Cable terminal discharge early warning method, device and system based on dual-channel waveform identification

Abstract

The invention relates to the field of data signal processing, in particular to a network algorithm cable terminal discharge early warning method, device and system based on dual-channel waveform identification. The method comprises the following steps of: firstly, acquiring a periodic ultrasonic signal and an ultrahigh frequency electromagnetic field signal generated by discharging of a cable terminal, converting the two signals into a corresponding time-frequency diagram by utilizing short-time Fourier transform, and performing normalization processing to serve as input of a subsequent network; a dual-channel high-order feature extraction network formed by cascading a ConvNeXt backbone network and a convolution block attention module is constructed, and is used for extracting time-frequency diagram detail fault features of ultrasonic signals and high-frequency electromagnetic field signals. And finally, by constructing an adaptive feature fusion network, fusing the fault features of the two channels, and obtaining cross-channel fault joint features by using an adaptive weight coefficient vector, thereby outputting a diagnosis result, and realizing accurate evaluation of a defect state of an adjacent cable terminal.

Description

Technical Field

[0001] The present invention relates to the field of data signal processing, and in particular to a network algorithm cable terminal discharge early warning method, device and system based on dual-channel waveform recognition. Background Art

[0002] The development of urban economy has made urban land resources increasingly scarce. In order to improve the utilization rate of urban land, a large number of new lines or lines in operation have been transformed to operate in an overhead-cable hybrid mode, and the cable rate has increased year by year. Cable terminals are important media for connecting cables and overhead transmission conductors. They play the role of shielding, insulation and stress control, and are indispensable when hybrid transmission lines are put into operation.

[0003] The cable terminal joint has a complex structure and serious electric field distortion. Its insulation performance is lower than that of the cable body, and it is a weak point in the cable system. According to statistics from the State Grid Corporation of China, among the operational failures of cable lines, cable accessory failures account for more than 70%; among the cable accessory failures, insulation failures account for about 40%. Therefore, timely detection of defects in cable terminal joints and repairs are of great significance to improving their operational reliability.

[0004] At present, partial discharge detection technology has become an ideal detection method for cable terminal insulation defect detection. Among them, the most widely used methods mainly include pulse current method, ultra-high frequency method, ultrasonic detection method, infrared detection method, etc. Among them, the pulse current method obtains discharge information such as apparent discharge amount and discharge phase by measuring the pulse current on the cable terminal grounding wire. It has high measurement sensitivity, but is mostly used for offline measurement; the ultra-high frequency method and ultrasonic method have strong discharge detection ability by measuring the UHF and ultrasonic signals generated when the cable terminal discharges, and can realize online monitoring, but it is difficult to distinguish the strength of the discharge when diagnosing alone, there are a lot of false alarms, and it is impossible to locate the specific phase of the specific loop line, which may cause misjudgment; the infrared detection method receives the infrared rays radiated by the cable terminal to measure the abnormal temperature rise caused by the discharge. It has strong anti-interference ability, but there is a monitoring blind spot and poor economic efficiency.

[0005] When there are insulation defects in the cable terminal, a long-term partial discharge phenomenon will occur, accompanied by significant changes in ultrasonic, UHF, and temperature signals. Ultrasonic and UHF signals will diffuse into space with air as the medium, and temperature signals can cause abnormal temperature rise on the terminal surface. Existing cable terminal partial discharge detection methods generally use a single parameter for separate monitoring, which has the problem of multi-system islanding, making it difficult to eliminate a large number of false alarms and difficult to locate specific terminals. Summary of the invention

[0006] This patent provides a network algorithm cable terminal discharge warning method, equipment and system based on dual-channel waveform recognition, which can realize the simultaneous monitoring of ultrasonic and UHF signals, realize the defect status assessment of adjacent cable terminals, and combine with the surface temperature measurement of the cable terminal to realize the accurate positioning of the specific loop, specific point and specific phase discharge point of the cable terminal.

