A power grid security situation awareness method

Abstract

The application discloses a power grid security situation awareness method, and particularly relates to the technical field of power grid security situation awareness, and comprises the following steps: acquiring a state data set of power grid safe operation, analyzing the connection mode of each node in the state data set, identifying all normal propagation paths, eliminating the paths with topological coincidence relationship, outputting a test path set, constraining the intensity range of a disturbance signal injected into each test path, screening out the disturbance signal intensity conforming to all test paths, and outputting an amplitude parameter range; constructing a state response path through active injection of a disturbance signal, eliminating an unrecoverable path and accessing a stable channel according to abnormal segment identification and simulation repair scoring, and completing power grid unobservable situation dynamic completion and safety modeling based on edge data transmission under an information physical system.

Description

Technical Field

[0001] This invention relates to the field of power grid security situation awareness technology, and more specifically, to a power grid security situation awareness method. Background Technology

[0002] Existing power grid situation awareness methods are based on an observability-driven technical architecture, which assumes that abnormal states in the power grid will inevitably cause fluctuations in electrical parameters. After sensing through telemetry equipment or PMU, situation judgment and response control are completed. Mainstream models generally adopt fluctuation feature extraction and trend comparison methods, construct situation evolution paths based on abrupt changes in variables such as voltage, current, and frequency, and rely on significant data offsets to trigger risk identification.

[0003] With the widespread deployment of long-chain power transmission structures, series compensation devices, and flexible interconnection equipment, some disturbance sources actually occur at the physical layer. However, due to the propagation path passing through high-impedance branches, power compensation devices, or phase angle matching structures, the spectral signal is weakened during transmission and ultimately does not manifest as quantifiable fluctuations at the sampling level. The fluctuation amplitude is lower than the sensing threshold, and the state evolution process cannot establish an effective trajectory in the situation model. The logical judgment result is a stable operating state.

[0004] The current perception logic uses "whether there is significant fluctuation" as the basis for judging the existence of a state. This results in the physical layer of the cyber-physical system being completely silent at the data layer during the edge data transmission stage. The sensing link does not trigger a perception response, the model does not establish a risk mapping, and the control system does not implement protective measures. Such state disturbances are imperceptible, unpredictable, and unverifiable under the cyber-physical system framework and edge data transmission mechanism. They constitute a blind spot in the stability modeling of the existing situational awareness system and further expose the structural defect that the state mapping mechanism has no ability to respond to unobservable states. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a power grid security situation awareness method. This method constructs a state response path by actively injecting disturbance signals, and eliminates unrecoverable paths and connects to stable channels based on abnormal segment identification and simulated repair scoring. This completes the dynamic completion and security modeling of the unobservable state of the power grid under the cyber-physical system.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a power grid security situation awareness method, comprising:

[0007] S1: Obtain the set of state data for safe operation of the power grid, analyze the connection mode of each node in the state data set, identify all normal propagation paths, remove paths with topological overlap, output the set of test paths, constrain the intensity range of the disturbance signal injected into each test path, filter out the disturbance signal intensity that meets all test paths, and output the amplitude parameter range.

[0008] S2: Based on the selected amplitude parameter range, perform disturbance signal strength injection debugging on each test path, collect the response changes of each test path after the injected disturbance signal, output feedback data, identify typical response characteristics at multiple sampling times, and output the main response feature set;

[0009] S3: Compare the main response feature set with the feedback data point by point on the sampling time axis, quantify the response offset of each path, and extract the path segments that are higher than the preset response offset threshold at multiple consecutive sampling times. Output the abnormal path segment set, input the abnormal path segment set into the trained autoencoder, perform simulated repair and scoring processing of abnormal path segments, filter the paths with a score clustering degree higher than the preset clustering degree threshold, and output the abnormal path set.

[0010] S4: Perform a joint judgment on the abnormal path set, and output the broken path group and the retained path set;

[0011] S5: Remove the broken path group from the debugging path set, introduce the path that has not participated in the disturbance debugging from the normal propagation path set, perform disturbance signal strength debugging, screen out the path that can stably transmit the disturbance signal, and output the reliable path set.

[0012] In a preferred embodiment, in S1, a set of state data for safe operation of the power grid is acquired, and by analyzing the connection mode of each node in the state data set, a normal propagation path is identified and a set of normal propagation paths is output.

[0013] Based on the topological overlap between paths in the normal propagation path set, an exclusion process is performed, and a test path set is output.

[0014] Perform perturbation amplitude injection constraint calculation on each path in the test path set, constrain the intensity range of the injected perturbation signal for each path, and output a set of amplitude parameters;

[0015] Based on the amplitude parameter set, the intensity of the disturbance signal injected into all test paths is uniformly screened, and the intensity of the injected disturbance signal that meets the requirements of all test paths is selected, and the amplitude parameter range is output.

[0016] In a preferred embodiment, S1 further includes the calculation of the disturbance amplitude injection limit by extracting the node numbers of the propagation start and end points of each path from the test path set, and reading the voltage phase difference and node connection resistance between all nodes in the path.

[0017] Based on the path propagation direction, the connection segments between adjacent nodes in the path are traversed segment by segment. The energy attenuation caused by signal propagation on the electrical connection of each connection segment is calculated. The original intensity of the disturbance signal injected at the starting point of the path is subtracted from the energy attenuation of each segment in turn to generate a disturbance intensity attenuation sequence propagating along the path direction. The portion of all non-negative disturbance intensity values ​​that is not higher than the preset propagation tolerance value is extracted from the disturbance intensity attenuation sequence as a reference for the upper limit of disturbance intensity that can be reached on the current path.

