Partial discharge identification method and device, computer equipment and readable storage medium

Abstract

The invention relates to a partial discharge identification method and device, computer equipment and a readable storage medium. The method comprises the following steps: in a process of moving in a transformer substation according to a target path, acquiring a signal generated by power equipment in the transformer substation through a sensor array carried on self-moving equipment; if it is identified that a target signal corresponding to partial discharge exists in the signal, determining a target device corresponding to the target signal, performing feature extraction on the target signal to obtain a feature vector corresponding to the target signal, and matching the feature vector with each standard feature vector in a target feature library to obtain a target device corresponding to the target signal; obtaining a target standard feature vector matched with the feature vector; and obtaining a target partial discharge type corresponding to the target standard feature vector, and analyzing the target signal based on the target partial discharge type to obtain partial discharge information corresponding to the target equipment. By adopting the method, the partial discharge condition can be found in time, and the accuracy and reliability of identifying the partial discharge condition can be improved.

Description

Technical Field

[0001] This application relates to the field of equipment testing technology, and in particular to a partial discharge identification method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Technology

[0002] In the stable operation of the social power grid, substations play a vital role as the heart and hub. The numerous high-voltage power equipment within them, such as switchgear, transformers, and instrument transformers, are like organs in the human body; their health directly affects the safety of the entire power grid. Under long-term exposure to high voltage, high temperature, and mechanical vibration, the internal insulation structures of these devices inevitably age and deteriorate.

[0003] Partial discharge (PD) is the most important and typical "early symptom" in this aging process. It can be simply understood as "micro-sparks" or "micro-arcs" inside or on the surface of insulating materials. Although the energy of a single discharge is very small, its long-term, repeated occurrences will continuously erode the insulating medium like an "anthill," eventually leading to complete insulation breakdown. This can cause catastrophic failures such as short circuits, explosions, and fires, resulting in widespread power outages and huge economic losses.

[0004] Therefore, effective monitoring of partial discharge is equivalent to performing a "health check" on electrical equipment, and is a key technical means to prevent major accidents and achieve predictive maintenance. However, current methods typically involve manually inspecting the partial discharge status of various electrical devices within the substation, resulting in low detection efficiency, long inspection cycles, and an inability to detect sudden defects in a timely manner. Summary of the Invention

[0005] Therefore, it is necessary to provide a partial discharge identification method, device, computer equipment, computer-readable storage medium, and computer program product that can promptly detect partial discharge and improve the accuracy and reliability of partial discharge identification, in order to address the aforementioned technical problems.

[0006] In a first aspect, this application provides a partial discharge identification method, including:

[0007] While moving within the substation along the target path, the system acquires signals generated by the power equipment within the substation through a sensor array mounted on the self-moving device.

[0008] If a target signal corresponding to partial discharge is identified in the signal, the target device corresponding to the target signal is determined, the target signal is used to extract features to obtain the feature vector corresponding to the target signal, the feature vector is matched with each standard feature vector in the target feature library to obtain the target standard feature vector that matches the feature vector; the target partial discharge type corresponding to the target standard feature vector is obtained, the target signal is analyzed based on the target partial discharge type to obtain the partial discharge information corresponding to the target device;

[0009] The target feature library is used to store the correlation between standard feature vectors and partial discharge types. The target feature library is obtained by collecting various types of partial discharge training signals in the target environment and analyzing the feature vectors of various types of partial discharge training signals.

[0010] In one embodiment, before identifying a target signal corresponding to partial discharge in the signal, the method further includes: processing the signal to obtain amplitude information of the signal; if the amplitude information exceeds a preset range, determining that a target signal corresponding to partial discharge exists in the signal, and extracting the target signal from the signal based on the amplitude information.

[0011] In one embodiment, the target signal includes a target filtered signal, and feature extraction of the target signal includes: obtaining a first filter and a second filter; inputting the target signal into the first filter and the second filter to obtain a first processed signal, the first processed signal including a first signal and a second signal corresponding to the target signal; inputting the first signal corresponding to the first processed signal into the first filter and the second filter to obtain a second processed signal, the second processed signal including a first signal and a second signal corresponding to the first processed signal; repeating the above steps until the first signal and the second signal corresponding to the Nth processed signal are obtained; determining a target threshold based on the first signal corresponding to the target signal and the second signals corresponding to the first to Nth processed signals; processing the second signals corresponding to the first to Nth processed signals respectively based on the target threshold to obtain detail signals corresponding to the first to Nth processed signals; reconstructing the target filtered signal based on each detail signal and the first signal corresponding to the target signal; and performing feature extraction on the target filtered signal.

[0012] In one embodiment, reconstructing the target filtering signal based on each detail signal and the first signal corresponding to the target signal includes: reconstructing the approximation coefficients corresponding to the (N-1)th processed signal based on the detail signal corresponding to the Nth processed signal and the first signal; reconstructing the approximation coefficients corresponding to the (N-2)th processed signal based on the approximation coefficients corresponding to the (N-1)th processed signal and the detail signal corresponding to the (N-1)th processed signal; and so on until the target filtering signal is reconstructed based on the approximation coefficients corresponding to the first processed signal and the detail signal corresponding to the first processed signal.

[0013] In one embodiment, feature extraction is performed on the target signal to obtain a feature vector corresponding to the target signal, including: acquiring time-domain information corresponding to the target signal, extracting features from the time-domain information to obtain time-domain features corresponding to the target signal; acquiring frequency-domain information corresponding to the target signal, extracting features from the frequency-domain information to obtain frequency-domain features corresponding to the target signal; obtaining a target spectrum based on the power frequency phase of the target signal, and obtaining spectral features corresponding to the target signal based on the target spectrum; and fusing the time-domain features, frequency-domain features, and spectral features to obtain a feature vector corresponding to the target signal.

[0014] In one embodiment, the feature vector is matched with each standard feature vector in the target feature library to obtain a target standard feature vector that matches the feature vector. This includes: calculating the Euclidean distance between the feature vector and each standard feature vector; determining a preset number of undetermined standard feature vectors that match the feature vector based on each Euclidean distance; determining a target partial discharge type that meets the target conditions based on the partial discharge type corresponding to the preset number of undetermined standard feature vectors; and determining a target partial discharge type that matches the target signal based on the target partial discharge type.

