Method for detecting internal interface discharge of vehicle-mounted high-voltage cable terminal and related device

Abstract

The application discloses a detection method for internal interface discharge of a vehicle-mounted high-voltage cable terminal and related devices, and relates to the field of high-voltage cable insulation detection. The method comprises the following steps: obtaining near-surface electric field time-varying data synchronized with a power frequency phase; performing spatial difference operation on data of at least two non-grounded measuring points to obtain near-surface electric field gradient distribution data; performing phase resolution analysis on the gradient data to extract external electric field differential characteristic quantities; inputting the characteristic quantities into an internal discharge stage recognition model trained based on a labeled database constructed based on internal local discharge true value signals and external electric field signals to obtain a current evolution stage determination result of internal interface discharge. The application can overcome strong electromagnetic interference and shielding effect without disassembling and stopping a train, and can realize accurate online diagnosis and early warning of the evolution stage of internal interface discharge of a vehicle-mounted high-voltage cable terminal.

Description

Technical Field

[0001] This invention relates to the field of high-voltage cable insulation testing, specifically to a method and related apparatus for detecting internal interface discharge in vehicle-mounted high-voltage cable terminals. Background Technology

[0002] Onboard cable terminals are core components of the train traction power supply system, responsible for transmitting electrical energy from the traction substation to the train's traction motors. Unlike fixed cable terminals in substations, onboard cable terminals operate in a complex, high-speed environment, enduring multiple stress coupling effects such as high-speed airflow erosion, rapid temperature fluctuations, and frequent overvoltage surges from traction system operations. Under these harsh operating conditions, partial discharge is highly likely to occur at the internal insulation interface of the onboard cable terminal. This discharge phenomenon can gradually develop and evolve, eventually leading to insulation breakdown and causing a train power outage, seriously threatening the operational safety of high-speed railways.

[0003] Currently, the main technical method for evaluating the insulation performance of vehicle-mounted cable terminals is partial discharge testing. This method assesses the insulation status by detecting the partial discharge signal generated during the application of a test voltage to the cable terminal. However, existing partial discharge testing techniques have fundamental limitations:

[0004] During testing, the onboard cable terminals must be completely removed from the train and transferred to a dedicated testing site equipped with an electromagnetic shielding room to avoid the influence of external electromagnetic interference on the detection of weak discharge signals. This testing method necessitates the train's shutdown, severely disrupting the operation of high-speed trains. Furthermore, the testing cycle is long and costly, and online monitoring during operation is not feasible. Summary of the Invention

[0005] In view of the above problems, this application provides a method and related device for detecting internal interface discharge of vehicle-mounted high-voltage cable terminals, so as to achieve real-time and accurate diagnosis of the development state of internal insulation defects in cable terminals under current vehicle operating conditions. The specific solution is as follows:

[0006] The first aspect of this application provides a method for detecting internal interface discharge in a vehicle-mounted high-voltage cable terminal, comprising:

[0007] Obtain time-varying electric field data near the surface of the cable terminal that is synchronized with the power frequency phase;

[0008] Spatial difference calculation is performed on the time-varying electric field data collected at at least two ungrounded measuring points to obtain the near-surface electric field intensity difference value between each ungrounded measuring point. Based on the spatial positional relationship of the ungrounded measuring points, the near-surface electric field intensity difference value is arranged into near-surface electric field gradient distribution data. The ungrounded measuring points are located on the outer insulating surface at the external electric field leakage window of the vehicle-mounted high-voltage cable terminal.

[0009] Phase-resolved analysis was performed on the near-surface electric field gradient distribution data to extract the differential characteristic quantities of the external electric field;

[0010] The differential characteristic of the external electric field is input into the pre-trained internal discharge stage identification model to obtain the current evolution stage determination result of the internal interface discharge of the cable terminal.

[0011] In one possible implementation, the external electric field leakage window includes the axial projection area of ​​the stress cone of the cable terminal and the root of the shed; and / or, the time-varying electric field data is acquired by an electric field sensor array disposed on the insulating outer surface of the external electric field leakage window of the vehicle-mounted high-voltage cable terminal.

[0012] In one possible implementation, the at least two ungrounded measurement points include: a first measurement point and a second measurement point;

[0013] The step involves performing spatial difference calculations on time-varying electric field data collected at at least two ungrounded measurement points to obtain near-surface electric field intensity difference values ​​between each ungrounded measurement point, and arranging these near-surface electric field intensity difference values ​​into near-surface electric field gradient distribution data based on the spatial positional relationship of the ungrounded measurement points. This includes:

[0014] Spatial difference calculation is performed on the time-varying electric field data collected from the first and second measuring points to obtain the near-surface electric field intensity difference value between the first and second measuring points. Based on the spatial positional relationship between the first and second measuring points, the near-surface electric field intensity difference value is arranged into near-surface electric field gradient distribution data.

[0015] In one possible implementation, the first measuring point is located on the outer surface of the insulation of the stress cone axial projection area of ​​the cable terminal, and the second measuring point is located on the outer surface of the insulation of the shed root.

[0016] In one possible implementation, the step of performing phase-resolved analysis on the near-surface electric field gradient distribution data to extract the differential features of the external electric field includes:

[0017] Using the contact network voltage of the vehicle-mounted high-voltage cable or the secondary voltage of the vehicle-mounted transformer as the phase reference, the near-surface electric field gradient distribution data is mapped to a power frequency phase window from 0° to 360° to obtain phase-resolved electric field gradient data.

[0018] The phase-resolved electric field gradient data is decomposed into tangential and normal components;

[0019] The tangential and normal components are respectively subjected to first- to third-order numerical differentiation to obtain the tangential component differential and the normal component differential;

[0020] Numerical filtering and noise reduction processing is performed on the differentials of the tangential and normal components;

[0021] The tangential and normal component derivatives after numerical filtering and noise reduction are grouped according to tangential and normal directions, and phase alignment is performed using dynamic time warping.

