Special cable fault positioning method based on AI identification
By collecting and analyzing the temperature and acoustic signal data of special cables and combining it with a deep learning network, the data fusion problem in special cable fault location is solved, accurate detection and location of various fault types are achieved, and the accuracy and efficiency of cable fault diagnosis are improved.
Patent Information
- Application Number
- CN202510815063.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-18
- Publication Date
- 2025-09-05
AI Technical Summary
Existing technologies are unable to effectively integrate multiple heterogeneous data, resulting in difficulties in accurately locating the types and locations of special cable faults, especially in complex environments where misjudgment and missed judgments are serious.
By collecting temperature distribution information, acoustic signal data and cable layout data, feature extraction and classification are performed, and support vector machines and deep learning networks are used to perform fault correlation analysis and location, generate a fault distribution map and calculate the coordinates of the fault point.
It achieves accurate detection and positioning of various fault types of special cables, improves the accuracy and efficiency of fault diagnosis, and provides comprehensive analytical support for power system maintenance.
Smart Images
Figure CN120595019A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of cable fault location, and in particular to a special cable fault location method based on AI recognition. Background Art
[0002] As a core component of power transmission and communications, specialty cables play a vital role in industrial production, energy supply, and national defense security. Their operational stability and reliability are directly related to the overall safety of the system. However, cables often fail during use due to environmental factors, aging, or external damage. If these failures cannot be accurately located and repaired in a timely manner, they can lead to significant economic losses or even safety accidents. Therefore, the development of efficient and accurate fault location methods is crucial.
[0003] Currently, traditional cable fault location methods rely on single detection methods, such as pulse reflection. While these methods can determine the approximate fault location to a certain extent, they often lack comprehensive assessment of fault type and severity, resulting in insufficient positioning accuracy and reliability. Misdiagnosis and missed detection are common, especially in complex environments or with multiple faults. Against this backdrop, the field of specialty cable fault location faces significant technical challenges. The primary challenge is integrating data from multiple sources, such as thermal imaging and acoustic detection sensors. These data vary greatly in form and characteristics, making unified analysis difficult. This difficulty in data integration further results in incomplete fault feature extraction, which fails to accurately reflect the true fault condition and, in turn, hinders the precise determination of fault type and location. This multi-layered challenge, from data fusion to feature extraction to precise location, presents a significant obstacle to improving fault location effectiveness. Therefore, effectively fusing multi-source heterogeneous data and extracting key fault features to accurately determine the type, severity, and location of specialty cable faults, has become a critical issue that needs to be addressed. Summary of the Invention
[0004] To solve the above technical problems, the present invention proposes a special cable fault location method based on AI recognition, comprising:
[0005] Collect sensor data, perform feature extraction, and obtain a comprehensive feature vector set; wherein the sensor data includes temperature distribution information, acoustic signal data, and cable layout data;
[0006] Performing feature classification on the comprehensive feature vector set to obtain a preliminary fault classification result;
[0007] Conduct fault correlation analysis on the preliminary fault classification results to obtain the fault distribution map;
[0008] Based on the fault distribution map, the coordinates of the fault point are calculated to obtain a positioning result.
[0009] Preferably, feature extraction is performed to obtain a comprehensive feature vector set, including:
[0010] Based on the sensor data, extracting peak features and frequency anomaly features through a feature extraction algorithm;
[0011] The extracted features are vector-combined to obtain a comprehensive feature vector set.
[0012] Preferably, performing feature classification on the comprehensive feature vector set to obtain a preliminary fault classification result includes:
[0013] A support vector machine classification algorithm is used to perform feature classification on the comprehensive feature vector set to obtain temperature anomaly peak features and acoustic wave signal frequency anomaly features;
[0014] Pattern matching is performed on the abnormal temperature peak characteristics and the abnormal acoustic wave signal frequency characteristics to obtain a preliminary fault classification result.
[0015] Preferably, based on the preliminary fault classification results, a fault correlation analysis is performed, including:
[0016] Inputting the preliminary fault classification results into a preset fault correlation analysis model to obtain potential thermal failure points and potential mechanical damage points;
[0017] A temporal and spatial correlation calculation is performed on the potential thermal failure points and the potential mechanical damage points to obtain a failure distribution map.
