Cable fault detection method, device, computer equipment, readable storage medium and program product
By acquiring and processing historical fault data of cables, and using an autoencoder network for dimensionality reduction and feature fusion, a multi-view detection model is constructed, which solves the problem of low accuracy in early cable fault detection in existing technologies and achieves high-precision identification of early cable faults.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ELECTRIC POWER RES INST CHINA SOUTHERN POWER GRID CO LTD
- Filing Date
- 2025-12-11
- Publication Date
- 2026-07-21
AI Technical Summary
Existing technologies struggle to effectively identify the weak signal characteristics of early cable faults, leading to a decrease in fault detection accuracy and hindering the shift from reactive emergency repairs to proactive preventative maintenance.
By acquiring historical fault data of cables, preprocessing it, determining time series features, artificial time domain features, and two-dimensional image features, using an autoencoder network for dimensionality reduction, fusing multi-view features, and constructing a DELM-KELM-Adaboost fault detection model to achieve complementarity of multi-view information.
It improves the accuracy of cable fault detection, can fully express the complex non-stationary and transient characteristics of early cable faults, enhances the ability to express fault information features, and improves the accuracy of detection.
Smart Images
Figure CN121686129B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of fault detection technology, and in particular to a cable fault detection method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Technology
[0002] With the widespread application of power cables in urban power distribution networks, their operational reliability has become crucial for ensuring public safety and livelihoods. However, cables are prone to developing highly concealed early-stage faults due to factors such as insulation aging. These faults exhibit extremely weak and complex signal characteristics before evolving into permanent power outages, making them difficult for existing detection technologies to effectively identify. Therefore, developing fault detection technologies capable of accurately and promptly identifying early-stage cable faults, and shifting from reactive emergency repairs to proactive preventative maintenance, is of urgent practical need and significant application value for improving power grid safety and power supply reliability.
[0003] In related technologies, existing technologies mostly focus on characterizing fault information from a single view or dimension. Whether based on artificial time-domain features, original time-series waveforms, or two-dimensional image transformations of signals, single-view features are only a partial representation of fault information and are insufficient to comprehensively express the fault information of early cable faults. They are also difficult to fully characterize the complex non-stationary and transient characteristics of early faults. Therefore, when faced with weak early fault signals and insignificant fault features, the accuracy of fault detection may decrease. Summary of the Invention
[0004] Therefore, it is necessary to provide a cable fault detection method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can improve the accuracy of cable fault detection in response to the above-mentioned technical problems.
[0005] In a first aspect, this application provides a cable fault detection method, including:
[0006] Acquire historical fault data of the cable, including fault-free data, cable half-cycle early fault data, cable multi-cycle early fault data, and corresponding tags;
[0007] The historical fault data is preprocessed to obtain processed historical fault data;
[0008] Based on the processed historical fault data, the time series characteristics, artificial time domain characteristics, and two-dimensional image characteristics of the cable are determined;
[0009] Based on an autoencoder network, the time series features, the artificial time-domain features, and the two-dimensional image features are dimensionality-reduced to obtain three dimensionality-reduced depth features.
[0010] The reduced-dimensional depth features are fused to obtain multi-view fusion features;
[0011] Based on the multi-view fusion features and the labels, a training set for fault detection is constructed;
[0012] The training set is input into a preset detection model for training. When the detection accuracy of the preset detection model reaches a preset level, the trained preset detection model is obtained.
[0013] Obtain target detection data of the cable, input the target detection data into the trained preset detection model, and obtain the fault detection result of the cable.
[0014] In one embodiment, determining the artificial time-domain characteristics of the cable based on the processed historical fault data includes:
[0015] Based on the fault-free data, the cable half-cycle early fault data, and the cable multi-cycle early fault data, calculate the time-domain characteristics corresponding to each fault data.
[0016] The feature vector composed of the time-domain features corresponding to the three types of fault data is used as the artificial time-domain feature of the cable.
[0017] In one embodiment, the time-domain features include peak-to-peak value, absolute mean, variance, standard deviation, root mean square, skewness, kurtosis of the current, and waveform factor of the current.
[0018] In one embodiment, determining the two-dimensional image features of the cable based on the processed historical fault data includes:
[0019] The current sequence in the historical fault data is normalized to obtain the normalized current sequence.
[0020] The normalized current sequence is mapped to a polar coordinate system, and a two-dimensional image matrix is generated;
[0021] The current sequence in the historical fault data is combined with the two-dimensional image matrix to obtain the two-dimensional image features of the cable.
[0022] In one embodiment, fusing the dimensionality-reduced depth features to obtain multi-view fusion features includes:
[0023] Calculate the contribution scores of the three dimensionality-reduced deep features respectively;
[0024] Based on the contribution score, the weight of each dimensionality reduction depth feature is determined;
[0025] Based on the weights, the reduced-dimensional depth features are weighted and summed to obtain the multi-view fusion features.