[0022] Principal component analysis is used to perform joint dimensionality reduction on the phase-aligned tangential and normal component differentials to extract high-variance principal components. The top k principal components whose cumulative variance contribution rate exceeds a preset threshold are used as the differential characteristic quantities of the external electric field.

[0023] In one possible implementation, the internal discharge stage identification model is trained on a tagged database constructed by synchronously collecting internal partial discharge truth signals and external electric field signals on a cable terminal test specimen with prefabricated interface defects. The tagged database uses the discharge evolution stage determined by the internal partial discharge truth signals as a label; and / or, the discharge evolution stage includes an initial stage, a development stage, and a severe degradation stage.

[0024] In one possible implementation, before inputting the differential characteristic of the external electric field into the internal discharge stage identification model, the following is also included:

[0025] Obtain the current ambient temperature and load current of the cable terminal;

[0026] Based on the pre-established temperature-load-electric field baseline drift model, temperature compensation and load compensation are performed on the differential characteristic of the external electric field.

[0027] The temperature-load-electric field baseline drift model is a compensation model constructed based on multivariate polynomial regression. Its inputs are ambient temperature and load current, and its output is the baseline offset of the differential characteristic of the external electric field.

[0028] The second aspect of this application provides a detection system for internal interface discharge of a vehicle-mounted high-voltage cable terminal, comprising:

[0029] The data acquisition unit is configured to acquire time-varying electric field data of the near-surface of the cable terminal that is synchronized with the power frequency phase.

[0030] The processing unit is configured to perform spatial difference operations on time-varying electric field data collected at at least two ungrounded measurement points to obtain near-surface electric field intensity difference values ​​between each ungrounded measurement point, and arrange the near-surface electric field intensity difference values ​​into near-surface electric field gradient distribution data based on the spatial positional relationship of the ungrounded measurement points, wherein the ungrounded measurement points are located on the insulating outer surface at the external electric field leakage window of the vehicle-mounted high-voltage cable terminal; perform phase-resolved analysis on the near-surface electric field gradient distribution data to extract the differential feature quantity of the external electric field; input the differential feature quantity of the external electric field into a pre-trained internal discharge stage identification model to obtain the current evolution stage determination result of the internal interface discharge of the cable terminal.

[0031] A third aspect of this application provides a computer program product, including computer-readable instructions, which, when executed on an electronic device, cause the electronic device to implement the detection method for internal interface discharge of a vehicle-mounted high-voltage cable terminal as described in the first aspect or any implementation thereof.

[0032] The fourth aspect of this application provides a computer storage medium carrying one or more computer programs, which, when executed by an electronic device, enable the electronic device to implement the detection method for internal interface discharge of a vehicle-mounted high-voltage cable terminal as described in the first aspect or any implementation thereof.

[0033] This invention provides a method and related apparatus for detecting internal interface discharge in vehicle-mounted high-voltage cable terminals. It can obtain time-varying electric field data near the cable terminal surface, synchronized with the power frequency phase. This non-invasive method acquires raw electric field information reflecting the internal insulation state. Spatial difference calculations are then performed on the time-varying electric field data collected at at least two ungrounded measurement points to obtain the near-surface electric field intensity difference values ​​between each measurement point. Based on the spatial relationship of the measurement points, the difference values ​​are arranged into near-surface electric field gradient distribution data. Spatial difference processing effectively suppresses background common-mode interference and highlights the local electric field distortion characteristics caused by internal interface defects. Furthermore, phase-resolved analysis is performed on the electric field gradient distribution data to extract differential features of the external electric field. Through phase synchronization and multi-dimensional differential calculations, noise is further removed and the transient response law of the electric field with phase change is quantified, thereby enhancing the sensitivity of the features to weak discharge signals. Subsequently, the extracted differential features of the external electric field are input into a pre-trained internal discharge stage identification model to finally obtain the current evolution stage determination result of the internal interface discharge of the cable terminal. Therefore, this solution effectively solves the problem that existing on-board high-voltage cable terminals must be disassembled off the train before they can be tested in a dedicated area. It avoids train downtime and disruption to operational order caused by disassembly and testing, and enables real-time and accurate diagnosis of the development of internal insulation defects in cable terminals under current train conditions, thereby improving the level of intelligence and safety assurance capabilities of high-speed rail operation and maintenance. Attached Figure Description

[0034] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0035] Figure 1 A flowchart of a method for detecting internal interface discharge of a vehicle-mounted high-voltage cable terminal provided in this application;

[0036] Figure 2 This is a schematic diagram of a vehicle-mounted cable terminal provided in this application.

[0037] The annotations in the attached figures are explained as follows:

[0038] Stress cone axial projection area 001, umbrella skirt root 002, stress cone structure 003, umbrella skirt 004, insulating outer surface 005, first measuring point 006, second measuring point 007, grounding flange 008, XLPE insulation layer 009, high voltage conductor 010. Detailed Implementation

[0039] The embodiments of this application are described below with reference to the accompanying drawings. The terminology used in the implementation section of this application is for explaining specific embodiments only and is not intended to limit the scope of this application.

[0040] As will be known to those skilled in the art, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0041] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms are interchangeable where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of elements is not necessarily limited to those elements, but may include other elements not explicitly listed or inherent to those processes, methods, products, or apparatuses.

[0042] like Figure 1 As shown in the embodiment of this application, a method for detecting internal interface discharge of a vehicle-mounted high-voltage cable terminal may include the following steps:

[0043] Step S101: Obtain time-varying electric field data near the surface of the cable terminal that is synchronized with the power frequency phase.

[0044] Optionally, the time-varying electric field data can be acquired by an electric field sensor array, wherein the electric field sensor array can be disposed on the insulating outer surface at the external electric field leakage window of the vehicle-mounted high-voltage cable terminal.

