A method for power load monitoring and control based on digital twin

Abstract

The present application provides a power load monitoring and control method based on digital twin, including: evaluating the security and importance of the labeled electronic devices existing in each node and edge of the power grid to obtain the sensitivity weight values of each node and edge; analyzing the load changes in the power grid according to the monitoring data, and when it is identified that the load change is greater than a preset threshold, starting an electromagnetic wave propagation simulation module, where the electromagnetic wave propagation simulation module is used to simulate the electromagnetic wave propagation of the load change; if the interference area is identified, generating a risk value for the labeled electronic devices in the interference area according to the sensitivity weight values of each node and edge in the interference area, and when the risk value is greater than a preset threshold, generating an instruction to evacuate the labeled electronic devices or adjust the power load.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a power load monitoring and control method based on digital twins. Background Art

[0002] With the rapid development of social economy and the transformation of energy structure, smart grid has become the key to improving the flexibility, reliability and efficiency of power grid. Smart grid requires the realization of high informatization, automation and interaction in all aspects of power system. Among them, real-time monitoring, prediction and control of power grid operation status are one of its core elements. There are electromagnetic compatibility problems in power system. The influx of large current will generate strong electromagnetic interference in the power grid, and its propagation process is complex and changeable. When electromagnetic waves propagate in different media, their propagation characteristics are different and will be affected by environmental and weather factors. The topological structure of the power grid is intricate, and a large number of precision electronic equipment are distributed at different nodes and edges. They are sensitive to electromagnetic interference to different degrees. When electromagnetic waves propagate near sensitive equipment, they may have a serious impact on their performance, and even cause equipment damage, affecting the safe and stable operation of the power grid. Summary of the invention

[0003] The present invention provides a method for monitoring and controlling electric loads based on digital twins, which mainly includes:

[0004] Acquire historical load data of the power grid and monitor the power grid data, establish a digital twin model of the power grid based on the acquired historical load data and monitoring data, map the physical topology of the power grid to the digital twin model, and include attribute information of each node and edge corresponding to the power grid in the digital twin model;

[0005] Conduct safety and importance assessments on the tagged electronic devices at each node and edge in the power grid, and obtain sensitivity weight values ​​for each node and edge;

[0006] Analyzing the load change in the power grid according to the monitoring data, and when it is identified that the load change is greater than a preset threshold, starting an electromagnetic wave propagation simulation module, wherein the electromagnetic wave propagation simulation module is used to simulate the electromagnetic wave propagation of the load change;

[0007] Acquire the real-time operation data of the power grid, determine the range and size of the load change through a load state estimation algorithm based on Kalman filtering, acquire the power grid topology data and material properties in the digital twin model, and simulate the propagation process and spatial distribution of electromagnetic waves in the power grid based on the range and size of the load change, combined with the power grid topology data and the material properties, and using an electromagnetic field simulation engine;

[0008] In the process of simulating electromagnetic wave propagation, a time-frequency domain analysis method is used to calculate in real time the power load of the power grid and the energy density distribution data of the simulated electromagnetic waves at each node and edge, and an electromagnetic interference assessment model is constructed in combination with the sensitivity weight values ​​of each node and edge, and the interference area where the electromagnetic wave energy density exceeds the preset threshold is identified according to the electromagnetic interference assessment model;

[0009] If the interference area is identified, a risk value for the marked electronic device in the interference area is generated according to the sensitivity weight values ​​of each node and edge in the interference area, and when the risk value is greater than a preset threshold, an instruction to evacuate the marked electronic device or adjust the power load is generated;

[0010] According to the monitoring data, a high-load area where the power load is higher than a threshold is identified; if the high-load area exists for a duration greater than a time threshold and there are multiple tagged electronic devices within it, the tagged electronic devices are prioritized based on the sensitivity differences of the nodes and edges associated with the tagged electronic devices, and a corresponding interference handling strategy is generated based on the priority of the tagged electronic devices.

[0011] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0012] The present invention discloses a method for monitoring and controlling electric loads based on digital twins, which can immediately trigger electromagnetic wave propagation simulation, quickly identify potential abnormal situations, and improve the response speed and early warning accuracy of the power grid to abnormal load changes;

[0013] The safety and importance of electronic equipment at each node and edge in the power grid are evaluated, so that power grid management can focus on the most critical and vulnerable parts, and achieve refined and targeted risk management; the electromagnetic field simulation engine is used to simulate the propagation of electromagnetic waves in the power grid. This process can not only predict the impact range of load changes, but also deeply analyze the spatiotemporal characteristics of electromagnetic interference, providing a scientific basis for power grid operation; it can generate risk control instructions in a timely manner, effectively prevent or reduce the impact of electromagnetic interference on key electronic equipment, and protect the safety of power grid infrastructure.

[0014] In general, the present invention effectively identifies and prevents electromagnetic interference, improves the safety and stability of the power grid, and improves the operational flexibility and ability to respond to emergencies of the power grid; through real-time monitoring and early warning of electromagnetic interference, it significantly improves the safety of the power grid and the protection effect of electronic equipment; through the high simulation and real-time updating of the digital twin model, it ensures the efficiency and accuracy of power grid management. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a flow chart of a method for monitoring and controlling electric loads based on digital twins of the present invention.

[0016] Figure 2 It is a schematic diagram of a method for monitoring and controlling electric loads based on digital twins of the present invention.

[0017] Figure 3 This is another schematic diagram of a method for monitoring and controlling electric loads based on digital twins of the present invention. DETAILED DESCRIPTION

[0018] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] like Figure 1-3 In this embodiment, a method for monitoring and controlling electric load based on digital twins may specifically include:

[0020] S101. Acquire historical load data of the power grid and monitor the power grid data. Based on the acquired historical load data and monitoring data, establish a digital twin model of the power grid, map the physical topology of the power grid to the digital twin model, and include attribute information of each node and edge corresponding to the power grid in the digital twin model.

[0021] According to the physical topology of the power grid, the digital twin model is constructed, and the nodes and edges in the physical topology of the power grid are mapped to the digital twin model through a graph database method to obtain a virtual power grid model corresponding to the physical power grid. Using the mapping data of the nodes and edges, the attribute information of each node in the physical topology of the power grid is obtained, and associated with the corresponding node in the digital twin model, so as to achieve synchronization of the node attribute information with the digital twin model. Through the nodes mapped to the model, the attribute information of each edge in the physical topology of the power grid is obtained, and associated with the corresponding edge in the digital twin model, so as to achieve synchronization of the edge attribute information with the digital twin model. Through the synchronization of the node attribute information and the edge attribute information, the monitoring data is associated with the digital twin model to obtain real-time updates and displays in the digital twin model. Obtain the association results between the monitoring data and the digital twin model, analyze the historical load data, explore the association relationship between the historical load data and the power grid topology structure, map the historical load data to the corresponding nodes and edges in the digital twin model, and obtain the digital twin model with topology structure, parameter attributes, real-time status and historical data.

[0022] Exemplarily, the historical load data and real-time monitoring data of the power grid are obtained, and the data are preprocessed to ensure the accuracy and integrity of the data through data cleaning, data integration and other methods. The preprocessed data is analyzed, and the features related to the topology and attributes of the power grid are extracted through feature engineering technology to prepare data and features for the subsequent establishment of the digital twin model of the power grid. When constructing the digital twin model of the power grid, the attribute information of each node in the physical topology of the power grid is obtained from the nameplate information, equipment archives, monitoring system and other channels of the power grid equipment, such as the parameter information of equipment such as substations and transmission lines, and these attribute information are associated with the corresponding nodes in the digital twin model to enrich the node attributes of the digital twin model. At the same time, the attribute information of each edge in the physical topology of the power grid is obtained, such as the length, resistance, capacitance and reactance of the transmission line, and these attribute information are associated with the corresponding edges in the digital twin model to improve the edge attributes of the digital twin model so that the digital twin model can truly reflect the characteristics of the physical power grid. The real-time monitoring data of the power grid is associated with the digital twin model. Through data mapping technology, the unique identifiers such as the measuring point number and equipment ID are used to match the measuring point information in the real-time monitoring data with the corresponding nodes and edges in the digital twin model, so as to realize the real-time update and display of the real-time data in the digital twin model. The historical load data of the power grid is analyzed. Through data mining technologies such as cluster analysis and association rule mining, the correlation between the historical load data and the topological structure of the power grid is found, the distribution law of the load data under different topological structures is found, and the historical load data is mapped to the corresponding nodes and edges in the digital twin model to enrich the historical data dimension of the digital twin model. Based on the established digital twin model of the power grid, the real-time monitoring data and historical load data are comprehensively utilized, and the operation status of the power grid is modeled and predicted through machine learning algorithms such as support vector machines and random forests. The real-time and historical data of the digital twin model are used to optimize the power grid dispatch by solving the optimal power flow equation to improve the operation efficiency of the power grid; fault diagnosis is carried out by matching fault information with historical cases to achieve rapid positioning and processing, and improve the reliability of the power grid. Finally, the application of the digital twin model of the power grid in optimization scheduling, fault diagnosis and other aspects is realized, providing support for the safe and stable operation of the power grid. In the data preprocessing stage, the mean filter algorithm can be used to denoise the historical load data of the power grid. The sliding window size is set to 10, the data points in the window are arithmetic averaged, the value of the center point is replaced, the curve is smoothed, and the abnormal points are removed. At the same time, the 5σ principle is used to calculate the difference between the data point and the mean. If it exceeds 5 times the standard deviation, it is regarded as an abnormal value and replaced by the mean of the two normal points before and after. When extracting features, the statistical characteristics of the load data, such as mean, variance, peak-to-valley difference, etc., can be calculated, and the shape characteristics of the load curve, such as slope, curvature, etc., can be extracted. The characteristic coefficients at different frequency scales are extracted using wavelet transform to form a high-dimensional feature vector.When constructing the digital twin model, the Neo4j graph database is used to abstract power grid equipment such as substations, buses, switches, and lines as nodes, and the connection relationships between them as edges to form a network topology. The Cypher query language is used to realize the attribute mapping of nodes and edges, such as substation capacity, bus voltage level, line length and other parameters. When mapping real-time data, the device ID is matched with the node ID in the digital twin model, and the real-time current, voltage, power and other measurement point data collected by the SCADA system are stored in the attributes of the corresponding node with the timestamp as the index to realize the update of real-time data. When analyzing historical data, the K-means clustering algorithm is used to divide the load data into several categories according to similarity, analyze the distribution law of each category in the power grid topology, and find the influence of power supply area, equipment type and other factors on load characteristics. The April association rule mining algorithm is used to mine frequent patterns and association rules between load data and topology structure with support 0.05 and confidence 0.8 as thresholds. Finally, in the application of the digital twin model, the support vector regression algorithm is used to predict the power grid load, and external factors such as temperature, humidity, and holidays are selected as input features, and load data is used as output labels to train the model. In the optimization scheduling, the goal is to minimize line loss, and the node voltage and line power are used as constraints to build the optimal power flow model. The interior point method is used to solve it and obtain the optimal power grid operation state. In fault diagnosis, the decision tree algorithm is used, with the fault feature as the non-leaf node and the fault type as the leaf node, to train the classification model to achieve rapid identification and location of the fault.

