Contact network power supply high-voltage system insulation detection method, device, equipment and medium

Abstract

The invention relates to a contact network power supply high-voltage system insulation detection method, device and equipment and a medium. The method comprises the following steps: constructing a network topology model based on electrical connection data of a convergence area, and generating a topology matrix for describing structural characteristics; accurately simulating a test signal propagation path by using the matrix in combination with a time domain reflection method, and quantifying reflection and attenuation characteristics to form a signal propagation matrix; based on this, the insulation sensitivity of each component is calculated, and a high-priority monitoring area set is screened; iteratively generating a least observation point layout scheme meeting monitoring requirements through a coverage optimization algorithm; a virtual sensor is deployed, measurement data are dynamically processed by adopting a particle filtering algorithm, and the insulation degradation trend is tracked in real time; and finally, when an anomaly is detected, accurately determining a fault position through a triangulation positioning principle by using the time delay difference data of the multiple observation points. According to the technology, the fault positioning precision and the response speed in a dynamic environment are remarkably improved, so that the system resource utilization efficiency is optimized while the detection comprehensiveness is guaranteed.

Description

Technical Field

[0001] This invention belongs to the field of power supply system fault detection technology, and in particular relates to a method, device, equipment and medium for insulation detection of high voltage systems for overhead contact line power supply. Background Technology

[0002] The overhead contact line power supply high-voltage system's busbar area serves as a convergence point for multiple power lines, making reliable insulation condition testing crucial for ensuring the normal operation of the power system. The complex electrical connections and dynamic operating environment of the busbar area necessitate high precision and adaptability in insulation testing to cope with diverse operating scenarios and potential fault risks.

[0003] Traditional insulation testing techniques have significant limitations in this field. Existing insulation testing methods mainly rely on conventional voltage or current measurements, which are difficult to adapt to the complex electrical topology of busbar areas. These methods typically assume simple electrical connections and ignore the complex interactions between busbars and segment insulators, resulting in incomplete test results, especially under dynamic loads or environmental interference, which can easily lead to misjudgments or missed detections. For example, in busbar areas where multiple lines intersect, testing equipment may not be able to accurately distinguish insulation faults between different lines, affecting the accuracy of fault location.

[0004] The unique characteristics of the busbar region present significant technical challenges, the primary one being the complexity of signal propagation characteristics. The electrical connections of multiple lines and segmented insulators within the busbar region form a complex network, causing test signals to be affected by multiple reflections and attenuation during propagation, making it difficult to accurately capture changes in insulation status. Furthermore, this complexity leads to another critical issue: how to obtain comprehensive insulation information from a limited number of observation points. The selection of observation points directly determines whether the detection system can cover the insulation status of all critical components, and improper layout may lead to the overlooking of weak points in insulation in some areas. For example, in actual operation, a minor defect in a segmented insulator may go undetected due to an improperly positioned observation point, thus creating a potential safety hazard.

[0005] Therefore, how to optimize the layout and number of observation points based on the electrical connection topology and signal propagation characteristics of the merging region, so as to achieve comprehensive monitoring of the insulation status in complex networks with the fewest observation points, has become a key problem that needs to be solved in this study. Summary of the Invention

[0006] Therefore, it is necessary to provide a method, device, equipment, and medium for insulation testing of high-voltage systems for overhead contact line power supply, addressing the aforementioned technical problems.

[0007] In a first aspect, this application provides a method for insulation testing of a high-voltage system for overhead contact line power supply, comprising:

[0008] S1. Based on the electrical connection data of the busbar area and the location information of the segmented insulators of the overhead contact power supply network, a network topology model is constructed using graph theory to generate a topology characteristic description matrix;

[0009] S2. Based on the topology characteristic description matrix, the time-domain reflection method is used to simulate the propagation path of the test signal in the contact power supply network. The reflection characteristics and attenuation characteristics of the signal are calculated based on the propagation path, and the signal propagation characteristic matrix is ​​constructed based on the reflection characteristics and attenuation characteristics.

[0010] S3. Based on the signal propagation characteristic matrix, calculate the sensitivity of each component in the overhead contact power supply network to insulation changes; mark the areas corresponding to components with sensitivity higher than the threshold as high-priority monitoring areas to obtain a set of high-priority areas;

[0011] S4. Based on the high-priority region set, the observation point layout scheme is iteratively calculated through the coverage optimization algorithm. When the coverage meets the monitoring requirements, the minimum observation point layout scheme is determined.

[0012] S5. Based on the minimum observation point layout scheme, deploy virtual sensors at the corresponding locations; use particle filter algorithm to process the measurement data collected by the virtual sensors to obtain the real-time insulation change trend;

[0013] S6. Based on the real-time insulation change trend, when an insulation abnormality is detected, the signal arrival time delay difference data of each observation point is collected; based on the signal arrival time delay difference data, the fault location coordinates are determined by the triangulation principle and the fault location result is generated.

[0014] Secondly, this application also provides an insulation detection device for a high-voltage system of overhead contact line power supply, used to implement the method described in the first aspect, the device comprising:

[0015] The power grid structure characterization module is used to construct a network topology model based on the electrical connection data of the busbar area and the location information of the segmented insulators of the overhead contact line power supply network, and generate a topology characteristic description matrix.

[0016] The electromagnetic propagation simulation module is used to simulate the propagation path of test signals in the contact power supply network based on the topology characteristic description matrix and the time-domain reflection method. Based on the propagation path, the reflection characteristics and attenuation characteristics of the signal are calculated, and the signal propagation characteristic matrix is ​​constructed according to the reflection characteristics and attenuation characteristics.

[0017] The risk sensitivity assessment module is used to calculate the sensitivity of each component in the overhead contact power supply network to insulation changes based on the signal propagation characteristic matrix; the areas corresponding to components with sensitivity higher than the threshold are marked as high-priority monitoring areas, thus obtaining a set of high-priority areas;

[0018] The optimal monitoring point planning module is used to iteratively calculate the observation point layout scheme based on the high-priority area set and through the coverage optimization algorithm. When the coverage meets the monitoring requirements, the minimum observation point layout scheme is determined.

[0019] The real-time insulation status sensing module is used to deploy virtual sensors at corresponding locations according to the minimum observation point layout scheme; the particle filter algorithm is used to process the measurement data collected by the virtual sensors to obtain the real-time insulation change trend.

[0020] The fault location execution module is used to collect signal arrival time delay difference data at each observation point when an insulation abnormality is detected, based on the real-time insulation change trend; and to determine the fault location coordinates and generate fault location results based on the signal arrival time delay difference data using the triangulation principle.

[0021] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement an insulation detection method for a high-voltage contact network power supply system as described in the first aspect.

[0022] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements an insulation detection method for a high-voltage power supply system for overhead contact lines as described in the first aspect.

[0023] The aforementioned method, apparatus, equipment, and medium for insulation detection in a high-voltage system of overhead contact line power supply constructs a network topology model based on electrical connection data of the busbar area, generating a topology matrix describing structural characteristics. This matrix, combined with time-domain reflectometry, is then used to accurately simulate the propagation path of test signals, quantifying reflection and attenuation characteristics to form a signal propagation matrix. Based on this, the insulation sensitivity of each component is calculated, and a set of high-priority monitoring areas is selected. A coverage optimization algorithm iteratively generates a minimum observation point layout scheme that meets monitoring requirements. Virtual sensors are deployed accordingly, and a particle filter algorithm is used to dynamically process measurement data, tracking insulation degradation trends in real time. Finally, when an anomaly is detected, the fault location is accurately determined using the triangulation principle based on the time delay difference data from multiple observation points. This technology overcomes the detection blind spots of traditional methods under the complexity of the electrical structure of the busbar area through topology modeling and time-domain reflectometry analysis; achieves full-area key area monitoring with minimal resource consumption through sensitivity selection and coverage optimization; and significantly improves fault location accuracy and response speed in dynamic environments through particle filter anti-interference processing and multi-source time-difference positioning fusion, thereby optimizing system resource utilization efficiency while ensuring comprehensive detection. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 A flowchart illustrating an insulation detection method for a high-voltage power supply system for overhead contact lines provided by this invention;

[0026] Figure 2 This is a schematic diagram of the process for obtaining a high-priority region set in an optional embodiment of the present invention;

[0027] Figure 3 This is a schematic diagram of the structure of an insulation detection device for a high-voltage power supply system of a contact network, provided by the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0029] refer to Figure 1 The document presents a flowchart illustrating an insulation testing method for a high-voltage contact network power supply system provided in this application. The method includes the following steps:

[0030] S1. Based on the electrical connection data of the busbar area and the location information of the segmented insulators of the overhead contact line power supply network, a network topology model is constructed using graph theory to generate a topology characteristic description matrix.

