Grounding loop test method and system in operation state of high-voltage cable

Abstract

According to the grounding loop test method and system in the high-voltage cable operation state, the current response signal and the voltage response signal generated by responding to the excitation signal are analyzed through vector spectrum, and tiny but key changes along with signal instability in an electric power system can be detected more sensitively; a basis is provided for a waveform anomaly detection task; the voltage response three-dimensional spectrogram semantic feature vector and the current response three-dimensional spectrogram semantic feature vector are subjected to optimization analysis through external knowledge in combination with a self-attention mechanism, and while the semantic expression of the self-response three-dimensional spectrogram semantic feature vector is improved, the self-attention mechanism of the three-dimensional spectrogram semantic feature vector is improved. The method is beneficial for capturing interaction between spectrogram semantics in the current and voltage response signals, can perform subsequent single-phase resistance test under the condition of ensuring waveform stability, and is beneficial for improving accuracy and reliability of a grounding loop resistance test result.

Description

Technical Field

[0001] The present invention belongs to the technical field related to ground loop testing, and in particular relates to a ground loop testing method and system under a high-voltage cable operating state. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] In power transmission systems, high-voltage cables are key components for reliable power transmission. To ensure the safe and stable operation of power systems, regular testing of the ground loop resistance of high-voltage cables is essential. Ground loop resistance refers to the resistance between the equipment and the ground. A low and stable ground loop resistance effectively reduces hazards to personnel and equipment in the event of electrical equipment failures and ensures that protective devices (such as circuit breakers) respond accurately. Therefore, timely and effective ground loop resistance testing provides a crucial guarantee for the safe operation of the power grid.

[0004] However, traditional ground loop resistance testing methods typically require disconnecting the cable or equipment before measurement. This not only affects power supply continuity and leads to financial losses, but can also be difficult to implement in certain situations, such as for critical infrastructure and high-voltage power transmission cables. Furthermore, testing under high voltage conditions while high-voltage cables are in operation is inherently dangerous. When working on live systems, even simple connections or disconnections can pose a risk of electric shock.

[0005] Another major drawback of traditional ground loop resistance testing solutions is the lack or incompleteness of waveform verification and confirmation steps. Specifically, ground loop resistance testing places strict demands on signal quality and the measurement environment. If the waveform stability (non-oscillatory) is not guaranteed during testing, the resulting results may be unstable or even misleading. Traditional testing may not have adequate mechanisms to identify such instability, leading to continued analysis based on unreliable data, compromising the accuracy and reliability of ground loop resistance test results. Summary of the Invention

[0006] In order to overcome the deficiencies of the above-mentioned prior art, the present invention provides a ground loop testing method and system under the operation state of a high-voltage cable. The method adopts full vector spectrum analysis, combines the self-attention mechanism and external knowledge to capture the semantic interaction information between the response signals of the two, thereby performing waveform anomaly detection. Subsequent single-phase resistance testing can be performed while ensuring the stability of the waveform, thereby helping to improve the accuracy and reliability of the ground loop resistance test results.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] In a first aspect, the present invention provides a method for testing a ground loop in a high-voltage cable in operation, comprising:

[0009] Acquire a current response signal and a voltage response signal generated by the ground loop in response to an excitation signal;

[0010] Performing full vector spectrum analysis on the current response signal and the voltage response signal respectively to obtain a current response three-dimensional spectrum and a voltage response three-dimensional spectrum, and performing spectrum semantic feature extraction on the current response three-dimensional spectrum and the voltage response three-dimensional spectrum to obtain a current response three-dimensional spectrum semantic feature vector and a voltage response three-dimensional spectrum semantic feature vector;

[0011] Combined with the self-attention mechanism, the semantic feature vectors of the voltage response three-dimensional spectrum and the semantic feature vectors of the current response three-dimensional spectrum are subjected to feature interaction analysis based on fine-grained optimization based on external knowledge, thereby obtaining a fine-grained interaction feature representation between the semantics of the voltage-current response spectrum;

[0012] A single-phase test result is determined based on the fine-grained interaction feature representation between the semantics of the voltage-current response spectrum, and the test result is used to indicate whether there is an abnormality in the waveform.

[0013] In a second aspect, the present invention provides a ground loop testing system for a high-voltage cable in operation, comprising:

[0014] An acquisition module is configured to: acquire a current response signal and a voltage response signal generated by the ground loop in response to an excitation signal;

[0015] a feature extraction module configured to: perform full vector spectrum analysis on the current response signal and the voltage response signal respectively to obtain a three-dimensional current response spectrum and a three-dimensional voltage response spectrum, and perform spectrum semantic feature extraction on the three-dimensional current response spectrum and the three-dimensional voltage response spectrum to obtain a semantic feature vector of the three-dimensional current response spectrum and a semantic feature vector of the three-dimensional voltage response spectrum;

[0016] a feature interaction analysis module configured to: combine a self-attention mechanism to perform feature interaction analysis on the semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum, using fine-grained feature interaction analysis optimized based on external knowledge, to obtain a fine-grained interaction feature representation between the semantic features of the voltage-current response spectrum;

[0017] The prediction module is configured to determine a single-phase test result based on the fine-grained interaction feature representation between the semantics of the voltage-current response spectrum, wherein the test result is used to indicate whether there is an abnormality in the waveform.

[0018] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein the computer instructions, when executed by the processor, perform the method described in the first aspect.

[0019] In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, wherein when the computer instructions are executed by a processor, the method described in the first aspect is performed.

[0020] One or more of the above technical solutions have the following beneficial effects:

[0021] In the present invention, the current response signal and voltage response signal generated by the response excitation signal are analyzed using the full vector spectrum, which can more sensitively detect the subtle but critical changes accompanied by signal instability in the power system, and provide a basis for the waveform anomaly detection task; the voltage response three-dimensional spectrum semantic feature vector and the current response three-dimensional spectrum semantic feature vector are combined with the self-attention mechanism and optimized through external knowledge. While improving the semantic expression of the self-response three-dimensional spectrum semantic feature vector, it helps to capture the interaction between the spectrum semantics in the current and voltage response signals, and can perform subsequent single-phase resistance testing while ensuring the stability of the waveform, thereby helping to improve the accuracy and reliability of the ground loop resistance test results.

[0022] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0024] Figure 1 Flowchart of a ground loop testing method for a high-voltage cable in operation according to a first embodiment of the present application;

[0025] Figure 2 Schematic diagram of data flow of a ground loop testing method for a high-voltage cable in operation according to a first embodiment of the present application;

[0026] Figure 3 Flowchart of sub-step S5 of the ground loop testing method in the high-voltage cable operating state according to the first embodiment of the present application;

[0027] Figure 4 This is a block diagram of a ground loop testing system for a high-voltage cable in operation according to a second embodiment of the present application;

[0028] Figure 5A diagram showing a high-voltage cable grounding current detection according to an embodiment of the present application;

[0029] Figure 6 This is a diagram showing the diagnosis basis for live detection of the high-voltage cable grounding loop resistance according to an embodiment of the present application. DETAILED DESCRIPTION

[0030] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0031] It should be noted that the terms used herein are for describing particular embodiments only and are not intended to limit the exemplary embodiments according to the present invention.