[0015] In one embodiment, analyzing the target signal based on the target partial discharge type to obtain the partial discharge information corresponding to the target device includes: acquiring data items to be analyzed based on the target partial discharge type; the data items to be analyzed include defect type, location, signal strength, and risk level; analyzing the target signal to obtain analysis data corresponding to the data items to be analyzed; and obtaining the partial discharge information corresponding to the target device based on the analysis data.

[0016] Secondly, this application also provides a partial discharge identification device, comprising:

[0017] The acquisition module is used to acquire signals generated by electrical equipment in the substation through a sensor array mounted on the self-moving device while moving within the substation according to the target path;

[0018] The identification module is used to identify the target device corresponding to the target signal if a target signal corresponding to partial discharge is detected in the signal; extract features from the target signal to obtain the feature vector corresponding to the target signal; match the feature vector with each standard feature vector in the target feature library to obtain the target standard feature vector that matches the feature vector; obtain the target partial discharge type corresponding to the target standard feature vector; analyze the target signal based on the target partial discharge type to obtain the partial discharge information corresponding to the target device.

[0019] The target feature library is used to store the correlation between standard feature vectors and partial discharge types. The target feature library is obtained by collecting various types of partial discharge training signals in the target environment and analyzing the feature vectors of various types of partial discharge training signals.

[0020] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0021] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above-described method embodiments.

[0022] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0023] The aforementioned partial discharge identification methods, devices, computer equipment, computer-readable storage media, and computer program products utilize self-moving devices equipped with sensor arrays to inspect along the target path. This dynamic movement and holographic perception overcome the limitations of fixed monitoring, achieving comprehensive, blind-spot-free coverage of substation equipment. Subsequently, target signals are efficiently filtered from massive amounts of data and immediately associated with specific equipment, enabling timely problem detection and accurate problem location. Furthermore, by transforming complex signals into standardized feature vectors and matching them with a target feature library constructed within a specific substation environment, partial discharge types can be identified with extremely high accuracy and objectivity. Finally, based on the target partial discharge type, the target signal is analyzed to obtain the corresponding partial discharge information for the target equipment. This automated and standardized intelligent diagnostic approach improves the operational efficiency and safety of substations. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a diagram illustrating the application environment of the partial discharge identification method in one embodiment;

[0026] Figure 2 This is a flowchart illustrating a partial discharge identification method in one embodiment;

[0027] Figure 3This is a flowchart illustrating the partial discharge identification method in another embodiment;

[0028] Figure 4 This is a flowchart illustrating the partial discharge identification method in yet another embodiment;

[0029] Figure 5 This is a structural block diagram of a partial discharge identification device in one embodiment;

[0030] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0032] The partial discharge identification method provided in this application embodiment can be applied to, for example... Figure 1 The application environment shown. In this environment, terminal 102 (e.g., reference) Figure 2The self-moving device (POS) communicates with server 104 via a network. As the POS moves along a target path within the substation, it acquires signals generated by the power equipment within the substation through a sensor array mounted on the POS. If a target signal corresponding to partial discharge is detected, the target device corresponding to the target signal is identified. Feature extraction is performed on the target signal to obtain a feature vector corresponding to the target signal. This feature vector is then matched with various standard feature vectors in a target feature library to obtain a target standard feature vector that matches the feature vector. The target partial discharge type corresponding to the target standard feature vector is obtained, and the target signal is analyzed based on the target partial discharge type to obtain the partial discharge information corresponding to the target device. Afterward, terminal 102 can send this partial discharge information to server 104, which then sends the partial discharge information corresponding to the target device to the user terminal 106 of relevant maintenance personnel, enabling them to obtain the health status of the power equipment within the substation through the user terminal 106. Terminal 102 can be, but is not limited to, various robots, personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle systems, and projection equipment. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted displays. Head-mounted displays can be virtual reality (VR) devices, augmented reality (AR) devices, and smart glasses. Server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0033] In one exemplary embodiment, such as Figure 2 As shown, a partial discharge identification method is provided, which can be applied to... Figure 1 Taking terminal 102 as an example, the explanation includes the following steps S202 to S204. Wherein:

[0034] S202, while moving within the substation along the target path, acquires signals generated by the power equipment within the substation through a sensor array mounted on the self-moving device.

[0035] In this context, self-moving devices refer to robots or platforms with autonomous navigation and mobility capabilities, such as inspection robots. Sensor arrays are collections of multiple sensors (such as UHF sensors, ultrasonic sensors, and HFCT current transformers) used to collect signals from different angles or dimensions. The target path is a pre-planned inspection route for the self-moving device to navigate within the substation.

[0036] In other words, after receiving an inspection task, the self-propelled robot moves within the substation according to the target path corresponding to the inspection task. During the movement, the self-propelled robot collects various signals generated by the various electrical devices in the substation through its onboard sensor array. These signals include electromagnetic signals, acoustic signals, and optical signals.

[0037] S204, if a target signal corresponding to partial discharge is identified in the signal, the target device corresponding to the target signal is determined, the target signal is feature extracted to obtain the feature vector corresponding to the target signal, the feature vector is matched with each standard feature vector in the target feature library to obtain the target standard feature vector that matches the feature vector; the target partial discharge type corresponding to the target standard feature vector is obtained, the target signal is analyzed based on the target partial discharge type to obtain the partial discharge information corresponding to the target device.

[0038] The target signal is a specific signal identified from all acquired signals as potentially caused by partial discharge. The target device is the specific electrical equipment (such as transformers, switchgear, etc.) within the substation that generates the target signal. The feature vector is a numerical vector used to mathematically represent the key features of the target signal (such as waveform, frequency, phase, etc.). The target feature library is a pre-built database storing standard feature vectors for various known partial discharge types and establishing a one-to-one correspondence between standard feature vectors and discharge types. The target partial discharge type is the known partial discharge type (such as air gap discharge, surface discharge, corona discharge, etc.) that best matches the characteristics of the current target signal. Partial discharge information is detailed information about partial discharges occurring on the target device, such as type, severity, and location. In some embodiments, the self-moving device can also receive partial discharge reports to send to the user terminals of relevant maintenance personnel.