[0031] Specifically, the electrical connection data includes the busbar material parameters, cross-sectional dimensions, line length, and connection method parameters. The segment insulator location information can be collected with precise coordinates by positioning equipment, and a Cartesian coordinate system is established with the geometric center of the busbar area as the origin. At the same time, the electrical parameters corresponding to the segment insulator model are recorded.

[0032] When constructing the model using graph theory, electrical components within the bus region are abstracted as nodes. Node types include feeder inlets, bus nodes, segment insulators, and load outlets. Each node is assigned an attribute vector containing a type identifier, rated voltage, rated current, and location coordinates. Electrical connections between nodes are abstracted as edges. The edge weight vector includes the actual line length, DC resistance, distributed inductance, and distributed capacitance. The DC resistance is determined by the formula... Calculation. In this formula, R represents the DC resistance of the bus, ρ represents the resistivity of the bus material, L represents the actual length of the bus, and S represents the cross-sectional area of ​​the bus. Distributed inductance can be calculated using finite element simulation software. After determining the distributed inductance per unit length based on operating conditions such as line spacing and ground clearance, the total distributed inductance is obtained based on the line length. Distributed capacitance is calculated using the parallel conductor capacitance formula, and subsequently, an edge characteristic matrix is ​​constructed based on the edge weight vector.

[0033] The topology description matrix includes an adjacency matrix, an attribute matrix, and an edge characteristic matrix. The adjacency matrix is ​​n×n (n is the total number of nodes), where a value of 1 indicates a direct connection between the two nodes, and a value of 0 indicates no direct connection. The attribute matrix is ​​n×4, with each row representing the attribute vector of a node. The edge characteristic matrix is ​​m×5 (m is the total number of edges), with each row recording the starting node number, ending node number, and weight vector parameters of each edge. After the matrices are constructed, topology validity is verified using specialized software to check for isolated nodes or redundant edges, ensuring the model accurately reflects the electrical connections in the confluence region.

[0034] S2. Based on the topology characteristic description matrix, the time-domain reflection method is used to simulate the propagation path of the test signal in the contact power supply network. The reflection characteristics and attenuation characteristics of the signal are calculated based on the propagation path, and the signal propagation characteristic matrix is ​​constructed based on the reflection characteristics and attenuation characteristics.

[0035] Specifically, when conducting time-domain reflection (TDR) simulations based on the topology characteristic description matrix, the technical parameters of the test signal source are first determined, including the signal rise time, amplitude, and duration. The signal duration covers the time it takes for the signal to travel back and forth along the longest path, so as to fully record the reflection process.

[0036] The propagation path simulation employs a depth-first search algorithm, using the main power supply node in the merging region as the signal injection point. Starting from this node, all adjacent edges are traversed, and the node sequence for each path is recorded. For each path, the signal propagation speed is calculated based on the actual line length, distributed inductance, and distributed capacitance parameters in the edge characteristic matrix, using the following formula:

[0037]

[0038] In the formula, v represents the speed at which the signal propagates in the line, L0 represents the distributed inductance per unit length of the line, and C0 represents the distributed capacitance per unit length of the line.

[0039] Signal attenuation characteristics are calculated using the attenuation constant, and the formula for the attenuation constant is:

[0040]

[0041] In this formula, α represents the signal attenuation constant, R represents the DC resistance of the line, L0 represents the distributed inductance per unit length of the line, C0 represents the distributed capacitance per unit length of the line, and G represents the leakage conductance of the line. Since high-voltage lines have good insulation performance, the leakage conductance can be ignored. In this case, the formula can be simplified to a form that only includes the resistance term, and then the attenuation of the signal after propagation can be calculated based on the line length.

[0042] The calculation of reflection characteristics focuses on impedance discontinuities and is reflected by the reflection coefficient, the formula for which is:

[0043]

[0044] In the formula, Γ represents the reflection coefficient, and Z L Z0 represents the equivalent impedance at the load end, and Z0 represents the characteristic impedance of the line at the incident end, where the characteristic impedance of the line is obtained through... Calculation. The amplitude of the reflected signal can be further determined based on the reflection coefficient, and the time it takes for the reflected signal to reach the injection point can be calculated by combining the signal propagation speed and the line length.

[0045] The signal propagation characteristic matrix can adopt an m×8 structure (where m is the number of propagation paths). Each row contains the path number, starting node, ending node, propagation speed, attenuation constant, reflection coefficient magnitude, reflection delay, incident signal amplitude, and reflected signal amplitude. After the matrix is ​​constructed, a model is built using circuit simulation software. The matrix parameters are input for time-domain simulation to verify the consistency between the reflected waveform and the calculated results, ensuring the accuracy of the propagation characteristic description.

[0046] S3. Based on the signal propagation characteristic matrix, calculate the sensitivity of each component in the overhead contact power supply network to insulation changes; mark the areas corresponding to components with sensitivity higher than the threshold as high-priority monitoring areas to obtain a set of high-priority areas.

[0047] Specifically, when calculating the sensitivity of each component to insulation changes, a "component" is defined as an independent electrical unit such as a bus section, segment insulator, or feeder branch section. Sensitivity is defined as the ratio of the change in the characteristic quantity of the detection signal caused by a change in insulation resistance to the change in insulation resistance itself, as shown in the following formula:

[0048]

[0049] In this formula, Sens represents the sensitivity of the component to changes in insulation, ΔS represents the change in the characteristic quantity of the detected signal (the change in reflection coefficient ΔΓ is used here), and ΔR represents the change in insulation resistance.

[0050] The insulation degradation simulation employs a stepped degradation model, setting normal insulation resistance and post-degradation insulation resistance; the difference between the two represents the change in insulation resistance. For each component, circuit simulation is used to model the changes in signal characteristics after insulation degradation. Taking a segmented insulator as an example, insulation degradation manifests as a change in its equivalent impedance, leading to a change in the reflection coefficient. The sensitivity can be calculated by combining this with the change in insulation resistance. For a busbar section, insulation degradation manifests as a decrease in insulation resistance to ground. By analyzing the impact of changes in insulation resistance to ground on the line's equivalent characteristic impedance and attenuation constant, the change in reflection coefficient or attenuation is determined, and then the sensitivity is calculated.

[0051] The sensitivity threshold can be determined statistically from experimental data. Multiple samples with different degrees of insulation degradation are selected, and the sensitivity distribution of each component is calculated. The lower limit of a specific confidence interval is taken as the threshold. The physical areas corresponding to components with sensitivity higher than this threshold are marked as high-priority monitoring areas. The scope of each high-priority monitoring area is defined, forming a set of high-priority areas.

[0052] S4. Based on the high-priority area set, the observation point layout scheme is iteratively calculated through the coverage optimization algorithm. When the coverage meets the monitoring requirements, the minimum observation point layout scheme is determined.

[0053] Specifically, when performing coverage optimization based on a set of high-priority areas, a greedy algorithm for the set coverage problem is adopted. The goal is to cover all high-priority areas with the fewest observation points. The coverage range of the observation points is determined by the detection capability of the virtual sensor, specifically based on the amplitude attenuation characteristics of the TDR signal. When the signal propagation distance is within a certain range, the measurement error of the signal characteristic quantity must be controlled within the allowable range. This propagation distance range is the coverage range of a single observation point.

[0054] First, the high-priority regions are discretized into a set of spatial points. Each high-priority region is then divided according to a fixed grid density, resulting in a total set containing discrete points from all high-priority regions. The candidate location set for observation points is determined based on the electrical nodes and line routing of the merging region, including all merging nodes, segment insulator installation points, and feeder midpoints. For each candidate point, the set of discrete points it covers is calculated, which is all points within the coverage area of ​​the candidate point in the total discrete point set.

[0055] The greedy algorithm iterative process is as follows: First, calculate the coverage efficiency of each candidate point. The coverage efficiency is the number of new discrete points covered by the candidate point (i.e., the number of discrete points not covered by the selected observation points). Select the candidate point with the highest coverage efficiency and add it to the observation point set. At the same time, update the covered point set and the uncovered point set. Second, repeat the operation of the first step, calculate the coverage efficiency of the remaining candidate points on the uncovered point set, select the candidate point with the highest coverage efficiency and add it to the observation point set, until the uncovered point set is empty, that is, the coverage rate meets the monitoring requirements.

