A power transmission line mutual inductance synchronous sampling intelligent filtering monitoring method and system

Abstract

The application discloses a kind of transmission line mutual inductance synchronous sampling intelligent filter monitoring method and system, it is related to transmission line state monitoring technical field, the application is by synchronous acquisition line three-phase voltage and current and mutual inductor temperature gradient and leakage flux density, into heat magnetic coupling transfer function generation phase shift distribution sequence, and accordingly execute fractional order delay interpolation resampling generation phase correction matrix;It is reconstructed into phase space evolution trajectory by delay coordinate mapping, according to node neighborhood manifold curvature cut-off random wideband noise, retain core attractor manifold;And it is converted into topological relationship diagram, calculates geodesic distance and state transition probability distribution connected edge weight, executes graph Laplace smoothing processing and generates synchronous measurement electric signal after reverse projection;Extract transient distortion feature vector and calculate its and historical benchmark set Mahalanobis distance generation electrical state deviation degree, when electrical state deviation degree is out of limit, generate abnormal early warning instruction, guarantee the accuracy of micro-fault monitoring.

Description

Technical Field

[0001] This invention relates to the field of power transmission line condition monitoring technology, and in particular to a method and system for intelligent filtering monitoring of power transmission line mutual inductance synchronous sampling. Background Technology

[0002] As the core hub of the power system, the operational status of transmission lines directly affects the safety and stability of the entire power grid. With the continuous expansion of the power grid and the increasing complexity of its topology, the requirements for power supply reliability and power quality are constantly rising. Transmission lines typically traverse complex and variable geographical environments, are exposed to harsh natural conditions for extended periods, and are susceptible to the impact of extreme weather and complex electromagnetic interference. Therefore, real-time, accurate, synchronous sampling and online monitoring of transmission line operating parameters such as voltage and current are fundamental to achieving holistic power grid awareness, preventing large-scale power outages, and ensuring the stable operation of the power system.

[0003] However, in the complex real-world power grid environment, existing monitoring and signal acquisition schemes face numerous technical bottlenecks. In the instrument transformer signal sampling stage, traditional instrument transformers experience changes in core permeability and parasitic parameters when subjected to sudden temperature changes and complex leakage magnetic field interference. This leads to significant physical phase shifts in multi-channel synchronous sampling, making it difficult to meet the strict time synchronization requirements of transient fault analysis. In the signal processing stage, signals acquired at the transmission line site are often mixed with broadband non-stationary electromagnetic noise. Existing filtering methods mostly employ fixed frequency band division or conventional time-domain smoothing operations, which cannot effectively remove high-frequency nonlinear noise overlapping with transient characteristic frequency bands. While filtering out interference, this can easily cause the loss of effective transient change characteristics or waveform distortion, leading to false alarms or missed alarms in subsequent assessments of minor line hazards. Therefore, it is still necessary to provide an intelligent filtering monitoring method for synchronous sampling of transmission line instrument transformers to solve the problems of sampling phase drift and non-stationary strong noise interference in complex physical environments, ensuring the accuracy of micro-fault monitoring. Summary of the Invention

[0004] In view of the above-mentioned prior art, the present invention provides a method and system for intelligent filtering monitoring of synchronous sampling of mutual inductance in transmission lines, which mainly solves the technical problems existing in the background art.

[0005] To achieve the above objectives, the technical solution of this invention is implemented as follows: In a first aspect, the present invention provides a method for intelligent filtering monitoring of synchronous sampling of mutual inductance in transmission lines, the method comprising the following steps: S1. Collect the three-phase voltage and current sequence of the transmission line and the temperature gradient and leakage flux density of the transformer, and substitute them into the thermomagnetic coupling transfer function to obtain the phase shift distribution sequence. Based on the phase shift distribution sequence, perform fractional delay interpolation resampling on the three-phase voltage and current sequence to obtain the phase correction matrix. S2. The phase correction matrix is ​​reconstructed into a phase space evolution trajectory through delayed coordinate mapping. The curvature of the neighborhood manifold of each node in the phase space evolution trajectory is calculated. Based on the curvature of the neighborhood manifold, nodes belonging to random broadband noise components are removed, and the core attractor manifold corresponding to the fundamental harmonic components is retained. S3. Convert the core attractor manifold into a topological graph, calculate the geodesic distance and state transition probability between nodes and assign connected edge weights, perform graph Laplace smoothing on the topological graph according to the connected edge weights, and project it inversely to a one-dimensional time domain to obtain the synchronous measurement electrical signal. S4. Extract the transient distortion feature vector of the synchronously measured electrical signal, calculate the Mahalanobis distance between the transient distortion feature vector and the historical steady-state feature benchmark set, and obtain the electrical state deviation. When the electrical state deviation is greater than the dynamic health baseline threshold, generate a transmission line abnormal monitoring and early warning command.

[0006] As a preferred embodiment of the present invention, the specific process of substituting the temperature gradient and leakage flux density into the thermomagnetic coupling transfer function to calculate the physical phase shift at the sampling point and obtain the phase shift distribution sequence in step S1 is as follows: Extract the temperature gradient amplitude and leakage flux density vector components in the spatial coordinate system on the surface of the transformer core, and combine the temperature gradient amplitude and leakage flux density with the corresponding vector components to construct an environmental disturbance vector. The environmental disturbance vector is input into the preset thermomagnetic coupling transfer function, and the phase shift angle at the corresponding sampling time is calculated by the nonlinear mapping operator of the thermomagnetic coupling transfer function. By splicing the phase shift angles in chronological order of sampling time, a phase shift distribution sequence is obtained.

[0007] As a preferred embodiment of the present invention, the specific process of performing fractional-order delay interpolation resampling on the three-phase voltage and current sequence according to the phase shift distribution sequence to obtain the phase correction matrix is ​​as follows: Extract the time delay corresponding to each sampling point in the phase shift distribution sequence, separate the integer delay part and the non-integer delay part of the time delay, and dynamically adjust the tap coefficients of the finite-length impulse response filter based on the non-integer delay part to obtain a set of fractional interpolation coefficients. The three-phase voltage and current sequences are subjected to time-domain convolution interpolation based on the fractional interpolation coefficient set to obtain a synchronously aligned state sequence. The synchronously aligned state sequences of multiple channels are combined to obtain a phase correction matrix.

[0008] As a preferred embodiment of the present invention, the specific process of reconstructing the phase correction matrix into the phase space evolution trajectory through delayed coordinate mapping in step S2 is as follows: Extracting the mutual information minimum points of the phase correction matrix determines the phase space delay time variable; The false nearest neighbor criterion is used to identify the projection folding rate of the phase correction matrix in different dimensions, and the dimension value corresponding to the projection folding rate being lower than the set limit is used as the embedded dimension variable. The phase correction matrix is ​​expanded by coordinate delay based on the phase space delay time variable and the embedded dimension variable to construct a multidimensional state vector set. The multidimensional state vector set is then spliced ​​together in chronological order to obtain the phase space evolution trajectory.