[0023] S102: Perform security and importance assessment on the marked electronic devices at each node and edge in the power grid to obtain sensitivity weight values ​​of each node and edge.

[0024] The power grid topology data and the real-time operation data of the power grid are obtained, and the key nodes and key edges of the power grid are determined according to the power grid topology data; the reliability, data integrity, communication security, and power supply reliability of the marked electronic device are used as evaluation indicators to construct an evaluation indicator system; the importance of each indicator in the evaluation indicator system is compared pairwise by using the hierarchical analysis method, and a judgment matrix is ​​constructed, and the weight of each indicator is obtained by performing matrix operation on the judgment matrix; the weight of each indicator is weighted and summed with the actual evaluation value of the key node and the key edge, and a comprehensive evaluation model is constructed to obtain the comprehensive score of the security and importance of each key node and each key edge; according to the comprehensive score, the sensitivity level of the key node and the key edge is divided to obtain the key nodes and key edges with high sensitivity, medium sensitivity, and low sensitivity; the sensitivity level of the key node and the key edge is dynamically combined with the real-time operation data of the power grid, and when the load rate and failure rate of the key node or the key edge change, the sensitivity weight value of the key node and the key edge is adjusted accordingly to obtain the sensitivity weight value of each key node and key edge in the power grid.

[0025] For example, by analyzing the topological structure of the power grid, the key nodes and key edges in the power grid are determined. According to the type, quantity, importance and other factors of the marked electronic devices connected to the nodes and edges, a security and importance evaluation index system is constructed. The index system includes multiple specific index items such as equipment failure rate, data encryption strength, communication link redundancy, load importance level, and the calculation method of each index is given. The hierarchical analysis method is used to calculate the index weight. First, the index hierarchy is constructed and the judgment matrix is ​​designed. The matrix elements use a 1-9 scale to represent the relative importance of each pair of indicators. The maximum eigenvalue and the corresponding eigenvector of the judgment matrix are calculated by the eigenvalue method. The eigenvector is normalized to obtain the index weight, and a consistency test is performed to make the consistency rate <0.1. Considering the nature of the evaluation indicators and the data distribution, a suitable comprehensive evaluation model is selected, and the weighted geometric mean method is used to calculate the comprehensive score of the security and importance of the nodes and edges. The higher the score, the higher the sensitivity of the node or edge. For the situation where the data distribution of some indicators is uneven, other evaluation models such as TOPS IS and entropy weight method can be considered. The comprehensive evaluation results are analyzed, and the sensitivity levels are divided according to the percentile of the comprehensive score. The top 10% of the highest scores are highly sensitive nodes or edges, 10%-30% are medium sensitive, and more than 30% are low sensitive. Differentiated safety protection measures and control strategies are formulated according to the sensitivity level, and high-sensitive nodes or edges are given priority protection. Real-time monitoring of the power grid operation status, acquisition of key operating indicator data such as load rate and failure rate of nodes and edges, setting reasonable thresholds, and triggering the dynamic adjustment mechanism of sensitivity weights when the actual value exceeds the threshold. Fuzzy control theory is used to calculate the fuzzy membership of weight adjustment according to the degree of over-limit of operating indicators, generate weight adjustment amount, and realize dynamic optimization of sensitivity weights. In the power grid dispatching and control system, sensitivity weights are introduced into power flow calculation, safety verification, fault analysis and other links as quantitative indicators of node and edge importance. In power flow calculation, the voltage amplitude and phase angle deviation of high-weight nodes are set to be smaller than that of low-weight nodes. In safety verification, the transmission power limit of high-weight nodes and edges is preferentially met. In fault analysis, the scope and degree of fault impact of high-weight nodes or edges are evaluated. In the planning and construction of power grids, sensitivity weight is used as one of the evaluation indicators and incorporated into the decision-making process of power grid development planning, site selection of new projects, and technical transformation. For weak nodes or edges with high weights, priority is given to upgrading and transformation to improve power supply reliability. When selecting sites for new projects, priority is given to high-weight load areas. When formulating power grid development plans, focus on analyzing the "bottleneck" risks of high-weight nodes and edges, and formulate targeted preventive measures.When constructing the safety and importance evaluation index system, the Delphi method can be used to consult 10 experts in the power grid field, and each expert is required to score the preliminary index items from 0 to 5 points, and add or delete index items according to the expert's suggestions. After three rounds of consultation, 15 index items are finally determined, including equipment failure rate, data encryption strength, communication link redundancy, load importance level, etc. For the equipment failure rate index, the historical failure data of the past three years can be counted to calculate the average failure rate λ=0.02 times / year. When calculating the index weights by the hierarchical analysis method, 5 experts can be invited to fill in the judgment matrix, using a 1-9 scale. If the ratio of the importance of the equipment failure rate to the load importance level is 5:1, the judgment matrix element is 5. The maximum eigenvalue λ_max=5.12 of the judgment matrix is ​​calculated by MATLAB software, corresponding to the eigenvector W=[0.48,0.24,0.16,0.08,0.04]. The weight vector is obtained after normalization, and the consistency ratio CR=0.03<0.1, passing the consistency test. When constructing a comprehensive evaluation model, considering that some qualitative indicators are difficult to quantify directly, the fuzzy comprehensive evaluation method can be used to set 5 comment levels, let V = {v1, v2, v3, v4, v5} = {high, relatively high, medium, relatively low, low}, and map the qualitative indicators to the 0-1 interval through the membership function, and then substitute them into the weighted geometric mean model together with the quantitative indicators to calculate the comprehensive score. When dividing the sensitivity level, SPSS software can be used to perform statistical analysis on the comprehensive score to obtain the percentile of the score. With P10 = 0.85 and P30 = 0.70 as the threshold, the sensitivity is divided into three levels: high, medium, and low. When dynamically adjusting the sensitivity weight, based on fuzzy control theory, we can set the load rate over-limit fuzzy set A = {low, medium, high} = {0.2, 0.5, 0.8}, and the weight adjustment fuzzy set B = {small, medium, large} = {0.1, 0.3, 0.5}, and establish a fuzzy control rule base between the over-limit degree and the weight adjustment amount, such as "if the load rate over-limit degree is medium, the weight adjustment amount is medium", and calculate the weight adjustment amount δw = 0.3 through fuzzy reasoning and defuzzification. Introducing sensitivity weights in power flow calculations, the voltage amplitude deviation of high-weight nodes can be set to no more than ±3% of the rated value in the node voltage constraint, and the voltage amplitude deviation of low-weight nodes can be set to no more than ±5% of the rated value. When selecting a site for a new project, the sum of the sensitivity weights within 5km around each candidate site can be calculated, and the site with the highest sum of weights can be selected as the construction site.

[0026] S103, analyzing the load change in the power grid according to the monitoring data, and when it is identified that the load change is greater than a preset threshold, starting an electromagnetic wave propagation simulation module, wherein the electromagnetic wave propagation simulation module is used to simulate the electromagnetic wave propagation of the load change.

[0027] The load parameters are obtained by real-time monitoring of each load node in the power grid, and the load parameters include voltage, current and power; a sliding window algorithm is used to calculate the change rate of the load parameters; when the change rate is greater than a preset threshold, it is determined that a sudden load has occurred, and the electromagnetic wave propagation simulation module is started; a propagation model of electromagnetic waves in the power grid is constructed according to the topological structure and conductor parameters of the power grid; a time-domain finite difference method is used to numerically simulate the electromagnetic wave propagation process, and the propagation model combines the influence of the resistance, inductance and capacitance of the conductor on the electromagnetic wave propagation characteristics, and the influence of the power grid topology on the electromagnetic wave propagation path; in the propagation model, the influencing factors of the power grid environment are introduced, and the influencing factors include temperature, humidity and wind speed, and the propagation model parameters are dynamically adjusted using the monitoring data; at the same time, combined with the electromagnetic interference source in the power grid, the electromagnetic interference power supply is used as the boundary condition of the electromagnetic wave propagation model; through the simulation calculation of the propagation model, the characteristic parameters of the electromagnetic wave propagation in the power grid are obtained, and the Characteristic parameters include time delay, attenuation loss and velocity distribution; the location and propagation range of the load change are calculated using the characteristic parameters; the result of the electromagnetic wave propagation simulation is compared with the actual monitoring data, and a mapping relationship between the load change and the electromagnetic wave propagation characteristics is established using a machine learning algorithm to achieve load change positioning and prediction based on the electromagnetic wave propagation simulation; during the electromagnetic wave propagation simulation, a parallel computing method is used to divide the power grid topology into multiple sub-areas, and an independent electromagnetic wave propagation simulation is performed in each sub-area; information interaction and synchronization between the sub-areas are achieved through the electromagnetic field continuity condition on the regional boundary; the electromagnetic wave propagation simulation module is integrated into the power grid monitoring system; when it is monitored that the load change is greater than a threshold, the electromagnetic wave propagation simulation is triggered to generate the load change positioning result, and the location and influence range of the load change are displayed on the power grid topology through a visualization method, and the power grid is dispatched and fault processing is performed according to the location and influence range of the load change.