[0026] In one embodiment, the preprocessing of the historical fault data to obtain processed historical fault data includes:
[0027] The historical fault data is subjected to outlier removal, duplicate data deletion and noise filtering to obtain cleaned historical fault data;
[0028] The historical fault data after cleaning is resampled to obtain processed historical fault data; the number of sampling points for fault-free data, cable half-cycle early fault data and cable multi-cycle early fault data are the same in the processed historical fault data.
[0029] Secondly, this application also provides a cable fault detection device, comprising:
[0030] The acquisition module is used to acquire historical fault data of the cable, which includes fault-free data, cable half-cycle early fault data, cable multi-cycle early fault data, and corresponding tags.
[0031] The processing module is used to preprocess the historical fault data to obtain processed historical fault data;
[0032] The determination module is used to determine the time series characteristics, artificial time domain characteristics, and two-dimensional image characteristics of the cable based on the processed historical fault data;
[0033] The dimensionality reduction module is used to reduce the dimensionality of the time series features, the artificial time-domain features, and the two-dimensional image features based on the autoencoder network, to obtain three dimensionality-reduced depth features.
[0034] The fusion module is used to fuse the dimensionality-reduced depth features to obtain multi-view fusion features;
[0035] A construction module is used to construct a training set for fault detection based on the multi-view fusion features and the labels;
[0036] The training module is used to input the training set into the preset detection model for training, and to obtain the trained preset detection model when the detection accuracy of the preset detection model reaches a preset level.
[0037] The acquisition module is also used to acquire target detection data of the cable, input the target detection data into the trained preset detection model, and obtain the fault detection result of the cable.
[0038] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:
[0039] Acquire historical fault data of the cable, including fault-free data, cable half-cycle early fault data, cable multi-cycle early fault data, and corresponding tags;
[0040] The historical fault data is preprocessed to obtain processed historical fault data;
[0041] Based on the processed historical fault data, the time series characteristics, artificial time domain characteristics, and two-dimensional image characteristics of the cable are determined;
[0042] Based on an autoencoder network, the time series features, the artificial time-domain features, and the two-dimensional image features are dimensionality-reduced to obtain three dimensionality-reduced depth features.
[0043] The reduced-dimensional depth features are fused to obtain multi-view fusion features;
[0044] Based on the multi-view fusion features and the labels, a training set for fault detection is constructed;
[0045] The training set is input into a preset detection model for training. When the detection accuracy of the preset detection model reaches a preset level, the trained preset detection model is obtained.
[0046] Obtain target detection data of the cable, input the target detection data into the trained preset detection model, and obtain the fault detection result of the cable.
[0047] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:
[0048] Acquire historical fault data of the cable, including fault-free data, cable half-cycle early fault data, cable multi-cycle early fault data, and corresponding tags;
[0049] The historical fault data is preprocessed to obtain processed historical fault data;
[0050] Based on the processed historical fault data, the time series characteristics, artificial time domain characteristics, and two-dimensional image characteristics of the cable are determined;
[0051] Based on an autoencoder network, the time series features, the artificial time-domain features, and the two-dimensional image features are dimensionality-reduced to obtain three dimensionality-reduced depth features.
[0052] The reduced-dimensional depth features are fused to obtain multi-view fusion features;
[0053] Based on the multi-view fusion features and the labels, a training set for fault detection is constructed;
[0054] The training set is input into a preset detection model for training. When the detection accuracy of the preset detection model reaches a preset level, the trained preset detection model is obtained.
[0055] Obtain target detection data of the cable, input the target detection data into the trained preset detection model, and obtain the fault detection result of the cable.
[0056] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:
[0057] Acquire historical fault data of the cable, including fault-free data, cable half-cycle early fault data, cable multi-cycle early fault data, and corresponding tags;
[0058] The historical fault data is preprocessed to obtain processed historical fault data;
[0059] Based on the processed historical fault data, the time series characteristics, artificial time domain characteristics, and two-dimensional image characteristics of the cable are determined;
[0060] Based on an autoencoder network, the time series features, the artificial time-domain features, and the two-dimensional image features are dimensionality-reduced to obtain three dimensionality-reduced depth features.
[0061] The reduced-dimensional depth features are fused to obtain multi-view fusion features;
[0062] Based on the multi-view fusion features and the labels, a training set for fault detection is constructed;
[0063] The training set is input into a preset detection model for training. When the detection accuracy of the preset detection model reaches a preset level, the trained preset detection model is obtained.
[0064] Obtain target detection data of the cable, input the target detection data into the trained preset detection model, and obtain the fault detection result of the cable.