[0045] The electric field sensor array can be a group of highly sensitive electric field detection units deployed in a specific area of ​​the cable terminal to non-contactly sense changes in the electric field intensity on the insulation surface. The external electric field leakage window can be a detectable area where internal electric field distortions in the cable terminal can effectively penetrate the insulation layer and propagate to the outside; this typically includes areas where the electric field is concentrated or prone to distortion, such as the axial projection area of ​​the stress cone and the root of the shed. The time-varying electric field data synchronized with the power frequency phase can be a sequence of electric field intensity that is phase-locked with the contact network voltage of the vehicle-mounted high-voltage cable or the secondary voltage of the vehicle-mounted transformer and varies continuously over time.

[0046] Optionally, the electric field sensor array can be constructed using at least one of optical electric field sensors, miniature capacitive electric field probes, or MEMS electric field sensor arrays, with these sensors tightly fitted or fixedly mounted at the leakage window location on the outer surface of the cable termination insulation. For example... Figure 2 As shown in the figure, this is a schematic diagram of a vehicle-mounted high-voltage cable terminal. The terminal may include: a stress cone structure 003, a shed 004, an outer insulating surface 005, an XLPE insulation layer 009, a grounding flange 008, and a high-voltage conductor 010. The sensor array is positioned at the critical leakage window of the outer insulating surface 005 to capture weak electric field fluctuations caused by internal discharge. During data acquisition, the system uses the power grid frequency signal as a reference to trigger sampling, ensuring that the acquired electric field data has clear phase information, enabling subsequent phase-resolved analysis. For example, when a small interface defect occurs inside the cable terminal, the resulting local electric field distortion will penetrate the insulation layer and form a specific electric field distribution pattern on the outside. The sensor array records the trajectory of these patterns changing with the power frequency cycle in real time. Through this multi-point synchronous acquisition method, the dynamic behavior of the electric field near the surface of the cable terminal can be completely reproduced, providing the original data foundation for subsequent spatial differential processing.

[0047] It should be noted that, as Figure 2As shown, the electric field leakage window includes the axial projection area 001 of the stress cone at the cable terminal and the root of the shed 002. The axial projection area of ​​the stress cone is the mapping range of the core region of the cable terminal where the internal electric field is concentrated and interface slip discharge is prone to occur, projected onto the outer surface of the insulation. Its location corresponds to the axial extension of the internal stress cone structure. The root of the shed is the recessed area where the outer insulation shed of the cable terminal connects to the main insulation body. This area easily accumulates dirt and moisture and generates electric field distortion, making it a sensitive point for external flashover and internal defect signal leakage. For example, in actual arrangements, the axial projection area of ​​the stress cone can be defined as an annular strip extending 50mm to 100mm above and below the end of the stress cone, while the root of the shed is selected from the root groove surface of the first to third sheds closest to the high-voltage conductor. Optical electric field sensors, based on the Paulcke effect, are devices that convert electric field intensity into optical signal changes using electro-optic crystals, exhibiting complete electrical isolation. Miniature capacitive electric field probes acquire electric field information through tiny capacitive coupling between the sensing electrode and ground, characterized by small size and fast response. MEMS electric field sensor arrays integrate multiple miniature electric field sensing units on a silicon substrate using micro-nano fabrication technology, enabling high spatial resolution electric field distribution measurement. In vehicle environments with strong electromagnetic interference and high voltage levels, optical electric field sensors can be used as the primary acquisition element, attached to the insulating surface of the stress cone's axial projection area, and using optical fiber to transmit signals to completely eliminate electric field disturbances introduced by metal leads. Alternatively, in the space-constrained base of the skirt, a millimeter-sized MEMS electric field sensor array can be densely arranged to capture subtle changes in electric field gradient. These sensor types all possess characteristics of being ungrounded, miniaturized, and highly resistant to interference, meeting the long-term stable monitoring requirements under harsh conditions such as train operation vibration and temperature changes.

[0048] Based on the above, by defining the electric field leakage window as the axial projection area of ​​the stress cone and the root of the skirt, and combining optical, miniature capacitive, or MEMS electric field sensors, the detection scheme achieves synergistic optimization in spatial positioning and hardware adaptation. The axial projection area of ​​the stress cone and the root of the skirt, as the shortest path and strongest radiation area for the outward propagation of internal defect signals, provide the optimal placement for the sensors. Optical or MEMS sensors, with their non-contact, non-metallic disturbance, and high sensitivity characteristics, can accurately capture the weak time-varying electric field information within these specific areas. The combined use of these two types of sensors not only avoids the problem of traditional metal sensors altering the local electric field distribution due to their own presence, but also significantly improves the ability to capture early characteristics of internal interface discharge under high-speed train movement conditions, thereby ensuring the diagnostic accuracy of retrieving the internal discharge evolution stage based on external electric field data.

[0049] Step S102: Perform spatial difference calculation on the time-varying electric field data collected at at least two ungrounded measuring points to obtain the near-surface electric field intensity difference value between each ungrounded measuring point, and arrange the near-surface electric field intensity difference value into near-surface electric field gradient distribution data based on the spatial position relationship of the ungrounded measuring points, wherein the ungrounded measuring points are located on the outer surface of the insulation at the external electric field leakage window of the vehicle-mounted high-voltage cable terminal.

[0050] Among them, ungrounded measuring points refer to measurement locations directly arranged on the surface of the insulating medium and not directly short-circuited to the ground potential, used to sense changes in the electric field caused by floating potential or local charge accumulation. Spatial differential calculation refers to using data from adjacent or spatially distributed measuring points to calculate the rate of change of electric field intensity in the spatial domain, used to eliminate background common-mode interference and highlight local anomalous signals. Near-surface electric field gradient distribution data is a two-dimensional or three-dimensional field distribution matrix formed by recombining the calculated differential values ​​according to the actual physical coordinates of the measuring points on the cable terminal surface, reflecting the steepness and directionality of the electric field intensity in space.