[0056] Next, we verify the minimum number of observation points and check if there is a single observation point that can cover all high-priority areas. If not, the current set of observation points is the minimum observation point layout scheme; if it exists, we need to adjust the selection of observation points and finally determine the minimum observation point layout scheme.

[0057] S5. Based on the minimum observation point layout scheme, deploy virtual sensors at the corresponding locations; use a particle filter algorithm to process the measurement data collected by the virtual sensors to obtain the real-time insulation change trend.

[0058] Specifically, when deploying virtual sensors at the locations of the minimum observation point layout scheme, the virtual sensors are constructed based on physical sensor data and data fusion technology. For observation points without physical sensors, the TDR signal at that location is calculated using interpolation algorithms based on data collected from surrounding physical sensors; for observation points with physical sensors, data such as voltage, current, and TDR reflection signal amplitude are directly collected from the physical sensors.

[0059] The particle filter algorithm is used to process the amplitude data of TDR reflected signals acquired by virtual sensors, estimate the real-time insulation resistance value, and thus obtain the insulation change trend. The core steps of the algorithm include initializing the particle set, state prediction, observation update, weight update, resampling, and state estimation. State prediction is based on an insulation degradation model, and the formula is:

[0060] R(t+1)=R(t)-λΔt

[0061] In the formula, R(t) represents the estimated insulation resistance at time t, R(t+1) represents the predicted insulation resistance at time t+1, λ represents the insulation degradation rate, and Δt represents the sampling interval.

[0062] In the observation update phase, a mapping relationship between the amplitude of the TDR reflected signal and the insulation resistance is established, and the observation likelihood probability is calculated based on this mapping relationship. In the weight update phase, the prior weight of each particle is multiplied by the observation likelihood probability to obtain the posterior weight, which is then normalized. In the resampling phase, particles with excessively low weights are removed, and particles with excessively high weights are copied, while keeping the total number of particles unchanged. In the state estimation phase, the estimated value of the insulation resistance at time t+1 is calculated by weighted summation, using the following formula:

[0063]

[0064] In this formula, R est (t+1) represents the final estimated value of the insulation resistance at time t+1, and N represents the total number of particles. The normalized weights of the i-th particle. This represents the state value of the i-th particle at time t+1.

[0065] Through continuous iterative calculations, the estimated insulation resistance values ​​at different times are obtained, thereby forming a real-time insulation change trend, which can intuitively reflect the deterioration process of insulation performance.

[0066] S6. Based on the real-time insulation change trend, when an insulation abnormality is detected, the signal arrival time delay difference data of each observation point is collected; based on the signal arrival time delay difference data, the fault location coordinates are determined by the triangulation principle and the fault location result is generated.

[0067] Specifically, the insulation anomaly detection uses a dual-criteria system based on real-time insulation change trends. An insulation anomaly is identified when the estimated insulation resistance is below the fault threshold or when the rate of change of the estimated insulation resistance exceeds the sudden deterioration threshold. After triggering the insulation anomaly alarm, the fault location process is initiated, collecting signal arrival time delay difference data from each observation point—that is, the time difference between the arrival of the partial discharge pulse signal generated by the fault at different observation points.

[0068] The propagation speed of a signal in a high-voltage line is calculated based on the line's distributed inductance and capacitance, using the same formula as the signal propagation speed formula in step S2. The distance difference is calculated based on the time delay difference, using the following formula:

[0069] Δd ij =v×Δt ij

[0070] In the formula, Δd ij Δt represents the difference in distance from observation point i and observation point j to the fault point, v represents the signal propagation speed, and Δt represents the signal propagation speed. ij This represents the time delay difference between the arrival of the signal at observation point i and observation point j.

[0071] Based on the principle of triangulation, let the coordinates of the fault point be (x, y), and the coordinates of the observation point i be (x, y). i ,y i The coordinates of observation point j are (x j ,y j If the distance from observation point i to the fault point is calculated as follows:

[0072]

[0073] The formula for the distance from observation point j to the fault point is:

[0074]

[0075] Combining distance difference Δd ij =d i -d jThis yields an equation relating the fault point coordinates (x, y). By selecting three or more observation points, multiple equations are established to form a system of equations. The system of equations is then solved using a numerical algorithm (such as Newton's iteration method) to obtain the fault point coordinates.

[0076] The fault location results can include the fault location coordinates, insulation resistance value at the time of the fault, fault type, and fault level. At the same time, it generates a visual marker of the fault location on the topology map of the bus region, which facilitates maintenance personnel to quickly locate the fault and carry out maintenance work.

[0077] The aforementioned method for insulation detection in a high-voltage overhead contact line power supply system constructs a network topology model based on electrical connection data of the busbar area, generating a topology matrix describing structural characteristics. This matrix, combined with time-domain reflectometry, is then used to accurately simulate the propagation path of test signals, quantifying reflection and attenuation characteristics to form a signal propagation matrix. Based on this, the insulation sensitivity of each component is calculated, and a set of high-priority monitoring areas is selected. A coverage optimization algorithm iteratively generates a minimum observation point layout scheme that meets monitoring requirements. Virtual sensors are deployed accordingly, and a particle filter algorithm is used to dynamically process measurement data, tracking insulation degradation trends in real time. Finally, when an anomaly is detected, the fault location is accurately determined using the triangulation principle based on the time delay difference data from multiple observation points. This technology overcomes the detection blind spots of traditional methods under the complexity of the electrical structure in the busbar area through topology modeling and time-domain reflectometry analysis; achieves full-area key area monitoring with minimal resource consumption through sensitivity selection and coverage optimization; and significantly improves fault location accuracy and response speed in dynamic environments through particle filter anti-interference processing and multi-source time-difference positioning fusion, thereby optimizing system resource utilization efficiency while ensuring comprehensive detection.

[0078] refer to Figure 2 In one optional embodiment, based on the signal propagation characteristic matrix, the sensitivity of each component in the contact network to insulation changes is calculated; the areas corresponding to components with sensitivity higher than a threshold are marked as high-priority monitoring areas, resulting in a set of high-priority areas, including the following steps:

[0079] S11. Based on the signal propagation characteristic matrix, preset insulation degradation disturbance parameters are injected into the simulation environment to simulate the signal reflection characteristic changes when the insulation resistance of each component continuously decreases, generating a multi-dimensional response dataset including reflection intensity offset, time delay deviation, and frequency response fluctuation.

[0080] Specifically, before conducting insulation degradation simulation based on the signal propagation characteristic matrix, a simulation environment equivalent to the actual contact power supply network is first constructed. This simulation environment is based on parameters such as propagation path, reflection coefficient, and attenuation constant in the signal propagation characteristic matrix. An equivalent circuit model including components such as busbars, segmented insulators, and feeders is built using circuit simulation software to ensure that the model can reproduce the signal propagation and reflection laws in the real network.

[0081] The injected preset insulation degradation disturbance parameter is defined as the continuous change process of the insulation resistance of each component, specifically manifested as a dynamic sequence in which the insulation resistance gradually decreases from the normal state value to the degradation threshold value. During the simulation, this disturbance parameter is applied individually to each independent component (such as a bus section or a segmented insulator) to avoid the aliasing of response signals caused by simultaneous disturbances of multiple components, ensuring accurate location of the signal change corresponding to the insulation degradation of a single component.

[0082] During the simulation, real-time data on changes in signal reflection characteristics are collected to generate a multidimensional response dataset. Among these, the reflection intensity offset refers to the difference between the signal reflection amplitude after insulation degradation and the amplitude under normal conditions, reflecting the impact of decreased insulation resistance on signal energy reflection; the time delay deviation refers to the difference between the arrival time of the reflected signal at the observation point under degradation and the arrival time under normal conditions, stemming from the change in the equivalent impedance of the line caused by insulation degradation, which in turn alters the signal propagation speed; and the frequency response fluctuation refers to the difference in the reflection coefficient of the signal at different frequency components. Since different frequency signals have different sensitivities to changes in insulation condition, this parameter can capture the frequency-dependent characteristics of insulation degradation.