[0028] For example, the sliding window algorithm is used to calculate the rate of change of the load parameters, the window size is set to 10 sampling points, the data is normalized, and the preset threshold of the rate of change is 0.2; according to the topological structure of the power grid and the parameters of the conductor, the propagation model of electromagnetic waves in the power grid is constructed, and the finite difference time domain method is used to numerically simulate the electromagnetic wave propagation process. The finite difference time domain method discretizes the Maxwell I equation into a difference equation, uses the central difference format to differentiate time and space, and iteratively calculates the evolution process of the electromagnetic field. The model considers the influence of the resistance, inductance, capacitance and other parameters of the conductor on the electromagnetic wave propagation characteristics, as well as the influence of the topological structure of the power grid on the propagation path of the electromagnetic wave. When constructing the electromagnetic wave propagation model, the idea of ​​parallel computing and regional decomposition is introduced, the topological structure of the power grid is divided into multiple sub-regions, and the Jacobi iterative algorithm is used in each sub-region for parallel computing. Through the electromagnetic field continuity condition on the regional boundary, information interaction and synchronization between sub-regions are realized, and the computational efficiency of electromagnetic wave propagation simulation is improved. In the electromagnetic wave propagation model, the influencing factors of the power grid environment, such as temperature, humidity, wind speed, etc., are introduced. The environmental monitoring data is quantified into the medium parameters and boundary conditions of the electromagnetic wave propagation model by using interpolation mapping and vector decomposition methods, and the model parameters are dynamically adjusted to improve the accuracy of the electromagnetic wave propagation simulation. At the same time, the electromagnetic interference sources in the power grid, such as harmonics and transient interference generated by equipment such as transformers and motors, are considered as the excitation source of the electromagnetic wave propagation model. Through the simulation calculation of the electromagnetic wave propagation model, the characteristic parameters such as time delay, attenuation loss, and velocity distribution of electromagnetic waves propagating in the power grid are obtained. The distance between the load change and the observation point is estimated according to the time delay and propagation speed, and the propagation path of the electromagnetic wave is estimated according to the attenuation loss, and then the location and impact range of the load change are deduced. The results of the electromagnetic wave propagation simulation are compared with the actual monitoring data, and the voltage, current waveform and spectrum characteristics when the load changes, as well as the time delay, attenuation and other characteristics of electromagnetic wave propagation are extracted. The feature vector is constructed, and the load change positioning model is trained using the support vector machine algorithm. The cross-validation method is used to optimize the model hyperparameters, improve the generalization performance of the model, and realize the load change positioning and prediction based on the electromagnetic wave propagation characteristics. The electromagnetic wave propagation simulation module is integrated into the power grid monitoring system. When the load change exceeds the threshold, the electromagnetic wave propagation simulation is automatically triggered to generate the load change positioning result. The spatial distribution of the electromagnetic field intensity is displayed through cloud maps, and the propagation process of the load change is demonstrated through animation. The visualization results are integrated with the power grid topology information to intuitively present the location and impact range of the load change, providing auxiliary decision support for power grid dispatching and fault handling.When real-time monitoring of power grid load is performed, PMU and other equipment can be used to obtain load voltage and current data at a sampling frequency of 100Hz, and the collected data can be preprocessed, such as removing invalid data points and filtering outliers, and then a sliding window with a length of 10 sampling points is used to calculate the change rate of the load active power in the window. When the change rate exceeds 0.2, it is determined to be a sudden load and the electromagnetic wave propagation simulation is started. When constructing an electromagnetic wave propagation model, the FDTD algorithm can be used to discretize the power grid in three-dimensional space, dividing the power grid into 100100100 grids, each with a size of 1m. According to the topological structure of the power grid and the material parameters of the conductor, the electromagnetic parameters such as the dielectric constant and permeability of each grid are calculated, and the Maxwell equation is discretized by difference using the Yee grid to form the FDTD iterative calculation formula, with a calculation time step of 1ns and a spatial step of 1m. When introducing parallel computing, the MPI parallel library can be used to divide the power grid into 10 sub-areas, each of which is assigned to a CPU core for FDTD calculation. Tang boundary conditions are used on the regional boundaries to handle the continuity of the electromagnetic field, realize data communication and synchronization between sub-areas, and dynamically adjust the size and computing tasks of the sub-areas through the load balancing algorithm to improve parallel efficiency. When considering environmental impact, the power grid geographic information system GIS can be used to obtain environmental parameters such as temperature, humidity, and wind speed in the area covered by the power grid. The environmental parameters are mapped to the FDTD grid through the IDW interpolation algorithm, and the conductivity of the medium is adjusted according to the temperature and humidity. The anisotropy of the magnetic permeability of the medium is adjusted according to the wind speed and wind direction to achieve the coupling of electromagnetic wave propagation and environmental conditions. When analyzing the propagation characteristics of electromagnetic waves, the electric field intensity data in the FDTD simulation results can be extracted, and the electric field intensity at different times and locations can be statistically analyzed. The time decay constant and spatial decay index of the electric field energy are calculated, and the propagation distance and direction of the electromagnetic wave are estimated. Combined with the topological structure of the power grid, the location and impact range of the load change can be inferred. When training the load change positioning model, 100 typical load change cases, corresponding voltage, current waveforms and electromagnetic wave propagation characteristics can be selected as training samples. The 10-fold cross validation method is used to optimize the hyperparameters of the SVM model such as C and g. The grid search algorithm is used for parameter optimization. The model performance evaluation index is the root mean square of the positioning error. When the root mean square error is less than 10%, it is considered that the positioning model meets the practical requirements. Finally, in the power grid monitoring system, when a sudden load occurs, the electromagnetic wave propagation simulation and load change positioning are triggered. The simulation calculation results are presented in the form of 3D visualization, with different colors representing the electromagnetic field distribution at different times, and the propagation process of the electromagnetic wave is displayed in the form of animation. The load change positioning results are associated with functional modules such as flow calculation and state estimation, providing intuitive and comprehensive auxiliary decision-making information for power grid dispatchers.

[0029] S104. Acquire the real-time operation data of the power grid, determine the range and size of the load change through a load state estimation algorithm based on Kalman filtering, acquire the power grid topology data and material properties in the digital twin model, and simulate the propagation process and spatial distribution of electromagnetic waves in the power grid based on the range and size of the load change, combined with the power grid topology data and the material properties, and using an electromagnetic field simulation engine.

[0030] When the load change amplitude is greater than the preset load threshold, or the change range is greater than the preset range threshold node, the electromagnetic wave propagation simulation task is triggered; the power grid topology data includes the connection relationship and geometric parameters between the bus, line and transformer; the grid model required by the electromagnetic field simulation engine is generated according to the power grid topology data, and an unstructured tetrahedral grid is used. The size of the unstructured tetrahedral grid is adaptively adjusted according to the wavelength, and the unstructured tetrahedral grid meets the numerical discrete accuracy requirements; the attribute parameters of the power grid equipment and materials in the digital twin model are obtained, and the attribute parameters include the resistivity and dielectric constant of the conductor, and the insulation material The magnetic permeability and dielectric loss factor of the load are mapped to the material property library of the electromagnetic field simulation engine to construct a simulation model of multi-physical field coupling; according to the location and range of the load change, an excitation source for electromagnetic wave propagation is set in the electromagnetic field simulation engine, and an ideal current source or voltage source is used as a drive, and the excitation source generates an excitation signal; the frequency, amplitude and phase of the excitation signal are generated according to the actual load change data, and the time domain-frequency domain conversion is realized through Fourier transform; the electromagnetic wave propagation process is simulated in the time domain-frequency domain in the electromagnetic field simulation engine, and the Maxwell finite difference method (FDTD) is used to solve the Maxwell finite difference method (FDTD) problem. ll equation group, calculate the evolution of the electromagnetic field in space and time; the time step is adaptively adjusted according to the CFL stability condition, the space step meets the grid accuracy requirement, and the boundary condition adopts the PML absorption boundary to avoid false reflection; extract the result data of the electromagnetic field simulation, including the time series and spatial distribution of the electric field intensity, magnetic induction intensity and current density, and generate the electromagnetic field cloud map, isosurface map and streamline map through visualization technology; analyze the propagation characteristics of the electromagnetic wave in the power grid, and the propagation characteristics include propagation speed, attenuation law and standing wave distribution; associate the electromagnetic field simulation results with the power grid topology and equipment parameters in the digital twin model to realize the mapping of the electromagnetic wave propagation process and the physical structure of the power grid; establish the mapping relationship between the electromagnetic field distribution and the load change through the convolutional neural network, and realize the positioning and prediction of the load change.