[0007] The technical solution of the present invention is as follows:

[0008] The cable terminal discharge early warning method based on dual-channel waveform recognition is characterized by comprising the following steps:

[0009] S1: The hybrid line panoramic perception and rapid inspection system collects the periodic ultrasonic signal and UHF signal generated by the cable terminal discharge, and then performs STFT transformation on the collected ultrasonic signal and UHF signal to obtain their corresponding time-frequency diagrams, which are normalized and used as the input of the overall network;

[0010] S2: Construct a dual-channel high-order feature extraction network for ultrasonic signals and UHF signals by cascading the ConvNeXt backbone network and the convolutional block attention module. Each channel obtains the detailed fault features of the corresponding time-frequency graph. and

[0011] S3: Construct an adaptive feature fusion network to obtain fault features and The adaptive weight coefficient vector is integrated to obtain the cross-channel fault joint features, output the diagnosis results, and realize the defect status assessment of the adjacent cable terminal;

[0012]

[0013] Where F is the fused cross-channel fault joint feature; α and β are adaptive weight coefficient vectors, which are determined by the adaptive weight network composed of the global pooling layer, the fully connected layer and the Sigmoid activation function, satisfying α i ∈[0,1],β i ∈[0,1],α i +β i =1;

[0014] The process of obtaining the adaptive weight coefficient vectors α and β is as follows: First, the fault characteristics and Perform global pooling to compress the fault features into the depth dimension to obtain the global fault feature G I and G U , as shown in formula (4);

[0015]

[0016] Then the global feature G I and G U Merge, and then pass through the fully connected layer and Sigmoid activation function to transform the global fault feature G I and G U Converted into weights, as shown in formula (5).

[0017]

[0018] In the formula, H represents the number of extracted feature column vectors, It is d Depth, d = 1, 2, ..., D; G I ,G U ∈R 1×1×D ; FC represents the fully connected layer function that implements feature weighting.

[0019] The cable terminal discharge early warning method based on dual-channel waveform recognition measures the abnormal temperature rise caused by the cable terminal discharge, and determines the accurate location of the defective discharge position in conjunction with the adjacent cable terminal defect status assessment result output in step S3.

[0020] The cable terminal discharge early warning method based on dual-channel waveform recognition, the ultrasonic signal, the detection frequency band range is 20kHz~100MHz, and the acquisition is scheduled; the UHF signal, the detection frequency band range is 300MHz~3GHz, and the acquisition is synchronized with the ultrasonic sensor; the temperature measurement range is -40℃~125℃, and the acquisition is synchronized with the ultrasonic sensor.

[0021] In the cable terminal discharge warning method based on dual-channel waveform recognition, in step S1, the Fourier STFT transform implementation process is:

[0022] ① Select a suitable window function such as Hamming window, Hanning window, rectangular window, etc. to intercept the ultrasound

[0023] Signal and short period of high frequency electromagnetic field signal waveform;

[0024] ② Divide the input signal waveform into frames according to the window length and overlap rate;

[0025] ③ Perform fast Fourier transform FFT on each frame of signal waveform and generate a time-frequency diagram.

[0026] The power spectrum shows the strength of the signal waveform at different times and frequencies;

[0027] The expression of the time domain signal s(t) after STFT transformation is:

[0028]

[0029] Where f is the frequency, S(t,f) is the output of the STFT transform, s(τ) is the original input signal, and w(t) is a window function with a fixed length.

[0030] In the cable terminal discharge warning method based on dual-channel waveform recognition, in step S1, the normalization processing is to adjust the time-frequency graph data to a distribution with a mean of 0 and a standard deviation of 1 through the Z-score normalization method; the implementation process is as follows:

[0031]

[0032] Where X is the original time-frequency graph data, X' is the normalized data, μ is the mean, and σ is the standard deviation.

[0033] In the cable terminal discharge warning method based on dual-channel waveform recognition, in step S2,

[0034] The ConvNeXt backbone network includes the following architecture:

[0035] ① Input layer. It is used to receive the time-frequency image data of ultrasound and UHF signals after STFT transformation and normalization, usually a 224×224×3 RGB image with 3 channels;

[0036] ② Preliminary processing block. Perform preliminary convolution operations to extract low-level features. Use convolution with a step size of 2 for downsampling, and use the GELU activation function to transform the 224×224×3 feature map into a 56×56×96 feature map with 96 channels;

[0037] ③ Convolutional block. The convolutional block is the core component of the network, consisting of multiple convolutional layers and optimized modules. The convolution operation is used for feature extraction. Nonlinearity is introduced through the GELU activation function, and then the feature map is standardized to alleviate the gradient vanishing problem in training. A jump connection similar to ResNet is used to add the input and output to promote gradient flow, and some paths are randomly discarded to enhance the robustness of the model. After convolution and activation, the feature map shape is converted to 7×7×768, and the number of channels is 768;

[0038] ④ Bottleneck layer: Use 1*1 convolution to perform channel compression, reduce the number of channels and spatial dimensions to effectively reduce the amount of calculation, thereby improving calculation efficiency, and the number of feature map channels remains unchanged.