[0018] Preferably, calculating the coordinates of the fault point based on the fault distribution map and obtaining the positioning result includes:
[0019] Input the fault distribution map into the preset deep learning network for enhancement processing to obtain the fault distribution matrix;
[0020] Based on the fault distribution matrix, the coordinates of the fault point are calculated to obtain a positioning result.
[0021] Preferably, the fault distribution map is input into a preset deep learning network for enhancement processing, including:
[0022] Extracting multiple fault scenario information based on the fault distribution map; wherein the multiple fault scenario information includes a distribution pattern of fault points, a density of fault points, and a relationship between fault points;
[0023] Inputting the multiple fault scenario information into a preset deep learning network model to perform feature enhancement and obtain a dense area of fault points;
[0024] A weighted analysis is performed based on the fault point dense area to obtain an enhanced fault distribution matrix.
[0025] Preferably, the multiple fault scenario information is input into a preset deep learning network model for feature enhancement, including:
[0026] Extracting shallow features and deep features from the multiple fault scenario information;
[0027] The shallow features and deep features are fused through a feature pyramid network to obtain a dense area of fault points;
[0028] The weights of the fault point dense areas are learned by the spatial attention method to enhance the features of the fault point dense areas.
[0029] Preferably, the calculation of the coordinates of the fault point is performed based on the fault distribution matrix, including:
[0030] Based on the fault distribution matrix, a position regression algorithm is used to perform position coordinate mapping to obtain an initial fault coordinate set;
[0031] Based on the initial fault coordinate set and the cable layout data, the specific position coordinates of the fault point are calculated to obtain a positioning result.
[0032] Preferably, based on the fault distribution matrix, a position regression algorithm is used to perform position coordinate mapping, including:
[0033] Performing feature extraction on the fault distribution matrix to obtain fault points;
[0034] Input the fault point in the fault distribution matrix into the preset position regression model to obtain the initial physical coordinates of the fault point;
[0035] The initial physical coordinates of the fault point are fused to obtain an initial fault coordinate set.
[0036] Compared with the prior art, the present invention has the following advantages and technical effects:
[0037] The present invention discloses a special cable fault location method based on AI recognition, comprising: collecting sensor data, performing feature extraction, and obtaining a comprehensive feature vector set; wherein the sensor data includes temperature distribution information, acoustic signal data, and cable layout data; performing feature classification on the comprehensive feature vector set to obtain a preliminary fault classification result; performing fault correlation analysis on the preliminary fault classification result to obtain a fault distribution map; and calculating the coordinates of the fault point based on the fault distribution map to obtain a positioning result; the present invention realizes comprehensive detection and precise positioning of various fault types of special cables, and provides comprehensive fault analysis and visual decision support for power system maintenance through algorithm, namely AI model analysis, effectively improving the accuracy and efficiency of special cable fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:
[0039] Figure 1 The figure is a flowchart of a method for locating a special cable fault based on AI recognition according to an embodiment of the present invention. DETAILED DESCRIPTION
[0040] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0041] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.
[0042] like Figure 1 As shown, the present invention proposes a special cable fault location method based on AI identification, including:
[0043] Collect sensor data, perform feature extraction, and obtain a comprehensive feature vector set; the sensor data includes temperature distribution information, acoustic signal data, and cable layout data;
[0044] Perform feature classification on the comprehensive feature vector set to obtain preliminary fault classification results;
[0045] Conduct fault correlation analysis on the preliminary fault classification results to obtain the fault distribution map;
[0046] Based on the fault distribution map, the coordinates of the fault point are calculated to obtain the positioning result.
[0047] Specifically, comparing ordinary cables with specialty cables, ordinary cables primarily encounter common faults such as low resistance, short circuits, and open circuits. These fault types are relatively simple, and the methods for locating them are relatively mature. Specialty cables, however, due to their complex operating environments, may face complex fault types such as high resistance faults, intermittent faults, and partial discharge. For example, in high-voltage environments, high-resistance faults in specialty cables require specialized high-voltage pulse technology to locate.
[0048] Furthermore, feature extraction is performed to obtain a comprehensive feature vector set, including:
[0049] Based on the sensor data, the peak features and frequency anomaly features are extracted through feature extraction algorithms;
[0050] The extracted features are vector-combined to obtain a comprehensive feature vector set.
