Cable carbonization occurrence position identification method and device

Abstract

The invention discloses a cable carbonization occurrence position identification method and device, and the method comprises the steps: respectively extracting a cable carbonization one-dimensional line selection feature and a cable carbonization two-dimensional line selection feature according to current signal data and voltage signal data in a normal operation stage and a cable carbonization continuous stage; and performing spatial-temporal feature extraction and fusion on the cable carbonization one-dimensional line selection feature and the cable carbonization two-dimensional line selection feature by adopting a preset multi-modal feature fusion model, constructing a target multi-modal feature fusion model, and performing cable carbonization occurrence position identification on a to-be-identified cable according to the target multi-modal feature fusion model. According to the method, the cable carbonization position can be accurately and effectively positioned, the cable fault can be quickly removed, an effective fire prevention solution is provided for a low-voltage power distribution system, and the technical problems that an existing cable carbonization position identification method is difficult to quickly position the cable carbonization position and cannot quickly remove the cable fault are solved.

Description

Technical Field

[0001] This invention relates to the field of electrical safety technology, and in particular to a method and apparatus for identifying the location of cable carbonization. Background Technology

[0002] During the rapid development of power systems, the increased limiting current of equipment leads to highly destructive electric arcs generated during short-circuit faults, which can easily ignite flammable materials and cause fires. Approximately 60% of electrical fires are directly caused by cable carbonization, especially in residential areas and power stations. Carbonization of series cables caused by aging lines, damaged insulation, or poor contact is difficult to detect effectively with traditional protection devices due to its small current amplitude and high concealment, posing a significant hidden danger to electrical safety.

[0003] Current methods for identifying the location of cable carbonization are insufficient for quickly locating the carbonization site, hindering rapid fault removal and impacting the reliability and safety of the power system. Summary of the Invention

[0004] This invention provides a method and apparatus for identifying the location of cable carbonization, which solves the technical problem that existing methods for identifying the location of cable carbonization are unable to quickly locate the location of cable carbonization and thus cannot quickly cut off cable faults.

[0005] In view of this, the first aspect of the present invention provides a method for identifying the location of cable carbonization, comprising:

[0006] Current signal data containing both the normal operation phase and the continuous carbonization phase of the cable were collected from the established cable carbonization experimental platform.

[0007] Calculate the one-dimensional cable carbonization line selection characteristics based on the current signal data, and construct a one-dimensional cable carbonization line selection characteristic database;

[0008] Voltage signal data containing the normal operation phase and the continuous carbonization phase of the cable were collected from the established cable carbonization experimental platform, and the cable carbonization effect current signal data were obtained based on the current signal data.

[0009] The voltage signal data and the cable carbonization effect current signal are processed to obtain a binary pixel map of the VI trajectory, and a two-dimensional cable carbonization line selection feature database is constructed.

[0010] Based on the one-dimensional cable carbonization line selection feature database and the two-dimensional cable carbonization line selection feature database, the pre-set multimodal feature fusion model is trained to obtain the target multimodal feature fusion model.

[0011] Based on the target multimodal feature fusion model, the location of cable carbonization is identified for the cable to be identified.

[0012] Optionally, a one-dimensional cable carbonization line selection feature database is constructed based on the current signal data, including:

[0013] The current signal data is divided into time windows based on a complete sine wave cycle. The one-dimensional cable carbonization line selection characteristics within each time window are calculated to construct a one-dimensional cable carbonization line selection characteristic database. The one-dimensional cable carbonization line selection characteristics include current waveform factor, margin factor, kurtosis, peak value, RMS value, average value, variance waveform factor, current harmonic amplitude in the frequency domain, and arc characteristics of a preset frequency band in the time and frequency domain.

[0014] Optionally, the arc characteristics of the preset frequency band in the time-frequency domain are the time-frequency domain characteristics obtained by wavelet packet transformation of the current signal.

[0015] Optionally, the voltage signal data and the cable carbonization effect current signal are processed to obtain a binary pixel map of the VI trajectory, and a two-dimensional cable carbonization line selection feature database is constructed, including:

[0016] The voltage signal data is divided into time windows based on a complete sine wave cycle to obtain the loop voltage signal within the time window.