[0039] The convolutional block attention module is an attention mechanism for enhancing the performance of convolutional neural networks. First, channel attention is applied to the time-frequency graph to obtain key channel information; then spatial attention is applied to mark the key areas of the time-frequency graph, based on which the detailed features of the time-frequency graph corresponding to each channel can be accurately obtained.

[0040] In the cable terminal discharge warning method based on dual-channel waveform recognition, in step S2, the convolution block attention module is implemented as follows:

[0041] ① Channel attention module: Perform global average pooling and global maximum pooling on the input time-frequency map to generate two channel descriptors, pass these two descriptors through a shared multi-layer perceptron to generate a channel attention map, and apply attention weights by element-by-element multiplication with the original feature map. After the channel attention module, the feature map shape remains at 7×7×768, but the features of each channel have been weighted according to the channel attention;

[0042] ② Spatial attention module: Global average pooling and global maximum pooling are performed on the output of each channel attention module to generate a spatial descriptor. The two descriptors are combined to generate a spatial attention map, which is also multiplied with the feature map by element-by-element multiplication to apply spatial attention. After the spatial attention module, the feature map shape remains at 7×7×768, but the features of each spatial position have been weighted according to the spatial attention.

[0043] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the cable terminal discharge warning method based on a dual-channel synchronous waveform recognition network algorithm as described in any of the above items are implemented.

[0044] The cable terminal discharge warning system based on dual-channel waveform recognition is installed at the cable terminal, and includes a central control system, which is characterized by including a sensor module, a short-distance communication module, a data preprocessing module, a data analysis module, a data fusion module, a fault diagnosis module and an integrated communication module;

[0045] The sensing module includes an ultrasonic sensor, an ultra-high frequency sensor, and a wireless temperature sensor. The ultrasonic sensor and the ultra-high frequency sensor are integrated inside the system. The ultrasonic sensor is used to measure the periodic ultrasonic signal generated by the discharge of the cable terminal. The ultra-high frequency sensor is used to measure the periodic UHF signal generated by the discharge of the cable terminal. The wireless temperature sensor adopts a strap-type infrared temperature measurement technology, and integrates an optical system and a photoelectric converter inside. It is installed on the surface of the cable terminal and is used to measure the abnormal temperature rise generated by the discharge of the cable terminal.

[0046] The short-distance communication module adopts a wireless transmitting and receiving module to realize short-distance communication between the wireless temperature measurement sensor and the system;

[0047] The data preprocessing module includes a filtering circuit and a signal amplifying circuit, which respectively perform basic noise filtering and signal amplification on the collected ultrasonic, UHF and temperature signals for subsequent data analysis and data fusion;

[0048] The data analysis module performs preliminary classification of the collected data according to the time-frequency characteristics of the collected signals, and distinguishes the collected ultrasound, UHF, and temperature signals into valid and invalid signals, discards the invalid signals, and stores and uploads the valid signals;

[0049] The data fusion module integrates the data from the ultrasonic sensor, the ultra-high frequency sensor, and the temperature sensor to ensure the consistency of data from different sources, process the differences in data format and timestamp, enhance the system's fault tolerance to data loss or sensor failure, and ensure the stable operation of the system;

[0050] The data storage module stores the valid data collected at regular intervals and supports real-time retrieval and viewing;

[0051] The integrated communication module aggregates the data of each module inside the device and realizes external communication with the monitoring center, cloud platform or other terminal devices to perform more complex data processing and artificial intelligence intelligent training, and realize accurate early warning of cable terminal defect discharge.

[0052] The cable terminal discharge warning system based on dual-channel waveform recognition is characterized in that:

[0053] The ultrasonic sensor has a detection frequency band range of 20kHz to 100MHz;

[0054] The detection frequency band of the UHF sensor is 300MHz~3GHz;

[0055] The temperature measurement range of the wireless temperature sensor is -40℃~125℃;

[0056] The short-distance communication module described above adopts a 433MHz wireless transmitting and receiving module.

[0057] UHF signal refers to ultra-high frequency electromagnetic field signal.

[0058] The present invention first collects the periodic ultrasonic signal and UHF signal generated by the cable terminal discharge. These two signals respectively reflect the different physical characteristics of the cable terminal discharge. The ultrasonic signal mainly reflects the mechanical vibration effect of the discharge, while the UHF signal reflects the electromagnetic radiation generated by the discharge.

