is pushed to the scheduling terminal and operation and maintenance platform through a distributed message queue. The warning information includes the risk location, impact range, and development trend. Simultaneously, a digital twin simulation engine is invoked, based on the real-time power grid... The system simulates risk propagation paths in the topology and communication network (such as the line tripping sequence that may be caused by the loss of protection signals after a communication fiber optic cable break). While generating early warning information, based on a genetic algorithm, it automatically searches for the optimal risk control strategy under the premise of meeting power grid safety constraints such as power flow not exceeding limits and voltage within acceptable ranges. This includes backup link switching schemes for the communication network, load transfer paths for the power grid, and the order of emergency power supply activation. It also calculates the implementation cost and risk reduction rate of each scheme, providing dispatchers with quantitative decision-making references. For example, scheme A can reduce risk by 60% but requires 15 minutes to implement, while scheme B reduces risk by 40% but only requires 5 minutes to implement. This helps dispatchers quickly select the execution scheme based on the actual situation.

[0060] At the model level, the method in this embodiment continuously optimizes model performance through a closed-loop feedback mechanism. After each early warning event is processed, the actual fault development process and the model prediction results are analyzed to extract new fault features. The model parameters are updated through an incremental learning algorithm, so that the accuracy of risk prediction continues to improve with the increase of system running time, gradually reaching a prediction accuracy of over 95%, forming a complete closed loop of data acquisition, model evaluation, early warning decision-making, and model optimization.

[0061] At the algorithmic level, this embodiment employs a multi-layered collaborative algorithm architecture. In the data preprocessing stage, isolated forest and Kalman filtering are integrated to ensure outlier identification accuracy ≥95% and missing data completion error ≤3%. In the risk assessment stage, a hybrid model combining matrix calculation and Bayesian networks is used. The LSTM network's prediction error for fault propagation probability is controlled within 5%, while the Bayesian network increases the dimension of coupling relationship identification by 40% through latent variable layer expansion. In the dynamic adjustment stage, a reward function mechanism based on reinforcement learning is introduced, aiming to minimize risk prediction error and achieving second-level updates of parameter weights. In the decision support stage, Monte Carlo simulation using digital twins and genetic algorithms are combined to improve the efficiency of emergency repair path planning by 50%.

[0062] At the system level, this embodiment constructs a three-layer software architecture of data platform, evaluation engine, and application terminal. The data platform integrates heterogeneous data interfaces, supports 12 types of data sources such as OTN (Optical Transport Network), SDH (Synchronous Digital Hierarchy), and PMU (Phasor Measurement Unit), and a spatiotemporal alignment engine, meeting the requirements of timestamp synchronization accuracy ≤1ms and blockchain encryption module to ensure data immutability. The evaluation engine includes a dynamic coupling matrix generator, supports real-time calculation at a scale of at least 1000 nodes, a parameter adaptive module, can respond to power grid operating condition changes with a delay ≤2s, and a risk level classifier, and adopts the fuzzy comprehensive evaluation method. The application terminal deploys a three-dimensional visualization component, supports dynamic risk evolution display, and an early warning rule engine, which can configure at least 100 early warning scenarios and a decision scheme generator, and outputs an execution list with priority ranking.

[0063] Please see Figure 2 Another embodiment of the present invention also proposes a joint risk assessment and early warning system for power communication networks and power grids, comprising: The two-way risk transmission path matrix construction module 201 is used to construct a two-way risk transmission path matrix based on the positive and negative two-way transmission relationship between power communication network faults and power grid faults. The scene coupling coefficient calculation module 202 is used to calculate the scene coupling coefficient using grey relational analysis and entropy weight correction method; The implicit coupling probability acquisition module 203 is used to quantify the implicit coupling relationship through the observation sequence of the hidden Markov model and obtain the implicit coupling probability. The graph neural network model building module 204 is used to construct a graph neural network model by using the scene coupling coefficient and implicit coupling probability as constraints and fusing the bidirectional risk transmission path matrix. The data layering mapping module 205 is used to calibrate the scenario coupling coefficient through historical fault data, update the implicit coupling probability by using real-time communication-power grid data as the observation sequence of the hidden Markov model, and generate the node features and edge weights of the graph neural network model using topology-load data. The data dynamic driving parameter update module 206 is used to update the observation sequence of the hidden Markov model using power grid short-circuit data and adjust the scene coupling coefficient; and to update the attention weight of the graph neural network model using communication node load data to strengthen the features of key nodes. The feature fusion and risk quantification module 207 is used to complete the cross-network feature fusion between the power communication network and the power grid based on the graph neural network model after the scene coupling coefficient is adjusted and the attention weight is updated. It uses graph convolution and attention mechanism to convert the coupling features obtained by cross-network feature fusion into dynamic coupling risk values ​​through normalization calculation. The evaluation model dynamic adjustment module 208 is used to execute a dynamic parameter adaptive adjustment mechanism based on the dynamic coupling risk value; construct a reward function by minimizing the risk prediction deviation; and dynamically correct the weight coefficients of each evaluation indicator in the coupled model by adjusting the weights of each indicator. The early warning and decision-making module 209 is used to recalculate the final dynamic coupling risk value based on the dynamically corrected weight coefficients, classify the early warning level according to the magnitude of the final dynamic coupling risk value, and carry out risk early warning and decision-making.

