A method and system for automatic identification and location detection of power cable fault type

Abstract

The application discloses a kind of power cable fault type automatic identification and positioning detection method and system, it is related to fault detection technical field, including: the electrical parameter signal of collection power cable and the electromagnetic field distribution data outside line, construct the double-domain feature space of cable operating state and execute space-time feature decomposition, form electrical characteristic sequence and electromagnetic field characteristic spectrum, and respectively input fault type identification module and fault location prediction module;The fault type identification result generated by fault type identification module and the fault location candidate area determined by fault location prediction module are integrated by feature cascading network, generate fusion decision result, determine fault type and location.The application solves the technical problem that the recognition ability of traditional single-domain information detection is limited and the positioning accuracy is unstable in multiple fault scenarios, through parallel processing and fusion decision of electrical parameters and electromagnetic field information, accurate identification and high-precision positioning of fault type are realized.

Description

Technical Field

[0001] This invention relates to the field of fault detection technology, specifically to a method and system for automatic identification and location detection of power cable fault types. Background Technology

[0002] For a long time, power cable fault detection has mainly relied on two types of technical methods: fault identification methods based on electrical parameter analysis and fault location methods based on traveling wave theory. Electrical parameter analysis methods determine the fault type by monitoring changes in the current and voltage waveforms of the cable, specifically including transient overvoltage analysis, zero-sequence current monitoring, and harmonic analysis. Traveling wave location methods, on the other hand, determine the fault location by capturing the propagation characteristics of the electromagnetic waves generated by the fault in the cable, mainly including pulse reflection methods, correlation methods, and traveling wave velocity methods. With the increasing complexity of power systems, cable faults are becoming more diverse and concealed, posing a serious challenge to traditional single-information-source detection methods. Domestic and international research institutions have proposed various improvement schemes to address this problem, such as transient signal analysis based on wavelet transform, fault mode recognition based on support vector machines, and multi-sensor collaborative localization. However, these methods still do not fundamentally overcome the limitations of a single information domain, lack effective integration of electrical parameters and electromagnetic field information, and cannot construct a complete dual-domain feature space.

[0003] Existing technologies for power cable fault detection have significant shortcomings. First, fault identification methods based on single electrical parameters have limited ability to identify faults in various fault scenarios, especially for special types such as high-resistance faults and multi-point faults, where the accuracy rate drops significantly. Their identification principle mainly relies on the matching degree between fault current characteristics and preset patterns; when fault types exhibit similar electrical characteristics, misjudgment is easily caused. Second, the accuracy of traditional traveling wave location methods is unstable due to various factors, including environmental interference, signal attenuation, and wave velocity changes. Especially at cable joints, the reflected signal aliasing caused by changes in wave impedance significantly reduces location accuracy. Third, existing technologies have limited ability to decompose the spatiotemporal features of power signals, making it difficult to effectively distinguish between transient and steady-state components, as well as zero-sequence and positive / negative-sequence components, thus affecting the accuracy of fault feature extraction. Furthermore, existing methods generally employ a sequential processing mode, where fault type is determined first, followed by selection of an appropriate localization algorithm. This approach not only prolongs fault diagnosis time but also makes the localization results overly reliant on the accuracy of the fault type determination. If the fault type is incorrectly identified, the localization algorithm selection will be flawed, ultimately affecting the overall detection performance. More critically, existing technologies lack an effective feature cascade network mechanism, failing to optimize and integrate fault type identification results with fault location candidate area information, thus hindering the full utilization of the complementarity between the two types of information and impacting the accuracy and reliability of fault detection. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the present invention provides an automatic identification and location detection method and system for power cable fault types, which can solve the problems mentioned in the background art.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: an automatic identification and location detection method and system for power cable fault types, comprising: collecting electrical parameter signals inside the power cable and electromagnetic field distribution data from the outside, constructing a dual-domain feature space of the cable's operating state; performing spatiotemporal feature decomposition on the dual-domain feature space to form an electrical feature sequence and an electromagnetic field feature map; inputting the electrical feature sequence into a fault type identification module and the electromagnetic field feature map into a fault location prediction module; the fault type identification module and the fault location prediction module constitute a dual discrimination system; integrating the fault type identification result generated by the fault type identification module and the fault location candidate region determined by the fault location prediction module through a feature concatenation network to generate a fusion decision result, determining the fault type and fault location; the feature concatenation network optimizes and integrates the results of the fault type identification module and the fault location prediction module through weighted information fusion.

[0007] As a preferred embodiment of the automatic identification and location detection method for power cable fault types described in this invention, the construction of the dual-domain feature space for cable operating status includes: analyzing the temporal characteristics of electrical parameter signals, identifying feature change points, and dividing the electrical parameter signals into multiple feature segments; performing feature extraction on each of the multiple feature segments, and connecting the features of each feature segment through a continuity constraint function to form a complete electrical parameter feature expression; synchronously acquiring electromagnetic field distribution data corresponding to the electrical parameter signals, and constructing an electromagnetic field parameter feature expression; and aligning the electrical parameter feature expression and the electromagnetic field parameter feature expression in a four-dimensional tensor structure in time and space to form a dual-domain feature space.

[0008] As a preferred embodiment of the automatic identification and location detection method for power cable fault types described in this invention, the method includes: performing spatiotemporal feature decomposition on the dual domain feature space, comprising: performing a short-time Fourier transform on the electrical parameter signal to obtain a frequency domain feature spectrum, and extracting the dominant frequency component to form a spectral feature set; determining a feature subspace based on the spectral feature set, and performing non-negative tensor decomposition on the feature subspace to obtain basic electrical domain features; determining the associated region in the electromagnetic field distribution data based on the basic electrical domain features, performing a short-time Fourier transform on the associated region to obtain electromagnetic domain spectral features; and combining the basic electrical domain features and the electromagnetic domain spectral features to form a complete spatiotemporal feature representation, thereby forming the electrical feature sequence and the electromagnetic field feature map.