[0032] In the absence of conflict, the embodiments of the present invention and the features thereof may be combined with each other.

[0033] Example 1

[0034] This embodiment discloses a ground loop testing method for a high-voltage cable in operation, including:

[0035] Acquire a current response signal and a voltage response signal generated by the ground loop in response to an excitation signal;

[0036] Performing full vector spectrum analysis on the current response signal and the voltage response signal respectively to obtain a current response three-dimensional spectrum and a voltage response three-dimensional spectrum, and performing spectrum semantic feature extraction on the current response three-dimensional spectrum and the voltage response three-dimensional spectrum to obtain a current response three-dimensional spectrum semantic feature vector and a voltage response three-dimensional spectrum semantic feature vector;

[0037] Combined with the self-attention mechanism, the semantic feature vectors of the voltage response three-dimensional spectrum and the semantic feature vectors of the current response three-dimensional spectrum are subjected to feature interaction analysis based on fine-grained optimization based on external knowledge, thereby obtaining a fine-grained interaction feature representation between the semantics of the voltage-current response spectrum;

[0038] A single-phase test result is determined based on the fine-grained interaction feature representation between the semantics of the voltage-current response spectrum, and the test result is used to indicate whether there is an abnormality in the waveform.

[0039] Before performing the ground loop testing method for a high-voltage cable in operation provided in this embodiment, the following preparations are required:

[0040] S1: Wire the test equipment based on the type of grounding box.

[0041] Technicians must identify the specific model and type of grounding box to be tested. Different types of grounding boxes may have different electrical characteristics, interface standards, or internal structures, which directly affect how to properly connect the test equipment. For example, some grounding boxes may use a single-phase grounding system, while others may use a three-phase or multi-point grounding system. Understanding this information is crucial for correctly setting up the test equipment.

[0042] S2: Make sure the ground wire of the test equipment is grounded.

[0043] Ensuring that the ground wire of the test equipment is properly and reliably grounded is an essential safety measure when performing ground loop resistance tests on high-voltage cables. This step not only ensures the safety of the tester but also ensures the accuracy of the test results.

[0044] In one example, before actual operation, the condition of the existing grounding system needs to be inspected to ensure it is in good condition. This involves a comprehensive assessment of the condition of the grounding electrodes (e.g., copper rods, plates), grounding conductors, and connection points to ensure there is no corrosion, damage, or looseness. Selecting the right grounding material is also key, and low-resistivity copper wire or specialized grounding cable is generally recommended due to its excellent electrical conductivity.

[0045] After completing the preliminary preparations, the next step is to connect the test equipment's grounding cable to the selected grounding system. First, identify the grounding terminal or connector on the test equipment, which is usually labeled "GND" or "Earth." If any oxide or dirt is found on the terminal, clean it with appropriate tools to ensure good contact. Then, secure one end of the treated grounding cable to the test equipment's grounding terminal, securing it securely by screwing, welding, or crimping. The other end should be connected to a proven grounding system, such as the building's foundation steel or a dedicated ground grid. This connection must be strong and durable to ensure a good ground connection during the test. After the connection is complete, verify the effectiveness of the entire ground path using specialized measuring instruments, such as a ground resistance tester. Ideally, ground resistance should not exceed 0.5 ohms to 1 ohm, depending on the standard. If the measured resistance is too high, investigate the cause and implement corrective measures until satisfactory results are achieved. For example, the resistance can be reduced by adding additional ground electrodes or improving soil moisture. Multi-point grounding—installing ground electrodes at more than one location—is also an effective method to improve the efficiency of the grounding system.

[0046] S3: Output terminal wiring of the grounding box.

[0047] Correct wiring of the grounding box output terminal is one of the key steps to ensure that the ground loop resistance test can be carried out safely and accurately when the high-voltage cable is in operation.

[0048] S4: Power on the test equipment and turn on the power switch of the test equipment.

[0049] Before powering up the test equipment, be sure to confirm that the power conditions provided on site match the equipment's requirements. This includes checking that the power outlet meets the equipment's required voltage level (e.g., 110V / 220V), frequency (50Hz / 60Hz), and plug type. If necessary, use a tool such as a multimeter to measure the actual supply voltage to ensure that it fluctuates within the allowable range. For some high-precision test equipment, a stable uninterruptible power supply (UPS) may also be required to protect the equipment from transient power grid conditions. In addition, check the quality and length of the power cables to ensure that they are intact and long enough for easy routing. Verifying power conditions is a prerequisite for ensuring the normal operation of the test equipment and is also an important step in ensuring operator safety.

[0050] After confirming the appropriate power supply conditions, the next step is to connect the test equipment's power cord to a power outlet. Depending on the equipment's design, the power cord may plug directly into a standard wall outlet or connect to a dedicated distribution box via an industrial-grade connector. Regardless, ensure that the plug is fully inserted into the outlet and that all contacts are securely connected. For power cords protected by a fuse or circuit breaker, check that the specifications meet the equipment requirements and confirm that the fuse is not blown or the circuit breaker is in the closed position. Proper power cord connection not only ensures a stable power supply but also provides a reliable foundation for subsequent operations. Once all preparations are complete, you can begin powering on the test equipment.

[0051] First, ensure that all external interfaces and connections on the test equipment are properly connected as specified. Then, slowly plug the power cord into a power outlet and observe any unusual indicators or display on the equipment. If everything appears normal, press the power switch to start the test equipment. During this process, remain alert and listen for any sounds, smells, or visual signals emitted by the equipment. Any unusual behavior may indicate a malfunction; immediately stop operation and investigate the issue. After powering on the equipment, immediately check its operating status. Most modern test equipment automatically performs a self-test upon startup, displaying current status information or prompting the user for further setup. At this point, the technician can adjust various equipment parameters, such as measurement mode and range selection, according to the on-screen prompts or instructions in the user manual. Ensure that all settings meet the requirements of the test plan and record the initial status information for later reference. Checking the equipment status is not only necessary to confirm that the equipment is functioning properly but also ensures the accuracy and reliability of the test results.

[0052] S5: Set equipment parameters and start the single-phase test program to obtain single-phase test results.

[0053] It should be understood that the process of starting the single-phase test program to conduct single-phase testing is crucial. The main step is to effectively detect the situation of unstable waveform, that is, non-oscillating waveform, to ensure that the subsequent ground loop resistance test results are more accurate and reliable.

[0054] like Figure 1-Figure 3 As shown, the ground loop testing method for a high-voltage cable in operation provided by this embodiment includes the following steps:

[0055] S51: Send an excitation signal using the test equipment.

[0056] Specifically, the test equipment sends an excitation signal. The excitation signal is an electrical signal of a specific frequency or pattern generated by the test equipment to stimulate a response in the system under test. For ground loop resistance testing, the excitation signal typically consists of both current and voltage. These signals propagate through the cable, generating corresponding current and voltage responses in the grounding system. By analyzing these response signals, the performance of the grounding system can be assessed and potential problems detected.

[0057] S52: Receive the returned current response signal and voltage response signal.

[0058] Specifically, the returned current and voltage response signals are received. The primary purpose of receiving these signals after transmitting the stimulus signal is to obtain the true response of the system under test. These response signals reflect the behavioral characteristics of the high-voltage cable and its grounding system under specific stimulus conditions.