[0009] As a preferred embodiment of the present invention, in step S2, the specific process of calculating the neighborhood manifold curvature of each node in the phase space evolution trajectory, and cutting off nodes belonging to random broadband noise components based on the neighborhood manifold curvature while retaining the core attractor manifold corresponding to the fundamental harmonic components is as follows: Extract the target node from the phase space evolution trajectory and search for the K nearest neighbors of the target node. Combine the target node with the K nearest neighbors to construct the local spatial covariance matrix. Singular value decomposition is performed on the local spatial covariance matrix to extract the eigenvector direction corresponding to the minimum singular value and the curvature of the neighborhood manifold is inferred based on the minimum singular value. When the curvature of the neighborhood manifold is greater than the smoothness classification boundary, the corresponding target node is marked as a random broadband noise component; when the curvature of the neighborhood manifold is less than or equal to the smoothness classification boundary, the corresponding target node is marked as a fundamental harmonic component. Remove nodes labeled as random broadband noise components from the phase space evolution trajectory, while retaining the core attractor manifold corresponding to the fundamental harmonic components.

[0010] As a preferred embodiment of the present invention, the specific process of converting the core attractor manifold into a topological graph in step S3, calculating the geodesic distance and state transition probability between nodes in the topological graph, and assigning weights to connected edges is as follows: In the core attractor manifold, calculate the multidimensional Euclidean distance between any two nodes. When the multidimensional Euclidean distance between any two nodes is less than the neighborhood connection distance, establish a connected edge between the corresponding two nodes to generate a topological graph. Calculate the shortest path length between nodes along the connected edges of the topological graph and use the shortest path length as the geodesic distance; The geodesic distance is transformed into a spatial manifold probability by using a heat core distribution mapping to obtain the state transition probability, and the state transition probability is assigned to the corresponding connected edge as the connected edge weight.

[0011] As a preferred embodiment of the present invention, the specific process of performing graph Laplacian smoothing on the topological graph according to the connected edge weights in step S3, and then projecting it inversely to a one-dimensional time domain space to obtain the synchronous measurement electrical signal is as follows: The sum of the weights of the connected edges of each node in the topological graph is calculated to obtain the degree distribution matrix. The degree distribution matrix and the adjacency matrix of the topological graph are then combined to construct the graph Laplacian operator. The node coordinates of the topological graph are subjected to spatial manifold regularization iterative processing by the graph Laplacian operator, and the replacement node corresponding to the cut-off node is generated by interpolation. Extract the first-dimensional feature components of each node from the topological graph containing replacement nodes, and reorganize the first-dimensional feature components according to the original evolution sequence of the nodes to achieve inverse dimensionality reduction, thereby obtaining the synchronous measurement electrical signal.

[0012] As a preferred embodiment of the present invention, the specific process of extracting the transient distortion feature vector of the synchronously measured electrical signal within a set power frequency period window in step S4, and calculating the Mahalanobis distance between the transient distortion feature vector and the historical steady-state feature reference set to obtain the electrical state deviation is as follows: According to the set power frequency cycle window, the synchronous measurement electrical signal is time-series segmented to extract single-cycle time-domain waveform segments. The permutation entropy values ​​of the single-cycle time-domain waveform segments under different coarse-grained scales are calculated, and the permutation entropy values ​​under different coarse-grained scales are spliced ​​and combined to obtain the transient distortion feature vector. Obtain the set of historical feature vectors during fault-free operation as the historical steady-state feature benchmark set; The Mahalanobis distance is calculated from the transient distortion feature vector and the historical steady-state feature reference set, and then used as the electrical state deviation.

[0013] As a preferred embodiment of the present invention, the specific process of generating a transmission line anomaly monitoring and early warning command when the electrical state deviation is greater than the dynamic health baseline threshold in step S4 is as follows: Acquire historical electrical state deviation data sets during periods without abnormal alarms, fit a probability density distribution curve based on the historical electrical state deviation data sets, determine the deviation value boundary corresponding to the preset cumulative probability on the probability density distribution curve, and calibrate the deviation value boundary as the dynamic health baseline threshold. By comparing the real-time measured electrical state deviation with the dynamic health baseline threshold, when the real-time measured electrical state deviation is greater than the dynamic health baseline threshold, the over-limit triggering time parameter and transient distortion feature vector are encapsulated to generate a transmission line abnormality monitoring and early warning command.

[0014] Secondly, the present invention also provides a transmission line mutual inductance synchronous sampling intelligent filtering monitoring system, applied to the aforementioned transmission line mutual inductance synchronous sampling intelligent filtering monitoring method, the system comprising: The phase correction module is used to collect the three-phase voltage and current sequence of the transmission line and the temperature gradient and leakage flux density of the transformer, substitute them into the thermomagnetic coupling transfer function to obtain the phase shift distribution sequence, and perform fractional delay interpolation and resampling on the three-phase voltage and current sequence according to the phase shift distribution sequence to obtain the phase correction matrix. The manifold noise reduction module is used to reconstruct the phase correction matrix into a phase space evolution trajectory through delayed coordinate mapping, calculate the neighborhood manifold curvature of each node in the phase space evolution trajectory, and cut off the nodes belonging to the random broadband noise component according to the neighborhood manifold curvature, while retaining the core attractor manifold corresponding to the fundamental harmonic component. The topology reconstruction module is used to convert the core attractor manifold into a topological graph, calculate the geodesic distance and state transition probability between nodes and assign connected edge weights, perform graph Laplace smoothing on the topological graph according to the connected edge weights, and project it inversely to a one-dimensional time domain space to obtain the synchronous measurement electrical signal. The status early warning module is used to extract the transient distortion feature vector of the synchronously measured electrical signal, calculate the Mahalanobis distance between the transient distortion feature vector and the historical steady-state feature benchmark set, obtain the electrical status deviation, and generate a transmission line abnormal monitoring early warning command when the electrical status deviation is greater than the dynamic health baseline threshold.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) This invention acquires the temperature gradient and leakage flux density of the transformer and substitutes them into the thermomagnetic coupling transfer function. After calculating the physical phase shift, fractional-order delay interpolation resampling is performed to achieve precise alignment of the time axis of the multi-channel signal. Then, the corrected signal is reconstructed into the phase space evolution trajectory through delay coordinate mapping. The curvature of the neighborhood manifold is used as the discrimination criterion to directly remove non-stationary broadband noise in the geometric dimension. Thus, the hardware-level sampling phase shift problem caused by the external physical environment is solved, and the defect of traditional frequency domain filtering that cannot separate non-stationary noise overlapping with the frequency band of transient signals is overcome. The absolute alignment of multi-channel synchronous sampling is guaranteed at the source, and the precise removal of complex electromagnetic interference is achieved at the multi-dimensional phase space level.