[0051] Specifically, in this embodiment, the temperature distribution information and acoustic signal data are screened for the distribution of abnormal peaks and frequency anomalies by using a preset threshold comparison method. If the proportion of abnormal points exceeds the preset threshold, the abnormal points are marked to obtain the marked feature set. For the marked feature set, the marked abnormal points are removed by using data cleaning technology to obtain a cleaned feature data set. Based on the cleaned feature data set, the spatial interpolation technology is used to smooth and optimize the data for the temperature distribution information to obtain an optimized temperature distribution feature set. For the optimized temperature distribution feature set, the changing trend of the data is monitored by a time series analysis tool. If the changing trend exceeds the preset range, the abnormal trend is recorded to obtain a trend monitoring result. Based on the trend monitoring result, the frequency distribution characteristics of the signal are extracted by using a spectrum analysis method for the frequency abnormality characteristics of the acoustic signal, the fluctuation range of the signal is judged, and the final frequency feature analysis result is obtained. For the final frequency feature analysis result, the temperature distribution characteristics and the acoustic signal characteristics are integrated by using data fusion technology to determine the final distribution state of the comprehensive characteristics and obtain the integrated data result;
[0052] Furthermore, the comprehensive feature vector set is subjected to feature classification to obtain preliminary fault classification results, including:
[0053] The support vector machine classification algorithm is used to classify the comprehensive feature vector set to obtain the temperature anomaly peak characteristics and the acoustic signal frequency anomaly characteristics;
[0054] Pattern matching is performed on the abnormal temperature peak characteristics and the abnormal frequency characteristics of the acoustic wave signal to obtain preliminary fault classification results.
[0055] Specifically, in this embodiment, a pre-established support vector machine model is used to perform preliminary classification of the abnormal peaks and frequency anomalies of the temperature distribution information and acoustic signal data of the special cable to obtain the abnormal temperature peak characteristics and the abnormal frequency characteristics of the acoustic signal;
[0056] Specifically, in this embodiment, based on the extracted characteristic classification information, the support vector machine (SVM) classification model used is analyzed through a classification algorithm, i.e., an AI model. It is a powerful classification algorithm that classifies data by finding the optimal segmentation hyperplane in a high-dimensional space. When processing the temperature anomaly peak characteristics and the acoustic signal frequency anomaly characteristics, a suitable kernel function can be selected. For nonlinearly separable data, commonly used kernel functions include radial basis functions (RBF) and polynomial kernel functions. The RBF kernel function is in the form of K(x,y)=exp(-γ∥xy∥2), where γ is the kernel parameter, which can map the data to a high-dimensional space and make the data linearly separable in the high-dimensional space. The polynomial kernel function is in the form of K(x,y)=(γx·y+r)d, where γ, r, and d are kernel parameters, which are suitable for situations where the data has an obvious polynomial relationship; the temperature anomaly peak characteristics and the acoustic signal frequency anomaly characteristics are pattern matched by the AI model, i.e., the support vector machine (SVM) classification model.
[0057] Furthermore, based on the preliminary fault classification results, fault correlation analysis is performed, including:
[0058] Input the preliminary fault classification results into the preset fault correlation analysis model to obtain potential thermal failure points and potential mechanical damage points;
[0059] The temporal and spatial correlation of potential thermal failure points and potential mechanical damage points is calculated to obtain a failure distribution map.
[0060] Specifically, in this embodiment, the preset fault association analysis model uses a decision tree model; for example, based on historical data and domain knowledge, fault association rules are established. If the same device experiences both abnormal temperature and abnormal acoustic signals within a short period of time, it may indicate a complex fault in the device. If multiple devices experience similar faults within a similar period of time, it may indicate a systemic problem.
[0061] Fault correlation analysis can be implemented using rule-based systems or machine learning models (such as decision trees and random forests). Common AI models include: Machine learning models: such as decision trees, support vector machines (SVMs), and random forests. Deep learning models: such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), and their variants (LSTMs and GRUs).