[0059] S53: performing full vector spectrum analysis on the current response signal and the voltage response signal respectively to obtain a current response three-dimensional spectrum and a voltage response three-dimensional spectrum;

[0060] Specifically, full vector spectrum analysis is performed on the current response signal and the voltage response signal to obtain a current response three-dimensional spectrum graph and a voltage response three-dimensional spectrum graph.

[0061] The implicit correlation between the current response signal and the voltage response signal can be further explored by performing spectrum analysis on the two. Traditional spectrum analysis technology usually only focuses on the frequency or amplitude of the signal, but ignores the phase information and the correlation between them. This may lead to insufficient understanding and judgment of certain waveform instability situations.

[0062] Therefore, in this embodiment, full vector spectrum analysis is further performed on the current response signal and the voltage response signal to obtain a three-dimensional current response spectrum and a three-dimensional voltage response spectrum. In particular, it is worth mentioning that full vector spectrum analysis is an improved spectrum analysis technology. In full vector spectrum analysis, the two sensors or signal channels are usually orthogonal to each other, so that a complete vector description can be constructed. As two channels, current and voltage can reflect the dynamic system and have certain complementary information. It has significant advantages in processing current response signals and voltage response signals. Unlike traditional spectrum analysis, full vector spectrum not only considers the frequency and amplitude (amplitude) of the signal, but also retains phase information, which is particularly important for processing complex response signals in power systems. In addition, since signal instability in power systems is often accompanied by subtle but critical changes, such as phase offset or amplitude fluctuations at specific frequencies, full vector spectrum analysis can more sensitively detect these changes, providing a basis for waveform anomaly detection tasks. The three-dimensional spectrum obtained by full vector spectrum analysis of the current response signal and the voltage response signal can simultaneously display the relationship between frequency, amplitude, and phase on the same chart, providing a more comprehensive perspective to observe the response signal characteristics. This makes it easier to understand the overall behavior of the signal and helps to intuitively observe which frequencies have obvious phase differences or which frequency bands have significant amplitude changes. This is crucial for identifying non-oscillatory waveforms or expected oscillation characteristics.

[0063] S54: performing spectrum semantic feature extraction on the current response three-dimensional spectrum and the voltage response three-dimensional spectrum respectively to obtain a current response three-dimensional spectrum semantic feature vector and a voltage response three-dimensional spectrum semantic feature vector.

[0064] Because current and voltage signals in power systems are typically non-stationary and time-varying, the three-dimensional spectrograms of current and voltage responses generated by spectral analysis contain a wealth of information about the relationships between signal frequency, amplitude, and phase. Traditional feature extraction methods are unable to fully capture this information.

[0065] Therefore, in order to automatically extract the most representative and discriminative features from complex signal data, in this embodiment, the current response 3D spectrogram and the voltage response 3D spectrogram are further passed through a spectral feature extractor based on a multi-scale convolutional LSTM to obtain a semantic feature vector for the current response 3D spectrogram and a semantic feature vector for the voltage response 3D spectrogram. The multi-scale convolutional LSTM is a deep learning model that combines the advantages of convolutional neural networks (CNNs) and long short-term memory networks (LSTMs) and is primarily used to process data with spatiotemporal characteristics.

[0066] The way to combine convolutional neural network and LSTM for feature extraction:

[0067] The multi-scale convolutional LSTM model usually consists of three main parts: the multi-scale convolutional feature extraction module, namely the CNN part: it uses multiple convolution operations of different scales, namely different convolution kernel sizes, to extract local spatial features from the input data; the time series modeling module, namely the convolutional LSTM part: it applies convolutional LSTM units to the extracted multi-scale feature sequence to capture the dynamic changes and dependencies of the data in the time dimension; the output mapping module: it maps the spatiotemporal features to the output of the final task, such as predicting the image, classification label or regression value at the next moment.

[0068] The detailed process of the multi-scale convolutional LSTM model from input to output:

[0069] (1) Input layer: The input data size can be expressed as T×H×W×C, where T represents the number of time steps, H×W represents the spatial size, and C represents the number of channels.

[0070] (2) Multi-scale convolutional feature extraction, also known as the CNN module: To capture local features of different sizes, multiple parallel convolution operations are performed at the same time step, with each convolution branch using a different convolution kernel size. Each branch performs convolution, activation (e.g., ReLU) and pooling operations on the input data to generate feature maps at different scales. These feature maps reflect the local patterns of the input data at different receptive fields and can capture both fine-grained and global information.

[0071] Feature fusion: Convolutional features of different scales can be fused through splicing or weighted summation. The fused feature tensor maintains the spatial structure while containing multi-scale information in the channel dimension.

[0072] (3) Temporal modeling module: The fused multi-scale feature map sequence is fed into the convolutional LSTM layer.

[0073] The difference between convolutional LSTM and traditional LSTM is that it uses convolution operations in the calculation of each gate, namely the input gate, forget gate, and output gate, thereby preserving spatial information.

[0074] In the formula, assuming the input feature is X t , the internal state of the convolutional LSTM is H t-1 and C t-1 , then the calculation process:

[0075] i t =σ(W xi *X t +W hi *H t-1 +b i )

[0076] f t =σ(W xf *X t +W hf *H t-1 +b f )

[0077] o t =σ(W xo *X t +W ho *H t-1 +b o )

[0078]

[0079] H t =o t ⊙tanh(C t )

[0080] Among them, “*” represents the convolution operation, which ensures that the spatial information is not lost in the temporal propagation. xi 、W xf 、W xo 、W xc 、W hi 、W hf 、W ho is the convolution kernel matrix, b i 、b f 、b o 、b v is the bias term, σ is the activation function, and tanh is the hyperbolic tangent activation function.

[0081] To further improve the model's expressiveness, multiple convolutional LSTM layers can be stacked, with each layer receiving the output of the previous layer to form a higher level of spatiotemporal feature abstraction. This multi-layered structure helps capture multi-level features, from local details to global dynamics.

[0082] (4) Output mapping module: On the output of the last convolutional LSTM layer, high-dimensional spatiotemporal features can be mapped to the output of a specific task through additional convolutional layers, fully connected layers, or deconvolution layers.

[0083] It should be understood that the multi-scale convolutional LSTM can effectively model complex patterns in the time-frequency domain by combining the advantages of convolutional neural networks in local feature extraction with the strengths of LSTM in processing sequential data. Therefore, by using a spectral feature extractor based on a multi-scale convolutional LSTM, not only can local spatial structural features at different scales be extracted from the three-dimensional current and voltage response spectra, including but not limited to frequency components, amplitude changes, and phase differences, but also the long-term dependencies between these local features, that is, the dynamic changes over time, can be captured. By processing the three-dimensional current and voltage response spectra through a spectral feature extractor based on a multi-scale convolutional LSTM, it is possible to understand the multi-scale local spatial structure and temporal dynamic changes in the signal frequency domain. This is crucial for understanding the time series characteristics of signals, especially when detecting waveform stability, and can help identify unstable oscillations or other abnormal behavior.

[0084] S55: Perform fine-grained feature interaction analysis based on external knowledge optimization on the semantic feature vectors of the voltage response three-dimensional spectrum and the current response three-dimensional spectrum to obtain a fine-grained interaction feature representation between the semantics of the voltage-current response spectrum.