[0064] According to the technical solutions described in the above embodiments, on the one hand, this invention breaks through the cross-system assessment barrier by introducing a Bayesian network with latent variables to quantify implicit coupling relationships such as electromagnetic interference for the first time, solving the problem of the lack of assessment of bidirectional risk transmission between "communication and power grid" in traditional methods, and improving the coupling risk identification rate by 60%. On the other hand, this invention proposes a dynamic parameter adaptive mechanism, which realizes the real-time adaptation of the assessment model to dynamic operating conditions such as new energy fluctuations and load mutations based on reinforcement learning, reducing the false alarm rate in complex scenarios from 30% to 8%. On the other hand, this invention enhances the depth of multi-source data fusion by transforming communication network performance data, power grid electrical quantities, and environmental parameters into a unified risk assessment dimension through spatiotemporal alignment and blockchain encryption technology, making up for the shortcomings of traditional methods in assessing physical layer faults such as latent damage to optical cables. On the other hand, this invention upgrades the intelligent decision support capability by integrating digital twins and genetic algorithms to realize the integration of fault propagation simulation and emergency repair path optimization, improving the response speed by 50% compared with manual decision-making, and significantly enhancing the initiative and accuracy of power system risk prevention and control.

[0065] According to the technical solutions described in the above embodiments, this invention employs a collaborative algorithm combining local outlier identification, sliding window mean filtering for noise reduction, and LSTM time-series completion. Combined with GPS timestamps and power grid topology node encoding, it achieves standardized processing and spatial correlation mapping of multi-source data from power communication networks, power grids, and the environment, providing a high-quality data foundation for subsequent coupling assessment. Secondly, by improving grey relational analysis and entropy weighting methods to quantify coupling coefficients in different laying scenarios, it captures implicit coupling relationships using a hidden Markov model, and constructs a complete bidirectional risk transmission model by fusing topological features and attention mechanisms through graph neural networks. Risk value quantification is achieved through multi-step mathematical formula derivation. Simultaneously, this invention uses dynamic coupling risk values... Using the core benchmark, deep reinforcement learning is triggered by linked power grid operating data to achieve adaptive correction of model index weights and early warning thresholds. The linkage logic between coupling results and dynamic adjustments is clearly defined, solving the problem of poor adaptability in traditional static early warning systems. Finally, this invention integrates risk early warning classification, digital twin simulation, and improved genetic algorithm decision support, combined with incremental learning to achieve iterative optimization of model parameters, forming a complete closed loop of "data acquisition - coupled evaluation - dynamic adjustment - early warning decision - model update".

[0066] Another embodiment of the present invention also proposes an electronic device, including a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement the joint risk assessment and early warning method of the power communication network and the power grid.

[0067] Another embodiment of the present invention provides a computer-readable storage medium storing at least one instruction that, when executed by a processor, implements the joint risk assessment and early warning method for the power communication network and the power grid.

[0068] The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable storage medium can include any entity or device capable of carrying the computer program code, a medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory, a random access memory, an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals. For ease of explanation, the above content only shows the parts related to the embodiments of the present invention; for specific technical details not disclosed, please refer to the method section of the embodiments of the present invention. This computer-readable storage medium is non-transitory and can be stored in storage devices formed by various electronic devices, enabling the execution process described in the method of the embodiments of the present invention.