[0017] Subtract the current signal data during the continuous carbonization phase from the current signal data during the normal operation phase to obtain the current signal of the cable carbonization effect.

[0018] The loop voltage signal and the cable carbonization effect current signal within the time window are normalized to construct a binary VI trajectory matrix;

[0019] Based on the binary VI trajectory matrix, a binary pixel map of the VI trajectory is drawn, and a two-dimensional line selection feature database of cable carbonization is constructed.

[0020] Optionally, the pre-set multimodal feature fusion model includes a CNN-SE model, a BiLSTM-T model, and a multimodal feature fusion module;

[0021] The CNN-SE model is used to process the binary pixel map of the VI trajectory to obtain spatial features;

[0022] The BiLSTM-T model is used to process the one-dimensional line selection features of the cable carbonization to obtain time-series features;

[0023] The multimodal feature fusion module is used to fuse the spatial features and the temporal features to output the identification result of the location of cable carbonization.

[0024] Optionally, the multimodal feature fusion module employs an adaptive weighted fusion strategy to fuse the spatial features and the temporal features, with the fusion formula being:

[0025]

[0026] Where F is the fusion function, Let GELUE be the time-series weight matrix, and GELUE() be the activation function. Let b be the spatial weight matrix and b be the bias matrix.

[0027] Optionally, the normalization formulas for the loop voltage signal and the cable carbonization effect current signal within the time window are:

[0028]

[0029]

[0030] in, Here, k represents the normalized voltage, m represents the number of sampling points, and N represents the size of the two-dimensional image. The maximum value in the voltage sequence. It is the minimum value in the voltage sequence. This is the minimum value in the current sequence. The maximum value in the current sequence. This is the normalized current.

[0031] A second aspect of the present invention provides a cable carbonization location identification device, comprising:

[0032] The first signal acquisition module is used to collect current signal data containing the normal operation phase and the continuous carbonization phase of the cable from the established cable carbonization experimental platform.

[0033] A one-dimensional feature extraction module is used to calculate the one-dimensional cable carbonization line selection features based on the current signal data and construct a one-dimensional cable carbonization line selection feature database.

[0034] The second signal acquisition module is used to collect voltage signal data containing the normal operation stage and the continuous stage of cable carbonization from the established cable carbonization experimental platform, and to obtain cable carbonization effect current signal data based on the current signal data.

[0035] A two-dimensional feature extraction module is used to process the voltage signal data and the cable carbonization effect current signal to obtain a binary pixel map of the VI trajectory and construct a two-dimensional cable carbonization line selection feature database.

[0036] The feature fusion model construction module is used to train a pre-set multimodal feature fusion model based on the one-dimensional cable carbonization line selection feature database and the two-dimensional cable carbonization line selection feature database to obtain a target multimodal feature fusion model.

[0037] The identification module is used to identify the location of cable carbonization in the cable to be identified based on the target multimodal feature fusion model.

[0038] Optionally, the one-dimensional feature extraction module is specifically used for:

[0039] The current signal data is divided into time windows based on a complete sine wave cycle. The one-dimensional cable carbonization line selection characteristics within each time window are calculated to construct a one-dimensional cable carbonization line selection characteristic database. The one-dimensional cable carbonization line selection characteristics include current waveform factor, margin factor, kurtosis, peak value, RMS value, average value, variance waveform factor, current harmonic amplitude in the frequency domain, and arc characteristics of a preset frequency band in the time and frequency domain.

[0040] Optionally, the two-dimensional feature extraction module is specifically used for:

[0041] The voltage signal data is divided into time windows based on a complete sine wave cycle to obtain the loop voltage signal within the time window.

[0042] Subtract the current signal data during the continuous carbonization phase from the current signal data during the normal operation phase to obtain the current signal of the cable carbonization effect.

[0043] The loop voltage signal and the cable carbonization effect current signal within the time window are normalized to construct a binary VI trajectory matrix;

[0044] Based on the binary VI trajectory matrix, a binary pixel map of the VI trajectory is drawn, and a two-dimensional line selection feature database of cable carbonization is constructed.

