Hydroelectric generating set operation state voiceprint online monitoring method and related device

Abstract

The invention discloses a hydroelectric generating set operation state voiceprint online monitoring method, a hydroelectric generating set operation state voiceprint online monitoring device, a hydroelectric generating set operation state voiceprint online monitoring device and a computer readable storage medium. According to the method, firstly, operation acoustic signals of key equipment such as a water turbine, a generator and a transformer are collected through a distributed optical fiber sensing array, and then noise reduction processing and feature transformation are carried out on the collected acoustic signals to extract voiceprint feature vectors; and a pre-trained fault diagnosis model is used to perform reasoning on the voiceprint feature vector to obtain an equipment state diagnosis result, the fault diagnosis model is formed by fusion training based on historical voiceprint data and maintenance experience data, and finally trend prediction and comprehensive evaluation are performed according to the diagnosis result to generate graded early warning signals and corresponding maintenance suggestions. According to the invention, the anti-electromagnetic interference advantage of optical fiber sensing is fully utilized, and accurate sensing of the operation state of the hydroelectric generating set and intelligent fault identification are realized.

Description

Technical Field

[0001] This invention belongs to the field of monitoring technology for water conservancy and hydropower engineering equipment, and more specifically, relates to a method, device, and computer-readable storage medium for online monitoring of the operating status of hydropower units using acoustic signatures. Background Technology

[0002] With the continuous advancement of smart hydropower station construction, higher requirements have been placed on the real-time perception of the operating status of core equipment such as turbines, generators, transformers, GIS, and intake gates. As an effective means to reflect the mechanical and electrical status of equipment, acoustic fingerprint monitoring has gradually attracted industry attention.

[0003] Among related technologies, acoustic signature monitoring technology for hydropower units still has significant shortcomings. On the one hand, traditional acoustic acquisition schemes mostly use piezoelectric microphones or electronic pickups. These sensors are highly susceptible to power frequency and harmonic interference in the high electromagnetic interference environment of hydropower stations, resulting in a low signal-to-noise ratio. Furthermore, deploying active electronic sensors in high-voltage or confined areas such as GIS rooms and transformer interiors also faces challenges related to insulation safety and power supply. On the other hand, existing acoustic fault diagnosis methods are mostly based on simple decibel thresholds for alarm judgment, making it difficult to identify specific fault types. Moreover, the construction of diagnostic models often relies solely on data-driven approaches, lacking effective integration with the long-term maintenance experience and historical fault records accumulated by power plants. This leads to a high false alarm rate and an inability to provide targeted predictive maintenance recommendations for operation and maintenance personnel.

[0004] Therefore, how to develop an online acoustic fingerprint monitoring method that can adapt to the strong electromagnetic interference environment of hydropower stations and integrate traditional maintenance experience for intelligent diagnosis, so as to achieve accurate perception of the operating status of hydropower units and reliable identification of fault types, is a key issue of concern to those skilled in the art. Summary of the Invention

[0005] The purpose of this application is to provide a method, device, and computer-readable storage medium for online acoustic monitoring of the operating status of hydropower units. This method is capable of adapting to the strong electromagnetic interference environment of hydropower stations and integrating traditional maintenance experience for intelligent diagnosis, so as to achieve accurate perception of the operating status of hydropower units and reliable identification of fault types.

[0006] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention provides a method for online acoustic signature monitoring of the operating status of hydropower units, comprising: S1 collects operating acoustic signals through a distributed fiber optic sensor array deployed on key equipment of the hydropower unit, and converts the collected operating acoustic signals into multi-channel raw acoustic waveform data with timestamps and spatial location tags through photoelectric demodulation. S2, perform noise reduction, frame windowing, and feature transformation on the multi-channel raw acoustic waveform data to obtain the voiceprint feature vector; S3, a fault diagnosis model is trained based on historical voiceprint data and maintenance experience data; wherein, the fault diagnosis model includes a normal state model and a fault state model; S4, input the voiceprint feature vector into the fault diagnosis model for inference, and output the diagnosis result; wherein, the diagnosis result includes: real-time status of the equipment and fault type; S5. Based on the diagnostic results of continuous time series, perform trend prediction and comprehensive evaluation, and generate graded early warning signals and maintenance suggestions according to the comprehensive evaluation.