[0039] For example, a self-moving device travels along a preset path, its onboard sensor array continuously scanning and collecting electromagnetic, acoustic, or electrical signals generated by surrounding power equipment. It determines in real time whether the collected signals contain a target signal suspected of being a partial discharge. Once identified, the target device from which the signal originates is preliminarily determined by combining the self-moving device's current location and the sensor array's orientation information. Multi-dimensional feature extraction is performed on the target signal to generate a feature vector. Then, this vector is compared with all standard feature vectors in a target feature library to calculate similarity, finding the most matching standard vector to determine the corresponding target partial discharge type. Finally, based on the determined discharge type, the partial discharge information corresponding to the target device is obtained.

[0040] In the aforementioned partial discharge identification method, a sensor array mounted on a self-moving device is used to inspect along the target path. This dynamic movement and holographic perception overcome the limitations of fixed monitoring, achieving comprehensive, blind-spot-free coverage of substation equipment. Subsequently, target signals are efficiently filtered from a massive amount of data and immediately associated with specific equipment, enabling timely detection and accurate problem location. Furthermore, by transforming complex signals into standardized feature vectors and matching them with a target feature library built within a specific substation environment, partial discharge types can be identified with extremely high accuracy and objectivity. Finally, based on the target partial discharge type, the target signal is analyzed to obtain the corresponding partial discharge information for the target equipment. This automated and standardized intelligent diagnostic approach improves the operational efficiency and safety of substations.

[0041] In some embodiments, before identifying a target signal corresponding to partial discharge in the signal, the method further includes: processing the signal to obtain amplitude information of the signal; if the amplitude information exceeds a preset range, determining that a target signal corresponding to partial discharge exists in the signal, and extracting the target signal from the signal based on the amplitude information.

[0042] Amplitude information refers to the signal's amplitude or intensity, serving as a direct indicator of whether abnormal discharge exists. The preset range is a pre-defined threshold interval. When the signal amplitude exceeds this range (usually exceeding the upper threshold), partial discharge is considered to be possible.

[0043] For example, the raw signal acquired by the sensor array is processed, such as by calculating the peak value, RMS value, or envelope amplitude of the signal using a sliding window. The calculated amplitude information is compared with a preset range. This preset range can be dynamically adjusted based on device type, historical data, or background noise level. When the amplitude of the signal is detected to exceed the preset range, the signal segment corresponding to the amplitude exceeding the preset range is extracted from the raw signal stream and used as the target signal to be analyzed.

[0044] Specifically, the raw signal stream acquired by the sensor array is processed in frames. For example, with each frame being 1 millisecond long, the peak or effective value of the signal in each frame is calculated in real time. Then, the peak or effective value of each frame is compared with a preset range. Optionally, this preset range can be automatically set at the start of the inspection by analyzing the background noise level of the previous few minutes, for example, set to 3 to 5 times the average amplitude of the background noise. When the amplitude of a acquired frame signal is detected to exceed this threshold, an anomaly is determined, and a recording module is triggered. This module will capture the complete signal waveform containing a period of time before and after the abnormal frame (e.g., the first 10ms and the last 40ms) as the target signal to be analyzed, while recording the precise timestamp and device location information. Then, it continues to monitor subsequent signals to achieve uninterrupted monitoring.

[0045] It is understandable that if the amplitude information does not exceed the preset range, it is determined that there is no target signal corresponding to partial discharge in the signal, and subsequent signals will continue to be collected.

[0046] In this embodiment, by quickly filtering out suspected abnormal signals from massive amounts of signal data, the computational load of subsequent complex feature extraction and matching algorithms can be reduced, improving the system's real-time performance and stability. A simple amplitude judgment triggers subsequent complex analysis processes, enabling on-demand allocation of computing resources.

[0047] In some embodiments, the target signal includes a target filtered signal, and feature extraction of the target signal includes: obtaining a first filter and a second filter; inputting the target signal into the first filter and the second filter to obtain a first processed signal, the first processed signal including a first signal and a second signal corresponding to the target signal; inputting the first signal corresponding to the first processed signal into the first filter and the second filter to obtain a second processed signal, the second processed signal including a first signal and a second signal corresponding to the first processed signal; repeating the above steps until the first signal and the second signal corresponding to the Nth processed signal are obtained; determining a target threshold based on the first signal corresponding to the target signal and the second signals corresponding to the first to Nth processed signals; processing the second signals corresponding to the first to Nth processed signals respectively based on the target threshold to obtain detail signals corresponding to the first to Nth processed signals; reconstructing the target filtered signal based on each detail signal and the first signal corresponding to the target signal; and performing feature extraction on the target filtered signal.

[0048] The target filtered signal is the cleaner target signal after noise reduction and filtering. The first filter and second filter are a pair of complementary filters, typically referring to the low-pass filter (which decomposes the approximate signal, i.e., the first signal) and high-pass filter (which decomposes the detail signal, i.e., the second signal) in wavelet transform. The first signal is the low-frequency component of the signal, representing the main outline and trend of the signal, also known as the approximation coefficients. The second signal is the high-frequency component of the signal, containing the details, abrupt changes, and noise, also known as the detail coefficients. The Nth processed signal is the signal obtained after N levels of filtering decomposition. N can be preset. The target threshold is a threshold used to quantize the detail signal, distinguishing between valid signals and noise. The detail signal is the detail coefficients after thresholding, removing some noise.

[0049] It can be understood that this embodiment is a key preprocessing step before feature extraction of the target signal, namely signal noise reduction.

