A power equipment voiceprint online monitoring method and device

Abstract

The application discloses a kind of power equipment voiceprint online monitoring methods, including at least one group of voiceprint collection equipment to target power equipment and establish the communication connection with edge gateway or collection host;Acoustic signal is continuously collected in the process of power equipment operation and forms voiceprint data stream in multiple channels;Voiceprint data stream is carried out anti-interference processing to suppress background noise and obtain pure voiceprint;Pure voiceprint is carried out feature extraction and vectorization, and is carried out dimension reduction processing;Abnormal detection is carried out based on the feature vector after processing to identify abnormal voiceprint mode;When identifying abnormal voiceprint mode, trigger early warning mechanism, generate the alarm type and state information corresponding to abnormal voiceprint segment;Abnormal voiceprint segment, alarm type and state information are uploaded to monitoring platform or server and are stored and shown.The application can realize continuous collection, anti-interference processing, feature extraction, abnormal detection, defect identification and platform management to power equipment voiceprint.

Description

Technical Field

[0001] This invention relates to the field of online voiceprint monitoring technology, specifically to a method and device for online voiceprint monitoring of power equipment. Background Technology

[0002] Existing substation fault detection methods in remote intelligent inspection scenarios mainly rely on monitoring state quantities such as dissolved gas in oil, infrared thermal imaging, and partial discharge after a fault occurs. These methods generally suffer from problems such as accident lag, high monitoring costs, and the need for manual assistance. Furthermore, traditional detection methods are not suitable for detection during operation, and power outage detection cannot meet the requirements for power supply reliability. In addition, in field applications, there are also pain points such as the aliasing of operating noise from different equipment leading to inaccurate extraction of target acoustic signals, the scarcity of negative samples leading to weak generalization ability of recognition models, low confidence of state assessment under single acoustic data, and insufficient reliability of online monitoring devices in complex environments. These methods are insufficient to meet the needs of real-time, live, and full-coverage monitoring of power grid equipment. Summary of the Invention

[0003] The problem to be solved by the present invention is to provide a method and equipment for online monitoring of the acoustic signature of power equipment, which can realize continuous acquisition, anti-interference processing, feature extraction, anomaly detection, defect identification and platform-based management of the acoustic signature of power equipment.

[0004] The technical solution provided by this invention to solve the above problems is: a method for online monitoring of acoustic signatures in power equipment, the method comprising the following steps: At least one set of voiceprint acquisition devices shall be deployed on the target power equipment and a communication connection shall be established with the edge gateway or acquisition host. During the operation of power equipment, acoustic signals are continuously acquired through multiple channels and formed into an acoustic data stream. The voiceprint data stream is subjected to anti-interference processing to suppress background noise and obtain a clean voiceprint; The pure voiceprint is feature extracted and vectorized, and then dimensionality reduced. Anomaly detection is performed based on the processed feature vectors to identify abnormal voiceprint patterns; When an abnormal voiceprint pattern is detected, an early warning mechanism is triggered to generate an alarm type and status information corresponding to the abnormal voiceprint segment. The abnormal soundprint fragments, alarm types, and status information are uploaded to the monitoring platform or server for storage and display, so as to realize online monitoring of the operating status of power equipment and fault identification.

[0005] Preferably, multi-channel uninterrupted acquisition includes continuous voiceprint monitoring using at least 14 channels.

[0006] Preferably, the anti-interference processing includes: directional enhancement of sound sources where the operating sounds of different equipment on site overlap, and blind source separation performed by combining time-frequency spectra and feature clustering to remove background noise from the power equipment.

[0007] Preferably, the anti-interference processing further includes a combined noise reduction process of spectral subtraction and wavelet packet decomposition, used to locate the frequency band and time of environmental noise interference and suppress it.

[0008] Preferably, the anomaly detection includes abnormal voiceprint state recognition based on feature distance, so as to achieve rapid identification and localization of abnormal voiceprint signals.

[0009] Preferably, the fault identification includes constructing a deep learning model and introducing an attention mechanism to improve the generalization ability in scenarios where negative voiceprint samples are scarce, and to achieve accurate diagnosis of different defect types; in response to the scarcity of voiceprint fault samples, adversarial generative networks are used to perform voiceprint sample expansion to enrich the voiceprint database.