[0009] As a preferred embodiment of the automatic identification and location detection method for power cable fault types according to the present invention, the implementation of the dual discrimination system includes: the fault type identification module initially identifies the fault type based on the electrical feature sequence and generates a fault feature template; the fault location prediction module receives the fault feature template and performs matching analysis with the electromagnetic field feature spectrum to determine the region with the highest matching degree as a fault location candidate region; the fault location prediction module feeds back the spatial distribution characteristics of the fault location candidate region to the fault type identification module; the fault type identification module confirms or corrects the fault type determination through feature consistency verification based on the feedback spatial distribution characteristics and the phase relationship and harmonic component distribution at the corresponding positions in the electrical feature sequence, thereby forming a fault type identification result.

[0010] As a preferred embodiment of the automatic identification and location detection method for power cable fault types according to the present invention, the feature cascaded network includes an input layer, an intermediate layer, a decision layer, and a residual connection channel connecting the intermediate layer and the decision layer; the residual connection channel transmits the fault type identification result and the fault location candidate area to the decision layer, so that the decision layer can generate the fusion decision result.

[0011] As a preferred embodiment of the automatic identification and location detection method for power cable fault types according to the present invention, the residual connection channel transmits the fault type identification result and the fault location candidate area to the decision layer, including: extracting transient component features and steady-state component features from the fault type identification result, and extracting zero-sequence component features and positive and negative-sequence component features from the fault location candidate area; calculating the symmetric component correlation between the transient component features and the zero-sequence component features, and comparing it with a first preset threshold; when the symmetric component correlation is lower than the first preset threshold, determining the module with a larger ratio of zero-sequence current to positive-sequence current between the fault type identification module and the fault location prediction module, and transmitting the corresponding result generated by this module to the decision layer, while simultaneously filtering feature data containing traveling wave characteristic frequencies from another module and transmitting it to the decision layer; when the symmetric component correlation is higher than or equal to the first preset threshold, transmitting the fault type identification result and the fault location candidate area to the decision layer.

[0012] As a preferred embodiment of the automatic identification and location detection method for power cable fault types according to the present invention, the decision layer, after receiving the fault type identification result and the fault location candidate area, performs the following processing to generate the fusion decision result: constructing a fault type feature space and a fault location feature space, respectively mapping the fault type identification result and the fault location candidate area; for single-phase grounding faults, constructing a three-component feature space containing zero-sequence components, positive-sequence components, and negative-sequence components, calculating the similarity between the current fault features and the standard single-phase grounding fault template in the three-component feature space, and determining the specific fault type; for phase-to-phase short-circuit faults, constructing a dual-gradient feature space containing electromagnetic field gradient and electric field gradient, calculating the similarity between the current fault features and the standard phase-to-phase short-circuit fault template in the dual-gradient feature space, and determining the specific fault location; establishing an association mapping function between the fault type feature space and the fault location feature space, wherein the association mapping function describes the spatial distribution characteristics corresponding to different fault types; using the association mapping function to calculate the fusion decision result, and outputting the fault type and fault location.

[0013] To further address the aforementioned technical problems, this invention provides the following technical solution: an automatic identification and location detection system for power cable fault types, comprising: a data acquisition module for acquiring electrical parameter signals inside the power cable and electromagnetic field distribution data outside the cable to construct a dual-domain feature space of the cable's operating state; a feature decomposition module for performing spatiotemporal feature decomposition on the dual-domain feature space to form an electrical feature sequence and an electromagnetic field feature map, inputting the electrical feature sequence into a fault type identification module and the electromagnetic field feature map into a fault location prediction module; and a feature concatenation module for integrating the fault type identification result generated by the fault type identification module with the fault location candidate area determined by the fault location prediction module through a feature concatenation network to generate a fusion decision result and determine the fault type and fault location.

[0014] A computer device includes a memory and a processor, the memory storing a computer program, characterized in that the processor executes the computer program to implement the steps of the automatic identification and location detection method for power cable fault types as described above.

[0015] 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 steps of the power cable fault type automatic identification and location detection method as described above.

[0016] The beneficial effects of this invention are as follows: This invention solves the technical problems of limited identification capability and unstable positioning accuracy of traditional single-domain information detection in multi-type fault scenarios. By parallel processing and fusion decision-making of electrical parameters and electromagnetic field information, it achieves accurate identification and high-precision positioning of fault types, and has the following technical effects: On the one hand, the dual-domain feature space provides more comprehensive fault characterization information, enabling this invention to distinguish similar fault types that are difficult to distinguish using traditional methods; on the other hand, the integration of the feature cascade network optimizes and fuses information from the electrical and electromagnetic domains, improving the anti-interference capability of fault positioning; in addition, in typical difficult-to-identify scenarios such as high-resistance faults, multi-point faults, and cable joint faults, this invention effectively shortens the fault diagnosis time and reduces power system outage losses. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the overall process of an automatic identification and location detection method for power cable fault types proposed in this invention;

[0019] Figure 2 This is a diagram of the computer equipment used in the automatic identification and location detection method for power cable fault types proposed in this invention. Detailed Implementation

[0020] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0021] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0022] Example 1, referring to Figure 1 This invention provides an embodiment of a method for automatic identification and location detection of power cable fault types.