[0065] The aforementioned cable fault detection method, apparatus, computer equipment, computer-readable storage medium, and computer program product first acquire historical fault data of the cable, preprocess the historical fault data to obtain processed historical fault data; based on the processed historical fault data, determine the cable's time-series features, artificial time-domain features, and two-dimensional image features; based on an autoencoder network, reduce the dimensionality of the time-series features, artificial time-domain features, and two-dimensional image features to obtain three dimensionality-reduced depth features; fuse the dimensionality-reduced depth features to obtain multi-view fusion features; construct a fault detection training set based on the multi-view fusion features and labels; input the training set into a preset detection model for training, and obtain a trained preset detection model when the detection accuracy of the preset detection model reaches a preset level; acquire the cable's target detection data, input the target detection data into the trained preset detection model, and obtain the cable fault detection result. Thus, based on the fault current signal, features from three different views are constructed: time-series features corresponding to the original three-phase current signal, artificial time-domain features, and two-dimensional image features. This provides a more comprehensive reflection of the fault characteristics of early cable faults from multiple views, enhancing the feature representation capability of fault information. Furthermore, a feature fusion framework based on the combination of autoencoder, Fisher score and Softmax function was designed to effectively fuse the features of the three heterogeneous views, resulting in a fused feature that integrates fault information from the three views. This fused feature integrates the advantages of multiple views, achieves complementarity of multiple view information, and can improve the accuracy of subsequent fault detection. Attached Figure Description
[0066] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0067] Figure 1 This is a diagram illustrating the application environment of a cable fault detection method in one embodiment.
[0068] Figure 2 This is a flowchart illustrating a cable fault detection method in one embodiment;
[0069] Figure 3 This is a structural block diagram of a cable fault detection device in one embodiment;
[0070] Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0071] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0072] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.
[0073] The cable fault detection method provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or located on the cloud or other network servers. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, drones, low-altitude aircraft, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, projection devices, etc. Portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, etc. Head-mounted devices can be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc. Server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.
[0074] In one exemplary embodiment, such as Figure 2 As shown, a cable fault detection method is provided, which can be applied to... Figure 1 Taking terminal 102 as an example, the explanation includes the following steps 202 to 216. Wherein:
[0075] Step 202: Obtain historical fault data for the cable.
[0076] The historical fault data includes fault-free data, cable half-cycle early fault data, and cable multi-cycle early fault data, along with corresponding tags.
[0077] For example, historical fault data that can characterize early cable fault data is collected to determine the three-phase current data under fault-free conditions, cable half-cycle early fault, and cable multi-cycle early fault, i.e. fault-free data, cable half-cycle early fault data, and cable multi-cycle early fault data, respectively labeled as fault-free, cable half-cycle early fault, and cable multi-cycle early fault.
[0078] In one embodiment, the historical fault data of the cable can also be used as early fault simulation data of the cable to build an early fault simulation platform for simulating early faults of the cable in a low-voltage distribution network scenario.
[0079] Step 204: Preprocess the historical fault data to obtain processed historical fault data.
[0080] Optionally, the historical fault data can be preprocessed to obtain processed historical fault data.
[0081] Step 206: Based on the processed historical fault data, determine the time series characteristics, artificial time domain characteristics, and two-dimensional image characteristics of the cable.
[0082] For example, based on the processed historical fault data, the time series characteristics corresponding to the original three-phase current signals are determined. Artificial time-domain characteristics calculated from three-phase current signals Two-dimensional image features obtained by converting three-phase current signals using Gram angle field technology. .
[0083] Among them, the three-phase current signals are the current signals in the fault-free data, the cable half-cycle early fault data, and the cable multi-cycle early fault data.
[0084] In one embodiment, the time series feature consists of the raw signal data of the original three-phase current, which has three channels and is a feature with a dimension of 3*N, as shown in formula (1):
[0085] (1)
[0086] in, These represent the currents of phase A, phase B, and phase C, respectively, and each contains... Each time sampling point.
[0087] Step 208: Based on the autoencoder network, the time series features, artificial time-domain features, and two-dimensional image features are dimensionality reduced to obtain three dimensionality-reduced deep features.
[0088] Optionally, three autoencoder networks are constructed to reduce the dimensionality of time series features, artificial time-domain features, and two-dimensional image features. Each autoencoder contains an encoder network and a decoder network. After the features are input into the encoder network for dimensionality reduction, three deep features are obtained. The specific formulas are shown in formulas (2)-(4):
[0089] (2)
[0090] (3)
[0091] (4)
[0092] in, , and These represent the encoder networks corresponding to the three autoencoder models. , and These are the three deep features obtained after dimensionality reduction by the encoder, and their corresponding dimensions are all... .