[0010] Preferably, the feature extraction and recognition process includes: vectorizing the voiceprint information, performing PCA dimensionality reduction processing, and inputting it into a residual neural network to obtain the recognition result output.

[0011] Preferably, the acquisition of voiceprint data adopts a fully digital signal transmission system and follows the IEEE 802.3 standard protocol to achieve high-fidelity transmission without signal line length limitations and improve the system networking scale.

[0012] Preferably, the alarm types include at least one or more of the following: transformer abnormality, short circuit impact, DC bias, partial discharge, loose clamps, heavy overload, and winding deformation.

[0013] The method comprises a multi-channel acoustic fingerprint sensor interface module, a signal conditioning and analog-to-digital conversion module, an anti-interference processing module, an edge computing processing module, a communication module, and a power supply module. The multi-channel acoustic fingerprint sensor interface module is used to connect at least 14 channels of sensors and acquire acoustic fingerprint data. The anti-interference processing module performs background noise suppression and blind source separation to obtain clean acoustic fingerprints. The edge computing processing module performs feature extraction, dimensionality reduction, and model inference on the clean acoustic fingerprints to output anomaly detection results and fault types. The communication module uploads abnormal acoustic fingerprint segments and alarm information to the platform and supports remote access. The power supply module adopts an electrical isolation and protection design to adapt to strong electromagnetic interference and continuous operation in outdoor environments. The method for online acoustic fingerprint monitoring of power equipment, as described in section 1, is characterized in that the monitoring platform supports the storage of abnormal acoustic fingerprint segments and corresponding alarm types, and supports real-time monitoring, recording download and playback, historical alarm query, dynamic management of multiple gateways and devices, and map display of device location layout.

[0014] This invention also discloses a voiceprint acquisition device, applied in the above-described online voiceprint monitoring method for power equipment. The device comprises a cabinet and, within the cabinet, a multi-channel voiceprint sensor interface module, a signal conditioning and analog-to-digital conversion module, an anti-interference processing module, an edge computing processing module, a communication module, and a power supply module. The multi-channel voiceprint sensor interface module is used to connect at least 14 channels of sensors and acquire voiceprint data. The anti-interference processing module performs background noise suppression and blind source separation to obtain clean voiceprints. The edge computing processing module performs feature extraction, dimensionality reduction, and model inference on the clean voiceprints to output anomaly detection results and fault types. The communication module uploads abnormal voiceprint segments and alarm information to the platform and supports remote access. The power supply module employs electrical isolation and protection design to adapt to strong electromagnetic interference and continuous operation in outdoor environments.

[0015] Compared with existing technologies, the advantages of this invention are: through multi-channel continuous acquisition, anti-interference processing combining directional enhancement and blind source separation, feature representation based on vectorization and dimensionality reduction, anomaly detection and defect identification combining adversarial generative sample expansion and deep learning models, and early warning and operation and maintenance management mechanisms linked with the platform, this invention realizes real-time online monitoring and intelligent fault diagnosis of power equipment in energized operation. It can effectively improve problems such as noise aliasing, sample scarcity, insufficient confidence and insufficient device reliability, and has technical effects such as high accuracy, fast response, scalability and adaptability to complex environments. Attached Figure Description

[0016] The accompanying drawings, which are provided to further illustrate the invention and constitute a part of this invention, are illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention.

[0017] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a three-dimensional structural diagram of the voiceprint acquisition device of the present invention; Figure 3 This is a front view of the cabinet door of the present invention; Figure 4 This is a front view of the cabinet of the present invention; Figure 5 This is a front sectional view of the cabinet of the present invention; Figure 6 yes Figure 5 Enlarged view of point A in the middle; Figure 7 yes Figure 5 Enlarged view of point B in the middle; Figure 8 This is a top sectional view of the sealing ring of the present invention; Figure 9 This is a top view of the liquid discharge control component of the present invention.