[0023] In related technologies, fault identification methods based on single electrical parameters have limited ability to identify faults in multiple fault scenarios. The identification principle mainly relies on the matching degree between fault current characteristics and preset patterns, which can easily lead to misjudgment when fault types exhibit similar electrical characteristics. Secondly, the accuracy of traditional traveling wave location methods is unstable due to various factors, including environmental interference, signal attenuation, and wave velocity changes. Especially at cable joints, the reflected signal aliasing caused by changes in wave impedance significantly reduces location accuracy. Thirdly, existing technologies have limited ability to decompose the spatiotemporal features of power signals, making it difficult to effectively distinguish between transient and steady-state components, as well as zero-sequence and positive / negative-sequence components, affecting the accuracy of fault feature extraction. Furthermore, existing methods generally employ a serial processing mode, i.e., first determining the fault type, and then selecting an appropriate location algorithm based on the fault type. This approach not only prolongs fault diagnosis time but also makes the location results overly dependent on the accuracy of the fault type determination. If the fault type is incorrectly identified, it will lead to an incorrect location algorithm selection, ultimately affecting the overall detection effect. More importantly, existing technologies lack effective feature cascade network mechanisms, making it impossible to optimize and integrate fault type identification results with fault location candidate area information for decision-making. This makes it difficult to fully utilize the complementarity of the two types of information, thus affecting the accuracy and reliability of fault detection.

[0024] This application provides a method that can effectively solve the problems mentioned above. The following will describe in detail how to realize the automatic identification and location detection method of power cable fault types with multiple embodiments.

[0025] Figure 1 This diagram illustrates the overall process of an automatic identification and location detection method for power cable fault types, including the following steps:

[0026] S1: Collect electrical parameter signals inside the power cable and electromagnetic field distribution data outside to construct a dual-domain feature space of the cable's operating status.

[0027] Specifically, the dual-domain feature space contains two complementary information sets: the electrical domain and the electromagnetic domain.

[0028] Specifically, the construction of the dual-domain feature space for cable operating status includes:

[0029] Analyze the timing characteristics of electrical parameter signals, identify characteristic change points, and divide the electrical parameter signals into multiple characteristic segments;

[0030] Feature extraction is performed on multiple feature segments separately, and the features of each feature segment are connected by a continuity constraint function to form a complete electrical parameter feature expression;

[0031] Simultaneously acquire electromagnetic field distribution data corresponding to electrical parameter signals, and construct electromagnetic field parameter feature representations;

[0032] The electrical parameter feature representation and the electromagnetic field parameter feature representation are spatiotemporally aligned in a four-dimensional tensor structure to form a dual-domain feature space.

[0033] The technical solution of the present invention will be described in detail below with reference to specific embodiments.

[0034] In step S1, electrical parameter signals inside the power cable and electromagnetic field distribution data from the outside are collected to construct a dual-domain feature space of the cable's operating status. The dual-domain feature space contains two complementary information sets: an electrical domain and an electromagnetic domain. The electrical parameter signals mainly include voltage and current waveform data at both ends of the cable, acquired through high-precision sampling devices installed at both ends of the cable line. The electromagnetic field distribution data is collected through an array of electromagnetic field sensors arranged along the cable's route, including spatial distribution information of electric and magnetic field strengths.

[0035] The dual-domain feature space is represented by a four-dimensional tensor structure, including the time dimension, the cable spatial location dimension, the electrical parameter dimension corresponding to the electrical domain, and the electromagnetic field parameter dimension corresponding to the electromagnetic domain. The electrical parameter dimension corresponds to the electrical parameter signal, and the electromagnetic field parameter dimension corresponds to the electromagnetic field distribution data.

[0036] Constructing the two-domain feature space of cable operating status involves the following four steps:

[0037] The first step is to analyze the timing characteristics of the electrical parameter signals, identify key change points, and divide the electrical parameter signals into multiple characteristic segments. When a power cable experiences a fault, its electrical parameter signals (such as current and voltage) typically exhibit significant changes. Through a designed key change point detection algorithm, these change points can be automatically identified, thereby segmenting the continuous electrical parameter signals into multiple characteristic segments with different characteristics.

[0038] In this embodiment, the feature change point detection employs a sliding window method to calculate the rate of change of local statistical features of the electrical parameter signal. When the rate of change exceeds a preset threshold, that moment is marked as a feature change point. The specific calculation formula is as follows:

[0039] ;

[0040] in, This represents the statistical feature vector calculated within time window i, including statistical measures such as the signal's mean, variance, skewness, and kurtosis. This represents the normalized difference between adjacent windows; This is the system's minimum precision value. When... Greater than the preset threshold At that time, the boundary between window i and i+1 is marked as the characteristic change point, thereby dividing the electrical parameter signal into multiple characteristic segments.

[0041] The second step involves performing feature extraction on multiple feature segments separately, and then connecting the features of each feature segment using a continuity constraint function to form a complete electrical parameter feature representation. Different feature extraction strategies are adopted for different feature segments: for normal operation segments, steady-state features are mainly extracted; for transient change segments, transient features are mainly extracted; and for steady-state fault segments, both steady-state features and transient decay features after the fault are considered.

[0042] To ensure continuity and smooth transition between different feature segments, this invention employs a continuity constraint function, the expression of which is:

[0043] ;

[0044] in, and These represent the characteristic functions of adjacent feature segments; and These represent the end time of the previous segment and the start time of the next segment, respectively. and These are weighting coefficients used to balance the importance of positional continuity and gradient continuity, and are usually adjusted according to the specific application scenario. This represents the gradient operator. By minimizing the continuity constraint function, the optimal connection method of each feature segment can be obtained, forming a complete expression of electrical parameter features.

[0045] The third step involves simultaneously acquiring electromagnetic field distribution data corresponding to the electrical parameter signals to construct a characteristic representation of the electromagnetic field parameters. The electromagnetic field distribution generated during a power cable fault exhibits distinct spatial characteristics, which can provide crucial information for fault type identification and location. Electromagnetic field distribution data is acquired through a multi-point electromagnetic sensor array deployed along the cable, with each sensor measuring the local electric and magnetic field strengths.