[0031] For example, the real-time operation data of the power grid is obtained, and the range and size of the load change are determined by the load state estimation algorithm based on Kalman filtering. The Kalman filter uses the dynamic model of the load as the state equation, and the measured voltage and current data as the observation equation. The covariance matrix of the process noise and the measurement noise is set according to the statistical characteristics of the load. When the load change amplitude estimated by the filter exceeds 5% of the rated load, or the change range is greater than 10 nodes, the electromagnetic wave propagation simulation task is triggered. The threshold is determined according to the safety margin and sensitivity analysis of the power grid. The physical topological structure data of the power grid is extracted from the digital twin model, and the Del aunay triangulation algorithm is used to generate an unstructured tetrahedral mesh model. According to the frequency characteristics of the electromagnetic wave and the electromagnetic parameters of the propagation medium, the CFL stability criterion is used to adaptively adjust the mesh size, so as to improve the simulation efficiency while ensuring the calculation accuracy. The attribute parameters of the power grid equipment and materials in the digital twin model are obtained, and these parameters are mapped to the material attribute library of the electromagnetic field simulation engine. The multi-physical field coupling method is used to consider the mutual influence of various physical factors such as electric field, magnetic field, and thermal field to build a high-fidelity simulation model. According to the location and range of load change, the excitation source of electromagnetic wave propagation is set in the electromagnetic field simulation engine. According to the impedance characteristics of load change, the ideal current source or voltage source is selected as the excitation source. The Fourier transform results of load change data are used to extract the amplitude, phase, frequency and other parameters of load current or voltage, and construct an excitation signal consistent with the actual load change. In the electromagnetic field simulation engine, the electromagnetic wave propagation process is simulated in the time domain and frequency domain. The Maxwell equations are discretized in time and space by using the finite difference time domain method (FDTD). The electric field and magnetic field components are staggered by using the Yee grid. The distribution of the electromagnetic field at each node in space is calculated by iteration. The time step is adaptively adjusted according to the CFL numerical stability condition. The space step meets the grid accuracy requirements. The boundary condition uses the perfect matching layer (PML) absorption boundary treatment to avoid false reflection of electromagnetic waves. The result data of electromagnetic field simulation are extracted, including the time series and spatial distribution of physical quantities such as electric field intensity, magnetic induction intensity, and current density. Visualization algorithms such as streamline tracing and isosurface extraction are used, and visualization tools such as ParaView and VTK are used to generate intuitive graphics such as electromagnetic field cloud maps, isosurface maps, and streamline maps. The propagation characteristics of electromagnetic waves in the power grid, such as field strength distribution, energy attenuation law, standing wave distribution, etc., are analyzed, and the impact of electromagnetic waves on power grid equipment and materials is evaluated.The electromagnetic field simulation results are associated with the grid topology, equipment parameters and other data in the digital twin model, and key feature quantities such as electromagnetic field energy distribution and spectrum characteristics are extracted. The mapping relationship dataset between electromagnetic field distribution and load change position and size is constructed. A three-dimensional convolutional neural network (3DCNN) model is used, with electromagnetic field distribution as input and load change position and size as output. Through end-to-end deep learning, the mapping relationship between electromagnetic field distribution and load change is established. The network adopts classic structures such as VGG16, the loss function adopts the weighted sum of mean square error and cross entropy, and the optimization algorithm adopts Adam. The generalization performance of the model is improved through transfer learning and data enhancement. The trained model is deployed in the grid dispatching control system to realize real-time positioning and prediction of load changes, providing a basis for intelligent dispatching decisions. When obtaining real-time operation data of the power grid, the power quality monitoring devices and micro PMUs deployed at key nodes such as substations and feeder terminals can measure parameters such as voltage, current, and power in real time at a sampling frequency of up to 1000 times per second, and upload them to the data center in real time through the optical fiber communication network. When applying the Kalman filter algorithm to estimate the load state, the day-ahead forecast value of the load can be selected as the initial state value, and the time-varying Auto-Regressive Moving Average (ARMA) model of the load can be constructed as the state transition equation based on the typical daily load curve and influencing factors such as temperature and humidity. When extracting the physical topological structure data of the power grid, the multi-source heterogeneous data such as the equipment nameplate parameters, GIS geographic coordinates, and three-dimensional laser scanning point cloud in the power grid holographic database can be used to automatically generate the topological elements such as nodes, edges, and faces of the power grid through spatial topological relationship extraction algorithms, such as the minimum spanning tree algorithm based on graph theory. When generating a tetrahedral mesh model, the improved Bowyer-Watson algorithm can be used. By introducing the weighted Voronoi graph based on the De Launay triangulation, the adaptive encryption of power grid equipment and scenes can be realized, and the spatial indexing methods such as octree can be combined to improve the efficiency of mesh generation. When setting the electromagnetic wave excitation source, the first three frequency components that account for more than 90% can be extracted based on the power spectrum density analysis results of the load change, and the fourth power inverse attenuation model can be used to determine the field source distance, and the blocking attenuation of environmental factors such as terrain and buildings can be considered to simulate the propagation loss of electromagnetic waves in complex environments. When performing FDTD simulation calculations, appropriate medium parameters can be selected for electromagnetic waves in different frequency bands, such as using a quasi-static model from DC to very low frequency, a Drude dispersion model in medium and low frequency bands, and a Debye polarization model in high frequency bands. By introducing auxiliary differential equations to simplify the Maxwell I equation group, the computational efficiency and numerical stability of wide-band simulations can be improved.In the visualization analysis of electromagnetic fields, the cloud map can be generated by volume rendering algorithm, the isosurface can be extracted by ray casting algorithm, the electric field lines and magnetic lines can be drawn by combining streamline tracing algorithm and tensor interpolation algorithm, and the field intensity distribution can be mapped by pseudo-color coding method to generate intuitive and realistic visualization effects. When constructing load change positioning and prediction models, the 3DUnet structure can be used to integrate high-dimensional spatial features, and the frequency domain features and topological features can be introduced into the machine learning model structure to enhance the temporal and spatial correlation. The scene similarity constraint can be added to the loss function to improve the interpretation accuracy. Federated learning combined with blockchain can be used to realize the secure training and remote reasoning of the model, and the block hash value can be used as a digital fingerprint to verify the integrity of the model. The differential privacy technology can be used to protect sensitive information in the modeling process.

[0032] S105. During the simulated electromagnetic wave propagation process, a time-frequency domain analysis method is used to calculate in real time the power load of the power grid and the energy density distribution data of the simulated electromagnetic waves at each node and edge, and an electromagnetic interference assessment model is constructed in combination with the sensitivity weight values ​​of each node and edge. According to the electromagnetic interference assessment model, interference areas where the electromagnetic wave energy density exceeds the preset threshold are identified.

[0033] In the process of simulating electromagnetic wave propagation, an energy density monitoring model is established for each node and edge in the power grid according to the energy density distribution data, and the voltage level parameters, load power parameters and wire material parameters of the nodes, as well as the resistance parameters, reactance parameters and line length parameters of the edges are combined in the energy density monitoring model; the real-time operation data of the power grid is obtained, and the power load distribution of each node is obtained according to the power flow calculation results; the load distribution data is associated with the node attributes to form a node load attribute matrix; the power flow distribution of each edge is obtained according to the power grid operation state estimation result; the power flow distribution data is associated with the edge attributes The properties are associated to form an edge load attribute matrix; the power load data of the nodes and edges are matched with the electromagnetic wave energy density data, and the impact of power load changes on the electromagnetic wave energy density distribution is estimated in real time by using a data fusion algorithm; the correlation strength between load changes and electromagnetic wave energy density is quantified by establishing a load sensitivity model; the sensitivity weight value of each node and edge is obtained from the digital twin system; the sensitivity weight value is combined with the load sensitivity model to construct an electromagnetic interference assessment model; through the calculation of the assessment model, highly sensitive nodes and edges whose electromagnetic wave energy density exceeds the preset threshold are identified, and the impact area of ​​electromagnetic interference is determined.