[0007] Optionally, in S1, the distributed optical fiber sensing array uses a distributed optical fiber acoustic sensor or a fiber optic grating sensor, which is deployed on the turbine top cover, generator wind tunnel, main transformer tank wall, GIS pipeline and inlet gate hoist, with the timestamp synchronization accuracy of all measuring points reaching the microsecond level.

[0008] Optionally, in step S2, the noise reduction process uses wavelet thresholding or spectral subtraction to filter out background white noise and power frequency interference; the frame length of the frame segmentation and windowing is 20ms to 50ms, and the window function is a Hamming window; the feature transformation includes extracting Mel frequency cepstral coefficients or Log-Mel spectrograms.

[0009] Optionally, S3 includes: parsing historical maintenance logs using natural language processing technology, associating historical fault records with corresponding historical voiceprint data to generate a labeled training dataset; constructing the fault diagnosis model using a convolutional neural network or a long short-term memory network, and introducing an attention mechanism to perform weighted learning on feature regions that match the abnormal frequency bands described by expert experience; and training the fault diagnosis model using the training dataset to obtain the trained fault diagnosis model.

[0010] Optionally, S4 includes: inputting the voiceprint feature vector into the normal state model to calculate the deviation, and determining an abnormal state when the deviation exceeds a dynamic threshold; inputting the voiceprint feature vector determined to be in an abnormal state into the fault state model to identify the fault type, and calculating the confidence level of the diagnostic result.

[0011] Optionally, S3 further includes: establishing normal operating voiceprint reference models for different operating conditions, including start-up, shutdown, full load, and load shedding conditions; correspondingly, in S4, the corresponding voiceprint reference model is selected as the normal state model based on the current operating condition for deviation calculation.

[0012] Optionally, the trend prediction includes: analyzing the evolution trend of voiceprint features using a time series analysis algorithm; The graded early warning signal includes four levels: Level IV is the normal state, which does not trigger an early warning; Level III is the attention state, which generates a notice; Level II is the warning state, which generates an inspection suggestion; and Level I is the alarm state, which triggers emergency shutdown logic or an audible and visual alarm.

[0013] This application also provides an online acoustic monitoring device for the operating status of a hydropower unit, comprising: The acoustic signal acquisition module is used to acquire operating acoustic signals through a distributed fiber optic sensor array deployed in key equipment of the hydropower unit. The acquired operating acoustic signals are converted into multi-channel raw acoustic waveform data with timestamps and spatial location tags through photoelectric demodulation. The signal processing module is used to perform noise reduction, frame windowing, and feature transformation on the multi-channel raw acoustic waveform data to obtain the voiceprint feature vector. The model training module is used to train a fault diagnosis model based on historical voiceprint data and maintenance experience data; wherein, the fault diagnosis model includes a normal state model and a fault state model; The model inference module is used to input the voiceprint feature vector into the fault diagnosis model for inference and output the diagnosis result; wherein, the diagnosis result includes: the real-time status of the equipment and the fault type; The early warning information generation module is used to perform trend prediction and comprehensive evaluation based on the diagnostic results of continuous time series, and generate graded early warning signals and maintenance suggestions based on the comprehensive evaluation.

[0014] This application also provides an online acoustic monitoring device for the operating status of hydropower units, including: Memory, used to store computer programs; A processor is used to execute the computer program to implement the steps of the online acoustic fingerprint monitoring method for the operating status of hydropower units as described above.

[0015] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for online monitoring of the operating status of hydropower units using acoustic fingerprints.

[0016] This application provides a method for online acoustic signature monitoring of the operating status of a hydropower unit, comprising: S1, collecting operating acoustic signals through a distributed fiber optic sensor array deployed on key equipment of the hydropower unit, and converting the collected operating acoustic signals into multi-channel raw acoustic waveform data with timestamps and spatial location tags through photoelectric demodulation; S2, performing noise reduction, frame windowing, and feature transformation on the multi-channel raw acoustic waveform data to obtain acoustic signature feature vectors; S3, training a fault diagnosis model based on historical acoustic signature data and maintenance experience data; wherein the fault diagnosis model includes a normal state model and a fault state model; S4, inputting the acoustic signature feature vector into the fault diagnosis model for inference and outputting a diagnosis result; wherein the diagnosis result includes: real-time equipment status and fault type; S5, performing trend prediction and comprehensive evaluation based on the diagnosis result of continuous time series, and generating graded early warning signals and maintenance suggestions based on the comprehensive evaluation.