[0018] Attached diagram labels: 1. Cabinet body, 2. Cabinet door, 3. Annular sealing plate, 4. Mounting cavity, 5. Sealing ring, 6. Liquid inlet cavity, 7. Connecting pipe, 8. Vent pipe, 9. Piston cavity, 10. Spring 1, 11. Piston head, 12. Liquid storage cavity, 13. Mounting block, 14. Spring 2, 15. Baffle, 16. Liquid outlet hole, 17. Deformation rod, 18. Support block, 19. Deformation cavity, 20. Thinning groove. Detailed Implementation

[0019] The following will describe in detail the implementation of the present invention with reference to the accompanying drawings and embodiments, so that the process of how the present invention uses technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly.

[0020] In the description of this invention, it should be noted that the directional terms such as "center", "lateral", "longitudinal", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", and "counterclockwise" indicate the orientation and positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. They should not be construed as limiting the specific protection scope of this invention.

[0021] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features. Thus, the use of "first" and "second" to define a feature may explicitly or implicitly include one or more of that feature, and in the description of this invention, "a number" means two or more, unless otherwise explicitly specified.

[0022] In this invention, unless otherwise explicitly specified and limited, the terms "assembly," "connection," and "joining" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can also refer to a mechanical connection; they can refer to a direct connection or a connection through an intermediate medium; or they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0023] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0024] It should also be understood that the terminology used in this specification of embodiments of the invention is for the purpose of describing particular embodiments only and is not intended to limit the embodiments of the invention. As used in this specification of embodiments of the invention and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. Example 1

[0025] This embodiment provides a power equipment acoustic signature monitoring solution. Through a "collection end—edge end—monitoring / server end" architecture, it achieves continuous acquisition, anti-interference processing, feature extraction, anomaly detection, defect identification, and platform-based management of power equipment acoustic signatures. In this system, the collection end is used to pick up equipment operating acoustic signature information and perform anti-electromagnetic interference acoustic signature processing and fault identification. The edge end / gateway is used to aggregate data and connect to the monitoring end or backend server via 4G / 5G / RS485 / network cable, etc. The monitoring end is used for data storage and visualization, and provides support for operation and maintenance decisions.

[0026] In one embodiment, the monitoring method includes the following steps: deploying acoustic signature acquisition equipment on the target power equipment and establishing a communication connection with the acquisition host or edge gateway; continuously acquiring multi-channel acoustic / vibration signals of the equipment to form an acoustic signature data stream, and performing preprocessing on the acquisition side or edge side; performing background noise suppression and target acoustic signature enhancement to obtain pure acoustic signatures in response to complex noise and strong electromagnetic interference on site; inputting the pure acoustic signatures into the feature processing and recognition process, completing vectorization, dimensionality reduction, and model inference to obtain abnormal state identification results and defect type outputs; when an abnormal acoustic signature pattern is identified, an early warning mechanism is activated, and the abnormal acoustic signature fragments and corresponding alarm types are uploaded to the platform for storage, display, and querying to assist in maintenance decisions. The method enables live-line monitoring, real-time monitoring, and meets high monitoring accuracy requirements.

[0027] In terms of signal acquisition and device adaptability, the acquisition end can use a fixed acoustic signature monitoring device to achieve multi-channel continuous monitoring. For example, it supports 14-channel uninterrupted acoustic signature monitoring and large-span array beamforming to adapt to the directional sound pickup and enhancement of the target equipment area in the multi-sound source environment of the substation. The device adopts the design concept for the complex environment of industrial and power operation, and can operate continuously and stably at any time and capture acoustic signature changes to achieve all-weather monitoring and real-time analysis.

[0028] Regarding anti-interference and background noise removal, this invention addresses the problem of overlapping operating sounds from power equipment. First, it studies and implements directional enhancement of the voiceprint signal in the target equipment area. Second, it analyzes the time-frequency curves of each sound source, combines time-frequency spectra and feature clustering methods, and uses blind source separation to remove background noise from power equipment, obtaining a clean voiceprint. In the noise reduction process, a combination of spectral subtraction, speech enhancement, and wavelet packet decomposition can be used. By identifying the common frequency components of noise in the frequency domain distribution, the frequency band and time of environmental noise interference are located, and environmental noise interference is effectively suppressed to improve the quality of the original data for voiceprint recognition.