[0016] (2) This invention converts the denoised core attractor manifold into a topological graph, assigns connected edge weights based on geodesic distance and state transition probability, repairs trajectory breakpoints through graph Laplace smoothing, and then projects it backward to reconstruct a continuous synchronous measurement electrical signal. Furthermore, it extracts the multi-scale permutation entropy features within the power frequency period window and uses Mahalanobis distance to measure the deviation of the current state from the historical steady-state benchmark. This solves the problem of incomplete effective signals and temporal breaks caused by noise removal, losslessly reconstructs a complete and detailed real electrical signal, and enables highly sensitive deviation monitoring of the microscopic anomalies of transmission lines based on this real electrical signal, effectively avoiding the risk of missed detection of hidden minor faults by conventional threshold alarms. Attached Figure Description

[0017] Figure 1 This is a flowchart of an intelligent filtering monitoring method for synchronous sampling of mutual inductance in transmission lines according to the present invention. Figure 2 This is a schematic diagram of the phase space evolution trajectory and manifold noise reduction scatter plot of the present invention; Figure 3 This is a comparative schematic diagram of the reconstruction of electrical signals from one-dimensional time-domain synchronous measurement according to the present invention; Figure 4 This is a structural diagram of a synchronous sampling intelligent filtering monitoring system for mutual inductance in power transmission lines according to the present invention. Detailed Implementation

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. In the following description, the expression "some embodiments" describes a subset of all possible embodiments; however, it should be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with each other without conflict.

[0019] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.

[0020] It should be understood that the present invention can be embodied in various forms and should not be construed as being limited to the embodiments set forth herein. Rather, providing these embodiments will make the disclosure thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Furthermore, the terminology used herein is intended only to describe particular embodiments and is not intended to limit the invention. When used herein, the singular forms “a,” “an,” and “the” are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms “compose” and / or “comprising,” when used in this specification, identify the presence of the stated features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups. When used herein, the term “and / or” includes any and all combinations of the associated listed items.

[0021] It should also be noted that when an element is referred to as being "fixed to" another element, it can be directly attached to the other element or there may be an intervening element. When an element is referred to as being "connected to" another element, it can be directly connected to the other element or there may be an intervening element. The terms "vertical," "horizontal," "inner," "outer," "left," "right," and similar expressions used herein are for illustrative purposes only and do not represent the only possible implementation.

[0022] To fully understand this invention, a detailed structure will be presented in the following description to illustrate the technical solution proposed by this invention. Optional embodiments of the invention are described in detail below; however, in addition to these detailed descriptions, the invention may have other embodiments.

[0023] Firstly, this invention provides a method for synchronous sampling and intelligent filtering monitoring of mutual inductance in transmission lines. Please refer to the attached document for details. Figure 1 The method includes the following steps: S1. Collect the three-phase voltage and current sequences of the transmission line and the temperature gradient and leakage flux density of the transformer, and substitute them into the thermomagnetic coupling transfer function to obtain the phase shift distribution sequence. Based on the phase shift distribution sequence, perform fractional delay interpolation and resampling on the three-phase voltage and current sequences to obtain the phase correction matrix.

[0024] As a preferred embodiment of the present invention, the specific process for obtaining the phase shift distribution sequence is as follows: The absolute temperature difference between the inside and outside of the transformer core is directly acquired by attaching a hardware sensing unit to the transformer housing to extract the temperature difference amplitude. A spatially arranged three-axis Hall sensor array records the magnetic flux density in three orthogonal dimensions, directly forming vector components. The extracted temperature difference amplitude and vector components are then sequentially concatenated according to a fixed array dimension to form a one-dimensional environmental disturbance vector.

[0025] Due to the combined effects of external temperature rise and stray leakage magnetic field on the permeability and parasitic capacitance of the core material, nonlinear coupling drift characteristics will occur. To compensate for the impact of this drift on sampling accuracy, this embodiment constructs a preset thermomagnetic coupling transfer function. The constructed thermomagnetic coupling transfer function takes the environmental interference vector as input and the sampling phase shift angle as output. This transfer function essentially constructs a multidimensional nonlinear mapping relationship, which covers the temperature-independent action component, the magnetic field-independent action component, and the cross-coupling action component of temperature and magnetic field, thereby achieving accurate characterization of phase shift under complex environmental interference. The mapping relationship between the preset thermomagnetic coupling transfer function and the phase shift angle is shown in the formula:

[0026] in, This represents the output phase offset angle at any sampling time. This represents a nonlinear mapping operator from the environmental disturbance vector space to the phase offset scalar space. This represents the environmental interference vector input at any sampling time.

[0027] Its solution corresponds to the first The specific formula for calculating the phase offset angle at each sampling time is as follows:

[0028] In the formula, Indicates the first The phase offset angle of the output is calculated at each sampling time. The global index number representing the discrete sampling time; Indicates the first An environmental disturbance vector constructed by combining sampling times; This represents the weighted row vector of the first-order thermomagnetic physical response; This represents the exponentially decaying mapping operator with the natural constant as its base; This represents the decay diagonal matrix characterizing the temperature magnetic field saturation limit of the iron core; Represents the transpose of the environmental disturbance vector; A second-order nonlinear Hessian matrix characterizing the intensity of the physical interaction between temperature and leakage magnetic field cross-coupling.

[0029] Among them, for the second-order nonlinear Hessian matrix The determination method is as follows: During the factory hardware calibration stage, a high and low temperature alternating test chamber and a programmable Helmholtz coil are used to jointly apply a limiting step temperature and a distorted magnetic field to the transformer body. The equivalent impedance test data under all physical conditions are continuously recorded. The partial second derivatives of the equivalent impedance test data with respect to the temperature variable and the magnetic field strength variable are calculated to obtain the estimated value of the local Hessian matrix under each calibration condition. Then, the estimated values ​​under all calibration conditions are weighted and averaged to obtain the second-order nonlinear Hessian matrix. .

[0030] After calculating and obtaining the phase shift angles for the entire time period, all phase shift angles are spliced ​​together according to the chronological order of sampling time to obtain a complete phase shift distribution sequence.