[0093] The three decoder networks reconstruct the three corresponding dimensionality-reduced deep features to restore the original features, as shown in formulas (5)-(7):
[0094] (5)
[0095] (6)
[0096] (7)
[0097] in, , and These represent the three decoder networks corresponding to the three autoencoder models. , and These are the corresponding reconstructed features.
[0098] Based on the characteristics of reconstruction , and Calculation and three view features , and The reconstruction loss is used to train three autoencoders, and the three autoencoder models are saved after training.
[0099] Step 210: Fuse the dimensionality-reduced depth features to obtain multi-view fused features.
[0100] For example, the contribution score and corresponding weight of each dimensionality reduction depth feature are calculated, and a weighted sum is performed to obtain the multi-view fusion feature.
[0101] Step 212: Construct a training set for fault detection based on multi-view fusion features and labels.
[0102] Optionally, a training set for fault detection can be constructed based on multi-view fusion features and labels.
[0103] In one embodiment, a verification set for fault detection can also be constructed based on multi-view fusion features and labels.
[0104] Step 214: Input the training set into the preset detection model for training. When the detection accuracy of the preset detection model reaches the preset level, the trained preset detection model is obtained.
[0105] For example, the training set is input into the preset detection model for training, and the preset detection model is validated by the validation set. When the detection accuracy of the preset detection model reaches the preset level, the trained preset detection model is obtained.
[0106] The preset detection model can be a cable early fault detection model based on DELM-KELM-Adaboost. The construction of the DELM-KELM-Adaboost fault detection model includes two basic classification models: Deep Extreme Learning Machine (DELM) and Kernel Extreme Learning Machine (KELM). Using Adaboost's ensemble learning technology, DELM and KELM are integrated to obtain the DELM-KELM-Adaboost ensemble learning model.
[0107] Step 216: Obtain the target detection data of the cable, input the target detection data into the trained preset detection model, and obtain the fault detection results of the cable.
[0108] Optionally, target detection data of the cable can be acquired and input into a pre-trained detection model to obtain fault detection results of the cable.
[0109] In the aforementioned cable fault detection method, historical fault data of the cable is acquired, preprocessed, and then processed to obtain processed historical fault data. Based on the processed historical fault data, the time-series features, artificial time-domain features, and two-dimensional image features of the cable are determined. Using an autoencoder network, the time-series features, artificial time-domain features, and two-dimensional image features are dimensionality-reduced to obtain three dimensionality-reduced depth features. These dimensionality-reduced depth features are then fused to obtain multi-view fusion features. Based on the multi-view fusion features and labels, a training set for fault detection is constructed. The training set is input into a preset detection model for training. Once the detection accuracy of the preset detection model reaches a preset level, the trained preset detection model is obtained. Finally, target detection data of the cable is acquired and input into the trained preset detection model to obtain the cable fault detection result. Thus, features from three different views are constructed based on the fault current signal: time-series features corresponding to the original three-phase current signal, artificial time-domain features, and two-dimensional image features. This provides a more comprehensive reflection of the fault characteristics of early cable faults from multiple views, enhancing the feature representation capability of fault information. Furthermore, a feature fusion framework based on the combination of autoencoder, Fisher score and Softmax function was designed to effectively fuse the features of the three heterogeneous views, resulting in a fused feature that integrates fault information from the three views. This fused feature integrates the advantages of multiple views, achieves complementarity of multiple view information, and can improve the accuracy of subsequent fault detection.
[0110] In an exemplary embodiment, the artificial time-domain characteristics of the cable are determined based on the processed historical fault data, including: calculating the time-domain characteristics corresponding to each fault data based on fault-free data, cable half-cycle early fault data and cable multi-cycle early fault data; and using the feature vector composed of the time-domain characteristics corresponding to the three types of fault data as the artificial time-domain characteristics of the cable.
[0111] In practical implementation, artificial time-domain characteristics are calculated from the three-phase current signals. The time-domain characteristics of the raw signal for each phase of the current are calculated, and the feature vector composed of the time-domain characteristics corresponding to the three types of fault data is used as the artificial time-domain characteristics of the cable. .
[0112] In the above embodiments, by calculating different time-domain features, artificial time-domain features of the data are used to characterize the data, making the changes in the signal more intuitive.
[0113] In one exemplary embodiment, the time-domain characteristics include peak-to-peak value, absolute mean, variance, standard deviation, root mean square, skewness, kurtosis of the current, and waveform factor of the current.
[0114] In actual implementation, peak-to-peak value The calculation formula is shown in formula (8):
[0115] (8)
[0116] in, It is the maximum value in the original current signal. This represents the minimum value of the original current signal.