[0018] By combining the historical operation records of each node in the path in the power grid safe operation status data set, the disturbance intensity threshold that did not cause misjudgment in the past is statistically calculated, and the disturbance intensity value that is not lower than the preset disturbance identification threshold is extracted as the lower limit reference of the disturbance intensity of the current path. The upper limit reference of the disturbance intensity and the lower limit reference of the disturbance intensity are combined to form the disturbance intensity range value, which serves as the basis for calculating the path amplitude parameter set.

[0019] In a preferred embodiment, in S2, based on the selected amplitude parameter range, disturbance signal strength injection debugging is performed on each path in the test path set, and the debugging path set is output;

[0020] Response data acquisition is performed on the set of test paths. After the disturbance signal is injected, the response change of each test path is collected, and feedback data is output.

[0021] Extract the voltage and frequency change sequences of each test path from the feedback data, identify typical response characteristics at multiple sampling times, and output the main response feature set.

[0022] In a preferred embodiment, in S3, the main response feature set and the feedback data are compared point by point on the sampling time axis, the difference between voltage phasor, frequency fluctuation and timing features is calculated, the response offset of each path is quantified, and the offset quantization spectrum is output.

[0023] Extract path segments whose comprehensive response offset value is higher than the preset response offset threshold at multiple consecutive sampling times from the offset quantization map, and output a set of abnormal path segments;

[0024] The difference between the voltage phasor, frequency fluctuation and timing characteristics is obtained by extracting the voltage phasor value, frequency value and time index of each sampling moment in the feedback data for each test path, and simultaneously extracting the voltage phasor reference value, frequency reference value and timing reference point of the corresponding sampling moment in the main response feature set.

[0025] The voltage phasor value in the main response feature is subtracted from the voltage phasor value in the feedback data at each sampling point to obtain the voltage phasor difference. The frequency value is subtracted from the frequency reference value to obtain the frequency fluctuation difference. The difference between the time index and the time series reference point is calculated to obtain the response time series offset value. The voltage phasor difference, frequency fluctuation difference and response time series offset value are removed by their respective standardized reference amplitude, frequency reference value and time tolerance range to convert them into unitized offset values. For each sampling time, the three unitized offset values ​​are weighted and summed according to the preset offset importance ratio to obtain the comprehensive response offset value of the current sampling point.

[0026] Traverse all sampling times, record the comprehensive response offset value of each path, form a set of response offset sequences, arrange all response offset sequences in order of path number, align them according to sampling time, and generate an offset quantization map.

[0027] In a preferred embodiment, S3 further includes inputting the set of abnormal path segments into the trained autoencoder, performing simulated repair of the abnormal path segments, extracting the repair results and combining them with feedback data to perform scoring processing, and outputting a simulated repair score.

[0028] Based on the simulated repair score, the number and clustering density of abnormal segments with scores continuously higher than the score threshold in each path are counted, paths with a score clustering degree higher than the preset clustering degree threshold are filtered, and a set of abnormal paths is output.

[0029] The process involves simulating and repairing abnormal path segments, extracting the repair results, and combining them with feedback data to perform scoring and modeling. This is achieved by filling the abnormal path segments into a null response sequence on the sampling time axis according to the original sampling time sequence. The null response sequence is then input into the trained autoencoder. The autoencoder extracts the trend features and response structure patterns from the null response sequence and maps them into a latent representation vector. The autodecoder then generates the corresponding fitted response sequence based on the latent representation vector.

[0030] The voltage phasor value and frequency value at each sampling moment are read from the fitted response sequence. The difference between the voltage phasor value and frequency value at the corresponding path and time point in the feedback data is calculated point by point. The difference is then divided by the injected disturbance amplitude, the frequency reference value and the time index tolerance threshold, respectively, and converted into unit voltage error, unit frequency error and unit time offset error.

[0031] For the unit voltage error, unit frequency error, and unit time offset error generated within each sliding time window, the corresponding error values ​​are read sequentially according to the time index to form three error sequences of equal length. A fixed weight ratio is preset for each error sequence. For each time point, the unit voltage error, unit frequency error, and unit time offset error are multiplied by the corresponding weight ratio to generate the weighted error value for each time point.

[0032] The weighted error values ​​of all time points within each time window are accumulated, and the accumulated value is divided by the total number of time points contained in the time window to generate the anomaly score value of the current time window. All time windows of the fitted response sequence are traversed, and all anomaly score values ​​are arranged in chronological order to form an anomaly score sequence, which serves as the simulated repair score result of the anomaly path segment.

[0033] In a preferred embodiment, in S4, a joint judgment is made on the abnormal path set to determine whether the extreme value of the score in the simulated repair score sequence corresponding to the abnormal path is higher than a preset score boundary threshold, whether the unitized offset value of the abnormal path in the offset quantization map is higher than a preset response offset threshold, and whether the response frequency fluctuation amplitude of the abnormal path in the feedback data is higher than a preset frequency fluctuation threshold, as three judgment conditions.

[0034] If all three conditions are met, the broken path group is output; otherwise, the retained path set is output.

[0035] In a preferred embodiment, in S5, the broken path group is removed from the debugging path set, and the removal map is output.

[0036] The paths that did not participate in the disturbance debugging in the normal propagation path set are introduced into the broken path gaps in the elimination map, structural connectivity and response safety verification are performed, and the repair path set is output.

[0037] Each path in the repair path set is injected with the disturbance signal strength for adjustment, and compared with the main response feature set and offset quantization spectrum to select the path that can stably transmit the disturbance signal, and output a reliable path set.