[0031] Furthermore, the specific process of performing fractional-order delay interpolation resampling on the three-phase voltage and current sequences based on the phase shift distribution sequence to obtain the phase correction matrix in step S1 is as follows: The angle values ​​in the phase shift distribution sequence are equivalently converted into actual time delays based on the rated reference angular frequency of the power grid system. The conversion principle is based on the correspondence between sinusoidal wave phase and time. For a 50 Hz power frequency signal, a complete 360-degree phase cycle corresponds to a time length of 0.02 seconds. Therefore, the time delay corresponding to any phase shift angle is equal to the ratio between that phase shift angle and the corresponding angular frequency. For example, if the phase shift angle calculated for a certain sampling point is 0.1 radians, the corresponding time delay is approximately 0.000318 seconds (i.e., 318 microseconds).

[0032] The time delay is rounded down to remove the integer delay portion. The remaining fractional value spanning two sampling points is defined as the non-integer delay portion. For example, assuming a sampling period of 50 microseconds, dividing the above time delay (318 microseconds) by the sampling period of 50 microseconds yields a delay value of 6.36 sampling periods. Considering the discrete nature of the signal acquired by the analog-to-digital converter, the integer delay portion of 6.36 is rounded down to remove the integer delay portion, i.e., the integer part 6 represents an integer delay of 6 complete sampling periods; while retaining the fractional part 0.36, this fractional part is defined as the non-integer delay portion, indicating that an additional subsampling interval of less than one sampling period is required.

[0033] The integer delay portion is directly used to align the sampling index of the digital sequence, while the non-integer delay portion requires amplitude reconstruction of subsampled points through digital filtering. The center offset of the sine mapping function is then dynamically adjusted using the non-integer delay portion to generate fractional interpolation coefficients suitable for dynamic calibration at the current time. The specific formulas for constructing the tap coefficients and performing temporal convolutional interpolation are as follows:

[0034]

[0035] In the formula, Indicates the first The analysis point corresponding to the first Fractional interpolation coefficients for each tap position; Indicates the temporal position sequence number of the target signal reconstruction point; Indicates the internal tap position number of a finite-length impulse response filter; Indicates the first Each analysis point corresponds to the separated non-integer delay portion; Represents the sine mapping function; Pi is a constant. Represents the cosine mapping function; The parameter representing the total tap window length of a finite-length impulse response filter.

[0036] In the formula, This indicates that the synchronization-aligned state sequence output after temporal convolution interpolation is at the [number]th [time]. The accurate amplitude at each analysis point; This indicates that the index number in the original three-phase voltage and current sequence is The actual sampling amplitude at the location; Indicates the first Each analysis point corresponds to an integer delay portion, the value of which is the integer part of the time delay corresponding to the phase shift distribution sequence, used to shift the entire signal by an integer number of sampling periods.

[0037] Among them, the key parameter is the total tap window length parameter. The method for determining the frequency is as follows: the ratio of the highest effective transient harmonic frequency excited when a single-phase grounding fault occurs in the transmission line to the actual sampling frequency of the hardware is calculated, and the frequency is determined by looking up a table in combination with the lower limit standard of the stopband attenuation amplitude set in the industrial field, so as to ensure that the main lobe width during interpolation calculation completely covers the frequency band of the transient distortion signal to be monitored.

[0038] Because the physical installation locations of each instrument transformer on the transmission line differ, the temperature gradient distribution and leakage magnetic field strength they experience also vary. Therefore, the calculated phase shift distribution sequence, integer delay portion, and non-integer delay portion are all different. If a uniform delay correction amount is used, the relative timing misalignment introduced by the physical location differences between channels cannot be eliminated. Therefore, after independently interpolating the voltage and current of each phase in the three-phase voltage and current sequence to generate separate synchronously aligned state sequences, multi-dimensional column vectors are concatenated to finally construct a phase correction matrix that eliminates the timing misalignment between channels.

[0039] S2. The phase correction matrix is ​​reconstructed into a phase space evolution trajectory through delayed coordinate mapping. The curvature of the neighborhood manifold of each node in the phase space evolution trajectory is calculated. Based on the curvature of the neighborhood manifold, nodes belonging to random broadband noise components are removed, and the core attractor manifold corresponding to the fundamental harmonic components is retained.

[0040] Furthermore, the specific process of reconstructing the phase correction matrix into the phase space evolution trajectory through delayed coordinate mapping in step S2 is as follows: The purpose of reconstructing the phase correction matrix into a phase space evolution trajectory is to map the implicit transient dynamic characteristics of the power grid in the one-dimensional time series signal to a high-dimensional geometric space, so as to facilitate the subsequent identification and removal of noise components based on manifold curvature.

[0041] First, the mutual information function values ​​of the single-channel time series in the phase correction matrix are calculated under different time lags. As the lag gradually increases from zero, the mutual information function exhibits an oscillating decay trend. The lag corresponding to the first minimum value of the mutual information function is selected as the phase space delay time variable. The purpose of this selection method is that, under this lag, the information redundancy between adjacent state points is minimized, avoiding both excessive concentration of coordinate points in the diagonal region due to excessively small delays and loss of causal correlation between state points due to excessively large delays.

[0042] Simultaneously, the spurious nearest neighbor criterion is used to calculate the folding and unfolding state of spatial projection as the dimension increases. The basic principle is: project the phase correction matrix onto the dimension respectively. With dimension In the space, statistics in dimension Next adjacent but in dimension The proportion of points that are clearly separated is called the projection folding rate (projection folding rate refers to the phenomenon that points of different trajectories overlap or approach each other in a low-dimensional space due to insufficient spatial projection dimension; its quantitative value is the proportion of folded points to the total number of points). As the projection folding rate continuously decreases with increasing dimension, it reaches a certain point in a certain dimension... When the fold rate first falls below the preset limit, it is determined that the pseudo-adjacency relationships in the high-dimensional space projection have been sufficiently eliminated. At this point, the dimension is... Fixed as embedded dimension variable , The folding rate limit is set based on the following: During the factory calibration phase of the current transformer, a multi-dimensional projection folding rate test is performed on a noise-free, clean power frequency reference signal. As the dimension increases, the folding rate gradually decreases and tends to stabilize. When the relative change in folding rate after continuously increasing the dimension is lower than a preset threshold, the current folding rate value is recorded. After repeating the test multiple times, the average value of the recorded folding rate values ​​is taken as the set limit. When the folding rate is lower than the set limit, the current test dimension value is used as an embedded dimension variable.

[0043] Then, based on the phase space delay time variable and the embedding dimension variable, the phase correction matrix is ​​expanded by coordinate delay to construct a multidimensional state vector set. The specific formula for expanding by coordinate delay to construct the multidimensional state vector is as follows:

[0044] In the formula, Indicates the first A multidimensional state vector; Incrementing index representing the state point of temporal evolution; Indicates the phase correction matrix at index The signal sampling amplitude at that location; This represents the phase space delay time variable determined based on the minimum value of mutual information; This represents the embedded dimension variable defined by the projection folding rate, and its value is a positive integer; This represents the corresponding signal sampling amplitude extracted based on the maximum delay span.

