A Method and Electronic Equipment for Identifying High-Resistance Grounding Arc Features Based on Voiceprint Recognition

Abstract

This invention discloses a method and electronic device for identifying high-resistivity grounding arc features based on voiceprint recognition. The method includes: acquiring multimodal acoustic signals at monitoring points on the transmission line under test, wherein the multimodal acoustic signals include structure acoustic signals and airborne acoustic signals collected at the monitoring points during multiple sampling periods; identifying suspected fault transient disturbance events and corresponding target signal segments based on the structure acoustic signals in the multimodal acoustic signals, wherein the suspected fault transient disturbance event represents a transient disturbance event suspected to be caused by high-resistivity grounding arc discharge; extracting features from the target signal segments to obtain fused fault voiceprint features; and determining the high-resistivity grounding arc feature identification result of the transmission line under test based on the fused fault voiceprint features. This invention solves the technical problem that related technologies struggle to effectively distinguish between actual high-resistivity grounding arc discharge and normal system operation disturbances.

Description

Technical Field

[0001] This invention relates to the field of smart grids, and more specifically, to a method and electronic device for identifying high-resistivity grounding arc features based on voiceprint recognition. Background Technology

[0002] In the operation and maintenance of power systems, high-resistance grounding faults are a key issue affecting the safe and stable operation of the power grid due to their concealment and difficulty in detection. This type of fault typically occurs between transmission lines and the ground. Because of the high grounding resistance and weak fault current, the electrical characteristics are not obvious, making effective detection difficult using traditional electrical parameter monitoring methods. In recent years, researchers have attempted to introduce acoustic monitoring technology into the diagnosis of high-resistance grounding faults, using airborne acoustic sensors and structured acoustic sensors to capture the acoustic signals generated during a fault. The aim is to improve the sensitivity and reliability of fault diagnosis through indirect detection of non-electrical quantities. However, the application of single acoustic sensing technology still has significant limitations.

[0003] For detection methods based on airborne acoustic signals, the complex environment of transmission lines, with background noise such as wind noise, rain noise, traffic noise, and biological noise far exceeding the weak sound pressure generated by arc discharge, completely drowns out the effective signal, making extraction and identification difficult. Furthermore, the energy of airborne acoustic signals attenuates rapidly with propagation distance, requiring a large number of microphones to achieve full coverage of the line, which lacks economic efficiency and practicality in engineering implementation. On the other hand, while detection methods based on structured acoustic signals can effectively avoid environmental noise interference by utilizing the efficient waveguide characteristics of conductors, their features are limited, failing to effectively distinguish between genuine arc discharges and transient disturbances caused by lightning strikes, bird strikes, falling ice, or human operation. This results in a high false alarm rate, failing to meet the high-precision diagnostic results required for operational and maintenance decisions. Therefore, accurately identifying and distinguishing high-resistance grounding arc discharge events from normal system operation disturbances in complex field environments has become an urgent technical challenge. In particular, related technologies struggle to effectively distinguish between genuine high-resistance grounding arc discharges and normal system operation disturbances, hindering the achievement of high-reliability diagnostic results.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This invention provides a method and electronic device for identifying high-resistance grounding arc features based on voiceprint recognition, so as to at least solve the technical problem that related technologies have difficulty in effectively distinguishing between real high-resistance grounding arc discharge and system normal operation disturbances.

[0006] According to one aspect of the present invention, a method for identifying high-resistivity grounding arc features based on acoustic signature recognition is provided, comprising: acquiring multimodal acoustic signals at monitoring points on the transmission line under test, wherein the multimodal acoustic signals include structural acoustic signals and airborne acoustic signals collected at the monitoring points at multiple sampling periods, wherein the structural acoustic signals represent transient mechanical stress waves excited by arc discharge within the conductor medium, and the airborne acoustic signals represent pressure waves generated near the monitoring points by the interaction between the arc plasma channel and the surrounding air; identifying suspected fault transient disturbance events and target signal segments corresponding to the suspected fault transient disturbance events based on the structural acoustic signals in the multimodal acoustic signals, wherein the suspected fault transient disturbance events represent transient disturbance events suspected to be caused by high-resistivity grounding arc discharge; extracting features from the target signal segments to obtain fused fault acoustic signature features, wherein the fused fault acoustic signature features are obtained based on the structural acoustic fault features and airborne fault features of the target signal segments; and determining the high-resistivity grounding arc feature identification result of the transmission line under test based on the fused fault acoustic signature features.

[0007] According to another aspect of the present invention, a high-resistivity grounding arc feature identification device based on acoustic signature recognition is also provided, comprising: a signal acquisition module for acquiring multimodal acoustic signals at monitoring points on the transmission line under test, wherein the multimodal acoustic signals include structural acoustic signals and airborne acoustic signals collected at the monitoring points at multiple sampling periods, wherein the structural acoustic signals represent transient mechanical stress waves excited by arc discharge within the conductor medium, and the airborne acoustic signals represent pressure waves generated near the monitoring points by the interaction between the arc plasma channel and the surrounding air; a disturbance identification module for identifying suspected fault transient disturbance events and target signal segments corresponding to the suspected fault transient disturbance events based on the structural acoustic signals in the multimodal acoustic signals, wherein the suspected fault transient disturbance events represent transient disturbance events suspected to be caused by high-resistivity grounding arc discharge; a feature extraction module for extracting features from the target signal segments to obtain fused fault acoustic signature features, wherein the fused fault acoustic signature features are obtained based on the structural acoustic fault features and airborne fault features of the target signal segments; and a fault identification module for determining the high-resistivity grounding arc feature identification result of the transmission line under test based on the fused fault acoustic signature features.

[0008] According to another aspect of the present invention, a non-volatile storage medium is also provided, which stores multiple instructions, any one of which is adapted to be loaded and executed by a processor for a high-resistance grounding arc feature identification method based on voiceprint recognition.

[0009] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement any one of the following high-resistance grounding arc feature identification methods based on voiceprint recognition.

[0010] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps of any one of the methods for identifying high-resistance grounding arc features based on voiceprint recognition.