[0083] During the data acquisition process, a fixed sampling interval is set to ensure that for every gradient decrease in insulation resistance, a complete set of reflection intensity offset, time delay deviation, and frequency response fluctuation data is collected. This ultimately forms a multi-dimensional response dataset covering the entire degradation process of each component, providing basic data support for subsequent sensitivity calculations.

[0084] S12. Based on the multidimensional response dataset, the weighted evaluation values ​​of the reflection intensity change rate, time delay offset rate and frequency response distortion are calculated by comprehensively calculating the pre-constructed sensitivity quantification function to obtain the sensitivity coefficient of each component. The sensitivity coefficient is then normalized to obtain the normalized coefficient.

[0085] Specifically, the core idea of ​​the sensitivity quantification function is to comprehensively consider signal changes across three dimensions: reflection intensity, time delay, and frequency response. By using a weighted approach, these multi-dimensional parameters are transformed into a single sensitivity evaluation index to fully reflect the impact of component insulation degradation on the detection signal. The core formula of the sensitivity quantification function can be:

[0086] Sens k =ω1·ΔI k +ω2·ΔT k +ω3·ΔF k

[0087] In the formula, Sens kThe sensitivity coefficient represents the k-th component. ω1, ω2, and ω3 are the weighting coefficients for the rate of change of reflection intensity, the time delay offset rate, and the frequency response distortion, respectively. Their values ​​are determined based on the sensitivity of each parameter to insulation degradation. Parameters that respond more significantly to insulation changes are assigned higher weights, and ω1 + ω2 + ω3 = 1; ΔI k ΔT represents the rate of change of reflection intensity of the k-th component, i.e., the ratio of the reflection intensity shift to the reflection intensity under normal conditions; k ΔF represents the time delay offset rate of the k-th component, that is, the ratio of the time delay deviation to the reflection time delay under normal conditions; k This represents the frequency response distortion of the k-th component, which is the ratio of the maximum value of the frequency response fluctuation to the frequency response amplitude under normal conditions.

[0088] After calculating the sensitivity coefficients of each component, normalization is performed to eliminate the coefficient magnitude deviation caused by differences in basic parameters (such as reflection intensity and time delay under normal conditions) between different components. The normalization formula is:

[0089]

[0090] In the formula, Sens norm,k Sens represents the normalized coefficient of the k-th component. min Sens represents the minimum sensitivity coefficient among all components. max This represents the maximum value among all component sensitivity coefficients. This formula maps the sensitivity coefficients of all components to the [0,1] interval, facilitating subsequent threshold setting and comparison.

[0091] S13. Based on the statistical distribution of the normalization coefficients, the sensitivity threshold is dynamically set using quantile analysis. Components with normalization coefficients exceeding the threshold are mapped to their corresponding physical space coordinates to generate a set of high-priority regions.

[0092] Specifically, when setting sensitivity thresholds based on normalization coefficients, the core purpose of using quantile analysis is to dynamically adapt to the topological differences of different overhead contact power supply networks using statistical laws, thereby avoiding misjudgments or omissions caused by fixed thresholds.

[0093] First, a statistical distribution analysis is performed on the normalized coefficients of all components. By plotting frequency distribution histograms or cumulative distribution curves, the central tendency and dispersion of the coefficients are observed to determine whether the data distribution conforms to a normal or skewed distribution, thus providing a basis for quantile selection.

[0094] The quantile analysis method is implemented as follows: The normalized coefficients of all components are sorted in ascending order, and the coefficient value corresponding to a specific quantile is selected as the sensitivity threshold. The principle for selecting this quantile is to ensure coverage of critical, highly sensitive components while avoiding too many non-critical components being misclassified as high priority. Typically, a higher quantile (such as the 90th or 95th percentile) can be selected, meaning that only a few components with extremely high normalized coefficients will exceed this threshold, ensuring that the high-priority monitoring area focuses on the core components most sensitive to insulation changes.

[0095] Once the threshold is determined, components with normalization coefficients exceeding the threshold are mapped to physical spatial coordinates. The mapping process uses the network topology model constructed in step S1 as a bridge. By establishing the correspondence between components and physical locations in the model (such as the start and end coordinates of a busbar section corresponding to a component, and the installation coordinates of a segment insulator), the spatial range of the component in the actual contact network power supply is determined.

[0096] When generating the high-priority region set, the boundary coordinates of each high-priority region are clearly defined. For linear components (such as busbar sections), the coordinates of the component's start and end points are used as the basis, extending a certain range to both sides (determined based on the signal detection coverage range); for point-like components (such as segmented insulators), a circular or rectangular coverage area is delineated centered on the component's installation coordinates. Finally, the coordinate information of all high-priority regions is compiled into a set, forming the high-priority region set, providing a clear spatial target for subsequent observation point layout.

[0097] In this embodiment, a multi-dimensional response dataset is generated by injecting preset insulation degradation disturbance parameters. Combined with a sensitivity quantification function, the changes in multi-dimensional signals such as reflection, time delay, and frequency response are comprehensively evaluated. After normalization and dynamic threshold setting, the sensitivity of each component to insulation changes is accurately quantified, effectively avoiding sensitivity misjudgment caused by single signal features, accurately locking high-priority monitoring areas, providing accurate spatial targets for subsequent observation point layout, and greatly improving the targeting and priority differentiation capabilities of insulation detection.

[0098] In one optional embodiment, the sensitivity threshold is dynamically set using quantile analysis based on the statistical distribution of the normalized coefficients, including the following steps:

[0099] S21. Sort the normalized coefficients of all components in ascending order of their numerical values ​​to generate an ordered sensitivity sequence.

[0100] Specifically, before sorting the normalized coefficients of all components in ascending order, data preprocessing is performed to eliminate the interference of outliers on the sorting results. The preprocessing stage uses the 3σ criterion to identify and remove outliers from the normalized coefficients. This criterion is based on the normal distribution characteristics and considers values ​​exceeding the mean ± 3 times the standard deviation to be likely outliers (not reflecting the true sensitivity of the components, but mostly caused by data acquisition errors, simulation parameter disturbances, or calculation errors). In practice, the statistical characteristics of the normalized coefficients of all components are first calculated: the mean μ and the standard deviation σ. The mean μ is the arithmetic mean of all normalized coefficients, and the standard deviation σ measures the dispersion of the normalized coefficients. Then, an outlier determination interval [μ-3σ, μ+3σ] is constructed. Normalized coefficients falling outside this interval are marked as outliers and removed from the dataset, resulting in a set of valid normalized coefficients.

[0101] After preprocessing, the effective normalized coefficient set is sorted in ascending order using the Quick Sort algorithm. This algorithm divides the dataset into multiple sub-intervals using a divide-and-conquer strategy, recursively arranging the data within each sub-interval. It boasts advantages such as low average time complexity and adaptability to scenarios with multiple components (large sample sizes). During the sorting process, a mapping relationship is established between the normalized coefficients and their corresponding components. That is, each normalized coefficient retains the identification information of its respective component (such as component number and component type) after sorting, avoiding the loss of the correspondence between coefficients and components. After sorting, an ordered sensitivity sequence is generated, which is stored in array form, with array indices starting from 1, denoted as S = [s1, s2, ..., s]. n ], where n is the number of effective normalization coefficients (i.e., the total number of components after removing outliers), and satisfies s1≤s2≤...≤s n That is, the elements at the beginning of the array correspond to components with lower sensitivity, and the elements at the end correspond to components with higher sensitivity, providing an ordered data foundation for subsequent quantile location.

[0102] S22. For ordered sensitivity sequences, calculate the target quantile index value according to the preset monitoring accuracy requirements, and extract the normalization coefficient corresponding to the index value as the initial threshold benchmark.

[0103] Specifically, the core of calculating the target quantile index value for ordered sensitivity sequences is to determine the target quantile value based on the preset monitoring accuracy requirements, and then calculate the corresponding index through the quantile position formula to ensure that the initial threshold benchmark can cover the highly sensitive components that meet the monitoring accuracy.

[0104] Higher monitoring accuracy requirements mean minimizing the probability of highly sensitive components being missed, thus requiring the selection of higher quantile values ​​(i.e., later positions in the ordered sequence). Conversely, lower monitoring accuracy requirements allow for the selection of relatively lower quantile values. The target quantile value is denoted as p, and its range is (0,1). For example, if the preset monitoring accuracy requirement is "highly sensitive component miss rate ≤ 5%", then p is set to 0.95, meaning that only 5% of highly sensitive components are allowed to fall after the quantile, ensuring that 95% of highly sensitive components are covered.