[0045] Finally, the multidimensional state vectors are arranged sequentially according to the time index to form the phase space evolution trajectory.

[0046] Furthermore, in step S2, the specific process of calculating the neighborhood manifold curvature of each node in the phase space evolution trajectory, and removing nodes belonging to random broadband noise components while retaining the core attractor manifold corresponding to the fundamental harmonic components based on the neighborhood manifold curvature is as follows: The target node is extracted from the phase space evolution trajectory, and its K nearest neighbors are searched. The target node and its K nearest neighbors are combined to construct a local spatial covariance matrix. The specific formula for constructing the local spatial covariance matrix is ​​as follows:

[0047] In the formula, This represents the local spatial covariance matrix corresponding to the target node; This indicates the set number of nearest neighbor nodes to search; Indicates the local traversal sequence number of the nearest neighbor node; Represents the first element in the local space where the target node is located. Nearest neighbor state vectors; This represents the centroid coordinate vector within the local space, which contains the target node and all its nearest neighbor state vectors. The transpose operator for a matrix or vector; This represents the left singular orthogonal eigenvalue matrix obtained from the decomposition; This represents a diagonal eigenmatrix with singular value parameters arranged on its diagonal. This represents the transpose of the right singular orthogonal characteristic matrix.

[0048] Singular value decomposition is performed on the local spatial covariance matrix to extract the eigenvector directions corresponding to the minimum singular values. The curvature of the neighborhood manifold is then calculated based on these minimum singular values. The neighborhood manifold curvature measures the degree of curvature of the local geometric surface formed by the target node and its neighboring nodes in space; a larger value indicates a rougher and less smooth local surface. The specific formula for calculating the neighborhood manifold curvature based on the minimum singular values ​​is as follows:

[0049] In the formula, This represents the curvature of the neighborhood manifold of the target node, with a value ranging from 0 to 1, and is used to characterize the degree of curvature of the local manifold surface. Represents the diagonal characteristic matrix The smallest non-zero singular value parameter in; Indicates the dimension iteration index; Represents the diagonal characteristic matrix The Middle Each diagonal singular value parameter.

[0050] After calculating the curvature of the neighborhood manifold for each target node, it is compared with a preset smoothness classification boundary: if the curvature of the neighborhood manifold is greater than the preset smoothness classification boundary, the node is determined to belong to a random broadband noise component and is marked; otherwise, the node is determined to belong to a fundamental harmonic component and is retained. After traversing all nodes, all nodes marked as noise components are deleted from the phase space evolution trajectory, and the geometry formed by the remaining nodes is the core attractor manifold.

[0051] The method for determining the smoothness classification boundary is as follows: extract the phase space evolution trajectory of the transmission line under long-term operation without partial discharge and corona interference, calculate the curvature of the neighborhood manifold of all nodes in the trajectory, draw the probability density distribution map, and extract the curvature value corresponding to the cumulative probability distribution area reaching 99% as the smoothness classification boundary.

[0052] Please refer to the following: Figure 2The figure visually illustrates the physical geometric distribution of a one-dimensional time-series signal in a multi-dimensional spatial coordinate system after reconstruction by delayed coordinate mapping. Specifically, the dense, continuous, and regularly evolving geometric lattice entities in the figure represent the extracted core attractor manifold, which maps the nonlinear dynamic trajectories of the deterministic power frequency fundamental wave and transient effective harmonic components in the transmission line. The scattered geometric points outside the core attractor manifold, exhibiting a disordered and discrete state, represent random broadband noise components generated by external electromagnetic interference.

[0053] Using the curvature of the neighborhood manifold obtained from the local spatial covariance matrix as the separation boundary, random broadband noise components are removed, thus physically stripping away complex broadband noise from a multi-dimensional spatial geometric dimension without disrupting the core attractor manifold.

[0054] S3. Convert the core attractor manifold into a topological graph, calculate the geodesic distance and state transition probability between nodes and assign connected edge weights, perform graph Laplace smoothing on the topological graph according to the connected edge weights, and project it inversely to a one-dimensional time domain to obtain the synchronous measurement electrical signal.

[0055] Furthermore, the specific process of converting the core attractor manifold into a topological graph in step S3, calculating the geodesic distances and state transition probabilities between nodes in the topological graph, and assigning weights to connected edges is as follows: When quantifying the connectivity between geometric nodes in a multidimensional phase space, the first step is to calculate any spatial node in the core attractor manifold. With nodes Multidimensional Euclidean distance between The specific calculation formula is as follows:

[0056] in, For dimension index variables, For nodes exist Coordinate components in dimensional space, For nodes exist Coordinate components in the dimension.

[0057] Multidimensional Euclidean distance Distance from preset neighboring regions Numerical comparison was performed when the Euclidean distance was... Less than the preset neighborhood connection distance When two nodes are physically adjacent on a local manifold, an undirected connected edge is constructed between them, and a basic topological graph is generated after traversing all nodes.

[0058] Wherein, the neighborhood connection distance The method for determining the distance is as follows: The lower quartile of the Euclidean distance between all nodes in the core attractor manifold corresponding to the standard transient waveform without electromagnetic interference collected during the factory calibration phase is statistically analyzed. This lower quartile value is then used as the neighborhood connectivity distance. .

[0059] After establishing undirected connected edges, the nodes in the topological graph are calculated using the underlying graph traversal shortest path algorithm. With nodes The shortest actual path length measured along the surface of the core attractor manifold is used as the geodesic distance. The specific calculation formula is as follows:

[0060] in, Represents nodes in a topological graph With nodes The set of all possible paths, Represents any path in this set. Indicates in path upper connection node With nodes The sum of the Euclidean distances of all connected edges is the total length of the path.

[0061] Then, geodesic distance is mapped through the thermonuclear distribution. Perform spatial manifold probability transformation to map and generate state transition probabilities that characterize the tendency of the power grid's evolution state to shift and evolve on the core attractor manifold. The state transition probability is directly assigned as the weight of the corresponding connected edge, and the specific calculation formula is as follows:

[0062] In the formula, This represents the calculated state transition probability and the weights of the connected edges between corresponding nodes in the topological graph; Represents the natural constant exponent operator; Represents a node With nodes The geodesic distance calculated from the time interval; The bandwidth scale parameter represents the mapping of the thermonuclear distribution.

[0063] Among them, the bandwidth scale parameter The method for determining the value is as follows: extract the set of geodesic distances of all node pairs with connected edges in the current topology graph, calculate the global numerical variance of the set, and assign the variance value as a global scale constant to the bandwidth scale parameter.