[0050] Specifically, first, a suitable wavelet basis function (such as Daubechies or Symlet wavelets) and a decomposition level N are selected. Then, the target signal is input into a first-level filter bank, which contains a low-pass filter (i.e., the first filter) and a high-pass filter (i.e., the second filter), decomposing the signal into a first signal (i.e., approximation coefficients) representing the low-frequency profile and a second signal containing high-frequency details (i.e., detail coefficients), respectively. Next, the system uses the first-level first signal as new input to the second-level filter bank for the same decomposition, repeating this process N times to form a pyramid-shaped multi-resolution decomposition structure. When determining the target threshold, the system can use a heuristic method, such as calculating the standard deviation of the second signal (i.e., detail coefficients) at each level and then setting the threshold to a multiple of that standard deviation (e.g., 3 times), thus effectively estimating the noise level. Finally, the system uses a soft-thresholding function to process the second signal at each level, compressing coefficients with absolute values ​​less than the threshold towards zero, thereby suppressing noise while preserving the effective signal.

[0051] In this embodiment, by denoising the target signal, white noise and some periodic interference mixed in the partial discharge signal can be effectively removed, while the transient characteristics of the original discharge signal are preserved to the maximum extent, providing high-quality input for subsequent feature extraction. Furthermore, using the decomposed signal itself to determine the threshold allows the denoising process to adaptively adjust according to the actual noise level of the signal, resulting in stronger robustness.

[0052] In some embodiments, reconstructing the target filtering signal based on each detail signal and the first signal corresponding to the target signal includes: reconstructing the approximation coefficients corresponding to the (N-1)th processed signal based on the detail signal corresponding to the Nth processed signal and the first signal; reconstructing the approximation coefficients corresponding to the (N-2)th processed signal based on the approximation coefficients corresponding to the (N-1)th processed signal and the detail signal corresponding to the (N-1)th processed signal; and so on until the target filtering signal is reconstructed based on the approximation coefficients corresponding to the first processed signal and the detail signal corresponding to the first processed signal.

[0053] Specifically, the threshold-quantized detail signal of level N and the approximation coefficients of level N are input together into the corresponding wavelet reconstruction filter bank. This wavelet reconstruction filter bank contains a low-pass reconstruction filter and a high-pass reconstruction filter. These filters upsample and convolve the input coefficients, then sum the results to accurately reconstruct the approximation coefficients of level N-1. Subsequently, the system feeds this newly reconstructed level N-1 approximation coefficients and the threshold-quantized detail signal of level N-1 back into the reconstruction filter bank to generate the approximation coefficients of level N-2. This iterative process continues until the coefficients of level 1 are processed. Finally, the approximation coefficients obtained from the level 1 reconstruction are combined with the level 1 processed detail signal. Through a final reconstruction, a target filtered signal that has removed noise while retaining key transient features can be synthesized.

[0054] In this embodiment, by synthesizing the obtained detailed signals and approximate signals, it can be ensured that the useful components of the signal can be accurately recovered after noise removal.

[0055] In some embodiments, feature extraction is performed on the target signal to obtain a feature vector corresponding to the target signal, including: acquiring time-domain information corresponding to the target signal, extracting features from the time-domain information to obtain time-domain features corresponding to the target signal; acquiring frequency-domain information corresponding to the target signal, extracting features from the frequency-domain information to obtain frequency-domain features corresponding to the target signal; obtaining a target spectrum based on the power frequency phase of the target signal, and obtaining spectral features corresponding to the target signal based on the target spectrum; and fusing the time-domain features, frequency-domain features, and spectral features to obtain a feature vector corresponding to the target signal.

[0056] In this context, time-domain information / features refer to the signal's representation and statistical characteristics in the time dimension, such as pulse peak value, pulse width, rise time, and number of discharges. Frequency-domain information / features refer to the signal's characteristics in the frequency dimension after Fourier transform, such as dominant frequency, spectral width, and energy at specific frequency points. Power frequency phase refers to the periodic phase angle (0-360 degrees) of alternating current. Partial discharge typically occurs within a specific phase range of power frequency voltage. A target spectrum is a two-dimensional or three-dimensional map plotted by relating the number of partial discharge pulses, amplitude, or energy to the power frequency phase; the most typical example is the PRPD (Phase Resolved Partial Discharge) spectrum. Spectrum features are statistical or morphological features extracted from the target spectrum, such as the shape, symmetry, center position, and distribution dispersion of the spectrum.

[0057] For example, in terms of time-domain feature extraction, the system not only calculates basic parameters such as pulse peak value and pulse width, but also extracts more complex statistical features, such as the Weibull distribution parameters of the pulse interval time, the skewness and kurtosis of the signal waveform, to capture the randomness and intensity of the discharge. In terms of frequency-domain feature extraction, after performing a Fast Fourier Transform on the target signal, the system not only focuses on the dominant frequency but also analyzes the energy distribution of the spectrum, such as calculating the energy proportion of the spectrum in several key frequency bands (e.g., 300-600MHz, 600-1500MHz), forming a frequency-domain energy vector. In terms of spectral feature extraction, the system constructs a PRPD (Phase-Resolved Partial Discharge) map and employs image recognition technology, such as using a Convolutional Neural Network (CNN), to automatically learn and extract deep features that can distinguish different discharge types directly from the spectral image, such as the wing shape, symmetry, and grayscale distribution texture of the spectrum. Finally, during the fusion process, the system can first concatenate the time-domain and frequency-domain feature vectors, and then concatenate them again with the CNN feature vectors of the graph to form a high-dimensional comprehensive feature vector; or, a more advanced implementation is to design a fusion network that uses the three features as different input channels and automatically learns the optimal fusion weights between them to generate the final feature vector.

[0058] Specifically, for the target signal, the following 12 characteristic parameters can be calculated:

[0059] Time-domain characteristics (4): Pulse peak amplitude (V_peak), pulse rise time (t_rise), pulse duration (t_duration), and waveform energy.

[0060] Frequency domain features (4): Perform FFT transformation on the waveform to calculate its center frequency (f_center), root mean square frequency (f_rms), and the frequency band with the highest energy proportion (e.g., the ratio of energy in the 300-500MHz band to the total energy), and spectral entropy.