[0069] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0070] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0071] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0072] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A power communication network and power grid combined risk assessment and early warning method, characterized in that, include: Based on the bidirectional transmission relationship between power communication network faults and power grid faults, a bidirectional risk transmission path matrix is ​​constructed; The scene coupling coefficient is calculated using grey relational analysis and entropy weight correction method; The implicit coupling relationship is quantified by observing the sequence using a hidden Markov model, and the implicit coupling probability is obtained. By using the scenario coupling coefficient and implicit coupling probability as constraints, and integrating the bidirectional risk transmission path matrix, a graph neural network model is constructed. The scenario coupling coefficient is calibrated using historical fault data, the implicit coupling probability is updated using real-time communication-power grid data as the observation sequence of the hidden Markov model, and the node features and edge weights of the graph neural network model are generated using topology-load data. The hidden Markov model observation sequence is updated using power grid short-circuit data to adjust the scene coupling coefficient; the attention weights of the graph neural network model are updated using communication node load data. Based on the graph neural network model after adjusting the scene coupling coefficient and updating the attention weight, graph convolution and attention mechanisms are used to complete the cross-network feature fusion between the power communication network and the power grid. The coupling features obtained by cross-network feature fusion are transformed into dynamic coupling risk values ​​through normalization calculation. Early warning levels are determined based on the magnitude of the dynamically coupled risk value, enabling risk warning and decision-making.

2. The power communication network and power grid combined risk assessment and early warning method according to claim 1, characterized in that, It also includes steps for acquiring and preprocessing data related to the power communication network and the power grid. The power communication network data includes real-time performance data such as the bit error rate of optical transmission links, forwarding delay of routing equipment, and service bandwidth utilization. The power grid data includes synchronization phasor measurement data, switch status change information, power flow distribution data, and environmental parameters pushed by meteorological departments. The preprocessing steps for the power communication network and power grid data include: A density-based local outlier factor algorithm is used to identify outliers, and a sliding window mean filter is used to eliminate noise. For missing data caused by transmission interruption, a long short-term memory network is used for time-series prediction and completion. Based on the unified GPS timestamp and power grid topology node coding, communication links are spatially associated and mapped with corresponding transmission lines and substations to form a standardized dataset containing spatiotemporal labels.

3. The power communication network and power grid combined risk assessment and early warning method according to claim 1, characterized in that, In the step of constructing a bidirectional risk transmission path matrix based on the positive and negative bidirectional transmission relationship between power communication network faults and power grid faults, the expression of the bidirectional risk transmission path matrix is ​​as follows: In the formula, For a forward path, it represents the first... Class of communication failure triggers the first The transmission relationship of power grid-like faults; For the reverse path, it represents the first... Grid-like faults trigger the first The transmission relationship of communication failures.

4. The joint risk assessment and early warning method for power communication networks and power grids according to claim 1, characterized in that, In the step of calculating the scene coupling coefficient using grey relational analysis and entropy weight correction method, the path scene attributes are marked according to the laying method. This provides a basis for assigning values ​​to the scene coupling coefficient; Calculate the grey relation coefficient using the following expression: In the formula, Represents the communication network fault type number. Represents the type number of the power grid fault; , This indicates iterating through all communication fault numbers. With power grid fault number Find the global minimum and global maximum values ​​of the fault characteristic differences; For the corresponding scenario The difference in fault characteristics between communication and power grid. The resolution coefficient; Indicates the first In the first type of laying scenario, Class of communication failures and the first Grey relational coefficients directly related to grid-like faults; Scene coupling coefficient Calculate according to the following expression: In the formula, The scene weights obtained by the entropy weight method are... This represents the total number of fault combinations. It is the sum of the grey relational coefficients for all fault combinations.

5. The method for joint risk assessment and early warning of power communication network and power grid according to claim 4, characterized in that, In the step of quantifying implicit coupling relationships and obtaining implicit coupling probabilities through the observation sequence of the Hidden Markov Model, the expression for the Hidden Markov Model observation sequence is: In the formula, It is a set of hidden states; For observation sequence; Here is the state transition matrix. , In the formula , They are respectively , The implicit coupling state at time t, , This corresponds to the number of the hidden state; For the observation probability matrix, , In the formula for Observation sequence samples collected in real time. For the specific data corresponding to the observed samples; is the initial state probability vector of the hidden Markov model, used to characterize the initial occurrence probability of each implicit coupling condition at the initial moment of the system; Iterative optimization by the Expectation-Maximization (EM) algorithm , where the objective function is . In the formula,[[]] represents the optimal model parameters that maximize the objective function within all feasible model parameter ranges; is the independent variable that makes the function reach its maximum value; ]>[[]]represents taking the maximum value among all possible model parameters ; represents using the EM algorithm for iterative optimization to screen out the optimal model parameters with the highest matching degree to the actual monitoring data, fitting the implicit coupling correlation characteristics between the power communication network and the power grid, and making the accuracy of implicit coupling identification ≥ 90%.