[0017] It has the following beneficial effects: By deploying a distributed fiber optic sensor array to collect the acoustic signals of key equipment in hydropower units, the passive and electromagnetic interference-resistant characteristics of optical fibers effectively solve the problem of poor signal quality of traditional electronic sensors in the strong electromagnetic environment of hydropower stations. By performing noise reduction and feature transformation on the collected acoustic signals to extract acoustic feature vectors, and using a fault diagnosis model trained by fusing historical acoustic data and maintenance experience data for inference, the real-time status and specific fault types of equipment can be accurately identified, overcoming the limitation of traditional threshold alarm methods that cannot distinguish fault categories. By performing trend prediction and comprehensive evaluation of the diagnostic results and generating graded early warning signals and maintenance suggestions, the system provides maintenance personnel with complete decision support from status awareness to predictive maintenance. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1 A flowchart illustrating an online acoustic signature monitoring method for the operating status of a hydropower unit, provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of an online acoustic signature monitoring device for the operating status of a hydropower unit provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of the online acoustic monitoring device for the operating status of a hydropower unit provided in this embodiment of the application. Detailed Implementation

[0020] The purpose of this application is to provide a method, device, and computer-readable storage medium for online acoustic monitoring of the operating status of hydropower units. This method is capable of adapting to the strong electromagnetic interference environment of hydropower stations and integrating traditional maintenance experience for intelligent diagnosis, so as to achieve accurate perception of the operating status of hydropower units and reliable identification of fault types.

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0022] The following embodiment illustrates a method for online acoustic monitoring of the operating status of a hydropower unit provided in this application.

[0023] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating an online acoustic signature monitoring method for the operating status of a hydroelectric generator unit, as provided in an embodiment of this application.

[0024] In this embodiment, the method may include: S1 collects operating acoustic signals through a distributed fiber optic sensor array deployed on key equipment of the hydropower unit, and converts the collected operating acoustic signals into multi-channel raw acoustic waveform data with timestamps and spatial location tags through photoelectric demodulation. The core of this step is to use a distributed optical fiber sensor array to synchronously collect the acoustic signals generated by each key device of the hydropower unit during operation, and convert the collected optical signals into digital acoustic waveform data that is easy to process later.

[0025] In terms of sensor selection and deployment, this embodiment uses distributed fiber optic acoustic sensors or fiber optic grating sensors to construct the sensor array. Fiber optics serve as both a signal transmission medium and a sensing unit, possessing advantages such as being passive, resistant to electromagnetic interference, corrosion-resistant, and high-voltage resistant, making them suitable for the strong electromagnetic interference environment present in hydropower stations. Specific deployment locations of the sensor array include: the turbine roof area, for monitoring mechanical vibration and water flow impact noise during turbine operation; the generator wind tunnel area, for capturing airflow and electromagnetic vibration noise in the generator cooling duct; the main transformer tank wall, for sensing the operating sounds of the transformer core and windings; the surface of GIS pipelines, for monitoring potential partial discharge or mechanical noise within gas-insulated switchgear; and the inlet gate hoist, for monitoring mechanical movement and water flow noise during gate opening and closing.

[0026] During photoelectric demodulation, the demodulator's built-in laser emits pulsed laser light into the deployed optical fiber. When the acoustic waves generated by the device act on the fiber, they cause minute mechanical deformation. This deformation leads to a phase change or wavelength shift in the transmitted optical signal. The demodulator receives the backscattered Rayleigh light or Bragg reflected light returned from the fiber and, by analyzing the changes in the optical signal, restores it to the corresponding acoustic signal.

[0027] To ensure time consistency of data from multiple measurement points, this embodiment strictly synchronizes the acquisition timestamps of all measurement points, achieving a synchronization accuracy at the microsecond level. After photoelectric demodulation, the system outputs multi-channel raw acoustic waveform data with a unified timestamp and spatial location label. The timestamp marks the precise acquisition time of each sampling point, while the spatial location label identifies which device and measurement point the acoustic data originated from.