[0059] The collected signal is transformed by STFT to convert the time domain signal into a time-frequency diagram. STFT transformation can provide information about the signal in both the time and frequency domains, which is helpful for subsequent feature extraction. The time-frequency diagram is normalized to ensure the consistency of the input data and facilitate the processing of the subsequent network model.

[0060] The present invention constructs a dual-channel high-order feature extraction network composed of a cascade of a ConvNeXt backbone network and a convolutional block attention module. The ConvNeXt network is based on its powerful feature extraction capability, while the convolutional block attention module further enhances the network's ability to extract key features by introducing an attention mechanism.

[0061] The ultrasonic signal channel and the UHF signal channel are respectively input into the corresponding network branches. After a series of convolution operations and attention mechanism processing, the detailed fault features of the time-frequency graphs of their respective channels are extracted.

[0062] The present invention constructs an adaptive feature fusion network for fusing the fault features of the ultrasonic signal channel and the UHF signal channel. The network can dynamically adjust the weights of the two channel features to obtain more accurate cross-channel fault joint features.

[0063] The present invention outputs the discharge warning result and defect status evaluation of the cable terminal. By comparing the preset threshold value or model prediction result, it can be judged whether there is a discharge phenomenon at the cable terminal and the severity of the discharge, so as to take corresponding preventive measures.

[0064] The cable terminal discharge warning method based on dual-channel waveform recognition combines the dual detection methods of ultrasonic signals and UHF signals, and uses advanced deep learning networks for feature extraction and fusion, thus achieving accurate warning of cable terminal discharge phenomena and defect status assessment. This technical solution provides a strong technical guarantee for the safe operation of the power system.

[0065] The present invention collects ultrasonic signals and UHF signals generated by cable terminal discharge, and uses the ConvNeXt backbone network and convolutional block attention module to extract high-order features. Through the adaptive feature fusion network, the features of the two channels are fused to obtain the cross-channel fault joint feature. Finally, the discharge warning result and defect status assessment of the cable terminal are output based on the feature. The present invention improves the accuracy and reliability of cable terminal discharge warning, and provides a strong guarantee for the safe operation of the power system.

[0066] Aiming at the problems of hidden defects in cable terminals and islanding of multiple monitoring systems, the present invention provides a cable terminal discharge warning method based on dual-channel waveform identification. The method is installed on a cable terminal tower and has the functions of measuring ultrasonic signals, UHF signals and cable terminal surface temperature signals. Through data preprocessing, dual-channel high-order feature extraction, adaptive feature fusion and other processes, the cable terminal can be accurately identified in its healthy state, defective state and faulty state. Combined with the abnormal temperature rise signal of its surface monitoring, the defect position can be accurately located. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1It is a flow chart of the present invention.

[0068] Figure 2 This is a working principle diagram of the present invention.

[0069] Figure 3 It is the dual-channel synchronous waveform recognition network algorithm model architecture.

[0070] Figure 4 Build an example for a UHF signal spectrogram.

[0071] (a) Original UHF signal waveform - left; (b) Power spectrum after STFT transformation.

[0072] Figure 5 Build an example for a time-frequency plot of an ultrasound signal.

[0073] (a) Original ultrasonic signal waveform - left; (b) Power spectrum after STFT transformation.

[0074] Figure 6 It is the basic architecture of the dual-channel high-order feature extraction network.

[0075] Figure 7 It is an adaptive weight network. DETAILED DESCRIPTION

[0076] Example 1

[0077] See also Figure 2 The cable terminal discharge warning system based on dual-channel waveform recognition is installed on the cable terminal near the tower, including a central control system, including a sensor module, a short-distance communication module, a data preprocessing module, a data analysis module, a data fusion module, a fault diagnosis module and an integrated communication module;

[0078] The sensing module includes an ultrasonic sensor, an ultra-high frequency sensor, a wireless temperature sensor, and an ultrasonic sensor and an ultra-high frequency sensor are integrated inside the system.

[0079] (1) Ultrasonic sensor. It is made of piezoelectric ceramic material and integrated inside the device. It is used to measure the periodic ultrasonic signal generated by the discharge of the cable terminal. The detection frequency range is 20kHz to 100MHz. The data is collected regularly at a frequency of 1 time / h.

[0080] (2) Ultra-high frequency sensor. It uses ceramic, polymer and metal as the base material, is integrated inside the device, and is used to measure the periodic UHF signal generated by the discharge of the cable terminal. The detection frequency band ranges from 300MHz to 3GHz. It is synchronized with the ultrasonic sensor for regular collection, and the collection frequency is once / hour.