[0117] absolute average The calculation formula is shown in formula (9):
[0118] (9)
[0119] variance The calculation formula is shown in formula (10):
[0120] (10)
[0121] in, It is the average value of the original current signal. The current is the current at the nth sampling point of the original current signal.
[0122] Standard deviation The calculation formula is shown in formula (11):
[0123] (11)
[0124] Root Mean Square The calculation formula is shown in formula (12):
[0125] (12)
[0126] Skewness The calculation formula is shown in formula (13):
[0127] (13)
[0128] kurtosis of current The calculation formula is shown in formula (14):
[0129] (14)
[0130] Waveform factor of current The calculation formula is shown in formula (15):
[0131] (15)
[0132] In the above embodiments, by calculating multiple time-domain features, the features can provide the most intuitive measure of amplitude change for even extremely weak instantaneous pulses or amplitude abrupt changes that may be caused by early faults.
[0133] In an exemplary embodiment, determining the two-dimensional image features of the cable based on processed historical fault data includes: normalizing the current sequence in the historical fault data to obtain a normalized current sequence; mapping the normalized current sequence to a polar coordinate system and generating a two-dimensional image matrix; and combining the current sequence in the historical fault data with the two-dimensional image matrix to obtain the two-dimensional image features of the cable.
[0134] In practical implementation, Gram angle field technology is used to convert three-phase current signals into two-dimensional image features. The original signal of each phase current of the three-phase current is converted into a Gram angle field image, resulting in three... A two-dimensional image; three two-dimensional images constitute the features of a two-dimensional image. The corresponding dimension is .
[0135] The current sequence is normalized, and the specific calculation formula is shown in formula (16):
[0136] (16)
[0137] in, Represents the first in the current sequence One current sampling point, Represents the normalized i-th One sampling point, This indicates the operation of retrieving the maximum value. This indicates the operation of taking the minimum value.
[0138] The normalized current sequence is converted into polar coordinates, and the specific conversion formula is shown in formula (17):
[0139] (17)
[0140] in, express The corresponding polar coordinate angle, express The corresponding timestamp, It is a constant factor used to adjust the span of the polar coordinate system.
[0141] In one embodiment, Gram angle and field are utilized. And Graham's difference The Gram angle and field matrix are generated, which are the two-dimensional image matrix. The specific calculation formulas are shown in formulas (18) and (19):
[0142] (18)
[0143] (19)
[0144] In the above embodiments, the two-dimensional image matrix is determined by Gram corner field technology, realizing the conversion of key information from one-dimensional temporal domain to two-dimensional spatial domain. For common, weak, complex temporal correlations and transient patterns in early faults, image representation can enhance them into more easily identifiable visual structures, significantly improving the entire system's ability to perceive and identify hidden fault features.
[0145] In an exemplary embodiment, fusing dimensionality-reduced depth features to obtain multi-view fusion features includes: calculating the contribution scores of the three dimensionality-reduced depth features respectively; determining the weight of each dimensionality-reduced depth feature based on the contribution scores; and performing a weighted summation of the dimensionality-reduced depth features based on the weights to obtain the multi-view fusion features.
[0146] In practical implementation, the calculation features , and The Fisher score for each element is its contribution score. The first one inside The Fisher score for each element is: , The first one inside The Fisher score for each element is: , The first one inside The Fisher score for each element is: The specific formula for calculating Fisher's score is shown in formula (20):
[0147] (20)
[0148] in, It is the total number of categories. It is the first in the fault classification label kind, It is the first Number of samples in each category It is the first The first in the class The average of the elements, It is the first The first in the class The variance of each element.
[0149] Based on the contribution scores of different features, the weight of each dimension reduction depth feature is calculated, and the specific calculation formula is shown in formula (21):
[0150] (twenty one)
[0151] in, Indicates the first The weights of deep-level features, yes function, Indicates the first The first deep-level feature Fisher score for each element.
[0152] The multi-view fusion feature is obtained by weighted summation based on the weights and contribution scores of the three dimensionality-reduced deep features. The specific calculation formula is shown in formula (22):
[0153] (twenty two)
[0154] In the above embodiments, an intelligent weighted fusion mechanism based on Fisher scores and the Softmax function is introduced. This mechanism can automatically evaluate and assign differentiated weights to features of different views, so that features with strong discriminative power dominate the final fused features. The resulting fused features integrate the complementary advantages of multi-dimensional fault information such as temporal details, statistical regularities, and spatial structure to the greatest extent, overcoming the inherent defects of the limited representation capabilities of a single view.
[0155] In an exemplary embodiment, historical fault data is preprocessed to obtain processed historical fault data, including: removing outliers, deleting duplicate data, and filtering noise from the historical fault data to obtain cleaned historical fault data; resampling the cleaned historical fault data to obtain processed historical fault data; the number of sampling points for fault-free data, cable half-cycle early fault data, and cable multi-cycle early fault data in the processed historical fault data is the same.