[0029] In terms of feature extraction and intelligent recognition, this invention can establish a multi-dimensional classification sample library, which can include categories such as equipment failure, environmental noise, load samples, and switch samples, and further subdivide them, so as to continuously enrich the samples and improve the recognition accuracy as field applications grow; to address the problem of scarce voiceprint fault samples, voiceprint samples can be expanded based on adversarial generative networks to enrich the voiceprint database; anomaly detection can adopt an abnormal voiceprint state recognition method based on feature distance to quickly identify and locate abnormal voiceprint signals; defect recognition can build a deep learning model with strong generalization ability by combining attention mechanism on the basis of establishing different defect voiceprint databases, so as to achieve accurate diagnosis of different defect types.

[0030] In one feasible algorithmic chain, after the voiceprint information is input, it is first vectorized, then PCA is used for dimensionality reduction. The dimensionality-reduced features are then input into models such as residual neural networks to complete recognition and inference. Simultaneously, blind source separation processing is introduced before or during recognition to further reduce residual noise and output a clean voiceprint, thus obtaining the recognition result. This process can cover the recognition needs of abnormal / defect types such as three-phase imbalance, heavy overload, overvoltage, short-circuit impact, partial discharge, loose winding, loose installation, and DC bias.

[0031] In terms of early warning and platform management, when the device or edge-side model identifies an abnormal sound pattern, a timely early warning mechanism is activated to provide users with reaction time to take rapid countermeasures and ensure the stable operation of power equipment. The platform supports storing abnormal sound pattern segments and corresponding alarm types, and supports functions such as online device status visualization, real-time monitoring, recording download and playback, real-time automatic early warning and historical alarm query, dynamic management of multiple gateways and devices and dynamic addition of software, and map display of device location layout. It can also provide secondary development documentation to support third-party access and remote one-click upgrade / installation of engine programs and other expansion capabilities.

[0032] In terms of fault location and maintenance suggestion output, when the device has a built-in voiceprint diagnostic model for transformers, it can locate fault sources such as transformer anomalies, short circuit impacts, DC bias, partial discharge, loose clamps, heavy overload, and winding deformation through in-depth analysis of voiceprint patterns, and provide maintenance suggestions to users. For other types of power equipment, the model and rule base can be extended through secondary development to adapt to the voiceprint characteristics of different equipment.

[0033] Compared with traditional analog signal transmission and intermittent / sampling analysis schemes, the system of this invention can adopt a fully digital signal transmission architecture and use the IEEE 802.3 standard protocol. Sensors compatible with this protocol can be connected to achieve interference-resistant, high-fidelity transmission without signal line length limitations. The system has strong scalability, can directly network a large number of sensors and control sampling time errors, and supports the construction of ultra-large-scale acoustic arrays. In terms of working mode, it can achieve 24 / 7 uninterrupted continuous synchronous recording, calculation and analysis, and maintain a high data refresh rate, thereby ensuring the purity and accuracy of the original voiceprint data in complex electromagnetic environments and improving real-time performance and diagnostic reliability.

[0034] In summary, this invention achieves real-time online monitoring and intelligent fault diagnosis of power equipment in energized operation by combining multi-channel continuous acquisition, anti-interference processing with directional enhancement and blind source separation, feature representation based on vectorization and dimensionality reduction, anomaly detection and defect identification combining adversarial generative sample expansion and deep learning models, and an early warning and operation and maintenance management mechanism linked with the platform. It can effectively improve problems such as noise aliasing, sample scarcity, insufficient confidence and insufficient device reliability, and has technical effects such as high accuracy, fast response, scalability and adaptability to complex environments. Example 2

[0035] This embodiment discloses a voiceprint acquisition device. Specifically, the voiceprint acquisition device includes a cabinet and a multi-channel voiceprint sensor interface module, a signal conditioning and analog-to-digital conversion module, an anti-interference processing module, an edge computing processing module, a communication module, and a power supply module installed in the cabinet. The multi-channel voiceprint sensor interface module is used to connect to at least 14 channels of sensors and acquire voiceprint data. The anti-interference processing module is used to perform background noise suppression and blind source separation to obtain clean voiceprints. The edge computing processing module is used to perform feature extraction, dimensionality reduction, and model inference on the clean voiceprints to output anomaly detection results and fault types. The communication module is used to upload abnormal voiceprint segments and alarm information to the platform and supports remote access. The power supply module adopts an electrical isolation and protection design to adapt to strong electromagnetic interference and continuous operation in outdoor environments.