[0085] It should be understandable that since the voltage response three-dimensional spectrum semantic feature vector and the current response three-dimensional spectrum semantic feature vector represent the compact and expressive three-dimensional spectrum semantic feature representation extracted from the voltage response signal and the current response signal respectively after multi-scale convolutional LSTM processing, these two vectors contain spatiotemporal feature information and correlation relationships such as frequency, amplitude and phase of each signal.

[0086] Therefore, in order to deeply understand and extract the correlation and interaction information between the semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum, so as to provide a basis for the waveform anomaly detection task, in this embodiment, the semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum are further subjected to fine-grained feature interaction analysis based on external knowledge optimization to obtain a fine-grained interaction feature representation between the semantics of the voltage-current response spectrum.

[0087] Specifically, the specific process of performing fine-grained feature interaction analysis on the semantic feature vectors of the voltage response three-dimensional spectrum and the current response three-dimensional spectrum based on external knowledge optimization includes:

[0088] S551: Perform external knowledge optimization interaction analysis on the semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum to obtain a fine-grained feature interaction matrix between the semantics of the external knowledge optimized voltage-current response spectrum.

[0089] Specifically, first, the semantic feature vectors of the voltage response three-dimensional spectrum and the current response three-dimensional spectrum are input into the fine-grained feature interaction network to deeply explore the spectral semantic micro-level correlation between the voltage and current features, and construct a fine-grained feature interaction matrix between the spectral semantics to obtain the fine-grained feature interaction matrix between the voltage-current response spectrum semantics.

[0090] In this way, the model can understand the complex relationship between voltage and current responses, which is crucial for subsequent information fusion. For example, in power systems, current changes at certain frequencies may be closely related to corresponding voltage changes. This correlation is important for identifying non-oscillatory waveforms or expected oscillatory characteristics.

[0091] Then, the fine-grained feature interaction matrix between the semantics of the voltage-current response spectrum is input into the attention unit based on external knowledge to obtain the external knowledge optimized fine-grained feature interaction matrix between the semantics of the voltage-current response spectrum.

[0092] External knowledge can come from multiple sources, such as domain experts' understanding of the semantics of the power system's voltage and current response spectra and accumulated experience from historical data. This knowledge is effectively incorporated into the model through an attention mechanism, helping to further optimize the fine-grained feature interaction matrix between the semantics of the voltage and current response spectra. For example, if historical data shows that current fluctuations within a certain frequency range are often accompanied by specific types of voltage changes, this correlation can be used as external knowledge to guide the current analysis process. This not only enhances the model's ability to understand specific domain knowledge but also enables it to capture the potential connections between the semantic features of the three-dimensional current response spectra and the semantic features of the three-dimensional voltage response spectra on a larger scale.

[0093] Specifically, the fine-grained feature interaction network processes the semantic feature vectors of the voltage response 3D spectrum and the current response 3D spectrum to obtain a fine-grained feature interaction matrix between the spectrum semantics:

[0094]

[0095] Among them, V1 and V2 are the semantic feature vectors of the voltage response three-dimensional spectrum and the current response three-dimensional spectrum, respectively. p is the fine-grained feature interaction matrix between the semantics of the voltage-current response spectrum.

[0096] The fine-grained feature interaction matrix between the semantics of the voltage-current response spectrum is input into the attention unit based on external knowledge to obtain the fine-grained feature interaction matrix between the semantics of the voltage-current response spectrum optimized by external knowledge, specifically:

[0097]

[0098] Among them, M k and M v represents the first external knowledge attention learnable memory parameter matrix and the second external knowledge attention learnable memory parameter matrix, norm(·) represents the normalization function, M y Optimizing the fine-grained feature interaction matrix between semantics of voltage-current response spectra for external knowledge.

[0099] As an implementation method, the following external knowledge optimization interaction analysis formula is used to perform external knowledge optimization interaction analysis on the semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum to obtain the external knowledge optimized voltage-current response spectrum semantic fine-grained feature interaction matrix.

[0100] M k and M v They represent the first external knowledge attention learnable memory parameter matrix and the second external knowledge attention learnable memory parameter matrix respectively. The key matrix maps the current response features to the external knowledge space. The value matrix stores the interaction weights after knowledge enhancement.

[0101] Initialization formula:

[0102] in, Represents Gaussian distribution. The initial parameters are generated by having a mean of 0 and a variance adjusted by the input and output dimensions. k controls the expressiveness of the knowledge space, and m is the original dimension of the input features. The two together determine the complexity and flexibility of the knowledge mapping.

[0103] M is calculated by the following formula k and M v Incorporating the interaction matrix M y :

[0104]

[0105] in: Raw voltage-current interaction matrix. Softmax(·): Row normalization function, generating attention weights. ⊙: Element-wise multiplication.

[0106] Optimize M through backpropagation k and M v , the specific steps are as follows:

[0107] (1) Loss function definition.

[0108] The task is voltage-current association classification, and the loss function is cross entropy loss:

[0109]

[0110] Among them, y i is the one-hot encoding of the true category label, It is the category probability calculated by the model based on the optimized features, and the goal is to optimize M through gradient descent k and M v ,make Close to y i .

[0111] (2) Gradient calculation.

[0112] Calculate M by the chain rule k and M v Gradient:

[0113]

[0114] (3) Parameter update.

[0115] Update the parameters using gradient descent:

[0116]

[0117] Among them, η is the learning rate and t is the number of training steps.

[0118] Then, the fine-grained feature interaction matrix M between the semantics of the voltage-current response spectrum is optimized based on external knowledge. y , respectively, fine-grained modulation is performed on the semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum to obtain the optimized semantic feature vector of the voltage response three-dimensional spectrum and the optimized semantic feature vector of the current response three-dimensional spectrum.

[0119] The specific steps include: performing a linear transformation on the semantic feature vector of the voltage response three-dimensional spectrum to obtain a joint semantic query feature vector of the voltage response three-dimensional spectrum and a joint semantic value feature vector of the voltage response three-dimensional spectrum; using external knowledge to optimize the fine-grained feature interaction matrix between the semantics of the voltage-current response spectrum as a key matrix, and inputting the joint semantic query feature vector of the voltage response three-dimensional spectrum, the joint semantic value feature vector of the voltage response three-dimensional spectrum and the key matrix into a fine-grained modulation module, and the fine-grained modulation module adopts Transformer to obtain the optimized semantic feature vector of the voltage response three-dimensional spectrum; performing a linear transformation on the semantic feature vector of the current response three-dimensional spectrum to obtain a semantic query feature vector of the current response three-dimensional spectrum and a semantic value feature vector of the current response three-dimensional spectrum; using external knowledge to optimize the fine-grained feature interaction matrix between the semantics of the voltage-current response spectrum as a key matrix, and inputting the semantic query feature vector of the current response three-dimensional spectrum, the semantic value feature vector of the current response three-dimensional spectrum and the key matrix into a fine-grained modulation module based on the Transformer structure to obtain the optimized semantic feature vector of the current response three-dimensional spectrum.