[0081] (3) Wireless temperature sensor. It uses a strap-type infrared temperature measurement technology, which integrates an optical system and a photoelectric converter. It is installed on the surface of the cable terminal to measure the abnormal temperature rise caused by the discharge of the cable terminal. It combines the ultrasonic and UHF signal fusion diagnosis results to determine the specific terminal point where the defective discharge phenomenon exists. The temperature measurement range is -40℃ to 125℃. It is synchronized with the ultrasonic sensor for regular data collection, and the data collection frequency is 1 time / h.

[0082] (4) Short-distance communication module: A 433MHz wireless transmitting and receiving module is used to achieve short-distance communication between the wireless temperature sensor and the device body within a range of 200m.

[0083] (5) Data preprocessing module. It includes filtering circuit and signal amplification circuit, which respectively perform basic noise filtering and signal amplification on the collected ultrasonic, UHF and temperature signals for subsequent data analysis and data fusion.

[0084] (6) Data analysis module: Based on the time-frequency characteristics of the collected signals, the collected data are preliminarily classified, and the collected ultrasonic, UHF, and temperature signals are distinguished into valid and invalid signals. Invalid signals are discarded, and valid signals are stored and uploaded.

[0085] (7) Data fusion module: Integrate data from multiple sensors (ultrasonic sensors, ultra-high frequency sensors, temperature sensors), ensure the consistency of data from different sources, handle differences in data formats, timestamps, etc., enhance the system's fault tolerance to data loss or sensor failure, and ensure stable system operation.

[0086] (8) Data storage module: Stores valid data collected at regular intervals for no less than three months and supports real-time retrieval and viewing.

[0087] (9) Integrated communication module. The data of each module inside the equipment is aggregated, and external communication with the monitoring center, cloud platform or other terminal devices is realized to carry out more complex data processing and artificial intelligence intelligent training, so as to achieve accurate early warning of cable terminal defect discharge.

[0088] Example 2

[0089] See also Figure 1-Figure 7 The cable terminal discharge warning method based on the dual-channel synchronous waveform recognition network algorithm includes the following steps:

[0090] S1: Collect the periodic ultrasonic signal generated by the discharge at the cable terminal, measure the UHF signal generated by the discharge at the cable terminal, perform STFT transformation on the waveforms of the ultrasonic signal and the high-frequency electromagnetic field signal, and obtain the corresponding time-frequency diagram; and normalize it as the input of the overall network;

[0091] The time-frequency diagram of the present invention is constructed

[0092] A large number of theoretical studies and engineering practices have shown that when insulation defects occur at the cable terminal and continuous discharge occurs, the ultrasonic and UHF signal waveform data generated have different characteristics in the time domain and frequency domain. However, the actual operating state of the power system is complex and changeable, and the characteristics of the cable terminal defect discharge in the time domain and frequency domain are difficult to effectively extract; and considering that the line cable terminal discharge is that both ultrasonic and UHF signals have obvious change characteristics, and their defect information has certain similarities and complementarities, this patent uses STFT transformation to convert the ultrasonic signal and UHF signal waveform generated during the cable terminal discharge into a time-frequency diagram (time-frequency image) to express its characteristics in a higher dimension.

[0093] The idea of ​​STFT transformation is:

[0094] ① Select a suitable window function (such as Hamming window, Hanning window, rectangular window, etc.) to capture a short period of the signal;

[0095] ② Divide the input signal into frames according to the window length and overlap rate;

[0096] ③ Perform fast Fourier transform (FFT) on each frame of the signal and generate a time-frequency diagram. The power spectrum can be used to show the intensity of the signal at different times and frequencies.

[0097] The expression of the time domain signal s(t) after STFT transformation is:

[0098]

[0099] Where f is the frequency, S(t,f) is the output of the STFT transform, s(τ) is the original input signal, and w(t) is a window function with a fixed length.

[0100] Implementation effect display:

[0101] Then the time-frequency graph data is normalized, and the data is adjusted to a distribution with a mean of 0 and a standard deviation of 1 through the Z-score standardization method. The implementation process is as follows:

[0102]

[0103] Where X is the original time-frequency graph data, X' is the normalized data, μ is the mean, and σ is the standard deviation.