[0046] For the measured electromagnetic field data, its spatial distribution characteristics, gradient characteristics, spectral characteristics, and time-varying properties are extracted to construct a characteristic representation of the electromagnetic field parameters. The electromagnetic field gradient characteristics are calculated as follows:

[0047] ;

[0048] ;

[0049] in, Electric field strength; This is the magnetic field strength vector; Three-dimensional spatial coordinates; For gradient operators; , and They represent respectively to , , Partial derivatives in three directions. The characteristics of the electromagnetic field gradient directly reflect the spatial rate of change of the electromagnetic field, which is of great significance for fault location.

[0050] The fourth step involves spatiotemporally aligning the electrical parameter features with the electromagnetic field parameter features within a four-dimensional tensor structure, forming a dual-domain feature space. Spatiotemporal alignment is a crucial step in constructing this dual-domain feature space, ensuring the consistency of the electrical and electromagnetic field parameter features across both time and space dimensions.

[0051] When performing spatiotemporal alignment, the mapping relationship between the physical location of the cable and the electromagnetic field sampling points is first established. For spatial alignment, a piecewise linear interpolation method is used, as shown in the following expression:

[0052] ;

[0053] in, Indicates the location on the cable The parameter value at that location, and This indicates the location of adjacent sampling points. Using this interpolation method, electrical and electromagnetic field parameters collected from different spatial locations can be mapped to a unified spatial coordinate system, achieving spatial dimension alignment.

[0054] For time-dimensional alignment, it can be achieved through synchronized sampling clocks or post-processing. In practical applications, timestamp information is used to align data from different sampling sources onto a unified timeline.

[0055] Through the four steps described above, this invention constructs a dual-domain feature space for cable operating states. This feature space integrates information from the electrical and electromagnetic domains, providing rich feature representations for subsequent fault type identification and location analysis. Compared with traditional single-source methods, the dual-domain feature space can more comprehensively describe cable fault characteristics, improving the accuracy and reliability of fault detection.

[0056] S2: Perform spatiotemporal feature decomposition on the dual-domain feature space to form an electrical feature sequence and an electromagnetic field feature map. Input the electrical feature sequence into the fault type identification module and the electromagnetic field feature map into the fault location prediction module.

[0057] The fault type identification module and the fault location prediction module constitute a dual discrimination system.

[0058] Specifically, the implementation of the dual-discrimination system includes:

[0059] The fault type identification module initially identifies the fault type based on electrical feature sequences and generates a fault feature template.

[0060] The fault location prediction module receives a fault feature template, performs matching analysis with the electromagnetic field feature map, and determines the region with the highest matching degree as the candidate region for the fault location.

[0061] The fault location prediction module feeds back the spatial distribution characteristics of the fault location candidate area to the fault type identification module;

[0062] The fault type identification module confirms or corrects the fault type determination by verifying the feature consistency based on the spatial distribution characteristics of the feedback, combined with the phase relationship and harmonic component distribution at the corresponding positions in the electrical feature sequence, thus forming the fault type identification result.

[0063] The spatiotemporal feature decomposition method combines tensor decomposition and Fourier transform.

[0064] Specifically, spatiotemporal feature decomposition is performed on the dual-domain feature space, including:

[0065] Perform a short-time Fourier transform on the electrical parameter signal to obtain the frequency domain feature spectrum, and extract the main frequency component to form a spectral feature set;

[0066] The feature subspace is determined based on the spectral feature set, and non-negative tensor decomposition is performed in the feature subspace to obtain the basic features of the electrical domain.

[0067] Based on the fundamental characteristics of the electrical domain, the associated regions in the electromagnetic field distribution data are determined, and a short-time Fourier transform is performed on the associated regions to obtain the electromagnetic domain spectral characteristics.

[0068] The basic features of the electrical domain are combined with the spectral features of the electromagnetic domain to form a complete spatiotemporal feature representation, resulting in an electrical feature sequence and an electromagnetic field feature map.

[0069] The technical solution of the present invention will be described in detail below with reference to specific embodiments.

[0070] Spatiotemporal feature decomposition employs a combination of tensor decomposition and Fourier transform. Tensor decomposition is used to process multidimensional data structures in a two-domain feature space, while Fourier transform is used to extract the frequency domain features of the signal. This combined approach effectively captures the temporal and spatial characteristics of power cable faults.

[0071] Specifically, performing spatiotemporal feature decomposition on the dual-domain feature space includes the following four steps:

[0072] The first step is to perform a short-time Fourier transform on the electrical parameter signal to obtain the frequency domain feature spectrum, and extract the dominant frequency component to form a spectral feature set. The short-time Fourier transform can reflect the frequency characteristics of the signal at different time periods, and is particularly suitable for analyzing the transient process during cable faults.

[0073] The mathematical expression for performing a short-time Fourier transform on the electrical parameter signal x(t) is:

[0074] ;

[0075] in, This represents the signal x(t) at time t and angular frequency. The short-time Fourier transform result at the point, It is a window function centered at t. This is the Fourier kernel function. It is calculated... The time spectrum of the signal is obtained, and the frequency components with the most concentrated energy are extracted as the dominant frequency components to form a spectral feature set. ,in This represents the k-th dominant frequency component.

[0076] The second step involves determining a feature subspace based on the spectral feature set, and then performing nonnegative tensor decomposition within that subspace to obtain the fundamental features of the electrical domain. First, the feature subspace is determined based on the spectral feature set. Then, nonnegative tensor decomposition is performed within this subspace to extract the fundamental features of the electrical domain.

[0077] Let the bi-domain feature space be represented as a four-dimensional tensor. ,in , , and These represent the dimensions of the time dimension, spatial location dimension, electrical parameter dimension, and electromagnetic field parameter dimension, respectively. Based on the spectral feature set, the tensor can be determined. Feature subspace in Perform nonnegative tensor decomposition on the eigenspace:

[0078] ;

[0079] in, Represents the cross product of vectors. Let R represent the basis vector of the r-th component in the n-th dimension, and R be the rank of the decomposition. Through this decomposition, the fundamental feature representation of the electrical domain is obtained. .