[0120] That is, the semantic feature vector of the voltage response three-dimensional spectrum graph and the semantic feature vector of the current response three-dimensional spectrum graph are linearly transformed to generate query, key and value feature vectors; this embodiment uses the fine-grained feature interaction matrix optimized in the previous step as the key matrix Key to replace the linear transformation to obtain the key Key.

[0121] The external knowledge memory matrix of this embodiment plays an auxiliary role. y Domain prior information is introduced into the attention calculation process to guide the modulation of query and value vectors.

[0122] First, the semantic feature vectors of the three-dimensional spectrograms of voltage response and current response are linearly transformed to generate their respective query and value vectors. At the same time, external knowledge can be used to learn and memorize the parameter matrix M through external knowledge attention. k and M v The optimized fine-grained feature interaction matrix M y as a key matrix.

[0123] Specifically, the self-attention module within the Transformer structure calculates attention weights based on the query and key, and then performs a weighted summation of the values to generate an optimized feature vector. For the voltage response and current response paths, after passing through their respective Transformers, the output is the optimized semantic feature vector of the three-dimensional spectrogram of the voltage response and the three-dimensional spectrogram of the current response, respectively.

[0124] In this process, the self-attention mechanism enables information exchange and integration between the internal and external knowledge of the semantic features of the current response 3D spectrogram and the semantic features of the voltage response 3D spectrogram. This means that each semantic feature vector of the voltage response 3D spectrogram and the semantic feature vector of the current response 3D spectrogram can benefit from external knowledge and other features to improve the quality of its own semantic feature expression of the response 3D spectrogram.

[0125] However, the optimized semantic feature vectors of the voltage and current response 3D spectrograms are relatively well aligned in the high-dimensional feature space. This is because they both undergo the same external knowledge-based modulation process and share the same optimized fine-grained feature interaction matrix as the key matrix space. This alignment helps more accurately capture the interactions between the spectral semantics in the current and voltage response signals, providing a solid foundation for subsequent comprehensive analysis.

[0126] As an implementation method, the fine-grained feature interaction matrix between the semantics of the voltage-current response spectrum is optimized based on external knowledge, and the semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum are fine-grainedly modulated using the following fine-grained modulation formula to obtain the optimized semantic feature vector of the voltage response three-dimensional spectrum and the optimized semantic feature vector of the current response three-dimensional spectrum; wherein the fine-grained modulation formula is:

[0127]

[0128] Among them, W 1q and b 1q They represent the voltage response 3D spectrum graph joint semantic query weight matrix and the voltage response 3D spectrum graph joint semantic query bias vector, W 1v and b 1v They represent the voltage response three-dimensional spectrum joint semantic value weight matrix and the voltage response three-dimensional spectrum joint semantic value bias vector, is the matrix multiplication, V 1q and V 1v are the voltage response three-dimensional spectrum joint semantic query feature vector and the voltage response three-dimensional spectrum joint semantic value feature vector, (·) T is the matrix transpose, d is M y The scale is the width of the matrix multiplied by the height of the matrix, softmax(·) represents the softmax function, V1 ′ is the optimized semantic feature vector of the voltage response three-dimensional spectrum, W 2q and b 2q They represent the current response 3D spectrum semantic query weight matrix and the current response 3D spectrum semantic query bias vector, W 2v and b 2vThey represent the current response three-dimensional spectrum semantic value weight matrix and the current response three-dimensional spectrum semantic value bias vector, V 2q and V 2v are the semantic query feature vector of the current response three-dimensional spectrum and the semantic value feature vector of the current response three-dimensional spectrum, respectively, and V′2 is the optimized semantic feature vector of the current response three-dimensional spectrum.

[0129] Finally, semantic association encoding is performed on the optimized semantic feature vector of the voltage response three-dimensional spectrum and the optimized semantic feature vector of the current response three-dimensional spectrum to obtain the fine-grained interaction feature vector between the semantics of the voltage-current response spectrum as the fine-grained interaction feature representation between the semantics of the voltage-current response spectrum.

[0130] In this embodiment, a point-by-point division is calculated between the optimized semantic feature vectors of the voltage response 3D spectrum and the optimized semantic feature vectors of the current response 3D spectrum to obtain a fine-grained semantic interaction feature vector between the voltage and current response spectra. This fine-grained semantic interaction feature vector integrates information from the original voltage and current feature vectors, better reflecting the interaction between the two, particularly key features that indicate waveform stability.

[0131] More specifically, the semantic association encoding is performed on the optimized semantic feature vector of the voltage response three-dimensional spectrum and the optimized semantic feature vector of the current response three-dimensional spectrum using the following semantic association encoding formula to obtain a fine-grained interaction feature vector between the semantics of the voltage-current response spectrum; wherein the semantic association encoding formula is:

[0132]

[0133] Among them, V s is the fine-grained interaction feature vector between the semantics of the voltage-current response spectrum.

[0134] Based on this, the fine-grained interactive feature analysis process optimized based on external knowledge not only considers the inherent semantic features of the three-dimensional spectrograms of the voltage and current response signals, but also fully incorporates the guidance provided by external knowledge, thereby improving the accuracy of determining whether the waveform is non-oscillatory or meets the expected oscillation characteristics. This allows for more reliable resistance calculation after confirming waveform stability, ensuring the scientific and reliable nature of the entire testing process.

[0135] S56: Based on the fine-grained interactive feature representation between the semantics of the voltage-current response spectrum, determine the single-phase test result, which is used to indicate whether there is an abnormality in the waveform.

[0136] In this embodiment, the fine-grained interaction feature vectors between the voltage and current response spectrum semantics are passed through a classifier-based single-phase test result generator to generate single-phase test results. In other words, the fine-grained interaction feature information between the voltage and current response spectrum semantics is used for classification processing, thereby detecting waveform anomalies. This allows subsequent single-phase resistance testing to be performed while ensuring waveform stability, thereby helping to improve the accuracy and reliability of ground loop resistance test results.

[0137] Preferably, the step of passing the fine-grained interaction feature vector between the semantics of the voltage-current response spectrum through a classifier-based single-phase test result generator to obtain a single-phase test result comprises:

[0138] Calculate the distance between each pair of eigenvalues of the fine-grained interaction feature vector between the semantics of the voltage-current response spectrum, such as the L2 distance, and take the square root of the calculated distance to obtain the fine-grained interaction distance representation matrix between the semantics of the voltage-current response spectrum, that is:

[0139]

[0140] Among them, v i and v j The eigenvalue of the fine-grained interaction feature vector between the semantics of the voltage-current response spectrum, d(v i ,v j ) represents the distance between each pair of eigenvalues of the fine-grained interaction feature vector between the semantics of the voltage-current response spectrum, V represents the fine-grained interaction feature vector between the semantics of the voltage-current response spectrum, and D i,j Represents the fine-grained interaction distance representation matrix between semantics of voltage-current response spectra.

[0141] The voltage-current response spectrum semantic fine-grained interaction autocorrelation matrix of the voltage-current response spectrum semantic fine-grained interaction feature vector as the row vector is obtained, that is,

[0142] The voltage-current response spectrum semantics fine-grained interaction feature vector is matrix-multiplied with the voltage-current response spectrum semantics fine-grained interaction distance representation matrix to obtain the voltage-current response spectrum semantics fine-grained interaction first-level mapping vector, that is,

[0143] The fine-grained interaction multi-level mapping vector between the semantics of voltage-current response spectrum is obtained by matrix multiplication of the first-level mapping vector of fine-grained interaction between the semantics of voltage-current response spectrum and the matrix product of the fine-grained interaction distance representation matrix between the semantics of voltage-current response spectrum and the fine-grained interaction autocorrelation matrix between the semantics of voltage-current response spectrum.