[0011] In this embodiment of the invention, multimodal acoustic signals at monitoring points on the transmission line under test are acquired. These multimodal acoustic signals include structural acoustic signals and airborne acoustic signals collected at the monitoring points during multiple sampling periods. The structural acoustic signal represents the transient mechanical stress wave excited by arc discharge within the conductor medium, and the airborne acoustic signal represents the pressure wave generated near the monitoring point by the interaction between the arc plasma channel and the surrounding air. Based on the structural acoustic signal in the multimodal acoustic signal, suspected fault transient disturbance events and target signal segments corresponding to these events are identified. The suspected fault transient disturbance event represents a transient disturbance event suspected to be caused by a high-resistance grounding arc discharge. The target signal... Feature extraction is performed on the signal segment to obtain fused fault acoustic features. These fused fault acoustic features are based on the structural acoustic fault features and airborne acoustic fault features of the target signal segment. Based on the fused fault acoustic features, the high-resistance grounding arc feature identification result of the transmission line under test is determined. This achieves the goal of accurately identifying high-resistance grounding faults using a multi-modal acoustic feature identification method that integrates structural acoustic and airborne acoustic signals. This enables accurate, reliable, and low-false-alarm early detection and identification of high-resistance grounding faults, effectively improving the fault diagnosis capability and safety of the power system. Furthermore, it solves the technical problem that related technologies cannot effectively distinguish between real arc discharge in high-resistance grounding and disturbances during normal system operation. Attached Figure Description

[0012] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0013] Figure 1 This is a flowchart of a high-resistivity grounding arc feature identification method based on voiceprint recognition according to an embodiment of the present invention;

[0014] Figure 2 This is a schematic diagram of a high-resistivity grounding arc feature identification device based on voiceprint recognition according to an embodiment of the present invention. Detailed Implementation

[0015] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0016] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0017] According to an embodiment of the present invention, a method embodiment for identifying high-resistance grounding arc features based on voiceprint recognition is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0018] Figure 1 This is a flowchart of a high-resistivity grounding arc feature identification method based on voiceprint recognition according to an embodiment of the present invention, as follows: Figure 1 As shown, the method includes the following steps:

[0019] Step S102: Obtain multimodal acoustic signals at monitoring points on the transmission line under test. The multimodal acoustic signals include structural acoustic signals and airborne acoustic signals collected at monitoring points during multiple sampling periods. The structural acoustic signals represent transient mechanical stress waves excited by arc discharge within the conductor medium, and the airborne acoustic signals represent pressure waves generated near the monitoring point by the interaction between the arc plasma channel and the surrounding air.

[0020] Optionally, monitoring points can be selected on the transmission line to be inspected, and two types of sensors—structural acoustic sensors and airborne acoustic sensors—can be deployed to collect structural acoustic signals and airborne acoustic signals, respectively. Structural acoustic signals are mechanical stress waves propagating through the conductor medium, generated by arc discharge within the conductor; while airborne acoustic signals are pressure waves generated by the interaction between the arc plasma channel and the surrounding air, propagating in the air. This step utilizes the different propagation characteristics of the acoustic signals to capture acoustic evidence of arc discharge from multiple angles, providing rich raw data for subsequent feature analysis and fault diagnosis.

[0021] Step S104: Based on the structure acoustic signal in the multimodal acoustic signal, identify suspected fault transient disturbance events and the target signal segment corresponding to the suspected fault transient disturbance events. The suspected fault transient disturbance event refers to a transient disturbance event suspected to be caused by high-resistance grounding arc discharge.

[0022] Optionally, based on the structure-based acoustic signal obtained in step S102, a dynamic threshold is set by analyzing the short-time energy changes of the signal to identify transient disturbance events suspected to be caused by high-resistance grounding arc discharge. The target signal segment is the signal segment corresponding to the suspected fault transient disturbance event in the multi-modal acoustic signal, which can be the signal segment before and after the suspected fault transient disturbance event occurs, for more detailed analysis in subsequent steps. This approach can improve the ability to capture weak arc discharge events while reducing the possibility of false triggering.

[0023] In one optional embodiment, identifying suspected fault transient disturbance events and target signal segments corresponding to the suspected fault transient disturbance events based on structure acoustic signals in multimodal acoustic signals includes: dividing the structure acoustic signals in multimodal acoustic signals into multiple frames according to a preset sliding time window to obtain multiple frames of structure acoustic signals; determining the short-time energy corresponding to each of the multiple frames of structure acoustic signals, wherein the short-time energy represents the energy level of the structure acoustic signal within the preset sliding time window; comparing the short-time energy corresponding to each of the multiple frames of structure acoustic signals with the corresponding trigger threshold to obtain a comparison result; determining the suspected fault transient disturbance event and the trigger time corresponding to the suspected fault transient disturbance event based on the comparison result; and determining the target signal segment from the multimodal acoustic signals based on the trigger time.

[0024] Optionally, the acquired continuous structural acoustic signal is divided into multiple short-time segments, or "frames," using a sliding time window. Each frame represents acoustic activity information within a specific time window. The size of this time window and the sliding step are preset to capture the transient characteristics of the acoustic signal and facilitate subsequent feature extraction and analysis. For each frame of structural acoustic signal, its short-time energy is calculated. Short-time energy is a measure of the signal's energy over a certain time period, reflecting the signal's activity level within that interval. In power fault detection scenarios, especially for detecting high-resistance grounding arc discharge, changes in short-time energy are crucial indicators. When an arc discharge occurs, a significant energy release occurs within a short period, resulting in a significant jump in the short-time energy of the structural acoustic signal. Next, the short-time energy of each frame of structural acoustic signal is compared with a preset trigger threshold. The comparison result determines whether the current signal frame is considered a suspicious fault event signal. Through the above comparison, if the short-time energy of a signal frame exceeds the trigger threshold, this signal frame is marked as a suspected fault transient disturbance event, and the trigger time, i.e., the exact position of the signal frame on the time axis, is recorded. This moment is crucial for subsequent signal analysis because it indicates the starting point of the signal segment that needs further examination, helping to pinpoint the instant the fault occurred. Finally, a signal segment containing the trigger time is automatically extracted from a certain time range before and after the trigger time; this signal segment is defined as the target signal segment. The length of the target signal segment is preset; it contains complete acoustic information before and after the fault event, used for subsequent feature analysis and fault type confirmation. A key consideration in selecting the target signal segment is ensuring that enough information is captured to extract acoustic signals characterizing the fault from airborne and structure-based sound signals, while also controlling the signal segment length to avoid introducing too much irrelevant background noise information, which could affect the accuracy of subsequent analysis.

[0025] Using the above methods, it is possible to accurately identify transient disturbances suspected to be caused by high-resistivity grounding arc discharge from massive amounts of acoustic signal data, and to pinpoint key signal segments containing fault characteristics, providing a foundation for subsequent fault feature extraction and diagnostic decisions. This approach fully utilizes the locational advantages of structure acoustic signals on the time axis, combined with their sensitive characteristics in frequency and energy distribution, thereby improving the sensitivity and accuracy of fault detection.

[0026] In an optional embodiment, before comparing the short-time energy corresponding to each of the multi-frame structured acoustic signals with the corresponding trigger threshold to obtain the comparison result, the method further includes: obtaining the short-time energy corresponding to each of the predetermined frame structured acoustic signals preceding any frame structured acoustic signal; determining the mean and standard deviation of the short-time energy corresponding to each of the predetermined frame structured acoustic signals; obtaining the trigger threshold corresponding to any frame structured acoustic signal based on the mean and standard deviation; and obtaining the trigger threshold corresponding to each of the multi-frame structured acoustic signals by using the method of obtaining the trigger threshold corresponding to any frame structured acoustic signal.