[0062] Specifically, in this embodiment, spatiotemporal correlation calculation includes: temporal correlation calculation and spatial correlation calculation; temporal correlation calculation: calculating the time interval between fault points to determine whether faults occur continuously within a short period of time. For example, setting the time threshold to 10 minutes, if the time interval between two fault points is less than 10 minutes, they are considered to be temporally correlated; spatial correlation calculation: calculating the spatial distance between fault points to determine whether faults occur at similar locations. For example, using Euclidean distance to calculate the distance between devices, setting the distance threshold to 10 meters, if the distance between two fault points is less than 10 meters, they are considered to be spatially correlated. Fault distribution map generation: using a geographic information system (GIS) or a simple charting tool such as Matplotlib to generate a fault distribution map.
[0063] Furthermore, based on the fault distribution map, the coordinates of the fault point are calculated to obtain the positioning result, including:
[0064] Input the fault distribution map into the preset deep learning network for enhancement processing to obtain the fault distribution matrix;
[0065] Based on the fault distribution matrix, the coordinates of the fault point are calculated to obtain the positioning results.
[0066] Furthermore, the fault distribution map is input into a preset deep learning network for enhanced processing, including:
[0067] Extract multiple fault scenario information based on the fault distribution map; the multiple fault scenario information includes the distribution pattern of fault points, the density of fault points, and the relationship between fault points;
[0068] Input multiple fault scenario information into the preset deep learning network model to perform feature enhancement and obtain the dense fault point area;
[0069] A weighted analysis is performed based on the fault point density area to obtain the enhanced fault distribution matrix.
[0070] Furthermore, the multiple fault scenario information is input into a preset deep learning network model for feature enhancement, including: extracting shallow features and deep features from the multiple fault scenario information;
[0071] The shallow features and deep features are fused through the feature pyramid network to obtain the dense area of fault points;
[0072] The weights of the fault point dense areas are learned through the spatial attention method to enhance the features of the fault point dense areas.
[0073] Furthermore, based on the fault distribution matrix, the coordinates of the fault point are calculated, including:
[0074] Based on the fault distribution matrix, a position regression algorithm is used to map the position coordinates to obtain an initial fault coordinate set. Based on the initial fault coordinate set and special cable layout data, the specific position coordinates of the fault point are calculated to obtain the positioning result.
[0075] Specifically, in this embodiment, the position regression algorithm adopts a position regression model, which is a machine learning model used to map the position information of the fault point to the actual physical coordinates.
[0076] Specifically, in this embodiment, for example, there is a 3x3 fault distribution matrix, as shown in Table 1 below:
[0077] Table 1
[0078] Position 1 Position 2 Position 3 Type 1 0.1 0.2 0.3 Type 2 0.4 0.5 0.6 Type 3 0.7 0.8 0.9
[0079] This example has trained a position regression model that can map fault type and location to actual physical coordinates. Using this model, the initial physical coordinates of each fault point can be obtained as follows:
[0080] The initial physical coordinates of the fault point (type 1, position 1) are (x1, y1); the initial physical coordinates of the fault point (type 1, position 2) are (x2, y2); the initial physical coordinates of the fault point (type 1, position 3) are (x3, y3); the initial physical coordinates of the fault point (type 2, position 1) are (x4, y4); the initial physical coordinates of the fault point (type 2, position 2) are (x5, y5); the initial physical coordinates of the fault point (type 2, position 3) are (x6, y6); the initial physical coordinates of the fault point (type 1, position 2) are (x7, y8); the initial physical coordinates of the fault point (type 2, position 3) are (x8, y9); The initial physical coordinates of the fault point (type 3, position 1) are (x7, y7); the initial physical coordinates of the fault point (type 3, position 2) are (x8, y8); the initial physical coordinates of the fault point (type 3, position 3) are (x9, y9); these coordinates are collected to obtain the initial fault coordinate set: {(x1, y1), (x2, y2), (x3, y3), (x4, y4), (x5, y5), (x6, y6), (x7, y7), (x8, y8), (x9, y9)}.
[0081] Furthermore, based on the fault distribution matrix, a position regression algorithm is used to perform position coordinate mapping, including:
[0082] Perform feature extraction on the fault distribution matrix to obtain the fault point;
[0083] Input the fault point in the fault distribution matrix into the preset position regression model to obtain the initial physical coordinates of the fault point;
[0084] The initial physical coordinates of the fault point are fused to obtain an initial fault coordinate set.