[0045] As can be seen from the above technical solutions, the cable carbonization location identification method provided by the present invention has the following advantages:

[0046] The cable carbonization location identification method provided by this invention extracts one-dimensional and two-dimensional cable carbonization location features based on current and voltage signal data containing data from the normal operation phase and the continuous carbonization phase. A pre-set multimodal feature fusion model is used to extract and fuse the spatiotemporal features of these features, constructing a target multimodal feature fusion model. Based on this model, the location of cable carbonization is identified, enabling accurate and effective rapid location of cable carbonization. This facilitates rapid fault removal and provides an effective fire prevention solution for cable systems. It solves the technical problem that existing cable carbonization location identification methods struggle to quickly locate the location of cable carbonization and cannot rapidly remove cable faults. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart illustrating a method for identifying the location of cable carbonization in an embodiment of the present invention.

[0049] Figure 2 This is the cable carbonization experimental platform provided in the embodiments of the present invention;

[0050] Figure 3 This is a schematic diagram of the cable carbonization VI trajectory generation process provided in an embodiment of the present invention;

[0051] Figure 4 This is a diagram of the BiLSTM-T neural network structure provided in this embodiment of the invention;

[0052] Figure 5 This is a structural diagram of the pre-set multimodal feature fusion model provided in the embodiments of the present invention;

[0053] Figure 6 This is a graph showing the accuracy of cable carbonization location identification under different load conditions provided in this embodiment of the invention.

[0054] Figure 7 This is a comparison chart of the line selection accuracy under different features of different models provided in the embodiments of the present invention;

[0055] Figure 8 This is a schematic diagram of a cable carbonization location identification device provided in an embodiment of the present invention. Detailed Implementation

[0056] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] For easier understanding, please refer to Figure 1 This invention provides an embodiment of a method for identifying the location of cable carbonization, comprising:

[0058] Step 101: Collect current signal data containing the normal operation phase and the continuous carbonization phase of the cable from the established cable carbonization test platform.

[0059] It should be noted that a cable carbonization experimental platform should be set up first, such as... Figure 2 As shown, the cable carbonization experimental platform includes an AC power grid, a computer, an oscilloscope, carbonization equipment, carbonized cables, and multiple experimental loads. The AC power grid, carbonized cables, and experimental loads form a circuit. The AC power grid provides AC power to the carbonized cable circuit, and the carbonization equipment provides a fault arc for the carbonized cable. The oscilloscope acquires the voltage and current waveforms of the carbonized cable circuit. The computer identifies the location of cable carbonization. The multiple experimental loads include fluorescent lamps, a vacuum cleaner, an electric drill, and a switching power supply, all connected in parallel. Current measuring points measure the system current, voltage measuring points measure the system voltage, and arc voltage measuring points measure the arc voltage signal. Current signal data, including both the normal operation phase and the ongoing cable carbonization phase, are collected at the current measuring points using the cable carbonization experimental platform.

[0060] Step 102: Calculate the one-dimensional cable carbonization line selection characteristics based on the current signal data, and construct a one-dimensional cable carbonization line selection characteristic database.

[0061] It should be noted that one-dimensional cable carbonization line selection features are extracted from current signal data to construct a one-dimensional cable carbonization line selection feature database. In one embodiment, the current signal data is divided into time windows based on a complete sine wave cycle (20ms), and the one-dimensional cable carbonization line selection features within each time window are calculated to construct the one-dimensional cable carbonization line selection feature database. The one-dimensional cable carbonization line selection features include current waveform factor, margin factor, kurtosis, peak value, RMS value, average value, variance waveform factor, current harmonic amplitude in the frequency domain (2nd, 5th, and 8th harmonic amplitudes), and arc characteristics in the time-frequency domain within a preset frequency band (frequency band 1~10, excluding frequency band 4).

[0062] The formula for calculating the current waveform factor is:

[0063]

[0064] Where W is the current waveform factor, and i is the system current. This represents the effective value of the system current.

[0065] The formula for calculating the margin factor is:

[0066]

[0067] Where L is the margin factor, This is the peak value.

[0068] The formula for calculating kurtosis is:

[0069]

[0070] Where K is the kurtosis, n is the number of points in a certain time window, and N is the total number of points in the time window. The average value of the system current. For the system current timing point, Let Variance be the variance.