[0051] Optionally, the at least two ungrounded measuring points include: a first measuring point and a second measuring point; the step of performing spatial differential calculation on the time-varying electric field data collected at the at least two ungrounded measuring points to obtain the near-surface electric field intensity difference value between each measuring point, and arranging the difference value into near-surface electric field gradient distribution data based on the spatial positional relationship of the measuring points, includes: performing spatial differential calculation on the time-varying electric field data collected at the first measuring point and the second measuring point to obtain the near-surface electric field intensity difference value between the first measuring point and the second measuring point, and arranging the near-surface electric field intensity difference value into near-surface electric field gradient distribution data based on the spatial positional relationship between the first measuring point and the second measuring point. The first measuring point and the second measuring point are two specific monitoring locations arranged on the outer surface of the insulation at the external electric field leakage window of the vehicle-mounted high-voltage cable terminal, used as the basic nodes for constructing the spatial differential calculation. The first measuring point and the second measuring point have a definite relative positional relationship in space, and both are located in ungrounded areas to avoid shielding or interfering with weak discharge signals by ground potential. The selection of the first and second measuring points is based on the geometric characteristics of the cable terminal and the sensitive area for electric field distortion. For example, Figure 2 As shown, the first measuring point 006 is located on the outer surface of the insulation of the stress cone axial projection area of ​​the cable terminal, and the second measuring point 007 is located on the outer surface of the insulation of the shed root. The straight-line distance between the two points is set to between 50mm and 200mm according to the size ratio of the cable terminal. Optionally, there can be one or more second measuring points. By fixing the spatial coordinates of these two measuring points, a stable local electric field reference system can be established, ensuring the repeatability and comparability of subsequent differential calculation results.

[0052] The stress cone axial projection area refers to the vertical projection region of the internal stress cone structure of the vehicle-mounted high-voltage cable terminal onto the surface of the external insulation sheath. This region is where the electric field is most concentrated inside the cable terminal and is most sensitive to interface defects. The first measuring point is set in this region to capture the weak electric field distortion signal caused by the initial stage of internal interface discharge. The shed root refers to the annular region where the external insulation shed of the cable terminal connects to the main insulation body. This region is susceptible to mechanical stress concentration and environmental pollution accumulation, and is a critical path for discharge to extend to the surface. The second measuring point is set in this region to monitor the trend of electric field diffusion towards the outer edge during discharge evolution. The first and second measuring points are located in different functional areas of the cable terminal, forming a differential observation pair spanning the core high-field-strength region and the edge transition region. Through this spatial layout, the first measuring point mainly responds to the local field strength abrupt change caused by internal defects, while the second measuring point mainly responds to the propagation and attenuation characteristics of the field strength gradient. For example, when the internal interface is in the initial discharge stage, the differential electric field characteristics of the stress cone axial projection region will fluctuate significantly, while the changes at the root of the umbrella skirt will lag relatively. As the discharge enters the development stage, the difference in electric field intensity between the two measurement points will show a specific increasing trend. By selecting these two specific locations as ungrounded measurement points, it is possible to effectively construct a spatial electric field gradient distribution that reflects the evolution of the internal discharge from the core to the edge, thereby significantly improving the sensitivity of early weak defects and the ability to judge the direction of discharge development.

[0053] Based on the above, by explicitly defining the first and second measurement points as the specific execution objects of spatial difference operations, and specifying the arrangement of gradient distribution data based on the spatial positional relationship of these two points, the standardization and specification of electric field gradient calculation are achieved. The fixed configuration of the first and second measurement points eliminates the data inconsistency problem caused by the arbitrariness of measurement point selection, making the detection results of different batches or different vehicles horizontally comparable. By performing real-time difference operations on the data of these two points, not only is common-mode noise filtered out, but the spatial change rate of the local electric field is also directly quantified, transforming the abstract electric field distribution into a specific gradient numerical sequence. On this basis, the gradient distribution data is constructed by combining the spatial geometric relationship of the two points, enabling the subsequent feature extraction algorithm to accurately capture the directionality and intensity information of electric field distortion, thereby more accurately inverting the evolution state of internal interface discharge, effectively solving the problems of fuzzy gradient calculation and low diagnostic accuracy caused by unclear measurement point layout in existing technologies.

[0054] Step S103: Perform phase-resolved analysis on the near-surface electric field gradient distribution data to extract the differential characteristic quantity of the external electric field.

[0055] Phase resolution analysis refers to the process of mapping the electric field gradient data in the time domain to the power frequency phase domain (0° to 360°) and performing statistical or waveform analysis according to the phase window. The differential characteristic quantity of the external electric field refers to a high-dimensional characteristic vector that can characterize the rapid change characteristics of the discharge pulse after phase alignment, component decomposition and numerical differentiation. It usually contains the first to third order differential statistical values ​​of the tangential and normal components.

[0056] Optionally, the step of performing phase-resolved analysis on the electric field gradient distribution data to extract the differential features of the external electric field includes: using the contact network voltage of the vehicle-mounted high-voltage cable or the secondary voltage of the vehicle-mounted transformer as the phase reference, mapping the near-surface electric field gradient distribution data to a 0° to 360° power frequency phase window to obtain phase-resolved electric field gradient data; decomposing the phase-resolved electric field gradient data into tangential and normal components; performing first- to third-order numerical differentiation on the tangential and normal components respectively to obtain tangential component differentials and normal component differentials; performing numerical filtering and noise reduction processing on the tangential component differentials and normal component differentials; grouping the tangential component differentials and normal component differentials after numerical filtering and noise reduction according to tangential and normal directions respectively, and performing phase alignment using dynamic time warping; using principal component analysis to perform joint dimensionality reduction on the phase-aligned tangential component differentials and normal component differentials, extracting high-variance principal components, and using the top k principal components whose cumulative variance contribution rate exceeds a preset threshold as the differential features of the external electric field.