[0027] Optionally, since background noise levels may change with time, weather, or other external factors, a fixed threshold may not be flexible or adaptable enough, easily leading to over- or under-triggered triggering. Therefore, a dynamic trigger threshold is set. First, short-time energy of a series of predetermined frames (e.g., frames within the first few seconds) preceding each frame of structured acoustic signal is acquired to capture the trends in background noise and signal strength in the most recent time period. Next, the mean and standard deviation of the short-time energy of these predetermined frames are calculated. The mean reflects the average level of background noise and signal strength, while the standard deviation measures the range of signal strength fluctuations. A new trigger threshold is generated based on the calculated mean and standard deviation. This threshold setting mechanism can automatically adjust to changes in ambient noise, improving the robustness of fault detection. Subsequently, a similar process is used to generate a corresponding trigger threshold for each frame of structured acoustic signal. This means that the threshold used for each comparison is dynamically adjusted according to the latest signal background conditions, rather than being fixed.

[0028] Optionally, but not limited to, short-term energy can be obtained through the following methods: Where n is the index of the current frame structure acoustic signal, is a window function, and m is the index of the signal sampling point.

[0029] Optionally, but not limited to, the trigger threshold corresponding to any frame structure acoustic signal can be obtained through the following methods: ,in, This represents the trigger threshold corresponding to any frame of structural acoustic signal. This represents the historical average value calculated based on the short-time energy sequence of multiple preceding frames (i.e., the average short-time energy of each corresponding acoustic signal in the predetermined frame structure). This represents the standard deviation calculated based on the short-time energy sequence of multiple preceding frames (i.e., the standard deviation of the short-time energy corresponding to each of the predefined frame structure acoustic signals); k is a preset sensitivity coefficient used to balance detection sensitivity and anti-interference capability.

[0030] By employing the above methods, high-resistance grounding arc discharge events can be identified more accurately in constantly changing environments, while reducing the risk of false triggering. The dynamic threshold update strategy better adapts to environmental noise levels at different times and locations, ensuring that fault detection is both sensitive and reliable.

[0031] Step S106: Extract features from the target signal segment to obtain fused fault acoustic features, wherein the fused fault acoustic features are obtained based on the structural acoustic fault features and airborne acoustic fault features of the target signal segment.

[0032] Optionally, feature extraction includes independent analysis of structure-based acoustic signals and airborne acoustic signals, aiming to capture specific patterns associated with high-impedance grounding faults in both types of signals. For structure-based acoustic signals, features that can be extracted, but are not limited to, include short-time energy, zero-crossing rate, spectral centroid, and Mel-frequency cepstral coefficients (MFCCs), which describe the signal's time-domain, frequency-domain, and time-frequency-domain characteristics. For airborne acoustic signals, sound pressure level and harmonic spectral clarity (HCD) can be extracted as features; these metrics help identify whether the signal exhibits typical acoustic signatures of arcing. By fusing the extracted structure-based and airborne acoustic fault features, a fused fault acoustic signature feature vector is formed to comprehensively characterize the acoustic features of suspected fault events.

[0033] Optionally, structural acoustic fault features, airborne acoustic fault features, and fused fault acoustic features can all be in vector form.

[0034] In one optional embodiment, feature extraction is performed on the target signal segment to obtain fused fault acoustic signature features, including: feature extraction of the structured acoustic signal in the target signal segment to obtain a structured acoustic fault feature set, wherein the structured acoustic fault feature set includes short-time energy, zero-crossing rate, spectral centroid in the frequency domain, and Mel-frequency cepstral coefficients in the time-frequency domain. Short-time energy represents the energy level of the structured acoustic signal within a preset sliding time window, zero-crossing rate represents the number of times the structured acoustic signal crosses zero points per unit time, spectral centroid represents the center position of the frequency distribution of the structured acoustic signal, and Mel-frequency cepstral coefficients are used to capture the time-frequency characteristics of the structured acoustic signal; feature extraction is performed on the airborne acoustic signal in the target signal segment to obtain an airborne acoustic fault feature set, wherein the airborne acoustic fault feature set includes sound pressure level and spectral harmonic clarity, sound pressure level represents the decibel level of the airborne acoustic signal relative to the reference sound pressure, and spectral harmonic clarity represents the relative significance of harmonic components and fundamental components in the spectrum of the airborne acoustic signal; the structured acoustic fault feature set and the airborne acoustic fault feature set are spliced ​​together end-to-end in a preset order to obtain fused fault acoustic signature features.

[0035] Optionally, short-time energy represents the short-time energy level of the structural acoustic signal within a preset sliding window, reflecting the intensity change of transient mechanical stress waves caused by arc discharge, and can be considered a probability indicator of fault events. The zero-crossing rate indicator signal represents the number of times the signal crosses zero within a unit of time. By analyzing the zero-crossing rate of the structural acoustic signal, the activity level of the signal can be identified, thereby determining whether transient events such as arc discharge have occurred, as mechanical vibrations caused by arc discharge typically exhibit a high zero-crossing rate. The spectral centroid represents the center position of the frequency domain distribution of the structural acoustic signal, i.e., the frequency bias of the signal energy distribution, which helps distinguish different types of transient vibration events, as the mechanical stress waves excited by different fault events will have different frequency characteristics. Mel frequency cepstral coefficients (MFCC) represent a series of coefficients obtained through processes such as pre-emphasis, framing, Fourier transform, Mel filtering, and discrete cosine transform of the structural acoustic signal, used to capture the time-frequency characteristics of the signal. They are particularly suitable for characterizing nonlinear and non-stationary signal characteristics and help characterize the time-frequency features of complex transient events such as arc discharge. Sound pressure level (SPL) represents the decibel (dB) level of an airborne acoustic signal relative to a reference sound pressure level. It reflects the intensity of the pressure wave generated by the interaction between the arc plasma channel and the surrounding air, and is a crucial indicator for measuring the intensity of airborne acoustic signals, essential for determining whether an arc discharge event has occurred. Spectral harmonic clarity represents the relative significance of harmonic components to the fundamental component in the airborne acoustic signal spectrum, i.e., the degree of prominence of the signal's harmonic characteristics. Since the acoustic signals of arc discharge typically contain abundant harmonic components, this indicator helps to further confirm the authenticity and type of the fault event. The extracted structural acoustic fault feature sets and airborne acoustic fault feature sets are concatenated in a predetermined order to form a fused fault acoustic signature. This fusion process aims to create a comprehensive feature vector that contains complementary information from both acoustic signals, providing a more comprehensive description of the acoustic characteristics of potential fault events and offering a richer and more refined data foundation for subsequent classification and identification.

[0036] Optional, zero-crossing rate It can be obtained in the following way: Where N is the frame length of the target signal segment, sgn[ ] is the sign function, m is used to identify the m-th sampling point in the target signal segment, and m-1 is used to identify the (m-1)-th sampling point in the target signal segment.