[0034] For example, during the simulation of electromagnetic wave propagation, electromagnetic field sensors and energy metering devices are deployed on each node and edge in the power grid to measure the energy density distribution of electromagnetic waves on each node and edge in real time, and an energy density monitoring model is established. The model comprehensively considers the attribute parameters of the node such as voltage level, load power, and conductor material, as well as the attribute parameters of the edge such as resistance, reactance, and line length, and uses the time-frequency domain analysis method to process and extract features from the measured data. The real-time operation data of the power grid is obtained, and the steady-state operation characteristics of the power grid are analyzed through power flow calculation to obtain the voltage amplitude and phase angle of each node, as well as the active power and reactive power distribution of each branch, to form a node load attribute matrix. At the same time, through state estimation, the accurate operation state of the power grid, including node voltage, branch power, etc., is estimated as input for subsequent safety analysis and optimal scheduling to form an edge load attribute matrix. The power load data of the nodes and edges are fused with the electromagnetic wave energy density data, and the Dempster-Shafer evidence theory is used to consider the credibility of different data sources, and the load data and energy density data are weighted and combined to establish a load sensitivity model. Quantitative indicators such as load change rate and energy density change rate are introduced into the model. Through sensitivity analysis, the influence coefficient of load change on energy density is calculated, and the correlation strength between load change and electromagnetic wave energy density is quantified. The sensitivity weight values ​​of each node and edge are obtained from the digital twin system, and the weight values ​​are combined with the load sensitivity model to construct an electromagnetic interference assessment model. The threshold of electromagnetic wave energy density is set in the electromagnetic interference assessment model. Through model calculation, highly sensitive nodes and edges with energy density exceeding the threshold are identified, and an electromagnetic interference impact area map is generated. The electromagnetic interference level of each node and edge in the interference area is calculated. In the electromagnetic interference area map, the Kruska l minimum spanning tree algorithm is used to search for key nodes and key edges in the interference area with nodes as vertices and edges as connections to form a key path for interference propagation. At the same time, the degree centrality of each node and the betweenness centrality of the edge are calculated, and the nodes and edges with centrality indicators exceeding a certain threshold are selected as key nodes and key edges. Taking key nodes and key edges as the starting point, electromagnetic interference protection strategies are formulated, such as optimizing the load distribution of key nodes, adjusting the wire material of key edges, and adding electromagnetic shielding facilities on key nodes and edges. The electromagnetic interference protection strategy is simulated and verified. A power grid model is built in the electromagnetic simulation software. The parameters of the nodes and edges involved in the protection strategy are modified. Through time domain simulation and frequency domain analysis, the changes in the electromagnetic wave energy density before and after the implementation of the protection strategy are calculated to evaluate the effectiveness of the protection strategy. Evaluation indicators such as energy density peak value, average value, and exceedance rate are introduced. When these indicators are reduced below the threshold, the protection strategy is considered to be effective, and the protection strategy parameters are synchronized to the digital twin system and the physical power grid system. If the evaluation result of the protection strategy is not ideal, the protection strategy needs to be further optimized and the above simulation verification process is repeated until a satisfactory protection effect is obtained.In the case that the protection strategy cannot effectively reduce the electromagnetic interference level, it is necessary to activate the emergency plan, formulate emergency control measures such as load reduction and equipment withdrawal according to the safety and stability constraints of the power grid, and send warning information to the power grid dispatcher in time. At the same time, collect and save the electromagnetic wave spectrum data of each monitoring point, use spectrum analysis, energy clustering and other technical means to carry out electromagnetic interference source tracing analysis, search for high energy density areas, determine the location of electromagnetic interference sources, and provide a basis for subsequent electromagnetic environment management and interference source removal. Electromagnetic wave sensors and energy metering devices are installed on each substation and transmission line of the power grid. The sampling frequency of the sensor is set to 1MHz and the range is set to 0-100W / m. 2 ,The edge gateway is accessed through the RS485 bus, and the collected electromagnetic wave spectrum data and energy data are uploaded to the cloud server in real time.,In the energy density monitoring model, the voltage level of the substation is considered, such as 220kV, 500kV, the capacity of the main transformer such as 240MVA, 360MVA, the material of the bus such as copper bus and aluminum bus, and the conductor model of the transmission line such as LGJ-240 / 30, LGJ-400 / 35, the tower height such as 45m, 60m, the line length such as 10km, 20km, and other parameters.,The FDTD algorithm is used to simulate and analyze the propagation characteristics of electromagnetic waves in the time domain and frequency domain, and the energy density distribution curves at different frequencies are obtained. In the power flow calculation, the Newton-Raphson method is used to solve the power flow equations, and the voltage amplitude of each node, such as 0.95-1.05pu and the phase angle, such as -10°-10°, are obtained, as well as the active power of each branch, such as 50MW, 100MW, and the reactive power, such as 10MVar, 20MVar, to form a node load attribute matrix of 500500. In the state estimation, the weighted least squares method is used, with the node injection power and branch power flow as input, and the node voltage and branch power as state variables, to construct a state estimation model. By solving the optimization problem, the accurate operation state of the power grid is obtained, and a side load attribute matrix of 5001000 is formed. In the load sensitivity model, the DS evidence theory is used to treat the load data and energy density data as two independent evidence sources. The joint probability distribution of load change and energy density change is obtained through the product operation of the evidence body. Then, the correlation coefficient between the load change rate and the energy density change rate is calculated. When the correlation coefficient is greater than 0.6, it is considered that the load change has a significant impact on the energy density, and the load node is marked as a highly sensitive node. In the electromagnetic interference assessment model, the load sensitivity index is weighted and summed with the sensitivity weight value of the node and edge to obtain the electromagnetic interference sensitivity of each node and edge, and then the energy density threshold is set to 50W / m 2, cluster analysis is performed on the nodes and edges that exceed the threshold, and 5 interference impact areas are obtained, and the average interference degree of the nodes and edges in each area is calculated. In the search of key nodes and edges, the Kruska l algorithm is used to construct the minimum spanning tree of the interference area. Through iterative pruning operations, the key path covering 80% of the interference nodes and edges is obtained, and the degree centrality of the key nodes and the betweenness centrality of the key edges are calculated. When the centrality index is greater than 0.5, the nodes and edges are added to the key element set. In the electromagnetic interference simulation verification, the COMSOL software is used to build a power grid model, and the material, size and other parameters of the key nodes and key edges involved in the protection strategy are modified, such as changing the copper wire to an aluminum wire, and adding a 10mm thick shielding layer to the busbar, etc., and then the time domain transient analysis and frequency domain step response analysis are performed to obtain the time curve and spectrum of the electromagnetic wave energy density before and after protection, and calculate the average energy density in the 80MHz-300MHz frequency band, such as 30W / m 2 、10W / m 2 , maximum energy density such as 80W / m 2 , 20W / m 2 , exceeding standard rate such as 20%, 5% and other indicators, comprehensively evaluate the protection effect, when the average energy density is reduced by 70%, the maximum energy density is reduced by 80%, and the exceeding standard rate is reduced by 90%, the protection strategy is considered to be effective. In the electromagnetic interference source tracing analysis, the MU IC algorithm is used to estimate the spatial spectrum of the spectrum data of multiple monitoring points to obtain the spatial position of the interference source, and then the spectrum data is short-time Fourier transformed to extract the time-frequency characteristics of the interference signal. Through cluster analysis, the modulation method of the interference signal such as BPSK, QPSK, carrier frequency such as 900MHz, 1800MHz, bandwidth such as 5MHz, 10MHz and other parameters are identified, and the type of interference source such as radio transmitter and radar antenna is determined, providing a handle for subsequent electromagnetic environment management.

[0035] S106. If the interference area is identified, a risk value for the marked electronic device in the interference area is generated according to the sensitivity weight values ​​of each node and edge in the interference area. When the risk value is greater than a preset threshold, an instruction to evacuate the marked electronic device or adjust the power load is generated.

[0036] After identifying the interference area, the sensitivity weight values ​​of all nodes and edges in the interference area are obtained; based on the sensitivity weight values ​​of the nodes and edges, the risk value of each marked electronic device in the interference area is calculated by using the weighted average method; the risk value of the marked electronic device is compared with a preset threshold, and when the risk value exceeds the preset threshold, the marked electronic device is determined to be in a high-risk state; for high-risk marked electronic devices, a multi-objective optimization algorithm is used to solve and generate evacuation or load adjustment instructions; while generating the evacuation or load adjustment instructions, the safety check and stability calculation of the power grid are triggered; the evacuation or load adjustment instructions are sent to the corresponding model of the digital twin system for verification, and the changes in electromagnetic wave energy density of nodes and edges in the electromagnetic interference area are monitored in real time to evaluate the effectiveness of protective measures.

[0037] For example, after identifying the electromagnetic interference area, the sensitivity weight values ​​of all nodes and edges in the area are obtained. The sensitivity weight values ​​are calculated based on the electrical parameters, topological structure, equipment type and other attributes of the nodes and edges using the fuzzy comprehensive evaluation method. First, select indicators such as node voltage level, load rate, cable cross-sectional area, etc., construct a fuzzy evaluation matrix, then determine the membership function of each indicator, determine the indicator weight by expert scoring method, and finally obtain the sensitivity weight value of the node and edge through fuzzy operation. The weight value range is 0 to 1. The larger the value, the more serious the impact of electromagnetic interference on the node or edge. According to the sensitivity weight value of the node and edge, the risk value of each marked electronic device in the area is calculated. The risk value is calculated using the weighted average method, and the sensitivity weights of the nodes and edges connected to the equipment are used as weight coefficients. The importance of the equipment itself, the failure rate, the electromagnetic compatibility and other indicators are weighted and summed. Among them, the importance is determined according to the function of the equipment, load size and other factors, and the value is assigned to 0.2-0.5; the failure rate is obtained according to the historical operation data of the equipment, and the value is assigned to 0-0.1; the electromagnetic compatibility is determined according to the anti-interference level test results of the equipment, and the value is assigned to 0-0.3. The comprehensive risk value of the equipment is obtained by weighted average calculation, and the mathematical formula is: R i sk = ∑ (Wi × Index i), where Wi is the weight coefficient and Index i is the value of each index. The risk value of the marked electronic equipment is compared with the preset threshold. The threshold is set using the hierarchical analysis method. By constructing a judgment matrix, pairwise comparison and consistency test are performed to obtain the weights of the safety margin, stability margin, and reliability index. Then, according to the importance of the index and historical data, the threshold of each index is determined, and the comprehensive threshold is obtained by weighted summation. When the risk value of the equipment exceeds the comprehensive threshold, it is determined that the equipment is in a high-risk state and protective measures need to be taken. For high-risk equipment, evacuation or load adjustment instructions are generated. The generation of instructions adopts a multi-objective optimization model. The objective functions include minimizing the load adjustment amount and minimizing the number of equipment operations. The constraints include power balance constraints, line flow constraints, voltage amplitude constraints, etc. The improved genetic algorithm is used to solve the optimal scheduling plan. During the solution process, a new scheduling plan is generated through crossover mutation operations, and then the new plan is decoded, repaired, and evaluated. Individuals with higher fitness are selected to join the population, and individuals with lower fitness are eliminated. After multiple iterations, the optimal solution is obtained. After generating evacuation or load adjustment instructions, it is necessary to analyze and verify the safety and stability of the power grid, using a combination of N-1 verification and dynamic time domain simulation. The N-1 verification lists the failure scenarios of a single device, calculates the flow distribution after the failure, and analyzes whether the line power, voltage amplitude, etc. meet the safety operation requirements.Dynamic time-domain simulation uses a group of differential algebraic equations to describe the dynamic behavior of the power grid. The change process of the power grid state over time is calculated through numerical integration, and the transient stability and dynamic stability of the power grid are analyzed. The simulation time step is 0.01s, the total simulation time is 10s, and the initial condition is the steady-state operation point before the fault. After the command is sent to the digital twin system, a real-time simulation platform consisting of a simulation server, a real-time digital interface, and a physical controller is built through hardware-in-the-loop simulation technology. The command is converted into a control signal, and real-time data interaction is carried out with the digital twin model through a high-speed communication network. The execution effect of the command is simulated and verified with high precision at the millisecond level, and the key operating indicators such as the network loss rate and frequency deviation of the power grid are analyzed. If all indicators meet the requirements, the command is synchronized to the physical power grid system for execution. If the electromagnetic interference level still exceeds the threshold after the digital twin simulation verification, the regional electromagnetic environment governance process is initiated, and comprehensive protection measures such as isolation shielding, grounding filtering, and active suppression are taken. Among them, isolation shielding mainly adopts composite shielding materials with high conductivity and high magnetic permeability to build an electromagnetic isolation cover between the interference source and sensitive equipment, and adopts multi-point grounding and low impedance grounding to reduce the ground potential and loop impedance. Ground filtering mainly adopts a combination of capacitive absorption and LC trap. Capacitors and inductors are connected in series in the grounding line to form a resonant branch to absorb interference signals of specific frequencies. Active suppression mainly adopts a feedback-feedforward composite control strategy. The interference magnetic field is detected in real time by the sampling coil, and then the opposite magnetic field is generated by the secondary coil for active offset, thereby weakening the interference field strength. Finally, the protection effect is tracked and monitored in an all-round and long-term manner. Through spectrum analysis, energy detection and other methods, key indicators such as the attenuation rate and residual level of electromagnetic interference are evaluated to form a closed-loop control. When calculating the sensitivity weights of nodes and edges, node voltage levels such as 220kV and 500kV, load rates such as 0.6 and 0.8, and cable cross-sectional areas such as 240mm can be selected. 2 , 400mm 2 The membership function of the voltage level is constructed by using triangular fuzzy numbers and trapezoidal fuzzy numbers. For example, the membership function of the voltage level is μ(x) = (-x / 280+2, x∈[220,500]). Then, the Delphi method is used to conduct a questionnaire survey on 10 experts to calculate the weight vector of each indicator.