[0061] PRPD spectral statistical characteristics (4): 1000 pulses were plotted as PRPD spectral data based on their power frequency phase. The number of discharges (N), total discharge charge (Q), skewness, and kurtosis of the spectral data were calculated.

[0062] In this embodiment, by fusing multi-dimensional features, the characteristics of partial discharge signals can be comprehensively and three-dimensionally characterized, reducing the limitations and ambiguities that may exist due to a single feature dimension, thereby significantly improving the accuracy of subsequent type identification.

[0063] In some embodiments, matching the feature vector with each standard feature vector in the target feature library to obtain a target standard feature vector that matches the feature vector includes: calculating the Euclidean distance between the feature vector and each standard feature vector; determining a preset number of undetermined standard feature vectors that match the feature vector based on each Euclidean distance; and determining the partial discharge type that meets the target conditions based on the partial discharge type corresponding to the preset number of undetermined standard feature vectors, so as to determine the target partial discharge type that matches the target signal.

[0064] Euclidean distance is the straight-line distance between two points in multidimensional space, often used to measure the similarity between two vectors. The smaller the distance, the higher the similarity. The undetermined standard feature vectors are the top K standard feature vectors in the target feature library that have the smallest Euclidean distance to the current feature vector. The target condition is a decision rule used to select the final result from the K candidates.

[0065] For example, firstly, the Euclidean distance between the feature vector to be identified and all standard feature vectors in the target feature library is calculated. Then, all distances are sorted in ascending order, and the K vectors with the smallest distances are selected as the target standard feature vectors. The value of K can be optimized according to the size of the feature library and the complexity of the classification, for example, by determining an optimal value through cross-validation. When applying the target conditions for decision-making, in addition to the simple majority voting method, a weighted voting method can also be used. That is, the voting weight of each feature vector to be identified is inversely proportional to its Euclidean distance. The closer the distance, the greater its influence on the final decision. This can effectively avoid misclassification caused by uneven distribution of samples in the feature library. For example, if K=5, where 3 neighbors are of class A and 2 are of class B, but the distance between the 2 neighbors of class B is much smaller than that between the 3 neighbors of class A, the weighted voting method may ultimately determine it as class B, thereby improving the granularity and accuracy of the decision.

[0066] In this embodiment, compared to black-box models such as deep learning, the KNN algorithm has simple logic and controllable computational cost, making it very suitable for deployment on resource-constrained mobile devices. Furthermore, by introducing K nearest neighbors instead of a single nearest neighbor for voting decisions, misjudgments caused by individual noisy samples or imbalanced samples in the feature library can be effectively suppressed, improving the stability and generalization ability of the partial discharge type identification method.

[0067] In some embodiments, analyzing the target signal based on the target partial discharge type to obtain partial discharge information corresponding to the target device includes: acquiring data items to be analyzed based on the target partial discharge type; the data items to be analyzed include defect type, location, signal strength, and risk level; analyzing the target signal to obtain analysis data corresponding to the data items to be analyzed; and obtaining partial discharge information corresponding to the target device based on the analysis data.

[0068] Partial discharge information can be displayed in report form. When analyzing data items, key information dimensions requiring further analysis and output are determined based on the identified discharge type. Defect type refers to the specific type of partial discharge or a physical defect in the target device. Location: The specific location of the partial discharge inside or on the surface of the device. Risk level: The level of fault risk assessed based on the discharge type, intensity, and development trend (e.g., low, medium, high, urgent).

[0069] Specifically, when acquiring data items to be analyzed based on the target partial discharge type, the system queries a built-in diagnostic knowledge base. This knowledge base is indexed by the target partial discharge type and associates it with corresponding analysis templates and evaluation models. For example, when identified as surface discharge, the data items to be analyzed not only include general signal strength but also pay special attention to the discharge phase width and development trend. When analyzing the target signal, for location information, the system can utilize the precise time difference of arrival signals from the UHF sensor array, combined with the device's three-dimensional model, to achieve precise positioning through spatial geometric algorithms, with errors controllable to the centimeter level. For risk level, the system calls a multi-parameter evaluation model. The model's input includes the target partial discharge type, current signal strength, discharge pulse repetition rate, and growth trend in historical data. Through fuzzy logic or decision tree algorithms, it comprehensively outputs a quantified risk score and level (e.g., low, medium, high, critical). The final partial discharge information report will be presented in a combination of text and graphics, including the original waveform of the discharge signal, PRPD spectrum, the marked location on the device's 3D model, risk level, and preliminary handling suggestions based on the knowledge base, such as suggesting power outage maintenance or shortening the monitoring cycle, providing maintenance personnel with one-stop decision support.

[0070] In this embodiment, by combining the identification results with a structured knowledge base and risk assessment model, the diagnostic results are standardized and made more intelligent. This enables maintenance personnel of different levels to make consistent and scientific judgments based on the reports, thereby improving the professionalism and intelligence of substation maintenance.

[0071] In one exemplary embodiment, a target database may be constructed before partial discharge identification. (See reference...) Figure 3 For details on how to build the target database, please refer to the following:

[0072] 1. Build a simulation platform: First, in an electromagnetically pure and controllable environment such as a laboratory, build a physical simulation platform with the same or similar structure as the actual substation equipment (such as switchgear).

[0073] Specifically, refer to Figure 4The laboratory is a high-voltage shielded laboratory, with the ambient temperature controlled at 25°C ± 2°C and the relative humidity at 50% ± 5% RH. The discharge unit is, for example, a KYN28A-12 type metal-clad withdrawable switchgear. The high-voltage source can be a YDTW-50 / 100 type partial discharge-free test transformer, capable of outputting a continuously adjustable AC voltage of 0-100kV. A 100kV capacitive voltage divider is used for precise voltage measurement, and a 10kΩ protective resistor is connected in series in the circuit.

[0074] Furthermore, data acquisition can be performed using a PDB-P1 external UHF sensor with an effective frequency band of 300MHz ~ 1.5GHz, a Keysight Infiniium UXR series oscilloscope with a bandwidth of 2.5 GHz and a sampling rate of 20 GS / s, and a power frequency power supply synchronous trigger module can be used to obtain the relationship between the discharge signal and the power frequency phase (for generating PRPD spectra).