[0037] Spectral centroid in the frequency domain It can be obtained in the following way: ,in, It is the signal frame in the target signal segment The amplitude spectrum of the Discrete Fourier Transform (DFT). It is the frequency value corresponding to the frequency index k.

[0038] Sound pressure level (SPL) can be obtained as follows: ,in, It is the effective value of the signal frame. It is the reference sound pressure in the air, with a value of 20 micropascals (uPa).

[0039] The fundamental frequency can be found in the spectrum using peak detection algorithms. Then, the spectral harmonic clarity is obtained in the following way. ,in, It is the sum of energy within a narrowband region that is an integer multiple of the fundamental frequency. It is the total energy of the entire signal frame (i.e., the target signal segment). The higher the HCD, the more pronounced the harmonic characteristics of the signal.

[0040] By employing the above methods, the intrinsic characteristics of structured sound and airborne sound signals within the target signal segment can be deeply mined from multiple perspectives, including the time domain, frequency domain, and time-frequency domain. Furthermore, through feature fusion, a comprehensive fused fault acoustic signature feature representing high-resistance grounding arc discharge events can be constructed. This integrated feature not only improves the accuracy of fault type identification but also enhances the robustness and generalization ability of the algorithm, enabling it to effectively cope with complex and ever-changing transmission line environments, reduce misjudgments and omissions, and provide strong technical support for the safe operation of the power grid.

[0041] Step S108: Based on the fusion of fault acoustic signature features, determine the high-resistance grounding arc feature identification result of the transmission line under test.

[0042] Optionally, a pre-trained high-resistance grounding arc feature recognition model can be used to classify and identify the fused fault acoustic features obtained in step S106, thereby determining the high-resistance grounding arc feature recognition result of the transmission line under test. This high-resistance grounding arc feature recognition model can be trained based on a machine learning algorithm. After receiving the fused feature vector input, it can output a classification result and a relevant confidence score to determine whether the event is a high-resistance grounding arc discharge, ensuring the accuracy and reliability of the diagnostic results and effectively addressing the limitations of a single information source in fault diagnosis.

[0043] In one optional embodiment, the high-resistance grounding arc feature identification result of the transmission line under test is determined based on the fused fault acoustic features, including: based on the fused fault acoustic features, a high-resistance grounding arc feature identification model is used to obtain the high-resistance grounding arc feature identification result; wherein, the high-resistance grounding arc feature identification model is obtained by training multiple initial learners using the gradient boosting decision tree method based on multiple sets of historical fused fault acoustic features and corresponding fault feature identification results, and the corresponding fault feature identification result includes at least: normal operation state, multiple high-resistance grounding fault states, and multiple conventional interference states, wherein the conventional interference states are other interference states besides the high-resistance grounding fault state.

[0044] Optionally, Gradient Boosting Decision Tree (GBDT) is a powerful ensemble learning algorithm that constructs a series of weak learners (i.e., decision trees) and weights and combines the predictions of these decision trees to generate a final strong learning prediction model. This approach can handle high-dimensional features and performs well in complex tasks involving the integration of numerous different data sources, such as identifying potential fault modes in multimodal acoustic signals. The model is trained on a large amount of historical data, including fused fault acoustic signature features of multimodal acoustic signals, i.e., a combination of features from structured acoustic signals and airborne acoustic signals, covering signal samples collected under different fault and interference conditions. Each set of fused fault acoustic signature features is accompanied by its corresponding fault feature identification results. These results include not only signals under normal operating conditions but also signal samples under various high-resistivity grounding fault conditions, as well as signals generated by other interference sources (such as lightning strikes, bird activity, wind vibration, etc.). This comprehensive labeled data ensures that the model can learn effectively under various operating conditions. This high-resistivity grounding arc feature identification model is trained using the GBDT method. During training, the algorithm iteratively optimizes the structure of the decision tree to minimize prediction errors on the training data. By continuously correcting errors in the previous tree, GBDT can build highly complex models that achieve good performance even when there are highly nonlinear relationships between features. After training, the model can accept real-time acquired fused fault acoustic signature features as input and output a fault diagnosis result, including specific fault type identification (such as distinguishing between high-resistance grounding faults and normal operating conditions) and possible interference state classifications. This result is accompanied by a confidence score to evaluate the reliability of the diagnosis.

[0045] By employing a GBDT model trained on historical data as the high-resistivity grounding arc feature recognition model, an accurate mapping from fused fault acoustic signature features to high-resistivity grounding arc feature recognition results can be achieved, significantly improving the intelligence and accuracy of fault diagnosis. This approach can not only effectively distinguish various fault states but also identify and eliminate common interference sources.

[0046] Optionally, a comprehensive training sample library with operating condition labels can be collected and constructed. The operating condition samples should fully cover the normal operation of the line, as well as various high-resistance grounding fault states and common interference states formed by typical media such as dry branches, wet branches, and insulator flashover. For each sample in the sample library, the method described in the aforementioned embodiment is executed to extract its fused fault acoustic signature feature vector. This allows us to construct a training dataset consisting of N samples with precisely paired "features and labels". , where y i Let be the work condition label for the i-th sample. The GBDT algorithm is used for supervised learning on the training dataset D. This process iteratively constructs a series of weak learners (decision trees), each new learner aiming to fit the residuals of the ensemble model of all previous learners. Finally, a strong learner model M is generated through weighted summation, with the goal of minimizing the loss function. This leads to a parameter-fixed high-resistivity grounding arc feature recognition model that can accurately map from the fused fault acoustic signature feature vector to the specific operating condition category.

[0047] In one optional embodiment, based on the fused fault acoustic signature features, a high-resistivity grounding arc feature recognition model is used to obtain the high-resistivity grounding arc feature recognition result, including: based on the fused fault acoustic signature features, a high-resistivity grounding arc feature recognition model is used to obtain an initial fault recognition result and a corresponding confidence score; based on the initial fault recognition result, the confidence score, and the target signal segment, verification is performed to obtain a verification result; if the verification result indicates that the verification is successful, the initial fault recognition result is used as the high-resistivity grounding arc feature recognition result.

[0048] Optionally, the real-time acquired fused fault acoustic signature features are input into a pre-trained model. Based on its learned knowledge and patterns, the model outputs an initial fault identification result and a confidence score associated with that result. The initial fault identification result is a basic judgment of the fault type of the signal segment, while the confidence score reflects the model's confidence in this judgment. The confidence score can be a probability value between 0 and 1; a higher value indicates that the model is more confident that the identification result is correct. The subsequent validation phase aims to further ensure the reliability of the confidence result. This phase performs multi-dimensional cross-validation by combining the model's output initial fault identification result, the confidence score, and the previously identified target signal segment (including structured sound and airborne sound signals). The purpose of validation is to ensure that the fault identified by the model is indeed caused by arc discharge, rather than other interference sources, and that the features in the signal segment do indeed match the fault type identified by the model. If the validation result indicates that the validation has passed, it means that the model's initial fault identification result has been confirmed, and this identification result will be adopted as the final high-resistance grounding arc feature identification result. At this stage, the identification result is considered reliable and can be used for subsequent fault warning, location, and handling decisions. If the verification fails, the signal may need to be re-analyzed, or additional evidence may need to be awaited to make a final judgment in order to avoid taking unnecessary actions based on unreliable identification results.