[0080] The third step involves identifying the relevant regions in the electromagnetic field distribution data based on the fundamental characteristics of the electrical domain. A short-time Fourier transform is then performed on these regions to obtain the electromagnetic field spectrum characteristics. By utilizing the correlation between the fundamental characteristics of the electrical domain and the electromagnetic field distribution data, the regions in the electromagnetic field data most relevant to the fault are determined, and frequency domain analysis is performed on these regions.

[0081] The correlation between fundamental characteristics of the electrical domain and electromagnetic field data can be quantified by calculating the correlation coefficient:

[0082] ;

[0083] in, Indicates the spatial position on the cable. Indicates position Electromagnetic field data at the location, where Cov represents the covariance. and These represent the standard deviations of the fundamental characteristics of the electrical domain and the electromagnetic field data, respectively. Correlation coefficient. Greater than the threshold The area was identified as the associated area. .

[0084] Perform a short-time Fourier transform on the electromagnetic field data of the associated region to obtain the electromagnetic domain spectral characteristics. .

[0085] The fourth step involves combining the fundamental features of the electrical domain with the spectral features of the electromagnetic domain to form a complete spatiotemporal feature representation, resulting in an electrical feature sequence and an electromagnetic field feature map. This is achieved through feature fusion methods, combining features from the electrical and electromagnetic domains to form a complete spatiotemporal feature representation.

[0086] Electrical characteristic sequence ES Formed by chronologically arranged fundamental characteristics of the electrical domain, it reflects the characteristic changes of cable faults over time:

[0087] ;

[0088] in, Indicates time The fundamental characteristics of the electrical domain.

[0089] Electromagnetic field characteristic spectrum This is formed by the spatial distribution of electromagnetic spectrum characteristics, reflecting the spatial distribution of fault characteristics:

[0090] ;

[0091] in, Indicates position The electromagnetic spectrum characteristics at that location.

[0092] The fault type identification module and the fault location prediction module constitute a dual discrimination system. This dual discrimination system improves the accuracy of fault identification and location through the cooperation and feedback between the two modules.

[0093] Specifically, the implementation of the dual-discrimination system includes the following four steps:

[0094] The first step involves the fault type identification module initially identifying fault types based on electrical feature sequences and generating fault feature templates. By analyzing the feature patterns in the electrical feature sequences, possible fault types are preliminarily determined, and a corresponding feature template is generated for each possible fault type.

[0095] A fault feature template is an abstract representation of the typical behavior of a specific type of fault in the electrical feature space, containing key characteristic indicators and their variation patterns. For n possible fault types, the generated set of fault feature templates is... ,in The feature template represents the i-th type of fault.

[0096] The second step involves the fault location prediction module receiving a fault feature template and performing a matching analysis with the electromagnetic field feature map to determine the region with the highest matching degree as a candidate fault location area. The fault location prediction module then projects the fault feature template onto the electromagnetic field feature space, calculates the degree of matching with the electromagnetic field feature map, and determines the most likely fault location region.

[0097] For fault feature templates Its relationship with location The formula for calculating the matching degree of the electromagnetic field characteristics is:

[0098] ;

[0099] in, Indicates position Electromagnetic field characteristics at that location Represents the dot product of vectors. Represents the vector norm. Matching degree. The highest area was identified as a candidate fault location area. ,in This is the matching threshold.

[0100] The third step involves the fault location prediction module feeding back the spatial distribution characteristics of the fault location candidate area to the fault type identification module. These characteristics include the candidate area's location range, electromagnetic field intensity distribution, gradient changes, and other information, which provides crucial data for further confirmation of the fault type.

[0101] Spatial distribution characteristics can be represented by vectors It means that among them This represents the j-th spatial feature parameter of the candidate fault location region. These feature parameters are fed back to the fault type identification module for subsequent fault type confirmation.

[0102] The fourth step involves the fault type identification module confirming or correcting the fault type determination based on the spatial distribution characteristics fed back, combined with the phase relationship and harmonic component distribution at the corresponding positions in the electrical feature sequence, through feature consistency verification, thus forming the fault type identification result. By comparing the spatial distribution characteristics with the phase relationship and harmonic distribution in the electrical feature sequence, the module verifies whether the initially identified fault type is consistent with all observation data.

[0103] Feature consistency verification is based on the following criteria: First, the phase relationship and harmonic component distribution of the time period corresponding to the candidate fault location area are extracted from the electrical feature sequence. Then, the results are compared with the expected features in the fault feature template to calculate the consistency score. If the consistency score is higher than a preset threshold, the initially identified fault type is confirmed; otherwise, the fault type determination is corrected based on the consistency analysis results.

[0104] Through the four steps of the dual-discrimination system described above, this invention achieves mutual verification and optimization of fault type identification and fault location, improving the overall accuracy of identification and location. Compared with traditional unidirectional fault identification methods, the dual-discrimination system can more effectively utilize multi-source information and reduce the false positive rate.

[0105] S3: The fault type identification result generated by the fault type identification module and the fault location candidate region determined by the fault location prediction module are integrated through a feature concatenation network to generate a fusion decision result and determine the fault type and fault location.

[0106] Among them, the feature cascaded network optimizes and integrates the results of the fault type identification module and the fault location prediction module through weighted information fusion.

[0107] Specifically, the feature cascaded network includes an input layer, an intermediate layer, a decision layer, and a residual connection channel connecting the intermediate layer and the decision layer. The residual connection channel transmits the fault type identification result and the fault location candidate region to the decision layer, so that the decision layer can generate a fusion decision result.