[0075] 2. Implantation of Typical Defects: On this physical simulation platform, a single, known typical partial discharge defect is precisely implanted artificially. For example, a needle-tip electrode is installed to simulate "point discharge," a tiny air bubble is left inside the insulating plate to simulate "internal air gap discharge," or a free metal particle is placed to simulate "levitation discharge." Only one defect is introduced in each experiment, ensuring the uniqueness of the signal source.

[0076] Specifically, take two 100mm×100mm×5mm epoxy resin boards. Drill a flat-bottomed blind hole with a diameter of 5mm and a depth of 1mm in the center of one of the boards to simulate the air gap inside the insulation. Press the two boards together and seal the perimeter with epoxy resin to ensure the blind hole forms a sealed air gap. Attach a 50mm×50mm copper foil electrode to the top and bottom surfaces of each board. This completes the defect model. Then, place the defect model on the busbar support insulator inside the switchgear. Attach a UHF sensor to the observation window of the switchgear housing using a magnetic base, approximately 30cm away from the internal defect model.

[0077] 3. Multi-sensor signal acquisition: Apply operating voltage to the simulation platform to generate stable partial discharge. Use a multi-sensor array (such as an UHF antenna, acoustic sensors, etc.) of the same type as the one to be mounted on the inspection robot to synchronously and with high fidelity acquire the discharge signal.

[0078] Specifically, the high-voltage source is started, and the voltage is slowly and uniformly increased from 0 kV. Real-time monitoring is performed using an oscilloscope. When the voltage reaches 8.7 kV, a stable and repetitive partial discharge pulse signal is observed, which is considered the initiation voltage, and the voltage is maintained stable. For example, with the oscilloscope trigger threshold set to 5 mV, 1000 partial discharge UHF pulse waveforms are acquired over 10 minutes. Each waveform is recorded for 50 μs, including data before and after the pulse.

[0079] 4. Feature Extraction and Labeling: The acquired raw signal undergoes preprocessing such as digital filtering and noise reduction. Then, feature parameters characterizing the signal's essence are extracted from multiple dimensions, including time, frequency, and phase, forming a multi-dimensional feature vector. A crucial step is to uniquely bind and label this feature vector with a pre-defined defect type (e.g., "internal air gap discharge").

[0080] Specifically, the script `read_waveform.m` is written to batch read 1000 waveform data files in .csv format saved by the oscilloscope and store them in a 1000×N matrix `raw_data` (where N is the number of sampling points). Next, wavelet denoising is performed on each row of the `raw_data` matrix (i.e., each waveform). The specific parameters are: using the db5 (Daubechies 5) wavelet basis, performing a 5-level decomposition, and applying a soft thresholding rule for denoising. This yields the denoised data matrix `denoised_data`. This method can also be used to extract features from the target signal during partial discharge identification. Then, each row of the `denoised_data` matrix is ​​traversed, and the aforementioned 12 feature parameters are calculated. Finally, these 12 feature parameters are combined into a 1×12 feature vector. The average of the 1000 feature vectors generated from the 1000 waveforms is taken to obtain a standard feature vector that best represents the characteristics of the "internal air gap discharge" defect at 8.7kV.

[0081] 5. Construct a standard feature library: Repeat the above steps to simulate and characterize all typical defect types that need to be identified. Finally, store all the mapping relationships between "defect type and feature vector" to form a structured database, namely the target feature library. This library is equivalent to a "partial discharge fault atlas" that can be queried by a computer.

[0082] Specifically, the aforementioned standard feature vectors are stored in a database. The database table structure is (Defect_ID, Defect_Type, Voltage_Level, Feature_1, ..., Feature_12). In this example, the stored record is (1, 'Internal_Air_Gap', 8.7, value_1, ..., value_12).

[0083] It is understandable that, in order to continuously expand the target feature library, other defect models (such as tip discharge model, suspension discharge model) can be replaced, and steps 2 to 5 above can be repeated.

[0084] Combination Figure 3 and Figure 4 This involves analyzing partial discharge information of electrical equipment within a substation. For details, please refer to the following:

[0085] 1. While moving along the target path within the substation, signals generated by the power equipment within the substation are acquired through a sensor array mounted on the self-moving device.

[0086] The robot platform is a tracked autonomous mobile platform equipped with LiDAR and binocular vision for navigation. Sensors may include, for example, the same PDB-P1 UHF sensor used in the laboratory. The robot's main control chip is an ARK-3532 industrial computer (IPC) with an integrated Intel Core i7-10700E processor, responsible for signal processing and robot control. The robot is equipped with an NVIDIA Jetson AGX Xavier module for running intelligent recognition algorithms. The operating system is Ubuntu 20.04, integrating ROS (Robot Operating System) middleware. Diagnostic programs are programmed using a mix of C++ and Python and deployed on the main control chip. Furthermore, the aforementioned target feature library is pre-stored in the robot system.

[0087] Specifically, the robot autonomously navigates to the predetermined detection point in front of the target KYN28A-12 type switch cabinet (e.g., "#5 incoming line cabinet"). The onboard UHF sensor begins to continuously collect signals.

[0088] 2. If a target signal corresponding to partial discharge is identified in the signal, the target device corresponding to the target signal is determined, the target signal is feature extracted to obtain the feature vector corresponding to the target signal, the feature vector is matched with each standard feature vector in the target feature library to obtain the target standard feature vector that matches the feature vector; the target partial discharge type corresponding to the target standard feature vector is obtained, the target signal is analyzed based on the target partial discharge type to obtain the partial discharge information corresponding to the target device.