[0028] S2 performs noise reduction, frame windowing, and feature transformation on the multi-channel raw acoustic waveform data to obtain the voiceprint feature vector; This step preprocesses and transforms the multi-channel raw acoustic waveform data obtained in the previous step. The purpose is to eliminate noise interference and extract acoustic feature vectors that can characterize the operating status of the equipment.

[0029] In the noise reduction process, considering the background white noise present in the hydropower station environment and the inherent power frequency interference of the power system, this embodiment employs wavelet threshold denoising or spectral subtraction to filter the original acoustic data. Wavelet threshold denoising separates the signal components from the noise components by performing multi-scale wavelet decomposition on the signal, then applies a threshold to the wavelet coefficients to suppress noise, and finally obtains the denoised signal through wavelet reconstruction. Spectral subtraction processes the signal in the frequency domain, estimating the power spectrum of the noise and subtracting the noise component from the power spectrum of the noisy signal, thereby achieving noise reduction. After noise reduction, the characteristic frequency band signals generated by equipment operation are preserved, while background interference is effectively filtered out.

[0030] In the framing and windowing process, since the acoustic signal is time-varying, the continuous acoustic signal needs to be divided into a series of short frames for processing to facilitate subsequent frequency domain analysis. This embodiment sets the frame length to 20 to 50 milliseconds. This frame length range ensures the signal's stability within each frame while capturing sufficient frequency information. To reduce spectral leakage caused by signal truncation, a Hamming window function is applied to each frame for smoothing.

[0031] In the feature transformation stage, this embodiment extracts voiceprint features from three dimensions: time domain, frequency domain, and auditory perception. For time domain features, the zero-crossing rate and short-time energy of each frame are calculated. The zero-crossing rate reflects the frequency characteristics of the signal, while the short-time energy reflects the intensity changes. For frequency domain features, a Fast Fourier Transform is performed on the windowed frame signal to obtain the signal's spectral distribution. For auditory features, Mel-frequency cepstral coefficients or Log-Mel spectrograms are extracted. These features simulate the nonlinear perception characteristics of the human ear for sounds of different frequencies and have a good ability to distinguish abnormal device sounds. After the above feature transformations, the original time-domain waveform data is converted into a high-dimensional voiceprint feature vector, providing input for subsequent model inference.

[0032] S3, a fault diagnosis model is trained based on historical voiceprint data and maintenance experience data; the fault diagnosis model includes a normal state model and a fault state model; The core of this step is to build an artificial intelligence model that can identify the operating status and fault type of equipment. The training process of this model integrates historical voiceprint data and traditional maintenance experience.

[0033] In constructing the training dataset, this embodiment first collects historical acoustic signature data accumulated over long-term operation of the hydropower station, along with corresponding maintenance experience data. The latter includes unstructured information such as historical fault logs, maintenance logs, and expert descriptions of specific abnormal sounds. To transform this textual expert experience into labeled information usable for model training, this embodiment utilizes natural language processing technology to parse the historical maintenance logs, automatically extracting key information such as the time of the fault, the faulty equipment, the fault type, and expert descriptions of the fault sound characteristics. Then, the extracted fault records are correlated and matched with the corresponding historical acoustic signature data to generate training samples labeled with fault types. For example, if a maintenance log records a turbine cavitation fault on a certain day, the acoustic signature data of the turbine collected during that time period is labeled as "cavitation"; similarly, if a record describes abnormal vibration caused by a loose transformer core, the corresponding acoustic signature data is labeled as "loose core." In this way, traditional maintenance experience is digitized and integrated into the training dataset.

[0034] In terms of model architecture design, this embodiment uses convolutional neural networks or long short-term memory networks as the basic architecture of the fault diagnosis model. Convolutional neural networks are good at extracting local features and spatial patterns from acoustic signature maps, while long short-term memory networks are good at capturing the temporal dependencies of acoustic signals. The input layer of the model receives the acoustic signature feature vectors or acoustic signature maps extracted in the preprocessing step, the hidden layer learns the complex mapping relationship between acoustic signature features and device status through multiple nonlinear transformations, and the output layer provides the classification result of the device status. To enable the model to focus on abnormal frequency band features that are consistent with expert experience descriptions, this embodiment introduces an attention mechanism into the network structure. This mechanism can automatically learn the importance weights of features in different frequency bands, giving higher attention to feature regions that are strongly correlated with faults, thereby improving the model's sensitivity to abnormal acoustic signatures.