[0038] W = [0.2, 0.15, 0.1, 0.15, 0.05, 0.1, 0.15, 0.1], the sensitivity weights of nodes and edges are obtained by weighted average method, and the weight value range is [0.368, 0.912]. When calculating the risk value of equipment, the importance index is assigned to (0.2, 0.35, 0.5) according to the power capacity of the equipment (such as <1MW, 1-10MW, >10MW), the failure rate index is assigned to 0, 0.05, 0.1 according to the average number of failures per year of the equipment, such as 0 times, 1-2 times, 3-5 times, and the electromagnetic compatibility index is assigned to (0.1, 0.2, 0.3) according to the electromagnetic radiation level of the equipment, such as Class A, Class B, Class C. By formula

[0039] Risk = 0.2W_1 + 0.15W_2 + 0.1I_1 + 0.15I_2 + 0.05I_3 + 0.1C_1 + 0.15C_2 + 0.1C_3 to calculate the risk value, where W_i, I_i, C_i are weight coefficients and three index values ​​respectively. When determining the risk threshold, the third-order judgment matrix is ​​constructed by the AHP method, and the weights of the safety margin, stability margin, and reliability index are calculated to be 0.625, 0.238, and 0.137 respectively, and the consistency ratio CR = 0.017 < 0.1. The judgment matrix B = [1, 3, 5; 1 / 3, 1, 2; 1 / 5, 1 / 2, 1] is used to determine the thresholds of each index to be 0.1, 0.15, and 0.2 respectively, and the comprehensive threshold is 0.127 obtained by the weighted summation method. When generating equipment evacuation instructions, the multi-objective optimization model contains two objective functions min(F_1=∑|P_li-P_li'|) and min(F_2=∑x_i), and the constraints include ∑P_Gi-∑(P_Li+P_Di+ΔP_Ti)=0, S_i j≤S_max, etc., where P_li is the power before load adjustment, P_li' is the power after adjustment, x_i is the state variable of the equipment, P_Gi, P_Li, P_Di are generator output, load demand, and network loss respectively, ΔP_Ti is the parallel capacitor compensation power, and S_max is the upper limit of line current carrying capacity. The NSGA-II improved genetic algorithm is used for solution, and the population size is set to 100, the number of iterations is 200, the crossover probability is 0.8, and the mutation probability is 0.1. The Pareto non-inferior optimal solution set is obtained, and the compromise solution with F_1 and F_2 trade-off coefficients of 0.6 and 0.4 respectively is selected as the optimal scheduling solution. When conducting N-1 verification, 25 single equipment failure scenarios of 5 categories, including substation busbar, main transformer, feeder switch, etc., are listed, and the node voltage amplitude deviation and line power change percentage before and after the fault are calculated. If the deviation exceeds 10% or the change rate exceeds 30%, it is judged as a safety warning. When conducting dynamic simulation, the 6th-order synchronous generator model and ZIP comprehensive load model are used, the simulation step is set to 0.01s, the total simulation time is 10s, and a three-phase short-circuit fault is applied at t=1s. After 0.1s, the fault is removed. The rotor angle stability and voltage transient recovery level of the power grid are analyzed. If the rotor angle oscillation does not converge within 3s after the fault is removed or the voltage fluctuation exceeds 0.1pu, it is judged as an instability risk.When conducting digital twin simulation verification, a HIL real-time simulation platform consisting of an RTDS real-time simulator, a DSP physical controller, an optical fiber transceiver, etc. is built. Real-time data interaction across platforms is carried out through the IEC61850 communication protocol, and millisecond-level continuous simulation of power grid fault scenarios is performed. Key indicators such as network loss rate and frequency deviation during the fault period are analyzed and compared with the offline simulation results. If the error between the two is within 5% and the key indicators all meet the Class D power quality standards, the control instructions are determined to be feasible. When conducting regional electromagnetic environment management, for the high-voltage side of the substation, a metal shielding net is used to shield the transformer, busbar, reactor, etc. for electric field, and the shielding efficiency reaches 40dB; for the medium and low voltage side, a combined grounding method of annular grounding grid plus grounding electrode is used to reduce the grounding resistance to below 0.5Ω; for high-order harmonics, a 50Hz series resonant branch and a 250Hz parallel resonant branch are connected in series, and the total filtering capacity reaches 15% of the load; for high-power converters, a voltage source active filter is used, and the reverse harmonic output of the inverter circuit is controlled by PWM through real-time detection of compensation current, so that the harmonic content on the grid side is reduced to less than 3%. Through the above multi-means of comprehensive management, after 3 months of tracking and monitoring, the electromagnetic interference field strength has been reduced by 8dBμV / m, and the harmonic voltage content has been reduced by 5%, meeting the requirements of electromagnetic compatibility level 3.

[0040] S107. Based on the monitoring data, identify the high-load area where the power load is higher than the threshold; if the existence of the high-load area lasts longer than the time threshold and there are multiple marked electronic devices in it, prioritize the marked electronic devices based on the sensitivity differences of the nodes and edges associated with the marked electronic devices, and generate a corresponding interference handling strategy based on the priority of the marked electronic devices.

[0041] By deploying power load monitoring devices in substations, distribution rooms and user terminals, load data of each monitoring point is collected in real time, and the load data is uploaded to the main station database through a dedicated network for management; according to the safe and stable operation limit of the power grid, as well as the seasonal and periodic characteristics of the load, the load threshold is set by the piecewise function method; regional aggregation analysis is performed on the real-time load monitoring data, and a density clustering algorithm is used to identify high-load areas; for the identified high-load areas, the sliding time window method is used to count the duration of the high-load area, and if it is greater than the preset time threshold, it is determined to be a continuous high-load area; the continuous high-load area is obtained For all the marked electronic devices in the continuous high-load area, the attribute information of the nodes and edges associated with the devices is extracted, and the node sensitivity difference index and the edge sensitivity difference index are calculated by principal component analysis; the marked electronic devices in the continuous high-load area are sorted according to the node sensitivity difference index and the edge sensitivity difference index, and the devices are divided into three priorities: high, medium and low according to the sorting results; for the marked electronic devices with different priorities, the protection measures are matched from the policy knowledge base, and then the interference handling strategy is generated; the interference handling strategy is sent to the regional controller for execution, and the handling effect is tracked and evaluated to form a closed-loop control.

[0042] For example, by deploying power load monitoring devices at key nodes such as substations, distribution rooms, and user terminals, real-time data such as active power, reactive power, and apparent power are collected at each monitoring point. The sampling frequency is not less than 1 time / minute, and the measurement accuracy is better than level 1.0. The monitoring data is uploaded to the main station database through a fiber-optic private network for partition storage and management. According to the safe and stable operation limit of the power grid, as well as the seasonal and periodic characteristics of the load, the load threshold is set using the piecewise function method. The threshold function form is P_th = {P1, t∈[t1, t2]; P2, t∈(t2, t3]; ...; Pn, t∈(tn-1, tn]}, where t is a time variable, P1, P2, ..., Pn are the threshold values ​​of each segment, and t1, t2, ..., tn are the start and end times of each segment. The selection of segment points comprehensively considers factors such as the peak and valley characteristics of the load, the rate of increase and decrease, the temperature sensitivity coefficient, and the thermal capacity of the equipment. The real-time load monitoring data is partitioned. Domain aggregation analysis, set M consecutive sampling points exceeding the threshold to be judged as high load anomaly, and then use DBSCAN density clustering algorithm to identify high load areas. This algorithm defines the concepts of object density reachability and density connectivity, builds a cluster structure model, automatically divides the area with density higher than the threshold into a cluster, and marks the object with density lower than the threshold as a noise point. Through the process of iterative cluster expansion, the spatial distribution of high load density areas is finally obtained. For the identified high load areas, the sliding time window method is used to count their duration. The window length is dynamically adjusted according to the load characteristics of the power grid. The adjustment formula is T_wi n=f(σ,η,τ), where σ is the load fluctuation coefficient, η is the grid peak load coefficient, and τ is the equipment thermal stability time constant. The weights of each coefficient are determined by the hierarchical analysis method. If more than 90% of the monitoring points in the area are in a high load state in N consecutive time windows, and N is greater than the preset time threshold T, such as 60 minutes, the area is determined to be a continuous high load area. All the marked electronic devices in the continuous high load area are obtained, including micro-synchronous phasor measurement units (μPMUs), fault indicators (FTUs), distribution automation terminals (DTUs), etc. According to the topological association relationship, the attribute information of the nodes and edges associated with the equipment is extracted, including node voltage level, edge current carrying capacity, wire material, equipment type, etc., to construct the attribute matrix X, and then the principal component analysis (PCA) method is used to reduce the dimension of the attribute matrix to extract the first k principal components Z=PX, where P is the eigenvector matrix. The sensitivity weights of the nodes and edges are determined by the absolute value of the principal component loads, and the node sensitivity difference index I n and edge sensitivity difference index I l are obtained after normalization.For the marked electronic devices in the continuous high-load area, the priority is divided according to the node sensitivity difference index I n and the edge sensitivity difference index I l. The ABC classification method is used. According to the relative gap between the index and the regional mean, the devices with an index greater than 1.5 times the mean are classified as high priority (class A), the devices with an index less than 0.5 times the mean are classified as low priority (class C), and the rest are classified as medium priority (class B). In principle, the number ratio of A, B, and C devices is 2:3:5, and the classification results can be fine-tuned in combination with expert experience. For the marked electronic devices with different priorities, the optimal strategy combination is matched from the interference protection strategy knowledge base. The knowledge base is constructed using the ontology modeling method. The ontology concepts and relationships are defined from the three dimensions of protection objects, protection means, and performance indicators to form a strategy ontology model. Then, the strategy instances are extracted from multi-source heterogeneous data. Through semantic mapping, knowledge reasoning and other technologies, the instances are mapped to the ontology structure to achieve the standardized representation and storage of strategy knowledge. In the matching process, the semantic similarity-based reasoning algorithm is used to calculate the matching degree between the equipment features and the strategy features. At the same time, the hierarchical analysis method is used to evaluate the comprehensive benefits of the strategy from the aspects of economy, safety, reliability, etc., and finally the optimal strategy combination is selected. The interference handling strategy is decomposed into executable control instructions, which are sent to the regional controller through a dedicated communication network. The controller parses the instructions and distributes the control quantity to each execution unit through the field bus network, including the regulating transformer, controllable reactor, active power filter, etc. The execution unit adjusts the operating conditions according to the control quantity, including the transformer ratio, switching state, compensation current, etc. During the execution of the instructions, the field equipment uploads the feedback information such as the operating status and measurement value to the controller, and the controller synchronizes and coordinates the action sequence of each execution unit to ensure load balance, voltage stability, and harmonic compliance. The execution effect of the interference handling measures is tracked and evaluated, and a multi-dimensional and multi-indicator evaluation model is established. From the aspects of suppression effect, response speed, control accuracy, etc., key indicators such as load current distortion rate, voltage qualification rate, equipment loss, etc. are extracted. The weighted moving average method is used to calculate the comprehensive score of each indicator to form a quantitative effect evaluation conclusion. At the same time, data mining methods are used to analyze the collaborative characteristics of various devices under different working conditions, optimize the coordinated control strategy and parameters, and on this basis, improve the interference protection strategy knowledge base, realize dynamic optimization and updating of the strategy, and finally form an intelligent interference processing system with self-learning and self-adaptive capabilities. Intelligent power sensors are installed at locations such as the 10kV busbar of the substation, the 0.4kV busbar of the distribution room, and the user-end meter box. The sensors include current transformers, voltage transformers, Hall sensors, etc. The range covers 0-1000A, 0-500V, and the accuracy level is 0.2S. Data is uploaded through the RS-485 communication interface, and uploaded every 5s.The data of each monitoring point is transmitted to the main station through the fiber optic Ethernet ring network. The fiber bandwidth is 1000Mbps and the transmission delay is <1ms. The data is stored in the main station database according to the dimensions of substation, distribution line, and substation area. The Hadoop distributed file system is used to support PB-level massive data management. The load threshold is set to.