[0156] In practice, outlier removal, duplicate data deletion, and noise filtering are performed on historical fault data to obtain cleaned historical fault data. The current signal is then resampled using a cubic spline interpolation algorithm to ensure that the number of sampling points is the same for fault-free data, cable half-cycle early fault data, and cable multi-cycle early fault data.
[0157] In the above embodiments, data cleaning can filter out false signals and interference that are unrelated to early faults, preventing these noises from being extracted and amplified as valid features in subsequent steps, thereby improving the signal-to-noise ratio and reliability of subsequent feature construction.
[0158] To illustrate the cable fault detection method in this application in detail, an embodiment is described below. For example, this application describes a cable fault detection method in a specific scenario.
[0159] First, historical fault data that can characterize early cable faults are collected to determine the three-phase current data under fault-free conditions, cable half-cycle early faults, and cable multi-cycle early faults. These are labeled as fault-free data, cable half-cycle early fault data, and cable multi-cycle early fault data, respectively.
[0160] Historical fault data is preprocessed to obtain processed historical fault data. Based on the processed historical fault data, the time series characteristics corresponding to the original three-phase current signals are determined. Artificial time-domain characteristics calculated from three-phase current signals Two-dimensional image features obtained by converting three-phase current signals using Gram angle field technology. .
[0161] Three autoencoder networks are constructed to reduce the dimensionality of time series features, artificial time-domain features, and two-dimensional image features. Each autoencoder contains an encoder network and a decoder network. After the features are input into the encoder network for dimensionality reduction, three deep features are obtained.
[0162] Calculate the contribution score and corresponding weight of each dimensionality-reduced deep feature, and perform a weighted sum to obtain the multi-view fusion feature. Based on the multi-view fusion feature and labels, construct a training set for fault detection.
[0163] The training set is input into a preset detection model for training, and the preset detection model is validated using a validation set. Once the detection accuracy of the preset detection model reaches a preset level, the trained preset detection model is obtained. Target detection data for the cable is acquired and input into the trained preset detection model to obtain the cable fault detection results.
[0164] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.
[0165] Based on the same inventive concept, this application also provides a cable fault detection device for implementing the cable fault detection method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more cable fault detection device embodiments provided below can be found in the limitations of the cable fault detection method described above, and will not be repeated here.
[0166] In one exemplary embodiment, such as Figure 3 As shown, a cable fault detection device is provided, including: an acquisition module 301, a processing module 302, a determination module 303, a dimensionality reduction module 304, a fusion module 305, a construction module 306, and a training module 307, wherein:
[0167] The acquisition module is used to acquire historical fault data of the cable, which includes fault-free data, early half-cycle fault data and early multi-cycle fault data of the cable and corresponding tags.
[0168] The processing module is used to preprocess the historical fault data to obtain processed historical fault data.
[0169] The determination module is used to determine the time series characteristics, artificial time domain characteristics, and two-dimensional image characteristics of the cable based on the processed historical fault data.
[0170] The dimensionality reduction module is used to reduce the dimensionality of the time series features, the artificial time-domain features, and the two-dimensional image features based on the autoencoder network, to obtain three dimensionality-reduced depth features.
[0171] The fusion module is used to fuse the dimensionality-reduced depth features to obtain multi-view fused features.
[0172] A construction module is used to construct a training set for fault detection based on the multi-view fusion features and the labels.
[0173] The training module is used to input the training set into the preset detection model for training, and to obtain the trained preset detection model when the detection accuracy of the preset detection model reaches a preset level.
[0174] The acquisition module is also used to acquire target detection data of the cable, input the target detection data into the trained preset detection model, and obtain the fault detection result of the cable.
[0175] In one exemplary embodiment, the determining module is further configured to:
[0176] Based on the fault-free data, the cable half-cycle early fault data, and the cable multi-cycle early fault data, calculate the time-domain characteristics corresponding to each fault data.
[0177] The feature vector composed of the time-domain features corresponding to the three types of fault data is used as the artificial time-domain feature of the cable.
[0178] In one exemplary embodiment, the time-domain features include peak-to-peak value, absolute mean, variance, standard deviation, root mean square, skewness, kurtosis of the current, and waveform factor of the current.
[0179] In one exemplary embodiment, the determining module is further configured to:
[0180] The current sequence in the historical fault data is normalized to obtain the normalized current sequence.
[0181] The normalized current sequence is mapped to a polar coordinate system, and a two-dimensional image matrix is generated;
[0182] The current sequence in the historical fault data is combined with the two-dimensional image matrix to obtain the two-dimensional image features of the cable.