[0144] The multi-level mapping vector of the fine-grained interaction between the semantics of the voltage-current response spectrum is interpolated with the fine-grained interaction correlation eigenvector between the semantics of the voltage-current response spectrum composed of the eigenvalues of the fine-grained interaction autocorrelation matrix between the semantics of the voltage-current response spectrum to obtain the optimized fine-grained interaction feature vector between the semantics of the voltage-current response spectrum, where interpolation or zero padding is performed when the eigenvalue is insufficient.

[0145] The optimized fine-grained interaction feature vector between the semantics of the voltage-current response spectrum is passed through a classifier-based single-phase test result generator to obtain the single-phase test result.

[0146] Considering that the semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum represent the image semantic features of the three-dimensional spectra of the current response signal and the voltage response signal respectively, when performing fine-grained feature interaction based on external knowledge modulation, the insufficient correspondence of external knowledge modulation under the image semantic feature differences caused by the semantic differences of the image sources will lead to the lack of judgment of the interaction feature instance of the fine-grained interaction feature vector between the semantics of the voltage-current response spectrum, thereby affecting the accuracy of the single-phase test results obtained by the classifier-based single-phase test result generator.

[0147] Therefore, a linear target mapping representation based on the similarity distance representation matrix of the fine-grained interaction feature vectors between the semantics of the voltage-current response spectrum is used to instantiate the complete similarity of the self-correlation of the fine-grained interaction feature vectors between the semantics of the voltage-current response spectrum, and a secondary target mapping representation based on the multi-level distribution hierarchy is performed. The negative influence factor of the correlation mismatch is compensated by the association fusion kernel bias to improve the degree of eigenvalue instance determination of the fine-grained interaction feature vectors between the semantics of the voltage-current response spectrum under similarity constraints, that is, the degree of significance of the eigenvalue as an instance for classification and regression judgment, and to improve the accuracy of the single-phase test results obtained by the fine-grained interaction feature vectors between the semantics of the voltage-current response spectrum through the classifier-based single-phase test result generator.

[0148] After completing all single-phase tests, the test equipment displays the test results.

[0149] In particular, after obtaining a single-phase test result indicating whether there is an abnormality in the waveform, in response to the single-phase test result indicating that there is no abnormality in the waveform, it is confirmed that the measured waveform is a non-oscillating waveform or meets the expected oscillation characteristics, and the test conditions are considered good, and the resistance value calculation can be continued. At this time, the single-phase resistance value test can be performed by the following steps:

[0150] It's worth noting that the primary purpose of a single-phase resistance test is to evaluate the resistance between each individual phase (e.g., A, B, and C) and ground. This test helps identify differences in electrical characteristics between phases and detect potential problems. For a three-phase system, ensuring that each phase has appropriate ground resistance is crucial, as it directly impacts the safety and reliability of the entire system. Specifically, an appropriate excitation signal is first sent to each phase in sequence. Depending on specific requirements, a DC or AC signal can be selected, and parameters such as the signal amplitude and frequency can be set. After the excitation signal is successfully sent, the returned current and voltage response signals are received. Next, using Ohm's law (R = V / IR = V / I) (where R is resistance, V is voltage drop, and I is current), the resistance value of each phase is calculated. The voltage drop and current flowing through each phase are accurately recorded and then substituted into the formula to obtain the corresponding resistance value. This process is repeated three times, one for each phase A, B, and C. For a more accurate overall evaluation, the three single-phase resistance values can be averaged. Calculate the average of the three phase resistance values as the final single-phase resistance representative value. This method can effectively reduce the impact of random errors and provide a more stable and reliable reference standard.

[0151] The aforementioned ground loop testing method for operating high-voltage cables utilizes advanced signal processing techniques and machine learning algorithms to accurately measure ground loop resistance without disrupting the cable's normal power supply. Specifically, by injecting an excitation signal of a specific frequency into the cable and analyzing the resulting current and voltage responses, a deep learning model is employed to interpret the complex patterns and nonlinear relationships within these responses, enabling precise identification of waveform characteristics and timely warning of abnormalities.

[0152] When performing a loop resistance test, perform the following steps:

[0153] Step 1: Connect the equipment. If you are measuring a protective grounding box, you do not need to connect the current line, only the voltage output line is needed.

[0154] Step 2: The black grounding wire of the host is the working protection ground of the equipment and must be reliably grounded to the underground / tower grounding flat iron.

[0155] Step 3: Connect the output terminals in the grounding box. The requirement is to first connect the terminal block on the test host, then the grounding clamp, and finally the test clamp in the grounding box.

[0156] Step 4: Test the host power connection.

[0157] Depending on the environmental conditions, AC220V mobile power supply is generally used on site, and the power supply power is not less than 300W.

[0158] (2) After checking that all wiring is complete, connect the power cord, start the power supply, and turn on the power switch of the test host.

[0159] Step 5: Open the test tablet and connect to the host's LAN (the network starting with GS JIEDIDIANZU).

[0160] Step 6: Double-click "ControlJieDi-Shortcut" on the remote control tablet to open the software.

[0161] Step 7: Detect the internal wiring mode of the grounding box as needed and select a suitable program mode. This embodiment takes the cross-connection mode as an example.

[0162] Step 8: Parameter setting, click "Parameter setting" on the software interface, set the parameters, and click "OK" after setting.

[0163] Step 9: Once the settings are complete, click "Connect Measuring Instrument" and then "Start Measuring." You can check the connection status in the image below.

[0164] Step 10: Click "Start Measurement" to start the measurement program. After the instrument begins measuring, it will automatically charge and discharge according to the program. The host can clearly hear the internal switch activation sound. At this point, just wait. The first phase of the test will be completed in approximately 30-60 seconds.

[0165] Step 11: After the single-phase test is completed.

[0166] Step 12: After all three phases are tested, the device will automatically display the test results.

[0167] Step 13: After the test is completed, compare the test data with the judgment standard. The current reference standard for loop resistance is: Q / GDW 11223-2022 "Technical Specifications for High-voltage Cable Line Status Detection".

[0168] Step 14: After the test data is normal, you can restore it. The restoration and dismantling must be done in order.

[0169] First disconnect the tablet program connection;

[0170] Disconnect the device from the power source;

[0171] Remove the current / voltage clamp inside the grounding box;

[0172] Then remove the connection cable of the test host;

[0173] Restore the grounding box.

[0174] (6) It is strictly forbidden to remove the main unit's connection before removing the current clamp / voltage clamp in the grounding box.

[0175] It is worth mentioning that this method particularly emphasizes the importance of confirming whether the waveform meets the expected characteristics in the waveform inspection step, because this is directly related to the accuracy of the subsequent resistance value calculation. Only when the waveform is verified to be non-oscillatory or meets the expected oscillation characteristics will the next step of the resistance value calculation process be continued. The entire test process includes a series of steps from equipment wiring to parameter setting, to the start of the single-phase test program and waveform abnormality warning. Finally, the test results are displayed by the test equipment, realizing the effective monitoring and testing of the high-voltage cable grounding loop resistance. This is of great significance for evaluating the connection quality between the cable and the ground, detecting potential fault points, and predicting possible faults.