[0105] Based on the target quantile value p and the length n of the ordered sensitivity sequence, the index of the target quantile in the ordered sequence is calculated using the quantile position formula, as follows:

[0106]

[0107] In the formula, Q pos p is the index value of the target quantile (a positive integer to ensure it corresponds to a specific element in the ordered sequence), p is the preset target quantile value (determined based on monitoring accuracy), and n is the length of the ordered sensitivity sequence (i.e., the total number of effective components). This is a rounding function. The design logic of this formula is as follows: calculate the theoretical position of the quantile within a continuous interval using p×(n-1)+1, and then round up to ensure that the index value is not less than the theoretical position. This avoids a low threshold benchmark due to the quantile falling between two elements, thus preventing the omission of highly sensitive components. For example, if p=0.95 and n=100, the theoretical position is 0.95×99+1=95.05, and after rounding up, Q... pos =96, meaning the 96th element in the ordered sequence is selected as the initial threshold benchmark.

[0108] When truncating the initial threshold benchmark, directly extract the index Q from the ordered sensitivity sequence. pos The normalization coefficient is denoted as The physical meaning of this initial threshold benchmark is: under the preset monitoring accuracy requirements, all normalized coefficients greater than T... initial The component (corresponding to an index greater than Q in the ordered sequence) pos The elements can be preliminarily identified as highly sensitive components, and their number accounts for no more than 1-p, which meets the limit of monitoring accuracy on the false negative rate.

[0109] S23. Based on the initial threshold benchmark, the system noise tolerance compensation is superimposed to generate the dynamic sensitivity threshold.

[0110] Specifically, the system noise tolerance compensation is superimposed on the initial threshold benchmark. The core purpose is to offset the influence of system noise on the normalization coefficient measurement results, avoid the underestimation of the normalization coefficient of the real high-sensitivity component due to noise (and thus misjudgment as a non-high-sensitivity component), and ensure the reliability of the dynamic sensitivity threshold.

[0111] The main sources of system noise include random noise during virtual sensor data acquisition (such as measurement noise from voltage / current sensors), electromagnetic interference noise during signal transmission (such as interference from the electromagnetic environment surrounding the contact network on the TDR signal), and deviation noise between the simulation environment and the actual system (such as subtle differences between line parameters in the simulation model and actual parameters). This noise can cause the normalization coefficient of the actual measurement to be less than the true normalization coefficient of the component; therefore, a compensation amount is needed to correct the threshold.

[0112] The calculation of the system noise margin compensation is based on the statistical characteristics of noise and uses the following formula:

[0113] ΔT noise =k×σ noise

[0114] In the formula, ΔT noise σ represents the system noise margin compensation amount, where k is the safety factor, determined based on a preset confidence level, used to cover noise fluctuation ranges at different confidence levels. For example, a 95% confidence level corresponds to k = 1.96, and a 99% confidence level corresponds to k = 2.58. noise The standard deviation of the system noise is obtained through experimental measurement. Under normal insulation conditions of the overhead contact power supply network, multiple sets of normalized coefficients are continuously collected, and the standard deviation of this set of data is calculated, which is σ. noise It reflects the degree of dispersion of noise.

[0115] The formula for generating the dynamic sensitivity threshold is:

[0116] T dynamic =T initial +ΔT noise

[0117] In the formula, T dynamic For the final dynamic sensitivity threshold, T initial As the initial threshold benchmark, Δt noise This represents the system noise margin compensation amount. The design logic of this formula is: by superimposing the compensation amount, the threshold is shifted upwards to offset the effect of noise causing the normalized coefficient measurement value to be too low. Even if the measurement value of the actual highly sensitive component is suppressed by noise, it can still be higher than the dynamic threshold, thereby avoiding missed detection. For example, if the initial threshold reference t initial =0.8, noise standard deviation σ noise =0.02, safety factor k = 1.96, then the compensation amount ΔTnoise =1.96 × 0.02 = 0.0392, dynamic threshold t dynamic =0.8 + 0.0392 = 0.8392, ensuring that components with a true normalization coefficient of 0.83 (because the noise measurement value may be 0.81) can still be identified as highly sensitive components.

[0118] After the dynamic threshold is generated, its rationality is verified. By selecting a small number of standard components with known sensitivity (such as pre-calibrated high-sensitivity segmented insulators and low-sensitivity busbar sections), their normalized coefficients are compared with the dynamic threshold to check whether the judgment result is consistent with the known sensitivity. If there is a deviation, the value of the safety factor k is adjusted until the judgment accuracy meets the preset requirements (such as judgment accuracy ≥ 98%), and finally a reliable dynamic sensitivity threshold is determined.

[0119] In this embodiment, by sorting the normalized coefficients in ascending order and calculating the quantile index, the initial threshold benchmark that meets the monitoring accuracy requirements is accurately located. Then, the system noise tolerance compensation is superimposed to eliminate noise interference in data acquisition, signal transmission and other links, so that the sensitivity threshold can be dynamically adapted to the actual system characteristics, avoiding the omission / false judgment of highly sensitive components caused by fixed thresholds or noise, and significantly improving the reliability and adaptability of the judgment of highly sensitive components.

[0120] In one optional embodiment, based on a high-priority region set, an observation point layout scheme is iteratively calculated using a coverage optimization algorithm. When the coverage meets the monitoring requirements, the minimum observation point layout scheme is determined, including the following steps:

[0121] S31. Based on the spatial coordinate data of the high-priority area set, candidate observation points that meet the equipment installation conditions are selected to generate a set of physically deployable candidate points; for the candidate point set, the effective monitoring radius function of each point is calculated according to the signal propagation characteristic matrix to generate a set of coverage radius parameters.

[0122] Specifically, when screening candidate observation points based on the spatial coordinate data of high-priority areas, the constraints of equipment installation conditions must first be clarified, mainly including physical space constraints and electrical safety constraints. Physical space constraints require that the candidate point location has sufficient installation space (such as near busbar supports or reserved installation positions at both ends of segmented insulators) and that there are no obstructions affecting signal transmission. Electrical safety constraints require that the distance between the candidate point and high-voltage live conductors is not less than a safe distance threshold (determined based on the system's rated voltage) to avoid electromagnetic interference or physical contact risks. The screening process can be achieved through spatial coordinate comparison: the boundary coordinates of the high-priority area are intersected with the equipment installation permit area coordinates of the contact network power supply network, retaining the coordinates falling within the permit area to generate an initial candidate point set; then, collision detection is performed on the initial set to eliminate points that conflict with the spatial positions of existing equipment (such as supports and surge arresters), finally obtaining a physically deployable candidate point set, denoted as C = {c1, c2, ..., c...}. m}, where c i =(x i ,y i ,z i ) represents the three-dimensional spatial coordinates of the i-th candidate point.

[0123] When calculating the effective monitoring radius function for a set of candidate points, the core basis is the attenuation characteristic in the signal propagation characteristic matrix. The effective monitoring radius is defined as follows: within this radius, the attenuation of the signal propagating from the candidate point to the target location does not exceed a preset attenuation threshold, ensuring that the signal-to-noise ratio of the reflected signal meets the detection accuracy requirements. The formula for calculating the effective monitoring radius function is:

[0124] r i (d)=max{d|α i ·d≤A th}

[0125] In the formula, r i Let α be the effective monitoring radius of the i-th candidate point. i The signal attenuation constant at the location of the candidate point (taken from the attenuation parameter of the corresponding path in the signal propagation characteristic matrix), d is the signal propagation distance, and A is the signal attenuation constant at the location of the candidate point. th This is the preset maximum allowable attenuation (determined based on the signal-to-noise ratio requirements of the detection system). The physical meaning of this function is: for the i-th candidate point, when the propagation distance d satisfies the attenuation amount α... i ·d≤A th When the signal can be effectively detected, the corresponding maximum d is the effective monitoring radius. After calculating the effective monitoring radius of each candidate point, a coverage radius parameter set R = {r1, r2, ..., r} is formed. m This provides basic parameters for subsequent spatial coverage modeling.

[0126] S32. Based on the coverage radius parameter set, construct a spherical spatial model describing the monitoring range of candidate points to obtain the three-dimensional coverage domain mapping relationship; according to the three-dimensional coverage domain mapping relationship, use the graph theory set coverage algorithm to calculate the network coverage rate of candidate point combinations and generate a coverage efficiency matrix.