[0104] S2: Construct a dual-channel high-order feature extraction network for ultrasonic signals and high-frequency electromagnetic field signals by cascading the ConvNeXt backbone network and the convolutional block attention module. Each channel obtains the detailed fault features of its corresponding time-frequency graph. and

[0105] The basic architecture of the dual-channel high-order feature extraction network is as follows Figure 6 As shown. ConvNeXt is a convolutional neural network (CNN)-based architecture that aims to combine the design of modern convolutional networks with the self-attention mechanism to improve the performance of tasks such as image classification. It mainly consists of the following architectures:

[0106] ① Input layer. It is used to receive the time-frequency image data of ultrasound and UHF signals after STFT transformation and normalization, usually a 224×224×3 RGB image with 3 channels;

[0107] ② Preliminary processing block. Perform preliminary convolution operations to extract low-level features. Use convolution with a step size of 2 for downsampling, and use the GELU activation function to transform the 224×224×3 feature map into a 56×56×96 feature map with 96 channels;

[0108] ③ Convolutional block. The convolutional block is the core component of the network, consisting of multiple convolutional layers and optimized modules. The convolution operation is used for feature extraction. Nonlinearity is introduced through the GELU activation function, and then the feature map is standardized to alleviate the gradient vanishing problem in training. A jump connection similar to ResNet is used to add the input and output to promote gradient flow, and some paths are randomly discarded to enhance the robustness of the model. After convolution and activation, the feature map shape is converted to 7×7×768, and the number of channels is 768;

[0109] ④ Bottleneck layer: Use 1*1 convolution to perform channel compression, reduce the number of channels and spatial dimensions to effectively reduce the amount of calculation, thereby improving calculation efficiency, and the number of feature map channels remains unchanged.

[0110] The convolutional block attention module is an attention mechanism used to enhance the performance of convolutional neural networks. It first applies channel attention to the feature map to obtain key channel information; then it applies spatial attention to mark the key areas of the feature map. Based on this, the detailed features of the time-frequency graph corresponding to each channel can be accurately obtained.

[0111] The key points of its implementation are:

[0112] ① Channel attention module: Perform global average pooling and global maximum pooling on the input time-frequency map to generate two channel descriptors, pass these two descriptors through a shared multi-layer perceptron to generate a channel attention map, and apply attention weights by element-by-element multiplication with the original feature map. After the channel attention module, the feature map shape remains at 7×7×768, but the features of each channel have been weighted according to the channel attention;

[0113] ② Spatial attention module: Global average pooling and global maximum pooling are performed on the output of each channel attention module to generate a spatial descriptor. The two descriptors are combined to generate a spatial attention map, which is also multiplied with the feature map by element-by-element multiplication to apply spatial attention. After the spatial attention module, the feature map shape remains at 7×7×768, but the features of each spatial position have been weighted according to the spatial attention.

[0114] S3: Construct an adaptive feature fusion network and obtain the fault features of ultrasonic and UHF signals through a dual-channel high-order feature extraction network. and When using individual fault characteristics for diagnosis, they all have certain limitations, resulting in a large number of false alarms.

[0115] like Figure 7 As shown in the figure, in order to better characterize the fault characteristics, an adaptive weight network construction method is proposed to obtain the adaptive weight coefficient vector, obtain the joint characteristics of cross-channel faults, output the diagnosis results, and realize the defect status assessment of the adjacent cable terminal;

[0116]

[0117] Where F is the fused cross-channel fault joint feature; α and β are adaptive weight coefficient vectors, which are determined by the adaptive weight network composed of the global pooling layer, the fully connected layer and the Sigmoid activation function, satisfying α i ∈[0,1],β i ∈[0,1],α i +β i =1;

[0118] The process of obtaining the adaptive weight coefficient vectors α and β is as follows: First, the fault characteristics and Perform global pooling to compress the fault features into the depth dimension to obtain the global fault feature G I and G U , as shown in formula (4);

[0119]

[0120] Then the global feature G I and G U Merge, and then pass through the fully connected layer and Sigmoid activation function to transform the global fault feature G I and G U Converted into weights, as shown in formula (5).

[0121]

[0122] In the formula, H represents the number of extracted feature column vectors, It is d Depth, d = 1, 2, ..., D; G I ,G U ∈R 1×1×D ; FC represents the fully connected layer function that implements feature weighting.

[0123] The abnormal temperature rise caused by the discharge at the cable terminal is measured, and combined with the abnormal temperature rise signal monitored on the cable terminal surface, and the defect status assessment result of the adjacent cable terminal output in step S3, the accurate location of the defective discharge position is determined.