[0176] Example 2

[0177] The purpose of this embodiment is to provide a ground loop testing system for a high-voltage cable in operation, including:

[0178] An acquisition module is configured to: acquire a current response signal and a voltage response signal generated by the ground loop in response to an excitation signal;

[0179] a feature extraction module configured to: perform full vector spectrum analysis on the current response signal and the voltage response signal respectively to obtain a three-dimensional current response spectrum and a three-dimensional voltage response spectrum, and perform spectrum semantic feature extraction on the three-dimensional current response spectrum and the three-dimensional voltage response spectrum to obtain a semantic feature vector of the three-dimensional current response spectrum and a semantic feature vector of the three-dimensional voltage response spectrum;

[0180] a feature interaction analysis module configured to: combine a self-attention mechanism to perform feature interaction analysis on the semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum, using fine-grained feature interaction analysis optimized based on external knowledge, to obtain a fine-grained interaction feature representation between the semantic features of the voltage-current response spectrum;

[0181] The prediction module is configured to determine a single-phase test result based on the fine-grained interaction feature representation between the semantics of the voltage-current response spectrum, wherein the test result is used to indicate whether there is an abnormality in the waveform.

[0182] Figure 4 FIG is a block diagram of a ground loop test system for a high voltage cable in operation according to an embodiment of the present application. Figure 4As shown, according to the ground loop test system 300 of the high-voltage cable in the operating state of the embodiment of the present application, it includes: a signal sending module 310, which is used to send an excitation signal using the test equipment; a signal receiving module 320, which is used to receive the returned current response signal and voltage response signal; a full vector spectrum analysis module 330, which is used to perform full vector spectrum analysis on the current response signal and the voltage response signal to obtain a current response three-dimensional spectrum and a voltage response three-dimensional spectrum; a spectrum semantic feature extraction module 340, which is used to perform spectrum semantic feature extraction on the current response three-dimensional spectrum and the voltage response three-dimensional spectrum. Feature extraction is used to obtain the semantic feature vector of the current response three-dimensional spectrum and the semantic feature vector of the voltage response three-dimensional spectrum; a fine-grained feature interaction analysis module 350 is used to perform fine-grained feature interaction analysis on the semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum based on external knowledge optimization to obtain a fine-grained interaction feature representation between the semantics of the voltage-current response spectrum; a single-phase test result determination module 360 is used to determine the single-phase test result based on the fine-grained interaction feature representation between the semantics of the voltage-current response spectrum, and the single-phase test result is used to indicate whether there is an abnormality in the waveform.

[0183] As described above, the ground loop test system 300 for a high-voltage cable in operation according to an embodiment of the present application can be implemented in various wireless terminals, such as a server having a ground loop resistance test algorithm for a high-voltage cable in operation. In one possible implementation, the ground loop test system 300 for a high-voltage cable in operation according to an embodiment of the present application can be integrated into a wireless terminal as a software module and / or a hardware module. For example, the ground loop test system 300 for a high-voltage cable in operation can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the ground loop test system 300 for a high-voltage cable in operation can also be one of the many hardware modules of the wireless terminal.

[0184] Alternatively, in another example, the ground loop test system 300 when the high-voltage cable is in operation and the wireless terminal may also be separate devices, and the ground loop test system 300 when the high-voltage cable is in operation may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0185] In one example, when testing the ground loop resistance of a high-voltage cable in operation, the operator first carries a test rod and insulating gloves. The operator then tests the grounding box's outer shell and observes the outer box. The operator then records the parameters on the cable nameplate. The operator then records the on-site situation. The recorded situation includes information such as the grounding box, model, manufacturer, and date of manufacture. The operator then records relevant information based on the cable phase sequence tag, such as Figure 5 And use the flexible current clamp meter to measure the cable running current, as shown in Figure 6 Next, open the grounding box and observe the internal wiring and conditions. Take photos and record any signs of rust, moisture, or water damage. Based on the cable phase sequence, locate the corresponding coaxial cable in the grounding box and record the phase sequence and connection method of the copper busbar inside the grounding box. Then, use a FLUKE 319 clamp-on ammeter to locate the corresponding copper busbar inside the grounding box and perform current measurements according to the previously recorded phase sequence, taking photos and archiving the results.

[0186] When testing interlayer voltage, use the voltage range function of the Fluke 319 and connect a multimeter probe to measure the interlayer voltage within the grounding box. Specifically, test point 1: A core / B core / C core - voltage to ground; test point 2: A sheath / B sheath / C sheath - voltage to ground; test point 3: AC / BA / CB - voltage to ground; test point 4: A / B / C protectors - voltage to ground.

[0187] In further embodiments, there is also provided:

[0188] An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed by the processor. When the computer instructions are executed by the processor, the method described in Example 1 is performed. For the sake of brevity, no further details are given here.

[0189] It should be understood that in this embodiment, the processor may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), off-the-shelf field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0190] The memory may include a read-only memory and a random access memory, and provides instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0191] A computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the method described in embodiment 1 is performed.

[0192] The method in Example 1 can be directly implemented as being executed by a hardware processor, or by a combination of hardware and software modules within the processor. The software module can be located in a storage medium well-established in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. The storage medium is located in the memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not given here.

[0193] A computer program product includes a computer program, and when the computer program is executed by a processor, the method described in embodiment 1 is implemented.

[0194] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer-executable instructions, such as instructions contained in program modules, which are executed in a device on a real or virtual processor of a target to perform the process / method described above. Generally, program modules include routines, programs, libraries, objects, classes, components, data structures, etc. that perform specific tasks or implement specific abstract data types. In various embodiments, the functionality of program modules can be combined or divided between program modules as needed. The machine-executable instructions for the program modules can be executed in local or distributed devices. In distributed devices, program modules can be located in local and remote storage media.

[0195] The computer program code for implementing the method of the present invention can be written in one or more programming languages. These computer program codes can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the computer or other programmable data processing device, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on a computer, partially on a computer, as an independent software package, partially on a computer and partially on a remote computer, or entirely on a remote computer or server.

[0196] In the context of the present invention, computer program code or related data can be carried by any appropriate carrier to enable a device, apparatus, or processor to perform the various processes and operations described above. Examples of carriers include signals, computer-readable media, and the like. Examples of signals include electrical, optical, radio, acoustic, or other forms of propagated signals, such as carrier waves, infrared signals, and the like.

[0197] Those skilled in the art will appreciate that the units and algorithm steps of the various examples described in conjunction with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0198] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. A ground loop testing method for a high voltage cable in operation, characterized in that: include: Acquire a current response signal and a voltage response signal generated by the ground loop in response to an excitation signal; Performing full vector spectrum analysis on the current response signal and the voltage response signal respectively to obtain a current response three-dimensional spectrum and a voltage response three-dimensional spectrum, and performing spectrum semantic feature extraction on the current response three-dimensional spectrum and the voltage response three-dimensional spectrum to obtain a current response three-dimensional spectrum semantic feature vector and a voltage response three-dimensional spectrum semantic feature vector; Combined with the self-attention mechanism, the semantic feature vectors of the voltage response three-dimensional spectrum and the semantic feature vectors of the current response three-dimensional spectrum are subjected to feature interaction analysis based on fine-grained optimization based on external knowledge, thereby obtaining a fine-grained interaction feature representation between the semantics of the voltage-current response spectrum; A single-phase test result is determined based on the fine-grained interaction feature representation between the semantics of the voltage-current response spectrum, and the test result is used to indicate whether there is an abnormality in the waveform.