[0036] It should be noted that the reliability of the sealing performance of the cabinet and door structure of the voiceprint acquisition device of this invention is crucial during use. In rainy areas where winter temperatures easily drop below freezing, conventional static sealing rings will experience a decrease in sealing effectiveness due to the thermal expansion and contraction of the material. Specifically, low temperatures will cause the sealing ring to shrink, resulting in gaps between it and the mating surface, thus affecting the seal.

[0037] Therefore, in this solution, the cabinet further includes a cabinet body 1 and a cabinet door 2. The cabinet body 1 is provided with a mounting cavity 4 for installing electrical components. A sealing ring 5 is provided on the cavity wall of the mounting cavity 4, and an annular sealing plate 3 that cooperates with the sealing ring 5 is provided on the cabinet door 2. The cabinet 1 also includes an auxiliary sealing assembly for improving the sealing performance of the sealing ring 5. This auxiliary sealing assembly includes an inlet chamber 6, a piston chamber 9, a spring 10, a storage chamber 12, a piston head 11, and a deformation chamber 19. The inlet chamber 6, piston chamber 9, and storage chamber 12 are located within the cabinet 1. The inlet chamber 6 is located at the upper end of the cabinet 1 and is open. The storage chamber 12 is connected to the inlet chamber 6 via a connecting pipe 7. The upper surface of the storage chamber 12 is lower than the lower surface of the inlet chamber 6. The piston chamber 9 is located at the upper surface of the storage chamber 12 and is connected to it. The piston head 11 is installed within the piston chamber 9. One end of the spring 10 abuts against the top of the piston chamber 9, and the other end abuts against the piston head 11. The deformation chamber 19 is located within the sealing ring 5 and is connected to the piston via a vent pipe 8. The upper part of cavity 9 is connected; a thinning groove 20 is provided on the cavity wall of the deformation cavity 19 near the inner ring of the sealing ring 5, and a number of support blocks 18 are provided in the deformation cavity 19. One end of the support block 18 is integrally formed with the cavity wall of the deformation cavity 19; a liquid discharge control component is also provided in the liquid storage cavity 12. The liquid discharge control component includes a deformation rod 17, a baffle 15, a mounting block 13 and a second spring 14. A liquid outlet hole 16 is provided at the bottom of the liquid storage cavity 12. The baffle 15 is movably installed on the upper end of the liquid outlet hole 16. One end of the second spring 14 is fixedly connected to the baffle 15, and the other end is fixedly connected to the mounting block 13 installed at the bottom of the liquid storage cavity 12. The deformation rod 17 is installed at the bottom of the liquid storage cavity 12 and one end abuts against the baffle 15. The deformation rod 17 is made of (e.g., titanium metal shape memory alloy).

[0038] When it rains and the ambient temperature is still above freezing, rainwater enters through the open inlet chamber. Since the upper surface of the storage chamber is lower than the lower surface of the inlet chamber, the rainwater flows in by gravity through the connecting pipe and gradually fills the storage chamber. At this time, the deformation rod made of temperature-sensitive deformation material is in its original state with a higher temperature. One end of the rod abuts against the baffle, which keeps the baffle open against the tension of the second spring. The water in the storage chamber can be smoothly drained out through the outlet at the bottom. Therefore, rainwater will not continue to accumulate in the chamber during this stage, and the drive unit remains in standby mode.

[0039] As the ambient temperature gradually decreases and approaches freezing point, the temperature-sensing deformation rod undergoes a significant change in shape or stiffness due to the temperature drop. Specifically, it shortens, bends, or softens considerably, weakening or eliminating the resistance force applied to the baffle. Under the tension of spring two, the baffle automatically resets and slides, completely covering and sealing the outlet hole. The drainage path of the storage chamber is cut off, and rainwater that continues to flow in through the inlet chamber is blocked and accumulates in the storage chamber, forming a closed water body. At this time, the temperature continues to drop below freezing point, and the water in the storage chamber begins to freeze. During the phase change of water to ice, a strong volume expansion force is generated. This expansion force acts directly on the connection between the upper end of the storage chamber and the piston chamber, pushing the piston head installed in the piston chamber upward, overcoming spring one and compressing it.