[0038] The technical effects and advantages of this invention are as follows:

[0039] 1. This solution breaks the existing logic that no fluctuation means no risk by actively injecting disturbance signals into the test path set and collecting response feedback. It identifies the disturbance state of the edge data transmission layer in the cyber-physical system, which is silent at the data layer but real at the physical layer, and realizes explicit activation and risk exposure of unobservable states.

[0040] 2. By constructing a set of main response features and comparing them point by point with the feedback data on the time axis, an offset quantization map is generated, and path segments that are continuously higher than the preset response offset threshold are extracted to identify response distortion areas in the cyber-physical system and improve the extraction accuracy and trajectory reconstruction capability of abnormal path segments.

[0041] 3. Combine the autoencoder structure to perform simulated repair and scoring processing on abnormal path segments, extract densely clustered scoring paths and mark them as failure paths for abnormal feedback repair, so as to realize the structural expression of abnormal states, the judgment of repair feasibility and the path classification in cyber-physical systems;

[0042] 4. By introducing normal propagation paths that are not involved in disturbance debugging, structural connectivity and response security checks are performed and disturbance feedback checks are injected to rebuild the state feedback chain, thereby increasing the integrity of the power grid path structure and the stability of the situation evolution process. Attached Figure Description

[0043] Figure 1 This is a flowchart outlining the method steps of the present invention;

[0044] Figure 2 This is a flowchart of the path filtering scrambling process of the present invention;

[0045] Figure 3 This is a flowchart of the response acquisition and comparison process of the present invention;

[0046] Figure 4 This is a flowchart of the offset repair scoring process of the present invention;

[0047] Figure 5 This is a flowchart of the path anomaly determination process of the present invention;

[0048] Figure 6 This is a flowchart of the repair access screening process of the present invention. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] Refer to the instruction manual appendix Figure 1-6 An embodiment of the present invention provides a power grid security situation awareness method, comprising:

[0051] S1: Obtain the set of state data for safe operation of the power grid, analyze the connection mode of each node in the state data set, identify all normal propagation paths, remove paths with topological overlap, output the set of test paths, constrain the intensity range of the disturbance signal injected into each test path, filter out the disturbance signal intensity that meets all test paths, and output the amplitude parameter range.

[0052] S2: Based on the selected amplitude parameter range, perform disturbance signal strength injection debugging on each test path, collect the response changes of each test path after the injected disturbance signal, output feedback data, identify typical response characteristics at multiple sampling times, and output the main response feature set;

[0053] S3: Compare the main response feature set with the feedback data point by point on the sampling time axis, quantify the response offset of each path, and extract the path segments that are higher than the preset response offset threshold at multiple consecutive sampling times. Output the abnormal path segment set, input the abnormal path segment set into the trained autoencoder, perform simulated repair and scoring processing of abnormal path segments, filter the paths with a score clustering degree higher than the preset clustering degree threshold, and output the abnormal path set.

[0054] S4: Perform a joint judgment on the abnormal path set, and output the broken path group and the retained path set;

[0055] S5: Remove the broken path group from the debugging path set, introduce the path that has not participated in the disturbance debugging from the normal propagation path set, perform disturbance signal strength debugging, screen out the path that can stably transmit the disturbance signal, and output the reliable path set.

[0056] In S1, a set of state data for safe operation of the power grid is acquired. By analyzing the connection methods of each node in the state data set, normal propagation paths are identified, and a set of normal propagation paths is output. Identifying normal propagation paths involves reading the voltage phase angle information of each node in the state data set for safe operation of the power grid, analyzing the physical connection lines between nodes, constructing a signal propagation route map of the power grid, and calculating the electrical interaction strength and connectivity stability between nodes in each connection path. It is then determined whether each segment in the path meets the continuity condition, the resistance impedance is not higher than the propagation tolerance, and the voltage phase difference is within the allowable range. Paths that meet all conduction constraints are retained as normal propagation paths and integrated to form a set of normal propagation paths, which is used to support signal feedback analysis in disturbance injection and power grid safety situation awareness.

[0057] Based on the topological overlap between paths in the normal propagation path set, an exclusion process is performed to output a test path set. The exclusion process refers to removing paths that interfere with each other, have overlapping ranges, or have duplicate lines from the normal propagation path set, retaining only paths that can inject interference signals independently without affecting each other, forming an output test path set for disturbance injection and response assessment operations in power grid security situation awareness.

[0058] Perform perturbation amplitude injection constraint calculation on each path in the test path set, constrain the intensity range of the injected perturbation signal for each path, and output a set of amplitude parameters;

[0059] Based on the amplitude parameter set, the intensity of the injected disturbance signal is uniformly filtered for all test paths. The injected disturbance signal intensity that meets the requirements of all test paths is selected, and the amplitude parameter range is output. The uniform filtering of the injected disturbance signal intensity for all test paths involves extracting the upper and lower limits of the disturbance intensity range corresponding to each test path from the amplitude parameter set, constructing a set of upper and lower limit sequences of disturbance intensity for all test paths. For all upper limits in the set of upper and lower limit sequences, the disturbance intensity value that is not higher than the preset propagation tolerance is extracted as a unified upper limit reference for disturbance intensity. For all lower limits in the set of upper and lower limit sequences, the disturbance intensity value that is not lower than the preset disturbance identification threshold is extracted as a unified lower limit reference for disturbance intensity. It is then determined whether the unified lower limit reference for disturbance intensity is not higher than the unified upper limit reference for disturbance intensity. If the determination condition is met, the upper and lower limit reference values ​​are combined to form the amplitude parameter range, which is used for signal response evaluation operations in disturbance injection scheduling and power grid security situation awareness. If the determination condition is not met, the process returns to the amplitude parameter set to perform path exclusion or recalculation operations.