[0049] By employing the above methods, not only is fault identification based on machine learning models, but an additional verification step is also added to ensure the accuracy and reliability of the identification results. This approach, through multi-level decision-making logic, can significantly improve the ability to identify high-resistance grounding arc characteristics of transmission lines in complex environments, reducing false alarms and missed alarms.

[0050] In one optional embodiment, verification is performed based on the initial fault identification result, confidence score, and target signal segment to obtain a verification result, including: determining a first verification condition as follows: the effective sound pressure level of the airborne acoustic signal within the target signal segment reaches a predetermined significant level relative to the average effective sound pressure level under historical background noise levels; determining a second verification condition as follows: the start time of the transient disturbance detected in the structure acoustic signal and the airborne acoustic signal within the target signal segment is consistent within a preset error allowable range; determining a third verification condition as follows: the initial fault identification result is one of the preset fault categories, and the confidence score is greater than a preset confidence threshold; if the first verification condition, the second verification condition, and the third verification condition are all satisfied, the verification result is determined to be verified successfully; or if at least one of the first verification condition, the second verification condition, and the third verification condition is not satisfied, the verification result is determined to be verified unsuccessfully.

[0051] Optionally, the first verification condition can be understood as a significant sound pressure level verification (or energy jump verification), that is, verifying whether the effective sound pressure level of the airborne sound signal within the target signal segment is significantly higher than the average effective sound pressure level under historical background noise levels. This condition ensures that the detected sound signal has sufficient intensity to distinguish it from everyday environmental noise, indicating that an abnormal acoustic event has indeed occurred. By comparing the sound pressure level of the target signal segment with the background noise level, the possibility of false alarms can be initially ruled out, ensuring that the identification of acoustic events is based on significant changes in signal intensity. The first verification condition can be set in the following form: ,in, This represents the effective value of the airborne sound signal within the target signal segment. The effective baseline value represents the historical background noise.

[0052] The second verification condition can be understood as acoustic signal time alignment verification (or time concurrency verification), that is, confirming that the start time of the transient disturbance detected by the structure-based acoustic signal and the airborne acoustic signal within the target signal segment is consistent within a preset error allowable range. This step further verifies the physical consistency of the disturbance event by comparing the correlation of the two different modes of acoustic signals in the time series. Verifying the time consistency between the structure-based acoustic signal and the airborne acoustic signal is crucial to ensuring that the event is caused by the same physical source (such as arc discharge). If the time alignment of the two signals is good, the authenticity and correlation of the event are enhanced. The second verification condition can be set in the following form: ,in, and These are the start times of the structure-sound and air-sound transient signals, calculated using methods such as kurtosis maximization.

[0053] The third verification condition can be understood as model judgment and confidence threshold verification. This requires that the model's initial fault identification result belongs to one of the preset fault categories, and that the confidence score exceeds a preset confidence threshold. This condition ensures that the model's identification result is both within a reasonable range of fault types and has sufficient credibility, avoiding uncertainty or ambiguity in the model's output. The confidence score of the model output is an important indicator for evaluating the reliability of the identification result. A confidence score higher than the preset threshold means that the model is very confident in the identification result, increasing the robustness of the decision.

[0054] When all three verification conditions are met simultaneously, the verification result is considered passed. Combining the first, second, and third verification conditions, the final diagnostic conclusion of "effective high-resistance grounding fault" is determined as follows: ,in, This is the set of all fault types. Verification indicates that both the first and second verification conditions have been passed. This means that the target signal segment is not only significant in acoustic intensity but also highly consistent with the model's predictions in terms of time and fault type, further confirming the validity of the identification results. If any of the three verification conditions are not met, the verification result is deemed unsuccessful. This indicates that the information in the signal segment is insufficient to support the model's diagnosis, or that there are other interpretations, requiring further analysis or monitoring to avoid making decisions based on uncertain information.

[0055] By employing the above methods, the fault identification results of the model can be evaluated more meticulously and comprehensively, ensuring that the identification result is only adopted when the acoustic characteristics of the signal segment match the fault type predicted by the model and the confidence level is sufficiently high. This multi-condition verification mechanism can significantly enhance the accuracy and robustness of diagnosis, and reduce the risk of false alarms and false negatives.

[0056] Through the above steps S102 to S108, the goal of accurately determining high-resistance grounding faults can be achieved by using a multimodal acoustic fingerprint recognition method that integrates structured sound and airborne sound signals. This enables accurate, reliable, and low-false-alarm early detection and identification of high-resistance grounding faults, effectively improving the fault diagnosis capability and safety of the power system. Furthermore, it solves the technical problem that related technologies are unable to effectively distinguish between real arc discharge in high-resistance grounding and disturbances during normal system operation.

[0057] High-resistance grounding faults, due to their weak fault current and concealed electrical characteristics, are among the most difficult-to-detect persistent faults in power systems. They not only seriously threaten the lives and property of users but can also escalate into large-scale power outages, making them a key technological bottleneck that must be overcome to ensure the safe operation of the power grid. Existing detection technologies, whether relying on electrical quantity analysis or single acoustic sensing (airborne or structural sound), cannot effectively balance the trade-off between detection speed and reliability, leading to frequent false alarms and missed detections in complex field environments, thus failing to meet actual operation and maintenance needs.

[0058] To address the aforementioned problems, and based on the above embodiments and optional embodiments, this invention proposes an optional implementation method for identifying high-resistivity grounding arc features based on voiceprint recognition. This method includes:

[0059] Step S100: Synchronous acquisition and coupling of multimodal acoustic signals, specifically including:

[0060] One or more monitoring points are selected on the transmission line to be monitored, and a first sensing device and a second sensing device are deployed at each monitoring point. The first sensing device is a structure acoustic sensor, whose sensing end forms a stable mechanical coupling with the transmission line body or its suspension fasteners, designed to capture transient mechanical stress waves excited by arc discharge within the conductor medium, i.e., structure acoustic signals. The second sensing device is a micro-electromechanical system (MEMS) microphone, configured to collect pressure waves generated near the same monitoring point by the interaction between the arc plasma channel and the surrounding air, i.e., airborne acoustic signals. Through a synchronous data acquisition unit, the output signals of the first and second sensing devices are synchronously sampled with a unified sampling clock, generating a multimodal acoustic data stream with strictly aligned timestamps that can be accurately correlated in the time domain.