[0071] The formula for calculating the peak value is:

[0072]

[0073] The formula for calculating the effective value is:

[0074]

[0075] The formula for calculating the average is:

[0076]

[0077] The formula for calculating variance is:

[0078]

[0079] The formula for calculating the amplitude of current harmonics in the frequency domain is:

[0080]

[0081] in, For system current signal, The fundamental amplitude, The first harmonic angular frequency, For the kth harmonic phase, Let be the amplitude of the kth harmonic.

[0082] The arc characteristics in the preset frequency bands (bands 1 to 10, excluding band 4) in the time-frequency domain are obtained by wavelet packet transform of the current signal. A 6-level wavelet packet decomposition is performed using the rbio4.4 wavelet basis, and the calculation method is as follows:

[0083]

[0084] in, As an energy characteristic, These are wavelet coefficients.

[0085] Normalizing the calculated energy, we get:

[0086]

[0087] in, The normalized energy characteristics, Let i be the time series point of the energy characteristic, i be the time sequence number, and j be the frequency sequence number.

[0088] Step 103: Collect voltage signal data containing the normal operation stage and the continuous carbonization stage of the cable from the established cable carbonization experimental platform, and obtain the cable carbonization effect current signal data based on the current signal data.

[0089] It should be noted that voltage signal data, including both the normal operation phase and the continuous carbonization phase, were collected at voltage measurement points using a cable carbonization experimental platform. The voltage signal data was divided into time windows based on a complete sine wave cycle (20ms) to obtain the loop voltage signal within the time window.

[0090] The current signal data of the cable carbonization effect is obtained by decoupling using the current separation method. Essentially, it is the current signal data during normal operation minus the current signal data during the continuous carbonization phase of the cable.

[0091] The circuit voltage signal and cable carbonization effect current signal within the time window are normalized, converting the voltage and current into integers between 0 and N. The calculation formula is as follows:

[0092]

[0093]

[0094] in, The voltage is the normalized value, k is the number of sampling points, m is the number of sampling points, and N is the size of the two-dimensional image. In this embodiment of the invention, the size N of the two-dimensional image is 32. The maximum value in the voltage sequence. It is the minimum value in the voltage sequence. This is the minimum value in the current sequence. The maximum value in the current sequence. This is the normalized current.

[0095] Create an N×N matrix M, and store the transformed sampling points ( , Mapping this to matrix M, and iterating through all voltage-current sequences, let M( , =0, and obtain the binary VI trajectory matrix.

[0096] Step 104: Process the voltage signal data and the cable carbonization effect current signal to obtain the VI trajectory binary pixel map and construct a two-dimensional cable carbonization line selection feature database.

[0097] It should be noted that the binary pixel image of the VI trajectory is drawn based on the binary VI trajectory matrix. The image has 255 pixels and a size of N×N. The process of generating the VI trajectory for cable carbonization is as follows: Figure 3 As shown, a two-dimensional cable carbonization line selection feature database is constructed, consisting of binary pixel maps of VI trajectories.

[0098] Step 105: Based on the one-dimensional cable carbonization line selection feature database and the two-dimensional cable carbonization line selection feature database, train the pre-set multimodal feature fusion model to obtain the target multimodal feature fusion model.

[0099] It should be noted that a pre-built multimodal feature fusion model is constructed, which includes a CNN-SE model, a BiLSTM-T model, and a multimodal feature fusion module.

[0100] The CNN-SE model is used to process the binary pixel map of the VI trajectory to learn the potential correlation between loop voltage and arc effect current, thereby obtaining spatial features. The parameter configuration of the CNN-SE model is shown in Table 1.

[0101] Table 1

[0102]

[0103] The BiLSTM-T model is used to process one-dimensional line selection features of cable carbonization to obtain time-series features. The BiLSTM-T model mainly combines the structural features of BiLSTM and Transformer, integrating their advantages in time-series feature extraction and global key feature capture to learn the intrinsic relationship between cable carbonization effect current and load category. Its structure is as follows: Figure 4 As shown.