[0060] S1 also includes the calculation of the disturbance amplitude injection limit by extracting the node numbers of the propagation start and end points of each path from the test path set, and reading the voltage phase difference and node connection resistance between all nodes in the path;

[0061] Based on the path propagation direction, the connection segments between adjacent nodes in the path are traversed segment by segment. The energy attenuation caused by signal propagation on the electrical connection of each connection segment is calculated. The original intensity of the disturbance signal injected at the starting point of the path is subtracted from the energy attenuation of each segment in turn to generate a disturbance intensity attenuation sequence propagating along the path direction. The portion of all non-negative disturbance intensity values ​​that is not higher than the preset propagation tolerance value is extracted from the disturbance intensity attenuation sequence as a reference for the upper limit of disturbance intensity that can be reached on the current path.

[0062] By combining the historical operation records of each node in the path in the power grid safe operation status data set, the disturbance intensity threshold that did not cause misjudgment in the past is statistically calculated, and the disturbance intensity value that is not lower than the preset disturbance identification threshold is extracted as the lower limit reference of the disturbance intensity of the current path. The upper limit reference of the disturbance intensity and the lower limit reference of the disturbance intensity are combined to form the disturbance intensity range value, which serves as the basis for calculating the path amplitude parameter set and is used in the disturbance injection scheduling and feedback evaluation process.

[0063] In S2, based on the selected amplitude parameter range, disturbance signal strength injection debugging is performed on each path in the test path set, and the debugging path set is output as the source of the start signal for the active test response process in power grid security situation awareness;

[0064] The test path set is subjected to response data acquisition. After the disturbance signal is injected, the response changes of each test path are collected and feedback data is output. The response data acquisition refers to reading the termination node number of each path in the test path set, controlling the signal sensing device connected to the corresponding termination node, and recording the voltage change and frequency fluctuation data of the termination node according to the sampling time sequence after the disturbance signal is injected. This data serves as the response feedback data for each path and is used for reliability judgment of disturbance response and path distortion analysis in power grid security situation awareness.

[0065] The voltage and frequency change sequences of each test path are extracted from the feedback data. Typical response characteristics at multiple sampling times are identified, and a set of main response characteristics is output. Extracting the voltage and frequency change sequences of each test path and identifying typical response trends at multiple sampling times refers to reading the voltage and frequency changes recorded at the termination node of each test path at multiple consecutive sampling times from the feedback data, organizing them into voltage change sequences and frequency change sequences in chronological order, forming voltage change curves and frequency change curves. In multiple acquisition records of the same path, repeating voltage change curves and frequency change curves with similar shapes are found as typical response characteristics of the corresponding test path. All typical response characteristics are combined into a set of main response characteristics, which serves as a benchmark response template for anomaly identification and response deviation judgment in power grid security situation awareness.

[0066] In S3, the main response feature set and the feedback data are compared point by point on the sampling time axis. The difference between voltage phasor, frequency fluctuation and time series features is calculated, the response offset of each path is quantified, and the offset quantization spectrum is output.

[0067] The path segments whose comprehensive response offset values ​​are higher than the preset response offset threshold at multiple consecutive sampling times are extracted from the offset quantization map, and the abnormal path segment set is output. The extraction refers to reading the response offset sequence of each path from the offset quantization map, sliding to identify the offset segments that are continuously higher than the preset response offset threshold in the sampling time order, and extracting the path segments that meet the persistence condition as abnormal path segments, which are used to support the judgment of the repair and structural reconstruction of response interruption areas in the power grid security situation awareness.

[0068] The difference between the voltage phasor, frequency fluctuation and timing characteristics is obtained by extracting the voltage phasor value, frequency value and time index of each sampling moment in the feedback data for each test path, and simultaneously extracting the voltage phasor reference value, frequency reference value and timing reference point of the corresponding sampling moment in the main response feature set.

[0069] The voltage phasor value in the main response feature is subtracted from the voltage phasor value in the feedback data at each sampling point to obtain the voltage phasor difference. The frequency value is subtracted from the frequency reference value to obtain the frequency fluctuation difference. The difference between the time index and the time series reference point is calculated to obtain the response time series offset value. The voltage phasor difference, frequency fluctuation difference and response time series offset value are removed by their respective standardized reference amplitude, frequency reference value and time tolerance range to convert them into unitized offset values. For each sampling time, the three unitized offset values ​​are weighted and summed according to the preset offset importance ratio to obtain the comprehensive response offset value of the current sampling point.

[0070] Traverse all sampling times, record the comprehensive response offset value of each path, form a set of response offset sequences, arrange all response offset sequences in order of path number and align them according to sampling time, generate an offset quantization map, which is used to perform response offset identification and missing path extraction operations in power grid security situation awareness.

[0071] It should be noted that in the formula structure involved in this scheme, dimensionless terms can be used as proportional or structural adjustment factors. When combined with quantities with units, they only play a role in numerical scaling and do not introduce new physical dimensions. Therefore, they will not change or confuse the overall unit system. This combination of "dimensionless terms and terms with units" can be understood as a composite structural expression commonly used in mathematical physics modeling. It conforms to the principle of dimensional consistency and has a clear physical interpretation basis.