6. The method for joint risk assessment and early warning of power communication network and power grid according to claim 5, characterized in that, The graph neural network model is constructed as follows: Establish node characteristics: Topological betweenness centrality of communication networks Importance of power grid load To perform fusion, the expression is: ; Establish edge weights: Using bidirectional coupling strength as edge weights The calculation expression is: In the formula, The scene coupling coefficient, The implicit coupling probability is output by the Markov model; To extract associated features using a graph convolutional layer, the calculation expression is as follows: In the formula, This indicates the target node after graph convolutional layer aggregation, transformation, and normalization. Update the feature vector. () represents a nonlinear activation function. For the target node The adjacent nodes, For the target node The set of all adjacent nodes, Represents the target node The degree of the node is equal to that of the target node. The total number of connected adjacent nodes; Indicates adjacent nodes The degree of a node is equal to that of its neighboring nodes. The total number of connected adjacent nodes; Indicates adjacent nodes The original node feature vector, The graph neural network can learn the weight parameter matrix; An attention mechanism is established to strengthen key nodes, and the calculation expression is as follows: In the formula, Indicates adjacent nodes Relative to the target node The original attention association score, It is a nonlinear activation function with leakage, used for nonlinear fitting of node association features. For the transpose of the learnable attention weight vector, This represents the adjacent nodes after the output of the graph convolutional layer. The updated feature vector; Indicated by the natural index After summing and normalizing the neighboring nodes, the final attention weight coefficients are obtained; Based on the deep extraction of node-coupled features using graph convolution and attention mechanisms, the irregular abstract feature values ​​are linearly mapped to standard risk scores of 0-100 through max-min normalization. The calculation expression is as follows: In the formula, For the target node The quantified risk value obtained after normalization mapping is used to intuitively represent the magnitude of risk at a single node. After graph convolution feature extraction and attention mechanism weighting enhancement, the target node The final high-dimensional fused feature values ​​are output. For all target nodes in all power communication-grid coupling nodes The maximum value of the fused feature values, For all target nodes in all communication-grid coupled nodes The minimum value of the fused eigenvalues; Using the above maximum-minimum normalization expressions, the node risk scores resulting from the transmission of faults from the power communication network to the power grid are calculated respectively. The node risk score generated by the propagation of grid faults to the power communication network The combined risk value of the power communication network and the power grid is calculated by linear superposition, and the expression is as follows: 。 7. The method for joint risk assessment and early warning of power communication network and power grid according to claim 6, characterized in that, The expression for calibrating the scenario coupling coefficient using historical fault data is as follows: In the formula, This is historical fault data. This represents the mapping calculation function based on improved grey relational analysis combined with the entropy weight method; The expression for updating the implicit coupling probability using real-time communication-power grid data as an observation sequence in a Hidden Markov Model is as follows: In the formula, For real-time communication - power grid data, This represents the probabilistic inference operation function of a Hidden Markov Model; The steps for generating node features and edge weights of the graph neural network model using topology-load data are as follows: Utilizing topology-load data Calculate the topological betweenness centrality of communication nodes Importance of power grid node loads The node features of the graph neural network model are formed by concatenating vectors. Then, the explicit scenario coupling coefficient is determined based on the topological relationship between communication and power grid line laying. By combining the topological graph association data with the implicit coupling probability obtained through a Hidden Markov Model, the bidirectional coupling strength between nodes is obtained by multiplying the two probabilities, which is then used as the edge weights of the graph neural network. This completes the mapping of topology-load data to the node features and edge weights of the graph neural network model.

8. The method for joint risk assessment and early warning of power communication network and power grid according to claim 1, characterized in that, It also includes a dynamic parameter adaptive adjustment mechanism based on the dynamic coupling risk value; and a reward function constructed by minimizing the risk prediction deviation, with the weight of each evaluation indicator as the adjustment object, and the weight coefficient of each evaluation indicator dynamically adjusted. In the step of implementing a dynamic parameter adaptive adjustment mechanism based on dynamic coupling risk value, key power grid operation indicators associated with the dynamic coupling risk value are extracted. When it is detected that the output of new energy sources in the coupled risk associated area fluctuates by more than 20% within a set time range, or the load mutation rate reaches 5% per minute, the current power grid operation condition is determined to be abnormal, and the parameter adjustment process is triggered. The key power grid operation indicators include the fluctuation range of new energy output, the load mutation rate, and the system frequency deviation.