[0108] Specifically, the residual connection channel transmits the fault type identification results and fault location candidate regions to the decision layer, including:

[0109] Extract transient and steady-state component features from the fault type identification results, and extract zero-order and positive-negative-order component features from the fault location candidate region;

[0110] Calculate the symmetric component correlation between transient component features and zero-sequence component features, and compare it with a first preset threshold;

[0111] When the correlation of the symmetrical components is lower than the first preset threshold, the module with the larger ratio of zero-sequence current to positive-sequence current in the fault type identification module and the fault location prediction module is determined, and the corresponding result generated by the module is transmitted to the decision layer. At the same time, feature data containing traveling wave characteristic frequencies is selected from another module and transmitted to the decision layer.

[0112] When the correlation of the symmetric components is higher than or equal to the first preset threshold, the fault type identification result and the fault location candidate area are transmitted to the decision layer.

[0113] Specifically, after receiving the fault type identification result and the fault location candidate region, the decision-making layer performs the following processing to generate the fused decision result:

[0114] Construct a fault type feature space and a fault location feature space, and map the fault type identification result and the fault location candidate region, respectively;

[0115] For single-phase ground faults, a three-component feature space containing zero-sequence, positive-sequence, and negative-sequence components is constructed. The similarity between the current fault characteristics and the standard single-phase ground fault template is calculated within the three-component feature space to determine the specific type of fault.

[0116] For phase-to-phase short-circuit faults, a dual-gradient feature space containing electromagnetic field gradient and electric field gradient is constructed. The similarity between the current fault features and the standard phase-to-phase short-circuit fault template is calculated in the dual-gradient feature space to determine the specific location of the fault.

[0117] Establish a correlation mapping function between the fault type feature space and the fault location feature space. The correlation mapping function describes the spatial distribution characteristics corresponding to different fault types.

[0118] The fusion decision result is calculated using the correlation mapping function, and the fault type and fault location are output.

[0119] The technical solution of the present invention will be described in detail below with reference to specific embodiments.

[0120] Feature-cascaded networks optimize and integrate the results of the fault type identification module and the fault location prediction module through weighted information fusion. These networks not only integrate the results of the two modules but also dynamically adjust the weights of each module's results based on different fault conditions, thereby improving the accuracy of the final decision.

[0121] Specifically, the feature cascaded network includes an input layer, an intermediate layer, a decision layer, and a residual connection channel connecting the intermediate layer and the decision layer. The residual connection channel transmits the fault type identification result and the fault location candidate region to the decision layer, so that the decision layer can generate a fusion decision result.

[0122] The input layer receives the preliminary results from the fault type identification module and the fault location prediction module, and converts them into feature vectors that the network can process. The intermediate layer contains multiple processing units for extracting high-level features and performing preliminary fusion operations. The decision layer then generates the final fusion decision result based on all available information.

[0123] The residual connection channel is a key component of the feature cascade network. It ensures that the original fault type identification results and fault location candidate region information can be directly transmitted to the decision layer, avoiding information loss that may occur during intermediate layer processing. This design draws inspiration from the residual network structure in deep learning, effectively improving the network's performance and stability.

[0124] Specifically, the residual connection channel transmits the fault type identification results and fault location candidate regions to the decision layer, including the following steps:

[0125] The first step is to extract transient and steady-state features from the fault type identification results, and to extract zero-order and positive / negative-order features from the fault location candidate region.

[0126] Transient component characteristics mainly reflect the instantaneous changes in the early stage of a fault, including the initial fault waveform, rise time, and oscillation characteristics; steady-state component characteristics reflect the characteristics of the system when it reaches a relatively stable state after a fault occurs, including the fault steady-state current, voltage amplitude, and phase relationship.

[0127] Zero-sequence component characteristics refer to the zero-sequence components of three-phase current or voltage, which are mainly used to determine whether a ground fault exists; positive and negative sequence component characteristics reflect the symmetry of the three-phase system and are used to determine the type of phase-to-phase fault.

[0128] The second step is to calculate the correlation between the transient component features and the zero-sequence component features, and compare it with the first preset threshold.

[0129] Symmetric component correlation is an indicator that measures the correlation between transient component features and zero-sequence component features. It can be used to determine whether there is consistency between the fault type identification result and the fault location candidate region. When the symmetric component correlation is high, it indicates that the results of the two modules have good consistency; when the symmetric component correlation is low, it indicates that there may be inconsistencies in the results, which require further analysis and processing.

[0130] The third step is to identify the module with a larger ratio of zero-sequence current to positive-sequence current in the fault type identification module and the fault location prediction module when the correlation of the symmetrical components is lower than the first preset threshold. The corresponding result generated by this module is then transmitted to the decision layer. At the same time, feature data containing traveling wave characteristic frequencies is selected from another module and transmitted to the decision layer.

[0131] The ratio of zero-sequence current to positive-sequence current is an important indicator for judging the severity of grounding faults. When the correlation of symmetrical components is low, selecting the module result with a larger ratio of zero-sequence current to positive-sequence current as the primary reference can more accurately reflect the fault situation. Simultaneously, filtering feature data containing traveling wave characteristic frequencies from another module can provide supplementary information for fault location.

[0132] The characteristic frequency of a traveling wave refers to the characteristic frequency exhibited by the electromagnetic wave generated by a fault as it propagates in a cable. It is related to the distance from the fault point to the measurement point and is an important basis for fault location.

[0133] The fourth step is to transmit the fault type identification result and the fault location candidate area to the decision layer when the correlation of the symmetric components is higher than or equal to the first preset threshold.

[0134] When the correlation between the symmetric components is high, it indicates that the results of the fault type identification module and the fault location prediction module have good consistency. The results of the two modules can be transmitted to the decision-making layer at the same time, so that the decision-making layer can take into account them and generate the final fusion decision result.