[0127] Specifically, when constructing a spherical spatial model based on the coverage radius parameter set, each candidate point is considered as the center of a sphere, and its effective monitoring radius is the radius of the sphere, forming a coverage sphere in three-dimensional space. The three-dimensional coverage domain mapping relationship is used to describe the spatial intersection between each coverage sphere and the high-priority region, that is, the proportion of a high-priority region covered by candidate points. For any spatial point p = (x, y, z) in the high-priority region, if this point and candidate point c... i Euclidean distance || pc i ||≤r i If so, then the point is determined to be a candidate point c. i Coverage, where the Euclidean distance formula is:

[0128]

[0129] In the formula, (x,y,z) are the three-dimensional coordinates of a point within the high-priority region, (x... i ,y i ,z i ) represents the three-dimensional coordinates of the i-th candidate point.

[0130] The coverage ratio is calculated by discretizing high-priority regions. Each high-priority region is discretized into a three-dimensional point set according to a preset grid density (e.g., 1m×1m×0.5m). The ratio of the number of discrete points covered by candidate points to the total number of discrete points in the region is counted as the coverage of the candidate point to the region. Based on the coverage of all candidate points to all high-priority regions, a coverage efficiency matrix E is constructed. The matrix is ​​of order m×k (m is the number of candidate points, k is the number of high-priority regions), where the element E[i][j] represents the coverage of the i-th candidate point to the j-th high-priority region, and its value ranges from [0,1], where 1 represents complete coverage and 0 represents no coverage.

[0131] When using the graph theory set covering algorithm to calculate the network coverage of candidate point combinations, the set of high-priority regions is considered as the set of elements to be covered, and the candidate points are considered as subsets of the covered elements. The network coverage is defined as the proportion of elements covered by candidate point combinations (high-priority regions) out of the total number of elements, and the calculation formula is as follows:

[0132]

[0133] In the formula, S represents the combination of candidate points (i.e., a subset selected from the candidate point set), and max i∈SE[i][j] represents the maximum coverage of the j-th high-priority region in the combination S (taking the maximum coverage of all candidate points covering this region in the combination), where k is the total number of high-priority regions. This formula obtains the overall coverage rate of the combination by taking the maximum coverage of each region, summing them, and then dividing by the total number of regions, reflecting the coverage effect of the candidate point combination on the high-priority regions.

[0134] S33. Based on the coverage efficiency matrix, establish an optimization function with the objectives of minimizing the number of observation points and maximizing the coverage rate; based on the optimization function, dynamically adjust the candidate point coordinate combination through a greedy iterative algorithm, update the coverage efficiency matrix and calculate the incremental coverage rate in each iteration.

[0135] Specifically, when establishing the optimization function, the dual objectives of the number of observation points and coverage are balanced. A weighted summation method is used to transform the multi-objective function into a single-objective function, as shown in the following formula:

[0136] min f(S)=λ·|S|-(1-λ)·Coverage(S)

[0137] In the formula, f(S) is the objective function, |S| is the number of observation points in candidate point combination S, Coverage(S) is the network coverage of combination S, and λ is a weight coefficient (within the range of (0,1), used to adjust the priority between the number of observation points and the coverage; a larger λ indicates a higher priority in reducing the number of observation points). The function is designed as a minimization problem, and the objective function value is reduced by decreasing |S| and increasing Coverage(S) to achieve bi-objective optimization.

[0138] The implementation process of the greedy iterative algorithm is as follows: Initialize the candidate point combination S as an empty set, and the uncovered region set U as all high-priority regions; in the first iteration, calculate the value of each candidate point c. i After adding S, the incremental coverage rate of U (i.e. the proportion of newly covered areas) is calculated. The candidate point with the largest incremental coverage rate is selected and added to S, and the area covered by the candidate point is removed from U. The coverage efficiency matrix E is updated (only the columns related to uncovered areas are retained). The above process is repeated, that is, the incremental coverage rate of the remaining candidate points to the current U is calculated, the optimal candidate point is selected and added to S, until U is empty or the coverage rate reaches the preset requirement.

[0139] The formula for calculating incremental coverage is:

[0140] ΔCoverage(c i ,S)=Coverage(S∪{c i Coverage(S)

[0141] In the formula, ΔCoverage(c i S) represents the candidate point c iIncremental coverage after adding the current portfolio S, Coverage(S∪{c i}) represents the coverage of the new combination, and Coverage(S) represents the coverage of the original combination. This metric measures the contribution of a single candidate point to improving the coverage and is the core basis for the greedy algorithm to select candidate points.

[0142] S34. When the rate of change of incremental coverage is less than or equal to a preset threshold for N consecutive iterations, the algorithm is determined to converge and the current combination of observation points is locked. Based on the current combination of observation points, the optimal layout scheme that meets the coverage requirements and has the fewest number of observation points is set. Here, N is a preset value, and the optimal layout scheme includes the spatial coordinates and coverage radius of each observation point.

[0143] Specifically, the core of the algorithm's convergence determination is to monitor the rate of change of incremental coverage, and the formula for calculating the rate of change is:

[0144]

[0145] In the formula, η t Let ΔCoverage be the incremental coverage rate of the t-th iteration. t Let ΔCoverage be the incremental coverage in the t-th iteration. t-1 Let be the incremental coverage in the (t-1)th iteration. When η t When ≤∈ (∈ is a preset threshold reflecting the accuracy requirement for algorithm convergence), it indicates that the change in incremental coverage is small enough, and continued iteration has limited contribution to improving coverage.

[0146] When the rate of change η is over N consecutive iterations t When ≤∈, the algorithm is considered to have converged, and the current candidate point combination S is locked. At this time, it is verified whether the coverage of the combination meets the preset monitoring requirements (e.g., coverage ≥ 95%). If it does, the optimal layout scheme determination stage is entered; if it does not, the weight coefficient λ is adjusted (e.g., λ is reduced to prioritize improving coverage), and the iteration process is restarted.

[0147] When determining the optimal layout scheme, the number of observation points is minimized while still meeting coverage requirements. By comparing the convergent combination S with all intermediate combinations that meet coverage requirements during iteration, the combination with the fewest observation points is selected as the optimal scheme. The scheme includes the three-dimensional spatial coordinates of each observation point (taken from the coordinate data of the candidate point set) and the corresponding effective monitoring radius (taken from the coverage radius parameter set), while also marking the high-priority areas mainly covered by each observation point, providing clear guidance for actual deployment.

[0148] The final optimal layout scheme is verified through simulation. The observation point parameters in the scheme are input into the catenary power supply network simulation model to simulate signal propagation and coverage effect. The deviation between the actual coverage rate and the theoretical value is calculated. If the deviation is ≤ the preset error range (e.g., ±3%), the scheme is effective; otherwise, the observation point coordinates are fine-tuned or candidate points are added, and the scheme is re-optimized until the requirements are met.

[0149] In this embodiment, physical deployable candidate points are screened by combining equipment installation constraints and signal coverage characteristics. A three-dimensional coverage domain is constructed using a spherical space model. Then, the combination of observation points is optimized using a coverage efficiency matrix and a greedy iterative algorithm. At the same time, a convergence criterion is introduced to ensure efficient convergence of the algorithm. Finally, under the premise of meeting the monitoring coverage requirements, the number of observation points is minimized, balancing the comprehensiveness of insulation detection and deployment costs, and improving the rationality, economy and engineering feasibility of the observation point layout.

[0150] In one optional embodiment, based on signal arrival delay difference data, the fault location coordinates are determined using the triangulation principle and a fault location result is generated, including the following steps:

[0151] S41. Based on the signal arrival delay difference data collected from at least three observation points, calculate the delay difference sequence between each observation point and generate a delay difference vector set; for the delay difference vector set, construct a hyperboloid equation model with the observation points as the focus.

[0152] Specifically, let the set of observation points be O = {o1, o2, o3} (containing at least three non-collinear observation points to ensure unique positioning), where o i =(x i ,y i ,z i Let represent the three-dimensional spatial coordinates of the i-th observation point. The arrival times of the signal at each observation point are t1, t2, and t3, respectively. The time delay difference sequence is defined as the time difference between any two observation points, i.e., Δt. 12 =t1-t2, Δt 13 =t1-t3, Δt 23 =t2-t3, thus generating the time delay difference vector set Δt=[Δt 12 ,Δt 13 ,Δt 23 ] T .