[0072] Secondly, in the formula structure of this scheme, if multiple variables with different physical units are involved, including but not limited to time, mass or energy variables, their joint appearance is to express the collaborative modeling relationship of multiple physical mechanisms. Each variable can form a unified structure through function mapping, ratio combination or normalization adjustment, with clear units and clear meaning. The overall expression conforms to the principle of dimensional consistency and the conventional formula of engineering modeling.

[0073] In this scheme, constants, weights, adjustment factors, threshold parameters, proportional coefficients, etc., are all adjustable control parameters for different application environments. Their values ​​depend on the target equipment configuration, data input characteristics, and performance optimization goals. During the implementation phase, they are converged within a reasonable range through model verification, performance constraints, or engineering calibration. Although these parameters do not have a unique preset value, they have clear adjustment logic and calculation paths. They belong to the deterministic setting process in engineering implementation. The purpose of this setting is to ensure that the scheme is both universally adaptable and reproducible and operable, without affecting its technical clarity and feasibility.

[0074] S3 also includes inputting a set of abnormal path segments into a trained autoencoder to simulate the repair of abnormal path segments, extracting the repair results and combining them with feedback data to perform scoring processing, and outputting a simulated repair score.

[0075] Based on simulated repair scores, the number and clustering density of abnormal segments with scores continuously higher than the score threshold in each path are counted. Paths with score clustering degrees higher than a preset clustering degree threshold are filtered out, and an abnormal path set is output. The counting and filtering refer to traversing the simulated repair scores of each path in the order of sampling time, extracting score segments with multiple consecutive simulated repair score values ​​higher than the preset score threshold, defining them as abnormal score segments, counting the total number of abnormal score segments, and calculating the distribution interval between these abnormal score segments on the time axis. When multiple abnormal score segments appear consecutively multiple times in a short time interval, or are concentrated and the time interval is lower than the preset clustering interval threshold, it indicates that the score clustering degree is dense. It is then determined whether the score clustering degree is higher than the preset score clustering degree threshold, and paths that meet the clustering criteria are filtered out and summarized into an abnormal path set for reference in anomaly identification and repair in power grid security situation awareness.

[0076] The process involves simulating and repairing abnormal path segments, extracting the repair results, and combining them with feedback data to perform scoring and modeling. This is achieved by filling the abnormal path segments into a null response sequence on the sampling time axis according to the original sampling time sequence. The null response sequence is then input into the trained autoencoder. The autoencoder extracts the trend features and response structure patterns from the null response sequence and maps them into a latent representation vector. The autodecoder then generates the corresponding fitted response sequence based on the latent representation vector.

[0077] The voltage phasor value and frequency value at each sampling moment are read from the fitted response sequence. The difference between the voltage phasor value and frequency value at the corresponding path and time point in the feedback data is calculated point by point. The difference is then divided by the injected disturbance amplitude, the frequency reference value and the time index tolerance threshold, respectively, and converted into unit voltage error, unit frequency error and unit time offset error.

[0078] For the unit voltage error, unit frequency error, and unit time offset error generated within each sliding time window, the corresponding error values ​​are read sequentially according to the time index to form three error sequences of equal length. A fixed weight ratio is preset for each error sequence. For each time point, the unit voltage error, unit frequency error, and unit time offset error are multiplied by the corresponding weight ratio to generate the weighted error value for each time point.

[0079] The weighted error values ​​of all time points within each time window are accumulated, and the accumulated value is divided by the total number of time points contained in the time window to generate the anomaly score value of the current time window. All time windows of the fitted response sequence are traversed, and all anomaly score values ​​are arranged in chronological order to form an anomaly score sequence, which serves as the simulated repair score result of the anomaly path segment. This sequence is used for the reliability evaluation of the repair of anomaly path segments in power grid security situation awareness and as input for subsequent screening and judgment.

[0080] Define simulated repair score :

[0081]

[0082]

[0083]

[0084]

[0085]

[0086]

[0087] in Indicates abnormal path fragments At the point of time The fitted response value generated by the autoencoder decoding; abnormal path segment At the point of time The null response input vector, This represents the pseudo-input sequence in the missing state; This refers to the encoder function in an autoencoder; This represents the decoder function in an autoencoder; Indicates abnormal path fragments At the point of time The unit voltage error value; Indicates abnormal path fragments At the point of time Fit the generated voltage phasor values; Indicates abnormal path fragments At the point of time The true values ​​of voltage phasors collected from feedback data; Indicates abnormal path fragments The upper limit of the voltage injection amplitude corresponding to the disturbance signal. From the set of disturbance injection amplitude parameters; Indicates abnormal path fragments At the point of time The unit frequency error value; Indicates abnormal path fragments At the point of time The frequency values ​​generated by fitting; Indicates abnormal path fragments At the point of time The true frequency values ​​collected from the feedback data; Indicates the reference amplitude of the power grid frequency; Indicates abnormal path fragments At the point of time The unit time offset error value; Indicates abnormal path fragments At the point of time The main response reference timing point, From the main response feature set; Indicates abnormal path fragments At the point of time The weighted error value; This indicates the amount of adjustment that the voltage phasor error is incorporated into the weighted error calculation; This represents the fusion adjustment amount of frequency error in the weighted error calculation; This represents the fusion adjustment amount of the time offset error in the weighted error calculation; Indicates abnormal path fragments In the sliding time window Abnormal score values ​​within; Represents a sliding time window The number of time points included; Indicates a point in time Sliding time window The time period in; Indicates abnormal path fragments Abnormal scoring sequences; This represents the set of consecutive sliding time window numbers divided in chronological order, with a total of . ; Indicates abnormal path fragments In the sliding time window Normalized outlier scores within the range; Indicates abnormal path fragments The total number of time windows divided into sliding time windows on the sampling time axis; This represents the time series scoring fusion function;