[0035] Regarding model type classification, the fault diagnosis model constructed in this embodiment includes two sub-models: normal state model and fault state model. The normal state model is used to learn the voiceprint feature distribution of the equipment under normal operating conditions, serving as a benchmark for judging whether the equipment has malfunctioned; the fault state model is used to learn the voiceprint feature patterns of various known fault types, and is used to further identify the specific fault category after an anomaly is detected.

[0036] Considering the significant differences in the operating sounds of hydropower units under different operating conditions, this embodiment establishes separate acoustic signature reference models for normal operation under different conditions. Specifically, for start-up, shutdown, full-load, and load shedding conditions, normal operation acoustic signature data are collected for each corresponding condition, and normal state models for each condition are trained independently. The acoustic signature reference model for start-up learns the normal sound characteristics during the gradual increase in unit speed during startup; the acoustic signature reference model for shutdown learns the normal sound characteristics during the gradual decrease in unit speed during shutdown; the acoustic signature reference model for full-load conditions learns the normal sound characteristics of the unit during stable operation at rated power; and the acoustic signature reference model for load shedding learns the normal response sound characteristics of the unit when the load suddenly changes. By establishing acoustic signature reference models for multiple operating conditions, false alarms caused by switching between operating conditions can be avoided.

[0037] The model is trained using the constructed labeled training dataset. The model parameters are continuously optimized through the backpropagation algorithm until the classification accuracy of the model on the validation set reaches the preset requirement, and finally the trained fault diagnosis model is obtained.

[0038] S4. Input the voiceprint feature vector into the fault diagnosis model for inference and output the diagnosis result; wherein, the diagnosis result includes: the real-time status of the equipment and the fault type; This step involves inputting the real-time collected and pre-processed voiceprint feature vector into the trained fault diagnosis model for inference calculation, thereby enabling real-time monitoring of equipment operating status and automatic identification of fault types.

[0039] In the anomaly detection phase, the system first selects the corresponding acoustic signature reference model as the normal state model based on the current unit operating conditions. For example, when the unit is operating at full load, the full-load acoustic signature reference model is selected; when the unit is starting up, the start-up acoustic signature reference model is selected. Then, the real-time collected acoustic signature feature vector is input into the selected normal state model to calculate the deviation between the current acoustic signature feature and the normal acoustic signature distribution represented by the normal state model. This deviation is quantified as a numerical deviation degree. The larger the deviation degree, the more significant the difference between the current acoustic signature and the normal state. The system sets a dynamic threshold; when the calculated deviation degree exceeds this threshold, the current equipment is determined to be in an abnormal state; otherwise, it is determined to be in a normal state. The setting of the dynamic threshold comprehensively considers the statistical characteristics of historical operating data and actual operation and maintenance needs, and can be adjusted according to the on-site conditions.

[0040] In the fault classification stage, once the previous stage determines that the equipment is in an abnormal state, the system immediately inputs the abnormal sound signature feature vector into the fault state model for further analysis. Based on the learned fault sound signature patterns, the fault state model classifies and identifies the input abnormal sound signature and outputs the most likely fault type. For example, the model might identify specific fault types such as particle vibration noise inside the GIS, silt abrasion noise in the turbine runner chamber, or loosening of the transformer core.

[0041] In the confidence assessment phase, the system calculates the confidence probability of the diagnostic results output by the model. This probability value reflects the model's certainty about the current diagnostic conclusion. A higher confidence level indicates that the model is more confident in the diagnostic result, while a lower confidence level suggests that the diagnostic result is uncertain and may require verification using other methods.

[0042] After the above processing, the diagnostic results output in this step include two parts: the real-time status of the equipment and the type of fault, providing a basis for subsequent early warning decisions.