[0057] The phase reference can refer to a reference signal used to determine the starting point of the power frequency electric field change cycle, originating from the power supply end of the vehicle-mounted high-voltage system. Specifically, this phase reference can be obtained by directly acquiring the voltage signal of the contact network or by acquiring the voltage signal of the secondary side of the vehicle-mounted transformer. Both methods can provide a phase reference that is strictly synchronized with the internal discharge activity of the cable terminal. Mapping the electric field gradient distribution data to a 0°~360° power frequency phase window can refer to using the zero-crossing point or peak point of the phase reference as the origin of the time axis, and converting the electric field gradient data within a complete power frequency cycle into phase angle coordinates according to the time ratio, thereby forming phase-resolved electric field gradient data. For example, when the power frequency is 50Hz, one cycle is 20ms. If the time from the zero-crossing point of the phase reference at a certain moment is 5ms, then the corresponding phase angle is 90°. Through this mapping method, the electric field data that changes continuously with time can be reorganized into a data sequence distributed with phase angle. Phase synchronization ensures the consistency between electric field data and power supply cycle, facilitating subsequent identification of the clustering characteristics of discharge pulses in specific phase intervals and effectively eliminating data alignment difficulties caused by the randomness of sampling start time.

[0058] The tangential component refers to the electric field intensity component along the tangential direction of the cable terminal insulation surface, while the normal component refers to the electric field intensity component perpendicular to the cable terminal insulation surface. The decomposition process is based on vector decomposition theory. Utilizing the spatial orientation information of an array of electric field sensors arranged on the outer surface of the insulation, the obtained phase-resolved electric field gradient data is projected onto a predefined local coordinate system. Specifically, for each measurement point, based on its geometric position relative to the cable axis, an orthogonal coordinate system containing tangential and normal basis vectors is constructed. The total electric field gradient vector is then projected onto these two directions, resulting in independent sequences of tangential and normal components. For example, in the stress cone region, the tangential component mainly reflects the electric field distortion caused by surface slip discharge, while the normal component mainly reflects the electric field abrupt change caused by breakdown discharge perpendicular to the interface. By separating the two, the electric field response caused by different physical mechanisms can be distinguished. The independent decomposition of the tangential and normal components allows subsequent processing to extract more targeted features for different types of discharge modes, avoiding feature ambiguity caused by multi-directional signal aliasing.

[0059] Numerical differentiation refers to calculating the rate of change, acceleration, and jerk of electric field components with phase using discrete difference algorithms. Specifically, for the obtained tangential and normal component sequences, the rate of change is calculated using the first-order difference formula (first-order differential), the curvature change is calculated using the second-order difference formula (second-order differential), and the abrupt change is calculated using the third-order difference formula (third-order differential). Multi-order differential operations can significantly highlight the transient characteristics of electric field waveforms. The first-order differential reflects the increasing or decreasing trend of electric field intensity, the second-order differential captures the inflection point of electric field distortion, and the third-order differential has extremely high sensitivity to small partial discharge signals, thereby amplifying weak discharge signals submerged in background noise.

[0060] Numerical filtering denoising refers to using digital filters to remove high-frequency noise and environmental interference introduced during differentiation operations. Since differentiation amplifies high-frequency noise, the resulting tangential and normal derivatives must be smoothed. Specifically, wavelet thresholding, moving average filtering, or Butterworth low-pass filtering can be used, setting appropriate cutoff frequencies or thresholds to retain low-frequency and mid-frequency components relevant to partial discharge characteristics while filtering out random noise above the discharge characteristic frequency.

[0061] Dynamic Time Warping (DTW) is an algorithm used to measure the similarity between two time series. It achieves optimal matching of sequences by nonlinearly stretching or compressing the time axis. Specifically, the purified tangential component differential data are grouped into one group, and the normal component differential data into another. Within each group, a standard reference template sequence (such as the differential waveform of a typical initial discharge stage) is selected. The DTW algorithm calculates the optimal alignment path between the sequence to be processed and the reference template, and performs a nonlinear transformation on the phase axis of the sequence to be processed to ensure that its key feature points (such as pulse peaks and zero crossings) strictly correspond in phase with the reference template. For example, when the cable terminal is in a severely deteriorated stage, the discharge pulse may experience phase lag due to space charge accumulation, causing its peak to appear at a different phase angle than in the initial stage. Through DTW alignment, the lagging pulse can be pulled back to the standard phase position, eliminating this phase drift caused by different discharge development stages. This step overcomes the risk of misjudgment caused by the inconsistency of discharge phase between different samples, enabling subsequent feature extraction to be based on a unified phase benchmark, significantly improving the consistency and comparability of features.

[0062] Principal Component Analysis (PCA) is a statistical method that transforms a set of potentially correlated variables into a set of linearly uncorrelated variables through orthogonal transformations, known as principal components. Specifically, the aligned tangential and normal component differential data are merged into a high-dimensional feature matrix. The covariance matrix of this matrix is ​​calculated, and its eigenvalues ​​and eigenvectors are solved. The eigenvectors are arranged in descending order of eigenvalues ​​to obtain several principal components. The variance contribution rate and cumulative variance contribution rate of each principal component are calculated. A preset threshold (e.g., 95% or 98%) is set, and the first k principal components whose cumulative variance contribution rate first exceeds this threshold are selected. Joint dimensionality reduction using PCA can significantly reduce data dimensionality while preserving key discriminative information in the original data to the greatest extent, removing redundant information and secondary noise. This results in extracted external electric field differential features with higher representational power and discriminative power, providing high-quality feature input for subsequent input recognition models.