[0064] Furthermore, in step S3, the topological graph is smoothed using a graph Laplacian smoothing process based on the weights of the connected edges. The smoothed topological graph is then inversely projected onto a one-dimensional time domain to obtain the synchronous measurement electrical signal. The specific process is as follows: After the weighted construction of the topology graph is completed, the nodes marked as random broadband noise components were removed in step S2, resulting in data breaks in the phase space trajectory at the corresponding locations. Manifold repair is required for these broken regions. Specifically, a symmetric weight distribution matrix is ​​constructed by extracting the global connected edge weight information of the topology graph. The weight distribution matrix is ​​accumulated row by row. The diagonal elements within the matrix yield the degree distribution matrix. The main diagonal elements The degree distribution matrix With weight distribution matrix Perform matrix subtraction operations to construct the graph Laplacian operator. That is, the operation expression is The node coordinates of the topological graph are subjected to spatial manifold regularization iterative processing using the graph Laplacian operator. Then, the node coordinate vectors in the fractured region are forced to converge towards the physical centroid of their high-weight neighbor nodes. After multiple iterations, the coordinates of the cut node are interpolated from the coordinates of the surrounding retained nodes to generate a replacement node at the corresponding location of the cut node, thus repairing the fracture trajectory. The specific interpolation iterative repair formula is as follows:

[0065] In the formula, Indicates the first The full node multidimensional coordinate feature matrix after the next iteration step; Indicates the current iteration number; The regularization scaling factor represents the step size for smooth convergence of the spatial manifold. Its value ranges from 0 to 1 and is used to control the step size for the node coordinates to converge toward the center in each iteration. Indicates the first The full node multidimensional coordinate feature matrix after the next iteration step; The inverse of the degree distribution matrix is ​​represented by the matrix inverse.

[0066] Wherein, the regularization scaling factor The determination method is as follows: retrieve the remaining available floating-point operation cycle constant of the edge monitoring terminal's main control chip, and obtain a dynamically convergent value in the interval between 0 and 1 through inverse proportional mapping matching as the regularization scaling factor. .

[0067] After the repair is completed, all nodes in the topology graph now have complete multidimensional coordinate information. This information needs to be restored to a one-dimensional time-domain signal to support subsequent transient feature extraction. Specifically, the restoration method involves extracting the first element of each multidimensional coordinate state vector (i.e., the first-dimensional feature component) from the complete topology graph containing the replacement nodes. This element is then rearranged according to the original sampling time sequence of each node, completing the dimensionality reduction and outputting a synchronous measurement electrical signal free of any high-frequency glitches or broken regions.

[0068] Please refer to the following: Figure 3 The figure shows the comparison of one-dimensional inverse projection of time-series data before and after the spatial node removal operation. The light-colored undulating curve with many high-frequency spikes and local abrupt changes represents the original noisy acquisition signal without graph Laplacian smoothing and the time-domain incomplete signal generated after removing noisy nodes; while the dark solid curve that overlaps with it, showing a smooth transition and continuous waveform, represents the synchronous measurement electrical signal after removing high-frequency spikes and broken areas.

[0069] Depend on Figure 3 The comparative distribution shows that by combining the graph Laplacian operator with the state transition probability, the spatial manifold regularization iteration process was performed on the node coordinates of the fracture region, and the corresponding replacement nodes were generated by interpolation. This not only repaired the data breakpoints left by the cutting noise on the time domain scale, but also effectively filtered out the free broadband interference, and completely preserved the true low-frequency fundamental wave and transient distortion profile of the transmission line.

[0070] S4. Extract the transient distortion feature vector of the synchronously measured electrical signal, calculate the Mahalanobis distance between the transient distortion feature vector and the historical steady-state feature benchmark set, and obtain the electrical state deviation. When the electrical state deviation is greater than the dynamic health baseline threshold, generate a transmission line abnormal monitoring and early warning command.

[0071] Furthermore, in step S4, the transient distortion feature vector of the synchronously measured electrical signal is extracted, and the Mahalanobis distance between the transient distortion feature vector and the historical steady-state feature reference set is calculated to obtain the electrical state deviation. The specific process is as follows: After acquiring the synchronous measurement electrical signal, it is necessary to extract quantitative feature parameters that can sensitively reflect minute changes in the transmission line's state. These quantitative feature parameters include transient distortion feature vectors. This invention employs a multi-scale permutation entropy method to extract these transient distortion feature vectors. Specifically, the synchronous measurement electrical signal is time-sequentially segmented according to a set power frequency cycle window, extracting multiple independent single-cycle time-domain waveform segments. For each single-cycle time-domain waveform segment, mean downsampling is performed sequentially at multiple preset coarse-grained scales to obtain coarse-grained sequences corresponding to each scale. Phase space delay reconstruction is performed on each coarse-grained sequence, and the probability distribution of its element amplitude ascending order is statistically analyzed. The corresponding information entropy is calculated, thus obtaining the permutation entropy value at that scale. The permutation entropy values ​​at all preset scales are concatenated and combined in ascending order of scale to generate the transient distortion feature vector. The specific formula for calculating the corresponding information entropy is as follows:

[0072] in, Indicates the first The permutation entropy value corresponding to the coarse-grained sequence. Represents the set of all coarse-grained sequences. Indicates the first The probability of a coarse-grained sequence appearing in an ascending order.

[0073] Subsequently, the set of historical feature vectors during fault-free operation is obtained to construct a historical steady-state feature benchmark set; To eliminate interference from the correlation between numerical magnitudes and data dimensions among features of different scales, the Mahalanobis distance is calculated between the historical steady-state feature benchmark set and the transient distortion feature vector. The specific calculation formula is as follows:

[0074] In the formula, This indicates the degree of deviation of the calculated electrical state; the larger the value, the greater the degree of deviation of the current operating state from the normal operating condition. This represents the transient distortion feature vector extracted from the current single-cycle time-domain waveform segment; This represents the mean center distribution vector corresponding to the historical steady-state characteristic benchmark set; This represents the vector transpose operation; This represents the inverse matrix of the covariance distribution matrix corresponding to the historical steady-state characteristic benchmark set. By calculating the weighted projection of the inverse matrix of this covariance distribution matrix, the high-dimensional spatial geometric distance between the current transmission line transient waveform and the center of the normal operating condition baseline distribution can be objectively calculated, thereby quantifying the true degree of electrical state deviation.

[0075] Furthermore, the specific process for generating a transmission line anomaly monitoring and early warning command in step S4 when the electrical state deviation exceeds the dynamic health baseline threshold is as follows: After obtaining the real-time electrical status deviation, a dynamic baseline value needs to be set as the boundary between normal and abnormal conditions. This invention uses a probability density distribution function to fit the deviation distribution under historical normal conditions and determines the dynamic health baseline threshold accordingly. Specifically, it periodically extracts historical normal electrical status deviation data sets that have not experienced any abnormal alarm actions within a continuous rolling time window, and uses a probability density distribution function to fit the true probability density distribution curve of this data set. The specific calculation formula is as follows:

[0076] In the formula, This represents the probability density distribution function generated by the fit; The independent variable representing the degree of deviation from the electrical state; This indicates the total number of samples within the extracted historical normal electrical condition deviation data group; The smoothing bandwidth parameter represents the smoothness of the kernel function; The variable representing the traversal index of historical samples; This indicates the Gaussian kernel operator used; This represents the first data group in the historical normal electrical condition deviation data set. The actual measured value of the deviation of the historical electrical condition.