9. The method for joint risk assessment and early warning of power communication network and power grid according to claim 1, characterized in that, In the step of classifying early warning levels based on the magnitude of dynamic coupling risk values ​​and making risk warnings and decisions, a dynamic coupling risk value greater than or equal to 80 and less than or equal to 100 is a Level 1 warning, indicating the existence of system-level cascading failure risk; a dynamic coupling risk value greater than or equal to 60 and less than 80 is a Level 2 warning, indicating the existence of regional risk; a dynamic coupling risk value greater than or equal to 30 and less than 60 is a Level 3 warning, representing the existence of local equipment risk; a dynamic coupling risk value less than 30 is a normal state. For Level 1 and Level 2 early warnings, early warning information is pushed to the scheduling terminal and operation and maintenance platform through a distributed message queue. The early warning information includes the risk location, scope of impact, and development trend. At the same time, a digital twin simulation engine is invoked to simulate the risk diffusion path based on the real-time power grid topology and communication network status. While generating early warning information, the system searches for the optimal risk control strategy based on a genetic algorithm, under the premise of meeting the power grid safety constraints, and calculates the implementation cost and risk reduction rate of each risk control strategy, providing a quantitative decision-making reference.

10. A joint risk assessment and early warning system for power communication networks and power grids, characterized in that, include: The two-way risk transmission path matrix construction module is used to construct a two-way risk transmission path matrix based on the positive and negative two-way transmission relationship between power communication network faults and power grid faults. The scene coupling coefficient calculation module is used to calculate the scene coupling coefficient using grey relational analysis and entropy weight correction methods. The implicit coupling probability acquisition module is used to quantify implicit coupling relationships through the observation sequence of the Hidden Markov Model and obtain the implicit coupling probability. The graph neural network model building module is used to construct a graph neural network model by using the scene coupling coefficient and implicit coupling probability as constraints and fusing the bidirectional risk transmission path matrix. The data layering and mapping module is used to calibrate the scenario coupling coefficient through historical fault data, update the implicit coupling probability by using real-time communication-power grid data as the observation sequence of the hidden Markov model, and generate the node features and edge weights of the graph neural network model using topology-load data. The data-driven parameter update module is used to update the observation sequence of the hidden Markov model using power grid short-circuit data and adjust the scene coupling coefficient; and to update the attention weights of the graph neural network model using communication node load data to strengthen the features of key nodes. The feature fusion and risk quantification module is used to complete the cross-network feature fusion between the power communication network and the power grid based on the graph neural network model after the scene coupling coefficient is adjusted and the attention weight is updated. It uses graph convolution and attention mechanism to convert the coupling features obtained by cross-network feature fusion into dynamic coupling risk values ​​through normalization calculation. The early warning and decision-making module is used to classify early warning levels based on the magnitude of dynamically coupled risk values, and to conduct risk early warning and decision-making.

11. The joint risk assessment and early warning system for power communication network and power grid according to claim 10, characterized in that, It also includes an evaluation model dynamic adjustment module, which is used to execute a dynamic parameter adaptive adjustment mechanism based on the dynamic coupling risk value; and to construct a reward function based on minimizing risk prediction deviation, and to dynamically adjust the weight coefficients of each evaluation indicator based on the weight of each evaluation indicator. When implementing a dynamic parameter adaptive adjustment mechanism based on dynamic coupling risk values, key power grid operation indicators associated with the dynamic coupling risk values ​​are extracted. When it is detected that the output of new energy sources in the coupled risk-associated area fluctuates by more than 20% within a set time range, or the load mutation rate reaches 5% per minute, the current power grid operation condition is determined to be abnormal, and the parameter adjustment process is triggered. The key power grid operation indicators include the fluctuation range of new energy output, the load mutation rate, and the system frequency deviation.

12. An electronic device, characterized in that, It includes a processor and a memory, the processor being used to execute a computer program stored in the memory to implement the joint risk assessment and early warning method for power communication networks and power grids as described in any one of claims 1 to 9.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one instruction, which, when executed by a processor, implements the joint risk assessment and early warning method for power communication networks and power grids as described in any one of claims 1 to 9.

Images