[0063] It should be noted that before inputting the differential characteristic quantity of the external electric field into the internal discharge stage identification model, the method further includes: obtaining the current ambient temperature and load current of the cable terminal; and performing temperature compensation and load compensation on the differential characteristic quantity of the external electric field according to the pre-established temperature-load-electric field baseline drift model to eliminate the electric field baseline drift caused by ambient temperature and load current. The temperature-load-electric field baseline drift model is a compensation model constructed based on multivariate polynomial regression, with ambient temperature and load current as inputs and baseline offset of the differential characteristic quantity of the external electric field as output.

[0064] Ambient temperature refers to the real-time temperature of the air surrounding the cable terminal, reflecting the influence of the external thermal state on the dielectric constant and conductivity of the cable terminal insulation material. Load current refers to the real-time current flowing through the conductor of the vehicle-mounted high-voltage cable, determining the conductor's heating power and the resulting internal temperature field distribution. Ambient temperature and load current are acquired in real-time by temperature sensors located near the cable terminal and current transformers in the vehicle-mounted traction system. Ambient temperature and load current, as key operating parameters affecting the external electric field measurement benchmark, together form the basis of the input variables for subsequent compensation calculations. The temperature-load-electric field baseline drift model is a mathematical model used to predict and correct electric field reading deviations caused by non-discharge signals. This model is a compensation model based on multivariate polynomial regression, with the aforementioned ambient temperature and load current as input variables and the baseline deviation of the differential characteristic quantity of the external electric field as the output variable. Specifically, this model establishes a nonlinear mapping relationship between ambient temperature, load current, and the benchmark value of the differential characteristic quantity of the electric field by fitting a large amount of historical experimental or simulation data. When the ambient temperature rises, the polarization characteristics of the insulating material change, leading to a change in the background electric field. Similarly, when the load current increases, the thermal expansion and changes in dielectric properties caused by the conductor's temperature rise also result in distortion of the electric field distribution. The purpose of this model is to calculate these spurious electric field changes caused by changes in operating conditions rather than internal discharge, i.e., baseline offset.

[0065] Based on the above, the synergistic effect of the above steps achieves accurate extraction of highly discriminative differential features from raw electric field data. First, mapping is performed using the contact network voltage or transformer secondary voltage as a phase reference, establishing a time reference system for data analysis and ensuring the synchronization of electric field data with the discharge head. On this basis, orthogonal decomposition of tangential and normal directions decouples the complex three-dimensional electric field problem into two independent physical dimensions, capturing different response characteristics of surface slip and vertical breakdown respectively. Furthermore, by performing first- to third-order numerical differentiation on the two components and combining it with filtering and noise reduction, not only is the weak transient discharge signal amplified, but the noise amplification effect caused by differentiation is also effectively suppressed, highlighting the transient nature of the discharge. Subsequently, dynamic time warping technology is used to nonlinearly align the pulse phases at different development stages, eliminating phase drift interference caused by different degrees of insulation aging and ensuring the consistency of the feature extraction benchmark. Finally, principal component analysis is used to jointly reduce the dimensionality of the multidimensional differential data, eliminating redundant information and extracting the core principal component with the highest variance contribution rate as the external electric field differential feature. This series of interconnected processing steps not only significantly improves the robustness of feature extraction but also ensures that the extracted features accurately characterize the evolution of internal interface discharge, laying a solid data foundation for subsequent precise determination of the discharge stage. Simultaneously, by acquiring real-time ambient temperature and load current, and combining this with a temperature-load-electric field baseline drift model constructed based on multivariate polynomial regression, dynamic compensation of the differential features of the external electric field of the vehicle-mounted high-voltage cable terminal is achieved. Based on this, the baseline offset output by the model is used to correct the original features, thereby eliminating measurement errors caused by the frequently changing thermal-electric conditions during train operation. Furthermore, this compensation mechanism works closely with the aforementioned internal discharge stage identification model: the compensated features more purely reflect the physical nature of the internal interface discharge, enabling the identification model to maintain high sensitivity and accuracy under various extreme conditions, effectively solving the technical problem of unstable diagnostic results due to environmental interference in existing technologies.

[0066] Step S104: Input the differential characteristic quantity of the external electric field into the pre-trained internal discharge stage identification model to obtain the current evolution stage determination result of the internal interface discharge of the cable terminal.

[0067] Optionally, the internal discharge stage identification model is trained based on a tagged database constructed from the internal partial discharge true value signal and the external electric field signal synchronously collected on the cable terminal test specimen with prefabricated interface defects. The tagged database uses the discharge evolution stage determined by the internal partial discharge true value signal as the tag.

[0068] The internal discharge stage identification model is a classifier built based on machine learning or deep learning algorithms to establish a nonlinear mapping relationship between the differential characteristics of the external electric field and the internal discharge evolution state. The current evolution stage determination result can refer to the label output by the model regarding the specific stage of the discharge at the internal interface of the cable terminal, which typically includes the initial stage, the development stage, and the severe degradation stage (or the precursor stage of breakdown).

[0069] Optionally, the internal discharge stage identification model includes a nonlinear mapping model, which is constructed based on random forest, gradient boosting tree, or lightweight neural network.

[0070] Optionally, the discharge evolution stages include an initial stage, a development stage, and a severe degradation stage.

[0071] The internal discharge stage identification model refers to a computational architecture used to establish a complex nonlinear relationship between the differential features of the external electric field and the internal insulation state of the cable terminal. Specifically, this model includes a nonlinear mapping module, whose core function is to map the high-dimensional, nonlinear differential feature vector of the electric field to a discrete discharge state space. The nonlinear mapping model is trained using machine learning algorithms, and its source can be at least one of a random forest classifier, a gradient boosting tree (GBDT) classifier, or a lightweight neural network classifier built based on historical experimental data.