[0077] Wherein, the smooth bandwidth parameter The method for determining the standard deviation is as follows: extract the sample standard deviation of the historical normal electrical condition deviation data group, and multiply the sample standard deviation by the total number of samples as the smoothing bandwidth parameter. After fitting the probability density distribution curve, the independent variable along the electrical state deviation... The cumulative probability density distribution function is solved by the integral axis. The value of the upper limit of the integral variable corresponding to the integral area reaching the preset cumulative probability is used as the current dynamic health baseline threshold, wherein the preset cumulative probability is 99%.

[0078] In the real-time monitoring phase, the deviation of the electrical state calculated in real time is continuously compared with the dynamic health baseline threshold. When a limit is exceeded, the time information corresponding to the limit exceeding event, such as the end time of the power frequency cycle and the transient distortion feature vector calculated for the cycle, is used as associated data, encapsulated according to the preset communication data frame format, and combined to generate a transmission line abnormality monitoring and early warning instruction containing the limit exceeding time information and the transient distortion feature vector.

[0079] Secondly, the present invention also provides a transmission line mutual inductance synchronous sampling intelligent filtering monitoring system, applied to the aforementioned transmission line mutual inductance synchronous sampling intelligent filtering monitoring method. Please refer to [link to relevant documentation]. Figure 4 The system includes: The phase correction module is used to collect the three-phase voltage and current sequence of the transmission line and the temperature gradient and leakage flux density of the transformer, substitute them into the thermomagnetic coupling transfer function to obtain the phase shift distribution sequence, and perform fractional delay interpolation and resampling on the three-phase voltage and current sequence according to the phase shift distribution sequence to obtain the phase correction matrix. The manifold noise reduction module is used to reconstruct the phase correction matrix into a phase space evolution trajectory through delayed coordinate mapping, calculate the neighborhood manifold curvature of each node in the phase space evolution trajectory, and cut off the nodes belonging to the random broadband noise component according to the neighborhood manifold curvature, while retaining the core attractor manifold corresponding to the fundamental harmonic component. The topology reconstruction module is used to convert the core attractor manifold into a topological graph, calculate the geodesic distance and state transition probability between nodes and assign connected edge weights, perform graph Laplace smoothing on the topological graph according to the connected edge weights, and project it inversely to a one-dimensional time domain space to obtain the synchronous measurement electrical signal. The status early warning module is used to extract the transient distortion feature vector of the synchronously measured electrical signal, calculate the Mahalanobis distance between the transient distortion feature vector and the historical steady-state feature benchmark set, obtain the electrical status deviation, and generate a transmission line abnormal monitoring early warning command when the electrical status deviation is greater than the dynamic health baseline threshold.

[0080] The working principle of the system is the same as the implementation principle of the method described above, so it will not be repeated here.

[0081] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for synchronous sampling and intelligent filtering monitoring of mutual inductance in transmission lines, characterized in that, The method includes the following steps: S1. Collect the three-phase voltage and current sequence of the transmission line and the temperature gradient and leakage flux density of the transformer, and substitute them into the thermomagnetic coupling transfer function to obtain the phase shift distribution sequence. Based on the phase shift distribution sequence, perform fractional delay interpolation resampling on the three-phase voltage and current sequence to obtain the phase correction matrix. S2. The phase correction matrix is ​​reconstructed into a phase space evolution trajectory through delayed coordinate mapping. The curvature of the neighborhood manifold of each node in the phase space evolution trajectory is calculated. Based on the curvature of the neighborhood manifold, nodes belonging to random broadband noise components are removed, and the core attractor manifold corresponding to the fundamental harmonic components is retained. S3. Convert the core attractor manifold into a topological graph, calculate the geodesic distance and state transition probability between nodes and assign connected edge weights, perform graph Laplace smoothing on the topological graph according to the connected edge weights, and project it inversely to a one-dimensional time domain to obtain the synchronous measurement electrical signal. S4. Extract the transient distortion feature vector of the synchronously measured electrical signal, calculate the Mahalanobis distance between the transient distortion feature vector and the historical steady-state feature benchmark set, and obtain the electrical state deviation. When the electrical state deviation is greater than the dynamic health baseline threshold, generate a transmission line abnormal monitoring and early warning command.

2. The intelligent filtering monitoring method for synchronous sampling of mutual inductance in transmission lines according to claim 1, characterized in that, In step S1, the specific process of substituting the temperature gradient and leakage flux density into the thermomagnetic coupling transfer function to calculate the physical phase shift at the sampling point and obtain the phase shift distribution sequence is as follows: Extract the temperature gradient amplitude and leakage flux density vector components in the spatial coordinate system on the surface of the transformer core, and combine the temperature gradient amplitude and leakage flux density with the corresponding vector components to construct an environmental disturbance vector. The environmental disturbance vector is input into the preset thermomagnetic coupling transfer function, and the phase shift angle at the corresponding sampling time is calculated by the nonlinear mapping operator of the thermomagnetic coupling transfer function. By splicing the phase shift angles in chronological order of sampling time, a phase shift distribution sequence is obtained.

3. The intelligent filtering monitoring method for synchronous sampling of mutual inductance in transmission lines according to claim 1, characterized in that, In step S1, the specific process of performing fractional-order delay interpolation resampling on the three-phase voltage and current sequences based on the phase shift distribution sequence to obtain the phase correction matrix is ​​as follows: Extract the time delay corresponding to each sampling point in the phase shift distribution sequence, separate the integer delay part and the non-integer delay part of the time delay, and dynamically adjust the tap coefficients of the finite-length impulse response filter based on the non-integer delay part to obtain a set of fractional interpolation coefficients. The three-phase voltage and current sequences are subjected to time-domain convolution interpolation based on the fractional interpolation coefficient set to obtain a synchronously aligned state sequence. The synchronously aligned state sequences of multiple channels are combined to obtain a phase correction matrix.