[0043] P_th={0.8P_N,t∈[00:00,06:00];0.9P_N,t∈(06:00,08:00];0.95P_N,t∈(08:00,11:30];0.8P_N,t∈(11:30,16:30];0.95P_N,t∈(16:30,21:00];0.85P_N,t∈(21:00,24:00]}, where P_N is the rated capacity of the transformer. The selection of segmentation points comprehensively considers the load characteristics of daytime, nighttime, production period, and catering period. The real-time load data is clustered every 30 minutes. The DBSCAN algorithm parameters are set to ε=200m, M i nPts=5, that is, the radius is 200m as the neighborhood search range. If the number of samples in the neighborhood exceeds 5, it is marked as a core object. The core object is used as the seed point to expand in the density reachable direction until the density is lower than the threshold, forming a high-load cluster. The process is iterated until all sample points are visited. If more than 90% of the sample points in the cluster are continuously in a high-load state in 3 consecutive time windows, such as the window length T_wi n=30min, it is determined to be a continuous high-load area. Extract the FTU, DTU and other data acquisition terminals in the area, and obtain the bus voltage level associated with the terminal, which is 10kV / 0.4kV, the distribution transformer capacity ≤200kVA / >200kVA, and the cable cross section ≤120mm 2 / >120mm 2The fuzzy membership matrix R is constructed based on the attribute values. The PCA method is used for eigenvalue decomposition. The eigenvectors corresponding to the first two eigenvalues ​​are extracted as the principal components. The principal component expression Z = 0.6X_1 + 0.3X_2 + 0.1X_3 is calculated. The coefficients of each attribute are the sensitivity weights. After normalization, the node sensitivity index is obtained. The sensitivity index of the equipment is statistically analyzed. The mean index is 0.6. The equipment in the interval [0.9, 1] is classified as Class A, which is highly sensitive, the equipment in the interval [0.3, 0.6) is classified as Class B, which is medium sensitive, and the equipment in the interval [0, 0.3) is classified as Class C, which is low sensitive. For Class A equipment, the "comprehensive governance" strategy is matched, including harmonic governance such as active power filtering, reactive power governance such as dynamic reactive power compensation, load balancing such as multi-point coordinated control, etc. For Class B equipment, the "fixed-point governance" strategy is matched, such as installing energy storage devices at sensitive loads to alleviate local high loads; for Class C equipment, the "early warning monitoring" strategy is matched, the sampling frequency is encrypted, and the load data granularity is refined. The control strategy is converted into a GOOSE message and sent down. The information model of the message follows the IEC61850 standard, the interaction method adopts the MMS protocol, and the control cycle is 100ms. During the control process, the "start active filtering" command is sent from the master station. After receiving the command, the DTU outputs the PWM modulation signal to the IGBT power module. The IGBT power module changes the on and off of the switch tube according to the modulation signal, injects compensation current into the system, and suppresses harmonic current. When a short-circuit fault occurs on the bus side, the FTU detects that the short-circuit current exceeds twice the rated value and immediately sends a "trip" command to the local control unit. The control unit disconnects the circuit breaker within 2 cycles (40ms) to isolate the fault section and prevent the fault from spreading. The operating parameters of the protection device, such as IGBT switching frequency and power quality indicators such as U_THD harmonic content, are monitored in real time and compared with KPI assessment indicators such as harmonic content <5%. The assessment score is calculated, such as harmonic control score = 1-U_THD / 5%. The EWMA model weighting coefficient α = 0.2 is used to predict the assessment score at the next moment. If the predicted value is lower than 80 points, the strategy optimization process is triggered. The optimal control parameters such as IGBT dead time are explored through reinforcement learning algorithms such as Q-Learning to continuously improve the protection effect.

[0044] What is disclosed above is only a preferred embodiment of the present invention, which certainly cannot be used to limit the scope of rights of the present invention. A person skilled in the art can understand that all or part of the processes of the above embodiments and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. A method for monitoring and controlling electric loads based on digital twins, characterized in that: The method comprises: Acquire historical load data of the power grid and monitor the power grid data, establish a digital twin model of the power grid based on the acquired historical load data and monitoring data, map the physical topology of the power grid to the digital twin model, and include attribute information of each node and edge corresponding to the power grid in the digital twin model; Conduct safety and importance assessments on the tagged electronic devices at each node and edge in the power grid, and obtain sensitivity weight values ​​for each node and edge; Analyzing the load change in the power grid according to the monitoring data, and when it is identified that the load change is greater than a preset threshold, starting an electromagnetic wave propagation simulation module, wherein the electromagnetic wave propagation simulation module is used to simulate the electromagnetic wave propagation of the load change; Acquire real-time operation data of the power grid, determine the range and size of the load change through a load state estimation algorithm based on Kalman filtering, acquire the power grid topology data and material properties in the digital twin model, and simulate the propagation process and spatial distribution of electromagnetic waves in the power grid based on the range and size of the load change, combined with the power grid topology data and the material properties, and using an electromagnetic field simulation engine; In the process of simulating electromagnetic wave propagation, a time-frequency domain analysis method is used to calculate in real time the power load of the power grid and the energy density distribution data of the simulated electromagnetic waves at each node and edge, and an electromagnetic interference assessment model is constructed in combination with the sensitivity weight values ​​of each node and edge, and the interference area where the electromagnetic wave energy density exceeds the preset threshold is identified according to the electromagnetic interference assessment model; If the interference area is identified, a risk value for the marked electronic device in the interference area is generated according to the sensitivity weight values ​​of each node and edge in the interference area, and when the risk value is greater than a preset threshold, an instruction to evacuate the marked electronic device or adjust the power load is generated; According to the monitoring data, a high-load area where the power load is higher than a threshold is identified; if the high-load area exists for a duration greater than a time threshold and there are multiple tagged electronic devices within it, the tagged electronic devices are prioritized based on the sensitivity differences of the nodes and edges associated with the tagged electronic devices, and a corresponding interference handling strategy is generated based on the priority of the tagged electronic devices.

2. The method according to claim 1, wherein: The historical load data of the power grid is obtained, and the power grid data is monitored. According to the historical load data and the monitoring data obtained, a digital twin model of the power grid is established, and the physical topology of the power grid is mapped to the digital twin model. The digital twin model contains attribute information of each node and edge corresponding to the power grid, including: According to the physical topology of the power grid, the digital twin model is constructed, and nodes and edges in the physical topology of the power grid are mapped to the digital twin model through a graph database method to obtain a virtual power grid model corresponding to the physical power grid; Using the mapping data of the nodes and edges, the attribute information of each node in the physical topology of the power grid is obtained, and the attribute information is associated with the corresponding node in the digital twin model to achieve synchronization of the node attribute information and the digital twin model; Acquire the attribute information of each edge in the physical topology of the power grid through the nodes mapped to the model, associate them with the corresponding edges in the digital twin model, and synchronize the edge attribute information with the digital twin model; By synchronizing the node attribute information and the edge attribute information, the monitoring data is associated with the digital twin model to obtain real-time update and display in the digital twin model; Obtain the association results between the monitoring data and the digital twin model, analyze the historical load data, explore the association relationship between the historical load data and the power grid topology structure, map the historical load data to the corresponding nodes and edges in the digital twin model, and obtain the digital twin model with topology structure, parameter attributes, real-time status and historical data.