[0104] The multimodal feature fusion module is used to fuse spatial and temporal features. Finally, after passing through a fully connected layer and an activation function, it outputs the cable carbonization location identification result. The overall structure of the multimodal feature fusion module is as follows: Figure 5 As shown. The multimodal feature fusion module employs an adaptive weighted fusion strategy to fuse spatial and temporal features. The fusion formula is as follows:

[0105]

[0106] Where F is the fusion function, Let GELUE be the time-series weight matrix, and GELUE() be the activation function. Let b be the spatial weight matrix and b be the bias matrix.

[0107] Step 106: Based on the target multimodal feature fusion model, identify the location of cable carbonization in the cable to be identified.

[0108] It should be noted that, for cables whose carbonization locations need to be identified, current and voltage signals are collected using a cable carbonization experimental platform. One-dimensional and two-dimensional cable carbonization features are extracted from these signals. These features are then input into a target multimodal feature fusion model to obtain the cable carbonization location identification result output by the model. In one embodiment, the cable carbonization location can also be displayed on a computer user interface.

[0109] The cable carbonization location identification method provided by this invention extracts one-dimensional and two-dimensional cable carbonization location features based on current and voltage signal data containing data from the normal operation phase and the continuous carbonization phase of the cable. A pre-set multimodal feature fusion model is used to extract and fuse the spatiotemporal features of these features, constructing a target multimodal feature fusion model. Based on this model, the location of cable carbonization is identified, enabling accurate and effective rapid location of cable carbonization. This facilitates rapid fault removal and provides an effective fire prevention solution for low-voltage power distribution systems. It solves the technical problem that existing cable carbonization location identification methods struggle to quickly locate the location of cable carbonization and cannot rapidly remove cable faults.

[0110] To verify the effectiveness of the cable carbonization location identification method provided by this invention, model performance tests were conducted on this invention under different load levels. The accuracy results for cable carbonization location identification under different load levels are as follows: Figure 6 As shown, the average accuracy rate is 99.58% under single load; the line selection accuracy rate is slightly lower under two loads compared to single load, with an average accuracy rate of 96.39%; and the average line selection accuracy rate is 94.00% under three and four loads.

[0111] The multimodal feature fusion model of this invention was compared with BiLSTM, BiLSTM-T, CNN, and CNN-SE models, and the accuracy was as follows: Figure 7 As shown, the line selection accuracy of the multimodal feature fusion model of the present invention reaches 96.44%, which is significantly better than the accuracy of other models.

[0112] For easier understanding, please refer to Figure 8 This invention provides an embodiment of a cable carbonization location identification device, comprising:

[0113] The first signal acquisition module is used to collect current signal data containing the normal operation phase and the continuous carbonization phase of the cable from the established cable carbonization experimental platform.

[0114] A one-dimensional feature extraction module is used to calculate the one-dimensional cable carbonization line selection features based on the current signal data and construct a one-dimensional cable carbonization line selection feature database.

[0115] The second signal acquisition module is used to collect voltage signal data containing the normal operation stage and the continuous stage of cable carbonization from the established cable carbonization experimental platform, and to obtain cable carbonization effect current signal data based on the current signal data.

[0116] A two-dimensional feature extraction module is used to process the voltage signal data and the cable carbonization effect current signal to obtain a binary pixel map of the VI trajectory and construct a two-dimensional cable carbonization line selection feature database.

[0117] The feature fusion model construction module is used to train a pre-set multimodal feature fusion model based on the one-dimensional cable carbonization line selection feature database and the two-dimensional cable carbonization line selection feature database to obtain a target multimodal feature fusion model.

[0118] The identification module is used to identify the location of cable carbonization in the cable to be identified based on the target multimodal feature fusion model.

[0119] In one embodiment, the one-dimensional feature extraction module is specifically used for:

[0120] The current signal data is divided into time windows based on a complete sine wave cycle. The one-dimensional cable carbonization line selection characteristics within each time window are calculated to construct a one-dimensional cable carbonization line selection characteristic database. The one-dimensional cable carbonization line selection characteristics include current waveform factor, margin factor, kurtosis, peak value, RMS value, average value, variance waveform factor, current harmonic amplitude in the frequency domain, and arc characteristics of a preset frequency band in the time and frequency domain.