[0135] Specifically, after receiving the fault type identification result and the fault location candidate region, the decision-making layer performs the following processing to generate the fused decision result:

[0136] The first step is to construct a fault type feature space and a fault location feature space, which are used to map the fault type identification results and the fault location candidate regions, respectively.

[0137] The fault type feature space is a multi-dimensional feature space, where each dimension represents a characteristic parameter of a fault type, such as zero-sequence current amplitude, phase relationship, and harmonic content. Mapping the fault type identification results to this feature space can more intuitively represent the relationships and differences between different fault types.

[0138] The fault location feature space is also a multi-dimensional feature space, with each dimension representing a characteristic parameter related to the fault location, such as electromagnetic field strength, gradient distribution, and traveling wave delay. Mapping the fault location candidate region to this feature space allows for a more comprehensive analysis of the feature distribution of the fault location.

[0139] The second step is to construct a three-component feature space containing zero-sequence, positive-sequence, and negative-sequence components for single-phase grounding faults. The similarity between the current fault characteristics and the standard single-phase grounding fault template is calculated within the three-component feature space to determine the specific type of fault.

[0140] The three-component characteristic space is a characteristic space specifically designed for single-phase grounding faults. It comprehensively considers the characteristics of three symmetrical components: zero-sequence component, positive-sequence component, and negative-sequence component. In this characteristic space, different types of single-phase grounding faults (such as metallic grounding, arcing grounding, intermittent grounding, etc.) exhibit different characteristic distributions.

[0141] By calculating the similarity between the current fault characteristics and the standard single-phase grounding fault template, the specific type of single-phase grounding fault can be determined. The standard fault template is a representation of typical fault characteristics established based on a large amount of historical fault data, providing a reference standard for fault type determination.

[0142] The third step is to construct a dual-gradient feature space containing electromagnetic field gradient and electric field gradient for phase-to-phase short-circuit faults. The similarity between the current fault features and the standard phase-to-phase short-circuit fault template is calculated in the dual-gradient feature space to determine the specific location of the fault.

[0143] The dual-gradient feature space is a feature space specifically designed for phase-to-phase short-circuit faults, focusing on two key features: electromagnetic field gradient and electric field gradient. Phase-to-phase short-circuit faults generate significant changes in electromagnetic and electric field gradients near the fault point, and these gradient features are crucial for accurate fault location determination.

[0144] By calculating the similarity between the current fault characteristics and the standard phase-to-phase short-circuit fault template, the specific location of the fault can be determined more accurately. For different types of phase-to-phase short-circuit faults (such as two-phase short circuits, three-phase short circuits, etc.), different standard fault templates can be used for matching and comparison.

[0145] The fourth step is to establish a correlation mapping function between the fault type feature space and the fault location feature space. The correlation mapping function describes the spatial distribution characteristics corresponding to different fault types.

[0146] The correlation mapping function serves as a bridge connecting the fault type feature space and the fault location feature space, describing the spatial distribution characteristics of different fault types. For example, single-phase grounding faults typically exhibit concentrated zero-sequence current and relatively small changes in the electromagnetic field gradient; while phase-to-phase short-circuit faults are characterized by significant positive and negative sequence currents and drastic changes in the electromagnetic field gradient.

[0147] By establishing this correlation mapping relationship, we can gain a more comprehensive understanding of the intrinsic connection between fault type and fault location, providing theoretical support for the final fusion decision.

[0148] The fifth step is to use the correlation mapping function to calculate the fusion decision result and output the fault type and fault location.

[0149] Based on the correlation mapping function, the decision layer comprehensively considers information from the fault type feature space and the fault location feature space to calculate the final fusion decision result. This result includes two aspects: first, the determined fault type, such as single-phase grounding, two-phase short circuit, etc.; second, the determined fault location, usually expressed as the distance from the cable starting point.

[0150] Through these five processing steps of the feature cascade network, this invention achieves optimized integration of fault type identification results and fault location candidate regions, generating accurate fusion decision results and providing a scientific basis for rapid handling of power cable faults. Compared with traditional fault diagnosis methods, the method of this invention significantly improves the accuracy and reliability of fault identification and location through multi-source information fusion and dual-module collaborative optimization.

[0151] In summary, this invention solves the technical problems of limited identification capability and unstable positioning accuracy of traditional single-domain information detection in multi-type fault scenarios. By parallel processing and fusion decision-making of electrical parameters and electromagnetic field information, it achieves accurate identification and high-precision positioning of fault types, with the following technical effects: First, the dual-domain feature space provides more comprehensive fault characterization information, enabling this invention to distinguish similar fault types that are difficult to distinguish using traditional methods; second, the integration of the feature cascade network optimizes and fuses information from the electrical and electromagnetic domains, improving the anti-interference capability of fault positioning; furthermore, in typical difficult-to-identify scenarios such as high-resistance faults, multi-point faults, and cable joint faults, this invention effectively shortens fault diagnosis time and reduces power system outage losses.

[0152] Example 2, an embodiment of the present invention, provides an automatic identification and location detection system for power cable fault types, comprising:

[0153] Example 3, referring to Figure 2This is one embodiment of the present invention, which differs from the previous embodiment in that: if the function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing program code.

[0154] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0155] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0156] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0157] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for automatic identification and location detection of power cable fault types, characterized in that, include: By collecting electrical parameter signals inside the power cable and electromagnetic field distribution data from the outside, a dual-domain feature space for the cable's operating status is constructed. Spatiotemporal feature decomposition is performed on the dual-domain feature space to form an electrical feature sequence and an electromagnetic field feature map. The electrical feature sequence is input into the fault type identification module and the electromagnetic field feature map is input into the fault location prediction module. The fault type identification module and the fault location prediction module constitute a dual discrimination system. The fault type identification result generated by the fault type identification module and the fault location candidate region determined by the fault location prediction module are integrated through a feature concatenation network to generate a fusion decision result, which determines the fault type and fault location. The feature concatenation network achieves optimized integration of the results of the fault type identification module and the fault location prediction module through weighted information fusion.