3. The method according to claim 1, wherein: The security and importance of the marked electronic devices existing at each node and edge in the power grid are evaluated to obtain the sensitivity weight value of each node and edge, including: Acquire power grid topology data and real-time power grid operation data, and determine key nodes and key edges of the power grid based on the power grid topology data; The reliability, data integrity, communication security and power supply reliability of the electronic equipment are used as evaluation indicators to construct an evaluation indicator system; Using the hierarchical analysis method, the importance of each indicator in the evaluation indicator system is compared pairwise, a judgment matrix is ​​constructed, and the weight of each indicator is obtained by performing matrix operations on the judgment matrix; The weights of the various indicators are weighted and summed with the actual evaluation values ​​of the key nodes and the key edges, a comprehensive evaluation model is constructed, and a comprehensive score of the security and importance of each key node and each key edge is obtained; According to the comprehensive score, the sensitivity levels of the key nodes and the key edges are divided to obtain key nodes and key edges with high sensitivity, medium sensitivity and low sensitivity; The sensitivity levels of the key nodes and the key edges are dynamically combined with the real-time operation data of the power grid. When the load rate and failure rate of the key nodes or the key edges change, the sensitivity weight values ​​of the key nodes and the key edges are adjusted accordingly to obtain the sensitivity weight values ​​of each key node and key edge in the power grid.

4. The method according to claim 1, wherein: The load change in the power grid is analyzed according to the monitoring data, and when it is identified that the load change is greater than a preset threshold, an electromagnetic wave propagation simulation module is started, and the electromagnetic wave propagation simulation module is used to simulate the electromagnetic wave propagation of the load change, including: By real-time monitoring of each load node in the power grid, load parameters are obtained, wherein the load parameters include voltage, current and power; Using a sliding window algorithm to calculate the rate of change of the load parameter; When the change rate is greater than a preset threshold, it is determined that a sudden load has occurred, and the electromagnetic wave propagation simulation module is started; Constructing a propagation model of electromagnetic waves in the power grid according to the topological structure and conductor parameters of the power grid; The electromagnetic wave propagation process is numerically simulated using a finite difference time domain method, wherein the propagation model combines the influence of the resistance, inductance and capacitance of the conductor on the electromagnetic wave propagation characteristics, and the influence of the power grid topology on the electromagnetic wave propagation path; In the propagation model, influencing factors of the power grid environment are introduced, the influencing factors include temperature, humidity and wind speed, and the propagation model parameters are dynamically adjusted using the monitoring data; At the same time, in combination with the electromagnetic interference source in the power grid, the electromagnetic interference power source is used as the boundary condition of the propagation model; By means of simulation calculation of the propagation model, characteristic parameters of electromagnetic waves propagating in the power grid are obtained, wherein the characteristic parameters include time delay, attenuation loss and velocity distribution; Using the characteristic parameters, calculating the location and propagation range of the load change; The result of the electromagnetic wave propagation simulation is compared with the actual monitoring data, and a mapping relationship between the load change and the electromagnetic wave propagation characteristics is established by using a machine learning algorithm to achieve load change positioning and prediction based on the electromagnetic wave propagation simulation; In the electromagnetic wave propagation simulation process, a parallel computing method is used to divide the power grid topology into multiple sub-areas, and an independent electromagnetic wave propagation simulation is performed in each of the sub-areas; Information interaction and synchronization between the sub-regions are achieved through the electromagnetic field continuity condition on the region boundary; Integrating the electromagnetic wave propagation simulation module into a power grid monitoring system; When it is monitored that the load change is greater than a threshold, the electromagnetic wave propagation simulation is triggered to generate the load change positioning result, and the location and impact range of the load change are displayed on the power grid topology structure through a visualization method, and the power grid is dispatched and fault processing is performed according to the location and impact range of the load change.

5. The method according to claim 1, wherein: The real-time operation data of the power grid is obtained, the range and size of the load change are determined by a load state estimation algorithm based on Kalman filtering, the topological structure data and material properties of the power grid in the digital twin model are obtained, and according to the range and size of the load change, the power grid topological structure data and the material properties are combined, and an electromagnetic field simulation engine is used to simulate the propagation process and spatial distribution of electromagnetic waves in the power grid, including: When the load change amplitude is greater than a preset load threshold, or the change range is greater than a preset range threshold node, the electromagnetic wave propagation simulation task is triggered; The grid topology data includes the connection relationship and geometric parameters between busbars, lines and transformers; Generate a grid model required by an electromagnetic field simulation engine according to the power grid topology data, using an unstructured tetrahedral grid, wherein the size of the unstructured tetrahedral grid is adaptively adjusted according to the wavelength, and the unstructured tetrahedral grid meets the numerical discrete accuracy requirements; Acquire attribute parameters of power grid equipment and materials in the digital twin model, wherein the attribute parameters include the resistivity and dielectric constant of the conductor, and the magnetic permeability and dielectric loss factor of the insulating material; Mapping the property parameters to the material property library of the electromagnetic field simulation engine to construct a multi-physics field coupled simulation model; According to the position and range of the load change, an excitation source for electromagnetic wave propagation is set in the electromagnetic field simulation engine, and an ideal current source or voltage source is used as a drive, and the excitation source generates an excitation signal; The frequency, amplitude and phase of the excitation signal are generated according to the actual load change data, and the time domain-frequency domain conversion is realized through Fourier transform; Performing a time-domain-frequency-domain full-wave simulation on the electromagnetic wave propagation process in the electromagnetic field simulation engine, solving the Maxwell equations using a time-domain finite-difference method, and calculating the evolution of the electromagnetic field in space and time; The time step is adaptively adjusted according to the CFL stability condition, the space step meets the grid accuracy requirements, and the boundary condition adopts the PML absorption boundary to avoid false reflections; Extracting the result data of the electromagnetic field simulation, including the time series and spatial distribution of electric field intensity, magnetic induction intensity and current density, and generating electromagnetic field cloud maps, isosurface maps and streamline maps through visualization technology; Analyzing propagation characteristics of electromagnetic waves in the power grid, the propagation characteristics including propagation speed, attenuation law and standing wave distribution; Associating the electromagnetic field simulation results with the power grid topology and equipment parameters in the digital twin model to achieve mapping of the electromagnetic wave propagation process with the physical structure of the power grid; Through the convolutional neural network, a mapping relationship between the electromagnetic field distribution and the load change is established to achieve the positioning and prediction of the load change.

6. The method according to claim 1, wherein: In the process of simulating electromagnetic wave propagation, the power load of the power grid and the energy density distribution data of the simulated electromagnetic waves at each node and edge are calculated in real time by using a time-frequency domain analysis method, and an electromagnetic interference assessment model is constructed in combination with the sensitivity weight values ​​of each node and edge. According to the electromagnetic interference assessment model, an interference area where the electromagnetic wave energy density exceeds the preset threshold is identified, including: In the process of simulating electromagnetic wave propagation, an energy density monitoring model is established for each node and edge in the power grid according to the energy density distribution data, wherein the energy density monitoring model combines the voltage level parameters, load power parameters and wire material parameters of the nodes, and the resistance parameters, reactance parameters and line length parameters of the edges; Acquire the real-time operation data of the power grid, and obtain the power load distribution of each node according to the power flow calculation results; Associating the load distribution data with the node attributes to form a node load attribute matrix; According to the power grid operation state estimation result, the power flow distribution of each edge is obtained; Associating the power flow distribution data with the edge attributes to form an edge load attribute matrix; Matching the power load data of the nodes and edges with the electromagnetic wave energy density data, and using a data fusion algorithm to estimate in real time the impact of power load changes on the electromagnetic wave energy density distribution; By establishing a load sensitivity model, the correlation strength between load changes and electromagnetic wave energy density is quantified; Obtaining sensitivity weight values ​​of each of the nodes and edges from the digital twin system; Combining the sensitivity weight value with the load sensitivity model to construct an electromagnetic interference assessment model; Through calculation by the evaluation model, highly sensitive nodes and edges whose electromagnetic wave energy density exceeds the preset threshold are identified, and the impact area of ​​the electromagnetic interference is determined.

7. The method according to claim 6, wherein: If the interference area is identified, a risk value for the marked electronic device in the interference area is generated according to the sensitivity weight values ​​of each node and edge in the interference area, and when the risk value is greater than a preset threshold, an instruction to evacuate the marked electronic device or adjust the power load is generated, including: After identifying the interference area, obtaining sensitivity weight values ​​of all nodes and edges in the interference area; Calculate the risk value of each marked electronic device in the interference area by using a weighted average method according to the sensitivity weight values ​​of the nodes and edges; Comparing the risk value of the marked electronic device with a preset threshold, and determining that the marked electronic device is in a high-risk state when the risk value exceeds the preset threshold; For high-risk tagged electronic equipment, a multi-objective optimization algorithm is used to generate evacuation or load adjustment instructions; While generating the evacuation or load adjustment instruction, triggering the safety check and stability calculation of the power grid; The evacuation or load adjustment instruction is sent to the corresponding model of the digital twin system for verification, and the changes in the electromagnetic wave energy density of the nodes and edges in the interference area are monitored in real time to evaluate the effectiveness of the protection measures.

8. The method according to claim 1, wherein: According to the monitoring data, a high-load area where the power load is higher than a threshold is identified; if the existence of the high-load area lasts longer than a time threshold and there are multiple marked electronic devices in the area, the marked electronic devices are prioritized in combination with the sensitivity differences of the nodes and edges associated with the marked electronic devices, and a corresponding interference handling strategy is generated based on the priority of the marked electronic devices, including: By deploying power load monitoring devices in substations, distribution rooms and user terminals, load data of each monitoring point is collected in real time, and the load data is uploaded to the main station database through a dedicated network for management; According to the safe and stable operation limit of the power grid and the seasonal and periodic characteristics of the load, a load threshold is set by using a piecewise function method; Perform regional aggregation analysis on real-time load monitoring data and use density clustering algorithm to identify high-load areas; For the identified high-load area, a sliding time window method is used to count the duration of the high-load area, and if the duration is greater than a preset time threshold, it is determined to be a continuous high-load area; Acquire all marked electronic devices in the continuous high-load area, extract attribute information of nodes and edges associated with the devices, and calculate node sensitivity difference index and edge sensitivity difference index by principal component analysis method; Sorting the marked electronic devices in the continuous high-load area according to the node sensitivity difference index and the edge sensitivity difference index, and dividing the devices into three priorities of high, medium and low according to the sorting results; For the marked electronic devices with three priorities of high, medium and low, protection measures are matched from the policy knowledge base, thereby generating the interference handling policy; The interference handling strategy is sent to the regional controller for execution, and the handling effect is tracked and evaluated to form a closed-loop control.

Images