Claims

1. A cable terminal discharge warning method based on dual-channel waveform recognition, characterized in that: The following steps are involved: S1: Collect the periodic ultrasonic signal and ultra-high frequency electromagnetic field UHF signal generated by the cable terminal discharge, and then perform short-time Fourier transform (STFT) on the collected ultrasonic signal and ultra-high frequency electromagnetic field UHF signal to obtain their corresponding time-frequency diagrams, and normalize them as the input of the overall network; S2: Construct a dual-channel high-order feature extraction network for ultrasonic signals and ultra-high frequency electromagnetic field UHF signals by cascading the ConvNeXt backbone network and the convolutional block attention module. Each channel obtains the detailed fault features of the corresponding time-frequency graph. and S3: Construct an adaptive feature fusion network to obtain fault features and The adaptive weight coefficient vector is fused to obtain the cross-channel fault joint features, output the diagnosis results, and realize the defect status assessment of the adjacent cable terminal; Where F is the fused cross-channel fault joint feature; α and β are adaptive weight coefficient vectors, which are determined by the adaptive weight network composed of the global pooling layer, the fully connected layer and the Sigmoid activation function, satisfying α i ∈[0,1],β i ∈[0,1],α i +β i =1; The process of obtaining the adaptive weight coefficient vectors α and β is as follows: First, the fault characteristics of ultrasonic and ultra-high frequency electromagnetic field UHF signals are analyzed. and Perform global pooling to compress the fault features into the depth dimension to obtain the global fault feature G I and G U , as shown in formula (4); Then the global feature G I and G U Merge, and then pass through the fully connected layer and Sigmoid activation function to transform the global fault feature G I and G U Converted into weights, as shown in formula (5), In the formula, H represents the number of extracted feature column vectors, It is d Depth, d = 1, 2, ..., D; G I ,G U ∈R 1×1×D ; FC represents the fully connected layer function that implements feature weighting.

2. The cable terminal discharge early warning method based on dual-channel waveform recognition according to claim 1 is characterized in that: The abnormal temperature rise caused by the discharge at the cable terminal is measured, and the defect status assessment result of the adjacent cable terminal output in step S3 is used to determine the accurate location of the defective discharge position.

3. The cable terminal discharge early warning method based on dual-channel waveform recognition according to claim 1 is characterized in that: The ultrasonic signal has a detection frequency band range of 20kHz to 100MHz and is collected regularly. The UHF signal has a detection frequency band range of 300MHz to 3GHz and is collected regularly and synchronously with the ultrasonic sensor. The temperature measurement range is -40℃ to 125℃ and is collected regularly and synchronously with the ultrasonic sensor.

4. The cable terminal discharge warning method based on dual-channel waveform recognition according to claim 1 is characterized in that: In step S1, the STFT transformation implementation process is: ① Select appropriate window functions such as Hamming window, Hanning window, rectangular window, etc. to capture short periods of ultrasonic signal and UHF signal waveform; ② Divide the input signal waveform into frames according to the window length and overlap rate; ③ Perform fast Fourier transform on each frame of signal waveform and generate a time-frequency diagram, using the power spectrum to show the intensity of the signal waveform at different times and frequencies; The expression of the time domain signal s(t) after STFT transformation is: Where f is the frequency, S(t,f) is the output of the STFT transform, s(τ) is the original input signal, and w(t) is a window function with a fixed length.

5. The cable terminal discharge early warning method based on dual-channel waveform recognition according to claim 1 is characterized in that: In step S1, the normalization process is to adjust the time-frequency graph data to a distribution with a mean of 0 and a standard deviation of 1 by using the Z-score normalization method; the implementation process is as follows: Where X is the original time-frequency graph data, X' is the normalized data, μ is the mean, and σ is the standard deviation.