2. A ground loop testing method for a high-voltage cable in operation as claimed in claim 1, characterized in that: A multi-scale convolutional LSTM is used to extract spectral semantic features from the current response three-dimensional spectrum and the voltage response three-dimensional spectrum to obtain a current response three-dimensional spectrum semantic feature vector and a voltage response three-dimensional spectrum semantic feature vector.

3. A ground loop testing method for a high-voltage cable in operation as claimed in claim 1, characterized in that: Combined with the self-attention mechanism, the semantic feature vectors of the voltage response three-dimensional spectrum and the semantic feature vectors of the current response three-dimensional spectrum are subjected to feature interaction analysis based on fine-grained optimization based on external knowledge, and a fine-grained interaction feature representation between the semantics of the voltage-current response spectrum is obtained, specifically: According to the semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum, combined with the external knowledge attention learnable memory parameter matrix, an external knowledge optimized voltage-current response spectrum semantic fine-grained feature interaction matrix is obtained; Utilizing the external knowledge to optimize the fine-grained feature interaction matrix between the semantics of the voltage-current response spectrum, and combining the attention mechanism to optimize the semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum, respectively, to obtain optimized semantic feature vectors of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum; The optimized semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum are semantically associated and encoded to obtain a fine-grained interaction feature vector between the semantics of the voltage-current response spectrum.

4. A ground loop testing method for a high-voltage cable in operation as claimed in claim 3, characterized in that: The external knowledge is used to optimize the fine-grained feature interaction matrix between the semantics of the voltage-current response spectrum, and the attention mechanism is combined to optimize the semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum, respectively, to obtain the optimized semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum, specifically: Performing a linear transformation on the semantic feature vector of the voltage response three-dimensional spectrum to obtain a voltage response three-dimensional spectrum joint semantic query feature vector and a voltage response three-dimensional spectrum joint semantic value feature vector; The external knowledge is used to optimize the fine-grained feature interaction matrix between the semantics of the voltage-current response spectrum as a key matrix, and the voltage response three-dimensional spectrum joint semantic query feature vector, the voltage response three-dimensional spectrum joint semantic value feature vector, and the key matrix are calculated based on the self-attention mechanism in the Transformer to obtain an optimized voltage response three-dimensional spectrum semantic feature vector; Performing a linear transformation on the semantic feature vector of the current response three-dimensional spectrum to obtain a semantic query feature vector of the current response three-dimensional spectrum and a semantic value feature vector of the current response three-dimensional spectrum; The external knowledge-optimized voltage-current response spectrum semantic fine-grained feature interaction matrix is used as a key matrix, and the current response three-dimensional spectrum semantic query feature vector, the current response three-dimensional spectrum semantic value feature vector and the key matrix are calculated based on the self-attention mechanism in the Transformer to obtain the optimized current response three-dimensional spectrum semantic feature vector.

5. A ground loop testing method for a high-voltage cable in operation as claimed in claim 3, characterized in that: The optimized semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum are semantically associated and encoded to obtain a fine-grained interaction feature vector between the semantics of the voltage-current response spectrum, specifically: The optimized semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum are divided by position points to obtain a fine-grained interaction feature vector between the semantics of the voltage-current response spectrum.

6. A ground loop testing method for a high-voltage cable in operation as claimed in claim 3, characterized in that: The single-phase test result is determined based on the fine-grained interaction feature representation between the semantics of the voltage-current response spectrum, specifically: Calculating the distance between each pair of eigenvalues in the voltage-current response spectrum semantic fine-grained interaction eigenvector, performing square root processing on the calculated distance to obtain a voltage-current response spectrum semantic fine-grained interaction distance representation matrix; Performing matrix multiplication on the voltage-current response spectrum semantic fine-grained interaction feature vector and the voltage-current response spectrum semantic fine-grained interaction distance representation matrix to obtain a voltage-current response spectrum semantic fine-grained interaction first-level mapping vector; Performing matrix multiplication of the first-level mapping vector of the fine-grained interaction between the semantics of the voltage-current response spectrum and the matrix product of the fine-grained interaction distance representation matrix between the semantics of the voltage-current response spectrum and the fine-grained interaction autocorrelation matrix between the semantics of the voltage-current response spectrum to obtain a multi-level mapping vector of the fine-grained interaction between the semantics of the voltage-current response spectrum; Performing a dot-addition operation on the voltage-current response spectrum semantics fine-grained interaction multi-level mapping vector and the voltage-current response spectrum semantics fine-grained interaction correlation eigenvector composed of the eigenvalues of the voltage-current response spectrum semantics fine-grained interaction autocorrelation matrix to obtain an optimized voltage-current response spectrum semantics fine-grained interaction feature vector; The test results are obtained through a classifier based on the optimized fine-grained interaction feature vector between the semantics of the voltage-current response spectrum.

7. A ground loop testing method for a high-voltage cable in operation as claimed in claim 3, characterized in that: The initialization matrix of the external knowledge attention learnable memory parameter matrix conforms to the Gaussian distribution, and the external knowledge attention learnable memory parameter matrix is optimized using back propagation so that the optimized external knowledge attention learnable memory parameter matrix captures the potential connection between the semantic feature vector of the current response three-dimensional spectrum graph and the semantic feature vector of the voltage response three-dimensional spectrum graph.

8. A ground loop test system for a high voltage cable in operation, characterized in that: include: An acquisition module is configured to: acquire a current response signal and a voltage response signal generated by the ground loop in response to an excitation signal; a feature extraction module configured to: perform full vector spectrum analysis on the current response signal and the voltage response signal respectively to obtain a three-dimensional current response spectrum and a three-dimensional voltage response spectrum, and perform spectrum semantic feature extraction on the three-dimensional current response spectrum and the three-dimensional voltage response spectrum to obtain a semantic feature vector of the three-dimensional current response spectrum and a semantic feature vector of the three-dimensional voltage response spectrum; a feature interaction analysis module configured to: combine a self-attention mechanism to perform feature interaction analysis on the semantic feature vector of the voltage response three-dimensional spectrum and the semantic feature vector of the current response three-dimensional spectrum, using fine-grained feature interaction analysis optimized based on external knowledge, to obtain a fine-grained interaction feature representation between the semantic features of the voltage-current response spectrum; The prediction module is configured to determine a single-phase test result based on the fine-grained interaction feature representation between the semantics of the voltage-current response spectrum, wherein the test result is used to indicate whether there is an abnormality in the waveform.

9. An electronic device, characterized in that: The method comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method according to any one of claims 1 to 6 is completed.

10. A computer-readable storage medium, characterized in that Used to store computer instructions, which, when executed by a processor, complete the method according to any one of claims 1 to 6.

Images