[0153] The core basis for constructing the hyperboloid equation model is the relationship between signal propagation distance and time. Let the coordinates of the fault point be p = (x, y, z), and the signal propagation speed in space be v (determined based on the line characteristic impedance and the electromagnetic wave propagation law, similar to the previous steps). If the distance from the fault point to the observation point is... i The distance is d i =||poi || = v·t i (‖·‖ represents Euclidean distance). According to the definition of time delay difference, the distance difference satisfies d i -d j =v·Δt ij Corresponding to o in three-dimensional space i and o j Equation of a hyperboloid with focus:

[0154] ‖po i ‖-‖po j || = v·Δt ij

[0155] In the formula, Let be the Euclidean distance from the fault point to the i-th observation point. Let Δt be the Euclidean distance from the fault point to the j-th observation point. ij Let be the time delay difference between the arrival of the signal at observation points i and j, and v be the signal propagation speed.

[0156] For the three observation points, two independent sets of hyperboloid equations can be established (e.g., equations with foci o1 and o2 and equations with foci o1 and o3), forming a system of hyperboloid equations:

[0157]

[0158] The solution to this system of equations is the possible spatial location of the fault point, providing a mathematical model for subsequent numerical solutions.

[0159] S42. Using the hyperboloid equation system model, the least squares iterative method is used to solve the real number solution of the equation system and generate a set of candidate fault point coordinates.

[0160] Specifically, when solving the hyperboloid equations using the least squares iterative method, the nonlinear equations are first transformed into a linearized error model. Since the hyperboloid equations contain square root terms and are nonlinear, direct solutions are difficult; therefore, an iterative method is used to gradually approximate the true solution.

[0161] First, set the initial estimated value of the fault point coordinates as p0 = (x0, y0, z0) (the geometric center of the high priority area or the coordinates of a certain observation point can be selected as the initial value), and then linearize the equation at p0 through Taylor expansion.

[0162] Taking the first equation as an example, let f1(p) = ||p-o1|| - ||p-o2|| - v·Δt 12 =0, perform a Taylor expansion at p0, retaining the first-order terms:

[0163]

[0164] In the formula, is the gradient vector of f1 at p0, reflecting the sensitivity of the equation to coordinate changes.

[0165] After linearizing both equations, we obtain a system of linear equations:

[0166] J(p0)·Δp=-f(p0)

[0167] In the formula, J(p0) is the Jacobian matrix (composed of the gradient vectors of the two equations), Δp = p - p0 is the coordinate correction, and f(p0) = [f1(p0), f2(p0)]. T This is the equation error vector corresponding to the initial estimate.

[0168] The goal of least squares iteration is to minimize the sum of squared errors ||f(p)|| 2 By solving the above system of linear equations, Δp is obtained, and the estimated value p is updated. k+1 =p k +Δp, and repeat the linearization and solution process until the iteration converges (e.g., ‖Δp‖≤∈, where ∈ is a preset accuracy threshold). During the iteration process, multiple solutions may occur due to initial value selection or measurement noise. Therefore, all converged real solutions are retained, generating a candidate fault point coordinate set P = {p1, p2, ..., p...} n}

[0169] S43. Based on the set of candidate fault point coordinates, perform spatial geometric clustering analysis to filter dense point groups that meet the error tolerance threshold; for the dense point groups, calculate their spatial centroid coordinates as fault location estimates; generate fault location results based on the fault location estimates.

[0170] Specifically, the purpose of performing spatial geometric clustering analysis on the set of candidate fault point coordinates is to eliminate outliers caused by measurement noise or iteration errors, and retain dense point groups near the actual fault points. Density clustering algorithms (such as DBSCAN) are used to implement clustering: first, the neighborhood radius ∈ db (Based on signal delay measurement error and propagation speed calculation, reflecting the theoretical error range of fault location) and minimum number of points threshold N min (Ensure the dense point cluster is statistically significant); then iterate through the candidate point set, selecting points whose distance to a given point is less than ∈ [0, 1]. db The number of points exceeds N min The points are marked as core points, and a cluster is formed by the core point and all points in its neighborhood.

[0171] When selecting dense point clusters that meet the error tolerance threshold, the spatial dispersion of each cluster (e.g., the average distance between all points within the cluster and the cluster center) is calculated. Only clusters with dispersion less than the preset error tolerance threshold are retained, and these clusters are considered to correspond to the possible locations of the actual fault points. For the selected dense point clusters, their spatial centroid coordinates are calculated as the estimated fault location. The formula for centroid coordinates is:

[0172]

[0173] In the formula, Here, m represents the estimated fault location, and m is the number of candidate points contained in the dense point cluster. i ,y i ,z i Let be the coordinates of the i-th candidate point in the point group. By calculating the arithmetic mean, the centroid coordinates can offset random errors and improve positioning accuracy.

[0174] When generating fault location results, in addition to including centroid coordinates In addition, a location confidence index can be added. This index is calculated based on the dispersion and number of points in a dense cluster. The smaller the dispersion and the more points, the higher the confidence level. At the same time, the estimated fault location is mapped to the physical topology map of the overhead contact line power supply network, marking the component to which it belongs (such as a busbar section or a section insulator) and its relative position to surrounding equipment, providing maintenance personnel with an intuitive reference for fault location.

[0175] The final fault location results are verified on-site. The deviation between the location results and the actual fault point is compared. If the deviation is less than or equal to the preset location error (e.g., ±1m), the method is effective. Otherwise, the time delay measurement accuracy is optimized or the clustering parameters are adjusted until the engineering application requirements are met.

[0176] In this embodiment, a hyperboloid equation system is constructed based on the time delay difference of multiple observation points. Candidate fault points are solved by the least squares iterative method, and then outliers are eliminated by spatial geometric clustering and centroid coordinates are calculated. This effectively eliminates the positioning deviation caused by a single solution or system noise, and significantly improves the accuracy and reliability of fault location. At the same time, it outputs clear fault coordinates and positioning results, providing accurate basis for the rapid location and repair of insulation faults in the contact network power supply system, and shortening the fault investigation time.

[0177] The aforementioned method for insulation detection in a high-voltage overhead contact line power supply system constructs a network topology model based on electrical connection data of the busbar area, generating a topology matrix describing structural characteristics. This matrix, combined with time-domain reflectometry, is then used to accurately simulate the propagation path of test signals, quantifying reflection and attenuation characteristics to form a signal propagation matrix. Based on this, the insulation sensitivity of each component is calculated, and a set of high-priority monitoring areas is selected. A coverage optimization algorithm iteratively generates a minimum observation point layout scheme that meets monitoring requirements. Virtual sensors are deployed accordingly, and a particle filter algorithm is used to dynamically process measurement data, tracking insulation degradation trends in real time. Finally, when an anomaly is detected, the fault location is accurately determined using the triangulation principle based on the time delay difference data from multiple observation points. This technology overcomes the detection blind spots of traditional methods under the complexity of the electrical structure in the busbar area through topology modeling and time-domain reflectometry analysis; achieves full-area key area monitoring with minimal resource consumption through sensitivity selection and coverage optimization; and significantly improves fault location accuracy and response speed in dynamic environments through particle filter anti-interference processing and multi-source time-difference positioning fusion, thereby optimizing system resource utilization efficiency while ensuring comprehensive detection.

[0178] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0179] Based on the same inventive concept, this application also provides an apparatus for implementing the above-described method for insulation detection of a high-voltage contact power supply system. The solution provided by this apparatus is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the contact power supply high-voltage system insulation detection apparatus provided below can be found in the limitations of the contact power supply high-voltage system insulation detection method described above, and will not be repeated here.

[0180] In one exemplary embodiment, such as Figure 3 As shown, an insulation detection device 30 for a high-voltage power supply system of a contact network is provided to implement the methods in the above-described embodiments. The device includes:

[0181] The power grid structure characterization module 31 is used to construct a network topology model based on the electrical connection data of the busbar area and the location information of the segment insulators of the overhead contact power supply network, and generate a topology characteristic description matrix.

[0182] The electromagnetic propagation simulation module 32 is used to simulate the propagation path of test signals in the contact power supply network based on the topology characteristic description matrix and the time-domain reflection method. Based on the propagation path, the reflection characteristics and attenuation characteristics of the signal are calculated, and the signal propagation characteristic matrix is ​​constructed according to the reflection characteristics and attenuation characteristics.