[0089] Specifically, when the amplitude of the acquired signal exceeds a preset threshold (e.g., 3mV), the diagnostic program deployed on the robot is triggered. The diagnostic program automatically acquires a complete waveform containing pulses and immediately calls the same denoising and feature extraction algorithm used in the process of building the target feature library to calculate the 1×12-dimensional feature vector of the current signal. The calculated feature vector is sent to the NVIDIA Jetson AGX Xavier module. The Jetson module runs the target algorithm (e.g., a k-NN classification program) to calculate the Euclidean distance between the feature vector and all standard feature vectors in the target feature library, and finds the k (5) nearest neighbors. Based on the defect types of these 5 neighbors, a vote is taken, and the defect type with the most votes is output as the identification result. For example, the output result of the target algorithm is "Internal_Air_Gap" (internal air gap discharge) with a confidence level of 92%. The diagnostic program then determines the risk level of the partial discharge as "high risk" according to the built-in risk rule library (rule: IF Defect_Type == 'Internal_Air_Gap' THEN Risk_Level = 'High'). Subsequently, based on the obtained partial discharge information, an alarm message can be generated with the diagnosis result: "Internal air gap discharge, confidence level: 92%, risk level: high risk." This information is reported in real time to the back-end monitoring center (i.e., the user terminal of the relevant maintenance personnel) via the onboard 5G communication module and is highlighted on the robot's local screen.

[0090] The above steps significantly improve the accuracy and reliability of diagnosis. Existing technologies often rely on mixed and ambiguous field data for model training, resulting in inherent limitations in accuracy. This solution addresses this issue at its source, constructing a "pure" and absolutely accurate standard feature library through laboratory physical simulation. Diagnosis based on this library is like having a "standard answer," fundamentally eliminating the uncertainty of the data source and making defect type identification more accurate and reliable. Furthermore, it achieves "zero" latency and high timeliness in fault early warning. Diagnostic capabilities are directly integrated into the front end of the inspection robot, realizing a closed loop of "on-site data collection, on-site diagnosis, and on-site alarm." This reduces the warning time for high-risk faults from hours or even days to seconds, providing a crucial window of opportunity for emergency response and maintenance, effectively preventing faults from escalating into serious accidents. In addition, it standardizes diagnostic technology and enables a high degree of robot intelligence. The target feature library established in this solution is based on universal discharge physics mechanisms, possessing versatility and enabling robots equipped with this system to have replicable and scalable standardized diagnostic capabilities. It elevates robots from passive "data collection tools" to "intelligent diagnostic experts" capable of autonomous thinking and judgment, greatly improving the level of automation and intelligence in operation and maintenance.

[0091] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0092] Based on the same inventive concept, this application also provides a partial discharge identification device for implementing the partial discharge identification method described above. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations in one or more partial discharge identification device embodiments provided below can be found in the limitations of the partial discharge identification method described above, and will not be repeated here.

[0093] In one exemplary embodiment, such as Figure 5 As shown, a partial discharge identification device is provided, including: an acquisition module 501 and an identification module 502, wherein:

[0094] The acquisition module 501 is used to acquire signals generated by power equipment in the substation through a sensor array mounted on the self-moving device while moving within the substation according to the target path.

[0095] The identification module 502 is used to identify the target device corresponding to the target signal if a target signal corresponding to partial discharge is identified in the signal, extract features from the target signal to obtain the feature vector corresponding to the target signal, match the feature vector with each standard feature vector in the target feature library to obtain the target standard feature vector that matches the feature vector, obtain the target partial discharge type corresponding to the target standard feature vector, analyze the target signal based on the target partial discharge type, and obtain the partial discharge information corresponding to the target device.

[0096] In some embodiments, before identifying a target signal corresponding to partial discharge in the signal, the identification module 502 is further configured to: process the signal to obtain amplitude information of the signal; if the amplitude information exceeds a preset range, determine that a target signal corresponding to partial discharge exists in the signal, and extract the target signal from the signal based on the amplitude information.

[0097] In some embodiments, the target signal includes a target filter signal. In terms of feature extraction from the target signal, the identification module 502 is specifically configured to: acquire a first filter and a second filter; input the target signal to the first filter and the second filter to obtain a first processed signal, the first processed signal including a first signal and a second signal corresponding to the target signal; input the first signal corresponding to the first processed signal to the first filter and the second filter to obtain a second processed signal, the second processed signal including a first signal and a second signal corresponding to the first processed signal; repeat the above steps until the first signal and the second signal corresponding to the Nth processed signal are obtained; determine a target threshold based on the first signal corresponding to the target signal and the second signals corresponding to the first to Nth processed signals; process the second signals corresponding to the first to Nth processed signals respectively based on the target threshold to obtain detail signals corresponding to the first to Nth processed signals; reconstruct the target filter signal based on each detail signal and the first signal corresponding to the target signal; and perform feature extraction on the target filter signal.

[0098] In some embodiments, in reconstructing the target filtering signal based on each detail signal and the first signal corresponding to the target signal, the identification module 502 is specifically configured to: reconstruct the approximation coefficients corresponding to the (N-1)th processed signal based on the detail signal corresponding to the Nth processed signal and the first signal; reconstruct the approximation coefficients corresponding to the (N-2)th processed signal based on the approximation coefficients corresponding to the (N-1)th processed signal and the detail signal corresponding to the (N-1)th processed signal; until the target filtering signal is reconstructed based on the approximation coefficients corresponding to the first processed signal and the detail signal corresponding to the first processed signal.

[0099] In some embodiments, in terms of extracting features from the target signal to obtain a feature vector corresponding to the target signal, the identification module 502 is specifically used to: acquire time-domain information corresponding to the target signal, extract features from the time-domain information to obtain time-domain features corresponding to the target signal; acquire frequency-domain information corresponding to the target signal, extract features from the frequency-domain information to obtain frequency-domain features corresponding to the target signal; obtain a target spectrum based on the power frequency phase of the target signal, and obtain spectral features corresponding to the target signal based on the target spectrum; and perform fusion processing on the time-domain features, frequency-domain features, and spectral features to obtain a feature vector corresponding to the target signal.

[0100] In some embodiments, in matching the feature vector with each standard feature vector in the target feature library to obtain a target standard feature vector that matches the feature vector, the identification module 502 is specifically used to: calculate the Euclidean distance between the feature vector and each standard feature vector; determine a preset number of undetermined standard feature vectors that match the feature vector based on each Euclidean distance; and determine the partial discharge type that meets the target conditions based on the partial discharge type corresponding to the preset number of undetermined standard feature vectors, so as to determine the target partial discharge type that matches the target signal.