[0043] S5 performs trend prediction and comprehensive evaluation based on the diagnostic results of continuous time series, and generates graded early warning signals and maintenance suggestions based on the comprehensive evaluation.

[0044] This step performs trend analysis and comprehensive evaluation based on the continuous time series diagnostic results output in the previous step, and finally generates graded early warning signals and predictive maintenance recommendations.

[0045] In the trend prediction stage, this embodiment employs a time-series analysis algorithm to analyze the evolution trend of voiceprint features and diagnostic results. The system continuously records the voiceprint feature parameters and diagnostic results of the equipment, forming time-series data. By analyzing this historical data, the algorithm can identify the changing trends of the equipment's status and predict the possible deterioration trajectory of the equipment's status in the future. For example, if the voiceprint deviation of a certain device shows a gradually increasing trend, even if it has not yet exceeded the abnormal threshold, the system can predict that the device may fail at some point in the future, thereby providing decision support for advance scheduling of maintenance work.

[0046] In the comprehensive evaluation phase, the system comprehensively scores the health status of the equipment based on the current operating conditions. The comprehensive evaluation not only considers the acoustic signature diagnostic results but also incorporates operating parameters such as head and load to fully reflect the actual operating status of the equipment.

[0047] In the tiered early warning stage, the system generates tiered early warning signals based on the comprehensive evaluation results. This embodiment divides the early warning signals into four levels: Level IV is the normal state, indicating a high degree of matching between the device's acoustic signature and the baseline model, with all indicators within the normal range; the system does not trigger any warnings at this time. Level III is the attention state, indicating a slight deviation or unfavorable trend in the device's acoustic signature; the system generates a attention prompt to remind maintenance personnel to monitor subsequent changes in the device's status. Level II is the warning state, indicating that the device's acoustic signature has matched an early fault characteristic pattern; the system generates an inspection suggestion, recommending that maintenance personnel arrange on-site inspections to confirm the device's condition. Level I is the alarm state, indicating that the device's acoustic signature has matched a serious fault characteristic, such as a violent impact sound or a severe discharge sound; the system triggers an emergency response mechanism, including activating an audible and visual alarm device to alert on-site personnel and sending an emergency shutdown logic command to the control system to prevent the accident from escalating.

[0048] In addition to tiered early warning signals, the system also outputs maintenance recommendation reports based on diagnostic results and trend predictions. The reports include recommended maintenance time windows, maintenance items, and required spare parts, providing technical support for hydropower stations to shift from traditional reactive maintenance to predictive maintenance.

[0049] In summary, the online acoustic signature monitoring method for hydropower unit operation status in this embodiment achieves unified acoustic signature collection of key equipment across the entire hydropower unit through distributed optical fiber sensing technology, accurately identifies fault types using an artificial intelligence model that integrates expert experience, and provides timely and accurate decision support information for operation and maintenance personnel through trend analysis and a hierarchical early warning mechanism.

[0050] The following describes an online acoustic fingerprint monitoring device for the operating status of a hydropower unit provided in an embodiment of this application. The online acoustic fingerprint monitoring device and the online acoustic fingerprint monitoring method for the operating status of a hydropower unit described below can be referred to in correspondence with each other.

[0051] Please refer to Figure 2 , Figure 2 This is a schematic diagram of the structure of an online acoustic monitoring device for the operating status of a hydropower unit provided in an embodiment of this application.

[0052] In this embodiment, the device may include: The acoustic signal acquisition module 100 is used to acquire operating acoustic signals through a distributed optical fiber sensor array deployed in key equipment of the hydropower unit, and convert the acquired operating acoustic signals into multi-channel raw acoustic waveform data with timestamps and spatial location tags through photoelectric demodulation. Signal processing module 200 is used to perform noise reduction, frame windowing, and feature transformation on multi-channel raw acoustic waveform data to obtain acoustic feature vectors; The model training module 300 is used to train a fault diagnosis model based on historical voiceprint data and maintenance experience data; the fault diagnosis model includes a normal state model and a fault state model. The model inference module 400 is used to input the voiceprint feature vector into the fault diagnosis model for inference and output the diagnosis result; wherein, the diagnosis result includes: the real-time status of the equipment and the fault type; The early warning information generation module 500 is used to perform trend prediction and comprehensive evaluation based on the diagnostic results of continuous time series, and to generate graded early warning signals and maintenance suggestions based on the comprehensive evaluation.