[0085] Specifically, there are significant differences in fault location technology between special cables and ordinary cables. Special cables require more complex and advanced technologies to cope with complex fault types and working environments.
[0086] Specialty cables are used in aerospace, energy development, chemical plants, and other specialized applications. These environments place high demands on cable performance, including high-temperature resistance, corrosion resistance, and anti-interference capabilities. Therefore, fault location technology for specialty cables must adapt to these complex environments, such as using high-temperature-resistant cable fault location equipment in high-temperature environments.
[0087] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.
Claims
1. A special cable fault location method based on AI recognition, characterized in that: include: Collect sensor data, perform feature extraction, and obtain a comprehensive feature vector set; wherein the sensor data includes temperature distribution information, acoustic signal data, and cable layout data; Performing feature classification on the comprehensive feature vector set to obtain a preliminary fault classification result; Conduct fault correlation analysis on the preliminary fault classification results to obtain the fault distribution map; Based on the fault distribution map, the coordinates of the fault point are calculated to obtain a positioning result.
2. The special cable fault location method based on AI recognition according to claim 1 is characterized in that: Perform feature extraction to obtain a comprehensive feature vector set, including: Based on the sensor data, extracting peak features and frequency anomaly features through a feature extraction algorithm; The extracted features are vector-combined to obtain a comprehensive feature vector set.
3. The special cable fault location method based on AI identification according to claim 1 is characterized in that: Performing feature classification on the comprehensive feature vector set to obtain preliminary fault classification results, including: A support vector machine classification algorithm is used to perform feature classification on the comprehensive feature vector set to obtain temperature anomaly peak features and acoustic wave signal frequency anomaly features; Pattern matching is performed on the abnormal temperature peak characteristics and the abnormal acoustic wave signal frequency characteristics to obtain a preliminary fault classification result.
4. The special cable fault location method based on AI identification according to claim 1 is characterized in that: Based on the preliminary fault classification results, perform fault correlation analysis, including: Inputting the preliminary fault classification results into a preset fault correlation analysis model to obtain potential thermal failure points and potential mechanical damage points; A temporal and spatial correlation calculation is performed on the potential thermal failure points and the potential mechanical damage points to obtain a failure distribution map.
5. The special cable fault location method based on AI identification according to claim 1 is characterized in that: Calculating the coordinates of the fault point based on the fault distribution map to obtain a positioning result includes: Input the fault distribution map into the preset deep learning network for enhancement processing to obtain the fault distribution matrix; Based on the fault distribution matrix, the coordinates of the fault point are calculated to obtain a positioning result.
6. The special cable fault location method based on AI identification according to claim 5 is characterized in that: The fault distribution map is fed into a preset deep learning network for enhanced processing, including: Extracting multiple fault scenario information based on the fault distribution map; wherein the multiple fault scenario information includes a distribution pattern of fault points, a density of fault points, and a relationship between fault points; Inputting the multiple fault scenario information into a preset deep learning network model to perform feature enhancement and obtain a dense area of fault points; A weighted analysis is performed based on the fault point dense area to obtain an enhanced fault distribution matrix.
7. The special cable fault location method based on AI identification according to claim 6 is characterized in that: Input the multiple fault scenario information into a preset deep learning network model for feature enhancement, including: Extracting shallow features and deep features from the multiple fault scenario information; The shallow features and deep features are fused through a feature pyramid network to obtain a dense area of fault points; The weights of the fault point dense areas are learned by the spatial attention method to enhance the features of the fault point dense areas.
8. The special cable fault location method based on AI identification according to claim 6 is characterized in that: Calculating the coordinates of the fault point based on the fault distribution matrix includes: Based on the fault distribution matrix, a position regression algorithm is used to perform position coordinate mapping to obtain an initial fault coordinate set; Based on the initial fault coordinate set and the cable layout data, the specific position coordinates of the fault point are calculated to obtain a positioning result.
9. The special cable fault location method based on AI identification according to claim 8, characterized in that: Based on the fault distribution matrix, a position regression algorithm is used to perform position coordinate mapping, including: Performing feature extraction on the fault distribution matrix to obtain fault points; Input the fault point in the fault distribution matrix into the preset position regression model to obtain the initial physical coordinates of the fault point; The initial physical coordinates of the fault point are fused to obtain an initial fault coordinate set.