[0072] The initial stage is defined by detecting the first occurrence of partial discharge activity with low amplitude and infrequent frequency. Its function is to characterize the initial formation of microscopic defects at the insulation interface, at which point the differential characteristic quantity of the external electric field exhibits slight fluctuations. The development stage is determined by the increase in discharge amplitude, the expansion of the phase distribution range, and a significant increase in pulse frequency. Its function is to characterize the expansion of the defect region and the enhancement of the charge accumulation effect. At this point, the numerical gradient and rate of change of the differential characteristic quantity of the external electric field increase significantly. The severe degradation stage is determined by a sharp increase in discharge energy, the appearance of continuous discharge pulses, or the accompanying significant electric field distortion. Its function is to warn of impending insulation breakdown or flashover faults, at which point the differential characteristic quantity of the external electric field reaches a preset critical threshold. For example, when the model output is in the initial stage, it indicates that there is initial insulation aging inside the cable terminal, and planned maintenance can be arranged. When the output is in the severe degradation stage, it indicates that the cable terminal faces the risk of immediate breakdown, and immediate shutdown or load reduction measures are required. By quantifying the continuous physical degradation process into these three distinct discrete stages, it is possible to provide maintenance personnel with intuitive status indicators, which facilitates the development of graded response strategies and effectively avoids sudden power outages.

[0073] It is important to note that the construction of a tagged database is the cornerstone of model accuracy. In a laboratory environment, cable terminal test specimens containing typical interface defects (such as air gaps, impurities, and semiconductive layer protrusions) are prepared. Internal partial discharge truth signals are simultaneously acquired inside the test specimen or on the grounding wire using a high-frequency current transformer (HFCT) or ultrasonic sensor; this signal is considered the gold standard for determining the true state of the discharge. Simultaneously, the corresponding external electric field signal is acquired using an array of electric field sensors arranged on the outer surface of the test specimen. By analyzing the phase-resolved pulse sequence (PRPS) or phase-resolved partial discharge (PRPD) spectra of the internal truth signals, the discharge evolution stages (e.g., initiation, development, severe degradation) within each time period are manually or automatically labeled. These labeled stages are used as tags and paired with the simultaneously extracted differential features of the external electric field to form feature-tag sample pairs, which are then incorporated into the database.

[0074] Based on the above, by inputting the differential features of the external electric field into a nonlinear mapping model constructed using random forests, gradient boosting trees, or lightweight neural networks, and leveraging the powerful nonlinear fitting capabilities of these models, the mapping relationship between the electric field signal and the internal discharge state under complex operating conditions can be effectively decoupled. This overcomes the limitation of traditional linear methods in accurately characterizing high-frequency transient discharge features. Furthermore, by combining a three-classification output mechanism for the initial, development, and severe degradation stages, not only is the degree of insulation degradation quantified, but abstract electric field data is also transformed into maintenance instructions with clear engineering significance. This combination of model architecture and stage division enables the diagnostic system to operate efficiently on onboard edge computing devices and provide highly interpretable health status assessments, thereby significantly improving the operational safety and maintenance efficiency of the high-speed train high-voltage power supply system.

[0075] This application acquires power frequency phase synchronization data by deploying a sensor array at the electric field leakage window and combines it with spatial differential calculations at at least two ungrounded measurement points. This not only effectively suppresses strong electromagnetic background interference in the train operating environment but also significantly enhances the sensitivity to local electric field distortion, enabling the accurate capture of weak internal discharge signals. Based on this, phase-resolved analytical techniques are used to extract feature quantities containing multi-order differential information in the tangential and normal directions, fully exploring the deep evolution laws of the discharge pulse in the phase and time differential domains, solving the problem that traditional amplitude monitoring cannot distinguish discharge stages. Furthermore, by using an internal discharge stage identification model trained on the true value signal of a pre-fabricated defective sample, a reliable nonlinear mapping between the differential characteristics of the external electric field and the internal discharge evolution stages is established, overcoming the limitation of existing methods that require disassembly and inspection off-board. Ultimately, this method can assess the development degree of internal interface defects in cable terminals in real time and quantitatively without interfering with normal train operation, providing strong technical support for the safe maintenance of high-speed rail power supply systems.

[0076] This application also provides a detection system for internal interface discharge of a vehicle-mounted high-voltage cable terminal, which may include:

[0077] The data acquisition unit is configured to acquire time-varying electric field data of the near-surface of the cable terminal that is synchronized with the power frequency phase.

[0078] The processing unit is configured to perform spatial difference operations on time-varying electric field data collected at at least two ungrounded measurement points to obtain near-surface electric field intensity difference values ​​between each ungrounded measurement point, and arrange the near-surface electric field intensity difference values ​​into near-surface electric field gradient distribution data based on the spatial positional relationship of the ungrounded measurement points, wherein the ungrounded measurement points are located on the insulating outer surface at the external electric field leakage window of the vehicle-mounted high-voltage cable terminal; perform phase-resolved analysis on the near-surface electric field gradient distribution data to extract the differential feature quantity of the external electric field; input the differential feature quantity of the external electric field into a pre-trained internal discharge stage identification model to obtain the current evolution stage determination result of the internal interface discharge of the cable terminal.

[0079] This application also provides a computer program product, including computer-readable instructions, which, when executed on an electronic device, cause the electronic device to implement any of the methods for detecting internal interface discharge of vehicle-mounted high-voltage cable terminals provided in this application.

[0080] This application also provides a computer storage medium carrying one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any of the methods for detecting internal interface discharge of vehicle-mounted high-voltage cable terminals provided in this application.

[0081] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components 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 embodiment according to actual needs. In addition, in the device embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0082] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0083] In the above embodiments, the implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, in the form of a computer program product.