4. The intelligent filtering monitoring method for synchronous sampling of mutual inductance in transmission lines according to claim 3, characterized in that, In step S2, the specific process of reconstructing the phase correction matrix into the phase space evolution trajectory through delayed coordinate mapping is as follows: Extracting the mutual information minimum points of the phase correction matrix determines the phase space delay time variable; The false nearest neighbor criterion is used to identify the projection folding rate of the phase correction matrix in different dimensions, and the dimension value corresponding to the projection folding rate being lower than the set limit is used as the embedded dimension variable. The phase correction matrix is ​​expanded by coordinate delay based on the phase space delay time variable and the embedded dimension variable to construct a multidimensional state vector set. The multidimensional state vector set is then spliced ​​together in chronological order to obtain the phase space evolution trajectory.

5. The intelligent filtering monitoring method for synchronous sampling of mutual inductance in transmission lines according to claim 4, characterized in that, In step S2, the curvature of the neighborhood manifold of each node in the phase space evolution trajectory is calculated, and nodes belonging to random broadband noise components are removed based on the curvature of the neighborhood manifold, while retaining the core attractor manifold corresponding to the fundamental harmonic components. The specific process is as follows: Extract the target node from the phase space evolution trajectory and search for the K nearest neighbors of the target node. Combine the target node with the K nearest neighbors to construct the local spatial covariance matrix. Singular value decomposition is performed on the local spatial covariance matrix to extract the eigenvector direction corresponding to the minimum singular value and the curvature of the neighborhood manifold is inferred based on the minimum singular value. When the curvature of the neighborhood manifold is greater than the smoothness classification boundary, the corresponding target node is marked as a random broadband noise component; when the curvature of the neighborhood manifold is less than or equal to the smoothness classification boundary, the corresponding target node is marked as a fundamental harmonic component. Remove nodes labeled as random broadband noise components from the phase space evolution trajectory, while retaining the core attractor manifold corresponding to the fundamental harmonic components.

6. The intelligent filtering monitoring method for synchronous sampling of mutual inductance in transmission lines according to claim 5, characterized in that, The specific process of converting the core attractor manifold into a topological graph in step S3, calculating the geodesic distance and state transition probability between nodes in the topological graph, and assigning weights to connected edges is as follows: In the core attractor manifold, calculate the multidimensional Euclidean distance between any two nodes. When the multidimensional Euclidean distance between any two nodes is less than the neighborhood connection distance, establish a connected edge between the corresponding two nodes to generate a topological graph. Calculate the shortest path length between nodes along the connected edges of the topological graph and use the shortest path length as the geodesic distance; The geodesic distance is transformed into a spatial manifold probability by using a heat core distribution mapping to obtain the state transition probability, and the state transition probability is assigned to the corresponding connected edge as the connected edge weight.

7. The intelligent filtering monitoring method for synchronous sampling of mutual inductance in transmission lines according to claim 6, characterized in that, In step S3, the graph Laplacian smoothing process is performed on the topological graph according to the connected edge weights, and then inversely projected to a one-dimensional time domain space to obtain the synchronous measurement electrical signal. The specific process is as follows: The sum of the weights of the connected edges of each node in the topological graph is calculated to obtain the degree distribution matrix. The degree distribution matrix and the adjacency matrix of the topological graph are then combined to construct the graph Laplacian operator. The node coordinates of the topological graph are subjected to spatial manifold regularization iterative processing by the graph Laplacian operator, and the replacement node corresponding to the cut-off node is generated by interpolation. Extract the first-dimensional feature components of each node from the topological graph containing replacement nodes, and reorganize the first-dimensional feature components according to the original evolution sequence of the nodes to achieve inverse dimensionality reduction, thereby obtaining the synchronous measurement electrical signal.

8. The intelligent filtering monitoring method for synchronous sampling of mutual inductance in transmission lines according to claim 1, characterized in that, The specific process of extracting the transient distortion feature vector of the synchronous measurement electrical signal within a set power frequency period window in step S4, and calculating the Mahalanobis distance between the transient distortion feature vector and the historical steady-state feature reference set to obtain the electrical state deviation is as follows: According to the set power frequency cycle window, the synchronous measurement electrical signal is time-series segmented to extract single-cycle time-domain waveform segments. The permutation entropy values ​​of the single-cycle time-domain waveform segments under different coarse-grained scales are calculated, and the permutation entropy values ​​under different coarse-grained scales are spliced ​​and combined to obtain the transient distortion feature vector. Obtain the set of historical feature vectors during fault-free operation as the historical steady-state feature benchmark set; The Mahalanobis distance is calculated from the transient distortion feature vector and the historical steady-state feature reference set, and then used as the electrical state deviation.

9. The intelligent filtering monitoring method for synchronous sampling of mutual inductance in transmission lines according to claim 8, characterized in that, The specific process for generating a transmission line anomaly monitoring and early warning command in step S4 when the electrical state deviation exceeds the dynamic health baseline threshold is as follows: Acquire historical electrical state deviation data sets during periods without abnormal alarms, fit a probability density distribution curve based on the historical electrical state deviation data sets, determine the deviation value boundary corresponding to the preset cumulative probability on the probability density distribution curve, and calibrate the deviation value boundary as the dynamic health baseline threshold. By comparing the real-time measured electrical state deviation with the dynamic health baseline threshold, when the real-time measured electrical state deviation is greater than the dynamic health baseline threshold, the over-limit triggering time parameter and transient distortion feature vector are encapsulated to generate a transmission line abnormality monitoring and early warning command.

10. A transmission line mutual inductance synchronous sampling intelligent filtering monitoring system, used to execute the transmission line mutual inductance synchronous sampling intelligent filtering monitoring method as described in any one of claims 1-9, characterized in that, The system includes: The phase correction module is used to collect the three-phase voltage and current sequence of the transmission line and the temperature gradient and leakage flux density of the transformer, substitute them into the thermomagnetic coupling transfer function to obtain the phase shift distribution sequence, and perform fractional delay interpolation and resampling on the three-phase voltage and current sequence according to the phase shift distribution sequence to obtain the phase correction matrix. The manifold noise reduction module is used to reconstruct the phase correction matrix into a phase space evolution trajectory through delayed coordinate mapping, calculate the neighborhood manifold curvature of each node in the phase space evolution trajectory, and cut off the nodes belonging to the random broadband noise component according to the neighborhood manifold curvature, while retaining the core attractor manifold corresponding to the fundamental harmonic component. The topology reconstruction module is used to convert the core attractor manifold into a topological graph, calculate the geodesic distance and state transition probability between nodes and assign connected edge weights, perform graph Laplace smoothing on the topological graph according to the connected edge weights, and project it inversely to a one-dimensional time domain space to obtain the synchronous measurement electrical signal. The status early warning module is used to extract the transient distortion feature vector of the synchronously measured electrical signal, calculate the Mahalanobis distance between the transient distortion feature vector and the historical steady-state feature benchmark set, obtain the electrical status deviation, and generate a transmission line abnormal monitoring early warning command when the electrical status deviation is greater than the dynamic health baseline threshold.

Images