[0183] The risk sensitivity assessment module 33 is used to calculate the sensitivity of each component in the contact power supply network to insulation changes based on the signal propagation characteristic matrix; the area corresponding to the component with sensitivity higher than the threshold is marked as a high priority monitoring area, thus obtaining a set of high priority areas.

[0184] The optimal monitoring point planning module 34 is used to iteratively calculate the observation point layout scheme based on the high priority area set through the coverage optimization algorithm, and determine the minimum observation point layout scheme when the coverage meets the monitoring requirements.

[0185] The real-time insulation status sensing module 35 is used to deploy virtual sensors at corresponding locations according to the minimum observation point layout scheme; and to process the measurement data collected by the virtual sensors using a particle filter algorithm to obtain the real-time insulation change trend.

[0186] The fault location execution module 36 is used to collect signal arrival time delay difference data at each observation point when an insulation abnormality is detected, based on the real-time insulation change trend; and to determine the fault location coordinates and generate fault location results based on the signal arrival time delay difference data using the triangulation principle.

[0187] Embodiments of this application also provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the aforementioned method embodiments.

[0188] Embodiments of this application also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above-described method embodiments.

[0189] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0190] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for insulation testing of a high-voltage power supply system for overhead contact lines, characterized in that, The method includes: S1. Based on the electrical connection data of the busbar area and the location information of the segmented insulators of the overhead contact power supply network, a network topology model is constructed using graph theory to generate a topology characteristic description matrix; S2. Based on the topology characteristic description matrix, the propagation path of the test signal in the contact power supply network is simulated using the time-domain reflection method. The reflection characteristics and attenuation characteristics of the signal are calculated based on the propagation path. A signal propagation characteristic matrix is ​​constructed based on the reflection characteristics and the attenuation characteristics. S3. Based on the signal propagation characteristic matrix, calculate the sensitivity of each component in the contact power supply network to insulation changes; mark the areas corresponding to the components with sensitivity higher than the threshold as high-priority monitoring areas to obtain a set of high-priority areas; S4. Based on the set of high-priority areas, the observation point layout scheme is iteratively calculated using a coverage optimization algorithm. When the coverage meets the monitoring requirements, the minimum observation point layout scheme is determined. S5. Based on the minimum observation point layout scheme, deploy virtual sensors at the corresponding positions; use a particle filter algorithm to process the measurement data collected by the virtual sensors to obtain the real-time insulation change trend; S6. Based on the real-time insulation change trend, when an insulation abnormality is detected, the signal arrival delay difference data of each observation point is collected; based on the signal arrival delay difference data, the fault location coordinates are determined by the triangulation principle and a fault location result is generated.

2. The method according to claim 1, characterized in that, Based on the signal propagation characteristic matrix, the sensitivity of each component in the overhead contact power supply network to insulation changes is calculated. The regions corresponding to components with sensitivity higher than the threshold are marked as high-priority monitoring regions, resulting in a set of high-priority regions, including: S11. Based on the signal propagation characteristic matrix, a preset insulation degradation disturbance parameter is injected into the simulation environment to simulate the change in signal reflection characteristics when the insulation resistance of each component continuously decreases, and a multi-dimensional response dataset including reflection intensity offset, time delay deviation and frequency response fluctuation is generated. S12. Based on the multidimensional response dataset, the weighted evaluation values ​​of the reflection intensity change rate, time delay offset rate and frequency response distortion are calculated by comprehensively calculating the pre-constructed sensitivity quantification function to obtain the sensitivity coefficient of each component, and the sensitivity coefficient is normalized to obtain the normalized coefficient. S13. Based on the statistical distribution of the normalized coefficients, a sensitivity threshold is dynamically set using quantile analysis. Components whose normalized coefficients exceed the threshold are mapped to their corresponding physical space coordinates to generate the high-priority region set.

3. The method according to claim 2, characterized in that, The step of dynamically setting the sensitivity threshold using quantile analysis based on the statistical distribution of the normalized coefficients includes: S21. Sort the normalized coefficients of all components in ascending order of their numerical values ​​to generate an ordered sensitivity sequence; S22. For the ordered sensitivity sequence, calculate the target quantile index value according to the preset monitoring accuracy requirements, and extract the normalization coefficient corresponding to the index value as the initial threshold benchmark. S23. Based on the initial threshold benchmark, the system noise tolerance compensation is superimposed to generate the dynamic sensitivity threshold.

4. The method according to claim 1, characterized in that, Based on the set of high-priority regions, the observation point layout scheme is iteratively calculated using a coverage optimization algorithm. When the coverage meets the monitoring requirements, the minimum observation point layout scheme is determined, including: S31. Based on the spatial coordinate data of the high-priority area set, candidate observation points that meet the equipment installation conditions are selected to generate a set of physically deployable candidate points; for the candidate point set, the effective monitoring radius function of each point is calculated according to the signal propagation characteristic matrix to generate a set of coverage radius parameters. S32. Based on the coverage radius parameter set, construct a spherical space model describing the monitoring range of candidate points to obtain a three-dimensional coverage domain mapping relationship; according to the three-dimensional coverage domain mapping relationship, use a graph theory set coverage algorithm to calculate the network coverage rate of candidate point combinations and generate a coverage efficiency matrix. S33. Based on the coverage efficiency matrix, establish an optimization function with the objectives of minimizing the number of observation points and maximizing the coverage rate; based on the optimization function, dynamically adjust the candidate point coordinate combination using a greedy iterative algorithm, update the coverage efficiency matrix and calculate the incremental coverage rate in each iteration; S34. When the rate of change of the incremental coverage rate is less than or equal to a preset threshold for N consecutive iterations, the algorithm determines that the current combination of observation points has converged and is locked. Based on the current combination of observation points, an optimal layout scheme that meets the coverage requirements and has the fewest number of observation points is set. Wherein, N is a preset value, and the optimal layout scheme includes the spatial coordinates and coverage radius of each observation point.

5. The method according to any one of claims 1 to 4, characterized in that, The process of determining the fault location coordinates and generating a fault location result based on the signal arrival delay difference data using the triangulation principle includes: S41. Based on the signal arrival delay difference data collected from at least three observation points, calculate the delay difference sequence between each observation point and generate a delay difference vector group; for the delay difference vector group, construct a hyperboloid equation model with the observation points as the focus. S42. Using the hyperboloid equation system model, the least squares iterative method is used to solve the real number solution of the equation system to generate a set of candidate fault point coordinates. S43. Based on the set of candidate fault point coordinates, perform spatial geometric clustering analysis to filter dense point groups that meet the error tolerance threshold; for the dense point groups, calculate their spatial centroid coordinates as fault location estimates; and generate the fault location results based on the fault location estimates.

6. An insulation detection device for a high-voltage power supply system for overhead contact lines, used to implement the method according to any one of claims 1 to 5, characterized in that, The device includes: The power grid structure characterization module is used to construct a network topology model based on the electrical connection data of the busbar area and the location information of the segmented insulators of the overhead contact line power supply network, and generate a topology characteristic description matrix. The electromagnetic propagation simulation module is used to simulate the propagation path of the test signal in the contact power supply network based on the topology characteristic description matrix and the time-domain reflection method. Based on the propagation path, the reflection characteristics and attenuation characteristics of the signal are calculated, and a signal propagation characteristic matrix is ​​constructed according to the reflection characteristics and the attenuation characteristics. The risk sensitivity assessment module is used to calculate the sensitivity of each component in the overhead contact power supply network to insulation changes based on the signal propagation characteristic matrix; and to mark the areas corresponding to the components with sensitivity higher than the threshold as high-priority monitoring areas to obtain a set of high-priority areas. The optimal monitoring point planning module is used to iteratively calculate the observation point layout scheme based on the high priority area set through a coverage optimization algorithm, and determine the minimum observation point layout scheme when the coverage meets the monitoring requirements. The real-time insulation status sensing module is used to deploy virtual sensors at corresponding locations according to the minimum observation point layout scheme; and to process the measurement data collected by the virtual sensors using a particle filter algorithm to obtain the real-time insulation change trend. The fault location execution module is used to collect signal arrival time delay difference data at each observation point when an insulation abnormality is detected, based on the real-time insulation change trend; and to determine the fault location coordinates and generate fault location results based on the signal arrival time delay difference data using the triangulation principle.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 5.

Images