[0088] Autoencoders include variational autoencoders and denoising autoencoders;

[0089] In this scheme, if a variational autoencoder is applied to an autoencoder, the abnormal path segments can be padded into null response sequences and mapped to a latent representation distribution during the encoding stage. By introducing the changing trend features and response structure patterns in the disturbance response, a probability space expression is established. During the decoding stage, a fitted response sequence containing continuous structural features is reconstructed by sampling from the latent space. This adapts to the fluctuations in voltage phasor change amplitude, frequency drift, and inconsistencies in response delay, thereby realizing the simulated repair expression of abnormal path segments in power grid security situation perception and providing a stable input basis for scoring processing.

[0090] In this scheme, when a denoising autoencoder is applied to an autoencoder, the missing or severely disturbed time points in the abnormal path segment can be treated as high-noise points. The input is the response sequence after null values ​​are filled in. The encoder extracts the dominant change pattern under the interference background, and the noise-suppressed response sequence is reconstructed in the decoder. This increases the model's ability to faithfully restore the response contour of the abnormal path segment, realizes the stable recovery of voltage phasor and frequency change trends, and provides structural consistency support and comparison basis for the construction of anomaly scoring in power grid security situation awareness.

[0091] In S4, a joint judgment is made on the abnormal path set to determine whether the extreme value of the score in the simulated repair score sequence corresponding to the abnormal path is higher than the preset score boundary threshold, whether the unitized offset value of the abnormal path in the offset quantization map is higher than the preset response offset threshold, and whether the response frequency fluctuation amplitude of the abnormal path in the feedback data is higher than the preset frequency fluctuation threshold, as three judgment conditions.

[0092] If all three conditions are met, the broken path group is output as the final confirmation result of the failure to repair abnormal feedback in power grid security situation awareness; otherwise, the set of retained paths is output.

[0093] In S5, the broken path group is removed from the debugging path set and the removal map is output, which provides a structural basis for the access of alternative paths, the completion of situation structure and the reconstruction of power grid operation status.

[0094] Paths not involved in disturbance debugging from the normal propagation path set are introduced into the gaps in the elimination map of broken paths. Structural connectivity and response safety checks are performed, and a set of repair paths is output. The structural connectivity and response safety checks refer to reading the connection node sequence and electrical attribute parameters recorded in the normal propagation path set for each proposed path, verifying whether the starting node is connected to the starting gap of the broken path in the elimination map, whether the ending node is connected to the ending gap, and determining whether each connection structure in the path is continuous and without closure conflict. The disturbance propagation record of the path in the historical state data is read, and the voltage response amplitude, frequency fluctuation range, and transmission delay are analyzed to determine whether the voltage response amplitude is within the amplitude parameter range, whether the frequency fluctuation range meets the response judgment conditions, and whether the transmission delay is within the acceptable range. If the path connection structure is continuous, without closure conflict, and the disturbance response characteristics meet all response judgment conditions, it is determined to be a path that has passed the check and is included in the repair path set to complete the power grid state feedback chain.

[0095] Each path in the repair path set is injected with a disturbance signal intensity for adjustment, and compared with the main response feature set and the offset quantization map. Paths that can stably transmit disturbance signals are selected, and a reliable path set is output. Selecting a path that can stably transmit disturbance signals involves injecting and adjusting the disturbance signal intensity into each path in the repair path set, collecting the response voltage change sequence and frequency change sequence of each path after injection, comparing the response voltage change sequence and frequency change sequence with the typical response features of the corresponding path in the main response feature set, calculating the voltage phasor difference and frequency fluctuation difference at each sampling moment, and determining whether they are below the preset response similarity tolerance range. The response offset value of each path on the sampling time axis is compared point by point with the response offset value of the corresponding path in the offset quantization map to determine whether it is below the preset offset intensity threshold. When both comparisons meet the similarity requirement and the offset tolerance requirement, the corresponding path is selected as a path that can stably transmit disturbance signals. The reliable path set is then compiled and output for structural repair confirmation and transmission path stability determination in power grid security situation awareness.

[0096] It should be noted that, including but not limited to: traditional power grid situational awareness models use the presence of significant fluctuations as the basis for state judgment. Their core logic is based on signal responses triggered by observable data, and they generally rely on telemetry equipment, PMU equipment, etc., to capture the sudden changes in variables such as voltage, current, and frequency.

[0097] However, in modern complex power grid cyber-physical systems with high impedance structures, power compensation devices, or flexible interconnection equipment, a large number of real-world disturbance source signals are significantly weakened at the physical layer and further attenuated during edge data transmission. As a result, discernible fluctuations cannot be formed at the sensing level, leading to information masking and feedback silence problems in the state response logic.

[0098] To this end, this solution breaks through the passive architecture of response triggered by observable signals in logic, and instead constructs a new methodology based on active disturbance injection and response structure modeling. Its core mechanism is to actively stimulate the path response by applying external disturbance signals within the framework of cyber-physical systems and edge data transmission, and then identify hidden state shifts, filter distorted segments, perform simulation repair and credibility scoring, and finally reconstruct the power grid state feedback chain to achieve penetrating identification and structural perception of unobservable situations.