[0121] In one embodiment, the two-dimensional feature extraction module is specifically used for:

[0122] The voltage signal data is divided into time windows based on a complete sine wave cycle to obtain the loop voltage signal within the time window.

[0123] Subtract the current signal data during the continuous carbonization phase from the current signal data during the normal operation phase to obtain the current signal of the cable carbonization effect.

[0124] The loop voltage signal and the cable carbonization effect current signal within the time window are normalized to construct a binary VI trajectory matrix;

[0125] Based on the binary VI trajectory matrix, a binary pixel map of the VI trajectory is drawn, and a two-dimensional line selection feature database of cable carbonization is constructed.

[0126] In one embodiment, the pre-set multimodal feature fusion model includes a CNN-SE model, a BiLSTM-T model, and a multimodal feature fusion module;

[0127] The CNN-SE model is used to process the binary pixel map of the VI trajectory to obtain spatial features;

[0128] The BiLSTM-T model is used to process the one-dimensional line selection features of the cable carbonization to obtain time-series features;

[0129] The multimodal feature fusion module is used to fuse the spatial features and the temporal features to output the identification result of the location of cable carbonization.

[0130] In one embodiment, the multimodal feature fusion module employs an adaptive weighted fusion strategy to fuse the spatial features and the temporal features, with the fusion formula being:

[0131]

[0132] Where F is the fusion function, Let GELUE be the time-series weight matrix, and GELUE() be the activation function. Let b be the spatial weight matrix and b be the bias matrix.

[0133] In one embodiment, the normalization formulas for the loop voltage signal and the cable carbonization effect current signal within the time window are as follows:

[0134]

[0135] in, Here, k represents the normalized voltage, m represents the number of sampling points, and N represents the size of the two-dimensional image. The maximum value in the voltage sequence. It is the minimum value in the voltage sequence. This is the minimum value in the current sequence. The maximum value in the current sequence. This is the normalized current.

[0136] The cable carbonization location identification device provided in this invention is used to execute the cable carbonization location identification method provided in this invention. Its principle and the technical effects achieved are the same as those of the cable carbonization location identification method provided in this invention, and will not be repeated here.

[0137] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0138] The above-described 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying the location of cable carbonization, characterized in that, include: Current signal data containing both the normal operation phase and the continuous carbonization phase of the cable were collected from the established cable carbonization experimental platform. Calculate the one-dimensional cable carbonization line selection characteristics based on the current signal data, and construct a one-dimensional cable carbonization line selection characteristic database; Voltage signal data containing the normal operation phase and the continuous carbonization phase of the cable were collected from the established cable carbonization experimental platform, and the cable carbonization effect current signal data were obtained based on the current signal data. The voltage signal data and the cable carbonization effect current signal are processed to obtain a binary pixel map of the VI trajectory, and a two-dimensional cable carbonization line selection feature database is constructed. Based on the one-dimensional cable carbonization line selection feature database and the two-dimensional cable carbonization line selection feature database, the pre-set multimodal feature fusion model is trained to obtain the target multimodal feature fusion model. Based on the target multimodal feature fusion model, the location of cable carbonization is identified for the cable to be identified.

2. The method for identifying the location of cable carbonization according to claim 1, characterized in that, Based on the current signal data, calculate the one-dimensional cable carbonization line selection characteristics, and construct a one-dimensional cable carbonization line selection characteristic database, including: The current signal data is divided into time windows based on a complete sine wave cycle. The one-dimensional cable carbonization line selection characteristics within each time window are calculated to construct a one-dimensional cable carbonization line selection characteristic database. The one-dimensional cable carbonization line selection characteristics include current waveform factor, margin factor, kurtosis, peak value, RMS value, average value, variance waveform factor, current harmonic amplitude in the frequency domain, and arc characteristics of a preset frequency band in the time and frequency domain.

3. The method for identifying the location of cable carbonization according to claim 2, characterized in that, The arc characteristics of the preset frequency band in the time-frequency domain are the time-frequency domain characteristics obtained by wavelet packet transformation of the current signal.