[0061] Step S200: Fault transient disturbance triggering and data locking based on structure acoustic signals, specifically including:

[0062] Real-time online monitoring and analysis are performed on the structure-acoustic signals in the multimodal acoustic data stream acquired in step S100. This step aims to reliably identify transient disturbances suspected to be caused by faults from background vibrations. Specifically:

[0063] S201: Real-time calculation of short-time energy. Calculation of structure acoustic signals in real-time within a sliding time window. short-term energy The specific method for obtaining the short-term time setting is the same as in the aforementioned embodiments, and will not be repeated here.

[0064] S202: Dynamic threshold determination. An adaptive dynamic trigger threshold is used. This is to adapt to changing background noise. The specific process for obtaining the dynamic trigger threshold is the same as in the previous embodiments, and will not be repeated here.

[0065] S203: Trigger and Lock. When short-term energy is detected... Within a 50mm time window, its value continuously exceeds the dynamic trigger threshold. At that time, the system determined that a suspected transient disturbance event had been triggered. Subsequently, the system used the trigger time as a time reference to extract and lock a target signal segment with a length of 1 second from the multimodal acoustic data stream (0.2 seconds before the trigger time to 0.8 seconds after the trigger time). This data segment completely contains the original signal information of the two channels before and after the occurrence of the disturbance event.

[0066] Step S300: Extraction and construction of fused fault acoustic signature features. This step only targets the signal segment identified in step S200, aiming to construct a feature vector that comprehensively characterizes the physical nature of the transient disturbance. Specifically, it includes:

[0067] S301, Signal Preprocessing. The structure acoustic signal and airborne acoustic signal within the target signal segment are preprocessed separately using a fourth-order digital Butterworth bandpass filter with a passband of 1kHz-20kHz to filter out power frequency electromagnetic interference, low-frequency vibrations such as wind vibration, and high-frequency irrelevant noise.

[0068] S302, Extract the structural acoustic fault feature set. Extract the structural acoustic fault feature set from the preprocessed structural acoustic signal. The feature set includes short-time energy in the time domain. Zero crossing rate Spectral centroid in the frequency domain Mel-frequency cepstral coefficients (MFCC) in the time-frequency domain. Zero-crossing rate. The specific method for obtaining the spectral centroid in the frequency domain is the same as in the aforementioned embodiments, and will not be repeated here.

[0069] S303, Extract the airborne acoustic fault feature set. Extract the airborne acoustic fault feature set from the preprocessed airborne acoustic signal. The feature set specifically includes: Sound Pressure Level (SPL) and Spectral Harmonic Discreteness (HCD). The specific methods for obtaining SPL and HCD are the same as in the aforementioned embodiments and will not be repeated here.

[0070] S304, Feature Vector Fusion. Employing a fusion rule of direct feature vector concatenation, the structural acoustic fault feature set and the airborne acoustic fault feature set are concatenated end-to-end in a preset order to construct a fused fault acoustic signature feature vector V with defined dimensions and a unified structure. f .

[0071] Step S400: Offline training of the fault diagnosis model, specifically including:

[0072] S401, Sample Library Construction. Collect and construct a comprehensive training sample library with operating condition labels. The operating condition samples must fully cover the normal operation of the line, as well as various high-resistance grounding fault conditions and common interference conditions formed by typical media such as dry branches, wet branches, and insulator flashover.

[0073] S402, Training dataset generation. For each sample in the sample library, step S300 is performed to extract its fused fault voiceprint feature vector. This allows us to construct a training dataset consisting of N samples with precisely paired "features and labels". , where y i Let be the working condition label for the i-th sample.

[0074] S403, Model Training. The Gradient Boosting Decision Tree (GBDT) algorithm is used for supervised learning on the training dataset D. This process iteratively constructs a series of weak learners (decision trees), each new learner aiming to fit the residuals of the ensemble model of all previous learners. Finally, a strong learner model M is generated through a weighted summation, with the goal of minimizing the loss function. This allows us to obtain a fault diagnosis model with fixed parameters that can accurately map from the fused fault acoustic signature feature vector to specific operating condition categories.

[0075] Step S5: Based on collaborative decision-making using model inference and cross-verification of acoustic and vibration data, perform online diagnosis and decision-making for transient disturbance events captured in real time. This step is crucial for ensuring the high reliability of the diagnostic conclusions. Specifically, it includes:

[0076] S501, Model Inference. The fused fault acoustic signature feature vector V generated in real-time in step S300 is used... f The input is fed into the fault diagnosis model M generated in step S400 for forward reasoning to obtain a model containing a preliminary fault category determination y. pred and the corresponding confidence score p conf The diagnosis results.

[0077] S502, Perform the acoustic-vibration cross-verification procedure. In parallel, perform the acoustic-vibration cross-verification procedure on this event to verify whether there is a strong physical correlation between the captured structural vibrations and airborne sound. Verification passed. The condition for judgment is that both of the following logical judgments are true at the same time:

[0078] 1) Verification of the energy leap: .

[0079] 2) Time-based concurrent verification: .

[0080] S503, Collaborative Decision Making. Based on pre-defined decision logic, a final ruling is made on the preliminary diagnostic results. The conditions for a final diagnostic conclusion of "effective high-resistance grounding fault" are: .

[0081] Step S600: Generation and hierarchical reporting of structured fault diagnosis conclusions, specifically including:

[0082] Based on the final decision in step S500, a structured diagnostic conclusion that can guide operation and maintenance work is generated and reported. Specifically, this includes:

[0083] S601, Level 1 alarm generated. If the event is confirmed as a valid high-resistance grounding fault, a Level 1 alarm conclusion is generated. This conclusion explicitly includes: fault nature determination (high-resistance grounding), fault characteristic subtype determined by the model (e.g., tree branch sling discharge type), fault transient start and end timestamps, discharge intermittency and intensity assessment, and optionally, preliminary spatial orientation of the fault source calculated by the airborne acoustic sensor array.

[0084] S602, Level 2 warning generated. If an event does not meet all the conditions for a Level 1 alarm, it will be classified as a Level 2 warning event or a harmless interference event based on the sub-conditions it meets, and a corresponding status report will be generated for operation and maintenance personnel to pay attention to or file.

[0085] S603, Conclusion Reporting. Through the 5G wireless communication module, the conclusion of the first-level alarm or the report of the second-level early warning event is actively pushed to the remote alarm interface of the power grid dispatch automation system, namely the Supervisory Control and Data Acquisition (SCADA) system, in the form of encrypted data packets. This provides accurate and actionable decision-making intelligence for subsequent power grid dispatching, fault isolation, and line inspection and repair.