[0053] This application also provides an online acoustic monitoring device for the operating status of hydropower units; please refer to it. Figure 3 , Figure 3 This is a schematic diagram of the structure of the online acoustic fingerprint monitoring device for the operating status of a hydropower unit provided in this application embodiment. The online acoustic fingerprint monitoring device for the operating status of a hydropower unit may include: Memory, used to store computer programs; The processor, when executing a computer program, can implement the steps of any of the above-described methods for online acoustic monitoring of the operating status of hydropower units.

[0054] like Figure 3 The diagram shows the structural composition of an online acoustic monitoring device for hydropower unit operation. This device may include a processor 10, a memory 11, a communication interface 12, and a communication bus 13. The processor 10, memory 11, and communication interface 12 all communicate with each other via the communication bus 13.

[0055] In this embodiment, the processor 10 may be a central processing unit (CPU), an application-specific integrated circuit, a digital signal processor, a field-programmable gate array, or other programmable logic devices.

[0056] The processor 10 can call the program stored in the memory 11. Specifically, the processor 10 can execute the operations in the embodiment of the abnormal IP identification method.

[0057] The memory 11 is used to store one or more programs. The programs may include program code, which includes computer operation instructions. In this embodiment, the memory 11 stores at least a program for implementing the following functions: S1 collects operating acoustic signals through a distributed fiber optic sensor array deployed on key equipment of the hydropower unit, and converts the collected operating acoustic signals into multi-channel raw acoustic waveform data with timestamps and spatial location tags through photoelectric demodulation. S2 performs noise reduction, frame windowing, and feature transformation on the multi-channel raw acoustic waveform data to obtain the voiceprint feature vector; S3, a fault diagnosis model is trained based on historical voiceprint data and maintenance experience data; the fault diagnosis model includes a normal state model and a fault state model; S4. Input the voiceprint feature vector into the fault diagnosis model for inference and output the diagnosis result; wherein, the diagnosis result includes: the real-time status of the equipment and the fault type; S5 performs trend prediction and comprehensive evaluation based on the diagnostic results of continuous time series, and generates graded early warning signals and maintenance suggestions based on the comprehensive evaluation.

[0058] In one possible implementation, the memory 11 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and applications required for at least one function; and the data storage area may store data created during use.

[0059] In addition, memory 11 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device or other volatile solid-state storage device.

[0060] Communication interface 12 can be an interface for the communication module, used to connect with other devices or systems.

[0061] Of course, it should be noted that, Figure 3 The structure shown does not constitute a limitation on the online acoustic fingerprint monitoring device for the operating status of hydropower units in this application embodiment. In practical applications, the online acoustic fingerprint monitoring device for the operating status of hydropower units may include devices that are more advanced than those described above. Figure 3 More or fewer components as shown, or combinations of certain components.

[0062] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can implement the steps of any of the above-described methods for online monitoring of the operating status of hydropower units using acoustic fingerprints.

[0063] The computer-readable storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0064] For a description of the computer-readable storage medium provided in this application, please refer to the above method embodiments; further details will not be repeated here.

[0065] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0066] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0067] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0068] The foregoing has provided a detailed description of the online acoustic fingerprint monitoring method, device, equipment, and computer-readable storage medium for monitoring the operating status of hydropower units. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A method for online acoustic monitoring of the operating status of a hydropower unit, characterized in that, include: S1 collects operating acoustic signals through a distributed fiber optic sensor array deployed on key equipment of the hydropower unit, and converts the collected operating acoustic signals into multi-channel raw acoustic waveform data with timestamps and spatial location tags through photoelectric demodulation. S2, perform noise reduction, frame windowing, and feature transformation on the multi-channel raw acoustic waveform data to obtain the voiceprint feature vector; S3, a fault diagnosis model is trained based on historical voiceprint data and maintenance experience data; wherein, the fault diagnosis model includes a normal state model and a fault state model; S4, input the voiceprint feature vector into the fault diagnosis model for inference, and output the diagnosis result; wherein, the diagnosis result includes: real-time status of the equipment and fault type; S5. Based on the diagnostic results of continuous time series, perform trend prediction and comprehensive evaluation, and generate graded early warning signals and maintenance suggestions according to the comprehensive evaluation.