[0099] This solution includes a path filtering and injection configuration phase:

[0100] By reading the power grid state data set, the set of normal propagation paths that meet the conduction constraints is identified, and the topologically overlapping paths are eliminated to generate a set of structurally independent test paths. Based on the voltage phase difference, resistance and other parameters of the nodes in the test paths, a set of amplitude parameters that the disturbance signal can propagate is constructed. The uniform overlapping interval of the disturbance signal intensity is found in all test paths to obtain the range of legal amplitude parameters for global disturbance injection.

[0101] This path selection and injection setting stage is used to clarify the path structure basis and signal strength boundary of the disturbance injection, so as to avoid path interference and signal attenuation distortion.

[0102] This solution includes the main response extraction and feedback collection phases:

[0103] Based on the selected disturbance intensity, a disturbance signal is injected into each test path, and the voltage and frequency response sequences of the termination node are collected. By statistically analyzing the repeated voltage changes and frequency fluctuation patterns on the sampling time axis, the typical response behavior of each path is extracted, and a set of main response features is constructed as a reference for anomaly identification and structural comparison.

[0104] The main response extraction and feedback acquisition stage establishes a normal state expression of the path response, providing a reference structure for judging offset and distortion.

[0105] This solution includes an offset identification and repair scoring stage:

[0106] By comparing the feedback data with the main response feature set point by point, an offset quantization map is generated. Then, path segments that are continuously higher than the preset response offset threshold in the response offset sequence are extracted to generate an abnormal path segment set. The abnormal path segment set is input into the autoencoder to simulate and repair the missing structure. The scoring error at each time point is generated by combining the feedback data to construct an abnormal scoring sequence. Based on this, an abnormal path set with a high degree of scoring clustering is selected.

[0107] This offset identification and repair scoring stage is used to identify potential response fracture areas and to determine the credibility of the repair through simulated scoring, confirming whether there is a structural functional deficiency.

[0108] This plan includes the fracture determination and structural elimination stages:

[0109] Perform a joint judgment on the abnormal path set, and output the broken path group and the retained path set;

[0110] The fracture determination and structure elimination phase removes irreparable fracture regions from the current power grid path structure, preparing a structural entry point for alternative path access.

[0111] This plan includes a path substitution and stability screening phase:

[0112] Candidate paths connecting the two ends of the elimination region are selected from the normal propagation paths that have never participated in the debugging. Structural connectivity and response safety checks are performed to generate a set of repair paths. By re-injecting the perturbation signal and comparing the response results with the main response feature set and the offset quantization spectrum, paths that meet the conditions within the response similarity and offset tolerance are selected, and a set of reliable paths is output.

[0113] This path substitution and stabilization screening phase is used to repair structural pathways, ensuring that the selected alternative paths have stable disturbance propagation capabilities at the response level, and constructing a complete situational feedback structure.

[0114] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for power grid security situation awareness, characterized in that, include: S1: Obtain the set of state data for safe operation of the power grid, analyze the connection mode of each node in the state data set, identify all normal propagation paths, remove paths with topological overlap, output the set of test paths, constrain the intensity range of the disturbance signal injected into each test path, filter out the disturbance signal intensity that meets all test paths, and output the amplitude parameter range. S2: Based on the selected amplitude parameter range, perform disturbance signal strength injection debugging on each test path, collect the response changes of each test path after the injected disturbance signal, output feedback data, identify typical response characteristics at multiple sampling times, and output the main response feature set; S3: Compare the main response feature set with the feedback data point by point on the sampling time axis, quantify the response offset of each path, and extract the path segments that are higher than the preset response offset threshold at multiple consecutive sampling times. Output the abnormal path segment set, input the abnormal path segment set into the trained autoencoder, perform simulated repair and scoring processing of abnormal path segments, filter the paths with a score clustering degree higher than the preset clustering degree threshold, and output the abnormal path set. S4: Perform a joint judgment on the abnormal path set, and output the broken path group and the retained path set; S5: Remove the broken path group from the debugging path set, introduce the path that has not participated in the disturbance debugging from the normal propagation path set, perform disturbance signal strength debugging, screen out the path that can stably transmit the disturbance signal, and output the reliable path set.

2. The power grid security situation awareness method according to claim 1, characterized in that: In S1, the set of state data for safe operation of the power grid is acquired. By analyzing the connection mode of each node in the state data set, the normal propagation path is identified and the set of normal propagation paths is output. Based on the topological overlap between paths in the normal propagation path set, an exclusion process is performed, and a test path set is output. Perform perturbation amplitude injection constraint calculation on each path in the test path set, constrain the intensity range of the injected perturbation signal for each path, and output a set of amplitude parameters; Based on the amplitude parameter set, the intensity of the disturbance signal injected into all test paths is uniformly screened, and the intensity of the injected disturbance signal that meets all test paths is selected, and the amplitude parameter range is output.

3. The power grid security situation awareness method according to claim 2, characterized in that: S1 also includes the calculation of the disturbance amplitude injection limit by extracting the node numbers of the propagation start and end points of each path from the test path set, and reading the voltage phase difference and node connection resistance between all nodes in the path; Based on the path propagation direction, the connection segments between adjacent nodes in the path are traversed segment by segment. The energy attenuation caused by signal propagation on the electrical connection of each connection segment is calculated. The original intensity of the disturbance signal injected at the starting point of the path is subtracted from the energy attenuation of each segment in turn to generate a disturbance intensity attenuation sequence propagating along the path direction. The portion of all non-negative disturbance intensity values ​​that is not higher than the preset propagation tolerance value is extracted from the disturbance intensity attenuation sequence as a reference for the upper limit of disturbance intensity that can be reached on the current path. By combining the historical operation records of each node in the path in the power grid safe operation status data set, the disturbance intensity threshold that did not cause misjudgment in the past is statistically calculated, and the disturbance intensity value that is not lower than the preset disturbance identification threshold is extracted as the lower limit reference of the disturbance intensity of the current path. The upper limit reference of the disturbance intensity and the lower limit reference of the disturbance intensity are combined to form the disturbance intensity range value, which serves as the basis for calculating the path amplitude parameter set.