4. The method for identifying the location of cable carbonization according to claim 2, characterized in that, The voltage signal data and the cable carbonization effect current signal are processed to obtain a binary pixel map of the VI trajectory, and a two-dimensional cable carbonization line selection feature database is constructed, including: The voltage signal data is divided into time windows based on a complete sine wave cycle to obtain the loop voltage signal within the time window. Subtract the current signal data during the continuous carbonization phase from the current signal data during the normal operation phase to obtain the current signal of the cable carbonization effect. The loop voltage signal and the cable carbonization effect current signal within the time window are normalized to construct a binary VI trajectory matrix; Based on the binary VI trajectory matrix, a binary pixel map of the VI trajectory is drawn, and a two-dimensional line selection feature database of cable carbonization is constructed.

5. The method for identifying the location of cable carbonization according to claim 1, characterized in that, The pre-built multimodal feature fusion model includes a CNN-SE model, a BiLSTM-T model, and a multimodal feature fusion module; The CNN-SE model is used to process the binary pixel map of the VI trajectory to obtain spatial features; The BiLSTM-T model is used to process the one-dimensional line selection features of the cable carbonization to obtain time-series features; The multimodal feature fusion module is used to fuse the spatial features and the temporal features to output the identification result of the location of cable carbonization.

6. The method for identifying the location of cable carbonization according to claim 5, characterized in that, The multimodal feature fusion module employs an adaptive weighted fusion strategy to fuse the spatial features and the temporal features. The fusion formula is as follows: Where F is the fusion function, Let GELUE be the time-series weight matrix, and GELUE() be the activation function. Let be the spatial weight matrix, and b be the bias matrix.

7. The method for identifying the location of cable carbonization according to claim 4, characterized in that, The normalization formulas for the loop voltage signal and the cable carbonization effect current signal within the time window are as follows: in, Here, k represents the normalized voltage, m represents the number of sampling points, and N represents the size of the two-dimensional image. The maximum value in the voltage sequence. It is the minimum value in the voltage sequence. This is the minimum value in the current sequence. The maximum value in the current sequence. This is the normalized current.

8. A device for identifying the location of cable carbonization, characterized in that, include: The first signal acquisition module is used to collect current signal data containing the normal operation phase and the continuous carbonization phase of the cable from the established cable carbonization experimental platform. A one-dimensional feature extraction module is used to calculate the one-dimensional cable carbonization line selection features based on the current signal data and construct a one-dimensional cable carbonization line selection feature database. The second signal acquisition module is used to collect voltage signal data containing the normal operation stage and the continuous stage of cable carbonization from the established cable carbonization experimental platform, and to obtain cable carbonization effect current signal data based on the current signal data. A two-dimensional feature extraction module is used to process the voltage signal data and the cable carbonization effect current signal to obtain a binary pixel map of the VI trajectory and construct a two-dimensional cable carbonization line selection feature database. The feature fusion model construction module is used to train a pre-set multimodal feature fusion model based on the one-dimensional cable carbonization line selection feature database and the two-dimensional cable carbonization line selection feature database to obtain a target multimodal feature fusion model. The identification module is used to identify the location of cable carbonization in the cable to be identified based on the target multimodal feature fusion model.

9. The cable carbonization location identification device according to claim 8, characterized in that, The one-dimensional feature extraction module is specifically used for: The current signal data is divided into time windows based on a complete sine wave cycle. The one-dimensional cable carbonization line selection characteristics within each time window are calculated to construct a one-dimensional cable carbonization line selection characteristic database. The one-dimensional cable carbonization line selection characteristics include current waveform factor, margin factor, kurtosis, peak value, RMS value, average value, variance waveform factor, current harmonic amplitude in the frequency domain, and arc characteristics of a preset frequency band in the time and frequency domain.

10. The cable carbonization location identification device according to claim 9, characterized in that, The two-dimensional feature extraction module is specifically used for: The voltage signal data is divided into time windows based on a complete sine wave cycle to obtain the loop voltage signal within the time window. Subtract the current signal data during the continuous carbonization phase from the current signal data during the normal operation phase to obtain the current signal of the cable carbonization effect. The loop voltage signal and the cable carbonization effect current signal within the time window are normalized to construct a binary VI trajectory matrix; Based on the binary VI trajectory matrix, a binary pixel map of the VI trajectory is drawn, and a two-dimensional line selection feature database of cable carbonization is constructed.

Images