[0183] In one exemplary embodiment, the fusion module is further configured to:
[0184] Calculate the contribution scores of the three dimensionality-reduced deep features respectively;
[0185] Based on the contribution score, the weight of each dimensionality reduction depth feature is determined;
[0186] Based on the weights, the reduced-dimensional depth features are weighted and summed to obtain the multi-view fusion features.
[0187] In one exemplary embodiment, the above-described processing module is further configured to:
[0188] The historical fault data is subjected to outlier removal, duplicate data deletion and noise filtering to obtain cleaned historical fault data;
[0189] The historical fault data after cleaning is resampled to obtain processed historical fault data; the number of sampling points for fault-free data, cable half-cycle early fault data and cable multi-cycle early fault data are the same in the processed historical fault data.
[0190] Each module in the aforementioned cable fault detection device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.
[0191] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a cable fault detection method.
[0192] The display unit of this computer device is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of this computer device can be a touch layer covering the display screen, or buttons, a trackball, or a touchpad set on the casing of the computer device, or an external keyboard, touchpad, or mouse, etc.
[0193] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0194] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:
[0195] Acquire historical fault data of the cable, including fault-free data, cable half-cycle early fault data, cable multi-cycle early fault data, and corresponding tags;
[0196] The historical fault data is preprocessed to obtain processed historical fault data;
[0197] Based on the processed historical fault data, the time series characteristics, artificial time domain characteristics, and two-dimensional image characteristics of the cable are determined;
[0198] Based on an autoencoder network, the time series features, the artificial time-domain features, and the two-dimensional image features are dimensionality-reduced to obtain three dimensionality-reduced depth features.
[0199] The reduced-dimensional depth features are fused to obtain multi-view fusion features;
[0200] Based on the multi-view fusion features and the labels, a training set for fault detection is constructed;
[0201] The training set is input into a preset detection model for training. When the detection accuracy of the preset detection model reaches a preset level, the trained preset detection model is obtained.
[0202] Obtain target detection data of the cable, input the target detection data into the trained preset detection model, and obtain the fault detection result of the cable.
[0203] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:
[0204] Acquire historical fault data of the cable, including fault-free data, cable half-cycle early fault data, cable multi-cycle early fault data, and corresponding tags;
[0205] The historical fault data is preprocessed to obtain processed historical fault data;
[0206] Based on the processed historical fault data, the time series characteristics, artificial time domain characteristics, and two-dimensional image characteristics of the cable are determined;
[0207] Based on an autoencoder network, the time series features, the artificial time-domain features, and the two-dimensional image features are dimensionality-reduced to obtain three dimensionality-reduced depth features.
[0208] The reduced-dimensional depth features are fused to obtain multi-view fusion features;
[0209] Based on the multi-view fusion features and the labels, a training set for fault detection is constructed;
[0210] The training set is input into a preset detection model for training. When the detection accuracy of the preset detection model reaches a preset level, the trained preset detection model is obtained.
[0211] Obtain target detection data of the cable, input the target detection data into the trained preset detection model, and obtain the fault detection result of the cable.
[0212] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:
[0213] Acquire historical fault data of the cable, including fault-free data, cable half-cycle early fault data, cable multi-cycle early fault data, and corresponding tags;
[0214] The historical fault data is preprocessed to obtain processed historical fault data;
[0215] Based on the processed historical fault data, the time series characteristics, artificial time domain characteristics, and two-dimensional image characteristics of the cable are determined;
[0216] Based on an autoencoder network, the time series features, the artificial time-domain features, and the two-dimensional image features are dimensionality-reduced to obtain three dimensionality-reduced depth features.
[0217] The reduced-dimensional depth features are fused to obtain multi-view fusion features;
[0218] Based on the multi-view fusion features and the labels, a training set for fault detection is constructed;
[0219] The training set is input into a preset detection model for training. When the detection accuracy of the preset detection model reaches a preset level, the trained preset detection model is obtained.
[0220] Obtain target detection data of the cable, input the target detection data into the trained preset detection model, and obtain the fault detection result of the cable.
[0221] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.
[0222] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.
[0223] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.