2. The automatic identification and location detection method for power cable fault types as described in claim 1, characterized in that: The dual-domain feature space for constructing the cable's operating state includes: Analyze the timing characteristics of electrical parameter signals, identify characteristic change points, and divide the electrical parameter signals into multiple characteristic segments; Feature extraction is performed on each of the multiple feature segments, and the features of each feature segment are connected by a continuity constraint function to form a complete electrical parameter feature expression; Simultaneously acquire electromagnetic field distribution data corresponding to the electrical parameter signals, and construct an electromagnetic field parameter feature representation; The electrical parameter feature expression and the electromagnetic field parameter feature expression are spatiotemporally aligned in a four-dimensional tensor structure to form a dual-domain feature space.

3. The automatic identification and location detection method for power cable fault types as described in claim 2, characterized in that: Perform spatiotemporal feature decomposition on the dual domain feature space, including: Perform a short-time Fourier transform on the electrical parameter signal to obtain the frequency domain feature spectrum, and extract the main frequency component to form a spectral feature set; Based on the spectral feature set, a feature subspace is determined, and non-negative tensor decomposition is performed in the feature subspace to obtain the basic features of the electrical domain. Based on the fundamental characteristics of the electrical domain, the associated regions in the electromagnetic field distribution data are determined, and a short-time Fourier transform is performed on the associated regions to obtain the electromagnetic domain spectral characteristics. The electrical domain fundamental features and the electromagnetic domain spectral features are combined to form a complete spatiotemporal feature representation, thus forming the electrical feature sequence and the electromagnetic field feature spectrum.

4. The automatic identification and location detection method for power cable fault types as described in claim 3, characterized in that: The implementation of the dual-discrimination system includes: The fault type identification module initially identifies the fault type based on the electrical feature sequence and generates a fault feature template. The fault location prediction module receives the fault feature template and performs matching analysis with the electromagnetic field feature map to determine the region with the highest matching degree as the fault location candidate region. The fault location prediction module feeds back the spatial distribution characteristics of the fault location candidate area to the fault type identification module. The fault type identification module confirms or corrects the fault type determination by verifying the feature consistency based on the spatial distribution characteristics of the feedback and the phase relationship and harmonic component distribution at the corresponding positions in the electrical feature sequence, thus forming a fault type identification result.

5. The automatic identification and location detection method for power cable fault types as described in claim 4, characterized in that: The feature cascaded network includes an input layer, an intermediate layer, a decision layer, and a residual connection channel connecting the intermediate layer and the decision layer; the residual connection channel transmits the fault type identification result and the fault location candidate region to the decision layer, so that the decision layer can generate the fusion decision result.

6. The automatic identification and location detection method for power cable fault types as described in claim 5, characterized in that: The residual connection channel transmits the fault type identification result and the fault location candidate region to the decision layer, including: Extract transient and steady-state component features from the fault type identification results, and extract zero-order and positive-negative-order component features from the fault location candidate region; Calculate the symmetric component correlation between the transient component feature and the zero-order component feature, and compare it with a first preset threshold; When the correlation of the symmetric components is lower than the first preset threshold, the module with the larger ratio of zero-sequence current to positive-sequence current in the fault type identification module and the fault location prediction module is determined, and the corresponding result generated by the module is transmitted to the decision layer. At the same time, feature data containing traveling wave characteristic frequencies is selected from another module and transmitted to the decision layer. When the correlation of the symmetric component is higher than or equal to the first preset threshold, the fault type identification result and the fault location candidate region are transmitted to the decision layer.

7. The automatic identification and location detection method for power cable fault types as described in claim 6, characterized in that: After receiving the fault type identification result and the fault location candidate region, the decision layer performs the following processing to generate the fused decision result: Construct a fault type feature space and a fault location feature space, and map the fault type identification result and the fault location candidate region, respectively; For a single-phase ground fault, a three-component feature space containing zero-sequence component, positive-sequence component, and negative-sequence component is constructed. The similarity between the current fault characteristics and the standard single-phase ground fault template is calculated within the three-component feature space to determine the specific type of fault. For phase-to-phase short-circuit faults, a dual-gradient feature space containing electromagnetic field gradient and electric field gradient is constructed. The similarity between the current fault features and the standard phase-to-phase short-circuit fault template is calculated in the dual-gradient feature space to determine the specific location of the fault. Establish a correlation mapping function between the fault type feature space and the fault location feature space, wherein the correlation mapping function describes the spatial distribution characteristics corresponding to different fault types; The fusion decision result is calculated using the correlation mapping function, and the fault type and fault location are output.

8. An automatic identification and location detection system for power cable fault types, based on the automatic identification and location detection method for power cable fault types according to any one of claims 1 to 7, characterized in that: include, The data acquisition module is used to collect electrical parameter signals inside the power cable and electromagnetic field distribution data from the outside, and to construct a dual-domain feature space of the cable's operating status. The feature decomposition module is used to perform spatiotemporal feature decomposition on the dual-domain feature space to form an electrical feature sequence and an electromagnetic field feature map. The electrical feature sequence is input into the fault type identification module and the electromagnetic field feature map is input into the fault location prediction module. The feature concatenation module is used to integrate the fault type identification result generated by the fault type identification module with the fault location candidate region determined by the fault location prediction module through a feature concatenation network to generate a fusion decision result and determine the fault type and fault location.

9. 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 steps of the automatic identification and location detection method for power cable fault types as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the automatic identification and location detection method for power cable fault types as described in any one of claims 1 to 7.

Images