[0040] As the piston head moves upward, it compresses the air in the upper part of the piston chamber. The high-pressure gas is forced into the deformation chamber inside the sealing ring through the vent pipe. The deformation chamber expands due to the pressure. Because a thinning groove is pre-made on the cavity wall near the inner ring of the sealing ring, the overall expansion will preferentially bulge outward in the weaker direction of the thinning groove, which manifests as the inner ring of the sealing ring bulging outward on the side facing the annular sealing plate. This bulging action forcibly presses the sealing ring material, which has shrunk due to low temperature, back tightly onto the annular sealing plate of the cabinet door, completely filling the gap caused by cold contraction. The low-temperature sealing performance is improved rather than decreased.

[0041] Once the temperature rises above zero, the ice in the storage chamber and cylinder melts into liquid water, significantly reducing its volume and eliminating expansion force. The piston head returns to its original position under the rebound of spring one. The previously compressed air in the deformation chamber is drawn back through the vent pipe, and the expansion bulge of the inner ring of the sealing ring flattens, removing the additional pressure on the cabinet door sealing plate and preventing permanent deformation of the sealing ring material due to long-term excessive compression. Simultaneously, the temperature-sensing deformation rod returns to its initial shape or stiffness when the environment warms up, reapplying a pushing force to the baffle, causing it to move again and open the liquid outlet. The water accumulated during thawing in the chamber is discharged outside the cabinet under gravity. The entire auxiliary sealing assembly returns to its initial empty cavity state, awaiting the next low-temperature rainfall cycle to trigger a new round of compensation actions.

[0042] The above solution relies on the inherent physical properties of rainwater and temperature changes in nature to passively and automatically achieve adaptive compensation for sealing performance, transforming the usually harmful phenomenon of icing into a positive driving force for enhanced sealing. It integrates a series of actions, including temperature sensing, water storage, work done by ice expansion, pressure transmission, and function reset. The temperature-sensing deformation rod, acting as a maintenance-free, purely mechanical temperature switch, autonomously switches between water storage and drainage near the freezing point, eliminating the fragility and energy consumption of electronic control systems. In rainy winter environments, while the low temperature causes the sealing ring material to shrink and the sealing force to weaken, the expansion force of ice is triggered at this moment and acts on the sealing ring through air pressure, forcing it to expand in the opposite direction, perfectly filling the gap caused by cold contraction. This compensation resonates with external destructive factors, achieving a reverse environmental effect where the colder the environment, the more reliable the seal. The structure has a high degree of integration and a long service life. The entire auxiliary component is concealed and integrated within the cabinet, and the core pressure transmission is completed using a deformation cavity inside the sealing ring. This makes the sealing lip both a functional component and an actuator, resulting in a high degree of integration. Important protective mechanisms are incorporated through the design of springs and support blocks. The pneumatic drive method is a flexible force application that can automatically adapt to the unevenness of the sealing surface. A spring forcibly pulls the piston head back to its original position after the ice melts, achieving automatic pressure release and preventing permanent compression deformation of the sealing ring. The thinning groove and support block together constrain the expansion direction and amplitude of the deformation cavity, ensuring that the sealing is improved without damaging the sealing ring itself. The entire working cycle is stable and reversible, capable of withstanding countless freeze-thaw cycles without degradation.

[0043] The above description only illustrates the preferred embodiments of the present invention and should not be construed as limiting the scope of the claims. The present invention is not limited to the above embodiments, and variations in its specific structure are permitted. All modifications made within the scope of the independent claims of this invention are also within the scope of protection of this invention.

Claims

1. A method for online monitoring of acoustic signatures in power equipment, characterized in that, The method includes the following steps: At least one set of voiceprint acquisition devices shall be deployed on the target power equipment and a communication connection shall be established with the edge gateway or acquisition host. During the operation of power equipment, acoustic signals are continuously acquired through multiple channels and formed into an acoustic data stream. The voiceprint data stream is subjected to anti-interference processing to suppress background noise and obtain a clean voiceprint; The pure voiceprint is feature extracted and vectorized, and then dimensionality reduced. Anomaly detection is performed based on the processed feature vectors to identify abnormal voiceprint patterns; When an abnormal voiceprint pattern is detected, an early warning mechanism is triggered to generate an alarm type and status information corresponding to the abnormal voiceprint segment. The abnormal soundprint fragments, alarm types, and status information are uploaded to the monitoring platform or server for storage and display, so as to realize online monitoring of the operating status of power equipment and fault identification.