[0086] It should be noted that this embodiment proposes a method for distinguishing high-resistance grounding arc characteristics based on the fusion of structural acoustic and airborne acoustic signals. This method solves the problem of fault detection methods in related technologies that rely on a single electrical quantity or a single acoustic sensor, which are susceptible to interference from power grid background noise and non-fault events, leading to protection misjudgments or failures to operate. Its core lies in synchronously acquiring structural acoustic and airborne acoustic signals of the line, combined with dynamic threshold triggering technology, to reliably identify and lock the fault transient disturbance time window from the structural acoustic signal. By analyzing the fused fault acoustic signature characteristics of the two acoustic signals within this time window, a machine learning model is used to determine the fault type and characteristics. Furthermore, this embodiment addresses the difficulty of effectively distinguishing between actual arc discharge and normal system operation disturbances in related technologies. It proposes a collaborative decision-making mechanism that combines model reasoning with acoustic-vibration cross-verification, optimizing the feature extraction process, fusion strategy, and decision logic. This overcomes the shortcomings of information ambiguity in a single sensing dimension, effectively improving the accuracy and reliability of high-resistance grounding fault diagnosis. This method has the advantages of sensitive triggering, strong anti-interference ability, clear diagnostic criteria, and independence from power grid operation mode. It is widely applicable to the accurate identification and early warning of high-resistance grounding faults with high concealment and high risk in various power transmission and distribution networks.

[0087] This embodiment also provides a high-resistance grounding arc feature identification device based on voiceprint recognition. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the terms "module" and "device" can refer to a combination of software and / or hardware that performs a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0088] According to embodiments of the present invention, an apparatus embodiment for implementing the above-described high-resistivity grounding arc feature identification method based on voiceprint recognition is also provided. Figure 2 This is a schematic diagram of a high-resistance grounding arc feature identification device based on voiceprint recognition according to an embodiment of the present invention, as shown below. Figure 2 As shown, the above-mentioned high-resistance grounding arc feature identification device based on voiceprint recognition includes: a signal acquisition module 200, a disturbance identification module 202, a feature extraction module 204, and a fault identification module 206, wherein:

[0089] The signal acquisition module 200 is used to acquire multimodal acoustic signals at monitoring points on the transmission line under test. The multimodal acoustic signals include structural acoustic signals and airborne acoustic signals collected at the monitoring points at multiple sampling periods. The structural acoustic signals represent transient mechanical stress waves excited by arc discharge in the conductor medium, and the airborne acoustic signals represent pressure waves generated near the monitoring point by the interaction between the arc plasma channel and the surrounding air.

[0090] The disturbance identification module 202 is connected to the signal acquisition module 200 and is used to identify suspected fault transient disturbance events and the target signal segments corresponding to the suspected fault transient disturbance events based on the structure acoustic signals in the multimodal acoustic signals. The suspected fault transient disturbance event refers to a transient disturbance event suspected to be caused by high-resistance grounding arc discharge.

[0091] The feature extraction module 204 is connected to the disturbance recognition module 202 and is used to extract features from the target signal segment to obtain fused fault acoustic features. The fused fault acoustic features are obtained based on the structural acoustic fault features and airborne acoustic fault features of the target signal segment.

[0092] The fault identification module 206 is connected to the feature extraction module 204 and is used to determine the high-resistance grounding arc feature identification result of the transmission line under test based on the fused fault acoustic features.

[0093] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0094] It should be noted that the signal acquisition module 200, disturbance identification module 202, feature extraction module 204, and fault identification module 206 mentioned above correspond to steps S102 to S108 in the embodiments. The instances and application scenarios implemented by the above modules and their corresponding steps are the same, but they are not limited to the content disclosed in the above embodiments. It should be noted that the above modules, as part of the device, can run on a computer terminal.

[0095] It should be noted that the optional or preferred implementation methods of this embodiment can be found in the relevant descriptions in the embodiments, and will not be repeated here.

[0096] The aforementioned high-resistance grounding arc feature identification device based on voiceprint recognition may also include a processor and a memory. The aforementioned signal acquisition module 200, disturbance identification module 202, feature extraction module 204, fault identification module 206, etc., are all stored in the memory as program modules, and the processor executes the aforementioned program modules stored in the memory to realize the corresponding functions.

[0097] The processor contains a core that retrieves the corresponding program modules from memory. One or more cores may be configured. Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory includes at least one memory chip.

[0098] According to an embodiment of this application, an embodiment of a non-volatile storage medium is also provided. Optionally, in this embodiment, the non-volatile storage medium includes a stored program, wherein, when the program runs, it controls the device containing the non-volatile storage medium to execute any of the aforementioned high-resistance grounding arc feature identification methods based on voiceprint recognition.

[0099] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals, and the non-volatile storage medium includes stored programs.

[0100] Optionally, during program execution, a program controls the device containing the non-volatile storage medium to execute any of the steps of the high-resistance grounding arc feature identification method based on voiceprint recognition.

[0101] According to an embodiment of this application, an embodiment of a processor is also provided. Optionally, in this embodiment, the processor is used to run a program, wherein the program executes any of the above-described high-resistance grounding arc feature identification methods based on voiceprint recognition.

[0102] According to an embodiment of this application, an embodiment of a computer program product is also provided, which, when executed on a data processing device, is adapted to execute a program that initializes the steps of the high-resistivity grounding arc feature identification method based on voiceprint recognition, which includes any of the steps described above.

[0103] Optionally, when the above-mentioned computer program product is executed on a data processing device, it is suitable for executing a program that initializes the steps of the high-resistance grounding arc feature identification method based on voiceprint recognition, which includes any of the above steps.

[0104] This invention provides an electronic device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of any of the above-described methods for identifying high-resistance grounding arc features based on voiceprint recognition.

[0105] The order of the above embodiments of the present invention is merely for description and does not represent the superiority or inferiority of the embodiments.

[0106] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0107] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of modules described above can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between modules, and may be electrical or other forms.

[0108] The modules described above as separate components may or may not be physically separate. Similarly, the components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple modules. Some or all of the modules can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0109] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0110] If the aforementioned integrated modules are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable non-volatile storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a non-volatile storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned non-volatile storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0111] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for identifying high-resistivity grounding arc features based on voiceprint recognition, characterized in that, include: The multimodal acoustic signal at the monitoring point on the transmission line under test is acquired. The multimodal acoustic signal includes structural acoustic signal and airborne acoustic signal collected at the monitoring point at multiple sampling time periods. The structural acoustic signal represents the transient mechanical stress wave excited by the arc discharge in the conductor medium. The airborne acoustic signal represents the pressure wave generated near the monitoring point by the interaction between the arc plasma channel and the surrounding air. Based on the structure acoustic signal in the multimodal acoustic signal, suspected fault transient disturbance events and the target signal segment corresponding to the suspected fault transient disturbance events are identified, wherein the suspected fault transient disturbance event represents a transient disturbance event suspected to be caused by high-resistance grounding arc discharge. Feature extraction is performed on the target signal segment to obtain fused fault acoustic features, wherein the fused fault acoustic features are obtained based on the structure acoustic fault features and airborne acoustic fault features of the target signal segment; Based on the fused fault acoustic signature features, the high-resistance grounding arc feature identification result of the transmission line under test is determined, including: based on the fused fault acoustic signature features, a high-resistance grounding arc feature identification model is used to obtain the high-resistance grounding arc feature identification result; wherein, the high-resistance grounding arc feature identification model is obtained by training multiple initial learners using a gradient boosting decision tree method based on multiple sets of historical fused fault acoustic signature features and corresponding fault feature identification results, and the corresponding fault feature identification results include at least: normal operation state, multiple high-resistance grounding fault states, and multiple conventional interference states, wherein the conventional interference states are other interference states besides the high-resistance grounding fault state.