6. The cable terminal discharge early warning method based on dual-channel waveform recognition according to claim 1 is characterized in that: In step S2, The ConvNeXt backbone network includes the following architecture: ① Input layer, used to receive the time-frequency image data of ultrasound and UHF signals after STFT transformation and normalization, usually 224×224×3 RGB image with 3 channels; ②Preliminary processing block, performs preliminary convolution operations and extracts low-level features; A convolution with a stride of 2 is used for downsampling, and the GELU activation function is used to transform the 224×224×3 feature map into a feature map of shape 56×56×96 with 96 channels; ③ Convolutional block: The convolutional block is the core component of the network. It consists of multiple convolutional layers and optimized modules. The convolution operation is used for feature extraction. Nonlinearity is introduced through the GELU activation function, and then the feature map is standardized to alleviate the gradient vanishing problem in training. A jump connection similar to ResNet is used to add the input and output to promote gradient flow, and some paths are randomly discarded to enhance the robustness of the model. After convolution and activation, the feature map shape is converted to 7×7×768, and the number of channels is 768; ④ Bottleneck layer, using 1*1 convolution to perform channel compression, reducing the number of channels and spatial dimensions to effectively reduce the amount of calculation, thereby improving calculation efficiency, and the number of feature map channels remains unchanged; The convolutional block attention module is an attention mechanism for enhancing the performance of convolutional neural networks. First, channel attention is applied to the time-frequency graph to obtain key channel information; then spatial attention is applied to mark the key areas of the time-frequency graph, based on which the detailed features of the time-frequency graph corresponding to each channel can be accurately obtained.

7. The cable terminal discharge early warning method based on dual-channel waveform recognition according to claim 6 is characterized in that: In step S2, the convolutional block attention module is implemented as follows: ① Channel attention module: Perform global average pooling and global maximum pooling on the input time-frequency map to generate two channel descriptors. These two descriptors are passed through a shared multi-layer perceptron to generate a channel attention map. The attention weight is applied by element-by-element multiplication with the original feature map. After the channel attention module, the feature map shape remains at 7×7×768, but the features of each channel have been weighted according to the channel attention; ② Spatial attention module: Global average pooling and global maximum pooling are performed on the output of each channel attention module to generate a spatial descriptor. The two descriptors are combined to generate a spatial attention map, which is also multiplied with the feature map by element-by-element multiplication and spatial attention is applied. After the spatial attention module, the shape of the feature map remains at 7×7×768, but the features of each spatial position have been weighted according to the spatial attention.

8. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the cable terminal discharge warning method based on a dual-channel synchronous waveform recognition network algorithm as described in any one of claims 1 to 7 are implemented.

9. A cable terminal discharge warning system based on dual-channel waveform recognition, installed at the cable terminal, central control system, characterized in that: It includes a sensing module, a short-distance communication module, a data preprocessing module, a data analysis module, a data fusion module, a fault diagnosis module and an integrated communication module; The sensing module includes an ultrasonic sensor, an ultra-high frequency sensor, and a wireless temperature sensor. The ultrasonic sensor and the ultra-high frequency sensor are integrated inside the system. The ultrasonic sensor is used to measure the periodic ultrasonic signal generated by the discharge of the cable terminal. The ultra-high frequency sensor is used to measure the periodic UHF signal generated by the discharge of the cable terminal. The wireless temperature sensor adopts a strap-type infrared temperature measurement technology, and integrates an optical system and a photoelectric converter inside. It is installed on the surface of the cable terminal and is used to measure the abnormal temperature rise generated by the discharge of the cable terminal. The short-distance communication module adopts a wireless transmitting and receiving module to realize short-distance communication between the wireless temperature measurement sensor and the system; The data preprocessing module includes a filtering circuit and a signal amplifying circuit, which respectively perform basic noise filtering and signal amplification on the collected ultrasonic, UHF and temperature signals for subsequent data analysis and data fusion; The data analysis module performs preliminary classification of the collected data according to the time-frequency characteristics of the collected signals, and distinguishes the collected ultrasound, UHF, and temperature signals into valid and invalid signals, discards the invalid signals, and stores and uploads the valid signals; The data fusion module integrates the data from the ultrasonic sensor, the ultra-high frequency sensor, and the temperature sensor to ensure the consistency of data from different sources, process the differences in data format and timestamp, enhance the system's fault tolerance to data loss or sensor failure, and ensure the stable operation of the system; The data storage module stores the valid data collected at regular intervals and supports real-time retrieval and viewing; The integrated communication module aggregates the data of each module inside the device and realizes external communication with the monitoring center, cloud platform or other terminal devices to perform more complex data processing and artificial intelligence intelligent training, and realize accurate early warning of cable terminal defect discharge.

10. The cable terminal discharge warning system based on dual-channel waveform recognition according to claim 9 is characterized in that: The ultrasonic sensor has a detection frequency band range of 20kHz to 100MHz; The detection frequency band of the UHF sensor is 300MHz~3GHz; The temperature measurement range of the wireless temperature sensor is -40℃~125℃; The short-distance communication module described above adopts a 433MHz wireless transmitting and receiving module.

Images