2. The method for online monitoring of acoustic signatures of power equipment according to claim 1, characterized in that, Multi-channel uninterrupted acquisition includes continuous voiceprint monitoring using at least 14 channels.

3. The method for online monitoring of acoustic signatures of power equipment according to claim 1, characterized in that, The anti-interference processing includes: directional enhancement of sound sources where the operating sounds of different equipment overlap on site, and blind source separation by combining time-frequency spectra and feature clustering to remove background noise from power equipment.

4. The method for online monitoring of acoustic signatures of power equipment according to claim 1, characterized in that, The anti-interference processing further includes a combined noise reduction process of spectral subtraction and wavelet packet decomposition, used to locate the frequency band and time of environmental noise interference and suppress it.

5. The method for online monitoring of acoustic signatures of power equipment according to claim 1, characterized in that, The anomaly detection includes abnormal voiceprint state recognition based on feature distance, so as to achieve rapid identification and localization of abnormal voiceprint signals.

6. The method for online monitoring of acoustic signatures of power equipment according to claim 1, characterized in that, The fault identification includes constructing a deep learning model and introducing an attention mechanism to improve the generalization ability in scenarios where negative voiceprint samples are scarce, and to achieve accurate diagnosis of different defect types; in response to the scarcity of voiceprint fault samples, adversarial generative networks are used to perform voiceprint sample expansion to enrich the voiceprint database.

7. The method for online monitoring of acoustic signatures of power equipment according to claim 1, characterized in that, The feature extraction and recognition process includes: vectorizing the voiceprint information, performing PCA dimensionality reduction, and inputting it into a residual neural network to obtain the recognition result output.

8. The method for online acoustic signature monitoring of power equipment according to claim 1, characterized in that, The alarm types include at least one or more of the following: transformer abnormality, short circuit impact, DC bias, partial discharge, loose clamps, heavy overload, and winding deformation. The method comprises a multi-channel acoustic fingerprint sensor interface module, a signal conditioning and analog-to-digital conversion module, an anti-interference processing module, an edge computing processing module, a communication module, and a power supply module. The multi-channel acoustic fingerprint sensor interface module is used to connect at least 14 channels of sensors and acquire acoustic fingerprint data. The anti-interference processing module performs background noise suppression and blind source separation to obtain clean acoustic fingerprints. The edge computing processing module performs feature extraction, dimensionality reduction, and model inference on the clean acoustic fingerprints to output anomaly detection results and fault types. The communication module uploads abnormal acoustic fingerprint segments and alarm information to the platform and supports remote access. The power supply module adopts an electrical isolation and protection design to adapt to strong electromagnetic interference and continuous operation in outdoor environments. The method for online acoustic fingerprint monitoring of power equipment, as described in section 1, is characterized in that the monitoring platform supports the storage of abnormal acoustic fingerprint segments and corresponding alarm types, and supports real-time monitoring, recording download and playback, historical alarm query, dynamic management of multiple gateways and devices, and map display of device location layout.

9. A voiceprint acquisition device, applied in the online voiceprint monitoring method for power equipment as described in claim 1, characterized in that: The voiceprint acquisition device includes a cabinet and, within the cabinet, a multi-channel voiceprint sensor interface module, a signal conditioning and analog-to-digital conversion module, an anti-interference processing module, an edge computing processing module, a communication module, and a power supply module. The multi-channel voiceprint sensor interface module is used to connect to at least 14 channels of sensors and acquire voiceprint data. The anti-interference processing module performs background noise suppression and blind source separation to obtain clean voiceprints. The edge computing processing module performs feature extraction, dimensionality reduction, and model inference on the clean voiceprints to output anomaly detection results and fault types. The communication module uploads abnormal voiceprint segments and alarm information to the platform and supports remote access. The power supply module adopts an electrical isolation and protection design to adapt to strong electromagnetic interference and continuous operation in outdoor environments.

Images