[0084] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

Claims

1. A method for detecting internal interface discharge in a vehicle-mounted high-voltage cable terminal, characterized in that, include: Obtain time-varying electric field data near the surface of the cable terminal that is synchronized with the power frequency phase; Spatial difference calculation is performed on the time-varying electric field data collected at at least two ungrounded measuring points to obtain the near-surface electric field intensity difference value between each ungrounded measuring point. The near-surface electric field intensity difference value is arranged into near-surface electric field gradient distribution data based on the spatial position relationship of the ungrounded measuring points. The ungrounded measuring points are located on the outer insulating surface at the external electric field leakage window of the vehicle-mounted high-voltage cable terminal. Phase-resolved analysis was performed on the near-surface electric field gradient distribution data to extract the differential characteristic quantities of the external electric field; The differential characteristic of the external electric field is input into the pre-trained internal discharge stage identification model to obtain the current evolution stage determination result of the internal interface discharge of the cable terminal.

2. The method for detecting internal interface discharge of vehicle-mounted high-voltage cable terminals according to claim 1, characterized in that, The external electric field leakage window includes the axial projection area of ​​the stress cone of the cable terminal and the root of the shed; and / or, the time-varying electric field data is collected by an electric field sensor array, which is disposed on the insulating outer surface of the external electric field leakage window of the vehicle-mounted high-voltage cable terminal.

3. The method for detecting internal interface discharge of vehicle-mounted high-voltage cable terminals according to claim 1, characterized in that, The at least two ungrounded measurement points include: a first measurement point and a second measurement point; The step involves performing spatial difference calculations on time-varying electric field data collected at at least two ungrounded measurement points to obtain near-surface electric field intensity difference values ​​between each ungrounded measurement point, and arranging these near-surface electric field intensity difference values ​​into near-surface electric field gradient distribution data based on the spatial positional relationship of the ungrounded measurement points. This includes: Spatial difference calculation is performed on the time-varying electric field data collected from the first and second measuring points to obtain the near-surface electric field intensity difference value between the first and second measuring points. Based on the spatial positional relationship between the first and second measuring points, the near-surface electric field intensity difference value is arranged into near-surface electric field gradient distribution data.

4. The method for detecting internal interface discharge of vehicle-mounted high-voltage cable terminals according to claim 3, characterized in that, The first measuring point is located on the outer surface of the insulation of the stress cone axial projection area of ​​the cable terminal, and the second measuring point is located on the outer surface of the insulation of the umbel root.

5. The method for detecting internal interface discharge of vehicle-mounted high-voltage cable terminals according to claim 1, characterized in that, The step of performing phase-resolved analysis on the near-surface electric field gradient distribution data to extract the differential feature quantity of the external electric field includes: Using the contact network voltage of the vehicle-mounted high-voltage cable or the secondary voltage of the vehicle-mounted transformer as the phase reference, the near-surface electric field gradient distribution data is mapped to a power frequency phase window from 0° to 360° to obtain phase-resolved electric field gradient data. The phase-resolved electric field gradient data is decomposed into tangential and normal components; The tangential and normal components are respectively subjected to first- to third-order numerical differentiation to obtain the tangential component differential and the normal component differential; Numerical filtering and noise reduction processing is performed on the differentials of the tangential and normal components; The tangential and normal component derivatives after numerical filtering and noise reduction are grouped according to tangential and normal directions, and phase alignment is performed using dynamic time warping. Principal component analysis is used to perform joint dimensionality reduction on the phase-aligned tangential and normal component differentials to extract high-variance principal components. The top k principal components whose cumulative variance contribution rate exceeds a preset threshold are used as the differential characteristic quantities of the external electric field.

6. The method for detecting internal interface discharge of vehicle-mounted high-voltage cable terminals according to claim 1, characterized in that, The internal discharge stage identification model is trained on a tagged database constructed by synchronously collecting internal partial discharge true value signals and external electric field signals on a cable terminal test specimen with prefabricated interface defects. The tagged database uses the discharge evolution stage determined by the internal partial discharge true value signals as tags; and / or, the discharge evolution stage includes the initial stage, the development stage, and the severe degradation stage.

7. The method for detecting internal interface discharge of vehicle-mounted high-voltage cable terminals according to claim 1, characterized in that, Before inputting the differential characteristic of the external electric field into the internal discharge stage identification model, the method further includes: Obtain the current ambient temperature and load current of the cable terminal; Based on the pre-established temperature-load-electric field baseline drift model, temperature compensation and load compensation are performed on the differential characteristic of the external electric field. The temperature-load-electric field baseline drift model is a compensation model constructed based on multivariate polynomial regression. Its inputs are ambient temperature and load current, and its output is the baseline offset of the differential characteristic of the external electric field.

8. A detection system for internal interface discharge of a vehicle-mounted high-voltage cable terminal, characterized in that, include: The data acquisition unit is configured to acquire time-varying electric field data of the near-surface of the cable terminal that is synchronized with the power frequency phase. The processing unit is configured to perform spatial difference calculation on time-varying electric field data collected at at least two ungrounded measuring points to obtain near-surface electric field intensity difference values ​​between each ungrounded measuring point, and to arrange the near-surface electric field intensity difference values ​​into near-surface electric field gradient distribution data based on the spatial positional relationship of the ungrounded measuring points, wherein the ungrounded measuring points are located on the outer insulating surface at the external electric field leakage window of the vehicle-mounted high-voltage cable terminal; Phase-resolved analysis is performed on the near-surface electric field gradient distribution data to extract the differential feature quantity of the external electric field; the differential feature quantity of the external electric field is input into the pre-trained internal discharge stage identification model to obtain the current evolution stage determination result of the internal interface discharge of the cable terminal.

9. A computer program product, characterized in that, The method includes computer-readable instructions that, when executed on an electronic device, cause the electronic device to implement the method for detecting internal interface discharge of a vehicle-mounted high-voltage cable terminal as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The storage medium carries one or more computer programs, which, when executed by an electronic device, enable the electronic device to implement the detection method for internal interface discharge of a vehicle-mounted high-voltage cable terminal as described in any one of claims 1 to 7.

Images