4. The power grid security situation awareness method according to claim 3, characterized in that: In S2, based on the selected amplitude parameter range, disturbance signal strength injection debugging is performed on each path in the test path set, and the debugging path set is output. Response data acquisition is performed on the set of test paths. After the disturbance signal is injected, the response change of each test path is collected, and feedback data is output. Extract the voltage and frequency change sequences of each test path from the feedback data, identify typical response characteristics at multiple sampling times, and output the main response feature set.

5. The power grid security situation awareness method according to claim 4, characterized in that: In S3, the main response feature set and the feedback data are compared point by point on the sampling time axis. The difference between voltage phasor, frequency fluctuation and time series features is calculated, the response offset of each path is quantified, and the offset quantization spectrum is output. Extract path segments whose comprehensive response offset value is higher than the preset response offset threshold at multiple consecutive sampling times from the offset quantization map, and output a set of abnormal path segments; The difference between the voltage phasor, frequency fluctuation and timing characteristics is obtained by extracting the voltage phasor value, frequency value and time index of each sampling moment in the feedback data for each test path, and simultaneously extracting the voltage phasor reference value, frequency reference value and timing reference point of the corresponding sampling moment in the main response feature set. The voltage phasor value in the main response feature is subtracted from the voltage phasor value in the feedback data at each sampling point to obtain the voltage phasor difference. The frequency value is subtracted from the frequency reference value to obtain the frequency fluctuation difference. The difference between the time index and the time series reference point is calculated to obtain the response time series offset value. The voltage phasor difference, frequency fluctuation difference and response time series offset value are removed by their respective standardized reference amplitude, frequency reference value and time tolerance range to convert them into unitized offset values. For each sampling time, the three unitized offset values ​​are weighted and summed according to the preset offset importance ratio to obtain the comprehensive response offset value of the current sampling point. Traverse all sampling times, record the comprehensive response offset value of each path, form a set of response offset sequences, arrange all response offset sequences in order of path number, align them according to sampling time, and generate an offset quantization map.

6. The power grid security situation awareness method according to claim 5, characterized in that: S3 also includes inputting a set of abnormal path segments into a trained autoencoder to simulate the repair of abnormal path segments, extracting the repair results and combining them with feedback data to perform scoring processing, and outputting a simulated repair score. Based on the simulated repair score, the number and clustering density of abnormal segments with scores continuously higher than the score threshold in each path are counted, paths with a score clustering degree higher than the preset clustering degree threshold are filtered, and a set of abnormal paths is output. The process involves simulating and repairing abnormal path segments, extracting the repair results, and combining them with feedback data to perform scoring and modeling. This is achieved by filling the abnormal path segments into a null response sequence on the sampling time axis according to the original sampling time sequence. The null response sequence is then input into the trained autoencoder. The autoencoder extracts the trend features and response structure patterns from the null response sequence and maps them into a latent representation vector. The autodecoder then generates the corresponding fitted response sequence based on the latent representation vector. The voltage phasor value and frequency value at each sampling moment are read from the fitted response sequence. The difference between the voltage phasor value and frequency value at the corresponding path and time point in the feedback data is calculated point by point. The difference is then divided by the injected disturbance amplitude, the frequency reference value and the time index tolerance threshold, respectively, and converted into unit voltage error, unit frequency error and unit time offset error. For the unit voltage error, unit frequency error, and unit time offset error generated within each sliding time window, the corresponding error values ​​are read sequentially according to the time index to form three error sequences of equal length. A fixed weight ratio is preset for each error sequence. For each time point, the unit voltage error, unit frequency error, and unit time offset error are multiplied by the corresponding weight ratio to generate the weighted error value for each time point. The weighted error values ​​of all time points within each time window are accumulated, and the accumulated value is divided by the total number of time points contained in the time window to generate the anomaly score value of the current time window. All time windows of the fitted response sequence are traversed, and all anomaly score values ​​are arranged in chronological order to form an anomaly score sequence, which serves as the simulated repair score result of the anomaly path segment.

7. The power grid security situation awareness method according to claim 6, characterized in that: In S4, a joint judgment is made on the abnormal path set to determine whether the extreme value of the score in the simulated repair score sequence corresponding to the abnormal path is higher than the preset score boundary threshold, whether the unitized offset value of the abnormal path in the offset quantization map is higher than the preset response offset threshold, and whether the response frequency fluctuation amplitude of the abnormal path in the feedback data is higher than the preset frequency fluctuation threshold, as three judgment conditions. If all three conditions are met, the broken path group is output; otherwise, the retained path set is output.

8. The power grid security situation awareness method according to claim 7, characterized in that: In S5, the broken path group is removed from the debug path set, and the removal map is output. The paths that did not participate in the disturbance debugging in the normal propagation path set are introduced into the broken path gaps in the elimination map, structural connectivity and response safety verification are performed, and the repair path set is output. Each path in the repair path set is injected with the disturbance signal strength for adjustment, and compared with the main response feature set and offset quantization spectrum to select the path that can stably transmit the disturbance signal, and output a reliable path set.

Images