2. The method for online acoustic monitoring of the operating status of a hydropower unit according to claim 1, characterized in that, In S1, the distributed optical fiber sensing array uses a distributed optical fiber acoustic wave sensor or a fiber optic grating sensor, which is deployed on the turbine top cover, generator wind tunnel, main transformer oil tank wall, GIS pipeline and inlet gate hoist. The timestamp synchronization accuracy of all measuring points reaches the microsecond level.

3. The method for online acoustic monitoring of the operating status of hydropower units according to claim 1, characterized in that, In step S2, the noise reduction process uses wavelet thresholding or spectral subtraction to filter out background white noise and power frequency interference; the frame length of the frame segmentation and windowing is 20ms to 50ms, and the window function is a Hamming window; the feature transformation includes extracting Mel frequency cepstral coefficients or Log-Mel spectrograms.

4. The method for online acoustic monitoring of the operating status of a hydropower unit according to claim 1, characterized in that, S3 includes: parsing historical maintenance logs using natural language processing technology, associating historical fault records with corresponding historical voiceprint data to generate a labeled training dataset; constructing the fault diagnosis model using a convolutional neural network or a long short-term memory network, and introducing an attention mechanism to perform weighted learning on feature regions that match the abnormal frequency bands described by expert experience; and training the fault diagnosis model using the training dataset to obtain the trained fault diagnosis model.

5. The method for online acoustic monitoring of the operating status of a hydropower unit according to claim 1, characterized in that, S4 includes: inputting the voiceprint feature vector into the normal state model to calculate the deviation degree, and determining an abnormal state when the deviation degree exceeds a dynamic threshold; inputting the voiceprint feature vector determined to be in an abnormal state into the fault state model to identify the fault type, and calculating the confidence level of the diagnostic result.

6. The method for online acoustic monitoring of the operating status of a hydropower unit according to claim 1, characterized in that, S3 further includes: establishing normal operating voiceprint reference models for different operating conditions, including start-up, shutdown, full load, and load shedding conditions; correspondingly, in S4, the corresponding voiceprint reference model is selected as the normal state model based on the current operating condition for deviation calculation.

7. The method for online acoustic monitoring of the operating status of a hydropower unit according to claim 1, characterized in that, The trend prediction includes: analyzing the evolution trend of voiceprint features using time series analysis algorithms; The graded early warning signal includes four levels: Level IV is the normal state, which does not trigger an early warning; Level III is the attention state, which generates a notice; Level II is the warning state, which generates an inspection suggestion; and Level I is the alarm state, which triggers emergency shutdown logic or an audible and visual alarm.

8. A device for online monitoring of the operating status of a hydropower unit using acoustic signature, characterized in that, include: The acoustic signal acquisition module is used to acquire operating acoustic signals through a distributed fiber optic sensor array deployed in key equipment of the hydropower unit. The acquired operating acoustic signals are converted into multi-channel raw acoustic waveform data with timestamps and spatial location tags through photoelectric demodulation. The signal processing module is used to perform noise reduction, frame windowing, and feature transformation on the multi-channel raw acoustic waveform data to obtain the voiceprint feature vector. The model training module is used to train a fault diagnosis model based on historical voiceprint data and maintenance experience data; wherein, the fault diagnosis model includes a normal state model and a fault state model; The model inference module is used to input the voiceprint feature vector into the fault diagnosis model for inference and output the diagnosis result; wherein, the diagnosis result includes: the real-time status of the equipment and the fault type; The early warning information generation module is used to perform trend prediction and comprehensive evaluation based on the diagnostic results of continuous time series, and generate graded early warning signals and maintenance suggestions based on the comprehensive evaluation.

9. A device for online acoustic monitoring of the operating status of a hydropower unit, characterized in that, include: Memory, used to store computer programs; A processor is configured to execute the computer program to implement the steps of the online acoustic fingerprint monitoring method for the operating status of a hydropower unit as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the online acoustic fingerprint monitoring method for the operating status of hydropower units as described in any one of claims 1 to 7.

Images