2. The method according to claim 1, characterized in that, The step of identifying suspected fault transient disturbance events and the target signal segments corresponding to the suspected fault transient disturbance events based on the structure acoustic signals in the multimodal acoustic signals includes: The structure acoustic signal in the multimodal acoustic signal is divided into multiple frames according to a preset sliding time window to obtain multiple frames of structure acoustic signal; Determine the short-time energy corresponding to each of the multi-frame structured acoustic signals, wherein the short-time energy represents the energy level of the structured acoustic signal within the preset sliding time window; The short-time energy corresponding to each of the multi-frame structured acoustic signals is compared with the corresponding trigger threshold to obtain the comparison result; The suspected fault transient disturbance event and the corresponding trigger time of the suspected fault transient disturbance event are determined based on the comparison results. Based on the triggering time, the target signal segment is determined from the multimodal acoustic signal.

3. The method according to claim 2, characterized in that, Before comparing the short-time energy corresponding to each of the multi-frame structured acoustic signals with the corresponding trigger threshold to obtain the comparison result, the method further includes: Acquire the short-time energy of the pre-defined frame structure acoustic signal before any frame structure acoustic signal; Determine the mean and standard deviation of the short-time energy corresponding to each of the predetermined frame structure acoustic signals; Based on the mean and the standard deviation, the trigger threshold corresponding to any frame structured acoustic signal is obtained; The trigger thresholds corresponding to each of the multiple frame structured acoustic signals are obtained by using the method of obtaining the trigger threshold corresponding to any frame of the structured acoustic signal.

4. The method according to claim 1, characterized in that, The step of extracting features from the target signal segment to obtain fused fault acoustic signature features includes: Feature extraction is performed on the structure acoustic signal in the target signal segment to obtain a structure acoustic fault feature set. The structure acoustic fault feature set includes short-time energy, zero-crossing rate, spectral centroid in the frequency domain, and Mel-frequency cepstral coefficients in the time and frequency domain. The short-time energy represents the energy level of the structure acoustic signal within a preset sliding time window. The zero-crossing rate represents the number of times the structure acoustic signal crosses zero points per unit time. The spectral centroid represents the center position of the frequency domain distribution of the structure acoustic signal. The Mel-frequency cepstral coefficients are used to capture the time and frequency characteristics of the structure acoustic signal. Features are extracted from the airborne sound signal in the target signal segment to obtain an airborne sound fault feature set, wherein the airborne sound fault feature set includes sound pressure level and spectral harmonic clarity. The sound pressure level represents the decibel level of the airborne sound signal relative to the reference sound pressure, and the spectral harmonic clarity represents the relative significance of the harmonic components and the fundamental component in the spectrum of the airborne sound signal. The structural acoustic fault feature set and the airborne acoustic fault feature set are spliced ​​together end to end in a preset order to obtain the fused fault acoustic signature feature.

5. The method according to claim 1, characterized in that, The process of obtaining the high-resistance grounding arc feature recognition result based on the fused fault acoustic signature features using a high-resistance grounding arc feature recognition model includes: Based on the fused fault acoustic signature features, the high-resistivity grounding arc feature identification model is used to obtain the initial fault identification results and the corresponding confidence scores. Based on the initial fault identification results, the confidence score and the target signal segment are verified to obtain the verification results; If the verification result indicates that the verification is successful, the initial fault identification result shall be used as the high-resistance grounding arc feature identification result.

6. The method according to claim 5, characterized in that, The verification is performed based on the initial fault identification result, the confidence score, and the target signal segment to obtain the verification result, including: The first verification condition is determined as follows: the effective sound pressure level of the airborne sound signal in the target signal segment reaches a predetermined significant level relative to the average effective sound pressure level under the historical background noise level. The second verification condition is determined as follows: the start time of the transient disturbance detected in the structure acoustic signal and the airborne acoustic signal within the target signal segment is consistent within the preset error allowable range; The third verification condition is determined as follows: the initial fault identification result is one of the preset fault categories, and the confidence score is greater than the preset confidence threshold. If the first verification condition, the second verification condition, and the third verification condition are all met, the verification result is determined to be a successful verification; or If at least one of the first verification condition, the second verification condition, and the third verification condition is not met, the verification result is determined to be verification failure.

7. A high-resistivity grounding arc feature identification device based on voiceprint recognition, characterized in that, include: The signal acquisition module is used to acquire multimodal acoustic signals at monitoring points on the transmission line under test. The multimodal acoustic signals include structural acoustic signals and airborne acoustic signals collected at the monitoring points at multiple sampling time periods. The structural acoustic signals represent transient mechanical stress waves excited by arc discharge in the conductor medium, and the airborne acoustic signals represent pressure waves generated near the monitoring points by the interaction between the arc plasma channel and the surrounding air. The disturbance identification module is used to identify suspected fault transient disturbance events and target signal segments corresponding to the suspected fault transient disturbance events based on the structure acoustic signals in the multimodal acoustic signals, wherein the suspected fault transient disturbance event represents a transient disturbance event suspected to be caused by high-resistance grounding arc discharge. The feature extraction module is used to extract features from the target signal segment to obtain fused fault acoustic features, wherein the fused fault acoustic features are obtained based on the structure acoustic fault features and airborne acoustic fault features of the target signal segment; The fault identification module is used to determine the high-resistance grounding arc feature identification result of the transmission line under test based on the fused fault acoustic features. This includes: using a high-resistance grounding arc feature identification model based on the fused fault acoustic features to obtain the high-resistance grounding arc feature identification result; wherein the high-resistance grounding arc feature identification model is obtained by training multiple initial learners using a gradient boosting decision tree method based on multiple sets of historical fused fault acoustic features and corresponding fault feature identification results. The corresponding fault feature identification result includes at least: normal operation state, multiple high-resistance grounding fault states, and multiple conventional interference states, wherein the conventional interference states are interference states other than the high-resistance grounding fault state.

8. A non-volatile storage medium, characterized in that, The non-volatile storage medium stores multiple instructions, which are adapted to be loaded by a processor and executed by the high-resistance grounding arc feature identification method based on voiceprint recognition as described in any one of claims 1 to 6.

9. An electronic device, characterized in that, It includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the high-resistance grounding arc feature identification method based on voiceprint recognition as described in any one of claims 1 to 6.

Images