[0101] In some embodiments, in analyzing the target signal based on the target partial discharge type to obtain the partial discharge information corresponding to the target device, the identification module 502 is specifically used to: obtain the data item to be analyzed based on the target partial discharge type; the data item to be analyzed includes defect type, location, signal strength, and risk level; analyze the target signal to obtain analysis data corresponding to the data item to be analyzed; and obtain the partial discharge information corresponding to the target device based on the analysis data.

[0102] Each module in the aforementioned partial discharge identification device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0103] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 6As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a partial discharge method. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0104] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0105] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0106] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0107] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0108] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0109] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0110] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0111] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for identifying partial discharge, characterized in that, Applied to self-moving devices, the method includes: While moving within the substation along the target path, the sensor array mounted on the self-moving device acquires signals generated by the power equipment within the substation. If a target signal corresponding to partial discharge is identified in the signal, the target device corresponding to the target signal is determined, features are extracted from the target signal to obtain a feature vector corresponding to the target signal, and the feature vector is matched with each standard feature vector in the target feature library to obtain a target standard feature vector that matches the feature vector; the target partial discharge type corresponding to the target standard feature vector is obtained, and the target signal is analyzed based on the target partial discharge type to obtain the partial discharge information corresponding to the target device; The target feature library is used to store the correlation between standard feature vectors and partial discharge types. The target feature library is obtained by collecting various types of partial discharge training signals in the target environment and analyzing the feature vectors of the various types of partial discharge training signals.

2. The method according to claim 1, characterized in that, Before identifying a target signal corresponding to partial discharge in the signal, the method further includes: The signal is processed to obtain the amplitude information of the signal; If the amplitude information exceeds a preset range, it is determined that there is a target signal corresponding to partial discharge in the signal, and the target signal is extracted from the signal based on the amplitude information.

3. The method according to claim 1, characterized in that, The target signal includes a target filtering signal, and the feature extraction of the target signal includes: Obtain the first filter and the second filter; The target signal is input to the first filter and the second filter to obtain a first processed signal. The first processed signal includes a first signal and a second signal corresponding to the target signal. The first signal corresponding to the first processed signal is input to the first filter and the second filter to obtain a second processed signal. The second processed signal includes a first signal and a second signal corresponding to the first processed signal. The above steps are repeated until the first signal and the second signal corresponding to the Nth processed signal are obtained. The target threshold is determined based on the first signal corresponding to the target signal and the second signals corresponding to the first to the Nth processed signals. Based on the target threshold, the second signals corresponding to the first processed signal to the Nth processed signal are processed respectively to obtain the detail signals corresponding to the first processed signal to the Nth processed signal; Based on each of the detailed signals and the first signal corresponding to the target signal, the target filtering signal is reconstructed; Feature extraction is performed on the target filtered signal.

4. The method according to claim 3, characterized in that, The step of reconstructing the target filtering signal based on each of the detailed signals and the first signal corresponding to the target signal includes: Based on the detail signal and the first signal corresponding to the Nth processed signal, the approximation coefficients corresponding to the (N-1)th processed signal are reconstructed. Based on the approximation coefficients corresponding to the (N-1)th processed signal and the detail signal corresponding to the (N-1)th processed signal, the approximation coefficients corresponding to the (N-2)th processed signal are reconstructed. The target filtering signal is reconstructed based on the approximation coefficients and detail signals corresponding to the first processed signal.

5. The method according to claim 1, characterized in that, The step of extracting features from the target signal to obtain a feature vector corresponding to the target signal includes: Obtain the time-domain information corresponding to the target signal, and perform feature extraction on the time-domain information to obtain the time-domain features corresponding to the target signal; The frequency domain information corresponding to the target signal is obtained, and the frequency domain information is used to extract features to obtain the frequency domain features corresponding to the target signal. A target spectrum is obtained based on the power frequency phase of the target signal, and spectrum features corresponding to the target signal are obtained based on the target spectrum. The time-domain features, frequency-domain features, and spectral features are fused to obtain the feature vector corresponding to the target signal.

6. The method according to claim 1, characterized in that, The step of matching the feature vector with each standard feature vector in the target feature library to obtain a target standard feature vector that matches the feature vector includes: Calculate the Euclidean distance between the feature vector and each of the standard feature vectors; Based on each of the Euclidean distances, a preset number of undetermined standard feature vectors that match the feature vectors are determined; Based on the partial discharge types corresponding to the preset number of undetermined standard feature vectors, the partial discharge types that meet the target conditions are determined, so as to determine the target partial discharge type that matches the target signal.

7. The method according to claim 1, characterized in that, The step of analyzing the target signal based on the target partial discharge type to obtain the partial discharge information corresponding to the target device includes: The data items to be analyzed are obtained based on the target partial discharge type; the data items to be analyzed include defect type, location, signal strength, and risk level; The target signal is analyzed to obtain analysis data corresponding to the data item to be analyzed; The partial discharge information corresponding to the target device is obtained based on the analysis data.

8. A partial discharge identification device, characterized in that, Applied to self-moving devices, the device includes: The acquisition module is used to acquire signals generated by power equipment in the substation through a sensor array mounted on the self-moving device while moving within the substation according to the target path. The identification module is configured to: if a target signal corresponding to partial discharge is detected in the signal, determine the target device corresponding to the target signal; extract features from the target signal to obtain a feature vector corresponding to the target signal; match the feature vector with each standard feature vector in the target feature library to obtain a target standard feature vector that matches the feature vector; obtain the target partial discharge type corresponding to the target standard feature vector; analyze the target signal based on the target partial discharge type to obtain the partial discharge information corresponding to the target device. The target feature library is used to store the correlation between standard feature vectors and partial discharge types. The target feature library is obtained by collecting various types of partial discharge training signals in the target environment and analyzing the feature vectors of the various types of partial discharge training signals.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images