[0224] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A cable fault detection method, characterized in that, The method includes: Acquire historical fault data of the cable, including fault-free data, cable half-cycle early fault data, cable multi-cycle early fault data, and corresponding tags; The historical fault data is preprocessed to obtain processed historical fault data; Based on the processed historical fault data, the time series characteristics, artificial time domain characteristics, and two-dimensional image characteristics of the cable are determined; Based on an autoencoder network, the time series features, the artificial time-domain features, and the two-dimensional image features are dimensionality-reduced to obtain three dimensionality-reduced depth features. Calculate the contribution scores of the three dimensionality-reduced deep features respectively; The formula for calculating the contribution score is as follows: ； in, Let be the contribution score of the i-th element in the dimensionality-reduced deep feature. It is the total number of tag categories. It is the first in the tag kind, It is the first Number of samples in each category It is the first The first in the class The average of the elements, It is the first The first in the class The variance of each element; Based on the contribution score, the weight of each dimensionality reduction depth feature is determined; The formula for calculating the weight is: ； in, Indicates the first The weights of each dimension-reduced depth feature. yes function, Indicates the first The first dimensionality reduction depth feature The contribution score of each element, M is the number of all reduced-dimensionality deep feature elements; Based on the weights, the reduced depth features are weighted and summed to obtain the multi-view fusion features; The calculation formula for the multi-view fusion feature is as follows: ； in, For multi-view fusion features, For dimensionality reduction depth features The corresponding weights For dimensionality reduction depth features The corresponding weights For dimensionality reduction depth features The corresponding weights; Based on the multi-view fusion features and the labels, a training set for fault detection is constructed; The training set is input into a preset detection model for training. When the detection accuracy of the preset detection model reaches a preset level, the trained preset detection model is obtained. Obtain target detection data of the cable, input the target detection data into the trained preset detection model, and obtain the fault detection result of the cable.
2. The method according to claim 1, characterized in that, The determination of the artificial time-domain characteristics of the cable based on the processed historical fault data includes: Based on the fault-free data, the cable half-cycle early fault data, and the cable multi-cycle early fault data, calculate the time-domain characteristics corresponding to each fault data. The feature vector composed of the time-domain features corresponding to the three types of fault data is used as the artificial time-domain feature of the cable.
3. The method according to claim 2, characterized in that, The time-domain features include peak-to-peak value, absolute mean, variance, standard deviation, root mean square, skewness, kurtosis of current, and waveform factor of current.
4. The method according to claim 1, characterized in that, The step of determining the two-dimensional image features of the cable based on the processed historical fault data includes: The current sequence in the historical fault data is normalized to obtain the normalized current sequence. The normalized current sequence is mapped to a polar coordinate system, and a two-dimensional image matrix is generated; The current sequence in the historical fault data is combined with the two-dimensional image matrix to obtain the two-dimensional image features of the cable.
5. The method according to claim 1, characterized in that, The preprocessing of the historical fault data to obtain processed historical fault data includes: The historical fault data is subjected to outlier removal, duplicate data deletion and noise filtering to obtain cleaned historical fault data; The historical fault data after cleaning is resampled to obtain processed historical fault data; the number of sampling points for fault-free data, cable half-cycle early fault data and cable multi-cycle early fault data are the same in the processed historical fault data.
6. A cable fault detection device, characterized in that, The device includes: The acquisition module is used to acquire historical fault data of the cable, which includes fault-free data, cable half-cycle early fault data, cable multi-cycle early fault data, and corresponding tags. The processing module is used to preprocess the historical fault data to obtain processed historical fault data; The determination module is used to determine the time series characteristics, artificial time domain characteristics, and two-dimensional image characteristics of the cable based on the processed historical fault data; The dimensionality reduction module is used to reduce the dimensionality of the time series features, the artificial time-domain features, and the two-dimensional image features based on the autoencoder network, to obtain three dimensionality-reduced depth features. The fusion module is used to calculate the contribution scores of the three dimensionality-reduced deep features respectively; The formula for calculating the contribution score is as follows: ； in, Let be the contribution score of the i-th element in the dimensionality-reduced deep feature. It is the total number of tag categories. It is the first in the tag kind, It is the first Number of samples in each category It is the first The first in the class The average of the elements, It is the first The first in the class The variance of each element; Based on the contribution score, the weight of each dimensionality reduction depth feature is determined; The formula for calculating the weight is: ； in, Indicates the first The weights of each dimension-reduced depth feature. yes function, Indicates the first The first dimensionality reduction depth feature The contribution score of each element, M is the number of all reduced-dimensionality deep feature elements; Based on the weights, the reduced depth features are weighted and summed to obtain the multi-view fusion features; The calculation formula for the multi-view fusion feature is as follows: ； in, For multi-view fusion features, For dimensionality reduction depth features The corresponding weights For dimensionality reduction depth features The corresponding weights For dimensionality reduction depth features The corresponding weights; A construction module is used to construct a training set for fault detection based on the multi-view fusion features and the labels; The training module is used to input the training set into the preset detection model for training, and to obtain the trained preset detection model when the detection accuracy of the preset detection model reaches a preset level. The acquisition module is also used to acquire target detection data of the cable, input the target detection data into the trained preset detection model, and obtain the fault detection result of the cable.
7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.
9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.